{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.537174633998793, "eval_steps": 500, "global_step": 214000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 3.0000000000000004e-07, "loss": 2.5393, "step": 5 }, { "epoch": 0.0, "learning_rate": 8.000000000000001e-07, "loss": 2.5408, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.3e-06, "loss": 2.4964, "step": 15 }, { "epoch": 0.0, "learning_rate": 1.8e-06, "loss": 2.3205, "step": 20 }, { "epoch": 0.0, "learning_rate": 2.3e-06, "loss": 2.2572, "step": 25 }, { "epoch": 0.0, "learning_rate": 2.8000000000000003e-06, "loss": 2.1767, "step": 30 }, { "epoch": 0.0, "learning_rate": 3.3e-06, "loss": 2.0971, "step": 35 }, { "epoch": 0.0, "learning_rate": 3.8e-06, "loss": 1.9351, "step": 40 }, { "epoch": 0.0, "learning_rate": 4.2999999999999995e-06, "loss": 1.7317, "step": 45 }, { "epoch": 0.0, "learning_rate": 4.800000000000001e-06, "loss": 1.5445, "step": 50 }, { "epoch": 0.0, "learning_rate": 5.3e-06, "loss": 1.482, "step": 55 }, { "epoch": 0.0, "learning_rate": 5.8e-06, "loss": 1.4195, "step": 60 }, { "epoch": 0.0, "learning_rate": 6.300000000000001e-06, "loss": 1.3888, "step": 65 }, { "epoch": 0.0, "learning_rate": 6.800000000000001e-06, "loss": 1.3599, "step": 70 }, { "epoch": 0.0, "learning_rate": 7.2999999999999996e-06, "loss": 1.321, "step": 75 }, { "epoch": 0.0, "learning_rate": 7.8e-06, "loss": 1.2837, "step": 80 }, { "epoch": 0.0, "learning_rate": 8.3e-06, "loss": 1.2548, "step": 85 }, { "epoch": 0.0, "learning_rate": 8.8e-06, "loss": 1.2292, "step": 90 }, { "epoch": 0.0, "learning_rate": 9.3e-06, "loss": 1.2355, "step": 95 }, { "epoch": 0.0, "learning_rate": 9.800000000000001e-06, "loss": 1.2001, "step": 100 }, { "epoch": 0.0, "learning_rate": 1.03e-05, "loss": 1.1911, "step": 105 }, { "epoch": 0.0, "learning_rate": 1.08e-05, "loss": 1.1845, "step": 110 }, { "epoch": 0.0, "learning_rate": 1.13e-05, "loss": 1.1777, "step": 115 }, { "epoch": 0.0, "learning_rate": 1.18e-05, "loss": 1.1733, "step": 120 }, { "epoch": 0.0, "learning_rate": 1.23e-05, "loss": 1.1697, "step": 125 }, { "epoch": 0.0, "learning_rate": 1.2800000000000001e-05, "loss": 1.167, "step": 130 }, { "epoch": 0.0, "learning_rate": 1.3300000000000001e-05, "loss": 1.1612, "step": 135 }, { "epoch": 0.0, "learning_rate": 1.3800000000000002e-05, "loss": 1.1599, "step": 140 }, { "epoch": 0.0, "learning_rate": 1.43e-05, "loss": 1.157, "step": 145 }, { "epoch": 0.0, "learning_rate": 1.48e-05, "loss": 1.1557, "step": 150 }, { "epoch": 0.0, "learning_rate": 1.53e-05, "loss": 1.1531, "step": 155 }, { "epoch": 0.0, "learning_rate": 1.58e-05, "loss": 1.1511, "step": 160 }, { "epoch": 0.0, "learning_rate": 1.63e-05, "loss": 1.1488, "step": 165 }, { "epoch": 0.0, "learning_rate": 1.6800000000000002e-05, "loss": 1.1472, "step": 170 }, { "epoch": 0.0, "learning_rate": 1.73e-05, "loss": 1.1431, "step": 175 }, { "epoch": 0.0, "learning_rate": 1.78e-05, "loss": 1.1421, "step": 180 }, { "epoch": 0.0, "learning_rate": 1.83e-05, "loss": 1.1365, "step": 185 }, { "epoch": 0.0, "learning_rate": 1.88e-05, "loss": 1.1362, "step": 190 }, { "epoch": 0.0, "learning_rate": 1.93e-05, "loss": 1.1316, "step": 195 }, { "epoch": 0.0, "learning_rate": 1.9800000000000004e-05, "loss": 1.1287, "step": 200 }, { "epoch": 0.0, "learning_rate": 2.0300000000000002e-05, "loss": 1.1277, "step": 205 }, { "epoch": 0.0, "learning_rate": 2.08e-05, "loss": 1.1255, "step": 210 }, { "epoch": 0.0, "learning_rate": 2.13e-05, "loss": 1.1231, "step": 215 }, { "epoch": 0.0, "learning_rate": 2.18e-05, "loss": 1.1204, "step": 220 }, { "epoch": 0.0, "learning_rate": 2.23e-05, "loss": 1.143, "step": 225 }, { "epoch": 0.0, "learning_rate": 2.2800000000000002e-05, "loss": 1.1179, "step": 230 }, { "epoch": 0.0, "learning_rate": 2.3300000000000004e-05, "loss": 1.1175, "step": 235 }, { "epoch": 0.0, "learning_rate": 2.38e-05, "loss": 1.1387, "step": 240 }, { "epoch": 0.0, "learning_rate": 2.43e-05, "loss": 1.1155, "step": 245 }, { "epoch": 0.0, "learning_rate": 2.48e-05, "loss": 1.1152, "step": 250 }, { "epoch": 0.0, "learning_rate": 2.5300000000000002e-05, "loss": 1.1141, "step": 255 }, { "epoch": 0.0, "learning_rate": 2.58e-05, "loss": 1.1152, "step": 260 }, { "epoch": 0.0, "learning_rate": 2.6300000000000002e-05, "loss": 1.1113, "step": 265 }, { "epoch": 0.0, "learning_rate": 2.6800000000000004e-05, "loss": 1.1138, "step": 270 }, { "epoch": 0.0, "learning_rate": 2.7300000000000003e-05, "loss": 1.1125, "step": 275 }, { "epoch": 0.0, "learning_rate": 2.7800000000000005e-05, "loss": 1.1109, "step": 280 }, { "epoch": 0.0, "learning_rate": 2.83e-05, "loss": 1.111, "step": 285 }, { "epoch": 0.0, "learning_rate": 2.88e-05, "loss": 1.1087, "step": 290 }, { "epoch": 0.0, "learning_rate": 2.93e-05, "loss": 1.1098, "step": 295 }, { "epoch": 0.0, "learning_rate": 2.98e-05, "loss": 1.109, "step": 300 }, { "epoch": 0.0, "learning_rate": 3.03e-05, "loss": 1.1114, "step": 305 }, { "epoch": 0.0, "learning_rate": 3.08e-05, "loss": 1.1089, "step": 310 }, { "epoch": 0.0, "learning_rate": 3.13e-05, "loss": 1.1079, "step": 315 }, { "epoch": 0.0, "learning_rate": 3.18e-05, "loss": 1.1077, "step": 320 }, { "epoch": 0.0, "learning_rate": 3.2300000000000006e-05, "loss": 1.1066, "step": 325 }, { "epoch": 0.0, "learning_rate": 3.2800000000000004e-05, "loss": 1.1083, "step": 330 }, { "epoch": 0.0, "learning_rate": 3.33e-05, "loss": 1.1075, "step": 335 }, { "epoch": 0.0, "learning_rate": 3.3700000000000006e-05, "loss": 1.108, "step": 340 }, { "epoch": 0.0, "learning_rate": 3.4200000000000005e-05, "loss": 1.1075, "step": 345 }, { "epoch": 0.0, "learning_rate": 3.4699999999999996e-05, "loss": 1.1045, "step": 350 }, { "epoch": 0.0, "learning_rate": 3.52e-05, "loss": 1.1055, "step": 355 }, { "epoch": 0.0, "learning_rate": 3.57e-05, "loss": 1.1053, "step": 360 }, { "epoch": 0.0, "learning_rate": 3.62e-05, "loss": 1.1056, "step": 365 }, { "epoch": 0.0, "learning_rate": 3.6700000000000004e-05, "loss": 1.1048, "step": 370 }, { "epoch": 0.0, "learning_rate": 3.72e-05, "loss": 1.1063, "step": 375 }, { "epoch": 0.0, "learning_rate": 3.77e-05, "loss": 1.1057, "step": 380 }, { "epoch": 0.0, "learning_rate": 3.82e-05, "loss": 1.1043, "step": 385 }, { "epoch": 0.0, "learning_rate": 3.8700000000000006e-05, "loss": 1.1042, "step": 390 }, { "epoch": 0.0, "learning_rate": 3.9200000000000004e-05, "loss": 1.1041, "step": 395 }, { "epoch": 0.0, "learning_rate": 3.97e-05, "loss": 1.1055, "step": 400 }, { "epoch": 0.0, "learning_rate": 4.02e-05, "loss": 1.1031, "step": 405 }, { "epoch": 0.0, "learning_rate": 4.07e-05, "loss": 1.1033, "step": 410 }, { "epoch": 0.0, "learning_rate": 4.12e-05, "loss": 1.1034, "step": 415 }, { "epoch": 0.0, "learning_rate": 4.17e-05, "loss": 1.1023, "step": 420 }, { "epoch": 0.0, "learning_rate": 4.22e-05, "loss": 1.1054, "step": 425 }, { "epoch": 0.0, "learning_rate": 4.27e-05, "loss": 1.1009, "step": 430 }, { "epoch": 0.0, "learning_rate": 4.32e-05, "loss": 1.102, "step": 435 }, { "epoch": 0.0, "learning_rate": 4.3700000000000005e-05, "loss": 1.1018, "step": 440 }, { "epoch": 0.0, "learning_rate": 4.4200000000000004e-05, "loss": 1.1038, "step": 445 }, { "epoch": 0.0, "learning_rate": 4.47e-05, "loss": 1.1015, "step": 450 }, { "epoch": 0.0, "learning_rate": 4.52e-05, "loss": 1.1021, "step": 455 }, { "epoch": 0.0, "learning_rate": 4.5700000000000006e-05, "loss": 1.1043, "step": 460 }, { "epoch": 0.0, "learning_rate": 4.6200000000000005e-05, "loss": 1.1023, "step": 465 }, { "epoch": 0.0, "learning_rate": 4.6700000000000003e-05, "loss": 1.1019, "step": 470 }, { "epoch": 0.0, "learning_rate": 4.72e-05, "loss": 1.1028, "step": 475 }, { "epoch": 0.0, "learning_rate": 4.77e-05, "loss": 1.1004, "step": 480 }, { "epoch": 0.0, "learning_rate": 4.82e-05, "loss": 1.1022, "step": 485 }, { "epoch": 0.0, "learning_rate": 4.87e-05, "loss": 1.1009, "step": 490 }, { "epoch": 0.0, "learning_rate": 4.92e-05, "loss": 1.0997, "step": 495 }, { "epoch": 0.0, "learning_rate": 4.97e-05, "loss": 1.1019, "step": 500 }, { "epoch": 0.0, "learning_rate": 5.02e-05, "loss": 1.1004, "step": 505 }, { "epoch": 0.0, "learning_rate": 5.0700000000000006e-05, "loss": 1.1008, "step": 510 }, { "epoch": 0.0, "learning_rate": 5.1200000000000004e-05, "loss": 1.0999, "step": 515 }, { "epoch": 0.0, "learning_rate": 5.17e-05, "loss": 1.0997, "step": 520 }, { "epoch": 0.0, "learning_rate": 5.22e-05, "loss": 1.101, "step": 525 }, { "epoch": 0.0, "learning_rate": 5.270000000000001e-05, "loss": 1.0994, "step": 530 }, { "epoch": 0.0, "learning_rate": 5.3200000000000006e-05, "loss": 1.103, "step": 535 }, { "epoch": 0.0, "learning_rate": 5.3700000000000004e-05, "loss": 1.1005, "step": 540 }, { "epoch": 0.0, "learning_rate": 5.420000000000001e-05, "loss": 1.1009, "step": 545 }, { "epoch": 0.0, "learning_rate": 5.470000000000001e-05, "loss": 1.0998, "step": 550 }, { "epoch": 0.0, "learning_rate": 5.520000000000001e-05, "loss": 1.0999, "step": 555 }, { "epoch": 0.0, "learning_rate": 5.5700000000000005e-05, "loss": 1.1005, "step": 560 }, { "epoch": 0.0, "learning_rate": 5.620000000000001e-05, "loss": 1.1009, "step": 565 }, { "epoch": 0.0, "learning_rate": 5.6699999999999996e-05, "loss": 1.1001, "step": 570 }, { "epoch": 0.0, "learning_rate": 5.72e-05, "loss": 1.0985, "step": 575 }, { "epoch": 0.0, "learning_rate": 5.77e-05, "loss": 1.0975, "step": 580 }, { "epoch": 0.0, "learning_rate": 5.82e-05, "loss": 1.0991, "step": 585 }, { "epoch": 0.0, "learning_rate": 5.87e-05, "loss": 1.0993, "step": 590 }, { "epoch": 0.0, "learning_rate": 5.92e-05, "loss": 1.0974, "step": 595 }, { "epoch": 0.0, "learning_rate": 5.97e-05, "loss": 1.0993, "step": 600 }, { "epoch": 0.0, "learning_rate": 6.02e-05, "loss": 1.0994, "step": 605 }, { "epoch": 0.0, "learning_rate": 6.07e-05, "loss": 1.1005, "step": 610 }, { "epoch": 0.0, "learning_rate": 6.12e-05, "loss": 1.1013, "step": 615 }, { "epoch": 0.0, "learning_rate": 6.170000000000001e-05, "loss": 1.099, "step": 620 }, { "epoch": 0.0, "learning_rate": 6.220000000000001e-05, "loss": 1.0981, "step": 625 }, { "epoch": 0.0, "learning_rate": 6.27e-05, "loss": 1.0982, "step": 630 }, { "epoch": 0.0, "learning_rate": 6.32e-05, "loss": 1.1002, "step": 635 }, { "epoch": 0.0, "learning_rate": 6.37e-05, "loss": 1.0991, "step": 640 }, { "epoch": 0.0, "learning_rate": 6.42e-05, "loss": 1.0995, "step": 645 }, { "epoch": 0.0, "learning_rate": 6.47e-05, "loss": 1.0982, "step": 650 }, { "epoch": 0.0, "learning_rate": 6.52e-05, "loss": 1.097, "step": 655 }, { "epoch": 0.0, "learning_rate": 6.570000000000001e-05, "loss": 1.0974, "step": 660 }, { "epoch": 0.0, "learning_rate": 6.620000000000001e-05, "loss": 1.0967, "step": 665 }, { "epoch": 0.0, "learning_rate": 6.670000000000001e-05, "loss": 1.0967, "step": 670 }, { "epoch": 0.0, "learning_rate": 6.720000000000001e-05, "loss": 1.0978, "step": 675 }, { "epoch": 0.0, "learning_rate": 6.77e-05, "loss": 1.0993, "step": 680 }, { "epoch": 0.0, "learning_rate": 6.82e-05, "loss": 1.0978, "step": 685 }, { "epoch": 0.0, "learning_rate": 6.87e-05, "loss": 1.0971, "step": 690 }, { "epoch": 0.0, "learning_rate": 6.92e-05, "loss": 1.0973, "step": 695 }, { "epoch": 0.0, "learning_rate": 6.97e-05, "loss": 1.0979, "step": 700 }, { "epoch": 0.0, "learning_rate": 7.02e-05, "loss": 1.0981, "step": 705 }, { "epoch": 0.0, "learning_rate": 7.07e-05, "loss": 1.0981, "step": 710 }, { "epoch": 0.0, "learning_rate": 7.12e-05, "loss": 1.098, "step": 715 }, { "epoch": 0.0, "learning_rate": 7.17e-05, "loss": 1.0971, "step": 720 }, { "epoch": 0.0, "learning_rate": 7.22e-05, "loss": 1.0979, "step": 725 }, { "epoch": 0.0, "learning_rate": 7.27e-05, "loss": 1.0976, "step": 730 }, { "epoch": 0.0, "learning_rate": 7.32e-05, "loss": 1.098, "step": 735 }, { "epoch": 0.0, "learning_rate": 7.37e-05, "loss": 1.0971, "step": 740 }, { "epoch": 0.0, "learning_rate": 7.42e-05, "loss": 1.0978, "step": 745 }, { "epoch": 0.0, "learning_rate": 7.47e-05, "loss": 1.0953, "step": 750 }, { "epoch": 0.0, "learning_rate": 7.52e-05, "loss": 1.0982, "step": 755 }, { "epoch": 0.0, "learning_rate": 7.570000000000001e-05, "loss": 1.0962, "step": 760 }, { "epoch": 0.0, "learning_rate": 7.620000000000001e-05, "loss": 1.0965, "step": 765 }, { "epoch": 0.0, "learning_rate": 7.670000000000001e-05, "loss": 1.0964, "step": 770 }, { "epoch": 0.0, "learning_rate": 7.72e-05, "loss": 1.0959, "step": 775 }, { "epoch": 0.0, "learning_rate": 7.77e-05, "loss": 1.0963, "step": 780 }, { "epoch": 0.0, "learning_rate": 7.82e-05, "loss": 1.0969, "step": 785 }, { "epoch": 0.0, "learning_rate": 7.87e-05, "loss": 1.0956, "step": 790 }, { "epoch": 0.0, "learning_rate": 7.920000000000001e-05, "loss": 1.0965, "step": 795 }, { "epoch": 0.0, "learning_rate": 7.970000000000001e-05, "loss": 1.0972, "step": 800 }, { "epoch": 0.0, "learning_rate": 8.020000000000001e-05, "loss": 1.0975, "step": 805 }, { "epoch": 0.0, "learning_rate": 8.070000000000001e-05, "loss": 1.097, "step": 810 }, { "epoch": 0.0, "learning_rate": 8.120000000000001e-05, "loss": 1.0964, "step": 815 }, { "epoch": 0.0, "learning_rate": 8.17e-05, "loss": 1.0955, "step": 820 }, { "epoch": 0.0, "learning_rate": 8.22e-05, "loss": 1.1175, "step": 825 }, { "epoch": 0.0, "learning_rate": 8.27e-05, "loss": 1.0963, "step": 830 }, { "epoch": 0.0, "learning_rate": 8.32e-05, "loss": 1.0965, "step": 835 }, { "epoch": 0.0, "learning_rate": 8.37e-05, "loss": 1.0946, "step": 840 }, { "epoch": 0.0, "learning_rate": 8.42e-05, "loss": 1.0963, "step": 845 }, { "epoch": 0.0, "learning_rate": 8.47e-05, "loss": 1.0952, "step": 850 }, { "epoch": 0.0, "learning_rate": 8.52e-05, "loss": 1.0971, "step": 855 }, { "epoch": 0.0, "learning_rate": 8.57e-05, "loss": 1.096, "step": 860 }, { "epoch": 0.0, "learning_rate": 8.620000000000001e-05, "loss": 1.0948, "step": 865 }, { "epoch": 0.0, "learning_rate": 8.67e-05, "loss": 1.0968, "step": 870 }, { "epoch": 0.0, "learning_rate": 8.72e-05, "loss": 1.0949, "step": 875 }, { "epoch": 0.0, "learning_rate": 8.77e-05, "loss": 1.0952, "step": 880 }, { "epoch": 0.0, "learning_rate": 8.82e-05, "loss": 1.0953, "step": 885 }, { "epoch": 0.0, "learning_rate": 8.87e-05, "loss": 1.0954, "step": 890 }, { "epoch": 0.0, "learning_rate": 8.92e-05, "loss": 1.096, "step": 895 }, { "epoch": 0.0, "learning_rate": 8.970000000000001e-05, "loss": 1.096, "step": 900 }, { "epoch": 0.0, "learning_rate": 9.020000000000001e-05, "loss": 1.0964, "step": 905 }, { "epoch": 0.0, "learning_rate": 9.070000000000001e-05, "loss": 1.0971, "step": 910 }, { "epoch": 0.0, "learning_rate": 9.120000000000001e-05, "loss": 1.0942, "step": 915 }, { "epoch": 0.0, "learning_rate": 9.17e-05, "loss": 1.0969, "step": 920 }, { "epoch": 0.0, "learning_rate": 9.22e-05, "loss": 1.0949, "step": 925 }, { "epoch": 0.0, "learning_rate": 9.27e-05, "loss": 1.0947, "step": 930 }, { "epoch": 0.0, "learning_rate": 9.320000000000002e-05, "loss": 1.0945, "step": 935 }, { "epoch": 0.0, "learning_rate": 9.370000000000001e-05, "loss": 1.0966, "step": 940 }, { "epoch": 0.0, "learning_rate": 9.42e-05, "loss": 1.0944, "step": 945 }, { "epoch": 0.0, "learning_rate": 9.47e-05, "loss": 1.0968, "step": 950 }, { "epoch": 0.0, "learning_rate": 9.52e-05, "loss": 1.0967, "step": 955 }, { "epoch": 0.0, "learning_rate": 9.57e-05, "loss": 1.0926, "step": 960 }, { "epoch": 0.0, "learning_rate": 9.620000000000001e-05, "loss": 1.0948, "step": 965 }, { "epoch": 0.0, "learning_rate": 9.67e-05, "loss": 1.0946, "step": 970 }, { "epoch": 0.0, "learning_rate": 9.72e-05, "loss": 1.122, "step": 975 }, { "epoch": 0.0, "learning_rate": 9.77e-05, "loss": 1.0926, "step": 980 }, { "epoch": 0.0, "learning_rate": 9.82e-05, "loss": 1.0961, "step": 985 }, { "epoch": 0.0, "learning_rate": 9.87e-05, "loss": 1.0953, "step": 990 }, { "epoch": 0.0, "learning_rate": 9.92e-05, "loss": 1.0939, "step": 995 }, { "epoch": 0.0, "learning_rate": 9.970000000000001e-05, "loss": 1.0958, "step": 1000 }, { "epoch": 0.0, "learning_rate": 9.999949670340733e-05, "loss": 1.0933, "step": 1005 }, { "epoch": 0.0, "learning_rate": 9.999823846192562e-05, "loss": 1.0937, "step": 1010 }, { "epoch": 0.0, "learning_rate": 9.999698022044391e-05, "loss": 1.0941, "step": 1015 }, { "epoch": 0.0, "learning_rate": 9.99957219789622e-05, "loss": 1.0934, "step": 1020 }, { "epoch": 0.0, "learning_rate": 9.999446373748051e-05, "loss": 1.093, "step": 1025 }, { "epoch": 0.0, "learning_rate": 9.99932054959988e-05, "loss": 1.0935, "step": 1030 }, { "epoch": 0.0, "learning_rate": 9.999194725451709e-05, "loss": 1.0935, "step": 1035 }, { "epoch": 0.0, "learning_rate": 9.999068901303538e-05, "loss": 1.0927, "step": 1040 }, { "epoch": 0.0, "learning_rate": 9.998943077155368e-05, "loss": 1.0927, "step": 1045 }, { "epoch": 0.0, "learning_rate": 9.998817253007198e-05, "loss": 1.092, "step": 1050 }, { "epoch": 0.0, "learning_rate": 9.998691428859027e-05, "loss": 1.0928, "step": 1055 }, { "epoch": 0.0, "learning_rate": 9.998565604710856e-05, "loss": 1.0929, "step": 1060 }, { "epoch": 0.0, "learning_rate": 9.998439780562686e-05, "loss": 1.0912, "step": 1065 }, { "epoch": 0.0, "learning_rate": 9.998313956414516e-05, "loss": 1.0921, "step": 1070 }, { "epoch": 0.0, "learning_rate": 9.998188132266345e-05, "loss": 1.0898, "step": 1075 }, { "epoch": 0.0, "learning_rate": 9.998062308118174e-05, "loss": 1.0861, "step": 1080 }, { "epoch": 0.0, "learning_rate": 9.997936483970004e-05, "loss": 1.0849, "step": 1085 }, { "epoch": 0.0, "learning_rate": 9.997810659821834e-05, "loss": 1.0855, "step": 1090 }, { "epoch": 0.0, "learning_rate": 9.997684835673663e-05, "loss": 1.0821, "step": 1095 }, { "epoch": 0.0, "learning_rate": 9.997559011525492e-05, "loss": 1.0811, "step": 1100 }, { "epoch": 0.0, "learning_rate": 9.997433187377322e-05, "loss": 1.0781, "step": 1105 }, { "epoch": 0.0, "learning_rate": 9.997307363229151e-05, "loss": 1.1022, "step": 1110 }, { "epoch": 0.0, "learning_rate": 9.997181539080981e-05, "loss": 1.0776, "step": 1115 }, { "epoch": 0.0, "learning_rate": 9.99705571493281e-05, "loss": 1.0758, "step": 1120 }, { "epoch": 0.0, "learning_rate": 9.99692989078464e-05, "loss": 1.0762, "step": 1125 }, { "epoch": 0.0, "learning_rate": 9.996804066636469e-05, "loss": 1.0737, "step": 1130 }, { "epoch": 0.0, "learning_rate": 9.996678242488299e-05, "loss": 1.0696, "step": 1135 }, { "epoch": 0.0, "learning_rate": 9.996552418340128e-05, "loss": 1.0686, "step": 1140 }, { "epoch": 0.0, "learning_rate": 9.996426594191958e-05, "loss": 1.0686, "step": 1145 }, { "epoch": 0.0, "learning_rate": 9.996300770043787e-05, "loss": 1.0691, "step": 1150 }, { "epoch": 0.0, "learning_rate": 9.996174945895617e-05, "loss": 1.0674, "step": 1155 }, { "epoch": 0.0, "learning_rate": 9.996049121747446e-05, "loss": 1.0685, "step": 1160 }, { "epoch": 0.0, "learning_rate": 9.995923297599276e-05, "loss": 1.0695, "step": 1165 }, { "epoch": 0.0, "learning_rate": 9.995797473451105e-05, "loss": 1.067, "step": 1170 }, { "epoch": 0.0, "learning_rate": 9.995671649302934e-05, "loss": 1.0681, "step": 1175 }, { "epoch": 0.0, "learning_rate": 9.995545825154764e-05, "loss": 1.0653, "step": 1180 }, { "epoch": 0.0, "learning_rate": 9.995420001006594e-05, "loss": 1.0667, "step": 1185 }, { "epoch": 0.0, "learning_rate": 9.995294176858423e-05, "loss": 1.0683, "step": 1190 }, { "epoch": 0.0, "learning_rate": 9.995168352710252e-05, "loss": 1.0673, "step": 1195 }, { "epoch": 0.0, "learning_rate": 9.995042528562082e-05, "loss": 1.0628, "step": 1200 }, { "epoch": 0.0, "learning_rate": 9.994916704413912e-05, "loss": 1.0643, "step": 1205 }, { "epoch": 0.0, "learning_rate": 9.994790880265741e-05, "loss": 1.0666, "step": 1210 }, { "epoch": 0.0, "learning_rate": 9.99466505611757e-05, "loss": 1.0649, "step": 1215 }, { "epoch": 0.0, "learning_rate": 9.9945392319694e-05, "loss": 1.0639, "step": 1220 }, { "epoch": 0.0, "learning_rate": 9.99441340782123e-05, "loss": 1.0643, "step": 1225 }, { "epoch": 0.0, "learning_rate": 9.994287583673059e-05, "loss": 1.0657, "step": 1230 }, { "epoch": 0.0, "learning_rate": 9.994161759524888e-05, "loss": 1.0633, "step": 1235 }, { "epoch": 0.0, "learning_rate": 9.994035935376717e-05, "loss": 1.0635, "step": 1240 }, { "epoch": 0.0, "learning_rate": 9.993910111228548e-05, "loss": 1.0615, "step": 1245 }, { "epoch": 0.0, "learning_rate": 9.993784287080377e-05, "loss": 1.0638, "step": 1250 }, { "epoch": 0.0, "learning_rate": 9.993658462932206e-05, "loss": 1.0628, "step": 1255 }, { "epoch": 0.0, "learning_rate": 9.993532638784035e-05, "loss": 1.0632, "step": 1260 }, { "epoch": 0.0, "learning_rate": 9.993406814635866e-05, "loss": 1.0626, "step": 1265 }, { "epoch": 0.0, "learning_rate": 9.993280990487695e-05, "loss": 1.0625, "step": 1270 }, { "epoch": 0.0, "learning_rate": 9.993155166339524e-05, "loss": 1.0648, "step": 1275 }, { "epoch": 0.0, "learning_rate": 9.993029342191353e-05, "loss": 1.0642, "step": 1280 }, { "epoch": 0.0, "learning_rate": 9.992903518043184e-05, "loss": 1.0615, "step": 1285 }, { "epoch": 0.0, "learning_rate": 9.992777693895013e-05, "loss": 1.063, "step": 1290 }, { "epoch": 0.0, "learning_rate": 9.992651869746842e-05, "loss": 1.0851, "step": 1295 }, { "epoch": 0.0, "learning_rate": 9.992526045598671e-05, "loss": 1.0595, "step": 1300 }, { "epoch": 0.0, "learning_rate": 9.9924002214505e-05, "loss": 1.0612, "step": 1305 }, { "epoch": 0.0, "learning_rate": 9.99227439730233e-05, "loss": 1.0626, "step": 1310 }, { "epoch": 0.0, "learning_rate": 9.99214857315416e-05, "loss": 1.0606, "step": 1315 }, { "epoch": 0.0, "learning_rate": 9.992022749005989e-05, "loss": 1.0621, "step": 1320 }, { "epoch": 0.0, "learning_rate": 9.991896924857818e-05, "loss": 1.0632, "step": 1325 }, { "epoch": 0.0, "learning_rate": 9.991771100709649e-05, "loss": 1.061, "step": 1330 }, { "epoch": 0.0, "learning_rate": 9.991645276561478e-05, "loss": 1.0623, "step": 1335 }, { "epoch": 0.0, "learning_rate": 9.991519452413307e-05, "loss": 1.0609, "step": 1340 }, { "epoch": 0.0, "learning_rate": 9.991393628265136e-05, "loss": 1.0634, "step": 1345 }, { "epoch": 0.0, "learning_rate": 9.991267804116967e-05, "loss": 1.063, "step": 1350 }, { "epoch": 0.0, "learning_rate": 9.991141979968797e-05, "loss": 1.0606, "step": 1355 }, { "epoch": 0.0, "learning_rate": 9.991016155820626e-05, "loss": 1.0647, "step": 1360 }, { "epoch": 0.0, "learning_rate": 9.990890331672455e-05, "loss": 1.0627, "step": 1365 }, { "epoch": 0.0, "learning_rate": 9.990764507524285e-05, "loss": 1.0625, "step": 1370 }, { "epoch": 0.0, "learning_rate": 9.990638683376115e-05, "loss": 1.0613, "step": 1375 }, { "epoch": 0.0, "learning_rate": 9.990512859227944e-05, "loss": 1.0606, "step": 1380 }, { "epoch": 0.0, "learning_rate": 9.990387035079773e-05, "loss": 1.0597, "step": 1385 }, { "epoch": 0.0, "learning_rate": 9.990261210931603e-05, "loss": 1.0616, "step": 1390 }, { "epoch": 0.0, "learning_rate": 9.990135386783432e-05, "loss": 1.0626, "step": 1395 }, { "epoch": 0.0, "learning_rate": 9.990009562635262e-05, "loss": 1.0623, "step": 1400 }, { "epoch": 0.0, "learning_rate": 9.989883738487091e-05, "loss": 1.0611, "step": 1405 }, { "epoch": 0.0, "learning_rate": 9.98975791433892e-05, "loss": 1.0623, "step": 1410 }, { "epoch": 0.0, "learning_rate": 9.98963209019075e-05, "loss": 1.0599, "step": 1415 }, { "epoch": 0.0, "learning_rate": 9.98950626604258e-05, "loss": 1.061, "step": 1420 }, { "epoch": 0.0, "learning_rate": 9.98938044189441e-05, "loss": 1.0585, "step": 1425 }, { "epoch": 0.0, "learning_rate": 9.989254617746239e-05, "loss": 1.0601, "step": 1430 }, { "epoch": 0.0, "learning_rate": 9.989128793598068e-05, "loss": 1.0599, "step": 1435 }, { "epoch": 0.0, "learning_rate": 9.989002969449898e-05, "loss": 1.0617, "step": 1440 }, { "epoch": 0.0, "learning_rate": 9.988877145301727e-05, "loss": 1.0631, "step": 1445 }, { "epoch": 0.0, "learning_rate": 9.988751321153557e-05, "loss": 1.0596, "step": 1450 }, { "epoch": 0.0, "learning_rate": 9.988625497005386e-05, "loss": 1.0606, "step": 1455 }, { "epoch": 0.0, "learning_rate": 9.988499672857215e-05, "loss": 1.0592, "step": 1460 }, { "epoch": 0.0, "learning_rate": 9.988373848709045e-05, "loss": 1.0608, "step": 1465 }, { "epoch": 0.0, "learning_rate": 9.988248024560875e-05, "loss": 1.0597, "step": 1470 }, { "epoch": 0.0, "learning_rate": 9.988122200412704e-05, "loss": 1.0606, "step": 1475 }, { "epoch": 0.0, "learning_rate": 9.987996376264533e-05, "loss": 1.0601, "step": 1480 }, { "epoch": 0.0, "learning_rate": 9.987870552116363e-05, "loss": 1.0593, "step": 1485 }, { "epoch": 0.0, "learning_rate": 9.987744727968193e-05, "loss": 1.06, "step": 1490 }, { "epoch": 0.0, "learning_rate": 9.987618903820022e-05, "loss": 1.0582, "step": 1495 }, { "epoch": 0.0, "learning_rate": 9.987493079671851e-05, "loss": 1.0582, "step": 1500 }, { "epoch": 0.0, "learning_rate": 9.987367255523681e-05, "loss": 1.0585, "step": 1505 }, { "epoch": 0.0, "learning_rate": 9.98724143137551e-05, "loss": 1.0563, "step": 1510 }, { "epoch": 0.0, "learning_rate": 9.98711560722734e-05, "loss": 1.057, "step": 1515 }, { "epoch": 0.0, "learning_rate": 9.986989783079169e-05, "loss": 1.0554, "step": 1520 }, { "epoch": 0.0, "learning_rate": 9.986863958930998e-05, "loss": 1.0575, "step": 1525 }, { "epoch": 0.0, "learning_rate": 9.986738134782829e-05, "loss": 1.0566, "step": 1530 }, { "epoch": 0.0, "learning_rate": 9.986612310634658e-05, "loss": 1.0559, "step": 1535 }, { "epoch": 0.0, "learning_rate": 9.986486486486487e-05, "loss": 1.0574, "step": 1540 }, { "epoch": 0.0, "learning_rate": 9.986360662338316e-05, "loss": 1.056, "step": 1545 }, { "epoch": 0.0, "learning_rate": 9.986234838190147e-05, "loss": 1.0574, "step": 1550 }, { "epoch": 0.0, "learning_rate": 9.986109014041976e-05, "loss": 1.0571, "step": 1555 }, { "epoch": 0.0, "learning_rate": 9.985983189893805e-05, "loss": 1.0562, "step": 1560 }, { "epoch": 0.0, "learning_rate": 9.985857365745634e-05, "loss": 1.0544, "step": 1565 }, { "epoch": 0.0, "learning_rate": 9.985731541597465e-05, "loss": 1.0566, "step": 1570 }, { "epoch": 0.0, "learning_rate": 9.985605717449294e-05, "loss": 1.0536, "step": 1575 }, { "epoch": 0.0, "learning_rate": 9.985479893301123e-05, "loss": 1.057, "step": 1580 }, { "epoch": 0.0, "learning_rate": 9.985354069152952e-05, "loss": 1.055, "step": 1585 }, { "epoch": 0.0, "learning_rate": 9.985228245004781e-05, "loss": 1.0545, "step": 1590 }, { "epoch": 0.0, "learning_rate": 9.985102420856612e-05, "loss": 1.0539, "step": 1595 }, { "epoch": 0.0, "learning_rate": 9.984976596708441e-05, "loss": 1.0556, "step": 1600 }, { "epoch": 0.0, "learning_rate": 9.98485077256027e-05, "loss": 1.0556, "step": 1605 }, { "epoch": 0.0, "learning_rate": 9.984724948412099e-05, "loss": 1.0565, "step": 1610 }, { "epoch": 0.0, "learning_rate": 9.98459912426393e-05, "loss": 1.0537, "step": 1615 }, { "epoch": 0.0, "learning_rate": 9.984473300115759e-05, "loss": 1.0544, "step": 1620 }, { "epoch": 0.0, "learning_rate": 9.984347475967588e-05, "loss": 1.0745, "step": 1625 }, { "epoch": 0.0, "learning_rate": 9.984221651819417e-05, "loss": 1.0778, "step": 1630 }, { "epoch": 0.0, "learning_rate": 9.984095827671248e-05, "loss": 1.0528, "step": 1635 }, { "epoch": 0.0, "learning_rate": 9.983970003523077e-05, "loss": 1.0518, "step": 1640 }, { "epoch": 0.0, "learning_rate": 9.983844179374906e-05, "loss": 1.0626, "step": 1645 }, { "epoch": 0.0, "learning_rate": 9.983718355226735e-05, "loss": 1.0549, "step": 1650 }, { "epoch": 0.0, "learning_rate": 9.983592531078564e-05, "loss": 1.0529, "step": 1655 }, { "epoch": 0.0, "learning_rate": 9.983466706930395e-05, "loss": 1.052, "step": 1660 }, { "epoch": 0.0, "learning_rate": 9.983340882782224e-05, "loss": 1.0507, "step": 1665 }, { "epoch": 0.0, "learning_rate": 9.983215058634053e-05, "loss": 1.0727, "step": 1670 }, { "epoch": 0.0, "learning_rate": 9.983089234485882e-05, "loss": 1.0535, "step": 1675 }, { "epoch": 0.0, "learning_rate": 9.982963410337713e-05, "loss": 1.0529, "step": 1680 }, { "epoch": 0.0, "learning_rate": 9.982837586189542e-05, "loss": 1.0535, "step": 1685 }, { "epoch": 0.0, "learning_rate": 9.982711762041371e-05, "loss": 1.0522, "step": 1690 }, { "epoch": 0.0, "learning_rate": 9.9825859378932e-05, "loss": 1.0524, "step": 1695 }, { "epoch": 0.0, "learning_rate": 9.982460113745031e-05, "loss": 1.0532, "step": 1700 }, { "epoch": 0.0, "learning_rate": 9.98233428959686e-05, "loss": 1.0495, "step": 1705 }, { "epoch": 0.0, "learning_rate": 9.982208465448689e-05, "loss": 1.0536, "step": 1710 }, { "epoch": 0.0, "learning_rate": 9.982082641300518e-05, "loss": 1.0515, "step": 1715 }, { "epoch": 0.0, "learning_rate": 9.981956817152348e-05, "loss": 1.0488, "step": 1720 }, { "epoch": 0.0, "learning_rate": 9.981830993004178e-05, "loss": 1.0512, "step": 1725 }, { "epoch": 0.0, "learning_rate": 9.981705168856007e-05, "loss": 1.0506, "step": 1730 }, { "epoch": 0.0, "learning_rate": 9.981579344707836e-05, "loss": 1.0535, "step": 1735 }, { "epoch": 0.0, "learning_rate": 9.981453520559666e-05, "loss": 1.0524, "step": 1740 }, { "epoch": 0.0, "learning_rate": 9.981327696411496e-05, "loss": 1.0516, "step": 1745 }, { "epoch": 0.0, "learning_rate": 9.981201872263325e-05, "loss": 1.0518, "step": 1750 }, { "epoch": 0.0, "learning_rate": 9.981076048115154e-05, "loss": 1.0519, "step": 1755 }, { "epoch": 0.0, "learning_rate": 9.980950223966984e-05, "loss": 1.049, "step": 1760 }, { "epoch": 0.0, "learning_rate": 9.980824399818813e-05, "loss": 1.0516, "step": 1765 }, { "epoch": 0.0, "learning_rate": 9.980698575670643e-05, "loss": 1.0726, "step": 1770 }, { "epoch": 0.0, "learning_rate": 9.980572751522472e-05, "loss": 1.0511, "step": 1775 }, { "epoch": 0.0, "learning_rate": 9.980446927374302e-05, "loss": 1.0512, "step": 1780 }, { "epoch": 0.0, "learning_rate": 9.98032110322613e-05, "loss": 1.0505, "step": 1785 }, { "epoch": 0.0, "learning_rate": 9.980195279077961e-05, "loss": 1.048, "step": 1790 }, { "epoch": 0.0, "learning_rate": 9.98006945492979e-05, "loss": 1.0514, "step": 1795 }, { "epoch": 0.0, "learning_rate": 9.97994363078162e-05, "loss": 1.049, "step": 1800 }, { "epoch": 0.0, "learning_rate": 9.979817806633449e-05, "loss": 1.0498, "step": 1805 }, { "epoch": 0.0, "learning_rate": 9.979691982485279e-05, "loss": 1.0493, "step": 1810 }, { "epoch": 0.0, "learning_rate": 9.979566158337108e-05, "loss": 1.0514, "step": 1815 }, { "epoch": 0.0, "learning_rate": 9.979440334188937e-05, "loss": 1.0494, "step": 1820 }, { "epoch": 0.0, "learning_rate": 9.979314510040767e-05, "loss": 1.0474, "step": 1825 }, { "epoch": 0.0, "learning_rate": 9.979188685892596e-05, "loss": 1.0488, "step": 1830 }, { "epoch": 0.0, "learning_rate": 9.979062861744426e-05, "loss": 1.0483, "step": 1835 }, { "epoch": 0.0, "learning_rate": 9.978937037596255e-05, "loss": 1.0506, "step": 1840 }, { "epoch": 0.0, "learning_rate": 9.978811213448085e-05, "loss": 1.0487, "step": 1845 }, { "epoch": 0.0, "learning_rate": 9.978685389299914e-05, "loss": 1.0477, "step": 1850 }, { "epoch": 0.0, "learning_rate": 9.978559565151744e-05, "loss": 1.0491, "step": 1855 }, { "epoch": 0.0, "learning_rate": 9.978433741003575e-05, "loss": 1.0496, "step": 1860 }, { "epoch": 0.0, "learning_rate": 9.978307916855404e-05, "loss": 1.0491, "step": 1865 }, { "epoch": 0.0, "learning_rate": 9.978182092707233e-05, "loss": 1.0487, "step": 1870 }, { "epoch": 0.0, "learning_rate": 9.978056268559062e-05, "loss": 1.0502, "step": 1875 }, { "epoch": 0.0, "learning_rate": 9.977930444410893e-05, "loss": 1.0521, "step": 1880 }, { "epoch": 0.0, "learning_rate": 9.977804620262722e-05, "loss": 1.0498, "step": 1885 }, { "epoch": 0.0, "learning_rate": 9.977678796114551e-05, "loss": 1.0504, "step": 1890 }, { "epoch": 0.0, "learning_rate": 9.97755297196638e-05, "loss": 1.0479, "step": 1895 }, { "epoch": 0.0, "learning_rate": 9.977427147818211e-05, "loss": 1.0489, "step": 1900 }, { "epoch": 0.0, "learning_rate": 9.97730132367004e-05, "loss": 1.0486, "step": 1905 }, { "epoch": 0.0, "learning_rate": 9.977175499521869e-05, "loss": 1.047, "step": 1910 }, { "epoch": 0.0, "learning_rate": 9.977049675373698e-05, "loss": 1.049, "step": 1915 }, { "epoch": 0.0, "learning_rate": 9.976923851225527e-05, "loss": 1.0487, "step": 1920 }, { "epoch": 0.0, "learning_rate": 9.976798027077358e-05, "loss": 1.0463, "step": 1925 }, { "epoch": 0.0, "learning_rate": 9.976672202929187e-05, "loss": 1.0483, "step": 1930 }, { "epoch": 0.0, "learning_rate": 9.976546378781016e-05, "loss": 1.0485, "step": 1935 }, { "epoch": 0.0, "learning_rate": 9.976420554632845e-05, "loss": 1.0473, "step": 1940 }, { "epoch": 0.0, "learning_rate": 9.976294730484676e-05, "loss": 1.0469, "step": 1945 }, { "epoch": 0.0, "learning_rate": 9.976168906336505e-05, "loss": 1.0472, "step": 1950 }, { "epoch": 0.0, "learning_rate": 9.976043082188334e-05, "loss": 1.048, "step": 1955 }, { "epoch": 0.0, "learning_rate": 9.975917258040163e-05, "loss": 1.0491, "step": 1960 }, { "epoch": 0.0, "learning_rate": 9.975791433891994e-05, "loss": 1.0455, "step": 1965 }, { "epoch": 0.0, "learning_rate": 9.975665609743823e-05, "loss": 1.0485, "step": 1970 }, { "epoch": 0.0, "learning_rate": 9.975539785595652e-05, "loss": 1.0482, "step": 1975 }, { "epoch": 0.0, "learning_rate": 9.975413961447481e-05, "loss": 1.0467, "step": 1980 }, { "epoch": 0.0, "learning_rate": 9.97528813729931e-05, "loss": 1.0486, "step": 1985 }, { "epoch": 0.0, "learning_rate": 9.975162313151141e-05, "loss": 1.0495, "step": 1990 }, { "epoch": 0.01, "learning_rate": 9.97503648900297e-05, "loss": 1.0477, "step": 1995 }, { "epoch": 0.01, "learning_rate": 9.9749106648548e-05, "loss": 1.0473, "step": 2000 }, { "epoch": 0.01, "learning_rate": 9.974784840706629e-05, "loss": 1.046, "step": 2005 }, { "epoch": 0.01, "learning_rate": 9.974659016558459e-05, "loss": 1.0471, "step": 2010 }, { "epoch": 0.01, "learning_rate": 9.974533192410288e-05, "loss": 1.0473, "step": 2015 }, { "epoch": 0.01, "learning_rate": 9.974407368262117e-05, "loss": 1.0498, "step": 2020 }, { "epoch": 0.01, "learning_rate": 9.974281544113947e-05, "loss": 1.0464, "step": 2025 }, { "epoch": 0.01, "learning_rate": 9.974155719965777e-05, "loss": 1.0476, "step": 2030 }, { "epoch": 0.01, "learning_rate": 9.974029895817606e-05, "loss": 1.0459, "step": 2035 }, { "epoch": 0.01, "learning_rate": 9.973904071669435e-05, "loss": 1.0462, "step": 2040 }, { "epoch": 0.01, "learning_rate": 9.973778247521265e-05, "loss": 1.0482, "step": 2045 }, { "epoch": 0.01, "learning_rate": 9.973652423373094e-05, "loss": 1.0485, "step": 2050 }, { "epoch": 0.01, "learning_rate": 9.973526599224924e-05, "loss": 1.0476, "step": 2055 }, { "epoch": 0.01, "learning_rate": 9.973400775076753e-05, "loss": 1.072, "step": 2060 }, { "epoch": 0.01, "learning_rate": 9.973274950928583e-05, "loss": 1.0465, "step": 2065 }, { "epoch": 0.01, "learning_rate": 9.973149126780412e-05, "loss": 1.0465, "step": 2070 }, { "epoch": 0.01, "learning_rate": 9.973023302632242e-05, "loss": 1.0482, "step": 2075 }, { "epoch": 0.01, "learning_rate": 9.972897478484071e-05, "loss": 1.0486, "step": 2080 }, { "epoch": 0.01, "learning_rate": 9.9727716543359e-05, "loss": 1.0456, "step": 2085 }, { "epoch": 0.01, "learning_rate": 9.97264583018773e-05, "loss": 1.0466, "step": 2090 }, { "epoch": 0.01, "learning_rate": 9.97252000603956e-05, "loss": 1.0455, "step": 2095 }, { "epoch": 0.01, "learning_rate": 9.97239418189139e-05, "loss": 1.0468, "step": 2100 }, { "epoch": 0.01, "learning_rate": 9.972268357743219e-05, "loss": 1.0475, "step": 2105 }, { "epoch": 0.01, "learning_rate": 9.972142533595048e-05, "loss": 1.0461, "step": 2110 }, { "epoch": 0.01, "learning_rate": 9.972016709446877e-05, "loss": 1.0458, "step": 2115 }, { "epoch": 0.01, "learning_rate": 9.971890885298707e-05, "loss": 1.0481, "step": 2120 }, { "epoch": 0.01, "learning_rate": 9.971765061150537e-05, "loss": 1.0462, "step": 2125 }, { "epoch": 0.01, "learning_rate": 9.971639237002366e-05, "loss": 1.0485, "step": 2130 }, { "epoch": 0.01, "learning_rate": 9.971513412854195e-05, "loss": 1.0443, "step": 2135 }, { "epoch": 0.01, "learning_rate": 9.971387588706025e-05, "loss": 1.0431, "step": 2140 }, { "epoch": 0.01, "learning_rate": 9.971261764557855e-05, "loss": 1.0481, "step": 2145 }, { "epoch": 0.01, "learning_rate": 9.971135940409684e-05, "loss": 1.0466, "step": 2150 }, { "epoch": 0.01, "learning_rate": 9.971010116261513e-05, "loss": 1.0456, "step": 2155 }, { "epoch": 0.01, "learning_rate": 9.970884292113343e-05, "loss": 1.046, "step": 2160 }, { "epoch": 0.01, "learning_rate": 9.970758467965173e-05, "loss": 1.0465, "step": 2165 }, { "epoch": 0.01, "learning_rate": 9.970632643817002e-05, "loss": 1.0475, "step": 2170 }, { "epoch": 0.01, "learning_rate": 9.970506819668831e-05, "loss": 1.0479, "step": 2175 }, { "epoch": 0.01, "learning_rate": 9.97038099552066e-05, "loss": 1.0457, "step": 2180 }, { "epoch": 0.01, "learning_rate": 9.97025517137249e-05, "loss": 1.0466, "step": 2185 }, { "epoch": 0.01, "learning_rate": 9.97012934722432e-05, "loss": 1.0438, "step": 2190 }, { "epoch": 0.01, "learning_rate": 9.970003523076149e-05, "loss": 1.0475, "step": 2195 }, { "epoch": 0.01, "learning_rate": 9.969877698927978e-05, "loss": 1.0442, "step": 2200 }, { "epoch": 0.01, "learning_rate": 9.969751874779809e-05, "loss": 1.0468, "step": 2205 }, { "epoch": 0.01, "learning_rate": 9.969626050631638e-05, "loss": 1.0457, "step": 2210 }, { "epoch": 0.01, "learning_rate": 9.969500226483467e-05, "loss": 1.048, "step": 2215 }, { "epoch": 0.01, "learning_rate": 9.969374402335296e-05, "loss": 1.0469, "step": 2220 }, { "epoch": 0.01, "learning_rate": 9.969248578187127e-05, "loss": 1.0455, "step": 2225 }, { "epoch": 0.01, "learning_rate": 9.969122754038956e-05, "loss": 1.0471, "step": 2230 }, { "epoch": 0.01, "learning_rate": 9.968996929890785e-05, "loss": 1.0465, "step": 2235 }, { "epoch": 0.01, "learning_rate": 9.968871105742614e-05, "loss": 1.0433, "step": 2240 }, { "epoch": 0.01, "learning_rate": 9.968745281594443e-05, "loss": 1.0445, "step": 2245 }, { "epoch": 0.01, "learning_rate": 9.968619457446274e-05, "loss": 1.0471, "step": 2250 }, { "epoch": 0.01, "learning_rate": 9.968493633298103e-05, "loss": 1.0451, "step": 2255 }, { "epoch": 0.01, "learning_rate": 9.968367809149932e-05, "loss": 1.0471, "step": 2260 }, { "epoch": 0.01, "learning_rate": 9.968241985001761e-05, "loss": 1.0464, "step": 2265 }, { "epoch": 0.01, "learning_rate": 9.968116160853592e-05, "loss": 1.046, "step": 2270 }, { "epoch": 0.01, "learning_rate": 9.967990336705421e-05, "loss": 1.0481, "step": 2275 }, { "epoch": 0.01, "learning_rate": 9.96786451255725e-05, "loss": 1.0434, "step": 2280 }, { "epoch": 0.01, "learning_rate": 9.967738688409079e-05, "loss": 1.0462, "step": 2285 }, { "epoch": 0.01, "learning_rate": 9.96761286426091e-05, "loss": 1.0441, "step": 2290 }, { "epoch": 0.01, "learning_rate": 9.967487040112739e-05, "loss": 1.0457, "step": 2295 }, { "epoch": 0.01, "learning_rate": 9.967361215964568e-05, "loss": 1.0677, "step": 2300 }, { "epoch": 0.01, "learning_rate": 9.967235391816397e-05, "loss": 1.0468, "step": 2305 }, { "epoch": 0.01, "learning_rate": 9.967109567668226e-05, "loss": 1.0436, "step": 2310 }, { "epoch": 0.01, "learning_rate": 9.966983743520057e-05, "loss": 1.045, "step": 2315 }, { "epoch": 0.01, "learning_rate": 9.966857919371886e-05, "loss": 1.0477, "step": 2320 }, { "epoch": 0.01, "learning_rate": 9.966732095223715e-05, "loss": 1.0461, "step": 2325 }, { "epoch": 0.01, "learning_rate": 9.966606271075544e-05, "loss": 1.0453, "step": 2330 }, { "epoch": 0.01, "learning_rate": 9.966480446927375e-05, "loss": 1.0452, "step": 2335 }, { "epoch": 0.01, "learning_rate": 9.966354622779204e-05, "loss": 1.0467, "step": 2340 }, { "epoch": 0.01, "learning_rate": 9.966228798631033e-05, "loss": 1.0432, "step": 2345 }, { "epoch": 0.01, "learning_rate": 9.966102974482862e-05, "loss": 1.0459, "step": 2350 }, { "epoch": 0.01, "learning_rate": 9.965977150334693e-05, "loss": 1.0468, "step": 2355 }, { "epoch": 0.01, "learning_rate": 9.965851326186523e-05, "loss": 1.0439, "step": 2360 }, { "epoch": 0.01, "learning_rate": 9.965725502038352e-05, "loss": 1.045, "step": 2365 }, { "epoch": 0.01, "learning_rate": 9.965599677890182e-05, "loss": 1.0444, "step": 2370 }, { "epoch": 0.01, "learning_rate": 9.965473853742011e-05, "loss": 1.0443, "step": 2375 }, { "epoch": 0.01, "learning_rate": 9.96534802959384e-05, "loss": 1.0448, "step": 2380 }, { "epoch": 0.01, "learning_rate": 9.96522220544567e-05, "loss": 1.044, "step": 2385 }, { "epoch": 0.01, "learning_rate": 9.9650963812975e-05, "loss": 1.0471, "step": 2390 }, { "epoch": 0.01, "learning_rate": 9.964970557149329e-05, "loss": 1.0436, "step": 2395 }, { "epoch": 0.01, "learning_rate": 9.964844733001158e-05, "loss": 1.0419, "step": 2400 }, { "epoch": 0.01, "learning_rate": 9.964718908852988e-05, "loss": 1.0433, "step": 2405 }, { "epoch": 0.01, "learning_rate": 9.964593084704818e-05, "loss": 1.0453, "step": 2410 }, { "epoch": 0.01, "learning_rate": 9.964467260556647e-05, "loss": 1.0452, "step": 2415 }, { "epoch": 0.01, "learning_rate": 9.964341436408476e-05, "loss": 1.0449, "step": 2420 }, { "epoch": 0.01, "learning_rate": 9.964215612260306e-05, "loss": 1.0465, "step": 2425 }, { "epoch": 0.01, "learning_rate": 9.964089788112136e-05, "loss": 1.0425, "step": 2430 }, { "epoch": 0.01, "learning_rate": 9.963963963963965e-05, "loss": 1.0439, "step": 2435 }, { "epoch": 0.01, "learning_rate": 9.963838139815794e-05, "loss": 1.0447, "step": 2440 }, { "epoch": 0.01, "learning_rate": 9.963712315667623e-05, "loss": 1.0435, "step": 2445 }, { "epoch": 0.01, "learning_rate": 9.963586491519454e-05, "loss": 1.0444, "step": 2450 }, { "epoch": 0.01, "learning_rate": 9.963460667371283e-05, "loss": 1.0459, "step": 2455 }, { "epoch": 0.01, "learning_rate": 9.963334843223112e-05, "loss": 1.044, "step": 2460 }, { "epoch": 0.01, "learning_rate": 9.963209019074941e-05, "loss": 1.044, "step": 2465 }, { "epoch": 0.01, "learning_rate": 9.963083194926772e-05, "loss": 1.0443, "step": 2470 }, { "epoch": 0.01, "learning_rate": 9.962957370778601e-05, "loss": 1.0451, "step": 2475 }, { "epoch": 0.01, "learning_rate": 9.96283154663043e-05, "loss": 1.0437, "step": 2480 }, { "epoch": 0.01, "learning_rate": 9.962705722482259e-05, "loss": 1.0412, "step": 2485 }, { "epoch": 0.01, "learning_rate": 9.96257989833409e-05, "loss": 1.0441, "step": 2490 }, { "epoch": 0.01, "learning_rate": 9.962454074185919e-05, "loss": 1.0429, "step": 2495 }, { "epoch": 0.01, "learning_rate": 9.962328250037748e-05, "loss": 1.0442, "step": 2500 }, { "epoch": 0.01, "learning_rate": 9.962202425889577e-05, "loss": 1.0441, "step": 2505 }, { "epoch": 0.01, "learning_rate": 9.962076601741406e-05, "loss": 1.042, "step": 2510 }, { "epoch": 0.01, "learning_rate": 9.961950777593237e-05, "loss": 1.0424, "step": 2515 }, { "epoch": 0.01, "learning_rate": 9.961824953445066e-05, "loss": 1.0431, "step": 2520 }, { "epoch": 0.01, "learning_rate": 9.961699129296895e-05, "loss": 1.0452, "step": 2525 }, { "epoch": 0.01, "learning_rate": 9.961573305148724e-05, "loss": 1.0438, "step": 2530 }, { "epoch": 0.01, "learning_rate": 9.961447481000555e-05, "loss": 1.0448, "step": 2535 }, { "epoch": 0.01, "learning_rate": 9.961321656852384e-05, "loss": 1.0438, "step": 2540 }, { "epoch": 0.01, "learning_rate": 9.961195832704213e-05, "loss": 1.0672, "step": 2545 }, { "epoch": 0.01, "learning_rate": 9.961070008556042e-05, "loss": 1.0446, "step": 2550 }, { "epoch": 0.01, "learning_rate": 9.960944184407873e-05, "loss": 1.0429, "step": 2555 }, { "epoch": 0.01, "learning_rate": 9.960818360259702e-05, "loss": 1.0453, "step": 2560 }, { "epoch": 0.01, "learning_rate": 9.960692536111531e-05, "loss": 1.0422, "step": 2565 }, { "epoch": 0.01, "learning_rate": 9.96056671196336e-05, "loss": 1.0437, "step": 2570 }, { "epoch": 0.01, "learning_rate": 9.96044088781519e-05, "loss": 1.0454, "step": 2575 }, { "epoch": 0.01, "learning_rate": 9.96031506366702e-05, "loss": 1.0438, "step": 2580 }, { "epoch": 0.01, "learning_rate": 9.960189239518849e-05, "loss": 1.0435, "step": 2585 }, { "epoch": 0.01, "learning_rate": 9.960063415370678e-05, "loss": 1.0648, "step": 2590 }, { "epoch": 0.01, "learning_rate": 9.959937591222507e-05, "loss": 1.0439, "step": 2595 }, { "epoch": 0.01, "learning_rate": 9.959811767074338e-05, "loss": 1.0434, "step": 2600 }, { "epoch": 0.01, "learning_rate": 9.959685942926167e-05, "loss": 1.0457, "step": 2605 }, { "epoch": 0.01, "learning_rate": 9.959560118777996e-05, "loss": 1.0417, "step": 2610 }, { "epoch": 0.01, "learning_rate": 9.959434294629825e-05, "loss": 1.0431, "step": 2615 }, { "epoch": 0.01, "learning_rate": 9.959308470481656e-05, "loss": 1.0451, "step": 2620 }, { "epoch": 0.01, "learning_rate": 9.959182646333485e-05, "loss": 1.0435, "step": 2625 }, { "epoch": 0.01, "learning_rate": 9.959056822185314e-05, "loss": 1.044, "step": 2630 }, { "epoch": 0.01, "learning_rate": 9.958930998037143e-05, "loss": 1.0429, "step": 2635 }, { "epoch": 0.01, "learning_rate": 9.958805173888973e-05, "loss": 1.0449, "step": 2640 }, { "epoch": 0.01, "learning_rate": 9.958679349740803e-05, "loss": 1.0411, "step": 2645 }, { "epoch": 0.01, "learning_rate": 9.958553525592632e-05, "loss": 1.0449, "step": 2650 }, { "epoch": 0.01, "learning_rate": 9.958427701444461e-05, "loss": 1.0428, "step": 2655 }, { "epoch": 0.01, "learning_rate": 9.95830187729629e-05, "loss": 1.0457, "step": 2660 }, { "epoch": 0.01, "learning_rate": 9.958176053148121e-05, "loss": 1.0438, "step": 2665 }, { "epoch": 0.01, "learning_rate": 9.95805022899995e-05, "loss": 1.0447, "step": 2670 }, { "epoch": 0.01, "learning_rate": 9.95792440485178e-05, "loss": 1.043, "step": 2675 }, { "epoch": 0.01, "learning_rate": 9.957798580703609e-05, "loss": 1.0427, "step": 2680 }, { "epoch": 0.01, "learning_rate": 9.957672756555439e-05, "loss": 1.0442, "step": 2685 }, { "epoch": 0.01, "learning_rate": 9.957546932407268e-05, "loss": 1.0452, "step": 2690 }, { "epoch": 0.01, "learning_rate": 9.957421108259097e-05, "loss": 1.0426, "step": 2695 }, { "epoch": 0.01, "learning_rate": 9.957295284110927e-05, "loss": 1.0444, "step": 2700 }, { "epoch": 0.01, "learning_rate": 9.957169459962756e-05, "loss": 1.0436, "step": 2705 }, { "epoch": 0.01, "learning_rate": 9.957043635814586e-05, "loss": 1.0426, "step": 2710 }, { "epoch": 0.01, "learning_rate": 9.956917811666415e-05, "loss": 1.0452, "step": 2715 }, { "epoch": 0.01, "learning_rate": 9.956791987518245e-05, "loss": 1.0429, "step": 2720 }, { "epoch": 0.01, "learning_rate": 9.956666163370074e-05, "loss": 1.046, "step": 2725 }, { "epoch": 0.01, "learning_rate": 9.956540339221904e-05, "loss": 1.0437, "step": 2730 }, { "epoch": 0.01, "learning_rate": 9.956414515073733e-05, "loss": 1.0445, "step": 2735 }, { "epoch": 0.01, "learning_rate": 9.956288690925563e-05, "loss": 1.0413, "step": 2740 }, { "epoch": 0.01, "learning_rate": 9.956162866777392e-05, "loss": 1.0445, "step": 2745 }, { "epoch": 0.01, "learning_rate": 9.956037042629222e-05, "loss": 1.0443, "step": 2750 }, { "epoch": 0.01, "learning_rate": 9.955911218481051e-05, "loss": 1.0437, "step": 2755 }, { "epoch": 0.01, "learning_rate": 9.95578539433288e-05, "loss": 1.0436, "step": 2760 }, { "epoch": 0.01, "learning_rate": 9.95565957018471e-05, "loss": 1.0433, "step": 2765 }, { "epoch": 0.01, "learning_rate": 9.955533746036539e-05, "loss": 1.0427, "step": 2770 }, { "epoch": 0.01, "learning_rate": 9.95540792188837e-05, "loss": 1.042, "step": 2775 }, { "epoch": 0.01, "learning_rate": 9.955282097740199e-05, "loss": 1.0449, "step": 2780 }, { "epoch": 0.01, "learning_rate": 9.955156273592028e-05, "loss": 1.0425, "step": 2785 }, { "epoch": 0.01, "learning_rate": 9.955030449443857e-05, "loss": 1.0435, "step": 2790 }, { "epoch": 0.01, "learning_rate": 9.954904625295687e-05, "loss": 1.0427, "step": 2795 }, { "epoch": 0.01, "learning_rate": 9.954778801147517e-05, "loss": 1.0424, "step": 2800 }, { "epoch": 0.01, "learning_rate": 9.954652976999346e-05, "loss": 1.0429, "step": 2805 }, { "epoch": 0.01, "learning_rate": 9.954527152851175e-05, "loss": 1.0443, "step": 2810 }, { "epoch": 0.01, "learning_rate": 9.954401328703005e-05, "loss": 1.044, "step": 2815 }, { "epoch": 0.01, "learning_rate": 9.954275504554834e-05, "loss": 1.0432, "step": 2820 }, { "epoch": 0.01, "learning_rate": 9.954149680406664e-05, "loss": 1.0463, "step": 2825 }, { "epoch": 0.01, "learning_rate": 9.954023856258493e-05, "loss": 1.0466, "step": 2830 }, { "epoch": 0.01, "learning_rate": 9.953898032110322e-05, "loss": 1.0397, "step": 2835 }, { "epoch": 0.01, "learning_rate": 9.953772207962152e-05, "loss": 1.0436, "step": 2840 }, { "epoch": 0.01, "learning_rate": 9.953646383813982e-05, "loss": 1.0422, "step": 2845 }, { "epoch": 0.01, "learning_rate": 9.953520559665811e-05, "loss": 1.0433, "step": 2850 }, { "epoch": 0.01, "learning_rate": 9.953394735517641e-05, "loss": 1.0423, "step": 2855 }, { "epoch": 0.01, "learning_rate": 9.95326891136947e-05, "loss": 1.0435, "step": 2860 }, { "epoch": 0.01, "learning_rate": 9.953143087221301e-05, "loss": 1.0465, "step": 2865 }, { "epoch": 0.01, "learning_rate": 9.95301726307313e-05, "loss": 1.045, "step": 2870 }, { "epoch": 0.01, "learning_rate": 9.952891438924959e-05, "loss": 1.0445, "step": 2875 }, { "epoch": 0.01, "learning_rate": 9.952765614776788e-05, "loss": 1.0431, "step": 2880 }, { "epoch": 0.01, "learning_rate": 9.952639790628619e-05, "loss": 1.066, "step": 2885 }, { "epoch": 0.01, "learning_rate": 9.952513966480448e-05, "loss": 1.0439, "step": 2890 }, { "epoch": 0.01, "learning_rate": 9.952388142332277e-05, "loss": 1.0436, "step": 2895 }, { "epoch": 0.01, "learning_rate": 9.952262318184106e-05, "loss": 1.0431, "step": 2900 }, { "epoch": 0.01, "learning_rate": 9.952136494035936e-05, "loss": 1.0434, "step": 2905 }, { "epoch": 0.01, "learning_rate": 9.952010669887766e-05, "loss": 1.0436, "step": 2910 }, { "epoch": 0.01, "learning_rate": 9.951884845739595e-05, "loss": 1.0638, "step": 2915 }, { "epoch": 0.01, "learning_rate": 9.951759021591424e-05, "loss": 1.0429, "step": 2920 }, { "epoch": 0.01, "learning_rate": 9.951633197443254e-05, "loss": 1.0421, "step": 2925 }, { "epoch": 0.01, "learning_rate": 9.951507373295084e-05, "loss": 1.0424, "step": 2930 }, { "epoch": 0.01, "learning_rate": 9.951381549146913e-05, "loss": 1.0434, "step": 2935 }, { "epoch": 0.01, "learning_rate": 9.951255724998742e-05, "loss": 1.0441, "step": 2940 }, { "epoch": 0.01, "learning_rate": 9.951129900850572e-05, "loss": 1.0429, "step": 2945 }, { "epoch": 0.01, "learning_rate": 9.951004076702402e-05, "loss": 1.0463, "step": 2950 }, { "epoch": 0.01, "learning_rate": 9.950878252554231e-05, "loss": 1.0424, "step": 2955 }, { "epoch": 0.01, "learning_rate": 9.95075242840606e-05, "loss": 1.0451, "step": 2960 }, { "epoch": 0.01, "learning_rate": 9.95062660425789e-05, "loss": 1.0424, "step": 2965 }, { "epoch": 0.01, "learning_rate": 9.950500780109719e-05, "loss": 1.0439, "step": 2970 }, { "epoch": 0.01, "learning_rate": 9.950374955961549e-05, "loss": 1.0425, "step": 2975 }, { "epoch": 0.01, "learning_rate": 9.950249131813378e-05, "loss": 1.0443, "step": 2980 }, { "epoch": 0.01, "learning_rate": 9.950123307665208e-05, "loss": 1.0432, "step": 2985 }, { "epoch": 0.01, "learning_rate": 9.949997483517037e-05, "loss": 1.0422, "step": 2990 }, { "epoch": 0.01, "learning_rate": 9.949871659368867e-05, "loss": 1.0434, "step": 2995 }, { "epoch": 0.01, "learning_rate": 9.949745835220696e-05, "loss": 1.043, "step": 3000 }, { "epoch": 0.01, "learning_rate": 9.949620011072526e-05, "loss": 1.044, "step": 3005 }, { "epoch": 0.01, "learning_rate": 9.949494186924355e-05, "loss": 1.0422, "step": 3010 }, { "epoch": 0.01, "learning_rate": 9.949368362776185e-05, "loss": 1.042, "step": 3015 }, { "epoch": 0.01, "learning_rate": 9.949242538628014e-05, "loss": 1.0417, "step": 3020 }, { "epoch": 0.01, "learning_rate": 9.949116714479844e-05, "loss": 1.0415, "step": 3025 }, { "epoch": 0.01, "learning_rate": 9.948990890331673e-05, "loss": 1.0431, "step": 3030 }, { "epoch": 0.01, "learning_rate": 9.948865066183502e-05, "loss": 1.0429, "step": 3035 }, { "epoch": 0.01, "learning_rate": 9.948739242035332e-05, "loss": 1.042, "step": 3040 }, { "epoch": 0.01, "learning_rate": 9.948613417887162e-05, "loss": 1.0422, "step": 3045 }, { "epoch": 0.01, "learning_rate": 9.948487593738991e-05, "loss": 1.0427, "step": 3050 }, { "epoch": 0.01, "learning_rate": 9.94836176959082e-05, "loss": 1.0426, "step": 3055 }, { "epoch": 0.01, "learning_rate": 9.94823594544265e-05, "loss": 1.0421, "step": 3060 }, { "epoch": 0.01, "learning_rate": 9.94811012129448e-05, "loss": 1.0435, "step": 3065 }, { "epoch": 0.01, "learning_rate": 9.947984297146309e-05, "loss": 1.0433, "step": 3070 }, { "epoch": 0.01, "learning_rate": 9.947858472998138e-05, "loss": 1.0425, "step": 3075 }, { "epoch": 0.01, "learning_rate": 9.947732648849968e-05, "loss": 1.0432, "step": 3080 }, { "epoch": 0.01, "learning_rate": 9.947606824701798e-05, "loss": 1.0412, "step": 3085 }, { "epoch": 0.01, "learning_rate": 9.947481000553627e-05, "loss": 1.0436, "step": 3090 }, { "epoch": 0.01, "learning_rate": 9.947355176405456e-05, "loss": 1.0417, "step": 3095 }, { "epoch": 0.01, "learning_rate": 9.947229352257285e-05, "loss": 1.0426, "step": 3100 }, { "epoch": 0.01, "learning_rate": 9.947103528109116e-05, "loss": 1.0425, "step": 3105 }, { "epoch": 0.01, "learning_rate": 9.946977703960945e-05, "loss": 1.044, "step": 3110 }, { "epoch": 0.01, "learning_rate": 9.946851879812774e-05, "loss": 1.0414, "step": 3115 }, { "epoch": 0.01, "learning_rate": 9.946726055664603e-05, "loss": 1.0426, "step": 3120 }, { "epoch": 0.01, "learning_rate": 9.946600231516434e-05, "loss": 1.043, "step": 3125 }, { "epoch": 0.01, "learning_rate": 9.946474407368263e-05, "loss": 1.0419, "step": 3130 }, { "epoch": 0.01, "learning_rate": 9.946348583220092e-05, "loss": 1.0417, "step": 3135 }, { "epoch": 0.01, "learning_rate": 9.946222759071921e-05, "loss": 1.042, "step": 3140 }, { "epoch": 0.01, "learning_rate": 9.946096934923752e-05, "loss": 1.0427, "step": 3145 }, { "epoch": 0.01, "learning_rate": 9.945971110775581e-05, "loss": 1.0426, "step": 3150 }, { "epoch": 0.01, "learning_rate": 9.94584528662741e-05, "loss": 1.0434, "step": 3155 }, { "epoch": 0.01, "learning_rate": 9.945719462479239e-05, "loss": 1.0436, "step": 3160 }, { "epoch": 0.01, "learning_rate": 9.945593638331068e-05, "loss": 1.0421, "step": 3165 }, { "epoch": 0.01, "learning_rate": 9.945467814182899e-05, "loss": 1.0409, "step": 3170 }, { "epoch": 0.01, "learning_rate": 9.945341990034728e-05, "loss": 1.0421, "step": 3175 }, { "epoch": 0.01, "learning_rate": 9.945216165886557e-05, "loss": 1.0422, "step": 3180 }, { "epoch": 0.01, "learning_rate": 9.945090341738386e-05, "loss": 1.0423, "step": 3185 }, { "epoch": 0.01, "learning_rate": 9.944964517590217e-05, "loss": 1.0419, "step": 3190 }, { "epoch": 0.01, "learning_rate": 9.944838693442046e-05, "loss": 1.0443, "step": 3195 }, { "epoch": 0.01, "learning_rate": 9.944712869293875e-05, "loss": 1.0409, "step": 3200 }, { "epoch": 0.01, "learning_rate": 9.944587045145704e-05, "loss": 1.0433, "step": 3205 }, { "epoch": 0.01, "learning_rate": 9.944461220997535e-05, "loss": 1.044, "step": 3210 }, { "epoch": 0.01, "learning_rate": 9.944335396849364e-05, "loss": 1.0418, "step": 3215 }, { "epoch": 0.01, "learning_rate": 9.944209572701193e-05, "loss": 1.0434, "step": 3220 }, { "epoch": 0.01, "learning_rate": 9.944083748553022e-05, "loss": 1.0415, "step": 3225 }, { "epoch": 0.01, "learning_rate": 9.943957924404851e-05, "loss": 1.0413, "step": 3230 }, { "epoch": 0.01, "learning_rate": 9.943832100256682e-05, "loss": 1.0429, "step": 3235 }, { "epoch": 0.01, "learning_rate": 9.943706276108511e-05, "loss": 1.0418, "step": 3240 }, { "epoch": 0.01, "learning_rate": 9.94358045196034e-05, "loss": 1.0445, "step": 3245 }, { "epoch": 0.01, "learning_rate": 9.94345462781217e-05, "loss": 1.043, "step": 3250 }, { "epoch": 0.01, "learning_rate": 9.943328803664e-05, "loss": 1.0426, "step": 3255 }, { "epoch": 0.01, "learning_rate": 9.943202979515829e-05, "loss": 1.0411, "step": 3260 }, { "epoch": 0.01, "learning_rate": 9.943077155367658e-05, "loss": 1.0421, "step": 3265 }, { "epoch": 0.01, "learning_rate": 9.942951331219487e-05, "loss": 1.043, "step": 3270 }, { "epoch": 0.01, "learning_rate": 9.942825507071318e-05, "loss": 1.0422, "step": 3275 }, { "epoch": 0.01, "learning_rate": 9.942699682923147e-05, "loss": 1.042, "step": 3280 }, { "epoch": 0.01, "learning_rate": 9.942573858774976e-05, "loss": 1.0425, "step": 3285 }, { "epoch": 0.01, "learning_rate": 9.942448034626805e-05, "loss": 1.0411, "step": 3290 }, { "epoch": 0.01, "learning_rate": 9.942322210478634e-05, "loss": 1.0408, "step": 3295 }, { "epoch": 0.01, "learning_rate": 9.942196386330465e-05, "loss": 1.0406, "step": 3300 }, { "epoch": 0.01, "learning_rate": 9.942070562182294e-05, "loss": 1.044, "step": 3305 }, { "epoch": 0.01, "learning_rate": 9.941944738034123e-05, "loss": 1.0406, "step": 3310 }, { "epoch": 0.01, "learning_rate": 9.941818913885952e-05, "loss": 1.043, "step": 3315 }, { "epoch": 0.01, "learning_rate": 9.941693089737783e-05, "loss": 1.0418, "step": 3320 }, { "epoch": 0.01, "learning_rate": 9.941567265589612e-05, "loss": 1.0415, "step": 3325 }, { "epoch": 0.01, "learning_rate": 9.941441441441441e-05, "loss": 1.0428, "step": 3330 }, { "epoch": 0.01, "learning_rate": 9.94131561729327e-05, "loss": 1.0433, "step": 3335 }, { "epoch": 0.01, "learning_rate": 9.941189793145101e-05, "loss": 1.0424, "step": 3340 }, { "epoch": 0.01, "learning_rate": 9.94106396899693e-05, "loss": 1.0415, "step": 3345 }, { "epoch": 0.01, "learning_rate": 9.94093814484876e-05, "loss": 1.0431, "step": 3350 }, { "epoch": 0.01, "learning_rate": 9.94081232070059e-05, "loss": 1.0423, "step": 3355 }, { "epoch": 0.01, "learning_rate": 9.940686496552419e-05, "loss": 1.0421, "step": 3360 }, { "epoch": 0.01, "learning_rate": 9.940560672404248e-05, "loss": 1.0427, "step": 3365 }, { "epoch": 0.01, "learning_rate": 9.940434848256079e-05, "loss": 1.0447, "step": 3370 }, { "epoch": 0.01, "learning_rate": 9.940309024107908e-05, "loss": 1.0433, "step": 3375 }, { "epoch": 0.01, "learning_rate": 9.940183199959737e-05, "loss": 1.0423, "step": 3380 }, { "epoch": 0.01, "learning_rate": 9.940057375811566e-05, "loss": 1.0432, "step": 3385 }, { "epoch": 0.01, "learning_rate": 9.939931551663397e-05, "loss": 1.0411, "step": 3390 }, { "epoch": 0.01, "learning_rate": 9.939805727515226e-05, "loss": 1.0413, "step": 3395 }, { "epoch": 0.01, "learning_rate": 9.939679903367055e-05, "loss": 1.0415, "step": 3400 }, { "epoch": 0.01, "learning_rate": 9.939554079218884e-05, "loss": 1.0394, "step": 3405 }, { "epoch": 0.01, "learning_rate": 9.939428255070715e-05, "loss": 1.0411, "step": 3410 }, { "epoch": 0.01, "learning_rate": 9.939302430922544e-05, "loss": 1.0412, "step": 3415 }, { "epoch": 0.01, "learning_rate": 9.939176606774373e-05, "loss": 1.0409, "step": 3420 }, { "epoch": 0.01, "learning_rate": 9.939050782626202e-05, "loss": 1.0439, "step": 3425 }, { "epoch": 0.01, "learning_rate": 9.938924958478031e-05, "loss": 1.0411, "step": 3430 }, { "epoch": 0.01, "learning_rate": 9.938799134329862e-05, "loss": 1.0412, "step": 3435 }, { "epoch": 0.01, "learning_rate": 9.938673310181691e-05, "loss": 1.0454, "step": 3440 }, { "epoch": 0.01, "learning_rate": 9.93854748603352e-05, "loss": 1.0402, "step": 3445 }, { "epoch": 0.01, "learning_rate": 9.938421661885349e-05, "loss": 1.0421, "step": 3450 }, { "epoch": 0.01, "learning_rate": 9.93829583773718e-05, "loss": 1.0434, "step": 3455 }, { "epoch": 0.01, "learning_rate": 9.938170013589009e-05, "loss": 1.0429, "step": 3460 }, { "epoch": 0.01, "learning_rate": 9.938044189440838e-05, "loss": 1.042, "step": 3465 }, { "epoch": 0.01, "learning_rate": 9.937918365292667e-05, "loss": 1.0415, "step": 3470 }, { "epoch": 0.01, "learning_rate": 9.937792541144498e-05, "loss": 1.0427, "step": 3475 }, { "epoch": 0.01, "learning_rate": 9.937666716996327e-05, "loss": 1.0445, "step": 3480 }, { "epoch": 0.01, "learning_rate": 9.937540892848156e-05, "loss": 1.0419, "step": 3485 }, { "epoch": 0.01, "learning_rate": 9.937415068699985e-05, "loss": 1.0402, "step": 3490 }, { "epoch": 0.01, "learning_rate": 9.937289244551814e-05, "loss": 1.0421, "step": 3495 }, { "epoch": 0.01, "learning_rate": 9.937163420403645e-05, "loss": 1.0434, "step": 3500 }, { "epoch": 0.01, "learning_rate": 9.937037596255474e-05, "loss": 1.0412, "step": 3505 }, { "epoch": 0.01, "learning_rate": 9.936911772107303e-05, "loss": 1.042, "step": 3510 }, { "epoch": 0.01, "learning_rate": 9.936785947959132e-05, "loss": 1.06, "step": 3515 }, { "epoch": 0.01, "learning_rate": 9.936660123810963e-05, "loss": 1.0416, "step": 3520 }, { "epoch": 0.01, "learning_rate": 9.936534299662792e-05, "loss": 1.043, "step": 3525 }, { "epoch": 0.01, "learning_rate": 9.936408475514621e-05, "loss": 1.0584, "step": 3530 }, { "epoch": 0.01, "learning_rate": 9.93628265136645e-05, "loss": 1.0422, "step": 3535 }, { "epoch": 0.01, "learning_rate": 9.936156827218281e-05, "loss": 1.0432, "step": 3540 }, { "epoch": 0.01, "learning_rate": 9.93603100307011e-05, "loss": 1.0419, "step": 3545 }, { "epoch": 0.01, "learning_rate": 9.935905178921939e-05, "loss": 1.041, "step": 3550 }, { "epoch": 0.01, "learning_rate": 9.935779354773768e-05, "loss": 1.0417, "step": 3555 }, { "epoch": 0.01, "learning_rate": 9.935653530625598e-05, "loss": 1.0405, "step": 3560 }, { "epoch": 0.01, "learning_rate": 9.935527706477428e-05, "loss": 1.0413, "step": 3565 }, { "epoch": 0.01, "learning_rate": 9.935401882329257e-05, "loss": 1.0433, "step": 3570 }, { "epoch": 0.01, "learning_rate": 9.935276058181086e-05, "loss": 1.0414, "step": 3575 }, { "epoch": 0.01, "learning_rate": 9.935150234032916e-05, "loss": 1.0425, "step": 3580 }, { "epoch": 0.01, "learning_rate": 9.935024409884746e-05, "loss": 1.042, "step": 3585 }, { "epoch": 0.01, "learning_rate": 9.934898585736575e-05, "loss": 1.0432, "step": 3590 }, { "epoch": 0.01, "learning_rate": 9.934772761588404e-05, "loss": 1.0424, "step": 3595 }, { "epoch": 0.01, "learning_rate": 9.934646937440234e-05, "loss": 1.0413, "step": 3600 }, { "epoch": 0.01, "learning_rate": 9.934521113292064e-05, "loss": 1.0397, "step": 3605 }, { "epoch": 0.01, "learning_rate": 9.934395289143893e-05, "loss": 1.0426, "step": 3610 }, { "epoch": 0.01, "learning_rate": 9.934269464995722e-05, "loss": 1.0418, "step": 3615 }, { "epoch": 0.01, "learning_rate": 9.934143640847552e-05, "loss": 1.0424, "step": 3620 }, { "epoch": 0.01, "learning_rate": 9.934017816699381e-05, "loss": 1.0416, "step": 3625 }, { "epoch": 0.01, "learning_rate": 9.933891992551211e-05, "loss": 1.0414, "step": 3630 }, { "epoch": 0.01, "learning_rate": 9.93376616840304e-05, "loss": 1.0412, "step": 3635 }, { "epoch": 0.01, "learning_rate": 9.93364034425487e-05, "loss": 1.0441, "step": 3640 }, { "epoch": 0.01, "learning_rate": 9.933514520106699e-05, "loss": 1.0419, "step": 3645 }, { "epoch": 0.01, "learning_rate": 9.933388695958529e-05, "loss": 1.0412, "step": 3650 }, { "epoch": 0.01, "learning_rate": 9.933262871810358e-05, "loss": 1.0426, "step": 3655 }, { "epoch": 0.01, "learning_rate": 9.933137047662188e-05, "loss": 1.0435, "step": 3660 }, { "epoch": 0.01, "learning_rate": 9.933011223514017e-05, "loss": 1.0422, "step": 3665 }, { "epoch": 0.01, "learning_rate": 9.932885399365847e-05, "loss": 1.0396, "step": 3670 }, { "epoch": 0.01, "learning_rate": 9.932759575217676e-05, "loss": 1.0398, "step": 3675 }, { "epoch": 0.01, "learning_rate": 9.932633751069506e-05, "loss": 1.0615, "step": 3680 }, { "epoch": 0.01, "learning_rate": 9.932507926921335e-05, "loss": 1.0421, "step": 3685 }, { "epoch": 0.01, "learning_rate": 9.932382102773164e-05, "loss": 1.0444, "step": 3690 }, { "epoch": 0.01, "learning_rate": 9.932256278624994e-05, "loss": 1.044, "step": 3695 }, { "epoch": 0.01, "learning_rate": 9.932130454476824e-05, "loss": 1.0419, "step": 3700 }, { "epoch": 0.01, "learning_rate": 9.932004630328653e-05, "loss": 1.0402, "step": 3705 }, { "epoch": 0.01, "learning_rate": 9.931878806180482e-05, "loss": 1.0407, "step": 3710 }, { "epoch": 0.01, "learning_rate": 9.931752982032312e-05, "loss": 1.0415, "step": 3715 }, { "epoch": 0.01, "learning_rate": 9.931627157884142e-05, "loss": 1.0407, "step": 3720 }, { "epoch": 0.01, "learning_rate": 9.931501333735971e-05, "loss": 1.0417, "step": 3725 }, { "epoch": 0.01, "learning_rate": 9.9313755095878e-05, "loss": 1.0434, "step": 3730 }, { "epoch": 0.01, "learning_rate": 9.93124968543963e-05, "loss": 1.0432, "step": 3735 }, { "epoch": 0.01, "learning_rate": 9.93112386129146e-05, "loss": 1.0416, "step": 3740 }, { "epoch": 0.01, "learning_rate": 9.930998037143289e-05, "loss": 1.0431, "step": 3745 }, { "epoch": 0.01, "learning_rate": 9.930872212995118e-05, "loss": 1.0411, "step": 3750 }, { "epoch": 0.01, "learning_rate": 9.930746388846947e-05, "loss": 1.0402, "step": 3755 }, { "epoch": 0.01, "learning_rate": 9.930620564698778e-05, "loss": 1.0445, "step": 3760 }, { "epoch": 0.01, "learning_rate": 9.930494740550607e-05, "loss": 1.0406, "step": 3765 }, { "epoch": 0.01, "learning_rate": 9.930368916402436e-05, "loss": 1.0402, "step": 3770 }, { "epoch": 0.01, "learning_rate": 9.930243092254265e-05, "loss": 1.0415, "step": 3775 }, { "epoch": 0.01, "learning_rate": 9.930117268106096e-05, "loss": 1.0394, "step": 3780 }, { "epoch": 0.01, "learning_rate": 9.929991443957925e-05, "loss": 1.041, "step": 3785 }, { "epoch": 0.01, "learning_rate": 9.929865619809754e-05, "loss": 1.0419, "step": 3790 }, { "epoch": 0.01, "learning_rate": 9.929739795661583e-05, "loss": 1.0432, "step": 3795 }, { "epoch": 0.01, "learning_rate": 9.929613971513414e-05, "loss": 1.0415, "step": 3800 }, { "epoch": 0.01, "learning_rate": 9.929488147365243e-05, "loss": 1.0407, "step": 3805 }, { "epoch": 0.01, "learning_rate": 9.929362323217072e-05, "loss": 1.0444, "step": 3810 }, { "epoch": 0.01, "learning_rate": 9.929236499068901e-05, "loss": 1.0439, "step": 3815 }, { "epoch": 0.01, "learning_rate": 9.92911067492073e-05, "loss": 1.042, "step": 3820 }, { "epoch": 0.01, "learning_rate": 9.92898485077256e-05, "loss": 1.0424, "step": 3825 }, { "epoch": 0.01, "learning_rate": 9.92885902662439e-05, "loss": 1.0417, "step": 3830 }, { "epoch": 0.01, "learning_rate": 9.928733202476219e-05, "loss": 1.0407, "step": 3835 }, { "epoch": 0.01, "learning_rate": 9.928607378328048e-05, "loss": 1.0429, "step": 3840 }, { "epoch": 0.01, "learning_rate": 9.928481554179879e-05, "loss": 1.0411, "step": 3845 }, { "epoch": 0.01, "learning_rate": 9.928355730031709e-05, "loss": 1.0432, "step": 3850 }, { "epoch": 0.01, "learning_rate": 9.928229905883538e-05, "loss": 1.0422, "step": 3855 }, { "epoch": 0.01, "learning_rate": 9.928104081735367e-05, "loss": 1.0401, "step": 3860 }, { "epoch": 0.01, "learning_rate": 9.927978257587197e-05, "loss": 1.0417, "step": 3865 }, { "epoch": 0.01, "learning_rate": 9.927852433439027e-05, "loss": 1.043, "step": 3870 }, { "epoch": 0.01, "learning_rate": 9.927726609290856e-05, "loss": 1.0432, "step": 3875 }, { "epoch": 0.01, "learning_rate": 9.927600785142685e-05, "loss": 1.0423, "step": 3880 }, { "epoch": 0.01, "learning_rate": 9.927474960994515e-05, "loss": 1.0432, "step": 3885 }, { "epoch": 0.01, "learning_rate": 9.927349136846345e-05, "loss": 1.0408, "step": 3890 }, { "epoch": 0.01, "learning_rate": 9.927223312698174e-05, "loss": 1.0411, "step": 3895 }, { "epoch": 0.01, "learning_rate": 9.927097488550003e-05, "loss": 1.0399, "step": 3900 }, { "epoch": 0.01, "learning_rate": 9.926971664401833e-05, "loss": 1.0426, "step": 3905 }, { "epoch": 0.01, "learning_rate": 9.926845840253662e-05, "loss": 1.042, "step": 3910 }, { "epoch": 0.01, "learning_rate": 9.926720016105492e-05, "loss": 1.0406, "step": 3915 }, { "epoch": 0.01, "learning_rate": 9.926594191957321e-05, "loss": 1.0418, "step": 3920 }, { "epoch": 0.01, "learning_rate": 9.92646836780915e-05, "loss": 1.043, "step": 3925 }, { "epoch": 0.01, "learning_rate": 9.92634254366098e-05, "loss": 1.0427, "step": 3930 }, { "epoch": 0.01, "learning_rate": 9.92621671951281e-05, "loss": 1.0416, "step": 3935 }, { "epoch": 0.01, "learning_rate": 9.92609089536464e-05, "loss": 1.041, "step": 3940 }, { "epoch": 0.01, "learning_rate": 9.925965071216469e-05, "loss": 1.0427, "step": 3945 }, { "epoch": 0.01, "learning_rate": 9.925839247068298e-05, "loss": 1.0412, "step": 3950 }, { "epoch": 0.01, "learning_rate": 9.925713422920128e-05, "loss": 1.0424, "step": 3955 }, { "epoch": 0.01, "learning_rate": 9.925587598771957e-05, "loss": 1.0437, "step": 3960 }, { "epoch": 0.01, "learning_rate": 9.925461774623787e-05, "loss": 1.0413, "step": 3965 }, { "epoch": 0.01, "learning_rate": 9.925335950475616e-05, "loss": 1.0422, "step": 3970 }, { "epoch": 0.01, "learning_rate": 9.925210126327445e-05, "loss": 1.0419, "step": 3975 }, { "epoch": 0.01, "learning_rate": 9.925084302179275e-05, "loss": 1.042, "step": 3980 }, { "epoch": 0.01, "learning_rate": 9.924958478031105e-05, "loss": 1.0396, "step": 3985 }, { "epoch": 0.01, "learning_rate": 9.924832653882934e-05, "loss": 1.0393, "step": 3990 }, { "epoch": 0.01, "learning_rate": 9.924706829734763e-05, "loss": 1.0393, "step": 3995 }, { "epoch": 0.01, "learning_rate": 9.924581005586593e-05, "loss": 1.0416, "step": 4000 }, { "epoch": 0.01, "learning_rate": 9.924455181438423e-05, "loss": 1.0414, "step": 4005 }, { "epoch": 0.01, "learning_rate": 9.924329357290252e-05, "loss": 1.0436, "step": 4010 }, { "epoch": 0.01, "learning_rate": 9.924203533142081e-05, "loss": 1.0412, "step": 4015 }, { "epoch": 0.01, "learning_rate": 9.92407770899391e-05, "loss": 1.0403, "step": 4020 }, { "epoch": 0.01, "learning_rate": 9.92395188484574e-05, "loss": 1.0574, "step": 4025 }, { "epoch": 0.01, "learning_rate": 9.92382606069757e-05, "loss": 1.0411, "step": 4030 }, { "epoch": 0.01, "learning_rate": 9.923700236549399e-05, "loss": 1.0396, "step": 4035 }, { "epoch": 0.01, "learning_rate": 9.923574412401228e-05, "loss": 1.0409, "step": 4040 }, { "epoch": 0.01, "learning_rate": 9.923448588253059e-05, "loss": 1.0416, "step": 4045 }, { "epoch": 0.01, "learning_rate": 9.923322764104888e-05, "loss": 1.0408, "step": 4050 }, { "epoch": 0.01, "learning_rate": 9.923196939956717e-05, "loss": 1.0421, "step": 4055 }, { "epoch": 0.01, "learning_rate": 9.923071115808546e-05, "loss": 1.042, "step": 4060 }, { "epoch": 0.01, "learning_rate": 9.922945291660377e-05, "loss": 1.042, "step": 4065 }, { "epoch": 0.01, "learning_rate": 9.922819467512206e-05, "loss": 1.0421, "step": 4070 }, { "epoch": 0.01, "learning_rate": 9.922693643364035e-05, "loss": 1.042, "step": 4075 }, { "epoch": 0.01, "learning_rate": 9.922567819215864e-05, "loss": 1.0411, "step": 4080 }, { "epoch": 0.01, "learning_rate": 9.922441995067693e-05, "loss": 1.0441, "step": 4085 }, { "epoch": 0.01, "learning_rate": 9.922316170919524e-05, "loss": 1.0407, "step": 4090 }, { "epoch": 0.01, "learning_rate": 9.922190346771353e-05, "loss": 1.0417, "step": 4095 }, { "epoch": 0.01, "learning_rate": 9.922064522623182e-05, "loss": 1.0414, "step": 4100 }, { "epoch": 0.01, "learning_rate": 9.921938698475011e-05, "loss": 1.042, "step": 4105 }, { "epoch": 0.01, "learning_rate": 9.921812874326842e-05, "loss": 1.0407, "step": 4110 }, { "epoch": 0.01, "learning_rate": 9.921687050178671e-05, "loss": 1.0415, "step": 4115 }, { "epoch": 0.01, "learning_rate": 9.9215612260305e-05, "loss": 1.0418, "step": 4120 }, { "epoch": 0.01, "learning_rate": 9.921435401882329e-05, "loss": 1.0415, "step": 4125 }, { "epoch": 0.01, "learning_rate": 9.92130957773416e-05, "loss": 1.0418, "step": 4130 }, { "epoch": 0.01, "learning_rate": 9.921183753585989e-05, "loss": 1.0415, "step": 4135 }, { "epoch": 0.01, "learning_rate": 9.921057929437818e-05, "loss": 1.0426, "step": 4140 }, { "epoch": 0.01, "learning_rate": 9.920932105289647e-05, "loss": 1.0416, "step": 4145 }, { "epoch": 0.01, "learning_rate": 9.920806281141476e-05, "loss": 1.0445, "step": 4150 }, { "epoch": 0.01, "learning_rate": 9.920680456993307e-05, "loss": 1.0397, "step": 4155 }, { "epoch": 0.01, "learning_rate": 9.920554632845136e-05, "loss": 1.0416, "step": 4160 }, { "epoch": 0.01, "learning_rate": 9.920428808696965e-05, "loss": 1.0405, "step": 4165 }, { "epoch": 0.01, "learning_rate": 9.920302984548794e-05, "loss": 1.0411, "step": 4170 }, { "epoch": 0.01, "learning_rate": 9.920177160400625e-05, "loss": 1.0426, "step": 4175 }, { "epoch": 0.01, "learning_rate": 9.920051336252454e-05, "loss": 1.0425, "step": 4180 }, { "epoch": 0.01, "learning_rate": 9.919925512104283e-05, "loss": 1.0418, "step": 4185 }, { "epoch": 0.01, "learning_rate": 9.919799687956112e-05, "loss": 1.0416, "step": 4190 }, { "epoch": 0.01, "learning_rate": 9.919673863807943e-05, "loss": 1.0403, "step": 4195 }, { "epoch": 0.01, "learning_rate": 9.919548039659772e-05, "loss": 1.0415, "step": 4200 }, { "epoch": 0.01, "learning_rate": 9.919422215511601e-05, "loss": 1.0404, "step": 4205 }, { "epoch": 0.01, "learning_rate": 9.91929639136343e-05, "loss": 1.0414, "step": 4210 }, { "epoch": 0.01, "learning_rate": 9.91917056721526e-05, "loss": 1.041, "step": 4215 }, { "epoch": 0.01, "learning_rate": 9.91904474306709e-05, "loss": 1.0437, "step": 4220 }, { "epoch": 0.01, "learning_rate": 9.918918918918919e-05, "loss": 1.0442, "step": 4225 }, { "epoch": 0.01, "learning_rate": 9.918793094770748e-05, "loss": 1.0399, "step": 4230 }, { "epoch": 0.01, "learning_rate": 9.918667270622578e-05, "loss": 1.0396, "step": 4235 }, { "epoch": 0.01, "learning_rate": 9.918541446474408e-05, "loss": 1.0407, "step": 4240 }, { "epoch": 0.01, "learning_rate": 9.918415622326237e-05, "loss": 1.0415, "step": 4245 }, { "epoch": 0.01, "learning_rate": 9.918289798178066e-05, "loss": 1.0406, "step": 4250 }, { "epoch": 0.01, "learning_rate": 9.918163974029895e-05, "loss": 1.0399, "step": 4255 }, { "epoch": 0.01, "learning_rate": 9.918038149881726e-05, "loss": 1.0431, "step": 4260 }, { "epoch": 0.01, "learning_rate": 9.917912325733555e-05, "loss": 1.0404, "step": 4265 }, { "epoch": 0.01, "learning_rate": 9.917786501585384e-05, "loss": 1.0421, "step": 4270 }, { "epoch": 0.01, "learning_rate": 9.917660677437213e-05, "loss": 1.0403, "step": 4275 }, { "epoch": 0.01, "learning_rate": 9.917534853289043e-05, "loss": 1.0408, "step": 4280 }, { "epoch": 0.01, "learning_rate": 9.917409029140873e-05, "loss": 1.0427, "step": 4285 }, { "epoch": 0.01, "learning_rate": 9.917283204992702e-05, "loss": 1.0413, "step": 4290 }, { "epoch": 0.01, "learning_rate": 9.917157380844531e-05, "loss": 1.0415, "step": 4295 }, { "epoch": 0.01, "learning_rate": 9.91703155669636e-05, "loss": 1.0405, "step": 4300 }, { "epoch": 0.01, "learning_rate": 9.916905732548191e-05, "loss": 1.0402, "step": 4305 }, { "epoch": 0.01, "learning_rate": 9.91677990840002e-05, "loss": 1.0395, "step": 4310 }, { "epoch": 0.01, "learning_rate": 9.91665408425185e-05, "loss": 1.0389, "step": 4315 }, { "epoch": 0.01, "learning_rate": 9.916528260103679e-05, "loss": 1.0415, "step": 4320 }, { "epoch": 0.01, "learning_rate": 9.916402435955509e-05, "loss": 1.0403, "step": 4325 }, { "epoch": 0.01, "learning_rate": 9.916276611807338e-05, "loss": 1.0413, "step": 4330 }, { "epoch": 0.01, "learning_rate": 9.916150787659167e-05, "loss": 1.0405, "step": 4335 }, { "epoch": 0.01, "learning_rate": 9.916024963510997e-05, "loss": 1.0417, "step": 4340 }, { "epoch": 0.01, "learning_rate": 9.915899139362826e-05, "loss": 1.042, "step": 4345 }, { "epoch": 0.01, "learning_rate": 9.915773315214658e-05, "loss": 1.0414, "step": 4350 }, { "epoch": 0.01, "learning_rate": 9.915647491066487e-05, "loss": 1.042, "step": 4355 }, { "epoch": 0.01, "learning_rate": 9.915521666918316e-05, "loss": 1.0412, "step": 4360 }, { "epoch": 0.01, "learning_rate": 9.915395842770145e-05, "loss": 1.0387, "step": 4365 }, { "epoch": 0.01, "learning_rate": 9.915270018621974e-05, "loss": 1.059, "step": 4370 }, { "epoch": 0.01, "learning_rate": 9.915144194473805e-05, "loss": 1.0418, "step": 4375 }, { "epoch": 0.01, "learning_rate": 9.915018370325634e-05, "loss": 1.04, "step": 4380 }, { "epoch": 0.01, "learning_rate": 9.914892546177463e-05, "loss": 1.0414, "step": 4385 }, { "epoch": 0.01, "learning_rate": 9.914766722029292e-05, "loss": 1.0412, "step": 4390 }, { "epoch": 0.01, "learning_rate": 9.914640897881123e-05, "loss": 1.0402, "step": 4395 }, { "epoch": 0.01, "learning_rate": 9.914515073732952e-05, "loss": 1.0409, "step": 4400 }, { "epoch": 0.01, "learning_rate": 9.914389249584781e-05, "loss": 1.0409, "step": 4405 }, { "epoch": 0.01, "learning_rate": 9.91426342543661e-05, "loss": 1.0412, "step": 4410 }, { "epoch": 0.01, "learning_rate": 9.914137601288441e-05, "loss": 1.0394, "step": 4415 }, { "epoch": 0.01, "learning_rate": 9.91401177714027e-05, "loss": 1.0405, "step": 4420 }, { "epoch": 0.01, "learning_rate": 9.913885952992099e-05, "loss": 1.0417, "step": 4425 }, { "epoch": 0.01, "learning_rate": 9.913760128843928e-05, "loss": 1.0384, "step": 4430 }, { "epoch": 0.01, "learning_rate": 9.913634304695757e-05, "loss": 1.0416, "step": 4435 }, { "epoch": 0.01, "learning_rate": 9.913508480547588e-05, "loss": 1.0376, "step": 4440 }, { "epoch": 0.01, "learning_rate": 9.913382656399417e-05, "loss": 1.0421, "step": 4445 }, { "epoch": 0.01, "learning_rate": 9.913256832251246e-05, "loss": 1.0403, "step": 4450 }, { "epoch": 0.01, "learning_rate": 9.913131008103075e-05, "loss": 1.0417, "step": 4455 }, { "epoch": 0.01, "learning_rate": 9.913005183954906e-05, "loss": 1.0398, "step": 4460 }, { "epoch": 0.01, "learning_rate": 9.912879359806735e-05, "loss": 1.0407, "step": 4465 }, { "epoch": 0.01, "learning_rate": 9.912753535658564e-05, "loss": 1.0399, "step": 4470 }, { "epoch": 0.01, "learning_rate": 9.912627711510393e-05, "loss": 1.0424, "step": 4475 }, { "epoch": 0.01, "learning_rate": 9.912501887362224e-05, "loss": 1.0429, "step": 4480 }, { "epoch": 0.01, "learning_rate": 9.912376063214053e-05, "loss": 1.0406, "step": 4485 }, { "epoch": 0.01, "learning_rate": 9.912250239065882e-05, "loss": 1.0398, "step": 4490 }, { "epoch": 0.01, "learning_rate": 9.912124414917711e-05, "loss": 1.0418, "step": 4495 }, { "epoch": 0.01, "learning_rate": 9.91199859076954e-05, "loss": 1.0412, "step": 4500 }, { "epoch": 0.01, "learning_rate": 9.911872766621371e-05, "loss": 1.0403, "step": 4505 }, { "epoch": 0.01, "learning_rate": 9.9117469424732e-05, "loss": 1.0397, "step": 4510 }, { "epoch": 0.01, "learning_rate": 9.91162111832503e-05, "loss": 1.0415, "step": 4515 }, { "epoch": 0.01, "learning_rate": 9.911495294176859e-05, "loss": 1.0436, "step": 4520 }, { "epoch": 0.01, "learning_rate": 9.911369470028689e-05, "loss": 1.0411, "step": 4525 }, { "epoch": 0.01, "learning_rate": 9.911243645880518e-05, "loss": 1.0405, "step": 4530 }, { "epoch": 0.01, "learning_rate": 9.911117821732347e-05, "loss": 1.0421, "step": 4535 }, { "epoch": 0.01, "learning_rate": 9.910991997584177e-05, "loss": 1.0383, "step": 4540 }, { "epoch": 0.01, "learning_rate": 9.910866173436007e-05, "loss": 1.0466, "step": 4545 }, { "epoch": 0.01, "learning_rate": 9.910740349287836e-05, "loss": 1.0398, "step": 4550 }, { "epoch": 0.01, "learning_rate": 9.910614525139665e-05, "loss": 1.0411, "step": 4555 }, { "epoch": 0.01, "learning_rate": 9.910488700991495e-05, "loss": 1.0413, "step": 4560 }, { "epoch": 0.01, "learning_rate": 9.910362876843324e-05, "loss": 1.0405, "step": 4565 }, { "epoch": 0.01, "learning_rate": 9.910237052695154e-05, "loss": 1.04, "step": 4570 }, { "epoch": 0.01, "learning_rate": 9.910111228546983e-05, "loss": 1.0426, "step": 4575 }, { "epoch": 0.01, "learning_rate": 9.909985404398813e-05, "loss": 1.0395, "step": 4580 }, { "epoch": 0.01, "learning_rate": 9.909859580250642e-05, "loss": 1.0398, "step": 4585 }, { "epoch": 0.01, "learning_rate": 9.909733756102472e-05, "loss": 1.0421, "step": 4590 }, { "epoch": 0.01, "learning_rate": 9.909607931954301e-05, "loss": 1.041, "step": 4595 }, { "epoch": 0.01, "learning_rate": 9.90948210780613e-05, "loss": 1.0394, "step": 4600 }, { "epoch": 0.01, "learning_rate": 9.90935628365796e-05, "loss": 1.0415, "step": 4605 }, { "epoch": 0.01, "learning_rate": 9.90923045950979e-05, "loss": 1.0431, "step": 4610 }, { "epoch": 0.01, "learning_rate": 9.90910463536162e-05, "loss": 1.0413, "step": 4615 }, { "epoch": 0.01, "learning_rate": 9.908978811213449e-05, "loss": 1.0426, "step": 4620 }, { "epoch": 0.01, "learning_rate": 9.908852987065278e-05, "loss": 1.0419, "step": 4625 }, { "epoch": 0.01, "learning_rate": 9.908727162917107e-05, "loss": 1.0424, "step": 4630 }, { "epoch": 0.01, "learning_rate": 9.908601338768937e-05, "loss": 1.0406, "step": 4635 }, { "epoch": 0.01, "learning_rate": 9.908475514620767e-05, "loss": 1.043, "step": 4640 }, { "epoch": 0.01, "learning_rate": 9.908349690472596e-05, "loss": 1.0423, "step": 4645 }, { "epoch": 0.01, "learning_rate": 9.908223866324425e-05, "loss": 1.0375, "step": 4650 }, { "epoch": 0.01, "learning_rate": 9.908098042176255e-05, "loss": 1.0413, "step": 4655 }, { "epoch": 0.01, "learning_rate": 9.907972218028085e-05, "loss": 1.0417, "step": 4660 }, { "epoch": 0.01, "learning_rate": 9.907846393879914e-05, "loss": 1.0405, "step": 4665 }, { "epoch": 0.01, "learning_rate": 9.907720569731743e-05, "loss": 1.0431, "step": 4670 }, { "epoch": 0.01, "learning_rate": 9.907594745583573e-05, "loss": 1.04, "step": 4675 }, { "epoch": 0.01, "learning_rate": 9.907468921435403e-05, "loss": 1.038, "step": 4680 }, { "epoch": 0.01, "learning_rate": 9.907343097287232e-05, "loss": 1.0424, "step": 4685 }, { "epoch": 0.01, "learning_rate": 9.907217273139061e-05, "loss": 1.0411, "step": 4690 }, { "epoch": 0.01, "learning_rate": 9.90709144899089e-05, "loss": 1.0415, "step": 4695 }, { "epoch": 0.01, "learning_rate": 9.90696562484272e-05, "loss": 1.0399, "step": 4700 }, { "epoch": 0.01, "learning_rate": 9.90683980069455e-05, "loss": 1.0385, "step": 4705 }, { "epoch": 0.01, "learning_rate": 9.906713976546379e-05, "loss": 1.0397, "step": 4710 }, { "epoch": 0.01, "learning_rate": 9.906588152398208e-05, "loss": 1.0403, "step": 4715 }, { "epoch": 0.01, "learning_rate": 9.906462328250039e-05, "loss": 1.0411, "step": 4720 }, { "epoch": 0.01, "learning_rate": 9.906336504101868e-05, "loss": 1.0413, "step": 4725 }, { "epoch": 0.01, "learning_rate": 9.906210679953697e-05, "loss": 1.0408, "step": 4730 }, { "epoch": 0.01, "learning_rate": 9.906084855805526e-05, "loss": 1.0409, "step": 4735 }, { "epoch": 0.01, "learning_rate": 9.905959031657355e-05, "loss": 1.0404, "step": 4740 }, { "epoch": 0.01, "learning_rate": 9.905833207509186e-05, "loss": 1.0401, "step": 4745 }, { "epoch": 0.01, "learning_rate": 9.905707383361015e-05, "loss": 1.0384, "step": 4750 }, { "epoch": 0.01, "learning_rate": 9.905581559212844e-05, "loss": 1.042, "step": 4755 }, { "epoch": 0.01, "learning_rate": 9.905455735064673e-05, "loss": 1.043, "step": 4760 }, { "epoch": 0.01, "learning_rate": 9.905329910916504e-05, "loss": 1.0395, "step": 4765 }, { "epoch": 0.01, "learning_rate": 9.905204086768333e-05, "loss": 1.0407, "step": 4770 }, { "epoch": 0.01, "learning_rate": 9.905078262620162e-05, "loss": 1.0393, "step": 4775 }, { "epoch": 0.01, "learning_rate": 9.904952438471991e-05, "loss": 1.0388, "step": 4780 }, { "epoch": 0.01, "learning_rate": 9.904826614323822e-05, "loss": 1.0402, "step": 4785 }, { "epoch": 0.01, "learning_rate": 9.904700790175651e-05, "loss": 1.065, "step": 4790 }, { "epoch": 0.01, "learning_rate": 9.90457496602748e-05, "loss": 1.0396, "step": 4795 }, { "epoch": 0.01, "learning_rate": 9.904449141879309e-05, "loss": 1.0406, "step": 4800 }, { "epoch": 0.01, "learning_rate": 9.904323317731138e-05, "loss": 1.0423, "step": 4805 }, { "epoch": 0.01, "learning_rate": 9.904197493582969e-05, "loss": 1.0398, "step": 4810 }, { "epoch": 0.01, "learning_rate": 9.904071669434798e-05, "loss": 1.0394, "step": 4815 }, { "epoch": 0.01, "learning_rate": 9.903945845286627e-05, "loss": 1.0399, "step": 4820 }, { "epoch": 0.01, "learning_rate": 9.903820021138456e-05, "loss": 1.0408, "step": 4825 }, { "epoch": 0.01, "learning_rate": 9.903694196990287e-05, "loss": 1.0424, "step": 4830 }, { "epoch": 0.01, "learning_rate": 9.903568372842116e-05, "loss": 1.042, "step": 4835 }, { "epoch": 0.01, "learning_rate": 9.903442548693945e-05, "loss": 1.0406, "step": 4840 }, { "epoch": 0.01, "learning_rate": 9.903316724545774e-05, "loss": 1.0415, "step": 4845 }, { "epoch": 0.01, "learning_rate": 9.903190900397605e-05, "loss": 1.04, "step": 4850 }, { "epoch": 0.01, "learning_rate": 9.903065076249435e-05, "loss": 1.0407, "step": 4855 }, { "epoch": 0.01, "learning_rate": 9.902939252101264e-05, "loss": 1.0399, "step": 4860 }, { "epoch": 0.01, "learning_rate": 9.902813427953094e-05, "loss": 1.0404, "step": 4865 }, { "epoch": 0.01, "learning_rate": 9.902687603804923e-05, "loss": 1.0416, "step": 4870 }, { "epoch": 0.01, "learning_rate": 9.902561779656753e-05, "loss": 1.0411, "step": 4875 }, { "epoch": 0.01, "learning_rate": 9.902435955508582e-05, "loss": 1.0416, "step": 4880 }, { "epoch": 0.01, "learning_rate": 9.902310131360412e-05, "loss": 1.0421, "step": 4885 }, { "epoch": 0.01, "learning_rate": 9.902184307212241e-05, "loss": 1.0421, "step": 4890 }, { "epoch": 0.01, "learning_rate": 9.90205848306407e-05, "loss": 1.0393, "step": 4895 }, { "epoch": 0.01, "learning_rate": 9.9019326589159e-05, "loss": 1.0411, "step": 4900 }, { "epoch": 0.01, "learning_rate": 9.90180683476773e-05, "loss": 1.0404, "step": 4905 }, { "epoch": 0.01, "learning_rate": 9.901681010619559e-05, "loss": 1.0431, "step": 4910 }, { "epoch": 0.01, "learning_rate": 9.901555186471388e-05, "loss": 1.0384, "step": 4915 }, { "epoch": 0.01, "learning_rate": 9.901429362323218e-05, "loss": 1.0397, "step": 4920 }, { "epoch": 0.01, "learning_rate": 9.901303538175048e-05, "loss": 1.0383, "step": 4925 }, { "epoch": 0.01, "learning_rate": 9.901177714026877e-05, "loss": 1.0395, "step": 4930 }, { "epoch": 0.01, "learning_rate": 9.901051889878706e-05, "loss": 1.0399, "step": 4935 }, { "epoch": 0.01, "learning_rate": 9.900926065730536e-05, "loss": 1.0412, "step": 4940 }, { "epoch": 0.01, "learning_rate": 9.900800241582366e-05, "loss": 1.04, "step": 4945 }, { "epoch": 0.01, "learning_rate": 9.900674417434195e-05, "loss": 1.0389, "step": 4950 }, { "epoch": 0.01, "learning_rate": 9.900548593286024e-05, "loss": 1.0399, "step": 4955 }, { "epoch": 0.01, "learning_rate": 9.900422769137853e-05, "loss": 1.0408, "step": 4960 }, { "epoch": 0.01, "learning_rate": 9.900296944989684e-05, "loss": 1.0398, "step": 4965 }, { "epoch": 0.01, "learning_rate": 9.900171120841513e-05, "loss": 1.0406, "step": 4970 }, { "epoch": 0.01, "learning_rate": 9.900045296693342e-05, "loss": 1.0433, "step": 4975 }, { "epoch": 0.01, "learning_rate": 9.899919472545171e-05, "loss": 1.0414, "step": 4980 }, { "epoch": 0.01, "learning_rate": 9.899793648397002e-05, "loss": 1.0411, "step": 4985 }, { "epoch": 0.01, "learning_rate": 9.899667824248831e-05, "loss": 1.043, "step": 4990 }, { "epoch": 0.01, "learning_rate": 9.89954200010066e-05, "loss": 1.0418, "step": 4995 }, { "epoch": 0.01, "learning_rate": 9.899416175952489e-05, "loss": 1.0412, "step": 5000 }, { "epoch": 0.01, "learning_rate": 9.89929035180432e-05, "loss": 1.0426, "step": 5005 }, { "epoch": 0.01, "learning_rate": 9.899164527656149e-05, "loss": 1.0401, "step": 5010 }, { "epoch": 0.01, "learning_rate": 9.899038703507978e-05, "loss": 1.0393, "step": 5015 }, { "epoch": 0.01, "learning_rate": 9.898912879359807e-05, "loss": 1.0421, "step": 5020 }, { "epoch": 0.01, "learning_rate": 9.898787055211636e-05, "loss": 1.0421, "step": 5025 }, { "epoch": 0.01, "learning_rate": 9.898661231063467e-05, "loss": 1.0408, "step": 5030 }, { "epoch": 0.01, "learning_rate": 9.898535406915296e-05, "loss": 1.0403, "step": 5035 }, { "epoch": 0.01, "learning_rate": 9.898409582767125e-05, "loss": 1.0406, "step": 5040 }, { "epoch": 0.01, "learning_rate": 9.898283758618954e-05, "loss": 1.0403, "step": 5045 }, { "epoch": 0.01, "learning_rate": 9.898157934470785e-05, "loss": 1.0398, "step": 5050 }, { "epoch": 0.01, "learning_rate": 9.898032110322614e-05, "loss": 1.0414, "step": 5055 }, { "epoch": 0.01, "learning_rate": 9.897906286174443e-05, "loss": 1.0415, "step": 5060 }, { "epoch": 0.01, "learning_rate": 9.897780462026272e-05, "loss": 1.0403, "step": 5065 }, { "epoch": 0.01, "learning_rate": 9.897654637878103e-05, "loss": 1.0417, "step": 5070 }, { "epoch": 0.01, "learning_rate": 9.897528813729932e-05, "loss": 1.0434, "step": 5075 }, { "epoch": 0.01, "learning_rate": 9.897402989581761e-05, "loss": 1.058, "step": 5080 }, { "epoch": 0.01, "learning_rate": 9.89727716543359e-05, "loss": 1.0392, "step": 5085 }, { "epoch": 0.01, "learning_rate": 9.89715134128542e-05, "loss": 1.0404, "step": 5090 }, { "epoch": 0.01, "learning_rate": 9.89702551713725e-05, "loss": 1.0396, "step": 5095 }, { "epoch": 0.01, "learning_rate": 9.896899692989079e-05, "loss": 1.0393, "step": 5100 }, { "epoch": 0.01, "learning_rate": 9.896773868840908e-05, "loss": 1.0397, "step": 5105 }, { "epoch": 0.01, "learning_rate": 9.896648044692737e-05, "loss": 1.0406, "step": 5110 }, { "epoch": 0.01, "learning_rate": 9.896522220544568e-05, "loss": 1.0407, "step": 5115 }, { "epoch": 0.01, "learning_rate": 9.896396396396397e-05, "loss": 1.0412, "step": 5120 }, { "epoch": 0.01, "learning_rate": 9.896270572248226e-05, "loss": 1.0407, "step": 5125 }, { "epoch": 0.01, "learning_rate": 9.896144748100055e-05, "loss": 1.039, "step": 5130 }, { "epoch": 0.01, "learning_rate": 9.896018923951886e-05, "loss": 1.0415, "step": 5135 }, { "epoch": 0.01, "learning_rate": 9.895893099803715e-05, "loss": 1.0429, "step": 5140 }, { "epoch": 0.01, "learning_rate": 9.895767275655544e-05, "loss": 1.0408, "step": 5145 }, { "epoch": 0.01, "learning_rate": 9.895641451507373e-05, "loss": 1.0422, "step": 5150 }, { "epoch": 0.01, "learning_rate": 9.895515627359203e-05, "loss": 1.0658, "step": 5155 }, { "epoch": 0.01, "learning_rate": 9.895389803211033e-05, "loss": 1.0396, "step": 5160 }, { "epoch": 0.01, "learning_rate": 9.895263979062862e-05, "loss": 1.0411, "step": 5165 }, { "epoch": 0.01, "learning_rate": 9.895138154914691e-05, "loss": 1.0408, "step": 5170 }, { "epoch": 0.01, "learning_rate": 9.89501233076652e-05, "loss": 1.0395, "step": 5175 }, { "epoch": 0.01, "learning_rate": 9.894886506618351e-05, "loss": 1.0411, "step": 5180 }, { "epoch": 0.01, "learning_rate": 9.89476068247018e-05, "loss": 1.0409, "step": 5185 }, { "epoch": 0.01, "learning_rate": 9.89463485832201e-05, "loss": 1.0403, "step": 5190 }, { "epoch": 0.01, "learning_rate": 9.894509034173839e-05, "loss": 1.0401, "step": 5195 }, { "epoch": 0.01, "learning_rate": 9.894383210025669e-05, "loss": 1.0384, "step": 5200 }, { "epoch": 0.01, "learning_rate": 9.894257385877498e-05, "loss": 1.0417, "step": 5205 }, { "epoch": 0.01, "learning_rate": 9.894131561729327e-05, "loss": 1.0385, "step": 5210 }, { "epoch": 0.01, "learning_rate": 9.894005737581157e-05, "loss": 1.0584, "step": 5215 }, { "epoch": 0.01, "learning_rate": 9.893879913432986e-05, "loss": 1.0392, "step": 5220 }, { "epoch": 0.01, "learning_rate": 9.893754089284816e-05, "loss": 1.041, "step": 5225 }, { "epoch": 0.01, "learning_rate": 9.893628265136645e-05, "loss": 1.0377, "step": 5230 }, { "epoch": 0.01, "learning_rate": 9.893502440988475e-05, "loss": 1.0394, "step": 5235 }, { "epoch": 0.01, "learning_rate": 9.893376616840304e-05, "loss": 1.0409, "step": 5240 }, { "epoch": 0.01, "learning_rate": 9.893250792692134e-05, "loss": 1.0385, "step": 5245 }, { "epoch": 0.01, "learning_rate": 9.893124968543963e-05, "loss": 1.0404, "step": 5250 }, { "epoch": 0.01, "learning_rate": 9.892999144395793e-05, "loss": 1.0404, "step": 5255 }, { "epoch": 0.01, "learning_rate": 9.892873320247622e-05, "loss": 1.0392, "step": 5260 }, { "epoch": 0.01, "learning_rate": 9.892747496099452e-05, "loss": 1.039, "step": 5265 }, { "epoch": 0.01, "learning_rate": 9.892621671951281e-05, "loss": 1.0394, "step": 5270 }, { "epoch": 0.01, "learning_rate": 9.89249584780311e-05, "loss": 1.0394, "step": 5275 }, { "epoch": 0.01, "learning_rate": 9.89237002365494e-05, "loss": 1.0406, "step": 5280 }, { "epoch": 0.01, "learning_rate": 9.892244199506769e-05, "loss": 1.0401, "step": 5285 }, { "epoch": 0.01, "learning_rate": 9.8921183753586e-05, "loss": 1.0417, "step": 5290 }, { "epoch": 0.01, "learning_rate": 9.891992551210428e-05, "loss": 1.0403, "step": 5295 }, { "epoch": 0.01, "learning_rate": 9.891866727062258e-05, "loss": 1.0372, "step": 5300 }, { "epoch": 0.01, "learning_rate": 9.891740902914087e-05, "loss": 1.0409, "step": 5305 }, { "epoch": 0.01, "learning_rate": 9.891615078765917e-05, "loss": 1.0404, "step": 5310 }, { "epoch": 0.01, "learning_rate": 9.891489254617746e-05, "loss": 1.0411, "step": 5315 }, { "epoch": 0.01, "learning_rate": 9.891363430469576e-05, "loss": 1.0402, "step": 5320 }, { "epoch": 0.01, "learning_rate": 9.891237606321405e-05, "loss": 1.0406, "step": 5325 }, { "epoch": 0.01, "learning_rate": 9.891111782173235e-05, "loss": 1.0396, "step": 5330 }, { "epoch": 0.01, "learning_rate": 9.890985958025064e-05, "loss": 1.0381, "step": 5335 }, { "epoch": 0.01, "learning_rate": 9.890860133876894e-05, "loss": 1.0398, "step": 5340 }, { "epoch": 0.01, "learning_rate": 9.890734309728723e-05, "loss": 1.0394, "step": 5345 }, { "epoch": 0.01, "learning_rate": 9.890608485580553e-05, "loss": 1.0389, "step": 5350 }, { "epoch": 0.01, "learning_rate": 9.890482661432382e-05, "loss": 1.0391, "step": 5355 }, { "epoch": 0.01, "learning_rate": 9.890356837284213e-05, "loss": 1.04, "step": 5360 }, { "epoch": 0.01, "learning_rate": 9.890231013136042e-05, "loss": 1.0392, "step": 5365 }, { "epoch": 0.01, "learning_rate": 9.890105188987871e-05, "loss": 1.042, "step": 5370 }, { "epoch": 0.01, "learning_rate": 9.8899793648397e-05, "loss": 1.0415, "step": 5375 }, { "epoch": 0.01, "learning_rate": 9.889853540691531e-05, "loss": 1.0414, "step": 5380 }, { "epoch": 0.01, "learning_rate": 9.88972771654336e-05, "loss": 1.0404, "step": 5385 }, { "epoch": 0.01, "learning_rate": 9.889601892395189e-05, "loss": 1.0385, "step": 5390 }, { "epoch": 0.01, "learning_rate": 9.889476068247018e-05, "loss": 1.0423, "step": 5395 }, { "epoch": 0.01, "learning_rate": 9.889350244098849e-05, "loss": 1.0385, "step": 5400 }, { "epoch": 0.01, "learning_rate": 9.889224419950678e-05, "loss": 1.0618, "step": 5405 }, { "epoch": 0.01, "learning_rate": 9.889098595802507e-05, "loss": 1.0385, "step": 5410 }, { "epoch": 0.01, "learning_rate": 9.888972771654336e-05, "loss": 1.0398, "step": 5415 }, { "epoch": 0.01, "learning_rate": 9.888846947506166e-05, "loss": 1.0411, "step": 5420 }, { "epoch": 0.01, "learning_rate": 9.888721123357996e-05, "loss": 1.0393, "step": 5425 }, { "epoch": 0.01, "learning_rate": 9.888595299209825e-05, "loss": 1.0398, "step": 5430 }, { "epoch": 0.01, "learning_rate": 9.888469475061654e-05, "loss": 1.039, "step": 5435 }, { "epoch": 0.01, "learning_rate": 9.888343650913484e-05, "loss": 1.057, "step": 5440 }, { "epoch": 0.01, "learning_rate": 9.888217826765314e-05, "loss": 1.0399, "step": 5445 }, { "epoch": 0.01, "learning_rate": 9.888092002617143e-05, "loss": 1.0416, "step": 5450 }, { "epoch": 0.01, "learning_rate": 9.887966178468972e-05, "loss": 1.0396, "step": 5455 }, { "epoch": 0.01, "learning_rate": 9.887840354320802e-05, "loss": 1.0381, "step": 5460 }, { "epoch": 0.01, "learning_rate": 9.887714530172632e-05, "loss": 1.0411, "step": 5465 }, { "epoch": 0.01, "learning_rate": 9.887588706024461e-05, "loss": 1.0408, "step": 5470 }, { "epoch": 0.01, "learning_rate": 9.88746288187629e-05, "loss": 1.0398, "step": 5475 }, { "epoch": 0.01, "learning_rate": 9.88733705772812e-05, "loss": 1.041, "step": 5480 }, { "epoch": 0.01, "learning_rate": 9.887211233579949e-05, "loss": 1.0432, "step": 5485 }, { "epoch": 0.01, "learning_rate": 9.887085409431779e-05, "loss": 1.0411, "step": 5490 }, { "epoch": 0.01, "learning_rate": 9.886959585283608e-05, "loss": 1.0393, "step": 5495 }, { "epoch": 0.01, "learning_rate": 9.886833761135438e-05, "loss": 1.0422, "step": 5500 }, { "epoch": 0.01, "learning_rate": 9.886707936987267e-05, "loss": 1.0419, "step": 5505 }, { "epoch": 0.01, "learning_rate": 9.886582112839097e-05, "loss": 1.0403, "step": 5510 }, { "epoch": 0.01, "learning_rate": 9.886456288690926e-05, "loss": 1.0393, "step": 5515 }, { "epoch": 0.01, "learning_rate": 9.886330464542756e-05, "loss": 1.0416, "step": 5520 }, { "epoch": 0.01, "learning_rate": 9.886204640394585e-05, "loss": 1.039, "step": 5525 }, { "epoch": 0.01, "learning_rate": 9.886078816246415e-05, "loss": 1.041, "step": 5530 }, { "epoch": 0.01, "learning_rate": 9.885952992098244e-05, "loss": 1.0398, "step": 5535 }, { "epoch": 0.01, "learning_rate": 9.885827167950074e-05, "loss": 1.0389, "step": 5540 }, { "epoch": 0.01, "learning_rate": 9.885701343801903e-05, "loss": 1.0406, "step": 5545 }, { "epoch": 0.01, "learning_rate": 9.885575519653732e-05, "loss": 1.0405, "step": 5550 }, { "epoch": 0.01, "learning_rate": 9.885449695505562e-05, "loss": 1.0423, "step": 5555 }, { "epoch": 0.01, "learning_rate": 9.885323871357392e-05, "loss": 1.0396, "step": 5560 }, { "epoch": 0.01, "learning_rate": 9.885198047209221e-05, "loss": 1.04, "step": 5565 }, { "epoch": 0.01, "learning_rate": 9.88507222306105e-05, "loss": 1.0406, "step": 5570 }, { "epoch": 0.01, "learning_rate": 9.88494639891288e-05, "loss": 1.0389, "step": 5575 }, { "epoch": 0.01, "learning_rate": 9.88482057476471e-05, "loss": 1.0424, "step": 5580 }, { "epoch": 0.01, "learning_rate": 9.884694750616539e-05, "loss": 1.0413, "step": 5585 }, { "epoch": 0.01, "learning_rate": 9.884568926468368e-05, "loss": 1.041, "step": 5590 }, { "epoch": 0.01, "learning_rate": 9.884443102320198e-05, "loss": 1.0396, "step": 5595 }, { "epoch": 0.01, "learning_rate": 9.884317278172028e-05, "loss": 1.0419, "step": 5600 }, { "epoch": 0.01, "learning_rate": 9.884191454023857e-05, "loss": 1.041, "step": 5605 }, { "epoch": 0.01, "learning_rate": 9.884065629875686e-05, "loss": 1.039, "step": 5610 }, { "epoch": 0.01, "learning_rate": 9.883939805727515e-05, "loss": 1.0383, "step": 5615 }, { "epoch": 0.01, "learning_rate": 9.883813981579346e-05, "loss": 1.0388, "step": 5620 }, { "epoch": 0.01, "learning_rate": 9.883688157431175e-05, "loss": 1.0417, "step": 5625 }, { "epoch": 0.01, "learning_rate": 9.883562333283004e-05, "loss": 1.0392, "step": 5630 }, { "epoch": 0.01, "learning_rate": 9.883436509134833e-05, "loss": 1.0391, "step": 5635 }, { "epoch": 0.01, "learning_rate": 9.883310684986664e-05, "loss": 1.065, "step": 5640 }, { "epoch": 0.01, "learning_rate": 9.883184860838493e-05, "loss": 1.0397, "step": 5645 }, { "epoch": 0.01, "learning_rate": 9.883059036690322e-05, "loss": 1.0376, "step": 5650 }, { "epoch": 0.01, "learning_rate": 9.882933212542151e-05, "loss": 1.0388, "step": 5655 }, { "epoch": 0.01, "learning_rate": 9.882807388393982e-05, "loss": 1.0385, "step": 5660 }, { "epoch": 0.01, "learning_rate": 9.882681564245811e-05, "loss": 1.0381, "step": 5665 }, { "epoch": 0.01, "learning_rate": 9.88255574009764e-05, "loss": 1.0379, "step": 5670 }, { "epoch": 0.01, "learning_rate": 9.882429915949469e-05, "loss": 1.0427, "step": 5675 }, { "epoch": 0.01, "learning_rate": 9.882304091801298e-05, "loss": 1.041, "step": 5680 }, { "epoch": 0.01, "learning_rate": 9.882178267653129e-05, "loss": 1.0385, "step": 5685 }, { "epoch": 0.01, "learning_rate": 9.882052443504958e-05, "loss": 1.0405, "step": 5690 }, { "epoch": 0.01, "learning_rate": 9.881926619356787e-05, "loss": 1.0397, "step": 5695 }, { "epoch": 0.01, "learning_rate": 9.881800795208616e-05, "loss": 1.0402, "step": 5700 }, { "epoch": 0.01, "learning_rate": 9.881674971060447e-05, "loss": 1.0412, "step": 5705 }, { "epoch": 0.01, "learning_rate": 9.881549146912276e-05, "loss": 1.0396, "step": 5710 }, { "epoch": 0.01, "learning_rate": 9.881423322764105e-05, "loss": 1.039, "step": 5715 }, { "epoch": 0.01, "learning_rate": 9.881297498615934e-05, "loss": 1.0617, "step": 5720 }, { "epoch": 0.01, "learning_rate": 9.881171674467765e-05, "loss": 1.0412, "step": 5725 }, { "epoch": 0.01, "learning_rate": 9.881045850319594e-05, "loss": 1.0405, "step": 5730 }, { "epoch": 0.01, "learning_rate": 9.880920026171423e-05, "loss": 1.0401, "step": 5735 }, { "epoch": 0.01, "learning_rate": 9.880794202023252e-05, "loss": 1.0396, "step": 5740 }, { "epoch": 0.01, "learning_rate": 9.880668377875081e-05, "loss": 1.0648, "step": 5745 }, { "epoch": 0.01, "learning_rate": 9.880542553726912e-05, "loss": 1.0394, "step": 5750 }, { "epoch": 0.01, "learning_rate": 9.880416729578741e-05, "loss": 1.0391, "step": 5755 }, { "epoch": 0.01, "learning_rate": 9.88029090543057e-05, "loss": 1.0401, "step": 5760 }, { "epoch": 0.01, "learning_rate": 9.880165081282399e-05, "loss": 1.0381, "step": 5765 }, { "epoch": 0.01, "learning_rate": 9.88003925713423e-05, "loss": 1.0399, "step": 5770 }, { "epoch": 0.01, "learning_rate": 9.879913432986059e-05, "loss": 1.0354, "step": 5775 }, { "epoch": 0.01, "learning_rate": 9.879787608837888e-05, "loss": 1.0395, "step": 5780 }, { "epoch": 0.01, "learning_rate": 9.879661784689717e-05, "loss": 1.0385, "step": 5785 }, { "epoch": 0.01, "learning_rate": 9.879535960541548e-05, "loss": 1.0417, "step": 5790 }, { "epoch": 0.01, "learning_rate": 9.879410136393377e-05, "loss": 1.0402, "step": 5795 }, { "epoch": 0.01, "learning_rate": 9.879284312245206e-05, "loss": 1.0403, "step": 5800 }, { "epoch": 0.01, "learning_rate": 9.879158488097035e-05, "loss": 1.0409, "step": 5805 }, { "epoch": 0.01, "learning_rate": 9.879032663948864e-05, "loss": 1.0627, "step": 5810 }, { "epoch": 0.01, "learning_rate": 9.878906839800695e-05, "loss": 1.0389, "step": 5815 }, { "epoch": 0.01, "learning_rate": 9.878781015652524e-05, "loss": 1.0415, "step": 5820 }, { "epoch": 0.01, "learning_rate": 9.878655191504353e-05, "loss": 1.0386, "step": 5825 }, { "epoch": 0.01, "learning_rate": 9.878529367356182e-05, "loss": 1.0417, "step": 5830 }, { "epoch": 0.01, "learning_rate": 9.878403543208013e-05, "loss": 1.0381, "step": 5835 }, { "epoch": 0.01, "learning_rate": 9.878277719059842e-05, "loss": 1.0385, "step": 5840 }, { "epoch": 0.01, "learning_rate": 9.878151894911671e-05, "loss": 1.0427, "step": 5845 }, { "epoch": 0.01, "learning_rate": 9.878026070763502e-05, "loss": 1.0616, "step": 5850 }, { "epoch": 0.01, "learning_rate": 9.877900246615331e-05, "loss": 1.0393, "step": 5855 }, { "epoch": 0.01, "learning_rate": 9.877774422467161e-05, "loss": 1.0386, "step": 5860 }, { "epoch": 0.01, "learning_rate": 9.87764859831899e-05, "loss": 1.0397, "step": 5865 }, { "epoch": 0.01, "learning_rate": 9.87752277417082e-05, "loss": 1.0399, "step": 5870 }, { "epoch": 0.01, "learning_rate": 9.877396950022649e-05, "loss": 1.0399, "step": 5875 }, { "epoch": 0.01, "learning_rate": 9.877271125874478e-05, "loss": 1.0396, "step": 5880 }, { "epoch": 0.01, "learning_rate": 9.877145301726309e-05, "loss": 1.0411, "step": 5885 }, { "epoch": 0.01, "learning_rate": 9.877019477578138e-05, "loss": 1.0416, "step": 5890 }, { "epoch": 0.01, "learning_rate": 9.876893653429967e-05, "loss": 1.0392, "step": 5895 }, { "epoch": 0.01, "learning_rate": 9.876767829281796e-05, "loss": 1.0408, "step": 5900 }, { "epoch": 0.01, "learning_rate": 9.876642005133627e-05, "loss": 1.0411, "step": 5905 }, { "epoch": 0.01, "learning_rate": 9.876516180985456e-05, "loss": 1.0387, "step": 5910 }, { "epoch": 0.01, "learning_rate": 9.876390356837285e-05, "loss": 1.0389, "step": 5915 }, { "epoch": 0.01, "learning_rate": 9.876264532689114e-05, "loss": 1.0407, "step": 5920 }, { "epoch": 0.01, "learning_rate": 9.876138708540945e-05, "loss": 1.0399, "step": 5925 }, { "epoch": 0.01, "learning_rate": 9.876012884392774e-05, "loss": 1.0385, "step": 5930 }, { "epoch": 0.01, "learning_rate": 9.875887060244603e-05, "loss": 1.0391, "step": 5935 }, { "epoch": 0.01, "learning_rate": 9.875761236096432e-05, "loss": 1.0384, "step": 5940 }, { "epoch": 0.01, "learning_rate": 9.875635411948261e-05, "loss": 1.0391, "step": 5945 }, { "epoch": 0.01, "learning_rate": 9.875509587800092e-05, "loss": 1.0403, "step": 5950 }, { "epoch": 0.01, "learning_rate": 9.875383763651921e-05, "loss": 1.0416, "step": 5955 }, { "epoch": 0.01, "learning_rate": 9.87525793950375e-05, "loss": 1.0375, "step": 5960 }, { "epoch": 0.01, "learning_rate": 9.875132115355579e-05, "loss": 1.0395, "step": 5965 }, { "epoch": 0.01, "learning_rate": 9.87500629120741e-05, "loss": 1.0408, "step": 5970 }, { "epoch": 0.01, "learning_rate": 9.874880467059239e-05, "loss": 1.0398, "step": 5975 }, { "epoch": 0.02, "learning_rate": 9.874754642911068e-05, "loss": 1.0399, "step": 5980 }, { "epoch": 0.02, "learning_rate": 9.874628818762897e-05, "loss": 1.0385, "step": 5985 }, { "epoch": 0.02, "learning_rate": 9.874502994614728e-05, "loss": 1.0415, "step": 5990 }, { "epoch": 0.02, "learning_rate": 9.874377170466557e-05, "loss": 1.0386, "step": 5995 }, { "epoch": 0.02, "learning_rate": 9.874251346318386e-05, "loss": 1.043, "step": 6000 }, { "epoch": 0.02, "learning_rate": 9.874125522170215e-05, "loss": 1.0411, "step": 6005 }, { "epoch": 0.02, "learning_rate": 9.873999698022044e-05, "loss": 1.0397, "step": 6010 }, { "epoch": 0.02, "learning_rate": 9.873873873873875e-05, "loss": 1.0393, "step": 6015 }, { "epoch": 0.02, "learning_rate": 9.873748049725704e-05, "loss": 1.0385, "step": 6020 }, { "epoch": 0.02, "learning_rate": 9.873622225577533e-05, "loss": 1.0395, "step": 6025 }, { "epoch": 0.02, "learning_rate": 9.873496401429362e-05, "loss": 1.0375, "step": 6030 }, { "epoch": 0.02, "learning_rate": 9.873370577281193e-05, "loss": 1.039, "step": 6035 }, { "epoch": 0.02, "learning_rate": 9.873244753133022e-05, "loss": 1.0391, "step": 6040 }, { "epoch": 0.02, "learning_rate": 9.873118928984851e-05, "loss": 1.0399, "step": 6045 }, { "epoch": 0.02, "learning_rate": 9.87299310483668e-05, "loss": 1.0404, "step": 6050 }, { "epoch": 0.02, "learning_rate": 9.872867280688511e-05, "loss": 1.0404, "step": 6055 }, { "epoch": 0.02, "learning_rate": 9.87274145654034e-05, "loss": 1.0429, "step": 6060 }, { "epoch": 0.02, "learning_rate": 9.872615632392169e-05, "loss": 1.0414, "step": 6065 }, { "epoch": 0.02, "learning_rate": 9.872489808243998e-05, "loss": 1.0393, "step": 6070 }, { "epoch": 0.02, "learning_rate": 9.872363984095828e-05, "loss": 1.0401, "step": 6075 }, { "epoch": 0.02, "learning_rate": 9.872238159947658e-05, "loss": 1.0399, "step": 6080 }, { "epoch": 0.02, "learning_rate": 9.872112335799487e-05, "loss": 1.0398, "step": 6085 }, { "epoch": 0.02, "learning_rate": 9.871986511651316e-05, "loss": 1.0391, "step": 6090 }, { "epoch": 0.02, "learning_rate": 9.871860687503146e-05, "loss": 1.0404, "step": 6095 }, { "epoch": 0.02, "learning_rate": 9.871734863354976e-05, "loss": 1.0388, "step": 6100 }, { "epoch": 0.02, "learning_rate": 9.871609039206805e-05, "loss": 1.0412, "step": 6105 }, { "epoch": 0.02, "learning_rate": 9.871483215058634e-05, "loss": 1.0393, "step": 6110 }, { "epoch": 0.02, "learning_rate": 9.871357390910464e-05, "loss": 1.0398, "step": 6115 }, { "epoch": 0.02, "learning_rate": 9.871231566762294e-05, "loss": 1.0393, "step": 6120 }, { "epoch": 0.02, "learning_rate": 9.871105742614123e-05, "loss": 1.0399, "step": 6125 }, { "epoch": 0.02, "learning_rate": 9.870979918465952e-05, "loss": 1.0404, "step": 6130 }, { "epoch": 0.02, "learning_rate": 9.870854094317782e-05, "loss": 1.0383, "step": 6135 }, { "epoch": 0.02, "learning_rate": 9.870728270169611e-05, "loss": 1.0398, "step": 6140 }, { "epoch": 0.02, "learning_rate": 9.870602446021441e-05, "loss": 1.0382, "step": 6145 }, { "epoch": 0.02, "learning_rate": 9.87047662187327e-05, "loss": 1.0378, "step": 6150 }, { "epoch": 0.02, "learning_rate": 9.8703507977251e-05, "loss": 1.0402, "step": 6155 }, { "epoch": 0.02, "learning_rate": 9.870224973576929e-05, "loss": 1.0387, "step": 6160 }, { "epoch": 0.02, "learning_rate": 9.870099149428759e-05, "loss": 1.0407, "step": 6165 }, { "epoch": 0.02, "learning_rate": 9.869973325280588e-05, "loss": 1.0379, "step": 6170 }, { "epoch": 0.02, "learning_rate": 9.869847501132418e-05, "loss": 1.0413, "step": 6175 }, { "epoch": 0.02, "learning_rate": 9.869721676984247e-05, "loss": 1.0393, "step": 6180 }, { "epoch": 0.02, "learning_rate": 9.869595852836077e-05, "loss": 1.0402, "step": 6185 }, { "epoch": 0.02, "learning_rate": 9.869470028687906e-05, "loss": 1.0394, "step": 6190 }, { "epoch": 0.02, "learning_rate": 9.869344204539736e-05, "loss": 1.04, "step": 6195 }, { "epoch": 0.02, "learning_rate": 9.869218380391565e-05, "loss": 1.0406, "step": 6200 }, { "epoch": 0.02, "learning_rate": 9.869092556243394e-05, "loss": 1.0382, "step": 6205 }, { "epoch": 0.02, "learning_rate": 9.868966732095224e-05, "loss": 1.0396, "step": 6210 }, { "epoch": 0.02, "learning_rate": 9.868840907947054e-05, "loss": 1.0391, "step": 6215 }, { "epoch": 0.02, "learning_rate": 9.868715083798883e-05, "loss": 1.0387, "step": 6220 }, { "epoch": 0.02, "learning_rate": 9.868589259650712e-05, "loss": 1.036, "step": 6225 }, { "epoch": 0.02, "learning_rate": 9.868463435502542e-05, "loss": 1.041, "step": 6230 }, { "epoch": 0.02, "learning_rate": 9.868337611354372e-05, "loss": 1.0416, "step": 6235 }, { "epoch": 0.02, "learning_rate": 9.8682117872062e-05, "loss": 1.0421, "step": 6240 }, { "epoch": 0.02, "learning_rate": 9.86808596305803e-05, "loss": 1.0388, "step": 6245 }, { "epoch": 0.02, "learning_rate": 9.86796013890986e-05, "loss": 1.0378, "step": 6250 }, { "epoch": 0.02, "learning_rate": 9.86783431476169e-05, "loss": 1.0389, "step": 6255 }, { "epoch": 0.02, "learning_rate": 9.867708490613519e-05, "loss": 1.0403, "step": 6260 }, { "epoch": 0.02, "learning_rate": 9.867607831294982e-05, "loss": 1.039, "step": 6265 }, { "epoch": 0.02, "learning_rate": 9.867482007146813e-05, "loss": 1.039, "step": 6270 }, { "epoch": 0.02, "learning_rate": 9.867356182998642e-05, "loss": 1.0383, "step": 6275 }, { "epoch": 0.02, "learning_rate": 9.867230358850471e-05, "loss": 1.0389, "step": 6280 }, { "epoch": 0.02, "learning_rate": 9.8671045347023e-05, "loss": 1.0415, "step": 6285 }, { "epoch": 0.02, "learning_rate": 9.86697871055413e-05, "loss": 1.0415, "step": 6290 }, { "epoch": 0.02, "learning_rate": 9.86685288640596e-05, "loss": 1.0395, "step": 6295 }, { "epoch": 0.02, "learning_rate": 9.866727062257789e-05, "loss": 1.0385, "step": 6300 }, { "epoch": 0.02, "learning_rate": 9.866601238109618e-05, "loss": 1.038, "step": 6305 }, { "epoch": 0.02, "learning_rate": 9.866475413961447e-05, "loss": 1.0397, "step": 6310 }, { "epoch": 0.02, "learning_rate": 9.866349589813278e-05, "loss": 1.037, "step": 6315 }, { "epoch": 0.02, "learning_rate": 9.866223765665107e-05, "loss": 1.0394, "step": 6320 }, { "epoch": 0.02, "learning_rate": 9.866097941516936e-05, "loss": 1.0428, "step": 6325 }, { "epoch": 0.02, "learning_rate": 9.865972117368765e-05, "loss": 1.0383, "step": 6330 }, { "epoch": 0.02, "learning_rate": 9.865846293220596e-05, "loss": 1.041, "step": 6335 }, { "epoch": 0.02, "learning_rate": 9.865720469072425e-05, "loss": 1.0408, "step": 6340 }, { "epoch": 0.02, "learning_rate": 9.865594644924254e-05, "loss": 1.0396, "step": 6345 }, { "epoch": 0.02, "learning_rate": 9.865468820776083e-05, "loss": 1.0423, "step": 6350 }, { "epoch": 0.02, "learning_rate": 9.865342996627913e-05, "loss": 1.0422, "step": 6355 }, { "epoch": 0.02, "learning_rate": 9.865217172479743e-05, "loss": 1.0388, "step": 6360 }, { "epoch": 0.02, "learning_rate": 9.865091348331572e-05, "loss": 1.0564, "step": 6365 }, { "epoch": 0.02, "learning_rate": 9.864965524183401e-05, "loss": 1.0406, "step": 6370 }, { "epoch": 0.02, "learning_rate": 9.86483970003523e-05, "loss": 1.0406, "step": 6375 }, { "epoch": 0.02, "learning_rate": 9.864713875887061e-05, "loss": 1.0406, "step": 6380 }, { "epoch": 0.02, "learning_rate": 9.86458805173889e-05, "loss": 1.0384, "step": 6385 }, { "epoch": 0.02, "learning_rate": 9.86446222759072e-05, "loss": 1.0396, "step": 6390 }, { "epoch": 0.02, "learning_rate": 9.864336403442549e-05, "loss": 1.0387, "step": 6395 }, { "epoch": 0.02, "learning_rate": 9.864210579294379e-05, "loss": 1.0371, "step": 6400 }, { "epoch": 0.02, "learning_rate": 9.864084755146208e-05, "loss": 1.0392, "step": 6405 }, { "epoch": 0.02, "learning_rate": 9.863958930998037e-05, "loss": 1.0411, "step": 6410 }, { "epoch": 0.02, "learning_rate": 9.863833106849867e-05, "loss": 1.04, "step": 6415 }, { "epoch": 0.02, "learning_rate": 9.863707282701696e-05, "loss": 1.0372, "step": 6420 }, { "epoch": 0.02, "learning_rate": 9.863581458553526e-05, "loss": 1.0388, "step": 6425 }, { "epoch": 0.02, "learning_rate": 9.863455634405355e-05, "loss": 1.0399, "step": 6430 }, { "epoch": 0.02, "learning_rate": 9.863329810257185e-05, "loss": 1.0388, "step": 6435 }, { "epoch": 0.02, "learning_rate": 9.863203986109014e-05, "loss": 1.0601, "step": 6440 }, { "epoch": 0.02, "learning_rate": 9.863078161960844e-05, "loss": 1.059, "step": 6445 }, { "epoch": 0.02, "learning_rate": 9.862952337812673e-05, "loss": 1.0383, "step": 6450 }, { "epoch": 0.02, "learning_rate": 9.862826513664503e-05, "loss": 1.0394, "step": 6455 }, { "epoch": 0.02, "learning_rate": 9.862700689516332e-05, "loss": 1.0398, "step": 6460 }, { "epoch": 0.02, "learning_rate": 9.862574865368162e-05, "loss": 1.0394, "step": 6465 }, { "epoch": 0.02, "learning_rate": 9.862449041219991e-05, "loss": 1.0405, "step": 6470 }, { "epoch": 0.02, "learning_rate": 9.86232321707182e-05, "loss": 1.0384, "step": 6475 }, { "epoch": 0.02, "learning_rate": 9.86219739292365e-05, "loss": 1.0385, "step": 6480 }, { "epoch": 0.02, "learning_rate": 9.862071568775479e-05, "loss": 1.0389, "step": 6485 }, { "epoch": 0.02, "learning_rate": 9.86194574462731e-05, "loss": 1.0375, "step": 6490 }, { "epoch": 0.02, "learning_rate": 9.861819920479139e-05, "loss": 1.0391, "step": 6495 }, { "epoch": 0.02, "learning_rate": 9.861694096330968e-05, "loss": 1.0374, "step": 6500 }, { "epoch": 0.02, "learning_rate": 9.861568272182797e-05, "loss": 1.0404, "step": 6505 }, { "epoch": 0.02, "learning_rate": 9.861442448034627e-05, "loss": 1.0411, "step": 6510 }, { "epoch": 0.02, "learning_rate": 9.861316623886457e-05, "loss": 1.042, "step": 6515 }, { "epoch": 0.02, "learning_rate": 9.861190799738286e-05, "loss": 1.0388, "step": 6520 }, { "epoch": 0.02, "learning_rate": 9.861064975590115e-05, "loss": 1.0412, "step": 6525 }, { "epoch": 0.02, "learning_rate": 9.860939151441945e-05, "loss": 1.0392, "step": 6530 }, { "epoch": 0.02, "learning_rate": 9.860813327293775e-05, "loss": 1.0388, "step": 6535 }, { "epoch": 0.02, "learning_rate": 9.860687503145604e-05, "loss": 1.0408, "step": 6540 }, { "epoch": 0.02, "learning_rate": 9.860561678997433e-05, "loss": 1.0407, "step": 6545 }, { "epoch": 0.02, "learning_rate": 9.860435854849263e-05, "loss": 1.0384, "step": 6550 }, { "epoch": 0.02, "learning_rate": 9.860310030701092e-05, "loss": 1.038, "step": 6555 }, { "epoch": 0.02, "learning_rate": 9.860184206552923e-05, "loss": 1.0407, "step": 6560 }, { "epoch": 0.02, "learning_rate": 9.860058382404752e-05, "loss": 1.0379, "step": 6565 }, { "epoch": 0.02, "learning_rate": 9.859932558256581e-05, "loss": 1.0373, "step": 6570 }, { "epoch": 0.02, "learning_rate": 9.85980673410841e-05, "loss": 1.0413, "step": 6575 }, { "epoch": 0.02, "learning_rate": 9.859680909960241e-05, "loss": 1.0398, "step": 6580 }, { "epoch": 0.02, "learning_rate": 9.85955508581207e-05, "loss": 1.0393, "step": 6585 }, { "epoch": 0.02, "learning_rate": 9.8594292616639e-05, "loss": 1.0407, "step": 6590 }, { "epoch": 0.02, "learning_rate": 9.859303437515728e-05, "loss": 1.0368, "step": 6595 }, { "epoch": 0.02, "learning_rate": 9.859177613367559e-05, "loss": 1.039, "step": 6600 }, { "epoch": 0.02, "learning_rate": 9.859051789219388e-05, "loss": 1.0384, "step": 6605 }, { "epoch": 0.02, "learning_rate": 9.858925965071217e-05, "loss": 1.0377, "step": 6610 }, { "epoch": 0.02, "learning_rate": 9.858800140923046e-05, "loss": 1.0635, "step": 6615 }, { "epoch": 0.02, "learning_rate": 9.858674316774876e-05, "loss": 1.0414, "step": 6620 }, { "epoch": 0.02, "learning_rate": 9.858548492626706e-05, "loss": 1.0393, "step": 6625 }, { "epoch": 0.02, "learning_rate": 9.858422668478535e-05, "loss": 1.04, "step": 6630 }, { "epoch": 0.02, "learning_rate": 9.858296844330364e-05, "loss": 1.0413, "step": 6635 }, { "epoch": 0.02, "learning_rate": 9.858171020182194e-05, "loss": 1.0378, "step": 6640 }, { "epoch": 0.02, "learning_rate": 9.858045196034024e-05, "loss": 1.0395, "step": 6645 }, { "epoch": 0.02, "learning_rate": 9.857919371885853e-05, "loss": 1.0402, "step": 6650 }, { "epoch": 0.02, "learning_rate": 9.857793547737682e-05, "loss": 1.0401, "step": 6655 }, { "epoch": 0.02, "learning_rate": 9.857667723589512e-05, "loss": 1.0383, "step": 6660 }, { "epoch": 0.02, "learning_rate": 9.857541899441342e-05, "loss": 1.0384, "step": 6665 }, { "epoch": 0.02, "learning_rate": 9.857416075293171e-05, "loss": 1.0389, "step": 6670 }, { "epoch": 0.02, "learning_rate": 9.857290251145e-05, "loss": 1.0381, "step": 6675 }, { "epoch": 0.02, "learning_rate": 9.85716442699683e-05, "loss": 1.0388, "step": 6680 }, { "epoch": 0.02, "learning_rate": 9.857038602848659e-05, "loss": 1.038, "step": 6685 }, { "epoch": 0.02, "learning_rate": 9.856912778700489e-05, "loss": 1.0389, "step": 6690 }, { "epoch": 0.02, "learning_rate": 9.856786954552318e-05, "loss": 1.0406, "step": 6695 }, { "epoch": 0.02, "learning_rate": 9.856661130404148e-05, "loss": 1.0404, "step": 6700 }, { "epoch": 0.02, "learning_rate": 9.856535306255977e-05, "loss": 1.0421, "step": 6705 }, { "epoch": 0.02, "learning_rate": 9.856409482107807e-05, "loss": 1.0417, "step": 6710 }, { "epoch": 0.02, "learning_rate": 9.856283657959636e-05, "loss": 1.0401, "step": 6715 }, { "epoch": 0.02, "learning_rate": 9.856157833811466e-05, "loss": 1.0403, "step": 6720 }, { "epoch": 0.02, "learning_rate": 9.856032009663295e-05, "loss": 1.0412, "step": 6725 }, { "epoch": 0.02, "learning_rate": 9.855906185515125e-05, "loss": 1.0408, "step": 6730 }, { "epoch": 0.02, "learning_rate": 9.855780361366954e-05, "loss": 1.0386, "step": 6735 }, { "epoch": 0.02, "learning_rate": 9.855654537218784e-05, "loss": 1.0397, "step": 6740 }, { "epoch": 0.02, "learning_rate": 9.855528713070613e-05, "loss": 1.0408, "step": 6745 }, { "epoch": 0.02, "learning_rate": 9.855402888922442e-05, "loss": 1.0379, "step": 6750 }, { "epoch": 0.02, "learning_rate": 9.855277064774272e-05, "loss": 1.04, "step": 6755 }, { "epoch": 0.02, "learning_rate": 9.855151240626102e-05, "loss": 1.0399, "step": 6760 }, { "epoch": 0.02, "learning_rate": 9.855025416477931e-05, "loss": 1.0396, "step": 6765 }, { "epoch": 0.02, "learning_rate": 9.85489959232976e-05, "loss": 1.0385, "step": 6770 }, { "epoch": 0.02, "learning_rate": 9.85477376818159e-05, "loss": 1.0392, "step": 6775 }, { "epoch": 0.02, "learning_rate": 9.85464794403342e-05, "loss": 1.0385, "step": 6780 }, { "epoch": 0.02, "learning_rate": 9.854522119885249e-05, "loss": 1.0375, "step": 6785 }, { "epoch": 0.02, "learning_rate": 9.854396295737078e-05, "loss": 1.0365, "step": 6790 }, { "epoch": 0.02, "learning_rate": 9.854270471588908e-05, "loss": 1.0385, "step": 6795 }, { "epoch": 0.02, "learning_rate": 9.854144647440738e-05, "loss": 1.0394, "step": 6800 }, { "epoch": 0.02, "learning_rate": 9.854018823292567e-05, "loss": 1.0392, "step": 6805 }, { "epoch": 0.02, "learning_rate": 9.853892999144396e-05, "loss": 1.0396, "step": 6810 }, { "epoch": 0.02, "learning_rate": 9.853767174996225e-05, "loss": 1.0404, "step": 6815 }, { "epoch": 0.02, "learning_rate": 9.853641350848056e-05, "loss": 1.0408, "step": 6820 }, { "epoch": 0.02, "learning_rate": 9.853515526699885e-05, "loss": 1.0402, "step": 6825 }, { "epoch": 0.02, "learning_rate": 9.853389702551714e-05, "loss": 1.0381, "step": 6830 }, { "epoch": 0.02, "learning_rate": 9.853263878403543e-05, "loss": 1.039, "step": 6835 }, { "epoch": 0.02, "learning_rate": 9.853138054255374e-05, "loss": 1.04, "step": 6840 }, { "epoch": 0.02, "learning_rate": 9.853012230107203e-05, "loss": 1.041, "step": 6845 }, { "epoch": 0.02, "learning_rate": 9.852886405959032e-05, "loss": 1.0391, "step": 6850 }, { "epoch": 0.02, "learning_rate": 9.852760581810861e-05, "loss": 1.0396, "step": 6855 }, { "epoch": 0.02, "learning_rate": 9.852634757662692e-05, "loss": 1.0401, "step": 6860 }, { "epoch": 0.02, "learning_rate": 9.852508933514521e-05, "loss": 1.0381, "step": 6865 }, { "epoch": 0.02, "learning_rate": 9.85238310936635e-05, "loss": 1.0401, "step": 6870 }, { "epoch": 0.02, "learning_rate": 9.852257285218179e-05, "loss": 1.0385, "step": 6875 }, { "epoch": 0.02, "learning_rate": 9.852131461070008e-05, "loss": 1.0406, "step": 6880 }, { "epoch": 0.02, "learning_rate": 9.852005636921839e-05, "loss": 1.0399, "step": 6885 }, { "epoch": 0.02, "learning_rate": 9.851879812773668e-05, "loss": 1.0386, "step": 6890 }, { "epoch": 0.02, "learning_rate": 9.851753988625497e-05, "loss": 1.0388, "step": 6895 }, { "epoch": 0.02, "learning_rate": 9.851628164477326e-05, "loss": 1.0376, "step": 6900 }, { "epoch": 0.02, "learning_rate": 9.851502340329157e-05, "loss": 1.0383, "step": 6905 }, { "epoch": 0.02, "learning_rate": 9.851376516180986e-05, "loss": 1.0417, "step": 6910 }, { "epoch": 0.02, "learning_rate": 9.851250692032815e-05, "loss": 1.0391, "step": 6915 }, { "epoch": 0.02, "learning_rate": 9.851124867884644e-05, "loss": 1.054, "step": 6920 }, { "epoch": 0.02, "learning_rate": 9.850999043736475e-05, "loss": 1.0377, "step": 6925 }, { "epoch": 0.02, "learning_rate": 9.850873219588304e-05, "loss": 1.0392, "step": 6930 }, { "epoch": 0.02, "learning_rate": 9.850747395440133e-05, "loss": 1.0416, "step": 6935 }, { "epoch": 0.02, "learning_rate": 9.850621571291962e-05, "loss": 1.0393, "step": 6940 }, { "epoch": 0.02, "learning_rate": 9.850495747143791e-05, "loss": 1.0399, "step": 6945 }, { "epoch": 0.02, "learning_rate": 9.850369922995622e-05, "loss": 1.0536, "step": 6950 }, { "epoch": 0.02, "learning_rate": 9.850244098847451e-05, "loss": 1.0388, "step": 6955 }, { "epoch": 0.02, "learning_rate": 9.85011827469928e-05, "loss": 1.0383, "step": 6960 }, { "epoch": 0.02, "learning_rate": 9.84999245055111e-05, "loss": 1.0399, "step": 6965 }, { "epoch": 0.02, "learning_rate": 9.84986662640294e-05, "loss": 1.0384, "step": 6970 }, { "epoch": 0.02, "learning_rate": 9.849740802254769e-05, "loss": 1.0401, "step": 6975 }, { "epoch": 0.02, "learning_rate": 9.849614978106598e-05, "loss": 1.037, "step": 6980 }, { "epoch": 0.02, "learning_rate": 9.849489153958427e-05, "loss": 1.0385, "step": 6985 }, { "epoch": 0.02, "learning_rate": 9.849363329810258e-05, "loss": 1.0589, "step": 6990 }, { "epoch": 0.02, "learning_rate": 9.849237505662087e-05, "loss": 1.0365, "step": 6995 }, { "epoch": 0.02, "learning_rate": 9.849111681513916e-05, "loss": 1.04, "step": 7000 }, { "epoch": 0.02, "learning_rate": 9.848985857365745e-05, "loss": 1.0382, "step": 7005 }, { "epoch": 0.02, "learning_rate": 9.848860033217574e-05, "loss": 1.0379, "step": 7010 }, { "epoch": 0.02, "learning_rate": 9.848734209069405e-05, "loss": 1.0398, "step": 7015 }, { "epoch": 0.02, "learning_rate": 9.848608384921234e-05, "loss": 1.0388, "step": 7020 }, { "epoch": 0.02, "learning_rate": 9.848482560773063e-05, "loss": 1.038, "step": 7025 }, { "epoch": 0.02, "learning_rate": 9.848356736624892e-05, "loss": 1.0379, "step": 7030 }, { "epoch": 0.02, "learning_rate": 9.848230912476723e-05, "loss": 1.0405, "step": 7035 }, { "epoch": 0.02, "learning_rate": 9.848105088328552e-05, "loss": 1.0386, "step": 7040 }, { "epoch": 0.02, "learning_rate": 9.847979264180383e-05, "loss": 1.0589, "step": 7045 }, { "epoch": 0.02, "learning_rate": 9.847853440032212e-05, "loss": 1.0393, "step": 7050 }, { "epoch": 0.02, "learning_rate": 9.847727615884041e-05, "loss": 1.0399, "step": 7055 }, { "epoch": 0.02, "learning_rate": 9.847601791735872e-05, "loss": 1.0379, "step": 7060 }, { "epoch": 0.02, "learning_rate": 9.8474759675877e-05, "loss": 1.04, "step": 7065 }, { "epoch": 0.02, "learning_rate": 9.84735014343953e-05, "loss": 1.0393, "step": 7070 }, { "epoch": 0.02, "learning_rate": 9.847224319291359e-05, "loss": 1.0394, "step": 7075 }, { "epoch": 0.02, "learning_rate": 9.847098495143188e-05, "loss": 1.0404, "step": 7080 }, { "epoch": 0.02, "learning_rate": 9.846972670995019e-05, "loss": 1.0404, "step": 7085 }, { "epoch": 0.02, "learning_rate": 9.846846846846848e-05, "loss": 1.0382, "step": 7090 }, { "epoch": 0.02, "learning_rate": 9.846721022698677e-05, "loss": 1.0399, "step": 7095 }, { "epoch": 0.02, "learning_rate": 9.846595198550506e-05, "loss": 1.0378, "step": 7100 }, { "epoch": 0.02, "learning_rate": 9.846469374402337e-05, "loss": 1.0392, "step": 7105 }, { "epoch": 0.02, "learning_rate": 9.846343550254166e-05, "loss": 1.0391, "step": 7110 }, { "epoch": 0.02, "learning_rate": 9.846217726105995e-05, "loss": 1.0635, "step": 7115 }, { "epoch": 0.02, "learning_rate": 9.846091901957824e-05, "loss": 1.0391, "step": 7120 }, { "epoch": 0.02, "learning_rate": 9.845966077809655e-05, "loss": 1.0389, "step": 7125 }, { "epoch": 0.02, "learning_rate": 9.845840253661484e-05, "loss": 1.0384, "step": 7130 }, { "epoch": 0.02, "learning_rate": 9.845714429513313e-05, "loss": 1.0384, "step": 7135 }, { "epoch": 0.02, "learning_rate": 9.845588605365142e-05, "loss": 1.039, "step": 7140 }, { "epoch": 0.02, "learning_rate": 9.845462781216971e-05, "loss": 1.0384, "step": 7145 }, { "epoch": 0.02, "learning_rate": 9.845336957068802e-05, "loss": 1.0375, "step": 7150 }, { "epoch": 0.02, "learning_rate": 9.845211132920631e-05, "loss": 1.0393, "step": 7155 }, { "epoch": 0.02, "learning_rate": 9.84508530877246e-05, "loss": 1.0377, "step": 7160 }, { "epoch": 0.02, "learning_rate": 9.844959484624289e-05, "loss": 1.0404, "step": 7165 }, { "epoch": 0.02, "learning_rate": 9.84483366047612e-05, "loss": 1.039, "step": 7170 }, { "epoch": 0.02, "learning_rate": 9.844707836327949e-05, "loss": 1.0591, "step": 7175 }, { "epoch": 0.02, "learning_rate": 9.844582012179778e-05, "loss": 1.0398, "step": 7180 }, { "epoch": 0.02, "learning_rate": 9.844456188031607e-05, "loss": 1.0383, "step": 7185 }, { "epoch": 0.02, "learning_rate": 9.844330363883438e-05, "loss": 1.0401, "step": 7190 }, { "epoch": 0.02, "learning_rate": 9.844204539735267e-05, "loss": 1.0382, "step": 7195 }, { "epoch": 0.02, "learning_rate": 9.844078715587096e-05, "loss": 1.0396, "step": 7200 }, { "epoch": 0.02, "learning_rate": 9.843952891438925e-05, "loss": 1.0395, "step": 7205 }, { "epoch": 0.02, "learning_rate": 9.843827067290754e-05, "loss": 1.0382, "step": 7210 }, { "epoch": 0.02, "learning_rate": 9.843701243142585e-05, "loss": 1.0395, "step": 7215 }, { "epoch": 0.02, "learning_rate": 9.843575418994414e-05, "loss": 1.0403, "step": 7220 }, { "epoch": 0.02, "learning_rate": 9.843449594846243e-05, "loss": 1.0377, "step": 7225 }, { "epoch": 0.02, "learning_rate": 9.843323770698072e-05, "loss": 1.053, "step": 7230 }, { "epoch": 0.02, "learning_rate": 9.843197946549903e-05, "loss": 1.0377, "step": 7235 }, { "epoch": 0.02, "learning_rate": 9.843072122401732e-05, "loss": 1.056, "step": 7240 }, { "epoch": 0.02, "learning_rate": 9.842946298253561e-05, "loss": 1.0362, "step": 7245 }, { "epoch": 0.02, "learning_rate": 9.84282047410539e-05, "loss": 1.0381, "step": 7250 }, { "epoch": 0.02, "learning_rate": 9.842694649957221e-05, "loss": 1.0402, "step": 7255 }, { "epoch": 0.02, "learning_rate": 9.84256882580905e-05, "loss": 1.038, "step": 7260 }, { "epoch": 0.02, "learning_rate": 9.842443001660879e-05, "loss": 1.0393, "step": 7265 }, { "epoch": 0.02, "learning_rate": 9.842317177512708e-05, "loss": 1.0375, "step": 7270 }, { "epoch": 0.02, "learning_rate": 9.842191353364538e-05, "loss": 1.0382, "step": 7275 }, { "epoch": 0.02, "learning_rate": 9.842065529216368e-05, "loss": 1.0388, "step": 7280 }, { "epoch": 0.02, "learning_rate": 9.841939705068197e-05, "loss": 1.0381, "step": 7285 }, { "epoch": 0.02, "learning_rate": 9.841813880920026e-05, "loss": 1.0415, "step": 7290 }, { "epoch": 0.02, "learning_rate": 9.841688056771856e-05, "loss": 1.0366, "step": 7295 }, { "epoch": 0.02, "learning_rate": 9.841562232623686e-05, "loss": 1.0375, "step": 7300 }, { "epoch": 0.02, "learning_rate": 9.841436408475515e-05, "loss": 1.0371, "step": 7305 }, { "epoch": 0.02, "learning_rate": 9.841310584327344e-05, "loss": 1.037, "step": 7310 }, { "epoch": 0.02, "learning_rate": 9.841184760179174e-05, "loss": 1.0486, "step": 7315 }, { "epoch": 0.02, "learning_rate": 9.841058936031004e-05, "loss": 1.0375, "step": 7320 }, { "epoch": 0.02, "learning_rate": 9.840933111882833e-05, "loss": 1.0373, "step": 7325 }, { "epoch": 0.02, "learning_rate": 9.840807287734662e-05, "loss": 1.0387, "step": 7330 }, { "epoch": 0.02, "learning_rate": 9.840681463586492e-05, "loss": 1.0376, "step": 7335 }, { "epoch": 0.02, "learning_rate": 9.840555639438321e-05, "loss": 1.0391, "step": 7340 }, { "epoch": 0.02, "learning_rate": 9.840429815290151e-05, "loss": 1.0394, "step": 7345 }, { "epoch": 0.02, "learning_rate": 9.84030399114198e-05, "loss": 1.0393, "step": 7350 }, { "epoch": 0.02, "learning_rate": 9.84017816699381e-05, "loss": 1.0398, "step": 7355 }, { "epoch": 0.02, "learning_rate": 9.840052342845639e-05, "loss": 1.0389, "step": 7360 }, { "epoch": 0.02, "learning_rate": 9.839926518697469e-05, "loss": 1.0359, "step": 7365 }, { "epoch": 0.02, "learning_rate": 9.839800694549298e-05, "loss": 1.0377, "step": 7370 }, { "epoch": 0.02, "learning_rate": 9.839674870401128e-05, "loss": 1.0385, "step": 7375 }, { "epoch": 0.02, "learning_rate": 9.839549046252957e-05, "loss": 1.041, "step": 7380 }, { "epoch": 0.02, "learning_rate": 9.839423222104787e-05, "loss": 1.0379, "step": 7385 }, { "epoch": 0.02, "learning_rate": 9.839297397956616e-05, "loss": 1.0393, "step": 7390 }, { "epoch": 0.02, "learning_rate": 9.839171573808446e-05, "loss": 1.0409, "step": 7395 }, { "epoch": 0.02, "learning_rate": 9.839045749660275e-05, "loss": 1.0375, "step": 7400 }, { "epoch": 0.02, "learning_rate": 9.838919925512104e-05, "loss": 1.0384, "step": 7405 }, { "epoch": 0.02, "learning_rate": 9.838794101363934e-05, "loss": 1.0368, "step": 7410 }, { "epoch": 0.02, "learning_rate": 9.838668277215764e-05, "loss": 1.0394, "step": 7415 }, { "epoch": 0.02, "learning_rate": 9.838542453067593e-05, "loss": 1.04, "step": 7420 }, { "epoch": 0.02, "learning_rate": 9.838416628919422e-05, "loss": 1.0388, "step": 7425 }, { "epoch": 0.02, "learning_rate": 9.838290804771252e-05, "loss": 1.0378, "step": 7430 }, { "epoch": 0.02, "learning_rate": 9.838164980623082e-05, "loss": 1.0402, "step": 7435 }, { "epoch": 0.02, "learning_rate": 9.838039156474911e-05, "loss": 1.0396, "step": 7440 }, { "epoch": 0.02, "learning_rate": 9.83791333232674e-05, "loss": 1.04, "step": 7445 }, { "epoch": 0.02, "learning_rate": 9.83778750817857e-05, "loss": 1.04, "step": 7450 }, { "epoch": 0.02, "learning_rate": 9.8376616840304e-05, "loss": 1.0377, "step": 7455 }, { "epoch": 0.02, "learning_rate": 9.837535859882229e-05, "loss": 1.0381, "step": 7460 }, { "epoch": 0.02, "learning_rate": 9.837410035734058e-05, "loss": 1.0411, "step": 7465 }, { "epoch": 0.02, "learning_rate": 9.837284211585887e-05, "loss": 1.038, "step": 7470 }, { "epoch": 0.02, "learning_rate": 9.837158387437718e-05, "loss": 1.0376, "step": 7475 }, { "epoch": 0.02, "learning_rate": 9.837032563289547e-05, "loss": 1.0415, "step": 7480 }, { "epoch": 0.02, "learning_rate": 9.836906739141376e-05, "loss": 1.0405, "step": 7485 }, { "epoch": 0.02, "learning_rate": 9.836780914993205e-05, "loss": 1.0392, "step": 7490 }, { "epoch": 0.02, "learning_rate": 9.836655090845036e-05, "loss": 1.0373, "step": 7495 }, { "epoch": 0.02, "learning_rate": 9.836529266696865e-05, "loss": 1.0396, "step": 7500 }, { "epoch": 0.02, "learning_rate": 9.836403442548694e-05, "loss": 1.0402, "step": 7505 }, { "epoch": 0.02, "learning_rate": 9.836277618400523e-05, "loss": 1.039, "step": 7510 }, { "epoch": 0.02, "learning_rate": 9.836151794252354e-05, "loss": 1.0379, "step": 7515 }, { "epoch": 0.02, "learning_rate": 9.836025970104183e-05, "loss": 1.0404, "step": 7520 }, { "epoch": 0.02, "learning_rate": 9.835900145956012e-05, "loss": 1.0382, "step": 7525 }, { "epoch": 0.02, "learning_rate": 9.835774321807841e-05, "loss": 1.0379, "step": 7530 }, { "epoch": 0.02, "learning_rate": 9.83564849765967e-05, "loss": 1.0367, "step": 7535 }, { "epoch": 0.02, "learning_rate": 9.8355226735115e-05, "loss": 1.0391, "step": 7540 }, { "epoch": 0.02, "learning_rate": 9.835396849363331e-05, "loss": 1.0397, "step": 7545 }, { "epoch": 0.02, "learning_rate": 9.83527102521516e-05, "loss": 1.0389, "step": 7550 }, { "epoch": 0.02, "learning_rate": 9.83514520106699e-05, "loss": 1.041, "step": 7555 }, { "epoch": 0.02, "learning_rate": 9.835019376918819e-05, "loss": 1.0402, "step": 7560 }, { "epoch": 0.02, "learning_rate": 9.834893552770649e-05, "loss": 1.0397, "step": 7565 }, { "epoch": 0.02, "learning_rate": 9.834767728622478e-05, "loss": 1.037, "step": 7570 }, { "epoch": 0.02, "learning_rate": 9.834641904474307e-05, "loss": 1.0389, "step": 7575 }, { "epoch": 0.02, "learning_rate": 9.834516080326137e-05, "loss": 1.0362, "step": 7580 }, { "epoch": 0.02, "learning_rate": 9.834390256177967e-05, "loss": 1.0385, "step": 7585 }, { "epoch": 0.02, "learning_rate": 9.834264432029796e-05, "loss": 1.0372, "step": 7590 }, { "epoch": 0.02, "learning_rate": 9.834138607881625e-05, "loss": 1.0396, "step": 7595 }, { "epoch": 0.02, "learning_rate": 9.834012783733455e-05, "loss": 1.039, "step": 7600 }, { "epoch": 0.02, "learning_rate": 9.833886959585285e-05, "loss": 1.0383, "step": 7605 }, { "epoch": 0.02, "learning_rate": 9.833761135437114e-05, "loss": 1.0391, "step": 7610 }, { "epoch": 0.02, "learning_rate": 9.833635311288943e-05, "loss": 1.0366, "step": 7615 }, { "epoch": 0.02, "learning_rate": 9.833509487140773e-05, "loss": 1.0404, "step": 7620 }, { "epoch": 0.02, "learning_rate": 9.833383662992602e-05, "loss": 1.0387, "step": 7625 }, { "epoch": 0.02, "learning_rate": 9.833257838844432e-05, "loss": 1.0377, "step": 7630 }, { "epoch": 0.02, "learning_rate": 9.833132014696261e-05, "loss": 1.0398, "step": 7635 }, { "epoch": 0.02, "learning_rate": 9.83300619054809e-05, "loss": 1.0359, "step": 7640 }, { "epoch": 0.02, "learning_rate": 9.83288036639992e-05, "loss": 1.0384, "step": 7645 }, { "epoch": 0.02, "learning_rate": 9.83275454225175e-05, "loss": 1.0385, "step": 7650 }, { "epoch": 0.02, "learning_rate": 9.83262871810358e-05, "loss": 1.0373, "step": 7655 }, { "epoch": 0.02, "learning_rate": 9.832502893955409e-05, "loss": 1.0594, "step": 7660 }, { "epoch": 0.02, "learning_rate": 9.832377069807238e-05, "loss": 1.0405, "step": 7665 }, { "epoch": 0.02, "learning_rate": 9.832251245659068e-05, "loss": 1.0383, "step": 7670 }, { "epoch": 0.02, "learning_rate": 9.832125421510897e-05, "loss": 1.04, "step": 7675 }, { "epoch": 0.02, "learning_rate": 9.831999597362727e-05, "loss": 1.0581, "step": 7680 }, { "epoch": 0.02, "learning_rate": 9.831873773214556e-05, "loss": 1.0389, "step": 7685 }, { "epoch": 0.02, "learning_rate": 9.831747949066385e-05, "loss": 1.0389, "step": 7690 }, { "epoch": 0.02, "learning_rate": 9.831622124918215e-05, "loss": 1.038, "step": 7695 }, { "epoch": 0.02, "learning_rate": 9.831496300770045e-05, "loss": 1.0389, "step": 7700 }, { "epoch": 0.02, "learning_rate": 9.831370476621874e-05, "loss": 1.0372, "step": 7705 }, { "epoch": 0.02, "learning_rate": 9.831244652473703e-05, "loss": 1.0416, "step": 7710 }, { "epoch": 0.02, "learning_rate": 9.831118828325533e-05, "loss": 1.0385, "step": 7715 }, { "epoch": 0.02, "learning_rate": 9.830993004177363e-05, "loss": 1.0374, "step": 7720 }, { "epoch": 0.02, "learning_rate": 9.830867180029192e-05, "loss": 1.0394, "step": 7725 }, { "epoch": 0.02, "learning_rate": 9.830741355881021e-05, "loss": 1.0375, "step": 7730 }, { "epoch": 0.02, "learning_rate": 9.83061553173285e-05, "loss": 1.0388, "step": 7735 }, { "epoch": 0.02, "learning_rate": 9.83048970758468e-05, "loss": 1.0396, "step": 7740 }, { "epoch": 0.02, "learning_rate": 9.83036388343651e-05, "loss": 1.0392, "step": 7745 }, { "epoch": 0.02, "learning_rate": 9.830238059288339e-05, "loss": 1.0373, "step": 7750 }, { "epoch": 0.02, "learning_rate": 9.830112235140168e-05, "loss": 1.0399, "step": 7755 }, { "epoch": 0.02, "learning_rate": 9.829986410991999e-05, "loss": 1.0368, "step": 7760 }, { "epoch": 0.02, "learning_rate": 9.829860586843828e-05, "loss": 1.039, "step": 7765 }, { "epoch": 0.02, "learning_rate": 9.829734762695657e-05, "loss": 1.0376, "step": 7770 }, { "epoch": 0.02, "learning_rate": 9.829608938547486e-05, "loss": 1.0377, "step": 7775 }, { "epoch": 0.02, "learning_rate": 9.829483114399317e-05, "loss": 1.0401, "step": 7780 }, { "epoch": 0.02, "learning_rate": 9.829357290251146e-05, "loss": 1.0386, "step": 7785 }, { "epoch": 0.02, "learning_rate": 9.829231466102975e-05, "loss": 1.0403, "step": 7790 }, { "epoch": 0.02, "learning_rate": 9.829105641954804e-05, "loss": 1.0365, "step": 7795 }, { "epoch": 0.02, "learning_rate": 9.828979817806633e-05, "loss": 1.0369, "step": 7800 }, { "epoch": 0.02, "learning_rate": 9.828853993658464e-05, "loss": 1.0379, "step": 7805 }, { "epoch": 0.02, "learning_rate": 9.828728169510293e-05, "loss": 1.0384, "step": 7810 }, { "epoch": 0.02, "learning_rate": 9.828602345362122e-05, "loss": 1.0381, "step": 7815 }, { "epoch": 0.02, "learning_rate": 9.828476521213951e-05, "loss": 1.0405, "step": 7820 }, { "epoch": 0.02, "learning_rate": 9.828350697065782e-05, "loss": 1.0398, "step": 7825 }, { "epoch": 0.02, "learning_rate": 9.828224872917611e-05, "loss": 1.0378, "step": 7830 }, { "epoch": 0.02, "learning_rate": 9.82809904876944e-05, "loss": 1.0379, "step": 7835 }, { "epoch": 0.02, "learning_rate": 9.827973224621269e-05, "loss": 1.0412, "step": 7840 }, { "epoch": 0.02, "learning_rate": 9.8278474004731e-05, "loss": 1.0388, "step": 7845 }, { "epoch": 0.02, "learning_rate": 9.827721576324929e-05, "loss": 1.0383, "step": 7850 }, { "epoch": 0.02, "learning_rate": 9.827595752176758e-05, "loss": 1.0397, "step": 7855 }, { "epoch": 0.02, "learning_rate": 9.827469928028587e-05, "loss": 1.0384, "step": 7860 }, { "epoch": 0.02, "learning_rate": 9.827344103880416e-05, "loss": 1.0376, "step": 7865 }, { "epoch": 0.02, "learning_rate": 9.827218279732247e-05, "loss": 1.0376, "step": 7870 }, { "epoch": 0.02, "learning_rate": 9.827092455584076e-05, "loss": 1.0385, "step": 7875 }, { "epoch": 0.02, "learning_rate": 9.826966631435905e-05, "loss": 1.0366, "step": 7880 }, { "epoch": 0.02, "learning_rate": 9.826840807287734e-05, "loss": 1.0365, "step": 7885 }, { "epoch": 0.02, "learning_rate": 9.826714983139565e-05, "loss": 1.0377, "step": 7890 }, { "epoch": 0.02, "learning_rate": 9.826589158991394e-05, "loss": 1.0384, "step": 7895 }, { "epoch": 0.02, "learning_rate": 9.826463334843223e-05, "loss": 1.0387, "step": 7900 }, { "epoch": 0.02, "learning_rate": 9.826337510695052e-05, "loss": 1.0356, "step": 7905 }, { "epoch": 0.02, "learning_rate": 9.826211686546883e-05, "loss": 1.0378, "step": 7910 }, { "epoch": 0.02, "learning_rate": 9.826085862398712e-05, "loss": 1.0386, "step": 7915 }, { "epoch": 0.02, "learning_rate": 9.825960038250541e-05, "loss": 1.0406, "step": 7920 }, { "epoch": 0.02, "learning_rate": 9.82583421410237e-05, "loss": 1.0382, "step": 7925 }, { "epoch": 0.02, "learning_rate": 9.8257083899542e-05, "loss": 1.0386, "step": 7930 }, { "epoch": 0.02, "learning_rate": 9.82558256580603e-05, "loss": 1.0396, "step": 7935 }, { "epoch": 0.02, "learning_rate": 9.825456741657859e-05, "loss": 1.0367, "step": 7940 }, { "epoch": 0.02, "learning_rate": 9.825330917509688e-05, "loss": 1.0369, "step": 7945 }, { "epoch": 0.02, "learning_rate": 9.825205093361518e-05, "loss": 1.0379, "step": 7950 }, { "epoch": 0.02, "learning_rate": 9.825079269213348e-05, "loss": 1.04, "step": 7955 }, { "epoch": 0.02, "learning_rate": 9.824953445065177e-05, "loss": 1.0391, "step": 7960 }, { "epoch": 0.02, "learning_rate": 9.824827620917006e-05, "loss": 1.0394, "step": 7965 }, { "epoch": 0.02, "learning_rate": 9.824701796768836e-05, "loss": 1.0388, "step": 7970 }, { "epoch": 0.02, "learning_rate": 9.824575972620666e-05, "loss": 1.0372, "step": 7975 }, { "epoch": 0.02, "learning_rate": 9.824450148472495e-05, "loss": 1.0379, "step": 7980 }, { "epoch": 0.02, "learning_rate": 9.824324324324324e-05, "loss": 1.0382, "step": 7985 }, { "epoch": 0.02, "learning_rate": 9.824198500176154e-05, "loss": 1.0398, "step": 7990 }, { "epoch": 0.02, "learning_rate": 9.824072676027983e-05, "loss": 1.0393, "step": 7995 }, { "epoch": 0.02, "learning_rate": 9.823946851879813e-05, "loss": 1.0367, "step": 8000 }, { "epoch": 0.02, "learning_rate": 9.823821027731642e-05, "loss": 1.0359, "step": 8005 }, { "epoch": 0.02, "learning_rate": 9.823695203583471e-05, "loss": 1.0369, "step": 8010 }, { "epoch": 0.02, "learning_rate": 9.8235693794353e-05, "loss": 1.0388, "step": 8015 }, { "epoch": 0.02, "learning_rate": 9.823443555287131e-05, "loss": 1.0383, "step": 8020 }, { "epoch": 0.02, "learning_rate": 9.82331773113896e-05, "loss": 1.036, "step": 8025 }, { "epoch": 0.02, "learning_rate": 9.82319190699079e-05, "loss": 1.0392, "step": 8030 }, { "epoch": 0.02, "learning_rate": 9.823066082842619e-05, "loss": 1.0375, "step": 8035 }, { "epoch": 0.02, "learning_rate": 9.822940258694449e-05, "loss": 1.0407, "step": 8040 }, { "epoch": 0.02, "learning_rate": 9.82281443454628e-05, "loss": 1.0382, "step": 8045 }, { "epoch": 0.02, "learning_rate": 9.822688610398109e-05, "loss": 1.0391, "step": 8050 }, { "epoch": 0.02, "learning_rate": 9.822562786249938e-05, "loss": 1.0382, "step": 8055 }, { "epoch": 0.02, "learning_rate": 9.822436962101767e-05, "loss": 1.0394, "step": 8060 }, { "epoch": 0.02, "learning_rate": 9.822311137953598e-05, "loss": 1.0384, "step": 8065 }, { "epoch": 0.02, "learning_rate": 9.822185313805427e-05, "loss": 1.0379, "step": 8070 }, { "epoch": 0.02, "learning_rate": 9.822059489657256e-05, "loss": 1.0376, "step": 8075 }, { "epoch": 0.02, "learning_rate": 9.821933665509085e-05, "loss": 1.038, "step": 8080 }, { "epoch": 0.02, "learning_rate": 9.821807841360914e-05, "loss": 1.0612, "step": 8085 }, { "epoch": 0.02, "learning_rate": 9.821682017212745e-05, "loss": 1.0384, "step": 8090 }, { "epoch": 0.02, "learning_rate": 9.821556193064574e-05, "loss": 1.0376, "step": 8095 }, { "epoch": 0.02, "learning_rate": 9.821430368916403e-05, "loss": 1.0368, "step": 8100 }, { "epoch": 0.02, "learning_rate": 9.821304544768232e-05, "loss": 1.035, "step": 8105 }, { "epoch": 0.02, "learning_rate": 9.821178720620063e-05, "loss": 1.0392, "step": 8110 }, { "epoch": 0.02, "learning_rate": 9.821052896471892e-05, "loss": 1.0371, "step": 8115 }, { "epoch": 0.02, "learning_rate": 9.820927072323721e-05, "loss": 1.0391, "step": 8120 }, { "epoch": 0.02, "learning_rate": 9.82080124817555e-05, "loss": 1.038, "step": 8125 }, { "epoch": 0.02, "learning_rate": 9.820675424027381e-05, "loss": 1.0373, "step": 8130 }, { "epoch": 0.02, "learning_rate": 9.82054959987921e-05, "loss": 1.0376, "step": 8135 }, { "epoch": 0.02, "learning_rate": 9.820423775731039e-05, "loss": 1.0397, "step": 8140 }, { "epoch": 0.02, "learning_rate": 9.820297951582868e-05, "loss": 1.0388, "step": 8145 }, { "epoch": 0.02, "learning_rate": 9.820172127434697e-05, "loss": 1.038, "step": 8150 }, { "epoch": 0.02, "learning_rate": 9.820046303286528e-05, "loss": 1.0397, "step": 8155 }, { "epoch": 0.02, "learning_rate": 9.819920479138357e-05, "loss": 1.0369, "step": 8160 }, { "epoch": 0.02, "learning_rate": 9.819794654990186e-05, "loss": 1.0361, "step": 8165 }, { "epoch": 0.02, "learning_rate": 9.819668830842015e-05, "loss": 1.0393, "step": 8170 }, { "epoch": 0.02, "learning_rate": 9.819543006693846e-05, "loss": 1.0397, "step": 8175 }, { "epoch": 0.02, "learning_rate": 9.819417182545675e-05, "loss": 1.0372, "step": 8180 }, { "epoch": 0.02, "learning_rate": 9.819291358397504e-05, "loss": 1.0406, "step": 8185 }, { "epoch": 0.02, "learning_rate": 9.819165534249333e-05, "loss": 1.0393, "step": 8190 }, { "epoch": 0.02, "learning_rate": 9.819039710101164e-05, "loss": 1.0429, "step": 8195 }, { "epoch": 0.02, "learning_rate": 9.818913885952993e-05, "loss": 1.0367, "step": 8200 }, { "epoch": 0.02, "learning_rate": 9.818788061804822e-05, "loss": 1.0405, "step": 8205 }, { "epoch": 0.02, "learning_rate": 9.818662237656651e-05, "loss": 1.0383, "step": 8210 }, { "epoch": 0.02, "learning_rate": 9.81853641350848e-05, "loss": 1.0377, "step": 8215 }, { "epoch": 0.02, "learning_rate": 9.818410589360311e-05, "loss": 1.0382, "step": 8220 }, { "epoch": 0.02, "learning_rate": 9.81828476521214e-05, "loss": 1.0395, "step": 8225 }, { "epoch": 0.02, "learning_rate": 9.81815894106397e-05, "loss": 1.0373, "step": 8230 }, { "epoch": 0.02, "learning_rate": 9.818033116915799e-05, "loss": 1.0401, "step": 8235 }, { "epoch": 0.02, "learning_rate": 9.817907292767629e-05, "loss": 1.0376, "step": 8240 }, { "epoch": 0.02, "learning_rate": 9.817781468619458e-05, "loss": 1.0406, "step": 8245 }, { "epoch": 0.02, "learning_rate": 9.817655644471287e-05, "loss": 1.0381, "step": 8250 }, { "epoch": 0.02, "learning_rate": 9.817529820323117e-05, "loss": 1.0374, "step": 8255 }, { "epoch": 0.02, "learning_rate": 9.817403996174947e-05, "loss": 1.0381, "step": 8260 }, { "epoch": 0.02, "learning_rate": 9.817278172026776e-05, "loss": 1.0395, "step": 8265 }, { "epoch": 0.02, "learning_rate": 9.817152347878605e-05, "loss": 1.0405, "step": 8270 }, { "epoch": 0.02, "learning_rate": 9.817026523730435e-05, "loss": 1.039, "step": 8275 }, { "epoch": 0.02, "learning_rate": 9.816900699582264e-05, "loss": 1.0389, "step": 8280 }, { "epoch": 0.02, "learning_rate": 9.816774875434094e-05, "loss": 1.0394, "step": 8285 }, { "epoch": 0.02, "learning_rate": 9.816649051285923e-05, "loss": 1.0368, "step": 8290 }, { "epoch": 0.02, "learning_rate": 9.816523227137753e-05, "loss": 1.0381, "step": 8295 }, { "epoch": 0.02, "learning_rate": 9.816397402989582e-05, "loss": 1.0399, "step": 8300 }, { "epoch": 0.02, "learning_rate": 9.816271578841412e-05, "loss": 1.0391, "step": 8305 }, { "epoch": 0.02, "learning_rate": 9.816145754693241e-05, "loss": 1.0372, "step": 8310 }, { "epoch": 0.02, "learning_rate": 9.81601993054507e-05, "loss": 1.0403, "step": 8315 }, { "epoch": 0.02, "learning_rate": 9.8158941063969e-05, "loss": 1.0391, "step": 8320 }, { "epoch": 0.02, "learning_rate": 9.81576828224873e-05, "loss": 1.0396, "step": 8325 }, { "epoch": 0.02, "learning_rate": 9.81564245810056e-05, "loss": 1.0395, "step": 8330 }, { "epoch": 0.02, "learning_rate": 9.815516633952389e-05, "loss": 1.0399, "step": 8335 }, { "epoch": 0.02, "learning_rate": 9.815390809804218e-05, "loss": 1.0385, "step": 8340 }, { "epoch": 0.02, "learning_rate": 9.815264985656047e-05, "loss": 1.0384, "step": 8345 }, { "epoch": 0.02, "learning_rate": 9.815139161507877e-05, "loss": 1.0392, "step": 8350 }, { "epoch": 0.02, "learning_rate": 9.815013337359707e-05, "loss": 1.0385, "step": 8355 }, { "epoch": 0.02, "learning_rate": 9.814887513211536e-05, "loss": 1.0389, "step": 8360 }, { "epoch": 0.02, "learning_rate": 9.814761689063365e-05, "loss": 1.0411, "step": 8365 }, { "epoch": 0.02, "learning_rate": 9.814635864915195e-05, "loss": 1.0377, "step": 8370 }, { "epoch": 0.02, "learning_rate": 9.814510040767025e-05, "loss": 1.0388, "step": 8375 }, { "epoch": 0.02, "learning_rate": 9.814384216618854e-05, "loss": 1.0381, "step": 8380 }, { "epoch": 0.02, "learning_rate": 9.814258392470683e-05, "loss": 1.0384, "step": 8385 }, { "epoch": 0.02, "learning_rate": 9.814132568322513e-05, "loss": 1.0387, "step": 8390 }, { "epoch": 0.02, "learning_rate": 9.814006744174343e-05, "loss": 1.0409, "step": 8395 }, { "epoch": 0.02, "learning_rate": 9.813880920026172e-05, "loss": 1.0381, "step": 8400 }, { "epoch": 0.02, "learning_rate": 9.813755095878001e-05, "loss": 1.0378, "step": 8405 }, { "epoch": 0.02, "learning_rate": 9.81362927172983e-05, "loss": 1.0367, "step": 8410 }, { "epoch": 0.02, "learning_rate": 9.81350344758166e-05, "loss": 1.0375, "step": 8415 }, { "epoch": 0.02, "learning_rate": 9.81337762343349e-05, "loss": 1.04, "step": 8420 }, { "epoch": 0.02, "learning_rate": 9.813251799285319e-05, "loss": 1.0422, "step": 8425 }, { "epoch": 0.02, "learning_rate": 9.813125975137148e-05, "loss": 1.0378, "step": 8430 }, { "epoch": 0.02, "learning_rate": 9.813000150988979e-05, "loss": 1.0382, "step": 8435 }, { "epoch": 0.02, "learning_rate": 9.812874326840808e-05, "loss": 1.0388, "step": 8440 }, { "epoch": 0.02, "learning_rate": 9.812748502692637e-05, "loss": 1.0387, "step": 8445 }, { "epoch": 0.02, "learning_rate": 9.812622678544466e-05, "loss": 1.039, "step": 8450 }, { "epoch": 0.02, "learning_rate": 9.812496854396295e-05, "loss": 1.0384, "step": 8455 }, { "epoch": 0.02, "learning_rate": 9.812371030248126e-05, "loss": 1.0384, "step": 8460 }, { "epoch": 0.02, "learning_rate": 9.812245206099955e-05, "loss": 1.0382, "step": 8465 }, { "epoch": 0.02, "learning_rate": 9.812119381951784e-05, "loss": 1.0378, "step": 8470 }, { "epoch": 0.02, "learning_rate": 9.811993557803613e-05, "loss": 1.0381, "step": 8475 }, { "epoch": 0.02, "learning_rate": 9.811867733655444e-05, "loss": 1.0374, "step": 8480 }, { "epoch": 0.02, "learning_rate": 9.811741909507273e-05, "loss": 1.0821, "step": 8485 }, { "epoch": 0.02, "learning_rate": 9.811616085359102e-05, "loss": 1.0385, "step": 8490 }, { "epoch": 0.02, "learning_rate": 9.811490261210931e-05, "loss": 1.0387, "step": 8495 }, { "epoch": 0.02, "learning_rate": 9.811364437062762e-05, "loss": 1.0384, "step": 8500 }, { "epoch": 0.02, "learning_rate": 9.811238612914591e-05, "loss": 1.0376, "step": 8505 }, { "epoch": 0.02, "learning_rate": 9.81111278876642e-05, "loss": 1.037, "step": 8510 }, { "epoch": 0.02, "learning_rate": 9.810986964618249e-05, "loss": 1.0383, "step": 8515 }, { "epoch": 0.02, "learning_rate": 9.810861140470078e-05, "loss": 1.0384, "step": 8520 }, { "epoch": 0.02, "learning_rate": 9.810735316321909e-05, "loss": 1.0361, "step": 8525 }, { "epoch": 0.02, "learning_rate": 9.810609492173738e-05, "loss": 1.0398, "step": 8530 }, { "epoch": 0.02, "learning_rate": 9.810483668025567e-05, "loss": 1.0395, "step": 8535 }, { "epoch": 0.02, "learning_rate": 9.810357843877396e-05, "loss": 1.0381, "step": 8540 }, { "epoch": 0.02, "learning_rate": 9.810232019729227e-05, "loss": 1.038, "step": 8545 }, { "epoch": 0.02, "learning_rate": 9.810106195581057e-05, "loss": 1.038, "step": 8550 }, { "epoch": 0.02, "learning_rate": 9.809980371432887e-05, "loss": 1.038, "step": 8555 }, { "epoch": 0.02, "learning_rate": 9.809854547284716e-05, "loss": 1.0389, "step": 8560 }, { "epoch": 0.02, "learning_rate": 9.809728723136545e-05, "loss": 1.0394, "step": 8565 }, { "epoch": 0.02, "learning_rate": 9.809602898988375e-05, "loss": 1.0372, "step": 8570 }, { "epoch": 0.02, "learning_rate": 9.809477074840204e-05, "loss": 1.0365, "step": 8575 }, { "epoch": 0.02, "learning_rate": 9.809351250692034e-05, "loss": 1.0378, "step": 8580 }, { "epoch": 0.02, "learning_rate": 9.809225426543863e-05, "loss": 1.0397, "step": 8585 }, { "epoch": 0.02, "learning_rate": 9.809099602395693e-05, "loss": 1.039, "step": 8590 }, { "epoch": 0.02, "learning_rate": 9.808973778247522e-05, "loss": 1.0597, "step": 8595 }, { "epoch": 0.02, "learning_rate": 9.808847954099352e-05, "loss": 1.0363, "step": 8600 }, { "epoch": 0.02, "learning_rate": 9.808722129951181e-05, "loss": 1.0373, "step": 8605 }, { "epoch": 0.02, "learning_rate": 9.80859630580301e-05, "loss": 1.0377, "step": 8610 }, { "epoch": 0.02, "learning_rate": 9.80847048165484e-05, "loss": 1.0374, "step": 8615 }, { "epoch": 0.02, "learning_rate": 9.80834465750667e-05, "loss": 1.037, "step": 8620 }, { "epoch": 0.02, "learning_rate": 9.808218833358499e-05, "loss": 1.0385, "step": 8625 }, { "epoch": 0.02, "learning_rate": 9.808093009210328e-05, "loss": 1.0379, "step": 8630 }, { "epoch": 0.02, "learning_rate": 9.807967185062158e-05, "loss": 1.0383, "step": 8635 }, { "epoch": 0.02, "learning_rate": 9.807841360913988e-05, "loss": 1.0419, "step": 8640 }, { "epoch": 0.02, "learning_rate": 9.807715536765817e-05, "loss": 1.0401, "step": 8645 }, { "epoch": 0.02, "learning_rate": 9.807589712617646e-05, "loss": 1.0376, "step": 8650 }, { "epoch": 0.02, "learning_rate": 9.807463888469476e-05, "loss": 1.039, "step": 8655 }, { "epoch": 0.02, "learning_rate": 9.807338064321306e-05, "loss": 1.0379, "step": 8660 }, { "epoch": 0.02, "learning_rate": 9.807212240173135e-05, "loss": 1.0359, "step": 8665 }, { "epoch": 0.02, "learning_rate": 9.807086416024964e-05, "loss": 1.0393, "step": 8670 }, { "epoch": 0.02, "learning_rate": 9.806960591876793e-05, "loss": 1.0354, "step": 8675 }, { "epoch": 0.02, "learning_rate": 9.806834767728624e-05, "loss": 1.0407, "step": 8680 }, { "epoch": 0.02, "learning_rate": 9.806708943580453e-05, "loss": 1.0392, "step": 8685 }, { "epoch": 0.02, "learning_rate": 9.806583119432282e-05, "loss": 1.0389, "step": 8690 }, { "epoch": 0.02, "learning_rate": 9.806457295284111e-05, "loss": 1.0384, "step": 8695 }, { "epoch": 0.02, "learning_rate": 9.806331471135942e-05, "loss": 1.0358, "step": 8700 }, { "epoch": 0.02, "learning_rate": 9.806205646987771e-05, "loss": 1.0364, "step": 8705 }, { "epoch": 0.02, "learning_rate": 9.8060798228396e-05, "loss": 1.0383, "step": 8710 }, { "epoch": 0.02, "learning_rate": 9.805953998691429e-05, "loss": 1.0373, "step": 8715 }, { "epoch": 0.02, "learning_rate": 9.80582817454326e-05, "loss": 1.0379, "step": 8720 }, { "epoch": 0.02, "learning_rate": 9.805702350395089e-05, "loss": 1.0349, "step": 8725 }, { "epoch": 0.02, "learning_rate": 9.805576526246918e-05, "loss": 1.0376, "step": 8730 }, { "epoch": 0.02, "learning_rate": 9.805450702098747e-05, "loss": 1.0384, "step": 8735 }, { "epoch": 0.02, "learning_rate": 9.805324877950576e-05, "loss": 1.0388, "step": 8740 }, { "epoch": 0.02, "learning_rate": 9.805199053802407e-05, "loss": 1.0382, "step": 8745 }, { "epoch": 0.02, "learning_rate": 9.805073229654236e-05, "loss": 1.0387, "step": 8750 }, { "epoch": 0.02, "learning_rate": 9.804947405506065e-05, "loss": 1.0383, "step": 8755 }, { "epoch": 0.02, "learning_rate": 9.804821581357894e-05, "loss": 1.0586, "step": 8760 }, { "epoch": 0.02, "learning_rate": 9.804695757209725e-05, "loss": 1.0389, "step": 8765 }, { "epoch": 0.02, "learning_rate": 9.804569933061554e-05, "loss": 1.038, "step": 8770 }, { "epoch": 0.02, "learning_rate": 9.804444108913383e-05, "loss": 1.0393, "step": 8775 }, { "epoch": 0.02, "learning_rate": 9.804318284765212e-05, "loss": 1.0391, "step": 8780 }, { "epoch": 0.02, "learning_rate": 9.804192460617043e-05, "loss": 1.0356, "step": 8785 }, { "epoch": 0.02, "learning_rate": 9.804066636468872e-05, "loss": 1.0572, "step": 8790 }, { "epoch": 0.02, "learning_rate": 9.803940812320701e-05, "loss": 1.0382, "step": 8795 }, { "epoch": 0.02, "learning_rate": 9.80381498817253e-05, "loss": 1.038, "step": 8800 }, { "epoch": 0.02, "learning_rate": 9.80368916402436e-05, "loss": 1.0368, "step": 8805 }, { "epoch": 0.02, "learning_rate": 9.80356333987619e-05, "loss": 1.0375, "step": 8810 }, { "epoch": 0.02, "learning_rate": 9.803437515728019e-05, "loss": 1.0382, "step": 8815 }, { "epoch": 0.02, "learning_rate": 9.803311691579848e-05, "loss": 1.0383, "step": 8820 }, { "epoch": 0.02, "learning_rate": 9.803185867431677e-05, "loss": 1.0384, "step": 8825 }, { "epoch": 0.02, "learning_rate": 9.803060043283508e-05, "loss": 1.0373, "step": 8830 }, { "epoch": 0.02, "learning_rate": 9.802934219135337e-05, "loss": 1.0369, "step": 8835 }, { "epoch": 0.02, "learning_rate": 9.802808394987166e-05, "loss": 1.0383, "step": 8840 }, { "epoch": 0.02, "learning_rate": 9.802682570838995e-05, "loss": 1.0376, "step": 8845 }, { "epoch": 0.02, "learning_rate": 9.802556746690826e-05, "loss": 1.0597, "step": 8850 }, { "epoch": 0.02, "learning_rate": 9.802430922542655e-05, "loss": 1.0391, "step": 8855 }, { "epoch": 0.02, "learning_rate": 9.802305098394484e-05, "loss": 1.0362, "step": 8860 }, { "epoch": 0.02, "learning_rate": 9.802179274246313e-05, "loss": 1.0372, "step": 8865 }, { "epoch": 0.02, "learning_rate": 9.802053450098143e-05, "loss": 1.0353, "step": 8870 }, { "epoch": 0.02, "learning_rate": 9.801927625949973e-05, "loss": 1.0381, "step": 8875 }, { "epoch": 0.02, "learning_rate": 9.801801801801802e-05, "loss": 1.0372, "step": 8880 }, { "epoch": 0.02, "learning_rate": 9.801675977653631e-05, "loss": 1.0386, "step": 8885 }, { "epoch": 0.02, "learning_rate": 9.80155015350546e-05, "loss": 1.0382, "step": 8890 }, { "epoch": 0.02, "learning_rate": 9.801424329357291e-05, "loss": 1.0346, "step": 8895 }, { "epoch": 0.02, "learning_rate": 9.80129850520912e-05, "loss": 1.0372, "step": 8900 }, { "epoch": 0.02, "learning_rate": 9.80117268106095e-05, "loss": 1.0407, "step": 8905 }, { "epoch": 0.02, "learning_rate": 9.801046856912779e-05, "loss": 1.0375, "step": 8910 }, { "epoch": 0.02, "learning_rate": 9.800921032764609e-05, "loss": 1.0372, "step": 8915 }, { "epoch": 0.02, "learning_rate": 9.800795208616438e-05, "loss": 1.0402, "step": 8920 }, { "epoch": 0.02, "learning_rate": 9.800669384468267e-05, "loss": 1.0395, "step": 8925 }, { "epoch": 0.02, "learning_rate": 9.800543560320097e-05, "loss": 1.0362, "step": 8930 }, { "epoch": 0.02, "learning_rate": 9.800417736171926e-05, "loss": 1.0524, "step": 8935 }, { "epoch": 0.02, "learning_rate": 9.800291912023756e-05, "loss": 1.0376, "step": 8940 }, { "epoch": 0.02, "learning_rate": 9.800166087875585e-05, "loss": 1.0381, "step": 8945 }, { "epoch": 0.02, "learning_rate": 9.800040263727415e-05, "loss": 1.038, "step": 8950 }, { "epoch": 0.02, "learning_rate": 9.799914439579244e-05, "loss": 1.037, "step": 8955 }, { "epoch": 0.02, "learning_rate": 9.799788615431074e-05, "loss": 1.0371, "step": 8960 }, { "epoch": 0.02, "learning_rate": 9.799662791282903e-05, "loss": 1.0386, "step": 8965 }, { "epoch": 0.02, "learning_rate": 9.799536967134733e-05, "loss": 1.0387, "step": 8970 }, { "epoch": 0.02, "learning_rate": 9.799411142986562e-05, "loss": 1.0377, "step": 8975 }, { "epoch": 0.02, "learning_rate": 9.799285318838392e-05, "loss": 1.0381, "step": 8980 }, { "epoch": 0.02, "learning_rate": 9.799159494690221e-05, "loss": 1.0385, "step": 8985 }, { "epoch": 0.02, "learning_rate": 9.79903367054205e-05, "loss": 1.0356, "step": 8990 }, { "epoch": 0.02, "learning_rate": 9.79890784639388e-05, "loss": 1.0357, "step": 8995 }, { "epoch": 0.02, "learning_rate": 9.798782022245709e-05, "loss": 1.0359, "step": 9000 }, { "epoch": 0.02, "learning_rate": 9.79865619809754e-05, "loss": 1.0378, "step": 9005 }, { "epoch": 0.02, "learning_rate": 9.798530373949368e-05, "loss": 1.0584, "step": 9010 }, { "epoch": 0.02, "learning_rate": 9.798404549801198e-05, "loss": 1.0391, "step": 9015 }, { "epoch": 0.02, "learning_rate": 9.798278725653027e-05, "loss": 1.0374, "step": 9020 }, { "epoch": 0.02, "learning_rate": 9.798152901504857e-05, "loss": 1.0391, "step": 9025 }, { "epoch": 0.02, "learning_rate": 9.798027077356686e-05, "loss": 1.0373, "step": 9030 }, { "epoch": 0.02, "learning_rate": 9.797901253208516e-05, "loss": 1.0368, "step": 9035 }, { "epoch": 0.02, "learning_rate": 9.797775429060345e-05, "loss": 1.0364, "step": 9040 }, { "epoch": 0.02, "learning_rate": 9.797649604912175e-05, "loss": 1.0372, "step": 9045 }, { "epoch": 0.02, "learning_rate": 9.797523780764006e-05, "loss": 1.0598, "step": 9050 }, { "epoch": 0.02, "learning_rate": 9.797397956615835e-05, "loss": 1.0363, "step": 9055 }, { "epoch": 0.02, "learning_rate": 9.797272132467664e-05, "loss": 1.036, "step": 9060 }, { "epoch": 0.02, "learning_rate": 9.797146308319493e-05, "loss": 1.036, "step": 9065 }, { "epoch": 0.02, "learning_rate": 9.797020484171322e-05, "loss": 1.037, "step": 9070 }, { "epoch": 0.02, "learning_rate": 9.796894660023153e-05, "loss": 1.0373, "step": 9075 }, { "epoch": 0.02, "learning_rate": 9.796768835874982e-05, "loss": 1.039, "step": 9080 }, { "epoch": 0.02, "learning_rate": 9.796643011726811e-05, "loss": 1.0363, "step": 9085 }, { "epoch": 0.02, "learning_rate": 9.79651718757864e-05, "loss": 1.04, "step": 9090 }, { "epoch": 0.02, "learning_rate": 9.796391363430471e-05, "loss": 1.0377, "step": 9095 }, { "epoch": 0.02, "learning_rate": 9.7962655392823e-05, "loss": 1.0392, "step": 9100 }, { "epoch": 0.02, "learning_rate": 9.796139715134129e-05, "loss": 1.0369, "step": 9105 }, { "epoch": 0.02, "learning_rate": 9.796013890985958e-05, "loss": 1.0378, "step": 9110 }, { "epoch": 0.02, "learning_rate": 9.795888066837789e-05, "loss": 1.0367, "step": 9115 }, { "epoch": 0.02, "learning_rate": 9.795762242689618e-05, "loss": 1.0365, "step": 9120 }, { "epoch": 0.02, "learning_rate": 9.795636418541447e-05, "loss": 1.0375, "step": 9125 }, { "epoch": 0.02, "learning_rate": 9.795510594393276e-05, "loss": 1.0454, "step": 9130 }, { "epoch": 0.02, "learning_rate": 9.795384770245106e-05, "loss": 1.0387, "step": 9135 }, { "epoch": 0.02, "learning_rate": 9.795258946096936e-05, "loss": 1.0365, "step": 9140 }, { "epoch": 0.02, "learning_rate": 9.795133121948765e-05, "loss": 1.0385, "step": 9145 }, { "epoch": 0.02, "learning_rate": 9.795007297800594e-05, "loss": 1.0358, "step": 9150 }, { "epoch": 0.02, "learning_rate": 9.794881473652424e-05, "loss": 1.0376, "step": 9155 }, { "epoch": 0.02, "learning_rate": 9.794755649504254e-05, "loss": 1.0372, "step": 9160 }, { "epoch": 0.02, "learning_rate": 9.794629825356083e-05, "loss": 1.0378, "step": 9165 }, { "epoch": 0.02, "learning_rate": 9.794504001207912e-05, "loss": 1.041, "step": 9170 }, { "epoch": 0.02, "learning_rate": 9.794378177059742e-05, "loss": 1.0353, "step": 9175 }, { "epoch": 0.02, "learning_rate": 9.794252352911572e-05, "loss": 1.0383, "step": 9180 }, { "epoch": 0.02, "learning_rate": 9.794126528763401e-05, "loss": 1.0364, "step": 9185 }, { "epoch": 0.02, "learning_rate": 9.79400070461523e-05, "loss": 1.0371, "step": 9190 }, { "epoch": 0.02, "learning_rate": 9.79387488046706e-05, "loss": 1.0382, "step": 9195 }, { "epoch": 0.02, "learning_rate": 9.793749056318889e-05, "loss": 1.0407, "step": 9200 }, { "epoch": 0.02, "learning_rate": 9.793623232170719e-05, "loss": 1.0368, "step": 9205 }, { "epoch": 0.02, "learning_rate": 9.793497408022548e-05, "loss": 1.0386, "step": 9210 }, { "epoch": 0.02, "learning_rate": 9.793371583874378e-05, "loss": 1.0371, "step": 9215 }, { "epoch": 0.02, "learning_rate": 9.793245759726207e-05, "loss": 1.0378, "step": 9220 }, { "epoch": 0.02, "learning_rate": 9.793119935578037e-05, "loss": 1.0382, "step": 9225 }, { "epoch": 0.02, "learning_rate": 9.792994111429866e-05, "loss": 1.0384, "step": 9230 }, { "epoch": 0.02, "learning_rate": 9.792868287281696e-05, "loss": 1.0615, "step": 9235 }, { "epoch": 0.02, "learning_rate": 9.792742463133525e-05, "loss": 1.0365, "step": 9240 }, { "epoch": 0.02, "learning_rate": 9.792616638985355e-05, "loss": 1.0373, "step": 9245 }, { "epoch": 0.02, "learning_rate": 9.792490814837184e-05, "loss": 1.0385, "step": 9250 }, { "epoch": 0.02, "learning_rate": 9.792364990689014e-05, "loss": 1.0393, "step": 9255 }, { "epoch": 0.02, "learning_rate": 9.792239166540843e-05, "loss": 1.0376, "step": 9260 }, { "epoch": 0.02, "learning_rate": 9.792113342392672e-05, "loss": 1.0379, "step": 9265 }, { "epoch": 0.02, "learning_rate": 9.791987518244502e-05, "loss": 1.0384, "step": 9270 }, { "epoch": 0.02, "learning_rate": 9.791861694096332e-05, "loss": 1.038, "step": 9275 }, { "epoch": 0.02, "learning_rate": 9.791735869948161e-05, "loss": 1.0386, "step": 9280 }, { "epoch": 0.02, "learning_rate": 9.79161004579999e-05, "loss": 1.0387, "step": 9285 }, { "epoch": 0.02, "learning_rate": 9.79148422165182e-05, "loss": 1.0387, "step": 9290 }, { "epoch": 0.02, "learning_rate": 9.79135839750365e-05, "loss": 1.0368, "step": 9295 }, { "epoch": 0.02, "learning_rate": 9.791232573355479e-05, "loss": 1.0369, "step": 9300 }, { "epoch": 0.02, "learning_rate": 9.791106749207308e-05, "loss": 1.0389, "step": 9305 }, { "epoch": 0.02, "learning_rate": 9.790980925059138e-05, "loss": 1.0397, "step": 9310 }, { "epoch": 0.02, "learning_rate": 9.790855100910968e-05, "loss": 1.0388, "step": 9315 }, { "epoch": 0.02, "learning_rate": 9.790729276762797e-05, "loss": 1.0361, "step": 9320 }, { "epoch": 0.02, "learning_rate": 9.790603452614626e-05, "loss": 1.0373, "step": 9325 }, { "epoch": 0.02, "learning_rate": 9.790477628466455e-05, "loss": 1.0376, "step": 9330 }, { "epoch": 0.02, "learning_rate": 9.790351804318286e-05, "loss": 1.041, "step": 9335 }, { "epoch": 0.02, "learning_rate": 9.790225980170115e-05, "loss": 1.0389, "step": 9340 }, { "epoch": 0.02, "learning_rate": 9.790100156021944e-05, "loss": 1.0366, "step": 9345 }, { "epoch": 0.02, "learning_rate": 9.789974331873773e-05, "loss": 1.0385, "step": 9350 }, { "epoch": 0.02, "learning_rate": 9.789848507725604e-05, "loss": 1.0404, "step": 9355 }, { "epoch": 0.02, "learning_rate": 9.789722683577433e-05, "loss": 1.038, "step": 9360 }, { "epoch": 0.02, "learning_rate": 9.789596859429262e-05, "loss": 1.0393, "step": 9365 }, { "epoch": 0.02, "learning_rate": 9.789471035281091e-05, "loss": 1.0376, "step": 9370 }, { "epoch": 0.02, "learning_rate": 9.789345211132922e-05, "loss": 1.0376, "step": 9375 }, { "epoch": 0.02, "learning_rate": 9.789219386984751e-05, "loss": 1.0397, "step": 9380 }, { "epoch": 0.02, "learning_rate": 9.78909356283658e-05, "loss": 1.0392, "step": 9385 }, { "epoch": 0.02, "learning_rate": 9.788967738688409e-05, "loss": 1.0381, "step": 9390 }, { "epoch": 0.02, "learning_rate": 9.788841914540238e-05, "loss": 1.0377, "step": 9395 }, { "epoch": 0.02, "learning_rate": 9.788716090392069e-05, "loss": 1.0401, "step": 9400 }, { "epoch": 0.02, "learning_rate": 9.788590266243898e-05, "loss": 1.0358, "step": 9405 }, { "epoch": 0.02, "learning_rate": 9.788464442095727e-05, "loss": 1.036, "step": 9410 }, { "epoch": 0.02, "learning_rate": 9.788338617947556e-05, "loss": 1.0378, "step": 9415 }, { "epoch": 0.02, "learning_rate": 9.788212793799387e-05, "loss": 1.0359, "step": 9420 }, { "epoch": 0.02, "learning_rate": 9.788086969651216e-05, "loss": 1.0374, "step": 9425 }, { "epoch": 0.02, "learning_rate": 9.787961145503045e-05, "loss": 1.0396, "step": 9430 }, { "epoch": 0.02, "learning_rate": 9.787835321354874e-05, "loss": 1.0382, "step": 9435 }, { "epoch": 0.02, "learning_rate": 9.787709497206705e-05, "loss": 1.0501, "step": 9440 }, { "epoch": 0.02, "learning_rate": 9.787583673058534e-05, "loss": 1.0391, "step": 9445 }, { "epoch": 0.02, "learning_rate": 9.787457848910363e-05, "loss": 1.0384, "step": 9450 }, { "epoch": 0.02, "learning_rate": 9.787332024762192e-05, "loss": 1.0363, "step": 9455 }, { "epoch": 0.02, "learning_rate": 9.787206200614021e-05, "loss": 1.0406, "step": 9460 }, { "epoch": 0.02, "learning_rate": 9.787080376465852e-05, "loss": 1.0377, "step": 9465 }, { "epoch": 0.02, "learning_rate": 9.786954552317681e-05, "loss": 1.0377, "step": 9470 }, { "epoch": 0.02, "learning_rate": 9.78682872816951e-05, "loss": 1.0395, "step": 9475 }, { "epoch": 0.02, "learning_rate": 9.78670290402134e-05, "loss": 1.0385, "step": 9480 }, { "epoch": 0.02, "learning_rate": 9.78657707987317e-05, "loss": 1.0371, "step": 9485 }, { "epoch": 0.02, "learning_rate": 9.786451255724999e-05, "loss": 1.0377, "step": 9490 }, { "epoch": 0.02, "learning_rate": 9.786325431576828e-05, "loss": 1.0393, "step": 9495 }, { "epoch": 0.02, "learning_rate": 9.786199607428657e-05, "loss": 1.0358, "step": 9500 }, { "epoch": 0.02, "learning_rate": 9.786073783280488e-05, "loss": 1.0364, "step": 9505 }, { "epoch": 0.02, "learning_rate": 9.785947959132317e-05, "loss": 1.0367, "step": 9510 }, { "epoch": 0.02, "learning_rate": 9.785822134984146e-05, "loss": 1.0384, "step": 9515 }, { "epoch": 0.02, "learning_rate": 9.785696310835975e-05, "loss": 1.0394, "step": 9520 }, { "epoch": 0.02, "learning_rate": 9.785570486687804e-05, "loss": 1.0391, "step": 9525 }, { "epoch": 0.02, "learning_rate": 9.785444662539635e-05, "loss": 1.0359, "step": 9530 }, { "epoch": 0.02, "learning_rate": 9.785318838391464e-05, "loss": 1.0373, "step": 9535 }, { "epoch": 0.02, "learning_rate": 9.785193014243293e-05, "loss": 1.0403, "step": 9540 }, { "epoch": 0.02, "learning_rate": 9.785067190095124e-05, "loss": 1.0375, "step": 9545 }, { "epoch": 0.02, "learning_rate": 9.784941365946953e-05, "loss": 1.0374, "step": 9550 }, { "epoch": 0.02, "learning_rate": 9.784815541798784e-05, "loss": 1.0385, "step": 9555 }, { "epoch": 0.02, "learning_rate": 9.784689717650613e-05, "loss": 1.038, "step": 9560 }, { "epoch": 0.02, "learning_rate": 9.784563893502442e-05, "loss": 1.0382, "step": 9565 }, { "epoch": 0.02, "learning_rate": 9.784438069354271e-05, "loss": 1.0383, "step": 9570 }, { "epoch": 0.02, "learning_rate": 9.784312245206101e-05, "loss": 1.0372, "step": 9575 }, { "epoch": 0.02, "learning_rate": 9.78418642105793e-05, "loss": 1.0363, "step": 9580 }, { "epoch": 0.02, "learning_rate": 9.78406059690976e-05, "loss": 1.0368, "step": 9585 }, { "epoch": 0.02, "learning_rate": 9.783934772761589e-05, "loss": 1.0379, "step": 9590 }, { "epoch": 0.02, "learning_rate": 9.783808948613418e-05, "loss": 1.0354, "step": 9595 }, { "epoch": 0.02, "learning_rate": 9.783683124465249e-05, "loss": 1.0388, "step": 9600 }, { "epoch": 0.02, "learning_rate": 9.783557300317078e-05, "loss": 1.0393, "step": 9605 }, { "epoch": 0.02, "learning_rate": 9.783431476168907e-05, "loss": 1.0393, "step": 9610 }, { "epoch": 0.02, "learning_rate": 9.783305652020736e-05, "loss": 1.0369, "step": 9615 }, { "epoch": 0.02, "learning_rate": 9.783179827872567e-05, "loss": 1.0357, "step": 9620 }, { "epoch": 0.02, "learning_rate": 9.783054003724396e-05, "loss": 1.0371, "step": 9625 }, { "epoch": 0.02, "learning_rate": 9.782928179576225e-05, "loss": 1.0376, "step": 9630 }, { "epoch": 0.02, "learning_rate": 9.782802355428054e-05, "loss": 1.0368, "step": 9635 }, { "epoch": 0.02, "learning_rate": 9.782676531279885e-05, "loss": 1.0371, "step": 9640 }, { "epoch": 0.02, "learning_rate": 9.782550707131714e-05, "loss": 1.0348, "step": 9645 }, { "epoch": 0.02, "learning_rate": 9.782424882983543e-05, "loss": 1.0406, "step": 9650 }, { "epoch": 0.02, "learning_rate": 9.782299058835372e-05, "loss": 1.0385, "step": 9655 }, { "epoch": 0.02, "learning_rate": 9.782173234687201e-05, "loss": 1.0365, "step": 9660 }, { "epoch": 0.02, "learning_rate": 9.782047410539032e-05, "loss": 1.0367, "step": 9665 }, { "epoch": 0.02, "learning_rate": 9.781921586390861e-05, "loss": 1.038, "step": 9670 }, { "epoch": 0.02, "learning_rate": 9.78179576224269e-05, "loss": 1.0388, "step": 9675 }, { "epoch": 0.02, "learning_rate": 9.781669938094519e-05, "loss": 1.0374, "step": 9680 }, { "epoch": 0.02, "learning_rate": 9.78154411394635e-05, "loss": 1.0368, "step": 9685 }, { "epoch": 0.02, "learning_rate": 9.781418289798179e-05, "loss": 1.0401, "step": 9690 }, { "epoch": 0.02, "learning_rate": 9.781292465650008e-05, "loss": 1.0381, "step": 9695 }, { "epoch": 0.02, "learning_rate": 9.781166641501837e-05, "loss": 1.0363, "step": 9700 }, { "epoch": 0.02, "learning_rate": 9.781040817353668e-05, "loss": 1.039, "step": 9705 }, { "epoch": 0.02, "learning_rate": 9.780914993205497e-05, "loss": 1.0374, "step": 9710 }, { "epoch": 0.02, "learning_rate": 9.780789169057326e-05, "loss": 1.0367, "step": 9715 }, { "epoch": 0.02, "learning_rate": 9.780663344909155e-05, "loss": 1.0373, "step": 9720 }, { "epoch": 0.02, "learning_rate": 9.780537520760984e-05, "loss": 1.038, "step": 9725 }, { "epoch": 0.02, "learning_rate": 9.780411696612815e-05, "loss": 1.0391, "step": 9730 }, { "epoch": 0.02, "learning_rate": 9.780285872464644e-05, "loss": 1.0391, "step": 9735 }, { "epoch": 0.02, "learning_rate": 9.780160048316473e-05, "loss": 1.0389, "step": 9740 }, { "epoch": 0.02, "learning_rate": 9.780034224168302e-05, "loss": 1.0363, "step": 9745 }, { "epoch": 0.02, "learning_rate": 9.779908400020133e-05, "loss": 1.0376, "step": 9750 }, { "epoch": 0.02, "learning_rate": 9.779782575871962e-05, "loss": 1.0364, "step": 9755 }, { "epoch": 0.02, "learning_rate": 9.779656751723791e-05, "loss": 1.0395, "step": 9760 }, { "epoch": 0.02, "learning_rate": 9.77953092757562e-05, "loss": 1.0387, "step": 9765 }, { "epoch": 0.02, "learning_rate": 9.779405103427451e-05, "loss": 1.0375, "step": 9770 }, { "epoch": 0.02, "learning_rate": 9.77927927927928e-05, "loss": 1.0383, "step": 9775 }, { "epoch": 0.02, "learning_rate": 9.779153455131109e-05, "loss": 1.0361, "step": 9780 }, { "epoch": 0.02, "learning_rate": 9.779027630982938e-05, "loss": 1.0362, "step": 9785 }, { "epoch": 0.02, "learning_rate": 9.778901806834768e-05, "loss": 1.0377, "step": 9790 }, { "epoch": 0.02, "learning_rate": 9.778775982686598e-05, "loss": 1.0349, "step": 9795 }, { "epoch": 0.02, "learning_rate": 9.778650158538427e-05, "loss": 1.0356, "step": 9800 }, { "epoch": 0.02, "learning_rate": 9.778524334390256e-05, "loss": 1.0365, "step": 9805 }, { "epoch": 0.02, "learning_rate": 9.778398510242086e-05, "loss": 1.0359, "step": 9810 }, { "epoch": 0.02, "learning_rate": 9.778272686093916e-05, "loss": 1.0378, "step": 9815 }, { "epoch": 0.02, "learning_rate": 9.778146861945745e-05, "loss": 1.0378, "step": 9820 }, { "epoch": 0.02, "learning_rate": 9.778021037797574e-05, "loss": 1.0361, "step": 9825 }, { "epoch": 0.02, "learning_rate": 9.777895213649404e-05, "loss": 1.0371, "step": 9830 }, { "epoch": 0.02, "learning_rate": 9.777769389501234e-05, "loss": 1.0379, "step": 9835 }, { "epoch": 0.02, "learning_rate": 9.777643565353063e-05, "loss": 1.0376, "step": 9840 }, { "epoch": 0.02, "learning_rate": 9.777517741204892e-05, "loss": 1.0369, "step": 9845 }, { "epoch": 0.02, "learning_rate": 9.777391917056722e-05, "loss": 1.0383, "step": 9850 }, { "epoch": 0.02, "learning_rate": 9.777266092908551e-05, "loss": 1.0394, "step": 9855 }, { "epoch": 0.02, "learning_rate": 9.777140268760381e-05, "loss": 1.0384, "step": 9860 }, { "epoch": 0.02, "learning_rate": 9.77701444461221e-05, "loss": 1.0389, "step": 9865 }, { "epoch": 0.02, "learning_rate": 9.77688862046404e-05, "loss": 1.037, "step": 9870 }, { "epoch": 0.02, "learning_rate": 9.776762796315869e-05, "loss": 1.037, "step": 9875 }, { "epoch": 0.02, "learning_rate": 9.776636972167699e-05, "loss": 1.0397, "step": 9880 }, { "epoch": 0.02, "learning_rate": 9.776511148019528e-05, "loss": 1.0377, "step": 9885 }, { "epoch": 0.02, "learning_rate": 9.776385323871358e-05, "loss": 1.0365, "step": 9890 }, { "epoch": 0.02, "learning_rate": 9.776259499723187e-05, "loss": 1.037, "step": 9895 }, { "epoch": 0.02, "learning_rate": 9.776133675575017e-05, "loss": 1.0366, "step": 9900 }, { "epoch": 0.02, "learning_rate": 9.776007851426846e-05, "loss": 1.0371, "step": 9905 }, { "epoch": 0.02, "learning_rate": 9.775882027278676e-05, "loss": 1.0362, "step": 9910 }, { "epoch": 0.02, "learning_rate": 9.775756203130505e-05, "loss": 1.0365, "step": 9915 }, { "epoch": 0.02, "learning_rate": 9.775630378982334e-05, "loss": 1.037, "step": 9920 }, { "epoch": 0.02, "learning_rate": 9.775504554834164e-05, "loss": 1.0371, "step": 9925 }, { "epoch": 0.02, "learning_rate": 9.775378730685994e-05, "loss": 1.0364, "step": 9930 }, { "epoch": 0.02, "learning_rate": 9.775252906537823e-05, "loss": 1.0391, "step": 9935 }, { "epoch": 0.02, "learning_rate": 9.775127082389652e-05, "loss": 1.0376, "step": 9940 }, { "epoch": 0.02, "learning_rate": 9.775001258241482e-05, "loss": 1.0363, "step": 9945 }, { "epoch": 0.02, "learning_rate": 9.774875434093312e-05, "loss": 1.037, "step": 9950 }, { "epoch": 0.02, "learning_rate": 9.77474960994514e-05, "loss": 1.0379, "step": 9955 }, { "epoch": 0.03, "learning_rate": 9.77462378579697e-05, "loss": 1.0381, "step": 9960 }, { "epoch": 0.03, "learning_rate": 9.7744979616488e-05, "loss": 1.037, "step": 9965 }, { "epoch": 0.03, "learning_rate": 9.77437213750063e-05, "loss": 1.0368, "step": 9970 }, { "epoch": 0.03, "learning_rate": 9.774246313352459e-05, "loss": 1.036, "step": 9975 }, { "epoch": 0.03, "learning_rate": 9.774120489204288e-05, "loss": 1.036, "step": 9980 }, { "epoch": 0.03, "learning_rate": 9.773994665056117e-05, "loss": 1.0388, "step": 9985 }, { "epoch": 0.03, "learning_rate": 9.773868840907948e-05, "loss": 1.0361, "step": 9990 }, { "epoch": 0.03, "learning_rate": 9.773743016759777e-05, "loss": 1.039, "step": 9995 }, { "epoch": 0.03, "learning_rate": 9.773617192611606e-05, "loss": 1.0385, "step": 10000 }, { "epoch": 0.03, "learning_rate": 9.773491368463435e-05, "loss": 1.0369, "step": 10005 }, { "epoch": 0.03, "learning_rate": 9.773365544315265e-05, "loss": 1.0358, "step": 10010 }, { "epoch": 0.03, "learning_rate": 9.773239720167095e-05, "loss": 1.0393, "step": 10015 }, { "epoch": 0.03, "learning_rate": 9.773113896018924e-05, "loss": 1.0361, "step": 10020 }, { "epoch": 0.03, "learning_rate": 9.772988071870753e-05, "loss": 1.0382, "step": 10025 }, { "epoch": 0.03, "learning_rate": 9.772862247722583e-05, "loss": 1.0374, "step": 10030 }, { "epoch": 0.03, "learning_rate": 9.772736423574413e-05, "loss": 1.0371, "step": 10035 }, { "epoch": 0.03, "learning_rate": 9.772610599426243e-05, "loss": 1.0607, "step": 10040 }, { "epoch": 0.03, "learning_rate": 9.772484775278072e-05, "loss": 1.0379, "step": 10045 }, { "epoch": 0.03, "learning_rate": 9.772358951129901e-05, "loss": 1.04, "step": 10050 }, { "epoch": 0.03, "learning_rate": 9.77223312698173e-05, "loss": 1.0382, "step": 10055 }, { "epoch": 0.03, "learning_rate": 9.772107302833561e-05, "loss": 1.0368, "step": 10060 }, { "epoch": 0.03, "learning_rate": 9.77198147868539e-05, "loss": 1.0359, "step": 10065 }, { "epoch": 0.03, "learning_rate": 9.771880819366853e-05, "loss": 1.0389, "step": 10070 }, { "epoch": 0.03, "learning_rate": 9.771754995218683e-05, "loss": 1.06, "step": 10075 }, { "epoch": 0.03, "learning_rate": 9.771629171070512e-05, "loss": 1.0362, "step": 10080 }, { "epoch": 0.03, "learning_rate": 9.771503346922341e-05, "loss": 1.0383, "step": 10085 }, { "epoch": 0.03, "learning_rate": 9.77137752277417e-05, "loss": 1.0365, "step": 10090 }, { "epoch": 0.03, "learning_rate": 9.771251698626001e-05, "loss": 1.0371, "step": 10095 }, { "epoch": 0.03, "learning_rate": 9.77112587447783e-05, "loss": 1.0373, "step": 10100 }, { "epoch": 0.03, "learning_rate": 9.77100005032966e-05, "loss": 1.0353, "step": 10105 }, { "epoch": 0.03, "learning_rate": 9.770874226181489e-05, "loss": 1.0381, "step": 10110 }, { "epoch": 0.03, "learning_rate": 9.770748402033319e-05, "loss": 1.0376, "step": 10115 }, { "epoch": 0.03, "learning_rate": 9.770622577885148e-05, "loss": 1.0373, "step": 10120 }, { "epoch": 0.03, "learning_rate": 9.770496753736977e-05, "loss": 1.0375, "step": 10125 }, { "epoch": 0.03, "learning_rate": 9.770370929588807e-05, "loss": 1.0602, "step": 10130 }, { "epoch": 0.03, "learning_rate": 9.770245105440636e-05, "loss": 1.0361, "step": 10135 }, { "epoch": 0.03, "learning_rate": 9.770119281292466e-05, "loss": 1.0369, "step": 10140 }, { "epoch": 0.03, "learning_rate": 9.769993457144295e-05, "loss": 1.0386, "step": 10145 }, { "epoch": 0.03, "learning_rate": 9.769867632996125e-05, "loss": 1.0557, "step": 10150 }, { "epoch": 0.03, "learning_rate": 9.769741808847954e-05, "loss": 1.0373, "step": 10155 }, { "epoch": 0.03, "learning_rate": 9.769615984699784e-05, "loss": 1.0401, "step": 10160 }, { "epoch": 0.03, "learning_rate": 9.769490160551613e-05, "loss": 1.0388, "step": 10165 }, { "epoch": 0.03, "learning_rate": 9.769364336403443e-05, "loss": 1.0375, "step": 10170 }, { "epoch": 0.03, "learning_rate": 9.769238512255272e-05, "loss": 1.0363, "step": 10175 }, { "epoch": 0.03, "learning_rate": 9.769112688107102e-05, "loss": 1.0372, "step": 10180 }, { "epoch": 0.03, "learning_rate": 9.768986863958931e-05, "loss": 1.0389, "step": 10185 }, { "epoch": 0.03, "learning_rate": 9.76886103981076e-05, "loss": 1.0381, "step": 10190 }, { "epoch": 0.03, "learning_rate": 9.76873521566259e-05, "loss": 1.0384, "step": 10195 }, { "epoch": 0.03, "learning_rate": 9.768609391514419e-05, "loss": 1.0396, "step": 10200 }, { "epoch": 0.03, "learning_rate": 9.76848356736625e-05, "loss": 1.0387, "step": 10205 }, { "epoch": 0.03, "learning_rate": 9.768357743218079e-05, "loss": 1.0376, "step": 10210 }, { "epoch": 0.03, "learning_rate": 9.768231919069908e-05, "loss": 1.0366, "step": 10215 }, { "epoch": 0.03, "learning_rate": 9.768106094921737e-05, "loss": 1.0353, "step": 10220 }, { "epoch": 0.03, "learning_rate": 9.767980270773567e-05, "loss": 1.0364, "step": 10225 }, { "epoch": 0.03, "learning_rate": 9.767854446625397e-05, "loss": 1.0387, "step": 10230 }, { "epoch": 0.03, "learning_rate": 9.767728622477226e-05, "loss": 1.0385, "step": 10235 }, { "epoch": 0.03, "learning_rate": 9.767602798329055e-05, "loss": 1.0374, "step": 10240 }, { "epoch": 0.03, "learning_rate": 9.767476974180885e-05, "loss": 1.0371, "step": 10245 }, { "epoch": 0.03, "learning_rate": 9.767351150032716e-05, "loss": 1.0375, "step": 10250 }, { "epoch": 0.03, "learning_rate": 9.767225325884545e-05, "loss": 1.0381, "step": 10255 }, { "epoch": 0.03, "learning_rate": 9.767099501736374e-05, "loss": 1.037, "step": 10260 }, { "epoch": 0.03, "learning_rate": 9.766973677588203e-05, "loss": 1.0378, "step": 10265 }, { "epoch": 0.03, "learning_rate": 9.766847853440033e-05, "loss": 1.0379, "step": 10270 }, { "epoch": 0.03, "learning_rate": 9.766722029291863e-05, "loss": 1.0339, "step": 10275 }, { "epoch": 0.03, "learning_rate": 9.766596205143692e-05, "loss": 1.0393, "step": 10280 }, { "epoch": 0.03, "learning_rate": 9.766470380995521e-05, "loss": 1.0606, "step": 10285 }, { "epoch": 0.03, "learning_rate": 9.76634455684735e-05, "loss": 1.0377, "step": 10290 }, { "epoch": 0.03, "learning_rate": 9.766218732699181e-05, "loss": 1.0361, "step": 10295 }, { "epoch": 0.03, "learning_rate": 9.76609290855101e-05, "loss": 1.0364, "step": 10300 }, { "epoch": 0.03, "learning_rate": 9.76596708440284e-05, "loss": 1.0372, "step": 10305 }, { "epoch": 0.03, "learning_rate": 9.765841260254668e-05, "loss": 1.0348, "step": 10310 }, { "epoch": 0.03, "learning_rate": 9.765715436106499e-05, "loss": 1.039, "step": 10315 }, { "epoch": 0.03, "learning_rate": 9.765589611958328e-05, "loss": 1.0365, "step": 10320 }, { "epoch": 0.03, "learning_rate": 9.765463787810157e-05, "loss": 1.0377, "step": 10325 }, { "epoch": 0.03, "learning_rate": 9.765337963661986e-05, "loss": 1.0366, "step": 10330 }, { "epoch": 0.03, "learning_rate": 9.765212139513816e-05, "loss": 1.0366, "step": 10335 }, { "epoch": 0.03, "learning_rate": 9.765086315365646e-05, "loss": 1.0527, "step": 10340 }, { "epoch": 0.03, "learning_rate": 9.764960491217475e-05, "loss": 1.037, "step": 10345 }, { "epoch": 0.03, "learning_rate": 9.764834667069304e-05, "loss": 1.0376, "step": 10350 }, { "epoch": 0.03, "learning_rate": 9.764708842921134e-05, "loss": 1.0353, "step": 10355 }, { "epoch": 0.03, "learning_rate": 9.764583018772964e-05, "loss": 1.0379, "step": 10360 }, { "epoch": 0.03, "learning_rate": 9.764457194624793e-05, "loss": 1.0377, "step": 10365 }, { "epoch": 0.03, "learning_rate": 9.764331370476622e-05, "loss": 1.0376, "step": 10370 }, { "epoch": 0.03, "learning_rate": 9.764205546328452e-05, "loss": 1.0353, "step": 10375 }, { "epoch": 0.03, "learning_rate": 9.764079722180282e-05, "loss": 1.0369, "step": 10380 }, { "epoch": 0.03, "learning_rate": 9.763953898032111e-05, "loss": 1.0369, "step": 10385 }, { "epoch": 0.03, "learning_rate": 9.76382807388394e-05, "loss": 1.0382, "step": 10390 }, { "epoch": 0.03, "learning_rate": 9.76370224973577e-05, "loss": 1.0568, "step": 10395 }, { "epoch": 0.03, "learning_rate": 9.763576425587599e-05, "loss": 1.0373, "step": 10400 }, { "epoch": 0.03, "learning_rate": 9.763450601439429e-05, "loss": 1.0385, "step": 10405 }, { "epoch": 0.03, "learning_rate": 9.763324777291258e-05, "loss": 1.0387, "step": 10410 }, { "epoch": 0.03, "learning_rate": 9.763198953143088e-05, "loss": 1.038, "step": 10415 }, { "epoch": 0.03, "learning_rate": 9.763073128994917e-05, "loss": 1.0385, "step": 10420 }, { "epoch": 0.03, "learning_rate": 9.762947304846747e-05, "loss": 1.0633, "step": 10425 }, { "epoch": 0.03, "learning_rate": 9.762821480698576e-05, "loss": 1.0383, "step": 10430 }, { "epoch": 0.03, "learning_rate": 9.762695656550406e-05, "loss": 1.038, "step": 10435 }, { "epoch": 0.03, "learning_rate": 9.762569832402235e-05, "loss": 1.0364, "step": 10440 }, { "epoch": 0.03, "learning_rate": 9.762444008254065e-05, "loss": 1.0579, "step": 10445 }, { "epoch": 0.03, "learning_rate": 9.762318184105894e-05, "loss": 1.0373, "step": 10450 }, { "epoch": 0.03, "learning_rate": 9.762192359957724e-05, "loss": 1.0362, "step": 10455 }, { "epoch": 0.03, "learning_rate": 9.762066535809553e-05, "loss": 1.0385, "step": 10460 }, { "epoch": 0.03, "learning_rate": 9.761940711661382e-05, "loss": 1.0504, "step": 10465 }, { "epoch": 0.03, "learning_rate": 9.761814887513212e-05, "loss": 1.0377, "step": 10470 }, { "epoch": 0.03, "learning_rate": 9.761689063365042e-05, "loss": 1.0363, "step": 10475 }, { "epoch": 0.03, "learning_rate": 9.761563239216871e-05, "loss": 1.0351, "step": 10480 }, { "epoch": 0.03, "learning_rate": 9.7614374150687e-05, "loss": 1.037, "step": 10485 }, { "epoch": 0.03, "learning_rate": 9.76131159092053e-05, "loss": 1.0362, "step": 10490 }, { "epoch": 0.03, "learning_rate": 9.76118576677236e-05, "loss": 1.0382, "step": 10495 }, { "epoch": 0.03, "learning_rate": 9.761059942624189e-05, "loss": 1.0366, "step": 10500 }, { "epoch": 0.03, "learning_rate": 9.760934118476018e-05, "loss": 1.0371, "step": 10505 }, { "epoch": 0.03, "learning_rate": 9.760808294327848e-05, "loss": 1.0364, "step": 10510 }, { "epoch": 0.03, "learning_rate": 9.760682470179678e-05, "loss": 1.0375, "step": 10515 }, { "epoch": 0.03, "learning_rate": 9.760556646031507e-05, "loss": 1.0604, "step": 10520 }, { "epoch": 0.03, "learning_rate": 9.760430821883336e-05, "loss": 1.0387, "step": 10525 }, { "epoch": 0.03, "learning_rate": 9.760304997735165e-05, "loss": 1.0369, "step": 10530 }, { "epoch": 0.03, "learning_rate": 9.760179173586996e-05, "loss": 1.0372, "step": 10535 }, { "epoch": 0.03, "learning_rate": 9.760053349438825e-05, "loss": 1.0365, "step": 10540 }, { "epoch": 0.03, "learning_rate": 9.759927525290654e-05, "loss": 1.0374, "step": 10545 }, { "epoch": 0.03, "learning_rate": 9.759801701142483e-05, "loss": 1.0379, "step": 10550 }, { "epoch": 0.03, "learning_rate": 9.759675876994314e-05, "loss": 1.0389, "step": 10555 }, { "epoch": 0.03, "learning_rate": 9.759550052846143e-05, "loss": 1.0364, "step": 10560 }, { "epoch": 0.03, "learning_rate": 9.759424228697972e-05, "loss": 1.0363, "step": 10565 }, { "epoch": 0.03, "learning_rate": 9.759298404549801e-05, "loss": 1.0365, "step": 10570 }, { "epoch": 0.03, "learning_rate": 9.759172580401632e-05, "loss": 1.0368, "step": 10575 }, { "epoch": 0.03, "learning_rate": 9.759046756253461e-05, "loss": 1.0363, "step": 10580 }, { "epoch": 0.03, "learning_rate": 9.75892093210529e-05, "loss": 1.0361, "step": 10585 }, { "epoch": 0.03, "learning_rate": 9.758795107957119e-05, "loss": 1.0384, "step": 10590 }, { "epoch": 0.03, "learning_rate": 9.758669283808948e-05, "loss": 1.0366, "step": 10595 }, { "epoch": 0.03, "learning_rate": 9.758543459660779e-05, "loss": 1.0385, "step": 10600 }, { "epoch": 0.03, "learning_rate": 9.758417635512608e-05, "loss": 1.0371, "step": 10605 }, { "epoch": 0.03, "learning_rate": 9.758291811364437e-05, "loss": 1.0379, "step": 10610 }, { "epoch": 0.03, "learning_rate": 9.758165987216266e-05, "loss": 1.0377, "step": 10615 }, { "epoch": 0.03, "learning_rate": 9.758040163068097e-05, "loss": 1.0378, "step": 10620 }, { "epoch": 0.03, "learning_rate": 9.757914338919926e-05, "loss": 1.0356, "step": 10625 }, { "epoch": 0.03, "learning_rate": 9.757788514771755e-05, "loss": 1.0357, "step": 10630 }, { "epoch": 0.03, "learning_rate": 9.757662690623584e-05, "loss": 1.0391, "step": 10635 }, { "epoch": 0.03, "learning_rate": 9.757536866475415e-05, "loss": 1.0361, "step": 10640 }, { "epoch": 0.03, "learning_rate": 9.757411042327244e-05, "loss": 1.0372, "step": 10645 }, { "epoch": 0.03, "learning_rate": 9.757285218179073e-05, "loss": 1.036, "step": 10650 }, { "epoch": 0.03, "learning_rate": 9.757159394030902e-05, "loss": 1.0382, "step": 10655 }, { "epoch": 0.03, "learning_rate": 9.757033569882731e-05, "loss": 1.0371, "step": 10660 }, { "epoch": 0.03, "learning_rate": 9.756907745734562e-05, "loss": 1.0373, "step": 10665 }, { "epoch": 0.03, "learning_rate": 9.756781921586391e-05, "loss": 1.0395, "step": 10670 }, { "epoch": 0.03, "learning_rate": 9.75665609743822e-05, "loss": 1.0371, "step": 10675 }, { "epoch": 0.03, "learning_rate": 9.75653027329005e-05, "loss": 1.0373, "step": 10680 }, { "epoch": 0.03, "learning_rate": 9.75640444914188e-05, "loss": 1.0383, "step": 10685 }, { "epoch": 0.03, "learning_rate": 9.756278624993709e-05, "loss": 1.0385, "step": 10690 }, { "epoch": 0.03, "learning_rate": 9.756152800845538e-05, "loss": 1.0397, "step": 10695 }, { "epoch": 0.03, "learning_rate": 9.756026976697367e-05, "loss": 1.0359, "step": 10700 }, { "epoch": 0.03, "learning_rate": 9.755901152549198e-05, "loss": 1.0382, "step": 10705 }, { "epoch": 0.03, "learning_rate": 9.755775328401027e-05, "loss": 1.0382, "step": 10710 }, { "epoch": 0.03, "learning_rate": 9.755649504252856e-05, "loss": 1.035, "step": 10715 }, { "epoch": 0.03, "learning_rate": 9.755523680104685e-05, "loss": 1.0361, "step": 10720 }, { "epoch": 0.03, "learning_rate": 9.755397855956515e-05, "loss": 1.0366, "step": 10725 }, { "epoch": 0.03, "learning_rate": 9.755272031808345e-05, "loss": 1.0387, "step": 10730 }, { "epoch": 0.03, "learning_rate": 9.755146207660174e-05, "loss": 1.0365, "step": 10735 }, { "epoch": 0.03, "learning_rate": 9.755020383512005e-05, "loss": 1.0389, "step": 10740 }, { "epoch": 0.03, "learning_rate": 9.754894559363834e-05, "loss": 1.0382, "step": 10745 }, { "epoch": 0.03, "learning_rate": 9.754768735215663e-05, "loss": 1.0345, "step": 10750 }, { "epoch": 0.03, "learning_rate": 9.754642911067494e-05, "loss": 1.0367, "step": 10755 }, { "epoch": 0.03, "learning_rate": 9.754517086919323e-05, "loss": 1.0385, "step": 10760 }, { "epoch": 0.03, "learning_rate": 9.754391262771152e-05, "loss": 1.0367, "step": 10765 }, { "epoch": 0.03, "learning_rate": 9.754265438622981e-05, "loss": 1.0376, "step": 10770 }, { "epoch": 0.03, "learning_rate": 9.754139614474812e-05, "loss": 1.0623, "step": 10775 }, { "epoch": 0.03, "learning_rate": 9.754013790326641e-05, "loss": 1.0363, "step": 10780 }, { "epoch": 0.03, "learning_rate": 9.75388796617847e-05, "loss": 1.0386, "step": 10785 }, { "epoch": 0.03, "learning_rate": 9.753762142030299e-05, "loss": 1.0374, "step": 10790 }, { "epoch": 0.03, "learning_rate": 9.753636317882128e-05, "loss": 1.0374, "step": 10795 }, { "epoch": 0.03, "learning_rate": 9.753510493733959e-05, "loss": 1.0365, "step": 10800 }, { "epoch": 0.03, "learning_rate": 9.753384669585788e-05, "loss": 1.0373, "step": 10805 }, { "epoch": 0.03, "learning_rate": 9.753258845437617e-05, "loss": 1.036, "step": 10810 }, { "epoch": 0.03, "learning_rate": 9.753133021289446e-05, "loss": 1.0601, "step": 10815 }, { "epoch": 0.03, "learning_rate": 9.753007197141277e-05, "loss": 1.0355, "step": 10820 }, { "epoch": 0.03, "learning_rate": 9.752881372993106e-05, "loss": 1.0361, "step": 10825 }, { "epoch": 0.03, "learning_rate": 9.752755548844935e-05, "loss": 1.0367, "step": 10830 }, { "epoch": 0.03, "learning_rate": 9.752629724696764e-05, "loss": 1.0355, "step": 10835 }, { "epoch": 0.03, "learning_rate": 9.752503900548595e-05, "loss": 1.0379, "step": 10840 }, { "epoch": 0.03, "learning_rate": 9.752378076400424e-05, "loss": 1.0376, "step": 10845 }, { "epoch": 0.03, "learning_rate": 9.752252252252253e-05, "loss": 1.0349, "step": 10850 }, { "epoch": 0.03, "learning_rate": 9.752126428104082e-05, "loss": 1.0382, "step": 10855 }, { "epoch": 0.03, "learning_rate": 9.752000603955911e-05, "loss": 1.0381, "step": 10860 }, { "epoch": 0.03, "learning_rate": 9.751874779807742e-05, "loss": 1.0376, "step": 10865 }, { "epoch": 0.03, "learning_rate": 9.751748955659571e-05, "loss": 1.0391, "step": 10870 }, { "epoch": 0.03, "learning_rate": 9.7516231315114e-05, "loss": 1.0357, "step": 10875 }, { "epoch": 0.03, "learning_rate": 9.751497307363229e-05, "loss": 1.0366, "step": 10880 }, { "epoch": 0.03, "learning_rate": 9.75137148321506e-05, "loss": 1.0349, "step": 10885 }, { "epoch": 0.03, "learning_rate": 9.751245659066889e-05, "loss": 1.0377, "step": 10890 }, { "epoch": 0.03, "learning_rate": 9.751119834918718e-05, "loss": 1.0375, "step": 10895 }, { "epoch": 0.03, "learning_rate": 9.750994010770547e-05, "loss": 1.0365, "step": 10900 }, { "epoch": 0.03, "learning_rate": 9.750868186622378e-05, "loss": 1.037, "step": 10905 }, { "epoch": 0.03, "learning_rate": 9.750742362474207e-05, "loss": 1.0384, "step": 10910 }, { "epoch": 0.03, "learning_rate": 9.750616538326036e-05, "loss": 1.0382, "step": 10915 }, { "epoch": 0.03, "learning_rate": 9.750490714177865e-05, "loss": 1.0359, "step": 10920 }, { "epoch": 0.03, "learning_rate": 9.750364890029694e-05, "loss": 1.0389, "step": 10925 }, { "epoch": 0.03, "learning_rate": 9.750239065881525e-05, "loss": 1.0357, "step": 10930 }, { "epoch": 0.03, "learning_rate": 9.750113241733354e-05, "loss": 1.038, "step": 10935 }, { "epoch": 0.03, "learning_rate": 9.749987417585183e-05, "loss": 1.0359, "step": 10940 }, { "epoch": 0.03, "learning_rate": 9.749861593437012e-05, "loss": 1.0369, "step": 10945 }, { "epoch": 0.03, "learning_rate": 9.749735769288843e-05, "loss": 1.0376, "step": 10950 }, { "epoch": 0.03, "learning_rate": 9.749609945140672e-05, "loss": 1.0349, "step": 10955 }, { "epoch": 0.03, "learning_rate": 9.749484120992501e-05, "loss": 1.0384, "step": 10960 }, { "epoch": 0.03, "learning_rate": 9.74935829684433e-05, "loss": 1.0364, "step": 10965 }, { "epoch": 0.03, "learning_rate": 9.749232472696161e-05, "loss": 1.0363, "step": 10970 }, { "epoch": 0.03, "learning_rate": 9.74910664854799e-05, "loss": 1.038, "step": 10975 }, { "epoch": 0.03, "learning_rate": 9.748980824399819e-05, "loss": 1.0381, "step": 10980 }, { "epoch": 0.03, "learning_rate": 9.748855000251648e-05, "loss": 1.0374, "step": 10985 }, { "epoch": 0.03, "learning_rate": 9.748729176103478e-05, "loss": 1.0376, "step": 10990 }, { "epoch": 0.03, "learning_rate": 9.748603351955308e-05, "loss": 1.0353, "step": 10995 }, { "epoch": 0.03, "learning_rate": 9.748477527807137e-05, "loss": 1.0385, "step": 11000 }, { "epoch": 0.03, "learning_rate": 9.748351703658966e-05, "loss": 1.0359, "step": 11005 }, { "epoch": 0.03, "learning_rate": 9.748225879510796e-05, "loss": 1.0349, "step": 11010 }, { "epoch": 0.03, "learning_rate": 9.748100055362626e-05, "loss": 1.0372, "step": 11015 }, { "epoch": 0.03, "learning_rate": 9.747974231214455e-05, "loss": 1.036, "step": 11020 }, { "epoch": 0.03, "learning_rate": 9.747848407066284e-05, "loss": 1.0348, "step": 11025 }, { "epoch": 0.03, "learning_rate": 9.747722582918114e-05, "loss": 1.0384, "step": 11030 }, { "epoch": 0.03, "learning_rate": 9.747596758769944e-05, "loss": 1.0368, "step": 11035 }, { "epoch": 0.03, "learning_rate": 9.747470934621773e-05, "loss": 1.0365, "step": 11040 }, { "epoch": 0.03, "learning_rate": 9.747345110473602e-05, "loss": 1.0368, "step": 11045 }, { "epoch": 0.03, "learning_rate": 9.747219286325432e-05, "loss": 1.0376, "step": 11050 }, { "epoch": 0.03, "learning_rate": 9.747093462177261e-05, "loss": 1.0357, "step": 11055 }, { "epoch": 0.03, "learning_rate": 9.746967638029091e-05, "loss": 1.0394, "step": 11060 }, { "epoch": 0.03, "learning_rate": 9.74684181388092e-05, "loss": 1.0371, "step": 11065 }, { "epoch": 0.03, "learning_rate": 9.74671598973275e-05, "loss": 1.0618, "step": 11070 }, { "epoch": 0.03, "learning_rate": 9.746590165584579e-05, "loss": 1.0373, "step": 11075 }, { "epoch": 0.03, "learning_rate": 9.746464341436409e-05, "loss": 1.0374, "step": 11080 }, { "epoch": 0.03, "learning_rate": 9.746338517288238e-05, "loss": 1.0364, "step": 11085 }, { "epoch": 0.03, "learning_rate": 9.746212693140068e-05, "loss": 1.035, "step": 11090 }, { "epoch": 0.03, "learning_rate": 9.746086868991897e-05, "loss": 1.0371, "step": 11095 }, { "epoch": 0.03, "learning_rate": 9.745961044843727e-05, "loss": 1.0382, "step": 11100 }, { "epoch": 0.03, "learning_rate": 9.745835220695556e-05, "loss": 1.0358, "step": 11105 }, { "epoch": 0.03, "learning_rate": 9.745709396547386e-05, "loss": 1.0362, "step": 11110 }, { "epoch": 0.03, "learning_rate": 9.745583572399215e-05, "loss": 1.0339, "step": 11115 }, { "epoch": 0.03, "learning_rate": 9.745457748251044e-05, "loss": 1.0354, "step": 11120 }, { "epoch": 0.03, "learning_rate": 9.745331924102874e-05, "loss": 1.035, "step": 11125 }, { "epoch": 0.03, "learning_rate": 9.745206099954704e-05, "loss": 1.038, "step": 11130 }, { "epoch": 0.03, "learning_rate": 9.745080275806533e-05, "loss": 1.0364, "step": 11135 }, { "epoch": 0.03, "learning_rate": 9.744954451658362e-05, "loss": 1.0359, "step": 11140 }, { "epoch": 0.03, "learning_rate": 9.744828627510192e-05, "loss": 1.0376, "step": 11145 }, { "epoch": 0.03, "learning_rate": 9.744702803362022e-05, "loss": 1.0372, "step": 11150 }, { "epoch": 0.03, "learning_rate": 9.744576979213851e-05, "loss": 1.0369, "step": 11155 }, { "epoch": 0.03, "learning_rate": 9.74445115506568e-05, "loss": 1.038, "step": 11160 }, { "epoch": 0.03, "learning_rate": 9.74432533091751e-05, "loss": 1.0383, "step": 11165 }, { "epoch": 0.03, "learning_rate": 9.74419950676934e-05, "loss": 1.0368, "step": 11170 }, { "epoch": 0.03, "learning_rate": 9.744073682621169e-05, "loss": 1.035, "step": 11175 }, { "epoch": 0.03, "learning_rate": 9.743947858472998e-05, "loss": 1.038, "step": 11180 }, { "epoch": 0.03, "learning_rate": 9.743822034324827e-05, "loss": 1.038, "step": 11185 }, { "epoch": 0.03, "learning_rate": 9.743696210176658e-05, "loss": 1.0379, "step": 11190 }, { "epoch": 0.03, "learning_rate": 9.743570386028487e-05, "loss": 1.0383, "step": 11195 }, { "epoch": 0.03, "learning_rate": 9.743444561880316e-05, "loss": 1.0368, "step": 11200 }, { "epoch": 0.03, "learning_rate": 9.743318737732145e-05, "loss": 1.0365, "step": 11205 }, { "epoch": 0.03, "learning_rate": 9.743192913583976e-05, "loss": 1.0372, "step": 11210 }, { "epoch": 0.03, "learning_rate": 9.743067089435805e-05, "loss": 1.0372, "step": 11215 }, { "epoch": 0.03, "learning_rate": 9.742941265287634e-05, "loss": 1.0384, "step": 11220 }, { "epoch": 0.03, "learning_rate": 9.742815441139463e-05, "loss": 1.0365, "step": 11225 }, { "epoch": 0.03, "learning_rate": 9.742689616991294e-05, "loss": 1.058, "step": 11230 }, { "epoch": 0.03, "learning_rate": 9.742563792843123e-05, "loss": 1.0367, "step": 11235 }, { "epoch": 0.03, "learning_rate": 9.742437968694953e-05, "loss": 1.0453, "step": 11240 }, { "epoch": 0.03, "learning_rate": 9.742312144546782e-05, "loss": 1.0369, "step": 11245 }, { "epoch": 0.03, "learning_rate": 9.742186320398612e-05, "loss": 1.0357, "step": 11250 }, { "epoch": 0.03, "learning_rate": 9.742060496250442e-05, "loss": 1.0378, "step": 11255 }, { "epoch": 0.03, "learning_rate": 9.741934672102271e-05, "loss": 1.0391, "step": 11260 }, { "epoch": 0.03, "learning_rate": 9.7418088479541e-05, "loss": 1.0389, "step": 11265 }, { "epoch": 0.03, "learning_rate": 9.74168302380593e-05, "loss": 1.0358, "step": 11270 }, { "epoch": 0.03, "learning_rate": 9.741557199657759e-05, "loss": 1.0369, "step": 11275 }, { "epoch": 0.03, "learning_rate": 9.741431375509589e-05, "loss": 1.0358, "step": 11280 }, { "epoch": 0.03, "learning_rate": 9.741305551361418e-05, "loss": 1.0378, "step": 11285 }, { "epoch": 0.03, "learning_rate": 9.741179727213248e-05, "loss": 1.0367, "step": 11290 }, { "epoch": 0.03, "learning_rate": 9.741053903065077e-05, "loss": 1.0367, "step": 11295 }, { "epoch": 0.03, "learning_rate": 9.740928078916907e-05, "loss": 1.0375, "step": 11300 }, { "epoch": 0.03, "learning_rate": 9.740802254768736e-05, "loss": 1.0375, "step": 11305 }, { "epoch": 0.03, "learning_rate": 9.740676430620565e-05, "loss": 1.0353, "step": 11310 }, { "epoch": 0.03, "learning_rate": 9.740550606472395e-05, "loss": 1.0361, "step": 11315 }, { "epoch": 0.03, "learning_rate": 9.740424782324225e-05, "loss": 1.0361, "step": 11320 }, { "epoch": 0.03, "learning_rate": 9.740298958176054e-05, "loss": 1.0393, "step": 11325 }, { "epoch": 0.03, "learning_rate": 9.740173134027883e-05, "loss": 1.037, "step": 11330 }, { "epoch": 0.03, "learning_rate": 9.740047309879713e-05, "loss": 1.0382, "step": 11335 }, { "epoch": 0.03, "learning_rate": 9.739921485731542e-05, "loss": 1.0347, "step": 11340 }, { "epoch": 0.03, "learning_rate": 9.739795661583372e-05, "loss": 1.0384, "step": 11345 }, { "epoch": 0.03, "learning_rate": 9.739669837435201e-05, "loss": 1.0359, "step": 11350 }, { "epoch": 0.03, "learning_rate": 9.73954401328703e-05, "loss": 1.0372, "step": 11355 }, { "epoch": 0.03, "learning_rate": 9.73941818913886e-05, "loss": 1.0376, "step": 11360 }, { "epoch": 0.03, "learning_rate": 9.73929236499069e-05, "loss": 1.0356, "step": 11365 }, { "epoch": 0.03, "learning_rate": 9.73916654084252e-05, "loss": 1.036, "step": 11370 }, { "epoch": 0.03, "learning_rate": 9.739040716694349e-05, "loss": 1.0353, "step": 11375 }, { "epoch": 0.03, "learning_rate": 9.738914892546178e-05, "loss": 1.0383, "step": 11380 }, { "epoch": 0.03, "learning_rate": 9.738789068398008e-05, "loss": 1.0387, "step": 11385 }, { "epoch": 0.03, "learning_rate": 9.738663244249837e-05, "loss": 1.0367, "step": 11390 }, { "epoch": 0.03, "learning_rate": 9.738537420101667e-05, "loss": 1.0368, "step": 11395 }, { "epoch": 0.03, "learning_rate": 9.738411595953496e-05, "loss": 1.0344, "step": 11400 }, { "epoch": 0.03, "learning_rate": 9.738285771805325e-05, "loss": 1.0365, "step": 11405 }, { "epoch": 0.03, "learning_rate": 9.738159947657155e-05, "loss": 1.0381, "step": 11410 }, { "epoch": 0.03, "learning_rate": 9.738034123508985e-05, "loss": 1.036, "step": 11415 }, { "epoch": 0.03, "learning_rate": 9.737908299360814e-05, "loss": 1.0368, "step": 11420 }, { "epoch": 0.03, "learning_rate": 9.737782475212643e-05, "loss": 1.0343, "step": 11425 }, { "epoch": 0.03, "learning_rate": 9.737656651064473e-05, "loss": 1.0373, "step": 11430 }, { "epoch": 0.03, "learning_rate": 9.737530826916303e-05, "loss": 1.037, "step": 11435 }, { "epoch": 0.03, "learning_rate": 9.737405002768132e-05, "loss": 1.0425, "step": 11440 }, { "epoch": 0.03, "learning_rate": 9.737279178619961e-05, "loss": 1.037, "step": 11445 }, { "epoch": 0.03, "learning_rate": 9.73715335447179e-05, "loss": 1.0453, "step": 11450 }, { "epoch": 0.03, "learning_rate": 9.73702753032362e-05, "loss": 1.0354, "step": 11455 }, { "epoch": 0.03, "learning_rate": 9.73690170617545e-05, "loss": 1.0371, "step": 11460 }, { "epoch": 0.03, "learning_rate": 9.736775882027279e-05, "loss": 1.0614, "step": 11465 }, { "epoch": 0.03, "learning_rate": 9.736650057879108e-05, "loss": 1.0381, "step": 11470 }, { "epoch": 0.03, "learning_rate": 9.736524233730939e-05, "loss": 1.038, "step": 11475 }, { "epoch": 0.03, "learning_rate": 9.736398409582768e-05, "loss": 1.0364, "step": 11480 }, { "epoch": 0.03, "learning_rate": 9.736272585434597e-05, "loss": 1.0347, "step": 11485 }, { "epoch": 0.03, "learning_rate": 9.736146761286426e-05, "loss": 1.0377, "step": 11490 }, { "epoch": 0.03, "learning_rate": 9.736020937138257e-05, "loss": 1.0354, "step": 11495 }, { "epoch": 0.03, "learning_rate": 9.735895112990086e-05, "loss": 1.0348, "step": 11500 }, { "epoch": 0.03, "learning_rate": 9.735769288841915e-05, "loss": 1.0355, "step": 11505 }, { "epoch": 0.03, "learning_rate": 9.735643464693744e-05, "loss": 1.0361, "step": 11510 }, { "epoch": 0.03, "learning_rate": 9.735517640545573e-05, "loss": 1.0388, "step": 11515 }, { "epoch": 0.03, "learning_rate": 9.735391816397404e-05, "loss": 1.0365, "step": 11520 }, { "epoch": 0.03, "learning_rate": 9.735265992249233e-05, "loss": 1.0351, "step": 11525 }, { "epoch": 0.03, "learning_rate": 9.735140168101062e-05, "loss": 1.0389, "step": 11530 }, { "epoch": 0.03, "learning_rate": 9.735014343952891e-05, "loss": 1.0362, "step": 11535 }, { "epoch": 0.03, "learning_rate": 9.734888519804722e-05, "loss": 1.0397, "step": 11540 }, { "epoch": 0.03, "learning_rate": 9.734762695656551e-05, "loss": 1.0362, "step": 11545 }, { "epoch": 0.03, "learning_rate": 9.73463687150838e-05, "loss": 1.036, "step": 11550 }, { "epoch": 0.03, "learning_rate": 9.734511047360209e-05, "loss": 1.0395, "step": 11555 }, { "epoch": 0.03, "learning_rate": 9.73438522321204e-05, "loss": 1.0385, "step": 11560 }, { "epoch": 0.03, "learning_rate": 9.734259399063869e-05, "loss": 1.0358, "step": 11565 }, { "epoch": 0.03, "learning_rate": 9.734133574915698e-05, "loss": 1.0368, "step": 11570 }, { "epoch": 0.03, "learning_rate": 9.734007750767527e-05, "loss": 1.0371, "step": 11575 }, { "epoch": 0.03, "learning_rate": 9.733881926619356e-05, "loss": 1.0388, "step": 11580 }, { "epoch": 0.03, "learning_rate": 9.733756102471187e-05, "loss": 1.0374, "step": 11585 }, { "epoch": 0.03, "learning_rate": 9.733630278323016e-05, "loss": 1.0355, "step": 11590 }, { "epoch": 0.03, "learning_rate": 9.733504454174845e-05, "loss": 1.0377, "step": 11595 }, { "epoch": 0.03, "learning_rate": 9.733378630026674e-05, "loss": 1.0373, "step": 11600 }, { "epoch": 0.03, "learning_rate": 9.733252805878505e-05, "loss": 1.0383, "step": 11605 }, { "epoch": 0.03, "learning_rate": 9.733126981730334e-05, "loss": 1.0371, "step": 11610 }, { "epoch": 0.03, "learning_rate": 9.733001157582163e-05, "loss": 1.0355, "step": 11615 }, { "epoch": 0.03, "learning_rate": 9.732875333433992e-05, "loss": 1.0383, "step": 11620 }, { "epoch": 0.03, "learning_rate": 9.732749509285823e-05, "loss": 1.0364, "step": 11625 }, { "epoch": 0.03, "learning_rate": 9.732623685137652e-05, "loss": 1.037, "step": 11630 }, { "epoch": 0.03, "learning_rate": 9.732497860989481e-05, "loss": 1.0372, "step": 11635 }, { "epoch": 0.03, "learning_rate": 9.73237203684131e-05, "loss": 1.0379, "step": 11640 }, { "epoch": 0.03, "learning_rate": 9.73224621269314e-05, "loss": 1.0375, "step": 11645 }, { "epoch": 0.03, "learning_rate": 9.73212038854497e-05, "loss": 1.0363, "step": 11650 }, { "epoch": 0.03, "learning_rate": 9.731994564396799e-05, "loss": 1.0377, "step": 11655 }, { "epoch": 0.03, "learning_rate": 9.731868740248628e-05, "loss": 1.0349, "step": 11660 }, { "epoch": 0.03, "learning_rate": 9.731742916100458e-05, "loss": 1.0353, "step": 11665 }, { "epoch": 0.03, "learning_rate": 9.731617091952288e-05, "loss": 1.0363, "step": 11670 }, { "epoch": 0.03, "learning_rate": 9.731491267804117e-05, "loss": 1.0358, "step": 11675 }, { "epoch": 0.03, "learning_rate": 9.731365443655946e-05, "loss": 1.0355, "step": 11680 }, { "epoch": 0.03, "learning_rate": 9.731239619507776e-05, "loss": 1.037, "step": 11685 }, { "epoch": 0.03, "learning_rate": 9.731113795359606e-05, "loss": 1.0394, "step": 11690 }, { "epoch": 0.03, "learning_rate": 9.730987971211435e-05, "loss": 1.0348, "step": 11695 }, { "epoch": 0.03, "learning_rate": 9.730862147063264e-05, "loss": 1.0384, "step": 11700 }, { "epoch": 0.03, "learning_rate": 9.730736322915094e-05, "loss": 1.0356, "step": 11705 }, { "epoch": 0.03, "learning_rate": 9.730610498766923e-05, "loss": 1.0359, "step": 11710 }, { "epoch": 0.03, "learning_rate": 9.730484674618753e-05, "loss": 1.0343, "step": 11715 }, { "epoch": 0.03, "learning_rate": 9.730358850470582e-05, "loss": 1.0378, "step": 11720 }, { "epoch": 0.03, "learning_rate": 9.730233026322412e-05, "loss": 1.0382, "step": 11725 }, { "epoch": 0.03, "learning_rate": 9.73010720217424e-05, "loss": 1.036, "step": 11730 }, { "epoch": 0.03, "learning_rate": 9.729981378026071e-05, "loss": 1.0369, "step": 11735 }, { "epoch": 0.03, "learning_rate": 9.729855553877902e-05, "loss": 1.0379, "step": 11740 }, { "epoch": 0.03, "learning_rate": 9.729729729729731e-05, "loss": 1.0378, "step": 11745 }, { "epoch": 0.03, "learning_rate": 9.72960390558156e-05, "loss": 1.038, "step": 11750 }, { "epoch": 0.03, "learning_rate": 9.729478081433389e-05, "loss": 1.0373, "step": 11755 }, { "epoch": 0.03, "learning_rate": 9.72935225728522e-05, "loss": 1.0343, "step": 11760 }, { "epoch": 0.03, "learning_rate": 9.729226433137049e-05, "loss": 1.0371, "step": 11765 }, { "epoch": 0.03, "learning_rate": 9.729100608988878e-05, "loss": 1.0372, "step": 11770 }, { "epoch": 0.03, "learning_rate": 9.728974784840707e-05, "loss": 1.0355, "step": 11775 }, { "epoch": 0.03, "learning_rate": 9.728848960692538e-05, "loss": 1.0339, "step": 11780 }, { "epoch": 0.03, "learning_rate": 9.728723136544367e-05, "loss": 1.0372, "step": 11785 }, { "epoch": 0.03, "learning_rate": 9.728597312396196e-05, "loss": 1.037, "step": 11790 }, { "epoch": 0.03, "learning_rate": 9.728471488248025e-05, "loss": 1.0361, "step": 11795 }, { "epoch": 0.03, "learning_rate": 9.728345664099854e-05, "loss": 1.0357, "step": 11800 }, { "epoch": 0.03, "learning_rate": 9.728219839951685e-05, "loss": 1.0352, "step": 11805 }, { "epoch": 0.03, "learning_rate": 9.728094015803514e-05, "loss": 1.0363, "step": 11810 }, { "epoch": 0.03, "learning_rate": 9.727968191655343e-05, "loss": 1.038, "step": 11815 }, { "epoch": 0.03, "learning_rate": 9.727842367507172e-05, "loss": 1.0527, "step": 11820 }, { "epoch": 0.03, "learning_rate": 9.727716543359003e-05, "loss": 1.0368, "step": 11825 }, { "epoch": 0.03, "learning_rate": 9.727590719210832e-05, "loss": 1.0347, "step": 11830 }, { "epoch": 0.03, "learning_rate": 9.727464895062661e-05, "loss": 1.0363, "step": 11835 }, { "epoch": 0.03, "learning_rate": 9.72733907091449e-05, "loss": 1.0357, "step": 11840 }, { "epoch": 0.03, "learning_rate": 9.727213246766321e-05, "loss": 1.0382, "step": 11845 }, { "epoch": 0.03, "learning_rate": 9.72708742261815e-05, "loss": 1.0357, "step": 11850 }, { "epoch": 0.03, "learning_rate": 9.726961598469979e-05, "loss": 1.0352, "step": 11855 }, { "epoch": 0.03, "learning_rate": 9.726835774321808e-05, "loss": 1.0365, "step": 11860 }, { "epoch": 0.03, "learning_rate": 9.726709950173637e-05, "loss": 1.037, "step": 11865 }, { "epoch": 0.03, "learning_rate": 9.726584126025468e-05, "loss": 1.036, "step": 11870 }, { "epoch": 0.03, "learning_rate": 9.726458301877297e-05, "loss": 1.0391, "step": 11875 }, { "epoch": 0.03, "learning_rate": 9.726332477729126e-05, "loss": 1.0572, "step": 11880 }, { "epoch": 0.03, "learning_rate": 9.726206653580955e-05, "loss": 1.0371, "step": 11885 }, { "epoch": 0.03, "learning_rate": 9.726080829432786e-05, "loss": 1.0362, "step": 11890 }, { "epoch": 0.03, "learning_rate": 9.725955005284615e-05, "loss": 1.0366, "step": 11895 }, { "epoch": 0.03, "learning_rate": 9.725829181136444e-05, "loss": 1.0357, "step": 11900 }, { "epoch": 0.03, "learning_rate": 9.725703356988273e-05, "loss": 1.0368, "step": 11905 }, { "epoch": 0.03, "learning_rate": 9.725577532840104e-05, "loss": 1.0614, "step": 11910 }, { "epoch": 0.03, "learning_rate": 9.725451708691933e-05, "loss": 1.0366, "step": 11915 }, { "epoch": 0.03, "learning_rate": 9.725325884543762e-05, "loss": 1.0373, "step": 11920 }, { "epoch": 0.03, "learning_rate": 9.725200060395591e-05, "loss": 1.037, "step": 11925 }, { "epoch": 0.03, "learning_rate": 9.72507423624742e-05, "loss": 1.0393, "step": 11930 }, { "epoch": 0.03, "learning_rate": 9.724948412099251e-05, "loss": 1.0377, "step": 11935 }, { "epoch": 0.03, "learning_rate": 9.72482258795108e-05, "loss": 1.0368, "step": 11940 }, { "epoch": 0.03, "learning_rate": 9.72469676380291e-05, "loss": 1.0346, "step": 11945 }, { "epoch": 0.03, "learning_rate": 9.724570939654739e-05, "loss": 1.0379, "step": 11950 }, { "epoch": 0.03, "learning_rate": 9.724445115506569e-05, "loss": 1.0354, "step": 11955 }, { "epoch": 0.03, "learning_rate": 9.724319291358398e-05, "loss": 1.0365, "step": 11960 }, { "epoch": 0.03, "learning_rate": 9.724193467210227e-05, "loss": 1.0369, "step": 11965 }, { "epoch": 0.03, "learning_rate": 9.724067643062057e-05, "loss": 1.036, "step": 11970 }, { "epoch": 0.03, "learning_rate": 9.723941818913887e-05, "loss": 1.0359, "step": 11975 }, { "epoch": 0.03, "learning_rate": 9.723815994765716e-05, "loss": 1.0358, "step": 11980 }, { "epoch": 0.03, "learning_rate": 9.723690170617545e-05, "loss": 1.0355, "step": 11985 }, { "epoch": 0.03, "learning_rate": 9.723564346469375e-05, "loss": 1.0355, "step": 11990 }, { "epoch": 0.03, "learning_rate": 9.723438522321204e-05, "loss": 1.0359, "step": 11995 }, { "epoch": 0.03, "learning_rate": 9.723312698173034e-05, "loss": 1.0343, "step": 12000 }, { "epoch": 0.03, "learning_rate": 9.723186874024863e-05, "loss": 1.0363, "step": 12005 }, { "epoch": 0.03, "learning_rate": 9.723061049876693e-05, "loss": 1.0366, "step": 12010 }, { "epoch": 0.03, "learning_rate": 9.722935225728522e-05, "loss": 1.0381, "step": 12015 }, { "epoch": 0.03, "learning_rate": 9.722809401580352e-05, "loss": 1.0756, "step": 12020 }, { "epoch": 0.03, "learning_rate": 9.722683577432181e-05, "loss": 1.054, "step": 12025 }, { "epoch": 0.03, "learning_rate": 9.72255775328401e-05, "loss": 1.0364, "step": 12030 }, { "epoch": 0.03, "learning_rate": 9.72243192913584e-05, "loss": 1.0343, "step": 12035 }, { "epoch": 0.03, "learning_rate": 9.72230610498767e-05, "loss": 1.0558, "step": 12040 }, { "epoch": 0.03, "learning_rate": 9.7221802808395e-05, "loss": 1.0357, "step": 12045 }, { "epoch": 0.03, "learning_rate": 9.722054456691329e-05, "loss": 1.0388, "step": 12050 }, { "epoch": 0.03, "learning_rate": 9.721928632543158e-05, "loss": 1.0374, "step": 12055 }, { "epoch": 0.03, "learning_rate": 9.721802808394987e-05, "loss": 1.0355, "step": 12060 }, { "epoch": 0.03, "learning_rate": 9.721676984246817e-05, "loss": 1.0367, "step": 12065 }, { "epoch": 0.03, "learning_rate": 9.721551160098647e-05, "loss": 1.0354, "step": 12070 }, { "epoch": 0.03, "learning_rate": 9.721425335950476e-05, "loss": 1.0386, "step": 12075 }, { "epoch": 0.03, "learning_rate": 9.721299511802305e-05, "loss": 1.0362, "step": 12080 }, { "epoch": 0.03, "learning_rate": 9.721173687654135e-05, "loss": 1.035, "step": 12085 }, { "epoch": 0.03, "learning_rate": 9.721047863505965e-05, "loss": 1.0373, "step": 12090 }, { "epoch": 0.03, "learning_rate": 9.720922039357794e-05, "loss": 1.0375, "step": 12095 }, { "epoch": 0.03, "learning_rate": 9.720796215209623e-05, "loss": 1.0367, "step": 12100 }, { "epoch": 0.03, "learning_rate": 9.720670391061453e-05, "loss": 1.0372, "step": 12105 }, { "epoch": 0.03, "learning_rate": 9.720544566913283e-05, "loss": 1.0366, "step": 12110 }, { "epoch": 0.03, "learning_rate": 9.720418742765112e-05, "loss": 1.0356, "step": 12115 }, { "epoch": 0.03, "learning_rate": 9.720292918616941e-05, "loss": 1.0372, "step": 12120 }, { "epoch": 0.03, "learning_rate": 9.72016709446877e-05, "loss": 1.0345, "step": 12125 }, { "epoch": 0.03, "learning_rate": 9.7200412703206e-05, "loss": 1.0372, "step": 12130 }, { "epoch": 0.03, "learning_rate": 9.71991544617243e-05, "loss": 1.0366, "step": 12135 }, { "epoch": 0.03, "learning_rate": 9.719789622024259e-05, "loss": 1.038, "step": 12140 }, { "epoch": 0.03, "learning_rate": 9.719663797876088e-05, "loss": 1.0375, "step": 12145 }, { "epoch": 0.03, "learning_rate": 9.719537973727919e-05, "loss": 1.0377, "step": 12150 }, { "epoch": 0.03, "learning_rate": 9.719412149579748e-05, "loss": 1.034, "step": 12155 }, { "epoch": 0.03, "learning_rate": 9.719286325431577e-05, "loss": 1.0355, "step": 12160 }, { "epoch": 0.03, "learning_rate": 9.719160501283406e-05, "loss": 1.0358, "step": 12165 }, { "epoch": 0.03, "learning_rate": 9.719034677135235e-05, "loss": 1.0374, "step": 12170 }, { "epoch": 0.03, "learning_rate": 9.718908852987066e-05, "loss": 1.0373, "step": 12175 }, { "epoch": 0.03, "learning_rate": 9.718783028838895e-05, "loss": 1.0361, "step": 12180 }, { "epoch": 0.03, "learning_rate": 9.718657204690724e-05, "loss": 1.0365, "step": 12185 }, { "epoch": 0.03, "learning_rate": 9.718531380542553e-05, "loss": 1.0362, "step": 12190 }, { "epoch": 0.03, "learning_rate": 9.718405556394384e-05, "loss": 1.0386, "step": 12195 }, { "epoch": 0.03, "learning_rate": 9.718279732246213e-05, "loss": 1.0378, "step": 12200 }, { "epoch": 0.03, "learning_rate": 9.718153908098042e-05, "loss": 1.0371, "step": 12205 }, { "epoch": 0.03, "learning_rate": 9.718028083949871e-05, "loss": 1.0358, "step": 12210 }, { "epoch": 0.03, "learning_rate": 9.717902259801702e-05, "loss": 1.0357, "step": 12215 }, { "epoch": 0.03, "learning_rate": 9.717776435653531e-05, "loss": 1.0373, "step": 12220 }, { "epoch": 0.03, "learning_rate": 9.71765061150536e-05, "loss": 1.0372, "step": 12225 }, { "epoch": 0.03, "learning_rate": 9.717524787357189e-05, "loss": 1.0361, "step": 12230 }, { "epoch": 0.03, "learning_rate": 9.717398963209018e-05, "loss": 1.0377, "step": 12235 }, { "epoch": 0.03, "learning_rate": 9.71727313906085e-05, "loss": 1.0347, "step": 12240 }, { "epoch": 0.03, "learning_rate": 9.71714731491268e-05, "loss": 1.0373, "step": 12245 }, { "epoch": 0.03, "learning_rate": 9.717021490764509e-05, "loss": 1.0564, "step": 12250 }, { "epoch": 0.03, "learning_rate": 9.716895666616338e-05, "loss": 1.0377, "step": 12255 }, { "epoch": 0.03, "learning_rate": 9.716769842468167e-05, "loss": 1.0356, "step": 12260 }, { "epoch": 0.03, "learning_rate": 9.716644018319997e-05, "loss": 1.0357, "step": 12265 }, { "epoch": 0.03, "learning_rate": 9.716518194171827e-05, "loss": 1.0395, "step": 12270 }, { "epoch": 0.03, "learning_rate": 9.716392370023656e-05, "loss": 1.036, "step": 12275 }, { "epoch": 0.03, "learning_rate": 9.716266545875485e-05, "loss": 1.0361, "step": 12280 }, { "epoch": 0.03, "learning_rate": 9.716140721727315e-05, "loss": 1.0369, "step": 12285 }, { "epoch": 0.03, "learning_rate": 9.716014897579145e-05, "loss": 1.0362, "step": 12290 }, { "epoch": 0.03, "learning_rate": 9.715889073430974e-05, "loss": 1.0361, "step": 12295 }, { "epoch": 0.03, "learning_rate": 9.715763249282803e-05, "loss": 1.0367, "step": 12300 }, { "epoch": 0.03, "learning_rate": 9.715637425134633e-05, "loss": 1.0357, "step": 12305 }, { "epoch": 0.03, "learning_rate": 9.715511600986462e-05, "loss": 1.0364, "step": 12310 }, { "epoch": 0.03, "learning_rate": 9.715385776838292e-05, "loss": 1.0361, "step": 12315 }, { "epoch": 0.03, "learning_rate": 9.715259952690121e-05, "loss": 1.0354, "step": 12320 }, { "epoch": 0.03, "learning_rate": 9.71513412854195e-05, "loss": 1.035, "step": 12325 }, { "epoch": 0.03, "learning_rate": 9.71500830439378e-05, "loss": 1.035, "step": 12330 }, { "epoch": 0.03, "learning_rate": 9.71488248024561e-05, "loss": 1.0367, "step": 12335 }, { "epoch": 0.03, "learning_rate": 9.714756656097439e-05, "loss": 1.0365, "step": 12340 }, { "epoch": 0.03, "learning_rate": 9.714630831949268e-05, "loss": 1.0359, "step": 12345 }, { "epoch": 0.03, "learning_rate": 9.714505007801098e-05, "loss": 1.0404, "step": 12350 }, { "epoch": 0.03, "learning_rate": 9.714379183652928e-05, "loss": 1.0359, "step": 12355 }, { "epoch": 0.03, "learning_rate": 9.714253359504757e-05, "loss": 1.0349, "step": 12360 }, { "epoch": 0.03, "learning_rate": 9.714127535356586e-05, "loss": 1.0366, "step": 12365 }, { "epoch": 0.03, "learning_rate": 9.714001711208416e-05, "loss": 1.0362, "step": 12370 }, { "epoch": 0.03, "learning_rate": 9.713875887060246e-05, "loss": 1.0361, "step": 12375 }, { "epoch": 0.03, "learning_rate": 9.713750062912075e-05, "loss": 1.0345, "step": 12380 }, { "epoch": 0.03, "learning_rate": 9.713624238763904e-05, "loss": 1.0365, "step": 12385 }, { "epoch": 0.03, "learning_rate": 9.713498414615733e-05, "loss": 1.0373, "step": 12390 }, { "epoch": 0.03, "learning_rate": 9.713372590467564e-05, "loss": 1.0348, "step": 12395 }, { "epoch": 0.03, "learning_rate": 9.713246766319393e-05, "loss": 1.0363, "step": 12400 }, { "epoch": 0.03, "learning_rate": 9.713120942171222e-05, "loss": 1.0369, "step": 12405 }, { "epoch": 0.03, "learning_rate": 9.712995118023051e-05, "loss": 1.0357, "step": 12410 }, { "epoch": 0.03, "learning_rate": 9.712869293874882e-05, "loss": 1.0376, "step": 12415 }, { "epoch": 0.03, "learning_rate": 9.712743469726711e-05, "loss": 1.0366, "step": 12420 }, { "epoch": 0.03, "learning_rate": 9.71261764557854e-05, "loss": 1.0366, "step": 12425 }, { "epoch": 0.03, "learning_rate": 9.712491821430369e-05, "loss": 1.0363, "step": 12430 }, { "epoch": 0.03, "learning_rate": 9.7123659972822e-05, "loss": 1.0368, "step": 12435 }, { "epoch": 0.03, "learning_rate": 9.712240173134029e-05, "loss": 1.0365, "step": 12440 }, { "epoch": 0.03, "learning_rate": 9.712114348985858e-05, "loss": 1.0378, "step": 12445 }, { "epoch": 0.03, "learning_rate": 9.711988524837687e-05, "loss": 1.0347, "step": 12450 }, { "epoch": 0.03, "learning_rate": 9.711862700689516e-05, "loss": 1.0378, "step": 12455 }, { "epoch": 0.03, "learning_rate": 9.711736876541347e-05, "loss": 1.0369, "step": 12460 }, { "epoch": 0.03, "learning_rate": 9.711611052393176e-05, "loss": 1.0363, "step": 12465 }, { "epoch": 0.03, "learning_rate": 9.711485228245005e-05, "loss": 1.0348, "step": 12470 }, { "epoch": 0.03, "learning_rate": 9.711359404096834e-05, "loss": 1.0338, "step": 12475 }, { "epoch": 0.03, "learning_rate": 9.711233579948665e-05, "loss": 1.0369, "step": 12480 }, { "epoch": 0.03, "learning_rate": 9.711107755800494e-05, "loss": 1.0358, "step": 12485 }, { "epoch": 0.03, "learning_rate": 9.710981931652323e-05, "loss": 1.0381, "step": 12490 }, { "epoch": 0.03, "learning_rate": 9.710856107504152e-05, "loss": 1.0372, "step": 12495 }, { "epoch": 0.03, "learning_rate": 9.710730283355983e-05, "loss": 1.0366, "step": 12500 }, { "epoch": 0.03, "learning_rate": 9.710604459207812e-05, "loss": 1.0356, "step": 12505 }, { "epoch": 0.03, "learning_rate": 9.710478635059641e-05, "loss": 1.054, "step": 12510 }, { "epoch": 0.03, "learning_rate": 9.71035281091147e-05, "loss": 1.0363, "step": 12515 }, { "epoch": 0.03, "learning_rate": 9.7102269867633e-05, "loss": 1.0354, "step": 12520 }, { "epoch": 0.03, "learning_rate": 9.71010116261513e-05, "loss": 1.0367, "step": 12525 }, { "epoch": 0.03, "learning_rate": 9.709975338466959e-05, "loss": 1.0342, "step": 12530 }, { "epoch": 0.03, "learning_rate": 9.709849514318788e-05, "loss": 1.038, "step": 12535 }, { "epoch": 0.03, "learning_rate": 9.709723690170617e-05, "loss": 1.0359, "step": 12540 }, { "epoch": 0.03, "learning_rate": 9.709597866022448e-05, "loss": 1.0355, "step": 12545 }, { "epoch": 0.03, "learning_rate": 9.709472041874277e-05, "loss": 1.0365, "step": 12550 }, { "epoch": 0.03, "learning_rate": 9.709346217726106e-05, "loss": 1.0367, "step": 12555 }, { "epoch": 0.03, "learning_rate": 9.709220393577935e-05, "loss": 1.036, "step": 12560 }, { "epoch": 0.03, "learning_rate": 9.709094569429766e-05, "loss": 1.0355, "step": 12565 }, { "epoch": 0.03, "learning_rate": 9.708968745281595e-05, "loss": 1.0374, "step": 12570 }, { "epoch": 0.03, "learning_rate": 9.708842921133424e-05, "loss": 1.035, "step": 12575 }, { "epoch": 0.03, "learning_rate": 9.708717096985253e-05, "loss": 1.0349, "step": 12580 }, { "epoch": 0.03, "learning_rate": 9.708591272837083e-05, "loss": 1.0354, "step": 12585 }, { "epoch": 0.03, "learning_rate": 9.708465448688913e-05, "loss": 1.037, "step": 12590 }, { "epoch": 0.03, "learning_rate": 9.708339624540742e-05, "loss": 1.0371, "step": 12595 }, { "epoch": 0.03, "learning_rate": 9.708213800392571e-05, "loss": 1.038, "step": 12600 }, { "epoch": 0.03, "learning_rate": 9.7080879762444e-05, "loss": 1.0347, "step": 12605 }, { "epoch": 0.03, "learning_rate": 9.707962152096231e-05, "loss": 1.037, "step": 12610 }, { "epoch": 0.03, "learning_rate": 9.70783632794806e-05, "loss": 1.0366, "step": 12615 }, { "epoch": 0.03, "learning_rate": 9.70771050379989e-05, "loss": 1.0375, "step": 12620 }, { "epoch": 0.03, "learning_rate": 9.707584679651719e-05, "loss": 1.0583, "step": 12625 }, { "epoch": 0.03, "learning_rate": 9.707458855503549e-05, "loss": 1.0375, "step": 12630 }, { "epoch": 0.03, "learning_rate": 9.707333031355378e-05, "loss": 1.0363, "step": 12635 }, { "epoch": 0.03, "learning_rate": 9.707207207207207e-05, "loss": 1.0375, "step": 12640 }, { "epoch": 0.03, "learning_rate": 9.707081383059037e-05, "loss": 1.035, "step": 12645 }, { "epoch": 0.03, "learning_rate": 9.706955558910866e-05, "loss": 1.0374, "step": 12650 }, { "epoch": 0.03, "learning_rate": 9.706829734762696e-05, "loss": 1.0369, "step": 12655 }, { "epoch": 0.03, "learning_rate": 9.706703910614525e-05, "loss": 1.037, "step": 12660 }, { "epoch": 0.03, "learning_rate": 9.706578086466355e-05, "loss": 1.037, "step": 12665 }, { "epoch": 0.03, "learning_rate": 9.706452262318184e-05, "loss": 1.0359, "step": 12670 }, { "epoch": 0.03, "learning_rate": 9.706326438170014e-05, "loss": 1.0348, "step": 12675 }, { "epoch": 0.03, "learning_rate": 9.706200614021843e-05, "loss": 1.0368, "step": 12680 }, { "epoch": 0.03, "learning_rate": 9.706074789873673e-05, "loss": 1.0362, "step": 12685 }, { "epoch": 0.03, "learning_rate": 9.705948965725502e-05, "loss": 1.0372, "step": 12690 }, { "epoch": 0.03, "learning_rate": 9.705823141577332e-05, "loss": 1.0378, "step": 12695 }, { "epoch": 0.03, "learning_rate": 9.705697317429161e-05, "loss": 1.0359, "step": 12700 }, { "epoch": 0.03, "learning_rate": 9.70557149328099e-05, "loss": 1.0338, "step": 12705 }, { "epoch": 0.03, "learning_rate": 9.70544566913282e-05, "loss": 1.035, "step": 12710 }, { "epoch": 0.03, "learning_rate": 9.705319844984649e-05, "loss": 1.0353, "step": 12715 }, { "epoch": 0.03, "learning_rate": 9.70519402083648e-05, "loss": 1.0363, "step": 12720 }, { "epoch": 0.03, "learning_rate": 9.705068196688309e-05, "loss": 1.0349, "step": 12725 }, { "epoch": 0.03, "learning_rate": 9.704942372540138e-05, "loss": 1.0353, "step": 12730 }, { "epoch": 0.03, "learning_rate": 9.704816548391967e-05, "loss": 1.0351, "step": 12735 }, { "epoch": 0.03, "learning_rate": 9.704690724243797e-05, "loss": 1.0359, "step": 12740 }, { "epoch": 0.03, "learning_rate": 9.704564900095628e-05, "loss": 1.0599, "step": 12745 }, { "epoch": 0.03, "learning_rate": 9.704439075947457e-05, "loss": 1.0366, "step": 12750 }, { "epoch": 0.03, "learning_rate": 9.704313251799286e-05, "loss": 1.0371, "step": 12755 }, { "epoch": 0.03, "learning_rate": 9.704187427651115e-05, "loss": 1.0369, "step": 12760 }, { "epoch": 0.03, "learning_rate": 9.704061603502946e-05, "loss": 1.0355, "step": 12765 }, { "epoch": 0.03, "learning_rate": 9.703935779354775e-05, "loss": 1.0363, "step": 12770 }, { "epoch": 0.03, "learning_rate": 9.703809955206604e-05, "loss": 1.0369, "step": 12775 }, { "epoch": 0.03, "learning_rate": 9.703684131058433e-05, "loss": 1.0347, "step": 12780 }, { "epoch": 0.03, "learning_rate": 9.703558306910262e-05, "loss": 1.0343, "step": 12785 }, { "epoch": 0.03, "learning_rate": 9.703432482762093e-05, "loss": 1.0358, "step": 12790 }, { "epoch": 0.03, "learning_rate": 9.703306658613922e-05, "loss": 1.0341, "step": 12795 }, { "epoch": 0.03, "learning_rate": 9.703180834465751e-05, "loss": 1.0355, "step": 12800 }, { "epoch": 0.03, "learning_rate": 9.70305501031758e-05, "loss": 1.036, "step": 12805 }, { "epoch": 0.03, "learning_rate": 9.702929186169411e-05, "loss": 1.036, "step": 12810 }, { "epoch": 0.03, "learning_rate": 9.70280336202124e-05, "loss": 1.034, "step": 12815 }, { "epoch": 0.03, "learning_rate": 9.702677537873069e-05, "loss": 1.0356, "step": 12820 }, { "epoch": 0.03, "learning_rate": 9.702551713724898e-05, "loss": 1.0379, "step": 12825 }, { "epoch": 0.03, "learning_rate": 9.702425889576729e-05, "loss": 1.0368, "step": 12830 }, { "epoch": 0.03, "learning_rate": 9.702300065428558e-05, "loss": 1.0359, "step": 12835 }, { "epoch": 0.03, "learning_rate": 9.702174241280387e-05, "loss": 1.0338, "step": 12840 }, { "epoch": 0.03, "learning_rate": 9.702048417132216e-05, "loss": 1.0369, "step": 12845 }, { "epoch": 0.03, "learning_rate": 9.701922592984046e-05, "loss": 1.0352, "step": 12850 }, { "epoch": 0.03, "learning_rate": 9.701796768835876e-05, "loss": 1.0365, "step": 12855 }, { "epoch": 0.03, "learning_rate": 9.701670944687705e-05, "loss": 1.0366, "step": 12860 }, { "epoch": 0.03, "learning_rate": 9.701545120539534e-05, "loss": 1.0362, "step": 12865 }, { "epoch": 0.03, "learning_rate": 9.701419296391364e-05, "loss": 1.0373, "step": 12870 }, { "epoch": 0.03, "learning_rate": 9.701293472243194e-05, "loss": 1.0353, "step": 12875 }, { "epoch": 0.03, "learning_rate": 9.701167648095023e-05, "loss": 1.0342, "step": 12880 }, { "epoch": 0.03, "learning_rate": 9.701041823946852e-05, "loss": 1.0345, "step": 12885 }, { "epoch": 0.03, "learning_rate": 9.700915999798682e-05, "loss": 1.0359, "step": 12890 }, { "epoch": 0.03, "learning_rate": 9.700790175650512e-05, "loss": 1.0365, "step": 12895 }, { "epoch": 0.03, "learning_rate": 9.700664351502341e-05, "loss": 1.0375, "step": 12900 }, { "epoch": 0.03, "learning_rate": 9.70053852735417e-05, "loss": 1.0368, "step": 12905 }, { "epoch": 0.03, "learning_rate": 9.700412703206e-05, "loss": 1.0365, "step": 12910 }, { "epoch": 0.03, "learning_rate": 9.700286879057829e-05, "loss": 1.0363, "step": 12915 }, { "epoch": 0.03, "learning_rate": 9.700161054909659e-05, "loss": 1.0575, "step": 12920 }, { "epoch": 0.03, "learning_rate": 9.700035230761488e-05, "loss": 1.035, "step": 12925 }, { "epoch": 0.03, "learning_rate": 9.699909406613318e-05, "loss": 1.0369, "step": 12930 }, { "epoch": 0.03, "learning_rate": 9.699783582465147e-05, "loss": 1.0377, "step": 12935 }, { "epoch": 0.03, "learning_rate": 9.699657758316977e-05, "loss": 1.0368, "step": 12940 }, { "epoch": 0.03, "learning_rate": 9.699531934168806e-05, "loss": 1.0366, "step": 12945 }, { "epoch": 0.03, "learning_rate": 9.699406110020636e-05, "loss": 1.0349, "step": 12950 }, { "epoch": 0.03, "learning_rate": 9.699280285872465e-05, "loss": 1.0362, "step": 12955 }, { "epoch": 0.03, "learning_rate": 9.699154461724295e-05, "loss": 1.0398, "step": 12960 }, { "epoch": 0.03, "learning_rate": 9.699028637576124e-05, "loss": 1.0562, "step": 12965 }, { "epoch": 0.03, "learning_rate": 9.698902813427954e-05, "loss": 1.0365, "step": 12970 }, { "epoch": 0.03, "learning_rate": 9.698776989279783e-05, "loss": 1.0359, "step": 12975 }, { "epoch": 0.03, "learning_rate": 9.698651165131612e-05, "loss": 1.0361, "step": 12980 }, { "epoch": 0.03, "learning_rate": 9.698525340983442e-05, "loss": 1.0379, "step": 12985 }, { "epoch": 0.03, "learning_rate": 9.698399516835272e-05, "loss": 1.0376, "step": 12990 }, { "epoch": 0.03, "learning_rate": 9.698273692687101e-05, "loss": 1.0345, "step": 12995 }, { "epoch": 0.03, "learning_rate": 9.69814786853893e-05, "loss": 1.0361, "step": 13000 }, { "epoch": 0.03, "learning_rate": 9.69802204439076e-05, "loss": 1.0331, "step": 13005 }, { "epoch": 0.03, "learning_rate": 9.69789622024259e-05, "loss": 1.0356, "step": 13010 }, { "epoch": 0.03, "learning_rate": 9.697770396094419e-05, "loss": 1.0343, "step": 13015 }, { "epoch": 0.03, "learning_rate": 9.697644571946248e-05, "loss": 1.0367, "step": 13020 }, { "epoch": 0.03, "learning_rate": 9.697518747798078e-05, "loss": 1.0373, "step": 13025 }, { "epoch": 0.03, "learning_rate": 9.697392923649908e-05, "loss": 1.0379, "step": 13030 }, { "epoch": 0.03, "learning_rate": 9.697267099501737e-05, "loss": 1.0357, "step": 13035 }, { "epoch": 0.03, "learning_rate": 9.697141275353566e-05, "loss": 1.0351, "step": 13040 }, { "epoch": 0.03, "learning_rate": 9.697015451205395e-05, "loss": 1.0369, "step": 13045 }, { "epoch": 0.03, "learning_rate": 9.696889627057226e-05, "loss": 1.0347, "step": 13050 }, { "epoch": 0.03, "learning_rate": 9.696763802909055e-05, "loss": 1.0348, "step": 13055 }, { "epoch": 0.03, "learning_rate": 9.696637978760884e-05, "loss": 1.0356, "step": 13060 }, { "epoch": 0.03, "learning_rate": 9.696512154612713e-05, "loss": 1.0364, "step": 13065 }, { "epoch": 0.03, "learning_rate": 9.696386330464544e-05, "loss": 1.0358, "step": 13070 }, { "epoch": 0.03, "learning_rate": 9.696260506316373e-05, "loss": 1.0353, "step": 13075 }, { "epoch": 0.03, "learning_rate": 9.696134682168202e-05, "loss": 1.0375, "step": 13080 }, { "epoch": 0.03, "learning_rate": 9.696008858020031e-05, "loss": 1.0367, "step": 13085 }, { "epoch": 0.03, "learning_rate": 9.695883033871862e-05, "loss": 1.0341, "step": 13090 }, { "epoch": 0.03, "learning_rate": 9.695757209723691e-05, "loss": 1.0364, "step": 13095 }, { "epoch": 0.03, "learning_rate": 9.69563138557552e-05, "loss": 1.0358, "step": 13100 }, { "epoch": 0.03, "learning_rate": 9.695505561427349e-05, "loss": 1.0368, "step": 13105 }, { "epoch": 0.03, "learning_rate": 9.695379737279178e-05, "loss": 1.0355, "step": 13110 }, { "epoch": 0.03, "learning_rate": 9.695253913131009e-05, "loss": 1.0364, "step": 13115 }, { "epoch": 0.03, "learning_rate": 9.695128088982838e-05, "loss": 1.0362, "step": 13120 }, { "epoch": 0.03, "learning_rate": 9.695002264834667e-05, "loss": 1.0354, "step": 13125 }, { "epoch": 0.03, "learning_rate": 9.694876440686496e-05, "loss": 1.0363, "step": 13130 }, { "epoch": 0.03, "learning_rate": 9.694750616538327e-05, "loss": 1.0367, "step": 13135 }, { "epoch": 0.03, "learning_rate": 9.694624792390156e-05, "loss": 1.0351, "step": 13140 }, { "epoch": 0.03, "learning_rate": 9.694498968241985e-05, "loss": 1.0363, "step": 13145 }, { "epoch": 0.03, "learning_rate": 9.694373144093814e-05, "loss": 1.0553, "step": 13150 }, { "epoch": 0.03, "learning_rate": 9.694247319945645e-05, "loss": 1.0338, "step": 13155 }, { "epoch": 0.03, "learning_rate": 9.694121495797474e-05, "loss": 1.0359, "step": 13160 }, { "epoch": 0.03, "learning_rate": 9.693995671649303e-05, "loss": 1.0381, "step": 13165 }, { "epoch": 0.03, "learning_rate": 9.693869847501132e-05, "loss": 1.0361, "step": 13170 }, { "epoch": 0.03, "learning_rate": 9.693744023352961e-05, "loss": 1.0348, "step": 13175 }, { "epoch": 0.03, "learning_rate": 9.693618199204792e-05, "loss": 1.0354, "step": 13180 }, { "epoch": 0.03, "learning_rate": 9.693492375056621e-05, "loss": 1.0349, "step": 13185 }, { "epoch": 0.03, "learning_rate": 9.69336655090845e-05, "loss": 1.0339, "step": 13190 }, { "epoch": 0.03, "learning_rate": 9.69324072676028e-05, "loss": 1.0366, "step": 13195 }, { "epoch": 0.03, "learning_rate": 9.69311490261211e-05, "loss": 1.0375, "step": 13200 }, { "epoch": 0.03, "learning_rate": 9.692989078463939e-05, "loss": 1.0366, "step": 13205 }, { "epoch": 0.03, "learning_rate": 9.692863254315768e-05, "loss": 1.0352, "step": 13210 }, { "epoch": 0.03, "learning_rate": 9.692737430167597e-05, "loss": 1.0356, "step": 13215 }, { "epoch": 0.03, "learning_rate": 9.692611606019428e-05, "loss": 1.0357, "step": 13220 }, { "epoch": 0.03, "learning_rate": 9.692485781871257e-05, "loss": 1.0362, "step": 13225 }, { "epoch": 0.03, "learning_rate": 9.692359957723086e-05, "loss": 1.0372, "step": 13230 }, { "epoch": 0.03, "learning_rate": 9.692234133574915e-05, "loss": 1.0367, "step": 13235 }, { "epoch": 0.03, "learning_rate": 9.692108309426746e-05, "loss": 1.036, "step": 13240 }, { "epoch": 0.03, "learning_rate": 9.691982485278575e-05, "loss": 1.037, "step": 13245 }, { "epoch": 0.03, "learning_rate": 9.691856661130406e-05, "loss": 1.035, "step": 13250 }, { "epoch": 0.03, "learning_rate": 9.691730836982235e-05, "loss": 1.0363, "step": 13255 }, { "epoch": 0.03, "learning_rate": 9.691605012834064e-05, "loss": 1.0382, "step": 13260 }, { "epoch": 0.03, "learning_rate": 9.691479188685893e-05, "loss": 1.0341, "step": 13265 }, { "epoch": 0.03, "learning_rate": 9.691353364537724e-05, "loss": 1.0365, "step": 13270 }, { "epoch": 0.03, "learning_rate": 9.691227540389553e-05, "loss": 1.0371, "step": 13275 }, { "epoch": 0.03, "learning_rate": 9.691101716241382e-05, "loss": 1.037, "step": 13280 }, { "epoch": 0.03, "learning_rate": 9.690975892093211e-05, "loss": 1.0357, "step": 13285 }, { "epoch": 0.03, "learning_rate": 9.690850067945042e-05, "loss": 1.0378, "step": 13290 }, { "epoch": 0.03, "learning_rate": 9.69072424379687e-05, "loss": 1.036, "step": 13295 }, { "epoch": 0.03, "learning_rate": 9.6905984196487e-05, "loss": 1.0345, "step": 13300 }, { "epoch": 0.03, "learning_rate": 9.690472595500529e-05, "loss": 1.0363, "step": 13305 }, { "epoch": 0.03, "learning_rate": 9.690346771352358e-05, "loss": 1.0362, "step": 13310 }, { "epoch": 0.03, "learning_rate": 9.690220947204189e-05, "loss": 1.0362, "step": 13315 }, { "epoch": 0.03, "learning_rate": 9.690095123056018e-05, "loss": 1.035, "step": 13320 }, { "epoch": 0.03, "learning_rate": 9.689969298907847e-05, "loss": 1.0358, "step": 13325 }, { "epoch": 0.03, "learning_rate": 9.689843474759676e-05, "loss": 1.0357, "step": 13330 }, { "epoch": 0.03, "learning_rate": 9.689717650611507e-05, "loss": 1.039, "step": 13335 }, { "epoch": 0.03, "learning_rate": 9.689591826463336e-05, "loss": 1.0363, "step": 13340 }, { "epoch": 0.03, "learning_rate": 9.689466002315165e-05, "loss": 1.0371, "step": 13345 }, { "epoch": 0.03, "learning_rate": 9.689340178166994e-05, "loss": 1.0369, "step": 13350 }, { "epoch": 0.03, "learning_rate": 9.689214354018825e-05, "loss": 1.0355, "step": 13355 }, { "epoch": 0.03, "learning_rate": 9.689088529870654e-05, "loss": 1.0366, "step": 13360 }, { "epoch": 0.03, "learning_rate": 9.688962705722483e-05, "loss": 1.0348, "step": 13365 }, { "epoch": 0.03, "learning_rate": 9.688836881574312e-05, "loss": 1.0344, "step": 13370 }, { "epoch": 0.03, "learning_rate": 9.688711057426141e-05, "loss": 1.0359, "step": 13375 }, { "epoch": 0.03, "learning_rate": 9.688585233277972e-05, "loss": 1.0361, "step": 13380 }, { "epoch": 0.03, "learning_rate": 9.688459409129801e-05, "loss": 1.0347, "step": 13385 }, { "epoch": 0.03, "learning_rate": 9.68833358498163e-05, "loss": 1.0359, "step": 13390 }, { "epoch": 0.03, "learning_rate": 9.688207760833459e-05, "loss": 1.0333, "step": 13395 }, { "epoch": 0.03, "learning_rate": 9.68808193668529e-05, "loss": 1.0368, "step": 13400 }, { "epoch": 0.03, "learning_rate": 9.687956112537119e-05, "loss": 1.0352, "step": 13405 }, { "epoch": 0.03, "learning_rate": 9.687830288388948e-05, "loss": 1.0359, "step": 13410 }, { "epoch": 0.03, "learning_rate": 9.687704464240777e-05, "loss": 1.0354, "step": 13415 }, { "epoch": 0.03, "learning_rate": 9.687578640092608e-05, "loss": 1.0334, "step": 13420 }, { "epoch": 0.03, "learning_rate": 9.687452815944437e-05, "loss": 1.0356, "step": 13425 }, { "epoch": 0.03, "learning_rate": 9.687326991796266e-05, "loss": 1.0346, "step": 13430 }, { "epoch": 0.03, "learning_rate": 9.687201167648095e-05, "loss": 1.0367, "step": 13435 }, { "epoch": 0.03, "learning_rate": 9.687075343499924e-05, "loss": 1.0386, "step": 13440 }, { "epoch": 0.03, "learning_rate": 9.686949519351755e-05, "loss": 1.0576, "step": 13445 }, { "epoch": 0.03, "learning_rate": 9.686823695203584e-05, "loss": 1.036, "step": 13450 }, { "epoch": 0.03, "learning_rate": 9.686697871055413e-05, "loss": 1.0363, "step": 13455 }, { "epoch": 0.03, "learning_rate": 9.686572046907242e-05, "loss": 1.0328, "step": 13460 }, { "epoch": 0.03, "learning_rate": 9.686446222759073e-05, "loss": 1.036, "step": 13465 }, { "epoch": 0.03, "learning_rate": 9.686320398610902e-05, "loss": 1.0358, "step": 13470 }, { "epoch": 0.03, "learning_rate": 9.686194574462731e-05, "loss": 1.0351, "step": 13475 }, { "epoch": 0.03, "learning_rate": 9.68606875031456e-05, "loss": 1.0357, "step": 13480 }, { "epoch": 0.03, "learning_rate": 9.685942926166391e-05, "loss": 1.0376, "step": 13485 }, { "epoch": 0.03, "learning_rate": 9.68581710201822e-05, "loss": 1.0349, "step": 13490 }, { "epoch": 0.03, "learning_rate": 9.685691277870049e-05, "loss": 1.0372, "step": 13495 }, { "epoch": 0.03, "learning_rate": 9.685565453721878e-05, "loss": 1.0351, "step": 13500 }, { "epoch": 0.03, "learning_rate": 9.685439629573708e-05, "loss": 1.0361, "step": 13505 }, { "epoch": 0.03, "learning_rate": 9.685313805425538e-05, "loss": 1.0365, "step": 13510 }, { "epoch": 0.03, "learning_rate": 9.685187981277367e-05, "loss": 1.034, "step": 13515 }, { "epoch": 0.03, "learning_rate": 9.685062157129196e-05, "loss": 1.0373, "step": 13520 }, { "epoch": 0.03, "learning_rate": 9.684936332981026e-05, "loss": 1.0347, "step": 13525 }, { "epoch": 0.03, "learning_rate": 9.684810508832856e-05, "loss": 1.0365, "step": 13530 }, { "epoch": 0.03, "learning_rate": 9.684684684684685e-05, "loss": 1.0347, "step": 13535 }, { "epoch": 0.03, "learning_rate": 9.684558860536514e-05, "loss": 1.0347, "step": 13540 }, { "epoch": 0.03, "learning_rate": 9.684433036388344e-05, "loss": 1.0344, "step": 13545 }, { "epoch": 0.03, "learning_rate": 9.684307212240174e-05, "loss": 1.0356, "step": 13550 }, { "epoch": 0.03, "learning_rate": 9.684181388092003e-05, "loss": 1.0347, "step": 13555 }, { "epoch": 0.03, "learning_rate": 9.684055563943832e-05, "loss": 1.0371, "step": 13560 }, { "epoch": 0.03, "learning_rate": 9.683929739795662e-05, "loss": 1.0375, "step": 13565 }, { "epoch": 0.03, "learning_rate": 9.683803915647491e-05, "loss": 1.0345, "step": 13570 }, { "epoch": 0.03, "learning_rate": 9.683678091499321e-05, "loss": 1.0356, "step": 13575 }, { "epoch": 0.03, "learning_rate": 9.68355226735115e-05, "loss": 1.0576, "step": 13580 }, { "epoch": 0.03, "learning_rate": 9.68342644320298e-05, "loss": 1.037, "step": 13585 }, { "epoch": 0.03, "learning_rate": 9.683300619054809e-05, "loss": 1.0375, "step": 13590 }, { "epoch": 0.03, "learning_rate": 9.683174794906639e-05, "loss": 1.0389, "step": 13595 }, { "epoch": 0.03, "learning_rate": 9.683048970758468e-05, "loss": 1.0358, "step": 13600 }, { "epoch": 0.03, "learning_rate": 9.682923146610298e-05, "loss": 1.0344, "step": 13605 }, { "epoch": 0.03, "learning_rate": 9.682797322462127e-05, "loss": 1.0345, "step": 13610 }, { "epoch": 0.03, "learning_rate": 9.682671498313957e-05, "loss": 1.0373, "step": 13615 }, { "epoch": 0.03, "learning_rate": 9.682545674165786e-05, "loss": 1.0359, "step": 13620 }, { "epoch": 0.03, "learning_rate": 9.682419850017616e-05, "loss": 1.036, "step": 13625 }, { "epoch": 0.03, "learning_rate": 9.682294025869445e-05, "loss": 1.036, "step": 13630 }, { "epoch": 0.03, "learning_rate": 9.682168201721274e-05, "loss": 1.0344, "step": 13635 }, { "epoch": 0.03, "learning_rate": 9.682042377573104e-05, "loss": 1.0382, "step": 13640 }, { "epoch": 0.03, "learning_rate": 9.681916553424934e-05, "loss": 1.0355, "step": 13645 }, { "epoch": 0.03, "learning_rate": 9.681790729276763e-05, "loss": 1.0338, "step": 13650 }, { "epoch": 0.03, "learning_rate": 9.681664905128592e-05, "loss": 1.0356, "step": 13655 }, { "epoch": 0.03, "learning_rate": 9.681539080980422e-05, "loss": 1.0382, "step": 13660 }, { "epoch": 0.03, "learning_rate": 9.681413256832252e-05, "loss": 1.0352, "step": 13665 }, { "epoch": 0.03, "learning_rate": 9.681287432684081e-05, "loss": 1.0372, "step": 13670 }, { "epoch": 0.03, "learning_rate": 9.68116160853591e-05, "loss": 1.0363, "step": 13675 }, { "epoch": 0.03, "learning_rate": 9.68103578438774e-05, "loss": 1.0362, "step": 13680 }, { "epoch": 0.03, "learning_rate": 9.68090996023957e-05, "loss": 1.0358, "step": 13685 }, { "epoch": 0.03, "learning_rate": 9.680784136091399e-05, "loss": 1.0356, "step": 13690 }, { "epoch": 0.03, "learning_rate": 9.680658311943228e-05, "loss": 1.0344, "step": 13695 }, { "epoch": 0.03, "learning_rate": 9.680532487795057e-05, "loss": 1.0365, "step": 13700 }, { "epoch": 0.03, "learning_rate": 9.680406663646888e-05, "loss": 1.0369, "step": 13705 }, { "epoch": 0.03, "learning_rate": 9.680280839498717e-05, "loss": 1.0356, "step": 13710 }, { "epoch": 0.03, "learning_rate": 9.680155015350546e-05, "loss": 1.0382, "step": 13715 }, { "epoch": 0.03, "learning_rate": 9.680029191202375e-05, "loss": 1.0615, "step": 13720 }, { "epoch": 0.03, "learning_rate": 9.679903367054206e-05, "loss": 1.0377, "step": 13725 }, { "epoch": 0.03, "learning_rate": 9.679777542906035e-05, "loss": 1.038, "step": 13730 }, { "epoch": 0.03, "learning_rate": 9.679651718757864e-05, "loss": 1.0351, "step": 13735 }, { "epoch": 0.03, "learning_rate": 9.679525894609694e-05, "loss": 1.0347, "step": 13740 }, { "epoch": 0.03, "learning_rate": 9.679400070461524e-05, "loss": 1.0376, "step": 13745 }, { "epoch": 0.03, "learning_rate": 9.679274246313354e-05, "loss": 1.0319, "step": 13750 }, { "epoch": 0.03, "learning_rate": 9.679148422165183e-05, "loss": 1.0348, "step": 13755 }, { "epoch": 0.03, "learning_rate": 9.679022598017012e-05, "loss": 1.0357, "step": 13760 }, { "epoch": 0.03, "learning_rate": 9.678896773868841e-05, "loss": 1.0357, "step": 13765 }, { "epoch": 0.03, "learning_rate": 9.67877094972067e-05, "loss": 1.0332, "step": 13770 }, { "epoch": 0.03, "learning_rate": 9.678645125572501e-05, "loss": 1.039, "step": 13775 }, { "epoch": 0.03, "learning_rate": 9.67851930142433e-05, "loss": 1.0366, "step": 13780 }, { "epoch": 0.03, "learning_rate": 9.67839347727616e-05, "loss": 1.0349, "step": 13785 }, { "epoch": 0.03, "learning_rate": 9.678267653127989e-05, "loss": 1.0354, "step": 13790 }, { "epoch": 0.03, "learning_rate": 9.678141828979819e-05, "loss": 1.0361, "step": 13795 }, { "epoch": 0.03, "learning_rate": 9.678016004831648e-05, "loss": 1.0326, "step": 13800 }, { "epoch": 0.03, "learning_rate": 9.677890180683477e-05, "loss": 1.0363, "step": 13805 }, { "epoch": 0.03, "learning_rate": 9.677764356535307e-05, "loss": 1.0347, "step": 13810 }, { "epoch": 0.03, "learning_rate": 9.677638532387137e-05, "loss": 1.0361, "step": 13815 }, { "epoch": 0.03, "learning_rate": 9.677512708238966e-05, "loss": 1.0336, "step": 13820 }, { "epoch": 0.03, "learning_rate": 9.677386884090795e-05, "loss": 1.0353, "step": 13825 }, { "epoch": 0.03, "learning_rate": 9.677261059942625e-05, "loss": 1.0355, "step": 13830 }, { "epoch": 0.03, "learning_rate": 9.677135235794454e-05, "loss": 1.0343, "step": 13835 }, { "epoch": 0.03, "learning_rate": 9.677009411646284e-05, "loss": 1.0354, "step": 13840 }, { "epoch": 0.03, "learning_rate": 9.676883587498113e-05, "loss": 1.0361, "step": 13845 }, { "epoch": 0.03, "learning_rate": 9.676757763349943e-05, "loss": 1.0358, "step": 13850 }, { "epoch": 0.03, "learning_rate": 9.676631939201772e-05, "loss": 1.0351, "step": 13855 }, { "epoch": 0.03, "learning_rate": 9.676506115053602e-05, "loss": 1.0361, "step": 13860 }, { "epoch": 0.03, "learning_rate": 9.676380290905431e-05, "loss": 1.0354, "step": 13865 }, { "epoch": 0.03, "learning_rate": 9.67625446675726e-05, "loss": 1.0339, "step": 13870 }, { "epoch": 0.03, "learning_rate": 9.67612864260909e-05, "loss": 1.035, "step": 13875 }, { "epoch": 0.03, "learning_rate": 9.67600281846092e-05, "loss": 1.0372, "step": 13880 }, { "epoch": 0.03, "learning_rate": 9.67587699431275e-05, "loss": 1.0366, "step": 13885 }, { "epoch": 0.03, "learning_rate": 9.675751170164579e-05, "loss": 1.0366, "step": 13890 }, { "epoch": 0.03, "learning_rate": 9.675625346016408e-05, "loss": 1.0362, "step": 13895 }, { "epoch": 0.03, "learning_rate": 9.675499521868237e-05, "loss": 1.036, "step": 13900 }, { "epoch": 0.03, "learning_rate": 9.675373697720067e-05, "loss": 1.051, "step": 13905 }, { "epoch": 0.03, "learning_rate": 9.675247873571897e-05, "loss": 1.0366, "step": 13910 }, { "epoch": 0.03, "learning_rate": 9.675122049423726e-05, "loss": 1.0357, "step": 13915 }, { "epoch": 0.03, "learning_rate": 9.674996225275555e-05, "loss": 1.0331, "step": 13920 }, { "epoch": 0.03, "learning_rate": 9.674870401127385e-05, "loss": 1.0344, "step": 13925 }, { "epoch": 0.03, "learning_rate": 9.674744576979215e-05, "loss": 1.0354, "step": 13930 }, { "epoch": 0.03, "learning_rate": 9.674618752831044e-05, "loss": 1.0345, "step": 13935 }, { "epoch": 0.03, "learning_rate": 9.674492928682873e-05, "loss": 1.0517, "step": 13940 }, { "epoch": 0.04, "learning_rate": 9.674367104534703e-05, "loss": 1.0368, "step": 13945 }, { "epoch": 0.04, "learning_rate": 9.674241280386533e-05, "loss": 1.0377, "step": 13950 }, { "epoch": 0.04, "learning_rate": 9.674115456238362e-05, "loss": 1.0355, "step": 13955 }, { "epoch": 0.04, "learning_rate": 9.673989632090191e-05, "loss": 1.0347, "step": 13960 }, { "epoch": 0.04, "learning_rate": 9.67386380794202e-05, "loss": 1.0344, "step": 13965 }, { "epoch": 0.04, "learning_rate": 9.67373798379385e-05, "loss": 1.0349, "step": 13970 }, { "epoch": 0.04, "learning_rate": 9.67361215964568e-05, "loss": 1.0359, "step": 13975 }, { "epoch": 0.04, "learning_rate": 9.673486335497509e-05, "loss": 1.0386, "step": 13980 }, { "epoch": 0.04, "learning_rate": 9.673360511349338e-05, "loss": 1.0351, "step": 13985 }, { "epoch": 0.04, "learning_rate": 9.673234687201169e-05, "loss": 1.0357, "step": 13990 }, { "epoch": 0.04, "learning_rate": 9.673108863052998e-05, "loss": 1.0352, "step": 13995 }, { "epoch": 0.04, "learning_rate": 9.672983038904827e-05, "loss": 1.0353, "step": 14000 }, { "epoch": 0.04, "learning_rate": 9.672857214756656e-05, "loss": 1.0359, "step": 14005 }, { "epoch": 0.04, "learning_rate": 9.672731390608487e-05, "loss": 1.0363, "step": 14010 }, { "epoch": 0.04, "learning_rate": 9.672605566460316e-05, "loss": 1.036, "step": 14015 }, { "epoch": 0.04, "learning_rate": 9.672479742312145e-05, "loss": 1.0346, "step": 14020 }, { "epoch": 0.04, "learning_rate": 9.672353918163974e-05, "loss": 1.0363, "step": 14025 }, { "epoch": 0.04, "learning_rate": 9.672228094015803e-05, "loss": 1.0359, "step": 14030 }, { "epoch": 0.04, "learning_rate": 9.672102269867634e-05, "loss": 1.0332, "step": 14035 }, { "epoch": 0.04, "learning_rate": 9.671976445719463e-05, "loss": 1.0369, "step": 14040 }, { "epoch": 0.04, "learning_rate": 9.671850621571292e-05, "loss": 1.036, "step": 14045 }, { "epoch": 0.04, "learning_rate": 9.671724797423121e-05, "loss": 1.0356, "step": 14050 }, { "epoch": 0.04, "learning_rate": 9.671598973274952e-05, "loss": 1.0369, "step": 14055 }, { "epoch": 0.04, "learning_rate": 9.671473149126781e-05, "loss": 1.0357, "step": 14060 }, { "epoch": 0.04, "learning_rate": 9.67134732497861e-05, "loss": 1.0338, "step": 14065 }, { "epoch": 0.04, "learning_rate": 9.671221500830439e-05, "loss": 1.0329, "step": 14070 }, { "epoch": 0.04, "learning_rate": 9.67109567668227e-05, "loss": 1.0361, "step": 14075 }, { "epoch": 0.04, "learning_rate": 9.670969852534099e-05, "loss": 1.0363, "step": 14080 }, { "epoch": 0.04, "learning_rate": 9.670844028385928e-05, "loss": 1.0371, "step": 14085 }, { "epoch": 0.04, "learning_rate": 9.670718204237757e-05, "loss": 1.0378, "step": 14090 }, { "epoch": 0.04, "learning_rate": 9.670592380089586e-05, "loss": 1.0363, "step": 14095 }, { "epoch": 0.04, "learning_rate": 9.670466555941417e-05, "loss": 1.0366, "step": 14100 }, { "epoch": 0.04, "learning_rate": 9.670340731793246e-05, "loss": 1.0339, "step": 14105 }, { "epoch": 0.04, "learning_rate": 9.670214907645075e-05, "loss": 1.0351, "step": 14110 }, { "epoch": 0.04, "learning_rate": 9.670089083496904e-05, "loss": 1.0333, "step": 14115 }, { "epoch": 0.04, "learning_rate": 9.669963259348735e-05, "loss": 1.0366, "step": 14120 }, { "epoch": 0.04, "learning_rate": 9.669837435200564e-05, "loss": 1.0431, "step": 14125 }, { "epoch": 0.04, "learning_rate": 9.669711611052393e-05, "loss": 1.0358, "step": 14130 }, { "epoch": 0.04, "learning_rate": 9.669585786904222e-05, "loss": 1.0336, "step": 14135 }, { "epoch": 0.04, "learning_rate": 9.669459962756053e-05, "loss": 1.0355, "step": 14140 }, { "epoch": 0.04, "learning_rate": 9.669334138607882e-05, "loss": 1.0365, "step": 14145 }, { "epoch": 0.04, "learning_rate": 9.669208314459711e-05, "loss": 1.0351, "step": 14150 }, { "epoch": 0.04, "learning_rate": 9.66908249031154e-05, "loss": 1.0388, "step": 14155 }, { "epoch": 0.04, "learning_rate": 9.66895666616337e-05, "loss": 1.0338, "step": 14160 }, { "epoch": 0.04, "learning_rate": 9.6688308420152e-05, "loss": 1.0348, "step": 14165 }, { "epoch": 0.04, "learning_rate": 9.668705017867029e-05, "loss": 1.0367, "step": 14170 }, { "epoch": 0.04, "learning_rate": 9.668579193718858e-05, "loss": 1.0363, "step": 14175 }, { "epoch": 0.04, "learning_rate": 9.668453369570688e-05, "loss": 1.0347, "step": 14180 }, { "epoch": 0.04, "learning_rate": 9.668327545422518e-05, "loss": 1.0367, "step": 14185 }, { "epoch": 0.04, "learning_rate": 9.668201721274347e-05, "loss": 1.0385, "step": 14190 }, { "epoch": 0.04, "learning_rate": 9.668101061955811e-05, "loss": 1.0338, "step": 14195 }, { "epoch": 0.04, "learning_rate": 9.66797523780764e-05, "loss": 1.035, "step": 14200 }, { "epoch": 0.04, "learning_rate": 9.66784941365947e-05, "loss": 1.0368, "step": 14205 }, { "epoch": 0.04, "learning_rate": 9.6677235895113e-05, "loss": 1.0329, "step": 14210 }, { "epoch": 0.04, "learning_rate": 9.667597765363129e-05, "loss": 1.0361, "step": 14215 }, { "epoch": 0.04, "learning_rate": 9.667471941214958e-05, "loss": 1.0352, "step": 14220 }, { "epoch": 0.04, "learning_rate": 9.667346117066788e-05, "loss": 1.0339, "step": 14225 }, { "epoch": 0.04, "learning_rate": 9.667220292918618e-05, "loss": 1.0339, "step": 14230 }, { "epoch": 0.04, "learning_rate": 9.667094468770447e-05, "loss": 1.0366, "step": 14235 }, { "epoch": 0.04, "learning_rate": 9.666968644622276e-05, "loss": 1.0343, "step": 14240 }, { "epoch": 0.04, "learning_rate": 9.666842820474105e-05, "loss": 1.0369, "step": 14245 }, { "epoch": 0.04, "learning_rate": 9.666716996325936e-05, "loss": 1.0354, "step": 14250 }, { "epoch": 0.04, "learning_rate": 9.666591172177765e-05, "loss": 1.0346, "step": 14255 }, { "epoch": 0.04, "learning_rate": 9.666465348029594e-05, "loss": 1.0376, "step": 14260 }, { "epoch": 0.04, "learning_rate": 9.666339523881423e-05, "loss": 1.0353, "step": 14265 }, { "epoch": 0.04, "learning_rate": 9.666213699733254e-05, "loss": 1.0355, "step": 14270 }, { "epoch": 0.04, "learning_rate": 9.666087875585083e-05, "loss": 1.0384, "step": 14275 }, { "epoch": 0.04, "learning_rate": 9.665962051436912e-05, "loss": 1.0349, "step": 14280 }, { "epoch": 0.04, "learning_rate": 9.665836227288741e-05, "loss": 1.0339, "step": 14285 }, { "epoch": 0.04, "learning_rate": 9.665710403140572e-05, "loss": 1.0355, "step": 14290 }, { "epoch": 0.04, "learning_rate": 9.665584578992401e-05, "loss": 1.0344, "step": 14295 }, { "epoch": 0.04, "learning_rate": 9.66545875484423e-05, "loss": 1.0352, "step": 14300 }, { "epoch": 0.04, "learning_rate": 9.665332930696059e-05, "loss": 1.0357, "step": 14305 }, { "epoch": 0.04, "learning_rate": 9.665207106547888e-05, "loss": 1.0368, "step": 14310 }, { "epoch": 0.04, "learning_rate": 9.665081282399719e-05, "loss": 1.0365, "step": 14315 }, { "epoch": 0.04, "learning_rate": 9.664955458251548e-05, "loss": 1.0364, "step": 14320 }, { "epoch": 0.04, "learning_rate": 9.664829634103377e-05, "loss": 1.035, "step": 14325 }, { "epoch": 0.04, "learning_rate": 9.664703809955206e-05, "loss": 1.037, "step": 14330 }, { "epoch": 0.04, "learning_rate": 9.664577985807037e-05, "loss": 1.0368, "step": 14335 }, { "epoch": 0.04, "learning_rate": 9.664452161658866e-05, "loss": 1.035, "step": 14340 }, { "epoch": 0.04, "learning_rate": 9.664326337510695e-05, "loss": 1.0329, "step": 14345 }, { "epoch": 0.04, "learning_rate": 9.664200513362524e-05, "loss": 1.0332, "step": 14350 }, { "epoch": 0.04, "learning_rate": 9.664074689214355e-05, "loss": 1.0374, "step": 14355 }, { "epoch": 0.04, "learning_rate": 9.663948865066184e-05, "loss": 1.0358, "step": 14360 }, { "epoch": 0.04, "learning_rate": 9.663823040918013e-05, "loss": 1.0386, "step": 14365 }, { "epoch": 0.04, "learning_rate": 9.663697216769842e-05, "loss": 1.0339, "step": 14370 }, { "epoch": 0.04, "learning_rate": 9.663571392621671e-05, "loss": 1.0369, "step": 14375 }, { "epoch": 0.04, "learning_rate": 9.663445568473502e-05, "loss": 1.0576, "step": 14380 }, { "epoch": 0.04, "learning_rate": 9.663319744325331e-05, "loss": 1.0354, "step": 14385 }, { "epoch": 0.04, "learning_rate": 9.66319392017716e-05, "loss": 1.0367, "step": 14390 }, { "epoch": 0.04, "learning_rate": 9.66306809602899e-05, "loss": 1.0359, "step": 14395 }, { "epoch": 0.04, "learning_rate": 9.66294227188082e-05, "loss": 1.0365, "step": 14400 }, { "epoch": 0.04, "learning_rate": 9.662816447732649e-05, "loss": 1.036, "step": 14405 }, { "epoch": 0.04, "learning_rate": 9.662690623584478e-05, "loss": 1.0351, "step": 14410 }, { "epoch": 0.04, "learning_rate": 9.662564799436307e-05, "loss": 1.0351, "step": 14415 }, { "epoch": 0.04, "learning_rate": 9.662438975288138e-05, "loss": 1.0368, "step": 14420 }, { "epoch": 0.04, "learning_rate": 9.662313151139967e-05, "loss": 1.0351, "step": 14425 }, { "epoch": 0.04, "learning_rate": 9.662187326991796e-05, "loss": 1.0359, "step": 14430 }, { "epoch": 0.04, "learning_rate": 9.662061502843625e-05, "loss": 1.0367, "step": 14435 }, { "epoch": 0.04, "learning_rate": 9.661935678695456e-05, "loss": 1.0363, "step": 14440 }, { "epoch": 0.04, "learning_rate": 9.661809854547286e-05, "loss": 1.0382, "step": 14445 }, { "epoch": 0.04, "learning_rate": 9.661684030399116e-05, "loss": 1.0349, "step": 14450 }, { "epoch": 0.04, "learning_rate": 9.661558206250945e-05, "loss": 1.0362, "step": 14455 }, { "epoch": 0.04, "learning_rate": 9.661432382102774e-05, "loss": 1.0345, "step": 14460 }, { "epoch": 0.04, "learning_rate": 9.661306557954603e-05, "loss": 1.035, "step": 14465 }, { "epoch": 0.04, "learning_rate": 9.661180733806434e-05, "loss": 1.0339, "step": 14470 }, { "epoch": 0.04, "learning_rate": 9.661054909658263e-05, "loss": 1.0347, "step": 14475 }, { "epoch": 0.04, "learning_rate": 9.660929085510092e-05, "loss": 1.0355, "step": 14480 }, { "epoch": 0.04, "learning_rate": 9.660803261361921e-05, "loss": 1.0361, "step": 14485 }, { "epoch": 0.04, "learning_rate": 9.660677437213752e-05, "loss": 1.036, "step": 14490 }, { "epoch": 0.04, "learning_rate": 9.660551613065581e-05, "loss": 1.0354, "step": 14495 }, { "epoch": 0.04, "learning_rate": 9.66042578891741e-05, "loss": 1.0366, "step": 14500 }, { "epoch": 0.04, "learning_rate": 9.660299964769239e-05, "loss": 1.0362, "step": 14505 }, { "epoch": 0.04, "learning_rate": 9.660174140621068e-05, "loss": 1.037, "step": 14510 }, { "epoch": 0.04, "learning_rate": 9.660048316472899e-05, "loss": 1.0345, "step": 14515 }, { "epoch": 0.04, "learning_rate": 9.659922492324728e-05, "loss": 1.0335, "step": 14520 }, { "epoch": 0.04, "learning_rate": 9.659796668176557e-05, "loss": 1.0368, "step": 14525 }, { "epoch": 0.04, "learning_rate": 9.659670844028386e-05, "loss": 1.035, "step": 14530 }, { "epoch": 0.04, "learning_rate": 9.659545019880217e-05, "loss": 1.0339, "step": 14535 }, { "epoch": 0.04, "learning_rate": 9.659419195732046e-05, "loss": 1.0361, "step": 14540 }, { "epoch": 0.04, "learning_rate": 9.659293371583875e-05, "loss": 1.0362, "step": 14545 }, { "epoch": 0.04, "learning_rate": 9.659167547435704e-05, "loss": 1.0356, "step": 14550 }, { "epoch": 0.04, "learning_rate": 9.659041723287535e-05, "loss": 1.0343, "step": 14555 }, { "epoch": 0.04, "learning_rate": 9.658915899139364e-05, "loss": 1.0348, "step": 14560 }, { "epoch": 0.04, "learning_rate": 9.658790074991193e-05, "loss": 1.0376, "step": 14565 }, { "epoch": 0.04, "learning_rate": 9.658664250843022e-05, "loss": 1.0364, "step": 14570 }, { "epoch": 0.04, "learning_rate": 9.658538426694851e-05, "loss": 1.0357, "step": 14575 }, { "epoch": 0.04, "learning_rate": 9.658412602546682e-05, "loss": 1.0367, "step": 14580 }, { "epoch": 0.04, "learning_rate": 9.658286778398511e-05, "loss": 1.0347, "step": 14585 }, { "epoch": 0.04, "learning_rate": 9.65816095425034e-05, "loss": 1.0362, "step": 14590 }, { "epoch": 0.04, "learning_rate": 9.658035130102169e-05, "loss": 1.0329, "step": 14595 }, { "epoch": 0.04, "learning_rate": 9.657909305954e-05, "loss": 1.0341, "step": 14600 }, { "epoch": 0.04, "learning_rate": 9.657783481805829e-05, "loss": 1.0354, "step": 14605 }, { "epoch": 0.04, "learning_rate": 9.657657657657658e-05, "loss": 1.0359, "step": 14610 }, { "epoch": 0.04, "learning_rate": 9.657531833509487e-05, "loss": 1.0374, "step": 14615 }, { "epoch": 0.04, "learning_rate": 9.657406009361318e-05, "loss": 1.0348, "step": 14620 }, { "epoch": 0.04, "learning_rate": 9.657280185213147e-05, "loss": 1.0356, "step": 14625 }, { "epoch": 0.04, "learning_rate": 9.657154361064976e-05, "loss": 1.0358, "step": 14630 }, { "epoch": 0.04, "learning_rate": 9.657028536916805e-05, "loss": 1.035, "step": 14635 }, { "epoch": 0.04, "learning_rate": 9.656902712768634e-05, "loss": 1.039, "step": 14640 }, { "epoch": 0.04, "learning_rate": 9.656776888620465e-05, "loss": 1.0357, "step": 14645 }, { "epoch": 0.04, "learning_rate": 9.656651064472294e-05, "loss": 1.0341, "step": 14650 }, { "epoch": 0.04, "learning_rate": 9.656525240324123e-05, "loss": 1.0358, "step": 14655 }, { "epoch": 0.04, "learning_rate": 9.656399416175952e-05, "loss": 1.0364, "step": 14660 }, { "epoch": 0.04, "learning_rate": 9.656273592027783e-05, "loss": 1.0359, "step": 14665 }, { "epoch": 0.04, "learning_rate": 9.656147767879612e-05, "loss": 1.0353, "step": 14670 }, { "epoch": 0.04, "learning_rate": 9.656021943731441e-05, "loss": 1.0351, "step": 14675 }, { "epoch": 0.04, "learning_rate": 9.65589611958327e-05, "loss": 1.0349, "step": 14680 }, { "epoch": 0.04, "learning_rate": 9.655770295435101e-05, "loss": 1.0367, "step": 14685 }, { "epoch": 0.04, "learning_rate": 9.65564447128693e-05, "loss": 1.0367, "step": 14690 }, { "epoch": 0.04, "learning_rate": 9.655518647138759e-05, "loss": 1.035, "step": 14695 }, { "epoch": 0.04, "learning_rate": 9.655392822990588e-05, "loss": 1.0351, "step": 14700 }, { "epoch": 0.04, "learning_rate": 9.655266998842418e-05, "loss": 1.034, "step": 14705 }, { "epoch": 0.04, "learning_rate": 9.655141174694248e-05, "loss": 1.0356, "step": 14710 }, { "epoch": 0.04, "learning_rate": 9.655015350546077e-05, "loss": 1.0354, "step": 14715 }, { "epoch": 0.04, "learning_rate": 9.654889526397906e-05, "loss": 1.0353, "step": 14720 }, { "epoch": 0.04, "learning_rate": 9.654763702249736e-05, "loss": 1.0347, "step": 14725 }, { "epoch": 0.04, "learning_rate": 9.654637878101566e-05, "loss": 1.0355, "step": 14730 }, { "epoch": 0.04, "learning_rate": 9.654512053953395e-05, "loss": 1.0349, "step": 14735 }, { "epoch": 0.04, "learning_rate": 9.654386229805224e-05, "loss": 1.0355, "step": 14740 }, { "epoch": 0.04, "learning_rate": 9.654260405657054e-05, "loss": 1.0356, "step": 14745 }, { "epoch": 0.04, "learning_rate": 9.654134581508884e-05, "loss": 1.0354, "step": 14750 }, { "epoch": 0.04, "learning_rate": 9.654008757360713e-05, "loss": 1.0358, "step": 14755 }, { "epoch": 0.04, "learning_rate": 9.653882933212542e-05, "loss": 1.0338, "step": 14760 }, { "epoch": 0.04, "learning_rate": 9.653757109064372e-05, "loss": 1.0344, "step": 14765 }, { "epoch": 0.04, "learning_rate": 9.653631284916201e-05, "loss": 1.036, "step": 14770 }, { "epoch": 0.04, "learning_rate": 9.653505460768031e-05, "loss": 1.0355, "step": 14775 }, { "epoch": 0.04, "learning_rate": 9.65337963661986e-05, "loss": 1.035, "step": 14780 }, { "epoch": 0.04, "learning_rate": 9.65325381247169e-05, "loss": 1.0344, "step": 14785 }, { "epoch": 0.04, "learning_rate": 9.653127988323519e-05, "loss": 1.0364, "step": 14790 }, { "epoch": 0.04, "learning_rate": 9.653002164175349e-05, "loss": 1.036, "step": 14795 }, { "epoch": 0.04, "learning_rate": 9.652876340027178e-05, "loss": 1.0331, "step": 14800 }, { "epoch": 0.04, "learning_rate": 9.652750515879008e-05, "loss": 1.0362, "step": 14805 }, { "epoch": 0.04, "learning_rate": 9.652624691730837e-05, "loss": 1.0355, "step": 14810 }, { "epoch": 0.04, "learning_rate": 9.652498867582667e-05, "loss": 1.0351, "step": 14815 }, { "epoch": 0.04, "learning_rate": 9.652373043434496e-05, "loss": 1.0341, "step": 14820 }, { "epoch": 0.04, "learning_rate": 9.652247219286326e-05, "loss": 1.0357, "step": 14825 }, { "epoch": 0.04, "learning_rate": 9.652121395138155e-05, "loss": 1.0357, "step": 14830 }, { "epoch": 0.04, "learning_rate": 9.651995570989984e-05, "loss": 1.0366, "step": 14835 }, { "epoch": 0.04, "learning_rate": 9.651869746841814e-05, "loss": 1.0343, "step": 14840 }, { "epoch": 0.04, "learning_rate": 9.651743922693644e-05, "loss": 1.035, "step": 14845 }, { "epoch": 0.04, "learning_rate": 9.651618098545473e-05, "loss": 1.0576, "step": 14850 }, { "epoch": 0.04, "learning_rate": 9.651492274397302e-05, "loss": 1.0356, "step": 14855 }, { "epoch": 0.04, "learning_rate": 9.651366450249132e-05, "loss": 1.036, "step": 14860 }, { "epoch": 0.04, "learning_rate": 9.651240626100962e-05, "loss": 1.0348, "step": 14865 }, { "epoch": 0.04, "learning_rate": 9.651114801952791e-05, "loss": 1.0362, "step": 14870 }, { "epoch": 0.04, "learning_rate": 9.65098897780462e-05, "loss": 1.0339, "step": 14875 }, { "epoch": 0.04, "learning_rate": 9.65086315365645e-05, "loss": 1.0348, "step": 14880 }, { "epoch": 0.04, "learning_rate": 9.65073732950828e-05, "loss": 1.0359, "step": 14885 }, { "epoch": 0.04, "learning_rate": 9.650611505360109e-05, "loss": 1.0342, "step": 14890 }, { "epoch": 0.04, "learning_rate": 9.650485681211938e-05, "loss": 1.0355, "step": 14895 }, { "epoch": 0.04, "learning_rate": 9.650359857063767e-05, "loss": 1.0363, "step": 14900 }, { "epoch": 0.04, "learning_rate": 9.650234032915598e-05, "loss": 1.0366, "step": 14905 }, { "epoch": 0.04, "learning_rate": 9.650108208767427e-05, "loss": 1.0344, "step": 14910 }, { "epoch": 0.04, "learning_rate": 9.649982384619256e-05, "loss": 1.0345, "step": 14915 }, { "epoch": 0.04, "learning_rate": 9.649856560471085e-05, "loss": 1.0337, "step": 14920 }, { "epoch": 0.04, "learning_rate": 9.649730736322916e-05, "loss": 1.0354, "step": 14925 }, { "epoch": 0.04, "learning_rate": 9.649604912174745e-05, "loss": 1.0351, "step": 14930 }, { "epoch": 0.04, "learning_rate": 9.649479088026575e-05, "loss": 1.0359, "step": 14935 }, { "epoch": 0.04, "learning_rate": 9.649353263878404e-05, "loss": 1.0359, "step": 14940 }, { "epoch": 0.04, "learning_rate": 9.649227439730234e-05, "loss": 1.035, "step": 14945 }, { "epoch": 0.04, "learning_rate": 9.649101615582064e-05, "loss": 1.0354, "step": 14950 }, { "epoch": 0.04, "learning_rate": 9.648975791433893e-05, "loss": 1.0355, "step": 14955 }, { "epoch": 0.04, "learning_rate": 9.648849967285722e-05, "loss": 1.0351, "step": 14960 }, { "epoch": 0.04, "learning_rate": 9.648724143137552e-05, "loss": 1.0352, "step": 14965 }, { "epoch": 0.04, "learning_rate": 9.648598318989382e-05, "loss": 1.0519, "step": 14970 }, { "epoch": 0.04, "learning_rate": 9.648472494841211e-05, "loss": 1.0359, "step": 14975 }, { "epoch": 0.04, "learning_rate": 9.64834667069304e-05, "loss": 1.0359, "step": 14980 }, { "epoch": 0.04, "learning_rate": 9.64822084654487e-05, "loss": 1.0558, "step": 14985 }, { "epoch": 0.04, "learning_rate": 9.648095022396699e-05, "loss": 1.0352, "step": 14990 }, { "epoch": 0.04, "learning_rate": 9.647969198248529e-05, "loss": 1.0375, "step": 14995 }, { "epoch": 0.04, "learning_rate": 9.647843374100358e-05, "loss": 1.0362, "step": 15000 }, { "epoch": 0.04, "learning_rate": 9.647717549952188e-05, "loss": 1.0338, "step": 15005 }, { "epoch": 0.04, "learning_rate": 9.647591725804017e-05, "loss": 1.0353, "step": 15010 }, { "epoch": 0.04, "learning_rate": 9.647465901655847e-05, "loss": 1.0336, "step": 15015 }, { "epoch": 0.04, "learning_rate": 9.647340077507676e-05, "loss": 1.0369, "step": 15020 }, { "epoch": 0.04, "learning_rate": 9.647214253359506e-05, "loss": 1.0368, "step": 15025 }, { "epoch": 0.04, "learning_rate": 9.647088429211335e-05, "loss": 1.0356, "step": 15030 }, { "epoch": 0.04, "learning_rate": 9.646962605063165e-05, "loss": 1.0349, "step": 15035 }, { "epoch": 0.04, "learning_rate": 9.646861945744627e-05, "loss": 1.0337, "step": 15040 }, { "epoch": 0.04, "learning_rate": 9.646736121596457e-05, "loss": 1.0362, "step": 15045 }, { "epoch": 0.04, "learning_rate": 9.646610297448286e-05, "loss": 1.0372, "step": 15050 }, { "epoch": 0.04, "learning_rate": 9.646484473300116e-05, "loss": 1.0373, "step": 15055 }, { "epoch": 0.04, "learning_rate": 9.646358649151945e-05, "loss": 1.0364, "step": 15060 }, { "epoch": 0.04, "learning_rate": 9.646257989833409e-05, "loss": 1.0556, "step": 15065 }, { "epoch": 0.04, "learning_rate": 9.646132165685238e-05, "loss": 1.0341, "step": 15070 }, { "epoch": 0.04, "learning_rate": 9.646006341537069e-05, "loss": 1.0345, "step": 15075 }, { "epoch": 0.04, "learning_rate": 9.645880517388898e-05, "loss": 1.0327, "step": 15080 }, { "epoch": 0.04, "learning_rate": 9.645754693240727e-05, "loss": 1.0358, "step": 15085 }, { "epoch": 0.04, "learning_rate": 9.645628869092556e-05, "loss": 1.0356, "step": 15090 }, { "epoch": 0.04, "learning_rate": 9.645503044944387e-05, "loss": 1.0356, "step": 15095 }, { "epoch": 0.04, "learning_rate": 9.645377220796216e-05, "loss": 1.0356, "step": 15100 }, { "epoch": 0.04, "learning_rate": 9.645251396648045e-05, "loss": 1.0365, "step": 15105 }, { "epoch": 0.04, "learning_rate": 9.645125572499874e-05, "loss": 1.0345, "step": 15110 }, { "epoch": 0.04, "learning_rate": 9.644999748351705e-05, "loss": 1.0363, "step": 15115 }, { "epoch": 0.04, "learning_rate": 9.644873924203534e-05, "loss": 1.0355, "step": 15120 }, { "epoch": 0.04, "learning_rate": 9.644748100055363e-05, "loss": 1.0329, "step": 15125 }, { "epoch": 0.04, "learning_rate": 9.644622275907192e-05, "loss": 1.0359, "step": 15130 }, { "epoch": 0.04, "learning_rate": 9.644496451759021e-05, "loss": 1.0378, "step": 15135 }, { "epoch": 0.04, "learning_rate": 9.644370627610852e-05, "loss": 1.0353, "step": 15140 }, { "epoch": 0.04, "learning_rate": 9.644244803462681e-05, "loss": 1.0365, "step": 15145 }, { "epoch": 0.04, "learning_rate": 9.64411897931451e-05, "loss": 1.0338, "step": 15150 }, { "epoch": 0.04, "learning_rate": 9.643993155166339e-05, "loss": 1.0334, "step": 15155 }, { "epoch": 0.04, "learning_rate": 9.64386733101817e-05, "loss": 1.0375, "step": 15160 }, { "epoch": 0.04, "learning_rate": 9.643741506869999e-05, "loss": 1.0366, "step": 15165 }, { "epoch": 0.04, "learning_rate": 9.643615682721828e-05, "loss": 1.0365, "step": 15170 }, { "epoch": 0.04, "learning_rate": 9.643489858573657e-05, "loss": 1.0344, "step": 15175 }, { "epoch": 0.04, "learning_rate": 9.643364034425488e-05, "loss": 1.0357, "step": 15180 }, { "epoch": 0.04, "learning_rate": 9.643238210277317e-05, "loss": 1.0372, "step": 15185 }, { "epoch": 0.04, "learning_rate": 9.643112386129146e-05, "loss": 1.0341, "step": 15190 }, { "epoch": 0.04, "learning_rate": 9.642986561980975e-05, "loss": 1.0354, "step": 15195 }, { "epoch": 0.04, "learning_rate": 9.642860737832804e-05, "loss": 1.0362, "step": 15200 }, { "epoch": 0.04, "learning_rate": 9.642734913684635e-05, "loss": 1.0363, "step": 15205 }, { "epoch": 0.04, "learning_rate": 9.642609089536464e-05, "loss": 1.0342, "step": 15210 }, { "epoch": 0.04, "learning_rate": 9.642483265388293e-05, "loss": 1.0356, "step": 15215 }, { "epoch": 0.04, "learning_rate": 9.642357441240122e-05, "loss": 1.0354, "step": 15220 }, { "epoch": 0.04, "learning_rate": 9.642231617091953e-05, "loss": 1.035, "step": 15225 }, { "epoch": 0.04, "learning_rate": 9.642105792943782e-05, "loss": 1.0352, "step": 15230 }, { "epoch": 0.04, "learning_rate": 9.641979968795611e-05, "loss": 1.034, "step": 15235 }, { "epoch": 0.04, "learning_rate": 9.64185414464744e-05, "loss": 1.0338, "step": 15240 }, { "epoch": 0.04, "learning_rate": 9.641728320499271e-05, "loss": 1.0352, "step": 15245 }, { "epoch": 0.04, "learning_rate": 9.6416024963511e-05, "loss": 1.0362, "step": 15250 }, { "epoch": 0.04, "learning_rate": 9.641476672202929e-05, "loss": 1.0344, "step": 15255 }, { "epoch": 0.04, "learning_rate": 9.641350848054758e-05, "loss": 1.0347, "step": 15260 }, { "epoch": 0.04, "learning_rate": 9.641225023906588e-05, "loss": 1.0329, "step": 15265 }, { "epoch": 0.04, "learning_rate": 9.641099199758418e-05, "loss": 1.0364, "step": 15270 }, { "epoch": 0.04, "learning_rate": 9.640973375610247e-05, "loss": 1.0346, "step": 15275 }, { "epoch": 0.04, "learning_rate": 9.640847551462076e-05, "loss": 1.0365, "step": 15280 }, { "epoch": 0.04, "learning_rate": 9.640721727313906e-05, "loss": 1.0359, "step": 15285 }, { "epoch": 0.04, "learning_rate": 9.640595903165736e-05, "loss": 1.0375, "step": 15290 }, { "epoch": 0.04, "learning_rate": 9.640470079017565e-05, "loss": 1.0378, "step": 15295 }, { "epoch": 0.04, "learning_rate": 9.640344254869394e-05, "loss": 1.0323, "step": 15300 }, { "epoch": 0.04, "learning_rate": 9.640218430721224e-05, "loss": 1.0354, "step": 15305 }, { "epoch": 0.04, "learning_rate": 9.640092606573054e-05, "loss": 1.0373, "step": 15310 }, { "epoch": 0.04, "learning_rate": 9.639966782424883e-05, "loss": 1.0335, "step": 15315 }, { "epoch": 0.04, "learning_rate": 9.639840958276712e-05, "loss": 1.0354, "step": 15320 }, { "epoch": 0.04, "learning_rate": 9.639715134128542e-05, "loss": 1.0356, "step": 15325 }, { "epoch": 0.04, "learning_rate": 9.639589309980371e-05, "loss": 1.0371, "step": 15330 }, { "epoch": 0.04, "learning_rate": 9.639463485832201e-05, "loss": 1.0349, "step": 15335 }, { "epoch": 0.04, "learning_rate": 9.639337661684032e-05, "loss": 1.0334, "step": 15340 }, { "epoch": 0.04, "learning_rate": 9.639211837535861e-05, "loss": 1.0341, "step": 15345 }, { "epoch": 0.04, "learning_rate": 9.63908601338769e-05, "loss": 1.0339, "step": 15350 }, { "epoch": 0.04, "learning_rate": 9.638960189239519e-05, "loss": 1.0344, "step": 15355 }, { "epoch": 0.04, "learning_rate": 9.63883436509135e-05, "loss": 1.0355, "step": 15360 }, { "epoch": 0.04, "learning_rate": 9.638708540943179e-05, "loss": 1.0353, "step": 15365 }, { "epoch": 0.04, "learning_rate": 9.638582716795008e-05, "loss": 1.0375, "step": 15370 }, { "epoch": 0.04, "learning_rate": 9.638456892646837e-05, "loss": 1.0352, "step": 15375 }, { "epoch": 0.04, "learning_rate": 9.638331068498668e-05, "loss": 1.0347, "step": 15380 }, { "epoch": 0.04, "learning_rate": 9.638205244350497e-05, "loss": 1.0353, "step": 15385 }, { "epoch": 0.04, "learning_rate": 9.638079420202326e-05, "loss": 1.0356, "step": 15390 }, { "epoch": 0.04, "learning_rate": 9.637953596054155e-05, "loss": 1.0343, "step": 15395 }, { "epoch": 0.04, "learning_rate": 9.637827771905984e-05, "loss": 1.0331, "step": 15400 }, { "epoch": 0.04, "learning_rate": 9.637701947757815e-05, "loss": 1.0323, "step": 15405 }, { "epoch": 0.04, "learning_rate": 9.637576123609644e-05, "loss": 1.0384, "step": 15410 }, { "epoch": 0.04, "learning_rate": 9.637450299461473e-05, "loss": 1.0359, "step": 15415 }, { "epoch": 0.04, "learning_rate": 9.637324475313302e-05, "loss": 1.0384, "step": 15420 }, { "epoch": 0.04, "learning_rate": 9.637198651165133e-05, "loss": 1.0364, "step": 15425 }, { "epoch": 0.04, "learning_rate": 9.637072827016962e-05, "loss": 1.0356, "step": 15430 }, { "epoch": 0.04, "learning_rate": 9.636947002868791e-05, "loss": 1.0326, "step": 15435 }, { "epoch": 0.04, "learning_rate": 9.63682117872062e-05, "loss": 1.0345, "step": 15440 }, { "epoch": 0.04, "learning_rate": 9.636695354572451e-05, "loss": 1.0347, "step": 15445 }, { "epoch": 0.04, "learning_rate": 9.63656953042428e-05, "loss": 1.0335, "step": 15450 }, { "epoch": 0.04, "learning_rate": 9.636443706276109e-05, "loss": 1.0578, "step": 15455 }, { "epoch": 0.04, "learning_rate": 9.636317882127938e-05, "loss": 1.0338, "step": 15460 }, { "epoch": 0.04, "learning_rate": 9.636192057979768e-05, "loss": 1.0353, "step": 15465 }, { "epoch": 0.04, "learning_rate": 9.636066233831598e-05, "loss": 1.0326, "step": 15470 }, { "epoch": 0.04, "learning_rate": 9.635940409683427e-05, "loss": 1.0347, "step": 15475 }, { "epoch": 0.04, "learning_rate": 9.635814585535256e-05, "loss": 1.0393, "step": 15480 }, { "epoch": 0.04, "learning_rate": 9.635688761387086e-05, "loss": 1.0336, "step": 15485 }, { "epoch": 0.04, "learning_rate": 9.635562937238916e-05, "loss": 1.0352, "step": 15490 }, { "epoch": 0.04, "learning_rate": 9.635437113090745e-05, "loss": 1.0362, "step": 15495 }, { "epoch": 0.04, "learning_rate": 9.635311288942574e-05, "loss": 1.0338, "step": 15500 }, { "epoch": 0.04, "learning_rate": 9.635185464794404e-05, "loss": 1.0338, "step": 15505 }, { "epoch": 0.04, "learning_rate": 9.635059640646234e-05, "loss": 1.0342, "step": 15510 }, { "epoch": 0.04, "learning_rate": 9.634933816498063e-05, "loss": 1.0346, "step": 15515 }, { "epoch": 0.04, "learning_rate": 9.634807992349892e-05, "loss": 1.0347, "step": 15520 }, { "epoch": 0.04, "learning_rate": 9.634682168201722e-05, "loss": 1.0349, "step": 15525 }, { "epoch": 0.04, "learning_rate": 9.634556344053551e-05, "loss": 1.0326, "step": 15530 }, { "epoch": 0.04, "learning_rate": 9.634430519905381e-05, "loss": 1.0339, "step": 15535 }, { "epoch": 0.04, "learning_rate": 9.63430469575721e-05, "loss": 1.0361, "step": 15540 }, { "epoch": 0.04, "learning_rate": 9.63417887160904e-05, "loss": 1.0359, "step": 15545 }, { "epoch": 0.04, "learning_rate": 9.634053047460869e-05, "loss": 1.0348, "step": 15550 }, { "epoch": 0.04, "learning_rate": 9.633927223312699e-05, "loss": 1.0343, "step": 15555 }, { "epoch": 0.04, "learning_rate": 9.633801399164528e-05, "loss": 1.0354, "step": 15560 }, { "epoch": 0.04, "learning_rate": 9.633675575016358e-05, "loss": 1.0365, "step": 15565 }, { "epoch": 0.04, "learning_rate": 9.633549750868187e-05, "loss": 1.0348, "step": 15570 }, { "epoch": 0.04, "learning_rate": 9.633423926720017e-05, "loss": 1.0352, "step": 15575 }, { "epoch": 0.04, "learning_rate": 9.633298102571846e-05, "loss": 1.0344, "step": 15580 }, { "epoch": 0.04, "learning_rate": 9.633172278423676e-05, "loss": 1.0566, "step": 15585 }, { "epoch": 0.04, "learning_rate": 9.633046454275505e-05, "loss": 1.035, "step": 15590 }, { "epoch": 0.04, "learning_rate": 9.632920630127334e-05, "loss": 1.036, "step": 15595 }, { "epoch": 0.04, "learning_rate": 9.632794805979164e-05, "loss": 1.0349, "step": 15600 }, { "epoch": 0.04, "learning_rate": 9.632668981830994e-05, "loss": 1.0344, "step": 15605 }, { "epoch": 0.04, "learning_rate": 9.632543157682823e-05, "loss": 1.0568, "step": 15610 }, { "epoch": 0.04, "learning_rate": 9.632417333534652e-05, "loss": 1.034, "step": 15615 }, { "epoch": 0.04, "learning_rate": 9.632291509386482e-05, "loss": 1.0356, "step": 15620 }, { "epoch": 0.04, "learning_rate": 9.632165685238312e-05, "loss": 1.0366, "step": 15625 }, { "epoch": 0.04, "learning_rate": 9.63203986109014e-05, "loss": 1.0356, "step": 15630 }, { "epoch": 0.04, "learning_rate": 9.63191403694197e-05, "loss": 1.0354, "step": 15635 }, { "epoch": 0.04, "learning_rate": 9.6317882127938e-05, "loss": 1.0341, "step": 15640 }, { "epoch": 0.04, "learning_rate": 9.63166238864563e-05, "loss": 1.0348, "step": 15645 }, { "epoch": 0.04, "learning_rate": 9.631536564497459e-05, "loss": 1.0352, "step": 15650 }, { "epoch": 0.04, "learning_rate": 9.631410740349288e-05, "loss": 1.0355, "step": 15655 }, { "epoch": 0.04, "learning_rate": 9.631284916201117e-05, "loss": 1.0346, "step": 15660 }, { "epoch": 0.04, "learning_rate": 9.631159092052947e-05, "loss": 1.0356, "step": 15665 }, { "epoch": 0.04, "learning_rate": 9.631033267904777e-05, "loss": 1.0343, "step": 15670 }, { "epoch": 0.04, "learning_rate": 9.630907443756606e-05, "loss": 1.0358, "step": 15675 }, { "epoch": 0.04, "learning_rate": 9.630781619608435e-05, "loss": 1.0357, "step": 15680 }, { "epoch": 0.04, "learning_rate": 9.630655795460265e-05, "loss": 1.0344, "step": 15685 }, { "epoch": 0.04, "learning_rate": 9.630529971312095e-05, "loss": 1.0346, "step": 15690 }, { "epoch": 0.04, "learning_rate": 9.630404147163924e-05, "loss": 1.0354, "step": 15695 }, { "epoch": 0.04, "learning_rate": 9.630278323015753e-05, "loss": 1.0323, "step": 15700 }, { "epoch": 0.04, "learning_rate": 9.630152498867583e-05, "loss": 1.0335, "step": 15705 }, { "epoch": 0.04, "learning_rate": 9.630026674719413e-05, "loss": 1.0362, "step": 15710 }, { "epoch": 0.04, "learning_rate": 9.629900850571242e-05, "loss": 1.0341, "step": 15715 }, { "epoch": 0.04, "learning_rate": 9.629775026423071e-05, "loss": 1.0351, "step": 15720 }, { "epoch": 0.04, "learning_rate": 9.6296492022749e-05, "loss": 1.0345, "step": 15725 }, { "epoch": 0.04, "learning_rate": 9.62952337812673e-05, "loss": 1.0355, "step": 15730 }, { "epoch": 0.04, "learning_rate": 9.62939755397856e-05, "loss": 1.035, "step": 15735 }, { "epoch": 0.04, "learning_rate": 9.629271729830389e-05, "loss": 1.0368, "step": 15740 }, { "epoch": 0.04, "learning_rate": 9.629145905682218e-05, "loss": 1.0352, "step": 15745 }, { "epoch": 0.04, "learning_rate": 9.629020081534049e-05, "loss": 1.0368, "step": 15750 }, { "epoch": 0.04, "learning_rate": 9.628894257385878e-05, "loss": 1.0366, "step": 15755 }, { "epoch": 0.04, "learning_rate": 9.628768433237707e-05, "loss": 1.0323, "step": 15760 }, { "epoch": 0.04, "learning_rate": 9.628642609089536e-05, "loss": 1.0334, "step": 15765 }, { "epoch": 0.04, "learning_rate": 9.628516784941367e-05, "loss": 1.0352, "step": 15770 }, { "epoch": 0.04, "learning_rate": 9.628390960793196e-05, "loss": 1.0344, "step": 15775 }, { "epoch": 0.04, "learning_rate": 9.628265136645025e-05, "loss": 1.0347, "step": 15780 }, { "epoch": 0.04, "learning_rate": 9.628139312496854e-05, "loss": 1.0391, "step": 15785 }, { "epoch": 0.04, "learning_rate": 9.628013488348683e-05, "loss": 1.0321, "step": 15790 }, { "epoch": 0.04, "learning_rate": 9.627887664200514e-05, "loss": 1.0344, "step": 15795 }, { "epoch": 0.04, "learning_rate": 9.627761840052343e-05, "loss": 1.0343, "step": 15800 }, { "epoch": 0.04, "learning_rate": 9.627636015904172e-05, "loss": 1.0344, "step": 15805 }, { "epoch": 0.04, "learning_rate": 9.627510191756001e-05, "loss": 1.0353, "step": 15810 }, { "epoch": 0.04, "learning_rate": 9.627384367607832e-05, "loss": 1.0356, "step": 15815 }, { "epoch": 0.04, "learning_rate": 9.627258543459661e-05, "loss": 1.0363, "step": 15820 }, { "epoch": 0.04, "learning_rate": 9.62713271931149e-05, "loss": 1.0364, "step": 15825 }, { "epoch": 0.04, "learning_rate": 9.627006895163319e-05, "loss": 1.0366, "step": 15830 }, { "epoch": 0.04, "learning_rate": 9.62688107101515e-05, "loss": 1.0357, "step": 15835 }, { "epoch": 0.04, "learning_rate": 9.62675524686698e-05, "loss": 1.0353, "step": 15840 }, { "epoch": 0.04, "learning_rate": 9.62662942271881e-05, "loss": 1.0352, "step": 15845 }, { "epoch": 0.04, "learning_rate": 9.626503598570639e-05, "loss": 1.0328, "step": 15850 }, { "epoch": 0.04, "learning_rate": 9.626377774422468e-05, "loss": 1.0355, "step": 15855 }, { "epoch": 0.04, "learning_rate": 9.626251950274298e-05, "loss": 1.0356, "step": 15860 }, { "epoch": 0.04, "learning_rate": 9.626126126126127e-05, "loss": 1.0348, "step": 15865 }, { "epoch": 0.04, "learning_rate": 9.626000301977957e-05, "loss": 1.0327, "step": 15870 }, { "epoch": 0.04, "learning_rate": 9.625874477829786e-05, "loss": 1.0353, "step": 15875 }, { "epoch": 0.04, "learning_rate": 9.625748653681615e-05, "loss": 1.0356, "step": 15880 }, { "epoch": 0.04, "learning_rate": 9.625622829533445e-05, "loss": 1.0342, "step": 15885 }, { "epoch": 0.04, "learning_rate": 9.625497005385275e-05, "loss": 1.0346, "step": 15890 }, { "epoch": 0.04, "learning_rate": 9.625371181237104e-05, "loss": 1.0337, "step": 15895 }, { "epoch": 0.04, "learning_rate": 9.625245357088933e-05, "loss": 1.0348, "step": 15900 }, { "epoch": 0.04, "learning_rate": 9.625119532940763e-05, "loss": 1.0357, "step": 15905 }, { "epoch": 0.04, "learning_rate": 9.624993708792593e-05, "loss": 1.0368, "step": 15910 }, { "epoch": 0.04, "learning_rate": 9.624867884644422e-05, "loss": 1.0361, "step": 15915 }, { "epoch": 0.04, "learning_rate": 9.624742060496251e-05, "loss": 1.0344, "step": 15920 }, { "epoch": 0.04, "learning_rate": 9.624616236348081e-05, "loss": 1.0353, "step": 15925 }, { "epoch": 0.04, "learning_rate": 9.624515577029544e-05, "loss": 1.034, "step": 15930 }, { "epoch": 0.04, "learning_rate": 9.624389752881373e-05, "loss": 1.034, "step": 15935 }, { "epoch": 0.04, "learning_rate": 9.624263928733202e-05, "loss": 1.0345, "step": 15940 }, { "epoch": 0.04, "learning_rate": 9.624138104585032e-05, "loss": 1.0355, "step": 15945 }, { "epoch": 0.04, "learning_rate": 9.624012280436862e-05, "loss": 1.0365, "step": 15950 }, { "epoch": 0.04, "learning_rate": 9.623886456288691e-05, "loss": 1.0343, "step": 15955 }, { "epoch": 0.04, "learning_rate": 9.62376063214052e-05, "loss": 1.0346, "step": 15960 }, { "epoch": 0.04, "learning_rate": 9.62363480799235e-05, "loss": 1.0344, "step": 15965 }, { "epoch": 0.04, "learning_rate": 9.62350898384418e-05, "loss": 1.0372, "step": 15970 }, { "epoch": 0.04, "learning_rate": 9.623383159696009e-05, "loss": 1.0359, "step": 15975 }, { "epoch": 0.04, "learning_rate": 9.623257335547838e-05, "loss": 1.035, "step": 15980 }, { "epoch": 0.04, "learning_rate": 9.623131511399668e-05, "loss": 1.0346, "step": 15985 }, { "epoch": 0.04, "learning_rate": 9.623005687251498e-05, "loss": 1.035, "step": 15990 }, { "epoch": 0.04, "learning_rate": 9.622879863103327e-05, "loss": 1.0346, "step": 15995 }, { "epoch": 0.04, "learning_rate": 9.622754038955156e-05, "loss": 1.0338, "step": 16000 }, { "epoch": 0.04, "learning_rate": 9.622628214806985e-05, "loss": 1.0376, "step": 16005 }, { "epoch": 0.04, "learning_rate": 9.622502390658816e-05, "loss": 1.0353, "step": 16010 }, { "epoch": 0.04, "learning_rate": 9.622376566510645e-05, "loss": 1.0325, "step": 16015 }, { "epoch": 0.04, "learning_rate": 9.622250742362474e-05, "loss": 1.0352, "step": 16020 }, { "epoch": 0.04, "learning_rate": 9.622124918214303e-05, "loss": 1.036, "step": 16025 }, { "epoch": 0.04, "learning_rate": 9.621999094066134e-05, "loss": 1.036, "step": 16030 }, { "epoch": 0.04, "learning_rate": 9.621873269917963e-05, "loss": 1.0374, "step": 16035 }, { "epoch": 0.04, "learning_rate": 9.621747445769793e-05, "loss": 1.0365, "step": 16040 }, { "epoch": 0.04, "learning_rate": 9.621621621621622e-05, "loss": 1.0342, "step": 16045 }, { "epoch": 0.04, "learning_rate": 9.621495797473452e-05, "loss": 1.0339, "step": 16050 }, { "epoch": 0.04, "learning_rate": 9.621369973325282e-05, "loss": 1.0347, "step": 16055 }, { "epoch": 0.04, "learning_rate": 9.621244149177111e-05, "loss": 1.0348, "step": 16060 }, { "epoch": 0.04, "learning_rate": 9.62111832502894e-05, "loss": 1.0341, "step": 16065 }, { "epoch": 0.04, "learning_rate": 9.62099250088077e-05, "loss": 1.0353, "step": 16070 }, { "epoch": 0.04, "learning_rate": 9.620866676732599e-05, "loss": 1.0348, "step": 16075 }, { "epoch": 0.04, "learning_rate": 9.620740852584429e-05, "loss": 1.0359, "step": 16080 }, { "epoch": 0.04, "learning_rate": 9.620615028436258e-05, "loss": 1.0355, "step": 16085 }, { "epoch": 0.04, "learning_rate": 9.620489204288088e-05, "loss": 1.038, "step": 16090 }, { "epoch": 0.04, "learning_rate": 9.620363380139917e-05, "loss": 1.0344, "step": 16095 }, { "epoch": 0.04, "learning_rate": 9.620237555991747e-05, "loss": 1.034, "step": 16100 }, { "epoch": 0.04, "learning_rate": 9.620111731843576e-05, "loss": 1.0398, "step": 16105 }, { "epoch": 0.04, "learning_rate": 9.619985907695406e-05, "loss": 1.0346, "step": 16110 }, { "epoch": 0.04, "learning_rate": 9.619860083547235e-05, "loss": 1.0351, "step": 16115 }, { "epoch": 0.04, "learning_rate": 9.619734259399065e-05, "loss": 1.0329, "step": 16120 }, { "epoch": 0.04, "learning_rate": 9.619608435250894e-05, "loss": 1.0332, "step": 16125 }, { "epoch": 0.04, "learning_rate": 9.619482611102724e-05, "loss": 1.034, "step": 16130 }, { "epoch": 0.04, "learning_rate": 9.619356786954553e-05, "loss": 1.034, "step": 16135 }, { "epoch": 0.04, "learning_rate": 9.619230962806382e-05, "loss": 1.0328, "step": 16140 }, { "epoch": 0.04, "learning_rate": 9.619105138658212e-05, "loss": 1.0336, "step": 16145 }, { "epoch": 0.04, "learning_rate": 9.618979314510042e-05, "loss": 1.0374, "step": 16150 }, { "epoch": 0.04, "learning_rate": 9.618853490361871e-05, "loss": 1.035, "step": 16155 }, { "epoch": 0.04, "learning_rate": 9.6187276662137e-05, "loss": 1.0334, "step": 16160 }, { "epoch": 0.04, "learning_rate": 9.61860184206553e-05, "loss": 1.0352, "step": 16165 }, { "epoch": 0.04, "learning_rate": 9.61847601791736e-05, "loss": 1.0351, "step": 16170 }, { "epoch": 0.04, "learning_rate": 9.618350193769189e-05, "loss": 1.0354, "step": 16175 }, { "epoch": 0.04, "learning_rate": 9.618224369621018e-05, "loss": 1.034, "step": 16180 }, { "epoch": 0.04, "learning_rate": 9.618098545472848e-05, "loss": 1.0354, "step": 16185 }, { "epoch": 0.04, "learning_rate": 9.617972721324678e-05, "loss": 1.0365, "step": 16190 }, { "epoch": 0.04, "learning_rate": 9.617846897176507e-05, "loss": 1.0349, "step": 16195 }, { "epoch": 0.04, "learning_rate": 9.617721073028336e-05, "loss": 1.0348, "step": 16200 }, { "epoch": 0.04, "learning_rate": 9.617595248880165e-05, "loss": 1.0346, "step": 16205 }, { "epoch": 0.04, "learning_rate": 9.617469424731996e-05, "loss": 1.0339, "step": 16210 }, { "epoch": 0.04, "learning_rate": 9.617343600583825e-05, "loss": 1.0356, "step": 16215 }, { "epoch": 0.04, "learning_rate": 9.617217776435654e-05, "loss": 1.0369, "step": 16220 }, { "epoch": 0.04, "learning_rate": 9.617091952287483e-05, "loss": 1.0364, "step": 16225 }, { "epoch": 0.04, "learning_rate": 9.616966128139314e-05, "loss": 1.0342, "step": 16230 }, { "epoch": 0.04, "learning_rate": 9.616840303991143e-05, "loss": 1.0337, "step": 16235 }, { "epoch": 0.04, "learning_rate": 9.616714479842972e-05, "loss": 1.0346, "step": 16240 }, { "epoch": 0.04, "learning_rate": 9.616588655694801e-05, "loss": 1.0361, "step": 16245 }, { "epoch": 0.04, "learning_rate": 9.616462831546632e-05, "loss": 1.0347, "step": 16250 }, { "epoch": 0.04, "learning_rate": 9.616337007398461e-05, "loss": 1.0369, "step": 16255 }, { "epoch": 0.04, "learning_rate": 9.61621118325029e-05, "loss": 1.0349, "step": 16260 }, { "epoch": 0.04, "learning_rate": 9.616085359102119e-05, "loss": 1.0602, "step": 16265 }, { "epoch": 0.04, "learning_rate": 9.615959534953948e-05, "loss": 1.0332, "step": 16270 }, { "epoch": 0.04, "learning_rate": 9.615833710805779e-05, "loss": 1.036, "step": 16275 }, { "epoch": 0.04, "learning_rate": 9.615707886657608e-05, "loss": 1.0506, "step": 16280 }, { "epoch": 0.04, "learning_rate": 9.615582062509437e-05, "loss": 1.033, "step": 16285 }, { "epoch": 0.04, "learning_rate": 9.615456238361266e-05, "loss": 1.0358, "step": 16290 }, { "epoch": 0.04, "learning_rate": 9.615330414213097e-05, "loss": 1.0361, "step": 16295 }, { "epoch": 0.04, "learning_rate": 9.615204590064926e-05, "loss": 1.0352, "step": 16300 }, { "epoch": 0.04, "learning_rate": 9.615078765916755e-05, "loss": 1.0366, "step": 16305 }, { "epoch": 0.04, "learning_rate": 9.614952941768584e-05, "loss": 1.0349, "step": 16310 }, { "epoch": 0.04, "learning_rate": 9.614827117620415e-05, "loss": 1.0349, "step": 16315 }, { "epoch": 0.04, "learning_rate": 9.614701293472244e-05, "loss": 1.0343, "step": 16320 }, { "epoch": 0.04, "learning_rate": 9.614575469324073e-05, "loss": 1.0361, "step": 16325 }, { "epoch": 0.04, "learning_rate": 9.614449645175902e-05, "loss": 1.0332, "step": 16330 }, { "epoch": 0.04, "learning_rate": 9.614323821027731e-05, "loss": 1.0354, "step": 16335 }, { "epoch": 0.04, "learning_rate": 9.614197996879562e-05, "loss": 1.0366, "step": 16340 }, { "epoch": 0.04, "learning_rate": 9.614072172731391e-05, "loss": 1.0338, "step": 16345 }, { "epoch": 0.04, "learning_rate": 9.61394634858322e-05, "loss": 1.0335, "step": 16350 }, { "epoch": 0.04, "learning_rate": 9.61382052443505e-05, "loss": 1.0358, "step": 16355 }, { "epoch": 0.04, "learning_rate": 9.61369470028688e-05, "loss": 1.0346, "step": 16360 }, { "epoch": 0.04, "learning_rate": 9.613568876138709e-05, "loss": 1.0356, "step": 16365 }, { "epoch": 0.04, "learning_rate": 9.613443051990538e-05, "loss": 1.0361, "step": 16370 }, { "epoch": 0.04, "learning_rate": 9.613317227842367e-05, "loss": 1.0336, "step": 16375 }, { "epoch": 0.04, "learning_rate": 9.613191403694198e-05, "loss": 1.0347, "step": 16380 }, { "epoch": 0.04, "learning_rate": 9.613065579546027e-05, "loss": 1.0357, "step": 16385 }, { "epoch": 0.04, "learning_rate": 9.612939755397856e-05, "loss": 1.0379, "step": 16390 }, { "epoch": 0.04, "learning_rate": 9.612813931249685e-05, "loss": 1.0346, "step": 16395 }, { "epoch": 0.04, "learning_rate": 9.612688107101514e-05, "loss": 1.0354, "step": 16400 }, { "epoch": 0.04, "learning_rate": 9.612562282953345e-05, "loss": 1.0342, "step": 16405 }, { "epoch": 0.04, "learning_rate": 9.612436458805174e-05, "loss": 1.0398, "step": 16410 }, { "epoch": 0.04, "learning_rate": 9.612310634657003e-05, "loss": 1.0365, "step": 16415 }, { "epoch": 0.04, "learning_rate": 9.612184810508832e-05, "loss": 1.0338, "step": 16420 }, { "epoch": 0.04, "learning_rate": 9.612058986360663e-05, "loss": 1.0351, "step": 16425 }, { "epoch": 0.04, "learning_rate": 9.611933162212492e-05, "loss": 1.0353, "step": 16430 }, { "epoch": 0.04, "learning_rate": 9.611807338064321e-05, "loss": 1.0342, "step": 16435 }, { "epoch": 0.04, "learning_rate": 9.61168151391615e-05, "loss": 1.0368, "step": 16440 }, { "epoch": 0.04, "learning_rate": 9.611555689767981e-05, "loss": 1.0592, "step": 16445 }, { "epoch": 0.04, "learning_rate": 9.61142986561981e-05, "loss": 1.0331, "step": 16450 }, { "epoch": 0.04, "learning_rate": 9.611304041471639e-05, "loss": 1.0356, "step": 16455 }, { "epoch": 0.04, "learning_rate": 9.611178217323468e-05, "loss": 1.0335, "step": 16460 }, { "epoch": 0.04, "learning_rate": 9.611052393175298e-05, "loss": 1.0336, "step": 16465 }, { "epoch": 0.04, "learning_rate": 9.610926569027128e-05, "loss": 1.033, "step": 16470 }, { "epoch": 0.04, "learning_rate": 9.610800744878957e-05, "loss": 1.0351, "step": 16475 }, { "epoch": 0.04, "learning_rate": 9.610674920730786e-05, "loss": 1.0357, "step": 16480 }, { "epoch": 0.04, "learning_rate": 9.610549096582616e-05, "loss": 1.0352, "step": 16485 }, { "epoch": 0.04, "learning_rate": 9.610423272434446e-05, "loss": 1.0354, "step": 16490 }, { "epoch": 0.04, "learning_rate": 9.610297448286275e-05, "loss": 1.0335, "step": 16495 }, { "epoch": 0.04, "learning_rate": 9.610171624138104e-05, "loss": 1.0346, "step": 16500 }, { "epoch": 0.04, "learning_rate": 9.610045799989934e-05, "loss": 1.0352, "step": 16505 }, { "epoch": 0.04, "learning_rate": 9.609919975841764e-05, "loss": 1.0353, "step": 16510 }, { "epoch": 0.04, "learning_rate": 9.609794151693593e-05, "loss": 1.032, "step": 16515 }, { "epoch": 0.04, "learning_rate": 9.609668327545422e-05, "loss": 1.0348, "step": 16520 }, { "epoch": 0.04, "learning_rate": 9.609542503397252e-05, "loss": 1.035, "step": 16525 }, { "epoch": 0.04, "learning_rate": 9.609416679249081e-05, "loss": 1.0578, "step": 16530 }, { "epoch": 0.04, "learning_rate": 9.609290855100913e-05, "loss": 1.0338, "step": 16535 }, { "epoch": 0.04, "learning_rate": 9.609165030952742e-05, "loss": 1.0368, "step": 16540 }, { "epoch": 0.04, "learning_rate": 9.609039206804571e-05, "loss": 1.0363, "step": 16545 }, { "epoch": 0.04, "learning_rate": 9.6089133826564e-05, "loss": 1.0328, "step": 16550 }, { "epoch": 0.04, "learning_rate": 9.608787558508229e-05, "loss": 1.035, "step": 16555 }, { "epoch": 0.04, "learning_rate": 9.60866173436006e-05, "loss": 1.0329, "step": 16560 }, { "epoch": 0.04, "learning_rate": 9.608535910211889e-05, "loss": 1.0368, "step": 16565 }, { "epoch": 0.04, "learning_rate": 9.608410086063718e-05, "loss": 1.0376, "step": 16570 }, { "epoch": 0.04, "learning_rate": 9.608284261915547e-05, "loss": 1.0362, "step": 16575 }, { "epoch": 0.04, "learning_rate": 9.608158437767378e-05, "loss": 1.0373, "step": 16580 }, { "epoch": 0.04, "learning_rate": 9.608032613619207e-05, "loss": 1.0543, "step": 16585 }, { "epoch": 0.04, "learning_rate": 9.607906789471036e-05, "loss": 1.0334, "step": 16590 }, { "epoch": 0.04, "learning_rate": 9.607780965322865e-05, "loss": 1.0354, "step": 16595 }, { "epoch": 0.04, "learning_rate": 9.607655141174696e-05, "loss": 1.0347, "step": 16600 }, { "epoch": 0.04, "learning_rate": 9.607529317026525e-05, "loss": 1.0351, "step": 16605 }, { "epoch": 0.04, "learning_rate": 9.607403492878354e-05, "loss": 1.0357, "step": 16610 }, { "epoch": 0.04, "learning_rate": 9.607277668730183e-05, "loss": 1.0344, "step": 16615 }, { "epoch": 0.04, "learning_rate": 9.607151844582012e-05, "loss": 1.0342, "step": 16620 }, { "epoch": 0.04, "learning_rate": 9.607026020433843e-05, "loss": 1.0335, "step": 16625 }, { "epoch": 0.04, "learning_rate": 9.606900196285672e-05, "loss": 1.0351, "step": 16630 }, { "epoch": 0.04, "learning_rate": 9.606774372137501e-05, "loss": 1.0348, "step": 16635 }, { "epoch": 0.04, "learning_rate": 9.60664854798933e-05, "loss": 1.0375, "step": 16640 }, { "epoch": 0.04, "learning_rate": 9.606522723841161e-05, "loss": 1.0362, "step": 16645 }, { "epoch": 0.04, "learning_rate": 9.60639689969299e-05, "loss": 1.0346, "step": 16650 }, { "epoch": 0.04, "learning_rate": 9.606271075544819e-05, "loss": 1.0339, "step": 16655 }, { "epoch": 0.04, "learning_rate": 9.606145251396648e-05, "loss": 1.0348, "step": 16660 }, { "epoch": 0.04, "learning_rate": 9.606019427248479e-05, "loss": 1.0336, "step": 16665 }, { "epoch": 0.04, "learning_rate": 9.605893603100308e-05, "loss": 1.0355, "step": 16670 }, { "epoch": 0.04, "learning_rate": 9.605767778952137e-05, "loss": 1.0345, "step": 16675 }, { "epoch": 0.04, "learning_rate": 9.605641954803966e-05, "loss": 1.0325, "step": 16680 }, { "epoch": 0.04, "learning_rate": 9.605516130655796e-05, "loss": 1.0352, "step": 16685 }, { "epoch": 0.04, "learning_rate": 9.605390306507626e-05, "loss": 1.0352, "step": 16690 }, { "epoch": 0.04, "learning_rate": 9.605264482359455e-05, "loss": 1.0346, "step": 16695 }, { "epoch": 0.04, "learning_rate": 9.605138658211284e-05, "loss": 1.0315, "step": 16700 }, { "epoch": 0.04, "learning_rate": 9.605012834063114e-05, "loss": 1.0326, "step": 16705 }, { "epoch": 0.04, "learning_rate": 9.604887009914944e-05, "loss": 1.034, "step": 16710 }, { "epoch": 0.04, "learning_rate": 9.604761185766773e-05, "loss": 1.035, "step": 16715 }, { "epoch": 0.04, "learning_rate": 9.604635361618602e-05, "loss": 1.0364, "step": 16720 }, { "epoch": 0.04, "learning_rate": 9.604509537470432e-05, "loss": 1.0337, "step": 16725 }, { "epoch": 0.04, "learning_rate": 9.604383713322261e-05, "loss": 1.0359, "step": 16730 }, { "epoch": 0.04, "learning_rate": 9.604257889174091e-05, "loss": 1.0334, "step": 16735 }, { "epoch": 0.04, "learning_rate": 9.60413206502592e-05, "loss": 1.0344, "step": 16740 }, { "epoch": 0.04, "learning_rate": 9.60400624087775e-05, "loss": 1.0336, "step": 16745 }, { "epoch": 0.04, "learning_rate": 9.603880416729579e-05, "loss": 1.0333, "step": 16750 }, { "epoch": 0.04, "learning_rate": 9.603754592581409e-05, "loss": 1.0351, "step": 16755 }, { "epoch": 0.04, "learning_rate": 9.603628768433238e-05, "loss": 1.0359, "step": 16760 }, { "epoch": 0.04, "learning_rate": 9.603502944285068e-05, "loss": 1.0343, "step": 16765 }, { "epoch": 0.04, "learning_rate": 9.603377120136897e-05, "loss": 1.0333, "step": 16770 }, { "epoch": 0.04, "learning_rate": 9.603251295988727e-05, "loss": 1.0333, "step": 16775 }, { "epoch": 0.04, "learning_rate": 9.603125471840556e-05, "loss": 1.0353, "step": 16780 }, { "epoch": 0.04, "learning_rate": 9.602999647692386e-05, "loss": 1.0352, "step": 16785 }, { "epoch": 0.04, "learning_rate": 9.602873823544215e-05, "loss": 1.0353, "step": 16790 }, { "epoch": 0.04, "learning_rate": 9.602747999396044e-05, "loss": 1.0324, "step": 16795 }, { "epoch": 0.04, "learning_rate": 9.602622175247874e-05, "loss": 1.0347, "step": 16800 }, { "epoch": 0.04, "learning_rate": 9.602496351099704e-05, "loss": 1.0331, "step": 16805 }, { "epoch": 0.04, "learning_rate": 9.602370526951533e-05, "loss": 1.0355, "step": 16810 }, { "epoch": 0.04, "learning_rate": 9.602244702803362e-05, "loss": 1.0352, "step": 16815 }, { "epoch": 0.04, "learning_rate": 9.602118878655192e-05, "loss": 1.0343, "step": 16820 }, { "epoch": 0.04, "learning_rate": 9.601993054507022e-05, "loss": 1.0363, "step": 16825 }, { "epoch": 0.04, "learning_rate": 9.601867230358851e-05, "loss": 1.0353, "step": 16830 }, { "epoch": 0.04, "learning_rate": 9.60174140621068e-05, "loss": 1.0334, "step": 16835 }, { "epoch": 0.04, "learning_rate": 9.60161558206251e-05, "loss": 1.0358, "step": 16840 }, { "epoch": 0.04, "learning_rate": 9.60148975791434e-05, "loss": 1.035, "step": 16845 }, { "epoch": 0.04, "learning_rate": 9.601363933766169e-05, "loss": 1.0344, "step": 16850 }, { "epoch": 0.04, "learning_rate": 9.601238109617998e-05, "loss": 1.0343, "step": 16855 }, { "epoch": 0.04, "learning_rate": 9.601112285469827e-05, "loss": 1.0353, "step": 16860 }, { "epoch": 0.04, "learning_rate": 9.600986461321658e-05, "loss": 1.0348, "step": 16865 }, { "epoch": 0.04, "learning_rate": 9.600860637173487e-05, "loss": 1.0323, "step": 16870 }, { "epoch": 0.04, "learning_rate": 9.600734813025316e-05, "loss": 1.0353, "step": 16875 }, { "epoch": 0.04, "learning_rate": 9.600608988877145e-05, "loss": 1.0349, "step": 16880 }, { "epoch": 0.04, "learning_rate": 9.600483164728976e-05, "loss": 1.0343, "step": 16885 }, { "epoch": 0.04, "learning_rate": 9.600357340580805e-05, "loss": 1.0352, "step": 16890 }, { "epoch": 0.04, "learning_rate": 9.600231516432634e-05, "loss": 1.035, "step": 16895 }, { "epoch": 0.04, "learning_rate": 9.600105692284463e-05, "loss": 1.0351, "step": 16900 }, { "epoch": 0.04, "learning_rate": 9.599979868136294e-05, "loss": 1.0371, "step": 16905 }, { "epoch": 0.04, "learning_rate": 9.599854043988123e-05, "loss": 1.0357, "step": 16910 }, { "epoch": 0.04, "learning_rate": 9.599728219839952e-05, "loss": 1.0326, "step": 16915 }, { "epoch": 0.04, "learning_rate": 9.599602395691781e-05, "loss": 1.0338, "step": 16920 }, { "epoch": 0.04, "learning_rate": 9.59947657154361e-05, "loss": 1.0354, "step": 16925 }, { "epoch": 0.04, "learning_rate": 9.59935074739544e-05, "loss": 1.0359, "step": 16930 }, { "epoch": 0.04, "learning_rate": 9.59922492324727e-05, "loss": 1.0351, "step": 16935 }, { "epoch": 0.04, "learning_rate": 9.599099099099099e-05, "loss": 1.0361, "step": 16940 }, { "epoch": 0.04, "learning_rate": 9.598973274950928e-05, "loss": 1.049, "step": 16945 }, { "epoch": 0.04, "learning_rate": 9.598847450802759e-05, "loss": 1.0363, "step": 16950 }, { "epoch": 0.04, "learning_rate": 9.598721626654588e-05, "loss": 1.0337, "step": 16955 }, { "epoch": 0.04, "learning_rate": 9.598595802506417e-05, "loss": 1.0335, "step": 16960 }, { "epoch": 0.04, "learning_rate": 9.598469978358246e-05, "loss": 1.034, "step": 16965 }, { "epoch": 0.04, "learning_rate": 9.598344154210077e-05, "loss": 1.0344, "step": 16970 }, { "epoch": 0.04, "learning_rate": 9.598218330061906e-05, "loss": 1.0358, "step": 16975 }, { "epoch": 0.04, "learning_rate": 9.598092505913735e-05, "loss": 1.0357, "step": 16980 }, { "epoch": 0.04, "learning_rate": 9.597966681765564e-05, "loss": 1.0334, "step": 16985 }, { "epoch": 0.04, "learning_rate": 9.597840857617393e-05, "loss": 1.0347, "step": 16990 }, { "epoch": 0.04, "learning_rate": 9.597715033469224e-05, "loss": 1.0344, "step": 16995 }, { "epoch": 0.04, "learning_rate": 9.597589209321053e-05, "loss": 1.0341, "step": 17000 }, { "epoch": 0.04, "learning_rate": 9.597463385172882e-05, "loss": 1.0588, "step": 17005 }, { "epoch": 0.04, "learning_rate": 9.597337561024711e-05, "loss": 1.0322, "step": 17010 }, { "epoch": 0.04, "learning_rate": 9.597211736876542e-05, "loss": 1.0346, "step": 17015 }, { "epoch": 0.04, "learning_rate": 9.597085912728371e-05, "loss": 1.0347, "step": 17020 }, { "epoch": 0.04, "learning_rate": 9.5969600885802e-05, "loss": 1.0593, "step": 17025 }, { "epoch": 0.04, "learning_rate": 9.596834264432029e-05, "loss": 1.0341, "step": 17030 }, { "epoch": 0.04, "learning_rate": 9.59670844028386e-05, "loss": 1.0355, "step": 17035 }, { "epoch": 0.04, "learning_rate": 9.59658261613569e-05, "loss": 1.0357, "step": 17040 }, { "epoch": 0.04, "learning_rate": 9.59645679198752e-05, "loss": 1.0354, "step": 17045 }, { "epoch": 0.04, "learning_rate": 9.596330967839349e-05, "loss": 1.0359, "step": 17050 }, { "epoch": 0.04, "learning_rate": 9.596205143691178e-05, "loss": 1.0345, "step": 17055 }, { "epoch": 0.04, "learning_rate": 9.596079319543008e-05, "loss": 1.0333, "step": 17060 }, { "epoch": 0.04, "learning_rate": 9.595953495394837e-05, "loss": 1.0351, "step": 17065 }, { "epoch": 0.04, "learning_rate": 9.595827671246667e-05, "loss": 1.0364, "step": 17070 }, { "epoch": 0.04, "learning_rate": 9.595701847098496e-05, "loss": 1.0329, "step": 17075 }, { "epoch": 0.04, "learning_rate": 9.595576022950325e-05, "loss": 1.0492, "step": 17080 }, { "epoch": 0.04, "learning_rate": 9.595450198802155e-05, "loss": 1.0357, "step": 17085 }, { "epoch": 0.04, "learning_rate": 9.595324374653985e-05, "loss": 1.0358, "step": 17090 }, { "epoch": 0.04, "learning_rate": 9.595198550505814e-05, "loss": 1.0334, "step": 17095 }, { "epoch": 0.04, "learning_rate": 9.595072726357643e-05, "loss": 1.0351, "step": 17100 }, { "epoch": 0.04, "learning_rate": 9.594946902209473e-05, "loss": 1.0345, "step": 17105 }, { "epoch": 0.04, "learning_rate": 9.594821078061303e-05, "loss": 1.0333, "step": 17110 }, { "epoch": 0.04, "learning_rate": 9.594695253913132e-05, "loss": 1.0344, "step": 17115 }, { "epoch": 0.04, "learning_rate": 9.594569429764961e-05, "loss": 1.0342, "step": 17120 }, { "epoch": 0.04, "learning_rate": 9.594443605616791e-05, "loss": 1.0365, "step": 17125 }, { "epoch": 0.04, "learning_rate": 9.59431778146862e-05, "loss": 1.0353, "step": 17130 }, { "epoch": 0.04, "learning_rate": 9.59419195732045e-05, "loss": 1.0363, "step": 17135 }, { "epoch": 0.04, "learning_rate": 9.594066133172279e-05, "loss": 1.0349, "step": 17140 }, { "epoch": 0.04, "learning_rate": 9.593940309024108e-05, "loss": 1.0352, "step": 17145 }, { "epoch": 0.04, "learning_rate": 9.593814484875939e-05, "loss": 1.0353, "step": 17150 }, { "epoch": 0.04, "learning_rate": 9.593688660727768e-05, "loss": 1.0329, "step": 17155 }, { "epoch": 0.04, "learning_rate": 9.593562836579597e-05, "loss": 1.0364, "step": 17160 }, { "epoch": 0.04, "learning_rate": 9.593437012431426e-05, "loss": 1.0348, "step": 17165 }, { "epoch": 0.04, "learning_rate": 9.593311188283257e-05, "loss": 1.0361, "step": 17170 }, { "epoch": 0.04, "learning_rate": 9.593185364135086e-05, "loss": 1.0353, "step": 17175 }, { "epoch": 0.04, "learning_rate": 9.593059539986915e-05, "loss": 1.033, "step": 17180 }, { "epoch": 0.04, "learning_rate": 9.592933715838744e-05, "loss": 1.0328, "step": 17185 }, { "epoch": 0.04, "learning_rate": 9.592807891690575e-05, "loss": 1.0352, "step": 17190 }, { "epoch": 0.04, "learning_rate": 9.592682067542404e-05, "loss": 1.0335, "step": 17195 }, { "epoch": 0.04, "learning_rate": 9.592556243394233e-05, "loss": 1.0357, "step": 17200 }, { "epoch": 0.04, "learning_rate": 9.592430419246062e-05, "loss": 1.036, "step": 17205 }, { "epoch": 0.04, "learning_rate": 9.592304595097891e-05, "loss": 1.0348, "step": 17210 }, { "epoch": 0.04, "learning_rate": 9.592178770949722e-05, "loss": 1.0373, "step": 17215 }, { "epoch": 0.04, "learning_rate": 9.592052946801551e-05, "loss": 1.0358, "step": 17220 }, { "epoch": 0.04, "learning_rate": 9.59192712265338e-05, "loss": 1.0351, "step": 17225 }, { "epoch": 0.04, "learning_rate": 9.591801298505209e-05, "loss": 1.0358, "step": 17230 }, { "epoch": 0.04, "learning_rate": 9.59167547435704e-05, "loss": 1.0358, "step": 17235 }, { "epoch": 0.04, "learning_rate": 9.591549650208869e-05, "loss": 1.0345, "step": 17240 }, { "epoch": 0.04, "learning_rate": 9.591423826060698e-05, "loss": 1.0483, "step": 17245 }, { "epoch": 0.04, "learning_rate": 9.591298001912527e-05, "loss": 1.0374, "step": 17250 }, { "epoch": 0.04, "learning_rate": 9.591172177764358e-05, "loss": 1.0363, "step": 17255 }, { "epoch": 0.04, "learning_rate": 9.591046353616187e-05, "loss": 1.0337, "step": 17260 }, { "epoch": 0.04, "learning_rate": 9.590920529468016e-05, "loss": 1.035, "step": 17265 }, { "epoch": 0.04, "learning_rate": 9.590794705319845e-05, "loss": 1.0349, "step": 17270 }, { "epoch": 0.04, "learning_rate": 9.590668881171674e-05, "loss": 1.0366, "step": 17275 }, { "epoch": 0.04, "learning_rate": 9.590543057023505e-05, "loss": 1.0337, "step": 17280 }, { "epoch": 0.04, "learning_rate": 9.590417232875334e-05, "loss": 1.0373, "step": 17285 }, { "epoch": 0.04, "learning_rate": 9.590291408727163e-05, "loss": 1.032, "step": 17290 }, { "epoch": 0.04, "learning_rate": 9.590165584578992e-05, "loss": 1.0362, "step": 17295 }, { "epoch": 0.04, "learning_rate": 9.590039760430823e-05, "loss": 1.0343, "step": 17300 }, { "epoch": 0.04, "learning_rate": 9.589913936282652e-05, "loss": 1.034, "step": 17305 }, { "epoch": 0.04, "learning_rate": 9.589788112134481e-05, "loss": 1.0354, "step": 17310 }, { "epoch": 0.04, "learning_rate": 9.58966228798631e-05, "loss": 1.0342, "step": 17315 }, { "epoch": 0.04, "learning_rate": 9.589536463838141e-05, "loss": 1.0334, "step": 17320 }, { "epoch": 0.04, "learning_rate": 9.58941063968997e-05, "loss": 1.0345, "step": 17325 }, { "epoch": 0.04, "learning_rate": 9.589284815541799e-05, "loss": 1.0338, "step": 17330 }, { "epoch": 0.04, "learning_rate": 9.589158991393628e-05, "loss": 1.0357, "step": 17335 }, { "epoch": 0.04, "learning_rate": 9.589033167245458e-05, "loss": 1.0344, "step": 17340 }, { "epoch": 0.04, "learning_rate": 9.588907343097288e-05, "loss": 1.0354, "step": 17345 }, { "epoch": 0.04, "learning_rate": 9.588781518949117e-05, "loss": 1.0343, "step": 17350 }, { "epoch": 0.04, "learning_rate": 9.588655694800946e-05, "loss": 1.0358, "step": 17355 }, { "epoch": 0.04, "learning_rate": 9.588529870652776e-05, "loss": 1.0345, "step": 17360 }, { "epoch": 0.04, "learning_rate": 9.588404046504606e-05, "loss": 1.0345, "step": 17365 }, { "epoch": 0.04, "learning_rate": 9.588278222356435e-05, "loss": 1.0333, "step": 17370 }, { "epoch": 0.04, "learning_rate": 9.588152398208264e-05, "loss": 1.0343, "step": 17375 }, { "epoch": 0.04, "learning_rate": 9.588026574060093e-05, "loss": 1.0348, "step": 17380 }, { "epoch": 0.04, "learning_rate": 9.587900749911923e-05, "loss": 1.0336, "step": 17385 }, { "epoch": 0.04, "learning_rate": 9.587774925763753e-05, "loss": 1.0343, "step": 17390 }, { "epoch": 0.04, "learning_rate": 9.587649101615582e-05, "loss": 1.0347, "step": 17395 }, { "epoch": 0.04, "learning_rate": 9.587523277467411e-05, "loss": 1.0345, "step": 17400 }, { "epoch": 0.04, "learning_rate": 9.58739745331924e-05, "loss": 1.0319, "step": 17405 }, { "epoch": 0.04, "learning_rate": 9.587271629171071e-05, "loss": 1.0336, "step": 17410 }, { "epoch": 0.04, "learning_rate": 9.5871458050229e-05, "loss": 1.0344, "step": 17415 }, { "epoch": 0.04, "learning_rate": 9.58701998087473e-05, "loss": 1.0341, "step": 17420 }, { "epoch": 0.04, "learning_rate": 9.586894156726559e-05, "loss": 1.0339, "step": 17425 }, { "epoch": 0.04, "learning_rate": 9.586768332578389e-05, "loss": 1.0366, "step": 17430 }, { "epoch": 0.04, "learning_rate": 9.586642508430218e-05, "loss": 1.0346, "step": 17435 }, { "epoch": 0.04, "learning_rate": 9.586516684282047e-05, "loss": 1.0367, "step": 17440 }, { "epoch": 0.04, "learning_rate": 9.586390860133877e-05, "loss": 1.0359, "step": 17445 }, { "epoch": 0.04, "learning_rate": 9.586265035985706e-05, "loss": 1.0337, "step": 17450 }, { "epoch": 0.04, "learning_rate": 9.586139211837536e-05, "loss": 1.0359, "step": 17455 }, { "epoch": 0.04, "learning_rate": 9.586013387689365e-05, "loss": 1.0351, "step": 17460 }, { "epoch": 0.04, "learning_rate": 9.585887563541195e-05, "loss": 1.0338, "step": 17465 }, { "epoch": 0.04, "learning_rate": 9.585761739393024e-05, "loss": 1.0355, "step": 17470 }, { "epoch": 0.04, "learning_rate": 9.585635915244854e-05, "loss": 1.0336, "step": 17475 }, { "epoch": 0.04, "learning_rate": 9.585510091096683e-05, "loss": 1.0345, "step": 17480 }, { "epoch": 0.04, "learning_rate": 9.585384266948513e-05, "loss": 1.0352, "step": 17485 }, { "epoch": 0.04, "learning_rate": 9.585258442800342e-05, "loss": 1.0357, "step": 17490 }, { "epoch": 0.04, "learning_rate": 9.585132618652172e-05, "loss": 1.0343, "step": 17495 }, { "epoch": 0.04, "learning_rate": 9.585006794504001e-05, "loss": 1.0361, "step": 17500 }, { "epoch": 0.04, "learning_rate": 9.58488097035583e-05, "loss": 1.0328, "step": 17505 }, { "epoch": 0.04, "learning_rate": 9.58475514620766e-05, "loss": 1.036, "step": 17510 }, { "epoch": 0.04, "learning_rate": 9.584629322059489e-05, "loss": 1.0322, "step": 17515 }, { "epoch": 0.04, "learning_rate": 9.58450349791132e-05, "loss": 1.0566, "step": 17520 }, { "epoch": 0.04, "learning_rate": 9.584377673763149e-05, "loss": 1.0346, "step": 17525 }, { "epoch": 0.04, "learning_rate": 9.584251849614978e-05, "loss": 1.0341, "step": 17530 }, { "epoch": 0.04, "learning_rate": 9.584126025466808e-05, "loss": 1.0372, "step": 17535 }, { "epoch": 0.04, "learning_rate": 9.584000201318637e-05, "loss": 1.0349, "step": 17540 }, { "epoch": 0.04, "learning_rate": 9.583874377170468e-05, "loss": 1.0355, "step": 17545 }, { "epoch": 0.04, "learning_rate": 9.583748553022297e-05, "loss": 1.0364, "step": 17550 }, { "epoch": 0.04, "learning_rate": 9.583622728874126e-05, "loss": 1.0339, "step": 17555 }, { "epoch": 0.04, "learning_rate": 9.583496904725955e-05, "loss": 1.034, "step": 17560 }, { "epoch": 0.04, "learning_rate": 9.583371080577786e-05, "loss": 1.0336, "step": 17565 }, { "epoch": 0.04, "learning_rate": 9.583245256429615e-05, "loss": 1.0336, "step": 17570 }, { "epoch": 0.04, "learning_rate": 9.583119432281444e-05, "loss": 1.0357, "step": 17575 }, { "epoch": 0.04, "learning_rate": 9.582993608133273e-05, "loss": 1.0365, "step": 17580 }, { "epoch": 0.04, "learning_rate": 9.582867783985104e-05, "loss": 1.0376, "step": 17585 }, { "epoch": 0.04, "learning_rate": 9.582741959836933e-05, "loss": 1.0343, "step": 17590 }, { "epoch": 0.04, "learning_rate": 9.582616135688762e-05, "loss": 1.0344, "step": 17595 }, { "epoch": 0.04, "learning_rate": 9.582490311540591e-05, "loss": 1.0328, "step": 17600 }, { "epoch": 0.04, "learning_rate": 9.58236448739242e-05, "loss": 1.0327, "step": 17605 }, { "epoch": 0.04, "learning_rate": 9.582238663244251e-05, "loss": 1.0341, "step": 17610 }, { "epoch": 0.04, "learning_rate": 9.58211283909608e-05, "loss": 1.0355, "step": 17615 }, { "epoch": 0.04, "learning_rate": 9.58198701494791e-05, "loss": 1.0356, "step": 17620 }, { "epoch": 0.04, "learning_rate": 9.581861190799739e-05, "loss": 1.0361, "step": 17625 }, { "epoch": 0.04, "learning_rate": 9.581735366651569e-05, "loss": 1.0339, "step": 17630 }, { "epoch": 0.04, "learning_rate": 9.581609542503398e-05, "loss": 1.0331, "step": 17635 }, { "epoch": 0.04, "learning_rate": 9.581483718355227e-05, "loss": 1.0591, "step": 17640 }, { "epoch": 0.04, "learning_rate": 9.581357894207057e-05, "loss": 1.0325, "step": 17645 }, { "epoch": 0.04, "learning_rate": 9.581232070058887e-05, "loss": 1.0356, "step": 17650 }, { "epoch": 0.04, "learning_rate": 9.581106245910716e-05, "loss": 1.0361, "step": 17655 }, { "epoch": 0.04, "learning_rate": 9.580980421762545e-05, "loss": 1.0327, "step": 17660 }, { "epoch": 0.04, "learning_rate": 9.580854597614375e-05, "loss": 1.0332, "step": 17665 }, { "epoch": 0.04, "learning_rate": 9.580728773466204e-05, "loss": 1.0367, "step": 17670 }, { "epoch": 0.04, "learning_rate": 9.580602949318034e-05, "loss": 1.0335, "step": 17675 }, { "epoch": 0.04, "learning_rate": 9.580477125169863e-05, "loss": 1.0364, "step": 17680 }, { "epoch": 0.04, "learning_rate": 9.580351301021693e-05, "loss": 1.0369, "step": 17685 }, { "epoch": 0.04, "learning_rate": 9.580225476873522e-05, "loss": 1.0343, "step": 17690 }, { "epoch": 0.04, "learning_rate": 9.580099652725352e-05, "loss": 1.0341, "step": 17695 }, { "epoch": 0.04, "learning_rate": 9.579973828577181e-05, "loss": 1.0354, "step": 17700 }, { "epoch": 0.04, "learning_rate": 9.57984800442901e-05, "loss": 1.0363, "step": 17705 }, { "epoch": 0.04, "learning_rate": 9.57972218028084e-05, "loss": 1.0347, "step": 17710 }, { "epoch": 0.04, "learning_rate": 9.57959635613267e-05, "loss": 1.0365, "step": 17715 }, { "epoch": 0.04, "learning_rate": 9.5794705319845e-05, "loss": 1.0352, "step": 17720 }, { "epoch": 0.04, "learning_rate": 9.579344707836329e-05, "loss": 1.034, "step": 17725 }, { "epoch": 0.04, "learning_rate": 9.579218883688158e-05, "loss": 1.0339, "step": 17730 }, { "epoch": 0.04, "learning_rate": 9.579093059539987e-05, "loss": 1.0334, "step": 17735 }, { "epoch": 0.04, "learning_rate": 9.578967235391817e-05, "loss": 1.0348, "step": 17740 }, { "epoch": 0.04, "learning_rate": 9.578841411243647e-05, "loss": 1.0361, "step": 17745 }, { "epoch": 0.04, "learning_rate": 9.578715587095476e-05, "loss": 1.0354, "step": 17750 }, { "epoch": 0.04, "learning_rate": 9.578589762947305e-05, "loss": 1.0321, "step": 17755 }, { "epoch": 0.04, "learning_rate": 9.578463938799135e-05, "loss": 1.0349, "step": 17760 }, { "epoch": 0.04, "learning_rate": 9.578338114650965e-05, "loss": 1.0339, "step": 17765 }, { "epoch": 0.04, "learning_rate": 9.578212290502794e-05, "loss": 1.0375, "step": 17770 }, { "epoch": 0.04, "learning_rate": 9.578086466354623e-05, "loss": 1.0335, "step": 17775 }, { "epoch": 0.04, "learning_rate": 9.577960642206453e-05, "loss": 1.0355, "step": 17780 }, { "epoch": 0.04, "learning_rate": 9.577834818058283e-05, "loss": 1.0349, "step": 17785 }, { "epoch": 0.04, "learning_rate": 9.577708993910112e-05, "loss": 1.0367, "step": 17790 }, { "epoch": 0.04, "learning_rate": 9.577583169761941e-05, "loss": 1.0333, "step": 17795 }, { "epoch": 0.04, "learning_rate": 9.57745734561377e-05, "loss": 1.0363, "step": 17800 }, { "epoch": 0.04, "learning_rate": 9.5773315214656e-05, "loss": 1.0353, "step": 17805 }, { "epoch": 0.04, "learning_rate": 9.57720569731743e-05, "loss": 1.0333, "step": 17810 }, { "epoch": 0.04, "learning_rate": 9.577079873169259e-05, "loss": 1.0348, "step": 17815 }, { "epoch": 0.04, "learning_rate": 9.576954049021088e-05, "loss": 1.0339, "step": 17820 }, { "epoch": 0.04, "learning_rate": 9.576828224872919e-05, "loss": 1.0346, "step": 17825 }, { "epoch": 0.04, "learning_rate": 9.576702400724748e-05, "loss": 1.0332, "step": 17830 }, { "epoch": 0.04, "learning_rate": 9.576576576576577e-05, "loss": 1.0353, "step": 17835 }, { "epoch": 0.04, "learning_rate": 9.576450752428406e-05, "loss": 1.032, "step": 17840 }, { "epoch": 0.04, "learning_rate": 9.576324928280237e-05, "loss": 1.0312, "step": 17845 }, { "epoch": 0.04, "learning_rate": 9.576199104132066e-05, "loss": 1.0356, "step": 17850 }, { "epoch": 0.04, "learning_rate": 9.576073279983895e-05, "loss": 1.033, "step": 17855 }, { "epoch": 0.04, "learning_rate": 9.575947455835724e-05, "loss": 1.0309, "step": 17860 }, { "epoch": 0.04, "learning_rate": 9.575821631687553e-05, "loss": 1.0335, "step": 17865 }, { "epoch": 0.04, "learning_rate": 9.575695807539384e-05, "loss": 1.0348, "step": 17870 }, { "epoch": 0.04, "learning_rate": 9.575569983391213e-05, "loss": 1.0344, "step": 17875 }, { "epoch": 0.04, "learning_rate": 9.575444159243042e-05, "loss": 1.0347, "step": 17880 }, { "epoch": 0.04, "learning_rate": 9.575318335094871e-05, "loss": 1.0313, "step": 17885 }, { "epoch": 0.04, "learning_rate": 9.575192510946702e-05, "loss": 1.035, "step": 17890 }, { "epoch": 0.04, "learning_rate": 9.575066686798531e-05, "loss": 1.033, "step": 17895 }, { "epoch": 0.04, "learning_rate": 9.57494086265036e-05, "loss": 1.0336, "step": 17900 }, { "epoch": 0.04, "learning_rate": 9.574815038502189e-05, "loss": 1.035, "step": 17905 }, { "epoch": 0.04, "learning_rate": 9.57468921435402e-05, "loss": 1.0347, "step": 17910 }, { "epoch": 0.04, "learning_rate": 9.574563390205849e-05, "loss": 1.0333, "step": 17915 }, { "epoch": 0.04, "learning_rate": 9.574437566057678e-05, "loss": 1.0329, "step": 17920 }, { "epoch": 0.04, "learning_rate": 9.574311741909507e-05, "loss": 1.032, "step": 17925 }, { "epoch": 0.05, "learning_rate": 9.574185917761336e-05, "loss": 1.036, "step": 17930 }, { "epoch": 0.05, "learning_rate": 9.574060093613167e-05, "loss": 1.0341, "step": 17935 }, { "epoch": 0.05, "learning_rate": 9.573934269464996e-05, "loss": 1.0353, "step": 17940 }, { "epoch": 0.05, "learning_rate": 9.573808445316825e-05, "loss": 1.0347, "step": 17945 }, { "epoch": 0.05, "learning_rate": 9.573682621168654e-05, "loss": 1.0364, "step": 17950 }, { "epoch": 0.05, "learning_rate": 9.573556797020485e-05, "loss": 1.0333, "step": 17955 }, { "epoch": 0.05, "learning_rate": 9.573430972872314e-05, "loss": 1.0345, "step": 17960 }, { "epoch": 0.05, "learning_rate": 9.573305148724143e-05, "loss": 1.0446, "step": 17965 }, { "epoch": 0.05, "learning_rate": 9.573179324575972e-05, "loss": 1.0358, "step": 17970 }, { "epoch": 0.05, "learning_rate": 9.573053500427803e-05, "loss": 1.0327, "step": 17975 }, { "epoch": 0.05, "learning_rate": 9.572927676279632e-05, "loss": 1.0346, "step": 17980 }, { "epoch": 0.05, "learning_rate": 9.572801852131461e-05, "loss": 1.0348, "step": 17985 }, { "epoch": 0.05, "learning_rate": 9.57267602798329e-05, "loss": 1.0367, "step": 17990 }, { "epoch": 0.05, "learning_rate": 9.57255020383512e-05, "loss": 1.0336, "step": 17995 }, { "epoch": 0.05, "learning_rate": 9.57242437968695e-05, "loss": 1.0328, "step": 18000 }, { "epoch": 0.05, "learning_rate": 9.572298555538779e-05, "loss": 1.0358, "step": 18005 }, { "epoch": 0.05, "learning_rate": 9.572172731390608e-05, "loss": 1.0343, "step": 18010 }, { "epoch": 0.05, "learning_rate": 9.572046907242437e-05, "loss": 1.0331, "step": 18015 }, { "epoch": 0.05, "learning_rate": 9.571921083094268e-05, "loss": 1.0356, "step": 18020 }, { "epoch": 0.05, "learning_rate": 9.571795258946097e-05, "loss": 1.0322, "step": 18025 }, { "epoch": 0.05, "learning_rate": 9.571669434797926e-05, "loss": 1.035, "step": 18030 }, { "epoch": 0.05, "learning_rate": 9.571543610649757e-05, "loss": 1.0337, "step": 18035 }, { "epoch": 0.05, "learning_rate": 9.571417786501586e-05, "loss": 1.0353, "step": 18040 }, { "epoch": 0.05, "learning_rate": 9.571291962353416e-05, "loss": 1.0344, "step": 18045 }, { "epoch": 0.05, "learning_rate": 9.571166138205246e-05, "loss": 1.0329, "step": 18050 }, { "epoch": 0.05, "learning_rate": 9.571040314057075e-05, "loss": 1.032, "step": 18055 }, { "epoch": 0.05, "learning_rate": 9.570914489908904e-05, "loss": 1.0344, "step": 18060 }, { "epoch": 0.05, "learning_rate": 9.570788665760733e-05, "loss": 1.0355, "step": 18065 }, { "epoch": 0.05, "learning_rate": 9.570662841612564e-05, "loss": 1.0342, "step": 18070 }, { "epoch": 0.05, "learning_rate": 9.570537017464393e-05, "loss": 1.0348, "step": 18075 }, { "epoch": 0.05, "learning_rate": 9.570411193316222e-05, "loss": 1.0352, "step": 18080 }, { "epoch": 0.05, "learning_rate": 9.570285369168051e-05, "loss": 1.0337, "step": 18085 }, { "epoch": 0.05, "learning_rate": 9.570159545019882e-05, "loss": 1.0348, "step": 18090 }, { "epoch": 0.05, "learning_rate": 9.570033720871711e-05, "loss": 1.0331, "step": 18095 }, { "epoch": 0.05, "learning_rate": 9.56990789672354e-05, "loss": 1.0338, "step": 18100 }, { "epoch": 0.05, "learning_rate": 9.569782072575369e-05, "loss": 1.0346, "step": 18105 }, { "epoch": 0.05, "learning_rate": 9.5696562484272e-05, "loss": 1.034, "step": 18110 }, { "epoch": 0.05, "learning_rate": 9.569530424279029e-05, "loss": 1.0336, "step": 18115 }, { "epoch": 0.05, "learning_rate": 9.569404600130858e-05, "loss": 1.033, "step": 18120 }, { "epoch": 0.05, "learning_rate": 9.569278775982687e-05, "loss": 1.0346, "step": 18125 }, { "epoch": 0.05, "learning_rate": 9.569152951834516e-05, "loss": 1.0351, "step": 18130 }, { "epoch": 0.05, "learning_rate": 9.569027127686347e-05, "loss": 1.0339, "step": 18135 }, { "epoch": 0.05, "learning_rate": 9.568901303538176e-05, "loss": 1.0338, "step": 18140 }, { "epoch": 0.05, "learning_rate": 9.568775479390005e-05, "loss": 1.0341, "step": 18145 }, { "epoch": 0.05, "learning_rate": 9.568649655241834e-05, "loss": 1.0352, "step": 18150 }, { "epoch": 0.05, "learning_rate": 9.568523831093665e-05, "loss": 1.0355, "step": 18155 }, { "epoch": 0.05, "learning_rate": 9.568398006945494e-05, "loss": 1.0321, "step": 18160 }, { "epoch": 0.05, "learning_rate": 9.568272182797323e-05, "loss": 1.0347, "step": 18165 }, { "epoch": 0.05, "learning_rate": 9.568146358649152e-05, "loss": 1.032, "step": 18170 }, { "epoch": 0.05, "learning_rate": 9.568020534500983e-05, "loss": 1.0348, "step": 18175 }, { "epoch": 0.05, "learning_rate": 9.567894710352812e-05, "loss": 1.0339, "step": 18180 }, { "epoch": 0.05, "learning_rate": 9.567768886204641e-05, "loss": 1.0356, "step": 18185 }, { "epoch": 0.05, "learning_rate": 9.56764306205647e-05, "loss": 1.0342, "step": 18190 }, { "epoch": 0.05, "learning_rate": 9.5675172379083e-05, "loss": 1.0361, "step": 18195 }, { "epoch": 0.05, "learning_rate": 9.56739141376013e-05, "loss": 1.0336, "step": 18200 }, { "epoch": 0.05, "learning_rate": 9.567265589611959e-05, "loss": 1.0343, "step": 18205 }, { "epoch": 0.05, "learning_rate": 9.567139765463788e-05, "loss": 1.0341, "step": 18210 }, { "epoch": 0.05, "learning_rate": 9.567013941315617e-05, "loss": 1.0358, "step": 18215 }, { "epoch": 0.05, "learning_rate": 9.566888117167448e-05, "loss": 1.0355, "step": 18220 }, { "epoch": 0.05, "learning_rate": 9.566762293019277e-05, "loss": 1.0314, "step": 18225 }, { "epoch": 0.05, "learning_rate": 9.566636468871106e-05, "loss": 1.0347, "step": 18230 }, { "epoch": 0.05, "learning_rate": 9.566510644722935e-05, "loss": 1.0344, "step": 18235 }, { "epoch": 0.05, "learning_rate": 9.566384820574766e-05, "loss": 1.0372, "step": 18240 }, { "epoch": 0.05, "learning_rate": 9.566258996426595e-05, "loss": 1.0342, "step": 18245 }, { "epoch": 0.05, "learning_rate": 9.566133172278424e-05, "loss": 1.0346, "step": 18250 }, { "epoch": 0.05, "learning_rate": 9.566007348130253e-05, "loss": 1.0374, "step": 18255 }, { "epoch": 0.05, "learning_rate": 9.565881523982083e-05, "loss": 1.0332, "step": 18260 }, { "epoch": 0.05, "learning_rate": 9.565755699833913e-05, "loss": 1.0352, "step": 18265 }, { "epoch": 0.05, "learning_rate": 9.565629875685742e-05, "loss": 1.0345, "step": 18270 }, { "epoch": 0.05, "learning_rate": 9.565504051537571e-05, "loss": 1.0353, "step": 18275 }, { "epoch": 0.05, "learning_rate": 9.5653782273894e-05, "loss": 1.0343, "step": 18280 }, { "epoch": 0.05, "learning_rate": 9.565252403241231e-05, "loss": 1.0336, "step": 18285 }, { "epoch": 0.05, "learning_rate": 9.56512657909306e-05, "loss": 1.0355, "step": 18290 }, { "epoch": 0.05, "learning_rate": 9.56500075494489e-05, "loss": 1.0448, "step": 18295 }, { "epoch": 0.05, "learning_rate": 9.564874930796719e-05, "loss": 1.0325, "step": 18300 }, { "epoch": 0.05, "learning_rate": 9.564749106648549e-05, "loss": 1.0321, "step": 18305 }, { "epoch": 0.05, "learning_rate": 9.564623282500378e-05, "loss": 1.0346, "step": 18310 }, { "epoch": 0.05, "learning_rate": 9.564497458352207e-05, "loss": 1.0335, "step": 18315 }, { "epoch": 0.05, "learning_rate": 9.564371634204037e-05, "loss": 1.0334, "step": 18320 }, { "epoch": 0.05, "learning_rate": 9.564245810055866e-05, "loss": 1.0334, "step": 18325 }, { "epoch": 0.05, "learning_rate": 9.564119985907696e-05, "loss": 1.0327, "step": 18330 }, { "epoch": 0.05, "learning_rate": 9.563994161759525e-05, "loss": 1.0338, "step": 18335 }, { "epoch": 0.05, "learning_rate": 9.563868337611355e-05, "loss": 1.0332, "step": 18340 }, { "epoch": 0.05, "learning_rate": 9.563742513463184e-05, "loss": 1.0354, "step": 18345 }, { "epoch": 0.05, "learning_rate": 9.563616689315014e-05, "loss": 1.0342, "step": 18350 }, { "epoch": 0.05, "learning_rate": 9.563490865166843e-05, "loss": 1.0337, "step": 18355 }, { "epoch": 0.05, "learning_rate": 9.563365041018673e-05, "loss": 1.0513, "step": 18360 }, { "epoch": 0.05, "learning_rate": 9.563239216870502e-05, "loss": 1.035, "step": 18365 }, { "epoch": 0.05, "learning_rate": 9.563113392722332e-05, "loss": 1.0337, "step": 18370 }, { "epoch": 0.05, "learning_rate": 9.562987568574161e-05, "loss": 1.034, "step": 18375 }, { "epoch": 0.05, "learning_rate": 9.56286174442599e-05, "loss": 1.0334, "step": 18380 }, { "epoch": 0.05, "learning_rate": 9.56273592027782e-05, "loss": 1.0356, "step": 18385 }, { "epoch": 0.05, "learning_rate": 9.562610096129649e-05, "loss": 1.0334, "step": 18390 }, { "epoch": 0.05, "learning_rate": 9.56248427198148e-05, "loss": 1.0317, "step": 18395 }, { "epoch": 0.05, "learning_rate": 9.562358447833308e-05, "loss": 1.0356, "step": 18400 }, { "epoch": 0.05, "learning_rate": 9.562232623685138e-05, "loss": 1.0355, "step": 18405 }, { "epoch": 0.05, "learning_rate": 9.562106799536967e-05, "loss": 1.0338, "step": 18410 }, { "epoch": 0.05, "learning_rate": 9.561980975388797e-05, "loss": 1.0331, "step": 18415 }, { "epoch": 0.05, "learning_rate": 9.561855151240626e-05, "loss": 1.0321, "step": 18420 }, { "epoch": 0.05, "learning_rate": 9.561729327092456e-05, "loss": 1.0342, "step": 18425 }, { "epoch": 0.05, "learning_rate": 9.561603502944285e-05, "loss": 1.035, "step": 18430 }, { "epoch": 0.05, "learning_rate": 9.561477678796115e-05, "loss": 1.0324, "step": 18435 }, { "epoch": 0.05, "learning_rate": 9.561351854647944e-05, "loss": 1.0335, "step": 18440 }, { "epoch": 0.05, "learning_rate": 9.561226030499774e-05, "loss": 1.0342, "step": 18445 }, { "epoch": 0.05, "learning_rate": 9.561100206351603e-05, "loss": 1.0339, "step": 18450 }, { "epoch": 0.05, "learning_rate": 9.560974382203432e-05, "loss": 1.0332, "step": 18455 }, { "epoch": 0.05, "learning_rate": 9.560848558055262e-05, "loss": 1.0327, "step": 18460 }, { "epoch": 0.05, "learning_rate": 9.560722733907092e-05, "loss": 1.0351, "step": 18465 }, { "epoch": 0.05, "learning_rate": 9.560596909758921e-05, "loss": 1.0348, "step": 18470 }, { "epoch": 0.05, "learning_rate": 9.56047108561075e-05, "loss": 1.0345, "step": 18475 }, { "epoch": 0.05, "learning_rate": 9.56034526146258e-05, "loss": 1.035, "step": 18480 }, { "epoch": 0.05, "learning_rate": 9.56021943731441e-05, "loss": 1.0348, "step": 18485 }, { "epoch": 0.05, "learning_rate": 9.560093613166239e-05, "loss": 1.0343, "step": 18490 }, { "epoch": 0.05, "learning_rate": 9.559967789018068e-05, "loss": 1.0347, "step": 18495 }, { "epoch": 0.05, "learning_rate": 9.559841964869898e-05, "loss": 1.0344, "step": 18500 }, { "epoch": 0.05, "learning_rate": 9.559716140721728e-05, "loss": 1.0342, "step": 18505 }, { "epoch": 0.05, "learning_rate": 9.559590316573557e-05, "loss": 1.035, "step": 18510 }, { "epoch": 0.05, "learning_rate": 9.559464492425386e-05, "loss": 1.0335, "step": 18515 }, { "epoch": 0.05, "learning_rate": 9.559338668277215e-05, "loss": 1.0373, "step": 18520 }, { "epoch": 0.05, "learning_rate": 9.559212844129046e-05, "loss": 1.0343, "step": 18525 }, { "epoch": 0.05, "learning_rate": 9.559087019980875e-05, "loss": 1.0357, "step": 18530 }, { "epoch": 0.05, "learning_rate": 9.558961195832705e-05, "loss": 1.0344, "step": 18535 }, { "epoch": 0.05, "learning_rate": 9.558835371684534e-05, "loss": 1.0336, "step": 18540 }, { "epoch": 0.05, "learning_rate": 9.558709547536364e-05, "loss": 1.0348, "step": 18545 }, { "epoch": 0.05, "learning_rate": 9.558583723388194e-05, "loss": 1.0337, "step": 18550 }, { "epoch": 0.05, "learning_rate": 9.558457899240023e-05, "loss": 1.0339, "step": 18555 }, { "epoch": 0.05, "learning_rate": 9.558332075091852e-05, "loss": 1.0327, "step": 18560 }, { "epoch": 0.05, "learning_rate": 9.558206250943682e-05, "loss": 1.034, "step": 18565 }, { "epoch": 0.05, "learning_rate": 9.558080426795512e-05, "loss": 1.0351, "step": 18570 }, { "epoch": 0.05, "learning_rate": 9.557954602647341e-05, "loss": 1.0327, "step": 18575 }, { "epoch": 0.05, "learning_rate": 9.55782877849917e-05, "loss": 1.0358, "step": 18580 }, { "epoch": 0.05, "learning_rate": 9.557702954351e-05, "loss": 1.0323, "step": 18585 }, { "epoch": 0.05, "learning_rate": 9.557577130202829e-05, "loss": 1.0364, "step": 18590 }, { "epoch": 0.05, "learning_rate": 9.557451306054659e-05, "loss": 1.0315, "step": 18595 }, { "epoch": 0.05, "learning_rate": 9.557325481906488e-05, "loss": 1.0354, "step": 18600 }, { "epoch": 0.05, "learning_rate": 9.557199657758318e-05, "loss": 1.0354, "step": 18605 }, { "epoch": 0.05, "learning_rate": 9.557073833610147e-05, "loss": 1.0341, "step": 18610 }, { "epoch": 0.05, "learning_rate": 9.556948009461977e-05, "loss": 1.0334, "step": 18615 }, { "epoch": 0.05, "learning_rate": 9.556822185313806e-05, "loss": 1.0338, "step": 18620 }, { "epoch": 0.05, "learning_rate": 9.556696361165636e-05, "loss": 1.0341, "step": 18625 }, { "epoch": 0.05, "learning_rate": 9.556570537017465e-05, "loss": 1.0339, "step": 18630 }, { "epoch": 0.05, "learning_rate": 9.556444712869295e-05, "loss": 1.0351, "step": 18635 }, { "epoch": 0.05, "learning_rate": 9.556318888721124e-05, "loss": 1.0328, "step": 18640 }, { "epoch": 0.05, "learning_rate": 9.556193064572954e-05, "loss": 1.0341, "step": 18645 }, { "epoch": 0.05, "learning_rate": 9.556067240424783e-05, "loss": 1.0349, "step": 18650 }, { "epoch": 0.05, "learning_rate": 9.555941416276612e-05, "loss": 1.0329, "step": 18655 }, { "epoch": 0.05, "learning_rate": 9.555815592128442e-05, "loss": 1.0325, "step": 18660 }, { "epoch": 0.05, "learning_rate": 9.555689767980272e-05, "loss": 1.0347, "step": 18665 }, { "epoch": 0.05, "learning_rate": 9.555563943832101e-05, "loss": 1.0349, "step": 18670 }, { "epoch": 0.05, "learning_rate": 9.55543811968393e-05, "loss": 1.0337, "step": 18675 }, { "epoch": 0.05, "learning_rate": 9.55531229553576e-05, "loss": 1.0324, "step": 18680 }, { "epoch": 0.05, "learning_rate": 9.55518647138759e-05, "loss": 1.0321, "step": 18685 }, { "epoch": 0.05, "learning_rate": 9.555060647239419e-05, "loss": 1.0334, "step": 18690 }, { "epoch": 0.05, "learning_rate": 9.554934823091248e-05, "loss": 1.0342, "step": 18695 }, { "epoch": 0.05, "learning_rate": 9.554808998943078e-05, "loss": 1.036, "step": 18700 }, { "epoch": 0.05, "learning_rate": 9.554683174794908e-05, "loss": 1.0345, "step": 18705 }, { "epoch": 0.05, "learning_rate": 9.554557350646737e-05, "loss": 1.0329, "step": 18710 }, { "epoch": 0.05, "learning_rate": 9.554431526498566e-05, "loss": 1.0347, "step": 18715 }, { "epoch": 0.05, "learning_rate": 9.554305702350395e-05, "loss": 1.0341, "step": 18720 }, { "epoch": 0.05, "learning_rate": 9.554179878202226e-05, "loss": 1.0338, "step": 18725 }, { "epoch": 0.05, "learning_rate": 9.554054054054055e-05, "loss": 1.0341, "step": 18730 }, { "epoch": 0.05, "learning_rate": 9.553928229905884e-05, "loss": 1.0337, "step": 18735 }, { "epoch": 0.05, "learning_rate": 9.553802405757713e-05, "loss": 1.0329, "step": 18740 }, { "epoch": 0.05, "learning_rate": 9.553676581609544e-05, "loss": 1.0335, "step": 18745 }, { "epoch": 0.05, "learning_rate": 9.553550757461373e-05, "loss": 1.0328, "step": 18750 }, { "epoch": 0.05, "learning_rate": 9.553424933313202e-05, "loss": 1.034, "step": 18755 }, { "epoch": 0.05, "learning_rate": 9.553299109165031e-05, "loss": 1.0345, "step": 18760 }, { "epoch": 0.05, "learning_rate": 9.553173285016862e-05, "loss": 1.0327, "step": 18765 }, { "epoch": 0.05, "learning_rate": 9.553047460868691e-05, "loss": 1.0361, "step": 18770 }, { "epoch": 0.05, "learning_rate": 9.55292163672052e-05, "loss": 1.0338, "step": 18775 }, { "epoch": 0.05, "learning_rate": 9.552795812572349e-05, "loss": 1.033, "step": 18780 }, { "epoch": 0.05, "learning_rate": 9.552669988424178e-05, "loss": 1.0359, "step": 18785 }, { "epoch": 0.05, "learning_rate": 9.552544164276009e-05, "loss": 1.033, "step": 18790 }, { "epoch": 0.05, "learning_rate": 9.552443504957472e-05, "loss": 1.0587, "step": 18795 }, { "epoch": 0.05, "learning_rate": 9.552317680809301e-05, "loss": 1.0342, "step": 18800 }, { "epoch": 0.05, "learning_rate": 9.55219185666113e-05, "loss": 1.0331, "step": 18805 }, { "epoch": 0.05, "learning_rate": 9.552066032512961e-05, "loss": 1.0358, "step": 18810 }, { "epoch": 0.05, "learning_rate": 9.55194020836479e-05, "loss": 1.037, "step": 18815 }, { "epoch": 0.05, "learning_rate": 9.55181438421662e-05, "loss": 1.0343, "step": 18820 }, { "epoch": 0.05, "learning_rate": 9.551688560068449e-05, "loss": 1.0325, "step": 18825 }, { "epoch": 0.05, "learning_rate": 9.551562735920279e-05, "loss": 1.0346, "step": 18830 }, { "epoch": 0.05, "learning_rate": 9.551436911772108e-05, "loss": 1.0345, "step": 18835 }, { "epoch": 0.05, "learning_rate": 9.551311087623937e-05, "loss": 1.035, "step": 18840 }, { "epoch": 0.05, "learning_rate": 9.551185263475767e-05, "loss": 1.0338, "step": 18845 }, { "epoch": 0.05, "learning_rate": 9.551059439327597e-05, "loss": 1.0326, "step": 18850 }, { "epoch": 0.05, "learning_rate": 9.550933615179426e-05, "loss": 1.0339, "step": 18855 }, { "epoch": 0.05, "learning_rate": 9.550807791031255e-05, "loss": 1.0336, "step": 18860 }, { "epoch": 0.05, "learning_rate": 9.550681966883085e-05, "loss": 1.0323, "step": 18865 }, { "epoch": 0.05, "learning_rate": 9.550556142734914e-05, "loss": 1.0345, "step": 18870 }, { "epoch": 0.05, "learning_rate": 9.550430318586744e-05, "loss": 1.0351, "step": 18875 }, { "epoch": 0.05, "learning_rate": 9.550304494438573e-05, "loss": 1.0319, "step": 18880 }, { "epoch": 0.05, "learning_rate": 9.550178670290403e-05, "loss": 1.0358, "step": 18885 }, { "epoch": 0.05, "learning_rate": 9.550052846142232e-05, "loss": 1.0323, "step": 18890 }, { "epoch": 0.05, "learning_rate": 9.549927021994062e-05, "loss": 1.0345, "step": 18895 }, { "epoch": 0.05, "learning_rate": 9.549801197845891e-05, "loss": 1.0339, "step": 18900 }, { "epoch": 0.05, "learning_rate": 9.54967537369772e-05, "loss": 1.0317, "step": 18905 }, { "epoch": 0.05, "learning_rate": 9.54954954954955e-05, "loss": 1.0335, "step": 18910 }, { "epoch": 0.05, "learning_rate": 9.54942372540138e-05, "loss": 1.0335, "step": 18915 }, { "epoch": 0.05, "learning_rate": 9.54929790125321e-05, "loss": 1.0338, "step": 18920 }, { "epoch": 0.05, "learning_rate": 9.549172077105039e-05, "loss": 1.034, "step": 18925 }, { "epoch": 0.05, "learning_rate": 9.549046252956868e-05, "loss": 1.035, "step": 18930 }, { "epoch": 0.05, "learning_rate": 9.548920428808697e-05, "loss": 1.0349, "step": 18935 }, { "epoch": 0.05, "learning_rate": 9.548794604660527e-05, "loss": 1.0336, "step": 18940 }, { "epoch": 0.05, "learning_rate": 9.548668780512357e-05, "loss": 1.0342, "step": 18945 }, { "epoch": 0.05, "learning_rate": 9.548542956364186e-05, "loss": 1.0354, "step": 18950 }, { "epoch": 0.05, "learning_rate": 9.548417132216015e-05, "loss": 1.0365, "step": 18955 }, { "epoch": 0.05, "learning_rate": 9.548291308067845e-05, "loss": 1.0364, "step": 18960 }, { "epoch": 0.05, "learning_rate": 9.548165483919675e-05, "loss": 1.0362, "step": 18965 }, { "epoch": 0.05, "learning_rate": 9.548039659771504e-05, "loss": 1.0318, "step": 18970 }, { "epoch": 0.05, "learning_rate": 9.547913835623333e-05, "loss": 1.0333, "step": 18975 }, { "epoch": 0.05, "learning_rate": 9.547788011475163e-05, "loss": 1.035, "step": 18980 }, { "epoch": 0.05, "learning_rate": 9.547662187326993e-05, "loss": 1.0331, "step": 18985 }, { "epoch": 0.05, "learning_rate": 9.547536363178822e-05, "loss": 1.0329, "step": 18990 }, { "epoch": 0.05, "learning_rate": 9.547410539030651e-05, "loss": 1.0358, "step": 18995 }, { "epoch": 0.05, "learning_rate": 9.54728471488248e-05, "loss": 1.0339, "step": 19000 }, { "epoch": 0.05, "learning_rate": 9.54715889073431e-05, "loss": 1.034, "step": 19005 }, { "epoch": 0.05, "learning_rate": 9.54703306658614e-05, "loss": 1.0328, "step": 19010 }, { "epoch": 0.05, "learning_rate": 9.546907242437969e-05, "loss": 1.0332, "step": 19015 }, { "epoch": 0.05, "learning_rate": 9.546781418289798e-05, "loss": 1.0327, "step": 19020 }, { "epoch": 0.05, "learning_rate": 9.546655594141629e-05, "loss": 1.033, "step": 19025 }, { "epoch": 0.05, "learning_rate": 9.546529769993458e-05, "loss": 1.0372, "step": 19030 }, { "epoch": 0.05, "learning_rate": 9.546403945845287e-05, "loss": 1.0321, "step": 19035 }, { "epoch": 0.05, "learning_rate": 9.546278121697116e-05, "loss": 1.0337, "step": 19040 }, { "epoch": 0.05, "learning_rate": 9.546152297548947e-05, "loss": 1.0347, "step": 19045 }, { "epoch": 0.05, "learning_rate": 9.546026473400776e-05, "loss": 1.033, "step": 19050 }, { "epoch": 0.05, "learning_rate": 9.545900649252605e-05, "loss": 1.0352, "step": 19055 }, { "epoch": 0.05, "learning_rate": 9.545774825104434e-05, "loss": 1.0337, "step": 19060 }, { "epoch": 0.05, "learning_rate": 9.545649000956263e-05, "loss": 1.0355, "step": 19065 }, { "epoch": 0.05, "learning_rate": 9.545523176808094e-05, "loss": 1.0347, "step": 19070 }, { "epoch": 0.05, "learning_rate": 9.545397352659923e-05, "loss": 1.034, "step": 19075 }, { "epoch": 0.05, "learning_rate": 9.545271528511752e-05, "loss": 1.0471, "step": 19080 }, { "epoch": 0.05, "learning_rate": 9.545145704363581e-05, "loss": 1.0365, "step": 19085 }, { "epoch": 0.05, "learning_rate": 9.545019880215412e-05, "loss": 1.0352, "step": 19090 }, { "epoch": 0.05, "learning_rate": 9.544894056067241e-05, "loss": 1.0329, "step": 19095 }, { "epoch": 0.05, "learning_rate": 9.54476823191907e-05, "loss": 1.0332, "step": 19100 }, { "epoch": 0.05, "learning_rate": 9.544642407770899e-05, "loss": 1.0357, "step": 19105 }, { "epoch": 0.05, "learning_rate": 9.54451658362273e-05, "loss": 1.0315, "step": 19110 }, { "epoch": 0.05, "learning_rate": 9.544390759474559e-05, "loss": 1.0347, "step": 19115 }, { "epoch": 0.05, "learning_rate": 9.544264935326388e-05, "loss": 1.0347, "step": 19120 }, { "epoch": 0.05, "learning_rate": 9.544139111178217e-05, "loss": 1.0346, "step": 19125 }, { "epoch": 0.05, "learning_rate": 9.544013287030046e-05, "loss": 1.0368, "step": 19130 }, { "epoch": 0.05, "learning_rate": 9.543887462881877e-05, "loss": 1.035, "step": 19135 }, { "epoch": 0.05, "learning_rate": 9.543761638733706e-05, "loss": 1.0353, "step": 19140 }, { "epoch": 0.05, "learning_rate": 9.543635814585535e-05, "loss": 1.035, "step": 19145 }, { "epoch": 0.05, "learning_rate": 9.543509990437364e-05, "loss": 1.0351, "step": 19150 }, { "epoch": 0.05, "learning_rate": 9.543384166289195e-05, "loss": 1.0617, "step": 19155 }, { "epoch": 0.05, "learning_rate": 9.543258342141024e-05, "loss": 1.0367, "step": 19160 }, { "epoch": 0.05, "learning_rate": 9.543132517992853e-05, "loss": 1.0338, "step": 19165 }, { "epoch": 0.05, "learning_rate": 9.543006693844682e-05, "loss": 1.0361, "step": 19170 }, { "epoch": 0.05, "learning_rate": 9.542880869696513e-05, "loss": 1.0348, "step": 19175 }, { "epoch": 0.05, "learning_rate": 9.542755045548342e-05, "loss": 1.0348, "step": 19180 }, { "epoch": 0.05, "learning_rate": 9.542629221400171e-05, "loss": 1.0306, "step": 19185 }, { "epoch": 0.05, "learning_rate": 9.542503397252e-05, "loss": 1.0355, "step": 19190 }, { "epoch": 0.05, "learning_rate": 9.54237757310383e-05, "loss": 1.0348, "step": 19195 }, { "epoch": 0.05, "learning_rate": 9.54225174895566e-05, "loss": 1.0345, "step": 19200 }, { "epoch": 0.05, "learning_rate": 9.542125924807489e-05, "loss": 1.0342, "step": 19205 }, { "epoch": 0.05, "learning_rate": 9.542000100659318e-05, "loss": 1.0323, "step": 19210 }, { "epoch": 0.05, "learning_rate": 9.541874276511147e-05, "loss": 1.035, "step": 19215 }, { "epoch": 0.05, "learning_rate": 9.541748452362978e-05, "loss": 1.0362, "step": 19220 }, { "epoch": 0.05, "learning_rate": 9.541622628214807e-05, "loss": 1.0322, "step": 19225 }, { "epoch": 0.05, "learning_rate": 9.541496804066636e-05, "loss": 1.0367, "step": 19230 }, { "epoch": 0.05, "learning_rate": 9.541370979918467e-05, "loss": 1.0342, "step": 19235 }, { "epoch": 0.05, "learning_rate": 9.541245155770296e-05, "loss": 1.0325, "step": 19240 }, { "epoch": 0.05, "learning_rate": 9.541119331622126e-05, "loss": 1.0346, "step": 19245 }, { "epoch": 0.05, "learning_rate": 9.540993507473956e-05, "loss": 1.034, "step": 19250 }, { "epoch": 0.05, "learning_rate": 9.540867683325785e-05, "loss": 1.0352, "step": 19255 }, { "epoch": 0.05, "learning_rate": 9.540741859177614e-05, "loss": 1.0344, "step": 19260 }, { "epoch": 0.05, "learning_rate": 9.540616035029443e-05, "loss": 1.0351, "step": 19265 }, { "epoch": 0.05, "learning_rate": 9.540490210881274e-05, "loss": 1.0331, "step": 19270 }, { "epoch": 0.05, "learning_rate": 9.540364386733103e-05, "loss": 1.0369, "step": 19275 }, { "epoch": 0.05, "learning_rate": 9.540238562584932e-05, "loss": 1.0337, "step": 19280 }, { "epoch": 0.05, "learning_rate": 9.540112738436761e-05, "loss": 1.033, "step": 19285 }, { "epoch": 0.05, "learning_rate": 9.539986914288592e-05, "loss": 1.0353, "step": 19290 }, { "epoch": 0.05, "learning_rate": 9.539861090140421e-05, "loss": 1.0318, "step": 19295 }, { "epoch": 0.05, "learning_rate": 9.53973526599225e-05, "loss": 1.0512, "step": 19300 }, { "epoch": 0.05, "learning_rate": 9.539609441844079e-05, "loss": 1.0355, "step": 19305 }, { "epoch": 0.05, "learning_rate": 9.53948361769591e-05, "loss": 1.0356, "step": 19310 }, { "epoch": 0.05, "learning_rate": 9.539357793547739e-05, "loss": 1.0361, "step": 19315 }, { "epoch": 0.05, "learning_rate": 9.539231969399568e-05, "loss": 1.0338, "step": 19320 }, { "epoch": 0.05, "learning_rate": 9.539106145251397e-05, "loss": 1.0321, "step": 19325 }, { "epoch": 0.05, "learning_rate": 9.538980321103226e-05, "loss": 1.033, "step": 19330 }, { "epoch": 0.05, "learning_rate": 9.538854496955057e-05, "loss": 1.0333, "step": 19335 }, { "epoch": 0.05, "learning_rate": 9.538728672806886e-05, "loss": 1.0331, "step": 19340 }, { "epoch": 0.05, "learning_rate": 9.538602848658715e-05, "loss": 1.0341, "step": 19345 }, { "epoch": 0.05, "learning_rate": 9.538477024510544e-05, "loss": 1.0361, "step": 19350 }, { "epoch": 0.05, "learning_rate": 9.538351200362375e-05, "loss": 1.0355, "step": 19355 }, { "epoch": 0.05, "learning_rate": 9.538225376214204e-05, "loss": 1.0354, "step": 19360 }, { "epoch": 0.05, "learning_rate": 9.538099552066033e-05, "loss": 1.0347, "step": 19365 }, { "epoch": 0.05, "learning_rate": 9.537973727917862e-05, "loss": 1.0346, "step": 19370 }, { "epoch": 0.05, "learning_rate": 9.537847903769693e-05, "loss": 1.036, "step": 19375 }, { "epoch": 0.05, "learning_rate": 9.537722079621522e-05, "loss": 1.0331, "step": 19380 }, { "epoch": 0.05, "learning_rate": 9.537596255473351e-05, "loss": 1.0322, "step": 19385 }, { "epoch": 0.05, "learning_rate": 9.53747043132518e-05, "loss": 1.0342, "step": 19390 }, { "epoch": 0.05, "learning_rate": 9.53734460717701e-05, "loss": 1.0331, "step": 19395 }, { "epoch": 0.05, "learning_rate": 9.53721878302884e-05, "loss": 1.0334, "step": 19400 }, { "epoch": 0.05, "learning_rate": 9.537092958880669e-05, "loss": 1.0339, "step": 19405 }, { "epoch": 0.05, "learning_rate": 9.536967134732498e-05, "loss": 1.0345, "step": 19410 }, { "epoch": 0.05, "learning_rate": 9.536841310584327e-05, "loss": 1.0359, "step": 19415 }, { "epoch": 0.05, "learning_rate": 9.536715486436158e-05, "loss": 1.0337, "step": 19420 }, { "epoch": 0.05, "learning_rate": 9.536589662287987e-05, "loss": 1.0337, "step": 19425 }, { "epoch": 0.05, "learning_rate": 9.536463838139816e-05, "loss": 1.0335, "step": 19430 }, { "epoch": 0.05, "learning_rate": 9.536338013991645e-05, "loss": 1.057, "step": 19435 }, { "epoch": 0.05, "learning_rate": 9.536212189843476e-05, "loss": 1.0338, "step": 19440 }, { "epoch": 0.05, "learning_rate": 9.536086365695305e-05, "loss": 1.0328, "step": 19445 }, { "epoch": 0.05, "learning_rate": 9.535960541547134e-05, "loss": 1.0336, "step": 19450 }, { "epoch": 0.05, "learning_rate": 9.535834717398963e-05, "loss": 1.0328, "step": 19455 }, { "epoch": 0.05, "learning_rate": 9.535708893250793e-05, "loss": 1.0347, "step": 19460 }, { "epoch": 0.05, "learning_rate": 9.535583069102623e-05, "loss": 1.033, "step": 19465 }, { "epoch": 0.05, "learning_rate": 9.535457244954452e-05, "loss": 1.0325, "step": 19470 }, { "epoch": 0.05, "learning_rate": 9.535331420806281e-05, "loss": 1.0352, "step": 19475 }, { "epoch": 0.05, "learning_rate": 9.53520559665811e-05, "loss": 1.0351, "step": 19480 }, { "epoch": 0.05, "learning_rate": 9.535079772509941e-05, "loss": 1.0341, "step": 19485 }, { "epoch": 0.05, "learning_rate": 9.53495394836177e-05, "loss": 1.0322, "step": 19490 }, { "epoch": 0.05, "learning_rate": 9.5348281242136e-05, "loss": 1.0337, "step": 19495 }, { "epoch": 0.05, "learning_rate": 9.534702300065429e-05, "loss": 1.0355, "step": 19500 }, { "epoch": 0.05, "learning_rate": 9.534576475917259e-05, "loss": 1.0341, "step": 19505 }, { "epoch": 0.05, "learning_rate": 9.534450651769088e-05, "loss": 1.0366, "step": 19510 }, { "epoch": 0.05, "learning_rate": 9.534324827620917e-05, "loss": 1.0318, "step": 19515 }, { "epoch": 0.05, "learning_rate": 9.534199003472747e-05, "loss": 1.0362, "step": 19520 }, { "epoch": 0.05, "learning_rate": 9.534073179324576e-05, "loss": 1.033, "step": 19525 }, { "epoch": 0.05, "learning_rate": 9.533947355176406e-05, "loss": 1.0352, "step": 19530 }, { "epoch": 0.05, "learning_rate": 9.533821531028235e-05, "loss": 1.0333, "step": 19535 }, { "epoch": 0.05, "learning_rate": 9.533695706880065e-05, "loss": 1.0326, "step": 19540 }, { "epoch": 0.05, "learning_rate": 9.533569882731894e-05, "loss": 1.033, "step": 19545 }, { "epoch": 0.05, "learning_rate": 9.533444058583724e-05, "loss": 1.0344, "step": 19550 }, { "epoch": 0.05, "learning_rate": 9.533318234435553e-05, "loss": 1.0323, "step": 19555 }, { "epoch": 0.05, "learning_rate": 9.533192410287383e-05, "loss": 1.0348, "step": 19560 }, { "epoch": 0.05, "learning_rate": 9.533066586139212e-05, "loss": 1.033, "step": 19565 }, { "epoch": 0.05, "learning_rate": 9.532940761991042e-05, "loss": 1.0345, "step": 19570 }, { "epoch": 0.05, "learning_rate": 9.532814937842871e-05, "loss": 1.035, "step": 19575 }, { "epoch": 0.05, "learning_rate": 9.5326891136947e-05, "loss": 1.0335, "step": 19580 }, { "epoch": 0.05, "learning_rate": 9.53256328954653e-05, "loss": 1.0304, "step": 19585 }, { "epoch": 0.05, "learning_rate": 9.532437465398359e-05, "loss": 1.0352, "step": 19590 }, { "epoch": 0.05, "learning_rate": 9.53231164125019e-05, "loss": 1.0337, "step": 19595 }, { "epoch": 0.05, "learning_rate": 9.532185817102019e-05, "loss": 1.0358, "step": 19600 }, { "epoch": 0.05, "learning_rate": 9.532059992953848e-05, "loss": 1.0338, "step": 19605 }, { "epoch": 0.05, "learning_rate": 9.531934168805677e-05, "loss": 1.0339, "step": 19610 }, { "epoch": 0.05, "learning_rate": 9.531808344657507e-05, "loss": 1.0318, "step": 19615 }, { "epoch": 0.05, "learning_rate": 9.531682520509337e-05, "loss": 1.0334, "step": 19620 }, { "epoch": 0.05, "learning_rate": 9.531556696361166e-05, "loss": 1.0319, "step": 19625 }, { "epoch": 0.05, "learning_rate": 9.531430872212995e-05, "loss": 1.0347, "step": 19630 }, { "epoch": 0.05, "learning_rate": 9.531305048064825e-05, "loss": 1.0323, "step": 19635 }, { "epoch": 0.05, "learning_rate": 9.531179223916655e-05, "loss": 1.035, "step": 19640 }, { "epoch": 0.05, "learning_rate": 9.531053399768484e-05, "loss": 1.0331, "step": 19645 }, { "epoch": 0.05, "learning_rate": 9.530927575620313e-05, "loss": 1.0332, "step": 19650 }, { "epoch": 0.05, "learning_rate": 9.530801751472142e-05, "loss": 1.0338, "step": 19655 }, { "epoch": 0.05, "learning_rate": 9.530675927323973e-05, "loss": 1.0335, "step": 19660 }, { "epoch": 0.05, "learning_rate": 9.530550103175802e-05, "loss": 1.0333, "step": 19665 }, { "epoch": 0.05, "learning_rate": 9.530424279027631e-05, "loss": 1.0307, "step": 19670 }, { "epoch": 0.05, "learning_rate": 9.53029845487946e-05, "loss": 1.0322, "step": 19675 }, { "epoch": 0.05, "learning_rate": 9.53017263073129e-05, "loss": 1.0333, "step": 19680 }, { "epoch": 0.05, "learning_rate": 9.53004680658312e-05, "loss": 1.0339, "step": 19685 }, { "epoch": 0.05, "learning_rate": 9.529920982434949e-05, "loss": 1.0338, "step": 19690 }, { "epoch": 0.05, "learning_rate": 9.529795158286778e-05, "loss": 1.0337, "step": 19695 }, { "epoch": 0.05, "learning_rate": 9.529669334138608e-05, "loss": 1.0321, "step": 19700 }, { "epoch": 0.05, "learning_rate": 9.529543509990438e-05, "loss": 1.0336, "step": 19705 }, { "epoch": 0.05, "learning_rate": 9.529417685842267e-05, "loss": 1.0346, "step": 19710 }, { "epoch": 0.05, "learning_rate": 9.529291861694096e-05, "loss": 1.0327, "step": 19715 }, { "epoch": 0.05, "learning_rate": 9.529166037545925e-05, "loss": 1.0346, "step": 19720 }, { "epoch": 0.05, "learning_rate": 9.529040213397756e-05, "loss": 1.0319, "step": 19725 }, { "epoch": 0.05, "learning_rate": 9.528914389249585e-05, "loss": 1.053, "step": 19730 }, { "epoch": 0.05, "learning_rate": 9.528788565101415e-05, "loss": 1.0324, "step": 19735 }, { "epoch": 0.05, "learning_rate": 9.528662740953244e-05, "loss": 1.0329, "step": 19740 }, { "epoch": 0.05, "learning_rate": 9.528536916805074e-05, "loss": 1.0343, "step": 19745 }, { "epoch": 0.05, "learning_rate": 9.528411092656904e-05, "loss": 1.0348, "step": 19750 }, { "epoch": 0.05, "learning_rate": 9.528285268508733e-05, "loss": 1.0327, "step": 19755 }, { "epoch": 0.05, "learning_rate": 9.528159444360562e-05, "loss": 1.0369, "step": 19760 }, { "epoch": 0.05, "learning_rate": 9.528033620212392e-05, "loss": 1.0336, "step": 19765 }, { "epoch": 0.05, "learning_rate": 9.527907796064222e-05, "loss": 1.0346, "step": 19770 }, { "epoch": 0.05, "learning_rate": 9.527781971916051e-05, "loss": 1.0339, "step": 19775 }, { "epoch": 0.05, "learning_rate": 9.52765614776788e-05, "loss": 1.0308, "step": 19780 }, { "epoch": 0.05, "learning_rate": 9.52753032361971e-05, "loss": 1.0295, "step": 19785 }, { "epoch": 0.05, "learning_rate": 9.527404499471539e-05, "loss": 1.0349, "step": 19790 }, { "epoch": 0.05, "learning_rate": 9.527278675323369e-05, "loss": 1.0333, "step": 19795 }, { "epoch": 0.05, "learning_rate": 9.527152851175198e-05, "loss": 1.0335, "step": 19800 }, { "epoch": 0.05, "learning_rate": 9.527027027027028e-05, "loss": 1.0347, "step": 19805 }, { "epoch": 0.05, "learning_rate": 9.526901202878857e-05, "loss": 1.0587, "step": 19810 }, { "epoch": 0.05, "learning_rate": 9.526775378730687e-05, "loss": 1.0354, "step": 19815 }, { "epoch": 0.05, "learning_rate": 9.526649554582516e-05, "loss": 1.0311, "step": 19820 }, { "epoch": 0.05, "learning_rate": 9.526523730434346e-05, "loss": 1.0362, "step": 19825 }, { "epoch": 0.05, "learning_rate": 9.526397906286175e-05, "loss": 1.034, "step": 19830 }, { "epoch": 0.05, "learning_rate": 9.526272082138005e-05, "loss": 1.0343, "step": 19835 }, { "epoch": 0.05, "learning_rate": 9.526146257989834e-05, "loss": 1.0332, "step": 19840 }, { "epoch": 0.05, "learning_rate": 9.526020433841664e-05, "loss": 1.0338, "step": 19845 }, { "epoch": 0.05, "learning_rate": 9.525894609693493e-05, "loss": 1.0339, "step": 19850 }, { "epoch": 0.05, "learning_rate": 9.525768785545322e-05, "loss": 1.0341, "step": 19855 }, { "epoch": 0.05, "learning_rate": 9.525642961397152e-05, "loss": 1.0333, "step": 19860 }, { "epoch": 0.05, "learning_rate": 9.525517137248982e-05, "loss": 1.035, "step": 19865 }, { "epoch": 0.05, "learning_rate": 9.525391313100811e-05, "loss": 1.0345, "step": 19870 }, { "epoch": 0.05, "learning_rate": 9.52526548895264e-05, "loss": 1.0353, "step": 19875 }, { "epoch": 0.05, "learning_rate": 9.52513966480447e-05, "loss": 1.0345, "step": 19880 }, { "epoch": 0.05, "learning_rate": 9.5250138406563e-05, "loss": 1.0348, "step": 19885 }, { "epoch": 0.05, "learning_rate": 9.524888016508129e-05, "loss": 1.0358, "step": 19890 }, { "epoch": 0.05, "learning_rate": 9.524762192359958e-05, "loss": 1.036, "step": 19895 }, { "epoch": 0.05, "learning_rate": 9.524636368211788e-05, "loss": 1.0336, "step": 19900 }, { "epoch": 0.05, "learning_rate": 9.524510544063618e-05, "loss": 1.0329, "step": 19905 }, { "epoch": 0.05, "learning_rate": 9.524384719915447e-05, "loss": 1.0332, "step": 19910 }, { "epoch": 0.05, "learning_rate": 9.524258895767276e-05, "loss": 1.0336, "step": 19915 }, { "epoch": 0.05, "learning_rate": 9.524133071619105e-05, "loss": 1.0322, "step": 19920 }, { "epoch": 0.05, "learning_rate": 9.524007247470936e-05, "loss": 1.0342, "step": 19925 }, { "epoch": 0.05, "learning_rate": 9.523881423322765e-05, "loss": 1.034, "step": 19930 }, { "epoch": 0.05, "learning_rate": 9.523755599174594e-05, "loss": 1.033, "step": 19935 }, { "epoch": 0.05, "learning_rate": 9.523629775026423e-05, "loss": 1.0332, "step": 19940 }, { "epoch": 0.05, "learning_rate": 9.523503950878254e-05, "loss": 1.0344, "step": 19945 }, { "epoch": 0.05, "learning_rate": 9.523378126730083e-05, "loss": 1.0333, "step": 19950 }, { "epoch": 0.05, "learning_rate": 9.523252302581912e-05, "loss": 1.0352, "step": 19955 }, { "epoch": 0.05, "learning_rate": 9.523126478433741e-05, "loss": 1.0344, "step": 19960 }, { "epoch": 0.05, "learning_rate": 9.523000654285572e-05, "loss": 1.0344, "step": 19965 }, { "epoch": 0.05, "learning_rate": 9.522874830137401e-05, "loss": 1.0334, "step": 19970 }, { "epoch": 0.05, "learning_rate": 9.52274900598923e-05, "loss": 1.0342, "step": 19975 }, { "epoch": 0.05, "learning_rate": 9.522623181841059e-05, "loss": 1.033, "step": 19980 }, { "epoch": 0.05, "learning_rate": 9.522497357692888e-05, "loss": 1.0342, "step": 19985 }, { "epoch": 0.05, "learning_rate": 9.522371533544719e-05, "loss": 1.0345, "step": 19990 }, { "epoch": 0.05, "learning_rate": 9.522245709396548e-05, "loss": 1.0341, "step": 19995 }, { "epoch": 0.05, "learning_rate": 9.522119885248377e-05, "loss": 1.0362, "step": 20000 }, { "epoch": 0.05, "learning_rate": 9.521994061100206e-05, "loss": 1.0331, "step": 20005 }, { "epoch": 0.05, "learning_rate": 9.521868236952037e-05, "loss": 1.032, "step": 20010 }, { "epoch": 0.05, "learning_rate": 9.521742412803866e-05, "loss": 1.0331, "step": 20015 }, { "epoch": 0.05, "learning_rate": 9.521616588655695e-05, "loss": 1.0314, "step": 20020 }, { "epoch": 0.05, "learning_rate": 9.521490764507524e-05, "loss": 1.0338, "step": 20025 }, { "epoch": 0.05, "learning_rate": 9.521364940359355e-05, "loss": 1.0334, "step": 20030 }, { "epoch": 0.05, "learning_rate": 9.521239116211184e-05, "loss": 1.0317, "step": 20035 }, { "epoch": 0.05, "learning_rate": 9.521113292063013e-05, "loss": 1.0344, "step": 20040 }, { "epoch": 0.05, "learning_rate": 9.520987467914842e-05, "loss": 1.0351, "step": 20045 }, { "epoch": 0.05, "learning_rate": 9.520861643766671e-05, "loss": 1.0362, "step": 20050 }, { "epoch": 0.05, "learning_rate": 9.520735819618502e-05, "loss": 1.0331, "step": 20055 }, { "epoch": 0.05, "learning_rate": 9.520609995470331e-05, "loss": 1.0343, "step": 20060 }, { "epoch": 0.05, "learning_rate": 9.52048417132216e-05, "loss": 1.0351, "step": 20065 }, { "epoch": 0.05, "learning_rate": 9.52035834717399e-05, "loss": 1.0349, "step": 20070 }, { "epoch": 0.05, "learning_rate": 9.52023252302582e-05, "loss": 1.0335, "step": 20075 }, { "epoch": 0.05, "learning_rate": 9.520106698877649e-05, "loss": 1.0313, "step": 20080 }, { "epoch": 0.05, "learning_rate": 9.519980874729478e-05, "loss": 1.0354, "step": 20085 }, { "epoch": 0.05, "learning_rate": 9.519855050581307e-05, "loss": 1.0327, "step": 20090 }, { "epoch": 0.05, "learning_rate": 9.519729226433138e-05, "loss": 1.0346, "step": 20095 }, { "epoch": 0.05, "learning_rate": 9.519603402284967e-05, "loss": 1.0339, "step": 20100 }, { "epoch": 0.05, "learning_rate": 9.519477578136796e-05, "loss": 1.033, "step": 20105 }, { "epoch": 0.05, "learning_rate": 9.519351753988625e-05, "loss": 1.0341, "step": 20110 }, { "epoch": 0.05, "learning_rate": 9.519225929840454e-05, "loss": 1.0338, "step": 20115 }, { "epoch": 0.05, "learning_rate": 9.519100105692285e-05, "loss": 1.0346, "step": 20120 }, { "epoch": 0.05, "learning_rate": 9.518974281544114e-05, "loss": 1.034, "step": 20125 }, { "epoch": 0.05, "learning_rate": 9.518848457395943e-05, "loss": 1.0323, "step": 20130 }, { "epoch": 0.05, "learning_rate": 9.518722633247772e-05, "loss": 1.0336, "step": 20135 }, { "epoch": 0.05, "learning_rate": 9.518596809099603e-05, "loss": 1.033, "step": 20140 }, { "epoch": 0.05, "learning_rate": 9.518470984951432e-05, "loss": 1.0345, "step": 20145 }, { "epoch": 0.05, "learning_rate": 9.518345160803261e-05, "loss": 1.0318, "step": 20150 }, { "epoch": 0.05, "learning_rate": 9.51821933665509e-05, "loss": 1.033, "step": 20155 }, { "epoch": 0.05, "learning_rate": 9.518093512506921e-05, "loss": 1.0346, "step": 20160 }, { "epoch": 0.05, "learning_rate": 9.51796768835875e-05, "loss": 1.0363, "step": 20165 }, { "epoch": 0.05, "learning_rate": 9.517841864210579e-05, "loss": 1.0333, "step": 20170 }, { "epoch": 0.05, "learning_rate": 9.517716040062408e-05, "loss": 1.0335, "step": 20175 }, { "epoch": 0.05, "learning_rate": 9.517590215914238e-05, "loss": 1.0325, "step": 20180 }, { "epoch": 0.05, "learning_rate": 9.517464391766068e-05, "loss": 1.0334, "step": 20185 }, { "epoch": 0.05, "learning_rate": 9.517338567617897e-05, "loss": 1.0352, "step": 20190 }, { "epoch": 0.05, "learning_rate": 9.517212743469726e-05, "loss": 1.0329, "step": 20195 }, { "epoch": 0.05, "learning_rate": 9.517086919321556e-05, "loss": 1.0343, "step": 20200 }, { "epoch": 0.05, "learning_rate": 9.516961095173386e-05, "loss": 1.033, "step": 20205 }, { "epoch": 0.05, "learning_rate": 9.516835271025215e-05, "loss": 1.0332, "step": 20210 }, { "epoch": 0.05, "learning_rate": 9.516709446877044e-05, "loss": 1.0343, "step": 20215 }, { "epoch": 0.05, "learning_rate": 9.516583622728874e-05, "loss": 1.0325, "step": 20220 }, { "epoch": 0.05, "learning_rate": 9.516457798580704e-05, "loss": 1.0318, "step": 20225 }, { "epoch": 0.05, "learning_rate": 9.516331974432533e-05, "loss": 1.0332, "step": 20230 }, { "epoch": 0.05, "learning_rate": 9.516206150284364e-05, "loss": 1.0331, "step": 20235 }, { "epoch": 0.05, "learning_rate": 9.516080326136193e-05, "loss": 1.0341, "step": 20240 }, { "epoch": 0.05, "learning_rate": 9.515954501988022e-05, "loss": 1.0337, "step": 20245 }, { "epoch": 0.05, "learning_rate": 9.515828677839853e-05, "loss": 1.0305, "step": 20250 }, { "epoch": 0.05, "learning_rate": 9.515702853691682e-05, "loss": 1.0306, "step": 20255 }, { "epoch": 0.05, "learning_rate": 9.515577029543511e-05, "loss": 1.0342, "step": 20260 }, { "epoch": 0.05, "learning_rate": 9.51545120539534e-05, "loss": 1.0352, "step": 20265 }, { "epoch": 0.05, "learning_rate": 9.515325381247169e-05, "loss": 1.0528, "step": 20270 }, { "epoch": 0.05, "learning_rate": 9.515199557099e-05, "loss": 1.0347, "step": 20275 }, { "epoch": 0.05, "learning_rate": 9.515073732950829e-05, "loss": 1.0319, "step": 20280 }, { "epoch": 0.05, "learning_rate": 9.514947908802658e-05, "loss": 1.0328, "step": 20285 }, { "epoch": 0.05, "learning_rate": 9.514822084654487e-05, "loss": 1.0529, "step": 20290 }, { "epoch": 0.05, "learning_rate": 9.514696260506318e-05, "loss": 1.0314, "step": 20295 }, { "epoch": 0.05, "learning_rate": 9.514570436358147e-05, "loss": 1.0343, "step": 20300 }, { "epoch": 0.05, "learning_rate": 9.514444612209976e-05, "loss": 1.0345, "step": 20305 }, { "epoch": 0.05, "learning_rate": 9.514318788061805e-05, "loss": 1.0327, "step": 20310 }, { "epoch": 0.05, "learning_rate": 9.514192963913636e-05, "loss": 1.0309, "step": 20315 }, { "epoch": 0.05, "learning_rate": 9.514067139765465e-05, "loss": 1.0335, "step": 20320 }, { "epoch": 0.05, "learning_rate": 9.513941315617294e-05, "loss": 1.0333, "step": 20325 }, { "epoch": 0.05, "learning_rate": 9.513815491469123e-05, "loss": 1.0327, "step": 20330 }, { "epoch": 0.05, "learning_rate": 9.513689667320952e-05, "loss": 1.0353, "step": 20335 }, { "epoch": 0.05, "learning_rate": 9.513563843172783e-05, "loss": 1.0341, "step": 20340 }, { "epoch": 0.05, "learning_rate": 9.513438019024612e-05, "loss": 1.0338, "step": 20345 }, { "epoch": 0.05, "learning_rate": 9.513312194876441e-05, "loss": 1.0363, "step": 20350 }, { "epoch": 0.05, "learning_rate": 9.51318637072827e-05, "loss": 1.0317, "step": 20355 }, { "epoch": 0.05, "learning_rate": 9.513060546580101e-05, "loss": 1.0344, "step": 20360 }, { "epoch": 0.05, "learning_rate": 9.51293472243193e-05, "loss": 1.0528, "step": 20365 }, { "epoch": 0.05, "learning_rate": 9.512808898283759e-05, "loss": 1.0354, "step": 20370 }, { "epoch": 0.05, "learning_rate": 9.512683074135588e-05, "loss": 1.0349, "step": 20375 }, { "epoch": 0.05, "learning_rate": 9.512557249987419e-05, "loss": 1.0334, "step": 20380 }, { "epoch": 0.05, "learning_rate": 9.512431425839248e-05, "loss": 1.0316, "step": 20385 }, { "epoch": 0.05, "learning_rate": 9.512305601691077e-05, "loss": 1.0328, "step": 20390 }, { "epoch": 0.05, "learning_rate": 9.512179777542906e-05, "loss": 1.0331, "step": 20395 }, { "epoch": 0.05, "learning_rate": 9.512053953394736e-05, "loss": 1.0339, "step": 20400 }, { "epoch": 0.05, "learning_rate": 9.511928129246566e-05, "loss": 1.0329, "step": 20405 }, { "epoch": 0.05, "learning_rate": 9.511802305098395e-05, "loss": 1.0318, "step": 20410 }, { "epoch": 0.05, "learning_rate": 9.511676480950224e-05, "loss": 1.0368, "step": 20415 }, { "epoch": 0.05, "learning_rate": 9.511550656802054e-05, "loss": 1.0342, "step": 20420 }, { "epoch": 0.05, "learning_rate": 9.511424832653884e-05, "loss": 1.0334, "step": 20425 }, { "epoch": 0.05, "learning_rate": 9.511299008505713e-05, "loss": 1.0351, "step": 20430 }, { "epoch": 0.05, "learning_rate": 9.511173184357542e-05, "loss": 1.0323, "step": 20435 }, { "epoch": 0.05, "learning_rate": 9.511047360209372e-05, "loss": 1.0343, "step": 20440 }, { "epoch": 0.05, "learning_rate": 9.510921536061201e-05, "loss": 1.0352, "step": 20445 }, { "epoch": 0.05, "learning_rate": 9.510795711913031e-05, "loss": 1.0343, "step": 20450 }, { "epoch": 0.05, "learning_rate": 9.51066988776486e-05, "loss": 1.0324, "step": 20455 }, { "epoch": 0.05, "learning_rate": 9.51054406361669e-05, "loss": 1.0355, "step": 20460 }, { "epoch": 0.05, "learning_rate": 9.510418239468519e-05, "loss": 1.0321, "step": 20465 }, { "epoch": 0.05, "learning_rate": 9.510292415320349e-05, "loss": 1.0333, "step": 20470 }, { "epoch": 0.05, "learning_rate": 9.510166591172178e-05, "loss": 1.034, "step": 20475 }, { "epoch": 0.05, "learning_rate": 9.510040767024008e-05, "loss": 1.0337, "step": 20480 }, { "epoch": 0.05, "learning_rate": 9.509914942875837e-05, "loss": 1.0342, "step": 20485 }, { "epoch": 0.05, "learning_rate": 9.509789118727667e-05, "loss": 1.0328, "step": 20490 }, { "epoch": 0.05, "learning_rate": 9.509663294579496e-05, "loss": 1.0344, "step": 20495 }, { "epoch": 0.05, "learning_rate": 9.509537470431326e-05, "loss": 1.034, "step": 20500 }, { "epoch": 0.05, "learning_rate": 9.509411646283155e-05, "loss": 1.0341, "step": 20505 }, { "epoch": 0.05, "learning_rate": 9.509285822134984e-05, "loss": 1.0339, "step": 20510 }, { "epoch": 0.05, "learning_rate": 9.509159997986814e-05, "loss": 1.035, "step": 20515 }, { "epoch": 0.05, "learning_rate": 9.509034173838644e-05, "loss": 1.0343, "step": 20520 }, { "epoch": 0.05, "learning_rate": 9.508908349690473e-05, "loss": 1.0332, "step": 20525 }, { "epoch": 0.05, "learning_rate": 9.508782525542302e-05, "loss": 1.0366, "step": 20530 }, { "epoch": 0.05, "learning_rate": 9.508656701394132e-05, "loss": 1.0356, "step": 20535 }, { "epoch": 0.05, "learning_rate": 9.508530877245962e-05, "loss": 1.0331, "step": 20540 }, { "epoch": 0.05, "learning_rate": 9.508405053097791e-05, "loss": 1.0347, "step": 20545 }, { "epoch": 0.05, "learning_rate": 9.50827922894962e-05, "loss": 1.0334, "step": 20550 }, { "epoch": 0.05, "learning_rate": 9.50815340480145e-05, "loss": 1.0345, "step": 20555 }, { "epoch": 0.05, "learning_rate": 9.50802758065328e-05, "loss": 1.0336, "step": 20560 }, { "epoch": 0.05, "learning_rate": 9.507901756505109e-05, "loss": 1.0331, "step": 20565 }, { "epoch": 0.05, "learning_rate": 9.507775932356938e-05, "loss": 1.0345, "step": 20570 }, { "epoch": 0.05, "learning_rate": 9.507650108208767e-05, "loss": 1.0329, "step": 20575 }, { "epoch": 0.05, "learning_rate": 9.507524284060598e-05, "loss": 1.0329, "step": 20580 }, { "epoch": 0.05, "learning_rate": 9.507398459912427e-05, "loss": 1.0343, "step": 20585 }, { "epoch": 0.05, "learning_rate": 9.507272635764256e-05, "loss": 1.0334, "step": 20590 }, { "epoch": 0.05, "learning_rate": 9.507146811616085e-05, "loss": 1.035, "step": 20595 }, { "epoch": 0.05, "learning_rate": 9.507020987467916e-05, "loss": 1.0332, "step": 20600 }, { "epoch": 0.05, "learning_rate": 9.506895163319745e-05, "loss": 1.0356, "step": 20605 }, { "epoch": 0.05, "learning_rate": 9.506769339171574e-05, "loss": 1.0322, "step": 20610 }, { "epoch": 0.05, "learning_rate": 9.506643515023403e-05, "loss": 1.0334, "step": 20615 }, { "epoch": 0.05, "learning_rate": 9.506517690875234e-05, "loss": 1.032, "step": 20620 }, { "epoch": 0.05, "learning_rate": 9.506391866727063e-05, "loss": 1.0332, "step": 20625 }, { "epoch": 0.05, "learning_rate": 9.506266042578892e-05, "loss": 1.034, "step": 20630 }, { "epoch": 0.05, "learning_rate": 9.506140218430721e-05, "loss": 1.0331, "step": 20635 }, { "epoch": 0.05, "learning_rate": 9.50601439428255e-05, "loss": 1.0339, "step": 20640 }, { "epoch": 0.05, "learning_rate": 9.50588857013438e-05, "loss": 1.0344, "step": 20645 }, { "epoch": 0.05, "learning_rate": 9.50576274598621e-05, "loss": 1.0326, "step": 20650 }, { "epoch": 0.05, "learning_rate": 9.505636921838039e-05, "loss": 1.0584, "step": 20655 }, { "epoch": 0.05, "learning_rate": 9.505511097689868e-05, "loss": 1.0344, "step": 20660 }, { "epoch": 0.05, "learning_rate": 9.505385273541699e-05, "loss": 1.0342, "step": 20665 }, { "epoch": 0.05, "learning_rate": 9.505259449393528e-05, "loss": 1.0334, "step": 20670 }, { "epoch": 0.05, "learning_rate": 9.505133625245357e-05, "loss": 1.0324, "step": 20675 }, { "epoch": 0.05, "learning_rate": 9.505007801097186e-05, "loss": 1.033, "step": 20680 }, { "epoch": 0.05, "learning_rate": 9.504881976949017e-05, "loss": 1.0339, "step": 20685 }, { "epoch": 0.05, "learning_rate": 9.504756152800846e-05, "loss": 1.0352, "step": 20690 }, { "epoch": 0.05, "learning_rate": 9.504630328652675e-05, "loss": 1.0349, "step": 20695 }, { "epoch": 0.05, "learning_rate": 9.504504504504504e-05, "loss": 1.0564, "step": 20700 }, { "epoch": 0.05, "learning_rate": 9.504378680356333e-05, "loss": 1.034, "step": 20705 }, { "epoch": 0.05, "learning_rate": 9.504252856208164e-05, "loss": 1.0335, "step": 20710 }, { "epoch": 0.05, "learning_rate": 9.504127032059993e-05, "loss": 1.0347, "step": 20715 }, { "epoch": 0.05, "learning_rate": 9.504001207911822e-05, "loss": 1.0331, "step": 20720 }, { "epoch": 0.05, "learning_rate": 9.503875383763651e-05, "loss": 1.0345, "step": 20725 }, { "epoch": 0.05, "learning_rate": 9.503749559615482e-05, "loss": 1.0351, "step": 20730 }, { "epoch": 0.05, "learning_rate": 9.503623735467312e-05, "loss": 1.0328, "step": 20735 }, { "epoch": 0.05, "learning_rate": 9.503497911319141e-05, "loss": 1.0338, "step": 20740 }, { "epoch": 0.05, "learning_rate": 9.50337208717097e-05, "loss": 1.0324, "step": 20745 }, { "epoch": 0.05, "learning_rate": 9.5032462630228e-05, "loss": 1.0342, "step": 20750 }, { "epoch": 0.05, "learning_rate": 9.50312043887463e-05, "loss": 1.0351, "step": 20755 }, { "epoch": 0.05, "learning_rate": 9.50299461472646e-05, "loss": 1.0327, "step": 20760 }, { "epoch": 0.05, "learning_rate": 9.502868790578289e-05, "loss": 1.0338, "step": 20765 }, { "epoch": 0.05, "learning_rate": 9.502742966430118e-05, "loss": 1.0344, "step": 20770 }, { "epoch": 0.05, "learning_rate": 9.502617142281948e-05, "loss": 1.0329, "step": 20775 }, { "epoch": 0.05, "learning_rate": 9.502491318133777e-05, "loss": 1.0325, "step": 20780 }, { "epoch": 0.05, "learning_rate": 9.502365493985607e-05, "loss": 1.0336, "step": 20785 }, { "epoch": 0.05, "learning_rate": 9.502239669837436e-05, "loss": 1.0357, "step": 20790 }, { "epoch": 0.05, "learning_rate": 9.502113845689265e-05, "loss": 1.0329, "step": 20795 }, { "epoch": 0.05, "learning_rate": 9.501988021541095e-05, "loss": 1.035, "step": 20800 }, { "epoch": 0.05, "learning_rate": 9.501862197392925e-05, "loss": 1.0334, "step": 20805 }, { "epoch": 0.05, "learning_rate": 9.501736373244754e-05, "loss": 1.0322, "step": 20810 }, { "epoch": 0.05, "learning_rate": 9.501610549096583e-05, "loss": 1.0337, "step": 20815 }, { "epoch": 0.05, "learning_rate": 9.501484724948413e-05, "loss": 1.0337, "step": 20820 }, { "epoch": 0.05, "learning_rate": 9.501358900800243e-05, "loss": 1.0337, "step": 20825 }, { "epoch": 0.05, "learning_rate": 9.501233076652072e-05, "loss": 1.0333, "step": 20830 }, { "epoch": 0.05, "learning_rate": 9.501107252503901e-05, "loss": 1.0338, "step": 20835 }, { "epoch": 0.05, "learning_rate": 9.500981428355731e-05, "loss": 1.0335, "step": 20840 }, { "epoch": 0.05, "learning_rate": 9.50085560420756e-05, "loss": 1.0341, "step": 20845 }, { "epoch": 0.05, "learning_rate": 9.50072978005939e-05, "loss": 1.0349, "step": 20850 }, { "epoch": 0.05, "learning_rate": 9.500603955911219e-05, "loss": 1.0343, "step": 20855 }, { "epoch": 0.05, "learning_rate": 9.500478131763048e-05, "loss": 1.0327, "step": 20860 }, { "epoch": 0.05, "learning_rate": 9.500352307614879e-05, "loss": 1.0318, "step": 20865 }, { "epoch": 0.05, "learning_rate": 9.500226483466708e-05, "loss": 1.0345, "step": 20870 }, { "epoch": 0.05, "learning_rate": 9.500100659318537e-05, "loss": 1.0361, "step": 20875 }, { "epoch": 0.05, "learning_rate": 9.499974835170366e-05, "loss": 1.0347, "step": 20880 }, { "epoch": 0.05, "learning_rate": 9.499849011022197e-05, "loss": 1.0345, "step": 20885 }, { "epoch": 0.05, "learning_rate": 9.499723186874026e-05, "loss": 1.0348, "step": 20890 }, { "epoch": 0.05, "learning_rate": 9.499597362725855e-05, "loss": 1.0339, "step": 20895 }, { "epoch": 0.05, "learning_rate": 9.499471538577684e-05, "loss": 1.0324, "step": 20900 }, { "epoch": 0.05, "learning_rate": 9.499345714429515e-05, "loss": 1.0334, "step": 20905 }, { "epoch": 0.05, "learning_rate": 9.499219890281344e-05, "loss": 1.0341, "step": 20910 }, { "epoch": 0.05, "learning_rate": 9.499094066133173e-05, "loss": 1.0347, "step": 20915 }, { "epoch": 0.05, "learning_rate": 9.498968241985002e-05, "loss": 1.036, "step": 20920 }, { "epoch": 0.05, "learning_rate": 9.498842417836831e-05, "loss": 1.0329, "step": 20925 }, { "epoch": 0.05, "learning_rate": 9.498716593688662e-05, "loss": 1.0335, "step": 20930 }, { "epoch": 0.05, "learning_rate": 9.498590769540491e-05, "loss": 1.0324, "step": 20935 }, { "epoch": 0.05, "learning_rate": 9.49846494539232e-05, "loss": 1.0344, "step": 20940 }, { "epoch": 0.05, "learning_rate": 9.498339121244149e-05, "loss": 1.0333, "step": 20945 }, { "epoch": 0.05, "learning_rate": 9.49821329709598e-05, "loss": 1.0329, "step": 20950 }, { "epoch": 0.05, "learning_rate": 9.498087472947809e-05, "loss": 1.0571, "step": 20955 }, { "epoch": 0.05, "learning_rate": 9.497961648799638e-05, "loss": 1.0332, "step": 20960 }, { "epoch": 0.05, "learning_rate": 9.497835824651467e-05, "loss": 1.0317, "step": 20965 }, { "epoch": 0.05, "learning_rate": 9.497710000503298e-05, "loss": 1.0322, "step": 20970 }, { "epoch": 0.05, "learning_rate": 9.497584176355127e-05, "loss": 1.034, "step": 20975 }, { "epoch": 0.05, "learning_rate": 9.497458352206956e-05, "loss": 1.0331, "step": 20980 }, { "epoch": 0.05, "learning_rate": 9.497332528058785e-05, "loss": 1.0317, "step": 20985 }, { "epoch": 0.05, "learning_rate": 9.497206703910614e-05, "loss": 1.0335, "step": 20990 }, { "epoch": 0.05, "learning_rate": 9.497080879762445e-05, "loss": 1.0341, "step": 20995 }, { "epoch": 0.05, "learning_rate": 9.496955055614274e-05, "loss": 1.0363, "step": 21000 }, { "epoch": 0.05, "learning_rate": 9.496829231466103e-05, "loss": 1.0321, "step": 21005 }, { "epoch": 0.05, "learning_rate": 9.496703407317932e-05, "loss": 1.0324, "step": 21010 }, { "epoch": 0.05, "learning_rate": 9.496577583169763e-05, "loss": 1.0342, "step": 21015 }, { "epoch": 0.05, "learning_rate": 9.496451759021592e-05, "loss": 1.0334, "step": 21020 }, { "epoch": 0.05, "learning_rate": 9.496325934873421e-05, "loss": 1.0335, "step": 21025 }, { "epoch": 0.05, "learning_rate": 9.49620011072525e-05, "loss": 1.0324, "step": 21030 }, { "epoch": 0.05, "learning_rate": 9.496074286577081e-05, "loss": 1.0338, "step": 21035 }, { "epoch": 0.05, "learning_rate": 9.49594846242891e-05, "loss": 1.0324, "step": 21040 }, { "epoch": 0.05, "learning_rate": 9.495822638280739e-05, "loss": 1.0346, "step": 21045 }, { "epoch": 0.05, "learning_rate": 9.495696814132568e-05, "loss": 1.0325, "step": 21050 }, { "epoch": 0.05, "learning_rate": 9.495570989984398e-05, "loss": 1.0347, "step": 21055 }, { "epoch": 0.05, "learning_rate": 9.495445165836228e-05, "loss": 1.0333, "step": 21060 }, { "epoch": 0.05, "learning_rate": 9.495319341688057e-05, "loss": 1.0352, "step": 21065 }, { "epoch": 0.05, "learning_rate": 9.495193517539886e-05, "loss": 1.0346, "step": 21070 }, { "epoch": 0.05, "learning_rate": 9.495067693391716e-05, "loss": 1.033, "step": 21075 }, { "epoch": 0.05, "learning_rate": 9.494941869243546e-05, "loss": 1.0361, "step": 21080 }, { "epoch": 0.05, "learning_rate": 9.494816045095375e-05, "loss": 1.0345, "step": 21085 }, { "epoch": 0.05, "learning_rate": 9.494690220947204e-05, "loss": 1.0337, "step": 21090 }, { "epoch": 0.05, "learning_rate": 9.494564396799034e-05, "loss": 1.0333, "step": 21095 }, { "epoch": 0.05, "learning_rate": 9.494438572650863e-05, "loss": 1.0325, "step": 21100 }, { "epoch": 0.05, "learning_rate": 9.494312748502693e-05, "loss": 1.0326, "step": 21105 }, { "epoch": 0.05, "learning_rate": 9.494186924354522e-05, "loss": 1.0329, "step": 21110 }, { "epoch": 0.05, "learning_rate": 9.494061100206351e-05, "loss": 1.0348, "step": 21115 }, { "epoch": 0.05, "learning_rate": 9.49393527605818e-05, "loss": 1.0364, "step": 21120 }, { "epoch": 0.05, "learning_rate": 9.493809451910011e-05, "loss": 1.0344, "step": 21125 }, { "epoch": 0.05, "learning_rate": 9.49368362776184e-05, "loss": 1.0346, "step": 21130 }, { "epoch": 0.05, "learning_rate": 9.49355780361367e-05, "loss": 1.0327, "step": 21135 }, { "epoch": 0.05, "learning_rate": 9.493431979465499e-05, "loss": 1.0326, "step": 21140 }, { "epoch": 0.05, "learning_rate": 9.493306155317329e-05, "loss": 1.0325, "step": 21145 }, { "epoch": 0.05, "learning_rate": 9.493180331169158e-05, "loss": 1.0354, "step": 21150 }, { "epoch": 0.05, "learning_rate": 9.493054507020987e-05, "loss": 1.0344, "step": 21155 }, { "epoch": 0.05, "learning_rate": 9.492928682872817e-05, "loss": 1.0352, "step": 21160 }, { "epoch": 0.05, "learning_rate": 9.492802858724646e-05, "loss": 1.0341, "step": 21165 }, { "epoch": 0.05, "learning_rate": 9.492677034576476e-05, "loss": 1.0346, "step": 21170 }, { "epoch": 0.05, "learning_rate": 9.492551210428305e-05, "loss": 1.0323, "step": 21175 }, { "epoch": 0.05, "learning_rate": 9.492425386280135e-05, "loss": 1.0322, "step": 21180 }, { "epoch": 0.05, "learning_rate": 9.492299562131964e-05, "loss": 1.0514, "step": 21185 }, { "epoch": 0.05, "learning_rate": 9.492173737983794e-05, "loss": 1.0363, "step": 21190 }, { "epoch": 0.05, "learning_rate": 9.492047913835623e-05, "loss": 1.0335, "step": 21195 }, { "epoch": 0.05, "learning_rate": 9.491922089687453e-05, "loss": 1.0311, "step": 21200 }, { "epoch": 0.05, "learning_rate": 9.491796265539282e-05, "loss": 1.0309, "step": 21205 }, { "epoch": 0.05, "learning_rate": 9.491670441391112e-05, "loss": 1.0326, "step": 21210 }, { "epoch": 0.05, "learning_rate": 9.491544617242941e-05, "loss": 1.0325, "step": 21215 }, { "epoch": 0.05, "learning_rate": 9.49141879309477e-05, "loss": 1.0329, "step": 21220 }, { "epoch": 0.05, "learning_rate": 9.4912929689466e-05, "loss": 1.0327, "step": 21225 }, { "epoch": 0.05, "learning_rate": 9.49116714479843e-05, "loss": 1.0334, "step": 21230 }, { "epoch": 0.05, "learning_rate": 9.491041320650261e-05, "loss": 1.0566, "step": 21235 }, { "epoch": 0.05, "learning_rate": 9.49091549650209e-05, "loss": 1.0323, "step": 21240 }, { "epoch": 0.05, "learning_rate": 9.490789672353919e-05, "loss": 1.0338, "step": 21245 }, { "epoch": 0.05, "learning_rate": 9.490663848205748e-05, "loss": 1.0334, "step": 21250 }, { "epoch": 0.05, "learning_rate": 9.490538024057577e-05, "loss": 1.0338, "step": 21255 }, { "epoch": 0.05, "learning_rate": 9.490412199909408e-05, "loss": 1.0351, "step": 21260 }, { "epoch": 0.05, "learning_rate": 9.490286375761237e-05, "loss": 1.0352, "step": 21265 }, { "epoch": 0.05, "learning_rate": 9.490160551613066e-05, "loss": 1.0333, "step": 21270 }, { "epoch": 0.05, "learning_rate": 9.490034727464895e-05, "loss": 1.0313, "step": 21275 }, { "epoch": 0.05, "learning_rate": 9.489908903316726e-05, "loss": 1.0349, "step": 21280 }, { "epoch": 0.05, "learning_rate": 9.489783079168555e-05, "loss": 1.0354, "step": 21285 }, { "epoch": 0.05, "learning_rate": 9.489657255020384e-05, "loss": 1.0333, "step": 21290 }, { "epoch": 0.05, "learning_rate": 9.489531430872213e-05, "loss": 1.0351, "step": 21295 }, { "epoch": 0.05, "learning_rate": 9.489405606724044e-05, "loss": 1.0326, "step": 21300 }, { "epoch": 0.05, "learning_rate": 9.489279782575873e-05, "loss": 1.035, "step": 21305 }, { "epoch": 0.05, "learning_rate": 9.489153958427702e-05, "loss": 1.0336, "step": 21310 }, { "epoch": 0.05, "learning_rate": 9.489028134279531e-05, "loss": 1.0312, "step": 21315 }, { "epoch": 0.05, "learning_rate": 9.48890231013136e-05, "loss": 1.0325, "step": 21320 }, { "epoch": 0.05, "learning_rate": 9.488776485983191e-05, "loss": 1.0351, "step": 21325 }, { "epoch": 0.05, "learning_rate": 9.48865066183502e-05, "loss": 1.0315, "step": 21330 }, { "epoch": 0.05, "learning_rate": 9.48852483768685e-05, "loss": 1.0326, "step": 21335 }, { "epoch": 0.05, "learning_rate": 9.488399013538679e-05, "loss": 1.0339, "step": 21340 }, { "epoch": 0.05, "learning_rate": 9.488273189390509e-05, "loss": 1.0594, "step": 21345 }, { "epoch": 0.05, "learning_rate": 9.488147365242338e-05, "loss": 1.0343, "step": 21350 }, { "epoch": 0.05, "learning_rate": 9.488021541094167e-05, "loss": 1.0343, "step": 21355 }, { "epoch": 0.05, "learning_rate": 9.487895716945997e-05, "loss": 1.0566, "step": 21360 }, { "epoch": 0.05, "learning_rate": 9.487769892797827e-05, "loss": 1.0323, "step": 21365 }, { "epoch": 0.05, "learning_rate": 9.487644068649656e-05, "loss": 1.0335, "step": 21370 }, { "epoch": 0.05, "learning_rate": 9.487518244501485e-05, "loss": 1.0331, "step": 21375 }, { "epoch": 0.05, "learning_rate": 9.487392420353315e-05, "loss": 1.0349, "step": 21380 }, { "epoch": 0.05, "learning_rate": 9.487266596205144e-05, "loss": 1.0318, "step": 21385 }, { "epoch": 0.05, "learning_rate": 9.487140772056974e-05, "loss": 1.034, "step": 21390 }, { "epoch": 0.05, "learning_rate": 9.487014947908803e-05, "loss": 1.0311, "step": 21395 }, { "epoch": 0.05, "learning_rate": 9.486889123760633e-05, "loss": 1.0344, "step": 21400 }, { "epoch": 0.05, "learning_rate": 9.486763299612462e-05, "loss": 1.0321, "step": 21405 }, { "epoch": 0.05, "learning_rate": 9.486637475464292e-05, "loss": 1.0335, "step": 21410 }, { "epoch": 0.05, "learning_rate": 9.486511651316121e-05, "loss": 1.0351, "step": 21415 }, { "epoch": 0.05, "learning_rate": 9.48638582716795e-05, "loss": 1.0335, "step": 21420 }, { "epoch": 0.05, "learning_rate": 9.48626000301978e-05, "loss": 1.0332, "step": 21425 }, { "epoch": 0.05, "learning_rate": 9.48613417887161e-05, "loss": 1.0344, "step": 21430 }, { "epoch": 0.05, "learning_rate": 9.48600835472344e-05, "loss": 1.0313, "step": 21435 }, { "epoch": 0.05, "learning_rate": 9.485882530575269e-05, "loss": 1.0343, "step": 21440 }, { "epoch": 0.05, "learning_rate": 9.485756706427098e-05, "loss": 1.0318, "step": 21445 }, { "epoch": 0.05, "learning_rate": 9.485630882278927e-05, "loss": 1.0334, "step": 21450 }, { "epoch": 0.05, "learning_rate": 9.485505058130757e-05, "loss": 1.0334, "step": 21455 }, { "epoch": 0.05, "learning_rate": 9.485379233982587e-05, "loss": 1.0319, "step": 21460 }, { "epoch": 0.05, "learning_rate": 9.485253409834416e-05, "loss": 1.0314, "step": 21465 }, { "epoch": 0.05, "learning_rate": 9.485127585686245e-05, "loss": 1.0338, "step": 21470 }, { "epoch": 0.05, "learning_rate": 9.485001761538075e-05, "loss": 1.0336, "step": 21475 }, { "epoch": 0.05, "learning_rate": 9.484875937389905e-05, "loss": 1.0337, "step": 21480 }, { "epoch": 0.05, "learning_rate": 9.484750113241734e-05, "loss": 1.0341, "step": 21485 }, { "epoch": 0.05, "learning_rate": 9.484624289093563e-05, "loss": 1.0317, "step": 21490 }, { "epoch": 0.05, "learning_rate": 9.484498464945393e-05, "loss": 1.0362, "step": 21495 }, { "epoch": 0.05, "learning_rate": 9.484372640797223e-05, "loss": 1.0336, "step": 21500 }, { "epoch": 0.05, "learning_rate": 9.484246816649052e-05, "loss": 1.0326, "step": 21505 }, { "epoch": 0.05, "learning_rate": 9.484120992500881e-05, "loss": 1.0328, "step": 21510 }, { "epoch": 0.05, "learning_rate": 9.48399516835271e-05, "loss": 1.0324, "step": 21515 }, { "epoch": 0.05, "learning_rate": 9.48386934420454e-05, "loss": 1.0358, "step": 21520 }, { "epoch": 0.05, "learning_rate": 9.48374352005637e-05, "loss": 1.034, "step": 21525 }, { "epoch": 0.05, "learning_rate": 9.483617695908199e-05, "loss": 1.0332, "step": 21530 }, { "epoch": 0.05, "learning_rate": 9.483491871760028e-05, "loss": 1.0353, "step": 21535 }, { "epoch": 0.05, "learning_rate": 9.483366047611859e-05, "loss": 1.0341, "step": 21540 }, { "epoch": 0.05, "learning_rate": 9.483240223463688e-05, "loss": 1.0325, "step": 21545 }, { "epoch": 0.05, "learning_rate": 9.483114399315517e-05, "loss": 1.0357, "step": 21550 }, { "epoch": 0.05, "learning_rate": 9.482988575167346e-05, "loss": 1.0329, "step": 21555 }, { "epoch": 0.05, "learning_rate": 9.482862751019177e-05, "loss": 1.0318, "step": 21560 }, { "epoch": 0.05, "learning_rate": 9.482736926871006e-05, "loss": 1.034, "step": 21565 }, { "epoch": 0.05, "learning_rate": 9.482611102722835e-05, "loss": 1.0326, "step": 21570 }, { "epoch": 0.05, "learning_rate": 9.482485278574664e-05, "loss": 1.033, "step": 21575 }, { "epoch": 0.05, "learning_rate": 9.482359454426493e-05, "loss": 1.0336, "step": 21580 }, { "epoch": 0.05, "learning_rate": 9.482233630278324e-05, "loss": 1.0342, "step": 21585 }, { "epoch": 0.05, "learning_rate": 9.482107806130153e-05, "loss": 1.0326, "step": 21590 }, { "epoch": 0.05, "learning_rate": 9.481981981981982e-05, "loss": 1.0334, "step": 21595 }, { "epoch": 0.05, "learning_rate": 9.481856157833811e-05, "loss": 1.0506, "step": 21600 }, { "epoch": 0.05, "learning_rate": 9.481730333685642e-05, "loss": 1.0325, "step": 21605 }, { "epoch": 0.05, "learning_rate": 9.481604509537471e-05, "loss": 1.0304, "step": 21610 }, { "epoch": 0.05, "learning_rate": 9.4814786853893e-05, "loss": 1.0331, "step": 21615 }, { "epoch": 0.05, "learning_rate": 9.481352861241129e-05, "loss": 1.0318, "step": 21620 }, { "epoch": 0.05, "learning_rate": 9.48122703709296e-05, "loss": 1.0349, "step": 21625 }, { "epoch": 0.05, "learning_rate": 9.481101212944789e-05, "loss": 1.0356, "step": 21630 }, { "epoch": 0.05, "learning_rate": 9.480975388796618e-05, "loss": 1.0335, "step": 21635 }, { "epoch": 0.05, "learning_rate": 9.480849564648447e-05, "loss": 1.0335, "step": 21640 }, { "epoch": 0.05, "learning_rate": 9.480723740500276e-05, "loss": 1.0313, "step": 21645 }, { "epoch": 0.05, "learning_rate": 9.480597916352107e-05, "loss": 1.0323, "step": 21650 }, { "epoch": 0.05, "learning_rate": 9.480472092203936e-05, "loss": 1.0322, "step": 21655 }, { "epoch": 0.05, "learning_rate": 9.480346268055765e-05, "loss": 1.0324, "step": 21660 }, { "epoch": 0.05, "learning_rate": 9.480220443907594e-05, "loss": 1.057, "step": 21665 }, { "epoch": 0.05, "learning_rate": 9.480094619759425e-05, "loss": 1.0351, "step": 21670 }, { "epoch": 0.05, "learning_rate": 9.479968795611254e-05, "loss": 1.0341, "step": 21675 }, { "epoch": 0.05, "learning_rate": 9.479842971463083e-05, "loss": 1.0342, "step": 21680 }, { "epoch": 0.05, "learning_rate": 9.479717147314912e-05, "loss": 1.032, "step": 21685 }, { "epoch": 0.05, "learning_rate": 9.479591323166743e-05, "loss": 1.0325, "step": 21690 }, { "epoch": 0.05, "learning_rate": 9.479465499018572e-05, "loss": 1.0357, "step": 21695 }, { "epoch": 0.05, "learning_rate": 9.479339674870401e-05, "loss": 1.0331, "step": 21700 }, { "epoch": 0.05, "learning_rate": 9.47921385072223e-05, "loss": 1.0341, "step": 21705 }, { "epoch": 0.05, "learning_rate": 9.47908802657406e-05, "loss": 1.0348, "step": 21710 }, { "epoch": 0.05, "learning_rate": 9.47896220242589e-05, "loss": 1.0327, "step": 21715 }, { "epoch": 0.05, "learning_rate": 9.478836378277719e-05, "loss": 1.034, "step": 21720 }, { "epoch": 0.05, "learning_rate": 9.478710554129548e-05, "loss": 1.0336, "step": 21725 }, { "epoch": 0.05, "learning_rate": 9.478584729981379e-05, "loss": 1.0341, "step": 21730 }, { "epoch": 0.05, "learning_rate": 9.478458905833208e-05, "loss": 1.0332, "step": 21735 }, { "epoch": 0.05, "learning_rate": 9.478333081685038e-05, "loss": 1.034, "step": 21740 }, { "epoch": 0.05, "learning_rate": 9.478207257536868e-05, "loss": 1.0318, "step": 21745 }, { "epoch": 0.05, "learning_rate": 9.478081433388697e-05, "loss": 1.0356, "step": 21750 }, { "epoch": 0.05, "learning_rate": 9.477955609240526e-05, "loss": 1.0345, "step": 21755 }, { "epoch": 0.05, "learning_rate": 9.477829785092356e-05, "loss": 1.0341, "step": 21760 }, { "epoch": 0.05, "learning_rate": 9.477703960944186e-05, "loss": 1.034, "step": 21765 }, { "epoch": 0.05, "learning_rate": 9.477578136796015e-05, "loss": 1.0349, "step": 21770 }, { "epoch": 0.05, "learning_rate": 9.477452312647844e-05, "loss": 1.0343, "step": 21775 }, { "epoch": 0.05, "learning_rate": 9.477326488499673e-05, "loss": 1.0343, "step": 21780 }, { "epoch": 0.05, "learning_rate": 9.477200664351504e-05, "loss": 1.0332, "step": 21785 }, { "epoch": 0.05, "learning_rate": 9.477074840203333e-05, "loss": 1.032, "step": 21790 }, { "epoch": 0.05, "learning_rate": 9.476949016055162e-05, "loss": 1.0324, "step": 21795 }, { "epoch": 0.05, "learning_rate": 9.476823191906991e-05, "loss": 1.0331, "step": 21800 }, { "epoch": 0.05, "learning_rate": 9.476697367758822e-05, "loss": 1.0306, "step": 21805 }, { "epoch": 0.05, "learning_rate": 9.476571543610651e-05, "loss": 1.0301, "step": 21810 }, { "epoch": 0.05, "learning_rate": 9.47644571946248e-05, "loss": 1.0326, "step": 21815 }, { "epoch": 0.05, "learning_rate": 9.476319895314309e-05, "loss": 1.0329, "step": 21820 }, { "epoch": 0.05, "learning_rate": 9.47619407116614e-05, "loss": 1.0318, "step": 21825 }, { "epoch": 0.05, "learning_rate": 9.476068247017969e-05, "loss": 1.0327, "step": 21830 }, { "epoch": 0.05, "learning_rate": 9.475942422869798e-05, "loss": 1.0328, "step": 21835 }, { "epoch": 0.05, "learning_rate": 9.475816598721627e-05, "loss": 1.0339, "step": 21840 }, { "epoch": 0.05, "learning_rate": 9.475690774573456e-05, "loss": 1.0325, "step": 21845 }, { "epoch": 0.05, "learning_rate": 9.475564950425287e-05, "loss": 1.0323, "step": 21850 }, { "epoch": 0.05, "learning_rate": 9.475439126277116e-05, "loss": 1.0336, "step": 21855 }, { "epoch": 0.05, "learning_rate": 9.475313302128945e-05, "loss": 1.0328, "step": 21860 }, { "epoch": 0.05, "learning_rate": 9.475187477980774e-05, "loss": 1.0342, "step": 21865 }, { "epoch": 0.05, "learning_rate": 9.475061653832605e-05, "loss": 1.0337, "step": 21870 }, { "epoch": 0.05, "learning_rate": 9.474935829684434e-05, "loss": 1.0314, "step": 21875 }, { "epoch": 0.05, "learning_rate": 9.474810005536263e-05, "loss": 1.034, "step": 21880 }, { "epoch": 0.05, "learning_rate": 9.474684181388092e-05, "loss": 1.0305, "step": 21885 }, { "epoch": 0.05, "learning_rate": 9.474558357239923e-05, "loss": 1.0326, "step": 21890 }, { "epoch": 0.05, "learning_rate": 9.474432533091752e-05, "loss": 1.0341, "step": 21895 }, { "epoch": 0.05, "learning_rate": 9.474306708943581e-05, "loss": 1.0329, "step": 21900 }, { "epoch": 0.05, "learning_rate": 9.47418088479541e-05, "loss": 1.0332, "step": 21905 }, { "epoch": 0.05, "learning_rate": 9.47405506064724e-05, "loss": 1.0348, "step": 21910 }, { "epoch": 0.06, "learning_rate": 9.47392923649907e-05, "loss": 1.0347, "step": 21915 }, { "epoch": 0.06, "learning_rate": 9.473803412350899e-05, "loss": 1.0304, "step": 21920 }, { "epoch": 0.06, "learning_rate": 9.473677588202728e-05, "loss": 1.0341, "step": 21925 }, { "epoch": 0.06, "learning_rate": 9.473551764054557e-05, "loss": 1.0342, "step": 21930 }, { "epoch": 0.06, "learning_rate": 9.473425939906388e-05, "loss": 1.0303, "step": 21935 }, { "epoch": 0.06, "learning_rate": 9.473300115758217e-05, "loss": 1.0336, "step": 21940 }, { "epoch": 0.06, "learning_rate": 9.473174291610046e-05, "loss": 1.0331, "step": 21945 }, { "epoch": 0.06, "learning_rate": 9.473048467461875e-05, "loss": 1.0325, "step": 21950 }, { "epoch": 0.06, "learning_rate": 9.472922643313706e-05, "loss": 1.0323, "step": 21955 }, { "epoch": 0.06, "learning_rate": 9.472796819165535e-05, "loss": 1.0343, "step": 21960 }, { "epoch": 0.06, "learning_rate": 9.472670995017364e-05, "loss": 1.0342, "step": 21965 }, { "epoch": 0.06, "learning_rate": 9.472545170869193e-05, "loss": 1.0304, "step": 21970 }, { "epoch": 0.06, "learning_rate": 9.472419346721023e-05, "loss": 1.0319, "step": 21975 }, { "epoch": 0.06, "learning_rate": 9.472293522572853e-05, "loss": 1.0308, "step": 21980 }, { "epoch": 0.06, "learning_rate": 9.472167698424682e-05, "loss": 1.0334, "step": 21985 }, { "epoch": 0.06, "learning_rate": 9.472041874276511e-05, "loss": 1.0327, "step": 21990 }, { "epoch": 0.06, "learning_rate": 9.47191605012834e-05, "loss": 1.0347, "step": 21995 }, { "epoch": 0.06, "learning_rate": 9.471790225980171e-05, "loss": 1.0347, "step": 22000 }, { "epoch": 0.06, "learning_rate": 9.471664401832e-05, "loss": 1.0366, "step": 22005 }, { "epoch": 0.06, "learning_rate": 9.47153857768383e-05, "loss": 1.0345, "step": 22010 }, { "epoch": 0.06, "learning_rate": 9.471412753535659e-05, "loss": 1.032, "step": 22015 }, { "epoch": 0.06, "learning_rate": 9.471286929387489e-05, "loss": 1.0336, "step": 22020 }, { "epoch": 0.06, "learning_rate": 9.471161105239318e-05, "loss": 1.0331, "step": 22025 }, { "epoch": 0.06, "learning_rate": 9.471035281091147e-05, "loss": 1.0342, "step": 22030 }, { "epoch": 0.06, "learning_rate": 9.470909456942977e-05, "loss": 1.0336, "step": 22035 }, { "epoch": 0.06, "learning_rate": 9.470783632794806e-05, "loss": 1.0337, "step": 22040 }, { "epoch": 0.06, "learning_rate": 9.470657808646636e-05, "loss": 1.0333, "step": 22045 }, { "epoch": 0.06, "learning_rate": 9.470531984498465e-05, "loss": 1.0321, "step": 22050 }, { "epoch": 0.06, "learning_rate": 9.470406160350295e-05, "loss": 1.032, "step": 22055 }, { "epoch": 0.06, "learning_rate": 9.470280336202124e-05, "loss": 1.0338, "step": 22060 }, { "epoch": 0.06, "learning_rate": 9.470154512053954e-05, "loss": 1.0338, "step": 22065 }, { "epoch": 0.06, "learning_rate": 9.470028687905783e-05, "loss": 1.0326, "step": 22070 }, { "epoch": 0.06, "learning_rate": 9.469902863757613e-05, "loss": 1.0348, "step": 22075 }, { "epoch": 0.06, "learning_rate": 9.469777039609442e-05, "loss": 1.0323, "step": 22080 }, { "epoch": 0.06, "learning_rate": 9.469651215461272e-05, "loss": 1.0336, "step": 22085 }, { "epoch": 0.06, "learning_rate": 9.469525391313101e-05, "loss": 1.0349, "step": 22090 }, { "epoch": 0.06, "learning_rate": 9.46939956716493e-05, "loss": 1.0326, "step": 22095 }, { "epoch": 0.06, "learning_rate": 9.46927374301676e-05, "loss": 1.0342, "step": 22100 }, { "epoch": 0.06, "learning_rate": 9.469147918868589e-05, "loss": 1.0366, "step": 22105 }, { "epoch": 0.06, "learning_rate": 9.46902209472042e-05, "loss": 1.0335, "step": 22110 }, { "epoch": 0.06, "learning_rate": 9.468896270572249e-05, "loss": 1.0352, "step": 22115 }, { "epoch": 0.06, "learning_rate": 9.468770446424078e-05, "loss": 1.0335, "step": 22120 }, { "epoch": 0.06, "learning_rate": 9.468644622275907e-05, "loss": 1.0331, "step": 22125 }, { "epoch": 0.06, "learning_rate": 9.468518798127737e-05, "loss": 1.0339, "step": 22130 }, { "epoch": 0.06, "learning_rate": 9.468392973979566e-05, "loss": 1.0334, "step": 22135 }, { "epoch": 0.06, "learning_rate": 9.468267149831396e-05, "loss": 1.0322, "step": 22140 }, { "epoch": 0.06, "learning_rate": 9.468141325683225e-05, "loss": 1.0322, "step": 22145 }, { "epoch": 0.06, "learning_rate": 9.468015501535055e-05, "loss": 1.0327, "step": 22150 }, { "epoch": 0.06, "learning_rate": 9.467889677386884e-05, "loss": 1.0325, "step": 22155 }, { "epoch": 0.06, "learning_rate": 9.467763853238714e-05, "loss": 1.0329, "step": 22160 }, { "epoch": 0.06, "learning_rate": 9.467638029090543e-05, "loss": 1.0355, "step": 22165 }, { "epoch": 0.06, "learning_rate": 9.467512204942372e-05, "loss": 1.0343, "step": 22170 }, { "epoch": 0.06, "learning_rate": 9.467386380794202e-05, "loss": 1.0356, "step": 22175 }, { "epoch": 0.06, "learning_rate": 9.467260556646032e-05, "loss": 1.0321, "step": 22180 }, { "epoch": 0.06, "learning_rate": 9.467134732497861e-05, "loss": 1.0313, "step": 22185 }, { "epoch": 0.06, "learning_rate": 9.46700890834969e-05, "loss": 1.0352, "step": 22190 }, { "epoch": 0.06, "learning_rate": 9.46688308420152e-05, "loss": 1.031, "step": 22195 }, { "epoch": 0.06, "learning_rate": 9.46675726005335e-05, "loss": 1.0325, "step": 22200 }, { "epoch": 0.06, "learning_rate": 9.466631435905179e-05, "loss": 1.0336, "step": 22205 }, { "epoch": 0.06, "learning_rate": 9.466505611757008e-05, "loss": 1.0322, "step": 22210 }, { "epoch": 0.06, "learning_rate": 9.466379787608838e-05, "loss": 1.0356, "step": 22215 }, { "epoch": 0.06, "learning_rate": 9.466253963460668e-05, "loss": 1.0338, "step": 22220 }, { "epoch": 0.06, "learning_rate": 9.466128139312497e-05, "loss": 1.0557, "step": 22225 }, { "epoch": 0.06, "learning_rate": 9.466002315164327e-05, "loss": 1.0337, "step": 22230 }, { "epoch": 0.06, "learning_rate": 9.465876491016156e-05, "loss": 1.0309, "step": 22235 }, { "epoch": 0.06, "learning_rate": 9.465750666867986e-05, "loss": 1.0326, "step": 22240 }, { "epoch": 0.06, "learning_rate": 9.465624842719816e-05, "loss": 1.0321, "step": 22245 }, { "epoch": 0.06, "learning_rate": 9.465499018571645e-05, "loss": 1.0295, "step": 22250 }, { "epoch": 0.06, "learning_rate": 9.465373194423474e-05, "loss": 1.0318, "step": 22255 }, { "epoch": 0.06, "learning_rate": 9.465247370275304e-05, "loss": 1.0536, "step": 22260 }, { "epoch": 0.06, "learning_rate": 9.465121546127134e-05, "loss": 1.0337, "step": 22265 }, { "epoch": 0.06, "learning_rate": 9.464995721978963e-05, "loss": 1.0336, "step": 22270 }, { "epoch": 0.06, "learning_rate": 9.464869897830792e-05, "loss": 1.0323, "step": 22275 }, { "epoch": 0.06, "learning_rate": 9.464744073682622e-05, "loss": 1.0327, "step": 22280 }, { "epoch": 0.06, "learning_rate": 9.464618249534452e-05, "loss": 1.0328, "step": 22285 }, { "epoch": 0.06, "learning_rate": 9.464492425386281e-05, "loss": 1.0319, "step": 22290 }, { "epoch": 0.06, "learning_rate": 9.46436660123811e-05, "loss": 1.0316, "step": 22295 }, { "epoch": 0.06, "learning_rate": 9.46424077708994e-05, "loss": 1.0335, "step": 22300 }, { "epoch": 0.06, "learning_rate": 9.464114952941769e-05, "loss": 1.0342, "step": 22305 }, { "epoch": 0.06, "learning_rate": 9.463989128793599e-05, "loss": 1.0343, "step": 22310 }, { "epoch": 0.06, "learning_rate": 9.463863304645428e-05, "loss": 1.0321, "step": 22315 }, { "epoch": 0.06, "learning_rate": 9.463737480497258e-05, "loss": 1.0346, "step": 22320 }, { "epoch": 0.06, "learning_rate": 9.463611656349087e-05, "loss": 1.0336, "step": 22325 }, { "epoch": 0.06, "learning_rate": 9.463485832200917e-05, "loss": 1.0324, "step": 22330 }, { "epoch": 0.06, "learning_rate": 9.463360008052746e-05, "loss": 1.0341, "step": 22335 }, { "epoch": 0.06, "learning_rate": 9.463234183904576e-05, "loss": 1.0298, "step": 22340 }, { "epoch": 0.06, "learning_rate": 9.463108359756405e-05, "loss": 1.0336, "step": 22345 }, { "epoch": 0.06, "learning_rate": 9.462982535608235e-05, "loss": 1.0318, "step": 22350 }, { "epoch": 0.06, "learning_rate": 9.462856711460064e-05, "loss": 1.032, "step": 22355 }, { "epoch": 0.06, "learning_rate": 9.462730887311894e-05, "loss": 1.0339, "step": 22360 }, { "epoch": 0.06, "learning_rate": 9.462605063163723e-05, "loss": 1.0338, "step": 22365 }, { "epoch": 0.06, "learning_rate": 9.462479239015552e-05, "loss": 1.034, "step": 22370 }, { "epoch": 0.06, "learning_rate": 9.462353414867382e-05, "loss": 1.0331, "step": 22375 }, { "epoch": 0.06, "learning_rate": 9.462227590719212e-05, "loss": 1.0326, "step": 22380 }, { "epoch": 0.06, "learning_rate": 9.462101766571041e-05, "loss": 1.0328, "step": 22385 }, { "epoch": 0.06, "learning_rate": 9.46197594242287e-05, "loss": 1.0334, "step": 22390 }, { "epoch": 0.06, "learning_rate": 9.4618501182747e-05, "loss": 1.0557, "step": 22395 }, { "epoch": 0.06, "learning_rate": 9.46172429412653e-05, "loss": 1.0344, "step": 22400 }, { "epoch": 0.06, "learning_rate": 9.461598469978359e-05, "loss": 1.0342, "step": 22405 }, { "epoch": 0.06, "learning_rate": 9.461472645830188e-05, "loss": 1.034, "step": 22410 }, { "epoch": 0.06, "learning_rate": 9.461346821682018e-05, "loss": 1.0319, "step": 22415 }, { "epoch": 0.06, "learning_rate": 9.461220997533848e-05, "loss": 1.0326, "step": 22420 }, { "epoch": 0.06, "learning_rate": 9.461095173385677e-05, "loss": 1.0505, "step": 22425 }, { "epoch": 0.06, "learning_rate": 9.460969349237506e-05, "loss": 1.0334, "step": 22430 }, { "epoch": 0.06, "learning_rate": 9.460843525089335e-05, "loss": 1.0337, "step": 22435 }, { "epoch": 0.06, "learning_rate": 9.460717700941166e-05, "loss": 1.0311, "step": 22440 }, { "epoch": 0.06, "learning_rate": 9.460591876792995e-05, "loss": 1.0344, "step": 22445 }, { "epoch": 0.06, "learning_rate": 9.460466052644824e-05, "loss": 1.0334, "step": 22450 }, { "epoch": 0.06, "learning_rate": 9.460340228496653e-05, "loss": 1.0351, "step": 22455 }, { "epoch": 0.06, "learning_rate": 9.460214404348484e-05, "loss": 1.0317, "step": 22460 }, { "epoch": 0.06, "learning_rate": 9.460088580200313e-05, "loss": 1.034, "step": 22465 }, { "epoch": 0.06, "learning_rate": 9.459962756052142e-05, "loss": 1.0327, "step": 22470 }, { "epoch": 0.06, "learning_rate": 9.459836931903971e-05, "loss": 1.0343, "step": 22475 }, { "epoch": 0.06, "learning_rate": 9.459711107755802e-05, "loss": 1.0328, "step": 22480 }, { "epoch": 0.06, "learning_rate": 9.459585283607631e-05, "loss": 1.0341, "step": 22485 }, { "epoch": 0.06, "learning_rate": 9.45945945945946e-05, "loss": 1.0356, "step": 22490 }, { "epoch": 0.06, "learning_rate": 9.459333635311289e-05, "loss": 1.035, "step": 22495 }, { "epoch": 0.06, "learning_rate": 9.459207811163118e-05, "loss": 1.0345, "step": 22500 }, { "epoch": 0.06, "learning_rate": 9.459081987014949e-05, "loss": 1.0336, "step": 22505 }, { "epoch": 0.06, "learning_rate": 9.458956162866778e-05, "loss": 1.0316, "step": 22510 }, { "epoch": 0.06, "learning_rate": 9.458830338718607e-05, "loss": 1.0315, "step": 22515 }, { "epoch": 0.06, "learning_rate": 9.458704514570436e-05, "loss": 1.0318, "step": 22520 }, { "epoch": 0.06, "learning_rate": 9.458578690422267e-05, "loss": 1.0523, "step": 22525 }, { "epoch": 0.06, "learning_rate": 9.458452866274096e-05, "loss": 1.034, "step": 22530 }, { "epoch": 0.06, "learning_rate": 9.458327042125925e-05, "loss": 1.0322, "step": 22535 }, { "epoch": 0.06, "learning_rate": 9.458201217977754e-05, "loss": 1.0312, "step": 22540 }, { "epoch": 0.06, "learning_rate": 9.458075393829585e-05, "loss": 1.0324, "step": 22545 }, { "epoch": 0.06, "learning_rate": 9.457949569681414e-05, "loss": 1.0343, "step": 22550 }, { "epoch": 0.06, "learning_rate": 9.457823745533243e-05, "loss": 1.0308, "step": 22555 }, { "epoch": 0.06, "learning_rate": 9.457697921385072e-05, "loss": 1.0338, "step": 22560 }, { "epoch": 0.06, "learning_rate": 9.457572097236901e-05, "loss": 1.036, "step": 22565 }, { "epoch": 0.06, "learning_rate": 9.457446273088732e-05, "loss": 1.033, "step": 22570 }, { "epoch": 0.06, "learning_rate": 9.457320448940561e-05, "loss": 1.035, "step": 22575 }, { "epoch": 0.06, "learning_rate": 9.45719462479239e-05, "loss": 1.0328, "step": 22580 }, { "epoch": 0.06, "learning_rate": 9.45706880064422e-05, "loss": 1.0342, "step": 22585 }, { "epoch": 0.06, "learning_rate": 9.45694297649605e-05, "loss": 1.0335, "step": 22590 }, { "epoch": 0.06, "learning_rate": 9.456817152347879e-05, "loss": 1.034, "step": 22595 }, { "epoch": 0.06, "learning_rate": 9.456691328199708e-05, "loss": 1.031, "step": 22600 }, { "epoch": 0.06, "learning_rate": 9.456565504051537e-05, "loss": 1.0331, "step": 22605 }, { "epoch": 0.06, "learning_rate": 9.456439679903368e-05, "loss": 1.0344, "step": 22610 }, { "epoch": 0.06, "learning_rate": 9.456313855755197e-05, "loss": 1.0326, "step": 22615 }, { "epoch": 0.06, "learning_rate": 9.456188031607026e-05, "loss": 1.0325, "step": 22620 }, { "epoch": 0.06, "learning_rate": 9.456062207458855e-05, "loss": 1.0327, "step": 22625 }, { "epoch": 0.06, "learning_rate": 9.455936383310684e-05, "loss": 1.0333, "step": 22630 }, { "epoch": 0.06, "learning_rate": 9.455810559162515e-05, "loss": 1.0341, "step": 22635 }, { "epoch": 0.06, "learning_rate": 9.455684735014344e-05, "loss": 1.034, "step": 22640 }, { "epoch": 0.06, "learning_rate": 9.455558910866173e-05, "loss": 1.0345, "step": 22645 }, { "epoch": 0.06, "learning_rate": 9.455433086718002e-05, "loss": 1.0332, "step": 22650 }, { "epoch": 0.06, "learning_rate": 9.455307262569833e-05, "loss": 1.0346, "step": 22655 }, { "epoch": 0.06, "learning_rate": 9.455181438421662e-05, "loss": 1.034, "step": 22660 }, { "epoch": 0.06, "learning_rate": 9.455055614273491e-05, "loss": 1.0317, "step": 22665 }, { "epoch": 0.06, "learning_rate": 9.45492979012532e-05, "loss": 1.0342, "step": 22670 }, { "epoch": 0.06, "learning_rate": 9.454803965977151e-05, "loss": 1.033, "step": 22675 }, { "epoch": 0.06, "learning_rate": 9.45467814182898e-05, "loss": 1.031, "step": 22680 }, { "epoch": 0.06, "learning_rate": 9.454552317680809e-05, "loss": 1.0335, "step": 22685 }, { "epoch": 0.06, "learning_rate": 9.454426493532638e-05, "loss": 1.033, "step": 22690 }, { "epoch": 0.06, "learning_rate": 9.454300669384468e-05, "loss": 1.0319, "step": 22695 }, { "epoch": 0.06, "learning_rate": 9.454174845236298e-05, "loss": 1.0338, "step": 22700 }, { "epoch": 0.06, "learning_rate": 9.454049021088127e-05, "loss": 1.0317, "step": 22705 }, { "epoch": 0.06, "learning_rate": 9.453923196939956e-05, "loss": 1.0315, "step": 22710 }, { "epoch": 0.06, "learning_rate": 9.453797372791786e-05, "loss": 1.0577, "step": 22715 }, { "epoch": 0.06, "learning_rate": 9.453671548643616e-05, "loss": 1.0344, "step": 22720 }, { "epoch": 0.06, "learning_rate": 9.453545724495445e-05, "loss": 1.0312, "step": 22725 }, { "epoch": 0.06, "learning_rate": 9.453419900347276e-05, "loss": 1.0497, "step": 22730 }, { "epoch": 0.06, "learning_rate": 9.453294076199105e-05, "loss": 1.0342, "step": 22735 }, { "epoch": 0.06, "learning_rate": 9.453168252050934e-05, "loss": 1.0317, "step": 22740 }, { "epoch": 0.06, "learning_rate": 9.453042427902765e-05, "loss": 1.0303, "step": 22745 }, { "epoch": 0.06, "learning_rate": 9.452916603754594e-05, "loss": 1.0319, "step": 22750 }, { "epoch": 0.06, "learning_rate": 9.452790779606423e-05, "loss": 1.0326, "step": 22755 }, { "epoch": 0.06, "learning_rate": 9.452664955458252e-05, "loss": 1.0333, "step": 22760 }, { "epoch": 0.06, "learning_rate": 9.452539131310081e-05, "loss": 1.0351, "step": 22765 }, { "epoch": 0.06, "learning_rate": 9.452413307161912e-05, "loss": 1.0355, "step": 22770 }, { "epoch": 0.06, "learning_rate": 9.452287483013741e-05, "loss": 1.0336, "step": 22775 }, { "epoch": 0.06, "learning_rate": 9.45216165886557e-05, "loss": 1.0313, "step": 22780 }, { "epoch": 0.06, "learning_rate": 9.452035834717399e-05, "loss": 1.0301, "step": 22785 }, { "epoch": 0.06, "learning_rate": 9.45191001056923e-05, "loss": 1.0334, "step": 22790 }, { "epoch": 0.06, "learning_rate": 9.451784186421059e-05, "loss": 1.0352, "step": 22795 }, { "epoch": 0.06, "learning_rate": 9.451658362272888e-05, "loss": 1.0333, "step": 22800 }, { "epoch": 0.06, "learning_rate": 9.451532538124717e-05, "loss": 1.0327, "step": 22805 }, { "epoch": 0.06, "learning_rate": 9.451406713976548e-05, "loss": 1.032, "step": 22810 }, { "epoch": 0.06, "learning_rate": 9.451280889828377e-05, "loss": 1.0333, "step": 22815 }, { "epoch": 0.06, "learning_rate": 9.451155065680206e-05, "loss": 1.0339, "step": 22820 }, { "epoch": 0.06, "learning_rate": 9.451029241532035e-05, "loss": 1.0354, "step": 22825 }, { "epoch": 0.06, "learning_rate": 9.450903417383864e-05, "loss": 1.0327, "step": 22830 }, { "epoch": 0.06, "learning_rate": 9.450777593235695e-05, "loss": 1.0347, "step": 22835 }, { "epoch": 0.06, "learning_rate": 9.450651769087524e-05, "loss": 1.0303, "step": 22840 }, { "epoch": 0.06, "learning_rate": 9.450525944939353e-05, "loss": 1.0341, "step": 22845 }, { "epoch": 0.06, "learning_rate": 9.450400120791182e-05, "loss": 1.0337, "step": 22850 }, { "epoch": 0.06, "learning_rate": 9.450274296643013e-05, "loss": 1.0342, "step": 22855 }, { "epoch": 0.06, "learning_rate": 9.450148472494842e-05, "loss": 1.0339, "step": 22860 }, { "epoch": 0.06, "learning_rate": 9.450022648346671e-05, "loss": 1.0542, "step": 22865 }, { "epoch": 0.06, "learning_rate": 9.4498968241985e-05, "loss": 1.0333, "step": 22870 }, { "epoch": 0.06, "learning_rate": 9.449771000050331e-05, "loss": 1.0331, "step": 22875 }, { "epoch": 0.06, "learning_rate": 9.44964517590216e-05, "loss": 1.0328, "step": 22880 }, { "epoch": 0.06, "learning_rate": 9.449519351753989e-05, "loss": 1.0338, "step": 22885 }, { "epoch": 0.06, "learning_rate": 9.449393527605818e-05, "loss": 1.0351, "step": 22890 }, { "epoch": 0.06, "learning_rate": 9.449267703457648e-05, "loss": 1.0326, "step": 22895 }, { "epoch": 0.06, "learning_rate": 9.449141879309478e-05, "loss": 1.0323, "step": 22900 }, { "epoch": 0.06, "learning_rate": 9.449016055161307e-05, "loss": 1.0358, "step": 22905 }, { "epoch": 0.06, "learning_rate": 9.448890231013136e-05, "loss": 1.0329, "step": 22910 }, { "epoch": 0.06, "learning_rate": 9.448764406864966e-05, "loss": 1.0342, "step": 22915 }, { "epoch": 0.06, "learning_rate": 9.448638582716796e-05, "loss": 1.0355, "step": 22920 }, { "epoch": 0.06, "learning_rate": 9.448512758568625e-05, "loss": 1.0325, "step": 22925 }, { "epoch": 0.06, "learning_rate": 9.448386934420454e-05, "loss": 1.0342, "step": 22930 }, { "epoch": 0.06, "learning_rate": 9.448261110272284e-05, "loss": 1.0541, "step": 22935 }, { "epoch": 0.06, "learning_rate": 9.448135286124114e-05, "loss": 1.0333, "step": 22940 }, { "epoch": 0.06, "learning_rate": 9.448009461975943e-05, "loss": 1.031, "step": 22945 }, { "epoch": 0.06, "learning_rate": 9.447883637827772e-05, "loss": 1.0347, "step": 22950 }, { "epoch": 0.06, "learning_rate": 9.447757813679602e-05, "loss": 1.032, "step": 22955 }, { "epoch": 0.06, "learning_rate": 9.447631989531431e-05, "loss": 1.0324, "step": 22960 }, { "epoch": 0.06, "learning_rate": 9.447506165383261e-05, "loss": 1.0336, "step": 22965 }, { "epoch": 0.06, "learning_rate": 9.44738034123509e-05, "loss": 1.0326, "step": 22970 }, { "epoch": 0.06, "learning_rate": 9.44725451708692e-05, "loss": 1.0336, "step": 22975 }, { "epoch": 0.06, "learning_rate": 9.447128692938749e-05, "loss": 1.0338, "step": 22980 }, { "epoch": 0.06, "learning_rate": 9.447002868790579e-05, "loss": 1.0312, "step": 22985 }, { "epoch": 0.06, "learning_rate": 9.446877044642408e-05, "loss": 1.0319, "step": 22990 }, { "epoch": 0.06, "learning_rate": 9.446751220494238e-05, "loss": 1.0335, "step": 22995 }, { "epoch": 0.06, "learning_rate": 9.446625396346067e-05, "loss": 1.0321, "step": 23000 }, { "epoch": 0.06, "learning_rate": 9.446499572197897e-05, "loss": 1.0339, "step": 23005 }, { "epoch": 0.06, "learning_rate": 9.446373748049726e-05, "loss": 1.0338, "step": 23010 }, { "epoch": 0.06, "learning_rate": 9.446247923901556e-05, "loss": 1.0333, "step": 23015 }, { "epoch": 0.06, "learning_rate": 9.446122099753385e-05, "loss": 1.0323, "step": 23020 }, { "epoch": 0.06, "learning_rate": 9.445996275605214e-05, "loss": 1.0312, "step": 23025 }, { "epoch": 0.06, "learning_rate": 9.445870451457044e-05, "loss": 1.0343, "step": 23030 }, { "epoch": 0.06, "learning_rate": 9.445744627308874e-05, "loss": 1.0333, "step": 23035 }, { "epoch": 0.06, "learning_rate": 9.445618803160703e-05, "loss": 1.0343, "step": 23040 }, { "epoch": 0.06, "learning_rate": 9.445492979012532e-05, "loss": 1.0364, "step": 23045 }, { "epoch": 0.06, "learning_rate": 9.445367154864362e-05, "loss": 1.0331, "step": 23050 }, { "epoch": 0.06, "learning_rate": 9.445241330716192e-05, "loss": 1.0336, "step": 23055 }, { "epoch": 0.06, "learning_rate": 9.445115506568021e-05, "loss": 1.0556, "step": 23060 }, { "epoch": 0.06, "learning_rate": 9.44498968241985e-05, "loss": 1.0326, "step": 23065 }, { "epoch": 0.06, "learning_rate": 9.44486385827168e-05, "loss": 1.0305, "step": 23070 }, { "epoch": 0.06, "learning_rate": 9.44473803412351e-05, "loss": 1.0328, "step": 23075 }, { "epoch": 0.06, "learning_rate": 9.444612209975339e-05, "loss": 1.0344, "step": 23080 }, { "epoch": 0.06, "learning_rate": 9.444486385827168e-05, "loss": 1.0344, "step": 23085 }, { "epoch": 0.06, "learning_rate": 9.444360561678997e-05, "loss": 1.0339, "step": 23090 }, { "epoch": 0.06, "learning_rate": 9.444234737530828e-05, "loss": 1.034, "step": 23095 }, { "epoch": 0.06, "learning_rate": 9.444108913382657e-05, "loss": 1.0345, "step": 23100 }, { "epoch": 0.06, "learning_rate": 9.443983089234486e-05, "loss": 1.0331, "step": 23105 }, { "epoch": 0.06, "learning_rate": 9.443857265086315e-05, "loss": 1.0543, "step": 23110 }, { "epoch": 0.06, "learning_rate": 9.443731440938146e-05, "loss": 1.0318, "step": 23115 }, { "epoch": 0.06, "learning_rate": 9.443605616789975e-05, "loss": 1.0329, "step": 23120 }, { "epoch": 0.06, "learning_rate": 9.443479792641804e-05, "loss": 1.0324, "step": 23125 }, { "epoch": 0.06, "learning_rate": 9.443353968493633e-05, "loss": 1.0328, "step": 23130 }, { "epoch": 0.06, "learning_rate": 9.443228144345463e-05, "loss": 1.0317, "step": 23135 }, { "epoch": 0.06, "learning_rate": 9.443102320197293e-05, "loss": 1.0322, "step": 23140 }, { "epoch": 0.06, "learning_rate": 9.442976496049122e-05, "loss": 1.0318, "step": 23145 }, { "epoch": 0.06, "learning_rate": 9.442850671900951e-05, "loss": 1.0324, "step": 23150 }, { "epoch": 0.06, "learning_rate": 9.44272484775278e-05, "loss": 1.0345, "step": 23155 }, { "epoch": 0.06, "learning_rate": 9.44259902360461e-05, "loss": 1.0335, "step": 23160 }, { "epoch": 0.06, "learning_rate": 9.44247319945644e-05, "loss": 1.0315, "step": 23165 }, { "epoch": 0.06, "learning_rate": 9.442347375308269e-05, "loss": 1.0357, "step": 23170 }, { "epoch": 0.06, "learning_rate": 9.442221551160098e-05, "loss": 1.0334, "step": 23175 }, { "epoch": 0.06, "learning_rate": 9.442095727011929e-05, "loss": 1.0323, "step": 23180 }, { "epoch": 0.06, "learning_rate": 9.441969902863758e-05, "loss": 1.0345, "step": 23185 }, { "epoch": 0.06, "learning_rate": 9.441844078715587e-05, "loss": 1.0323, "step": 23190 }, { "epoch": 0.06, "learning_rate": 9.441718254567416e-05, "loss": 1.0349, "step": 23195 }, { "epoch": 0.06, "learning_rate": 9.441592430419247e-05, "loss": 1.0307, "step": 23200 }, { "epoch": 0.06, "learning_rate": 9.441466606271076e-05, "loss": 1.0312, "step": 23205 }, { "epoch": 0.06, "learning_rate": 9.441340782122905e-05, "loss": 1.0333, "step": 23210 }, { "epoch": 0.06, "learning_rate": 9.441214957974734e-05, "loss": 1.0338, "step": 23215 }, { "epoch": 0.06, "learning_rate": 9.441089133826563e-05, "loss": 1.0339, "step": 23220 }, { "epoch": 0.06, "learning_rate": 9.440963309678394e-05, "loss": 1.0322, "step": 23225 }, { "epoch": 0.06, "learning_rate": 9.440837485530224e-05, "loss": 1.0326, "step": 23230 }, { "epoch": 0.06, "learning_rate": 9.440711661382053e-05, "loss": 1.0359, "step": 23235 }, { "epoch": 0.06, "learning_rate": 9.440585837233883e-05, "loss": 1.0339, "step": 23240 }, { "epoch": 0.06, "learning_rate": 9.440460013085712e-05, "loss": 1.0334, "step": 23245 }, { "epoch": 0.06, "learning_rate": 9.440334188937542e-05, "loss": 1.0331, "step": 23250 }, { "epoch": 0.06, "learning_rate": 9.440208364789371e-05, "loss": 1.0337, "step": 23255 }, { "epoch": 0.06, "learning_rate": 9.4400825406412e-05, "loss": 1.0307, "step": 23260 }, { "epoch": 0.06, "learning_rate": 9.43995671649303e-05, "loss": 1.0325, "step": 23265 }, { "epoch": 0.06, "learning_rate": 9.43983089234486e-05, "loss": 1.032, "step": 23270 }, { "epoch": 0.06, "learning_rate": 9.43970506819669e-05, "loss": 1.0322, "step": 23275 }, { "epoch": 0.06, "learning_rate": 9.439579244048519e-05, "loss": 1.0349, "step": 23280 }, { "epoch": 0.06, "learning_rate": 9.439453419900348e-05, "loss": 1.0326, "step": 23285 }, { "epoch": 0.06, "learning_rate": 9.439327595752178e-05, "loss": 1.0342, "step": 23290 }, { "epoch": 0.06, "learning_rate": 9.439201771604007e-05, "loss": 1.0343, "step": 23295 }, { "epoch": 0.06, "learning_rate": 9.439075947455837e-05, "loss": 1.0519, "step": 23300 }, { "epoch": 0.06, "learning_rate": 9.438950123307666e-05, "loss": 1.0318, "step": 23305 }, { "epoch": 0.06, "learning_rate": 9.438824299159495e-05, "loss": 1.0301, "step": 23310 }, { "epoch": 0.06, "learning_rate": 9.438698475011325e-05, "loss": 1.0316, "step": 23315 }, { "epoch": 0.06, "learning_rate": 9.438572650863155e-05, "loss": 1.0331, "step": 23320 }, { "epoch": 0.06, "learning_rate": 9.438446826714984e-05, "loss": 1.0318, "step": 23325 }, { "epoch": 0.06, "learning_rate": 9.438321002566813e-05, "loss": 1.034, "step": 23330 }, { "epoch": 0.06, "learning_rate": 9.438195178418643e-05, "loss": 1.0344, "step": 23335 }, { "epoch": 0.06, "learning_rate": 9.438069354270473e-05, "loss": 1.0328, "step": 23340 }, { "epoch": 0.06, "learning_rate": 9.437943530122302e-05, "loss": 1.0337, "step": 23345 }, { "epoch": 0.06, "learning_rate": 9.437817705974131e-05, "loss": 1.0333, "step": 23350 }, { "epoch": 0.06, "learning_rate": 9.437691881825961e-05, "loss": 1.0316, "step": 23355 }, { "epoch": 0.06, "learning_rate": 9.43756605767779e-05, "loss": 1.0334, "step": 23360 }, { "epoch": 0.06, "learning_rate": 9.43744023352962e-05, "loss": 1.0345, "step": 23365 }, { "epoch": 0.06, "learning_rate": 9.437314409381449e-05, "loss": 1.0344, "step": 23370 }, { "epoch": 0.06, "learning_rate": 9.437188585233278e-05, "loss": 1.0339, "step": 23375 }, { "epoch": 0.06, "learning_rate": 9.437062761085109e-05, "loss": 1.033, "step": 23380 }, { "epoch": 0.06, "learning_rate": 9.436936936936938e-05, "loss": 1.0333, "step": 23385 }, { "epoch": 0.06, "learning_rate": 9.436811112788767e-05, "loss": 1.0334, "step": 23390 }, { "epoch": 0.06, "learning_rate": 9.436685288640596e-05, "loss": 1.0347, "step": 23395 }, { "epoch": 0.06, "learning_rate": 9.436559464492427e-05, "loss": 1.0332, "step": 23400 }, { "epoch": 0.06, "learning_rate": 9.436433640344256e-05, "loss": 1.0324, "step": 23405 }, { "epoch": 0.06, "learning_rate": 9.436307816196085e-05, "loss": 1.0323, "step": 23410 }, { "epoch": 0.06, "learning_rate": 9.436181992047914e-05, "loss": 1.0322, "step": 23415 }, { "epoch": 0.06, "learning_rate": 9.436056167899743e-05, "loss": 1.033, "step": 23420 }, { "epoch": 0.06, "learning_rate": 9.435930343751574e-05, "loss": 1.0534, "step": 23425 }, { "epoch": 0.06, "learning_rate": 9.435804519603403e-05, "loss": 1.0339, "step": 23430 }, { "epoch": 0.06, "learning_rate": 9.435678695455232e-05, "loss": 1.0344, "step": 23435 }, { "epoch": 0.06, "learning_rate": 9.435552871307061e-05, "loss": 1.0306, "step": 23440 }, { "epoch": 0.06, "learning_rate": 9.435427047158892e-05, "loss": 1.0347, "step": 23445 }, { "epoch": 0.06, "learning_rate": 9.435301223010721e-05, "loss": 1.0319, "step": 23450 }, { "epoch": 0.06, "learning_rate": 9.43517539886255e-05, "loss": 1.0352, "step": 23455 }, { "epoch": 0.06, "learning_rate": 9.435049574714379e-05, "loss": 1.0339, "step": 23460 }, { "epoch": 0.06, "learning_rate": 9.43492375056621e-05, "loss": 1.0337, "step": 23465 }, { "epoch": 0.06, "learning_rate": 9.434797926418039e-05, "loss": 1.0314, "step": 23470 }, { "epoch": 0.06, "learning_rate": 9.434672102269868e-05, "loss": 1.0318, "step": 23475 }, { "epoch": 0.06, "learning_rate": 9.434546278121697e-05, "loss": 1.0335, "step": 23480 }, { "epoch": 0.06, "learning_rate": 9.434420453973526e-05, "loss": 1.0337, "step": 23485 }, { "epoch": 0.06, "learning_rate": 9.434294629825357e-05, "loss": 1.0579, "step": 23490 }, { "epoch": 0.06, "learning_rate": 9.434168805677186e-05, "loss": 1.0332, "step": 23495 }, { "epoch": 0.06, "learning_rate": 9.434042981529015e-05, "loss": 1.0327, "step": 23500 }, { "epoch": 0.06, "learning_rate": 9.433917157380844e-05, "loss": 1.0339, "step": 23505 }, { "epoch": 0.06, "learning_rate": 9.433791333232675e-05, "loss": 1.0336, "step": 23510 }, { "epoch": 0.06, "learning_rate": 9.433665509084504e-05, "loss": 1.0331, "step": 23515 }, { "epoch": 0.06, "learning_rate": 9.433539684936333e-05, "loss": 1.0334, "step": 23520 }, { "epoch": 0.06, "learning_rate": 9.433413860788162e-05, "loss": 1.0535, "step": 23525 }, { "epoch": 0.06, "learning_rate": 9.433288036639993e-05, "loss": 1.0319, "step": 23530 }, { "epoch": 0.06, "learning_rate": 9.433162212491822e-05, "loss": 1.0322, "step": 23535 }, { "epoch": 0.06, "learning_rate": 9.433036388343651e-05, "loss": 1.0518, "step": 23540 }, { "epoch": 0.06, "learning_rate": 9.43291056419548e-05, "loss": 1.0335, "step": 23545 }, { "epoch": 0.06, "learning_rate": 9.43278474004731e-05, "loss": 1.0517, "step": 23550 }, { "epoch": 0.06, "learning_rate": 9.43265891589914e-05, "loss": 1.0354, "step": 23555 }, { "epoch": 0.06, "learning_rate": 9.432533091750969e-05, "loss": 1.0329, "step": 23560 }, { "epoch": 0.06, "learning_rate": 9.432407267602798e-05, "loss": 1.0334, "step": 23565 }, { "epoch": 0.06, "learning_rate": 9.432281443454627e-05, "loss": 1.0331, "step": 23570 }, { "epoch": 0.06, "learning_rate": 9.432155619306458e-05, "loss": 1.0316, "step": 23575 }, { "epoch": 0.06, "learning_rate": 9.432029795158287e-05, "loss": 1.0336, "step": 23580 }, { "epoch": 0.06, "learning_rate": 9.431903971010116e-05, "loss": 1.033, "step": 23585 }, { "epoch": 0.06, "learning_rate": 9.431778146861945e-05, "loss": 1.0341, "step": 23590 }, { "epoch": 0.06, "learning_rate": 9.431652322713776e-05, "loss": 1.0339, "step": 23595 }, { "epoch": 0.06, "learning_rate": 9.431526498565605e-05, "loss": 1.0323, "step": 23600 }, { "epoch": 0.06, "learning_rate": 9.431400674417434e-05, "loss": 1.0567, "step": 23605 }, { "epoch": 0.06, "learning_rate": 9.431274850269263e-05, "loss": 1.0333, "step": 23610 }, { "epoch": 0.06, "learning_rate": 9.431149026121093e-05, "loss": 1.0327, "step": 23615 }, { "epoch": 0.06, "learning_rate": 9.431023201972923e-05, "loss": 1.0319, "step": 23620 }, { "epoch": 0.06, "learning_rate": 9.430897377824752e-05, "loss": 1.0349, "step": 23625 }, { "epoch": 0.06, "learning_rate": 9.430771553676581e-05, "loss": 1.0306, "step": 23630 }, { "epoch": 0.06, "learning_rate": 9.43064572952841e-05, "loss": 1.0326, "step": 23635 }, { "epoch": 0.06, "learning_rate": 9.430519905380241e-05, "loss": 1.0342, "step": 23640 }, { "epoch": 0.06, "learning_rate": 9.43039408123207e-05, "loss": 1.0333, "step": 23645 }, { "epoch": 0.06, "learning_rate": 9.4302682570839e-05, "loss": 1.0344, "step": 23650 }, { "epoch": 0.06, "learning_rate": 9.430142432935729e-05, "loss": 1.0347, "step": 23655 }, { "epoch": 0.06, "learning_rate": 9.430016608787559e-05, "loss": 1.0322, "step": 23660 }, { "epoch": 0.06, "learning_rate": 9.429890784639388e-05, "loss": 1.0339, "step": 23665 }, { "epoch": 0.06, "learning_rate": 9.429764960491217e-05, "loss": 1.032, "step": 23670 }, { "epoch": 0.06, "learning_rate": 9.429639136343047e-05, "loss": 1.0349, "step": 23675 }, { "epoch": 0.06, "learning_rate": 9.429513312194876e-05, "loss": 1.0332, "step": 23680 }, { "epoch": 0.06, "learning_rate": 9.429387488046706e-05, "loss": 1.0343, "step": 23685 }, { "epoch": 0.06, "learning_rate": 9.429261663898535e-05, "loss": 1.0329, "step": 23690 }, { "epoch": 0.06, "learning_rate": 9.429135839750365e-05, "loss": 1.0326, "step": 23695 }, { "epoch": 0.06, "learning_rate": 9.429010015602194e-05, "loss": 1.0323, "step": 23700 }, { "epoch": 0.06, "learning_rate": 9.428884191454024e-05, "loss": 1.0336, "step": 23705 }, { "epoch": 0.06, "learning_rate": 9.428758367305853e-05, "loss": 1.0337, "step": 23710 }, { "epoch": 0.06, "learning_rate": 9.428632543157683e-05, "loss": 1.0327, "step": 23715 }, { "epoch": 0.06, "learning_rate": 9.428506719009512e-05, "loss": 1.033, "step": 23720 }, { "epoch": 0.06, "learning_rate": 9.428380894861342e-05, "loss": 1.0329, "step": 23725 }, { "epoch": 0.06, "learning_rate": 9.428255070713173e-05, "loss": 1.0333, "step": 23730 }, { "epoch": 0.06, "learning_rate": 9.428129246565002e-05, "loss": 1.032, "step": 23735 }, { "epoch": 0.06, "learning_rate": 9.428003422416831e-05, "loss": 1.0315, "step": 23740 }, { "epoch": 0.06, "learning_rate": 9.42787759826866e-05, "loss": 1.0335, "step": 23745 }, { "epoch": 0.06, "learning_rate": 9.427751774120491e-05, "loss": 1.0328, "step": 23750 }, { "epoch": 0.06, "learning_rate": 9.42762594997232e-05, "loss": 1.0328, "step": 23755 }, { "epoch": 0.06, "learning_rate": 9.427500125824149e-05, "loss": 1.0335, "step": 23760 }, { "epoch": 0.06, "learning_rate": 9.427374301675978e-05, "loss": 1.033, "step": 23765 }, { "epoch": 0.06, "learning_rate": 9.427248477527807e-05, "loss": 1.0327, "step": 23770 }, { "epoch": 0.06, "learning_rate": 9.427122653379638e-05, "loss": 1.0332, "step": 23775 }, { "epoch": 0.06, "learning_rate": 9.426996829231467e-05, "loss": 1.033, "step": 23780 }, { "epoch": 0.06, "learning_rate": 9.426871005083296e-05, "loss": 1.0343, "step": 23785 }, { "epoch": 0.06, "learning_rate": 9.426745180935125e-05, "loss": 1.0335, "step": 23790 }, { "epoch": 0.06, "learning_rate": 9.426619356786956e-05, "loss": 1.0333, "step": 23795 }, { "epoch": 0.06, "learning_rate": 9.426518697468418e-05, "loss": 1.0563, "step": 23800 }, { "epoch": 0.06, "learning_rate": 9.426392873320247e-05, "loss": 1.0312, "step": 23805 }, { "epoch": 0.06, "learning_rate": 9.426267049172078e-05, "loss": 1.0347, "step": 23810 }, { "epoch": 0.06, "learning_rate": 9.426141225023907e-05, "loss": 1.0321, "step": 23815 }, { "epoch": 0.06, "learning_rate": 9.426015400875736e-05, "loss": 1.0343, "step": 23820 }, { "epoch": 0.06, "learning_rate": 9.425889576727565e-05, "loss": 1.0349, "step": 23825 }, { "epoch": 0.06, "learning_rate": 9.425763752579395e-05, "loss": 1.0356, "step": 23830 }, { "epoch": 0.06, "learning_rate": 9.425637928431225e-05, "loss": 1.0336, "step": 23835 }, { "epoch": 0.06, "learning_rate": 9.425512104283054e-05, "loss": 1.0342, "step": 23840 }, { "epoch": 0.06, "learning_rate": 9.425386280134883e-05, "loss": 1.0341, "step": 23845 }, { "epoch": 0.06, "learning_rate": 9.425260455986713e-05, "loss": 1.0343, "step": 23850 }, { "epoch": 0.06, "learning_rate": 9.425134631838543e-05, "loss": 1.0349, "step": 23855 }, { "epoch": 0.06, "learning_rate": 9.425008807690372e-05, "loss": 1.0347, "step": 23860 }, { "epoch": 0.06, "learning_rate": 9.424882983542201e-05, "loss": 1.0299, "step": 23865 }, { "epoch": 0.06, "learning_rate": 9.42475715939403e-05, "loss": 1.031, "step": 23870 }, { "epoch": 0.06, "learning_rate": 9.424631335245861e-05, "loss": 1.0326, "step": 23875 }, { "epoch": 0.06, "learning_rate": 9.42450551109769e-05, "loss": 1.0344, "step": 23880 }, { "epoch": 0.06, "learning_rate": 9.42437968694952e-05, "loss": 1.0349, "step": 23885 }, { "epoch": 0.06, "learning_rate": 9.424253862801348e-05, "loss": 1.0331, "step": 23890 }, { "epoch": 0.06, "learning_rate": 9.424128038653178e-05, "loss": 1.034, "step": 23895 }, { "epoch": 0.06, "learning_rate": 9.424002214505008e-05, "loss": 1.0337, "step": 23900 }, { "epoch": 0.06, "learning_rate": 9.423876390356837e-05, "loss": 1.0315, "step": 23905 }, { "epoch": 0.06, "learning_rate": 9.423750566208666e-05, "loss": 1.0335, "step": 23910 }, { "epoch": 0.06, "learning_rate": 9.423624742060496e-05, "loss": 1.0326, "step": 23915 }, { "epoch": 0.06, "learning_rate": 9.423498917912326e-05, "loss": 1.0326, "step": 23920 }, { "epoch": 0.06, "learning_rate": 9.423373093764155e-05, "loss": 1.0305, "step": 23925 }, { "epoch": 0.06, "learning_rate": 9.423247269615986e-05, "loss": 1.032, "step": 23930 }, { "epoch": 0.06, "learning_rate": 9.423121445467815e-05, "loss": 1.0339, "step": 23935 }, { "epoch": 0.06, "learning_rate": 9.423020786149279e-05, "loss": 1.0533, "step": 23940 }, { "epoch": 0.06, "learning_rate": 9.422894962001108e-05, "loss": 1.0343, "step": 23945 }, { "epoch": 0.06, "learning_rate": 9.422769137852937e-05, "loss": 1.0321, "step": 23950 }, { "epoch": 0.06, "learning_rate": 9.422643313704766e-05, "loss": 1.0299, "step": 23955 }, { "epoch": 0.06, "learning_rate": 9.422517489556597e-05, "loss": 1.0345, "step": 23960 }, { "epoch": 0.06, "learning_rate": 9.422391665408426e-05, "loss": 1.033, "step": 23965 }, { "epoch": 0.06, "learning_rate": 9.422265841260255e-05, "loss": 1.0342, "step": 23970 }, { "epoch": 0.06, "learning_rate": 9.422140017112084e-05, "loss": 1.0326, "step": 23975 }, { "epoch": 0.06, "learning_rate": 9.422014192963913e-05, "loss": 1.0332, "step": 23980 }, { "epoch": 0.06, "learning_rate": 9.421888368815744e-05, "loss": 1.0362, "step": 23985 }, { "epoch": 0.06, "learning_rate": 9.421762544667573e-05, "loss": 1.0319, "step": 23990 }, { "epoch": 0.06, "learning_rate": 9.421636720519402e-05, "loss": 1.0333, "step": 23995 }, { "epoch": 0.06, "learning_rate": 9.421510896371231e-05, "loss": 1.0559, "step": 24000 }, { "epoch": 0.06, "learning_rate": 9.421385072223062e-05, "loss": 1.0312, "step": 24005 }, { "epoch": 0.06, "learning_rate": 9.421259248074891e-05, "loss": 1.0302, "step": 24010 }, { "epoch": 0.06, "learning_rate": 9.42113342392672e-05, "loss": 1.034, "step": 24015 }, { "epoch": 0.06, "learning_rate": 9.421007599778549e-05, "loss": 1.0331, "step": 24020 }, { "epoch": 0.06, "learning_rate": 9.42088177563038e-05, "loss": 1.0335, "step": 24025 }, { "epoch": 0.06, "learning_rate": 9.420755951482209e-05, "loss": 1.0325, "step": 24030 }, { "epoch": 0.06, "learning_rate": 9.420630127334038e-05, "loss": 1.055, "step": 24035 }, { "epoch": 0.06, "learning_rate": 9.420504303185867e-05, "loss": 1.0336, "step": 24040 }, { "epoch": 0.06, "learning_rate": 9.420378479037696e-05, "loss": 1.031, "step": 24045 }, { "epoch": 0.06, "learning_rate": 9.420252654889527e-05, "loss": 1.0563, "step": 24050 }, { "epoch": 0.06, "learning_rate": 9.420126830741356e-05, "loss": 1.0327, "step": 24055 }, { "epoch": 0.06, "learning_rate": 9.420001006593185e-05, "loss": 1.0345, "step": 24060 }, { "epoch": 0.06, "learning_rate": 9.419875182445014e-05, "loss": 1.0327, "step": 24065 }, { "epoch": 0.06, "learning_rate": 9.419749358296845e-05, "loss": 1.0311, "step": 24070 }, { "epoch": 0.06, "learning_rate": 9.419623534148674e-05, "loss": 1.033, "step": 24075 }, { "epoch": 0.06, "learning_rate": 9.419497710000503e-05, "loss": 1.0328, "step": 24080 }, { "epoch": 0.06, "learning_rate": 9.419371885852332e-05, "loss": 1.0344, "step": 24085 }, { "epoch": 0.06, "learning_rate": 9.419246061704163e-05, "loss": 1.0339, "step": 24090 }, { "epoch": 0.06, "learning_rate": 9.419120237555992e-05, "loss": 1.0332, "step": 24095 }, { "epoch": 0.06, "learning_rate": 9.418994413407821e-05, "loss": 1.0325, "step": 24100 }, { "epoch": 0.06, "learning_rate": 9.41886858925965e-05, "loss": 1.0308, "step": 24105 }, { "epoch": 0.06, "learning_rate": 9.41874276511148e-05, "loss": 1.0332, "step": 24110 }, { "epoch": 0.06, "learning_rate": 9.41861694096331e-05, "loss": 1.0304, "step": 24115 }, { "epoch": 0.06, "learning_rate": 9.418491116815139e-05, "loss": 1.0328, "step": 24120 }, { "epoch": 0.06, "learning_rate": 9.418365292666968e-05, "loss": 1.0319, "step": 24125 }, { "epoch": 0.06, "learning_rate": 9.418239468518799e-05, "loss": 1.0307, "step": 24130 }, { "epoch": 0.06, "learning_rate": 9.418113644370628e-05, "loss": 1.0347, "step": 24135 }, { "epoch": 0.06, "learning_rate": 9.417987820222459e-05, "loss": 1.0328, "step": 24140 }, { "epoch": 0.06, "learning_rate": 9.417861996074288e-05, "loss": 1.0306, "step": 24145 }, { "epoch": 0.06, "learning_rate": 9.417736171926117e-05, "loss": 1.033, "step": 24150 }, { "epoch": 0.06, "learning_rate": 9.417610347777946e-05, "loss": 1.0501, "step": 24155 }, { "epoch": 0.06, "learning_rate": 9.417484523629777e-05, "loss": 1.0347, "step": 24160 }, { "epoch": 0.06, "learning_rate": 9.417358699481606e-05, "loss": 1.055, "step": 24165 }, { "epoch": 0.06, "learning_rate": 9.417232875333435e-05, "loss": 1.0377, "step": 24170 }, { "epoch": 0.06, "learning_rate": 9.417107051185264e-05, "loss": 1.033, "step": 24175 }, { "epoch": 0.06, "learning_rate": 9.416981227037093e-05, "loss": 1.0324, "step": 24180 }, { "epoch": 0.06, "learning_rate": 9.416855402888924e-05, "loss": 1.0334, "step": 24185 }, { "epoch": 0.06, "learning_rate": 9.416729578740753e-05, "loss": 1.0358, "step": 24190 }, { "epoch": 0.06, "learning_rate": 9.416603754592582e-05, "loss": 1.0322, "step": 24195 }, { "epoch": 0.06, "learning_rate": 9.416477930444411e-05, "loss": 1.0335, "step": 24200 }, { "epoch": 0.06, "learning_rate": 9.416352106296242e-05, "loss": 1.033, "step": 24205 }, { "epoch": 0.06, "learning_rate": 9.416226282148071e-05, "loss": 1.0332, "step": 24210 }, { "epoch": 0.06, "learning_rate": 9.4161004579999e-05, "loss": 1.0305, "step": 24215 }, { "epoch": 0.06, "learning_rate": 9.415974633851729e-05, "loss": 1.0335, "step": 24220 }, { "epoch": 0.06, "learning_rate": 9.41584880970356e-05, "loss": 1.032, "step": 24225 }, { "epoch": 0.06, "learning_rate": 9.415722985555389e-05, "loss": 1.0325, "step": 24230 }, { "epoch": 0.06, "learning_rate": 9.415597161407218e-05, "loss": 1.0324, "step": 24235 }, { "epoch": 0.06, "learning_rate": 9.415471337259047e-05, "loss": 1.033, "step": 24240 }, { "epoch": 0.06, "learning_rate": 9.415345513110876e-05, "loss": 1.0307, "step": 24245 }, { "epoch": 0.06, "learning_rate": 9.415219688962707e-05, "loss": 1.0338, "step": 24250 }, { "epoch": 0.06, "learning_rate": 9.415093864814536e-05, "loss": 1.0341, "step": 24255 }, { "epoch": 0.06, "learning_rate": 9.414968040666365e-05, "loss": 1.0299, "step": 24260 }, { "epoch": 0.06, "learning_rate": 9.414842216518194e-05, "loss": 1.0318, "step": 24265 }, { "epoch": 0.06, "learning_rate": 9.414716392370025e-05, "loss": 1.0331, "step": 24270 }, { "epoch": 0.06, "learning_rate": 9.414590568221854e-05, "loss": 1.0317, "step": 24275 }, { "epoch": 0.06, "learning_rate": 9.414464744073683e-05, "loss": 1.0331, "step": 24280 }, { "epoch": 0.06, "learning_rate": 9.414338919925512e-05, "loss": 1.0331, "step": 24285 }, { "epoch": 0.06, "learning_rate": 9.414213095777343e-05, "loss": 1.0323, "step": 24290 }, { "epoch": 0.06, "learning_rate": 9.414087271629172e-05, "loss": 1.0331, "step": 24295 }, { "epoch": 0.06, "learning_rate": 9.413961447481001e-05, "loss": 1.0316, "step": 24300 }, { "epoch": 0.06, "learning_rate": 9.41383562333283e-05, "loss": 1.0325, "step": 24305 }, { "epoch": 0.06, "learning_rate": 9.41370979918466e-05, "loss": 1.0323, "step": 24310 }, { "epoch": 0.06, "learning_rate": 9.41358397503649e-05, "loss": 1.033, "step": 24315 }, { "epoch": 0.06, "learning_rate": 9.413458150888319e-05, "loss": 1.0333, "step": 24320 }, { "epoch": 0.06, "learning_rate": 9.413332326740148e-05, "loss": 1.0333, "step": 24325 }, { "epoch": 0.06, "learning_rate": 9.413206502591977e-05, "loss": 1.0335, "step": 24330 }, { "epoch": 0.06, "learning_rate": 9.413080678443808e-05, "loss": 1.049, "step": 24335 }, { "epoch": 0.06, "learning_rate": 9.412954854295637e-05, "loss": 1.0312, "step": 24340 }, { "epoch": 0.06, "learning_rate": 9.412829030147466e-05, "loss": 1.0302, "step": 24345 }, { "epoch": 0.06, "learning_rate": 9.412703205999295e-05, "loss": 1.0339, "step": 24350 }, { "epoch": 0.06, "learning_rate": 9.412577381851126e-05, "loss": 1.0326, "step": 24355 }, { "epoch": 0.06, "learning_rate": 9.412451557702955e-05, "loss": 1.0328, "step": 24360 }, { "epoch": 0.06, "learning_rate": 9.412325733554784e-05, "loss": 1.0319, "step": 24365 }, { "epoch": 0.06, "learning_rate": 9.412199909406613e-05, "loss": 1.0303, "step": 24370 }, { "epoch": 0.06, "learning_rate": 9.412074085258443e-05, "loss": 1.0342, "step": 24375 }, { "epoch": 0.06, "learning_rate": 9.411948261110273e-05, "loss": 1.0347, "step": 24380 }, { "epoch": 0.06, "learning_rate": 9.411822436962102e-05, "loss": 1.0353, "step": 24385 }, { "epoch": 0.06, "learning_rate": 9.411696612813931e-05, "loss": 1.0301, "step": 24390 }, { "epoch": 0.06, "learning_rate": 9.41157078866576e-05, "loss": 1.0345, "step": 24395 }, { "epoch": 0.06, "learning_rate": 9.411444964517591e-05, "loss": 1.0314, "step": 24400 }, { "epoch": 0.06, "learning_rate": 9.41131914036942e-05, "loss": 1.0334, "step": 24405 }, { "epoch": 0.06, "learning_rate": 9.41119331622125e-05, "loss": 1.0331, "step": 24410 }, { "epoch": 0.06, "learning_rate": 9.411067492073079e-05, "loss": 1.0321, "step": 24415 }, { "epoch": 0.06, "learning_rate": 9.410941667924909e-05, "loss": 1.0335, "step": 24420 }, { "epoch": 0.06, "learning_rate": 9.410815843776738e-05, "loss": 1.0321, "step": 24425 }, { "epoch": 0.06, "learning_rate": 9.410690019628567e-05, "loss": 1.0307, "step": 24430 }, { "epoch": 0.06, "learning_rate": 9.410564195480397e-05, "loss": 1.0309, "step": 24435 }, { "epoch": 0.06, "learning_rate": 9.410438371332226e-05, "loss": 1.0309, "step": 24440 }, { "epoch": 0.06, "learning_rate": 9.410312547184056e-05, "loss": 1.0327, "step": 24445 }, { "epoch": 0.06, "learning_rate": 9.410186723035885e-05, "loss": 1.0312, "step": 24450 }, { "epoch": 0.06, "learning_rate": 9.410060898887715e-05, "loss": 1.0318, "step": 24455 }, { "epoch": 0.06, "learning_rate": 9.409935074739544e-05, "loss": 1.032, "step": 24460 }, { "epoch": 0.06, "learning_rate": 9.409809250591374e-05, "loss": 1.033, "step": 24465 }, { "epoch": 0.06, "learning_rate": 9.409683426443203e-05, "loss": 1.0321, "step": 24470 }, { "epoch": 0.06, "learning_rate": 9.409557602295033e-05, "loss": 1.0359, "step": 24475 }, { "epoch": 0.06, "learning_rate": 9.409431778146862e-05, "loss": 1.0326, "step": 24480 }, { "epoch": 0.06, "learning_rate": 9.409305953998692e-05, "loss": 1.0319, "step": 24485 }, { "epoch": 0.06, "learning_rate": 9.409180129850521e-05, "loss": 1.0333, "step": 24490 }, { "epoch": 0.06, "learning_rate": 9.40905430570235e-05, "loss": 1.0319, "step": 24495 }, { "epoch": 0.06, "learning_rate": 9.40892848155418e-05, "loss": 1.0327, "step": 24500 }, { "epoch": 0.06, "learning_rate": 9.408802657406009e-05, "loss": 1.0304, "step": 24505 }, { "epoch": 0.06, "learning_rate": 9.40867683325784e-05, "loss": 1.0329, "step": 24510 }, { "epoch": 0.06, "learning_rate": 9.408551009109669e-05, "loss": 1.0324, "step": 24515 }, { "epoch": 0.06, "learning_rate": 9.408425184961498e-05, "loss": 1.0337, "step": 24520 }, { "epoch": 0.06, "learning_rate": 9.408299360813327e-05, "loss": 1.0339, "step": 24525 }, { "epoch": 0.06, "learning_rate": 9.408173536665157e-05, "loss": 1.0333, "step": 24530 }, { "epoch": 0.06, "learning_rate": 9.408047712516987e-05, "loss": 1.0342, "step": 24535 }, { "epoch": 0.06, "learning_rate": 9.407921888368816e-05, "loss": 1.0332, "step": 24540 }, { "epoch": 0.06, "learning_rate": 9.407796064220645e-05, "loss": 1.0327, "step": 24545 }, { "epoch": 0.06, "learning_rate": 9.407670240072475e-05, "loss": 1.0306, "step": 24550 }, { "epoch": 0.06, "learning_rate": 9.407544415924305e-05, "loss": 1.0337, "step": 24555 }, { "epoch": 0.06, "learning_rate": 9.407418591776134e-05, "loss": 1.0308, "step": 24560 }, { "epoch": 0.06, "learning_rate": 9.407292767627963e-05, "loss": 1.0308, "step": 24565 }, { "epoch": 0.06, "learning_rate": 9.407166943479792e-05, "loss": 1.0349, "step": 24570 }, { "epoch": 0.06, "learning_rate": 9.407041119331623e-05, "loss": 1.0339, "step": 24575 }, { "epoch": 0.06, "learning_rate": 9.406915295183452e-05, "loss": 1.0322, "step": 24580 }, { "epoch": 0.06, "learning_rate": 9.406789471035281e-05, "loss": 1.0312, "step": 24585 }, { "epoch": 0.06, "learning_rate": 9.40666364688711e-05, "loss": 1.0326, "step": 24590 }, { "epoch": 0.06, "learning_rate": 9.40653782273894e-05, "loss": 1.0312, "step": 24595 }, { "epoch": 0.06, "learning_rate": 9.40641199859077e-05, "loss": 1.034, "step": 24600 }, { "epoch": 0.06, "learning_rate": 9.406286174442599e-05, "loss": 1.032, "step": 24605 }, { "epoch": 0.06, "learning_rate": 9.406160350294428e-05, "loss": 1.032, "step": 24610 }, { "epoch": 0.06, "learning_rate": 9.406034526146259e-05, "loss": 1.0322, "step": 24615 }, { "epoch": 0.06, "learning_rate": 9.405908701998088e-05, "loss": 1.0326, "step": 24620 }, { "epoch": 0.06, "learning_rate": 9.405782877849917e-05, "loss": 1.0328, "step": 24625 }, { "epoch": 0.06, "learning_rate": 9.405657053701747e-05, "loss": 1.0335, "step": 24630 }, { "epoch": 0.06, "learning_rate": 9.405531229553577e-05, "loss": 1.0307, "step": 24635 }, { "epoch": 0.06, "learning_rate": 9.405405405405407e-05, "loss": 1.0319, "step": 24640 }, { "epoch": 0.06, "learning_rate": 9.405279581257236e-05, "loss": 1.0561, "step": 24645 }, { "epoch": 0.06, "learning_rate": 9.405153757109065e-05, "loss": 1.033, "step": 24650 }, { "epoch": 0.06, "learning_rate": 9.405027932960895e-05, "loss": 1.035, "step": 24655 }, { "epoch": 0.06, "learning_rate": 9.404902108812724e-05, "loss": 1.0324, "step": 24660 }, { "epoch": 0.06, "learning_rate": 9.404776284664554e-05, "loss": 1.0337, "step": 24665 }, { "epoch": 0.06, "learning_rate": 9.404650460516383e-05, "loss": 1.0326, "step": 24670 }, { "epoch": 0.06, "learning_rate": 9.404524636368213e-05, "loss": 1.0303, "step": 24675 }, { "epoch": 0.06, "learning_rate": 9.404398812220042e-05, "loss": 1.0584, "step": 24680 }, { "epoch": 0.06, "learning_rate": 9.404272988071872e-05, "loss": 1.0339, "step": 24685 }, { "epoch": 0.06, "learning_rate": 9.404147163923701e-05, "loss": 1.0339, "step": 24690 }, { "epoch": 0.06, "learning_rate": 9.40402133977553e-05, "loss": 1.0331, "step": 24695 }, { "epoch": 0.06, "learning_rate": 9.40389551562736e-05, "loss": 1.0331, "step": 24700 }, { "epoch": 0.06, "learning_rate": 9.40376969147919e-05, "loss": 1.0332, "step": 24705 }, { "epoch": 0.06, "learning_rate": 9.40364386733102e-05, "loss": 1.0323, "step": 24710 }, { "epoch": 0.06, "learning_rate": 9.403518043182848e-05, "loss": 1.0313, "step": 24715 }, { "epoch": 0.06, "learning_rate": 9.403392219034678e-05, "loss": 1.0338, "step": 24720 }, { "epoch": 0.06, "learning_rate": 9.403266394886507e-05, "loss": 1.033, "step": 24725 }, { "epoch": 0.06, "learning_rate": 9.403140570738337e-05, "loss": 1.0335, "step": 24730 }, { "epoch": 0.06, "learning_rate": 9.403014746590166e-05, "loss": 1.0314, "step": 24735 }, { "epoch": 0.06, "learning_rate": 9.402888922441996e-05, "loss": 1.0328, "step": 24740 }, { "epoch": 0.06, "learning_rate": 9.402763098293825e-05, "loss": 1.0345, "step": 24745 }, { "epoch": 0.06, "learning_rate": 9.402637274145655e-05, "loss": 1.0328, "step": 24750 }, { "epoch": 0.06, "learning_rate": 9.402511449997484e-05, "loss": 1.0316, "step": 24755 }, { "epoch": 0.06, "learning_rate": 9.402385625849314e-05, "loss": 1.0333, "step": 24760 }, { "epoch": 0.06, "learning_rate": 9.402259801701143e-05, "loss": 1.0325, "step": 24765 }, { "epoch": 0.06, "learning_rate": 9.402133977552973e-05, "loss": 1.0338, "step": 24770 }, { "epoch": 0.06, "learning_rate": 9.402008153404802e-05, "loss": 1.0305, "step": 24775 }, { "epoch": 0.06, "learning_rate": 9.401882329256632e-05, "loss": 1.0337, "step": 24780 }, { "epoch": 0.06, "learning_rate": 9.401756505108461e-05, "loss": 1.0334, "step": 24785 }, { "epoch": 0.06, "learning_rate": 9.40163068096029e-05, "loss": 1.0342, "step": 24790 }, { "epoch": 0.06, "learning_rate": 9.40150485681212e-05, "loss": 1.0325, "step": 24795 }, { "epoch": 0.06, "learning_rate": 9.40137903266395e-05, "loss": 1.0313, "step": 24800 }, { "epoch": 0.06, "learning_rate": 9.401253208515779e-05, "loss": 1.0337, "step": 24805 }, { "epoch": 0.06, "learning_rate": 9.401127384367608e-05, "loss": 1.0335, "step": 24810 }, { "epoch": 0.06, "learning_rate": 9.401001560219438e-05, "loss": 1.0332, "step": 24815 }, { "epoch": 0.06, "learning_rate": 9.400875736071268e-05, "loss": 1.0335, "step": 24820 }, { "epoch": 0.06, "learning_rate": 9.400749911923097e-05, "loss": 1.0314, "step": 24825 }, { "epoch": 0.06, "learning_rate": 9.400624087774926e-05, "loss": 1.0338, "step": 24830 }, { "epoch": 0.06, "learning_rate": 9.400498263626756e-05, "loss": 1.033, "step": 24835 }, { "epoch": 0.06, "learning_rate": 9.400372439478586e-05, "loss": 1.0341, "step": 24840 }, { "epoch": 0.06, "learning_rate": 9.400246615330415e-05, "loss": 1.0325, "step": 24845 }, { "epoch": 0.06, "learning_rate": 9.400120791182244e-05, "loss": 1.0336, "step": 24850 }, { "epoch": 0.06, "learning_rate": 9.399994967034073e-05, "loss": 1.0335, "step": 24855 }, { "epoch": 0.06, "learning_rate": 9.399869142885904e-05, "loss": 1.0342, "step": 24860 }, { "epoch": 0.06, "learning_rate": 9.399743318737733e-05, "loss": 1.0326, "step": 24865 }, { "epoch": 0.06, "learning_rate": 9.399617494589562e-05, "loss": 1.0337, "step": 24870 }, { "epoch": 0.06, "learning_rate": 9.399491670441391e-05, "loss": 1.0314, "step": 24875 }, { "epoch": 0.06, "learning_rate": 9.399365846293222e-05, "loss": 1.0337, "step": 24880 }, { "epoch": 0.06, "learning_rate": 9.399240022145051e-05, "loss": 1.0335, "step": 24885 }, { "epoch": 0.06, "learning_rate": 9.39911419799688e-05, "loss": 1.0303, "step": 24890 }, { "epoch": 0.06, "learning_rate": 9.398988373848709e-05, "loss": 1.0328, "step": 24895 }, { "epoch": 0.06, "learning_rate": 9.398862549700538e-05, "loss": 1.0321, "step": 24900 }, { "epoch": 0.06, "learning_rate": 9.398736725552369e-05, "loss": 1.032, "step": 24905 }, { "epoch": 0.06, "learning_rate": 9.398610901404198e-05, "loss": 1.0331, "step": 24910 }, { "epoch": 0.06, "learning_rate": 9.398485077256027e-05, "loss": 1.0339, "step": 24915 }, { "epoch": 0.06, "learning_rate": 9.398359253107856e-05, "loss": 1.0332, "step": 24920 }, { "epoch": 0.06, "learning_rate": 9.398233428959687e-05, "loss": 1.0326, "step": 24925 }, { "epoch": 0.06, "learning_rate": 9.398107604811516e-05, "loss": 1.0324, "step": 24930 }, { "epoch": 0.06, "learning_rate": 9.397981780663345e-05, "loss": 1.0325, "step": 24935 }, { "epoch": 0.06, "learning_rate": 9.397855956515174e-05, "loss": 1.031, "step": 24940 }, { "epoch": 0.06, "learning_rate": 9.397730132367005e-05, "loss": 1.0331, "step": 24945 }, { "epoch": 0.06, "learning_rate": 9.397604308218834e-05, "loss": 1.0326, "step": 24950 }, { "epoch": 0.06, "learning_rate": 9.397478484070663e-05, "loss": 1.0338, "step": 24955 }, { "epoch": 0.06, "learning_rate": 9.397352659922492e-05, "loss": 1.032, "step": 24960 }, { "epoch": 0.06, "learning_rate": 9.397226835774321e-05, "loss": 1.0329, "step": 24965 }, { "epoch": 0.06, "learning_rate": 9.397101011626152e-05, "loss": 1.0308, "step": 24970 }, { "epoch": 0.06, "learning_rate": 9.396975187477981e-05, "loss": 1.0321, "step": 24975 }, { "epoch": 0.06, "learning_rate": 9.39684936332981e-05, "loss": 1.0332, "step": 24980 }, { "epoch": 0.06, "learning_rate": 9.39672353918164e-05, "loss": 1.0316, "step": 24985 }, { "epoch": 0.06, "learning_rate": 9.39659771503347e-05, "loss": 1.034, "step": 24990 }, { "epoch": 0.06, "learning_rate": 9.396471890885299e-05, "loss": 1.0307, "step": 24995 }, { "epoch": 0.06, "learning_rate": 9.396346066737128e-05, "loss": 1.0343, "step": 25000 }, { "epoch": 0.06, "learning_rate": 9.396220242588957e-05, "loss": 1.035, "step": 25005 }, { "epoch": 0.06, "learning_rate": 9.396094418440788e-05, "loss": 1.0338, "step": 25010 }, { "epoch": 0.06, "learning_rate": 9.395968594292617e-05, "loss": 1.0315, "step": 25015 }, { "epoch": 0.06, "learning_rate": 9.395842770144446e-05, "loss": 1.0323, "step": 25020 }, { "epoch": 0.06, "learning_rate": 9.395716945996275e-05, "loss": 1.0321, "step": 25025 }, { "epoch": 0.06, "learning_rate": 9.395591121848105e-05, "loss": 1.0324, "step": 25030 }, { "epoch": 0.06, "learning_rate": 9.395465297699935e-05, "loss": 1.0346, "step": 25035 }, { "epoch": 0.06, "learning_rate": 9.395339473551764e-05, "loss": 1.0329, "step": 25040 }, { "epoch": 0.06, "learning_rate": 9.395213649403593e-05, "loss": 1.0505, "step": 25045 }, { "epoch": 0.06, "learning_rate": 9.395087825255423e-05, "loss": 1.0315, "step": 25050 }, { "epoch": 0.06, "learning_rate": 9.394962001107253e-05, "loss": 1.0336, "step": 25055 }, { "epoch": 0.06, "learning_rate": 9.394836176959082e-05, "loss": 1.0311, "step": 25060 }, { "epoch": 0.06, "learning_rate": 9.394710352810911e-05, "loss": 1.0321, "step": 25065 }, { "epoch": 0.06, "learning_rate": 9.39458452866274e-05, "loss": 1.0311, "step": 25070 }, { "epoch": 0.06, "learning_rate": 9.394458704514571e-05, "loss": 1.0308, "step": 25075 }, { "epoch": 0.06, "learning_rate": 9.3943328803664e-05, "loss": 1.0324, "step": 25080 }, { "epoch": 0.06, "learning_rate": 9.39420705621823e-05, "loss": 1.0356, "step": 25085 }, { "epoch": 0.06, "learning_rate": 9.394081232070059e-05, "loss": 1.0315, "step": 25090 }, { "epoch": 0.06, "learning_rate": 9.393955407921888e-05, "loss": 1.0333, "step": 25095 }, { "epoch": 0.06, "learning_rate": 9.393829583773718e-05, "loss": 1.0313, "step": 25100 }, { "epoch": 0.06, "learning_rate": 9.393703759625547e-05, "loss": 1.0305, "step": 25105 }, { "epoch": 0.06, "learning_rate": 9.393577935477377e-05, "loss": 1.0317, "step": 25110 }, { "epoch": 0.06, "learning_rate": 9.393452111329206e-05, "loss": 1.0347, "step": 25115 }, { "epoch": 0.06, "learning_rate": 9.393326287181036e-05, "loss": 1.0296, "step": 25120 }, { "epoch": 0.06, "learning_rate": 9.393200463032867e-05, "loss": 1.0343, "step": 25125 }, { "epoch": 0.06, "learning_rate": 9.393074638884696e-05, "loss": 1.0321, "step": 25130 }, { "epoch": 0.06, "learning_rate": 9.392948814736525e-05, "loss": 1.0317, "step": 25135 }, { "epoch": 0.06, "learning_rate": 9.392822990588354e-05, "loss": 1.032, "step": 25140 }, { "epoch": 0.06, "learning_rate": 9.392697166440185e-05, "loss": 1.0308, "step": 25145 }, { "epoch": 0.06, "learning_rate": 9.392571342292014e-05, "loss": 1.0347, "step": 25150 }, { "epoch": 0.06, "learning_rate": 9.392445518143843e-05, "loss": 1.0322, "step": 25155 }, { "epoch": 0.06, "learning_rate": 9.392319693995672e-05, "loss": 1.0342, "step": 25160 }, { "epoch": 0.06, "learning_rate": 9.392193869847503e-05, "loss": 1.0312, "step": 25165 }, { "epoch": 0.06, "learning_rate": 9.392068045699332e-05, "loss": 1.0312, "step": 25170 }, { "epoch": 0.06, "learning_rate": 9.391942221551161e-05, "loss": 1.0313, "step": 25175 }, { "epoch": 0.06, "learning_rate": 9.39181639740299e-05, "loss": 1.0314, "step": 25180 }, { "epoch": 0.06, "learning_rate": 9.39169057325482e-05, "loss": 1.0315, "step": 25185 }, { "epoch": 0.06, "learning_rate": 9.39156474910665e-05, "loss": 1.0342, "step": 25190 }, { "epoch": 0.06, "learning_rate": 9.391438924958479e-05, "loss": 1.0514, "step": 25195 }, { "epoch": 0.06, "learning_rate": 9.391313100810308e-05, "loss": 1.0311, "step": 25200 }, { "epoch": 0.06, "learning_rate": 9.391187276662137e-05, "loss": 1.0337, "step": 25205 }, { "epoch": 0.06, "learning_rate": 9.391061452513968e-05, "loss": 1.032, "step": 25210 }, { "epoch": 0.06, "learning_rate": 9.390935628365797e-05, "loss": 1.0343, "step": 25215 }, { "epoch": 0.06, "learning_rate": 9.390809804217626e-05, "loss": 1.0322, "step": 25220 }, { "epoch": 0.06, "learning_rate": 9.390683980069455e-05, "loss": 1.0338, "step": 25225 }, { "epoch": 0.06, "learning_rate": 9.390558155921286e-05, "loss": 1.0343, "step": 25230 }, { "epoch": 0.06, "learning_rate": 9.390432331773115e-05, "loss": 1.0317, "step": 25235 }, { "epoch": 0.06, "learning_rate": 9.390306507624944e-05, "loss": 1.0328, "step": 25240 }, { "epoch": 0.06, "learning_rate": 9.390180683476773e-05, "loss": 1.0332, "step": 25245 }, { "epoch": 0.06, "learning_rate": 9.390054859328602e-05, "loss": 1.0328, "step": 25250 }, { "epoch": 0.06, "learning_rate": 9.389929035180433e-05, "loss": 1.0347, "step": 25255 }, { "epoch": 0.06, "learning_rate": 9.389803211032262e-05, "loss": 1.0311, "step": 25260 }, { "epoch": 0.06, "learning_rate": 9.389677386884091e-05, "loss": 1.0323, "step": 25265 }, { "epoch": 0.06, "learning_rate": 9.38955156273592e-05, "loss": 1.0334, "step": 25270 }, { "epoch": 0.06, "learning_rate": 9.389425738587751e-05, "loss": 1.0322, "step": 25275 }, { "epoch": 0.06, "learning_rate": 9.38929991443958e-05, "loss": 1.0327, "step": 25280 }, { "epoch": 0.06, "learning_rate": 9.389174090291409e-05, "loss": 1.0333, "step": 25285 }, { "epoch": 0.06, "learning_rate": 9.389048266143238e-05, "loss": 1.0318, "step": 25290 }, { "epoch": 0.06, "learning_rate": 9.388922441995069e-05, "loss": 1.0299, "step": 25295 }, { "epoch": 0.06, "learning_rate": 9.388796617846898e-05, "loss": 1.0324, "step": 25300 }, { "epoch": 0.06, "learning_rate": 9.388670793698727e-05, "loss": 1.0311, "step": 25305 }, { "epoch": 0.06, "learning_rate": 9.388544969550556e-05, "loss": 1.033, "step": 25310 }, { "epoch": 0.06, "learning_rate": 9.388419145402386e-05, "loss": 1.0343, "step": 25315 }, { "epoch": 0.06, "learning_rate": 9.388293321254216e-05, "loss": 1.0334, "step": 25320 }, { "epoch": 0.06, "learning_rate": 9.388167497106045e-05, "loss": 1.0525, "step": 25325 }, { "epoch": 0.06, "learning_rate": 9.388041672957874e-05, "loss": 1.0313, "step": 25330 }, { "epoch": 0.06, "learning_rate": 9.387915848809704e-05, "loss": 1.0333, "step": 25335 }, { "epoch": 0.06, "learning_rate": 9.387790024661534e-05, "loss": 1.0321, "step": 25340 }, { "epoch": 0.06, "learning_rate": 9.387664200513363e-05, "loss": 1.0315, "step": 25345 }, { "epoch": 0.06, "learning_rate": 9.387538376365192e-05, "loss": 1.0358, "step": 25350 }, { "epoch": 0.06, "learning_rate": 9.387412552217022e-05, "loss": 1.0349, "step": 25355 }, { "epoch": 0.06, "learning_rate": 9.387286728068852e-05, "loss": 1.0304, "step": 25360 }, { "epoch": 0.06, "learning_rate": 9.387160903920681e-05, "loss": 1.0307, "step": 25365 }, { "epoch": 0.06, "learning_rate": 9.38703507977251e-05, "loss": 1.0316, "step": 25370 }, { "epoch": 0.06, "learning_rate": 9.38690925562434e-05, "loss": 1.0321, "step": 25375 }, { "epoch": 0.06, "learning_rate": 9.386783431476169e-05, "loss": 1.0324, "step": 25380 }, { "epoch": 0.06, "learning_rate": 9.386657607327999e-05, "loss": 1.0301, "step": 25385 }, { "epoch": 0.06, "learning_rate": 9.386531783179828e-05, "loss": 1.0326, "step": 25390 }, { "epoch": 0.06, "learning_rate": 9.386405959031658e-05, "loss": 1.0331, "step": 25395 }, { "epoch": 0.06, "learning_rate": 9.386280134883487e-05, "loss": 1.0325, "step": 25400 }, { "epoch": 0.06, "learning_rate": 9.386154310735317e-05, "loss": 1.0319, "step": 25405 }, { "epoch": 0.06, "learning_rate": 9.386028486587146e-05, "loss": 1.0312, "step": 25410 }, { "epoch": 0.06, "learning_rate": 9.385902662438976e-05, "loss": 1.0328, "step": 25415 }, { "epoch": 0.06, "learning_rate": 9.385776838290805e-05, "loss": 1.0325, "step": 25420 }, { "epoch": 0.06, "learning_rate": 9.385651014142635e-05, "loss": 1.0335, "step": 25425 }, { "epoch": 0.06, "learning_rate": 9.385525189994464e-05, "loss": 1.0329, "step": 25430 }, { "epoch": 0.06, "learning_rate": 9.385399365846294e-05, "loss": 1.031, "step": 25435 }, { "epoch": 0.06, "learning_rate": 9.385273541698123e-05, "loss": 1.03, "step": 25440 }, { "epoch": 0.06, "learning_rate": 9.385147717549952e-05, "loss": 1.034, "step": 25445 }, { "epoch": 0.06, "learning_rate": 9.385021893401782e-05, "loss": 1.0318, "step": 25450 }, { "epoch": 0.06, "learning_rate": 9.384896069253612e-05, "loss": 1.0334, "step": 25455 }, { "epoch": 0.06, "learning_rate": 9.384770245105441e-05, "loss": 1.0327, "step": 25460 }, { "epoch": 0.06, "learning_rate": 9.38464442095727e-05, "loss": 1.0301, "step": 25465 }, { "epoch": 0.06, "learning_rate": 9.3845185968091e-05, "loss": 1.0304, "step": 25470 }, { "epoch": 0.06, "learning_rate": 9.38439277266093e-05, "loss": 1.0339, "step": 25475 }, { "epoch": 0.06, "learning_rate": 9.384266948512759e-05, "loss": 1.031, "step": 25480 }, { "epoch": 0.06, "learning_rate": 9.384141124364588e-05, "loss": 1.033, "step": 25485 }, { "epoch": 0.06, "learning_rate": 9.384015300216418e-05, "loss": 1.0308, "step": 25490 }, { "epoch": 0.06, "learning_rate": 9.383889476068248e-05, "loss": 1.0317, "step": 25495 }, { "epoch": 0.06, "learning_rate": 9.383763651920077e-05, "loss": 1.0316, "step": 25500 }, { "epoch": 0.06, "learning_rate": 9.383637827771906e-05, "loss": 1.0327, "step": 25505 }, { "epoch": 0.06, "learning_rate": 9.383512003623735e-05, "loss": 1.0312, "step": 25510 }, { "epoch": 0.06, "learning_rate": 9.383386179475566e-05, "loss": 1.0322, "step": 25515 }, { "epoch": 0.06, "learning_rate": 9.383260355327395e-05, "loss": 1.0314, "step": 25520 }, { "epoch": 0.06, "learning_rate": 9.383134531179224e-05, "loss": 1.0327, "step": 25525 }, { "epoch": 0.06, "learning_rate": 9.383008707031053e-05, "loss": 1.0342, "step": 25530 }, { "epoch": 0.06, "learning_rate": 9.382882882882884e-05, "loss": 1.0321, "step": 25535 }, { "epoch": 0.06, "learning_rate": 9.382757058734713e-05, "loss": 1.0297, "step": 25540 }, { "epoch": 0.06, "learning_rate": 9.382631234586542e-05, "loss": 1.033, "step": 25545 }, { "epoch": 0.06, "learning_rate": 9.382505410438371e-05, "loss": 1.0314, "step": 25550 }, { "epoch": 0.06, "learning_rate": 9.3823795862902e-05, "loss": 1.034, "step": 25555 }, { "epoch": 0.06, "learning_rate": 9.382253762142031e-05, "loss": 1.0318, "step": 25560 }, { "epoch": 0.06, "learning_rate": 9.38212793799386e-05, "loss": 1.0313, "step": 25565 }, { "epoch": 0.06, "learning_rate": 9.382002113845689e-05, "loss": 1.0307, "step": 25570 }, { "epoch": 0.06, "learning_rate": 9.381876289697518e-05, "loss": 1.0324, "step": 25575 }, { "epoch": 0.06, "learning_rate": 9.381750465549349e-05, "loss": 1.0326, "step": 25580 }, { "epoch": 0.06, "learning_rate": 9.381624641401178e-05, "loss": 1.0342, "step": 25585 }, { "epoch": 0.06, "learning_rate": 9.381498817253007e-05, "loss": 1.0324, "step": 25590 }, { "epoch": 0.06, "learning_rate": 9.381372993104836e-05, "loss": 1.0288, "step": 25595 }, { "epoch": 0.06, "learning_rate": 9.381247168956667e-05, "loss": 1.0316, "step": 25600 }, { "epoch": 0.06, "learning_rate": 9.381121344808496e-05, "loss": 1.0319, "step": 25605 }, { "epoch": 0.06, "learning_rate": 9.380995520660325e-05, "loss": 1.0334, "step": 25610 }, { "epoch": 0.06, "learning_rate": 9.380869696512154e-05, "loss": 1.0299, "step": 25615 }, { "epoch": 0.06, "learning_rate": 9.380743872363983e-05, "loss": 1.0309, "step": 25620 }, { "epoch": 0.06, "learning_rate": 9.380618048215815e-05, "loss": 1.0338, "step": 25625 }, { "epoch": 0.06, "learning_rate": 9.380492224067644e-05, "loss": 1.0323, "step": 25630 }, { "epoch": 0.06, "learning_rate": 9.380366399919474e-05, "loss": 1.0335, "step": 25635 }, { "epoch": 0.06, "learning_rate": 9.380240575771303e-05, "loss": 1.0306, "step": 25640 }, { "epoch": 0.06, "learning_rate": 9.380114751623132e-05, "loss": 1.0323, "step": 25645 }, { "epoch": 0.06, "learning_rate": 9.379988927474962e-05, "loss": 1.0308, "step": 25650 }, { "epoch": 0.06, "learning_rate": 9.379863103326792e-05, "loss": 1.0345, "step": 25655 }, { "epoch": 0.06, "learning_rate": 9.37973727917862e-05, "loss": 1.0313, "step": 25660 }, { "epoch": 0.06, "learning_rate": 9.37961145503045e-05, "loss": 1.0311, "step": 25665 }, { "epoch": 0.06, "learning_rate": 9.37948563088228e-05, "loss": 1.0328, "step": 25670 }, { "epoch": 0.06, "learning_rate": 9.37935980673411e-05, "loss": 1.0303, "step": 25675 }, { "epoch": 0.06, "learning_rate": 9.379233982585939e-05, "loss": 1.033, "step": 25680 }, { "epoch": 0.06, "learning_rate": 9.379108158437768e-05, "loss": 1.0319, "step": 25685 }, { "epoch": 0.06, "learning_rate": 9.378982334289598e-05, "loss": 1.0333, "step": 25690 }, { "epoch": 0.06, "learning_rate": 9.378856510141428e-05, "loss": 1.033, "step": 25695 }, { "epoch": 0.06, "learning_rate": 9.378730685993257e-05, "loss": 1.0355, "step": 25700 }, { "epoch": 0.06, "learning_rate": 9.378604861845086e-05, "loss": 1.033, "step": 25705 }, { "epoch": 0.06, "learning_rate": 9.378479037696915e-05, "loss": 1.0322, "step": 25710 }, { "epoch": 0.06, "learning_rate": 9.378353213548745e-05, "loss": 1.0313, "step": 25715 }, { "epoch": 0.06, "learning_rate": 9.378227389400575e-05, "loss": 1.0314, "step": 25720 }, { "epoch": 0.06, "learning_rate": 9.378101565252404e-05, "loss": 1.0322, "step": 25725 }, { "epoch": 0.06, "learning_rate": 9.377975741104233e-05, "loss": 1.0303, "step": 25730 }, { "epoch": 0.06, "learning_rate": 9.377849916956063e-05, "loss": 1.0313, "step": 25735 }, { "epoch": 0.06, "learning_rate": 9.377724092807893e-05, "loss": 1.0331, "step": 25740 }, { "epoch": 0.06, "learning_rate": 9.377598268659722e-05, "loss": 1.0318, "step": 25745 }, { "epoch": 0.06, "learning_rate": 9.377472444511551e-05, "loss": 1.0328, "step": 25750 }, { "epoch": 0.06, "learning_rate": 9.377346620363381e-05, "loss": 1.0321, "step": 25755 }, { "epoch": 0.06, "learning_rate": 9.37722079621521e-05, "loss": 1.0335, "step": 25760 }, { "epoch": 0.06, "learning_rate": 9.37709497206704e-05, "loss": 1.033, "step": 25765 }, { "epoch": 0.06, "learning_rate": 9.376969147918869e-05, "loss": 1.0332, "step": 25770 }, { "epoch": 0.06, "learning_rate": 9.376843323770698e-05, "loss": 1.032, "step": 25775 }, { "epoch": 0.06, "learning_rate": 9.376717499622529e-05, "loss": 1.0326, "step": 25780 }, { "epoch": 0.06, "learning_rate": 9.376591675474358e-05, "loss": 1.0322, "step": 25785 }, { "epoch": 0.06, "learning_rate": 9.376465851326187e-05, "loss": 1.0317, "step": 25790 }, { "epoch": 0.06, "learning_rate": 9.376340027178016e-05, "loss": 1.0321, "step": 25795 }, { "epoch": 0.06, "learning_rate": 9.376214203029847e-05, "loss": 1.0298, "step": 25800 }, { "epoch": 0.06, "learning_rate": 9.376088378881676e-05, "loss": 1.0336, "step": 25805 }, { "epoch": 0.06, "learning_rate": 9.375962554733505e-05, "loss": 1.0338, "step": 25810 }, { "epoch": 0.06, "learning_rate": 9.375836730585334e-05, "loss": 1.0317, "step": 25815 }, { "epoch": 0.06, "learning_rate": 9.375710906437165e-05, "loss": 1.0351, "step": 25820 }, { "epoch": 0.06, "learning_rate": 9.375585082288994e-05, "loss": 1.0316, "step": 25825 }, { "epoch": 0.06, "learning_rate": 9.375459258140823e-05, "loss": 1.0319, "step": 25830 }, { "epoch": 0.06, "learning_rate": 9.375333433992652e-05, "loss": 1.033, "step": 25835 }, { "epoch": 0.06, "learning_rate": 9.375207609844481e-05, "loss": 1.0309, "step": 25840 }, { "epoch": 0.06, "learning_rate": 9.375081785696312e-05, "loss": 1.033, "step": 25845 }, { "epoch": 0.06, "learning_rate": 9.374955961548141e-05, "loss": 1.0318, "step": 25850 }, { "epoch": 0.06, "learning_rate": 9.37483013739997e-05, "loss": 1.0322, "step": 25855 }, { "epoch": 0.06, "learning_rate": 9.374704313251799e-05, "loss": 1.0324, "step": 25860 }, { "epoch": 0.06, "learning_rate": 9.37457848910363e-05, "loss": 1.0313, "step": 25865 }, { "epoch": 0.06, "learning_rate": 9.374452664955459e-05, "loss": 1.0323, "step": 25870 }, { "epoch": 0.06, "learning_rate": 9.374326840807288e-05, "loss": 1.032, "step": 25875 }, { "epoch": 0.06, "learning_rate": 9.374201016659117e-05, "loss": 1.0325, "step": 25880 }, { "epoch": 0.06, "learning_rate": 9.374075192510948e-05, "loss": 1.0316, "step": 25885 }, { "epoch": 0.06, "learning_rate": 9.373949368362777e-05, "loss": 1.0333, "step": 25890 }, { "epoch": 0.07, "learning_rate": 9.373823544214606e-05, "loss": 1.0327, "step": 25895 }, { "epoch": 0.07, "learning_rate": 9.373697720066435e-05, "loss": 1.0336, "step": 25900 }, { "epoch": 0.07, "learning_rate": 9.373571895918264e-05, "loss": 1.0336, "step": 25905 }, { "epoch": 0.07, "learning_rate": 9.373446071770095e-05, "loss": 1.0318, "step": 25910 }, { "epoch": 0.07, "learning_rate": 9.373320247621924e-05, "loss": 1.0316, "step": 25915 }, { "epoch": 0.07, "learning_rate": 9.373194423473753e-05, "loss": 1.0314, "step": 25920 }, { "epoch": 0.07, "learning_rate": 9.373068599325582e-05, "loss": 1.0309, "step": 25925 }, { "epoch": 0.07, "learning_rate": 9.372942775177413e-05, "loss": 1.032, "step": 25930 }, { "epoch": 0.07, "learning_rate": 9.372816951029242e-05, "loss": 1.0322, "step": 25935 }, { "epoch": 0.07, "learning_rate": 9.372691126881071e-05, "loss": 1.0342, "step": 25940 }, { "epoch": 0.07, "learning_rate": 9.3725653027329e-05, "loss": 1.0317, "step": 25945 }, { "epoch": 0.07, "learning_rate": 9.372439478584731e-05, "loss": 1.0326, "step": 25950 }, { "epoch": 0.07, "learning_rate": 9.37231365443656e-05, "loss": 1.0337, "step": 25955 }, { "epoch": 0.07, "learning_rate": 9.372187830288389e-05, "loss": 1.0306, "step": 25960 }, { "epoch": 0.07, "learning_rate": 9.372062006140218e-05, "loss": 1.0313, "step": 25965 }, { "epoch": 0.07, "learning_rate": 9.371936181992048e-05, "loss": 1.0328, "step": 25970 }, { "epoch": 0.07, "learning_rate": 9.371810357843878e-05, "loss": 1.0342, "step": 25975 }, { "epoch": 0.07, "learning_rate": 9.371684533695707e-05, "loss": 1.0369, "step": 25980 }, { "epoch": 0.07, "learning_rate": 9.371558709547536e-05, "loss": 1.0322, "step": 25985 }, { "epoch": 0.07, "learning_rate": 9.371432885399366e-05, "loss": 1.0309, "step": 25990 }, { "epoch": 0.07, "learning_rate": 9.371307061251196e-05, "loss": 1.0325, "step": 25995 }, { "epoch": 0.07, "learning_rate": 9.371181237103025e-05, "loss": 1.0325, "step": 26000 }, { "epoch": 0.07, "learning_rate": 9.371055412954854e-05, "loss": 1.0327, "step": 26005 }, { "epoch": 0.07, "learning_rate": 9.370929588806684e-05, "loss": 1.0313, "step": 26010 }, { "epoch": 0.07, "learning_rate": 9.370803764658514e-05, "loss": 1.0319, "step": 26015 }, { "epoch": 0.07, "learning_rate": 9.370677940510343e-05, "loss": 1.0317, "step": 26020 }, { "epoch": 0.07, "learning_rate": 9.370552116362172e-05, "loss": 1.0325, "step": 26025 }, { "epoch": 0.07, "learning_rate": 9.370426292214002e-05, "loss": 1.0567, "step": 26030 }, { "epoch": 0.07, "learning_rate": 9.370300468065831e-05, "loss": 1.0345, "step": 26035 }, { "epoch": 0.07, "learning_rate": 9.370174643917661e-05, "loss": 1.0316, "step": 26040 }, { "epoch": 0.07, "learning_rate": 9.37004881976949e-05, "loss": 1.0329, "step": 26045 }, { "epoch": 0.07, "learning_rate": 9.36992299562132e-05, "loss": 1.0323, "step": 26050 }, { "epoch": 0.07, "learning_rate": 9.369797171473149e-05, "loss": 1.031, "step": 26055 }, { "epoch": 0.07, "learning_rate": 9.369671347324979e-05, "loss": 1.0347, "step": 26060 }, { "epoch": 0.07, "learning_rate": 9.369545523176808e-05, "loss": 1.0309, "step": 26065 }, { "epoch": 0.07, "learning_rate": 9.369419699028638e-05, "loss": 1.0319, "step": 26070 }, { "epoch": 0.07, "learning_rate": 9.369293874880467e-05, "loss": 1.0318, "step": 26075 }, { "epoch": 0.07, "learning_rate": 9.369168050732297e-05, "loss": 1.0342, "step": 26080 }, { "epoch": 0.07, "learning_rate": 9.369042226584126e-05, "loss": 1.0341, "step": 26085 }, { "epoch": 0.07, "learning_rate": 9.368916402435956e-05, "loss": 1.0299, "step": 26090 }, { "epoch": 0.07, "learning_rate": 9.368790578287785e-05, "loss": 1.0345, "step": 26095 }, { "epoch": 0.07, "learning_rate": 9.368664754139614e-05, "loss": 1.0339, "step": 26100 }, { "epoch": 0.07, "learning_rate": 9.368538929991444e-05, "loss": 1.031, "step": 26105 }, { "epoch": 0.07, "learning_rate": 9.368413105843274e-05, "loss": 1.0327, "step": 26110 }, { "epoch": 0.07, "learning_rate": 9.368287281695103e-05, "loss": 1.0316, "step": 26115 }, { "epoch": 0.07, "learning_rate": 9.368161457546932e-05, "loss": 1.0337, "step": 26120 }, { "epoch": 0.07, "learning_rate": 9.368035633398762e-05, "loss": 1.0315, "step": 26125 }, { "epoch": 0.07, "learning_rate": 9.367909809250593e-05, "loss": 1.0326, "step": 26130 }, { "epoch": 0.07, "learning_rate": 9.367783985102422e-05, "loss": 1.0318, "step": 26135 }, { "epoch": 0.07, "learning_rate": 9.367658160954251e-05, "loss": 1.0342, "step": 26140 }, { "epoch": 0.07, "learning_rate": 9.36753233680608e-05, "loss": 1.0328, "step": 26145 }, { "epoch": 0.07, "learning_rate": 9.367406512657911e-05, "loss": 1.0314, "step": 26150 }, { "epoch": 0.07, "learning_rate": 9.36728068850974e-05, "loss": 1.0323, "step": 26155 }, { "epoch": 0.07, "learning_rate": 9.367154864361569e-05, "loss": 1.0359, "step": 26160 }, { "epoch": 0.07, "learning_rate": 9.367029040213398e-05, "loss": 1.0325, "step": 26165 }, { "epoch": 0.07, "learning_rate": 9.366903216065227e-05, "loss": 1.0304, "step": 26170 }, { "epoch": 0.07, "learning_rate": 9.366777391917058e-05, "loss": 1.0319, "step": 26175 }, { "epoch": 0.07, "learning_rate": 9.366651567768887e-05, "loss": 1.033, "step": 26180 }, { "epoch": 0.07, "learning_rate": 9.366525743620716e-05, "loss": 1.0333, "step": 26185 }, { "epoch": 0.07, "learning_rate": 9.366399919472545e-05, "loss": 1.049, "step": 26190 }, { "epoch": 0.07, "learning_rate": 9.366274095324376e-05, "loss": 1.0324, "step": 26195 }, { "epoch": 0.07, "learning_rate": 9.366148271176205e-05, "loss": 1.0338, "step": 26200 }, { "epoch": 0.07, "learning_rate": 9.366022447028034e-05, "loss": 1.032, "step": 26205 }, { "epoch": 0.07, "learning_rate": 9.365896622879863e-05, "loss": 1.0314, "step": 26210 }, { "epoch": 0.07, "learning_rate": 9.365770798731694e-05, "loss": 1.0306, "step": 26215 }, { "epoch": 0.07, "learning_rate": 9.365644974583523e-05, "loss": 1.0312, "step": 26220 }, { "epoch": 0.07, "learning_rate": 9.365519150435352e-05, "loss": 1.0337, "step": 26225 }, { "epoch": 0.07, "learning_rate": 9.365393326287181e-05, "loss": 1.0536, "step": 26230 }, { "epoch": 0.07, "learning_rate": 9.36526750213901e-05, "loss": 1.0315, "step": 26235 }, { "epoch": 0.07, "learning_rate": 9.365141677990841e-05, "loss": 1.0301, "step": 26240 }, { "epoch": 0.07, "learning_rate": 9.36501585384267e-05, "loss": 1.032, "step": 26245 }, { "epoch": 0.07, "learning_rate": 9.3648900296945e-05, "loss": 1.0329, "step": 26250 }, { "epoch": 0.07, "learning_rate": 9.364764205546329e-05, "loss": 1.0344, "step": 26255 }, { "epoch": 0.07, "learning_rate": 9.364638381398159e-05, "loss": 1.0307, "step": 26260 }, { "epoch": 0.07, "learning_rate": 9.364512557249988e-05, "loss": 1.0345, "step": 26265 }, { "epoch": 0.07, "learning_rate": 9.364386733101817e-05, "loss": 1.0318, "step": 26270 }, { "epoch": 0.07, "learning_rate": 9.364260908953647e-05, "loss": 1.0332, "step": 26275 }, { "epoch": 0.07, "learning_rate": 9.364135084805477e-05, "loss": 1.032, "step": 26280 }, { "epoch": 0.07, "learning_rate": 9.364009260657306e-05, "loss": 1.0339, "step": 26285 }, { "epoch": 0.07, "learning_rate": 9.363883436509135e-05, "loss": 1.034, "step": 26290 }, { "epoch": 0.07, "learning_rate": 9.363757612360965e-05, "loss": 1.0323, "step": 26295 }, { "epoch": 0.07, "learning_rate": 9.363631788212794e-05, "loss": 1.0336, "step": 26300 }, { "epoch": 0.07, "learning_rate": 9.363505964064624e-05, "loss": 1.0299, "step": 26305 }, { "epoch": 0.07, "learning_rate": 9.363380139916453e-05, "loss": 1.0311, "step": 26310 }, { "epoch": 0.07, "learning_rate": 9.363254315768283e-05, "loss": 1.0357, "step": 26315 }, { "epoch": 0.07, "learning_rate": 9.363128491620112e-05, "loss": 1.0339, "step": 26320 }, { "epoch": 0.07, "learning_rate": 9.363002667471942e-05, "loss": 1.0324, "step": 26325 }, { "epoch": 0.07, "learning_rate": 9.362876843323771e-05, "loss": 1.0323, "step": 26330 }, { "epoch": 0.07, "learning_rate": 9.3627510191756e-05, "loss": 1.0316, "step": 26335 }, { "epoch": 0.07, "learning_rate": 9.36262519502743e-05, "loss": 1.0344, "step": 26340 }, { "epoch": 0.07, "learning_rate": 9.36249937087926e-05, "loss": 1.0316, "step": 26345 }, { "epoch": 0.07, "learning_rate": 9.36237354673109e-05, "loss": 1.0336, "step": 26350 }, { "epoch": 0.07, "learning_rate": 9.362247722582919e-05, "loss": 1.0316, "step": 26355 }, { "epoch": 0.07, "learning_rate": 9.362121898434748e-05, "loss": 1.0349, "step": 26360 }, { "epoch": 0.07, "learning_rate": 9.361996074286577e-05, "loss": 1.0329, "step": 26365 }, { "epoch": 0.07, "learning_rate": 9.361870250138407e-05, "loss": 1.0335, "step": 26370 }, { "epoch": 0.07, "learning_rate": 9.361744425990237e-05, "loss": 1.032, "step": 26375 }, { "epoch": 0.07, "learning_rate": 9.361618601842066e-05, "loss": 1.0327, "step": 26380 }, { "epoch": 0.07, "learning_rate": 9.361492777693895e-05, "loss": 1.0329, "step": 26385 }, { "epoch": 0.07, "learning_rate": 9.361366953545725e-05, "loss": 1.0336, "step": 26390 }, { "epoch": 0.07, "learning_rate": 9.361241129397555e-05, "loss": 1.0322, "step": 26395 }, { "epoch": 0.07, "learning_rate": 9.361115305249384e-05, "loss": 1.0329, "step": 26400 }, { "epoch": 0.07, "learning_rate": 9.360989481101213e-05, "loss": 1.0342, "step": 26405 }, { "epoch": 0.07, "learning_rate": 9.360863656953043e-05, "loss": 1.0321, "step": 26410 }, { "epoch": 0.07, "learning_rate": 9.360737832804873e-05, "loss": 1.0311, "step": 26415 }, { "epoch": 0.07, "learning_rate": 9.360612008656702e-05, "loss": 1.0312, "step": 26420 }, { "epoch": 0.07, "learning_rate": 9.360486184508531e-05, "loss": 1.0321, "step": 26425 }, { "epoch": 0.07, "learning_rate": 9.36036036036036e-05, "loss": 1.0331, "step": 26430 }, { "epoch": 0.07, "learning_rate": 9.36023453621219e-05, "loss": 1.031, "step": 26435 }, { "epoch": 0.07, "learning_rate": 9.36010871206402e-05, "loss": 1.0309, "step": 26440 }, { "epoch": 0.07, "learning_rate": 9.359982887915849e-05, "loss": 1.0333, "step": 26445 }, { "epoch": 0.07, "learning_rate": 9.359857063767678e-05, "loss": 1.0318, "step": 26450 }, { "epoch": 0.07, "learning_rate": 9.359731239619509e-05, "loss": 1.0315, "step": 26455 }, { "epoch": 0.07, "learning_rate": 9.359605415471338e-05, "loss": 1.0301, "step": 26460 }, { "epoch": 0.07, "learning_rate": 9.359479591323167e-05, "loss": 1.0328, "step": 26465 }, { "epoch": 0.07, "learning_rate": 9.359353767174996e-05, "loss": 1.0303, "step": 26470 }, { "epoch": 0.07, "learning_rate": 9.359227943026827e-05, "loss": 1.0312, "step": 26475 }, { "epoch": 0.07, "learning_rate": 9.359102118878656e-05, "loss": 1.0334, "step": 26480 }, { "epoch": 0.07, "learning_rate": 9.358976294730485e-05, "loss": 1.0338, "step": 26485 }, { "epoch": 0.07, "learning_rate": 9.358850470582314e-05, "loss": 1.0331, "step": 26490 }, { "epoch": 0.07, "learning_rate": 9.358724646434143e-05, "loss": 1.0318, "step": 26495 }, { "epoch": 0.07, "learning_rate": 9.358598822285974e-05, "loss": 1.0329, "step": 26500 }, { "epoch": 0.07, "learning_rate": 9.358472998137803e-05, "loss": 1.0322, "step": 26505 }, { "epoch": 0.07, "learning_rate": 9.358347173989632e-05, "loss": 1.0316, "step": 26510 }, { "epoch": 0.07, "learning_rate": 9.358221349841461e-05, "loss": 1.033, "step": 26515 }, { "epoch": 0.07, "learning_rate": 9.358095525693292e-05, "loss": 1.0338, "step": 26520 }, { "epoch": 0.07, "learning_rate": 9.357969701545121e-05, "loss": 1.0317, "step": 26525 }, { "epoch": 0.07, "learning_rate": 9.35784387739695e-05, "loss": 1.0322, "step": 26530 }, { "epoch": 0.07, "learning_rate": 9.357718053248779e-05, "loss": 1.0316, "step": 26535 }, { "epoch": 0.07, "learning_rate": 9.35759222910061e-05, "loss": 1.0332, "step": 26540 }, { "epoch": 0.07, "learning_rate": 9.357466404952439e-05, "loss": 1.0327, "step": 26545 }, { "epoch": 0.07, "learning_rate": 9.357340580804268e-05, "loss": 1.0325, "step": 26550 }, { "epoch": 0.07, "learning_rate": 9.357214756656097e-05, "loss": 1.0322, "step": 26555 }, { "epoch": 0.07, "learning_rate": 9.357088932507926e-05, "loss": 1.0313, "step": 26560 }, { "epoch": 0.07, "learning_rate": 9.356963108359757e-05, "loss": 1.0338, "step": 26565 }, { "epoch": 0.07, "learning_rate": 9.356837284211586e-05, "loss": 1.0338, "step": 26570 }, { "epoch": 0.07, "learning_rate": 9.356711460063415e-05, "loss": 1.0307, "step": 26575 }, { "epoch": 0.07, "learning_rate": 9.356585635915244e-05, "loss": 1.0327, "step": 26580 }, { "epoch": 0.07, "learning_rate": 9.356459811767075e-05, "loss": 1.0311, "step": 26585 }, { "epoch": 0.07, "learning_rate": 9.356333987618904e-05, "loss": 1.0485, "step": 26590 }, { "epoch": 0.07, "learning_rate": 9.356208163470733e-05, "loss": 1.0347, "step": 26595 }, { "epoch": 0.07, "learning_rate": 9.356082339322562e-05, "loss": 1.0328, "step": 26600 }, { "epoch": 0.07, "learning_rate": 9.355956515174393e-05, "loss": 1.032, "step": 26605 }, { "epoch": 0.07, "learning_rate": 9.355830691026222e-05, "loss": 1.0325, "step": 26610 }, { "epoch": 0.07, "learning_rate": 9.355704866878051e-05, "loss": 1.0485, "step": 26615 }, { "epoch": 0.07, "learning_rate": 9.35557904272988e-05, "loss": 1.0291, "step": 26620 }, { "epoch": 0.07, "learning_rate": 9.355453218581711e-05, "loss": 1.0313, "step": 26625 }, { "epoch": 0.07, "learning_rate": 9.35532739443354e-05, "loss": 1.032, "step": 26630 }, { "epoch": 0.07, "learning_rate": 9.35520157028537e-05, "loss": 1.0321, "step": 26635 }, { "epoch": 0.07, "learning_rate": 9.3550757461372e-05, "loss": 1.0312, "step": 26640 }, { "epoch": 0.07, "learning_rate": 9.354949921989029e-05, "loss": 1.0345, "step": 26645 }, { "epoch": 0.07, "learning_rate": 9.354824097840858e-05, "loss": 1.0349, "step": 26650 }, { "epoch": 0.07, "learning_rate": 9.354698273692689e-05, "loss": 1.032, "step": 26655 }, { "epoch": 0.07, "learning_rate": 9.354572449544518e-05, "loss": 1.0326, "step": 26660 }, { "epoch": 0.07, "learning_rate": 9.354446625396347e-05, "loss": 1.033, "step": 26665 }, { "epoch": 0.07, "learning_rate": 9.354320801248176e-05, "loss": 1.0327, "step": 26670 }, { "epoch": 0.07, "learning_rate": 9.354194977100007e-05, "loss": 1.0304, "step": 26675 }, { "epoch": 0.07, "learning_rate": 9.354069152951836e-05, "loss": 1.0332, "step": 26680 }, { "epoch": 0.07, "learning_rate": 9.353943328803665e-05, "loss": 1.0327, "step": 26685 }, { "epoch": 0.07, "learning_rate": 9.353817504655494e-05, "loss": 1.033, "step": 26690 }, { "epoch": 0.07, "learning_rate": 9.353691680507323e-05, "loss": 1.0312, "step": 26695 }, { "epoch": 0.07, "learning_rate": 9.353565856359154e-05, "loss": 1.0337, "step": 26700 }, { "epoch": 0.07, "learning_rate": 9.353440032210983e-05, "loss": 1.0335, "step": 26705 }, { "epoch": 0.07, "learning_rate": 9.353314208062812e-05, "loss": 1.0311, "step": 26710 }, { "epoch": 0.07, "learning_rate": 9.353188383914641e-05, "loss": 1.0309, "step": 26715 }, { "epoch": 0.07, "learning_rate": 9.353062559766472e-05, "loss": 1.0339, "step": 26720 }, { "epoch": 0.07, "learning_rate": 9.352936735618301e-05, "loss": 1.0334, "step": 26725 }, { "epoch": 0.07, "learning_rate": 9.35281091147013e-05, "loss": 1.0307, "step": 26730 }, { "epoch": 0.07, "learning_rate": 9.352685087321959e-05, "loss": 1.0336, "step": 26735 }, { "epoch": 0.07, "learning_rate": 9.35255926317379e-05, "loss": 1.0323, "step": 26740 }, { "epoch": 0.07, "learning_rate": 9.352433439025619e-05, "loss": 1.0305, "step": 26745 }, { "epoch": 0.07, "learning_rate": 9.352307614877448e-05, "loss": 1.0297, "step": 26750 }, { "epoch": 0.07, "learning_rate": 9.352181790729277e-05, "loss": 1.0338, "step": 26755 }, { "epoch": 0.07, "learning_rate": 9.352055966581106e-05, "loss": 1.0324, "step": 26760 }, { "epoch": 0.07, "learning_rate": 9.351930142432937e-05, "loss": 1.0302, "step": 26765 }, { "epoch": 0.07, "learning_rate": 9.351804318284766e-05, "loss": 1.0316, "step": 26770 }, { "epoch": 0.07, "learning_rate": 9.351678494136595e-05, "loss": 1.0309, "step": 26775 }, { "epoch": 0.07, "learning_rate": 9.351552669988424e-05, "loss": 1.035, "step": 26780 }, { "epoch": 0.07, "learning_rate": 9.351426845840255e-05, "loss": 1.0339, "step": 26785 }, { "epoch": 0.07, "learning_rate": 9.351301021692084e-05, "loss": 1.0323, "step": 26790 }, { "epoch": 0.07, "learning_rate": 9.351175197543913e-05, "loss": 1.0318, "step": 26795 }, { "epoch": 0.07, "learning_rate": 9.351049373395742e-05, "loss": 1.0304, "step": 26800 }, { "epoch": 0.07, "learning_rate": 9.350923549247573e-05, "loss": 1.0323, "step": 26805 }, { "epoch": 0.07, "learning_rate": 9.350797725099402e-05, "loss": 1.0313, "step": 26810 }, { "epoch": 0.07, "learning_rate": 9.350671900951231e-05, "loss": 1.0341, "step": 26815 }, { "epoch": 0.07, "learning_rate": 9.35054607680306e-05, "loss": 1.0332, "step": 26820 }, { "epoch": 0.07, "learning_rate": 9.35042025265489e-05, "loss": 1.0317, "step": 26825 }, { "epoch": 0.07, "learning_rate": 9.35029442850672e-05, "loss": 1.0302, "step": 26830 }, { "epoch": 0.07, "learning_rate": 9.350168604358549e-05, "loss": 1.0323, "step": 26835 }, { "epoch": 0.07, "learning_rate": 9.350042780210378e-05, "loss": 1.0281, "step": 26840 }, { "epoch": 0.07, "learning_rate": 9.349916956062207e-05, "loss": 1.0325, "step": 26845 }, { "epoch": 0.07, "learning_rate": 9.349791131914038e-05, "loss": 1.0317, "step": 26850 }, { "epoch": 0.07, "learning_rate": 9.349665307765867e-05, "loss": 1.0344, "step": 26855 }, { "epoch": 0.07, "learning_rate": 9.349539483617696e-05, "loss": 1.0349, "step": 26860 }, { "epoch": 0.07, "learning_rate": 9.349413659469525e-05, "loss": 1.0307, "step": 26865 }, { "epoch": 0.07, "learning_rate": 9.349287835321356e-05, "loss": 1.0323, "step": 26870 }, { "epoch": 0.07, "learning_rate": 9.349162011173185e-05, "loss": 1.0325, "step": 26875 }, { "epoch": 0.07, "learning_rate": 9.349036187025014e-05, "loss": 1.0302, "step": 26880 }, { "epoch": 0.07, "learning_rate": 9.348910362876843e-05, "loss": 1.0318, "step": 26885 }, { "epoch": 0.07, "learning_rate": 9.348784538728673e-05, "loss": 1.048, "step": 26890 }, { "epoch": 0.07, "learning_rate": 9.348658714580503e-05, "loss": 1.032, "step": 26895 }, { "epoch": 0.07, "learning_rate": 9.348532890432332e-05, "loss": 1.0327, "step": 26900 }, { "epoch": 0.07, "learning_rate": 9.348407066284161e-05, "loss": 1.0331, "step": 26905 }, { "epoch": 0.07, "learning_rate": 9.34828124213599e-05, "loss": 1.0307, "step": 26910 }, { "epoch": 0.07, "learning_rate": 9.348155417987821e-05, "loss": 1.0329, "step": 26915 }, { "epoch": 0.07, "learning_rate": 9.34802959383965e-05, "loss": 1.0321, "step": 26920 }, { "epoch": 0.07, "learning_rate": 9.34790376969148e-05, "loss": 1.0325, "step": 26925 }, { "epoch": 0.07, "learning_rate": 9.347777945543309e-05, "loss": 1.0328, "step": 26930 }, { "epoch": 0.07, "learning_rate": 9.347652121395139e-05, "loss": 1.0348, "step": 26935 }, { "epoch": 0.07, "learning_rate": 9.347526297246968e-05, "loss": 1.0336, "step": 26940 }, { "epoch": 0.07, "learning_rate": 9.347400473098797e-05, "loss": 1.0316, "step": 26945 }, { "epoch": 0.07, "learning_rate": 9.347274648950627e-05, "loss": 1.0324, "step": 26950 }, { "epoch": 0.07, "learning_rate": 9.347148824802456e-05, "loss": 1.0342, "step": 26955 }, { "epoch": 0.07, "learning_rate": 9.347023000654286e-05, "loss": 1.0315, "step": 26960 }, { "epoch": 0.07, "learning_rate": 9.346897176506115e-05, "loss": 1.0325, "step": 26965 }, { "epoch": 0.07, "learning_rate": 9.346771352357945e-05, "loss": 1.032, "step": 26970 }, { "epoch": 0.07, "learning_rate": 9.346645528209774e-05, "loss": 1.0318, "step": 26975 }, { "epoch": 0.07, "learning_rate": 9.346519704061604e-05, "loss": 1.032, "step": 26980 }, { "epoch": 0.07, "learning_rate": 9.346393879913433e-05, "loss": 1.0331, "step": 26985 }, { "epoch": 0.07, "learning_rate": 9.346268055765263e-05, "loss": 1.0326, "step": 26990 }, { "epoch": 0.07, "learning_rate": 9.346142231617092e-05, "loss": 1.0334, "step": 26995 }, { "epoch": 0.07, "learning_rate": 9.346016407468922e-05, "loss": 1.0324, "step": 27000 }, { "epoch": 0.07, "learning_rate": 9.345890583320751e-05, "loss": 1.0347, "step": 27005 }, { "epoch": 0.07, "learning_rate": 9.34576475917258e-05, "loss": 1.0307, "step": 27010 }, { "epoch": 0.07, "learning_rate": 9.34563893502441e-05, "loss": 1.0333, "step": 27015 }, { "epoch": 0.07, "learning_rate": 9.345513110876239e-05, "loss": 1.0332, "step": 27020 }, { "epoch": 0.07, "learning_rate": 9.34538728672807e-05, "loss": 1.032, "step": 27025 }, { "epoch": 0.07, "learning_rate": 9.345261462579899e-05, "loss": 1.0327, "step": 27030 }, { "epoch": 0.07, "learning_rate": 9.345135638431728e-05, "loss": 1.0331, "step": 27035 }, { "epoch": 0.07, "learning_rate": 9.345009814283557e-05, "loss": 1.0316, "step": 27040 }, { "epoch": 0.07, "learning_rate": 9.344883990135387e-05, "loss": 1.0309, "step": 27045 }, { "epoch": 0.07, "learning_rate": 9.344758165987217e-05, "loss": 1.0309, "step": 27050 }, { "epoch": 0.07, "learning_rate": 9.344632341839046e-05, "loss": 1.0326, "step": 27055 }, { "epoch": 0.07, "learning_rate": 9.344506517690875e-05, "loss": 1.0311, "step": 27060 }, { "epoch": 0.07, "learning_rate": 9.344380693542705e-05, "loss": 1.0306, "step": 27065 }, { "epoch": 0.07, "learning_rate": 9.344254869394535e-05, "loss": 1.0301, "step": 27070 }, { "epoch": 0.07, "learning_rate": 9.344129045246364e-05, "loss": 1.0324, "step": 27075 }, { "epoch": 0.07, "learning_rate": 9.344003221098193e-05, "loss": 1.0318, "step": 27080 }, { "epoch": 0.07, "learning_rate": 9.343877396950022e-05, "loss": 1.0509, "step": 27085 }, { "epoch": 0.07, "learning_rate": 9.343751572801853e-05, "loss": 1.0327, "step": 27090 }, { "epoch": 0.07, "learning_rate": 9.343625748653682e-05, "loss": 1.0327, "step": 27095 }, { "epoch": 0.07, "learning_rate": 9.343499924505511e-05, "loss": 1.0329, "step": 27100 }, { "epoch": 0.07, "learning_rate": 9.34337410035734e-05, "loss": 1.0316, "step": 27105 }, { "epoch": 0.07, "learning_rate": 9.34324827620917e-05, "loss": 1.0329, "step": 27110 }, { "epoch": 0.07, "learning_rate": 9.343122452061e-05, "loss": 1.0324, "step": 27115 }, { "epoch": 0.07, "learning_rate": 9.342996627912829e-05, "loss": 1.0316, "step": 27120 }, { "epoch": 0.07, "learning_rate": 9.34287080376466e-05, "loss": 1.0309, "step": 27125 }, { "epoch": 0.07, "learning_rate": 9.342744979616489e-05, "loss": 1.0327, "step": 27130 }, { "epoch": 0.07, "learning_rate": 9.342619155468319e-05, "loss": 1.0313, "step": 27135 }, { "epoch": 0.07, "learning_rate": 9.342493331320148e-05, "loss": 1.0345, "step": 27140 }, { "epoch": 0.07, "learning_rate": 9.342367507171977e-05, "loss": 1.0319, "step": 27145 }, { "epoch": 0.07, "learning_rate": 9.342241683023807e-05, "loss": 1.0285, "step": 27150 }, { "epoch": 0.07, "learning_rate": 9.342115858875636e-05, "loss": 1.0321, "step": 27155 }, { "epoch": 0.07, "learning_rate": 9.341990034727466e-05, "loss": 1.0517, "step": 27160 }, { "epoch": 0.07, "learning_rate": 9.341864210579295e-05, "loss": 1.0316, "step": 27165 }, { "epoch": 0.07, "learning_rate": 9.341738386431124e-05, "loss": 1.0318, "step": 27170 }, { "epoch": 0.07, "learning_rate": 9.341612562282954e-05, "loss": 1.0327, "step": 27175 }, { "epoch": 0.07, "learning_rate": 9.341486738134784e-05, "loss": 1.0311, "step": 27180 }, { "epoch": 0.07, "learning_rate": 9.341360913986613e-05, "loss": 1.03, "step": 27185 }, { "epoch": 0.07, "learning_rate": 9.341235089838442e-05, "loss": 1.0335, "step": 27190 }, { "epoch": 0.07, "learning_rate": 9.341109265690272e-05, "loss": 1.0301, "step": 27195 }, { "epoch": 0.07, "learning_rate": 9.340983441542102e-05, "loss": 1.0316, "step": 27200 }, { "epoch": 0.07, "learning_rate": 9.340857617393931e-05, "loss": 1.0548, "step": 27205 }, { "epoch": 0.07, "learning_rate": 9.34073179324576e-05, "loss": 1.0317, "step": 27210 }, { "epoch": 0.07, "learning_rate": 9.34060596909759e-05, "loss": 1.0343, "step": 27215 }, { "epoch": 0.07, "learning_rate": 9.340480144949419e-05, "loss": 1.0328, "step": 27220 }, { "epoch": 0.07, "learning_rate": 9.340354320801249e-05, "loss": 1.0285, "step": 27225 }, { "epoch": 0.07, "learning_rate": 9.340228496653078e-05, "loss": 1.0337, "step": 27230 }, { "epoch": 0.07, "learning_rate": 9.340102672504908e-05, "loss": 1.0328, "step": 27235 }, { "epoch": 0.07, "learning_rate": 9.339976848356737e-05, "loss": 1.0298, "step": 27240 }, { "epoch": 0.07, "learning_rate": 9.339851024208567e-05, "loss": 1.0318, "step": 27245 }, { "epoch": 0.07, "learning_rate": 9.339725200060396e-05, "loss": 1.0324, "step": 27250 }, { "epoch": 0.07, "learning_rate": 9.339599375912226e-05, "loss": 1.0318, "step": 27255 }, { "epoch": 0.07, "learning_rate": 9.339473551764055e-05, "loss": 1.0308, "step": 27260 }, { "epoch": 0.07, "learning_rate": 9.339347727615885e-05, "loss": 1.0338, "step": 27265 }, { "epoch": 0.07, "learning_rate": 9.339221903467714e-05, "loss": 1.0329, "step": 27270 }, { "epoch": 0.07, "learning_rate": 9.339096079319544e-05, "loss": 1.0304, "step": 27275 }, { "epoch": 0.07, "learning_rate": 9.338970255171373e-05, "loss": 1.031, "step": 27280 }, { "epoch": 0.07, "learning_rate": 9.338844431023202e-05, "loss": 1.0309, "step": 27285 }, { "epoch": 0.07, "learning_rate": 9.338718606875032e-05, "loss": 1.0326, "step": 27290 }, { "epoch": 0.07, "learning_rate": 9.338592782726862e-05, "loss": 1.0336, "step": 27295 }, { "epoch": 0.07, "learning_rate": 9.338466958578691e-05, "loss": 1.0317, "step": 27300 }, { "epoch": 0.07, "learning_rate": 9.33834113443052e-05, "loss": 1.0324, "step": 27305 }, { "epoch": 0.07, "learning_rate": 9.33821531028235e-05, "loss": 1.0335, "step": 27310 }, { "epoch": 0.07, "learning_rate": 9.33808948613418e-05, "loss": 1.0314, "step": 27315 }, { "epoch": 0.07, "learning_rate": 9.337963661986009e-05, "loss": 1.03, "step": 27320 }, { "epoch": 0.07, "learning_rate": 9.337837837837838e-05, "loss": 1.0307, "step": 27325 }, { "epoch": 0.07, "learning_rate": 9.337712013689668e-05, "loss": 1.0313, "step": 27330 }, { "epoch": 0.07, "learning_rate": 9.337586189541498e-05, "loss": 1.0316, "step": 27335 }, { "epoch": 0.07, "learning_rate": 9.337460365393327e-05, "loss": 1.0332, "step": 27340 }, { "epoch": 0.07, "learning_rate": 9.337334541245156e-05, "loss": 1.0318, "step": 27345 }, { "epoch": 0.07, "learning_rate": 9.337208717096985e-05, "loss": 1.0318, "step": 27350 }, { "epoch": 0.07, "learning_rate": 9.337082892948816e-05, "loss": 1.0318, "step": 27355 }, { "epoch": 0.07, "learning_rate": 9.336957068800645e-05, "loss": 1.0318, "step": 27360 }, { "epoch": 0.07, "learning_rate": 9.336831244652474e-05, "loss": 1.032, "step": 27365 }, { "epoch": 0.07, "learning_rate": 9.336705420504303e-05, "loss": 1.0277, "step": 27370 }, { "epoch": 0.07, "learning_rate": 9.336579596356134e-05, "loss": 1.0313, "step": 27375 }, { "epoch": 0.07, "learning_rate": 9.336453772207963e-05, "loss": 1.0332, "step": 27380 }, { "epoch": 0.07, "learning_rate": 9.336327948059792e-05, "loss": 1.0315, "step": 27385 }, { "epoch": 0.07, "learning_rate": 9.336202123911621e-05, "loss": 1.0309, "step": 27390 }, { "epoch": 0.07, "learning_rate": 9.336076299763452e-05, "loss": 1.0315, "step": 27395 }, { "epoch": 0.07, "learning_rate": 9.335950475615281e-05, "loss": 1.0481, "step": 27400 }, { "epoch": 0.07, "learning_rate": 9.33582465146711e-05, "loss": 1.0324, "step": 27405 }, { "epoch": 0.07, "learning_rate": 9.335698827318939e-05, "loss": 1.0326, "step": 27410 }, { "epoch": 0.07, "learning_rate": 9.335573003170768e-05, "loss": 1.0322, "step": 27415 }, { "epoch": 0.07, "learning_rate": 9.335447179022599e-05, "loss": 1.0297, "step": 27420 }, { "epoch": 0.07, "learning_rate": 9.335321354874428e-05, "loss": 1.0338, "step": 27425 }, { "epoch": 0.07, "learning_rate": 9.335195530726257e-05, "loss": 1.0309, "step": 27430 }, { "epoch": 0.07, "learning_rate": 9.335069706578086e-05, "loss": 1.0325, "step": 27435 }, { "epoch": 0.07, "learning_rate": 9.334943882429917e-05, "loss": 1.0323, "step": 27440 }, { "epoch": 0.07, "learning_rate": 9.334818058281746e-05, "loss": 1.0302, "step": 27445 }, { "epoch": 0.07, "learning_rate": 9.334692234133575e-05, "loss": 1.0325, "step": 27450 }, { "epoch": 0.07, "learning_rate": 9.334566409985404e-05, "loss": 1.0306, "step": 27455 }, { "epoch": 0.07, "learning_rate": 9.334440585837235e-05, "loss": 1.0311, "step": 27460 }, { "epoch": 0.07, "learning_rate": 9.334314761689064e-05, "loss": 1.0341, "step": 27465 }, { "epoch": 0.07, "learning_rate": 9.334188937540893e-05, "loss": 1.0309, "step": 27470 }, { "epoch": 0.07, "learning_rate": 9.334063113392722e-05, "loss": 1.0332, "step": 27475 }, { "epoch": 0.07, "learning_rate": 9.333937289244551e-05, "loss": 1.031, "step": 27480 }, { "epoch": 0.07, "learning_rate": 9.333811465096382e-05, "loss": 1.0321, "step": 27485 }, { "epoch": 0.07, "learning_rate": 9.333685640948211e-05, "loss": 1.0308, "step": 27490 }, { "epoch": 0.07, "learning_rate": 9.33355981680004e-05, "loss": 1.0336, "step": 27495 }, { "epoch": 0.07, "learning_rate": 9.33343399265187e-05, "loss": 1.0322, "step": 27500 }, { "epoch": 0.07, "learning_rate": 9.3333081685037e-05, "loss": 1.0335, "step": 27505 }, { "epoch": 0.07, "learning_rate": 9.333182344355529e-05, "loss": 1.0325, "step": 27510 }, { "epoch": 0.07, "learning_rate": 9.333056520207358e-05, "loss": 1.0299, "step": 27515 }, { "epoch": 0.07, "learning_rate": 9.332930696059187e-05, "loss": 1.0352, "step": 27520 }, { "epoch": 0.07, "learning_rate": 9.332804871911018e-05, "loss": 1.0327, "step": 27525 }, { "epoch": 0.07, "learning_rate": 9.332679047762847e-05, "loss": 1.0321, "step": 27530 }, { "epoch": 0.07, "learning_rate": 9.332553223614676e-05, "loss": 1.0341, "step": 27535 }, { "epoch": 0.07, "learning_rate": 9.332427399466505e-05, "loss": 1.0314, "step": 27540 }, { "epoch": 0.07, "learning_rate": 9.332301575318335e-05, "loss": 1.0316, "step": 27545 }, { "epoch": 0.07, "learning_rate": 9.332175751170165e-05, "loss": 1.0292, "step": 27550 }, { "epoch": 0.07, "learning_rate": 9.332049927021994e-05, "loss": 1.033, "step": 27555 }, { "epoch": 0.07, "learning_rate": 9.331924102873823e-05, "loss": 1.0491, "step": 27560 }, { "epoch": 0.07, "learning_rate": 9.331798278725653e-05, "loss": 1.0321, "step": 27565 }, { "epoch": 0.07, "learning_rate": 9.331672454577483e-05, "loss": 1.0332, "step": 27570 }, { "epoch": 0.07, "learning_rate": 9.331546630429312e-05, "loss": 1.0329, "step": 27575 }, { "epoch": 0.07, "learning_rate": 9.331420806281141e-05, "loss": 1.0325, "step": 27580 }, { "epoch": 0.07, "learning_rate": 9.33129498213297e-05, "loss": 1.0305, "step": 27585 }, { "epoch": 0.07, "learning_rate": 9.331169157984801e-05, "loss": 1.0313, "step": 27590 }, { "epoch": 0.07, "learning_rate": 9.33104333383663e-05, "loss": 1.033, "step": 27595 }, { "epoch": 0.07, "learning_rate": 9.33091750968846e-05, "loss": 1.0335, "step": 27600 }, { "epoch": 0.07, "learning_rate": 9.330791685540288e-05, "loss": 1.0312, "step": 27605 }, { "epoch": 0.07, "learning_rate": 9.330665861392118e-05, "loss": 1.0351, "step": 27610 }, { "epoch": 0.07, "learning_rate": 9.330540037243948e-05, "loss": 1.0335, "step": 27615 }, { "epoch": 0.07, "learning_rate": 9.330414213095777e-05, "loss": 1.0313, "step": 27620 }, { "epoch": 0.07, "learning_rate": 9.330288388947608e-05, "loss": 1.0321, "step": 27625 }, { "epoch": 0.07, "learning_rate": 9.330162564799437e-05, "loss": 1.0325, "step": 27630 }, { "epoch": 0.07, "learning_rate": 9.330036740651266e-05, "loss": 1.0307, "step": 27635 }, { "epoch": 0.07, "learning_rate": 9.329910916503097e-05, "loss": 1.0333, "step": 27640 }, { "epoch": 0.07, "learning_rate": 9.329785092354926e-05, "loss": 1.0331, "step": 27645 }, { "epoch": 0.07, "learning_rate": 9.329659268206755e-05, "loss": 1.0289, "step": 27650 }, { "epoch": 0.07, "learning_rate": 9.329533444058584e-05, "loss": 1.0334, "step": 27655 }, { "epoch": 0.07, "learning_rate": 9.329407619910415e-05, "loss": 1.0317, "step": 27660 }, { "epoch": 0.07, "learning_rate": 9.329281795762244e-05, "loss": 1.0328, "step": 27665 }, { "epoch": 0.07, "learning_rate": 9.329155971614073e-05, "loss": 1.0315, "step": 27670 }, { "epoch": 0.07, "learning_rate": 9.329030147465902e-05, "loss": 1.0313, "step": 27675 }, { "epoch": 0.07, "learning_rate": 9.328904323317733e-05, "loss": 1.0303, "step": 27680 }, { "epoch": 0.07, "learning_rate": 9.328778499169562e-05, "loss": 1.0321, "step": 27685 }, { "epoch": 0.07, "learning_rate": 9.328652675021391e-05, "loss": 1.0304, "step": 27690 }, { "epoch": 0.07, "learning_rate": 9.32852685087322e-05, "loss": 1.0338, "step": 27695 }, { "epoch": 0.07, "learning_rate": 9.328401026725049e-05, "loss": 1.032, "step": 27700 }, { "epoch": 0.07, "learning_rate": 9.32827520257688e-05, "loss": 1.0312, "step": 27705 }, { "epoch": 0.07, "learning_rate": 9.328149378428709e-05, "loss": 1.0288, "step": 27710 }, { "epoch": 0.07, "learning_rate": 9.328023554280538e-05, "loss": 1.0307, "step": 27715 }, { "epoch": 0.07, "learning_rate": 9.327897730132367e-05, "loss": 1.0318, "step": 27720 }, { "epoch": 0.07, "learning_rate": 9.327771905984198e-05, "loss": 1.0327, "step": 27725 }, { "epoch": 0.07, "learning_rate": 9.327646081836027e-05, "loss": 1.0322, "step": 27730 }, { "epoch": 0.07, "learning_rate": 9.327520257687856e-05, "loss": 1.0312, "step": 27735 }, { "epoch": 0.07, "learning_rate": 9.327394433539685e-05, "loss": 1.0532, "step": 27740 }, { "epoch": 0.07, "learning_rate": 9.327268609391516e-05, "loss": 1.032, "step": 27745 }, { "epoch": 0.07, "learning_rate": 9.327142785243345e-05, "loss": 1.0343, "step": 27750 }, { "epoch": 0.07, "learning_rate": 9.327016961095174e-05, "loss": 1.0307, "step": 27755 }, { "epoch": 0.07, "learning_rate": 9.326891136947003e-05, "loss": 1.0302, "step": 27760 }, { "epoch": 0.07, "learning_rate": 9.326765312798832e-05, "loss": 1.0536, "step": 27765 }, { "epoch": 0.07, "learning_rate": 9.326639488650663e-05, "loss": 1.0297, "step": 27770 }, { "epoch": 0.07, "learning_rate": 9.326513664502492e-05, "loss": 1.0536, "step": 27775 }, { "epoch": 0.07, "learning_rate": 9.326387840354321e-05, "loss": 1.0318, "step": 27780 }, { "epoch": 0.07, "learning_rate": 9.32626201620615e-05, "loss": 1.032, "step": 27785 }, { "epoch": 0.07, "learning_rate": 9.326136192057981e-05, "loss": 1.0318, "step": 27790 }, { "epoch": 0.07, "learning_rate": 9.32601036790981e-05, "loss": 1.0323, "step": 27795 }, { "epoch": 0.07, "learning_rate": 9.325884543761639e-05, "loss": 1.0311, "step": 27800 }, { "epoch": 0.07, "learning_rate": 9.325758719613468e-05, "loss": 1.0313, "step": 27805 }, { "epoch": 0.07, "learning_rate": 9.325632895465299e-05, "loss": 1.0324, "step": 27810 }, { "epoch": 0.07, "learning_rate": 9.325507071317128e-05, "loss": 1.033, "step": 27815 }, { "epoch": 0.07, "learning_rate": 9.325381247168957e-05, "loss": 1.0299, "step": 27820 }, { "epoch": 0.07, "learning_rate": 9.325255423020786e-05, "loss": 1.0304, "step": 27825 }, { "epoch": 0.07, "learning_rate": 9.325129598872616e-05, "loss": 1.0306, "step": 27830 }, { "epoch": 0.07, "learning_rate": 9.325003774724446e-05, "loss": 1.0317, "step": 27835 }, { "epoch": 0.07, "learning_rate": 9.324877950576275e-05, "loss": 1.0325, "step": 27840 }, { "epoch": 0.07, "learning_rate": 9.324752126428104e-05, "loss": 1.0304, "step": 27845 }, { "epoch": 0.07, "learning_rate": 9.324626302279934e-05, "loss": 1.0332, "step": 27850 }, { "epoch": 0.07, "learning_rate": 9.324500478131764e-05, "loss": 1.0316, "step": 27855 }, { "epoch": 0.07, "learning_rate": 9.324374653983593e-05, "loss": 1.031, "step": 27860 }, { "epoch": 0.07, "learning_rate": 9.324248829835422e-05, "loss": 1.0317, "step": 27865 }, { "epoch": 0.07, "learning_rate": 9.324123005687252e-05, "loss": 1.0326, "step": 27870 }, { "epoch": 0.07, "learning_rate": 9.323997181539081e-05, "loss": 1.0342, "step": 27875 }, { "epoch": 0.07, "learning_rate": 9.323871357390911e-05, "loss": 1.0326, "step": 27880 }, { "epoch": 0.07, "learning_rate": 9.32374553324274e-05, "loss": 1.0307, "step": 27885 }, { "epoch": 0.07, "learning_rate": 9.32361970909457e-05, "loss": 1.0328, "step": 27890 }, { "epoch": 0.07, "learning_rate": 9.323493884946399e-05, "loss": 1.0319, "step": 27895 }, { "epoch": 0.07, "learning_rate": 9.323368060798229e-05, "loss": 1.0307, "step": 27900 }, { "epoch": 0.07, "learning_rate": 9.323242236650058e-05, "loss": 1.0309, "step": 27905 }, { "epoch": 0.07, "learning_rate": 9.323116412501888e-05, "loss": 1.0294, "step": 27910 }, { "epoch": 0.07, "learning_rate": 9.322990588353717e-05, "loss": 1.0323, "step": 27915 }, { "epoch": 0.07, "learning_rate": 9.322864764205547e-05, "loss": 1.0318, "step": 27920 }, { "epoch": 0.07, "learning_rate": 9.322738940057376e-05, "loss": 1.0326, "step": 27925 }, { "epoch": 0.07, "learning_rate": 9.322613115909206e-05, "loss": 1.0315, "step": 27930 }, { "epoch": 0.07, "learning_rate": 9.322487291761035e-05, "loss": 1.0312, "step": 27935 }, { "epoch": 0.07, "learning_rate": 9.322361467612864e-05, "loss": 1.0328, "step": 27940 }, { "epoch": 0.07, "learning_rate": 9.322235643464694e-05, "loss": 1.0322, "step": 27945 }, { "epoch": 0.07, "learning_rate": 9.322109819316524e-05, "loss": 1.0328, "step": 27950 }, { "epoch": 0.07, "learning_rate": 9.321983995168353e-05, "loss": 1.0325, "step": 27955 }, { "epoch": 0.07, "learning_rate": 9.321858171020182e-05, "loss": 1.0561, "step": 27960 }, { "epoch": 0.07, "learning_rate": 9.321732346872012e-05, "loss": 1.0328, "step": 27965 }, { "epoch": 0.07, "learning_rate": 9.321606522723842e-05, "loss": 1.0337, "step": 27970 }, { "epoch": 0.07, "learning_rate": 9.321480698575671e-05, "loss": 1.0304, "step": 27975 }, { "epoch": 0.07, "learning_rate": 9.3213548744275e-05, "loss": 1.0308, "step": 27980 }, { "epoch": 0.07, "learning_rate": 9.32122905027933e-05, "loss": 1.0327, "step": 27985 }, { "epoch": 0.07, "learning_rate": 9.32110322613116e-05, "loss": 1.0291, "step": 27990 }, { "epoch": 0.07, "learning_rate": 9.320977401982989e-05, "loss": 1.0299, "step": 27995 }, { "epoch": 0.07, "learning_rate": 9.320851577834818e-05, "loss": 1.0314, "step": 28000 }, { "epoch": 0.07, "learning_rate": 9.320725753686647e-05, "loss": 1.0341, "step": 28005 }, { "epoch": 0.07, "learning_rate": 9.320599929538478e-05, "loss": 1.0323, "step": 28010 }, { "epoch": 0.07, "learning_rate": 9.320474105390307e-05, "loss": 1.0295, "step": 28015 }, { "epoch": 0.07, "learning_rate": 9.320348281242136e-05, "loss": 1.0305, "step": 28020 }, { "epoch": 0.07, "learning_rate": 9.320222457093965e-05, "loss": 1.0308, "step": 28025 }, { "epoch": 0.07, "learning_rate": 9.320096632945796e-05, "loss": 1.0318, "step": 28030 }, { "epoch": 0.07, "learning_rate": 9.319970808797625e-05, "loss": 1.0344, "step": 28035 }, { "epoch": 0.07, "learning_rate": 9.319844984649454e-05, "loss": 1.0341, "step": 28040 }, { "epoch": 0.07, "learning_rate": 9.319719160501283e-05, "loss": 1.0315, "step": 28045 }, { "epoch": 0.07, "learning_rate": 9.319593336353114e-05, "loss": 1.0308, "step": 28050 }, { "epoch": 0.07, "learning_rate": 9.319467512204943e-05, "loss": 1.0308, "step": 28055 }, { "epoch": 0.07, "learning_rate": 9.319341688056772e-05, "loss": 1.032, "step": 28060 }, { "epoch": 0.07, "learning_rate": 9.319215863908601e-05, "loss": 1.0322, "step": 28065 }, { "epoch": 0.07, "learning_rate": 9.31909003976043e-05, "loss": 1.0326, "step": 28070 }, { "epoch": 0.07, "learning_rate": 9.318964215612261e-05, "loss": 1.0316, "step": 28075 }, { "epoch": 0.07, "learning_rate": 9.31883839146409e-05, "loss": 1.033, "step": 28080 }, { "epoch": 0.07, "learning_rate": 9.318712567315919e-05, "loss": 1.0337, "step": 28085 }, { "epoch": 0.07, "learning_rate": 9.318586743167748e-05, "loss": 1.0315, "step": 28090 }, { "epoch": 0.07, "learning_rate": 9.318460919019579e-05, "loss": 1.0327, "step": 28095 }, { "epoch": 0.07, "learning_rate": 9.318335094871408e-05, "loss": 1.0291, "step": 28100 }, { "epoch": 0.07, "learning_rate": 9.318209270723237e-05, "loss": 1.0304, "step": 28105 }, { "epoch": 0.07, "learning_rate": 9.318083446575066e-05, "loss": 1.0316, "step": 28110 }, { "epoch": 0.07, "learning_rate": 9.317957622426897e-05, "loss": 1.0319, "step": 28115 }, { "epoch": 0.07, "learning_rate": 9.317831798278727e-05, "loss": 1.0313, "step": 28120 }, { "epoch": 0.07, "learning_rate": 9.317705974130556e-05, "loss": 1.0303, "step": 28125 }, { "epoch": 0.07, "learning_rate": 9.317580149982386e-05, "loss": 1.031, "step": 28130 }, { "epoch": 0.07, "learning_rate": 9.317454325834215e-05, "loss": 1.0334, "step": 28135 }, { "epoch": 0.07, "learning_rate": 9.317328501686045e-05, "loss": 1.0309, "step": 28140 }, { "epoch": 0.07, "learning_rate": 9.317202677537874e-05, "loss": 1.0313, "step": 28145 }, { "epoch": 0.07, "learning_rate": 9.317076853389704e-05, "loss": 1.0316, "step": 28150 }, { "epoch": 0.07, "learning_rate": 9.316951029241533e-05, "loss": 1.0298, "step": 28155 }, { "epoch": 0.07, "learning_rate": 9.316825205093362e-05, "loss": 1.0317, "step": 28160 }, { "epoch": 0.07, "learning_rate": 9.316699380945192e-05, "loss": 1.0316, "step": 28165 }, { "epoch": 0.07, "learning_rate": 9.316573556797021e-05, "loss": 1.0446, "step": 28170 }, { "epoch": 0.07, "learning_rate": 9.31644773264885e-05, "loss": 1.0316, "step": 28175 }, { "epoch": 0.07, "learning_rate": 9.31632190850068e-05, "loss": 1.0317, "step": 28180 }, { "epoch": 0.07, "learning_rate": 9.31619608435251e-05, "loss": 1.0307, "step": 28185 }, { "epoch": 0.07, "learning_rate": 9.31607026020434e-05, "loss": 1.033, "step": 28190 }, { "epoch": 0.07, "learning_rate": 9.315944436056169e-05, "loss": 1.0309, "step": 28195 }, { "epoch": 0.07, "learning_rate": 9.315818611907998e-05, "loss": 1.0314, "step": 28200 }, { "epoch": 0.07, "learning_rate": 9.315692787759828e-05, "loss": 1.03, "step": 28205 }, { "epoch": 0.07, "learning_rate": 9.315566963611657e-05, "loss": 1.0326, "step": 28210 }, { "epoch": 0.07, "learning_rate": 9.315441139463487e-05, "loss": 1.0315, "step": 28215 }, { "epoch": 0.07, "learning_rate": 9.315315315315316e-05, "loss": 1.0341, "step": 28220 }, { "epoch": 0.07, "learning_rate": 9.315189491167145e-05, "loss": 1.0351, "step": 28225 }, { "epoch": 0.07, "learning_rate": 9.315063667018975e-05, "loss": 1.0333, "step": 28230 }, { "epoch": 0.07, "learning_rate": 9.314937842870805e-05, "loss": 1.0339, "step": 28235 }, { "epoch": 0.07, "learning_rate": 9.314812018722634e-05, "loss": 1.0331, "step": 28240 }, { "epoch": 0.07, "learning_rate": 9.314686194574463e-05, "loss": 1.034, "step": 28245 }, { "epoch": 0.07, "learning_rate": 9.314560370426293e-05, "loss": 1.0325, "step": 28250 }, { "epoch": 0.07, "learning_rate": 9.314434546278123e-05, "loss": 1.0332, "step": 28255 }, { "epoch": 0.07, "learning_rate": 9.314308722129952e-05, "loss": 1.0336, "step": 28260 }, { "epoch": 0.07, "learning_rate": 9.314182897981781e-05, "loss": 1.0318, "step": 28265 }, { "epoch": 0.07, "learning_rate": 9.314057073833611e-05, "loss": 1.0301, "step": 28270 }, { "epoch": 0.07, "learning_rate": 9.31393124968544e-05, "loss": 1.0346, "step": 28275 }, { "epoch": 0.07, "learning_rate": 9.31380542553727e-05, "loss": 1.0315, "step": 28280 }, { "epoch": 0.07, "learning_rate": 9.313679601389099e-05, "loss": 1.033, "step": 28285 }, { "epoch": 0.07, "learning_rate": 9.313553777240928e-05, "loss": 1.0323, "step": 28290 }, { "epoch": 0.07, "learning_rate": 9.313427953092759e-05, "loss": 1.0309, "step": 28295 }, { "epoch": 0.07, "learning_rate": 9.313302128944588e-05, "loss": 1.0319, "step": 28300 }, { "epoch": 0.07, "learning_rate": 9.313176304796417e-05, "loss": 1.0328, "step": 28305 }, { "epoch": 0.07, "learning_rate": 9.313050480648246e-05, "loss": 1.0315, "step": 28310 }, { "epoch": 0.07, "learning_rate": 9.312924656500077e-05, "loss": 1.0323, "step": 28315 }, { "epoch": 0.07, "learning_rate": 9.312798832351906e-05, "loss": 1.0309, "step": 28320 }, { "epoch": 0.07, "learning_rate": 9.312673008203735e-05, "loss": 1.031, "step": 28325 }, { "epoch": 0.07, "learning_rate": 9.312547184055564e-05, "loss": 1.0317, "step": 28330 }, { "epoch": 0.07, "learning_rate": 9.312421359907395e-05, "loss": 1.028, "step": 28335 }, { "epoch": 0.07, "learning_rate": 9.312295535759224e-05, "loss": 1.0312, "step": 28340 }, { "epoch": 0.07, "learning_rate": 9.312169711611053e-05, "loss": 1.0318, "step": 28345 }, { "epoch": 0.07, "learning_rate": 9.312043887462882e-05, "loss": 1.0317, "step": 28350 }, { "epoch": 0.07, "learning_rate": 9.311918063314711e-05, "loss": 1.0305, "step": 28355 }, { "epoch": 0.07, "learning_rate": 9.311792239166542e-05, "loss": 1.0308, "step": 28360 }, { "epoch": 0.07, "learning_rate": 9.311666415018371e-05, "loss": 1.033, "step": 28365 }, { "epoch": 0.07, "learning_rate": 9.3115405908702e-05, "loss": 1.0333, "step": 28370 }, { "epoch": 0.07, "learning_rate": 9.311414766722029e-05, "loss": 1.0502, "step": 28375 }, { "epoch": 0.07, "learning_rate": 9.31128894257386e-05, "loss": 1.0331, "step": 28380 }, { "epoch": 0.07, "learning_rate": 9.311163118425689e-05, "loss": 1.0329, "step": 28385 }, { "epoch": 0.07, "learning_rate": 9.311037294277518e-05, "loss": 1.03, "step": 28390 }, { "epoch": 0.07, "learning_rate": 9.310911470129347e-05, "loss": 1.0315, "step": 28395 }, { "epoch": 0.07, "learning_rate": 9.310785645981178e-05, "loss": 1.0349, "step": 28400 }, { "epoch": 0.07, "learning_rate": 9.310659821833007e-05, "loss": 1.0317, "step": 28405 }, { "epoch": 0.07, "learning_rate": 9.310533997684836e-05, "loss": 1.0303, "step": 28410 }, { "epoch": 0.07, "learning_rate": 9.310408173536665e-05, "loss": 1.0324, "step": 28415 }, { "epoch": 0.07, "learning_rate": 9.310282349388494e-05, "loss": 1.0313, "step": 28420 }, { "epoch": 0.07, "learning_rate": 9.310156525240325e-05, "loss": 1.0304, "step": 28425 }, { "epoch": 0.07, "learning_rate": 9.310030701092154e-05, "loss": 1.0305, "step": 28430 }, { "epoch": 0.07, "learning_rate": 9.309904876943983e-05, "loss": 1.0317, "step": 28435 }, { "epoch": 0.07, "learning_rate": 9.309779052795812e-05, "loss": 1.0309, "step": 28440 }, { "epoch": 0.07, "learning_rate": 9.309653228647643e-05, "loss": 1.0312, "step": 28445 }, { "epoch": 0.07, "learning_rate": 9.309527404499472e-05, "loss": 1.0317, "step": 28450 }, { "epoch": 0.07, "learning_rate": 9.309401580351301e-05, "loss": 1.0317, "step": 28455 }, { "epoch": 0.07, "learning_rate": 9.30927575620313e-05, "loss": 1.0313, "step": 28460 }, { "epoch": 0.07, "learning_rate": 9.309149932054961e-05, "loss": 1.0304, "step": 28465 }, { "epoch": 0.07, "learning_rate": 9.30902410790679e-05, "loss": 1.0342, "step": 28470 }, { "epoch": 0.07, "learning_rate": 9.308898283758619e-05, "loss": 1.0314, "step": 28475 }, { "epoch": 0.07, "learning_rate": 9.308772459610448e-05, "loss": 1.0305, "step": 28480 }, { "epoch": 0.07, "learning_rate": 9.308646635462278e-05, "loss": 1.0295, "step": 28485 }, { "epoch": 0.07, "learning_rate": 9.308520811314108e-05, "loss": 1.0316, "step": 28490 }, { "epoch": 0.07, "learning_rate": 9.308394987165937e-05, "loss": 1.0309, "step": 28495 }, { "epoch": 0.07, "learning_rate": 9.308269163017766e-05, "loss": 1.0289, "step": 28500 }, { "epoch": 0.07, "learning_rate": 9.308143338869596e-05, "loss": 1.0322, "step": 28505 }, { "epoch": 0.07, "learning_rate": 9.308017514721426e-05, "loss": 1.0301, "step": 28510 }, { "epoch": 0.07, "learning_rate": 9.307891690573255e-05, "loss": 1.0318, "step": 28515 }, { "epoch": 0.07, "learning_rate": 9.307765866425084e-05, "loss": 1.0322, "step": 28520 }, { "epoch": 0.07, "learning_rate": 9.307640042276914e-05, "loss": 1.0333, "step": 28525 }, { "epoch": 0.07, "learning_rate": 9.307514218128743e-05, "loss": 1.0314, "step": 28530 }, { "epoch": 0.07, "learning_rate": 9.307388393980573e-05, "loss": 1.0319, "step": 28535 }, { "epoch": 0.07, "learning_rate": 9.307262569832402e-05, "loss": 1.0468, "step": 28540 }, { "epoch": 0.07, "learning_rate": 9.307136745684232e-05, "loss": 1.0355, "step": 28545 }, { "epoch": 0.07, "learning_rate": 9.30701092153606e-05, "loss": 1.0307, "step": 28550 }, { "epoch": 0.07, "learning_rate": 9.306885097387891e-05, "loss": 1.0299, "step": 28555 }, { "epoch": 0.07, "learning_rate": 9.30675927323972e-05, "loss": 1.0289, "step": 28560 }, { "epoch": 0.07, "learning_rate": 9.30663344909155e-05, "loss": 1.0307, "step": 28565 }, { "epoch": 0.07, "learning_rate": 9.306507624943379e-05, "loss": 1.0311, "step": 28570 }, { "epoch": 0.07, "learning_rate": 9.306381800795209e-05, "loss": 1.0303, "step": 28575 }, { "epoch": 0.07, "learning_rate": 9.306255976647038e-05, "loss": 1.032, "step": 28580 }, { "epoch": 0.07, "learning_rate": 9.306130152498868e-05, "loss": 1.0313, "step": 28585 }, { "epoch": 0.07, "learning_rate": 9.306004328350697e-05, "loss": 1.0322, "step": 28590 }, { "epoch": 0.07, "learning_rate": 9.305878504202526e-05, "loss": 1.0317, "step": 28595 }, { "epoch": 0.07, "learning_rate": 9.305752680054356e-05, "loss": 1.0314, "step": 28600 }, { "epoch": 0.07, "learning_rate": 9.305626855906186e-05, "loss": 1.0305, "step": 28605 }, { "epoch": 0.07, "learning_rate": 9.305501031758015e-05, "loss": 1.0297, "step": 28610 }, { "epoch": 0.07, "learning_rate": 9.305375207609844e-05, "loss": 1.0313, "step": 28615 }, { "epoch": 0.07, "learning_rate": 9.305249383461674e-05, "loss": 1.0343, "step": 28620 }, { "epoch": 0.07, "learning_rate": 9.305123559313505e-05, "loss": 1.0311, "step": 28625 }, { "epoch": 0.07, "learning_rate": 9.304997735165334e-05, "loss": 1.0334, "step": 28630 }, { "epoch": 0.07, "learning_rate": 9.304871911017163e-05, "loss": 1.033, "step": 28635 }, { "epoch": 0.07, "learning_rate": 9.304746086868992e-05, "loss": 1.0308, "step": 28640 }, { "epoch": 0.07, "learning_rate": 9.304620262720823e-05, "loss": 1.0325, "step": 28645 }, { "epoch": 0.07, "learning_rate": 9.304494438572652e-05, "loss": 1.0284, "step": 28650 }, { "epoch": 0.07, "learning_rate": 9.304368614424481e-05, "loss": 1.0306, "step": 28655 }, { "epoch": 0.07, "learning_rate": 9.30424279027631e-05, "loss": 1.0341, "step": 28660 }, { "epoch": 0.07, "learning_rate": 9.304116966128141e-05, "loss": 1.0325, "step": 28665 }, { "epoch": 0.07, "learning_rate": 9.30399114197997e-05, "loss": 1.0319, "step": 28670 }, { "epoch": 0.07, "learning_rate": 9.303865317831799e-05, "loss": 1.0307, "step": 28675 }, { "epoch": 0.07, "learning_rate": 9.303739493683628e-05, "loss": 1.0337, "step": 28680 }, { "epoch": 0.07, "learning_rate": 9.303613669535457e-05, "loss": 1.0316, "step": 28685 }, { "epoch": 0.07, "learning_rate": 9.303487845387288e-05, "loss": 1.0311, "step": 28690 }, { "epoch": 0.07, "learning_rate": 9.303362021239117e-05, "loss": 1.0309, "step": 28695 }, { "epoch": 0.07, "learning_rate": 9.303236197090946e-05, "loss": 1.0329, "step": 28700 }, { "epoch": 0.07, "learning_rate": 9.303110372942775e-05, "loss": 1.0296, "step": 28705 }, { "epoch": 0.07, "learning_rate": 9.302984548794606e-05, "loss": 1.0325, "step": 28710 }, { "epoch": 0.07, "learning_rate": 9.302858724646435e-05, "loss": 1.0294, "step": 28715 }, { "epoch": 0.07, "learning_rate": 9.302732900498264e-05, "loss": 1.0304, "step": 28720 }, { "epoch": 0.07, "learning_rate": 9.302607076350093e-05, "loss": 1.0301, "step": 28725 }, { "epoch": 0.07, "learning_rate": 9.302481252201924e-05, "loss": 1.029, "step": 28730 }, { "epoch": 0.07, "learning_rate": 9.302355428053753e-05, "loss": 1.0336, "step": 28735 }, { "epoch": 0.07, "learning_rate": 9.302229603905582e-05, "loss": 1.0323, "step": 28740 }, { "epoch": 0.07, "learning_rate": 9.302103779757411e-05, "loss": 1.0297, "step": 28745 }, { "epoch": 0.07, "learning_rate": 9.30197795560924e-05, "loss": 1.0301, "step": 28750 }, { "epoch": 0.07, "learning_rate": 9.301852131461071e-05, "loss": 1.0318, "step": 28755 }, { "epoch": 0.07, "learning_rate": 9.3017263073129e-05, "loss": 1.0306, "step": 28760 }, { "epoch": 0.07, "learning_rate": 9.30160048316473e-05, "loss": 1.0301, "step": 28765 }, { "epoch": 0.07, "learning_rate": 9.301474659016559e-05, "loss": 1.033, "step": 28770 }, { "epoch": 0.07, "learning_rate": 9.301348834868389e-05, "loss": 1.03, "step": 28775 }, { "epoch": 0.07, "learning_rate": 9.301223010720218e-05, "loss": 1.0321, "step": 28780 }, { "epoch": 0.07, "learning_rate": 9.301097186572047e-05, "loss": 1.0326, "step": 28785 }, { "epoch": 0.07, "learning_rate": 9.300971362423877e-05, "loss": 1.054, "step": 28790 }, { "epoch": 0.07, "learning_rate": 9.300845538275707e-05, "loss": 1.0327, "step": 28795 }, { "epoch": 0.07, "learning_rate": 9.300719714127536e-05, "loss": 1.0305, "step": 28800 }, { "epoch": 0.07, "learning_rate": 9.300593889979365e-05, "loss": 1.0328, "step": 28805 }, { "epoch": 0.07, "learning_rate": 9.300468065831195e-05, "loss": 1.0319, "step": 28810 }, { "epoch": 0.07, "learning_rate": 9.300342241683024e-05, "loss": 1.0303, "step": 28815 }, { "epoch": 0.07, "learning_rate": 9.300216417534854e-05, "loss": 1.0342, "step": 28820 }, { "epoch": 0.07, "learning_rate": 9.300090593386683e-05, "loss": 1.032, "step": 28825 }, { "epoch": 0.07, "learning_rate": 9.299964769238513e-05, "loss": 1.0315, "step": 28830 }, { "epoch": 0.07, "learning_rate": 9.299838945090342e-05, "loss": 1.0296, "step": 28835 }, { "epoch": 0.07, "learning_rate": 9.299713120942172e-05, "loss": 1.032, "step": 28840 }, { "epoch": 0.07, "learning_rate": 9.299587296794001e-05, "loss": 1.031, "step": 28845 }, { "epoch": 0.07, "learning_rate": 9.29946147264583e-05, "loss": 1.0293, "step": 28850 }, { "epoch": 0.07, "learning_rate": 9.29933564849766e-05, "loss": 1.0328, "step": 28855 }, { "epoch": 0.07, "learning_rate": 9.29920982434949e-05, "loss": 1.0327, "step": 28860 }, { "epoch": 0.07, "learning_rate": 9.29908400020132e-05, "loss": 1.0327, "step": 28865 }, { "epoch": 0.07, "learning_rate": 9.298958176053149e-05, "loss": 1.0318, "step": 28870 }, { "epoch": 0.07, "learning_rate": 9.298832351904978e-05, "loss": 1.0323, "step": 28875 }, { "epoch": 0.07, "learning_rate": 9.298706527756807e-05, "loss": 1.0317, "step": 28880 }, { "epoch": 0.07, "learning_rate": 9.298580703608637e-05, "loss": 1.0297, "step": 28885 }, { "epoch": 0.07, "learning_rate": 9.298454879460467e-05, "loss": 1.0289, "step": 28890 }, { "epoch": 0.07, "learning_rate": 9.298329055312296e-05, "loss": 1.0314, "step": 28895 }, { "epoch": 0.07, "learning_rate": 9.298203231164125e-05, "loss": 1.0321, "step": 28900 }, { "epoch": 0.07, "learning_rate": 9.298077407015955e-05, "loss": 1.033, "step": 28905 }, { "epoch": 0.07, "learning_rate": 9.297951582867785e-05, "loss": 1.031, "step": 28910 }, { "epoch": 0.07, "learning_rate": 9.297825758719614e-05, "loss": 1.0329, "step": 28915 }, { "epoch": 0.07, "learning_rate": 9.297699934571443e-05, "loss": 1.0541, "step": 28920 }, { "epoch": 0.07, "learning_rate": 9.297574110423273e-05, "loss": 1.0337, "step": 28925 }, { "epoch": 0.07, "learning_rate": 9.297448286275103e-05, "loss": 1.0339, "step": 28930 }, { "epoch": 0.07, "learning_rate": 9.297322462126932e-05, "loss": 1.0313, "step": 28935 }, { "epoch": 0.07, "learning_rate": 9.297196637978761e-05, "loss": 1.0483, "step": 28940 }, { "epoch": 0.07, "learning_rate": 9.29707081383059e-05, "loss": 1.0311, "step": 28945 }, { "epoch": 0.07, "learning_rate": 9.29694498968242e-05, "loss": 1.0484, "step": 28950 }, { "epoch": 0.07, "learning_rate": 9.29681916553425e-05, "loss": 1.032, "step": 28955 }, { "epoch": 0.07, "learning_rate": 9.296693341386079e-05, "loss": 1.0304, "step": 28960 }, { "epoch": 0.07, "learning_rate": 9.296567517237908e-05, "loss": 1.0301, "step": 28965 }, { "epoch": 0.07, "learning_rate": 9.296441693089739e-05, "loss": 1.033, "step": 28970 }, { "epoch": 0.07, "learning_rate": 9.296315868941568e-05, "loss": 1.0325, "step": 28975 }, { "epoch": 0.07, "learning_rate": 9.296190044793397e-05, "loss": 1.031, "step": 28980 }, { "epoch": 0.07, "learning_rate": 9.296064220645226e-05, "loss": 1.0314, "step": 28985 }, { "epoch": 0.07, "learning_rate": 9.295938396497057e-05, "loss": 1.0302, "step": 28990 }, { "epoch": 0.07, "learning_rate": 9.295812572348886e-05, "loss": 1.0304, "step": 28995 }, { "epoch": 0.07, "learning_rate": 9.295686748200715e-05, "loss": 1.0324, "step": 29000 }, { "epoch": 0.07, "learning_rate": 9.295560924052544e-05, "loss": 1.031, "step": 29005 }, { "epoch": 0.07, "learning_rate": 9.295435099904373e-05, "loss": 1.0308, "step": 29010 }, { "epoch": 0.07, "learning_rate": 9.295309275756204e-05, "loss": 1.031, "step": 29015 }, { "epoch": 0.07, "learning_rate": 9.295183451608033e-05, "loss": 1.0332, "step": 29020 }, { "epoch": 0.07, "learning_rate": 9.295057627459862e-05, "loss": 1.0537, "step": 29025 }, { "epoch": 0.07, "learning_rate": 9.294931803311691e-05, "loss": 1.0336, "step": 29030 }, { "epoch": 0.07, "learning_rate": 9.294805979163522e-05, "loss": 1.0347, "step": 29035 }, { "epoch": 0.07, "learning_rate": 9.294680155015351e-05, "loss": 1.0318, "step": 29040 }, { "epoch": 0.07, "learning_rate": 9.29455433086718e-05, "loss": 1.0308, "step": 29045 }, { "epoch": 0.07, "learning_rate": 9.294428506719009e-05, "loss": 1.0312, "step": 29050 }, { "epoch": 0.07, "learning_rate": 9.29430268257084e-05, "loss": 1.031, "step": 29055 }, { "epoch": 0.07, "learning_rate": 9.294176858422669e-05, "loss": 1.0302, "step": 29060 }, { "epoch": 0.07, "learning_rate": 9.294051034274498e-05, "loss": 1.0329, "step": 29065 }, { "epoch": 0.07, "learning_rate": 9.293925210126327e-05, "loss": 1.0304, "step": 29070 }, { "epoch": 0.07, "learning_rate": 9.293799385978156e-05, "loss": 1.0325, "step": 29075 }, { "epoch": 0.07, "learning_rate": 9.293673561829987e-05, "loss": 1.0305, "step": 29080 }, { "epoch": 0.07, "learning_rate": 9.293547737681816e-05, "loss": 1.0338, "step": 29085 }, { "epoch": 0.07, "learning_rate": 9.293421913533645e-05, "loss": 1.03, "step": 29090 }, { "epoch": 0.07, "learning_rate": 9.293296089385474e-05, "loss": 1.0315, "step": 29095 }, { "epoch": 0.07, "learning_rate": 9.293170265237305e-05, "loss": 1.0317, "step": 29100 }, { "epoch": 0.07, "learning_rate": 9.293044441089134e-05, "loss": 1.0305, "step": 29105 }, { "epoch": 0.07, "learning_rate": 9.292918616940963e-05, "loss": 1.0315, "step": 29110 }, { "epoch": 0.07, "learning_rate": 9.292792792792792e-05, "loss": 1.0324, "step": 29115 }, { "epoch": 0.07, "learning_rate": 9.292666968644623e-05, "loss": 1.03, "step": 29120 }, { "epoch": 0.07, "learning_rate": 9.292541144496453e-05, "loss": 1.0322, "step": 29125 }, { "epoch": 0.07, "learning_rate": 9.292415320348283e-05, "loss": 1.0327, "step": 29130 }, { "epoch": 0.07, "learning_rate": 9.292289496200112e-05, "loss": 1.0315, "step": 29135 }, { "epoch": 0.07, "learning_rate": 9.292163672051941e-05, "loss": 1.0319, "step": 29140 }, { "epoch": 0.07, "learning_rate": 9.29203784790377e-05, "loss": 1.0313, "step": 29145 }, { "epoch": 0.07, "learning_rate": 9.2919120237556e-05, "loss": 1.032, "step": 29150 }, { "epoch": 0.07, "learning_rate": 9.29178619960743e-05, "loss": 1.0307, "step": 29155 }, { "epoch": 0.07, "learning_rate": 9.291660375459259e-05, "loss": 1.0334, "step": 29160 }, { "epoch": 0.07, "learning_rate": 9.291534551311088e-05, "loss": 1.0304, "step": 29165 }, { "epoch": 0.07, "learning_rate": 9.291408727162918e-05, "loss": 1.0343, "step": 29170 }, { "epoch": 0.07, "learning_rate": 9.291282903014748e-05, "loss": 1.0323, "step": 29175 }, { "epoch": 0.07, "learning_rate": 9.291157078866577e-05, "loss": 1.0321, "step": 29180 }, { "epoch": 0.07, "learning_rate": 9.291031254718406e-05, "loss": 1.0305, "step": 29185 }, { "epoch": 0.07, "learning_rate": 9.290905430570236e-05, "loss": 1.0296, "step": 29190 }, { "epoch": 0.07, "learning_rate": 9.290779606422066e-05, "loss": 1.0325, "step": 29195 }, { "epoch": 0.07, "learning_rate": 9.290653782273895e-05, "loss": 1.031, "step": 29200 }, { "epoch": 0.07, "learning_rate": 9.290527958125724e-05, "loss": 1.0329, "step": 29205 }, { "epoch": 0.07, "learning_rate": 9.290402133977553e-05, "loss": 1.0316, "step": 29210 }, { "epoch": 0.07, "learning_rate": 9.290276309829384e-05, "loss": 1.0331, "step": 29215 }, { "epoch": 0.07, "learning_rate": 9.290150485681213e-05, "loss": 1.0301, "step": 29220 }, { "epoch": 0.07, "learning_rate": 9.290024661533042e-05, "loss": 1.0309, "step": 29225 }, { "epoch": 0.07, "learning_rate": 9.289898837384871e-05, "loss": 1.03, "step": 29230 }, { "epoch": 0.07, "learning_rate": 9.289773013236702e-05, "loss": 1.0296, "step": 29235 }, { "epoch": 0.07, "learning_rate": 9.289647189088531e-05, "loss": 1.0317, "step": 29240 }, { "epoch": 0.07, "learning_rate": 9.28952136494036e-05, "loss": 1.0307, "step": 29245 }, { "epoch": 0.07, "learning_rate": 9.289395540792189e-05, "loss": 1.0306, "step": 29250 }, { "epoch": 0.07, "learning_rate": 9.28926971664402e-05, "loss": 1.0346, "step": 29255 }, { "epoch": 0.07, "learning_rate": 9.289143892495849e-05, "loss": 1.0321, "step": 29260 }, { "epoch": 0.07, "learning_rate": 9.289018068347678e-05, "loss": 1.0315, "step": 29265 }, { "epoch": 0.07, "learning_rate": 9.288892244199507e-05, "loss": 1.0311, "step": 29270 }, { "epoch": 0.07, "learning_rate": 9.288766420051336e-05, "loss": 1.0302, "step": 29275 }, { "epoch": 0.07, "learning_rate": 9.288640595903167e-05, "loss": 1.0336, "step": 29280 }, { "epoch": 0.07, "learning_rate": 9.288514771754996e-05, "loss": 1.0295, "step": 29285 }, { "epoch": 0.07, "learning_rate": 9.288388947606825e-05, "loss": 1.032, "step": 29290 }, { "epoch": 0.07, "learning_rate": 9.288263123458654e-05, "loss": 1.032, "step": 29295 }, { "epoch": 0.07, "learning_rate": 9.288137299310485e-05, "loss": 1.0316, "step": 29300 }, { "epoch": 0.07, "learning_rate": 9.288011475162314e-05, "loss": 1.0327, "step": 29305 }, { "epoch": 0.07, "learning_rate": 9.287885651014143e-05, "loss": 1.0307, "step": 29310 }, { "epoch": 0.07, "learning_rate": 9.287759826865972e-05, "loss": 1.0294, "step": 29315 }, { "epoch": 0.07, "learning_rate": 9.287634002717803e-05, "loss": 1.0309, "step": 29320 }, { "epoch": 0.07, "learning_rate": 9.287508178569632e-05, "loss": 1.0323, "step": 29325 }, { "epoch": 0.07, "learning_rate": 9.287382354421461e-05, "loss": 1.0328, "step": 29330 }, { "epoch": 0.07, "learning_rate": 9.28725653027329e-05, "loss": 1.0302, "step": 29335 }, { "epoch": 0.07, "learning_rate": 9.28713070612512e-05, "loss": 1.0342, "step": 29340 }, { "epoch": 0.07, "learning_rate": 9.28700488197695e-05, "loss": 1.0309, "step": 29345 }, { "epoch": 0.07, "learning_rate": 9.286879057828779e-05, "loss": 1.0521, "step": 29350 }, { "epoch": 0.07, "learning_rate": 9.286753233680608e-05, "loss": 1.03, "step": 29355 }, { "epoch": 0.07, "learning_rate": 9.286627409532437e-05, "loss": 1.0305, "step": 29360 }, { "epoch": 0.07, "learning_rate": 9.286501585384268e-05, "loss": 1.0309, "step": 29365 }, { "epoch": 0.07, "learning_rate": 9.286375761236097e-05, "loss": 1.0319, "step": 29370 }, { "epoch": 0.07, "learning_rate": 9.286249937087926e-05, "loss": 1.0307, "step": 29375 }, { "epoch": 0.07, "learning_rate": 9.286124112939755e-05, "loss": 1.0318, "step": 29380 }, { "epoch": 0.07, "learning_rate": 9.285998288791586e-05, "loss": 1.0317, "step": 29385 }, { "epoch": 0.07, "learning_rate": 9.285872464643415e-05, "loss": 1.0334, "step": 29390 }, { "epoch": 0.07, "learning_rate": 9.285746640495244e-05, "loss": 1.0307, "step": 29395 }, { "epoch": 0.07, "learning_rate": 9.285620816347073e-05, "loss": 1.0283, "step": 29400 }, { "epoch": 0.07, "learning_rate": 9.285494992198903e-05, "loss": 1.0298, "step": 29405 }, { "epoch": 0.07, "learning_rate": 9.285369168050733e-05, "loss": 1.033, "step": 29410 }, { "epoch": 0.07, "learning_rate": 9.285243343902562e-05, "loss": 1.0316, "step": 29415 }, { "epoch": 0.07, "learning_rate": 9.285117519754391e-05, "loss": 1.0316, "step": 29420 }, { "epoch": 0.07, "learning_rate": 9.28499169560622e-05, "loss": 1.0295, "step": 29425 }, { "epoch": 0.07, "learning_rate": 9.284865871458051e-05, "loss": 1.032, "step": 29430 }, { "epoch": 0.07, "learning_rate": 9.28474004730988e-05, "loss": 1.0308, "step": 29435 }, { "epoch": 0.07, "learning_rate": 9.28461422316171e-05, "loss": 1.0557, "step": 29440 }, { "epoch": 0.07, "learning_rate": 9.284488399013539e-05, "loss": 1.0325, "step": 29445 }, { "epoch": 0.07, "learning_rate": 9.284362574865369e-05, "loss": 1.0316, "step": 29450 }, { "epoch": 0.07, "learning_rate": 9.284236750717198e-05, "loss": 1.0304, "step": 29455 }, { "epoch": 0.07, "learning_rate": 9.284110926569027e-05, "loss": 1.0329, "step": 29460 }, { "epoch": 0.07, "learning_rate": 9.283985102420857e-05, "loss": 1.031, "step": 29465 }, { "epoch": 0.07, "learning_rate": 9.283859278272686e-05, "loss": 1.0328, "step": 29470 }, { "epoch": 0.07, "learning_rate": 9.283733454124516e-05, "loss": 1.0328, "step": 29475 }, { "epoch": 0.07, "learning_rate": 9.283607629976345e-05, "loss": 1.0338, "step": 29480 }, { "epoch": 0.07, "learning_rate": 9.283481805828175e-05, "loss": 1.0312, "step": 29485 }, { "epoch": 0.07, "learning_rate": 9.283355981680004e-05, "loss": 1.0317, "step": 29490 }, { "epoch": 0.07, "learning_rate": 9.283230157531834e-05, "loss": 1.0304, "step": 29495 }, { "epoch": 0.07, "learning_rate": 9.283104333383663e-05, "loss": 1.03, "step": 29500 }, { "epoch": 0.07, "learning_rate": 9.282978509235493e-05, "loss": 1.0302, "step": 29505 }, { "epoch": 0.07, "learning_rate": 9.282852685087322e-05, "loss": 1.0325, "step": 29510 }, { "epoch": 0.07, "learning_rate": 9.282752025768787e-05, "loss": 1.0525, "step": 29515 }, { "epoch": 0.07, "learning_rate": 9.282626201620616e-05, "loss": 1.0316, "step": 29520 }, { "epoch": 0.07, "learning_rate": 9.282500377472445e-05, "loss": 1.0312, "step": 29525 }, { "epoch": 0.07, "learning_rate": 9.282374553324274e-05, "loss": 1.0307, "step": 29530 }, { "epoch": 0.07, "learning_rate": 9.282248729176105e-05, "loss": 1.0323, "step": 29535 }, { "epoch": 0.07, "learning_rate": 9.282122905027934e-05, "loss": 1.0325, "step": 29540 }, { "epoch": 0.07, "learning_rate": 9.281997080879763e-05, "loss": 1.0311, "step": 29545 }, { "epoch": 0.07, "learning_rate": 9.281871256731592e-05, "loss": 1.032, "step": 29550 }, { "epoch": 0.07, "learning_rate": 9.281745432583421e-05, "loss": 1.0291, "step": 29555 }, { "epoch": 0.07, "learning_rate": 9.281619608435252e-05, "loss": 1.0328, "step": 29560 }, { "epoch": 0.07, "learning_rate": 9.281493784287081e-05, "loss": 1.0323, "step": 29565 }, { "epoch": 0.07, "learning_rate": 9.28136796013891e-05, "loss": 1.0324, "step": 29570 }, { "epoch": 0.07, "learning_rate": 9.281242135990739e-05, "loss": 1.0331, "step": 29575 }, { "epoch": 0.07, "learning_rate": 9.28111631184257e-05, "loss": 1.0314, "step": 29580 }, { "epoch": 0.07, "learning_rate": 9.280990487694399e-05, "loss": 1.0564, "step": 29585 }, { "epoch": 0.07, "learning_rate": 9.280864663546228e-05, "loss": 1.029, "step": 29590 }, { "epoch": 0.07, "learning_rate": 9.280738839398057e-05, "loss": 1.0336, "step": 29595 }, { "epoch": 0.07, "learning_rate": 9.280613015249888e-05, "loss": 1.0325, "step": 29600 }, { "epoch": 0.07, "learning_rate": 9.280487191101717e-05, "loss": 1.0318, "step": 29605 }, { "epoch": 0.07, "learning_rate": 9.280361366953546e-05, "loss": 1.0297, "step": 29610 }, { "epoch": 0.07, "learning_rate": 9.280235542805375e-05, "loss": 1.0313, "step": 29615 }, { "epoch": 0.07, "learning_rate": 9.280109718657204e-05, "loss": 1.0319, "step": 29620 }, { "epoch": 0.07, "learning_rate": 9.279983894509035e-05, "loss": 1.0306, "step": 29625 }, { "epoch": 0.07, "learning_rate": 9.279858070360864e-05, "loss": 1.0314, "step": 29630 }, { "epoch": 0.07, "learning_rate": 9.279732246212693e-05, "loss": 1.0303, "step": 29635 }, { "epoch": 0.07, "learning_rate": 9.279606422064522e-05, "loss": 1.0322, "step": 29640 }, { "epoch": 0.07, "learning_rate": 9.279480597916353e-05, "loss": 1.0326, "step": 29645 }, { "epoch": 0.07, "learning_rate": 9.279354773768182e-05, "loss": 1.0334, "step": 29650 }, { "epoch": 0.07, "learning_rate": 9.279228949620011e-05, "loss": 1.0337, "step": 29655 }, { "epoch": 0.07, "learning_rate": 9.27910312547184e-05, "loss": 1.0303, "step": 29660 }, { "epoch": 0.07, "learning_rate": 9.278977301323671e-05, "loss": 1.0314, "step": 29665 }, { "epoch": 0.07, "learning_rate": 9.2788514771755e-05, "loss": 1.0337, "step": 29670 }, { "epoch": 0.07, "learning_rate": 9.278725653027329e-05, "loss": 1.0311, "step": 29675 }, { "epoch": 0.07, "learning_rate": 9.278599828879158e-05, "loss": 1.0317, "step": 29680 }, { "epoch": 0.07, "learning_rate": 9.278474004730988e-05, "loss": 1.0326, "step": 29685 }, { "epoch": 0.07, "learning_rate": 9.278348180582818e-05, "loss": 1.0332, "step": 29690 }, { "epoch": 0.07, "learning_rate": 9.278222356434647e-05, "loss": 1.0299, "step": 29695 }, { "epoch": 0.07, "learning_rate": 9.278096532286476e-05, "loss": 1.0319, "step": 29700 }, { "epoch": 0.07, "learning_rate": 9.277970708138306e-05, "loss": 1.0301, "step": 29705 }, { "epoch": 0.07, "learning_rate": 9.277844883990136e-05, "loss": 1.0298, "step": 29710 }, { "epoch": 0.07, "learning_rate": 9.277719059841965e-05, "loss": 1.0322, "step": 29715 }, { "epoch": 0.07, "learning_rate": 9.277593235693794e-05, "loss": 1.0325, "step": 29720 }, { "epoch": 0.07, "learning_rate": 9.277467411545624e-05, "loss": 1.0303, "step": 29725 }, { "epoch": 0.07, "learning_rate": 9.277341587397454e-05, "loss": 1.0335, "step": 29730 }, { "epoch": 0.07, "learning_rate": 9.277215763249283e-05, "loss": 1.0309, "step": 29735 }, { "epoch": 0.07, "learning_rate": 9.277089939101112e-05, "loss": 1.0334, "step": 29740 }, { "epoch": 0.07, "learning_rate": 9.276964114952942e-05, "loss": 1.0319, "step": 29745 }, { "epoch": 0.07, "learning_rate": 9.276838290804771e-05, "loss": 1.0309, "step": 29750 }, { "epoch": 0.07, "learning_rate": 9.276712466656601e-05, "loss": 1.0317, "step": 29755 }, { "epoch": 0.07, "learning_rate": 9.276611807338065e-05, "loss": 1.0537, "step": 29760 }, { "epoch": 0.07, "learning_rate": 9.276485983189894e-05, "loss": 1.0309, "step": 29765 }, { "epoch": 0.07, "learning_rate": 9.276360159041723e-05, "loss": 1.0322, "step": 29770 }, { "epoch": 0.07, "learning_rate": 9.276234334893554e-05, "loss": 1.0312, "step": 29775 }, { "epoch": 0.07, "learning_rate": 9.276108510745383e-05, "loss": 1.0304, "step": 29780 }, { "epoch": 0.07, "learning_rate": 9.275982686597212e-05, "loss": 1.0316, "step": 29785 }, { "epoch": 0.07, "learning_rate": 9.275856862449041e-05, "loss": 1.0316, "step": 29790 }, { "epoch": 0.07, "learning_rate": 9.275731038300872e-05, "loss": 1.0317, "step": 29795 }, { "epoch": 0.07, "learning_rate": 9.275605214152701e-05, "loss": 1.03, "step": 29800 }, { "epoch": 0.07, "learning_rate": 9.27547939000453e-05, "loss": 1.0308, "step": 29805 }, { "epoch": 0.07, "learning_rate": 9.275353565856359e-05, "loss": 1.0329, "step": 29810 }, { "epoch": 0.07, "learning_rate": 9.27522774170819e-05, "loss": 1.0297, "step": 29815 }, { "epoch": 0.07, "learning_rate": 9.275101917560019e-05, "loss": 1.0315, "step": 29820 }, { "epoch": 0.07, "learning_rate": 9.274976093411848e-05, "loss": 1.0312, "step": 29825 }, { "epoch": 0.07, "learning_rate": 9.274850269263677e-05, "loss": 1.0302, "step": 29830 }, { "epoch": 0.07, "learning_rate": 9.274724445115506e-05, "loss": 1.0307, "step": 29835 }, { "epoch": 0.07, "learning_rate": 9.274598620967337e-05, "loss": 1.0312, "step": 29840 }, { "epoch": 0.07, "learning_rate": 9.274472796819166e-05, "loss": 1.0325, "step": 29845 }, { "epoch": 0.07, "learning_rate": 9.274346972670995e-05, "loss": 1.0319, "step": 29850 }, { "epoch": 0.07, "learning_rate": 9.274221148522824e-05, "loss": 1.0314, "step": 29855 }, { "epoch": 0.07, "learning_rate": 9.274095324374655e-05, "loss": 1.0298, "step": 29860 }, { "epoch": 0.07, "learning_rate": 9.273969500226484e-05, "loss": 1.0307, "step": 29865 }, { "epoch": 0.07, "learning_rate": 9.273843676078313e-05, "loss": 1.0344, "step": 29870 }, { "epoch": 0.07, "learning_rate": 9.273717851930142e-05, "loss": 1.0295, "step": 29875 }, { "epoch": 0.08, "learning_rate": 9.273592027781973e-05, "loss": 1.0283, "step": 29880 }, { "epoch": 0.08, "learning_rate": 9.273466203633802e-05, "loss": 1.0325, "step": 29885 }, { "epoch": 0.08, "learning_rate": 9.273340379485631e-05, "loss": 1.0275, "step": 29890 }, { "epoch": 0.08, "learning_rate": 9.27321455533746e-05, "loss": 1.029, "step": 29895 }, { "epoch": 0.08, "learning_rate": 9.27308873118929e-05, "loss": 1.0314, "step": 29900 }, { "epoch": 0.08, "learning_rate": 9.27296290704112e-05, "loss": 1.0318, "step": 29905 }, { "epoch": 0.08, "learning_rate": 9.272837082892949e-05, "loss": 1.0305, "step": 29910 }, { "epoch": 0.08, "learning_rate": 9.272711258744778e-05, "loss": 1.0327, "step": 29915 }, { "epoch": 0.08, "learning_rate": 9.272585434596607e-05, "loss": 1.0313, "step": 29920 }, { "epoch": 0.08, "learning_rate": 9.272459610448438e-05, "loss": 1.032, "step": 29925 }, { "epoch": 0.08, "learning_rate": 9.272333786300267e-05, "loss": 1.0298, "step": 29930 }, { "epoch": 0.08, "learning_rate": 9.272207962152096e-05, "loss": 1.0313, "step": 29935 }, { "epoch": 0.08, "learning_rate": 9.272082138003925e-05, "loss": 1.0296, "step": 29940 }, { "epoch": 0.08, "learning_rate": 9.271956313855756e-05, "loss": 1.032, "step": 29945 }, { "epoch": 0.08, "learning_rate": 9.271830489707585e-05, "loss": 1.032, "step": 29950 }, { "epoch": 0.08, "learning_rate": 9.271704665559414e-05, "loss": 1.0304, "step": 29955 }, { "epoch": 0.08, "learning_rate": 9.271578841411243e-05, "loss": 1.03, "step": 29960 }, { "epoch": 0.08, "learning_rate": 9.271453017263073e-05, "loss": 1.0311, "step": 29965 }, { "epoch": 0.08, "learning_rate": 9.271327193114903e-05, "loss": 1.0315, "step": 29970 }, { "epoch": 0.08, "learning_rate": 9.271201368966732e-05, "loss": 1.0309, "step": 29975 }, { "epoch": 0.08, "learning_rate": 9.271075544818561e-05, "loss": 1.0323, "step": 29980 }, { "epoch": 0.08, "learning_rate": 9.27094972067039e-05, "loss": 1.0345, "step": 29985 }, { "epoch": 0.08, "learning_rate": 9.270823896522221e-05, "loss": 1.0302, "step": 29990 }, { "epoch": 0.08, "learning_rate": 9.27069807237405e-05, "loss": 1.0304, "step": 29995 }, { "epoch": 0.08, "learning_rate": 9.27057224822588e-05, "loss": 1.0314, "step": 30000 }, { "epoch": 0.08, "learning_rate": 9.270446424077709e-05, "loss": 1.0324, "step": 30005 }, { "epoch": 0.08, "learning_rate": 9.270320599929538e-05, "loss": 1.032, "step": 30010 }, { "epoch": 0.08, "learning_rate": 9.270194775781368e-05, "loss": 1.03, "step": 30015 }, { "epoch": 0.08, "learning_rate": 9.270068951633199e-05, "loss": 1.0304, "step": 30020 }, { "epoch": 0.08, "learning_rate": 9.269943127485028e-05, "loss": 1.031, "step": 30025 }, { "epoch": 0.08, "learning_rate": 9.269817303336857e-05, "loss": 1.0315, "step": 30030 }, { "epoch": 0.08, "learning_rate": 9.269691479188686e-05, "loss": 1.031, "step": 30035 }, { "epoch": 0.08, "learning_rate": 9.269565655040517e-05, "loss": 1.0337, "step": 30040 }, { "epoch": 0.08, "learning_rate": 9.269439830892346e-05, "loss": 1.0308, "step": 30045 }, { "epoch": 0.08, "learning_rate": 9.269314006744175e-05, "loss": 1.0316, "step": 30050 }, { "epoch": 0.08, "learning_rate": 9.269188182596004e-05, "loss": 1.0334, "step": 30055 }, { "epoch": 0.08, "learning_rate": 9.269062358447835e-05, "loss": 1.0286, "step": 30060 }, { "epoch": 0.08, "learning_rate": 9.268936534299664e-05, "loss": 1.0324, "step": 30065 }, { "epoch": 0.08, "learning_rate": 9.268810710151493e-05, "loss": 1.0325, "step": 30070 }, { "epoch": 0.08, "learning_rate": 9.268684886003322e-05, "loss": 1.0305, "step": 30075 }, { "epoch": 0.08, "learning_rate": 9.268559061855153e-05, "loss": 1.0328, "step": 30080 }, { "epoch": 0.08, "learning_rate": 9.268433237706982e-05, "loss": 1.0302, "step": 30085 }, { "epoch": 0.08, "learning_rate": 9.268307413558811e-05, "loss": 1.031, "step": 30090 }, { "epoch": 0.08, "learning_rate": 9.26818158941064e-05, "loss": 1.0307, "step": 30095 }, { "epoch": 0.08, "learning_rate": 9.26805576526247e-05, "loss": 1.0296, "step": 30100 }, { "epoch": 0.08, "learning_rate": 9.2679299411143e-05, "loss": 1.0327, "step": 30105 }, { "epoch": 0.08, "learning_rate": 9.267804116966129e-05, "loss": 1.0315, "step": 30110 }, { "epoch": 0.08, "learning_rate": 9.267678292817958e-05, "loss": 1.0306, "step": 30115 }, { "epoch": 0.08, "learning_rate": 9.267552468669787e-05, "loss": 1.0312, "step": 30120 }, { "epoch": 0.08, "learning_rate": 9.267426644521618e-05, "loss": 1.0547, "step": 30125 }, { "epoch": 0.08, "learning_rate": 9.267300820373447e-05, "loss": 1.0321, "step": 30130 }, { "epoch": 0.08, "learning_rate": 9.267174996225276e-05, "loss": 1.0306, "step": 30135 }, { "epoch": 0.08, "learning_rate": 9.267049172077105e-05, "loss": 1.0311, "step": 30140 }, { "epoch": 0.08, "learning_rate": 9.266923347928936e-05, "loss": 1.0312, "step": 30145 }, { "epoch": 0.08, "learning_rate": 9.266797523780765e-05, "loss": 1.029, "step": 30150 }, { "epoch": 0.08, "learning_rate": 9.266671699632594e-05, "loss": 1.0334, "step": 30155 }, { "epoch": 0.08, "learning_rate": 9.266545875484423e-05, "loss": 1.0542, "step": 30160 }, { "epoch": 0.08, "learning_rate": 9.266420051336253e-05, "loss": 1.0321, "step": 30165 }, { "epoch": 0.08, "learning_rate": 9.266294227188083e-05, "loss": 1.0294, "step": 30170 }, { "epoch": 0.08, "learning_rate": 9.266168403039912e-05, "loss": 1.0295, "step": 30175 }, { "epoch": 0.08, "learning_rate": 9.266042578891741e-05, "loss": 1.0304, "step": 30180 }, { "epoch": 0.08, "learning_rate": 9.26591675474357e-05, "loss": 1.0318, "step": 30185 }, { "epoch": 0.08, "learning_rate": 9.265790930595401e-05, "loss": 1.029, "step": 30190 }, { "epoch": 0.08, "learning_rate": 9.26566510644723e-05, "loss": 1.0327, "step": 30195 }, { "epoch": 0.08, "learning_rate": 9.26553928229906e-05, "loss": 1.0304, "step": 30200 }, { "epoch": 0.08, "learning_rate": 9.265413458150888e-05, "loss": 1.0317, "step": 30205 }, { "epoch": 0.08, "learning_rate": 9.265287634002719e-05, "loss": 1.0306, "step": 30210 }, { "epoch": 0.08, "learning_rate": 9.265161809854548e-05, "loss": 1.0315, "step": 30215 }, { "epoch": 0.08, "learning_rate": 9.265035985706377e-05, "loss": 1.0298, "step": 30220 }, { "epoch": 0.08, "learning_rate": 9.264910161558206e-05, "loss": 1.031, "step": 30225 }, { "epoch": 0.08, "learning_rate": 9.264784337410036e-05, "loss": 1.0324, "step": 30230 }, { "epoch": 0.08, "learning_rate": 9.264658513261866e-05, "loss": 1.0476, "step": 30235 }, { "epoch": 0.08, "learning_rate": 9.264532689113695e-05, "loss": 1.0321, "step": 30240 }, { "epoch": 0.08, "learning_rate": 9.264406864965524e-05, "loss": 1.0316, "step": 30245 }, { "epoch": 0.08, "learning_rate": 9.264281040817354e-05, "loss": 1.0312, "step": 30250 }, { "epoch": 0.08, "learning_rate": 9.264155216669184e-05, "loss": 1.0322, "step": 30255 }, { "epoch": 0.08, "learning_rate": 9.264029392521013e-05, "loss": 1.0309, "step": 30260 }, { "epoch": 0.08, "learning_rate": 9.263903568372842e-05, "loss": 1.0307, "step": 30265 }, { "epoch": 0.08, "learning_rate": 9.263777744224672e-05, "loss": 1.03, "step": 30270 }, { "epoch": 0.08, "learning_rate": 9.263651920076502e-05, "loss": 1.031, "step": 30275 }, { "epoch": 0.08, "learning_rate": 9.263526095928331e-05, "loss": 1.0317, "step": 30280 }, { "epoch": 0.08, "learning_rate": 9.26340027178016e-05, "loss": 1.0292, "step": 30285 }, { "epoch": 0.08, "learning_rate": 9.26327444763199e-05, "loss": 1.0287, "step": 30290 }, { "epoch": 0.08, "learning_rate": 9.263148623483819e-05, "loss": 1.0285, "step": 30295 }, { "epoch": 0.08, "learning_rate": 9.263022799335649e-05, "loss": 1.0311, "step": 30300 }, { "epoch": 0.08, "learning_rate": 9.262896975187478e-05, "loss": 1.032, "step": 30305 }, { "epoch": 0.08, "learning_rate": 9.262771151039308e-05, "loss": 1.0296, "step": 30310 }, { "epoch": 0.08, "learning_rate": 9.262645326891137e-05, "loss": 1.031, "step": 30315 }, { "epoch": 0.08, "learning_rate": 9.262519502742967e-05, "loss": 1.0327, "step": 30320 }, { "epoch": 0.08, "learning_rate": 9.262393678594796e-05, "loss": 1.0491, "step": 30325 }, { "epoch": 0.08, "learning_rate": 9.262267854446626e-05, "loss": 1.0307, "step": 30330 }, { "epoch": 0.08, "learning_rate": 9.262142030298455e-05, "loss": 1.0325, "step": 30335 }, { "epoch": 0.08, "learning_rate": 9.262016206150285e-05, "loss": 1.0306, "step": 30340 }, { "epoch": 0.08, "learning_rate": 9.261890382002114e-05, "loss": 1.0305, "step": 30345 }, { "epoch": 0.08, "learning_rate": 9.261764557853944e-05, "loss": 1.0314, "step": 30350 }, { "epoch": 0.08, "learning_rate": 9.261638733705773e-05, "loss": 1.0315, "step": 30355 }, { "epoch": 0.08, "learning_rate": 9.261512909557602e-05, "loss": 1.0314, "step": 30360 }, { "epoch": 0.08, "learning_rate": 9.261387085409432e-05, "loss": 1.0324, "step": 30365 }, { "epoch": 0.08, "learning_rate": 9.261261261261262e-05, "loss": 1.0297, "step": 30370 }, { "epoch": 0.08, "learning_rate": 9.261135437113091e-05, "loss": 1.0329, "step": 30375 }, { "epoch": 0.08, "learning_rate": 9.26100961296492e-05, "loss": 1.0293, "step": 30380 }, { "epoch": 0.08, "learning_rate": 9.26088378881675e-05, "loss": 1.0308, "step": 30385 }, { "epoch": 0.08, "learning_rate": 9.26075796466858e-05, "loss": 1.0328, "step": 30390 }, { "epoch": 0.08, "learning_rate": 9.260632140520409e-05, "loss": 1.0306, "step": 30395 }, { "epoch": 0.08, "learning_rate": 9.260506316372238e-05, "loss": 1.0303, "step": 30400 }, { "epoch": 0.08, "learning_rate": 9.260380492224068e-05, "loss": 1.0343, "step": 30405 }, { "epoch": 0.08, "learning_rate": 9.260254668075898e-05, "loss": 1.0328, "step": 30410 }, { "epoch": 0.08, "learning_rate": 9.260128843927727e-05, "loss": 1.0336, "step": 30415 }, { "epoch": 0.08, "learning_rate": 9.260003019779556e-05, "loss": 1.0296, "step": 30420 }, { "epoch": 0.08, "learning_rate": 9.259877195631385e-05, "loss": 1.0296, "step": 30425 }, { "epoch": 0.08, "learning_rate": 9.259751371483216e-05, "loss": 1.0299, "step": 30430 }, { "epoch": 0.08, "learning_rate": 9.259625547335045e-05, "loss": 1.0325, "step": 30435 }, { "epoch": 0.08, "learning_rate": 9.259499723186874e-05, "loss": 1.0309, "step": 30440 }, { "epoch": 0.08, "learning_rate": 9.259373899038703e-05, "loss": 1.0311, "step": 30445 }, { "epoch": 0.08, "learning_rate": 9.259248074890534e-05, "loss": 1.0322, "step": 30450 }, { "epoch": 0.08, "learning_rate": 9.259122250742363e-05, "loss": 1.0304, "step": 30455 }, { "epoch": 0.08, "learning_rate": 9.258996426594192e-05, "loss": 1.0321, "step": 30460 }, { "epoch": 0.08, "learning_rate": 9.258870602446021e-05, "loss": 1.0322, "step": 30465 }, { "epoch": 0.08, "learning_rate": 9.258744778297852e-05, "loss": 1.0297, "step": 30470 }, { "epoch": 0.08, "learning_rate": 9.258618954149681e-05, "loss": 1.0331, "step": 30475 }, { "epoch": 0.08, "learning_rate": 9.25849313000151e-05, "loss": 1.0291, "step": 30480 }, { "epoch": 0.08, "learning_rate": 9.258367305853339e-05, "loss": 1.031, "step": 30485 }, { "epoch": 0.08, "learning_rate": 9.258241481705168e-05, "loss": 1.0462, "step": 30490 }, { "epoch": 0.08, "learning_rate": 9.258115657556999e-05, "loss": 1.033, "step": 30495 }, { "epoch": 0.08, "learning_rate": 9.257989833408828e-05, "loss": 1.0325, "step": 30500 }, { "epoch": 0.08, "learning_rate": 9.257864009260657e-05, "loss": 1.0575, "step": 30505 }, { "epoch": 0.08, "learning_rate": 9.257738185112486e-05, "loss": 1.0311, "step": 30510 }, { "epoch": 0.08, "learning_rate": 9.257612360964317e-05, "loss": 1.0316, "step": 30515 }, { "epoch": 0.08, "learning_rate": 9.257486536816147e-05, "loss": 1.0315, "step": 30520 }, { "epoch": 0.08, "learning_rate": 9.257360712667976e-05, "loss": 1.0306, "step": 30525 }, { "epoch": 0.08, "learning_rate": 9.257234888519806e-05, "loss": 1.0327, "step": 30530 }, { "epoch": 0.08, "learning_rate": 9.257109064371635e-05, "loss": 1.0318, "step": 30535 }, { "epoch": 0.08, "learning_rate": 9.256983240223465e-05, "loss": 1.0299, "step": 30540 }, { "epoch": 0.08, "learning_rate": 9.256857416075294e-05, "loss": 1.0299, "step": 30545 }, { "epoch": 0.08, "learning_rate": 9.256731591927124e-05, "loss": 1.0315, "step": 30550 }, { "epoch": 0.08, "learning_rate": 9.256605767778953e-05, "loss": 1.0312, "step": 30555 }, { "epoch": 0.08, "learning_rate": 9.256479943630782e-05, "loss": 1.0306, "step": 30560 }, { "epoch": 0.08, "learning_rate": 9.256354119482612e-05, "loss": 1.0321, "step": 30565 }, { "epoch": 0.08, "learning_rate": 9.256228295334442e-05, "loss": 1.0312, "step": 30570 }, { "epoch": 0.08, "learning_rate": 9.256102471186271e-05, "loss": 1.0334, "step": 30575 }, { "epoch": 0.08, "learning_rate": 9.2559766470381e-05, "loss": 1.0291, "step": 30580 }, { "epoch": 0.08, "learning_rate": 9.25585082288993e-05, "loss": 1.029, "step": 30585 }, { "epoch": 0.08, "learning_rate": 9.25572499874176e-05, "loss": 1.03, "step": 30590 }, { "epoch": 0.08, "learning_rate": 9.255599174593589e-05, "loss": 1.0324, "step": 30595 }, { "epoch": 0.08, "learning_rate": 9.255473350445418e-05, "loss": 1.0309, "step": 30600 }, { "epoch": 0.08, "learning_rate": 9.255347526297248e-05, "loss": 1.0313, "step": 30605 }, { "epoch": 0.08, "learning_rate": 9.255221702149078e-05, "loss": 1.0305, "step": 30610 }, { "epoch": 0.08, "learning_rate": 9.255095878000907e-05, "loss": 1.0319, "step": 30615 }, { "epoch": 0.08, "learning_rate": 9.254970053852736e-05, "loss": 1.0498, "step": 30620 }, { "epoch": 0.08, "learning_rate": 9.254844229704565e-05, "loss": 1.0304, "step": 30625 }, { "epoch": 0.08, "learning_rate": 9.254718405556396e-05, "loss": 1.0301, "step": 30630 }, { "epoch": 0.08, "learning_rate": 9.254592581408225e-05, "loss": 1.0325, "step": 30635 }, { "epoch": 0.08, "learning_rate": 9.254466757260054e-05, "loss": 1.0322, "step": 30640 }, { "epoch": 0.08, "learning_rate": 9.254340933111883e-05, "loss": 1.0311, "step": 30645 }, { "epoch": 0.08, "learning_rate": 9.254215108963714e-05, "loss": 1.031, "step": 30650 }, { "epoch": 0.08, "learning_rate": 9.254089284815543e-05, "loss": 1.0314, "step": 30655 }, { "epoch": 0.08, "learning_rate": 9.253963460667372e-05, "loss": 1.0322, "step": 30660 }, { "epoch": 0.08, "learning_rate": 9.253837636519201e-05, "loss": 1.0556, "step": 30665 }, { "epoch": 0.08, "learning_rate": 9.253711812371032e-05, "loss": 1.0297, "step": 30670 }, { "epoch": 0.08, "learning_rate": 9.253585988222861e-05, "loss": 1.032, "step": 30675 }, { "epoch": 0.08, "learning_rate": 9.25346016407469e-05, "loss": 1.0324, "step": 30680 }, { "epoch": 0.08, "learning_rate": 9.253334339926519e-05, "loss": 1.0343, "step": 30685 }, { "epoch": 0.08, "learning_rate": 9.253208515778348e-05, "loss": 1.0306, "step": 30690 }, { "epoch": 0.08, "learning_rate": 9.253082691630179e-05, "loss": 1.0302, "step": 30695 }, { "epoch": 0.08, "learning_rate": 9.252956867482008e-05, "loss": 1.0315, "step": 30700 }, { "epoch": 0.08, "learning_rate": 9.252831043333837e-05, "loss": 1.0329, "step": 30705 }, { "epoch": 0.08, "learning_rate": 9.252705219185666e-05, "loss": 1.0312, "step": 30710 }, { "epoch": 0.08, "learning_rate": 9.252579395037497e-05, "loss": 1.032, "step": 30715 }, { "epoch": 0.08, "learning_rate": 9.252453570889326e-05, "loss": 1.0317, "step": 30720 }, { "epoch": 0.08, "learning_rate": 9.252327746741155e-05, "loss": 1.0295, "step": 30725 }, { "epoch": 0.08, "learning_rate": 9.252201922592984e-05, "loss": 1.0326, "step": 30730 }, { "epoch": 0.08, "learning_rate": 9.252076098444815e-05, "loss": 1.0292, "step": 30735 }, { "epoch": 0.08, "learning_rate": 9.251950274296644e-05, "loss": 1.0328, "step": 30740 }, { "epoch": 0.08, "learning_rate": 9.251824450148473e-05, "loss": 1.0304, "step": 30745 }, { "epoch": 0.08, "learning_rate": 9.251698626000302e-05, "loss": 1.031, "step": 30750 }, { "epoch": 0.08, "learning_rate": 9.251572801852131e-05, "loss": 1.0328, "step": 30755 }, { "epoch": 0.08, "learning_rate": 9.251446977703962e-05, "loss": 1.0301, "step": 30760 }, { "epoch": 0.08, "learning_rate": 9.251321153555791e-05, "loss": 1.0321, "step": 30765 }, { "epoch": 0.08, "learning_rate": 9.25119532940762e-05, "loss": 1.0285, "step": 30770 }, { "epoch": 0.08, "learning_rate": 9.251069505259449e-05, "loss": 1.0309, "step": 30775 }, { "epoch": 0.08, "learning_rate": 9.25094368111128e-05, "loss": 1.0309, "step": 30780 }, { "epoch": 0.08, "learning_rate": 9.250817856963109e-05, "loss": 1.031, "step": 30785 }, { "epoch": 0.08, "learning_rate": 9.250692032814938e-05, "loss": 1.0311, "step": 30790 }, { "epoch": 0.08, "learning_rate": 9.250566208666767e-05, "loss": 1.031, "step": 30795 }, { "epoch": 0.08, "learning_rate": 9.250440384518598e-05, "loss": 1.0302, "step": 30800 }, { "epoch": 0.08, "learning_rate": 9.250314560370427e-05, "loss": 1.0326, "step": 30805 }, { "epoch": 0.08, "learning_rate": 9.250188736222256e-05, "loss": 1.0293, "step": 30810 }, { "epoch": 0.08, "learning_rate": 9.250062912074085e-05, "loss": 1.056, "step": 30815 }, { "epoch": 0.08, "learning_rate": 9.249937087925914e-05, "loss": 1.0333, "step": 30820 }, { "epoch": 0.08, "learning_rate": 9.249811263777745e-05, "loss": 1.0321, "step": 30825 }, { "epoch": 0.08, "learning_rate": 9.249685439629574e-05, "loss": 1.0303, "step": 30830 }, { "epoch": 0.08, "learning_rate": 9.249559615481403e-05, "loss": 1.032, "step": 30835 }, { "epoch": 0.08, "learning_rate": 9.249433791333232e-05, "loss": 1.0324, "step": 30840 }, { "epoch": 0.08, "learning_rate": 9.249307967185063e-05, "loss": 1.0316, "step": 30845 }, { "epoch": 0.08, "learning_rate": 9.249182143036892e-05, "loss": 1.0319, "step": 30850 }, { "epoch": 0.08, "learning_rate": 9.249056318888721e-05, "loss": 1.0302, "step": 30855 }, { "epoch": 0.08, "learning_rate": 9.24893049474055e-05, "loss": 1.0315, "step": 30860 }, { "epoch": 0.08, "learning_rate": 9.248804670592381e-05, "loss": 1.0299, "step": 30865 }, { "epoch": 0.08, "learning_rate": 9.24867884644421e-05, "loss": 1.0304, "step": 30870 }, { "epoch": 0.08, "learning_rate": 9.248553022296039e-05, "loss": 1.0307, "step": 30875 }, { "epoch": 0.08, "learning_rate": 9.248427198147868e-05, "loss": 1.0289, "step": 30880 }, { "epoch": 0.08, "learning_rate": 9.248301373999698e-05, "loss": 1.03, "step": 30885 }, { "epoch": 0.08, "learning_rate": 9.248175549851528e-05, "loss": 1.0298, "step": 30890 }, { "epoch": 0.08, "learning_rate": 9.248049725703357e-05, "loss": 1.0301, "step": 30895 }, { "epoch": 0.08, "learning_rate": 9.247923901555186e-05, "loss": 1.0305, "step": 30900 }, { "epoch": 0.08, "learning_rate": 9.247798077407016e-05, "loss": 1.0299, "step": 30905 }, { "epoch": 0.08, "learning_rate": 9.247672253258846e-05, "loss": 1.0309, "step": 30910 }, { "epoch": 0.08, "learning_rate": 9.247546429110675e-05, "loss": 1.0298, "step": 30915 }, { "epoch": 0.08, "learning_rate": 9.247420604962504e-05, "loss": 1.0328, "step": 30920 }, { "epoch": 0.08, "learning_rate": 9.247294780814334e-05, "loss": 1.0306, "step": 30925 }, { "epoch": 0.08, "learning_rate": 9.247168956666164e-05, "loss": 1.0316, "step": 30930 }, { "epoch": 0.08, "learning_rate": 9.247043132517993e-05, "loss": 1.0296, "step": 30935 }, { "epoch": 0.08, "learning_rate": 9.246917308369822e-05, "loss": 1.0329, "step": 30940 }, { "epoch": 0.08, "learning_rate": 9.246791484221652e-05, "loss": 1.0283, "step": 30945 }, { "epoch": 0.08, "learning_rate": 9.246665660073481e-05, "loss": 1.0312, "step": 30950 }, { "epoch": 0.08, "learning_rate": 9.246539835925311e-05, "loss": 1.0322, "step": 30955 }, { "epoch": 0.08, "learning_rate": 9.24641401177714e-05, "loss": 1.0295, "step": 30960 }, { "epoch": 0.08, "learning_rate": 9.24628818762897e-05, "loss": 1.0334, "step": 30965 }, { "epoch": 0.08, "learning_rate": 9.246162363480799e-05, "loss": 1.03, "step": 30970 }, { "epoch": 0.08, "learning_rate": 9.246036539332629e-05, "loss": 1.0309, "step": 30975 }, { "epoch": 0.08, "learning_rate": 9.245910715184458e-05, "loss": 1.0342, "step": 30980 }, { "epoch": 0.08, "learning_rate": 9.245784891036288e-05, "loss": 1.0316, "step": 30985 }, { "epoch": 0.08, "learning_rate": 9.245659066888117e-05, "loss": 1.0307, "step": 30990 }, { "epoch": 0.08, "learning_rate": 9.245533242739947e-05, "loss": 1.0316, "step": 30995 }, { "epoch": 0.08, "learning_rate": 9.245407418591776e-05, "loss": 1.0316, "step": 31000 }, { "epoch": 0.08, "learning_rate": 9.245281594443606e-05, "loss": 1.03, "step": 31005 }, { "epoch": 0.08, "learning_rate": 9.245155770295435e-05, "loss": 1.0305, "step": 31010 }, { "epoch": 0.08, "learning_rate": 9.245029946147264e-05, "loss": 1.0305, "step": 31015 }, { "epoch": 0.08, "learning_rate": 9.244904121999094e-05, "loss": 1.0305, "step": 31020 }, { "epoch": 0.08, "learning_rate": 9.244778297850925e-05, "loss": 1.0321, "step": 31025 }, { "epoch": 0.08, "learning_rate": 9.244652473702754e-05, "loss": 1.0323, "step": 31030 }, { "epoch": 0.08, "learning_rate": 9.244526649554583e-05, "loss": 1.0293, "step": 31035 }, { "epoch": 0.08, "learning_rate": 9.244400825406412e-05, "loss": 1.0314, "step": 31040 }, { "epoch": 0.08, "learning_rate": 9.244275001258243e-05, "loss": 1.0323, "step": 31045 }, { "epoch": 0.08, "learning_rate": 9.244149177110072e-05, "loss": 1.029, "step": 31050 }, { "epoch": 0.08, "learning_rate": 9.244023352961901e-05, "loss": 1.0308, "step": 31055 }, { "epoch": 0.08, "learning_rate": 9.24389752881373e-05, "loss": 1.0309, "step": 31060 }, { "epoch": 0.08, "learning_rate": 9.243771704665561e-05, "loss": 1.0325, "step": 31065 }, { "epoch": 0.08, "learning_rate": 9.24364588051739e-05, "loss": 1.0303, "step": 31070 }, { "epoch": 0.08, "learning_rate": 9.243520056369219e-05, "loss": 1.0479, "step": 31075 }, { "epoch": 0.08, "learning_rate": 9.243394232221048e-05, "loss": 1.0318, "step": 31080 }, { "epoch": 0.08, "learning_rate": 9.243268408072878e-05, "loss": 1.032, "step": 31085 }, { "epoch": 0.08, "learning_rate": 9.243142583924708e-05, "loss": 1.0322, "step": 31090 }, { "epoch": 0.08, "learning_rate": 9.243016759776537e-05, "loss": 1.0322, "step": 31095 }, { "epoch": 0.08, "learning_rate": 9.242890935628366e-05, "loss": 1.0294, "step": 31100 }, { "epoch": 0.08, "learning_rate": 9.242765111480196e-05, "loss": 1.0326, "step": 31105 }, { "epoch": 0.08, "learning_rate": 9.242639287332026e-05, "loss": 1.0326, "step": 31110 }, { "epoch": 0.08, "learning_rate": 9.242513463183855e-05, "loss": 1.031, "step": 31115 }, { "epoch": 0.08, "learning_rate": 9.242387639035684e-05, "loss": 1.0319, "step": 31120 }, { "epoch": 0.08, "learning_rate": 9.242261814887514e-05, "loss": 1.0296, "step": 31125 }, { "epoch": 0.08, "learning_rate": 9.242135990739344e-05, "loss": 1.0302, "step": 31130 }, { "epoch": 0.08, "learning_rate": 9.242010166591173e-05, "loss": 1.0303, "step": 31135 }, { "epoch": 0.08, "learning_rate": 9.241884342443002e-05, "loss": 1.0291, "step": 31140 }, { "epoch": 0.08, "learning_rate": 9.241758518294832e-05, "loss": 1.0318, "step": 31145 }, { "epoch": 0.08, "learning_rate": 9.24163269414666e-05, "loss": 1.0318, "step": 31150 }, { "epoch": 0.08, "learning_rate": 9.241506869998491e-05, "loss": 1.0314, "step": 31155 }, { "epoch": 0.08, "learning_rate": 9.24138104585032e-05, "loss": 1.0322, "step": 31160 }, { "epoch": 0.08, "learning_rate": 9.24125522170215e-05, "loss": 1.0309, "step": 31165 }, { "epoch": 0.08, "learning_rate": 9.241129397553979e-05, "loss": 1.0295, "step": 31170 }, { "epoch": 0.08, "learning_rate": 9.241003573405809e-05, "loss": 1.0304, "step": 31175 }, { "epoch": 0.08, "learning_rate": 9.240877749257638e-05, "loss": 1.0299, "step": 31180 }, { "epoch": 0.08, "learning_rate": 9.240751925109468e-05, "loss": 1.0317, "step": 31185 }, { "epoch": 0.08, "learning_rate": 9.240626100961297e-05, "loss": 1.032, "step": 31190 }, { "epoch": 0.08, "learning_rate": 9.240500276813127e-05, "loss": 1.0308, "step": 31195 }, { "epoch": 0.08, "learning_rate": 9.240374452664956e-05, "loss": 1.0308, "step": 31200 }, { "epoch": 0.08, "learning_rate": 9.240248628516785e-05, "loss": 1.0323, "step": 31205 }, { "epoch": 0.08, "learning_rate": 9.240122804368615e-05, "loss": 1.0299, "step": 31210 }, { "epoch": 0.08, "learning_rate": 9.239996980220444e-05, "loss": 1.0325, "step": 31215 }, { "epoch": 0.08, "learning_rate": 9.239871156072274e-05, "loss": 1.0317, "step": 31220 }, { "epoch": 0.08, "learning_rate": 9.239745331924103e-05, "loss": 1.0314, "step": 31225 }, { "epoch": 0.08, "learning_rate": 9.239619507775933e-05, "loss": 1.032, "step": 31230 }, { "epoch": 0.08, "learning_rate": 9.239493683627762e-05, "loss": 1.0312, "step": 31235 }, { "epoch": 0.08, "learning_rate": 9.239367859479592e-05, "loss": 1.0333, "step": 31240 }, { "epoch": 0.08, "learning_rate": 9.239242035331421e-05, "loss": 1.0304, "step": 31245 }, { "epoch": 0.08, "learning_rate": 9.23911621118325e-05, "loss": 1.0302, "step": 31250 }, { "epoch": 0.08, "learning_rate": 9.23899038703508e-05, "loss": 1.0285, "step": 31255 }, { "epoch": 0.08, "learning_rate": 9.23886456288691e-05, "loss": 1.0305, "step": 31260 }, { "epoch": 0.08, "learning_rate": 9.23873873873874e-05, "loss": 1.03, "step": 31265 }, { "epoch": 0.08, "learning_rate": 9.238612914590569e-05, "loss": 1.033, "step": 31270 }, { "epoch": 0.08, "learning_rate": 9.238487090442398e-05, "loss": 1.0333, "step": 31275 }, { "epoch": 0.08, "learning_rate": 9.238361266294227e-05, "loss": 1.0307, "step": 31280 }, { "epoch": 0.08, "learning_rate": 9.238235442146057e-05, "loss": 1.0315, "step": 31285 }, { "epoch": 0.08, "learning_rate": 9.238109617997887e-05, "loss": 1.031, "step": 31290 }, { "epoch": 0.08, "learning_rate": 9.237983793849716e-05, "loss": 1.0319, "step": 31295 }, { "epoch": 0.08, "learning_rate": 9.237857969701545e-05, "loss": 1.0305, "step": 31300 }, { "epoch": 0.08, "learning_rate": 9.237732145553375e-05, "loss": 1.0298, "step": 31305 }, { "epoch": 0.08, "learning_rate": 9.237606321405205e-05, "loss": 1.029, "step": 31310 }, { "epoch": 0.08, "learning_rate": 9.237480497257034e-05, "loss": 1.0307, "step": 31315 }, { "epoch": 0.08, "learning_rate": 9.237354673108863e-05, "loss": 1.0327, "step": 31320 }, { "epoch": 0.08, "learning_rate": 9.237228848960693e-05, "loss": 1.0293, "step": 31325 }, { "epoch": 0.08, "learning_rate": 9.237103024812523e-05, "loss": 1.0321, "step": 31330 }, { "epoch": 0.08, "learning_rate": 9.236977200664352e-05, "loss": 1.031, "step": 31335 }, { "epoch": 0.08, "learning_rate": 9.236851376516181e-05, "loss": 1.0309, "step": 31340 }, { "epoch": 0.08, "learning_rate": 9.23672555236801e-05, "loss": 1.0289, "step": 31345 }, { "epoch": 0.08, "learning_rate": 9.23659972821984e-05, "loss": 1.0294, "step": 31350 }, { "epoch": 0.08, "learning_rate": 9.23647390407167e-05, "loss": 1.0299, "step": 31355 }, { "epoch": 0.08, "learning_rate": 9.236348079923499e-05, "loss": 1.0304, "step": 31360 }, { "epoch": 0.08, "learning_rate": 9.236222255775328e-05, "loss": 1.0279, "step": 31365 }, { "epoch": 0.08, "learning_rate": 9.236096431627159e-05, "loss": 1.0323, "step": 31370 }, { "epoch": 0.08, "learning_rate": 9.235970607478988e-05, "loss": 1.0294, "step": 31375 }, { "epoch": 0.08, "learning_rate": 9.235844783330817e-05, "loss": 1.0343, "step": 31380 }, { "epoch": 0.08, "learning_rate": 9.235718959182646e-05, "loss": 1.0317, "step": 31385 }, { "epoch": 0.08, "learning_rate": 9.235593135034477e-05, "loss": 1.031, "step": 31390 }, { "epoch": 0.08, "learning_rate": 9.235467310886306e-05, "loss": 1.0313, "step": 31395 }, { "epoch": 0.08, "learning_rate": 9.235341486738135e-05, "loss": 1.0497, "step": 31400 }, { "epoch": 0.08, "learning_rate": 9.235215662589964e-05, "loss": 1.0325, "step": 31405 }, { "epoch": 0.08, "learning_rate": 9.235089838441793e-05, "loss": 1.0294, "step": 31410 }, { "epoch": 0.08, "learning_rate": 9.234964014293624e-05, "loss": 1.0307, "step": 31415 }, { "epoch": 0.08, "learning_rate": 9.234838190145453e-05, "loss": 1.0319, "step": 31420 }, { "epoch": 0.08, "learning_rate": 9.234712365997282e-05, "loss": 1.0306, "step": 31425 }, { "epoch": 0.08, "learning_rate": 9.234586541849111e-05, "loss": 1.029, "step": 31430 }, { "epoch": 0.08, "learning_rate": 9.234460717700942e-05, "loss": 1.0337, "step": 31435 }, { "epoch": 0.08, "learning_rate": 9.234334893552771e-05, "loss": 1.0311, "step": 31440 }, { "epoch": 0.08, "learning_rate": 9.2342090694046e-05, "loss": 1.0502, "step": 31445 }, { "epoch": 0.08, "learning_rate": 9.234083245256429e-05, "loss": 1.0299, "step": 31450 }, { "epoch": 0.08, "learning_rate": 9.23395742110826e-05, "loss": 1.0299, "step": 31455 }, { "epoch": 0.08, "learning_rate": 9.233831596960089e-05, "loss": 1.0495, "step": 31460 }, { "epoch": 0.08, "learning_rate": 9.233705772811918e-05, "loss": 1.0301, "step": 31465 }, { "epoch": 0.08, "learning_rate": 9.233579948663747e-05, "loss": 1.0279, "step": 31470 }, { "epoch": 0.08, "learning_rate": 9.233454124515576e-05, "loss": 1.0305, "step": 31475 }, { "epoch": 0.08, "learning_rate": 9.233328300367407e-05, "loss": 1.0302, "step": 31480 }, { "epoch": 0.08, "learning_rate": 9.233202476219236e-05, "loss": 1.0515, "step": 31485 }, { "epoch": 0.08, "learning_rate": 9.233076652071065e-05, "loss": 1.033, "step": 31490 }, { "epoch": 0.08, "learning_rate": 9.232950827922894e-05, "loss": 1.0304, "step": 31495 }, { "epoch": 0.08, "learning_rate": 9.232825003774725e-05, "loss": 1.0306, "step": 31500 }, { "epoch": 0.08, "learning_rate": 9.232699179626554e-05, "loss": 1.0323, "step": 31505 }, { "epoch": 0.08, "learning_rate": 9.232573355478383e-05, "loss": 1.0297, "step": 31510 }, { "epoch": 0.08, "learning_rate": 9.232447531330212e-05, "loss": 1.0301, "step": 31515 }, { "epoch": 0.08, "learning_rate": 9.232321707182043e-05, "loss": 1.0308, "step": 31520 }, { "epoch": 0.08, "learning_rate": 9.232195883033873e-05, "loss": 1.0314, "step": 31525 }, { "epoch": 0.08, "learning_rate": 9.232070058885703e-05, "loss": 1.0305, "step": 31530 }, { "epoch": 0.08, "learning_rate": 9.231944234737532e-05, "loss": 1.0305, "step": 31535 }, { "epoch": 0.08, "learning_rate": 9.231818410589361e-05, "loss": 1.0311, "step": 31540 }, { "epoch": 0.08, "learning_rate": 9.231692586441191e-05, "loss": 1.0288, "step": 31545 }, { "epoch": 0.08, "learning_rate": 9.23156676229302e-05, "loss": 1.0309, "step": 31550 }, { "epoch": 0.08, "learning_rate": 9.23144093814485e-05, "loss": 1.0316, "step": 31555 }, { "epoch": 0.08, "learning_rate": 9.231315113996679e-05, "loss": 1.0308, "step": 31560 }, { "epoch": 0.08, "learning_rate": 9.231189289848508e-05, "loss": 1.0315, "step": 31565 }, { "epoch": 0.08, "learning_rate": 9.231063465700339e-05, "loss": 1.0319, "step": 31570 }, { "epoch": 0.08, "learning_rate": 9.230937641552168e-05, "loss": 1.0311, "step": 31575 }, { "epoch": 0.08, "learning_rate": 9.230811817403997e-05, "loss": 1.0465, "step": 31580 }, { "epoch": 0.08, "learning_rate": 9.230685993255826e-05, "loss": 1.0306, "step": 31585 }, { "epoch": 0.08, "learning_rate": 9.230560169107657e-05, "loss": 1.0288, "step": 31590 }, { "epoch": 0.08, "learning_rate": 9.230434344959486e-05, "loss": 1.0317, "step": 31595 }, { "epoch": 0.08, "learning_rate": 9.230308520811315e-05, "loss": 1.0312, "step": 31600 }, { "epoch": 0.08, "learning_rate": 9.230182696663144e-05, "loss": 1.0305, "step": 31605 }, { "epoch": 0.08, "learning_rate": 9.230056872514973e-05, "loss": 1.0327, "step": 31610 }, { "epoch": 0.08, "learning_rate": 9.229931048366804e-05, "loss": 1.0304, "step": 31615 }, { "epoch": 0.08, "learning_rate": 9.229805224218633e-05, "loss": 1.03, "step": 31620 }, { "epoch": 0.08, "learning_rate": 9.229679400070462e-05, "loss": 1.0305, "step": 31625 }, { "epoch": 0.08, "learning_rate": 9.229553575922291e-05, "loss": 1.0312, "step": 31630 }, { "epoch": 0.08, "learning_rate": 9.229427751774122e-05, "loss": 1.0311, "step": 31635 }, { "epoch": 0.08, "learning_rate": 9.229301927625951e-05, "loss": 1.0329, "step": 31640 }, { "epoch": 0.08, "learning_rate": 9.22917610347778e-05, "loss": 1.0308, "step": 31645 }, { "epoch": 0.08, "learning_rate": 9.229050279329609e-05, "loss": 1.0342, "step": 31650 }, { "epoch": 0.08, "learning_rate": 9.22892445518144e-05, "loss": 1.032, "step": 31655 }, { "epoch": 0.08, "learning_rate": 9.228798631033269e-05, "loss": 1.0291, "step": 31660 }, { "epoch": 0.08, "learning_rate": 9.228672806885098e-05, "loss": 1.0308, "step": 31665 }, { "epoch": 0.08, "learning_rate": 9.228546982736927e-05, "loss": 1.031, "step": 31670 }, { "epoch": 0.08, "learning_rate": 9.228421158588756e-05, "loss": 1.0305, "step": 31675 }, { "epoch": 0.08, "learning_rate": 9.228295334440587e-05, "loss": 1.0311, "step": 31680 }, { "epoch": 0.08, "learning_rate": 9.228169510292416e-05, "loss": 1.0334, "step": 31685 }, { "epoch": 0.08, "learning_rate": 9.228043686144245e-05, "loss": 1.0302, "step": 31690 }, { "epoch": 0.08, "learning_rate": 9.227917861996074e-05, "loss": 1.0298, "step": 31695 }, { "epoch": 0.08, "learning_rate": 9.227792037847905e-05, "loss": 1.0318, "step": 31700 }, { "epoch": 0.08, "learning_rate": 9.227666213699734e-05, "loss": 1.0301, "step": 31705 }, { "epoch": 0.08, "learning_rate": 9.227540389551563e-05, "loss": 1.032, "step": 31710 }, { "epoch": 0.08, "learning_rate": 9.227414565403392e-05, "loss": 1.0332, "step": 31715 }, { "epoch": 0.08, "learning_rate": 9.227288741255223e-05, "loss": 1.0313, "step": 31720 }, { "epoch": 0.08, "learning_rate": 9.227162917107052e-05, "loss": 1.0326, "step": 31725 }, { "epoch": 0.08, "learning_rate": 9.227037092958881e-05, "loss": 1.0318, "step": 31730 }, { "epoch": 0.08, "learning_rate": 9.22691126881071e-05, "loss": 1.0332, "step": 31735 }, { "epoch": 0.08, "learning_rate": 9.22678544466254e-05, "loss": 1.0314, "step": 31740 }, { "epoch": 0.08, "learning_rate": 9.22665962051437e-05, "loss": 1.0537, "step": 31745 }, { "epoch": 0.08, "learning_rate": 9.226533796366199e-05, "loss": 1.0305, "step": 31750 }, { "epoch": 0.08, "learning_rate": 9.226407972218028e-05, "loss": 1.032, "step": 31755 }, { "epoch": 0.08, "learning_rate": 9.226282148069857e-05, "loss": 1.0305, "step": 31760 }, { "epoch": 0.08, "learning_rate": 9.226156323921688e-05, "loss": 1.0295, "step": 31765 }, { "epoch": 0.08, "learning_rate": 9.226030499773517e-05, "loss": 1.0318, "step": 31770 }, { "epoch": 0.08, "learning_rate": 9.225904675625346e-05, "loss": 1.0305, "step": 31775 }, { "epoch": 0.08, "learning_rate": 9.225778851477175e-05, "loss": 1.0307, "step": 31780 }, { "epoch": 0.08, "learning_rate": 9.225653027329006e-05, "loss": 1.031, "step": 31785 }, { "epoch": 0.08, "learning_rate": 9.225527203180835e-05, "loss": 1.0317, "step": 31790 }, { "epoch": 0.08, "learning_rate": 9.225401379032664e-05, "loss": 1.0291, "step": 31795 }, { "epoch": 0.08, "learning_rate": 9.225275554884493e-05, "loss": 1.0332, "step": 31800 }, { "epoch": 0.08, "learning_rate": 9.225149730736323e-05, "loss": 1.032, "step": 31805 }, { "epoch": 0.08, "learning_rate": 9.225023906588153e-05, "loss": 1.0325, "step": 31810 }, { "epoch": 0.08, "learning_rate": 9.224898082439982e-05, "loss": 1.0302, "step": 31815 }, { "epoch": 0.08, "learning_rate": 9.224772258291811e-05, "loss": 1.0331, "step": 31820 }, { "epoch": 0.08, "learning_rate": 9.22464643414364e-05, "loss": 1.0314, "step": 31825 }, { "epoch": 0.08, "learning_rate": 9.224520609995471e-05, "loss": 1.032, "step": 31830 }, { "epoch": 0.08, "learning_rate": 9.2243947858473e-05, "loss": 1.029, "step": 31835 }, { "epoch": 0.08, "learning_rate": 9.22426896169913e-05, "loss": 1.0324, "step": 31840 }, { "epoch": 0.08, "learning_rate": 9.224143137550959e-05, "loss": 1.0283, "step": 31845 }, { "epoch": 0.08, "learning_rate": 9.224017313402789e-05, "loss": 1.0293, "step": 31850 }, { "epoch": 0.08, "learning_rate": 9.223891489254618e-05, "loss": 1.0541, "step": 31855 }, { "epoch": 0.08, "learning_rate": 9.223765665106447e-05, "loss": 1.031, "step": 31860 }, { "epoch": 0.08, "learning_rate": 9.223639840958277e-05, "loss": 1.0524, "step": 31865 }, { "epoch": 0.08, "learning_rate": 9.223514016810106e-05, "loss": 1.0305, "step": 31870 }, { "epoch": 0.08, "learning_rate": 9.223388192661936e-05, "loss": 1.0319, "step": 31875 }, { "epoch": 0.08, "learning_rate": 9.223262368513765e-05, "loss": 1.0322, "step": 31880 }, { "epoch": 0.08, "learning_rate": 9.223136544365595e-05, "loss": 1.0297, "step": 31885 }, { "epoch": 0.08, "learning_rate": 9.223010720217424e-05, "loss": 1.0328, "step": 31890 }, { "epoch": 0.08, "learning_rate": 9.222884896069254e-05, "loss": 1.0295, "step": 31895 }, { "epoch": 0.08, "learning_rate": 9.222759071921083e-05, "loss": 1.0306, "step": 31900 }, { "epoch": 0.08, "learning_rate": 9.222633247772913e-05, "loss": 1.0309, "step": 31905 }, { "epoch": 0.08, "learning_rate": 9.222507423624742e-05, "loss": 1.0322, "step": 31910 }, { "epoch": 0.08, "learning_rate": 9.222381599476572e-05, "loss": 1.0296, "step": 31915 }, { "epoch": 0.08, "learning_rate": 9.222255775328401e-05, "loss": 1.0284, "step": 31920 }, { "epoch": 0.08, "learning_rate": 9.22212995118023e-05, "loss": 1.0306, "step": 31925 }, { "epoch": 0.08, "learning_rate": 9.22200412703206e-05, "loss": 1.0316, "step": 31930 }, { "epoch": 0.08, "learning_rate": 9.221878302883889e-05, "loss": 1.0313, "step": 31935 }, { "epoch": 0.08, "learning_rate": 9.22175247873572e-05, "loss": 1.0317, "step": 31940 }, { "epoch": 0.08, "learning_rate": 9.221626654587549e-05, "loss": 1.0306, "step": 31945 }, { "epoch": 0.08, "learning_rate": 9.221500830439378e-05, "loss": 1.0308, "step": 31950 }, { "epoch": 0.08, "learning_rate": 9.221375006291207e-05, "loss": 1.0309, "step": 31955 }, { "epoch": 0.08, "learning_rate": 9.221249182143037e-05, "loss": 1.0306, "step": 31960 }, { "epoch": 0.08, "learning_rate": 9.221123357994867e-05, "loss": 1.0314, "step": 31965 }, { "epoch": 0.08, "learning_rate": 9.220997533846696e-05, "loss": 1.0316, "step": 31970 }, { "epoch": 0.08, "learning_rate": 9.220871709698525e-05, "loss": 1.0315, "step": 31975 }, { "epoch": 0.08, "learning_rate": 9.220745885550355e-05, "loss": 1.0314, "step": 31980 }, { "epoch": 0.08, "learning_rate": 9.220620061402185e-05, "loss": 1.0324, "step": 31985 }, { "epoch": 0.08, "learning_rate": 9.220494237254014e-05, "loss": 1.0285, "step": 31990 }, { "epoch": 0.08, "learning_rate": 9.220368413105843e-05, "loss": 1.0312, "step": 31995 }, { "epoch": 0.08, "learning_rate": 9.220242588957672e-05, "loss": 1.03, "step": 32000 }, { "epoch": 0.08, "learning_rate": 9.220116764809503e-05, "loss": 1.0315, "step": 32005 }, { "epoch": 0.08, "learning_rate": 9.219990940661332e-05, "loss": 1.0316, "step": 32010 }, { "epoch": 0.08, "learning_rate": 9.219865116513161e-05, "loss": 1.0313, "step": 32015 }, { "epoch": 0.08, "learning_rate": 9.219739292364991e-05, "loss": 1.0315, "step": 32020 }, { "epoch": 0.08, "learning_rate": 9.21961346821682e-05, "loss": 1.0318, "step": 32025 }, { "epoch": 0.08, "learning_rate": 9.219487644068651e-05, "loss": 1.0314, "step": 32030 }, { "epoch": 0.08, "learning_rate": 9.21936181992048e-05, "loss": 1.0332, "step": 32035 }, { "epoch": 0.08, "learning_rate": 9.21923599577231e-05, "loss": 1.0314, "step": 32040 }, { "epoch": 0.08, "learning_rate": 9.219110171624139e-05, "loss": 1.0315, "step": 32045 }, { "epoch": 0.08, "learning_rate": 9.218984347475969e-05, "loss": 1.0316, "step": 32050 }, { "epoch": 0.08, "learning_rate": 9.218858523327798e-05, "loss": 1.0301, "step": 32055 }, { "epoch": 0.08, "learning_rate": 9.218732699179627e-05, "loss": 1.0319, "step": 32060 }, { "epoch": 0.08, "learning_rate": 9.218606875031457e-05, "loss": 1.0301, "step": 32065 }, { "epoch": 0.08, "learning_rate": 9.218481050883287e-05, "loss": 1.033, "step": 32070 }, { "epoch": 0.08, "learning_rate": 9.218355226735116e-05, "loss": 1.0324, "step": 32075 }, { "epoch": 0.08, "learning_rate": 9.218229402586945e-05, "loss": 1.0293, "step": 32080 }, { "epoch": 0.08, "learning_rate": 9.218103578438775e-05, "loss": 1.0298, "step": 32085 }, { "epoch": 0.08, "learning_rate": 9.217977754290604e-05, "loss": 1.0303, "step": 32090 }, { "epoch": 0.08, "learning_rate": 9.217851930142434e-05, "loss": 1.031, "step": 32095 }, { "epoch": 0.08, "learning_rate": 9.217726105994263e-05, "loss": 1.0346, "step": 32100 }, { "epoch": 0.08, "learning_rate": 9.217600281846093e-05, "loss": 1.033, "step": 32105 }, { "epoch": 0.08, "learning_rate": 9.217474457697922e-05, "loss": 1.0318, "step": 32110 }, { "epoch": 0.08, "learning_rate": 9.217348633549752e-05, "loss": 1.0315, "step": 32115 }, { "epoch": 0.08, "learning_rate": 9.217222809401581e-05, "loss": 1.0299, "step": 32120 }, { "epoch": 0.08, "learning_rate": 9.21709698525341e-05, "loss": 1.0321, "step": 32125 }, { "epoch": 0.08, "learning_rate": 9.21697116110524e-05, "loss": 1.0317, "step": 32130 }, { "epoch": 0.08, "learning_rate": 9.21684533695707e-05, "loss": 1.0295, "step": 32135 }, { "epoch": 0.08, "learning_rate": 9.2167195128089e-05, "loss": 1.0325, "step": 32140 }, { "epoch": 0.08, "learning_rate": 9.216593688660729e-05, "loss": 1.0312, "step": 32145 }, { "epoch": 0.08, "learning_rate": 9.216467864512558e-05, "loss": 1.031, "step": 32150 }, { "epoch": 0.08, "learning_rate": 9.216342040364387e-05, "loss": 1.0301, "step": 32155 }, { "epoch": 0.08, "learning_rate": 9.216216216216217e-05, "loss": 1.0327, "step": 32160 }, { "epoch": 0.08, "learning_rate": 9.216090392068047e-05, "loss": 1.0308, "step": 32165 }, { "epoch": 0.08, "learning_rate": 9.215964567919876e-05, "loss": 1.0321, "step": 32170 }, { "epoch": 0.08, "learning_rate": 9.215838743771705e-05, "loss": 1.0325, "step": 32175 }, { "epoch": 0.08, "learning_rate": 9.215712919623535e-05, "loss": 1.0569, "step": 32180 }, { "epoch": 0.08, "learning_rate": 9.215587095475365e-05, "loss": 1.0321, "step": 32185 }, { "epoch": 0.08, "learning_rate": 9.215461271327194e-05, "loss": 1.0307, "step": 32190 }, { "epoch": 0.08, "learning_rate": 9.215335447179023e-05, "loss": 1.0308, "step": 32195 }, { "epoch": 0.08, "learning_rate": 9.215209623030853e-05, "loss": 1.0321, "step": 32200 }, { "epoch": 0.08, "learning_rate": 9.215083798882682e-05, "loss": 1.0302, "step": 32205 }, { "epoch": 0.08, "learning_rate": 9.214957974734512e-05, "loss": 1.0318, "step": 32210 }, { "epoch": 0.08, "learning_rate": 9.214832150586341e-05, "loss": 1.0308, "step": 32215 }, { "epoch": 0.08, "learning_rate": 9.21470632643817e-05, "loss": 1.032, "step": 32220 }, { "epoch": 0.08, "learning_rate": 9.21458050229e-05, "loss": 1.0289, "step": 32225 }, { "epoch": 0.08, "learning_rate": 9.21445467814183e-05, "loss": 1.0323, "step": 32230 }, { "epoch": 0.08, "learning_rate": 9.214328853993659e-05, "loss": 1.032, "step": 32235 }, { "epoch": 0.08, "learning_rate": 9.214203029845488e-05, "loss": 1.0319, "step": 32240 }, { "epoch": 0.08, "learning_rate": 9.214077205697318e-05, "loss": 1.0296, "step": 32245 }, { "epoch": 0.08, "learning_rate": 9.213951381549148e-05, "loss": 1.0314, "step": 32250 }, { "epoch": 0.08, "learning_rate": 9.213825557400977e-05, "loss": 1.0295, "step": 32255 }, { "epoch": 0.08, "learning_rate": 9.213699733252806e-05, "loss": 1.0321, "step": 32260 }, { "epoch": 0.08, "learning_rate": 9.213573909104636e-05, "loss": 1.0298, "step": 32265 }, { "epoch": 0.08, "learning_rate": 9.213448084956466e-05, "loss": 1.0316, "step": 32270 }, { "epoch": 0.08, "learning_rate": 9.213322260808295e-05, "loss": 1.0309, "step": 32275 }, { "epoch": 0.08, "learning_rate": 9.213196436660124e-05, "loss": 1.0305, "step": 32280 }, { "epoch": 0.08, "learning_rate": 9.213070612511953e-05, "loss": 1.031, "step": 32285 }, { "epoch": 0.08, "learning_rate": 9.212944788363784e-05, "loss": 1.0295, "step": 32290 }, { "epoch": 0.08, "learning_rate": 9.212818964215613e-05, "loss": 1.0302, "step": 32295 }, { "epoch": 0.08, "learning_rate": 9.212693140067442e-05, "loss": 1.0318, "step": 32300 }, { "epoch": 0.08, "learning_rate": 9.212567315919271e-05, "loss": 1.0336, "step": 32305 }, { "epoch": 0.08, "learning_rate": 9.212441491771102e-05, "loss": 1.0322, "step": 32310 }, { "epoch": 0.08, "learning_rate": 9.212315667622931e-05, "loss": 1.0329, "step": 32315 }, { "epoch": 0.08, "learning_rate": 9.21218984347476e-05, "loss": 1.0308, "step": 32320 }, { "epoch": 0.08, "learning_rate": 9.212064019326589e-05, "loss": 1.0313, "step": 32325 }, { "epoch": 0.08, "learning_rate": 9.211938195178418e-05, "loss": 1.0312, "step": 32330 }, { "epoch": 0.08, "learning_rate": 9.211812371030249e-05, "loss": 1.0324, "step": 32335 }, { "epoch": 0.08, "learning_rate": 9.211686546882078e-05, "loss": 1.0278, "step": 32340 }, { "epoch": 0.08, "learning_rate": 9.211560722733907e-05, "loss": 1.0286, "step": 32345 }, { "epoch": 0.08, "learning_rate": 9.211434898585736e-05, "loss": 1.031, "step": 32350 }, { "epoch": 0.08, "learning_rate": 9.211309074437567e-05, "loss": 1.033, "step": 32355 }, { "epoch": 0.08, "learning_rate": 9.211183250289396e-05, "loss": 1.0301, "step": 32360 }, { "epoch": 0.08, "learning_rate": 9.211057426141225e-05, "loss": 1.0298, "step": 32365 }, { "epoch": 0.08, "learning_rate": 9.210931601993054e-05, "loss": 1.0295, "step": 32370 }, { "epoch": 0.08, "learning_rate": 9.210805777844885e-05, "loss": 1.0305, "step": 32375 }, { "epoch": 0.08, "learning_rate": 9.210679953696714e-05, "loss": 1.0293, "step": 32380 }, { "epoch": 0.08, "learning_rate": 9.210554129548543e-05, "loss": 1.0294, "step": 32385 }, { "epoch": 0.08, "learning_rate": 9.210428305400372e-05, "loss": 1.0312, "step": 32390 }, { "epoch": 0.08, "learning_rate": 9.210302481252201e-05, "loss": 1.0329, "step": 32395 }, { "epoch": 0.08, "learning_rate": 9.210176657104032e-05, "loss": 1.0546, "step": 32400 }, { "epoch": 0.08, "learning_rate": 9.210050832955861e-05, "loss": 1.0304, "step": 32405 }, { "epoch": 0.08, "learning_rate": 9.20992500880769e-05, "loss": 1.0306, "step": 32410 }, { "epoch": 0.08, "learning_rate": 9.20979918465952e-05, "loss": 1.0312, "step": 32415 }, { "epoch": 0.08, "learning_rate": 9.20967336051135e-05, "loss": 1.0318, "step": 32420 }, { "epoch": 0.08, "learning_rate": 9.209547536363179e-05, "loss": 1.0304, "step": 32425 }, { "epoch": 0.08, "learning_rate": 9.209421712215008e-05, "loss": 1.0296, "step": 32430 }, { "epoch": 0.08, "learning_rate": 9.209295888066837e-05, "loss": 1.0306, "step": 32435 }, { "epoch": 0.08, "learning_rate": 9.209170063918668e-05, "loss": 1.0314, "step": 32440 }, { "epoch": 0.08, "learning_rate": 9.209044239770497e-05, "loss": 1.0319, "step": 32445 }, { "epoch": 0.08, "learning_rate": 9.208918415622326e-05, "loss": 1.0335, "step": 32450 }, { "epoch": 0.08, "learning_rate": 9.208792591474155e-05, "loss": 1.0331, "step": 32455 }, { "epoch": 0.08, "learning_rate": 9.208666767325985e-05, "loss": 1.0321, "step": 32460 }, { "epoch": 0.08, "learning_rate": 9.208540943177815e-05, "loss": 1.0308, "step": 32465 }, { "epoch": 0.08, "learning_rate": 9.208415119029644e-05, "loss": 1.0316, "step": 32470 }, { "epoch": 0.08, "learning_rate": 9.208289294881473e-05, "loss": 1.0323, "step": 32475 }, { "epoch": 0.08, "learning_rate": 9.208163470733303e-05, "loss": 1.0299, "step": 32480 }, { "epoch": 0.08, "learning_rate": 9.208037646585133e-05, "loss": 1.0314, "step": 32485 }, { "epoch": 0.08, "learning_rate": 9.207911822436962e-05, "loss": 1.029, "step": 32490 }, { "epoch": 0.08, "learning_rate": 9.207785998288791e-05, "loss": 1.0525, "step": 32495 }, { "epoch": 0.08, "learning_rate": 9.20766017414062e-05, "loss": 1.0334, "step": 32500 }, { "epoch": 0.08, "learning_rate": 9.207534349992451e-05, "loss": 1.0299, "step": 32505 }, { "epoch": 0.08, "learning_rate": 9.20740852584428e-05, "loss": 1.029, "step": 32510 }, { "epoch": 0.08, "learning_rate": 9.207282701696111e-05, "loss": 1.0309, "step": 32515 }, { "epoch": 0.08, "learning_rate": 9.20715687754794e-05, "loss": 1.0311, "step": 32520 }, { "epoch": 0.08, "learning_rate": 9.207031053399769e-05, "loss": 1.0326, "step": 32525 }, { "epoch": 0.08, "learning_rate": 9.2069052292516e-05, "loss": 1.0309, "step": 32530 }, { "epoch": 0.08, "learning_rate": 9.206779405103429e-05, "loss": 1.0295, "step": 32535 }, { "epoch": 0.08, "learning_rate": 9.206653580955258e-05, "loss": 1.031, "step": 32540 }, { "epoch": 0.08, "learning_rate": 9.206527756807087e-05, "loss": 1.0316, "step": 32545 }, { "epoch": 0.08, "learning_rate": 9.206401932658916e-05, "loss": 1.031, "step": 32550 }, { "epoch": 0.08, "learning_rate": 9.206276108510747e-05, "loss": 1.03, "step": 32555 }, { "epoch": 0.08, "learning_rate": 9.206150284362576e-05, "loss": 1.0301, "step": 32560 }, { "epoch": 0.08, "learning_rate": 9.206024460214405e-05, "loss": 1.0307, "step": 32565 }, { "epoch": 0.08, "learning_rate": 9.205898636066234e-05, "loss": 1.031, "step": 32570 }, { "epoch": 0.08, "learning_rate": 9.205772811918065e-05, "loss": 1.0285, "step": 32575 }, { "epoch": 0.08, "learning_rate": 9.205646987769894e-05, "loss": 1.0324, "step": 32580 }, { "epoch": 0.08, "learning_rate": 9.205521163621723e-05, "loss": 1.0302, "step": 32585 }, { "epoch": 0.08, "learning_rate": 9.205395339473552e-05, "loss": 1.0306, "step": 32590 }, { "epoch": 0.08, "learning_rate": 9.205269515325383e-05, "loss": 1.0306, "step": 32595 }, { "epoch": 0.08, "learning_rate": 9.205143691177212e-05, "loss": 1.0318, "step": 32600 }, { "epoch": 0.08, "learning_rate": 9.205017867029041e-05, "loss": 1.0313, "step": 32605 }, { "epoch": 0.08, "learning_rate": 9.20489204288087e-05, "loss": 1.0303, "step": 32610 }, { "epoch": 0.08, "learning_rate": 9.2047662187327e-05, "loss": 1.0304, "step": 32615 }, { "epoch": 0.08, "learning_rate": 9.20464039458453e-05, "loss": 1.0504, "step": 32620 }, { "epoch": 0.08, "learning_rate": 9.204514570436359e-05, "loss": 1.0309, "step": 32625 }, { "epoch": 0.08, "learning_rate": 9.204388746288188e-05, "loss": 1.0322, "step": 32630 }, { "epoch": 0.08, "learning_rate": 9.204262922140017e-05, "loss": 1.0299, "step": 32635 }, { "epoch": 0.08, "learning_rate": 9.204137097991848e-05, "loss": 1.033, "step": 32640 }, { "epoch": 0.08, "learning_rate": 9.204011273843677e-05, "loss": 1.0313, "step": 32645 }, { "epoch": 0.08, "learning_rate": 9.203885449695506e-05, "loss": 1.0304, "step": 32650 }, { "epoch": 0.08, "learning_rate": 9.203759625547335e-05, "loss": 1.0314, "step": 32655 }, { "epoch": 0.08, "learning_rate": 9.203633801399166e-05, "loss": 1.0314, "step": 32660 }, { "epoch": 0.08, "learning_rate": 9.203507977250995e-05, "loss": 1.0327, "step": 32665 }, { "epoch": 0.08, "learning_rate": 9.203382153102824e-05, "loss": 1.0296, "step": 32670 }, { "epoch": 0.08, "learning_rate": 9.203256328954653e-05, "loss": 1.0312, "step": 32675 }, { "epoch": 0.08, "learning_rate": 9.203130504806482e-05, "loss": 1.0294, "step": 32680 }, { "epoch": 0.08, "learning_rate": 9.203004680658313e-05, "loss": 1.0334, "step": 32685 }, { "epoch": 0.08, "learning_rate": 9.202878856510142e-05, "loss": 1.0324, "step": 32690 }, { "epoch": 0.08, "learning_rate": 9.202753032361971e-05, "loss": 1.0302, "step": 32695 }, { "epoch": 0.08, "learning_rate": 9.2026272082138e-05, "loss": 1.0302, "step": 32700 }, { "epoch": 0.08, "learning_rate": 9.202501384065631e-05, "loss": 1.0299, "step": 32705 }, { "epoch": 0.08, "learning_rate": 9.20237555991746e-05, "loss": 1.0319, "step": 32710 }, { "epoch": 0.08, "learning_rate": 9.202249735769289e-05, "loss": 1.031, "step": 32715 }, { "epoch": 0.08, "learning_rate": 9.202123911621118e-05, "loss": 1.0319, "step": 32720 }, { "epoch": 0.08, "learning_rate": 9.201998087472949e-05, "loss": 1.0316, "step": 32725 }, { "epoch": 0.08, "learning_rate": 9.201872263324778e-05, "loss": 1.0329, "step": 32730 }, { "epoch": 0.08, "learning_rate": 9.201746439176607e-05, "loss": 1.0285, "step": 32735 }, { "epoch": 0.08, "learning_rate": 9.201620615028436e-05, "loss": 1.0291, "step": 32740 }, { "epoch": 0.08, "learning_rate": 9.201494790880266e-05, "loss": 1.0328, "step": 32745 }, { "epoch": 0.08, "learning_rate": 9.201368966732096e-05, "loss": 1.0304, "step": 32750 }, { "epoch": 0.08, "learning_rate": 9.201243142583925e-05, "loss": 1.029, "step": 32755 }, { "epoch": 0.08, "learning_rate": 9.201117318435754e-05, "loss": 1.0313, "step": 32760 }, { "epoch": 0.08, "learning_rate": 9.200991494287584e-05, "loss": 1.0298, "step": 32765 }, { "epoch": 0.08, "learning_rate": 9.200865670139414e-05, "loss": 1.0281, "step": 32770 }, { "epoch": 0.08, "learning_rate": 9.200739845991243e-05, "loss": 1.0522, "step": 32775 }, { "epoch": 0.08, "learning_rate": 9.200614021843072e-05, "loss": 1.0312, "step": 32780 }, { "epoch": 0.08, "learning_rate": 9.200488197694902e-05, "loss": 1.0305, "step": 32785 }, { "epoch": 0.08, "learning_rate": 9.200362373546732e-05, "loss": 1.0301, "step": 32790 }, { "epoch": 0.08, "learning_rate": 9.200236549398561e-05, "loss": 1.032, "step": 32795 }, { "epoch": 0.08, "learning_rate": 9.20011072525039e-05, "loss": 1.031, "step": 32800 }, { "epoch": 0.08, "learning_rate": 9.19998490110222e-05, "loss": 1.0297, "step": 32805 }, { "epoch": 0.08, "learning_rate": 9.199859076954049e-05, "loss": 1.0315, "step": 32810 }, { "epoch": 0.08, "learning_rate": 9.199733252805879e-05, "loss": 1.0318, "step": 32815 }, { "epoch": 0.08, "learning_rate": 9.199607428657708e-05, "loss": 1.0297, "step": 32820 }, { "epoch": 0.08, "learning_rate": 9.199481604509538e-05, "loss": 1.0308, "step": 32825 }, { "epoch": 0.08, "learning_rate": 9.199355780361367e-05, "loss": 1.0294, "step": 32830 }, { "epoch": 0.08, "learning_rate": 9.199229956213197e-05, "loss": 1.0303, "step": 32835 }, { "epoch": 0.08, "learning_rate": 9.199104132065026e-05, "loss": 1.0303, "step": 32840 }, { "epoch": 0.08, "learning_rate": 9.198978307916856e-05, "loss": 1.0321, "step": 32845 }, { "epoch": 0.08, "learning_rate": 9.198852483768685e-05, "loss": 1.0303, "step": 32850 }, { "epoch": 0.08, "learning_rate": 9.198726659620515e-05, "loss": 1.0325, "step": 32855 }, { "epoch": 0.08, "learning_rate": 9.198600835472344e-05, "loss": 1.0304, "step": 32860 }, { "epoch": 0.08, "learning_rate": 9.198475011324174e-05, "loss": 1.0286, "step": 32865 }, { "epoch": 0.08, "learning_rate": 9.198349187176003e-05, "loss": 1.0547, "step": 32870 }, { "epoch": 0.08, "learning_rate": 9.198223363027832e-05, "loss": 1.0312, "step": 32875 }, { "epoch": 0.08, "learning_rate": 9.198097538879662e-05, "loss": 1.0302, "step": 32880 }, { "epoch": 0.08, "learning_rate": 9.197971714731492e-05, "loss": 1.0287, "step": 32885 }, { "epoch": 0.08, "learning_rate": 9.197845890583321e-05, "loss": 1.0296, "step": 32890 }, { "epoch": 0.08, "learning_rate": 9.19772006643515e-05, "loss": 1.0348, "step": 32895 }, { "epoch": 0.08, "learning_rate": 9.19759424228698e-05, "loss": 1.0309, "step": 32900 }, { "epoch": 0.08, "learning_rate": 9.19746841813881e-05, "loss": 1.0319, "step": 32905 }, { "epoch": 0.08, "learning_rate": 9.197342593990639e-05, "loss": 1.0292, "step": 32910 }, { "epoch": 0.08, "learning_rate": 9.197216769842468e-05, "loss": 1.0523, "step": 32915 }, { "epoch": 0.08, "learning_rate": 9.197090945694298e-05, "loss": 1.0296, "step": 32920 }, { "epoch": 0.08, "learning_rate": 9.196965121546128e-05, "loss": 1.031, "step": 32925 }, { "epoch": 0.08, "learning_rate": 9.196839297397957e-05, "loss": 1.0302, "step": 32930 }, { "epoch": 0.08, "learning_rate": 9.196713473249786e-05, "loss": 1.0336, "step": 32935 }, { "epoch": 0.08, "learning_rate": 9.196587649101615e-05, "loss": 1.0294, "step": 32940 }, { "epoch": 0.08, "learning_rate": 9.196461824953446e-05, "loss": 1.0318, "step": 32945 }, { "epoch": 0.08, "learning_rate": 9.196336000805275e-05, "loss": 1.0317, "step": 32950 }, { "epoch": 0.08, "learning_rate": 9.196210176657104e-05, "loss": 1.0298, "step": 32955 }, { "epoch": 0.08, "learning_rate": 9.196084352508933e-05, "loss": 1.0319, "step": 32960 }, { "epoch": 0.08, "learning_rate": 9.195958528360764e-05, "loss": 1.0491, "step": 32965 }, { "epoch": 0.08, "learning_rate": 9.195832704212593e-05, "loss": 1.0514, "step": 32970 }, { "epoch": 0.08, "learning_rate": 9.195706880064422e-05, "loss": 1.0301, "step": 32975 }, { "epoch": 0.08, "learning_rate": 9.195581055916251e-05, "loss": 1.0322, "step": 32980 }, { "epoch": 0.08, "learning_rate": 9.19545523176808e-05, "loss": 1.0522, "step": 32985 }, { "epoch": 0.08, "learning_rate": 9.195329407619911e-05, "loss": 1.0283, "step": 32990 }, { "epoch": 0.08, "learning_rate": 9.19520358347174e-05, "loss": 1.0305, "step": 32995 }, { "epoch": 0.08, "learning_rate": 9.195077759323569e-05, "loss": 1.0321, "step": 33000 }, { "epoch": 0.08, "learning_rate": 9.194951935175398e-05, "loss": 1.0294, "step": 33005 }, { "epoch": 0.08, "learning_rate": 9.194826111027229e-05, "loss": 1.0301, "step": 33010 }, { "epoch": 0.08, "learning_rate": 9.194700286879059e-05, "loss": 1.0508, "step": 33015 }, { "epoch": 0.08, "learning_rate": 9.194574462730888e-05, "loss": 1.0295, "step": 33020 }, { "epoch": 0.08, "learning_rate": 9.194448638582718e-05, "loss": 1.0309, "step": 33025 }, { "epoch": 0.08, "learning_rate": 9.194322814434547e-05, "loss": 1.0301, "step": 33030 }, { "epoch": 0.08, "learning_rate": 9.194196990286377e-05, "loss": 1.0313, "step": 33035 }, { "epoch": 0.08, "learning_rate": 9.194071166138206e-05, "loss": 1.0309, "step": 33040 }, { "epoch": 0.08, "learning_rate": 9.193945341990036e-05, "loss": 1.0309, "step": 33045 }, { "epoch": 0.08, "learning_rate": 9.193819517841865e-05, "loss": 1.0324, "step": 33050 }, { "epoch": 0.08, "learning_rate": 9.193693693693695e-05, "loss": 1.0311, "step": 33055 }, { "epoch": 0.08, "learning_rate": 9.193567869545524e-05, "loss": 1.0514, "step": 33060 }, { "epoch": 0.08, "learning_rate": 9.193442045397354e-05, "loss": 1.029, "step": 33065 }, { "epoch": 0.08, "learning_rate": 9.193316221249183e-05, "loss": 1.0294, "step": 33070 }, { "epoch": 0.08, "learning_rate": 9.193190397101012e-05, "loss": 1.0318, "step": 33075 }, { "epoch": 0.08, "learning_rate": 9.193064572952842e-05, "loss": 1.0282, "step": 33080 }, { "epoch": 0.08, "learning_rate": 9.192938748804672e-05, "loss": 1.0301, "step": 33085 }, { "epoch": 0.08, "learning_rate": 9.192812924656501e-05, "loss": 1.0556, "step": 33090 }, { "epoch": 0.08, "learning_rate": 9.19268710050833e-05, "loss": 1.032, "step": 33095 }, { "epoch": 0.08, "learning_rate": 9.19256127636016e-05, "loss": 1.029, "step": 33100 }, { "epoch": 0.08, "learning_rate": 9.19243545221199e-05, "loss": 1.028, "step": 33105 }, { "epoch": 0.08, "learning_rate": 9.192309628063819e-05, "loss": 1.0308, "step": 33110 }, { "epoch": 0.08, "learning_rate": 9.192183803915648e-05, "loss": 1.0286, "step": 33115 }, { "epoch": 0.08, "learning_rate": 9.192057979767478e-05, "loss": 1.0317, "step": 33120 }, { "epoch": 0.08, "learning_rate": 9.191932155619308e-05, "loss": 1.0309, "step": 33125 }, { "epoch": 0.08, "learning_rate": 9.191806331471137e-05, "loss": 1.0289, "step": 33130 }, { "epoch": 0.08, "learning_rate": 9.191680507322966e-05, "loss": 1.0514, "step": 33135 }, { "epoch": 0.08, "learning_rate": 9.191554683174795e-05, "loss": 1.0316, "step": 33140 }, { "epoch": 0.08, "learning_rate": 9.191428859026626e-05, "loss": 1.0309, "step": 33145 }, { "epoch": 0.08, "learning_rate": 9.191303034878455e-05, "loss": 1.0309, "step": 33150 }, { "epoch": 0.08, "learning_rate": 9.191177210730284e-05, "loss": 1.0324, "step": 33155 }, { "epoch": 0.08, "learning_rate": 9.191051386582113e-05, "loss": 1.0316, "step": 33160 }, { "epoch": 0.08, "learning_rate": 9.190925562433944e-05, "loss": 1.03, "step": 33165 }, { "epoch": 0.08, "learning_rate": 9.190799738285773e-05, "loss": 1.0294, "step": 33170 }, { "epoch": 0.08, "learning_rate": 9.190673914137602e-05, "loss": 1.03, "step": 33175 }, { "epoch": 0.08, "learning_rate": 9.190548089989431e-05, "loss": 1.0318, "step": 33180 }, { "epoch": 0.08, "learning_rate": 9.190422265841262e-05, "loss": 1.0312, "step": 33185 }, { "epoch": 0.08, "learning_rate": 9.19029644169309e-05, "loss": 1.0305, "step": 33190 }, { "epoch": 0.08, "learning_rate": 9.19017061754492e-05, "loss": 1.0302, "step": 33195 }, { "epoch": 0.08, "learning_rate": 9.190044793396749e-05, "loss": 1.0425, "step": 33200 }, { "epoch": 0.08, "learning_rate": 9.189918969248578e-05, "loss": 1.0307, "step": 33205 }, { "epoch": 0.08, "learning_rate": 9.189793145100409e-05, "loss": 1.0313, "step": 33210 }, { "epoch": 0.08, "learning_rate": 9.189667320952238e-05, "loss": 1.0298, "step": 33215 }, { "epoch": 0.08, "learning_rate": 9.189541496804067e-05, "loss": 1.0291, "step": 33220 }, { "epoch": 0.08, "learning_rate": 9.189415672655896e-05, "loss": 1.0288, "step": 33225 }, { "epoch": 0.08, "learning_rate": 9.189289848507727e-05, "loss": 1.03, "step": 33230 }, { "epoch": 0.08, "learning_rate": 9.189164024359556e-05, "loss": 1.0309, "step": 33235 }, { "epoch": 0.08, "learning_rate": 9.189038200211385e-05, "loss": 1.049, "step": 33240 }, { "epoch": 0.08, "learning_rate": 9.188912376063214e-05, "loss": 1.0325, "step": 33245 }, { "epoch": 0.08, "learning_rate": 9.188786551915045e-05, "loss": 1.0526, "step": 33250 }, { "epoch": 0.08, "learning_rate": 9.188660727766874e-05, "loss": 1.0318, "step": 33255 }, { "epoch": 0.08, "learning_rate": 9.188534903618703e-05, "loss": 1.0316, "step": 33260 }, { "epoch": 0.08, "learning_rate": 9.188409079470532e-05, "loss": 1.029, "step": 33265 }, { "epoch": 0.08, "learning_rate": 9.188283255322361e-05, "loss": 1.0317, "step": 33270 }, { "epoch": 0.08, "learning_rate": 9.188157431174192e-05, "loss": 1.0323, "step": 33275 }, { "epoch": 0.08, "learning_rate": 9.188031607026021e-05, "loss": 1.0323, "step": 33280 }, { "epoch": 0.08, "learning_rate": 9.18790578287785e-05, "loss": 1.0518, "step": 33285 }, { "epoch": 0.08, "learning_rate": 9.187779958729679e-05, "loss": 1.0297, "step": 33290 }, { "epoch": 0.08, "learning_rate": 9.18765413458151e-05, "loss": 1.0318, "step": 33295 }, { "epoch": 0.08, "learning_rate": 9.187528310433339e-05, "loss": 1.0316, "step": 33300 }, { "epoch": 0.08, "learning_rate": 9.187402486285168e-05, "loss": 1.0307, "step": 33305 }, { "epoch": 0.08, "learning_rate": 9.187276662136997e-05, "loss": 1.03, "step": 33310 }, { "epoch": 0.08, "learning_rate": 9.187150837988828e-05, "loss": 1.0301, "step": 33315 }, { "epoch": 0.08, "learning_rate": 9.187025013840657e-05, "loss": 1.0308, "step": 33320 }, { "epoch": 0.08, "learning_rate": 9.186899189692486e-05, "loss": 1.0317, "step": 33325 }, { "epoch": 0.08, "learning_rate": 9.186773365544315e-05, "loss": 1.0291, "step": 33330 }, { "epoch": 0.08, "learning_rate": 9.186647541396144e-05, "loss": 1.0321, "step": 33335 }, { "epoch": 0.08, "learning_rate": 9.186521717247975e-05, "loss": 1.0308, "step": 33340 }, { "epoch": 0.08, "learning_rate": 9.186395893099804e-05, "loss": 1.0322, "step": 33345 }, { "epoch": 0.08, "learning_rate": 9.186270068951633e-05, "loss": 1.0297, "step": 33350 }, { "epoch": 0.08, "learning_rate": 9.186144244803462e-05, "loss": 1.0308, "step": 33355 }, { "epoch": 0.08, "learning_rate": 9.186018420655293e-05, "loss": 1.0308, "step": 33360 }, { "epoch": 0.08, "learning_rate": 9.185892596507122e-05, "loss": 1.0318, "step": 33365 }, { "epoch": 0.08, "learning_rate": 9.185766772358951e-05, "loss": 1.0327, "step": 33370 }, { "epoch": 0.08, "learning_rate": 9.18564094821078e-05, "loss": 1.0314, "step": 33375 }, { "epoch": 0.08, "learning_rate": 9.185515124062611e-05, "loss": 1.0323, "step": 33380 }, { "epoch": 0.08, "learning_rate": 9.18538929991444e-05, "loss": 1.0299, "step": 33385 }, { "epoch": 0.08, "learning_rate": 9.185263475766269e-05, "loss": 1.0287, "step": 33390 }, { "epoch": 0.08, "learning_rate": 9.185137651618098e-05, "loss": 1.0295, "step": 33395 }, { "epoch": 0.08, "learning_rate": 9.185011827469928e-05, "loss": 1.0329, "step": 33400 }, { "epoch": 0.08, "learning_rate": 9.184886003321758e-05, "loss": 1.0326, "step": 33405 }, { "epoch": 0.08, "learning_rate": 9.184760179173587e-05, "loss": 1.0294, "step": 33410 }, { "epoch": 0.08, "learning_rate": 9.184634355025416e-05, "loss": 1.0295, "step": 33415 }, { "epoch": 0.08, "learning_rate": 9.184508530877246e-05, "loss": 1.0294, "step": 33420 }, { "epoch": 0.08, "learning_rate": 9.184382706729076e-05, "loss": 1.0307, "step": 33425 }, { "epoch": 0.08, "learning_rate": 9.184256882580905e-05, "loss": 1.031, "step": 33430 }, { "epoch": 0.08, "learning_rate": 9.184131058432734e-05, "loss": 1.0293, "step": 33435 }, { "epoch": 0.08, "learning_rate": 9.184005234284564e-05, "loss": 1.0506, "step": 33440 }, { "epoch": 0.08, "learning_rate": 9.183879410136394e-05, "loss": 1.0315, "step": 33445 }, { "epoch": 0.08, "learning_rate": 9.183753585988223e-05, "loss": 1.0286, "step": 33450 }, { "epoch": 0.08, "learning_rate": 9.183627761840052e-05, "loss": 1.0317, "step": 33455 }, { "epoch": 0.08, "learning_rate": 9.183501937691882e-05, "loss": 1.0303, "step": 33460 }, { "epoch": 0.08, "learning_rate": 9.183376113543711e-05, "loss": 1.0302, "step": 33465 }, { "epoch": 0.08, "learning_rate": 9.183250289395541e-05, "loss": 1.0333, "step": 33470 }, { "epoch": 0.08, "learning_rate": 9.18312446524737e-05, "loss": 1.0279, "step": 33475 }, { "epoch": 0.08, "learning_rate": 9.1829986410992e-05, "loss": 1.0309, "step": 33480 }, { "epoch": 0.08, "learning_rate": 9.182872816951029e-05, "loss": 1.0343, "step": 33485 }, { "epoch": 0.08, "learning_rate": 9.182746992802859e-05, "loss": 1.031, "step": 33490 }, { "epoch": 0.08, "learning_rate": 9.182621168654688e-05, "loss": 1.0319, "step": 33495 }, { "epoch": 0.08, "learning_rate": 9.182495344506518e-05, "loss": 1.0295, "step": 33500 }, { "epoch": 0.08, "learning_rate": 9.182369520358347e-05, "loss": 1.0311, "step": 33505 }, { "epoch": 0.08, "learning_rate": 9.182243696210177e-05, "loss": 1.0321, "step": 33510 }, { "epoch": 0.08, "learning_rate": 9.182117872062008e-05, "loss": 1.0298, "step": 33515 }, { "epoch": 0.08, "learning_rate": 9.181992047913837e-05, "loss": 1.0526, "step": 33520 }, { "epoch": 0.08, "learning_rate": 9.181866223765666e-05, "loss": 1.0287, "step": 33525 }, { "epoch": 0.08, "learning_rate": 9.181740399617495e-05, "loss": 1.0305, "step": 33530 }, { "epoch": 0.08, "learning_rate": 9.181614575469324e-05, "loss": 1.0286, "step": 33535 }, { "epoch": 0.08, "learning_rate": 9.181488751321155e-05, "loss": 1.0306, "step": 33540 }, { "epoch": 0.08, "learning_rate": 9.181362927172984e-05, "loss": 1.0316, "step": 33545 }, { "epoch": 0.08, "learning_rate": 9.181237103024813e-05, "loss": 1.0322, "step": 33550 }, { "epoch": 0.08, "learning_rate": 9.181111278876642e-05, "loss": 1.0312, "step": 33555 }, { "epoch": 0.08, "learning_rate": 9.180985454728473e-05, "loss": 1.0321, "step": 33560 }, { "epoch": 0.08, "learning_rate": 9.180859630580302e-05, "loss": 1.0313, "step": 33565 }, { "epoch": 0.08, "learning_rate": 9.180733806432131e-05, "loss": 1.0327, "step": 33570 }, { "epoch": 0.08, "learning_rate": 9.18060798228396e-05, "loss": 1.0321, "step": 33575 }, { "epoch": 0.08, "learning_rate": 9.180482158135791e-05, "loss": 1.0285, "step": 33580 }, { "epoch": 0.08, "learning_rate": 9.18035633398762e-05, "loss": 1.0302, "step": 33585 }, { "epoch": 0.08, "learning_rate": 9.180230509839449e-05, "loss": 1.0314, "step": 33590 }, { "epoch": 0.08, "learning_rate": 9.180104685691278e-05, "loss": 1.03, "step": 33595 }, { "epoch": 0.08, "learning_rate": 9.179978861543108e-05, "loss": 1.032, "step": 33600 }, { "epoch": 0.08, "learning_rate": 9.179853037394938e-05, "loss": 1.0292, "step": 33605 }, { "epoch": 0.08, "learning_rate": 9.179727213246767e-05, "loss": 1.0308, "step": 33610 }, { "epoch": 0.08, "learning_rate": 9.179601389098596e-05, "loss": 1.0299, "step": 33615 }, { "epoch": 0.08, "learning_rate": 9.179475564950426e-05, "loss": 1.0313, "step": 33620 }, { "epoch": 0.08, "learning_rate": 9.179349740802256e-05, "loss": 1.0302, "step": 33625 }, { "epoch": 0.08, "learning_rate": 9.179223916654085e-05, "loss": 1.0303, "step": 33630 }, { "epoch": 0.08, "learning_rate": 9.179098092505914e-05, "loss": 1.0307, "step": 33635 }, { "epoch": 0.08, "learning_rate": 9.178972268357744e-05, "loss": 1.0314, "step": 33640 }, { "epoch": 0.08, "learning_rate": 9.178846444209574e-05, "loss": 1.0314, "step": 33645 }, { "epoch": 0.08, "learning_rate": 9.178720620061403e-05, "loss": 1.0306, "step": 33650 }, { "epoch": 0.08, "learning_rate": 9.178594795913232e-05, "loss": 1.0304, "step": 33655 }, { "epoch": 0.08, "learning_rate": 9.178468971765061e-05, "loss": 1.0317, "step": 33660 }, { "epoch": 0.08, "learning_rate": 9.17834314761689e-05, "loss": 1.0304, "step": 33665 }, { "epoch": 0.08, "learning_rate": 9.178217323468721e-05, "loss": 1.0312, "step": 33670 }, { "epoch": 0.08, "learning_rate": 9.17809149932055e-05, "loss": 1.0293, "step": 33675 }, { "epoch": 0.08, "learning_rate": 9.17796567517238e-05, "loss": 1.0318, "step": 33680 }, { "epoch": 0.08, "learning_rate": 9.177839851024209e-05, "loss": 1.0299, "step": 33685 }, { "epoch": 0.08, "learning_rate": 9.177714026876039e-05, "loss": 1.0319, "step": 33690 }, { "epoch": 0.08, "learning_rate": 9.177588202727868e-05, "loss": 1.0296, "step": 33695 }, { "epoch": 0.08, "learning_rate": 9.177462378579697e-05, "loss": 1.0325, "step": 33700 }, { "epoch": 0.08, "learning_rate": 9.177336554431527e-05, "loss": 1.0308, "step": 33705 }, { "epoch": 0.08, "learning_rate": 9.177210730283357e-05, "loss": 1.0303, "step": 33710 }, { "epoch": 0.08, "learning_rate": 9.177084906135186e-05, "loss": 1.0296, "step": 33715 }, { "epoch": 0.08, "learning_rate": 9.176959081987015e-05, "loss": 1.034, "step": 33720 }, { "epoch": 0.08, "learning_rate": 9.176833257838845e-05, "loss": 1.0292, "step": 33725 }, { "epoch": 0.08, "learning_rate": 9.176707433690674e-05, "loss": 1.0311, "step": 33730 }, { "epoch": 0.08, "learning_rate": 9.176581609542504e-05, "loss": 1.0304, "step": 33735 }, { "epoch": 0.08, "learning_rate": 9.176455785394333e-05, "loss": 1.0409, "step": 33740 }, { "epoch": 0.08, "learning_rate": 9.176329961246163e-05, "loss": 1.0293, "step": 33745 }, { "epoch": 0.08, "learning_rate": 9.176204137097992e-05, "loss": 1.0312, "step": 33750 }, { "epoch": 0.08, "learning_rate": 9.176078312949822e-05, "loss": 1.0328, "step": 33755 }, { "epoch": 0.08, "learning_rate": 9.175952488801651e-05, "loss": 1.0313, "step": 33760 }, { "epoch": 0.08, "learning_rate": 9.17582666465348e-05, "loss": 1.031, "step": 33765 }, { "epoch": 0.08, "learning_rate": 9.17570084050531e-05, "loss": 1.0282, "step": 33770 }, { "epoch": 0.08, "learning_rate": 9.17557501635714e-05, "loss": 1.0308, "step": 33775 }, { "epoch": 0.08, "learning_rate": 9.17544919220897e-05, "loss": 1.0301, "step": 33780 }, { "epoch": 0.08, "learning_rate": 9.175323368060799e-05, "loss": 1.0307, "step": 33785 }, { "epoch": 0.08, "learning_rate": 9.175197543912628e-05, "loss": 1.0285, "step": 33790 }, { "epoch": 0.08, "learning_rate": 9.175071719764457e-05, "loss": 1.03, "step": 33795 }, { "epoch": 0.08, "learning_rate": 9.174945895616287e-05, "loss": 1.0312, "step": 33800 }, { "epoch": 0.08, "learning_rate": 9.174820071468117e-05, "loss": 1.0295, "step": 33805 }, { "epoch": 0.08, "learning_rate": 9.174694247319946e-05, "loss": 1.0311, "step": 33810 }, { "epoch": 0.08, "learning_rate": 9.174568423171775e-05, "loss": 1.0311, "step": 33815 }, { "epoch": 0.08, "learning_rate": 9.174442599023605e-05, "loss": 1.0311, "step": 33820 }, { "epoch": 0.08, "learning_rate": 9.174316774875435e-05, "loss": 1.0492, "step": 33825 }, { "epoch": 0.08, "learning_rate": 9.174190950727264e-05, "loss": 1.0323, "step": 33830 }, { "epoch": 0.08, "learning_rate": 9.174065126579093e-05, "loss": 1.0308, "step": 33835 }, { "epoch": 0.08, "learning_rate": 9.173939302430923e-05, "loss": 1.0307, "step": 33840 }, { "epoch": 0.08, "learning_rate": 9.173813478282753e-05, "loss": 1.03, "step": 33845 }, { "epoch": 0.08, "learning_rate": 9.173687654134582e-05, "loss": 1.0309, "step": 33850 }, { "epoch": 0.08, "learning_rate": 9.173561829986411e-05, "loss": 1.029, "step": 33855 }, { "epoch": 0.08, "learning_rate": 9.17343600583824e-05, "loss": 1.0293, "step": 33860 }, { "epoch": 0.09, "learning_rate": 9.17331018169007e-05, "loss": 1.0307, "step": 33865 }, { "epoch": 0.09, "learning_rate": 9.1731843575419e-05, "loss": 1.0328, "step": 33870 }, { "epoch": 0.09, "learning_rate": 9.173058533393729e-05, "loss": 1.0309, "step": 33875 }, { "epoch": 0.09, "learning_rate": 9.172932709245558e-05, "loss": 1.0283, "step": 33880 }, { "epoch": 0.09, "learning_rate": 9.172806885097389e-05, "loss": 1.0288, "step": 33885 }, { "epoch": 0.09, "learning_rate": 9.172681060949218e-05, "loss": 1.03, "step": 33890 }, { "epoch": 0.09, "learning_rate": 9.172555236801047e-05, "loss": 1.0302, "step": 33895 }, { "epoch": 0.09, "learning_rate": 9.172429412652876e-05, "loss": 1.0311, "step": 33900 }, { "epoch": 0.09, "learning_rate": 9.172303588504707e-05, "loss": 1.0311, "step": 33905 }, { "epoch": 0.09, "learning_rate": 9.172177764356536e-05, "loss": 1.0297, "step": 33910 }, { "epoch": 0.09, "learning_rate": 9.172051940208365e-05, "loss": 1.0299, "step": 33915 }, { "epoch": 0.09, "learning_rate": 9.171926116060194e-05, "loss": 1.0303, "step": 33920 }, { "epoch": 0.09, "learning_rate": 9.171800291912023e-05, "loss": 1.0319, "step": 33925 }, { "epoch": 0.09, "learning_rate": 9.171674467763854e-05, "loss": 1.0324, "step": 33930 }, { "epoch": 0.09, "learning_rate": 9.171548643615683e-05, "loss": 1.0307, "step": 33935 }, { "epoch": 0.09, "learning_rate": 9.171422819467512e-05, "loss": 1.0299, "step": 33940 }, { "epoch": 0.09, "learning_rate": 9.171296995319341e-05, "loss": 1.0304, "step": 33945 }, { "epoch": 0.09, "learning_rate": 9.171171171171172e-05, "loss": 1.0332, "step": 33950 }, { "epoch": 0.09, "learning_rate": 9.171045347023001e-05, "loss": 1.0312, "step": 33955 }, { "epoch": 0.09, "learning_rate": 9.17091952287483e-05, "loss": 1.0303, "step": 33960 }, { "epoch": 0.09, "learning_rate": 9.170793698726659e-05, "loss": 1.0317, "step": 33965 }, { "epoch": 0.09, "learning_rate": 9.17066787457849e-05, "loss": 1.0319, "step": 33970 }, { "epoch": 0.09, "learning_rate": 9.170542050430319e-05, "loss": 1.0282, "step": 33975 }, { "epoch": 0.09, "learning_rate": 9.170416226282148e-05, "loss": 1.031, "step": 33980 }, { "epoch": 0.09, "learning_rate": 9.170290402133977e-05, "loss": 1.0297, "step": 33985 }, { "epoch": 0.09, "learning_rate": 9.170164577985806e-05, "loss": 1.0274, "step": 33990 }, { "epoch": 0.09, "learning_rate": 9.170038753837637e-05, "loss": 1.0314, "step": 33995 }, { "epoch": 0.09, "learning_rate": 9.169912929689466e-05, "loss": 1.0298, "step": 34000 }, { "epoch": 0.09, "learning_rate": 9.169787105541295e-05, "loss": 1.0277, "step": 34005 }, { "epoch": 0.09, "learning_rate": 9.169661281393124e-05, "loss": 1.0302, "step": 34010 }, { "epoch": 0.09, "learning_rate": 9.169535457244955e-05, "loss": 1.0313, "step": 34015 }, { "epoch": 0.09, "learning_rate": 9.169409633096785e-05, "loss": 1.0309, "step": 34020 }, { "epoch": 0.09, "learning_rate": 9.169283808948615e-05, "loss": 1.0323, "step": 34025 }, { "epoch": 0.09, "learning_rate": 9.169157984800444e-05, "loss": 1.0315, "step": 34030 }, { "epoch": 0.09, "learning_rate": 9.169032160652273e-05, "loss": 1.0296, "step": 34035 }, { "epoch": 0.09, "learning_rate": 9.168906336504103e-05, "loss": 1.0298, "step": 34040 }, { "epoch": 0.09, "learning_rate": 9.168780512355933e-05, "loss": 1.0308, "step": 34045 }, { "epoch": 0.09, "learning_rate": 9.168654688207762e-05, "loss": 1.0308, "step": 34050 }, { "epoch": 0.09, "learning_rate": 9.168528864059591e-05, "loss": 1.0316, "step": 34055 }, { "epoch": 0.09, "learning_rate": 9.16840303991142e-05, "loss": 1.0297, "step": 34060 }, { "epoch": 0.09, "learning_rate": 9.16827721576325e-05, "loss": 1.0327, "step": 34065 }, { "epoch": 0.09, "learning_rate": 9.16815139161508e-05, "loss": 1.0313, "step": 34070 }, { "epoch": 0.09, "learning_rate": 9.168025567466909e-05, "loss": 1.0283, "step": 34075 }, { "epoch": 0.09, "learning_rate": 9.167899743318738e-05, "loss": 1.0306, "step": 34080 }, { "epoch": 0.09, "learning_rate": 9.167773919170569e-05, "loss": 1.0294, "step": 34085 }, { "epoch": 0.09, "learning_rate": 9.167648095022398e-05, "loss": 1.0313, "step": 34090 }, { "epoch": 0.09, "learning_rate": 9.167522270874227e-05, "loss": 1.0298, "step": 34095 }, { "epoch": 0.09, "learning_rate": 9.167396446726056e-05, "loss": 1.0315, "step": 34100 }, { "epoch": 0.09, "learning_rate": 9.167270622577887e-05, "loss": 1.0322, "step": 34105 }, { "epoch": 0.09, "learning_rate": 9.167144798429716e-05, "loss": 1.029, "step": 34110 }, { "epoch": 0.09, "learning_rate": 9.167018974281545e-05, "loss": 1.0287, "step": 34115 }, { "epoch": 0.09, "learning_rate": 9.166893150133374e-05, "loss": 1.0289, "step": 34120 }, { "epoch": 0.09, "learning_rate": 9.166767325985203e-05, "loss": 1.0308, "step": 34125 }, { "epoch": 0.09, "learning_rate": 9.166641501837034e-05, "loss": 1.031, "step": 34130 }, { "epoch": 0.09, "learning_rate": 9.166515677688863e-05, "loss": 1.0326, "step": 34135 }, { "epoch": 0.09, "learning_rate": 9.166389853540692e-05, "loss": 1.0306, "step": 34140 }, { "epoch": 0.09, "learning_rate": 9.166264029392521e-05, "loss": 1.0377, "step": 34145 }, { "epoch": 0.09, "learning_rate": 9.166138205244352e-05, "loss": 1.0298, "step": 34150 }, { "epoch": 0.09, "learning_rate": 9.166012381096181e-05, "loss": 1.0299, "step": 34155 }, { "epoch": 0.09, "learning_rate": 9.16588655694801e-05, "loss": 1.0329, "step": 34160 }, { "epoch": 0.09, "learning_rate": 9.165760732799839e-05, "loss": 1.0327, "step": 34165 }, { "epoch": 0.09, "learning_rate": 9.16563490865167e-05, "loss": 1.0299, "step": 34170 }, { "epoch": 0.09, "learning_rate": 9.165509084503499e-05, "loss": 1.0302, "step": 34175 }, { "epoch": 0.09, "learning_rate": 9.165383260355328e-05, "loss": 1.0306, "step": 34180 }, { "epoch": 0.09, "learning_rate": 9.165257436207157e-05, "loss": 1.0318, "step": 34185 }, { "epoch": 0.09, "learning_rate": 9.165131612058986e-05, "loss": 1.0308, "step": 34190 }, { "epoch": 0.09, "learning_rate": 9.165005787910817e-05, "loss": 1.0307, "step": 34195 }, { "epoch": 0.09, "learning_rate": 9.164879963762646e-05, "loss": 1.0303, "step": 34200 }, { "epoch": 0.09, "learning_rate": 9.164754139614475e-05, "loss": 1.0289, "step": 34205 }, { "epoch": 0.09, "learning_rate": 9.164628315466304e-05, "loss": 1.028, "step": 34210 }, { "epoch": 0.09, "learning_rate": 9.164502491318135e-05, "loss": 1.0287, "step": 34215 }, { "epoch": 0.09, "learning_rate": 9.164376667169964e-05, "loss": 1.0486, "step": 34220 }, { "epoch": 0.09, "learning_rate": 9.164250843021793e-05, "loss": 1.0296, "step": 34225 }, { "epoch": 0.09, "learning_rate": 9.164125018873622e-05, "loss": 1.0325, "step": 34230 }, { "epoch": 0.09, "learning_rate": 9.163999194725453e-05, "loss": 1.0301, "step": 34235 }, { "epoch": 0.09, "learning_rate": 9.163873370577282e-05, "loss": 1.0298, "step": 34240 }, { "epoch": 0.09, "learning_rate": 9.163747546429111e-05, "loss": 1.0306, "step": 34245 }, { "epoch": 0.09, "learning_rate": 9.16362172228094e-05, "loss": 1.0323, "step": 34250 }, { "epoch": 0.09, "learning_rate": 9.16349589813277e-05, "loss": 1.0307, "step": 34255 }, { "epoch": 0.09, "learning_rate": 9.1633700739846e-05, "loss": 1.0297, "step": 34260 }, { "epoch": 0.09, "learning_rate": 9.163244249836429e-05, "loss": 1.0287, "step": 34265 }, { "epoch": 0.09, "learning_rate": 9.163118425688258e-05, "loss": 1.0306, "step": 34270 }, { "epoch": 0.09, "learning_rate": 9.162992601540087e-05, "loss": 1.0302, "step": 34275 }, { "epoch": 0.09, "learning_rate": 9.162866777391918e-05, "loss": 1.0298, "step": 34280 }, { "epoch": 0.09, "learning_rate": 9.162740953243747e-05, "loss": 1.0301, "step": 34285 }, { "epoch": 0.09, "learning_rate": 9.162615129095576e-05, "loss": 1.0293, "step": 34290 }, { "epoch": 0.09, "learning_rate": 9.162489304947405e-05, "loss": 1.0301, "step": 34295 }, { "epoch": 0.09, "learning_rate": 9.162363480799236e-05, "loss": 1.0314, "step": 34300 }, { "epoch": 0.09, "learning_rate": 9.162237656651065e-05, "loss": 1.0308, "step": 34305 }, { "epoch": 0.09, "learning_rate": 9.162111832502894e-05, "loss": 1.0317, "step": 34310 }, { "epoch": 0.09, "learning_rate": 9.161986008354723e-05, "loss": 1.0297, "step": 34315 }, { "epoch": 0.09, "learning_rate": 9.161860184206553e-05, "loss": 1.0291, "step": 34320 }, { "epoch": 0.09, "learning_rate": 9.161734360058383e-05, "loss": 1.0296, "step": 34325 }, { "epoch": 0.09, "learning_rate": 9.161608535910212e-05, "loss": 1.0448, "step": 34330 }, { "epoch": 0.09, "learning_rate": 9.161482711762041e-05, "loss": 1.0325, "step": 34335 }, { "epoch": 0.09, "learning_rate": 9.16135688761387e-05, "loss": 1.0317, "step": 34340 }, { "epoch": 0.09, "learning_rate": 9.161231063465701e-05, "loss": 1.0478, "step": 34345 }, { "epoch": 0.09, "learning_rate": 9.16110523931753e-05, "loss": 1.0293, "step": 34350 }, { "epoch": 0.09, "learning_rate": 9.16097941516936e-05, "loss": 1.0307, "step": 34355 }, { "epoch": 0.09, "learning_rate": 9.160853591021189e-05, "loss": 1.0296, "step": 34360 }, { "epoch": 0.09, "learning_rate": 9.160727766873019e-05, "loss": 1.031, "step": 34365 }, { "epoch": 0.09, "learning_rate": 9.160601942724848e-05, "loss": 1.0303, "step": 34370 }, { "epoch": 0.09, "learning_rate": 9.160476118576677e-05, "loss": 1.0299, "step": 34375 }, { "epoch": 0.09, "learning_rate": 9.160350294428507e-05, "loss": 1.0316, "step": 34380 }, { "epoch": 0.09, "learning_rate": 9.160224470280336e-05, "loss": 1.0282, "step": 34385 }, { "epoch": 0.09, "learning_rate": 9.160098646132166e-05, "loss": 1.0303, "step": 34390 }, { "epoch": 0.09, "learning_rate": 9.159972821983995e-05, "loss": 1.031, "step": 34395 }, { "epoch": 0.09, "learning_rate": 9.159846997835825e-05, "loss": 1.0306, "step": 34400 }, { "epoch": 0.09, "learning_rate": 9.159721173687654e-05, "loss": 1.0291, "step": 34405 }, { "epoch": 0.09, "learning_rate": 9.159595349539484e-05, "loss": 1.0281, "step": 34410 }, { "epoch": 0.09, "learning_rate": 9.159469525391313e-05, "loss": 1.0305, "step": 34415 }, { "epoch": 0.09, "learning_rate": 9.159343701243143e-05, "loss": 1.0292, "step": 34420 }, { "epoch": 0.09, "learning_rate": 9.159217877094972e-05, "loss": 1.0308, "step": 34425 }, { "epoch": 0.09, "learning_rate": 9.159092052946802e-05, "loss": 1.0309, "step": 34430 }, { "epoch": 0.09, "learning_rate": 9.158966228798631e-05, "loss": 1.0288, "step": 34435 }, { "epoch": 0.09, "learning_rate": 9.15884040465046e-05, "loss": 1.0309, "step": 34440 }, { "epoch": 0.09, "learning_rate": 9.15871458050229e-05, "loss": 1.0308, "step": 34445 }, { "epoch": 0.09, "learning_rate": 9.158588756354119e-05, "loss": 1.0325, "step": 34450 }, { "epoch": 0.09, "learning_rate": 9.15846293220595e-05, "loss": 1.0321, "step": 34455 }, { "epoch": 0.09, "learning_rate": 9.158337108057779e-05, "loss": 1.03, "step": 34460 }, { "epoch": 0.09, "learning_rate": 9.158211283909608e-05, "loss": 1.03, "step": 34465 }, { "epoch": 0.09, "learning_rate": 9.158085459761437e-05, "loss": 1.0316, "step": 34470 }, { "epoch": 0.09, "learning_rate": 9.157959635613267e-05, "loss": 1.0315, "step": 34475 }, { "epoch": 0.09, "learning_rate": 9.157833811465097e-05, "loss": 1.0303, "step": 34480 }, { "epoch": 0.09, "learning_rate": 9.157707987316926e-05, "loss": 1.0321, "step": 34485 }, { "epoch": 0.09, "learning_rate": 9.157582163168755e-05, "loss": 1.0297, "step": 34490 }, { "epoch": 0.09, "learning_rate": 9.157456339020585e-05, "loss": 1.0304, "step": 34495 }, { "epoch": 0.09, "learning_rate": 9.157330514872415e-05, "loss": 1.0299, "step": 34500 }, { "epoch": 0.09, "learning_rate": 9.157204690724244e-05, "loss": 1.0313, "step": 34505 }, { "epoch": 0.09, "learning_rate": 9.157078866576073e-05, "loss": 1.0326, "step": 34510 }, { "epoch": 0.09, "learning_rate": 9.156953042427903e-05, "loss": 1.0291, "step": 34515 }, { "epoch": 0.09, "learning_rate": 9.156827218279734e-05, "loss": 1.0304, "step": 34520 }, { "epoch": 0.09, "learning_rate": 9.156701394131563e-05, "loss": 1.0554, "step": 34525 }, { "epoch": 0.09, "learning_rate": 9.156575569983392e-05, "loss": 1.0292, "step": 34530 }, { "epoch": 0.09, "learning_rate": 9.156449745835221e-05, "loss": 1.032, "step": 34535 }, { "epoch": 0.09, "learning_rate": 9.15632392168705e-05, "loss": 1.0303, "step": 34540 }, { "epoch": 0.09, "learning_rate": 9.156198097538881e-05, "loss": 1.0297, "step": 34545 }, { "epoch": 0.09, "learning_rate": 9.15607227339071e-05, "loss": 1.0278, "step": 34550 }, { "epoch": 0.09, "learning_rate": 9.15594644924254e-05, "loss": 1.0296, "step": 34555 }, { "epoch": 0.09, "learning_rate": 9.155820625094369e-05, "loss": 1.0324, "step": 34560 }, { "epoch": 0.09, "learning_rate": 9.155694800946199e-05, "loss": 1.0285, "step": 34565 }, { "epoch": 0.09, "learning_rate": 9.155568976798028e-05, "loss": 1.03, "step": 34570 }, { "epoch": 0.09, "learning_rate": 9.155443152649857e-05, "loss": 1.0313, "step": 34575 }, { "epoch": 0.09, "learning_rate": 9.155317328501687e-05, "loss": 1.0341, "step": 34580 }, { "epoch": 0.09, "learning_rate": 9.155191504353516e-05, "loss": 1.0303, "step": 34585 }, { "epoch": 0.09, "learning_rate": 9.155065680205346e-05, "loss": 1.0308, "step": 34590 }, { "epoch": 0.09, "learning_rate": 9.154939856057175e-05, "loss": 1.0313, "step": 34595 }, { "epoch": 0.09, "learning_rate": 9.154814031909005e-05, "loss": 1.0278, "step": 34600 }, { "epoch": 0.09, "learning_rate": 9.154688207760834e-05, "loss": 1.0303, "step": 34605 }, { "epoch": 0.09, "learning_rate": 9.154562383612664e-05, "loss": 1.0316, "step": 34610 }, { "epoch": 0.09, "learning_rate": 9.154436559464493e-05, "loss": 1.0312, "step": 34615 }, { "epoch": 0.09, "learning_rate": 9.154310735316323e-05, "loss": 1.0309, "step": 34620 }, { "epoch": 0.09, "learning_rate": 9.154184911168152e-05, "loss": 1.03, "step": 34625 }, { "epoch": 0.09, "learning_rate": 9.154059087019982e-05, "loss": 1.0278, "step": 34630 }, { "epoch": 0.09, "learning_rate": 9.153933262871811e-05, "loss": 1.0293, "step": 34635 }, { "epoch": 0.09, "learning_rate": 9.15380743872364e-05, "loss": 1.0327, "step": 34640 }, { "epoch": 0.09, "learning_rate": 9.15368161457547e-05, "loss": 1.0291, "step": 34645 }, { "epoch": 0.09, "learning_rate": 9.153555790427299e-05, "loss": 1.0301, "step": 34650 }, { "epoch": 0.09, "learning_rate": 9.15342996627913e-05, "loss": 1.0309, "step": 34655 }, { "epoch": 0.09, "learning_rate": 9.153304142130958e-05, "loss": 1.0304, "step": 34660 }, { "epoch": 0.09, "learning_rate": 9.153178317982788e-05, "loss": 1.0304, "step": 34665 }, { "epoch": 0.09, "learning_rate": 9.153052493834617e-05, "loss": 1.0296, "step": 34670 }, { "epoch": 0.09, "learning_rate": 9.152926669686447e-05, "loss": 1.0315, "step": 34675 }, { "epoch": 0.09, "learning_rate": 9.152800845538276e-05, "loss": 1.0299, "step": 34680 }, { "epoch": 0.09, "learning_rate": 9.152675021390106e-05, "loss": 1.0294, "step": 34685 }, { "epoch": 0.09, "learning_rate": 9.152549197241935e-05, "loss": 1.0297, "step": 34690 }, { "epoch": 0.09, "learning_rate": 9.152423373093765e-05, "loss": 1.0296, "step": 34695 }, { "epoch": 0.09, "learning_rate": 9.152297548945594e-05, "loss": 1.0314, "step": 34700 }, { "epoch": 0.09, "learning_rate": 9.152171724797424e-05, "loss": 1.0296, "step": 34705 }, { "epoch": 0.09, "learning_rate": 9.152045900649253e-05, "loss": 1.0304, "step": 34710 }, { "epoch": 0.09, "learning_rate": 9.151920076501082e-05, "loss": 1.0289, "step": 34715 }, { "epoch": 0.09, "learning_rate": 9.151794252352912e-05, "loss": 1.0306, "step": 34720 }, { "epoch": 0.09, "learning_rate": 9.151668428204742e-05, "loss": 1.03, "step": 34725 }, { "epoch": 0.09, "learning_rate": 9.151542604056571e-05, "loss": 1.0321, "step": 34730 }, { "epoch": 0.09, "learning_rate": 9.1514167799084e-05, "loss": 1.0496, "step": 34735 }, { "epoch": 0.09, "learning_rate": 9.15129095576023e-05, "loss": 1.0305, "step": 34740 }, { "epoch": 0.09, "learning_rate": 9.15116513161206e-05, "loss": 1.0295, "step": 34745 }, { "epoch": 0.09, "learning_rate": 9.151039307463889e-05, "loss": 1.0314, "step": 34750 }, { "epoch": 0.09, "learning_rate": 9.150913483315718e-05, "loss": 1.0307, "step": 34755 }, { "epoch": 0.09, "learning_rate": 9.150787659167548e-05, "loss": 1.0295, "step": 34760 }, { "epoch": 0.09, "learning_rate": 9.150661835019378e-05, "loss": 1.0319, "step": 34765 }, { "epoch": 0.09, "learning_rate": 9.150536010871207e-05, "loss": 1.0299, "step": 34770 }, { "epoch": 0.09, "learning_rate": 9.150410186723036e-05, "loss": 1.0529, "step": 34775 }, { "epoch": 0.09, "learning_rate": 9.150284362574865e-05, "loss": 1.029, "step": 34780 }, { "epoch": 0.09, "learning_rate": 9.150158538426696e-05, "loss": 1.0303, "step": 34785 }, { "epoch": 0.09, "learning_rate": 9.150032714278525e-05, "loss": 1.0291, "step": 34790 }, { "epoch": 0.09, "learning_rate": 9.149906890130354e-05, "loss": 1.0308, "step": 34795 }, { "epoch": 0.09, "learning_rate": 9.149781065982183e-05, "loss": 1.0314, "step": 34800 }, { "epoch": 0.09, "learning_rate": 9.149655241834014e-05, "loss": 1.0306, "step": 34805 }, { "epoch": 0.09, "learning_rate": 9.149529417685843e-05, "loss": 1.0289, "step": 34810 }, { "epoch": 0.09, "learning_rate": 9.149403593537672e-05, "loss": 1.0308, "step": 34815 }, { "epoch": 0.09, "learning_rate": 9.149277769389501e-05, "loss": 1.0294, "step": 34820 }, { "epoch": 0.09, "learning_rate": 9.149151945241332e-05, "loss": 1.0305, "step": 34825 }, { "epoch": 0.09, "learning_rate": 9.149026121093161e-05, "loss": 1.0308, "step": 34830 }, { "epoch": 0.09, "learning_rate": 9.14890029694499e-05, "loss": 1.0306, "step": 34835 }, { "epoch": 0.09, "learning_rate": 9.148774472796819e-05, "loss": 1.0287, "step": 34840 }, { "epoch": 0.09, "learning_rate": 9.148648648648648e-05, "loss": 1.0334, "step": 34845 }, { "epoch": 0.09, "learning_rate": 9.148522824500479e-05, "loss": 1.03, "step": 34850 }, { "epoch": 0.09, "learning_rate": 9.148397000352308e-05, "loss": 1.0311, "step": 34855 }, { "epoch": 0.09, "learning_rate": 9.148271176204137e-05, "loss": 1.0317, "step": 34860 }, { "epoch": 0.09, "learning_rate": 9.148145352055966e-05, "loss": 1.0318, "step": 34865 }, { "epoch": 0.09, "learning_rate": 9.148019527907797e-05, "loss": 1.0306, "step": 34870 }, { "epoch": 0.09, "learning_rate": 9.147893703759626e-05, "loss": 1.03, "step": 34875 }, { "epoch": 0.09, "learning_rate": 9.147767879611455e-05, "loss": 1.029, "step": 34880 }, { "epoch": 0.09, "learning_rate": 9.147642055463284e-05, "loss": 1.0315, "step": 34885 }, { "epoch": 0.09, "learning_rate": 9.147516231315115e-05, "loss": 1.033, "step": 34890 }, { "epoch": 0.09, "learning_rate": 9.147390407166944e-05, "loss": 1.0312, "step": 34895 }, { "epoch": 0.09, "learning_rate": 9.147264583018773e-05, "loss": 1.03, "step": 34900 }, { "epoch": 0.09, "learning_rate": 9.147138758870602e-05, "loss": 1.0297, "step": 34905 }, { "epoch": 0.09, "learning_rate": 9.147012934722431e-05, "loss": 1.0305, "step": 34910 }, { "epoch": 0.09, "learning_rate": 9.146887110574262e-05, "loss": 1.0313, "step": 34915 }, { "epoch": 0.09, "learning_rate": 9.146761286426091e-05, "loss": 1.0301, "step": 34920 }, { "epoch": 0.09, "learning_rate": 9.14663546227792e-05, "loss": 1.031, "step": 34925 }, { "epoch": 0.09, "learning_rate": 9.14650963812975e-05, "loss": 1.0321, "step": 34930 }, { "epoch": 0.09, "learning_rate": 9.14638381398158e-05, "loss": 1.0321, "step": 34935 }, { "epoch": 0.09, "learning_rate": 9.146257989833409e-05, "loss": 1.0294, "step": 34940 }, { "epoch": 0.09, "learning_rate": 9.146132165685238e-05, "loss": 1.0298, "step": 34945 }, { "epoch": 0.09, "learning_rate": 9.146006341537067e-05, "loss": 1.0323, "step": 34950 }, { "epoch": 0.09, "learning_rate": 9.145880517388898e-05, "loss": 1.0297, "step": 34955 }, { "epoch": 0.09, "learning_rate": 9.145754693240727e-05, "loss": 1.0295, "step": 34960 }, { "epoch": 0.09, "learning_rate": 9.145628869092556e-05, "loss": 1.0308, "step": 34965 }, { "epoch": 0.09, "learning_rate": 9.145503044944385e-05, "loss": 1.0287, "step": 34970 }, { "epoch": 0.09, "learning_rate": 9.145377220796215e-05, "loss": 1.028, "step": 34975 }, { "epoch": 0.09, "learning_rate": 9.145251396648045e-05, "loss": 1.0292, "step": 34980 }, { "epoch": 0.09, "learning_rate": 9.145125572499874e-05, "loss": 1.0306, "step": 34985 }, { "epoch": 0.09, "learning_rate": 9.144999748351703e-05, "loss": 1.0303, "step": 34990 }, { "epoch": 0.09, "learning_rate": 9.144873924203533e-05, "loss": 1.0287, "step": 34995 }, { "epoch": 0.09, "learning_rate": 9.144748100055363e-05, "loss": 1.0304, "step": 35000 }, { "epoch": 0.09, "learning_rate": 9.144622275907192e-05, "loss": 1.0306, "step": 35005 }, { "epoch": 0.09, "learning_rate": 9.144496451759021e-05, "loss": 1.0318, "step": 35010 }, { "epoch": 0.09, "learning_rate": 9.144370627610852e-05, "loss": 1.0302, "step": 35015 }, { "epoch": 0.09, "learning_rate": 9.144244803462681e-05, "loss": 1.0301, "step": 35020 }, { "epoch": 0.09, "learning_rate": 9.144118979314512e-05, "loss": 1.0303, "step": 35025 }, { "epoch": 0.09, "learning_rate": 9.143993155166341e-05, "loss": 1.0304, "step": 35030 }, { "epoch": 0.09, "learning_rate": 9.14386733101817e-05, "loss": 1.029, "step": 35035 }, { "epoch": 0.09, "learning_rate": 9.143741506869999e-05, "loss": 1.0315, "step": 35040 }, { "epoch": 0.09, "learning_rate": 9.14361568272183e-05, "loss": 1.0297, "step": 35045 }, { "epoch": 0.09, "learning_rate": 9.143489858573659e-05, "loss": 1.0297, "step": 35050 }, { "epoch": 0.09, "learning_rate": 9.143364034425488e-05, "loss": 1.0307, "step": 35055 }, { "epoch": 0.09, "learning_rate": 9.143238210277317e-05, "loss": 1.0306, "step": 35060 }, { "epoch": 0.09, "learning_rate": 9.143112386129146e-05, "loss": 1.0388, "step": 35065 }, { "epoch": 0.09, "learning_rate": 9.142986561980977e-05, "loss": 1.032, "step": 35070 }, { "epoch": 0.09, "learning_rate": 9.142860737832806e-05, "loss": 1.0287, "step": 35075 }, { "epoch": 0.09, "learning_rate": 9.142734913684635e-05, "loss": 1.0294, "step": 35080 }, { "epoch": 0.09, "learning_rate": 9.142609089536464e-05, "loss": 1.0298, "step": 35085 }, { "epoch": 0.09, "learning_rate": 9.142483265388295e-05, "loss": 1.028, "step": 35090 }, { "epoch": 0.09, "learning_rate": 9.142357441240124e-05, "loss": 1.0307, "step": 35095 }, { "epoch": 0.09, "learning_rate": 9.142231617091953e-05, "loss": 1.0311, "step": 35100 }, { "epoch": 0.09, "learning_rate": 9.142105792943782e-05, "loss": 1.0437, "step": 35105 }, { "epoch": 0.09, "learning_rate": 9.141979968795613e-05, "loss": 1.0296, "step": 35110 }, { "epoch": 0.09, "learning_rate": 9.141854144647442e-05, "loss": 1.0296, "step": 35115 }, { "epoch": 0.09, "learning_rate": 9.141728320499271e-05, "loss": 1.031, "step": 35120 }, { "epoch": 0.09, "learning_rate": 9.1416024963511e-05, "loss": 1.03, "step": 35125 }, { "epoch": 0.09, "learning_rate": 9.14147667220293e-05, "loss": 1.031, "step": 35130 }, { "epoch": 0.09, "learning_rate": 9.14135084805476e-05, "loss": 1.0317, "step": 35135 }, { "epoch": 0.09, "learning_rate": 9.141225023906589e-05, "loss": 1.0326, "step": 35140 }, { "epoch": 0.09, "learning_rate": 9.141099199758418e-05, "loss": 1.0314, "step": 35145 }, { "epoch": 0.09, "learning_rate": 9.140973375610247e-05, "loss": 1.0314, "step": 35150 }, { "epoch": 0.09, "learning_rate": 9.140847551462078e-05, "loss": 1.0309, "step": 35155 }, { "epoch": 0.09, "learning_rate": 9.140721727313907e-05, "loss": 1.031, "step": 35160 }, { "epoch": 0.09, "learning_rate": 9.140595903165736e-05, "loss": 1.0293, "step": 35165 }, { "epoch": 0.09, "learning_rate": 9.140470079017565e-05, "loss": 1.0306, "step": 35170 }, { "epoch": 0.09, "learning_rate": 9.140344254869396e-05, "loss": 1.0299, "step": 35175 }, { "epoch": 0.09, "learning_rate": 9.140218430721225e-05, "loss": 1.0306, "step": 35180 }, { "epoch": 0.09, "learning_rate": 9.140092606573054e-05, "loss": 1.0315, "step": 35185 }, { "epoch": 0.09, "learning_rate": 9.139966782424883e-05, "loss": 1.0292, "step": 35190 }, { "epoch": 0.09, "learning_rate": 9.139840958276712e-05, "loss": 1.0297, "step": 35195 }, { "epoch": 0.09, "learning_rate": 9.139715134128543e-05, "loss": 1.0299, "step": 35200 }, { "epoch": 0.09, "learning_rate": 9.139589309980372e-05, "loss": 1.0303, "step": 35205 }, { "epoch": 0.09, "learning_rate": 9.139463485832201e-05, "loss": 1.0306, "step": 35210 }, { "epoch": 0.09, "learning_rate": 9.13933766168403e-05, "loss": 1.0306, "step": 35215 }, { "epoch": 0.09, "learning_rate": 9.139211837535861e-05, "loss": 1.0293, "step": 35220 }, { "epoch": 0.09, "learning_rate": 9.13908601338769e-05, "loss": 1.0309, "step": 35225 }, { "epoch": 0.09, "learning_rate": 9.138960189239519e-05, "loss": 1.0301, "step": 35230 }, { "epoch": 0.09, "learning_rate": 9.138834365091348e-05, "loss": 1.0404, "step": 35235 }, { "epoch": 0.09, "learning_rate": 9.138708540943179e-05, "loss": 1.0301, "step": 35240 }, { "epoch": 0.09, "learning_rate": 9.138582716795008e-05, "loss": 1.0278, "step": 35245 }, { "epoch": 0.09, "learning_rate": 9.138456892646837e-05, "loss": 1.0309, "step": 35250 }, { "epoch": 0.09, "learning_rate": 9.138331068498666e-05, "loss": 1.0305, "step": 35255 }, { "epoch": 0.09, "learning_rate": 9.138205244350496e-05, "loss": 1.0439, "step": 35260 }, { "epoch": 0.09, "learning_rate": 9.138079420202326e-05, "loss": 1.0285, "step": 35265 }, { "epoch": 0.09, "learning_rate": 9.137953596054155e-05, "loss": 1.0296, "step": 35270 }, { "epoch": 0.09, "learning_rate": 9.137827771905984e-05, "loss": 1.0311, "step": 35275 }, { "epoch": 0.09, "learning_rate": 9.137701947757814e-05, "loss": 1.0293, "step": 35280 }, { "epoch": 0.09, "learning_rate": 9.137576123609644e-05, "loss": 1.0514, "step": 35285 }, { "epoch": 0.09, "learning_rate": 9.137450299461473e-05, "loss": 1.0273, "step": 35290 }, { "epoch": 0.09, "learning_rate": 9.137324475313302e-05, "loss": 1.0313, "step": 35295 }, { "epoch": 0.09, "learning_rate": 9.137198651165132e-05, "loss": 1.0331, "step": 35300 }, { "epoch": 0.09, "learning_rate": 9.137072827016961e-05, "loss": 1.0315, "step": 35305 }, { "epoch": 0.09, "learning_rate": 9.136947002868791e-05, "loss": 1.0295, "step": 35310 }, { "epoch": 0.09, "learning_rate": 9.13682117872062e-05, "loss": 1.0315, "step": 35315 }, { "epoch": 0.09, "learning_rate": 9.13669535457245e-05, "loss": 1.0298, "step": 35320 }, { "epoch": 0.09, "learning_rate": 9.136569530424279e-05, "loss": 1.0304, "step": 35325 }, { "epoch": 0.09, "learning_rate": 9.136443706276109e-05, "loss": 1.0301, "step": 35330 }, { "epoch": 0.09, "learning_rate": 9.136317882127938e-05, "loss": 1.0313, "step": 35335 }, { "epoch": 0.09, "learning_rate": 9.136192057979768e-05, "loss": 1.03, "step": 35340 }, { "epoch": 0.09, "learning_rate": 9.136066233831597e-05, "loss": 1.0292, "step": 35345 }, { "epoch": 0.09, "learning_rate": 9.135940409683427e-05, "loss": 1.0314, "step": 35350 }, { "epoch": 0.09, "learning_rate": 9.135814585535256e-05, "loss": 1.0305, "step": 35355 }, { "epoch": 0.09, "learning_rate": 9.135688761387086e-05, "loss": 1.0293, "step": 35360 }, { "epoch": 0.09, "learning_rate": 9.135562937238915e-05, "loss": 1.0309, "step": 35365 }, { "epoch": 0.09, "learning_rate": 9.135437113090744e-05, "loss": 1.0303, "step": 35370 }, { "epoch": 0.09, "learning_rate": 9.135311288942574e-05, "loss": 1.0292, "step": 35375 }, { "epoch": 0.09, "learning_rate": 9.135185464794404e-05, "loss": 1.0298, "step": 35380 }, { "epoch": 0.09, "learning_rate": 9.135059640646233e-05, "loss": 1.028, "step": 35385 }, { "epoch": 0.09, "learning_rate": 9.134933816498062e-05, "loss": 1.029, "step": 35390 }, { "epoch": 0.09, "learning_rate": 9.134807992349892e-05, "loss": 1.0291, "step": 35395 }, { "epoch": 0.09, "learning_rate": 9.134682168201722e-05, "loss": 1.0294, "step": 35400 }, { "epoch": 0.09, "learning_rate": 9.134556344053551e-05, "loss": 1.0299, "step": 35405 }, { "epoch": 0.09, "learning_rate": 9.13443051990538e-05, "loss": 1.0303, "step": 35410 }, { "epoch": 0.09, "learning_rate": 9.13430469575721e-05, "loss": 1.0296, "step": 35415 }, { "epoch": 0.09, "learning_rate": 9.13417887160904e-05, "loss": 1.0317, "step": 35420 }, { "epoch": 0.09, "learning_rate": 9.134053047460869e-05, "loss": 1.0301, "step": 35425 }, { "epoch": 0.09, "learning_rate": 9.133927223312698e-05, "loss": 1.0291, "step": 35430 }, { "epoch": 0.09, "learning_rate": 9.133801399164527e-05, "loss": 1.0301, "step": 35435 }, { "epoch": 0.09, "learning_rate": 9.133675575016358e-05, "loss": 1.0282, "step": 35440 }, { "epoch": 0.09, "learning_rate": 9.133549750868187e-05, "loss": 1.0277, "step": 35445 }, { "epoch": 0.09, "learning_rate": 9.133423926720016e-05, "loss": 1.0284, "step": 35450 }, { "epoch": 0.09, "learning_rate": 9.133298102571845e-05, "loss": 1.0331, "step": 35455 }, { "epoch": 0.09, "learning_rate": 9.133172278423676e-05, "loss": 1.0316, "step": 35460 }, { "epoch": 0.09, "learning_rate": 9.133046454275505e-05, "loss": 1.0314, "step": 35465 }, { "epoch": 0.09, "learning_rate": 9.132920630127334e-05, "loss": 1.032, "step": 35470 }, { "epoch": 0.09, "learning_rate": 9.132794805979163e-05, "loss": 1.0311, "step": 35475 }, { "epoch": 0.09, "learning_rate": 9.132668981830994e-05, "loss": 1.0283, "step": 35480 }, { "epoch": 0.09, "learning_rate": 9.132543157682823e-05, "loss": 1.0307, "step": 35485 }, { "epoch": 0.09, "learning_rate": 9.132417333534652e-05, "loss": 1.0289, "step": 35490 }, { "epoch": 0.09, "learning_rate": 9.132291509386481e-05, "loss": 1.0285, "step": 35495 }, { "epoch": 0.09, "learning_rate": 9.13216568523831e-05, "loss": 1.0282, "step": 35500 }, { "epoch": 0.09, "learning_rate": 9.132039861090141e-05, "loss": 1.0303, "step": 35505 }, { "epoch": 0.09, "learning_rate": 9.13191403694197e-05, "loss": 1.0312, "step": 35510 }, { "epoch": 0.09, "learning_rate": 9.1317882127938e-05, "loss": 1.0281, "step": 35515 }, { "epoch": 0.09, "learning_rate": 9.13166238864563e-05, "loss": 1.0302, "step": 35520 }, { "epoch": 0.09, "learning_rate": 9.131536564497459e-05, "loss": 1.031, "step": 35525 }, { "epoch": 0.09, "learning_rate": 9.131410740349289e-05, "loss": 1.0293, "step": 35530 }, { "epoch": 0.09, "learning_rate": 9.131284916201118e-05, "loss": 1.0308, "step": 35535 }, { "epoch": 0.09, "learning_rate": 9.131159092052948e-05, "loss": 1.0301, "step": 35540 }, { "epoch": 0.09, "learning_rate": 9.131033267904777e-05, "loss": 1.0837, "step": 35545 }, { "epoch": 0.09, "learning_rate": 9.130907443756607e-05, "loss": 1.0288, "step": 35550 }, { "epoch": 0.09, "learning_rate": 9.130781619608436e-05, "loss": 1.0308, "step": 35555 }, { "epoch": 0.09, "learning_rate": 9.130655795460266e-05, "loss": 1.031, "step": 35560 }, { "epoch": 0.09, "learning_rate": 9.130529971312095e-05, "loss": 1.0488, "step": 35565 }, { "epoch": 0.09, "learning_rate": 9.130404147163925e-05, "loss": 1.0312, "step": 35570 }, { "epoch": 0.09, "learning_rate": 9.130278323015754e-05, "loss": 1.0522, "step": 35575 }, { "epoch": 0.09, "learning_rate": 9.130152498867584e-05, "loss": 1.0316, "step": 35580 }, { "epoch": 0.09, "learning_rate": 9.130051839549046e-05, "loss": 1.0308, "step": 35585 }, { "epoch": 0.09, "learning_rate": 9.129926015400876e-05, "loss": 1.0305, "step": 35590 }, { "epoch": 0.09, "learning_rate": 9.129800191252705e-05, "loss": 1.0316, "step": 35595 }, { "epoch": 0.09, "learning_rate": 9.129674367104535e-05, "loss": 1.0286, "step": 35600 }, { "epoch": 0.09, "learning_rate": 9.129548542956364e-05, "loss": 1.0316, "step": 35605 }, { "epoch": 0.09, "learning_rate": 9.129422718808194e-05, "loss": 1.0543, "step": 35610 }, { "epoch": 0.09, "learning_rate": 9.129296894660023e-05, "loss": 1.0299, "step": 35615 }, { "epoch": 0.09, "learning_rate": 9.129171070511853e-05, "loss": 1.0303, "step": 35620 }, { "epoch": 0.09, "learning_rate": 9.129045246363682e-05, "loss": 1.0291, "step": 35625 }, { "epoch": 0.09, "learning_rate": 9.128919422215512e-05, "loss": 1.0301, "step": 35630 }, { "epoch": 0.09, "learning_rate": 9.128793598067341e-05, "loss": 1.0307, "step": 35635 }, { "epoch": 0.09, "learning_rate": 9.12866777391917e-05, "loss": 1.0307, "step": 35640 }, { "epoch": 0.09, "learning_rate": 9.128541949771e-05, "loss": 1.0286, "step": 35645 }, { "epoch": 0.09, "learning_rate": 9.128416125622829e-05, "loss": 1.0298, "step": 35650 }, { "epoch": 0.09, "learning_rate": 9.12829030147466e-05, "loss": 1.0311, "step": 35655 }, { "epoch": 0.09, "learning_rate": 9.128164477326489e-05, "loss": 1.0314, "step": 35660 }, { "epoch": 0.09, "learning_rate": 9.128038653178318e-05, "loss": 1.0287, "step": 35665 }, { "epoch": 0.09, "learning_rate": 9.127912829030147e-05, "loss": 1.0287, "step": 35670 }, { "epoch": 0.09, "learning_rate": 9.127787004881977e-05, "loss": 1.0312, "step": 35675 }, { "epoch": 0.09, "learning_rate": 9.127661180733807e-05, "loss": 1.0323, "step": 35680 }, { "epoch": 0.09, "learning_rate": 9.127535356585636e-05, "loss": 1.0294, "step": 35685 }, { "epoch": 0.09, "learning_rate": 9.127409532437465e-05, "loss": 1.0316, "step": 35690 }, { "epoch": 0.09, "learning_rate": 9.127283708289295e-05, "loss": 1.0463, "step": 35695 }, { "epoch": 0.09, "learning_rate": 9.127157884141125e-05, "loss": 1.0291, "step": 35700 }, { "epoch": 0.09, "learning_rate": 9.127032059992954e-05, "loss": 1.0293, "step": 35705 }, { "epoch": 0.09, "learning_rate": 9.126906235844783e-05, "loss": 1.0302, "step": 35710 }, { "epoch": 0.09, "learning_rate": 9.126780411696613e-05, "loss": 1.0286, "step": 35715 }, { "epoch": 0.09, "learning_rate": 9.126654587548444e-05, "loss": 1.0307, "step": 35720 }, { "epoch": 0.09, "learning_rate": 9.126528763400273e-05, "loss": 1.0283, "step": 35725 }, { "epoch": 0.09, "learning_rate": 9.126402939252102e-05, "loss": 1.0295, "step": 35730 }, { "epoch": 0.09, "learning_rate": 9.126277115103931e-05, "loss": 1.0303, "step": 35735 }, { "epoch": 0.09, "learning_rate": 9.12615129095576e-05, "loss": 1.0305, "step": 35740 }, { "epoch": 0.09, "learning_rate": 9.126025466807591e-05, "loss": 1.0294, "step": 35745 }, { "epoch": 0.09, "learning_rate": 9.12589964265942e-05, "loss": 1.0307, "step": 35750 }, { "epoch": 0.09, "learning_rate": 9.12577381851125e-05, "loss": 1.0302, "step": 35755 }, { "epoch": 0.09, "learning_rate": 9.125647994363079e-05, "loss": 1.0304, "step": 35760 }, { "epoch": 0.09, "learning_rate": 9.125522170214909e-05, "loss": 1.0304, "step": 35765 }, { "epoch": 0.09, "learning_rate": 9.125396346066738e-05, "loss": 1.0305, "step": 35770 }, { "epoch": 0.09, "learning_rate": 9.125270521918567e-05, "loss": 1.0294, "step": 35775 }, { "epoch": 0.09, "learning_rate": 9.125144697770397e-05, "loss": 1.0284, "step": 35780 }, { "epoch": 0.09, "learning_rate": 9.125018873622227e-05, "loss": 1.0291, "step": 35785 }, { "epoch": 0.09, "learning_rate": 9.124893049474056e-05, "loss": 1.0307, "step": 35790 }, { "epoch": 0.09, "learning_rate": 9.124767225325885e-05, "loss": 1.0297, "step": 35795 }, { "epoch": 0.09, "learning_rate": 9.124641401177715e-05, "loss": 1.0301, "step": 35800 }, { "epoch": 0.09, "learning_rate": 9.124515577029544e-05, "loss": 1.0286, "step": 35805 }, { "epoch": 0.09, "learning_rate": 9.124389752881374e-05, "loss": 1.0293, "step": 35810 }, { "epoch": 0.09, "learning_rate": 9.124263928733203e-05, "loss": 1.0296, "step": 35815 }, { "epoch": 0.09, "learning_rate": 9.124138104585033e-05, "loss": 1.0295, "step": 35820 }, { "epoch": 0.09, "learning_rate": 9.124012280436862e-05, "loss": 1.0329, "step": 35825 }, { "epoch": 0.09, "learning_rate": 9.123886456288692e-05, "loss": 1.0283, "step": 35830 }, { "epoch": 0.09, "learning_rate": 9.123760632140521e-05, "loss": 1.0301, "step": 35835 }, { "epoch": 0.09, "learning_rate": 9.12363480799235e-05, "loss": 1.0293, "step": 35840 }, { "epoch": 0.09, "learning_rate": 9.12350898384418e-05, "loss": 1.0287, "step": 35845 }, { "epoch": 0.09, "learning_rate": 9.12338315969601e-05, "loss": 1.0314, "step": 35850 }, { "epoch": 0.09, "learning_rate": 9.12325733554784e-05, "loss": 1.0288, "step": 35855 }, { "epoch": 0.09, "learning_rate": 9.123131511399669e-05, "loss": 1.0308, "step": 35860 }, { "epoch": 0.09, "learning_rate": 9.123005687251498e-05, "loss": 1.0288, "step": 35865 }, { "epoch": 0.09, "learning_rate": 9.122879863103327e-05, "loss": 1.0287, "step": 35870 }, { "epoch": 0.09, "learning_rate": 9.122754038955157e-05, "loss": 1.0323, "step": 35875 }, { "epoch": 0.09, "learning_rate": 9.122628214806987e-05, "loss": 1.0312, "step": 35880 }, { "epoch": 0.09, "learning_rate": 9.122502390658816e-05, "loss": 1.0303, "step": 35885 }, { "epoch": 0.09, "learning_rate": 9.122376566510645e-05, "loss": 1.0527, "step": 35890 }, { "epoch": 0.09, "learning_rate": 9.122250742362475e-05, "loss": 1.0302, "step": 35895 }, { "epoch": 0.09, "learning_rate": 9.122124918214305e-05, "loss": 1.0294, "step": 35900 }, { "epoch": 0.09, "learning_rate": 9.121999094066134e-05, "loss": 1.0283, "step": 35905 }, { "epoch": 0.09, "learning_rate": 9.121873269917963e-05, "loss": 1.0298, "step": 35910 }, { "epoch": 0.09, "learning_rate": 9.121747445769793e-05, "loss": 1.0315, "step": 35915 }, { "epoch": 0.09, "learning_rate": 9.121621621621623e-05, "loss": 1.0295, "step": 35920 }, { "epoch": 0.09, "learning_rate": 9.121495797473452e-05, "loss": 1.0315, "step": 35925 }, { "epoch": 0.09, "learning_rate": 9.121369973325281e-05, "loss": 1.0302, "step": 35930 }, { "epoch": 0.09, "learning_rate": 9.12124414917711e-05, "loss": 1.0302, "step": 35935 }, { "epoch": 0.09, "learning_rate": 9.12111832502894e-05, "loss": 1.0314, "step": 35940 }, { "epoch": 0.09, "learning_rate": 9.12099250088077e-05, "loss": 1.0294, "step": 35945 }, { "epoch": 0.09, "learning_rate": 9.120866676732599e-05, "loss": 1.0298, "step": 35950 }, { "epoch": 0.09, "learning_rate": 9.120740852584428e-05, "loss": 1.0299, "step": 35955 }, { "epoch": 0.09, "learning_rate": 9.120615028436258e-05, "loss": 1.0275, "step": 35960 }, { "epoch": 0.09, "learning_rate": 9.120489204288088e-05, "loss": 1.0316, "step": 35965 }, { "epoch": 0.09, "learning_rate": 9.120363380139917e-05, "loss": 1.0302, "step": 35970 }, { "epoch": 0.09, "learning_rate": 9.120237555991746e-05, "loss": 1.0302, "step": 35975 }, { "epoch": 0.09, "learning_rate": 9.120111731843576e-05, "loss": 1.0308, "step": 35980 }, { "epoch": 0.09, "learning_rate": 9.119985907695406e-05, "loss": 1.0293, "step": 35985 }, { "epoch": 0.09, "learning_rate": 9.119860083547235e-05, "loss": 1.03, "step": 35990 }, { "epoch": 0.09, "learning_rate": 9.119734259399064e-05, "loss": 1.0289, "step": 35995 }, { "epoch": 0.09, "learning_rate": 9.119608435250893e-05, "loss": 1.0276, "step": 36000 }, { "epoch": 0.09, "learning_rate": 9.119482611102724e-05, "loss": 1.0315, "step": 36005 }, { "epoch": 0.09, "learning_rate": 9.119356786954553e-05, "loss": 1.0297, "step": 36010 }, { "epoch": 0.09, "learning_rate": 9.119230962806382e-05, "loss": 1.029, "step": 36015 }, { "epoch": 0.09, "learning_rate": 9.119105138658211e-05, "loss": 1.0301, "step": 36020 }, { "epoch": 0.09, "learning_rate": 9.118979314510042e-05, "loss": 1.0295, "step": 36025 }, { "epoch": 0.09, "learning_rate": 9.118853490361871e-05, "loss": 1.0306, "step": 36030 }, { "epoch": 0.09, "learning_rate": 9.1187276662137e-05, "loss": 1.0306, "step": 36035 }, { "epoch": 0.09, "learning_rate": 9.118601842065529e-05, "loss": 1.0277, "step": 36040 }, { "epoch": 0.09, "learning_rate": 9.118476017917358e-05, "loss": 1.0278, "step": 36045 }, { "epoch": 0.09, "learning_rate": 9.118350193769189e-05, "loss": 1.0297, "step": 36050 }, { "epoch": 0.09, "learning_rate": 9.118224369621018e-05, "loss": 1.03, "step": 36055 }, { "epoch": 0.09, "learning_rate": 9.118098545472847e-05, "loss": 1.0288, "step": 36060 }, { "epoch": 0.09, "learning_rate": 9.117972721324676e-05, "loss": 1.0292, "step": 36065 }, { "epoch": 0.09, "learning_rate": 9.117846897176507e-05, "loss": 1.0291, "step": 36070 }, { "epoch": 0.09, "learning_rate": 9.117721073028336e-05, "loss": 1.0294, "step": 36075 }, { "epoch": 0.09, "learning_rate": 9.117595248880165e-05, "loss": 1.0315, "step": 36080 }, { "epoch": 0.09, "learning_rate": 9.117469424731994e-05, "loss": 1.0322, "step": 36085 }, { "epoch": 0.09, "learning_rate": 9.117343600583825e-05, "loss": 1.0302, "step": 36090 }, { "epoch": 0.09, "learning_rate": 9.117217776435654e-05, "loss": 1.0278, "step": 36095 }, { "epoch": 0.09, "learning_rate": 9.117091952287483e-05, "loss": 1.032, "step": 36100 }, { "epoch": 0.09, "learning_rate": 9.116966128139312e-05, "loss": 1.033, "step": 36105 }, { "epoch": 0.09, "learning_rate": 9.116840303991141e-05, "loss": 1.0304, "step": 36110 }, { "epoch": 0.09, "learning_rate": 9.116714479842972e-05, "loss": 1.0296, "step": 36115 }, { "epoch": 0.09, "learning_rate": 9.116588655694801e-05, "loss": 1.0312, "step": 36120 }, { "epoch": 0.09, "learning_rate": 9.11646283154663e-05, "loss": 1.0322, "step": 36125 }, { "epoch": 0.09, "learning_rate": 9.11633700739846e-05, "loss": 1.0306, "step": 36130 }, { "epoch": 0.09, "learning_rate": 9.11621118325029e-05, "loss": 1.028, "step": 36135 }, { "epoch": 0.09, "learning_rate": 9.116085359102119e-05, "loss": 1.0337, "step": 36140 }, { "epoch": 0.09, "learning_rate": 9.115959534953948e-05, "loss": 1.0328, "step": 36145 }, { "epoch": 0.09, "learning_rate": 9.115833710805777e-05, "loss": 1.0299, "step": 36150 }, { "epoch": 0.09, "learning_rate": 9.115707886657608e-05, "loss": 1.0306, "step": 36155 }, { "epoch": 0.09, "learning_rate": 9.115582062509437e-05, "loss": 1.0302, "step": 36160 }, { "epoch": 0.09, "learning_rate": 9.115456238361266e-05, "loss": 1.0295, "step": 36165 }, { "epoch": 0.09, "learning_rate": 9.115330414213095e-05, "loss": 1.0547, "step": 36170 }, { "epoch": 0.09, "learning_rate": 9.115204590064925e-05, "loss": 1.0311, "step": 36175 }, { "epoch": 0.09, "learning_rate": 9.115078765916755e-05, "loss": 1.0314, "step": 36180 }, { "epoch": 0.09, "learning_rate": 9.114952941768584e-05, "loss": 1.0319, "step": 36185 }, { "epoch": 0.09, "learning_rate": 9.114827117620413e-05, "loss": 1.029, "step": 36190 }, { "epoch": 0.09, "learning_rate": 9.114701293472243e-05, "loss": 1.0312, "step": 36195 }, { "epoch": 0.09, "learning_rate": 9.114575469324073e-05, "loss": 1.0302, "step": 36200 }, { "epoch": 0.09, "learning_rate": 9.114449645175902e-05, "loss": 1.0319, "step": 36205 }, { "epoch": 0.09, "learning_rate": 9.114323821027731e-05, "loss": 1.0299, "step": 36210 }, { "epoch": 0.09, "learning_rate": 9.114197996879562e-05, "loss": 1.0306, "step": 36215 }, { "epoch": 0.09, "learning_rate": 9.114072172731391e-05, "loss": 1.029, "step": 36220 }, { "epoch": 0.09, "learning_rate": 9.113946348583222e-05, "loss": 1.0288, "step": 36225 }, { "epoch": 0.09, "learning_rate": 9.113820524435051e-05, "loss": 1.0293, "step": 36230 }, { "epoch": 0.09, "learning_rate": 9.11369470028688e-05, "loss": 1.031, "step": 36235 }, { "epoch": 0.09, "learning_rate": 9.113568876138709e-05, "loss": 1.0299, "step": 36240 }, { "epoch": 0.09, "learning_rate": 9.11344305199054e-05, "loss": 1.0318, "step": 36245 }, { "epoch": 0.09, "learning_rate": 9.113317227842369e-05, "loss": 1.0291, "step": 36250 }, { "epoch": 0.09, "learning_rate": 9.113191403694198e-05, "loss": 1.0287, "step": 36255 }, { "epoch": 0.09, "learning_rate": 9.113065579546027e-05, "loss": 1.0297, "step": 36260 }, { "epoch": 0.09, "learning_rate": 9.112939755397856e-05, "loss": 1.0292, "step": 36265 }, { "epoch": 0.09, "learning_rate": 9.112813931249687e-05, "loss": 1.0307, "step": 36270 }, { "epoch": 0.09, "learning_rate": 9.112688107101516e-05, "loss": 1.0299, "step": 36275 }, { "epoch": 0.09, "learning_rate": 9.112562282953345e-05, "loss": 1.03, "step": 36280 }, { "epoch": 0.09, "learning_rate": 9.112436458805174e-05, "loss": 1.0292, "step": 36285 }, { "epoch": 0.09, "learning_rate": 9.112310634657005e-05, "loss": 1.0324, "step": 36290 }, { "epoch": 0.09, "learning_rate": 9.112184810508834e-05, "loss": 1.0297, "step": 36295 }, { "epoch": 0.09, "learning_rate": 9.112058986360663e-05, "loss": 1.0497, "step": 36300 }, { "epoch": 0.09, "learning_rate": 9.111933162212492e-05, "loss": 1.0314, "step": 36305 }, { "epoch": 0.09, "learning_rate": 9.111807338064323e-05, "loss": 1.0306, "step": 36310 }, { "epoch": 0.09, "learning_rate": 9.111681513916152e-05, "loss": 1.0314, "step": 36315 }, { "epoch": 0.09, "learning_rate": 9.111555689767981e-05, "loss": 1.0315, "step": 36320 }, { "epoch": 0.09, "learning_rate": 9.11142986561981e-05, "loss": 1.0297, "step": 36325 }, { "epoch": 0.09, "learning_rate": 9.11130404147164e-05, "loss": 1.0295, "step": 36330 }, { "epoch": 0.09, "learning_rate": 9.11117821732347e-05, "loss": 1.0287, "step": 36335 }, { "epoch": 0.09, "learning_rate": 9.111052393175299e-05, "loss": 1.0303, "step": 36340 }, { "epoch": 0.09, "learning_rate": 9.110926569027128e-05, "loss": 1.0297, "step": 36345 }, { "epoch": 0.09, "learning_rate": 9.110800744878957e-05, "loss": 1.0306, "step": 36350 }, { "epoch": 0.09, "learning_rate": 9.110674920730788e-05, "loss": 1.031, "step": 36355 }, { "epoch": 0.09, "learning_rate": 9.110549096582617e-05, "loss": 1.0303, "step": 36360 }, { "epoch": 0.09, "learning_rate": 9.110423272434446e-05, "loss": 1.0308, "step": 36365 }, { "epoch": 0.09, "learning_rate": 9.110297448286275e-05, "loss": 1.029, "step": 36370 }, { "epoch": 0.09, "learning_rate": 9.110171624138106e-05, "loss": 1.0305, "step": 36375 }, { "epoch": 0.09, "learning_rate": 9.110045799989935e-05, "loss": 1.0294, "step": 36380 }, { "epoch": 0.09, "learning_rate": 9.109919975841764e-05, "loss": 1.0282, "step": 36385 }, { "epoch": 0.09, "learning_rate": 9.109794151693593e-05, "loss": 1.0302, "step": 36390 }, { "epoch": 0.09, "learning_rate": 9.109668327545422e-05, "loss": 1.0287, "step": 36395 }, { "epoch": 0.09, "learning_rate": 9.109542503397253e-05, "loss": 1.031, "step": 36400 }, { "epoch": 0.09, "learning_rate": 9.109416679249082e-05, "loss": 1.0288, "step": 36405 }, { "epoch": 0.09, "learning_rate": 9.109290855100911e-05, "loss": 1.03, "step": 36410 }, { "epoch": 0.09, "learning_rate": 9.10916503095274e-05, "loss": 1.0271, "step": 36415 }, { "epoch": 0.09, "learning_rate": 9.109039206804571e-05, "loss": 1.0312, "step": 36420 }, { "epoch": 0.09, "learning_rate": 9.1089133826564e-05, "loss": 1.0299, "step": 36425 }, { "epoch": 0.09, "learning_rate": 9.10878755850823e-05, "loss": 1.0307, "step": 36430 }, { "epoch": 0.09, "learning_rate": 9.108661734360058e-05, "loss": 1.0297, "step": 36435 }, { "epoch": 0.09, "learning_rate": 9.108535910211889e-05, "loss": 1.0305, "step": 36440 }, { "epoch": 0.09, "learning_rate": 9.108410086063718e-05, "loss": 1.0303, "step": 36445 }, { "epoch": 0.09, "learning_rate": 9.108284261915547e-05, "loss": 1.0281, "step": 36450 }, { "epoch": 0.09, "learning_rate": 9.108158437767376e-05, "loss": 1.0292, "step": 36455 }, { "epoch": 0.09, "learning_rate": 9.108032613619206e-05, "loss": 1.0302, "step": 36460 }, { "epoch": 0.09, "learning_rate": 9.107906789471036e-05, "loss": 1.0323, "step": 36465 }, { "epoch": 0.09, "learning_rate": 9.107780965322865e-05, "loss": 1.029, "step": 36470 }, { "epoch": 0.09, "learning_rate": 9.107655141174694e-05, "loss": 1.0274, "step": 36475 }, { "epoch": 0.09, "learning_rate": 9.107529317026524e-05, "loss": 1.0307, "step": 36480 }, { "epoch": 0.09, "learning_rate": 9.107403492878354e-05, "loss": 1.0297, "step": 36485 }, { "epoch": 0.09, "learning_rate": 9.107277668730183e-05, "loss": 1.0288, "step": 36490 }, { "epoch": 0.09, "learning_rate": 9.107151844582012e-05, "loss": 1.0314, "step": 36495 }, { "epoch": 0.09, "learning_rate": 9.107026020433842e-05, "loss": 1.0298, "step": 36500 }, { "epoch": 0.09, "learning_rate": 9.106900196285672e-05, "loss": 1.029, "step": 36505 }, { "epoch": 0.09, "learning_rate": 9.106774372137501e-05, "loss": 1.0293, "step": 36510 }, { "epoch": 0.09, "learning_rate": 9.10664854798933e-05, "loss": 1.03, "step": 36515 }, { "epoch": 0.09, "learning_rate": 9.10652272384116e-05, "loss": 1.0296, "step": 36520 }, { "epoch": 0.09, "learning_rate": 9.106396899692989e-05, "loss": 1.0284, "step": 36525 }, { "epoch": 0.09, "learning_rate": 9.106271075544819e-05, "loss": 1.0306, "step": 36530 }, { "epoch": 0.09, "learning_rate": 9.106145251396648e-05, "loss": 1.0495, "step": 36535 }, { "epoch": 0.09, "learning_rate": 9.106019427248478e-05, "loss": 1.0325, "step": 36540 }, { "epoch": 0.09, "learning_rate": 9.105893603100307e-05, "loss": 1.0304, "step": 36545 }, { "epoch": 0.09, "learning_rate": 9.105767778952137e-05, "loss": 1.0298, "step": 36550 }, { "epoch": 0.09, "learning_rate": 9.105641954803966e-05, "loss": 1.0298, "step": 36555 }, { "epoch": 0.09, "learning_rate": 9.105516130655796e-05, "loss": 1.0317, "step": 36560 }, { "epoch": 0.09, "learning_rate": 9.105390306507625e-05, "loss": 1.0306, "step": 36565 }, { "epoch": 0.09, "learning_rate": 9.105264482359455e-05, "loss": 1.0292, "step": 36570 }, { "epoch": 0.09, "learning_rate": 9.105138658211284e-05, "loss": 1.03, "step": 36575 }, { "epoch": 0.09, "learning_rate": 9.105012834063114e-05, "loss": 1.0325, "step": 36580 }, { "epoch": 0.09, "learning_rate": 9.104887009914943e-05, "loss": 1.034, "step": 36585 }, { "epoch": 0.09, "learning_rate": 9.104761185766772e-05, "loss": 1.0302, "step": 36590 }, { "epoch": 0.09, "learning_rate": 9.104635361618602e-05, "loss": 1.0293, "step": 36595 }, { "epoch": 0.09, "learning_rate": 9.104509537470432e-05, "loss": 1.0282, "step": 36600 }, { "epoch": 0.09, "learning_rate": 9.104383713322261e-05, "loss": 1.0283, "step": 36605 }, { "epoch": 0.09, "learning_rate": 9.10425788917409e-05, "loss": 1.0285, "step": 36610 }, { "epoch": 0.09, "learning_rate": 9.10413206502592e-05, "loss": 1.0294, "step": 36615 }, { "epoch": 0.09, "learning_rate": 9.10400624087775e-05, "loss": 1.0303, "step": 36620 }, { "epoch": 0.09, "learning_rate": 9.103880416729579e-05, "loss": 1.0328, "step": 36625 }, { "epoch": 0.09, "learning_rate": 9.103754592581408e-05, "loss": 1.0303, "step": 36630 }, { "epoch": 0.09, "learning_rate": 9.103628768433238e-05, "loss": 1.0325, "step": 36635 }, { "epoch": 0.09, "learning_rate": 9.103502944285068e-05, "loss": 1.0299, "step": 36640 }, { "epoch": 0.09, "learning_rate": 9.103377120136897e-05, "loss": 1.0306, "step": 36645 }, { "epoch": 0.09, "learning_rate": 9.103251295988726e-05, "loss": 1.0286, "step": 36650 }, { "epoch": 0.09, "learning_rate": 9.103125471840555e-05, "loss": 1.0293, "step": 36655 }, { "epoch": 0.09, "learning_rate": 9.102999647692386e-05, "loss": 1.031, "step": 36660 }, { "epoch": 0.09, "learning_rate": 9.102873823544215e-05, "loss": 1.0288, "step": 36665 }, { "epoch": 0.09, "learning_rate": 9.102747999396044e-05, "loss": 1.0308, "step": 36670 }, { "epoch": 0.09, "learning_rate": 9.102622175247873e-05, "loss": 1.0315, "step": 36675 }, { "epoch": 0.09, "learning_rate": 9.102496351099704e-05, "loss": 1.0309, "step": 36680 }, { "epoch": 0.09, "learning_rate": 9.102370526951533e-05, "loss": 1.029, "step": 36685 }, { "epoch": 0.09, "learning_rate": 9.102244702803362e-05, "loss": 1.0304, "step": 36690 }, { "epoch": 0.09, "learning_rate": 9.102118878655191e-05, "loss": 1.0303, "step": 36695 }, { "epoch": 0.09, "learning_rate": 9.10199305450702e-05, "loss": 1.0313, "step": 36700 }, { "epoch": 0.09, "learning_rate": 9.101867230358851e-05, "loss": 1.0297, "step": 36705 }, { "epoch": 0.09, "learning_rate": 9.101741406210681e-05, "loss": 1.0288, "step": 36710 }, { "epoch": 0.09, "learning_rate": 9.10161558206251e-05, "loss": 1.0314, "step": 36715 }, { "epoch": 0.09, "learning_rate": 9.10148975791434e-05, "loss": 1.0296, "step": 36720 }, { "epoch": 0.09, "learning_rate": 9.101363933766169e-05, "loss": 1.0291, "step": 36725 }, { "epoch": 0.09, "learning_rate": 9.101238109617999e-05, "loss": 1.0301, "step": 36730 }, { "epoch": 0.09, "learning_rate": 9.101112285469828e-05, "loss": 1.0314, "step": 36735 }, { "epoch": 0.09, "learning_rate": 9.100986461321658e-05, "loss": 1.0297, "step": 36740 }, { "epoch": 0.09, "learning_rate": 9.100860637173487e-05, "loss": 1.0309, "step": 36745 }, { "epoch": 0.09, "learning_rate": 9.100734813025317e-05, "loss": 1.0307, "step": 36750 }, { "epoch": 0.09, "learning_rate": 9.100608988877146e-05, "loss": 1.0282, "step": 36755 }, { "epoch": 0.09, "learning_rate": 9.100483164728976e-05, "loss": 1.0314, "step": 36760 }, { "epoch": 0.09, "learning_rate": 9.100357340580805e-05, "loss": 1.0328, "step": 36765 }, { "epoch": 0.09, "learning_rate": 9.100231516432635e-05, "loss": 1.0298, "step": 36770 }, { "epoch": 0.09, "learning_rate": 9.100105692284464e-05, "loss": 1.0306, "step": 36775 }, { "epoch": 0.09, "learning_rate": 9.099979868136294e-05, "loss": 1.0316, "step": 36780 }, { "epoch": 0.09, "learning_rate": 9.099854043988123e-05, "loss": 1.0278, "step": 36785 }, { "epoch": 0.09, "learning_rate": 9.099728219839952e-05, "loss": 1.0297, "step": 36790 }, { "epoch": 0.09, "learning_rate": 9.099602395691782e-05, "loss": 1.031, "step": 36795 }, { "epoch": 0.09, "learning_rate": 9.099476571543612e-05, "loss": 1.03, "step": 36800 }, { "epoch": 0.09, "learning_rate": 9.099350747395441e-05, "loss": 1.0283, "step": 36805 }, { "epoch": 0.09, "learning_rate": 9.09922492324727e-05, "loss": 1.031, "step": 36810 }, { "epoch": 0.09, "learning_rate": 9.0990990990991e-05, "loss": 1.0285, "step": 36815 }, { "epoch": 0.09, "learning_rate": 9.09897327495093e-05, "loss": 1.0323, "step": 36820 }, { "epoch": 0.09, "learning_rate": 9.098847450802759e-05, "loss": 1.0322, "step": 36825 }, { "epoch": 0.09, "learning_rate": 9.098721626654588e-05, "loss": 1.0297, "step": 36830 }, { "epoch": 0.09, "learning_rate": 9.098595802506418e-05, "loss": 1.0289, "step": 36835 }, { "epoch": 0.09, "learning_rate": 9.098469978358248e-05, "loss": 1.0276, "step": 36840 }, { "epoch": 0.09, "learning_rate": 9.098344154210077e-05, "loss": 1.0302, "step": 36845 }, { "epoch": 0.09, "learning_rate": 9.098218330061906e-05, "loss": 1.0287, "step": 36850 }, { "epoch": 0.09, "learning_rate": 9.098092505913735e-05, "loss": 1.0302, "step": 36855 }, { "epoch": 0.09, "learning_rate": 9.097966681765566e-05, "loss": 1.0267, "step": 36860 }, { "epoch": 0.09, "learning_rate": 9.097840857617395e-05, "loss": 1.0289, "step": 36865 }, { "epoch": 0.09, "learning_rate": 9.097715033469224e-05, "loss": 1.0296, "step": 36870 }, { "epoch": 0.09, "learning_rate": 9.097589209321053e-05, "loss": 1.0296, "step": 36875 }, { "epoch": 0.09, "learning_rate": 9.097463385172884e-05, "loss": 1.0336, "step": 36880 }, { "epoch": 0.09, "learning_rate": 9.097337561024713e-05, "loss": 1.031, "step": 36885 }, { "epoch": 0.09, "learning_rate": 9.097211736876542e-05, "loss": 1.0285, "step": 36890 }, { "epoch": 0.09, "learning_rate": 9.097085912728371e-05, "loss": 1.0294, "step": 36895 }, { "epoch": 0.09, "learning_rate": 9.096960088580202e-05, "loss": 1.0321, "step": 36900 }, { "epoch": 0.09, "learning_rate": 9.09683426443203e-05, "loss": 1.0282, "step": 36905 }, { "epoch": 0.09, "learning_rate": 9.09670844028386e-05, "loss": 1.0307, "step": 36910 }, { "epoch": 0.09, "learning_rate": 9.096582616135689e-05, "loss": 1.0281, "step": 36915 }, { "epoch": 0.09, "learning_rate": 9.096456791987518e-05, "loss": 1.0304, "step": 36920 }, { "epoch": 0.09, "learning_rate": 9.096330967839349e-05, "loss": 1.0303, "step": 36925 }, { "epoch": 0.09, "learning_rate": 9.096205143691178e-05, "loss": 1.0299, "step": 36930 }, { "epoch": 0.09, "learning_rate": 9.096079319543007e-05, "loss": 1.0286, "step": 36935 }, { "epoch": 0.09, "learning_rate": 9.095953495394836e-05, "loss": 1.0308, "step": 36940 }, { "epoch": 0.09, "learning_rate": 9.095827671246667e-05, "loss": 1.0301, "step": 36945 }, { "epoch": 0.09, "learning_rate": 9.095701847098496e-05, "loss": 1.0292, "step": 36950 }, { "epoch": 0.09, "learning_rate": 9.095576022950325e-05, "loss": 1.029, "step": 36955 }, { "epoch": 0.09, "learning_rate": 9.095450198802154e-05, "loss": 1.0307, "step": 36960 }, { "epoch": 0.09, "learning_rate": 9.095324374653985e-05, "loss": 1.0304, "step": 36965 }, { "epoch": 0.09, "learning_rate": 9.095198550505814e-05, "loss": 1.0305, "step": 36970 }, { "epoch": 0.09, "learning_rate": 9.095072726357643e-05, "loss": 1.0319, "step": 36975 }, { "epoch": 0.09, "learning_rate": 9.094946902209472e-05, "loss": 1.0287, "step": 36980 }, { "epoch": 0.09, "learning_rate": 9.094821078061301e-05, "loss": 1.029, "step": 36985 }, { "epoch": 0.09, "learning_rate": 9.094695253913132e-05, "loss": 1.0305, "step": 36990 }, { "epoch": 0.09, "learning_rate": 9.094569429764961e-05, "loss": 1.0308, "step": 36995 }, { "epoch": 0.09, "learning_rate": 9.09444360561679e-05, "loss": 1.029, "step": 37000 }, { "epoch": 0.09, "learning_rate": 9.094317781468619e-05, "loss": 1.0294, "step": 37005 }, { "epoch": 0.09, "learning_rate": 9.09419195732045e-05, "loss": 1.0299, "step": 37010 }, { "epoch": 0.09, "learning_rate": 9.094066133172279e-05, "loss": 1.0308, "step": 37015 }, { "epoch": 0.09, "learning_rate": 9.093940309024108e-05, "loss": 1.0323, "step": 37020 }, { "epoch": 0.09, "learning_rate": 9.093814484875937e-05, "loss": 1.0324, "step": 37025 }, { "epoch": 0.09, "learning_rate": 9.093688660727768e-05, "loss": 1.0294, "step": 37030 }, { "epoch": 0.09, "learning_rate": 9.093562836579597e-05, "loss": 1.0305, "step": 37035 }, { "epoch": 0.09, "learning_rate": 9.093437012431426e-05, "loss": 1.0276, "step": 37040 }, { "epoch": 0.09, "learning_rate": 9.093311188283255e-05, "loss": 1.0302, "step": 37045 }, { "epoch": 0.09, "learning_rate": 9.093185364135084e-05, "loss": 1.0292, "step": 37050 }, { "epoch": 0.09, "learning_rate": 9.093059539986915e-05, "loss": 1.0292, "step": 37055 }, { "epoch": 0.09, "learning_rate": 9.092933715838744e-05, "loss": 1.0301, "step": 37060 }, { "epoch": 0.09, "learning_rate": 9.092807891690573e-05, "loss": 1.0292, "step": 37065 }, { "epoch": 0.09, "learning_rate": 9.092682067542402e-05, "loss": 1.0285, "step": 37070 }, { "epoch": 0.09, "learning_rate": 9.092556243394233e-05, "loss": 1.0286, "step": 37075 }, { "epoch": 0.09, "learning_rate": 9.092430419246062e-05, "loss": 1.0291, "step": 37080 }, { "epoch": 0.09, "learning_rate": 9.092304595097891e-05, "loss": 1.0296, "step": 37085 }, { "epoch": 0.09, "learning_rate": 9.09217877094972e-05, "loss": 1.0298, "step": 37090 }, { "epoch": 0.09, "learning_rate": 9.092052946801551e-05, "loss": 1.0295, "step": 37095 }, { "epoch": 0.09, "learning_rate": 9.09192712265338e-05, "loss": 1.0315, "step": 37100 }, { "epoch": 0.09, "learning_rate": 9.091801298505209e-05, "loss": 1.0327, "step": 37105 }, { "epoch": 0.09, "learning_rate": 9.091675474357038e-05, "loss": 1.03, "step": 37110 }, { "epoch": 0.09, "learning_rate": 9.091549650208868e-05, "loss": 1.0297, "step": 37115 }, { "epoch": 0.09, "learning_rate": 9.091423826060698e-05, "loss": 1.0319, "step": 37120 }, { "epoch": 0.09, "learning_rate": 9.091298001912527e-05, "loss": 1.0298, "step": 37125 }, { "epoch": 0.09, "learning_rate": 9.091172177764356e-05, "loss": 1.0305, "step": 37130 }, { "epoch": 0.09, "learning_rate": 9.091046353616186e-05, "loss": 1.029, "step": 37135 }, { "epoch": 0.09, "learning_rate": 9.090920529468016e-05, "loss": 1.0296, "step": 37140 }, { "epoch": 0.09, "learning_rate": 9.090794705319845e-05, "loss": 1.0544, "step": 37145 }, { "epoch": 0.09, "learning_rate": 9.090668881171674e-05, "loss": 1.0303, "step": 37150 }, { "epoch": 0.09, "learning_rate": 9.090543057023504e-05, "loss": 1.031, "step": 37155 }, { "epoch": 0.09, "learning_rate": 9.090417232875334e-05, "loss": 1.0308, "step": 37160 }, { "epoch": 0.09, "learning_rate": 9.090291408727163e-05, "loss": 1.0293, "step": 37165 }, { "epoch": 0.09, "learning_rate": 9.090165584578992e-05, "loss": 1.0295, "step": 37170 }, { "epoch": 0.09, "learning_rate": 9.090039760430822e-05, "loss": 1.0302, "step": 37175 }, { "epoch": 0.09, "learning_rate": 9.089913936282651e-05, "loss": 1.0286, "step": 37180 }, { "epoch": 0.09, "learning_rate": 9.089788112134481e-05, "loss": 1.0301, "step": 37185 }, { "epoch": 0.09, "learning_rate": 9.08966228798631e-05, "loss": 1.0314, "step": 37190 }, { "epoch": 0.09, "learning_rate": 9.08953646383814e-05, "loss": 1.0681, "step": 37195 }, { "epoch": 0.09, "learning_rate": 9.089410639689969e-05, "loss": 1.0295, "step": 37200 }, { "epoch": 0.09, "learning_rate": 9.089284815541799e-05, "loss": 1.0293, "step": 37205 }, { "epoch": 0.09, "learning_rate": 9.08915899139363e-05, "loss": 1.0297, "step": 37210 }, { "epoch": 0.09, "learning_rate": 9.089033167245459e-05, "loss": 1.0276, "step": 37215 }, { "epoch": 0.09, "learning_rate": 9.088907343097288e-05, "loss": 1.0305, "step": 37220 }, { "epoch": 0.09, "learning_rate": 9.088781518949117e-05, "loss": 1.0275, "step": 37225 }, { "epoch": 0.09, "learning_rate": 9.088655694800948e-05, "loss": 1.0286, "step": 37230 }, { "epoch": 0.09, "learning_rate": 9.088529870652777e-05, "loss": 1.0278, "step": 37235 }, { "epoch": 0.09, "learning_rate": 9.088404046504606e-05, "loss": 1.0322, "step": 37240 }, { "epoch": 0.09, "learning_rate": 9.088278222356435e-05, "loss": 1.0277, "step": 37245 }, { "epoch": 0.09, "learning_rate": 9.088152398208264e-05, "loss": 1.0292, "step": 37250 }, { "epoch": 0.09, "learning_rate": 9.088026574060095e-05, "loss": 1.0319, "step": 37255 }, { "epoch": 0.09, "learning_rate": 9.087900749911924e-05, "loss": 1.0323, "step": 37260 }, { "epoch": 0.09, "learning_rate": 9.087774925763753e-05, "loss": 1.0298, "step": 37265 }, { "epoch": 0.09, "learning_rate": 9.087649101615582e-05, "loss": 1.0307, "step": 37270 }, { "epoch": 0.09, "learning_rate": 9.087523277467413e-05, "loss": 1.0328, "step": 37275 }, { "epoch": 0.09, "learning_rate": 9.087397453319242e-05, "loss": 1.029, "step": 37280 }, { "epoch": 0.09, "learning_rate": 9.087271629171071e-05, "loss": 1.0294, "step": 37285 }, { "epoch": 0.09, "learning_rate": 9.0871458050229e-05, "loss": 1.0296, "step": 37290 }, { "epoch": 0.09, "learning_rate": 9.087019980874731e-05, "loss": 1.0287, "step": 37295 }, { "epoch": 0.09, "learning_rate": 9.08689415672656e-05, "loss": 1.0321, "step": 37300 }, { "epoch": 0.09, "learning_rate": 9.086768332578389e-05, "loss": 1.0314, "step": 37305 }, { "epoch": 0.09, "learning_rate": 9.086642508430218e-05, "loss": 1.0291, "step": 37310 }, { "epoch": 0.09, "learning_rate": 9.086516684282048e-05, "loss": 1.0305, "step": 37315 }, { "epoch": 0.09, "learning_rate": 9.086390860133878e-05, "loss": 1.0294, "step": 37320 }, { "epoch": 0.09, "learning_rate": 9.086265035985707e-05, "loss": 1.0529, "step": 37325 }, { "epoch": 0.09, "learning_rate": 9.086139211837536e-05, "loss": 1.0534, "step": 37330 }, { "epoch": 0.09, "learning_rate": 9.086013387689366e-05, "loss": 1.0292, "step": 37335 }, { "epoch": 0.09, "learning_rate": 9.085887563541196e-05, "loss": 1.0285, "step": 37340 }, { "epoch": 0.09, "learning_rate": 9.085761739393025e-05, "loss": 1.031, "step": 37345 }, { "epoch": 0.09, "learning_rate": 9.085635915244854e-05, "loss": 1.0299, "step": 37350 }, { "epoch": 0.09, "learning_rate": 9.085510091096684e-05, "loss": 1.032, "step": 37355 }, { "epoch": 0.09, "learning_rate": 9.085384266948514e-05, "loss": 1.0283, "step": 37360 }, { "epoch": 0.09, "learning_rate": 9.085258442800343e-05, "loss": 1.0292, "step": 37365 }, { "epoch": 0.09, "learning_rate": 9.085132618652172e-05, "loss": 1.0299, "step": 37370 }, { "epoch": 0.09, "learning_rate": 9.085006794504002e-05, "loss": 1.0316, "step": 37375 }, { "epoch": 0.09, "learning_rate": 9.08488097035583e-05, "loss": 1.0298, "step": 37380 }, { "epoch": 0.09, "learning_rate": 9.084755146207661e-05, "loss": 1.0285, "step": 37385 }, { "epoch": 0.09, "learning_rate": 9.08462932205949e-05, "loss": 1.0269, "step": 37390 }, { "epoch": 0.09, "learning_rate": 9.08450349791132e-05, "loss": 1.0285, "step": 37395 }, { "epoch": 0.09, "learning_rate": 9.084377673763149e-05, "loss": 1.0297, "step": 37400 }, { "epoch": 0.09, "learning_rate": 9.084251849614979e-05, "loss": 1.0299, "step": 37405 }, { "epoch": 0.09, "learning_rate": 9.084126025466808e-05, "loss": 1.0306, "step": 37410 }, { "epoch": 0.09, "learning_rate": 9.084000201318637e-05, "loss": 1.0277, "step": 37415 }, { "epoch": 0.09, "learning_rate": 9.083874377170467e-05, "loss": 1.0295, "step": 37420 }, { "epoch": 0.09, "learning_rate": 9.083748553022297e-05, "loss": 1.0317, "step": 37425 }, { "epoch": 0.09, "learning_rate": 9.083622728874126e-05, "loss": 1.0295, "step": 37430 }, { "epoch": 0.09, "learning_rate": 9.083496904725955e-05, "loss": 1.0287, "step": 37435 }, { "epoch": 0.09, "learning_rate": 9.083371080577785e-05, "loss": 1.0323, "step": 37440 }, { "epoch": 0.09, "learning_rate": 9.083245256429614e-05, "loss": 1.0277, "step": 37445 }, { "epoch": 0.09, "learning_rate": 9.083119432281444e-05, "loss": 1.0279, "step": 37450 }, { "epoch": 0.09, "learning_rate": 9.082993608133273e-05, "loss": 1.0298, "step": 37455 }, { "epoch": 0.09, "learning_rate": 9.082867783985103e-05, "loss": 1.0289, "step": 37460 }, { "epoch": 0.09, "learning_rate": 9.082741959836932e-05, "loss": 1.0297, "step": 37465 }, { "epoch": 0.09, "learning_rate": 9.082616135688762e-05, "loss": 1.0315, "step": 37470 }, { "epoch": 0.09, "learning_rate": 9.082490311540591e-05, "loss": 1.0481, "step": 37475 }, { "epoch": 0.09, "learning_rate": 9.08236448739242e-05, "loss": 1.0315, "step": 37480 }, { "epoch": 0.09, "learning_rate": 9.08223866324425e-05, "loss": 1.0292, "step": 37485 }, { "epoch": 0.09, "learning_rate": 9.08211283909608e-05, "loss": 1.029, "step": 37490 }, { "epoch": 0.09, "learning_rate": 9.08198701494791e-05, "loss": 1.0294, "step": 37495 }, { "epoch": 0.09, "learning_rate": 9.081861190799739e-05, "loss": 1.0308, "step": 37500 }, { "epoch": 0.09, "learning_rate": 9.081735366651568e-05, "loss": 1.0294, "step": 37505 }, { "epoch": 0.09, "learning_rate": 9.081609542503397e-05, "loss": 1.0283, "step": 37510 }, { "epoch": 0.09, "learning_rate": 9.081483718355227e-05, "loss": 1.0296, "step": 37515 }, { "epoch": 0.09, "learning_rate": 9.081357894207057e-05, "loss": 1.03, "step": 37520 }, { "epoch": 0.09, "learning_rate": 9.081232070058886e-05, "loss": 1.0311, "step": 37525 }, { "epoch": 0.09, "learning_rate": 9.081106245910715e-05, "loss": 1.0269, "step": 37530 }, { "epoch": 0.09, "learning_rate": 9.080980421762545e-05, "loss": 1.0269, "step": 37535 }, { "epoch": 0.09, "learning_rate": 9.080854597614375e-05, "loss": 1.0304, "step": 37540 }, { "epoch": 0.09, "learning_rate": 9.080728773466204e-05, "loss": 1.0281, "step": 37545 }, { "epoch": 0.09, "learning_rate": 9.080602949318033e-05, "loss": 1.0298, "step": 37550 }, { "epoch": 0.09, "learning_rate": 9.080477125169863e-05, "loss": 1.0292, "step": 37555 }, { "epoch": 0.09, "learning_rate": 9.080351301021693e-05, "loss": 1.0284, "step": 37560 }, { "epoch": 0.09, "learning_rate": 9.080225476873522e-05, "loss": 1.0295, "step": 37565 }, { "epoch": 0.09, "learning_rate": 9.080099652725351e-05, "loss": 1.0296, "step": 37570 }, { "epoch": 0.09, "learning_rate": 9.07997382857718e-05, "loss": 1.028, "step": 37575 }, { "epoch": 0.09, "learning_rate": 9.07984800442901e-05, "loss": 1.0295, "step": 37580 }, { "epoch": 0.09, "learning_rate": 9.07972218028084e-05, "loss": 1.0287, "step": 37585 }, { "epoch": 0.09, "learning_rate": 9.079596356132669e-05, "loss": 1.0307, "step": 37590 }, { "epoch": 0.09, "learning_rate": 9.079470531984498e-05, "loss": 1.0294, "step": 37595 }, { "epoch": 0.09, "learning_rate": 9.079344707836329e-05, "loss": 1.031, "step": 37600 }, { "epoch": 0.09, "learning_rate": 9.079218883688158e-05, "loss": 1.029, "step": 37605 }, { "epoch": 0.09, "learning_rate": 9.079093059539987e-05, "loss": 1.029, "step": 37610 }, { "epoch": 0.09, "learning_rate": 9.078967235391816e-05, "loss": 1.0294, "step": 37615 }, { "epoch": 0.09, "learning_rate": 9.078841411243647e-05, "loss": 1.0318, "step": 37620 }, { "epoch": 0.09, "learning_rate": 9.078715587095476e-05, "loss": 1.0291, "step": 37625 }, { "epoch": 0.09, "learning_rate": 9.078589762947305e-05, "loss": 1.0305, "step": 37630 }, { "epoch": 0.09, "learning_rate": 9.078463938799134e-05, "loss": 1.0315, "step": 37635 }, { "epoch": 0.09, "learning_rate": 9.078338114650963e-05, "loss": 1.0316, "step": 37640 }, { "epoch": 0.09, "learning_rate": 9.078212290502794e-05, "loss": 1.0284, "step": 37645 }, { "epoch": 0.09, "learning_rate": 9.078086466354623e-05, "loss": 1.0311, "step": 37650 }, { "epoch": 0.09, "learning_rate": 9.077960642206452e-05, "loss": 1.0285, "step": 37655 }, { "epoch": 0.09, "learning_rate": 9.077834818058281e-05, "loss": 1.028, "step": 37660 }, { "epoch": 0.09, "learning_rate": 9.077708993910112e-05, "loss": 1.0314, "step": 37665 }, { "epoch": 0.09, "learning_rate": 9.077583169761941e-05, "loss": 1.0295, "step": 37670 }, { "epoch": 0.09, "learning_rate": 9.07745734561377e-05, "loss": 1.0306, "step": 37675 }, { "epoch": 0.09, "learning_rate": 9.077331521465599e-05, "loss": 1.0304, "step": 37680 }, { "epoch": 0.09, "learning_rate": 9.07720569731743e-05, "loss": 1.0311, "step": 37685 }, { "epoch": 0.09, "learning_rate": 9.077079873169259e-05, "loss": 1.0307, "step": 37690 }, { "epoch": 0.09, "learning_rate": 9.076954049021088e-05, "loss": 1.0285, "step": 37695 }, { "epoch": 0.09, "learning_rate": 9.076828224872917e-05, "loss": 1.0305, "step": 37700 }, { "epoch": 0.09, "learning_rate": 9.076702400724746e-05, "loss": 1.0301, "step": 37705 }, { "epoch": 0.09, "learning_rate": 9.076576576576577e-05, "loss": 1.0292, "step": 37710 }, { "epoch": 0.09, "learning_rate": 9.076450752428407e-05, "loss": 1.0309, "step": 37715 }, { "epoch": 0.09, "learning_rate": 9.076324928280237e-05, "loss": 1.0309, "step": 37720 }, { "epoch": 0.09, "learning_rate": 9.076199104132066e-05, "loss": 1.0296, "step": 37725 }, { "epoch": 0.09, "learning_rate": 9.076073279983895e-05, "loss": 1.0291, "step": 37730 }, { "epoch": 0.09, "learning_rate": 9.075947455835725e-05, "loss": 1.0282, "step": 37735 }, { "epoch": 0.09, "learning_rate": 9.075821631687555e-05, "loss": 1.0297, "step": 37740 }, { "epoch": 0.09, "learning_rate": 9.075695807539384e-05, "loss": 1.031, "step": 37745 }, { "epoch": 0.09, "learning_rate": 9.075569983391213e-05, "loss": 1.0303, "step": 37750 }, { "epoch": 0.09, "learning_rate": 9.075444159243043e-05, "loss": 1.0522, "step": 37755 }, { "epoch": 0.09, "learning_rate": 9.075318335094873e-05, "loss": 1.0301, "step": 37760 }, { "epoch": 0.09, "learning_rate": 9.075192510946702e-05, "loss": 1.0276, "step": 37765 }, { "epoch": 0.09, "learning_rate": 9.075066686798531e-05, "loss": 1.0311, "step": 37770 }, { "epoch": 0.09, "learning_rate": 9.07494086265036e-05, "loss": 1.0296, "step": 37775 }, { "epoch": 0.09, "learning_rate": 9.07481503850219e-05, "loss": 1.0299, "step": 37780 }, { "epoch": 0.09, "learning_rate": 9.07468921435402e-05, "loss": 1.0312, "step": 37785 }, { "epoch": 0.09, "learning_rate": 9.074563390205849e-05, "loss": 1.0309, "step": 37790 }, { "epoch": 0.09, "learning_rate": 9.074437566057678e-05, "loss": 1.029, "step": 37795 }, { "epoch": 0.09, "learning_rate": 9.074311741909509e-05, "loss": 1.0313, "step": 37800 }, { "epoch": 0.09, "learning_rate": 9.074185917761338e-05, "loss": 1.0296, "step": 37805 }, { "epoch": 0.09, "learning_rate": 9.074060093613167e-05, "loss": 1.0289, "step": 37810 }, { "epoch": 0.09, "learning_rate": 9.073934269464996e-05, "loss": 1.0305, "step": 37815 }, { "epoch": 0.09, "learning_rate": 9.073808445316827e-05, "loss": 1.0277, "step": 37820 }, { "epoch": 0.09, "learning_rate": 9.073682621168656e-05, "loss": 1.0307, "step": 37825 }, { "epoch": 0.09, "learning_rate": 9.073556797020485e-05, "loss": 1.0272, "step": 37830 }, { "epoch": 0.09, "learning_rate": 9.073430972872314e-05, "loss": 1.0288, "step": 37835 }, { "epoch": 0.09, "learning_rate": 9.073305148724143e-05, "loss": 1.0289, "step": 37840 }, { "epoch": 0.09, "learning_rate": 9.073179324575974e-05, "loss": 1.0282, "step": 37845 }, { "epoch": 0.1, "learning_rate": 9.073053500427803e-05, "loss": 1.0302, "step": 37850 }, { "epoch": 0.1, "learning_rate": 9.072927676279632e-05, "loss": 1.0318, "step": 37855 }, { "epoch": 0.1, "learning_rate": 9.072801852131461e-05, "loss": 1.0318, "step": 37860 }, { "epoch": 0.1, "learning_rate": 9.072676027983292e-05, "loss": 1.0314, "step": 37865 }, { "epoch": 0.1, "learning_rate": 9.072550203835121e-05, "loss": 1.0321, "step": 37870 }, { "epoch": 0.1, "learning_rate": 9.07242437968695e-05, "loss": 1.0281, "step": 37875 }, { "epoch": 0.1, "learning_rate": 9.072298555538779e-05, "loss": 1.0306, "step": 37880 }, { "epoch": 0.1, "learning_rate": 9.07217273139061e-05, "loss": 1.0329, "step": 37885 }, { "epoch": 0.1, "learning_rate": 9.072046907242439e-05, "loss": 1.0286, "step": 37890 }, { "epoch": 0.1, "learning_rate": 9.071921083094268e-05, "loss": 1.0306, "step": 37895 }, { "epoch": 0.1, "learning_rate": 9.071795258946097e-05, "loss": 1.03, "step": 37900 }, { "epoch": 0.1, "learning_rate": 9.071669434797926e-05, "loss": 1.0306, "step": 37905 }, { "epoch": 0.1, "learning_rate": 9.071543610649757e-05, "loss": 1.0278, "step": 37910 }, { "epoch": 0.1, "learning_rate": 9.071417786501586e-05, "loss": 1.0299, "step": 37915 }, { "epoch": 0.1, "learning_rate": 9.071291962353415e-05, "loss": 1.0324, "step": 37920 }, { "epoch": 0.1, "learning_rate": 9.071166138205244e-05, "loss": 1.029, "step": 37925 }, { "epoch": 0.1, "learning_rate": 9.071040314057075e-05, "loss": 1.0319, "step": 37930 }, { "epoch": 0.1, "learning_rate": 9.070914489908904e-05, "loss": 1.0276, "step": 37935 }, { "epoch": 0.1, "learning_rate": 9.070788665760733e-05, "loss": 1.031, "step": 37940 }, { "epoch": 0.1, "learning_rate": 9.070662841612562e-05, "loss": 1.0292, "step": 37945 }, { "epoch": 0.1, "learning_rate": 9.070537017464393e-05, "loss": 1.0292, "step": 37950 }, { "epoch": 0.1, "learning_rate": 9.070411193316222e-05, "loss": 1.0305, "step": 37955 }, { "epoch": 0.1, "learning_rate": 9.070285369168051e-05, "loss": 1.0313, "step": 37960 }, { "epoch": 0.1, "learning_rate": 9.07015954501988e-05, "loss": 1.0285, "step": 37965 }, { "epoch": 0.1, "learning_rate": 9.07003372087171e-05, "loss": 1.0298, "step": 37970 }, { "epoch": 0.1, "learning_rate": 9.06990789672354e-05, "loss": 1.0299, "step": 37975 }, { "epoch": 0.1, "learning_rate": 9.069782072575369e-05, "loss": 1.0311, "step": 37980 }, { "epoch": 0.1, "learning_rate": 9.069656248427198e-05, "loss": 1.0282, "step": 37985 }, { "epoch": 0.1, "learning_rate": 9.069530424279027e-05, "loss": 1.0289, "step": 37990 }, { "epoch": 0.1, "learning_rate": 9.069404600130858e-05, "loss": 1.0289, "step": 37995 }, { "epoch": 0.1, "learning_rate": 9.069278775982687e-05, "loss": 1.0302, "step": 38000 }, { "epoch": 0.1, "learning_rate": 9.069152951834516e-05, "loss": 1.0296, "step": 38005 }, { "epoch": 0.1, "learning_rate": 9.069027127686345e-05, "loss": 1.0301, "step": 38010 }, { "epoch": 0.1, "learning_rate": 9.068901303538176e-05, "loss": 1.0292, "step": 38015 }, { "epoch": 0.1, "learning_rate": 9.068775479390005e-05, "loss": 1.0298, "step": 38020 }, { "epoch": 0.1, "learning_rate": 9.068649655241834e-05, "loss": 1.0281, "step": 38025 }, { "epoch": 0.1, "learning_rate": 9.068523831093663e-05, "loss": 1.0299, "step": 38030 }, { "epoch": 0.1, "learning_rate": 9.068398006945493e-05, "loss": 1.029, "step": 38035 }, { "epoch": 0.1, "learning_rate": 9.068272182797323e-05, "loss": 1.0273, "step": 38040 }, { "epoch": 0.1, "learning_rate": 9.068146358649152e-05, "loss": 1.0286, "step": 38045 }, { "epoch": 0.1, "learning_rate": 9.068020534500981e-05, "loss": 1.0301, "step": 38050 }, { "epoch": 0.1, "learning_rate": 9.06789471035281e-05, "loss": 1.0314, "step": 38055 }, { "epoch": 0.1, "learning_rate": 9.067768886204641e-05, "loss": 1.0296, "step": 38060 }, { "epoch": 0.1, "learning_rate": 9.06764306205647e-05, "loss": 1.0331, "step": 38065 }, { "epoch": 0.1, "learning_rate": 9.0675172379083e-05, "loss": 1.0311, "step": 38070 }, { "epoch": 0.1, "learning_rate": 9.067391413760129e-05, "loss": 1.0314, "step": 38075 }, { "epoch": 0.1, "learning_rate": 9.067265589611959e-05, "loss": 1.0312, "step": 38080 }, { "epoch": 0.1, "learning_rate": 9.067139765463788e-05, "loss": 1.0298, "step": 38085 }, { "epoch": 0.1, "learning_rate": 9.067013941315617e-05, "loss": 1.0323, "step": 38090 }, { "epoch": 0.1, "learning_rate": 9.066888117167447e-05, "loss": 1.029, "step": 38095 }, { "epoch": 0.1, "learning_rate": 9.066762293019276e-05, "loss": 1.03, "step": 38100 }, { "epoch": 0.1, "learning_rate": 9.066636468871106e-05, "loss": 1.0297, "step": 38105 }, { "epoch": 0.1, "learning_rate": 9.066510644722935e-05, "loss": 1.0283, "step": 38110 }, { "epoch": 0.1, "learning_rate": 9.066384820574765e-05, "loss": 1.0292, "step": 38115 }, { "epoch": 0.1, "learning_rate": 9.066258996426594e-05, "loss": 1.0317, "step": 38120 }, { "epoch": 0.1, "learning_rate": 9.066133172278424e-05, "loss": 1.0296, "step": 38125 }, { "epoch": 0.1, "learning_rate": 9.066007348130253e-05, "loss": 1.0307, "step": 38130 }, { "epoch": 0.1, "learning_rate": 9.065881523982083e-05, "loss": 1.0289, "step": 38135 }, { "epoch": 0.1, "learning_rate": 9.065755699833912e-05, "loss": 1.0317, "step": 38140 }, { "epoch": 0.1, "learning_rate": 9.065629875685742e-05, "loss": 1.0265, "step": 38145 }, { "epoch": 0.1, "learning_rate": 9.065504051537571e-05, "loss": 1.0297, "step": 38150 }, { "epoch": 0.1, "learning_rate": 9.0653782273894e-05, "loss": 1.0309, "step": 38155 }, { "epoch": 0.1, "learning_rate": 9.06525240324123e-05, "loss": 1.0278, "step": 38160 }, { "epoch": 0.1, "learning_rate": 9.065126579093059e-05, "loss": 1.0284, "step": 38165 }, { "epoch": 0.1, "learning_rate": 9.06500075494489e-05, "loss": 1.0304, "step": 38170 }, { "epoch": 0.1, "learning_rate": 9.064874930796719e-05, "loss": 1.0311, "step": 38175 }, { "epoch": 0.1, "learning_rate": 9.064749106648548e-05, "loss": 1.0302, "step": 38180 }, { "epoch": 0.1, "learning_rate": 9.064623282500377e-05, "loss": 1.0308, "step": 38185 }, { "epoch": 0.1, "learning_rate": 9.064497458352207e-05, "loss": 1.0293, "step": 38190 }, { "epoch": 0.1, "learning_rate": 9.064371634204037e-05, "loss": 1.0299, "step": 38195 }, { "epoch": 0.1, "learning_rate": 9.064245810055866e-05, "loss": 1.0286, "step": 38200 }, { "epoch": 0.1, "learning_rate": 9.064119985907695e-05, "loss": 1.03, "step": 38205 }, { "epoch": 0.1, "learning_rate": 9.063994161759525e-05, "loss": 1.029, "step": 38210 }, { "epoch": 0.1, "learning_rate": 9.063868337611356e-05, "loss": 1.0284, "step": 38215 }, { "epoch": 0.1, "learning_rate": 9.063742513463185e-05, "loss": 1.0286, "step": 38220 }, { "epoch": 0.1, "learning_rate": 9.063616689315014e-05, "loss": 1.0285, "step": 38225 }, { "epoch": 0.1, "learning_rate": 9.063490865166843e-05, "loss": 1.0303, "step": 38230 }, { "epoch": 0.1, "learning_rate": 9.063365041018674e-05, "loss": 1.0288, "step": 38235 }, { "epoch": 0.1, "learning_rate": 9.063239216870503e-05, "loss": 1.0301, "step": 38240 }, { "epoch": 0.1, "learning_rate": 9.063113392722332e-05, "loss": 1.0299, "step": 38245 }, { "epoch": 0.1, "learning_rate": 9.062987568574161e-05, "loss": 1.031, "step": 38250 }, { "epoch": 0.1, "learning_rate": 9.06286174442599e-05, "loss": 1.0295, "step": 38255 }, { "epoch": 0.1, "learning_rate": 9.062735920277821e-05, "loss": 1.03, "step": 38260 }, { "epoch": 0.1, "learning_rate": 9.06261009612965e-05, "loss": 1.0312, "step": 38265 }, { "epoch": 0.1, "learning_rate": 9.06248427198148e-05, "loss": 1.0289, "step": 38270 }, { "epoch": 0.1, "learning_rate": 9.062358447833309e-05, "loss": 1.0301, "step": 38275 }, { "epoch": 0.1, "learning_rate": 9.062232623685139e-05, "loss": 1.0264, "step": 38280 }, { "epoch": 0.1, "learning_rate": 9.062106799536968e-05, "loss": 1.0442, "step": 38285 }, { "epoch": 0.1, "learning_rate": 9.061980975388797e-05, "loss": 1.029, "step": 38290 }, { "epoch": 0.1, "learning_rate": 9.061855151240627e-05, "loss": 1.0291, "step": 38295 }, { "epoch": 0.1, "learning_rate": 9.061729327092456e-05, "loss": 1.0315, "step": 38300 }, { "epoch": 0.1, "learning_rate": 9.061603502944286e-05, "loss": 1.0297, "step": 38305 }, { "epoch": 0.1, "learning_rate": 9.061477678796115e-05, "loss": 1.0304, "step": 38310 }, { "epoch": 0.1, "learning_rate": 9.061351854647945e-05, "loss": 1.0311, "step": 38315 }, { "epoch": 0.1, "learning_rate": 9.061226030499774e-05, "loss": 1.0278, "step": 38320 }, { "epoch": 0.1, "learning_rate": 9.061100206351604e-05, "loss": 1.0316, "step": 38325 }, { "epoch": 0.1, "learning_rate": 9.060974382203433e-05, "loss": 1.0485, "step": 38330 }, { "epoch": 0.1, "learning_rate": 9.060848558055263e-05, "loss": 1.0293, "step": 38335 }, { "epoch": 0.1, "learning_rate": 9.060722733907092e-05, "loss": 1.0318, "step": 38340 }, { "epoch": 0.1, "learning_rate": 9.060596909758922e-05, "loss": 1.0276, "step": 38345 }, { "epoch": 0.1, "learning_rate": 9.060471085610751e-05, "loss": 1.0291, "step": 38350 }, { "epoch": 0.1, "learning_rate": 9.06034526146258e-05, "loss": 1.0287, "step": 38355 }, { "epoch": 0.1, "learning_rate": 9.06021943731441e-05, "loss": 1.0285, "step": 38360 }, { "epoch": 0.1, "learning_rate": 9.060093613166239e-05, "loss": 1.0298, "step": 38365 }, { "epoch": 0.1, "learning_rate": 9.05996778901807e-05, "loss": 1.029, "step": 38370 }, { "epoch": 0.1, "learning_rate": 9.059841964869899e-05, "loss": 1.0287, "step": 38375 }, { "epoch": 0.1, "learning_rate": 9.059716140721728e-05, "loss": 1.0287, "step": 38380 }, { "epoch": 0.1, "learning_rate": 9.059590316573557e-05, "loss": 1.0273, "step": 38385 }, { "epoch": 0.1, "learning_rate": 9.059464492425387e-05, "loss": 1.0298, "step": 38390 }, { "epoch": 0.1, "learning_rate": 9.059338668277217e-05, "loss": 1.0303, "step": 38395 }, { "epoch": 0.1, "learning_rate": 9.059212844129046e-05, "loss": 1.0304, "step": 38400 }, { "epoch": 0.1, "learning_rate": 9.059087019980875e-05, "loss": 1.0319, "step": 38405 }, { "epoch": 0.1, "learning_rate": 9.058961195832705e-05, "loss": 1.0276, "step": 38410 }, { "epoch": 0.1, "learning_rate": 9.058835371684534e-05, "loss": 1.0317, "step": 38415 }, { "epoch": 0.1, "learning_rate": 9.058709547536364e-05, "loss": 1.0284, "step": 38420 }, { "epoch": 0.1, "learning_rate": 9.058583723388193e-05, "loss": 1.0305, "step": 38425 }, { "epoch": 0.1, "learning_rate": 9.058457899240022e-05, "loss": 1.0293, "step": 38430 }, { "epoch": 0.1, "learning_rate": 9.058332075091852e-05, "loss": 1.0314, "step": 38435 }, { "epoch": 0.1, "learning_rate": 9.058206250943682e-05, "loss": 1.0482, "step": 38440 }, { "epoch": 0.1, "learning_rate": 9.058080426795511e-05, "loss": 1.0297, "step": 38445 }, { "epoch": 0.1, "learning_rate": 9.05795460264734e-05, "loss": 1.0305, "step": 38450 }, { "epoch": 0.1, "learning_rate": 9.05782877849917e-05, "loss": 1.0319, "step": 38455 }, { "epoch": 0.1, "learning_rate": 9.057702954351e-05, "loss": 1.0283, "step": 38460 }, { "epoch": 0.1, "learning_rate": 9.057577130202829e-05, "loss": 1.0286, "step": 38465 }, { "epoch": 0.1, "learning_rate": 9.057451306054658e-05, "loss": 1.0286, "step": 38470 }, { "epoch": 0.1, "learning_rate": 9.057325481906488e-05, "loss": 1.0309, "step": 38475 }, { "epoch": 0.1, "learning_rate": 9.057199657758318e-05, "loss": 1.0297, "step": 38480 }, { "epoch": 0.1, "learning_rate": 9.057073833610147e-05, "loss": 1.0329, "step": 38485 }, { "epoch": 0.1, "learning_rate": 9.056948009461976e-05, "loss": 1.0293, "step": 38490 }, { "epoch": 0.1, "learning_rate": 9.056822185313805e-05, "loss": 1.0284, "step": 38495 }, { "epoch": 0.1, "learning_rate": 9.056696361165636e-05, "loss": 1.0296, "step": 38500 }, { "epoch": 0.1, "learning_rate": 9.056570537017465e-05, "loss": 1.0307, "step": 38505 }, { "epoch": 0.1, "learning_rate": 9.056444712869294e-05, "loss": 1.0298, "step": 38510 }, { "epoch": 0.1, "learning_rate": 9.056318888721123e-05, "loss": 1.0297, "step": 38515 }, { "epoch": 0.1, "learning_rate": 9.056193064572954e-05, "loss": 1.0291, "step": 38520 }, { "epoch": 0.1, "learning_rate": 9.056067240424783e-05, "loss": 1.0282, "step": 38525 }, { "epoch": 0.1, "learning_rate": 9.055941416276612e-05, "loss": 1.031, "step": 38530 }, { "epoch": 0.1, "learning_rate": 9.055815592128441e-05, "loss": 1.0298, "step": 38535 }, { "epoch": 0.1, "learning_rate": 9.055689767980272e-05, "loss": 1.032, "step": 38540 }, { "epoch": 0.1, "learning_rate": 9.055563943832101e-05, "loss": 1.0253, "step": 38545 }, { "epoch": 0.1, "learning_rate": 9.05543811968393e-05, "loss": 1.0281, "step": 38550 }, { "epoch": 0.1, "learning_rate": 9.055312295535759e-05, "loss": 1.0287, "step": 38555 }, { "epoch": 0.1, "learning_rate": 9.055186471387588e-05, "loss": 1.0287, "step": 38560 }, { "epoch": 0.1, "learning_rate": 9.055060647239419e-05, "loss": 1.0304, "step": 38565 }, { "epoch": 0.1, "learning_rate": 9.054934823091248e-05, "loss": 1.0279, "step": 38570 }, { "epoch": 0.1, "learning_rate": 9.054808998943077e-05, "loss": 1.0299, "step": 38575 }, { "epoch": 0.1, "learning_rate": 9.054683174794906e-05, "loss": 1.0283, "step": 38580 }, { "epoch": 0.1, "learning_rate": 9.054557350646737e-05, "loss": 1.0315, "step": 38585 }, { "epoch": 0.1, "learning_rate": 9.054431526498566e-05, "loss": 1.0289, "step": 38590 }, { "epoch": 0.1, "learning_rate": 9.054305702350395e-05, "loss": 1.0299, "step": 38595 }, { "epoch": 0.1, "learning_rate": 9.054179878202224e-05, "loss": 1.0284, "step": 38600 }, { "epoch": 0.1, "learning_rate": 9.054054054054055e-05, "loss": 1.028, "step": 38605 }, { "epoch": 0.1, "learning_rate": 9.053928229905884e-05, "loss": 1.0277, "step": 38610 }, { "epoch": 0.1, "learning_rate": 9.053802405757713e-05, "loss": 1.0297, "step": 38615 }, { "epoch": 0.1, "learning_rate": 9.053676581609542e-05, "loss": 1.0304, "step": 38620 }, { "epoch": 0.1, "learning_rate": 9.053550757461371e-05, "loss": 1.0306, "step": 38625 }, { "epoch": 0.1, "learning_rate": 9.053424933313202e-05, "loss": 1.0288, "step": 38630 }, { "epoch": 0.1, "learning_rate": 9.053299109165031e-05, "loss": 1.0304, "step": 38635 }, { "epoch": 0.1, "learning_rate": 9.05317328501686e-05, "loss": 1.0304, "step": 38640 }, { "epoch": 0.1, "learning_rate": 9.05304746086869e-05, "loss": 1.0311, "step": 38645 }, { "epoch": 0.1, "learning_rate": 9.05292163672052e-05, "loss": 1.0303, "step": 38650 }, { "epoch": 0.1, "learning_rate": 9.052795812572349e-05, "loss": 1.03, "step": 38655 }, { "epoch": 0.1, "learning_rate": 9.052669988424178e-05, "loss": 1.0308, "step": 38660 }, { "epoch": 0.1, "learning_rate": 9.052544164276007e-05, "loss": 1.0268, "step": 38665 }, { "epoch": 0.1, "learning_rate": 9.052418340127838e-05, "loss": 1.0298, "step": 38670 }, { "epoch": 0.1, "learning_rate": 9.052292515979667e-05, "loss": 1.0289, "step": 38675 }, { "epoch": 0.1, "learning_rate": 9.052166691831496e-05, "loss": 1.0285, "step": 38680 }, { "epoch": 0.1, "learning_rate": 9.052040867683325e-05, "loss": 1.0532, "step": 38685 }, { "epoch": 0.1, "learning_rate": 9.051915043535155e-05, "loss": 1.0299, "step": 38690 }, { "epoch": 0.1, "learning_rate": 9.051789219386985e-05, "loss": 1.0294, "step": 38695 }, { "epoch": 0.1, "learning_rate": 9.051663395238814e-05, "loss": 1.0292, "step": 38700 }, { "epoch": 0.1, "learning_rate": 9.051537571090643e-05, "loss": 1.0298, "step": 38705 }, { "epoch": 0.1, "learning_rate": 9.051411746942474e-05, "loss": 1.0301, "step": 38710 }, { "epoch": 0.1, "learning_rate": 9.051285922794303e-05, "loss": 1.0299, "step": 38715 }, { "epoch": 0.1, "learning_rate": 9.051160098646134e-05, "loss": 1.0288, "step": 38720 }, { "epoch": 0.1, "learning_rate": 9.051034274497963e-05, "loss": 1.0511, "step": 38725 }, { "epoch": 0.1, "learning_rate": 9.050908450349792e-05, "loss": 1.0291, "step": 38730 }, { "epoch": 0.1, "learning_rate": 9.050782626201621e-05, "loss": 1.0298, "step": 38735 }, { "epoch": 0.1, "learning_rate": 9.050656802053452e-05, "loss": 1.0289, "step": 38740 }, { "epoch": 0.1, "learning_rate": 9.050530977905281e-05, "loss": 1.0295, "step": 38745 }, { "epoch": 0.1, "learning_rate": 9.05040515375711e-05, "loss": 1.0309, "step": 38750 }, { "epoch": 0.1, "learning_rate": 9.050279329608939e-05, "loss": 1.0305, "step": 38755 }, { "epoch": 0.1, "learning_rate": 9.05015350546077e-05, "loss": 1.03, "step": 38760 }, { "epoch": 0.1, "learning_rate": 9.050027681312599e-05, "loss": 1.0293, "step": 38765 }, { "epoch": 0.1, "learning_rate": 9.049901857164428e-05, "loss": 1.0281, "step": 38770 }, { "epoch": 0.1, "learning_rate": 9.049776033016257e-05, "loss": 1.0289, "step": 38775 }, { "epoch": 0.1, "learning_rate": 9.049650208868086e-05, "loss": 1.0314, "step": 38780 }, { "epoch": 0.1, "learning_rate": 9.049524384719917e-05, "loss": 1.0292, "step": 38785 }, { "epoch": 0.1, "learning_rate": 9.049398560571746e-05, "loss": 1.0302, "step": 38790 }, { "epoch": 0.1, "learning_rate": 9.049272736423575e-05, "loss": 1.0287, "step": 38795 }, { "epoch": 0.1, "learning_rate": 9.049146912275404e-05, "loss": 1.0323, "step": 38800 }, { "epoch": 0.1, "learning_rate": 9.049021088127235e-05, "loss": 1.0288, "step": 38805 }, { "epoch": 0.1, "learning_rate": 9.048895263979064e-05, "loss": 1.0303, "step": 38810 }, { "epoch": 0.1, "learning_rate": 9.048769439830893e-05, "loss": 1.0309, "step": 38815 }, { "epoch": 0.1, "learning_rate": 9.048643615682722e-05, "loss": 1.0293, "step": 38820 }, { "epoch": 0.1, "learning_rate": 9.048517791534553e-05, "loss": 1.0304, "step": 38825 }, { "epoch": 0.1, "learning_rate": 9.048391967386382e-05, "loss": 1.0296, "step": 38830 }, { "epoch": 0.1, "learning_rate": 9.048266143238211e-05, "loss": 1.0308, "step": 38835 }, { "epoch": 0.1, "learning_rate": 9.04814031909004e-05, "loss": 1.0299, "step": 38840 }, { "epoch": 0.1, "learning_rate": 9.04801449494187e-05, "loss": 1.0537, "step": 38845 }, { "epoch": 0.1, "learning_rate": 9.0478886707937e-05, "loss": 1.0311, "step": 38850 }, { "epoch": 0.1, "learning_rate": 9.047762846645529e-05, "loss": 1.0283, "step": 38855 }, { "epoch": 0.1, "learning_rate": 9.047637022497358e-05, "loss": 1.0297, "step": 38860 }, { "epoch": 0.1, "learning_rate": 9.047511198349187e-05, "loss": 1.0307, "step": 38865 }, { "epoch": 0.1, "learning_rate": 9.047385374201018e-05, "loss": 1.0277, "step": 38870 }, { "epoch": 0.1, "learning_rate": 9.047259550052847e-05, "loss": 1.0288, "step": 38875 }, { "epoch": 0.1, "learning_rate": 9.047133725904676e-05, "loss": 1.0308, "step": 38880 }, { "epoch": 0.1, "learning_rate": 9.047007901756505e-05, "loss": 1.0272, "step": 38885 }, { "epoch": 0.1, "learning_rate": 9.046882077608336e-05, "loss": 1.0317, "step": 38890 }, { "epoch": 0.1, "learning_rate": 9.046756253460165e-05, "loss": 1.0303, "step": 38895 }, { "epoch": 0.1, "learning_rate": 9.046630429311994e-05, "loss": 1.0308, "step": 38900 }, { "epoch": 0.1, "learning_rate": 9.046504605163823e-05, "loss": 1.0296, "step": 38905 }, { "epoch": 0.1, "learning_rate": 9.046378781015652e-05, "loss": 1.0282, "step": 38910 }, { "epoch": 0.1, "learning_rate": 9.046252956867483e-05, "loss": 1.0301, "step": 38915 }, { "epoch": 0.1, "learning_rate": 9.046127132719312e-05, "loss": 1.0298, "step": 38920 }, { "epoch": 0.1, "learning_rate": 9.046001308571141e-05, "loss": 1.029, "step": 38925 }, { "epoch": 0.1, "learning_rate": 9.04587548442297e-05, "loss": 1.0296, "step": 38930 }, { "epoch": 0.1, "learning_rate": 9.045749660274801e-05, "loss": 1.0291, "step": 38935 }, { "epoch": 0.1, "learning_rate": 9.04562383612663e-05, "loss": 1.0289, "step": 38940 }, { "epoch": 0.1, "learning_rate": 9.045498011978459e-05, "loss": 1.0291, "step": 38945 }, { "epoch": 0.1, "learning_rate": 9.045372187830288e-05, "loss": 1.0325, "step": 38950 }, { "epoch": 0.1, "learning_rate": 9.045246363682119e-05, "loss": 1.028, "step": 38955 }, { "epoch": 0.1, "learning_rate": 9.045120539533948e-05, "loss": 1.0303, "step": 38960 }, { "epoch": 0.1, "learning_rate": 9.044994715385777e-05, "loss": 1.0292, "step": 38965 }, { "epoch": 0.1, "learning_rate": 9.044868891237606e-05, "loss": 1.0295, "step": 38970 }, { "epoch": 0.1, "learning_rate": 9.044743067089436e-05, "loss": 1.0297, "step": 38975 }, { "epoch": 0.1, "learning_rate": 9.044617242941266e-05, "loss": 1.0303, "step": 38980 }, { "epoch": 0.1, "learning_rate": 9.044491418793095e-05, "loss": 1.0283, "step": 38985 }, { "epoch": 0.1, "learning_rate": 9.044365594644924e-05, "loss": 1.0283, "step": 38990 }, { "epoch": 0.1, "learning_rate": 9.044239770496754e-05, "loss": 1.0302, "step": 38995 }, { "epoch": 0.1, "learning_rate": 9.044113946348584e-05, "loss": 1.0286, "step": 39000 }, { "epoch": 0.1, "learning_rate": 9.043988122200413e-05, "loss": 1.0313, "step": 39005 }, { "epoch": 0.1, "learning_rate": 9.043862298052242e-05, "loss": 1.0286, "step": 39010 }, { "epoch": 0.1, "learning_rate": 9.043736473904072e-05, "loss": 1.0297, "step": 39015 }, { "epoch": 0.1, "learning_rate": 9.043610649755901e-05, "loss": 1.0294, "step": 39020 }, { "epoch": 0.1, "learning_rate": 9.043484825607731e-05, "loss": 1.0283, "step": 39025 }, { "epoch": 0.1, "learning_rate": 9.04335900145956e-05, "loss": 1.0291, "step": 39030 }, { "epoch": 0.1, "learning_rate": 9.04323317731139e-05, "loss": 1.0305, "step": 39035 }, { "epoch": 0.1, "learning_rate": 9.043107353163219e-05, "loss": 1.0282, "step": 39040 }, { "epoch": 0.1, "learning_rate": 9.042981529015049e-05, "loss": 1.0293, "step": 39045 }, { "epoch": 0.1, "learning_rate": 9.042855704866878e-05, "loss": 1.031, "step": 39050 }, { "epoch": 0.1, "learning_rate": 9.042729880718708e-05, "loss": 1.0312, "step": 39055 }, { "epoch": 0.1, "learning_rate": 9.042604056570537e-05, "loss": 1.0278, "step": 39060 }, { "epoch": 0.1, "learning_rate": 9.042478232422367e-05, "loss": 1.0285, "step": 39065 }, { "epoch": 0.1, "learning_rate": 9.042352408274196e-05, "loss": 1.0326, "step": 39070 }, { "epoch": 0.1, "learning_rate": 9.042226584126026e-05, "loss": 1.0296, "step": 39075 }, { "epoch": 0.1, "learning_rate": 9.042100759977855e-05, "loss": 1.0306, "step": 39080 }, { "epoch": 0.1, "learning_rate": 9.041974935829684e-05, "loss": 1.0308, "step": 39085 }, { "epoch": 0.1, "learning_rate": 9.041849111681514e-05, "loss": 1.0321, "step": 39090 }, { "epoch": 0.1, "learning_rate": 9.041723287533344e-05, "loss": 1.0303, "step": 39095 }, { "epoch": 0.1, "learning_rate": 9.041597463385173e-05, "loss": 1.0299, "step": 39100 }, { "epoch": 0.1, "learning_rate": 9.041471639237002e-05, "loss": 1.0292, "step": 39105 }, { "epoch": 0.1, "learning_rate": 9.041345815088832e-05, "loss": 1.0294, "step": 39110 }, { "epoch": 0.1, "learning_rate": 9.041219990940662e-05, "loss": 1.0328, "step": 39115 }, { "epoch": 0.1, "learning_rate": 9.041094166792491e-05, "loss": 1.0291, "step": 39120 }, { "epoch": 0.1, "learning_rate": 9.04096834264432e-05, "loss": 1.0323, "step": 39125 }, { "epoch": 0.1, "learning_rate": 9.04084251849615e-05, "loss": 1.03, "step": 39130 }, { "epoch": 0.1, "learning_rate": 9.04071669434798e-05, "loss": 1.0288, "step": 39135 }, { "epoch": 0.1, "learning_rate": 9.040590870199809e-05, "loss": 1.029, "step": 39140 }, { "epoch": 0.1, "learning_rate": 9.040465046051638e-05, "loss": 1.0296, "step": 39145 }, { "epoch": 0.1, "learning_rate": 9.040339221903467e-05, "loss": 1.0292, "step": 39150 }, { "epoch": 0.1, "learning_rate": 9.040213397755298e-05, "loss": 1.0278, "step": 39155 }, { "epoch": 0.1, "learning_rate": 9.040087573607127e-05, "loss": 1.0304, "step": 39160 }, { "epoch": 0.1, "learning_rate": 9.039961749458956e-05, "loss": 1.0293, "step": 39165 }, { "epoch": 0.1, "learning_rate": 9.039835925310785e-05, "loss": 1.0295, "step": 39170 }, { "epoch": 0.1, "learning_rate": 9.039710101162616e-05, "loss": 1.0295, "step": 39175 }, { "epoch": 0.1, "learning_rate": 9.039584277014445e-05, "loss": 1.0298, "step": 39180 }, { "epoch": 0.1, "learning_rate": 9.039458452866274e-05, "loss": 1.0282, "step": 39185 }, { "epoch": 0.1, "learning_rate": 9.039332628718103e-05, "loss": 1.0309, "step": 39190 }, { "epoch": 0.1, "learning_rate": 9.039206804569934e-05, "loss": 1.0293, "step": 39195 }, { "epoch": 0.1, "learning_rate": 9.039080980421763e-05, "loss": 1.0295, "step": 39200 }, { "epoch": 0.1, "learning_rate": 9.038955156273592e-05, "loss": 1.0274, "step": 39205 }, { "epoch": 0.1, "learning_rate": 9.038829332125422e-05, "loss": 1.0527, "step": 39210 }, { "epoch": 0.1, "learning_rate": 9.038703507977252e-05, "loss": 1.0311, "step": 39215 }, { "epoch": 0.1, "learning_rate": 9.038577683829082e-05, "loss": 1.0309, "step": 39220 }, { "epoch": 0.1, "learning_rate": 9.038451859680911e-05, "loss": 1.0298, "step": 39225 }, { "epoch": 0.1, "learning_rate": 9.03832603553274e-05, "loss": 1.0291, "step": 39230 }, { "epoch": 0.1, "learning_rate": 9.03820021138457e-05, "loss": 1.0295, "step": 39235 }, { "epoch": 0.1, "learning_rate": 9.038074387236399e-05, "loss": 1.0314, "step": 39240 }, { "epoch": 0.1, "learning_rate": 9.037948563088229e-05, "loss": 1.0286, "step": 39245 }, { "epoch": 0.1, "learning_rate": 9.037822738940058e-05, "loss": 1.0303, "step": 39250 }, { "epoch": 0.1, "learning_rate": 9.037696914791888e-05, "loss": 1.0306, "step": 39255 }, { "epoch": 0.1, "learning_rate": 9.037571090643717e-05, "loss": 1.0348, "step": 39260 }, { "epoch": 0.1, "learning_rate": 9.037445266495547e-05, "loss": 1.0285, "step": 39265 }, { "epoch": 0.1, "learning_rate": 9.037319442347376e-05, "loss": 1.0287, "step": 39270 }, { "epoch": 0.1, "learning_rate": 9.037193618199206e-05, "loss": 1.0303, "step": 39275 }, { "epoch": 0.1, "learning_rate": 9.037067794051035e-05, "loss": 1.0317, "step": 39280 }, { "epoch": 0.1, "learning_rate": 9.036941969902865e-05, "loss": 1.0307, "step": 39285 }, { "epoch": 0.1, "learning_rate": 9.036816145754694e-05, "loss": 1.0303, "step": 39290 }, { "epoch": 0.1, "learning_rate": 9.036690321606524e-05, "loss": 1.0293, "step": 39295 }, { "epoch": 0.1, "learning_rate": 9.036564497458353e-05, "loss": 1.0283, "step": 39300 }, { "epoch": 0.1, "learning_rate": 9.036438673310182e-05, "loss": 1.0294, "step": 39305 }, { "epoch": 0.1, "learning_rate": 9.036312849162012e-05, "loss": 1.0309, "step": 39310 }, { "epoch": 0.1, "learning_rate": 9.036187025013842e-05, "loss": 1.0277, "step": 39315 }, { "epoch": 0.1, "learning_rate": 9.036061200865671e-05, "loss": 1.0288, "step": 39320 }, { "epoch": 0.1, "learning_rate": 9.0359353767175e-05, "loss": 1.0321, "step": 39325 }, { "epoch": 0.1, "learning_rate": 9.03580955256933e-05, "loss": 1.0322, "step": 39330 }, { "epoch": 0.1, "learning_rate": 9.03568372842116e-05, "loss": 1.0288, "step": 39335 }, { "epoch": 0.1, "learning_rate": 9.035557904272989e-05, "loss": 1.0295, "step": 39340 }, { "epoch": 0.1, "learning_rate": 9.035432080124818e-05, "loss": 1.0278, "step": 39345 }, { "epoch": 0.1, "learning_rate": 9.035306255976648e-05, "loss": 1.0291, "step": 39350 }, { "epoch": 0.1, "learning_rate": 9.035180431828478e-05, "loss": 1.0324, "step": 39355 }, { "epoch": 0.1, "learning_rate": 9.035054607680307e-05, "loss": 1.0308, "step": 39360 }, { "epoch": 0.1, "learning_rate": 9.034928783532136e-05, "loss": 1.0283, "step": 39365 }, { "epoch": 0.1, "learning_rate": 9.034802959383965e-05, "loss": 1.0314, "step": 39370 }, { "epoch": 0.1, "learning_rate": 9.034677135235796e-05, "loss": 1.0305, "step": 39375 }, { "epoch": 0.1, "learning_rate": 9.034551311087625e-05, "loss": 1.0286, "step": 39380 }, { "epoch": 0.1, "learning_rate": 9.034425486939454e-05, "loss": 1.0302, "step": 39385 }, { "epoch": 0.1, "learning_rate": 9.034299662791283e-05, "loss": 1.0298, "step": 39390 }, { "epoch": 0.1, "learning_rate": 9.034173838643114e-05, "loss": 1.029, "step": 39395 }, { "epoch": 0.1, "learning_rate": 9.034048014494943e-05, "loss": 1.027, "step": 39400 }, { "epoch": 0.1, "learning_rate": 9.033922190346772e-05, "loss": 1.0284, "step": 39405 }, { "epoch": 0.1, "learning_rate": 9.033796366198601e-05, "loss": 1.0309, "step": 39410 }, { "epoch": 0.1, "learning_rate": 9.033670542050431e-05, "loss": 1.0289, "step": 39415 }, { "epoch": 0.1, "learning_rate": 9.03354471790226e-05, "loss": 1.0328, "step": 39420 }, { "epoch": 0.1, "learning_rate": 9.03341889375409e-05, "loss": 1.0285, "step": 39425 }, { "epoch": 0.1, "learning_rate": 9.033293069605919e-05, "loss": 1.0295, "step": 39430 }, { "epoch": 0.1, "learning_rate": 9.033167245457748e-05, "loss": 1.0301, "step": 39435 }, { "epoch": 0.1, "learning_rate": 9.033041421309579e-05, "loss": 1.0491, "step": 39440 }, { "epoch": 0.1, "learning_rate": 9.032915597161408e-05, "loss": 1.0296, "step": 39445 }, { "epoch": 0.1, "learning_rate": 9.032789773013237e-05, "loss": 1.0297, "step": 39450 }, { "epoch": 0.1, "learning_rate": 9.032663948865066e-05, "loss": 1.0286, "step": 39455 }, { "epoch": 0.1, "learning_rate": 9.032538124716897e-05, "loss": 1.0296, "step": 39460 }, { "epoch": 0.1, "learning_rate": 9.032412300568726e-05, "loss": 1.0303, "step": 39465 }, { "epoch": 0.1, "learning_rate": 9.032286476420555e-05, "loss": 1.0287, "step": 39470 }, { "epoch": 0.1, "learning_rate": 9.032160652272384e-05, "loss": 1.0299, "step": 39475 }, { "epoch": 0.1, "learning_rate": 9.032034828124215e-05, "loss": 1.0315, "step": 39480 }, { "epoch": 0.1, "learning_rate": 9.031909003976044e-05, "loss": 1.0303, "step": 39485 }, { "epoch": 0.1, "learning_rate": 9.031783179827873e-05, "loss": 1.0284, "step": 39490 }, { "epoch": 0.1, "learning_rate": 9.031657355679702e-05, "loss": 1.0298, "step": 39495 }, { "epoch": 0.1, "learning_rate": 9.031531531531531e-05, "loss": 1.0301, "step": 39500 }, { "epoch": 0.1, "learning_rate": 9.031405707383362e-05, "loss": 1.0298, "step": 39505 }, { "epoch": 0.1, "learning_rate": 9.031279883235191e-05, "loss": 1.0303, "step": 39510 }, { "epoch": 0.1, "learning_rate": 9.03115405908702e-05, "loss": 1.0297, "step": 39515 }, { "epoch": 0.1, "learning_rate": 9.031028234938849e-05, "loss": 1.029, "step": 39520 }, { "epoch": 0.1, "learning_rate": 9.03090241079068e-05, "loss": 1.0293, "step": 39525 }, { "epoch": 0.1, "learning_rate": 9.030776586642509e-05, "loss": 1.0291, "step": 39530 }, { "epoch": 0.1, "learning_rate": 9.030650762494338e-05, "loss": 1.0291, "step": 39535 }, { "epoch": 0.1, "learning_rate": 9.030524938346167e-05, "loss": 1.0298, "step": 39540 }, { "epoch": 0.1, "learning_rate": 9.030399114197998e-05, "loss": 1.0299, "step": 39545 }, { "epoch": 0.1, "learning_rate": 9.030273290049827e-05, "loss": 1.0305, "step": 39550 }, { "epoch": 0.1, "learning_rate": 9.030147465901656e-05, "loss": 1.0274, "step": 39555 }, { "epoch": 0.1, "learning_rate": 9.030021641753485e-05, "loss": 1.0289, "step": 39560 }, { "epoch": 0.1, "learning_rate": 9.029895817605314e-05, "loss": 1.03, "step": 39565 }, { "epoch": 0.1, "learning_rate": 9.029769993457145e-05, "loss": 1.0295, "step": 39570 }, { "epoch": 0.1, "learning_rate": 9.029644169308974e-05, "loss": 1.0293, "step": 39575 }, { "epoch": 0.1, "learning_rate": 9.029518345160803e-05, "loss": 1.0296, "step": 39580 }, { "epoch": 0.1, "learning_rate": 9.029392521012632e-05, "loss": 1.0297, "step": 39585 }, { "epoch": 0.1, "learning_rate": 9.029266696864463e-05, "loss": 1.0319, "step": 39590 }, { "epoch": 0.1, "learning_rate": 9.029140872716292e-05, "loss": 1.0286, "step": 39595 }, { "epoch": 0.1, "learning_rate": 9.029015048568121e-05, "loss": 1.0288, "step": 39600 }, { "epoch": 0.1, "learning_rate": 9.02888922441995e-05, "loss": 1.0295, "step": 39605 }, { "epoch": 0.1, "learning_rate": 9.028763400271781e-05, "loss": 1.0299, "step": 39610 }, { "epoch": 0.1, "learning_rate": 9.02863757612361e-05, "loss": 1.0306, "step": 39615 }, { "epoch": 0.1, "learning_rate": 9.028511751975439e-05, "loss": 1.0324, "step": 39620 }, { "epoch": 0.1, "learning_rate": 9.028385927827268e-05, "loss": 1.0302, "step": 39625 }, { "epoch": 0.1, "learning_rate": 9.028260103679098e-05, "loss": 1.0293, "step": 39630 }, { "epoch": 0.1, "learning_rate": 9.028134279530928e-05, "loss": 1.0292, "step": 39635 }, { "epoch": 0.1, "learning_rate": 9.028008455382757e-05, "loss": 1.0308, "step": 39640 }, { "epoch": 0.1, "learning_rate": 9.027882631234586e-05, "loss": 1.0271, "step": 39645 }, { "epoch": 0.1, "learning_rate": 9.027756807086416e-05, "loss": 1.0309, "step": 39650 }, { "epoch": 0.1, "learning_rate": 9.027630982938246e-05, "loss": 1.0304, "step": 39655 }, { "epoch": 0.1, "learning_rate": 9.027505158790075e-05, "loss": 1.03, "step": 39660 }, { "epoch": 0.1, "learning_rate": 9.027379334641904e-05, "loss": 1.0291, "step": 39665 }, { "epoch": 0.1, "learning_rate": 9.027253510493734e-05, "loss": 1.029, "step": 39670 }, { "epoch": 0.1, "learning_rate": 9.027127686345563e-05, "loss": 1.0325, "step": 39675 }, { "epoch": 0.1, "learning_rate": 9.027001862197393e-05, "loss": 1.0274, "step": 39680 }, { "epoch": 0.1, "learning_rate": 9.026876038049222e-05, "loss": 1.0295, "step": 39685 }, { "epoch": 0.1, "learning_rate": 9.026750213901052e-05, "loss": 1.0297, "step": 39690 }, { "epoch": 0.1, "learning_rate": 9.026624389752881e-05, "loss": 1.0314, "step": 39695 }, { "epoch": 0.1, "learning_rate": 9.026498565604711e-05, "loss": 1.0282, "step": 39700 }, { "epoch": 0.1, "learning_rate": 9.026372741456542e-05, "loss": 1.0323, "step": 39705 }, { "epoch": 0.1, "learning_rate": 9.026246917308371e-05, "loss": 1.0281, "step": 39710 }, { "epoch": 0.1, "learning_rate": 9.0261210931602e-05, "loss": 1.0295, "step": 39715 }, { "epoch": 0.1, "learning_rate": 9.025995269012029e-05, "loss": 1.0488, "step": 39720 }, { "epoch": 0.1, "learning_rate": 9.02586944486386e-05, "loss": 1.0283, "step": 39725 }, { "epoch": 0.1, "learning_rate": 9.025743620715689e-05, "loss": 1.0294, "step": 39730 }, { "epoch": 0.1, "learning_rate": 9.025617796567518e-05, "loss": 1.0279, "step": 39735 }, { "epoch": 0.1, "learning_rate": 9.025491972419347e-05, "loss": 1.0292, "step": 39740 }, { "epoch": 0.1, "learning_rate": 9.025366148271178e-05, "loss": 1.0527, "step": 39745 }, { "epoch": 0.1, "learning_rate": 9.025240324123007e-05, "loss": 1.0284, "step": 39750 }, { "epoch": 0.1, "learning_rate": 9.025114499974836e-05, "loss": 1.0277, "step": 39755 }, { "epoch": 0.1, "learning_rate": 9.024988675826665e-05, "loss": 1.0297, "step": 39760 }, { "epoch": 0.1, "learning_rate": 9.024862851678494e-05, "loss": 1.0285, "step": 39765 }, { "epoch": 0.1, "learning_rate": 9.024737027530325e-05, "loss": 1.0301, "step": 39770 }, { "epoch": 0.1, "learning_rate": 9.024611203382154e-05, "loss": 1.0298, "step": 39775 }, { "epoch": 0.1, "learning_rate": 9.024485379233983e-05, "loss": 1.0301, "step": 39780 }, { "epoch": 0.1, "learning_rate": 9.024359555085812e-05, "loss": 1.03, "step": 39785 }, { "epoch": 0.1, "learning_rate": 9.024233730937643e-05, "loss": 1.0315, "step": 39790 }, { "epoch": 0.1, "learning_rate": 9.024107906789472e-05, "loss": 1.0305, "step": 39795 }, { "epoch": 0.1, "learning_rate": 9.023982082641301e-05, "loss": 1.0291, "step": 39800 }, { "epoch": 0.1, "learning_rate": 9.02385625849313e-05, "loss": 1.0291, "step": 39805 }, { "epoch": 0.1, "learning_rate": 9.023730434344961e-05, "loss": 1.0292, "step": 39810 }, { "epoch": 0.1, "learning_rate": 9.02360461019679e-05, "loss": 1.0526, "step": 39815 }, { "epoch": 0.1, "learning_rate": 9.023478786048619e-05, "loss": 1.0281, "step": 39820 }, { "epoch": 0.1, "learning_rate": 9.023352961900448e-05, "loss": 1.0366, "step": 39825 }, { "epoch": 0.1, "learning_rate": 9.023227137752278e-05, "loss": 1.0307, "step": 39830 }, { "epoch": 0.1, "learning_rate": 9.023101313604108e-05, "loss": 1.029, "step": 39835 }, { "epoch": 0.1, "learning_rate": 9.022975489455937e-05, "loss": 1.0319, "step": 39840 }, { "epoch": 0.1, "learning_rate": 9.022849665307766e-05, "loss": 1.028, "step": 39845 }, { "epoch": 0.1, "learning_rate": 9.022723841159595e-05, "loss": 1.0264, "step": 39850 }, { "epoch": 0.1, "learning_rate": 9.022598017011426e-05, "loss": 1.0306, "step": 39855 }, { "epoch": 0.1, "learning_rate": 9.022472192863255e-05, "loss": 1.0272, "step": 39860 }, { "epoch": 0.1, "learning_rate": 9.022346368715084e-05, "loss": 1.0276, "step": 39865 }, { "epoch": 0.1, "learning_rate": 9.022220544566913e-05, "loss": 1.0273, "step": 39870 }, { "epoch": 0.1, "learning_rate": 9.022094720418744e-05, "loss": 1.0288, "step": 39875 }, { "epoch": 0.1, "learning_rate": 9.021968896270573e-05, "loss": 1.0282, "step": 39880 }, { "epoch": 0.1, "learning_rate": 9.021843072122402e-05, "loss": 1.0305, "step": 39885 }, { "epoch": 0.1, "learning_rate": 9.021717247974231e-05, "loss": 1.0313, "step": 39890 }, { "epoch": 0.1, "learning_rate": 9.02159142382606e-05, "loss": 1.0305, "step": 39895 }, { "epoch": 0.1, "learning_rate": 9.021465599677891e-05, "loss": 1.0299, "step": 39900 }, { "epoch": 0.1, "learning_rate": 9.02133977552972e-05, "loss": 1.0299, "step": 39905 }, { "epoch": 0.1, "learning_rate": 9.02121395138155e-05, "loss": 1.0288, "step": 39910 }, { "epoch": 0.1, "learning_rate": 9.021088127233379e-05, "loss": 1.0307, "step": 39915 }, { "epoch": 0.1, "learning_rate": 9.020962303085209e-05, "loss": 1.03, "step": 39920 }, { "epoch": 0.1, "learning_rate": 9.020836478937038e-05, "loss": 1.0272, "step": 39925 }, { "epoch": 0.1, "learning_rate": 9.020710654788867e-05, "loss": 1.0306, "step": 39930 }, { "epoch": 0.1, "learning_rate": 9.020584830640697e-05, "loss": 1.0319, "step": 39935 }, { "epoch": 0.1, "learning_rate": 9.020459006492527e-05, "loss": 1.0297, "step": 39940 }, { "epoch": 0.1, "learning_rate": 9.020333182344356e-05, "loss": 1.0292, "step": 39945 }, { "epoch": 0.1, "learning_rate": 9.020207358196185e-05, "loss": 1.0284, "step": 39950 }, { "epoch": 0.1, "learning_rate": 9.020081534048015e-05, "loss": 1.0304, "step": 39955 }, { "epoch": 0.1, "learning_rate": 9.019955709899844e-05, "loss": 1.0312, "step": 39960 }, { "epoch": 0.1, "learning_rate": 9.019829885751674e-05, "loss": 1.0285, "step": 39965 }, { "epoch": 0.1, "learning_rate": 9.019704061603503e-05, "loss": 1.0313, "step": 39970 }, { "epoch": 0.1, "learning_rate": 9.019578237455333e-05, "loss": 1.03, "step": 39975 }, { "epoch": 0.1, "learning_rate": 9.019452413307162e-05, "loss": 1.0292, "step": 39980 }, { "epoch": 0.1, "learning_rate": 9.019326589158992e-05, "loss": 1.0289, "step": 39985 }, { "epoch": 0.1, "learning_rate": 9.019200765010821e-05, "loss": 1.0281, "step": 39990 }, { "epoch": 0.1, "learning_rate": 9.01907494086265e-05, "loss": 1.0302, "step": 39995 }, { "epoch": 0.1, "learning_rate": 9.01894911671448e-05, "loss": 1.0297, "step": 40000 }, { "epoch": 0.1, "learning_rate": 9.01882329256631e-05, "loss": 1.0302, "step": 40005 }, { "epoch": 0.1, "learning_rate": 9.01869746841814e-05, "loss": 1.0301, "step": 40010 }, { "epoch": 0.1, "learning_rate": 9.018571644269969e-05, "loss": 1.0272, "step": 40015 }, { "epoch": 0.1, "learning_rate": 9.018445820121798e-05, "loss": 1.0307, "step": 40020 }, { "epoch": 0.1, "learning_rate": 9.018319995973627e-05, "loss": 1.0301, "step": 40025 }, { "epoch": 0.1, "learning_rate": 9.018194171825457e-05, "loss": 1.0322, "step": 40030 }, { "epoch": 0.1, "learning_rate": 9.018068347677287e-05, "loss": 1.0261, "step": 40035 }, { "epoch": 0.1, "learning_rate": 9.017942523529116e-05, "loss": 1.0291, "step": 40040 }, { "epoch": 0.1, "learning_rate": 9.017816699380945e-05, "loss": 1.0286, "step": 40045 }, { "epoch": 0.1, "learning_rate": 9.017690875232775e-05, "loss": 1.0288, "step": 40050 }, { "epoch": 0.1, "learning_rate": 9.017565051084605e-05, "loss": 1.0303, "step": 40055 }, { "epoch": 0.1, "learning_rate": 9.017439226936434e-05, "loss": 1.0296, "step": 40060 }, { "epoch": 0.1, "learning_rate": 9.017313402788263e-05, "loss": 1.0304, "step": 40065 }, { "epoch": 0.1, "learning_rate": 9.017187578640093e-05, "loss": 1.0309, "step": 40070 }, { "epoch": 0.1, "learning_rate": 9.017061754491923e-05, "loss": 1.03, "step": 40075 }, { "epoch": 0.1, "learning_rate": 9.016935930343752e-05, "loss": 1.0278, "step": 40080 }, { "epoch": 0.1, "learning_rate": 9.016810106195581e-05, "loss": 1.0272, "step": 40085 }, { "epoch": 0.1, "learning_rate": 9.01668428204741e-05, "loss": 1.0292, "step": 40090 }, { "epoch": 0.1, "learning_rate": 9.01655845789924e-05, "loss": 1.0297, "step": 40095 }, { "epoch": 0.1, "learning_rate": 9.01643263375107e-05, "loss": 1.0297, "step": 40100 }, { "epoch": 0.1, "learning_rate": 9.016306809602899e-05, "loss": 1.0309, "step": 40105 }, { "epoch": 0.1, "learning_rate": 9.016180985454728e-05, "loss": 1.032, "step": 40110 }, { "epoch": 0.1, "learning_rate": 9.016055161306559e-05, "loss": 1.0285, "step": 40115 }, { "epoch": 0.1, "learning_rate": 9.015929337158388e-05, "loss": 1.0305, "step": 40120 }, { "epoch": 0.1, "learning_rate": 9.015803513010217e-05, "loss": 1.0316, "step": 40125 }, { "epoch": 0.1, "learning_rate": 9.015677688862046e-05, "loss": 1.0298, "step": 40130 }, { "epoch": 0.1, "learning_rate": 9.015551864713877e-05, "loss": 1.0284, "step": 40135 }, { "epoch": 0.1, "learning_rate": 9.015426040565706e-05, "loss": 1.031, "step": 40140 }, { "epoch": 0.1, "learning_rate": 9.015300216417535e-05, "loss": 1.0272, "step": 40145 }, { "epoch": 0.1, "learning_rate": 9.015174392269364e-05, "loss": 1.0291, "step": 40150 }, { "epoch": 0.1, "learning_rate": 9.015048568121193e-05, "loss": 1.0294, "step": 40155 }, { "epoch": 0.1, "learning_rate": 9.014922743973024e-05, "loss": 1.0315, "step": 40160 }, { "epoch": 0.1, "learning_rate": 9.014796919824853e-05, "loss": 1.0285, "step": 40165 }, { "epoch": 0.1, "learning_rate": 9.014671095676682e-05, "loss": 1.0298, "step": 40170 }, { "epoch": 0.1, "learning_rate": 9.014545271528511e-05, "loss": 1.0292, "step": 40175 }, { "epoch": 0.1, "learning_rate": 9.014419447380342e-05, "loss": 1.0288, "step": 40180 }, { "epoch": 0.1, "learning_rate": 9.014293623232171e-05, "loss": 1.0283, "step": 40185 }, { "epoch": 0.1, "learning_rate": 9.014167799084e-05, "loss": 1.0296, "step": 40190 }, { "epoch": 0.1, "learning_rate": 9.014041974935829e-05, "loss": 1.0276, "step": 40195 }, { "epoch": 0.1, "learning_rate": 9.01391615078766e-05, "loss": 1.0312, "step": 40200 }, { "epoch": 0.1, "learning_rate": 9.01379032663949e-05, "loss": 1.0273, "step": 40205 }, { "epoch": 0.1, "learning_rate": 9.01366450249132e-05, "loss": 1.0286, "step": 40210 }, { "epoch": 0.1, "learning_rate": 9.013538678343149e-05, "loss": 1.0271, "step": 40215 }, { "epoch": 0.1, "learning_rate": 9.013412854194978e-05, "loss": 1.0303, "step": 40220 }, { "epoch": 0.1, "learning_rate": 9.013287030046807e-05, "loss": 1.0282, "step": 40225 }, { "epoch": 0.1, "learning_rate": 9.013161205898637e-05, "loss": 1.0302, "step": 40230 }, { "epoch": 0.1, "learning_rate": 9.013035381750467e-05, "loss": 1.0326, "step": 40235 }, { "epoch": 0.1, "learning_rate": 9.012909557602296e-05, "loss": 1.0289, "step": 40240 }, { "epoch": 0.1, "learning_rate": 9.012783733454125e-05, "loss": 1.0298, "step": 40245 }, { "epoch": 0.1, "learning_rate": 9.012657909305955e-05, "loss": 1.0285, "step": 40250 }, { "epoch": 0.1, "learning_rate": 9.012532085157785e-05, "loss": 1.0497, "step": 40255 }, { "epoch": 0.1, "learning_rate": 9.012406261009614e-05, "loss": 1.031, "step": 40260 }, { "epoch": 0.1, "learning_rate": 9.012280436861443e-05, "loss": 1.032, "step": 40265 }, { "epoch": 0.1, "learning_rate": 9.012154612713273e-05, "loss": 1.0299, "step": 40270 }, { "epoch": 0.1, "learning_rate": 9.012028788565103e-05, "loss": 1.0305, "step": 40275 }, { "epoch": 0.1, "learning_rate": 9.011902964416932e-05, "loss": 1.0296, "step": 40280 }, { "epoch": 0.1, "learning_rate": 9.011777140268761e-05, "loss": 1.0285, "step": 40285 }, { "epoch": 0.1, "learning_rate": 9.01165131612059e-05, "loss": 1.0289, "step": 40290 }, { "epoch": 0.1, "learning_rate": 9.01152549197242e-05, "loss": 1.0298, "step": 40295 }, { "epoch": 0.1, "learning_rate": 9.01139966782425e-05, "loss": 1.0295, "step": 40300 }, { "epoch": 0.1, "learning_rate": 9.011273843676079e-05, "loss": 1.0306, "step": 40305 }, { "epoch": 0.1, "learning_rate": 9.011148019527908e-05, "loss": 1.0291, "step": 40310 }, { "epoch": 0.1, "learning_rate": 9.011022195379739e-05, "loss": 1.0287, "step": 40315 }, { "epoch": 0.1, "learning_rate": 9.010896371231568e-05, "loss": 1.0289, "step": 40320 }, { "epoch": 0.1, "learning_rate": 9.010770547083397e-05, "loss": 1.0283, "step": 40325 }, { "epoch": 0.1, "learning_rate": 9.010644722935226e-05, "loss": 1.029, "step": 40330 }, { "epoch": 0.1, "learning_rate": 9.010518898787057e-05, "loss": 1.0293, "step": 40335 }, { "epoch": 0.1, "learning_rate": 9.010393074638886e-05, "loss": 1.0277, "step": 40340 }, { "epoch": 0.1, "learning_rate": 9.010267250490715e-05, "loss": 1.0286, "step": 40345 }, { "epoch": 0.1, "learning_rate": 9.010141426342544e-05, "loss": 1.0312, "step": 40350 }, { "epoch": 0.1, "learning_rate": 9.010015602194373e-05, "loss": 1.0278, "step": 40355 }, { "epoch": 0.1, "learning_rate": 9.009889778046204e-05, "loss": 1.0286, "step": 40360 }, { "epoch": 0.1, "learning_rate": 9.009763953898033e-05, "loss": 1.0302, "step": 40365 }, { "epoch": 0.1, "learning_rate": 9.009638129749862e-05, "loss": 1.0296, "step": 40370 }, { "epoch": 0.1, "learning_rate": 9.009512305601691e-05, "loss": 1.0272, "step": 40375 }, { "epoch": 0.1, "learning_rate": 9.009386481453522e-05, "loss": 1.0295, "step": 40380 }, { "epoch": 0.1, "learning_rate": 9.009260657305351e-05, "loss": 1.0297, "step": 40385 }, { "epoch": 0.1, "learning_rate": 9.00913483315718e-05, "loss": 1.0309, "step": 40390 }, { "epoch": 0.1, "learning_rate": 9.009009009009009e-05, "loss": 1.0315, "step": 40395 }, { "epoch": 0.1, "learning_rate": 9.00888318486084e-05, "loss": 1.0277, "step": 40400 }, { "epoch": 0.1, "learning_rate": 9.008757360712669e-05, "loss": 1.0296, "step": 40405 }, { "epoch": 0.1, "learning_rate": 9.008631536564498e-05, "loss": 1.0308, "step": 40410 }, { "epoch": 0.1, "learning_rate": 9.008505712416327e-05, "loss": 1.0274, "step": 40415 }, { "epoch": 0.1, "learning_rate": 9.008379888268156e-05, "loss": 1.0319, "step": 40420 }, { "epoch": 0.1, "learning_rate": 9.008254064119987e-05, "loss": 1.0306, "step": 40425 }, { "epoch": 0.1, "learning_rate": 9.008128239971816e-05, "loss": 1.0501, "step": 40430 }, { "epoch": 0.1, "learning_rate": 9.008002415823645e-05, "loss": 1.0305, "step": 40435 }, { "epoch": 0.1, "learning_rate": 9.007876591675474e-05, "loss": 1.0289, "step": 40440 }, { "epoch": 0.1, "learning_rate": 9.007750767527305e-05, "loss": 1.0295, "step": 40445 }, { "epoch": 0.1, "learning_rate": 9.007624943379134e-05, "loss": 1.0531, "step": 40450 }, { "epoch": 0.1, "learning_rate": 9.007499119230963e-05, "loss": 1.0295, "step": 40455 }, { "epoch": 0.1, "learning_rate": 9.007373295082792e-05, "loss": 1.0291, "step": 40460 }, { "epoch": 0.1, "learning_rate": 9.007247470934623e-05, "loss": 1.0319, "step": 40465 }, { "epoch": 0.1, "learning_rate": 9.007121646786452e-05, "loss": 1.0298, "step": 40470 }, { "epoch": 0.1, "learning_rate": 9.006995822638281e-05, "loss": 1.0273, "step": 40475 }, { "epoch": 0.1, "learning_rate": 9.00686999849011e-05, "loss": 1.0301, "step": 40480 }, { "epoch": 0.1, "learning_rate": 9.00674417434194e-05, "loss": 1.0326, "step": 40485 }, { "epoch": 0.1, "learning_rate": 9.00661835019377e-05, "loss": 1.029, "step": 40490 }, { "epoch": 0.1, "learning_rate": 9.006492526045599e-05, "loss": 1.0305, "step": 40495 }, { "epoch": 0.1, "learning_rate": 9.006366701897428e-05, "loss": 1.0295, "step": 40500 }, { "epoch": 0.1, "learning_rate": 9.006240877749257e-05, "loss": 1.0267, "step": 40505 }, { "epoch": 0.1, "learning_rate": 9.006115053601088e-05, "loss": 1.0294, "step": 40510 }, { "epoch": 0.1, "learning_rate": 9.005989229452917e-05, "loss": 1.0297, "step": 40515 }, { "epoch": 0.1, "learning_rate": 9.005863405304746e-05, "loss": 1.0284, "step": 40520 }, { "epoch": 0.1, "learning_rate": 9.005737581156575e-05, "loss": 1.03, "step": 40525 }, { "epoch": 0.1, "learning_rate": 9.005611757008406e-05, "loss": 1.0496, "step": 40530 }, { "epoch": 0.1, "learning_rate": 9.005485932860235e-05, "loss": 1.0283, "step": 40535 }, { "epoch": 0.1, "learning_rate": 9.005360108712064e-05, "loss": 1.0263, "step": 40540 }, { "epoch": 0.1, "learning_rate": 9.005234284563893e-05, "loss": 1.0283, "step": 40545 }, { "epoch": 0.1, "learning_rate": 9.005108460415723e-05, "loss": 1.0299, "step": 40550 }, { "epoch": 0.1, "learning_rate": 9.004982636267553e-05, "loss": 1.0282, "step": 40555 }, { "epoch": 0.1, "learning_rate": 9.004856812119382e-05, "loss": 1.0525, "step": 40560 }, { "epoch": 0.1, "learning_rate": 9.004730987971211e-05, "loss": 1.0285, "step": 40565 }, { "epoch": 0.1, "learning_rate": 9.00460516382304e-05, "loss": 1.0273, "step": 40570 }, { "epoch": 0.1, "learning_rate": 9.004479339674871e-05, "loss": 1.0284, "step": 40575 }, { "epoch": 0.1, "learning_rate": 9.0043535155267e-05, "loss": 1.0307, "step": 40580 }, { "epoch": 0.1, "learning_rate": 9.00422769137853e-05, "loss": 1.0295, "step": 40585 }, { "epoch": 0.1, "learning_rate": 9.004101867230359e-05, "loss": 1.0304, "step": 40590 }, { "epoch": 0.1, "learning_rate": 9.003976043082189e-05, "loss": 1.0292, "step": 40595 }, { "epoch": 0.1, "learning_rate": 9.003850218934018e-05, "loss": 1.0292, "step": 40600 }, { "epoch": 0.1, "learning_rate": 9.003724394785847e-05, "loss": 1.0288, "step": 40605 }, { "epoch": 0.1, "learning_rate": 9.003598570637677e-05, "loss": 1.0296, "step": 40610 }, { "epoch": 0.1, "learning_rate": 9.003472746489506e-05, "loss": 1.0297, "step": 40615 }, { "epoch": 0.1, "learning_rate": 9.003346922341336e-05, "loss": 1.0284, "step": 40620 }, { "epoch": 0.1, "learning_rate": 9.003221098193165e-05, "loss": 1.0296, "step": 40625 }, { "epoch": 0.1, "learning_rate": 9.003095274044995e-05, "loss": 1.0304, "step": 40630 }, { "epoch": 0.1, "learning_rate": 9.002969449896824e-05, "loss": 1.0288, "step": 40635 }, { "epoch": 0.1, "learning_rate": 9.002843625748654e-05, "loss": 1.0275, "step": 40640 }, { "epoch": 0.1, "learning_rate": 9.002717801600483e-05, "loss": 1.0291, "step": 40645 }, { "epoch": 0.1, "learning_rate": 9.002591977452313e-05, "loss": 1.0282, "step": 40650 }, { "epoch": 0.1, "learning_rate": 9.002466153304142e-05, "loss": 1.0276, "step": 40655 }, { "epoch": 0.1, "learning_rate": 9.002340329155972e-05, "loss": 1.0301, "step": 40660 }, { "epoch": 0.1, "learning_rate": 9.002214505007801e-05, "loss": 1.0305, "step": 40665 }, { "epoch": 0.1, "learning_rate": 9.00208868085963e-05, "loss": 1.0298, "step": 40670 }, { "epoch": 0.1, "learning_rate": 9.00196285671146e-05, "loss": 1.0293, "step": 40675 }, { "epoch": 0.1, "learning_rate": 9.001837032563289e-05, "loss": 1.0298, "step": 40680 }, { "epoch": 0.1, "learning_rate": 9.00171120841512e-05, "loss": 1.0267, "step": 40685 }, { "epoch": 0.1, "learning_rate": 9.001585384266949e-05, "loss": 1.0306, "step": 40690 }, { "epoch": 0.1, "learning_rate": 9.001459560118778e-05, "loss": 1.0292, "step": 40695 }, { "epoch": 0.1, "learning_rate": 9.001333735970607e-05, "loss": 1.0318, "step": 40700 }, { "epoch": 0.1, "learning_rate": 9.001207911822437e-05, "loss": 1.0288, "step": 40705 }, { "epoch": 0.1, "learning_rate": 9.001082087674268e-05, "loss": 1.0505, "step": 40710 }, { "epoch": 0.1, "learning_rate": 9.000956263526097e-05, "loss": 1.0287, "step": 40715 }, { "epoch": 0.1, "learning_rate": 9.000830439377926e-05, "loss": 1.0327, "step": 40720 }, { "epoch": 0.1, "learning_rate": 9.000704615229755e-05, "loss": 1.0317, "step": 40725 }, { "epoch": 0.1, "learning_rate": 9.000578791081586e-05, "loss": 1.0286, "step": 40730 }, { "epoch": 0.1, "learning_rate": 9.000452966933415e-05, "loss": 1.0275, "step": 40735 }, { "epoch": 0.1, "learning_rate": 9.000327142785244e-05, "loss": 1.0323, "step": 40740 }, { "epoch": 0.1, "learning_rate": 9.000201318637073e-05, "loss": 1.0292, "step": 40745 }, { "epoch": 0.1, "learning_rate": 9.000075494488903e-05, "loss": 1.0292, "step": 40750 }, { "epoch": 0.1, "learning_rate": 8.999949670340733e-05, "loss": 1.0297, "step": 40755 }, { "epoch": 0.1, "learning_rate": 8.999823846192562e-05, "loss": 1.0301, "step": 40760 }, { "epoch": 0.1, "learning_rate": 8.999698022044391e-05, "loss": 1.0275, "step": 40765 }, { "epoch": 0.1, "learning_rate": 8.99957219789622e-05, "loss": 1.0298, "step": 40770 }, { "epoch": 0.1, "learning_rate": 8.999446373748051e-05, "loss": 1.0287, "step": 40775 }, { "epoch": 0.1, "learning_rate": 8.99932054959988e-05, "loss": 1.0304, "step": 40780 }, { "epoch": 0.1, "learning_rate": 8.99919472545171e-05, "loss": 1.0286, "step": 40785 }, { "epoch": 0.1, "learning_rate": 8.999068901303539e-05, "loss": 1.0298, "step": 40790 }, { "epoch": 0.1, "learning_rate": 8.998943077155369e-05, "loss": 1.0293, "step": 40795 }, { "epoch": 0.1, "learning_rate": 8.998817253007198e-05, "loss": 1.0277, "step": 40800 }, { "epoch": 0.1, "learning_rate": 8.998691428859027e-05, "loss": 1.0306, "step": 40805 }, { "epoch": 0.1, "learning_rate": 8.998565604710857e-05, "loss": 1.0297, "step": 40810 }, { "epoch": 0.1, "learning_rate": 8.998439780562686e-05, "loss": 1.0294, "step": 40815 }, { "epoch": 0.1, "learning_rate": 8.998313956414516e-05, "loss": 1.0264, "step": 40820 }, { "epoch": 0.1, "learning_rate": 8.998188132266345e-05, "loss": 1.0307, "step": 40825 }, { "epoch": 0.1, "learning_rate": 8.998062308118175e-05, "loss": 1.0322, "step": 40830 }, { "epoch": 0.1, "learning_rate": 8.997936483970004e-05, "loss": 1.0298, "step": 40835 }, { "epoch": 0.1, "learning_rate": 8.997810659821834e-05, "loss": 1.0292, "step": 40840 }, { "epoch": 0.1, "learning_rate": 8.997684835673663e-05, "loss": 1.0288, "step": 40845 }, { "epoch": 0.1, "learning_rate": 8.997559011525493e-05, "loss": 1.0278, "step": 40850 }, { "epoch": 0.1, "learning_rate": 8.997433187377322e-05, "loss": 1.0278, "step": 40855 }, { "epoch": 0.1, "learning_rate": 8.997307363229152e-05, "loss": 1.029, "step": 40860 }, { "epoch": 0.1, "learning_rate": 8.997181539080981e-05, "loss": 1.032, "step": 40865 }, { "epoch": 0.1, "learning_rate": 8.99705571493281e-05, "loss": 1.0287, "step": 40870 }, { "epoch": 0.1, "learning_rate": 8.99692989078464e-05, "loss": 1.0303, "step": 40875 }, { "epoch": 0.1, "learning_rate": 8.996804066636469e-05, "loss": 1.0267, "step": 40880 }, { "epoch": 0.1, "learning_rate": 8.9966782424883e-05, "loss": 1.0308, "step": 40885 }, { "epoch": 0.1, "learning_rate": 8.996552418340128e-05, "loss": 1.03, "step": 40890 }, { "epoch": 0.1, "learning_rate": 8.996426594191958e-05, "loss": 1.03, "step": 40895 }, { "epoch": 0.1, "learning_rate": 8.996300770043787e-05, "loss": 1.0283, "step": 40900 }, { "epoch": 0.1, "learning_rate": 8.996174945895617e-05, "loss": 1.0305, "step": 40905 }, { "epoch": 0.1, "learning_rate": 8.996049121747446e-05, "loss": 1.0296, "step": 40910 }, { "epoch": 0.1, "learning_rate": 8.995923297599276e-05, "loss": 1.0287, "step": 40915 }, { "epoch": 0.1, "learning_rate": 8.995797473451105e-05, "loss": 1.0296, "step": 40920 }, { "epoch": 0.1, "learning_rate": 8.995671649302935e-05, "loss": 1.0305, "step": 40925 }, { "epoch": 0.1, "learning_rate": 8.995545825154764e-05, "loss": 1.0296, "step": 40930 }, { "epoch": 0.1, "learning_rate": 8.995420001006594e-05, "loss": 1.0281, "step": 40935 }, { "epoch": 0.1, "learning_rate": 8.995294176858423e-05, "loss": 1.0281, "step": 40940 }, { "epoch": 0.1, "learning_rate": 8.995168352710252e-05, "loss": 1.0305, "step": 40945 }, { "epoch": 0.1, "learning_rate": 8.995042528562082e-05, "loss": 1.0293, "step": 40950 }, { "epoch": 0.1, "learning_rate": 8.994916704413912e-05, "loss": 1.0273, "step": 40955 }, { "epoch": 0.1, "learning_rate": 8.994790880265741e-05, "loss": 1.03, "step": 40960 }, { "epoch": 0.1, "learning_rate": 8.99466505611757e-05, "loss": 1.0305, "step": 40965 }, { "epoch": 0.1, "learning_rate": 8.9945392319694e-05, "loss": 1.0289, "step": 40970 }, { "epoch": 0.1, "learning_rate": 8.99441340782123e-05, "loss": 1.0293, "step": 40975 }, { "epoch": 0.1, "learning_rate": 8.994287583673059e-05, "loss": 1.0301, "step": 40980 }, { "epoch": 0.1, "learning_rate": 8.994161759524888e-05, "loss": 1.0287, "step": 40985 }, { "epoch": 0.1, "learning_rate": 8.994035935376718e-05, "loss": 1.0282, "step": 40990 }, { "epoch": 0.1, "learning_rate": 8.993910111228548e-05, "loss": 1.0308, "step": 40995 }, { "epoch": 0.1, "learning_rate": 8.993784287080377e-05, "loss": 1.0291, "step": 41000 }, { "epoch": 0.1, "learning_rate": 8.993658462932206e-05, "loss": 1.0293, "step": 41005 }, { "epoch": 0.1, "learning_rate": 8.993532638784035e-05, "loss": 1.0317, "step": 41010 }, { "epoch": 0.1, "learning_rate": 8.993406814635866e-05, "loss": 1.0532, "step": 41015 }, { "epoch": 0.1, "learning_rate": 8.993280990487695e-05, "loss": 1.0295, "step": 41020 }, { "epoch": 0.1, "learning_rate": 8.993155166339524e-05, "loss": 1.0277, "step": 41025 }, { "epoch": 0.1, "learning_rate": 8.993029342191353e-05, "loss": 1.0273, "step": 41030 }, { "epoch": 0.1, "learning_rate": 8.992903518043184e-05, "loss": 1.032, "step": 41035 }, { "epoch": 0.1, "learning_rate": 8.992777693895013e-05, "loss": 1.0286, "step": 41040 }, { "epoch": 0.1, "learning_rate": 8.992651869746842e-05, "loss": 1.0294, "step": 41045 }, { "epoch": 0.1, "learning_rate": 8.992526045598671e-05, "loss": 1.031, "step": 41050 }, { "epoch": 0.1, "learning_rate": 8.992400221450502e-05, "loss": 1.0281, "step": 41055 }, { "epoch": 0.1, "learning_rate": 8.992274397302331e-05, "loss": 1.0299, "step": 41060 }, { "epoch": 0.1, "learning_rate": 8.99214857315416e-05, "loss": 1.0265, "step": 41065 }, { "epoch": 0.1, "learning_rate": 8.992022749005989e-05, "loss": 1.0291, "step": 41070 }, { "epoch": 0.1, "learning_rate": 8.991896924857818e-05, "loss": 1.0285, "step": 41075 }, { "epoch": 0.1, "learning_rate": 8.991771100709649e-05, "loss": 1.0278, "step": 41080 }, { "epoch": 0.1, "learning_rate": 8.991645276561478e-05, "loss": 1.0276, "step": 41085 }, { "epoch": 0.1, "learning_rate": 8.991519452413307e-05, "loss": 1.0318, "step": 41090 }, { "epoch": 0.1, "learning_rate": 8.991393628265136e-05, "loss": 1.025, "step": 41095 }, { "epoch": 0.1, "learning_rate": 8.991267804116967e-05, "loss": 1.0298, "step": 41100 }, { "epoch": 0.1, "learning_rate": 8.991141979968796e-05, "loss": 1.0276, "step": 41105 }, { "epoch": 0.1, "learning_rate": 8.991016155820625e-05, "loss": 1.0301, "step": 41110 }, { "epoch": 0.1, "learning_rate": 8.990890331672454e-05, "loss": 1.0298, "step": 41115 }, { "epoch": 0.1, "learning_rate": 8.990764507524285e-05, "loss": 1.03, "step": 41120 }, { "epoch": 0.1, "learning_rate": 8.990638683376114e-05, "loss": 1.0288, "step": 41125 }, { "epoch": 0.1, "learning_rate": 8.990512859227943e-05, "loss": 1.0283, "step": 41130 }, { "epoch": 0.1, "learning_rate": 8.990387035079772e-05, "loss": 1.0284, "step": 41135 }, { "epoch": 0.1, "learning_rate": 8.990261210931601e-05, "loss": 1.0296, "step": 41140 }, { "epoch": 0.1, "learning_rate": 8.990135386783432e-05, "loss": 1.0275, "step": 41145 }, { "epoch": 0.1, "learning_rate": 8.990009562635261e-05, "loss": 1.0306, "step": 41150 }, { "epoch": 0.1, "learning_rate": 8.98988373848709e-05, "loss": 1.0288, "step": 41155 }, { "epoch": 0.1, "learning_rate": 8.98975791433892e-05, "loss": 1.0316, "step": 41160 }, { "epoch": 0.1, "learning_rate": 8.98963209019075e-05, "loss": 1.0292, "step": 41165 }, { "epoch": 0.1, "learning_rate": 8.989506266042579e-05, "loss": 1.0279, "step": 41170 }, { "epoch": 0.1, "learning_rate": 8.989380441894408e-05, "loss": 1.0314, "step": 41175 }, { "epoch": 0.1, "learning_rate": 8.989254617746237e-05, "loss": 1.0303, "step": 41180 }, { "epoch": 0.1, "learning_rate": 8.989128793598068e-05, "loss": 1.0298, "step": 41185 }, { "epoch": 0.1, "learning_rate": 8.989002969449897e-05, "loss": 1.0273, "step": 41190 }, { "epoch": 0.1, "learning_rate": 8.988877145301726e-05, "loss": 1.0281, "step": 41195 }, { "epoch": 0.1, "learning_rate": 8.988751321153555e-05, "loss": 1.0283, "step": 41200 }, { "epoch": 0.1, "learning_rate": 8.988625497005386e-05, "loss": 1.0271, "step": 41205 }, { "epoch": 0.1, "learning_rate": 8.988499672857216e-05, "loss": 1.0506, "step": 41210 }, { "epoch": 0.1, "learning_rate": 8.988373848709046e-05, "loss": 1.0275, "step": 41215 }, { "epoch": 0.1, "learning_rate": 8.988248024560875e-05, "loss": 1.0302, "step": 41220 }, { "epoch": 0.1, "learning_rate": 8.988122200412704e-05, "loss": 1.0287, "step": 41225 }, { "epoch": 0.1, "learning_rate": 8.987996376264533e-05, "loss": 1.0297, "step": 41230 }, { "epoch": 0.1, "learning_rate": 8.987870552116364e-05, "loss": 1.0282, "step": 41235 }, { "epoch": 0.1, "learning_rate": 8.987744727968193e-05, "loss": 1.0282, "step": 41240 }, { "epoch": 0.1, "learning_rate": 8.987618903820022e-05, "loss": 1.0538, "step": 41245 }, { "epoch": 0.1, "learning_rate": 8.987493079671851e-05, "loss": 1.0302, "step": 41250 }, { "epoch": 0.1, "learning_rate": 8.987367255523682e-05, "loss": 1.0276, "step": 41255 }, { "epoch": 0.1, "learning_rate": 8.987241431375511e-05, "loss": 1.0322, "step": 41260 }, { "epoch": 0.1, "learning_rate": 8.98711560722734e-05, "loss": 1.029, "step": 41265 }, { "epoch": 0.1, "learning_rate": 8.986989783079169e-05, "loss": 1.0278, "step": 41270 }, { "epoch": 0.1, "learning_rate": 8.986863958931e-05, "loss": 1.0289, "step": 41275 }, { "epoch": 0.1, "learning_rate": 8.986738134782829e-05, "loss": 1.0303, "step": 41280 }, { "epoch": 0.1, "learning_rate": 8.986612310634658e-05, "loss": 1.0278, "step": 41285 }, { "epoch": 0.1, "learning_rate": 8.986486486486487e-05, "loss": 1.0277, "step": 41290 }, { "epoch": 0.1, "learning_rate": 8.986360662338316e-05, "loss": 1.0293, "step": 41295 }, { "epoch": 0.1, "learning_rate": 8.986234838190147e-05, "loss": 1.0299, "step": 41300 }, { "epoch": 0.1, "learning_rate": 8.986109014041976e-05, "loss": 1.0293, "step": 41305 }, { "epoch": 0.1, "learning_rate": 8.985983189893805e-05, "loss": 1.025, "step": 41310 }, { "epoch": 0.1, "learning_rate": 8.985857365745634e-05, "loss": 1.0305, "step": 41315 }, { "epoch": 0.1, "learning_rate": 8.985731541597465e-05, "loss": 1.0281, "step": 41320 }, { "epoch": 0.1, "learning_rate": 8.985605717449294e-05, "loss": 1.0285, "step": 41325 }, { "epoch": 0.1, "learning_rate": 8.985479893301123e-05, "loss": 1.0292, "step": 41330 }, { "epoch": 0.1, "learning_rate": 8.985354069152952e-05, "loss": 1.0325, "step": 41335 }, { "epoch": 0.1, "learning_rate": 8.985228245004781e-05, "loss": 1.0303, "step": 41340 }, { "epoch": 0.1, "learning_rate": 8.985102420856612e-05, "loss": 1.0292, "step": 41345 }, { "epoch": 0.1, "learning_rate": 8.984976596708441e-05, "loss": 1.0297, "step": 41350 }, { "epoch": 0.1, "learning_rate": 8.98485077256027e-05, "loss": 1.0289, "step": 41355 }, { "epoch": 0.1, "learning_rate": 8.984724948412099e-05, "loss": 1.0298, "step": 41360 }, { "epoch": 0.1, "learning_rate": 8.98459912426393e-05, "loss": 1.0279, "step": 41365 }, { "epoch": 0.1, "learning_rate": 8.984473300115759e-05, "loss": 1.03, "step": 41370 }, { "epoch": 0.1, "learning_rate": 8.984347475967588e-05, "loss": 1.0295, "step": 41375 }, { "epoch": 0.1, "learning_rate": 8.984221651819417e-05, "loss": 1.0289, "step": 41380 }, { "epoch": 0.1, "learning_rate": 8.984095827671248e-05, "loss": 1.0294, "step": 41385 }, { "epoch": 0.1, "learning_rate": 8.983970003523077e-05, "loss": 1.0287, "step": 41390 }, { "epoch": 0.1, "learning_rate": 8.983844179374906e-05, "loss": 1.0283, "step": 41395 }, { "epoch": 0.1, "learning_rate": 8.983718355226735e-05, "loss": 1.03, "step": 41400 }, { "epoch": 0.1, "learning_rate": 8.983592531078564e-05, "loss": 1.029, "step": 41405 }, { "epoch": 0.1, "learning_rate": 8.983466706930395e-05, "loss": 1.0291, "step": 41410 }, { "epoch": 0.1, "learning_rate": 8.983340882782224e-05, "loss": 1.0291, "step": 41415 }, { "epoch": 0.1, "learning_rate": 8.983215058634053e-05, "loss": 1.0296, "step": 41420 }, { "epoch": 0.1, "learning_rate": 8.983089234485882e-05, "loss": 1.0271, "step": 41425 }, { "epoch": 0.1, "learning_rate": 8.982963410337713e-05, "loss": 1.0321, "step": 41430 }, { "epoch": 0.1, "learning_rate": 8.982837586189542e-05, "loss": 1.0302, "step": 41435 }, { "epoch": 0.1, "learning_rate": 8.982711762041371e-05, "loss": 1.0283, "step": 41440 }, { "epoch": 0.1, "learning_rate": 8.9825859378932e-05, "loss": 1.0298, "step": 41445 }, { "epoch": 0.1, "learning_rate": 8.982460113745031e-05, "loss": 1.0289, "step": 41450 }, { "epoch": 0.1, "learning_rate": 8.98233428959686e-05, "loss": 1.0305, "step": 41455 }, { "epoch": 0.1, "learning_rate": 8.982208465448689e-05, "loss": 1.0305, "step": 41460 }, { "epoch": 0.1, "learning_rate": 8.982082641300518e-05, "loss": 1.0292, "step": 41465 }, { "epoch": 0.1, "learning_rate": 8.981956817152348e-05, "loss": 1.0288, "step": 41470 }, { "epoch": 0.1, "learning_rate": 8.981830993004178e-05, "loss": 1.0288, "step": 41475 }, { "epoch": 0.1, "learning_rate": 8.981705168856007e-05, "loss": 1.0306, "step": 41480 }, { "epoch": 0.1, "learning_rate": 8.981579344707836e-05, "loss": 1.0293, "step": 41485 }, { "epoch": 0.1, "learning_rate": 8.981453520559666e-05, "loss": 1.0295, "step": 41490 }, { "epoch": 0.1, "learning_rate": 8.981327696411496e-05, "loss": 1.0286, "step": 41495 }, { "epoch": 0.1, "learning_rate": 8.981201872263325e-05, "loss": 1.0275, "step": 41500 }, { "epoch": 0.1, "learning_rate": 8.981076048115154e-05, "loss": 1.0347, "step": 41505 }, { "epoch": 0.1, "learning_rate": 8.980950223966984e-05, "loss": 1.0298, "step": 41510 }, { "epoch": 0.1, "learning_rate": 8.980824399818814e-05, "loss": 1.0319, "step": 41515 }, { "epoch": 0.1, "learning_rate": 8.980698575670643e-05, "loss": 1.0285, "step": 41520 }, { "epoch": 0.1, "learning_rate": 8.980572751522472e-05, "loss": 1.0298, "step": 41525 }, { "epoch": 0.1, "learning_rate": 8.980446927374302e-05, "loss": 1.0472, "step": 41530 }, { "epoch": 0.1, "learning_rate": 8.980321103226131e-05, "loss": 1.0318, "step": 41535 }, { "epoch": 0.1, "learning_rate": 8.980195279077961e-05, "loss": 1.0284, "step": 41540 }, { "epoch": 0.1, "learning_rate": 8.98006945492979e-05, "loss": 1.0285, "step": 41545 }, { "epoch": 0.1, "learning_rate": 8.97994363078162e-05, "loss": 1.0292, "step": 41550 }, { "epoch": 0.1, "learning_rate": 8.979817806633449e-05, "loss": 1.0291, "step": 41555 }, { "epoch": 0.1, "learning_rate": 8.979691982485279e-05, "loss": 1.0295, "step": 41560 }, { "epoch": 0.1, "learning_rate": 8.979566158337108e-05, "loss": 1.0286, "step": 41565 }, { "epoch": 0.1, "learning_rate": 8.979440334188938e-05, "loss": 1.0295, "step": 41570 }, { "epoch": 0.1, "learning_rate": 8.979314510040767e-05, "loss": 1.0308, "step": 41575 }, { "epoch": 0.1, "learning_rate": 8.979188685892597e-05, "loss": 1.029, "step": 41580 }, { "epoch": 0.1, "learning_rate": 8.979062861744426e-05, "loss": 1.0427, "step": 41585 }, { "epoch": 0.1, "learning_rate": 8.978937037596256e-05, "loss": 1.0292, "step": 41590 }, { "epoch": 0.1, "learning_rate": 8.978811213448085e-05, "loss": 1.0311, "step": 41595 }, { "epoch": 0.1, "learning_rate": 8.978685389299914e-05, "loss": 1.0301, "step": 41600 }, { "epoch": 0.1, "learning_rate": 8.978559565151744e-05, "loss": 1.0297, "step": 41605 }, { "epoch": 0.1, "learning_rate": 8.978433741003574e-05, "loss": 1.031, "step": 41610 }, { "epoch": 0.1, "learning_rate": 8.978307916855403e-05, "loss": 1.0312, "step": 41615 }, { "epoch": 0.1, "learning_rate": 8.978182092707232e-05, "loss": 1.0277, "step": 41620 }, { "epoch": 0.1, "learning_rate": 8.978056268559062e-05, "loss": 1.0295, "step": 41625 }, { "epoch": 0.1, "learning_rate": 8.977930444410892e-05, "loss": 1.0298, "step": 41630 }, { "epoch": 0.1, "learning_rate": 8.977804620262721e-05, "loss": 1.0299, "step": 41635 }, { "epoch": 0.1, "learning_rate": 8.97767879611455e-05, "loss": 1.0275, "step": 41640 }, { "epoch": 0.1, "learning_rate": 8.97755297196638e-05, "loss": 1.0289, "step": 41645 }, { "epoch": 0.1, "learning_rate": 8.97742714781821e-05, "loss": 1.0308, "step": 41650 }, { "epoch": 0.1, "learning_rate": 8.977301323670039e-05, "loss": 1.0309, "step": 41655 }, { "epoch": 0.1, "learning_rate": 8.977175499521868e-05, "loss": 1.0298, "step": 41660 }, { "epoch": 0.1, "learning_rate": 8.977049675373697e-05, "loss": 1.0539, "step": 41665 }, { "epoch": 0.1, "learning_rate": 8.976923851225528e-05, "loss": 1.0283, "step": 41670 }, { "epoch": 0.1, "learning_rate": 8.976798027077357e-05, "loss": 1.0298, "step": 41675 }, { "epoch": 0.1, "learning_rate": 8.976672202929186e-05, "loss": 1.0287, "step": 41680 }, { "epoch": 0.1, "learning_rate": 8.976546378781015e-05, "loss": 1.0286, "step": 41685 }, { "epoch": 0.1, "learning_rate": 8.976420554632846e-05, "loss": 1.0294, "step": 41690 }, { "epoch": 0.1, "learning_rate": 8.976294730484675e-05, "loss": 1.0302, "step": 41695 }, { "epoch": 0.1, "learning_rate": 8.976168906336504e-05, "loss": 1.0295, "step": 41700 }, { "epoch": 0.1, "learning_rate": 8.976043082188334e-05, "loss": 1.029, "step": 41705 }, { "epoch": 0.1, "learning_rate": 8.975917258040164e-05, "loss": 1.0294, "step": 41710 }, { "epoch": 0.1, "learning_rate": 8.975791433891994e-05, "loss": 1.0287, "step": 41715 }, { "epoch": 0.1, "learning_rate": 8.975665609743823e-05, "loss": 1.0291, "step": 41720 }, { "epoch": 0.1, "learning_rate": 8.975539785595652e-05, "loss": 1.0304, "step": 41725 }, { "epoch": 0.1, "learning_rate": 8.975413961447482e-05, "loss": 1.0287, "step": 41730 }, { "epoch": 0.1, "learning_rate": 8.975288137299312e-05, "loss": 1.0282, "step": 41735 }, { "epoch": 0.1, "learning_rate": 8.975162313151141e-05, "loss": 1.0308, "step": 41740 }, { "epoch": 0.1, "learning_rate": 8.97503648900297e-05, "loss": 1.0283, "step": 41745 }, { "epoch": 0.1, "learning_rate": 8.9749106648548e-05, "loss": 1.0284, "step": 41750 }, { "epoch": 0.1, "learning_rate": 8.974784840706629e-05, "loss": 1.0298, "step": 41755 }, { "epoch": 0.1, "learning_rate": 8.974659016558459e-05, "loss": 1.0259, "step": 41760 }, { "epoch": 0.1, "learning_rate": 8.974533192410288e-05, "loss": 1.0292, "step": 41765 }, { "epoch": 0.1, "learning_rate": 8.974407368262118e-05, "loss": 1.0292, "step": 41770 }, { "epoch": 0.1, "learning_rate": 8.974281544113947e-05, "loss": 1.0304, "step": 41775 }, { "epoch": 0.1, "learning_rate": 8.974155719965777e-05, "loss": 1.0267, "step": 41780 }, { "epoch": 0.1, "learning_rate": 8.974029895817606e-05, "loss": 1.0299, "step": 41785 }, { "epoch": 0.1, "learning_rate": 8.973904071669436e-05, "loss": 1.0294, "step": 41790 }, { "epoch": 0.1, "learning_rate": 8.973778247521265e-05, "loss": 1.0291, "step": 41795 }, { "epoch": 0.1, "learning_rate": 8.973652423373095e-05, "loss": 1.0268, "step": 41800 }, { "epoch": 0.1, "learning_rate": 8.973526599224924e-05, "loss": 1.0289, "step": 41805 }, { "epoch": 0.1, "learning_rate": 8.973400775076754e-05, "loss": 1.0301, "step": 41810 }, { "epoch": 0.1, "learning_rate": 8.973274950928583e-05, "loss": 1.0309, "step": 41815 }, { "epoch": 0.1, "learning_rate": 8.973149126780412e-05, "loss": 1.0282, "step": 41820 }, { "epoch": 0.1, "learning_rate": 8.973023302632242e-05, "loss": 1.0287, "step": 41825 }, { "epoch": 0.11, "learning_rate": 8.972897478484072e-05, "loss": 1.0299, "step": 41830 }, { "epoch": 0.11, "learning_rate": 8.9727716543359e-05, "loss": 1.0271, "step": 41835 }, { "epoch": 0.11, "learning_rate": 8.97264583018773e-05, "loss": 1.0292, "step": 41840 }, { "epoch": 0.11, "learning_rate": 8.97252000603956e-05, "loss": 1.0298, "step": 41845 }, { "epoch": 0.11, "learning_rate": 8.97239418189139e-05, "loss": 1.0295, "step": 41850 }, { "epoch": 0.11, "learning_rate": 8.972268357743219e-05, "loss": 1.0288, "step": 41855 }, { "epoch": 0.11, "learning_rate": 8.972142533595048e-05, "loss": 1.0294, "step": 41860 }, { "epoch": 0.11, "learning_rate": 8.972016709446878e-05, "loss": 1.0296, "step": 41865 }, { "epoch": 0.11, "learning_rate": 8.971890885298707e-05, "loss": 1.0314, "step": 41870 }, { "epoch": 0.11, "learning_rate": 8.971765061150537e-05, "loss": 1.0302, "step": 41875 }, { "epoch": 0.11, "learning_rate": 8.971639237002366e-05, "loss": 1.031, "step": 41880 }, { "epoch": 0.11, "learning_rate": 8.971513412854195e-05, "loss": 1.0296, "step": 41885 }, { "epoch": 0.11, "learning_rate": 8.971387588706025e-05, "loss": 1.0311, "step": 41890 }, { "epoch": 0.11, "learning_rate": 8.971261764557855e-05, "loss": 1.0293, "step": 41895 }, { "epoch": 0.11, "learning_rate": 8.971135940409684e-05, "loss": 1.0284, "step": 41900 }, { "epoch": 0.11, "learning_rate": 8.971010116261513e-05, "loss": 1.0284, "step": 41905 }, { "epoch": 0.11, "learning_rate": 8.970884292113343e-05, "loss": 1.0513, "step": 41910 }, { "epoch": 0.11, "learning_rate": 8.970758467965173e-05, "loss": 1.0312, "step": 41915 }, { "epoch": 0.11, "learning_rate": 8.970632643817002e-05, "loss": 1.0297, "step": 41920 }, { "epoch": 0.11, "learning_rate": 8.970506819668831e-05, "loss": 1.031, "step": 41925 }, { "epoch": 0.11, "learning_rate": 8.970380995520661e-05, "loss": 1.0279, "step": 41930 }, { "epoch": 0.11, "learning_rate": 8.97025517137249e-05, "loss": 1.0293, "step": 41935 }, { "epoch": 0.11, "learning_rate": 8.97012934722432e-05, "loss": 1.0279, "step": 41940 }, { "epoch": 0.11, "learning_rate": 8.970003523076149e-05, "loss": 1.0296, "step": 41945 }, { "epoch": 0.11, "learning_rate": 8.969877698927978e-05, "loss": 1.0315, "step": 41950 }, { "epoch": 0.11, "learning_rate": 8.969751874779809e-05, "loss": 1.0295, "step": 41955 }, { "epoch": 0.11, "learning_rate": 8.969626050631638e-05, "loss": 1.0283, "step": 41960 }, { "epoch": 0.11, "learning_rate": 8.969500226483467e-05, "loss": 1.0288, "step": 41965 }, { "epoch": 0.11, "learning_rate": 8.969374402335296e-05, "loss": 1.0283, "step": 41970 }, { "epoch": 0.11, "learning_rate": 8.969248578187127e-05, "loss": 1.0287, "step": 41975 }, { "epoch": 0.11, "learning_rate": 8.969122754038956e-05, "loss": 1.0296, "step": 41980 }, { "epoch": 0.11, "learning_rate": 8.968996929890785e-05, "loss": 1.0287, "step": 41985 }, { "epoch": 0.11, "learning_rate": 8.968871105742614e-05, "loss": 1.0292, "step": 41990 }, { "epoch": 0.11, "learning_rate": 8.968745281594443e-05, "loss": 1.0275, "step": 41995 }, { "epoch": 0.11, "learning_rate": 8.968619457446274e-05, "loss": 1.0457, "step": 42000 }, { "epoch": 0.11, "learning_rate": 8.968493633298103e-05, "loss": 1.0423, "step": 42005 }, { "epoch": 0.11, "learning_rate": 8.968367809149932e-05, "loss": 1.0292, "step": 42010 }, { "epoch": 0.11, "learning_rate": 8.968241985001761e-05, "loss": 1.0257, "step": 42015 }, { "epoch": 0.11, "learning_rate": 8.968116160853592e-05, "loss": 1.0294, "step": 42020 }, { "epoch": 0.11, "learning_rate": 8.967990336705421e-05, "loss": 1.0298, "step": 42025 }, { "epoch": 0.11, "learning_rate": 8.96786451255725e-05, "loss": 1.0249, "step": 42030 }, { "epoch": 0.11, "learning_rate": 8.967738688409079e-05, "loss": 1.0276, "step": 42035 }, { "epoch": 0.11, "learning_rate": 8.96761286426091e-05, "loss": 1.0278, "step": 42040 }, { "epoch": 0.11, "learning_rate": 8.967487040112739e-05, "loss": 1.0308, "step": 42045 }, { "epoch": 0.11, "learning_rate": 8.967361215964568e-05, "loss": 1.0257, "step": 42050 }, { "epoch": 0.11, "learning_rate": 8.967235391816397e-05, "loss": 1.0291, "step": 42055 }, { "epoch": 0.11, "learning_rate": 8.967109567668226e-05, "loss": 1.0314, "step": 42060 }, { "epoch": 0.11, "learning_rate": 8.966983743520057e-05, "loss": 1.0288, "step": 42065 }, { "epoch": 0.11, "learning_rate": 8.966857919371886e-05, "loss": 1.0293, "step": 42070 }, { "epoch": 0.11, "learning_rate": 8.966732095223715e-05, "loss": 1.029, "step": 42075 }, { "epoch": 0.11, "learning_rate": 8.966606271075544e-05, "loss": 1.0289, "step": 42080 }, { "epoch": 0.11, "learning_rate": 8.966480446927375e-05, "loss": 1.0298, "step": 42085 }, { "epoch": 0.11, "learning_rate": 8.966354622779204e-05, "loss": 1.0281, "step": 42090 }, { "epoch": 0.11, "learning_rate": 8.966228798631033e-05, "loss": 1.0304, "step": 42095 }, { "epoch": 0.11, "learning_rate": 8.966102974482862e-05, "loss": 1.029, "step": 42100 }, { "epoch": 0.11, "learning_rate": 8.965977150334693e-05, "loss": 1.0302, "step": 42105 }, { "epoch": 0.11, "learning_rate": 8.965851326186522e-05, "loss": 1.03, "step": 42110 }, { "epoch": 0.11, "learning_rate": 8.965725502038351e-05, "loss": 1.0289, "step": 42115 }, { "epoch": 0.11, "learning_rate": 8.96559967789018e-05, "loss": 1.0296, "step": 42120 }, { "epoch": 0.11, "learning_rate": 8.96547385374201e-05, "loss": 1.0254, "step": 42125 }, { "epoch": 0.11, "learning_rate": 8.96534802959384e-05, "loss": 1.0315, "step": 42130 }, { "epoch": 0.11, "learning_rate": 8.965222205445669e-05, "loss": 1.0284, "step": 42135 }, { "epoch": 0.11, "learning_rate": 8.965096381297498e-05, "loss": 1.0276, "step": 42140 }, { "epoch": 0.11, "learning_rate": 8.964970557149328e-05, "loss": 1.0278, "step": 42145 }, { "epoch": 0.11, "learning_rate": 8.964844733001158e-05, "loss": 1.0292, "step": 42150 }, { "epoch": 0.11, "learning_rate": 8.964718908852987e-05, "loss": 1.029, "step": 42155 }, { "epoch": 0.11, "learning_rate": 8.964593084704816e-05, "loss": 1.0292, "step": 42160 }, { "epoch": 0.11, "learning_rate": 8.964467260556646e-05, "loss": 1.0272, "step": 42165 }, { "epoch": 0.11, "learning_rate": 8.964341436408476e-05, "loss": 1.0297, "step": 42170 }, { "epoch": 0.11, "learning_rate": 8.964215612260305e-05, "loss": 1.0257, "step": 42175 }, { "epoch": 0.11, "learning_rate": 8.964089788112134e-05, "loss": 1.0289, "step": 42180 }, { "epoch": 0.11, "learning_rate": 8.963963963963964e-05, "loss": 1.0293, "step": 42185 }, { "epoch": 0.11, "learning_rate": 8.963838139815793e-05, "loss": 1.0281, "step": 42190 }, { "epoch": 0.11, "learning_rate": 8.963712315667623e-05, "loss": 1.0273, "step": 42195 }, { "epoch": 0.11, "learning_rate": 8.963586491519452e-05, "loss": 1.0278, "step": 42200 }, { "epoch": 0.11, "learning_rate": 8.963460667371283e-05, "loss": 1.0299, "step": 42205 }, { "epoch": 0.11, "learning_rate": 8.963334843223112e-05, "loss": 1.0278, "step": 42210 }, { "epoch": 0.11, "learning_rate": 8.963209019074941e-05, "loss": 1.0265, "step": 42215 }, { "epoch": 0.11, "learning_rate": 8.963083194926772e-05, "loss": 1.0274, "step": 42220 }, { "epoch": 0.11, "learning_rate": 8.962957370778601e-05, "loss": 1.0284, "step": 42225 }, { "epoch": 0.11, "learning_rate": 8.96283154663043e-05, "loss": 1.032, "step": 42230 }, { "epoch": 0.11, "learning_rate": 8.962705722482259e-05, "loss": 1.07, "step": 42235 }, { "epoch": 0.11, "learning_rate": 8.96257989833409e-05, "loss": 1.0314, "step": 42240 }, { "epoch": 0.11, "learning_rate": 8.962454074185919e-05, "loss": 1.0287, "step": 42245 }, { "epoch": 0.11, "learning_rate": 8.962328250037748e-05, "loss": 1.027, "step": 42250 }, { "epoch": 0.11, "learning_rate": 8.962202425889577e-05, "loss": 1.027, "step": 42255 }, { "epoch": 0.11, "learning_rate": 8.962076601741408e-05, "loss": 1.03, "step": 42260 }, { "epoch": 0.11, "learning_rate": 8.961950777593237e-05, "loss": 1.0292, "step": 42265 }, { "epoch": 0.11, "learning_rate": 8.961824953445066e-05, "loss": 1.0285, "step": 42270 }, { "epoch": 0.11, "learning_rate": 8.961699129296895e-05, "loss": 1.0289, "step": 42275 }, { "epoch": 0.11, "learning_rate": 8.961573305148724e-05, "loss": 1.0299, "step": 42280 }, { "epoch": 0.11, "learning_rate": 8.961447481000555e-05, "loss": 1.0525, "step": 42285 }, { "epoch": 0.11, "learning_rate": 8.961321656852384e-05, "loss": 1.0276, "step": 42290 }, { "epoch": 0.11, "learning_rate": 8.961195832704213e-05, "loss": 1.0307, "step": 42295 }, { "epoch": 0.11, "learning_rate": 8.961070008556042e-05, "loss": 1.0288, "step": 42300 }, { "epoch": 0.11, "learning_rate": 8.960944184407873e-05, "loss": 1.0282, "step": 42305 }, { "epoch": 0.11, "learning_rate": 8.960818360259702e-05, "loss": 1.0287, "step": 42310 }, { "epoch": 0.11, "learning_rate": 8.960692536111531e-05, "loss": 1.0295, "step": 42315 }, { "epoch": 0.11, "learning_rate": 8.96056671196336e-05, "loss": 1.029, "step": 42320 }, { "epoch": 0.11, "learning_rate": 8.960440887815191e-05, "loss": 1.0289, "step": 42325 }, { "epoch": 0.11, "learning_rate": 8.96031506366702e-05, "loss": 1.0274, "step": 42330 }, { "epoch": 0.11, "learning_rate": 8.960189239518849e-05, "loss": 1.0284, "step": 42335 }, { "epoch": 0.11, "learning_rate": 8.960063415370678e-05, "loss": 1.0311, "step": 42340 }, { "epoch": 0.11, "learning_rate": 8.959937591222507e-05, "loss": 1.0505, "step": 42345 }, { "epoch": 0.11, "learning_rate": 8.959811767074338e-05, "loss": 1.0299, "step": 42350 }, { "epoch": 0.11, "learning_rate": 8.959685942926167e-05, "loss": 1.0361, "step": 42355 }, { "epoch": 0.11, "learning_rate": 8.959560118777996e-05, "loss": 1.029, "step": 42360 }, { "epoch": 0.11, "learning_rate": 8.959434294629825e-05, "loss": 1.0283, "step": 42365 }, { "epoch": 0.11, "learning_rate": 8.959308470481656e-05, "loss": 1.0306, "step": 42370 }, { "epoch": 0.11, "learning_rate": 8.959182646333485e-05, "loss": 1.0276, "step": 42375 }, { "epoch": 0.11, "learning_rate": 8.959056822185314e-05, "loss": 1.0297, "step": 42380 }, { "epoch": 0.11, "learning_rate": 8.958930998037143e-05, "loss": 1.05, "step": 42385 }, { "epoch": 0.11, "learning_rate": 8.958805173888974e-05, "loss": 1.0311, "step": 42390 }, { "epoch": 0.11, "learning_rate": 8.958679349740803e-05, "loss": 1.0293, "step": 42395 }, { "epoch": 0.11, "learning_rate": 8.958553525592632e-05, "loss": 1.0293, "step": 42400 }, { "epoch": 0.11, "learning_rate": 8.958427701444461e-05, "loss": 1.0275, "step": 42405 }, { "epoch": 0.11, "learning_rate": 8.95830187729629e-05, "loss": 1.0291, "step": 42410 }, { "epoch": 0.11, "learning_rate": 8.958176053148121e-05, "loss": 1.0281, "step": 42415 }, { "epoch": 0.11, "learning_rate": 8.95805022899995e-05, "loss": 1.0283, "step": 42420 }, { "epoch": 0.11, "learning_rate": 8.95792440485178e-05, "loss": 1.0303, "step": 42425 }, { "epoch": 0.11, "learning_rate": 8.957798580703609e-05, "loss": 1.0271, "step": 42430 }, { "epoch": 0.11, "learning_rate": 8.957672756555439e-05, "loss": 1.0318, "step": 42435 }, { "epoch": 0.11, "learning_rate": 8.957546932407268e-05, "loss": 1.0523, "step": 42440 }, { "epoch": 0.11, "learning_rate": 8.957421108259097e-05, "loss": 1.0313, "step": 42445 }, { "epoch": 0.11, "learning_rate": 8.957295284110927e-05, "loss": 1.029, "step": 42450 }, { "epoch": 0.11, "learning_rate": 8.957169459962757e-05, "loss": 1.0434, "step": 42455 }, { "epoch": 0.11, "learning_rate": 8.957043635814586e-05, "loss": 1.0285, "step": 42460 }, { "epoch": 0.11, "learning_rate": 8.956917811666415e-05, "loss": 1.0473, "step": 42465 }, { "epoch": 0.11, "learning_rate": 8.956791987518245e-05, "loss": 1.0287, "step": 42470 }, { "epoch": 0.11, "learning_rate": 8.956666163370074e-05, "loss": 1.0303, "step": 42475 }, { "epoch": 0.11, "learning_rate": 8.956540339221904e-05, "loss": 1.0496, "step": 42480 }, { "epoch": 0.11, "learning_rate": 8.956439679903368e-05, "loss": 1.0499, "step": 42485 }, { "epoch": 0.11, "learning_rate": 8.956313855755197e-05, "loss": 1.0299, "step": 42490 }, { "epoch": 0.11, "learning_rate": 8.956188031607026e-05, "loss": 1.0272, "step": 42495 }, { "epoch": 0.11, "learning_rate": 8.956062207458857e-05, "loss": 1.0282, "step": 42500 }, { "epoch": 0.11, "learning_rate": 8.955936383310686e-05, "loss": 1.0279, "step": 42505 }, { "epoch": 0.11, "learning_rate": 8.955810559162515e-05, "loss": 1.0296, "step": 42510 }, { "epoch": 0.11, "learning_rate": 8.955684735014344e-05, "loss": 1.0306, "step": 42515 }, { "epoch": 0.11, "learning_rate": 8.955558910866175e-05, "loss": 1.0526, "step": 42520 }, { "epoch": 0.11, "learning_rate": 8.955433086718004e-05, "loss": 1.0304, "step": 42525 }, { "epoch": 0.11, "learning_rate": 8.955307262569833e-05, "loss": 1.0277, "step": 42530 }, { "epoch": 0.11, "learning_rate": 8.955181438421662e-05, "loss": 1.0301, "step": 42535 }, { "epoch": 0.11, "learning_rate": 8.955055614273493e-05, "loss": 1.0277, "step": 42540 }, { "epoch": 0.11, "learning_rate": 8.954929790125322e-05, "loss": 1.0301, "step": 42545 }, { "epoch": 0.11, "learning_rate": 8.954803965977151e-05, "loss": 1.03, "step": 42550 }, { "epoch": 0.11, "learning_rate": 8.95467814182898e-05, "loss": 1.0284, "step": 42555 }, { "epoch": 0.11, "learning_rate": 8.95455231768081e-05, "loss": 1.029, "step": 42560 }, { "epoch": 0.11, "learning_rate": 8.95442649353264e-05, "loss": 1.03, "step": 42565 }, { "epoch": 0.11, "learning_rate": 8.954300669384469e-05, "loss": 1.0279, "step": 42570 }, { "epoch": 0.11, "learning_rate": 8.954174845236298e-05, "loss": 1.0275, "step": 42575 }, { "epoch": 0.11, "learning_rate": 8.954049021088127e-05, "loss": 1.0281, "step": 42580 }, { "epoch": 0.11, "learning_rate": 8.953923196939958e-05, "loss": 1.0283, "step": 42585 }, { "epoch": 0.11, "learning_rate": 8.953797372791787e-05, "loss": 1.0285, "step": 42590 }, { "epoch": 0.11, "learning_rate": 8.953671548643616e-05, "loss": 1.0282, "step": 42595 }, { "epoch": 0.11, "learning_rate": 8.953545724495445e-05, "loss": 1.0276, "step": 42600 }, { "epoch": 0.11, "learning_rate": 8.953419900347276e-05, "loss": 1.0311, "step": 42605 }, { "epoch": 0.11, "learning_rate": 8.953294076199105e-05, "loss": 1.0284, "step": 42610 }, { "epoch": 0.11, "learning_rate": 8.953168252050934e-05, "loss": 1.0289, "step": 42615 }, { "epoch": 0.11, "learning_rate": 8.953042427902763e-05, "loss": 1.0286, "step": 42620 }, { "epoch": 0.11, "learning_rate": 8.952916603754592e-05, "loss": 1.0274, "step": 42625 }, { "epoch": 0.11, "learning_rate": 8.952790779606423e-05, "loss": 1.0285, "step": 42630 }, { "epoch": 0.11, "learning_rate": 8.952664955458252e-05, "loss": 1.0317, "step": 42635 }, { "epoch": 0.11, "learning_rate": 8.952539131310081e-05, "loss": 1.0293, "step": 42640 }, { "epoch": 0.11, "learning_rate": 8.95241330716191e-05, "loss": 1.0307, "step": 42645 }, { "epoch": 0.11, "learning_rate": 8.952287483013741e-05, "loss": 1.0284, "step": 42650 }, { "epoch": 0.11, "learning_rate": 8.95216165886557e-05, "loss": 1.029, "step": 42655 }, { "epoch": 0.11, "learning_rate": 8.952035834717399e-05, "loss": 1.0283, "step": 42660 }, { "epoch": 0.11, "learning_rate": 8.951910010569228e-05, "loss": 1.0285, "step": 42665 }, { "epoch": 0.11, "learning_rate": 8.951784186421059e-05, "loss": 1.0301, "step": 42670 }, { "epoch": 0.11, "learning_rate": 8.951658362272888e-05, "loss": 1.0296, "step": 42675 }, { "epoch": 0.11, "learning_rate": 8.951532538124717e-05, "loss": 1.0277, "step": 42680 }, { "epoch": 0.11, "learning_rate": 8.951406713976546e-05, "loss": 1.0286, "step": 42685 }, { "epoch": 0.11, "learning_rate": 8.951280889828376e-05, "loss": 1.029, "step": 42690 }, { "epoch": 0.11, "learning_rate": 8.951155065680206e-05, "loss": 1.0313, "step": 42695 }, { "epoch": 0.11, "learning_rate": 8.951029241532035e-05, "loss": 1.0293, "step": 42700 }, { "epoch": 0.11, "learning_rate": 8.950903417383864e-05, "loss": 1.0278, "step": 42705 }, { "epoch": 0.11, "learning_rate": 8.950777593235694e-05, "loss": 1.0303, "step": 42710 }, { "epoch": 0.11, "learning_rate": 8.950651769087524e-05, "loss": 1.0299, "step": 42715 }, { "epoch": 0.11, "learning_rate": 8.950525944939353e-05, "loss": 1.0282, "step": 42720 }, { "epoch": 0.11, "learning_rate": 8.950400120791182e-05, "loss": 1.0303, "step": 42725 }, { "epoch": 0.11, "learning_rate": 8.950274296643012e-05, "loss": 1.0296, "step": 42730 }, { "epoch": 0.11, "learning_rate": 8.950148472494841e-05, "loss": 1.0293, "step": 42735 }, { "epoch": 0.11, "learning_rate": 8.950022648346671e-05, "loss": 1.0286, "step": 42740 }, { "epoch": 0.11, "learning_rate": 8.9498968241985e-05, "loss": 1.029, "step": 42745 }, { "epoch": 0.11, "learning_rate": 8.94977100005033e-05, "loss": 1.0318, "step": 42750 }, { "epoch": 0.11, "learning_rate": 8.949645175902159e-05, "loss": 1.0274, "step": 42755 }, { "epoch": 0.11, "learning_rate": 8.949519351753989e-05, "loss": 1.0302, "step": 42760 }, { "epoch": 0.11, "learning_rate": 8.949393527605818e-05, "loss": 1.031, "step": 42765 }, { "epoch": 0.11, "learning_rate": 8.949267703457648e-05, "loss": 1.0286, "step": 42770 }, { "epoch": 0.11, "learning_rate": 8.949141879309477e-05, "loss": 1.027, "step": 42775 }, { "epoch": 0.11, "learning_rate": 8.949016055161307e-05, "loss": 1.029, "step": 42780 }, { "epoch": 0.11, "learning_rate": 8.948890231013136e-05, "loss": 1.0278, "step": 42785 }, { "epoch": 0.11, "learning_rate": 8.948764406864966e-05, "loss": 1.0295, "step": 42790 }, { "epoch": 0.11, "learning_rate": 8.948638582716795e-05, "loss": 1.0279, "step": 42795 }, { "epoch": 0.11, "learning_rate": 8.948512758568624e-05, "loss": 1.0267, "step": 42800 }, { "epoch": 0.11, "learning_rate": 8.948386934420454e-05, "loss": 1.026, "step": 42805 }, { "epoch": 0.11, "learning_rate": 8.948261110272284e-05, "loss": 1.0299, "step": 42810 }, { "epoch": 0.11, "learning_rate": 8.948135286124113e-05, "loss": 1.0295, "step": 42815 }, { "epoch": 0.11, "learning_rate": 8.948009461975942e-05, "loss": 1.0271, "step": 42820 }, { "epoch": 0.11, "learning_rate": 8.947883637827772e-05, "loss": 1.0309, "step": 42825 }, { "epoch": 0.11, "learning_rate": 8.947757813679602e-05, "loss": 1.043, "step": 42830 }, { "epoch": 0.11, "learning_rate": 8.947631989531431e-05, "loss": 1.029, "step": 42835 }, { "epoch": 0.11, "learning_rate": 8.94750616538326e-05, "loss": 1.0306, "step": 42840 }, { "epoch": 0.11, "learning_rate": 8.94738034123509e-05, "loss": 1.0322, "step": 42845 }, { "epoch": 0.11, "learning_rate": 8.94725451708692e-05, "loss": 1.0276, "step": 42850 }, { "epoch": 0.11, "learning_rate": 8.947128692938749e-05, "loss": 1.031, "step": 42855 }, { "epoch": 0.11, "learning_rate": 8.947002868790578e-05, "loss": 1.0287, "step": 42860 }, { "epoch": 0.11, "learning_rate": 8.946877044642407e-05, "loss": 1.0501, "step": 42865 }, { "epoch": 0.11, "learning_rate": 8.946751220494238e-05, "loss": 1.0267, "step": 42870 }, { "epoch": 0.11, "learning_rate": 8.946625396346067e-05, "loss": 1.028, "step": 42875 }, { "epoch": 0.11, "learning_rate": 8.946499572197896e-05, "loss": 1.0285, "step": 42880 }, { "epoch": 0.11, "learning_rate": 8.946373748049725e-05, "loss": 1.0298, "step": 42885 }, { "epoch": 0.11, "learning_rate": 8.946247923901556e-05, "loss": 1.0303, "step": 42890 }, { "epoch": 0.11, "learning_rate": 8.946122099753385e-05, "loss": 1.0286, "step": 42895 }, { "epoch": 0.11, "learning_rate": 8.945996275605214e-05, "loss": 1.0295, "step": 42900 }, { "epoch": 0.11, "learning_rate": 8.945870451457044e-05, "loss": 1.0275, "step": 42905 }, { "epoch": 0.11, "learning_rate": 8.945744627308874e-05, "loss": 1.0281, "step": 42910 }, { "epoch": 0.11, "learning_rate": 8.945618803160704e-05, "loss": 1.027, "step": 42915 }, { "epoch": 0.11, "learning_rate": 8.945492979012533e-05, "loss": 1.0285, "step": 42920 }, { "epoch": 0.11, "learning_rate": 8.945367154864362e-05, "loss": 1.0288, "step": 42925 }, { "epoch": 0.11, "learning_rate": 8.945241330716192e-05, "loss": 1.0285, "step": 42930 }, { "epoch": 0.11, "learning_rate": 8.945115506568022e-05, "loss": 1.0307, "step": 42935 }, { "epoch": 0.11, "learning_rate": 8.944989682419851e-05, "loss": 1.0287, "step": 42940 }, { "epoch": 0.11, "learning_rate": 8.94486385827168e-05, "loss": 1.0286, "step": 42945 }, { "epoch": 0.11, "learning_rate": 8.94473803412351e-05, "loss": 1.0286, "step": 42950 }, { "epoch": 0.11, "learning_rate": 8.944612209975339e-05, "loss": 1.0293, "step": 42955 }, { "epoch": 0.11, "learning_rate": 8.944486385827169e-05, "loss": 1.0274, "step": 42960 }, { "epoch": 0.11, "learning_rate": 8.944360561678998e-05, "loss": 1.0282, "step": 42965 }, { "epoch": 0.11, "learning_rate": 8.944234737530828e-05, "loss": 1.0276, "step": 42970 }, { "epoch": 0.11, "learning_rate": 8.944108913382657e-05, "loss": 1.0297, "step": 42975 }, { "epoch": 0.11, "learning_rate": 8.943983089234487e-05, "loss": 1.0296, "step": 42980 }, { "epoch": 0.11, "learning_rate": 8.943857265086316e-05, "loss": 1.0297, "step": 42985 }, { "epoch": 0.11, "learning_rate": 8.943731440938146e-05, "loss": 1.0298, "step": 42990 }, { "epoch": 0.11, "learning_rate": 8.943605616789975e-05, "loss": 1.0495, "step": 42995 }, { "epoch": 0.11, "learning_rate": 8.943479792641805e-05, "loss": 1.0281, "step": 43000 }, { "epoch": 0.11, "learning_rate": 8.943353968493634e-05, "loss": 1.0312, "step": 43005 }, { "epoch": 0.11, "learning_rate": 8.943228144345464e-05, "loss": 1.0277, "step": 43010 }, { "epoch": 0.11, "learning_rate": 8.943102320197293e-05, "loss": 1.0294, "step": 43015 }, { "epoch": 0.11, "learning_rate": 8.942976496049122e-05, "loss": 1.0302, "step": 43020 }, { "epoch": 0.11, "learning_rate": 8.942850671900952e-05, "loss": 1.0291, "step": 43025 }, { "epoch": 0.11, "learning_rate": 8.942724847752782e-05, "loss": 1.0276, "step": 43030 }, { "epoch": 0.11, "learning_rate": 8.942599023604611e-05, "loss": 1.0276, "step": 43035 }, { "epoch": 0.11, "learning_rate": 8.94247319945644e-05, "loss": 1.0265, "step": 43040 }, { "epoch": 0.11, "learning_rate": 8.94234737530827e-05, "loss": 1.0286, "step": 43045 }, { "epoch": 0.11, "learning_rate": 8.9422215511601e-05, "loss": 1.0261, "step": 43050 }, { "epoch": 0.11, "learning_rate": 8.942095727011929e-05, "loss": 1.0305, "step": 43055 }, { "epoch": 0.11, "learning_rate": 8.941969902863758e-05, "loss": 1.0299, "step": 43060 }, { "epoch": 0.11, "learning_rate": 8.941844078715588e-05, "loss": 1.0284, "step": 43065 }, { "epoch": 0.11, "learning_rate": 8.941718254567418e-05, "loss": 1.0304, "step": 43070 }, { "epoch": 0.11, "learning_rate": 8.941592430419247e-05, "loss": 1.0306, "step": 43075 }, { "epoch": 0.11, "learning_rate": 8.941466606271076e-05, "loss": 1.0306, "step": 43080 }, { "epoch": 0.11, "learning_rate": 8.941340782122905e-05, "loss": 1.0296, "step": 43085 }, { "epoch": 0.11, "learning_rate": 8.941214957974736e-05, "loss": 1.0298, "step": 43090 }, { "epoch": 0.11, "learning_rate": 8.941089133826565e-05, "loss": 1.0298, "step": 43095 }, { "epoch": 0.11, "learning_rate": 8.940963309678394e-05, "loss": 1.0284, "step": 43100 }, { "epoch": 0.11, "learning_rate": 8.940837485530223e-05, "loss": 1.0511, "step": 43105 }, { "epoch": 0.11, "learning_rate": 8.940711661382054e-05, "loss": 1.0287, "step": 43110 }, { "epoch": 0.11, "learning_rate": 8.940585837233883e-05, "loss": 1.031, "step": 43115 }, { "epoch": 0.11, "learning_rate": 8.940460013085712e-05, "loss": 1.0291, "step": 43120 }, { "epoch": 0.11, "learning_rate": 8.940334188937541e-05, "loss": 1.0282, "step": 43125 }, { "epoch": 0.11, "learning_rate": 8.940208364789372e-05, "loss": 1.0282, "step": 43130 }, { "epoch": 0.11, "learning_rate": 8.9400825406412e-05, "loss": 1.0314, "step": 43135 }, { "epoch": 0.11, "learning_rate": 8.93995671649303e-05, "loss": 1.0282, "step": 43140 }, { "epoch": 0.11, "learning_rate": 8.939830892344859e-05, "loss": 1.0286, "step": 43145 }, { "epoch": 0.11, "learning_rate": 8.939705068196688e-05, "loss": 1.0298, "step": 43150 }, { "epoch": 0.11, "learning_rate": 8.939579244048519e-05, "loss": 1.0297, "step": 43155 }, { "epoch": 0.11, "learning_rate": 8.939453419900348e-05, "loss": 1.0291, "step": 43160 }, { "epoch": 0.11, "learning_rate": 8.939327595752177e-05, "loss": 1.0274, "step": 43165 }, { "epoch": 0.11, "learning_rate": 8.939201771604006e-05, "loss": 1.0273, "step": 43170 }, { "epoch": 0.11, "learning_rate": 8.939075947455837e-05, "loss": 1.0285, "step": 43175 }, { "epoch": 0.11, "learning_rate": 8.938950123307666e-05, "loss": 1.0331, "step": 43180 }, { "epoch": 0.11, "learning_rate": 8.938824299159495e-05, "loss": 1.0288, "step": 43185 }, { "epoch": 0.11, "learning_rate": 8.938698475011324e-05, "loss": 1.0298, "step": 43190 }, { "epoch": 0.11, "learning_rate": 8.938572650863155e-05, "loss": 1.0273, "step": 43195 }, { "epoch": 0.11, "learning_rate": 8.938446826714984e-05, "loss": 1.028, "step": 43200 }, { "epoch": 0.11, "learning_rate": 8.938321002566813e-05, "loss": 1.0271, "step": 43205 }, { "epoch": 0.11, "learning_rate": 8.938195178418642e-05, "loss": 1.0284, "step": 43210 }, { "epoch": 0.11, "learning_rate": 8.938069354270471e-05, "loss": 1.0294, "step": 43215 }, { "epoch": 0.11, "learning_rate": 8.937943530122302e-05, "loss": 1.0314, "step": 43220 }, { "epoch": 0.11, "learning_rate": 8.937817705974131e-05, "loss": 1.0283, "step": 43225 }, { "epoch": 0.11, "learning_rate": 8.93769188182596e-05, "loss": 1.0275, "step": 43230 }, { "epoch": 0.11, "learning_rate": 8.937566057677789e-05, "loss": 1.0289, "step": 43235 }, { "epoch": 0.11, "learning_rate": 8.93744023352962e-05, "loss": 1.0293, "step": 43240 }, { "epoch": 0.11, "learning_rate": 8.937314409381449e-05, "loss": 1.0301, "step": 43245 }, { "epoch": 0.11, "learning_rate": 8.937188585233278e-05, "loss": 1.0273, "step": 43250 }, { "epoch": 0.11, "learning_rate": 8.937062761085107e-05, "loss": 1.0284, "step": 43255 }, { "epoch": 0.11, "learning_rate": 8.936936936936938e-05, "loss": 1.0271, "step": 43260 }, { "epoch": 0.11, "learning_rate": 8.936811112788767e-05, "loss": 1.0279, "step": 43265 }, { "epoch": 0.11, "learning_rate": 8.936685288640596e-05, "loss": 1.0639, "step": 43270 }, { "epoch": 0.11, "learning_rate": 8.936559464492425e-05, "loss": 1.0267, "step": 43275 }, { "epoch": 0.11, "learning_rate": 8.936433640344254e-05, "loss": 1.0281, "step": 43280 }, { "epoch": 0.11, "learning_rate": 8.936307816196085e-05, "loss": 1.0298, "step": 43285 }, { "epoch": 0.11, "learning_rate": 8.936181992047914e-05, "loss": 1.0303, "step": 43290 }, { "epoch": 0.11, "learning_rate": 8.936056167899743e-05, "loss": 1.0287, "step": 43295 }, { "epoch": 0.11, "learning_rate": 8.935930343751572e-05, "loss": 1.0502, "step": 43300 }, { "epoch": 0.11, "learning_rate": 8.935804519603403e-05, "loss": 1.0295, "step": 43305 }, { "epoch": 0.11, "learning_rate": 8.935678695455232e-05, "loss": 1.0303, "step": 43310 }, { "epoch": 0.11, "learning_rate": 8.935552871307061e-05, "loss": 1.0313, "step": 43315 }, { "epoch": 0.11, "learning_rate": 8.93542704715889e-05, "loss": 1.0301, "step": 43320 }, { "epoch": 0.11, "learning_rate": 8.935301223010721e-05, "loss": 1.0274, "step": 43325 }, { "epoch": 0.11, "learning_rate": 8.93517539886255e-05, "loss": 1.0274, "step": 43330 }, { "epoch": 0.11, "learning_rate": 8.935049574714379e-05, "loss": 1.0297, "step": 43335 }, { "epoch": 0.11, "learning_rate": 8.934923750566208e-05, "loss": 1.0281, "step": 43340 }, { "epoch": 0.11, "learning_rate": 8.934797926418038e-05, "loss": 1.0251, "step": 43345 }, { "epoch": 0.11, "learning_rate": 8.934672102269868e-05, "loss": 1.0297, "step": 43350 }, { "epoch": 0.11, "learning_rate": 8.934546278121697e-05, "loss": 1.0272, "step": 43355 }, { "epoch": 0.11, "learning_rate": 8.934420453973526e-05, "loss": 1.0285, "step": 43360 }, { "epoch": 0.11, "learning_rate": 8.934294629825356e-05, "loss": 1.0268, "step": 43365 }, { "epoch": 0.11, "learning_rate": 8.934168805677186e-05, "loss": 1.0284, "step": 43370 }, { "epoch": 0.11, "learning_rate": 8.934042981529015e-05, "loss": 1.0284, "step": 43375 }, { "epoch": 0.11, "learning_rate": 8.933917157380844e-05, "loss": 1.0272, "step": 43380 }, { "epoch": 0.11, "learning_rate": 8.933791333232674e-05, "loss": 1.0306, "step": 43385 }, { "epoch": 0.11, "learning_rate": 8.933665509084503e-05, "loss": 1.029, "step": 43390 }, { "epoch": 0.11, "learning_rate": 8.933539684936333e-05, "loss": 1.0271, "step": 43395 }, { "epoch": 0.11, "learning_rate": 8.933413860788164e-05, "loss": 1.0284, "step": 43400 }, { "epoch": 0.11, "learning_rate": 8.933288036639993e-05, "loss": 1.0291, "step": 43405 }, { "epoch": 0.11, "learning_rate": 8.933162212491822e-05, "loss": 1.0283, "step": 43410 }, { "epoch": 0.11, "learning_rate": 8.933036388343651e-05, "loss": 1.0293, "step": 43415 }, { "epoch": 0.11, "learning_rate": 8.932910564195482e-05, "loss": 1.0265, "step": 43420 }, { "epoch": 0.11, "learning_rate": 8.932784740047311e-05, "loss": 1.0285, "step": 43425 }, { "epoch": 0.11, "learning_rate": 8.93265891589914e-05, "loss": 1.0296, "step": 43430 }, { "epoch": 0.11, "learning_rate": 8.932533091750969e-05, "loss": 1.0291, "step": 43435 }, { "epoch": 0.11, "learning_rate": 8.9324072676028e-05, "loss": 1.027, "step": 43440 }, { "epoch": 0.11, "learning_rate": 8.932281443454629e-05, "loss": 1.0275, "step": 43445 }, { "epoch": 0.11, "learning_rate": 8.932155619306458e-05, "loss": 1.0281, "step": 43450 }, { "epoch": 0.11, "learning_rate": 8.932029795158287e-05, "loss": 1.0286, "step": 43455 }, { "epoch": 0.11, "learning_rate": 8.931903971010118e-05, "loss": 1.0297, "step": 43460 }, { "epoch": 0.11, "learning_rate": 8.931778146861947e-05, "loss": 1.0292, "step": 43465 }, { "epoch": 0.11, "learning_rate": 8.931652322713776e-05, "loss": 1.0296, "step": 43470 }, { "epoch": 0.11, "learning_rate": 8.931526498565605e-05, "loss": 1.0289, "step": 43475 }, { "epoch": 0.11, "learning_rate": 8.931400674417434e-05, "loss": 1.0524, "step": 43480 }, { "epoch": 0.11, "learning_rate": 8.931274850269265e-05, "loss": 1.0289, "step": 43485 }, { "epoch": 0.11, "learning_rate": 8.931149026121094e-05, "loss": 1.0302, "step": 43490 }, { "epoch": 0.11, "learning_rate": 8.931023201972923e-05, "loss": 1.0304, "step": 43495 }, { "epoch": 0.11, "learning_rate": 8.930897377824752e-05, "loss": 1.0311, "step": 43500 }, { "epoch": 0.11, "learning_rate": 8.930771553676583e-05, "loss": 1.0481, "step": 43505 }, { "epoch": 0.11, "learning_rate": 8.930645729528412e-05, "loss": 1.0268, "step": 43510 }, { "epoch": 0.11, "learning_rate": 8.930519905380241e-05, "loss": 1.0293, "step": 43515 }, { "epoch": 0.11, "learning_rate": 8.93039408123207e-05, "loss": 1.0281, "step": 43520 }, { "epoch": 0.11, "learning_rate": 8.930268257083901e-05, "loss": 1.0292, "step": 43525 }, { "epoch": 0.11, "learning_rate": 8.93014243293573e-05, "loss": 1.0281, "step": 43530 }, { "epoch": 0.11, "learning_rate": 8.930016608787559e-05, "loss": 1.0285, "step": 43535 }, { "epoch": 0.11, "learning_rate": 8.929890784639388e-05, "loss": 1.0284, "step": 43540 }, { "epoch": 0.11, "learning_rate": 8.929764960491218e-05, "loss": 1.0286, "step": 43545 }, { "epoch": 0.11, "learning_rate": 8.929639136343048e-05, "loss": 1.0276, "step": 43550 }, { "epoch": 0.11, "learning_rate": 8.929513312194877e-05, "loss": 1.03, "step": 43555 }, { "epoch": 0.11, "learning_rate": 8.929387488046706e-05, "loss": 1.0276, "step": 43560 }, { "epoch": 0.11, "learning_rate": 8.929261663898536e-05, "loss": 1.0287, "step": 43565 }, { "epoch": 0.11, "learning_rate": 8.929135839750366e-05, "loss": 1.0303, "step": 43570 }, { "epoch": 0.11, "learning_rate": 8.929010015602195e-05, "loss": 1.0291, "step": 43575 }, { "epoch": 0.11, "learning_rate": 8.928884191454024e-05, "loss": 1.0274, "step": 43580 }, { "epoch": 0.11, "learning_rate": 8.928758367305854e-05, "loss": 1.0275, "step": 43585 }, { "epoch": 0.11, "learning_rate": 8.928632543157684e-05, "loss": 1.0292, "step": 43590 }, { "epoch": 0.11, "learning_rate": 8.928506719009513e-05, "loss": 1.0276, "step": 43595 }, { "epoch": 0.11, "learning_rate": 8.928380894861342e-05, "loss": 1.0265, "step": 43600 }, { "epoch": 0.11, "learning_rate": 8.928255070713171e-05, "loss": 1.0299, "step": 43605 }, { "epoch": 0.11, "learning_rate": 8.928129246565e-05, "loss": 1.028, "step": 43610 }, { "epoch": 0.11, "learning_rate": 8.928003422416831e-05, "loss": 1.0507, "step": 43615 }, { "epoch": 0.11, "learning_rate": 8.92787759826866e-05, "loss": 1.0304, "step": 43620 }, { "epoch": 0.11, "learning_rate": 8.92775177412049e-05, "loss": 1.0293, "step": 43625 }, { "epoch": 0.11, "learning_rate": 8.927625949972319e-05, "loss": 1.0275, "step": 43630 }, { "epoch": 0.11, "learning_rate": 8.927500125824149e-05, "loss": 1.0285, "step": 43635 }, { "epoch": 0.11, "learning_rate": 8.927374301675978e-05, "loss": 1.0269, "step": 43640 }, { "epoch": 0.11, "learning_rate": 8.927248477527807e-05, "loss": 1.0295, "step": 43645 }, { "epoch": 0.11, "learning_rate": 8.927122653379637e-05, "loss": 1.0326, "step": 43650 }, { "epoch": 0.11, "learning_rate": 8.926996829231467e-05, "loss": 1.0306, "step": 43655 }, { "epoch": 0.11, "learning_rate": 8.926871005083296e-05, "loss": 1.0276, "step": 43660 }, { "epoch": 0.11, "learning_rate": 8.926745180935125e-05, "loss": 1.0289, "step": 43665 }, { "epoch": 0.11, "learning_rate": 8.926619356786955e-05, "loss": 1.0271, "step": 43670 }, { "epoch": 0.11, "learning_rate": 8.926493532638784e-05, "loss": 1.0279, "step": 43675 }, { "epoch": 0.11, "learning_rate": 8.926367708490614e-05, "loss": 1.0264, "step": 43680 }, { "epoch": 0.11, "learning_rate": 8.926241884342443e-05, "loss": 1.0295, "step": 43685 }, { "epoch": 0.11, "learning_rate": 8.926116060194273e-05, "loss": 1.0296, "step": 43690 }, { "epoch": 0.11, "learning_rate": 8.925990236046102e-05, "loss": 1.029, "step": 43695 }, { "epoch": 0.11, "learning_rate": 8.925864411897932e-05, "loss": 1.0293, "step": 43700 }, { "epoch": 0.11, "learning_rate": 8.925738587749761e-05, "loss": 1.0305, "step": 43705 }, { "epoch": 0.11, "learning_rate": 8.92561276360159e-05, "loss": 1.0298, "step": 43710 }, { "epoch": 0.11, "learning_rate": 8.92548693945342e-05, "loss": 1.0282, "step": 43715 }, { "epoch": 0.11, "learning_rate": 8.92536111530525e-05, "loss": 1.0313, "step": 43720 }, { "epoch": 0.11, "learning_rate": 8.92523529115708e-05, "loss": 1.0281, "step": 43725 }, { "epoch": 0.11, "learning_rate": 8.925109467008909e-05, "loss": 1.0311, "step": 43730 }, { "epoch": 0.11, "learning_rate": 8.924983642860738e-05, "loss": 1.0299, "step": 43735 }, { "epoch": 0.11, "learning_rate": 8.924857818712567e-05, "loss": 1.0285, "step": 43740 }, { "epoch": 0.11, "learning_rate": 8.924731994564397e-05, "loss": 1.0289, "step": 43745 }, { "epoch": 0.11, "learning_rate": 8.924606170416227e-05, "loss": 1.0295, "step": 43750 }, { "epoch": 0.11, "learning_rate": 8.924480346268056e-05, "loss": 1.0268, "step": 43755 }, { "epoch": 0.11, "learning_rate": 8.924354522119885e-05, "loss": 1.0272, "step": 43760 }, { "epoch": 0.11, "learning_rate": 8.924228697971715e-05, "loss": 1.0281, "step": 43765 }, { "epoch": 0.11, "learning_rate": 8.924102873823545e-05, "loss": 1.0277, "step": 43770 }, { "epoch": 0.11, "learning_rate": 8.923977049675374e-05, "loss": 1.0291, "step": 43775 }, { "epoch": 0.11, "learning_rate": 8.923851225527203e-05, "loss": 1.0297, "step": 43780 }, { "epoch": 0.11, "learning_rate": 8.923725401379033e-05, "loss": 1.0273, "step": 43785 }, { "epoch": 0.11, "learning_rate": 8.923599577230863e-05, "loss": 1.0274, "step": 43790 }, { "epoch": 0.11, "learning_rate": 8.923473753082692e-05, "loss": 1.03, "step": 43795 }, { "epoch": 0.11, "learning_rate": 8.923347928934521e-05, "loss": 1.0292, "step": 43800 }, { "epoch": 0.11, "learning_rate": 8.92322210478635e-05, "loss": 1.0292, "step": 43805 }, { "epoch": 0.11, "learning_rate": 8.92309628063818e-05, "loss": 1.0502, "step": 43810 }, { "epoch": 0.11, "learning_rate": 8.92297045649001e-05, "loss": 1.0302, "step": 43815 }, { "epoch": 0.11, "learning_rate": 8.922844632341839e-05, "loss": 1.0279, "step": 43820 }, { "epoch": 0.11, "learning_rate": 8.922718808193668e-05, "loss": 1.0308, "step": 43825 }, { "epoch": 0.11, "learning_rate": 8.922592984045499e-05, "loss": 1.0289, "step": 43830 }, { "epoch": 0.11, "learning_rate": 8.922467159897328e-05, "loss": 1.0295, "step": 43835 }, { "epoch": 0.11, "learning_rate": 8.922341335749157e-05, "loss": 1.0312, "step": 43840 }, { "epoch": 0.11, "learning_rate": 8.922215511600986e-05, "loss": 1.0291, "step": 43845 }, { "epoch": 0.11, "learning_rate": 8.922089687452817e-05, "loss": 1.031, "step": 43850 }, { "epoch": 0.11, "learning_rate": 8.921963863304646e-05, "loss": 1.0326, "step": 43855 }, { "epoch": 0.11, "learning_rate": 8.921838039156475e-05, "loss": 1.0277, "step": 43860 }, { "epoch": 0.11, "learning_rate": 8.921712215008304e-05, "loss": 1.0299, "step": 43865 }, { "epoch": 0.11, "learning_rate": 8.921586390860133e-05, "loss": 1.03, "step": 43870 }, { "epoch": 0.11, "learning_rate": 8.921460566711964e-05, "loss": 1.028, "step": 43875 }, { "epoch": 0.11, "learning_rate": 8.921334742563793e-05, "loss": 1.0304, "step": 43880 }, { "epoch": 0.11, "learning_rate": 8.921208918415622e-05, "loss": 1.0294, "step": 43885 }, { "epoch": 0.11, "learning_rate": 8.921083094267451e-05, "loss": 1.0302, "step": 43890 }, { "epoch": 0.11, "learning_rate": 8.920957270119282e-05, "loss": 1.0278, "step": 43895 }, { "epoch": 0.11, "learning_rate": 8.920831445971112e-05, "loss": 1.05, "step": 43900 }, { "epoch": 0.11, "learning_rate": 8.920705621822941e-05, "loss": 1.0306, "step": 43905 }, { "epoch": 0.11, "learning_rate": 8.92057979767477e-05, "loss": 1.0292, "step": 43910 }, { "epoch": 0.11, "learning_rate": 8.9204539735266e-05, "loss": 1.0316, "step": 43915 }, { "epoch": 0.11, "learning_rate": 8.92032814937843e-05, "loss": 1.0294, "step": 43920 }, { "epoch": 0.11, "learning_rate": 8.92020232523026e-05, "loss": 1.0296, "step": 43925 }, { "epoch": 0.11, "learning_rate": 8.920076501082089e-05, "loss": 1.0284, "step": 43930 }, { "epoch": 0.11, "learning_rate": 8.919950676933918e-05, "loss": 1.0312, "step": 43935 }, { "epoch": 0.11, "learning_rate": 8.919824852785747e-05, "loss": 1.0303, "step": 43940 }, { "epoch": 0.11, "learning_rate": 8.919699028637577e-05, "loss": 1.0298, "step": 43945 }, { "epoch": 0.11, "learning_rate": 8.919573204489407e-05, "loss": 1.0298, "step": 43950 }, { "epoch": 0.11, "learning_rate": 8.919447380341236e-05, "loss": 1.0283, "step": 43955 }, { "epoch": 0.11, "learning_rate": 8.919321556193065e-05, "loss": 1.0291, "step": 43960 }, { "epoch": 0.11, "learning_rate": 8.919195732044895e-05, "loss": 1.0288, "step": 43965 }, { "epoch": 0.11, "learning_rate": 8.919069907896725e-05, "loss": 1.0277, "step": 43970 }, { "epoch": 0.11, "learning_rate": 8.918944083748554e-05, "loss": 1.0268, "step": 43975 }, { "epoch": 0.11, "learning_rate": 8.918818259600383e-05, "loss": 1.0307, "step": 43980 }, { "epoch": 0.11, "learning_rate": 8.918692435452213e-05, "loss": 1.0274, "step": 43985 }, { "epoch": 0.11, "learning_rate": 8.918566611304043e-05, "loss": 1.0297, "step": 43990 }, { "epoch": 0.11, "learning_rate": 8.918440787155872e-05, "loss": 1.0309, "step": 43995 }, { "epoch": 0.11, "learning_rate": 8.918314963007701e-05, "loss": 1.0271, "step": 44000 }, { "epoch": 0.11, "learning_rate": 8.91818913885953e-05, "loss": 1.0294, "step": 44005 }, { "epoch": 0.11, "learning_rate": 8.91806331471136e-05, "loss": 1.0279, "step": 44010 }, { "epoch": 0.11, "learning_rate": 8.91793749056319e-05, "loss": 1.026, "step": 44015 }, { "epoch": 0.11, "learning_rate": 8.917811666415019e-05, "loss": 1.0276, "step": 44020 }, { "epoch": 0.11, "learning_rate": 8.917685842266848e-05, "loss": 1.0308, "step": 44025 }, { "epoch": 0.11, "learning_rate": 8.917560018118679e-05, "loss": 1.0278, "step": 44030 }, { "epoch": 0.11, "learning_rate": 8.917434193970508e-05, "loss": 1.0284, "step": 44035 }, { "epoch": 0.11, "learning_rate": 8.917308369822337e-05, "loss": 1.0292, "step": 44040 }, { "epoch": 0.11, "learning_rate": 8.917182545674166e-05, "loss": 1.0287, "step": 44045 }, { "epoch": 0.11, "learning_rate": 8.917056721525997e-05, "loss": 1.03, "step": 44050 }, { "epoch": 0.11, "learning_rate": 8.916930897377826e-05, "loss": 1.03, "step": 44055 }, { "epoch": 0.11, "learning_rate": 8.916805073229655e-05, "loss": 1.0287, "step": 44060 }, { "epoch": 0.11, "learning_rate": 8.916679249081484e-05, "loss": 1.0289, "step": 44065 }, { "epoch": 0.11, "learning_rate": 8.916553424933313e-05, "loss": 1.0299, "step": 44070 }, { "epoch": 0.11, "learning_rate": 8.916427600785144e-05, "loss": 1.0286, "step": 44075 }, { "epoch": 0.11, "learning_rate": 8.916301776636973e-05, "loss": 1.0284, "step": 44080 }, { "epoch": 0.11, "learning_rate": 8.916175952488802e-05, "loss": 1.0297, "step": 44085 }, { "epoch": 0.11, "learning_rate": 8.916050128340631e-05, "loss": 1.0305, "step": 44090 }, { "epoch": 0.11, "learning_rate": 8.915924304192462e-05, "loss": 1.0267, "step": 44095 }, { "epoch": 0.11, "learning_rate": 8.915798480044291e-05, "loss": 1.0275, "step": 44100 }, { "epoch": 0.11, "learning_rate": 8.91567265589612e-05, "loss": 1.0288, "step": 44105 }, { "epoch": 0.11, "learning_rate": 8.915546831747949e-05, "loss": 1.0288, "step": 44110 }, { "epoch": 0.11, "learning_rate": 8.91542100759978e-05, "loss": 1.0292, "step": 44115 }, { "epoch": 0.11, "learning_rate": 8.915295183451609e-05, "loss": 1.028, "step": 44120 }, { "epoch": 0.11, "learning_rate": 8.915169359303438e-05, "loss": 1.0273, "step": 44125 }, { "epoch": 0.11, "learning_rate": 8.915043535155267e-05, "loss": 1.0294, "step": 44130 }, { "epoch": 0.11, "learning_rate": 8.914917711007096e-05, "loss": 1.028, "step": 44135 }, { "epoch": 0.11, "learning_rate": 8.914791886858927e-05, "loss": 1.0275, "step": 44140 }, { "epoch": 0.11, "learning_rate": 8.914666062710756e-05, "loss": 1.0284, "step": 44145 }, { "epoch": 0.11, "learning_rate": 8.914540238562585e-05, "loss": 1.0272, "step": 44150 }, { "epoch": 0.11, "learning_rate": 8.914414414414414e-05, "loss": 1.0304, "step": 44155 }, { "epoch": 0.11, "learning_rate": 8.914288590266245e-05, "loss": 1.0291, "step": 44160 }, { "epoch": 0.11, "learning_rate": 8.914162766118074e-05, "loss": 1.0274, "step": 44165 }, { "epoch": 0.11, "learning_rate": 8.914036941969903e-05, "loss": 1.0299, "step": 44170 }, { "epoch": 0.11, "learning_rate": 8.913911117821732e-05, "loss": 1.0286, "step": 44175 }, { "epoch": 0.11, "learning_rate": 8.913785293673563e-05, "loss": 1.0271, "step": 44180 }, { "epoch": 0.11, "learning_rate": 8.913659469525392e-05, "loss": 1.0296, "step": 44185 }, { "epoch": 0.11, "learning_rate": 8.913533645377221e-05, "loss": 1.0274, "step": 44190 }, { "epoch": 0.11, "learning_rate": 8.91340782122905e-05, "loss": 1.0283, "step": 44195 }, { "epoch": 0.11, "learning_rate": 8.91328199708088e-05, "loss": 1.029, "step": 44200 }, { "epoch": 0.11, "learning_rate": 8.91315617293271e-05, "loss": 1.0309, "step": 44205 }, { "epoch": 0.11, "learning_rate": 8.913030348784539e-05, "loss": 1.0299, "step": 44210 }, { "epoch": 0.11, "learning_rate": 8.912904524636368e-05, "loss": 1.0284, "step": 44215 }, { "epoch": 0.11, "learning_rate": 8.912778700488197e-05, "loss": 1.03, "step": 44220 }, { "epoch": 0.11, "learning_rate": 8.912652876340028e-05, "loss": 1.0281, "step": 44225 }, { "epoch": 0.11, "learning_rate": 8.912527052191857e-05, "loss": 1.0274, "step": 44230 }, { "epoch": 0.11, "learning_rate": 8.912401228043686e-05, "loss": 1.0299, "step": 44235 }, { "epoch": 0.11, "learning_rate": 8.912275403895515e-05, "loss": 1.03, "step": 44240 }, { "epoch": 0.11, "learning_rate": 8.912149579747346e-05, "loss": 1.0301, "step": 44245 }, { "epoch": 0.11, "learning_rate": 8.912023755599175e-05, "loss": 1.03, "step": 44250 }, { "epoch": 0.11, "learning_rate": 8.911897931451004e-05, "loss": 1.0307, "step": 44255 }, { "epoch": 0.11, "learning_rate": 8.911772107302833e-05, "loss": 1.0497, "step": 44260 }, { "epoch": 0.11, "learning_rate": 8.911646283154663e-05, "loss": 1.0282, "step": 44265 }, { "epoch": 0.11, "learning_rate": 8.911520459006493e-05, "loss": 1.0485, "step": 44270 }, { "epoch": 0.11, "learning_rate": 8.911394634858322e-05, "loss": 1.0267, "step": 44275 }, { "epoch": 0.11, "learning_rate": 8.911268810710151e-05, "loss": 1.0282, "step": 44280 }, { "epoch": 0.11, "learning_rate": 8.91114298656198e-05, "loss": 1.0274, "step": 44285 }, { "epoch": 0.11, "learning_rate": 8.911017162413811e-05, "loss": 1.0301, "step": 44290 }, { "epoch": 0.11, "learning_rate": 8.91089133826564e-05, "loss": 1.0297, "step": 44295 }, { "epoch": 0.11, "learning_rate": 8.91076551411747e-05, "loss": 1.0278, "step": 44300 }, { "epoch": 0.11, "learning_rate": 8.910639689969299e-05, "loss": 1.0291, "step": 44305 }, { "epoch": 0.11, "learning_rate": 8.910513865821129e-05, "loss": 1.0302, "step": 44310 }, { "epoch": 0.11, "learning_rate": 8.910388041672958e-05, "loss": 1.0308, "step": 44315 }, { "epoch": 0.11, "learning_rate": 8.910262217524787e-05, "loss": 1.0301, "step": 44320 }, { "epoch": 0.11, "learning_rate": 8.910136393376617e-05, "loss": 1.0293, "step": 44325 }, { "epoch": 0.11, "learning_rate": 8.910010569228446e-05, "loss": 1.0293, "step": 44330 }, { "epoch": 0.11, "learning_rate": 8.909884745080276e-05, "loss": 1.0508, "step": 44335 }, { "epoch": 0.11, "learning_rate": 8.909758920932105e-05, "loss": 1.0315, "step": 44340 }, { "epoch": 0.11, "learning_rate": 8.909633096783935e-05, "loss": 1.026, "step": 44345 }, { "epoch": 0.11, "learning_rate": 8.909507272635764e-05, "loss": 1.0301, "step": 44350 }, { "epoch": 0.11, "learning_rate": 8.909381448487594e-05, "loss": 1.0293, "step": 44355 }, { "epoch": 0.11, "learning_rate": 8.909255624339423e-05, "loss": 1.0272, "step": 44360 }, { "epoch": 0.11, "learning_rate": 8.909129800191253e-05, "loss": 1.0277, "step": 44365 }, { "epoch": 0.11, "learning_rate": 8.909003976043082e-05, "loss": 1.0296, "step": 44370 }, { "epoch": 0.11, "learning_rate": 8.908878151894912e-05, "loss": 1.0277, "step": 44375 }, { "epoch": 0.11, "learning_rate": 8.908752327746741e-05, "loss": 1.0305, "step": 44380 }, { "epoch": 0.11, "learning_rate": 8.90862650359857e-05, "loss": 1.028, "step": 44385 }, { "epoch": 0.11, "learning_rate": 8.9085006794504e-05, "loss": 1.0281, "step": 44390 }, { "epoch": 0.11, "learning_rate": 8.908374855302229e-05, "loss": 1.0298, "step": 44395 }, { "epoch": 0.11, "learning_rate": 8.90824903115406e-05, "loss": 1.0279, "step": 44400 }, { "epoch": 0.11, "learning_rate": 8.90812320700589e-05, "loss": 1.0282, "step": 44405 }, { "epoch": 0.11, "learning_rate": 8.907997382857719e-05, "loss": 1.0278, "step": 44410 }, { "epoch": 0.11, "learning_rate": 8.907871558709548e-05, "loss": 1.0265, "step": 44415 }, { "epoch": 0.11, "learning_rate": 8.907745734561377e-05, "loss": 1.0263, "step": 44420 }, { "epoch": 0.11, "learning_rate": 8.907619910413208e-05, "loss": 1.03, "step": 44425 }, { "epoch": 0.11, "learning_rate": 8.907494086265037e-05, "loss": 1.0282, "step": 44430 }, { "epoch": 0.11, "learning_rate": 8.907368262116866e-05, "loss": 1.0308, "step": 44435 }, { "epoch": 0.11, "learning_rate": 8.907242437968695e-05, "loss": 1.029, "step": 44440 }, { "epoch": 0.11, "learning_rate": 8.907116613820526e-05, "loss": 1.0324, "step": 44445 }, { "epoch": 0.11, "learning_rate": 8.906990789672355e-05, "loss": 1.0272, "step": 44450 }, { "epoch": 0.11, "learning_rate": 8.906864965524184e-05, "loss": 1.0272, "step": 44455 }, { "epoch": 0.11, "learning_rate": 8.906739141376013e-05, "loss": 1.029, "step": 44460 }, { "epoch": 0.11, "learning_rate": 8.906613317227843e-05, "loss": 1.0277, "step": 44465 }, { "epoch": 0.11, "learning_rate": 8.906487493079673e-05, "loss": 1.0282, "step": 44470 }, { "epoch": 0.11, "learning_rate": 8.906361668931502e-05, "loss": 1.0302, "step": 44475 }, { "epoch": 0.11, "learning_rate": 8.906235844783331e-05, "loss": 1.0278, "step": 44480 }, { "epoch": 0.11, "learning_rate": 8.90611002063516e-05, "loss": 1.0285, "step": 44485 }, { "epoch": 0.11, "learning_rate": 8.905984196486991e-05, "loss": 1.0293, "step": 44490 }, { "epoch": 0.11, "learning_rate": 8.90585837233882e-05, "loss": 1.0297, "step": 44495 }, { "epoch": 0.11, "learning_rate": 8.90573254819065e-05, "loss": 1.0275, "step": 44500 }, { "epoch": 0.11, "learning_rate": 8.905606724042479e-05, "loss": 1.0291, "step": 44505 }, { "epoch": 0.11, "learning_rate": 8.905480899894309e-05, "loss": 1.0293, "step": 44510 }, { "epoch": 0.11, "learning_rate": 8.905355075746138e-05, "loss": 1.0298, "step": 44515 }, { "epoch": 0.11, "learning_rate": 8.905229251597967e-05, "loss": 1.0275, "step": 44520 }, { "epoch": 0.11, "learning_rate": 8.905103427449797e-05, "loss": 1.0284, "step": 44525 }, { "epoch": 0.11, "learning_rate": 8.904977603301626e-05, "loss": 1.0277, "step": 44530 }, { "epoch": 0.11, "learning_rate": 8.904851779153456e-05, "loss": 1.0298, "step": 44535 }, { "epoch": 0.11, "learning_rate": 8.904725955005285e-05, "loss": 1.0304, "step": 44540 }, { "epoch": 0.11, "learning_rate": 8.904600130857115e-05, "loss": 1.0284, "step": 44545 }, { "epoch": 0.11, "learning_rate": 8.904474306708944e-05, "loss": 1.0302, "step": 44550 }, { "epoch": 0.11, "learning_rate": 8.904348482560774e-05, "loss": 1.0295, "step": 44555 }, { "epoch": 0.11, "learning_rate": 8.904222658412603e-05, "loss": 1.0294, "step": 44560 }, { "epoch": 0.11, "learning_rate": 8.904096834264433e-05, "loss": 1.0299, "step": 44565 }, { "epoch": 0.11, "learning_rate": 8.903971010116262e-05, "loss": 1.0273, "step": 44570 }, { "epoch": 0.11, "learning_rate": 8.903845185968092e-05, "loss": 1.0311, "step": 44575 }, { "epoch": 0.11, "learning_rate": 8.903719361819921e-05, "loss": 1.0293, "step": 44580 }, { "epoch": 0.11, "learning_rate": 8.90359353767175e-05, "loss": 1.0293, "step": 44585 }, { "epoch": 0.11, "learning_rate": 8.90346771352358e-05, "loss": 1.0282, "step": 44590 }, { "epoch": 0.11, "learning_rate": 8.903341889375409e-05, "loss": 1.0274, "step": 44595 }, { "epoch": 0.11, "learning_rate": 8.90321606522724e-05, "loss": 1.0277, "step": 44600 }, { "epoch": 0.11, "learning_rate": 8.903090241079068e-05, "loss": 1.0306, "step": 44605 }, { "epoch": 0.11, "learning_rate": 8.902964416930898e-05, "loss": 1.028, "step": 44610 }, { "epoch": 0.11, "learning_rate": 8.902838592782727e-05, "loss": 1.0292, "step": 44615 }, { "epoch": 0.11, "learning_rate": 8.902712768634557e-05, "loss": 1.0303, "step": 44620 }, { "epoch": 0.11, "learning_rate": 8.902586944486386e-05, "loss": 1.0277, "step": 44625 }, { "epoch": 0.11, "learning_rate": 8.902461120338216e-05, "loss": 1.0316, "step": 44630 }, { "epoch": 0.11, "learning_rate": 8.902335296190045e-05, "loss": 1.0273, "step": 44635 }, { "epoch": 0.11, "learning_rate": 8.902209472041875e-05, "loss": 1.0278, "step": 44640 }, { "epoch": 0.11, "learning_rate": 8.902083647893704e-05, "loss": 1.0305, "step": 44645 }, { "epoch": 0.11, "learning_rate": 8.901957823745534e-05, "loss": 1.0287, "step": 44650 }, { "epoch": 0.11, "learning_rate": 8.901831999597363e-05, "loss": 1.0284, "step": 44655 }, { "epoch": 0.11, "learning_rate": 8.901706175449192e-05, "loss": 1.0288, "step": 44660 }, { "epoch": 0.11, "learning_rate": 8.901580351301022e-05, "loss": 1.0282, "step": 44665 }, { "epoch": 0.11, "learning_rate": 8.901454527152852e-05, "loss": 1.0287, "step": 44670 }, { "epoch": 0.11, "learning_rate": 8.901328703004681e-05, "loss": 1.0262, "step": 44675 }, { "epoch": 0.11, "learning_rate": 8.90120287885651e-05, "loss": 1.0273, "step": 44680 }, { "epoch": 0.11, "learning_rate": 8.90107705470834e-05, "loss": 1.0292, "step": 44685 }, { "epoch": 0.11, "learning_rate": 8.90095123056017e-05, "loss": 1.0302, "step": 44690 }, { "epoch": 0.11, "learning_rate": 8.900825406411999e-05, "loss": 1.0301, "step": 44695 }, { "epoch": 0.11, "learning_rate": 8.900699582263828e-05, "loss": 1.0307, "step": 44700 }, { "epoch": 0.11, "learning_rate": 8.900573758115658e-05, "loss": 1.03, "step": 44705 }, { "epoch": 0.11, "learning_rate": 8.900447933967488e-05, "loss": 1.0256, "step": 44710 }, { "epoch": 0.11, "learning_rate": 8.900322109819317e-05, "loss": 1.0281, "step": 44715 }, { "epoch": 0.11, "learning_rate": 8.900196285671146e-05, "loss": 1.0281, "step": 44720 }, { "epoch": 0.11, "learning_rate": 8.900070461522975e-05, "loss": 1.0303, "step": 44725 }, { "epoch": 0.11, "learning_rate": 8.899944637374806e-05, "loss": 1.0457, "step": 44730 }, { "epoch": 0.11, "learning_rate": 8.899818813226635e-05, "loss": 1.0282, "step": 44735 }, { "epoch": 0.11, "learning_rate": 8.899692989078464e-05, "loss": 1.0277, "step": 44740 }, { "epoch": 0.11, "learning_rate": 8.899567164930293e-05, "loss": 1.0295, "step": 44745 }, { "epoch": 0.11, "learning_rate": 8.899441340782124e-05, "loss": 1.0279, "step": 44750 }, { "epoch": 0.11, "learning_rate": 8.899315516633953e-05, "loss": 1.0281, "step": 44755 }, { "epoch": 0.11, "learning_rate": 8.899189692485782e-05, "loss": 1.0296, "step": 44760 }, { "epoch": 0.11, "learning_rate": 8.899063868337611e-05, "loss": 1.0304, "step": 44765 }, { "epoch": 0.11, "learning_rate": 8.898938044189442e-05, "loss": 1.0291, "step": 44770 }, { "epoch": 0.11, "learning_rate": 8.898812220041271e-05, "loss": 1.028, "step": 44775 }, { "epoch": 0.11, "learning_rate": 8.8986863958931e-05, "loss": 1.028, "step": 44780 }, { "epoch": 0.11, "learning_rate": 8.898560571744929e-05, "loss": 1.0274, "step": 44785 }, { "epoch": 0.11, "learning_rate": 8.898434747596758e-05, "loss": 1.0262, "step": 44790 }, { "epoch": 0.11, "learning_rate": 8.898308923448589e-05, "loss": 1.0286, "step": 44795 }, { "epoch": 0.11, "learning_rate": 8.898183099300418e-05, "loss": 1.0295, "step": 44800 }, { "epoch": 0.11, "learning_rate": 8.898057275152247e-05, "loss": 1.0297, "step": 44805 }, { "epoch": 0.11, "learning_rate": 8.897931451004076e-05, "loss": 1.0284, "step": 44810 }, { "epoch": 0.11, "learning_rate": 8.897805626855907e-05, "loss": 1.0291, "step": 44815 }, { "epoch": 0.11, "learning_rate": 8.897679802707736e-05, "loss": 1.0305, "step": 44820 }, { "epoch": 0.11, "learning_rate": 8.897553978559565e-05, "loss": 1.0362, "step": 44825 }, { "epoch": 0.11, "learning_rate": 8.897428154411394e-05, "loss": 1.0279, "step": 44830 }, { "epoch": 0.11, "learning_rate": 8.897302330263225e-05, "loss": 1.0292, "step": 44835 }, { "epoch": 0.11, "learning_rate": 8.897176506115054e-05, "loss": 1.0292, "step": 44840 }, { "epoch": 0.11, "learning_rate": 8.897050681966883e-05, "loss": 1.0289, "step": 44845 }, { "epoch": 0.11, "learning_rate": 8.896924857818712e-05, "loss": 1.0304, "step": 44850 }, { "epoch": 0.11, "learning_rate": 8.896799033670541e-05, "loss": 1.027, "step": 44855 }, { "epoch": 0.11, "learning_rate": 8.896673209522372e-05, "loss": 1.029, "step": 44860 }, { "epoch": 0.11, "learning_rate": 8.896547385374201e-05, "loss": 1.0273, "step": 44865 }, { "epoch": 0.11, "learning_rate": 8.89642156122603e-05, "loss": 1.028, "step": 44870 }, { "epoch": 0.11, "learning_rate": 8.89629573707786e-05, "loss": 1.0287, "step": 44875 }, { "epoch": 0.11, "learning_rate": 8.89616991292969e-05, "loss": 1.0292, "step": 44880 }, { "epoch": 0.11, "learning_rate": 8.896044088781519e-05, "loss": 1.0278, "step": 44885 }, { "epoch": 0.11, "learning_rate": 8.895918264633348e-05, "loss": 1.0497, "step": 44890 }, { "epoch": 0.11, "learning_rate": 8.895792440485177e-05, "loss": 1.0522, "step": 44895 }, { "epoch": 0.11, "learning_rate": 8.895666616337008e-05, "loss": 1.0286, "step": 44900 }, { "epoch": 0.11, "learning_rate": 8.895540792188838e-05, "loss": 1.0262, "step": 44905 }, { "epoch": 0.11, "learning_rate": 8.895414968040668e-05, "loss": 1.0302, "step": 44910 }, { "epoch": 0.11, "learning_rate": 8.895289143892497e-05, "loss": 1.0297, "step": 44915 }, { "epoch": 0.11, "learning_rate": 8.895163319744326e-05, "loss": 1.0285, "step": 44920 }, { "epoch": 0.11, "learning_rate": 8.895037495596156e-05, "loss": 1.0288, "step": 44925 }, { "epoch": 0.11, "learning_rate": 8.894911671447986e-05, "loss": 1.0297, "step": 44930 }, { "epoch": 0.11, "learning_rate": 8.894785847299815e-05, "loss": 1.0283, "step": 44935 }, { "epoch": 0.11, "learning_rate": 8.894660023151644e-05, "loss": 1.0308, "step": 44940 }, { "epoch": 0.11, "learning_rate": 8.894534199003473e-05, "loss": 1.0275, "step": 44945 }, { "epoch": 0.11, "learning_rate": 8.894408374855304e-05, "loss": 1.0517, "step": 44950 }, { "epoch": 0.11, "learning_rate": 8.894282550707133e-05, "loss": 1.0295, "step": 44955 }, { "epoch": 0.11, "learning_rate": 8.894156726558962e-05, "loss": 1.0277, "step": 44960 }, { "epoch": 0.11, "learning_rate": 8.894030902410791e-05, "loss": 1.0284, "step": 44965 }, { "epoch": 0.11, "learning_rate": 8.893905078262622e-05, "loss": 1.0283, "step": 44970 }, { "epoch": 0.11, "learning_rate": 8.893779254114451e-05, "loss": 1.0474, "step": 44975 }, { "epoch": 0.11, "learning_rate": 8.89365342996628e-05, "loss": 1.0277, "step": 44980 }, { "epoch": 0.11, "learning_rate": 8.893527605818109e-05, "loss": 1.0298, "step": 44985 }, { "epoch": 0.11, "learning_rate": 8.89340178166994e-05, "loss": 1.0273, "step": 44990 }, { "epoch": 0.11, "learning_rate": 8.893275957521769e-05, "loss": 1.0301, "step": 44995 }, { "epoch": 0.11, "learning_rate": 8.893150133373598e-05, "loss": 1.0287, "step": 45000 }, { "epoch": 0.11, "learning_rate": 8.893024309225427e-05, "loss": 1.0297, "step": 45005 }, { "epoch": 0.11, "learning_rate": 8.892898485077256e-05, "loss": 1.0311, "step": 45010 }, { "epoch": 0.11, "learning_rate": 8.892772660929087e-05, "loss": 1.0311, "step": 45015 }, { "epoch": 0.11, "learning_rate": 8.892646836780916e-05, "loss": 1.0291, "step": 45020 }, { "epoch": 0.11, "learning_rate": 8.892521012632745e-05, "loss": 1.0262, "step": 45025 }, { "epoch": 0.11, "learning_rate": 8.892395188484574e-05, "loss": 1.0291, "step": 45030 }, { "epoch": 0.11, "learning_rate": 8.892269364336405e-05, "loss": 1.0302, "step": 45035 }, { "epoch": 0.11, "learning_rate": 8.892143540188234e-05, "loss": 1.0288, "step": 45040 }, { "epoch": 0.11, "learning_rate": 8.892017716040063e-05, "loss": 1.0284, "step": 45045 }, { "epoch": 0.11, "learning_rate": 8.891891891891892e-05, "loss": 1.0301, "step": 45050 }, { "epoch": 0.11, "learning_rate": 8.891766067743721e-05, "loss": 1.0279, "step": 45055 }, { "epoch": 0.11, "learning_rate": 8.891640243595552e-05, "loss": 1.0288, "step": 45060 }, { "epoch": 0.11, "learning_rate": 8.891514419447381e-05, "loss": 1.0292, "step": 45065 }, { "epoch": 0.11, "learning_rate": 8.89138859529921e-05, "loss": 1.0283, "step": 45070 }, { "epoch": 0.11, "learning_rate": 8.89126277115104e-05, "loss": 1.028, "step": 45075 }, { "epoch": 0.11, "learning_rate": 8.89113694700287e-05, "loss": 1.029, "step": 45080 }, { "epoch": 0.11, "learning_rate": 8.891011122854699e-05, "loss": 1.0283, "step": 45085 }, { "epoch": 0.11, "learning_rate": 8.890885298706528e-05, "loss": 1.0307, "step": 45090 }, { "epoch": 0.11, "learning_rate": 8.890759474558357e-05, "loss": 1.0285, "step": 45095 }, { "epoch": 0.11, "learning_rate": 8.890633650410188e-05, "loss": 1.0297, "step": 45100 }, { "epoch": 0.11, "learning_rate": 8.890507826262017e-05, "loss": 1.0283, "step": 45105 }, { "epoch": 0.11, "learning_rate": 8.890382002113846e-05, "loss": 1.03, "step": 45110 }, { "epoch": 0.11, "learning_rate": 8.890256177965675e-05, "loss": 1.0277, "step": 45115 }, { "epoch": 0.11, "learning_rate": 8.890130353817504e-05, "loss": 1.0294, "step": 45120 }, { "epoch": 0.11, "learning_rate": 8.890004529669335e-05, "loss": 1.0292, "step": 45125 }, { "epoch": 0.11, "learning_rate": 8.889878705521164e-05, "loss": 1.0273, "step": 45130 }, { "epoch": 0.11, "learning_rate": 8.889752881372993e-05, "loss": 1.0303, "step": 45135 }, { "epoch": 0.11, "learning_rate": 8.889627057224822e-05, "loss": 1.0255, "step": 45140 }, { "epoch": 0.11, "learning_rate": 8.889501233076653e-05, "loss": 1.0314, "step": 45145 }, { "epoch": 0.11, "learning_rate": 8.889375408928482e-05, "loss": 1.0481, "step": 45150 }, { "epoch": 0.11, "learning_rate": 8.889249584780311e-05, "loss": 1.0292, "step": 45155 }, { "epoch": 0.11, "learning_rate": 8.88912376063214e-05, "loss": 1.0295, "step": 45160 }, { "epoch": 0.11, "learning_rate": 8.888997936483971e-05, "loss": 1.0289, "step": 45165 }, { "epoch": 0.11, "learning_rate": 8.8888721123358e-05, "loss": 1.0313, "step": 45170 }, { "epoch": 0.11, "learning_rate": 8.888746288187629e-05, "loss": 1.0274, "step": 45175 }, { "epoch": 0.11, "learning_rate": 8.888620464039458e-05, "loss": 1.0302, "step": 45180 }, { "epoch": 0.11, "learning_rate": 8.888494639891288e-05, "loss": 1.0301, "step": 45185 }, { "epoch": 0.11, "learning_rate": 8.888368815743118e-05, "loss": 1.0315, "step": 45190 }, { "epoch": 0.11, "learning_rate": 8.888242991594947e-05, "loss": 1.0281, "step": 45195 }, { "epoch": 0.11, "learning_rate": 8.888117167446776e-05, "loss": 1.0298, "step": 45200 }, { "epoch": 0.11, "learning_rate": 8.887991343298606e-05, "loss": 1.0298, "step": 45205 }, { "epoch": 0.11, "learning_rate": 8.887865519150436e-05, "loss": 1.0292, "step": 45210 }, { "epoch": 0.11, "learning_rate": 8.887739695002265e-05, "loss": 1.0269, "step": 45215 }, { "epoch": 0.11, "learning_rate": 8.887613870854094e-05, "loss": 1.0273, "step": 45220 }, { "epoch": 0.11, "learning_rate": 8.887488046705924e-05, "loss": 1.0314, "step": 45225 }, { "epoch": 0.11, "learning_rate": 8.887362222557754e-05, "loss": 1.0299, "step": 45230 }, { "epoch": 0.11, "learning_rate": 8.887236398409583e-05, "loss": 1.0271, "step": 45235 }, { "epoch": 0.11, "learning_rate": 8.887110574261412e-05, "loss": 1.027, "step": 45240 }, { "epoch": 0.11, "learning_rate": 8.886984750113242e-05, "loss": 1.0295, "step": 45245 }, { "epoch": 0.11, "learning_rate": 8.886858925965071e-05, "loss": 1.0283, "step": 45250 }, { "epoch": 0.11, "learning_rate": 8.886733101816901e-05, "loss": 1.0281, "step": 45255 }, { "epoch": 0.11, "learning_rate": 8.88660727766873e-05, "loss": 1.0293, "step": 45260 }, { "epoch": 0.11, "learning_rate": 8.88648145352056e-05, "loss": 1.03, "step": 45265 }, { "epoch": 0.11, "learning_rate": 8.886355629372389e-05, "loss": 1.029, "step": 45270 }, { "epoch": 0.11, "learning_rate": 8.886229805224219e-05, "loss": 1.0302, "step": 45275 }, { "epoch": 0.11, "learning_rate": 8.886103981076048e-05, "loss": 1.0299, "step": 45280 }, { "epoch": 0.11, "learning_rate": 8.885978156927878e-05, "loss": 1.0286, "step": 45285 }, { "epoch": 0.11, "learning_rate": 8.885852332779707e-05, "loss": 1.0279, "step": 45290 }, { "epoch": 0.11, "learning_rate": 8.885726508631537e-05, "loss": 1.0263, "step": 45295 }, { "epoch": 0.11, "learning_rate": 8.885600684483366e-05, "loss": 1.0273, "step": 45300 }, { "epoch": 0.11, "learning_rate": 8.885474860335196e-05, "loss": 1.0299, "step": 45305 }, { "epoch": 0.11, "learning_rate": 8.885349036187025e-05, "loss": 1.0268, "step": 45310 }, { "epoch": 0.11, "learning_rate": 8.885223212038854e-05, "loss": 1.0279, "step": 45315 }, { "epoch": 0.11, "learning_rate": 8.885097387890684e-05, "loss": 1.0296, "step": 45320 }, { "epoch": 0.11, "learning_rate": 8.884971563742514e-05, "loss": 1.0292, "step": 45325 }, { "epoch": 0.11, "learning_rate": 8.884845739594343e-05, "loss": 1.0279, "step": 45330 }, { "epoch": 0.11, "learning_rate": 8.884719915446172e-05, "loss": 1.0294, "step": 45335 }, { "epoch": 0.11, "learning_rate": 8.884594091298002e-05, "loss": 1.0281, "step": 45340 }, { "epoch": 0.11, "learning_rate": 8.884468267149832e-05, "loss": 1.0282, "step": 45345 }, { "epoch": 0.11, "learning_rate": 8.884342443001661e-05, "loss": 1.0289, "step": 45350 }, { "epoch": 0.11, "learning_rate": 8.88421661885349e-05, "loss": 1.0296, "step": 45355 }, { "epoch": 0.11, "learning_rate": 8.88409079470532e-05, "loss": 1.0283, "step": 45360 }, { "epoch": 0.11, "learning_rate": 8.88396497055715e-05, "loss": 1.0283, "step": 45365 }, { "epoch": 0.11, "learning_rate": 8.883839146408979e-05, "loss": 1.0272, "step": 45370 }, { "epoch": 0.11, "learning_rate": 8.883713322260808e-05, "loss": 1.0268, "step": 45375 }, { "epoch": 0.11, "learning_rate": 8.883587498112637e-05, "loss": 1.028, "step": 45380 }, { "epoch": 0.11, "learning_rate": 8.883461673964468e-05, "loss": 1.029, "step": 45385 }, { "epoch": 0.11, "learning_rate": 8.883335849816297e-05, "loss": 1.0285, "step": 45390 }, { "epoch": 0.11, "learning_rate": 8.883210025668126e-05, "loss": 1.0285, "step": 45395 }, { "epoch": 0.11, "learning_rate": 8.883084201519956e-05, "loss": 1.0291, "step": 45400 }, { "epoch": 0.11, "learning_rate": 8.882958377371786e-05, "loss": 1.0281, "step": 45405 }, { "epoch": 0.11, "learning_rate": 8.882832553223616e-05, "loss": 1.0273, "step": 45410 }, { "epoch": 0.11, "learning_rate": 8.882706729075445e-05, "loss": 1.0278, "step": 45415 }, { "epoch": 0.11, "learning_rate": 8.882580904927274e-05, "loss": 1.0311, "step": 45420 }, { "epoch": 0.11, "learning_rate": 8.882455080779104e-05, "loss": 1.0277, "step": 45425 }, { "epoch": 0.11, "learning_rate": 8.882329256630934e-05, "loss": 1.0291, "step": 45430 }, { "epoch": 0.11, "learning_rate": 8.882203432482763e-05, "loss": 1.0295, "step": 45435 }, { "epoch": 0.11, "learning_rate": 8.882077608334592e-05, "loss": 1.0457, "step": 45440 }, { "epoch": 0.11, "learning_rate": 8.881951784186422e-05, "loss": 1.0257, "step": 45445 }, { "epoch": 0.11, "learning_rate": 8.881825960038252e-05, "loss": 1.0281, "step": 45450 }, { "epoch": 0.11, "learning_rate": 8.881700135890081e-05, "loss": 1.029, "step": 45455 }, { "epoch": 0.11, "learning_rate": 8.88157431174191e-05, "loss": 1.0294, "step": 45460 }, { "epoch": 0.11, "learning_rate": 8.88144848759374e-05, "loss": 1.0285, "step": 45465 }, { "epoch": 0.11, "learning_rate": 8.881322663445569e-05, "loss": 1.0296, "step": 45470 }, { "epoch": 0.11, "learning_rate": 8.881196839297399e-05, "loss": 1.0297, "step": 45475 }, { "epoch": 0.11, "learning_rate": 8.881071015149228e-05, "loss": 1.0291, "step": 45480 }, { "epoch": 0.11, "learning_rate": 8.880945191001058e-05, "loss": 1.0501, "step": 45485 }, { "epoch": 0.11, "learning_rate": 8.880819366852887e-05, "loss": 1.0289, "step": 45490 }, { "epoch": 0.11, "learning_rate": 8.880693542704717e-05, "loss": 1.0292, "step": 45495 }, { "epoch": 0.11, "learning_rate": 8.880567718556546e-05, "loss": 1.029, "step": 45500 }, { "epoch": 0.11, "learning_rate": 8.880441894408376e-05, "loss": 1.0298, "step": 45505 }, { "epoch": 0.11, "learning_rate": 8.880316070260205e-05, "loss": 1.026, "step": 45510 }, { "epoch": 0.11, "learning_rate": 8.880190246112035e-05, "loss": 1.0288, "step": 45515 }, { "epoch": 0.11, "learning_rate": 8.880064421963864e-05, "loss": 1.0294, "step": 45520 }, { "epoch": 0.11, "learning_rate": 8.879938597815694e-05, "loss": 1.0289, "step": 45525 }, { "epoch": 0.11, "learning_rate": 8.879812773667523e-05, "loss": 1.0256, "step": 45530 }, { "epoch": 0.11, "learning_rate": 8.879686949519352e-05, "loss": 1.0279, "step": 45535 }, { "epoch": 0.11, "learning_rate": 8.879561125371182e-05, "loss": 1.0302, "step": 45540 }, { "epoch": 0.11, "learning_rate": 8.879435301223012e-05, "loss": 1.028, "step": 45545 }, { "epoch": 0.11, "learning_rate": 8.87930947707484e-05, "loss": 1.0287, "step": 45550 }, { "epoch": 0.11, "learning_rate": 8.87918365292667e-05, "loss": 1.0291, "step": 45555 }, { "epoch": 0.11, "learning_rate": 8.8790578287785e-05, "loss": 1.0297, "step": 45560 }, { "epoch": 0.11, "learning_rate": 8.87893200463033e-05, "loss": 1.0288, "step": 45565 }, { "epoch": 0.11, "learning_rate": 8.878806180482159e-05, "loss": 1.0273, "step": 45570 }, { "epoch": 0.11, "learning_rate": 8.878680356333988e-05, "loss": 1.0266, "step": 45575 }, { "epoch": 0.11, "learning_rate": 8.878554532185818e-05, "loss": 1.028, "step": 45580 }, { "epoch": 0.11, "learning_rate": 8.878428708037648e-05, "loss": 1.0285, "step": 45585 }, { "epoch": 0.11, "learning_rate": 8.878302883889477e-05, "loss": 1.0275, "step": 45590 }, { "epoch": 0.11, "learning_rate": 8.878177059741306e-05, "loss": 1.029, "step": 45595 }, { "epoch": 0.11, "learning_rate": 8.878051235593135e-05, "loss": 1.0286, "step": 45600 }, { "epoch": 0.11, "learning_rate": 8.877925411444965e-05, "loss": 1.0296, "step": 45605 }, { "epoch": 0.11, "learning_rate": 8.877799587296795e-05, "loss": 1.0256, "step": 45610 }, { "epoch": 0.11, "learning_rate": 8.877673763148624e-05, "loss": 1.0287, "step": 45615 }, { "epoch": 0.11, "learning_rate": 8.877547939000453e-05, "loss": 1.0263, "step": 45620 }, { "epoch": 0.11, "learning_rate": 8.877422114852283e-05, "loss": 1.0302, "step": 45625 }, { "epoch": 0.11, "learning_rate": 8.877296290704113e-05, "loss": 1.0317, "step": 45630 }, { "epoch": 0.11, "learning_rate": 8.877170466555942e-05, "loss": 1.0283, "step": 45635 }, { "epoch": 0.11, "learning_rate": 8.877044642407771e-05, "loss": 1.0321, "step": 45640 }, { "epoch": 0.11, "learning_rate": 8.876918818259601e-05, "loss": 1.0292, "step": 45645 }, { "epoch": 0.11, "learning_rate": 8.87679299411143e-05, "loss": 1.0277, "step": 45650 }, { "epoch": 0.11, "learning_rate": 8.87666716996326e-05, "loss": 1.0293, "step": 45655 }, { "epoch": 0.11, "learning_rate": 8.876541345815089e-05, "loss": 1.0476, "step": 45660 }, { "epoch": 0.11, "learning_rate": 8.876415521666918e-05, "loss": 1.0321, "step": 45665 }, { "epoch": 0.11, "learning_rate": 8.876289697518749e-05, "loss": 1.0275, "step": 45670 }, { "epoch": 0.11, "learning_rate": 8.876163873370578e-05, "loss": 1.0279, "step": 45675 }, { "epoch": 0.11, "learning_rate": 8.876038049222407e-05, "loss": 1.0288, "step": 45680 }, { "epoch": 0.11, "learning_rate": 8.875912225074236e-05, "loss": 1.029, "step": 45685 }, { "epoch": 0.11, "learning_rate": 8.875786400926067e-05, "loss": 1.0299, "step": 45690 }, { "epoch": 0.11, "learning_rate": 8.875660576777896e-05, "loss": 1.0284, "step": 45695 }, { "epoch": 0.11, "learning_rate": 8.875534752629725e-05, "loss": 1.0288, "step": 45700 }, { "epoch": 0.11, "learning_rate": 8.875408928481554e-05, "loss": 1.0266, "step": 45705 }, { "epoch": 0.11, "learning_rate": 8.875283104333383e-05, "loss": 1.0273, "step": 45710 }, { "epoch": 0.11, "learning_rate": 8.875157280185214e-05, "loss": 1.0282, "step": 45715 }, { "epoch": 0.11, "learning_rate": 8.875031456037043e-05, "loss": 1.0275, "step": 45720 }, { "epoch": 0.11, "learning_rate": 8.874905631888872e-05, "loss": 1.0308, "step": 45725 }, { "epoch": 0.11, "learning_rate": 8.874779807740701e-05, "loss": 1.0281, "step": 45730 }, { "epoch": 0.11, "learning_rate": 8.874653983592532e-05, "loss": 1.0302, "step": 45735 }, { "epoch": 0.11, "learning_rate": 8.874528159444361e-05, "loss": 1.0274, "step": 45740 }, { "epoch": 0.11, "learning_rate": 8.87440233529619e-05, "loss": 1.0271, "step": 45745 }, { "epoch": 0.11, "learning_rate": 8.874276511148019e-05, "loss": 1.0281, "step": 45750 }, { "epoch": 0.11, "learning_rate": 8.87415068699985e-05, "loss": 1.0293, "step": 45755 }, { "epoch": 0.11, "learning_rate": 8.874024862851679e-05, "loss": 1.0291, "step": 45760 }, { "epoch": 0.11, "learning_rate": 8.873899038703508e-05, "loss": 1.0276, "step": 45765 }, { "epoch": 0.11, "learning_rate": 8.873773214555337e-05, "loss": 1.0291, "step": 45770 }, { "epoch": 0.11, "learning_rate": 8.873647390407166e-05, "loss": 1.0286, "step": 45775 }, { "epoch": 0.11, "learning_rate": 8.873521566258997e-05, "loss": 1.0303, "step": 45780 }, { "epoch": 0.11, "learning_rate": 8.873395742110826e-05, "loss": 1.0298, "step": 45785 }, { "epoch": 0.11, "learning_rate": 8.873269917962655e-05, "loss": 1.0283, "step": 45790 }, { "epoch": 0.11, "learning_rate": 8.873144093814484e-05, "loss": 1.0278, "step": 45795 }, { "epoch": 0.11, "learning_rate": 8.873018269666315e-05, "loss": 1.0286, "step": 45800 }, { "epoch": 0.11, "learning_rate": 8.872892445518144e-05, "loss": 1.0283, "step": 45805 }, { "epoch": 0.11, "learning_rate": 8.872766621369973e-05, "loss": 1.0267, "step": 45810 }, { "epoch": 0.12, "learning_rate": 8.872640797221802e-05, "loss": 1.0285, "step": 45815 }, { "epoch": 0.12, "learning_rate": 8.872514973073633e-05, "loss": 1.0273, "step": 45820 }, { "epoch": 0.12, "learning_rate": 8.872389148925462e-05, "loss": 1.0284, "step": 45825 }, { "epoch": 0.12, "learning_rate": 8.872263324777291e-05, "loss": 1.0289, "step": 45830 }, { "epoch": 0.12, "learning_rate": 8.87213750062912e-05, "loss": 1.0286, "step": 45835 }, { "epoch": 0.12, "learning_rate": 8.87201167648095e-05, "loss": 1.0299, "step": 45840 }, { "epoch": 0.12, "learning_rate": 8.87188585233278e-05, "loss": 1.0293, "step": 45845 }, { "epoch": 0.12, "learning_rate": 8.871760028184609e-05, "loss": 1.029, "step": 45850 }, { "epoch": 0.12, "learning_rate": 8.871634204036438e-05, "loss": 1.0271, "step": 45855 }, { "epoch": 0.12, "learning_rate": 8.871508379888268e-05, "loss": 1.0454, "step": 45860 }, { "epoch": 0.12, "learning_rate": 8.871382555740098e-05, "loss": 1.0301, "step": 45865 }, { "epoch": 0.12, "learning_rate": 8.871256731591927e-05, "loss": 1.0287, "step": 45870 }, { "epoch": 0.12, "learning_rate": 8.871130907443756e-05, "loss": 1.028, "step": 45875 }, { "epoch": 0.12, "learning_rate": 8.871005083295586e-05, "loss": 1.0277, "step": 45880 }, { "epoch": 0.12, "learning_rate": 8.870879259147416e-05, "loss": 1.029, "step": 45885 }, { "epoch": 0.12, "learning_rate": 8.870753434999245e-05, "loss": 1.0273, "step": 45890 }, { "epoch": 0.12, "learning_rate": 8.870627610851074e-05, "loss": 1.0274, "step": 45895 }, { "epoch": 0.12, "learning_rate": 8.870501786702905e-05, "loss": 1.0297, "step": 45900 }, { "epoch": 0.12, "learning_rate": 8.870375962554734e-05, "loss": 1.0302, "step": 45905 }, { "epoch": 0.12, "learning_rate": 8.870250138406565e-05, "loss": 1.026, "step": 45910 }, { "epoch": 0.12, "learning_rate": 8.870124314258394e-05, "loss": 1.0288, "step": 45915 }, { "epoch": 0.12, "learning_rate": 8.869998490110223e-05, "loss": 1.0284, "step": 45920 }, { "epoch": 0.12, "learning_rate": 8.869872665962052e-05, "loss": 1.0521, "step": 45925 }, { "epoch": 0.12, "learning_rate": 8.869746841813881e-05, "loss": 1.0278, "step": 45930 }, { "epoch": 0.12, "learning_rate": 8.869621017665712e-05, "loss": 1.0282, "step": 45935 }, { "epoch": 0.12, "learning_rate": 8.869495193517541e-05, "loss": 1.0292, "step": 45940 }, { "epoch": 0.12, "learning_rate": 8.86936936936937e-05, "loss": 1.0292, "step": 45945 }, { "epoch": 0.12, "learning_rate": 8.869243545221199e-05, "loss": 1.0293, "step": 45950 }, { "epoch": 0.12, "learning_rate": 8.86911772107303e-05, "loss": 1.0315, "step": 45955 }, { "epoch": 0.12, "learning_rate": 8.868991896924859e-05, "loss": 1.0291, "step": 45960 }, { "epoch": 0.12, "learning_rate": 8.868866072776688e-05, "loss": 1.0296, "step": 45965 }, { "epoch": 0.12, "learning_rate": 8.868740248628517e-05, "loss": 1.0276, "step": 45970 }, { "epoch": 0.12, "learning_rate": 8.868614424480348e-05, "loss": 1.0285, "step": 45975 }, { "epoch": 0.12, "learning_rate": 8.868488600332177e-05, "loss": 1.0282, "step": 45980 }, { "epoch": 0.12, "learning_rate": 8.868362776184006e-05, "loss": 1.029, "step": 45985 }, { "epoch": 0.12, "learning_rate": 8.868236952035835e-05, "loss": 1.0282, "step": 45990 }, { "epoch": 0.12, "learning_rate": 8.868111127887664e-05, "loss": 1.0301, "step": 45995 }, { "epoch": 0.12, "learning_rate": 8.867985303739495e-05, "loss": 1.0266, "step": 46000 }, { "epoch": 0.12, "learning_rate": 8.867859479591324e-05, "loss": 1.029, "step": 46005 }, { "epoch": 0.12, "learning_rate": 8.867733655443153e-05, "loss": 1.0281, "step": 46010 }, { "epoch": 0.12, "learning_rate": 8.867607831294982e-05, "loss": 1.0286, "step": 46015 }, { "epoch": 0.12, "learning_rate": 8.867482007146813e-05, "loss": 1.0277, "step": 46020 }, { "epoch": 0.12, "learning_rate": 8.867356182998642e-05, "loss": 1.0296, "step": 46025 }, { "epoch": 0.12, "learning_rate": 8.867230358850471e-05, "loss": 1.0299, "step": 46030 }, { "epoch": 0.12, "learning_rate": 8.8671045347023e-05, "loss": 1.0283, "step": 46035 }, { "epoch": 0.12, "learning_rate": 8.866978710554131e-05, "loss": 1.0292, "step": 46040 }, { "epoch": 0.12, "learning_rate": 8.86685288640596e-05, "loss": 1.027, "step": 46045 }, { "epoch": 0.12, "learning_rate": 8.866727062257789e-05, "loss": 1.0297, "step": 46050 }, { "epoch": 0.12, "learning_rate": 8.866601238109618e-05, "loss": 1.0311, "step": 46055 }, { "epoch": 0.12, "learning_rate": 8.866475413961447e-05, "loss": 1.0283, "step": 46060 }, { "epoch": 0.12, "learning_rate": 8.866349589813278e-05, "loss": 1.0291, "step": 46065 }, { "epoch": 0.12, "learning_rate": 8.866223765665107e-05, "loss": 1.0296, "step": 46070 }, { "epoch": 0.12, "learning_rate": 8.866097941516936e-05, "loss": 1.0284, "step": 46075 }, { "epoch": 0.12, "learning_rate": 8.865972117368765e-05, "loss": 1.0267, "step": 46080 }, { "epoch": 0.12, "learning_rate": 8.865846293220596e-05, "loss": 1.0291, "step": 46085 }, { "epoch": 0.12, "learning_rate": 8.865720469072425e-05, "loss": 1.026, "step": 46090 }, { "epoch": 0.12, "learning_rate": 8.865594644924254e-05, "loss": 1.028, "step": 46095 }, { "epoch": 0.12, "learning_rate": 8.865468820776083e-05, "loss": 1.0296, "step": 46100 }, { "epoch": 0.12, "learning_rate": 8.865342996627914e-05, "loss": 1.0284, "step": 46105 }, { "epoch": 0.12, "learning_rate": 8.865217172479743e-05, "loss": 1.0265, "step": 46110 }, { "epoch": 0.12, "learning_rate": 8.865091348331572e-05, "loss": 1.0289, "step": 46115 }, { "epoch": 0.12, "learning_rate": 8.864965524183401e-05, "loss": 1.0293, "step": 46120 }, { "epoch": 0.12, "learning_rate": 8.86483970003523e-05, "loss": 1.0302, "step": 46125 }, { "epoch": 0.12, "learning_rate": 8.864713875887061e-05, "loss": 1.0288, "step": 46130 }, { "epoch": 0.12, "learning_rate": 8.86458805173889e-05, "loss": 1.0307, "step": 46135 }, { "epoch": 0.12, "learning_rate": 8.86446222759072e-05, "loss": 1.0281, "step": 46140 }, { "epoch": 0.12, "learning_rate": 8.864336403442549e-05, "loss": 1.0293, "step": 46145 }, { "epoch": 0.12, "learning_rate": 8.864210579294379e-05, "loss": 1.0294, "step": 46150 }, { "epoch": 0.12, "learning_rate": 8.864084755146208e-05, "loss": 1.0282, "step": 46155 }, { "epoch": 0.12, "learning_rate": 8.863958930998037e-05, "loss": 1.029, "step": 46160 }, { "epoch": 0.12, "learning_rate": 8.863833106849867e-05, "loss": 1.0289, "step": 46165 }, { "epoch": 0.12, "learning_rate": 8.863707282701697e-05, "loss": 1.0288, "step": 46170 }, { "epoch": 0.12, "learning_rate": 8.863581458553526e-05, "loss": 1.0276, "step": 46175 }, { "epoch": 0.12, "learning_rate": 8.863455634405355e-05, "loss": 1.0288, "step": 46180 }, { "epoch": 0.12, "learning_rate": 8.863329810257185e-05, "loss": 1.0272, "step": 46185 }, { "epoch": 0.12, "learning_rate": 8.863203986109014e-05, "loss": 1.0302, "step": 46190 }, { "epoch": 0.12, "learning_rate": 8.863078161960844e-05, "loss": 1.0284, "step": 46195 }, { "epoch": 0.12, "learning_rate": 8.862952337812673e-05, "loss": 1.0279, "step": 46200 }, { "epoch": 0.12, "learning_rate": 8.862826513664503e-05, "loss": 1.0276, "step": 46205 }, { "epoch": 0.12, "learning_rate": 8.862700689516332e-05, "loss": 1.0303, "step": 46210 }, { "epoch": 0.12, "learning_rate": 8.862574865368162e-05, "loss": 1.0293, "step": 46215 }, { "epoch": 0.12, "learning_rate": 8.862449041219991e-05, "loss": 1.0259, "step": 46220 }, { "epoch": 0.12, "learning_rate": 8.86232321707182e-05, "loss": 1.0272, "step": 46225 }, { "epoch": 0.12, "learning_rate": 8.86219739292365e-05, "loss": 1.0283, "step": 46230 }, { "epoch": 0.12, "learning_rate": 8.86207156877548e-05, "loss": 1.0276, "step": 46235 }, { "epoch": 0.12, "learning_rate": 8.86194574462731e-05, "loss": 1.0279, "step": 46240 }, { "epoch": 0.12, "learning_rate": 8.861819920479139e-05, "loss": 1.0262, "step": 46245 }, { "epoch": 0.12, "learning_rate": 8.861694096330968e-05, "loss": 1.0284, "step": 46250 }, { "epoch": 0.12, "learning_rate": 8.861568272182797e-05, "loss": 1.0284, "step": 46255 }, { "epoch": 0.12, "learning_rate": 8.861442448034627e-05, "loss": 1.0301, "step": 46260 }, { "epoch": 0.12, "learning_rate": 8.861316623886457e-05, "loss": 1.0521, "step": 46265 }, { "epoch": 0.12, "learning_rate": 8.861190799738286e-05, "loss": 1.0269, "step": 46270 }, { "epoch": 0.12, "learning_rate": 8.861064975590115e-05, "loss": 1.03, "step": 46275 }, { "epoch": 0.12, "learning_rate": 8.860939151441945e-05, "loss": 1.0283, "step": 46280 }, { "epoch": 0.12, "learning_rate": 8.860813327293775e-05, "loss": 1.0283, "step": 46285 }, { "epoch": 0.12, "learning_rate": 8.860687503145604e-05, "loss": 1.0258, "step": 46290 }, { "epoch": 0.12, "learning_rate": 8.860561678997433e-05, "loss": 1.0279, "step": 46295 }, { "epoch": 0.12, "learning_rate": 8.860435854849263e-05, "loss": 1.0306, "step": 46300 }, { "epoch": 0.12, "learning_rate": 8.860310030701093e-05, "loss": 1.0283, "step": 46305 }, { "epoch": 0.12, "learning_rate": 8.860184206552922e-05, "loss": 1.0285, "step": 46310 }, { "epoch": 0.12, "learning_rate": 8.860058382404751e-05, "loss": 1.0262, "step": 46315 }, { "epoch": 0.12, "learning_rate": 8.85993255825658e-05, "loss": 1.028, "step": 46320 }, { "epoch": 0.12, "learning_rate": 8.85980673410841e-05, "loss": 1.0263, "step": 46325 }, { "epoch": 0.12, "learning_rate": 8.85968090996024e-05, "loss": 1.0282, "step": 46330 }, { "epoch": 0.12, "learning_rate": 8.859555085812069e-05, "loss": 1.0272, "step": 46335 }, { "epoch": 0.12, "learning_rate": 8.859429261663898e-05, "loss": 1.0292, "step": 46340 }, { "epoch": 0.12, "learning_rate": 8.859303437515729e-05, "loss": 1.0256, "step": 46345 }, { "epoch": 0.12, "learning_rate": 8.859177613367558e-05, "loss": 1.0299, "step": 46350 }, { "epoch": 0.12, "learning_rate": 8.859051789219387e-05, "loss": 1.0284, "step": 46355 }, { "epoch": 0.12, "learning_rate": 8.858925965071216e-05, "loss": 1.026, "step": 46360 }, { "epoch": 0.12, "learning_rate": 8.858800140923045e-05, "loss": 1.0267, "step": 46365 }, { "epoch": 0.12, "learning_rate": 8.858674316774876e-05, "loss": 1.0289, "step": 46370 }, { "epoch": 0.12, "learning_rate": 8.858548492626705e-05, "loss": 1.0286, "step": 46375 }, { "epoch": 0.12, "learning_rate": 8.858422668478534e-05, "loss": 1.0275, "step": 46380 }, { "epoch": 0.12, "learning_rate": 8.858296844330363e-05, "loss": 1.0304, "step": 46385 }, { "epoch": 0.12, "learning_rate": 8.858171020182194e-05, "loss": 1.0287, "step": 46390 }, { "epoch": 0.12, "learning_rate": 8.858045196034023e-05, "loss": 1.03, "step": 46395 }, { "epoch": 0.12, "learning_rate": 8.857919371885853e-05, "loss": 1.0293, "step": 46400 }, { "epoch": 0.12, "learning_rate": 8.857793547737683e-05, "loss": 1.0292, "step": 46405 }, { "epoch": 0.12, "learning_rate": 8.857667723589512e-05, "loss": 1.0289, "step": 46410 }, { "epoch": 0.12, "learning_rate": 8.857541899441342e-05, "loss": 1.028, "step": 46415 }, { "epoch": 0.12, "learning_rate": 8.857416075293171e-05, "loss": 1.0268, "step": 46420 }, { "epoch": 0.12, "learning_rate": 8.857290251145e-05, "loss": 1.0282, "step": 46425 }, { "epoch": 0.12, "learning_rate": 8.85716442699683e-05, "loss": 1.0308, "step": 46430 }, { "epoch": 0.12, "learning_rate": 8.85703860284866e-05, "loss": 1.0267, "step": 46435 }, { "epoch": 0.12, "learning_rate": 8.85691277870049e-05, "loss": 1.0269, "step": 46440 }, { "epoch": 0.12, "learning_rate": 8.856786954552319e-05, "loss": 1.0312, "step": 46445 }, { "epoch": 0.12, "learning_rate": 8.856661130404148e-05, "loss": 1.0279, "step": 46450 }, { "epoch": 0.12, "learning_rate": 8.856535306255977e-05, "loss": 1.029, "step": 46455 }, { "epoch": 0.12, "learning_rate": 8.856409482107807e-05, "loss": 1.0289, "step": 46460 }, { "epoch": 0.12, "learning_rate": 8.856283657959637e-05, "loss": 1.029, "step": 46465 }, { "epoch": 0.12, "learning_rate": 8.856157833811466e-05, "loss": 1.028, "step": 46470 }, { "epoch": 0.12, "learning_rate": 8.856032009663295e-05, "loss": 1.0272, "step": 46475 }, { "epoch": 0.12, "learning_rate": 8.855906185515125e-05, "loss": 1.03, "step": 46480 }, { "epoch": 0.12, "learning_rate": 8.855780361366955e-05, "loss": 1.0268, "step": 46485 }, { "epoch": 0.12, "learning_rate": 8.855654537218784e-05, "loss": 1.0293, "step": 46490 }, { "epoch": 0.12, "learning_rate": 8.855528713070613e-05, "loss": 1.0298, "step": 46495 }, { "epoch": 0.12, "learning_rate": 8.855402888922443e-05, "loss": 1.0264, "step": 46500 }, { "epoch": 0.12, "learning_rate": 8.855277064774273e-05, "loss": 1.0297, "step": 46505 }, { "epoch": 0.12, "learning_rate": 8.855151240626102e-05, "loss": 1.032, "step": 46510 }, { "epoch": 0.12, "learning_rate": 8.855025416477931e-05, "loss": 1.028, "step": 46515 }, { "epoch": 0.12, "learning_rate": 8.85489959232976e-05, "loss": 1.0269, "step": 46520 }, { "epoch": 0.12, "learning_rate": 8.85477376818159e-05, "loss": 1.0277, "step": 46525 }, { "epoch": 0.12, "learning_rate": 8.85464794403342e-05, "loss": 1.0266, "step": 46530 }, { "epoch": 0.12, "learning_rate": 8.854522119885249e-05, "loss": 1.0291, "step": 46535 }, { "epoch": 0.12, "learning_rate": 8.854396295737078e-05, "loss": 1.0273, "step": 46540 }, { "epoch": 0.12, "learning_rate": 8.854270471588909e-05, "loss": 1.0281, "step": 46545 }, { "epoch": 0.12, "learning_rate": 8.854144647440738e-05, "loss": 1.0294, "step": 46550 }, { "epoch": 0.12, "learning_rate": 8.854018823292567e-05, "loss": 1.028, "step": 46555 }, { "epoch": 0.12, "learning_rate": 8.853892999144396e-05, "loss": 1.0291, "step": 46560 }, { "epoch": 0.12, "learning_rate": 8.853767174996227e-05, "loss": 1.0271, "step": 46565 }, { "epoch": 0.12, "learning_rate": 8.853641350848056e-05, "loss": 1.0285, "step": 46570 }, { "epoch": 0.12, "learning_rate": 8.853515526699885e-05, "loss": 1.0291, "step": 46575 }, { "epoch": 0.12, "learning_rate": 8.853389702551714e-05, "loss": 1.0293, "step": 46580 }, { "epoch": 0.12, "learning_rate": 8.853263878403543e-05, "loss": 1.0284, "step": 46585 }, { "epoch": 0.12, "learning_rate": 8.853138054255374e-05, "loss": 1.0297, "step": 46590 }, { "epoch": 0.12, "learning_rate": 8.853012230107203e-05, "loss": 1.0278, "step": 46595 }, { "epoch": 0.12, "learning_rate": 8.852886405959032e-05, "loss": 1.0279, "step": 46600 }, { "epoch": 0.12, "learning_rate": 8.852760581810861e-05, "loss": 1.0281, "step": 46605 }, { "epoch": 0.12, "learning_rate": 8.852634757662692e-05, "loss": 1.0285, "step": 46610 }, { "epoch": 0.12, "learning_rate": 8.852508933514521e-05, "loss": 1.0274, "step": 46615 }, { "epoch": 0.12, "learning_rate": 8.85238310936635e-05, "loss": 1.0273, "step": 46620 }, { "epoch": 0.12, "learning_rate": 8.852257285218179e-05, "loss": 1.029, "step": 46625 }, { "epoch": 0.12, "learning_rate": 8.85213146107001e-05, "loss": 1.0282, "step": 46630 }, { "epoch": 0.12, "learning_rate": 8.852005636921839e-05, "loss": 1.0259, "step": 46635 }, { "epoch": 0.12, "learning_rate": 8.851879812773668e-05, "loss": 1.029, "step": 46640 }, { "epoch": 0.12, "learning_rate": 8.851753988625497e-05, "loss": 1.0302, "step": 46645 }, { "epoch": 0.12, "learning_rate": 8.851628164477326e-05, "loss": 1.0288, "step": 46650 }, { "epoch": 0.12, "learning_rate": 8.851502340329157e-05, "loss": 1.0298, "step": 46655 }, { "epoch": 0.12, "learning_rate": 8.851376516180986e-05, "loss": 1.0285, "step": 46660 }, { "epoch": 0.12, "learning_rate": 8.851250692032815e-05, "loss": 1.0285, "step": 46665 }, { "epoch": 0.12, "learning_rate": 8.851124867884644e-05, "loss": 1.0267, "step": 46670 }, { "epoch": 0.12, "learning_rate": 8.850999043736475e-05, "loss": 1.026, "step": 46675 }, { "epoch": 0.12, "learning_rate": 8.850873219588304e-05, "loss": 1.0275, "step": 46680 }, { "epoch": 0.12, "learning_rate": 8.850747395440133e-05, "loss": 1.0275, "step": 46685 }, { "epoch": 0.12, "learning_rate": 8.850621571291962e-05, "loss": 1.0279, "step": 46690 }, { "epoch": 0.12, "learning_rate": 8.850495747143793e-05, "loss": 1.028, "step": 46695 }, { "epoch": 0.12, "learning_rate": 8.850369922995622e-05, "loss": 1.0309, "step": 46700 }, { "epoch": 0.12, "learning_rate": 8.850244098847451e-05, "loss": 1.0271, "step": 46705 }, { "epoch": 0.12, "learning_rate": 8.85011827469928e-05, "loss": 1.029, "step": 46710 }, { "epoch": 0.12, "learning_rate": 8.84999245055111e-05, "loss": 1.0277, "step": 46715 }, { "epoch": 0.12, "learning_rate": 8.84986662640294e-05, "loss": 1.0287, "step": 46720 }, { "epoch": 0.12, "learning_rate": 8.849740802254769e-05, "loss": 1.0354, "step": 46725 }, { "epoch": 0.12, "learning_rate": 8.849614978106598e-05, "loss": 1.0257, "step": 46730 }, { "epoch": 0.12, "learning_rate": 8.849489153958427e-05, "loss": 1.0268, "step": 46735 }, { "epoch": 0.12, "learning_rate": 8.849363329810258e-05, "loss": 1.0272, "step": 46740 }, { "epoch": 0.12, "learning_rate": 8.849237505662087e-05, "loss": 1.0278, "step": 46745 }, { "epoch": 0.12, "learning_rate": 8.849111681513916e-05, "loss": 1.0275, "step": 46750 }, { "epoch": 0.12, "learning_rate": 8.848985857365745e-05, "loss": 1.0298, "step": 46755 }, { "epoch": 0.12, "learning_rate": 8.848860033217576e-05, "loss": 1.0296, "step": 46760 }, { "epoch": 0.12, "learning_rate": 8.848734209069405e-05, "loss": 1.03, "step": 46765 }, { "epoch": 0.12, "learning_rate": 8.848608384921234e-05, "loss": 1.0277, "step": 46770 }, { "epoch": 0.12, "learning_rate": 8.848482560773063e-05, "loss": 1.0303, "step": 46775 }, { "epoch": 0.12, "learning_rate": 8.848356736624893e-05, "loss": 1.0294, "step": 46780 }, { "epoch": 0.12, "learning_rate": 8.848230912476723e-05, "loss": 1.0286, "step": 46785 }, { "epoch": 0.12, "learning_rate": 8.848105088328552e-05, "loss": 1.0297, "step": 46790 }, { "epoch": 0.12, "learning_rate": 8.847979264180381e-05, "loss": 1.0287, "step": 46795 }, { "epoch": 0.12, "learning_rate": 8.84785344003221e-05, "loss": 1.0489, "step": 46800 }, { "epoch": 0.12, "learning_rate": 8.847727615884041e-05, "loss": 1.0304, "step": 46805 }, { "epoch": 0.12, "learning_rate": 8.84760179173587e-05, "loss": 1.031, "step": 46810 }, { "epoch": 0.12, "learning_rate": 8.8474759675877e-05, "loss": 1.0303, "step": 46815 }, { "epoch": 0.12, "learning_rate": 8.847350143439529e-05, "loss": 1.0277, "step": 46820 }, { "epoch": 0.12, "learning_rate": 8.847249484120994e-05, "loss": 1.0262, "step": 46825 }, { "epoch": 0.12, "learning_rate": 8.847123659972823e-05, "loss": 1.0285, "step": 46830 }, { "epoch": 0.12, "learning_rate": 8.846997835824652e-05, "loss": 1.0281, "step": 46835 }, { "epoch": 0.12, "learning_rate": 8.846872011676481e-05, "loss": 1.0292, "step": 46840 }, { "epoch": 0.12, "learning_rate": 8.846746187528312e-05, "loss": 1.028, "step": 46845 }, { "epoch": 0.12, "learning_rate": 8.84662036338014e-05, "loss": 1.0272, "step": 46850 }, { "epoch": 0.12, "learning_rate": 8.84649453923197e-05, "loss": 1.0269, "step": 46855 }, { "epoch": 0.12, "learning_rate": 8.846368715083799e-05, "loss": 1.048, "step": 46860 }, { "epoch": 0.12, "learning_rate": 8.846242890935628e-05, "loss": 1.0285, "step": 46865 }, { "epoch": 0.12, "learning_rate": 8.846117066787459e-05, "loss": 1.0276, "step": 46870 }, { "epoch": 0.12, "learning_rate": 8.845991242639288e-05, "loss": 1.0272, "step": 46875 }, { "epoch": 0.12, "learning_rate": 8.845865418491117e-05, "loss": 1.0258, "step": 46880 }, { "epoch": 0.12, "learning_rate": 8.845739594342946e-05, "loss": 1.0276, "step": 46885 }, { "epoch": 0.12, "learning_rate": 8.845613770194777e-05, "loss": 1.0284, "step": 46890 }, { "epoch": 0.12, "learning_rate": 8.845487946046606e-05, "loss": 1.0498, "step": 46895 }, { "epoch": 0.12, "learning_rate": 8.845362121898435e-05, "loss": 1.0291, "step": 46900 }, { "epoch": 0.12, "learning_rate": 8.845236297750264e-05, "loss": 1.0292, "step": 46905 }, { "epoch": 0.12, "learning_rate": 8.845110473602095e-05, "loss": 1.0305, "step": 46910 }, { "epoch": 0.12, "learning_rate": 8.844984649453924e-05, "loss": 1.0289, "step": 46915 }, { "epoch": 0.12, "learning_rate": 8.844858825305753e-05, "loss": 1.0289, "step": 46920 }, { "epoch": 0.12, "learning_rate": 8.844733001157582e-05, "loss": 1.0308, "step": 46925 }, { "epoch": 0.12, "learning_rate": 8.844607177009411e-05, "loss": 1.0279, "step": 46930 }, { "epoch": 0.12, "learning_rate": 8.844481352861242e-05, "loss": 1.0278, "step": 46935 }, { "epoch": 0.12, "learning_rate": 8.844355528713071e-05, "loss": 1.0265, "step": 46940 }, { "epoch": 0.12, "learning_rate": 8.8442297045649e-05, "loss": 1.0297, "step": 46945 }, { "epoch": 0.12, "learning_rate": 8.844103880416729e-05, "loss": 1.0285, "step": 46950 }, { "epoch": 0.12, "learning_rate": 8.84397805626856e-05, "loss": 1.0458, "step": 46955 }, { "epoch": 0.12, "learning_rate": 8.843852232120389e-05, "loss": 1.0264, "step": 46960 }, { "epoch": 0.12, "learning_rate": 8.843726407972218e-05, "loss": 1.0289, "step": 46965 }, { "epoch": 0.12, "learning_rate": 8.843600583824047e-05, "loss": 1.0273, "step": 46970 }, { "epoch": 0.12, "learning_rate": 8.843474759675878e-05, "loss": 1.0255, "step": 46975 }, { "epoch": 0.12, "learning_rate": 8.843348935527707e-05, "loss": 1.0283, "step": 46980 }, { "epoch": 0.12, "learning_rate": 8.843223111379536e-05, "loss": 1.0285, "step": 46985 }, { "epoch": 0.12, "learning_rate": 8.843097287231365e-05, "loss": 1.0438, "step": 46990 }, { "epoch": 0.12, "learning_rate": 8.842971463083194e-05, "loss": 1.0275, "step": 46995 }, { "epoch": 0.12, "learning_rate": 8.842845638935025e-05, "loss": 1.0478, "step": 47000 }, { "epoch": 0.12, "learning_rate": 8.842719814786854e-05, "loss": 1.0276, "step": 47005 }, { "epoch": 0.12, "learning_rate": 8.842593990638683e-05, "loss": 1.0288, "step": 47010 }, { "epoch": 0.12, "learning_rate": 8.842468166490512e-05, "loss": 1.0293, "step": 47015 }, { "epoch": 0.12, "learning_rate": 8.842342342342343e-05, "loss": 1.03, "step": 47020 }, { "epoch": 0.12, "learning_rate": 8.842216518194172e-05, "loss": 1.0273, "step": 47025 }, { "epoch": 0.12, "learning_rate": 8.842090694046001e-05, "loss": 1.0293, "step": 47030 }, { "epoch": 0.12, "learning_rate": 8.84196486989783e-05, "loss": 1.0303, "step": 47035 }, { "epoch": 0.12, "learning_rate": 8.841839045749661e-05, "loss": 1.0274, "step": 47040 }, { "epoch": 0.12, "learning_rate": 8.84171322160149e-05, "loss": 1.0287, "step": 47045 }, { "epoch": 0.12, "learning_rate": 8.841587397453319e-05, "loss": 1.0289, "step": 47050 }, { "epoch": 0.12, "learning_rate": 8.841461573305148e-05, "loss": 1.0276, "step": 47055 }, { "epoch": 0.12, "learning_rate": 8.841335749156978e-05, "loss": 1.0295, "step": 47060 }, { "epoch": 0.12, "learning_rate": 8.841209925008808e-05, "loss": 1.0526, "step": 47065 }, { "epoch": 0.12, "learning_rate": 8.841084100860637e-05, "loss": 1.0295, "step": 47070 }, { "epoch": 0.12, "learning_rate": 8.840958276712466e-05, "loss": 1.0272, "step": 47075 }, { "epoch": 0.12, "learning_rate": 8.840832452564296e-05, "loss": 1.0285, "step": 47080 }, { "epoch": 0.12, "learning_rate": 8.840706628416126e-05, "loss": 1.0282, "step": 47085 }, { "epoch": 0.12, "learning_rate": 8.840580804267955e-05, "loss": 1.027, "step": 47090 }, { "epoch": 0.12, "learning_rate": 8.840454980119784e-05, "loss": 1.0294, "step": 47095 }, { "epoch": 0.12, "learning_rate": 8.840329155971615e-05, "loss": 1.0274, "step": 47100 }, { "epoch": 0.12, "learning_rate": 8.840203331823444e-05, "loss": 1.0281, "step": 47105 }, { "epoch": 0.12, "learning_rate": 8.840077507675275e-05, "loss": 1.029, "step": 47110 }, { "epoch": 0.12, "learning_rate": 8.839951683527104e-05, "loss": 1.0278, "step": 47115 }, { "epoch": 0.12, "learning_rate": 8.839825859378933e-05, "loss": 1.0285, "step": 47120 }, { "epoch": 0.12, "learning_rate": 8.839700035230762e-05, "loss": 1.0275, "step": 47125 }, { "epoch": 0.12, "learning_rate": 8.839574211082591e-05, "loss": 1.0277, "step": 47130 }, { "epoch": 0.12, "learning_rate": 8.839448386934422e-05, "loss": 1.026, "step": 47135 }, { "epoch": 0.12, "learning_rate": 8.839322562786251e-05, "loss": 1.0269, "step": 47140 }, { "epoch": 0.12, "learning_rate": 8.83919673863808e-05, "loss": 1.0316, "step": 47145 }, { "epoch": 0.12, "learning_rate": 8.839070914489909e-05, "loss": 1.0262, "step": 47150 }, { "epoch": 0.12, "learning_rate": 8.83894509034174e-05, "loss": 1.0287, "step": 47155 }, { "epoch": 0.12, "learning_rate": 8.838819266193569e-05, "loss": 1.0263, "step": 47160 }, { "epoch": 0.12, "learning_rate": 8.838693442045398e-05, "loss": 1.0301, "step": 47165 }, { "epoch": 0.12, "learning_rate": 8.838567617897227e-05, "loss": 1.0298, "step": 47170 }, { "epoch": 0.12, "learning_rate": 8.838441793749058e-05, "loss": 1.0248, "step": 47175 }, { "epoch": 0.12, "learning_rate": 8.838315969600887e-05, "loss": 1.0518, "step": 47180 }, { "epoch": 0.12, "learning_rate": 8.838190145452716e-05, "loss": 1.0283, "step": 47185 }, { "epoch": 0.12, "learning_rate": 8.838064321304545e-05, "loss": 1.0289, "step": 47190 }, { "epoch": 0.12, "learning_rate": 8.837938497156374e-05, "loss": 1.0274, "step": 47195 }, { "epoch": 0.12, "learning_rate": 8.837812673008205e-05, "loss": 1.0284, "step": 47200 }, { "epoch": 0.12, "learning_rate": 8.837686848860034e-05, "loss": 1.0273, "step": 47205 }, { "epoch": 0.12, "learning_rate": 8.837561024711863e-05, "loss": 1.0282, "step": 47210 }, { "epoch": 0.12, "learning_rate": 8.837435200563692e-05, "loss": 1.0275, "step": 47215 }, { "epoch": 0.12, "learning_rate": 8.837309376415523e-05, "loss": 1.0267, "step": 47220 }, { "epoch": 0.12, "learning_rate": 8.837183552267352e-05, "loss": 1.0296, "step": 47225 }, { "epoch": 0.12, "learning_rate": 8.837057728119181e-05, "loss": 1.0772, "step": 47230 }, { "epoch": 0.12, "learning_rate": 8.83693190397101e-05, "loss": 1.0286, "step": 47235 }, { "epoch": 0.12, "learning_rate": 8.836806079822841e-05, "loss": 1.0309, "step": 47240 }, { "epoch": 0.12, "learning_rate": 8.83668025567467e-05, "loss": 1.0283, "step": 47245 }, { "epoch": 0.12, "learning_rate": 8.836554431526499e-05, "loss": 1.0298, "step": 47250 }, { "epoch": 0.12, "learning_rate": 8.836428607378328e-05, "loss": 1.0277, "step": 47255 }, { "epoch": 0.12, "learning_rate": 8.836302783230158e-05, "loss": 1.0273, "step": 47260 }, { "epoch": 0.12, "learning_rate": 8.836176959081988e-05, "loss": 1.029, "step": 47265 }, { "epoch": 0.12, "learning_rate": 8.836051134933817e-05, "loss": 1.0292, "step": 47270 }, { "epoch": 0.12, "learning_rate": 8.835925310785646e-05, "loss": 1.0283, "step": 47275 }, { "epoch": 0.12, "learning_rate": 8.835799486637476e-05, "loss": 1.029, "step": 47280 }, { "epoch": 0.12, "learning_rate": 8.835673662489306e-05, "loss": 1.0289, "step": 47285 }, { "epoch": 0.12, "learning_rate": 8.835547838341135e-05, "loss": 1.0276, "step": 47290 }, { "epoch": 0.12, "learning_rate": 8.835422014192964e-05, "loss": 1.0293, "step": 47295 }, { "epoch": 0.12, "learning_rate": 8.835296190044794e-05, "loss": 1.0289, "step": 47300 }, { "epoch": 0.12, "learning_rate": 8.835170365896624e-05, "loss": 1.0292, "step": 47305 }, { "epoch": 0.12, "learning_rate": 8.835044541748453e-05, "loss": 1.0266, "step": 47310 }, { "epoch": 0.12, "learning_rate": 8.834918717600282e-05, "loss": 1.0295, "step": 47315 }, { "epoch": 0.12, "learning_rate": 8.834792893452112e-05, "loss": 1.0317, "step": 47320 }, { "epoch": 0.12, "learning_rate": 8.83466706930394e-05, "loss": 1.0317, "step": 47325 }, { "epoch": 0.12, "learning_rate": 8.834541245155771e-05, "loss": 1.027, "step": 47330 }, { "epoch": 0.12, "learning_rate": 8.8344154210076e-05, "loss": 1.0293, "step": 47335 }, { "epoch": 0.12, "learning_rate": 8.83428959685943e-05, "loss": 1.03, "step": 47340 }, { "epoch": 0.12, "learning_rate": 8.834163772711259e-05, "loss": 1.0271, "step": 47345 }, { "epoch": 0.12, "learning_rate": 8.834037948563089e-05, "loss": 1.0296, "step": 47350 }, { "epoch": 0.12, "learning_rate": 8.833912124414918e-05, "loss": 1.0256, "step": 47355 }, { "epoch": 0.12, "learning_rate": 8.833786300266747e-05, "loss": 1.029, "step": 47360 }, { "epoch": 0.12, "learning_rate": 8.833660476118577e-05, "loss": 1.0283, "step": 47365 }, { "epoch": 0.12, "learning_rate": 8.833534651970407e-05, "loss": 1.0275, "step": 47370 }, { "epoch": 0.12, "learning_rate": 8.833408827822236e-05, "loss": 1.0292, "step": 47375 }, { "epoch": 0.12, "learning_rate": 8.833283003674065e-05, "loss": 1.0283, "step": 47380 }, { "epoch": 0.12, "learning_rate": 8.833157179525895e-05, "loss": 1.0302, "step": 47385 }, { "epoch": 0.12, "learning_rate": 8.833031355377724e-05, "loss": 1.0506, "step": 47390 }, { "epoch": 0.12, "learning_rate": 8.832905531229554e-05, "loss": 1.0283, "step": 47395 }, { "epoch": 0.12, "learning_rate": 8.832779707081383e-05, "loss": 1.0279, "step": 47400 }, { "epoch": 0.12, "learning_rate": 8.832653882933213e-05, "loss": 1.0278, "step": 47405 }, { "epoch": 0.12, "learning_rate": 8.832528058785042e-05, "loss": 1.0287, "step": 47410 }, { "epoch": 0.12, "learning_rate": 8.832402234636872e-05, "loss": 1.0271, "step": 47415 }, { "epoch": 0.12, "learning_rate": 8.832276410488701e-05, "loss": 1.0292, "step": 47420 }, { "epoch": 0.12, "learning_rate": 8.83215058634053e-05, "loss": 1.0284, "step": 47425 }, { "epoch": 0.12, "learning_rate": 8.83202476219236e-05, "loss": 1.0277, "step": 47430 }, { "epoch": 0.12, "learning_rate": 8.83189893804419e-05, "loss": 1.0255, "step": 47435 }, { "epoch": 0.12, "learning_rate": 8.83177311389602e-05, "loss": 1.0295, "step": 47440 }, { "epoch": 0.12, "learning_rate": 8.831647289747849e-05, "loss": 1.0271, "step": 47445 }, { "epoch": 0.12, "learning_rate": 8.831521465599678e-05, "loss": 1.0284, "step": 47450 }, { "epoch": 0.12, "learning_rate": 8.831395641451507e-05, "loss": 1.0288, "step": 47455 }, { "epoch": 0.12, "learning_rate": 8.831269817303337e-05, "loss": 1.0265, "step": 47460 }, { "epoch": 0.12, "learning_rate": 8.831143993155167e-05, "loss": 1.0262, "step": 47465 }, { "epoch": 0.12, "learning_rate": 8.831018169006996e-05, "loss": 1.0285, "step": 47470 }, { "epoch": 0.12, "learning_rate": 8.830892344858825e-05, "loss": 1.0274, "step": 47475 }, { "epoch": 0.12, "learning_rate": 8.830766520710655e-05, "loss": 1.0266, "step": 47480 }, { "epoch": 0.12, "learning_rate": 8.830640696562485e-05, "loss": 1.0294, "step": 47485 }, { "epoch": 0.12, "learning_rate": 8.830514872414314e-05, "loss": 1.0271, "step": 47490 }, { "epoch": 0.12, "learning_rate": 8.830389048266143e-05, "loss": 1.0285, "step": 47495 }, { "epoch": 0.12, "learning_rate": 8.830263224117973e-05, "loss": 1.029, "step": 47500 }, { "epoch": 0.12, "learning_rate": 8.830137399969803e-05, "loss": 1.0271, "step": 47505 }, { "epoch": 0.12, "learning_rate": 8.830011575821632e-05, "loss": 1.0284, "step": 47510 }, { "epoch": 0.12, "learning_rate": 8.829885751673461e-05, "loss": 1.0282, "step": 47515 }, { "epoch": 0.12, "learning_rate": 8.82975992752529e-05, "loss": 1.029, "step": 47520 }, { "epoch": 0.12, "learning_rate": 8.82963410337712e-05, "loss": 1.0271, "step": 47525 }, { "epoch": 0.12, "learning_rate": 8.82950827922895e-05, "loss": 1.0279, "step": 47530 }, { "epoch": 0.12, "learning_rate": 8.829382455080779e-05, "loss": 1.0273, "step": 47535 }, { "epoch": 0.12, "learning_rate": 8.829256630932608e-05, "loss": 1.0296, "step": 47540 }, { "epoch": 0.12, "learning_rate": 8.829130806784439e-05, "loss": 1.0279, "step": 47545 }, { "epoch": 0.12, "learning_rate": 8.829004982636268e-05, "loss": 1.0283, "step": 47550 }, { "epoch": 0.12, "learning_rate": 8.828879158488097e-05, "loss": 1.0292, "step": 47555 }, { "epoch": 0.12, "learning_rate": 8.828753334339926e-05, "loss": 1.0277, "step": 47560 }, { "epoch": 0.12, "learning_rate": 8.828627510191757e-05, "loss": 1.0291, "step": 47565 }, { "epoch": 0.12, "learning_rate": 8.828501686043586e-05, "loss": 1.0276, "step": 47570 }, { "epoch": 0.12, "learning_rate": 8.828375861895415e-05, "loss": 1.0272, "step": 47575 }, { "epoch": 0.12, "learning_rate": 8.828250037747244e-05, "loss": 1.0264, "step": 47580 }, { "epoch": 0.12, "learning_rate": 8.828124213599073e-05, "loss": 1.026, "step": 47585 }, { "epoch": 0.12, "learning_rate": 8.827998389450904e-05, "loss": 1.0287, "step": 47590 }, { "epoch": 0.12, "learning_rate": 8.827872565302734e-05, "loss": 1.0273, "step": 47595 }, { "epoch": 0.12, "learning_rate": 8.827746741154563e-05, "loss": 1.0275, "step": 47600 }, { "epoch": 0.12, "learning_rate": 8.827620917006393e-05, "loss": 1.0287, "step": 47605 }, { "epoch": 0.12, "learning_rate": 8.827495092858222e-05, "loss": 1.0294, "step": 47610 }, { "epoch": 0.12, "learning_rate": 8.827369268710052e-05, "loss": 1.0491, "step": 47615 }, { "epoch": 0.12, "learning_rate": 8.827243444561881e-05, "loss": 1.0288, "step": 47620 }, { "epoch": 0.12, "learning_rate": 8.82711762041371e-05, "loss": 1.0296, "step": 47625 }, { "epoch": 0.12, "learning_rate": 8.82699179626554e-05, "loss": 1.0265, "step": 47630 }, { "epoch": 0.12, "learning_rate": 8.82686597211737e-05, "loss": 1.0301, "step": 47635 }, { "epoch": 0.12, "learning_rate": 8.8267401479692e-05, "loss": 1.0287, "step": 47640 }, { "epoch": 0.12, "learning_rate": 8.826614323821029e-05, "loss": 1.0266, "step": 47645 }, { "epoch": 0.12, "learning_rate": 8.826488499672858e-05, "loss": 1.0301, "step": 47650 }, { "epoch": 0.12, "learning_rate": 8.826362675524687e-05, "loss": 1.0287, "step": 47655 }, { "epoch": 0.12, "learning_rate": 8.826236851376517e-05, "loss": 1.029, "step": 47660 }, { "epoch": 0.12, "learning_rate": 8.826111027228347e-05, "loss": 1.0282, "step": 47665 }, { "epoch": 0.12, "learning_rate": 8.825985203080176e-05, "loss": 1.0272, "step": 47670 }, { "epoch": 0.12, "learning_rate": 8.825859378932005e-05, "loss": 1.0253, "step": 47675 }, { "epoch": 0.12, "learning_rate": 8.825733554783835e-05, "loss": 1.0282, "step": 47680 }, { "epoch": 0.12, "learning_rate": 8.825607730635665e-05, "loss": 1.0268, "step": 47685 }, { "epoch": 0.12, "learning_rate": 8.825481906487494e-05, "loss": 1.0303, "step": 47690 }, { "epoch": 0.12, "learning_rate": 8.825356082339323e-05, "loss": 1.0297, "step": 47695 }, { "epoch": 0.12, "learning_rate": 8.825230258191153e-05, "loss": 1.0283, "step": 47700 }, { "epoch": 0.12, "learning_rate": 8.825104434042983e-05, "loss": 1.0289, "step": 47705 }, { "epoch": 0.12, "learning_rate": 8.824978609894812e-05, "loss": 1.0276, "step": 47710 }, { "epoch": 0.12, "learning_rate": 8.824852785746641e-05, "loss": 1.0298, "step": 47715 }, { "epoch": 0.12, "learning_rate": 8.82472696159847e-05, "loss": 1.0285, "step": 47720 }, { "epoch": 0.12, "learning_rate": 8.8246011374503e-05, "loss": 1.0284, "step": 47725 }, { "epoch": 0.12, "learning_rate": 8.82447531330213e-05, "loss": 1.0282, "step": 47730 }, { "epoch": 0.12, "learning_rate": 8.824349489153959e-05, "loss": 1.029, "step": 47735 }, { "epoch": 0.12, "learning_rate": 8.824223665005788e-05, "loss": 1.0295, "step": 47740 }, { "epoch": 0.12, "learning_rate": 8.824097840857619e-05, "loss": 1.0277, "step": 47745 }, { "epoch": 0.12, "learning_rate": 8.823972016709448e-05, "loss": 1.0282, "step": 47750 }, { "epoch": 0.12, "learning_rate": 8.823846192561277e-05, "loss": 1.0292, "step": 47755 }, { "epoch": 0.12, "learning_rate": 8.823720368413106e-05, "loss": 1.03, "step": 47760 }, { "epoch": 0.12, "learning_rate": 8.823594544264937e-05, "loss": 1.0289, "step": 47765 }, { "epoch": 0.12, "learning_rate": 8.823468720116766e-05, "loss": 1.0273, "step": 47770 }, { "epoch": 0.12, "learning_rate": 8.823342895968595e-05, "loss": 1.0297, "step": 47775 }, { "epoch": 0.12, "learning_rate": 8.823217071820424e-05, "loss": 1.0286, "step": 47780 }, { "epoch": 0.12, "learning_rate": 8.823091247672253e-05, "loss": 1.0301, "step": 47785 }, { "epoch": 0.12, "learning_rate": 8.822965423524084e-05, "loss": 1.0263, "step": 47790 }, { "epoch": 0.12, "learning_rate": 8.822839599375913e-05, "loss": 1.0282, "step": 47795 }, { "epoch": 0.12, "learning_rate": 8.822713775227742e-05, "loss": 1.0256, "step": 47800 }, { "epoch": 0.12, "learning_rate": 8.822587951079571e-05, "loss": 1.0295, "step": 47805 }, { "epoch": 0.12, "learning_rate": 8.822462126931402e-05, "loss": 1.0289, "step": 47810 }, { "epoch": 0.12, "learning_rate": 8.822336302783231e-05, "loss": 1.0295, "step": 47815 }, { "epoch": 0.12, "learning_rate": 8.82221047863506e-05, "loss": 1.0276, "step": 47820 }, { "epoch": 0.12, "learning_rate": 8.822084654486889e-05, "loss": 1.0277, "step": 47825 }, { "epoch": 0.12, "learning_rate": 8.82195883033872e-05, "loss": 1.0283, "step": 47830 }, { "epoch": 0.12, "learning_rate": 8.821833006190549e-05, "loss": 1.0273, "step": 47835 }, { "epoch": 0.12, "learning_rate": 8.821707182042378e-05, "loss": 1.0283, "step": 47840 }, { "epoch": 0.12, "learning_rate": 8.821581357894207e-05, "loss": 1.0289, "step": 47845 }, { "epoch": 0.12, "learning_rate": 8.821455533746036e-05, "loss": 1.03, "step": 47850 }, { "epoch": 0.12, "learning_rate": 8.821329709597867e-05, "loss": 1.027, "step": 47855 }, { "epoch": 0.12, "learning_rate": 8.821203885449696e-05, "loss": 1.028, "step": 47860 }, { "epoch": 0.12, "learning_rate": 8.821078061301525e-05, "loss": 1.0297, "step": 47865 }, { "epoch": 0.12, "learning_rate": 8.820952237153354e-05, "loss": 1.0293, "step": 47870 }, { "epoch": 0.12, "learning_rate": 8.820826413005185e-05, "loss": 1.0282, "step": 47875 }, { "epoch": 0.12, "learning_rate": 8.820700588857014e-05, "loss": 1.0279, "step": 47880 }, { "epoch": 0.12, "learning_rate": 8.820574764708843e-05, "loss": 1.0278, "step": 47885 }, { "epoch": 0.12, "learning_rate": 8.820448940560672e-05, "loss": 1.0281, "step": 47890 }, { "epoch": 0.12, "learning_rate": 8.820323116412503e-05, "loss": 1.027, "step": 47895 }, { "epoch": 0.12, "learning_rate": 8.820197292264332e-05, "loss": 1.0288, "step": 47900 }, { "epoch": 0.12, "learning_rate": 8.820071468116161e-05, "loss": 1.0267, "step": 47905 }, { "epoch": 0.12, "learning_rate": 8.81994564396799e-05, "loss": 1.0293, "step": 47910 }, { "epoch": 0.12, "learning_rate": 8.81981981981982e-05, "loss": 1.0292, "step": 47915 }, { "epoch": 0.12, "learning_rate": 8.81969399567165e-05, "loss": 1.0298, "step": 47920 }, { "epoch": 0.12, "learning_rate": 8.819568171523479e-05, "loss": 1.0268, "step": 47925 }, { "epoch": 0.12, "learning_rate": 8.819442347375308e-05, "loss": 1.028, "step": 47930 }, { "epoch": 0.12, "learning_rate": 8.819316523227137e-05, "loss": 1.0265, "step": 47935 }, { "epoch": 0.12, "learning_rate": 8.819190699078968e-05, "loss": 1.0268, "step": 47940 }, { "epoch": 0.12, "learning_rate": 8.819064874930797e-05, "loss": 1.0286, "step": 47945 }, { "epoch": 0.12, "learning_rate": 8.818939050782626e-05, "loss": 1.0283, "step": 47950 }, { "epoch": 0.12, "learning_rate": 8.818813226634455e-05, "loss": 1.0281, "step": 47955 }, { "epoch": 0.12, "learning_rate": 8.818687402486286e-05, "loss": 1.0275, "step": 47960 }, { "epoch": 0.12, "learning_rate": 8.818561578338115e-05, "loss": 1.027, "step": 47965 }, { "epoch": 0.12, "learning_rate": 8.818435754189944e-05, "loss": 1.0277, "step": 47970 }, { "epoch": 0.12, "learning_rate": 8.818309930041773e-05, "loss": 1.031, "step": 47975 }, { "epoch": 0.12, "learning_rate": 8.818184105893603e-05, "loss": 1.0275, "step": 47980 }, { "epoch": 0.12, "learning_rate": 8.818058281745433e-05, "loss": 1.0284, "step": 47985 }, { "epoch": 0.12, "learning_rate": 8.817932457597262e-05, "loss": 1.044, "step": 47990 }, { "epoch": 0.12, "learning_rate": 8.817806633449091e-05, "loss": 1.0289, "step": 47995 }, { "epoch": 0.12, "learning_rate": 8.81768080930092e-05, "loss": 1.0281, "step": 48000 }, { "epoch": 0.12, "learning_rate": 8.817554985152751e-05, "loss": 1.0295, "step": 48005 }, { "epoch": 0.12, "learning_rate": 8.81742916100458e-05, "loss": 1.0281, "step": 48010 }, { "epoch": 0.12, "learning_rate": 8.81730333685641e-05, "loss": 1.0296, "step": 48015 }, { "epoch": 0.12, "learning_rate": 8.817177512708239e-05, "loss": 1.0269, "step": 48020 }, { "epoch": 0.12, "learning_rate": 8.817051688560069e-05, "loss": 1.0292, "step": 48025 }, { "epoch": 0.12, "learning_rate": 8.816925864411898e-05, "loss": 1.0278, "step": 48030 }, { "epoch": 0.12, "learning_rate": 8.816800040263727e-05, "loss": 1.0282, "step": 48035 }, { "epoch": 0.12, "learning_rate": 8.816674216115557e-05, "loss": 1.0291, "step": 48040 }, { "epoch": 0.12, "learning_rate": 8.816548391967386e-05, "loss": 1.0263, "step": 48045 }, { "epoch": 0.12, "learning_rate": 8.816422567819216e-05, "loss": 1.0266, "step": 48050 }, { "epoch": 0.12, "learning_rate": 8.816296743671045e-05, "loss": 1.0276, "step": 48055 }, { "epoch": 0.12, "learning_rate": 8.816170919522875e-05, "loss": 1.0264, "step": 48060 }, { "epoch": 0.12, "learning_rate": 8.816045095374704e-05, "loss": 1.029, "step": 48065 }, { "epoch": 0.12, "learning_rate": 8.815919271226534e-05, "loss": 1.0285, "step": 48070 }, { "epoch": 0.12, "learning_rate": 8.815793447078363e-05, "loss": 1.0275, "step": 48075 }, { "epoch": 0.12, "learning_rate": 8.815667622930193e-05, "loss": 1.028, "step": 48080 }, { "epoch": 0.12, "learning_rate": 8.815541798782022e-05, "loss": 1.029, "step": 48085 }, { "epoch": 0.12, "learning_rate": 8.815415974633852e-05, "loss": 1.0497, "step": 48090 }, { "epoch": 0.12, "learning_rate": 8.815290150485683e-05, "loss": 1.0288, "step": 48095 }, { "epoch": 0.12, "learning_rate": 8.815164326337512e-05, "loss": 1.0267, "step": 48100 }, { "epoch": 0.12, "learning_rate": 8.815038502189341e-05, "loss": 1.0293, "step": 48105 }, { "epoch": 0.12, "learning_rate": 8.81491267804117e-05, "loss": 1.0262, "step": 48110 }, { "epoch": 0.12, "learning_rate": 8.814786853893e-05, "loss": 1.0267, "step": 48115 }, { "epoch": 0.12, "learning_rate": 8.81466102974483e-05, "loss": 1.028, "step": 48120 }, { "epoch": 0.12, "learning_rate": 8.814535205596659e-05, "loss": 1.0276, "step": 48125 }, { "epoch": 0.12, "learning_rate": 8.814409381448488e-05, "loss": 1.0298, "step": 48130 }, { "epoch": 0.12, "learning_rate": 8.814283557300317e-05, "loss": 1.049, "step": 48135 }, { "epoch": 0.12, "learning_rate": 8.814157733152148e-05, "loss": 1.0271, "step": 48140 }, { "epoch": 0.12, "learning_rate": 8.814031909003977e-05, "loss": 1.0261, "step": 48145 }, { "epoch": 0.12, "learning_rate": 8.813906084855806e-05, "loss": 1.028, "step": 48150 }, { "epoch": 0.12, "learning_rate": 8.813780260707635e-05, "loss": 1.0282, "step": 48155 }, { "epoch": 0.12, "learning_rate": 8.813654436559466e-05, "loss": 1.0284, "step": 48160 }, { "epoch": 0.12, "learning_rate": 8.813528612411295e-05, "loss": 1.0269, "step": 48165 }, { "epoch": 0.12, "learning_rate": 8.813402788263124e-05, "loss": 1.0288, "step": 48170 }, { "epoch": 0.12, "learning_rate": 8.813276964114953e-05, "loss": 1.0282, "step": 48175 }, { "epoch": 0.12, "learning_rate": 8.813151139966783e-05, "loss": 1.0281, "step": 48180 }, { "epoch": 0.12, "learning_rate": 8.813025315818613e-05, "loss": 1.0294, "step": 48185 }, { "epoch": 0.12, "learning_rate": 8.812899491670442e-05, "loss": 1.0258, "step": 48190 }, { "epoch": 0.12, "learning_rate": 8.812773667522271e-05, "loss": 1.0278, "step": 48195 }, { "epoch": 0.12, "learning_rate": 8.8126478433741e-05, "loss": 1.0277, "step": 48200 }, { "epoch": 0.12, "learning_rate": 8.812522019225931e-05, "loss": 1.0283, "step": 48205 }, { "epoch": 0.12, "learning_rate": 8.81239619507776e-05, "loss": 1.028, "step": 48210 }, { "epoch": 0.12, "learning_rate": 8.81227037092959e-05, "loss": 1.0293, "step": 48215 }, { "epoch": 0.12, "learning_rate": 8.812144546781419e-05, "loss": 1.0286, "step": 48220 }, { "epoch": 0.12, "learning_rate": 8.812018722633249e-05, "loss": 1.0271, "step": 48225 }, { "epoch": 0.12, "learning_rate": 8.811892898485078e-05, "loss": 1.0269, "step": 48230 }, { "epoch": 0.12, "learning_rate": 8.811767074336907e-05, "loss": 1.0277, "step": 48235 }, { "epoch": 0.12, "learning_rate": 8.811641250188737e-05, "loss": 1.0282, "step": 48240 }, { "epoch": 0.12, "learning_rate": 8.811515426040566e-05, "loss": 1.0269, "step": 48245 }, { "epoch": 0.12, "learning_rate": 8.811389601892396e-05, "loss": 1.0302, "step": 48250 }, { "epoch": 0.12, "learning_rate": 8.811263777744225e-05, "loss": 1.0309, "step": 48255 }, { "epoch": 0.12, "learning_rate": 8.811137953596055e-05, "loss": 1.0465, "step": 48260 }, { "epoch": 0.12, "learning_rate": 8.811012129447884e-05, "loss": 1.0292, "step": 48265 }, { "epoch": 0.12, "learning_rate": 8.810886305299714e-05, "loss": 1.03, "step": 48270 }, { "epoch": 0.12, "learning_rate": 8.810760481151543e-05, "loss": 1.0282, "step": 48275 }, { "epoch": 0.12, "learning_rate": 8.810634657003373e-05, "loss": 1.029, "step": 48280 }, { "epoch": 0.12, "learning_rate": 8.810508832855202e-05, "loss": 1.0274, "step": 48285 }, { "epoch": 0.12, "learning_rate": 8.810383008707032e-05, "loss": 1.0286, "step": 48290 }, { "epoch": 0.12, "learning_rate": 8.810257184558861e-05, "loss": 1.0287, "step": 48295 }, { "epoch": 0.12, "learning_rate": 8.81013136041069e-05, "loss": 1.028, "step": 48300 }, { "epoch": 0.12, "learning_rate": 8.81000553626252e-05, "loss": 1.0271, "step": 48305 }, { "epoch": 0.12, "learning_rate": 8.809879712114349e-05, "loss": 1.0277, "step": 48310 }, { "epoch": 0.12, "learning_rate": 8.80975388796618e-05, "loss": 1.0289, "step": 48315 }, { "epoch": 0.12, "learning_rate": 8.809628063818009e-05, "loss": 1.0286, "step": 48320 }, { "epoch": 0.12, "learning_rate": 8.809502239669838e-05, "loss": 1.026, "step": 48325 }, { "epoch": 0.12, "learning_rate": 8.809376415521667e-05, "loss": 1.0296, "step": 48330 }, { "epoch": 0.12, "learning_rate": 8.809250591373497e-05, "loss": 1.0278, "step": 48335 }, { "epoch": 0.12, "learning_rate": 8.809124767225327e-05, "loss": 1.0292, "step": 48340 }, { "epoch": 0.12, "learning_rate": 8.808998943077156e-05, "loss": 1.0272, "step": 48345 }, { "epoch": 0.12, "learning_rate": 8.808873118928985e-05, "loss": 1.0267, "step": 48350 }, { "epoch": 0.12, "learning_rate": 8.808747294780815e-05, "loss": 1.0295, "step": 48355 }, { "epoch": 0.12, "learning_rate": 8.808621470632644e-05, "loss": 1.0275, "step": 48360 }, { "epoch": 0.12, "learning_rate": 8.808495646484474e-05, "loss": 1.0281, "step": 48365 }, { "epoch": 0.12, "learning_rate": 8.808369822336303e-05, "loss": 1.0292, "step": 48370 }, { "epoch": 0.12, "learning_rate": 8.808243998188132e-05, "loss": 1.0284, "step": 48375 }, { "epoch": 0.12, "learning_rate": 8.808118174039962e-05, "loss": 1.0261, "step": 48380 }, { "epoch": 0.12, "learning_rate": 8.807992349891792e-05, "loss": 1.0291, "step": 48385 }, { "epoch": 0.12, "learning_rate": 8.807866525743621e-05, "loss": 1.0263, "step": 48390 }, { "epoch": 0.12, "learning_rate": 8.80774070159545e-05, "loss": 1.0266, "step": 48395 }, { "epoch": 0.12, "learning_rate": 8.80761487744728e-05, "loss": 1.0293, "step": 48400 }, { "epoch": 0.12, "learning_rate": 8.80748905329911e-05, "loss": 1.0261, "step": 48405 }, { "epoch": 0.12, "learning_rate": 8.807363229150939e-05, "loss": 1.0293, "step": 48410 }, { "epoch": 0.12, "learning_rate": 8.807237405002768e-05, "loss": 1.0279, "step": 48415 }, { "epoch": 0.12, "learning_rate": 8.807111580854598e-05, "loss": 1.0273, "step": 48420 }, { "epoch": 0.12, "learning_rate": 8.806985756706428e-05, "loss": 1.0257, "step": 48425 }, { "epoch": 0.12, "learning_rate": 8.806859932558257e-05, "loss": 1.0282, "step": 48430 }, { "epoch": 0.12, "learning_rate": 8.806734108410086e-05, "loss": 1.0423, "step": 48435 }, { "epoch": 0.12, "learning_rate": 8.806608284261915e-05, "loss": 1.0283, "step": 48440 }, { "epoch": 0.12, "learning_rate": 8.806482460113746e-05, "loss": 1.0286, "step": 48445 }, { "epoch": 0.12, "learning_rate": 8.806356635965575e-05, "loss": 1.0273, "step": 48450 }, { "epoch": 0.12, "learning_rate": 8.806230811817404e-05, "loss": 1.0283, "step": 48455 }, { "epoch": 0.12, "learning_rate": 8.806104987669233e-05, "loss": 1.0293, "step": 48460 }, { "epoch": 0.12, "learning_rate": 8.805979163521064e-05, "loss": 1.0289, "step": 48465 }, { "epoch": 0.12, "learning_rate": 8.805853339372893e-05, "loss": 1.0293, "step": 48470 }, { "epoch": 0.12, "learning_rate": 8.805727515224722e-05, "loss": 1.0289, "step": 48475 }, { "epoch": 0.12, "learning_rate": 8.805601691076551e-05, "loss": 1.0272, "step": 48480 }, { "epoch": 0.12, "learning_rate": 8.805475866928382e-05, "loss": 1.0241, "step": 48485 }, { "epoch": 0.12, "learning_rate": 8.805350042780211e-05, "loss": 1.0287, "step": 48490 }, { "epoch": 0.12, "learning_rate": 8.80522421863204e-05, "loss": 1.0276, "step": 48495 }, { "epoch": 0.12, "learning_rate": 8.805098394483869e-05, "loss": 1.029, "step": 48500 }, { "epoch": 0.12, "learning_rate": 8.804972570335698e-05, "loss": 1.0278, "step": 48505 }, { "epoch": 0.12, "learning_rate": 8.804846746187529e-05, "loss": 1.0274, "step": 48510 }, { "epoch": 0.12, "learning_rate": 8.804720922039358e-05, "loss": 1.026, "step": 48515 }, { "epoch": 0.12, "learning_rate": 8.804595097891187e-05, "loss": 1.0281, "step": 48520 }, { "epoch": 0.12, "learning_rate": 8.804469273743016e-05, "loss": 1.029, "step": 48525 }, { "epoch": 0.12, "learning_rate": 8.804343449594847e-05, "loss": 1.027, "step": 48530 }, { "epoch": 0.12, "learning_rate": 8.804217625446676e-05, "loss": 1.0275, "step": 48535 }, { "epoch": 0.12, "learning_rate": 8.804091801298505e-05, "loss": 1.0282, "step": 48540 }, { "epoch": 0.12, "learning_rate": 8.803965977150334e-05, "loss": 1.0311, "step": 48545 }, { "epoch": 0.12, "learning_rate": 8.803840153002165e-05, "loss": 1.0313, "step": 48550 }, { "epoch": 0.12, "learning_rate": 8.803714328853994e-05, "loss": 1.0278, "step": 48555 }, { "epoch": 0.12, "learning_rate": 8.803588504705823e-05, "loss": 1.0282, "step": 48560 }, { "epoch": 0.12, "learning_rate": 8.803462680557652e-05, "loss": 1.0284, "step": 48565 }, { "epoch": 0.12, "learning_rate": 8.803336856409481e-05, "loss": 1.0278, "step": 48570 }, { "epoch": 0.12, "learning_rate": 8.803211032261312e-05, "loss": 1.0268, "step": 48575 }, { "epoch": 0.12, "learning_rate": 8.803085208113141e-05, "loss": 1.0276, "step": 48580 }, { "epoch": 0.12, "learning_rate": 8.80295938396497e-05, "loss": 1.0298, "step": 48585 }, { "epoch": 0.12, "learning_rate": 8.8028335598168e-05, "loss": 1.0293, "step": 48590 }, { "epoch": 0.12, "learning_rate": 8.80270773566863e-05, "loss": 1.0285, "step": 48595 }, { "epoch": 0.12, "learning_rate": 8.80258191152046e-05, "loss": 1.0267, "step": 48600 }, { "epoch": 0.12, "learning_rate": 8.80245608737229e-05, "loss": 1.0264, "step": 48605 }, { "epoch": 0.12, "learning_rate": 8.802330263224119e-05, "loss": 1.0274, "step": 48610 }, { "epoch": 0.12, "learning_rate": 8.802204439075948e-05, "loss": 1.0273, "step": 48615 }, { "epoch": 0.12, "learning_rate": 8.802078614927778e-05, "loss": 1.0295, "step": 48620 }, { "epoch": 0.12, "learning_rate": 8.801952790779608e-05, "loss": 1.0278, "step": 48625 }, { "epoch": 0.12, "learning_rate": 8.801826966631437e-05, "loss": 1.0279, "step": 48630 }, { "epoch": 0.12, "learning_rate": 8.801701142483266e-05, "loss": 1.0261, "step": 48635 }, { "epoch": 0.12, "learning_rate": 8.801575318335096e-05, "loss": 1.0273, "step": 48640 }, { "epoch": 0.12, "learning_rate": 8.801449494186926e-05, "loss": 1.0275, "step": 48645 }, { "epoch": 0.12, "learning_rate": 8.801323670038755e-05, "loss": 1.0281, "step": 48650 }, { "epoch": 0.12, "learning_rate": 8.801197845890584e-05, "loss": 1.0268, "step": 48655 }, { "epoch": 0.12, "learning_rate": 8.801072021742413e-05, "loss": 1.0295, "step": 48660 }, { "epoch": 0.12, "learning_rate": 8.800946197594244e-05, "loss": 1.0509, "step": 48665 }, { "epoch": 0.12, "learning_rate": 8.800820373446073e-05, "loss": 1.0304, "step": 48670 }, { "epoch": 0.12, "learning_rate": 8.800694549297902e-05, "loss": 1.0285, "step": 48675 }, { "epoch": 0.12, "learning_rate": 8.800568725149731e-05, "loss": 1.028, "step": 48680 }, { "epoch": 0.12, "learning_rate": 8.800442901001562e-05, "loss": 1.0284, "step": 48685 }, { "epoch": 0.12, "learning_rate": 8.800317076853391e-05, "loss": 1.0274, "step": 48690 }, { "epoch": 0.12, "learning_rate": 8.80019125270522e-05, "loss": 1.028, "step": 48695 }, { "epoch": 0.12, "learning_rate": 8.800065428557049e-05, "loss": 1.0297, "step": 48700 }, { "epoch": 0.12, "learning_rate": 8.799939604408878e-05, "loss": 1.0273, "step": 48705 }, { "epoch": 0.12, "learning_rate": 8.799813780260709e-05, "loss": 1.027, "step": 48710 }, { "epoch": 0.12, "learning_rate": 8.799687956112538e-05, "loss": 1.029, "step": 48715 }, { "epoch": 0.12, "learning_rate": 8.799562131964367e-05, "loss": 1.0268, "step": 48720 }, { "epoch": 0.12, "learning_rate": 8.799436307816196e-05, "loss": 1.0301, "step": 48725 }, { "epoch": 0.12, "learning_rate": 8.799310483668027e-05, "loss": 1.0277, "step": 48730 }, { "epoch": 0.12, "learning_rate": 8.799184659519856e-05, "loss": 1.0277, "step": 48735 }, { "epoch": 0.12, "learning_rate": 8.799058835371685e-05, "loss": 1.028, "step": 48740 }, { "epoch": 0.12, "learning_rate": 8.798933011223514e-05, "loss": 1.0284, "step": 48745 }, { "epoch": 0.12, "learning_rate": 8.798807187075345e-05, "loss": 1.0299, "step": 48750 }, { "epoch": 0.12, "learning_rate": 8.798681362927174e-05, "loss": 1.0252, "step": 48755 }, { "epoch": 0.12, "learning_rate": 8.798555538779003e-05, "loss": 1.0286, "step": 48760 }, { "epoch": 0.12, "learning_rate": 8.798429714630832e-05, "loss": 1.0267, "step": 48765 }, { "epoch": 0.12, "learning_rate": 8.798303890482661e-05, "loss": 1.028, "step": 48770 }, { "epoch": 0.12, "learning_rate": 8.798178066334492e-05, "loss": 1.0291, "step": 48775 }, { "epoch": 0.12, "learning_rate": 8.798052242186321e-05, "loss": 1.0271, "step": 48780 }, { "epoch": 0.12, "learning_rate": 8.79792641803815e-05, "loss": 1.0282, "step": 48785 }, { "epoch": 0.12, "learning_rate": 8.79780059388998e-05, "loss": 1.0297, "step": 48790 }, { "epoch": 0.12, "learning_rate": 8.79767476974181e-05, "loss": 1.0266, "step": 48795 }, { "epoch": 0.12, "learning_rate": 8.797548945593639e-05, "loss": 1.0291, "step": 48800 }, { "epoch": 0.12, "learning_rate": 8.797423121445468e-05, "loss": 1.0284, "step": 48805 }, { "epoch": 0.12, "learning_rate": 8.797297297297297e-05, "loss": 1.0272, "step": 48810 }, { "epoch": 0.12, "learning_rate": 8.797171473149128e-05, "loss": 1.0292, "step": 48815 }, { "epoch": 0.12, "learning_rate": 8.797045649000957e-05, "loss": 1.027, "step": 48820 }, { "epoch": 0.12, "learning_rate": 8.796919824852786e-05, "loss": 1.028, "step": 48825 }, { "epoch": 0.12, "learning_rate": 8.796794000704615e-05, "loss": 1.0283, "step": 48830 }, { "epoch": 0.12, "learning_rate": 8.796668176556444e-05, "loss": 1.0298, "step": 48835 }, { "epoch": 0.12, "learning_rate": 8.796542352408275e-05, "loss": 1.0255, "step": 48840 }, { "epoch": 0.12, "learning_rate": 8.796416528260104e-05, "loss": 1.0275, "step": 48845 }, { "epoch": 0.12, "learning_rate": 8.796290704111933e-05, "loss": 1.0307, "step": 48850 }, { "epoch": 0.12, "learning_rate": 8.796164879963762e-05, "loss": 1.0265, "step": 48855 }, { "epoch": 0.12, "learning_rate": 8.796039055815593e-05, "loss": 1.0286, "step": 48860 }, { "epoch": 0.12, "learning_rate": 8.795913231667422e-05, "loss": 1.0261, "step": 48865 }, { "epoch": 0.12, "learning_rate": 8.795787407519251e-05, "loss": 1.0254, "step": 48870 }, { "epoch": 0.12, "learning_rate": 8.79566158337108e-05, "loss": 1.0267, "step": 48875 }, { "epoch": 0.12, "learning_rate": 8.795535759222911e-05, "loss": 1.0285, "step": 48880 }, { "epoch": 0.12, "learning_rate": 8.79540993507474e-05, "loss": 1.0276, "step": 48885 }, { "epoch": 0.12, "learning_rate": 8.795284110926569e-05, "loss": 1.0286, "step": 48890 }, { "epoch": 0.12, "learning_rate": 8.795158286778398e-05, "loss": 1.0282, "step": 48895 }, { "epoch": 0.12, "learning_rate": 8.795032462630228e-05, "loss": 1.0286, "step": 48900 }, { "epoch": 0.12, "learning_rate": 8.794906638482058e-05, "loss": 1.0285, "step": 48905 }, { "epoch": 0.12, "learning_rate": 8.794780814333887e-05, "loss": 1.0291, "step": 48910 }, { "epoch": 0.12, "learning_rate": 8.794654990185716e-05, "loss": 1.027, "step": 48915 }, { "epoch": 0.12, "learning_rate": 8.794529166037546e-05, "loss": 1.0268, "step": 48920 }, { "epoch": 0.12, "learning_rate": 8.794403341889376e-05, "loss": 1.0284, "step": 48925 }, { "epoch": 0.12, "learning_rate": 8.794277517741205e-05, "loss": 1.0277, "step": 48930 }, { "epoch": 0.12, "learning_rate": 8.794151693593034e-05, "loss": 1.0266, "step": 48935 }, { "epoch": 0.12, "learning_rate": 8.794025869444864e-05, "loss": 1.0305, "step": 48940 }, { "epoch": 0.12, "learning_rate": 8.793900045296694e-05, "loss": 1.0261, "step": 48945 }, { "epoch": 0.12, "learning_rate": 8.793774221148523e-05, "loss": 1.0263, "step": 48950 }, { "epoch": 0.12, "learning_rate": 8.793648397000352e-05, "loss": 1.0279, "step": 48955 }, { "epoch": 0.12, "learning_rate": 8.793522572852182e-05, "loss": 1.0281, "step": 48960 }, { "epoch": 0.12, "learning_rate": 8.793396748704011e-05, "loss": 1.0276, "step": 48965 }, { "epoch": 0.12, "learning_rate": 8.793270924555841e-05, "loss": 1.0274, "step": 48970 }, { "epoch": 0.12, "learning_rate": 8.79314510040767e-05, "loss": 1.0282, "step": 48975 }, { "epoch": 0.12, "learning_rate": 8.7930192762595e-05, "loss": 1.0273, "step": 48980 }, { "epoch": 0.12, "learning_rate": 8.792893452111329e-05, "loss": 1.0275, "step": 48985 }, { "epoch": 0.12, "learning_rate": 8.792767627963159e-05, "loss": 1.028, "step": 48990 }, { "epoch": 0.12, "learning_rate": 8.792641803814988e-05, "loss": 1.0295, "step": 48995 }, { "epoch": 0.12, "learning_rate": 8.792515979666818e-05, "loss": 1.0289, "step": 49000 }, { "epoch": 0.12, "learning_rate": 8.792390155518647e-05, "loss": 1.0284, "step": 49005 }, { "epoch": 0.12, "learning_rate": 8.792264331370477e-05, "loss": 1.0283, "step": 49010 }, { "epoch": 0.12, "learning_rate": 8.792138507222306e-05, "loss": 1.0268, "step": 49015 }, { "epoch": 0.12, "learning_rate": 8.792012683074136e-05, "loss": 1.0275, "step": 49020 }, { "epoch": 0.12, "learning_rate": 8.791886858925965e-05, "loss": 1.0268, "step": 49025 }, { "epoch": 0.12, "learning_rate": 8.791761034777794e-05, "loss": 1.0292, "step": 49030 }, { "epoch": 0.12, "learning_rate": 8.791635210629624e-05, "loss": 1.0272, "step": 49035 }, { "epoch": 0.12, "learning_rate": 8.791509386481454e-05, "loss": 1.0298, "step": 49040 }, { "epoch": 0.12, "learning_rate": 8.791383562333283e-05, "loss": 1.0285, "step": 49045 }, { "epoch": 0.12, "learning_rate": 8.791257738185112e-05, "loss": 1.0294, "step": 49050 }, { "epoch": 0.12, "learning_rate": 8.791131914036942e-05, "loss": 1.0274, "step": 49055 }, { "epoch": 0.12, "learning_rate": 8.791006089888772e-05, "loss": 1.0276, "step": 49060 }, { "epoch": 0.12, "learning_rate": 8.790880265740601e-05, "loss": 1.0263, "step": 49065 }, { "epoch": 0.12, "learning_rate": 8.790779606422064e-05, "loss": 1.0265, "step": 49070 }, { "epoch": 0.12, "learning_rate": 8.790653782273895e-05, "loss": 1.0285, "step": 49075 }, { "epoch": 0.12, "learning_rate": 8.790527958125724e-05, "loss": 1.0276, "step": 49080 }, { "epoch": 0.12, "learning_rate": 8.790402133977553e-05, "loss": 1.0304, "step": 49085 }, { "epoch": 0.12, "learning_rate": 8.790276309829382e-05, "loss": 1.0248, "step": 49090 }, { "epoch": 0.12, "learning_rate": 8.790150485681213e-05, "loss": 1.0269, "step": 49095 }, { "epoch": 0.12, "learning_rate": 8.790024661533042e-05, "loss": 1.0299, "step": 49100 }, { "epoch": 0.12, "learning_rate": 8.789898837384871e-05, "loss": 1.0271, "step": 49105 }, { "epoch": 0.12, "learning_rate": 8.7897730132367e-05, "loss": 1.0303, "step": 49110 }, { "epoch": 0.12, "learning_rate": 8.78964718908853e-05, "loss": 1.0281, "step": 49115 }, { "epoch": 0.12, "learning_rate": 8.78952136494036e-05, "loss": 1.0281, "step": 49120 }, { "epoch": 0.12, "learning_rate": 8.789395540792189e-05, "loss": 1.0295, "step": 49125 }, { "epoch": 0.12, "learning_rate": 8.789269716644018e-05, "loss": 1.0287, "step": 49130 }, { "epoch": 0.12, "learning_rate": 8.789143892495847e-05, "loss": 1.027, "step": 49135 }, { "epoch": 0.12, "learning_rate": 8.789018068347678e-05, "loss": 1.0267, "step": 49140 }, { "epoch": 0.12, "learning_rate": 8.788892244199507e-05, "loss": 1.0299, "step": 49145 }, { "epoch": 0.12, "learning_rate": 8.788766420051336e-05, "loss": 1.0276, "step": 49150 }, { "epoch": 0.12, "learning_rate": 8.788640595903165e-05, "loss": 1.0284, "step": 49155 }, { "epoch": 0.12, "learning_rate": 8.788514771754996e-05, "loss": 1.0269, "step": 49160 }, { "epoch": 0.12, "learning_rate": 8.788388947606825e-05, "loss": 1.0294, "step": 49165 }, { "epoch": 0.12, "learning_rate": 8.788263123458654e-05, "loss": 1.0274, "step": 49170 }, { "epoch": 0.12, "learning_rate": 8.788137299310483e-05, "loss": 1.0263, "step": 49175 }, { "epoch": 0.12, "learning_rate": 8.788011475162313e-05, "loss": 1.028, "step": 49180 }, { "epoch": 0.12, "learning_rate": 8.787885651014143e-05, "loss": 1.0267, "step": 49185 }, { "epoch": 0.12, "learning_rate": 8.787759826865972e-05, "loss": 1.0253, "step": 49190 }, { "epoch": 0.12, "learning_rate": 8.787634002717801e-05, "loss": 1.0283, "step": 49195 }, { "epoch": 0.12, "learning_rate": 8.78750817856963e-05, "loss": 1.0286, "step": 49200 }, { "epoch": 0.12, "learning_rate": 8.787382354421461e-05, "loss": 1.0286, "step": 49205 }, { "epoch": 0.12, "learning_rate": 8.78725653027329e-05, "loss": 1.0294, "step": 49210 }, { "epoch": 0.12, "learning_rate": 8.78713070612512e-05, "loss": 1.0297, "step": 49215 }, { "epoch": 0.12, "learning_rate": 8.787004881976949e-05, "loss": 1.0277, "step": 49220 }, { "epoch": 0.12, "learning_rate": 8.786879057828779e-05, "loss": 1.0288, "step": 49225 }, { "epoch": 0.12, "learning_rate": 8.786753233680608e-05, "loss": 1.0265, "step": 49230 }, { "epoch": 0.12, "learning_rate": 8.786627409532437e-05, "loss": 1.0241, "step": 49235 }, { "epoch": 0.12, "learning_rate": 8.786501585384267e-05, "loss": 1.0281, "step": 49240 }, { "epoch": 0.12, "learning_rate": 8.786375761236096e-05, "loss": 1.0287, "step": 49245 }, { "epoch": 0.12, "learning_rate": 8.786249937087926e-05, "loss": 1.0273, "step": 49250 }, { "epoch": 0.12, "learning_rate": 8.786124112939755e-05, "loss": 1.0508, "step": 49255 }, { "epoch": 0.12, "learning_rate": 8.785998288791585e-05, "loss": 1.0295, "step": 49260 }, { "epoch": 0.12, "learning_rate": 8.785872464643414e-05, "loss": 1.027, "step": 49265 }, { "epoch": 0.12, "learning_rate": 8.785746640495244e-05, "loss": 1.0288, "step": 49270 }, { "epoch": 0.12, "learning_rate": 8.785620816347073e-05, "loss": 1.0282, "step": 49275 }, { "epoch": 0.12, "learning_rate": 8.785494992198903e-05, "loss": 1.0282, "step": 49280 }, { "epoch": 0.12, "learning_rate": 8.785369168050732e-05, "loss": 1.0259, "step": 49285 }, { "epoch": 0.12, "learning_rate": 8.785243343902562e-05, "loss": 1.0263, "step": 49290 }, { "epoch": 0.12, "learning_rate": 8.785117519754393e-05, "loss": 1.0279, "step": 49295 }, { "epoch": 0.12, "learning_rate": 8.784991695606222e-05, "loss": 1.0282, "step": 49300 }, { "epoch": 0.12, "learning_rate": 8.784865871458051e-05, "loss": 1.0286, "step": 49305 }, { "epoch": 0.12, "learning_rate": 8.78474004730988e-05, "loss": 1.0284, "step": 49310 }, { "epoch": 0.12, "learning_rate": 8.784614223161711e-05, "loss": 1.028, "step": 49315 }, { "epoch": 0.12, "learning_rate": 8.78448839901354e-05, "loss": 1.0279, "step": 49320 }, { "epoch": 0.12, "learning_rate": 8.784362574865369e-05, "loss": 1.0269, "step": 49325 }, { "epoch": 0.12, "learning_rate": 8.784236750717198e-05, "loss": 1.0277, "step": 49330 }, { "epoch": 0.12, "learning_rate": 8.784110926569027e-05, "loss": 1.0276, "step": 49335 }, { "epoch": 0.12, "learning_rate": 8.783985102420858e-05, "loss": 1.0303, "step": 49340 }, { "epoch": 0.12, "learning_rate": 8.783859278272687e-05, "loss": 1.0276, "step": 49345 }, { "epoch": 0.12, "learning_rate": 8.783733454124516e-05, "loss": 1.0489, "step": 49350 }, { "epoch": 0.12, "learning_rate": 8.783607629976345e-05, "loss": 1.0301, "step": 49355 }, { "epoch": 0.12, "learning_rate": 8.783481805828176e-05, "loss": 1.0278, "step": 49360 }, { "epoch": 0.12, "learning_rate": 8.783355981680005e-05, "loss": 1.0285, "step": 49365 }, { "epoch": 0.12, "learning_rate": 8.783230157531834e-05, "loss": 1.0283, "step": 49370 }, { "epoch": 0.12, "learning_rate": 8.783104333383663e-05, "loss": 1.0287, "step": 49375 }, { "epoch": 0.12, "learning_rate": 8.782978509235494e-05, "loss": 1.0291, "step": 49380 }, { "epoch": 0.12, "learning_rate": 8.782852685087323e-05, "loss": 1.03, "step": 49385 }, { "epoch": 0.12, "learning_rate": 8.782726860939152e-05, "loss": 1.0255, "step": 49390 }, { "epoch": 0.12, "learning_rate": 8.782601036790981e-05, "loss": 1.027, "step": 49395 }, { "epoch": 0.12, "learning_rate": 8.78247521264281e-05, "loss": 1.0267, "step": 49400 }, { "epoch": 0.12, "learning_rate": 8.782349388494641e-05, "loss": 1.0279, "step": 49405 }, { "epoch": 0.12, "learning_rate": 8.78222356434647e-05, "loss": 1.028, "step": 49410 }, { "epoch": 0.12, "learning_rate": 8.7820977401983e-05, "loss": 1.027, "step": 49415 }, { "epoch": 0.12, "learning_rate": 8.781971916050129e-05, "loss": 1.0289, "step": 49420 }, { "epoch": 0.12, "learning_rate": 8.781846091901959e-05, "loss": 1.0287, "step": 49425 }, { "epoch": 0.12, "learning_rate": 8.781720267753788e-05, "loss": 1.0279, "step": 49430 }, { "epoch": 0.12, "learning_rate": 8.781594443605617e-05, "loss": 1.0281, "step": 49435 }, { "epoch": 0.12, "learning_rate": 8.781468619457447e-05, "loss": 1.0479, "step": 49440 }, { "epoch": 0.12, "learning_rate": 8.781342795309276e-05, "loss": 1.029, "step": 49445 }, { "epoch": 0.12, "learning_rate": 8.781216971161106e-05, "loss": 1.028, "step": 49450 }, { "epoch": 0.12, "learning_rate": 8.781091147012935e-05, "loss": 1.0283, "step": 49455 }, { "epoch": 0.12, "learning_rate": 8.780965322864765e-05, "loss": 1.0283, "step": 49460 }, { "epoch": 0.12, "learning_rate": 8.780839498716594e-05, "loss": 1.0298, "step": 49465 }, { "epoch": 0.12, "learning_rate": 8.780713674568424e-05, "loss": 1.0267, "step": 49470 }, { "epoch": 0.12, "learning_rate": 8.780587850420253e-05, "loss": 1.0289, "step": 49475 }, { "epoch": 0.12, "learning_rate": 8.780462026272083e-05, "loss": 1.0288, "step": 49480 }, { "epoch": 0.12, "learning_rate": 8.780336202123912e-05, "loss": 1.0308, "step": 49485 }, { "epoch": 0.12, "learning_rate": 8.780210377975742e-05, "loss": 1.0301, "step": 49490 }, { "epoch": 0.12, "learning_rate": 8.780084553827571e-05, "loss": 1.0462, "step": 49495 }, { "epoch": 0.12, "learning_rate": 8.7799587296794e-05, "loss": 1.0284, "step": 49500 }, { "epoch": 0.12, "learning_rate": 8.77983290553123e-05, "loss": 1.0305, "step": 49505 }, { "epoch": 0.12, "learning_rate": 8.779707081383059e-05, "loss": 1.0273, "step": 49510 }, { "epoch": 0.12, "learning_rate": 8.77958125723489e-05, "loss": 1.0274, "step": 49515 }, { "epoch": 0.12, "learning_rate": 8.779455433086719e-05, "loss": 1.0267, "step": 49520 }, { "epoch": 0.12, "learning_rate": 8.779329608938548e-05, "loss": 1.0297, "step": 49525 }, { "epoch": 0.12, "learning_rate": 8.779203784790377e-05, "loss": 1.0244, "step": 49530 }, { "epoch": 0.12, "learning_rate": 8.779077960642207e-05, "loss": 1.028, "step": 49535 }, { "epoch": 0.12, "learning_rate": 8.778952136494037e-05, "loss": 1.0278, "step": 49540 }, { "epoch": 0.12, "learning_rate": 8.778826312345866e-05, "loss": 1.0319, "step": 49545 }, { "epoch": 0.12, "learning_rate": 8.778700488197695e-05, "loss": 1.0289, "step": 49550 }, { "epoch": 0.12, "learning_rate": 8.778574664049525e-05, "loss": 1.0296, "step": 49555 }, { "epoch": 0.12, "learning_rate": 8.778448839901355e-05, "loss": 1.0282, "step": 49560 }, { "epoch": 0.12, "learning_rate": 8.778323015753184e-05, "loss": 1.0295, "step": 49565 }, { "epoch": 0.12, "learning_rate": 8.778197191605013e-05, "loss": 1.0277, "step": 49570 }, { "epoch": 0.12, "learning_rate": 8.778071367456842e-05, "loss": 1.0264, "step": 49575 }, { "epoch": 0.12, "learning_rate": 8.777945543308673e-05, "loss": 1.0301, "step": 49580 }, { "epoch": 0.12, "learning_rate": 8.777819719160502e-05, "loss": 1.0266, "step": 49585 }, { "epoch": 0.12, "learning_rate": 8.777693895012331e-05, "loss": 1.0267, "step": 49590 }, { "epoch": 0.12, "learning_rate": 8.77756807086416e-05, "loss": 1.0253, "step": 49595 }, { "epoch": 0.12, "learning_rate": 8.77744224671599e-05, "loss": 1.0268, "step": 49600 }, { "epoch": 0.12, "learning_rate": 8.77731642256782e-05, "loss": 1.0282, "step": 49605 }, { "epoch": 0.12, "learning_rate": 8.777190598419649e-05, "loss": 1.0302, "step": 49610 }, { "epoch": 0.12, "learning_rate": 8.777064774271478e-05, "loss": 1.0251, "step": 49615 }, { "epoch": 0.12, "learning_rate": 8.776938950123309e-05, "loss": 1.0287, "step": 49620 }, { "epoch": 0.12, "learning_rate": 8.776813125975138e-05, "loss": 1.0291, "step": 49625 }, { "epoch": 0.12, "learning_rate": 8.776687301826967e-05, "loss": 1.0285, "step": 49630 }, { "epoch": 0.12, "learning_rate": 8.776561477678796e-05, "loss": 1.0278, "step": 49635 }, { "epoch": 0.12, "learning_rate": 8.776435653530625e-05, "loss": 1.0286, "step": 49640 }, { "epoch": 0.12, "learning_rate": 8.776309829382456e-05, "loss": 1.0278, "step": 49645 }, { "epoch": 0.12, "learning_rate": 8.776184005234285e-05, "loss": 1.0292, "step": 49650 }, { "epoch": 0.12, "learning_rate": 8.776058181086114e-05, "loss": 1.0276, "step": 49655 }, { "epoch": 0.12, "learning_rate": 8.775932356937943e-05, "loss": 1.027, "step": 49660 }, { "epoch": 0.12, "learning_rate": 8.775806532789774e-05, "loss": 1.0269, "step": 49665 }, { "epoch": 0.12, "learning_rate": 8.775680708641603e-05, "loss": 1.0302, "step": 49670 }, { "epoch": 0.12, "learning_rate": 8.775554884493432e-05, "loss": 1.03, "step": 49675 }, { "epoch": 0.12, "learning_rate": 8.775429060345261e-05, "loss": 1.0283, "step": 49680 }, { "epoch": 0.12, "learning_rate": 8.775303236197092e-05, "loss": 1.0267, "step": 49685 }, { "epoch": 0.12, "learning_rate": 8.775177412048921e-05, "loss": 1.0291, "step": 49690 }, { "epoch": 0.12, "learning_rate": 8.77505158790075e-05, "loss": 1.0278, "step": 49695 }, { "epoch": 0.12, "learning_rate": 8.774925763752579e-05, "loss": 1.0277, "step": 49700 }, { "epoch": 0.12, "learning_rate": 8.774799939604408e-05, "loss": 1.028, "step": 49705 }, { "epoch": 0.12, "learning_rate": 8.774674115456239e-05, "loss": 1.0265, "step": 49710 }, { "epoch": 0.12, "learning_rate": 8.774548291308068e-05, "loss": 1.0284, "step": 49715 }, { "epoch": 0.12, "learning_rate": 8.774422467159897e-05, "loss": 1.0258, "step": 49720 }, { "epoch": 0.12, "learning_rate": 8.774296643011726e-05, "loss": 1.031, "step": 49725 }, { "epoch": 0.12, "learning_rate": 8.774170818863557e-05, "loss": 1.0277, "step": 49730 }, { "epoch": 0.12, "learning_rate": 8.774044994715386e-05, "loss": 1.0282, "step": 49735 }, { "epoch": 0.12, "learning_rate": 8.773919170567215e-05, "loss": 1.0271, "step": 49740 }, { "epoch": 0.12, "learning_rate": 8.773793346419044e-05, "loss": 1.0264, "step": 49745 }, { "epoch": 0.12, "learning_rate": 8.773667522270875e-05, "loss": 1.0285, "step": 49750 }, { "epoch": 0.12, "learning_rate": 8.773541698122704e-05, "loss": 1.0285, "step": 49755 }, { "epoch": 0.12, "learning_rate": 8.773415873974533e-05, "loss": 1.0281, "step": 49760 }, { "epoch": 0.12, "learning_rate": 8.773290049826362e-05, "loss": 1.0286, "step": 49765 }, { "epoch": 0.12, "learning_rate": 8.773164225678191e-05, "loss": 1.0272, "step": 49770 }, { "epoch": 0.12, "learning_rate": 8.773038401530022e-05, "loss": 1.0281, "step": 49775 }, { "epoch": 0.12, "learning_rate": 8.772912577381851e-05, "loss": 1.0288, "step": 49780 }, { "epoch": 0.12, "learning_rate": 8.77278675323368e-05, "loss": 1.0251, "step": 49785 }, { "epoch": 0.12, "learning_rate": 8.77266092908551e-05, "loss": 1.0296, "step": 49790 }, { "epoch": 0.12, "learning_rate": 8.77253510493734e-05, "loss": 1.0289, "step": 49795 }, { "epoch": 0.13, "learning_rate": 8.77240928078917e-05, "loss": 1.0293, "step": 49800 }, { "epoch": 0.13, "learning_rate": 8.772283456641e-05, "loss": 1.03, "step": 49805 }, { "epoch": 0.13, "learning_rate": 8.772157632492829e-05, "loss": 1.0285, "step": 49810 }, { "epoch": 0.13, "learning_rate": 8.772031808344658e-05, "loss": 1.0283, "step": 49815 }, { "epoch": 0.13, "learning_rate": 8.771905984196488e-05, "loss": 1.0272, "step": 49820 }, { "epoch": 0.13, "learning_rate": 8.771780160048318e-05, "loss": 1.0297, "step": 49825 }, { "epoch": 0.13, "learning_rate": 8.771654335900147e-05, "loss": 1.0274, "step": 49830 }, { "epoch": 0.13, "learning_rate": 8.771528511751976e-05, "loss": 1.027, "step": 49835 }, { "epoch": 0.13, "learning_rate": 8.771402687603806e-05, "loss": 1.0276, "step": 49840 }, { "epoch": 0.13, "learning_rate": 8.771276863455636e-05, "loss": 1.028, "step": 49845 }, { "epoch": 0.13, "learning_rate": 8.771151039307465e-05, "loss": 1.0271, "step": 49850 }, { "epoch": 0.13, "learning_rate": 8.771025215159294e-05, "loss": 1.0286, "step": 49855 }, { "epoch": 0.13, "learning_rate": 8.770899391011123e-05, "loss": 1.0264, "step": 49860 }, { "epoch": 0.13, "learning_rate": 8.770773566862954e-05, "loss": 1.0287, "step": 49865 }, { "epoch": 0.13, "learning_rate": 8.770647742714783e-05, "loss": 1.0276, "step": 49870 }, { "epoch": 0.13, "learning_rate": 8.770521918566612e-05, "loss": 1.028, "step": 49875 }, { "epoch": 0.13, "learning_rate": 8.770396094418441e-05, "loss": 1.0264, "step": 49880 }, { "epoch": 0.13, "learning_rate": 8.770270270270272e-05, "loss": 1.0304, "step": 49885 }, { "epoch": 0.13, "learning_rate": 8.770144446122101e-05, "loss": 1.0289, "step": 49890 }, { "epoch": 0.13, "learning_rate": 8.77001862197393e-05, "loss": 1.0288, "step": 49895 }, { "epoch": 0.13, "learning_rate": 8.769892797825759e-05, "loss": 1.0273, "step": 49900 }, { "epoch": 0.13, "learning_rate": 8.76976697367759e-05, "loss": 1.0279, "step": 49905 }, { "epoch": 0.13, "learning_rate": 8.769641149529419e-05, "loss": 1.0407, "step": 49910 }, { "epoch": 0.13, "learning_rate": 8.769515325381248e-05, "loss": 1.0306, "step": 49915 }, { "epoch": 0.13, "learning_rate": 8.769389501233077e-05, "loss": 1.0271, "step": 49920 }, { "epoch": 0.13, "learning_rate": 8.769263677084906e-05, "loss": 1.0277, "step": 49925 }, { "epoch": 0.13, "learning_rate": 8.769137852936737e-05, "loss": 1.028, "step": 49930 }, { "epoch": 0.13, "learning_rate": 8.769012028788566e-05, "loss": 1.0278, "step": 49935 }, { "epoch": 0.13, "learning_rate": 8.768886204640395e-05, "loss": 1.0272, "step": 49940 }, { "epoch": 0.13, "learning_rate": 8.768760380492224e-05, "loss": 1.0273, "step": 49945 }, { "epoch": 0.13, "learning_rate": 8.768634556344055e-05, "loss": 1.0262, "step": 49950 }, { "epoch": 0.13, "learning_rate": 8.768508732195884e-05, "loss": 1.0498, "step": 49955 }, { "epoch": 0.13, "learning_rate": 8.768382908047713e-05, "loss": 1.0289, "step": 49960 }, { "epoch": 0.13, "learning_rate": 8.768257083899542e-05, "loss": 1.0302, "step": 49965 }, { "epoch": 0.13, "learning_rate": 8.768131259751373e-05, "loss": 1.0261, "step": 49970 }, { "epoch": 0.13, "learning_rate": 8.768005435603202e-05, "loss": 1.0295, "step": 49975 }, { "epoch": 0.13, "learning_rate": 8.767879611455031e-05, "loss": 1.0296, "step": 49980 }, { "epoch": 0.13, "learning_rate": 8.76775378730686e-05, "loss": 1.0287, "step": 49985 }, { "epoch": 0.13, "learning_rate": 8.76762796315869e-05, "loss": 1.0262, "step": 49990 }, { "epoch": 0.13, "learning_rate": 8.76750213901052e-05, "loss": 1.0297, "step": 49995 }, { "epoch": 0.13, "learning_rate": 8.767376314862349e-05, "loss": 1.0282, "step": 50000 }, { "epoch": 0.13, "learning_rate": 8.767250490714178e-05, "loss": 1.0285, "step": 50005 }, { "epoch": 0.13, "learning_rate": 8.767124666566007e-05, "loss": 1.0279, "step": 50010 }, { "epoch": 0.13, "learning_rate": 8.766998842417838e-05, "loss": 1.0268, "step": 50015 }, { "epoch": 0.13, "learning_rate": 8.766873018269667e-05, "loss": 1.0291, "step": 50020 }, { "epoch": 0.13, "learning_rate": 8.766747194121496e-05, "loss": 1.0288, "step": 50025 }, { "epoch": 0.13, "learning_rate": 8.766621369973325e-05, "loss": 1.0297, "step": 50030 }, { "epoch": 0.13, "learning_rate": 8.766495545825156e-05, "loss": 1.0294, "step": 50035 }, { "epoch": 0.13, "learning_rate": 8.766369721676985e-05, "loss": 1.025, "step": 50040 }, { "epoch": 0.13, "learning_rate": 8.766243897528814e-05, "loss": 1.0257, "step": 50045 }, { "epoch": 0.13, "learning_rate": 8.766118073380643e-05, "loss": 1.0283, "step": 50050 }, { "epoch": 0.13, "learning_rate": 8.765992249232473e-05, "loss": 1.0269, "step": 50055 }, { "epoch": 0.13, "learning_rate": 8.765866425084303e-05, "loss": 1.0264, "step": 50060 }, { "epoch": 0.13, "learning_rate": 8.765740600936132e-05, "loss": 1.0303, "step": 50065 }, { "epoch": 0.13, "learning_rate": 8.765614776787961e-05, "loss": 1.0292, "step": 50070 }, { "epoch": 0.13, "learning_rate": 8.76548895263979e-05, "loss": 1.0264, "step": 50075 }, { "epoch": 0.13, "learning_rate": 8.765363128491621e-05, "loss": 1.0262, "step": 50080 }, { "epoch": 0.13, "learning_rate": 8.76523730434345e-05, "loss": 1.0279, "step": 50085 }, { "epoch": 0.13, "learning_rate": 8.76511148019528e-05, "loss": 1.0267, "step": 50090 }, { "epoch": 0.13, "learning_rate": 8.764985656047108e-05, "loss": 1.0283, "step": 50095 }, { "epoch": 0.13, "learning_rate": 8.764859831898939e-05, "loss": 1.0267, "step": 50100 }, { "epoch": 0.13, "learning_rate": 8.764734007750768e-05, "loss": 1.027, "step": 50105 }, { "epoch": 0.13, "learning_rate": 8.764608183602597e-05, "loss": 1.0292, "step": 50110 }, { "epoch": 0.13, "learning_rate": 8.764482359454426e-05, "loss": 1.0287, "step": 50115 }, { "epoch": 0.13, "learning_rate": 8.764356535306256e-05, "loss": 1.0263, "step": 50120 }, { "epoch": 0.13, "learning_rate": 8.764230711158086e-05, "loss": 1.0273, "step": 50125 }, { "epoch": 0.13, "learning_rate": 8.764104887009915e-05, "loss": 1.0265, "step": 50130 }, { "epoch": 0.13, "learning_rate": 8.763979062861744e-05, "loss": 1.0295, "step": 50135 }, { "epoch": 0.13, "learning_rate": 8.763853238713574e-05, "loss": 1.0281, "step": 50140 }, { "epoch": 0.13, "learning_rate": 8.763727414565404e-05, "loss": 1.0291, "step": 50145 }, { "epoch": 0.13, "learning_rate": 8.763601590417233e-05, "loss": 1.0288, "step": 50150 }, { "epoch": 0.13, "learning_rate": 8.763475766269062e-05, "loss": 1.0296, "step": 50155 }, { "epoch": 0.13, "learning_rate": 8.763349942120892e-05, "loss": 1.0289, "step": 50160 }, { "epoch": 0.13, "learning_rate": 8.763224117972721e-05, "loss": 1.0268, "step": 50165 }, { "epoch": 0.13, "learning_rate": 8.763098293824551e-05, "loss": 1.0284, "step": 50170 }, { "epoch": 0.13, "learning_rate": 8.76297246967638e-05, "loss": 1.0298, "step": 50175 }, { "epoch": 0.13, "learning_rate": 8.76284664552821e-05, "loss": 1.0276, "step": 50180 }, { "epoch": 0.13, "learning_rate": 8.762720821380039e-05, "loss": 1.0284, "step": 50185 }, { "epoch": 0.13, "learning_rate": 8.762594997231869e-05, "loss": 1.0276, "step": 50190 }, { "epoch": 0.13, "learning_rate": 8.762469173083698e-05, "loss": 1.0282, "step": 50195 }, { "epoch": 0.13, "learning_rate": 8.762343348935528e-05, "loss": 1.0286, "step": 50200 }, { "epoch": 0.13, "learning_rate": 8.762217524787357e-05, "loss": 1.0288, "step": 50205 }, { "epoch": 0.13, "learning_rate": 8.762091700639187e-05, "loss": 1.0285, "step": 50210 }, { "epoch": 0.13, "learning_rate": 8.761965876491016e-05, "loss": 1.0266, "step": 50215 }, { "epoch": 0.13, "learning_rate": 8.761840052342846e-05, "loss": 1.0265, "step": 50220 }, { "epoch": 0.13, "learning_rate": 8.761714228194675e-05, "loss": 1.0294, "step": 50225 }, { "epoch": 0.13, "learning_rate": 8.761588404046504e-05, "loss": 1.0282, "step": 50230 }, { "epoch": 0.13, "learning_rate": 8.761462579898334e-05, "loss": 1.0281, "step": 50235 }, { "epoch": 0.13, "learning_rate": 8.761336755750164e-05, "loss": 1.0289, "step": 50240 }, { "epoch": 0.13, "learning_rate": 8.761210931601993e-05, "loss": 1.0286, "step": 50245 }, { "epoch": 0.13, "learning_rate": 8.761085107453822e-05, "loss": 1.0295, "step": 50250 }, { "epoch": 0.13, "learning_rate": 8.760959283305652e-05, "loss": 1.027, "step": 50255 }, { "epoch": 0.13, "learning_rate": 8.760833459157482e-05, "loss": 1.0289, "step": 50260 }, { "epoch": 0.13, "learning_rate": 8.760707635009311e-05, "loss": 1.0294, "step": 50265 }, { "epoch": 0.13, "learning_rate": 8.76058181086114e-05, "loss": 1.0291, "step": 50270 }, { "epoch": 0.13, "learning_rate": 8.76045598671297e-05, "loss": 1.03, "step": 50275 }, { "epoch": 0.13, "learning_rate": 8.7603301625648e-05, "loss": 1.0267, "step": 50280 }, { "epoch": 0.13, "learning_rate": 8.760204338416629e-05, "loss": 1.0282, "step": 50285 }, { "epoch": 0.13, "learning_rate": 8.760078514268458e-05, "loss": 1.0293, "step": 50290 }, { "epoch": 0.13, "learning_rate": 8.759952690120288e-05, "loss": 1.0278, "step": 50295 }, { "epoch": 0.13, "learning_rate": 8.759826865972119e-05, "loss": 1.0268, "step": 50300 }, { "epoch": 0.13, "learning_rate": 8.759701041823948e-05, "loss": 1.0279, "step": 50305 }, { "epoch": 0.13, "learning_rate": 8.759575217675777e-05, "loss": 1.0263, "step": 50310 }, { "epoch": 0.13, "learning_rate": 8.759449393527606e-05, "loss": 1.0294, "step": 50315 }, { "epoch": 0.13, "learning_rate": 8.759323569379436e-05, "loss": 1.0267, "step": 50320 }, { "epoch": 0.13, "learning_rate": 8.759197745231266e-05, "loss": 1.0278, "step": 50325 }, { "epoch": 0.13, "learning_rate": 8.759071921083095e-05, "loss": 1.0283, "step": 50330 }, { "epoch": 0.13, "learning_rate": 8.758946096934924e-05, "loss": 1.0288, "step": 50335 }, { "epoch": 0.13, "learning_rate": 8.758820272786754e-05, "loss": 1.0282, "step": 50340 }, { "epoch": 0.13, "learning_rate": 8.758694448638584e-05, "loss": 1.0279, "step": 50345 }, { "epoch": 0.13, "learning_rate": 8.758568624490413e-05, "loss": 1.0287, "step": 50350 }, { "epoch": 0.13, "learning_rate": 8.758442800342242e-05, "loss": 1.0279, "step": 50355 }, { "epoch": 0.13, "learning_rate": 8.758316976194072e-05, "loss": 1.0286, "step": 50360 }, { "epoch": 0.13, "learning_rate": 8.758191152045902e-05, "loss": 1.0469, "step": 50365 }, { "epoch": 0.13, "learning_rate": 8.758065327897731e-05, "loss": 1.029, "step": 50370 }, { "epoch": 0.13, "learning_rate": 8.75793950374956e-05, "loss": 1.0279, "step": 50375 }, { "epoch": 0.13, "learning_rate": 8.75781367960139e-05, "loss": 1.0274, "step": 50380 }, { "epoch": 0.13, "learning_rate": 8.757687855453219e-05, "loss": 1.0277, "step": 50385 }, { "epoch": 0.13, "learning_rate": 8.757562031305049e-05, "loss": 1.0297, "step": 50390 }, { "epoch": 0.13, "learning_rate": 8.757436207156878e-05, "loss": 1.0301, "step": 50395 }, { "epoch": 0.13, "learning_rate": 8.757310383008708e-05, "loss": 1.0296, "step": 50400 }, { "epoch": 0.13, "learning_rate": 8.757184558860537e-05, "loss": 1.0286, "step": 50405 }, { "epoch": 0.13, "learning_rate": 8.757058734712367e-05, "loss": 1.0284, "step": 50410 }, { "epoch": 0.13, "learning_rate": 8.756932910564196e-05, "loss": 1.0268, "step": 50415 }, { "epoch": 0.13, "learning_rate": 8.756807086416026e-05, "loss": 1.028, "step": 50420 }, { "epoch": 0.13, "learning_rate": 8.756681262267855e-05, "loss": 1.0246, "step": 50425 }, { "epoch": 0.13, "learning_rate": 8.756555438119685e-05, "loss": 1.027, "step": 50430 }, { "epoch": 0.13, "learning_rate": 8.756429613971514e-05, "loss": 1.0298, "step": 50435 }, { "epoch": 0.13, "learning_rate": 8.756303789823344e-05, "loss": 1.0292, "step": 50440 }, { "epoch": 0.13, "learning_rate": 8.756177965675173e-05, "loss": 1.0302, "step": 50445 }, { "epoch": 0.13, "learning_rate": 8.756052141527002e-05, "loss": 1.0284, "step": 50450 }, { "epoch": 0.13, "learning_rate": 8.755926317378832e-05, "loss": 1.024, "step": 50455 }, { "epoch": 0.13, "learning_rate": 8.755800493230662e-05, "loss": 1.0278, "step": 50460 }, { "epoch": 0.13, "learning_rate": 8.755674669082491e-05, "loss": 1.0282, "step": 50465 }, { "epoch": 0.13, "learning_rate": 8.75554884493432e-05, "loss": 1.0288, "step": 50470 }, { "epoch": 0.13, "learning_rate": 8.75542302078615e-05, "loss": 1.0303, "step": 50475 }, { "epoch": 0.13, "learning_rate": 8.75529719663798e-05, "loss": 1.0273, "step": 50480 }, { "epoch": 0.13, "learning_rate": 8.755171372489809e-05, "loss": 1.0277, "step": 50485 }, { "epoch": 0.13, "learning_rate": 8.755045548341638e-05, "loss": 1.0278, "step": 50490 }, { "epoch": 0.13, "learning_rate": 8.754919724193468e-05, "loss": 1.0285, "step": 50495 }, { "epoch": 0.13, "learning_rate": 8.754793900045298e-05, "loss": 1.0274, "step": 50500 }, { "epoch": 0.13, "learning_rate": 8.754668075897127e-05, "loss": 1.0268, "step": 50505 }, { "epoch": 0.13, "learning_rate": 8.754542251748956e-05, "loss": 1.0287, "step": 50510 }, { "epoch": 0.13, "learning_rate": 8.754416427600785e-05, "loss": 1.0504, "step": 50515 }, { "epoch": 0.13, "learning_rate": 8.754290603452616e-05, "loss": 1.0289, "step": 50520 }, { "epoch": 0.13, "learning_rate": 8.754164779304445e-05, "loss": 1.0274, "step": 50525 }, { "epoch": 0.13, "learning_rate": 8.754038955156274e-05, "loss": 1.0262, "step": 50530 }, { "epoch": 0.13, "learning_rate": 8.753913131008103e-05, "loss": 1.0273, "step": 50535 }, { "epoch": 0.13, "learning_rate": 8.753787306859934e-05, "loss": 1.027, "step": 50540 }, { "epoch": 0.13, "learning_rate": 8.753661482711763e-05, "loss": 1.0278, "step": 50545 }, { "epoch": 0.13, "learning_rate": 8.753535658563592e-05, "loss": 1.0281, "step": 50550 }, { "epoch": 0.13, "learning_rate": 8.753409834415421e-05, "loss": 1.0278, "step": 50555 }, { "epoch": 0.13, "learning_rate": 8.753284010267252e-05, "loss": 1.0255, "step": 50560 }, { "epoch": 0.13, "learning_rate": 8.753158186119081e-05, "loss": 1.031, "step": 50565 }, { "epoch": 0.13, "learning_rate": 8.75303236197091e-05, "loss": 1.0272, "step": 50570 }, { "epoch": 0.13, "learning_rate": 8.752906537822739e-05, "loss": 1.0285, "step": 50575 }, { "epoch": 0.13, "learning_rate": 8.752780713674568e-05, "loss": 1.0289, "step": 50580 }, { "epoch": 0.13, "learning_rate": 8.752654889526399e-05, "loss": 1.0287, "step": 50585 }, { "epoch": 0.13, "learning_rate": 8.752529065378228e-05, "loss": 1.0267, "step": 50590 }, { "epoch": 0.13, "learning_rate": 8.752403241230057e-05, "loss": 1.0486, "step": 50595 }, { "epoch": 0.13, "learning_rate": 8.752277417081886e-05, "loss": 1.0276, "step": 50600 }, { "epoch": 0.13, "learning_rate": 8.752151592933717e-05, "loss": 1.0494, "step": 50605 }, { "epoch": 0.13, "learning_rate": 8.752025768785546e-05, "loss": 1.0304, "step": 50610 }, { "epoch": 0.13, "learning_rate": 8.751899944637375e-05, "loss": 1.0278, "step": 50615 }, { "epoch": 0.13, "learning_rate": 8.751774120489204e-05, "loss": 1.0274, "step": 50620 }, { "epoch": 0.13, "learning_rate": 8.751648296341035e-05, "loss": 1.0276, "step": 50625 }, { "epoch": 0.13, "learning_rate": 8.751522472192864e-05, "loss": 1.0298, "step": 50630 }, { "epoch": 0.13, "learning_rate": 8.751396648044693e-05, "loss": 1.0285, "step": 50635 }, { "epoch": 0.13, "learning_rate": 8.751270823896522e-05, "loss": 1.0279, "step": 50640 }, { "epoch": 0.13, "learning_rate": 8.751144999748351e-05, "loss": 1.0257, "step": 50645 }, { "epoch": 0.13, "learning_rate": 8.751019175600182e-05, "loss": 1.0257, "step": 50650 }, { "epoch": 0.13, "learning_rate": 8.750893351452011e-05, "loss": 1.0304, "step": 50655 }, { "epoch": 0.13, "learning_rate": 8.75076752730384e-05, "loss": 1.0282, "step": 50660 }, { "epoch": 0.13, "learning_rate": 8.750641703155669e-05, "loss": 1.0296, "step": 50665 }, { "epoch": 0.13, "learning_rate": 8.7505158790075e-05, "loss": 1.0286, "step": 50670 }, { "epoch": 0.13, "learning_rate": 8.750390054859329e-05, "loss": 1.0292, "step": 50675 }, { "epoch": 0.13, "learning_rate": 8.750264230711158e-05, "loss": 1.0266, "step": 50680 }, { "epoch": 0.13, "learning_rate": 8.750138406562987e-05, "loss": 1.0285, "step": 50685 }, { "epoch": 0.13, "learning_rate": 8.750012582414818e-05, "loss": 1.0285, "step": 50690 }, { "epoch": 0.13, "learning_rate": 8.749886758266647e-05, "loss": 1.0271, "step": 50695 }, { "epoch": 0.13, "learning_rate": 8.749760934118476e-05, "loss": 1.028, "step": 50700 }, { "epoch": 0.13, "learning_rate": 8.749635109970305e-05, "loss": 1.0277, "step": 50705 }, { "epoch": 0.13, "learning_rate": 8.749509285822134e-05, "loss": 1.0276, "step": 50710 }, { "epoch": 0.13, "learning_rate": 8.749383461673965e-05, "loss": 1.0276, "step": 50715 }, { "epoch": 0.13, "learning_rate": 8.749257637525794e-05, "loss": 1.0267, "step": 50720 }, { "epoch": 0.13, "learning_rate": 8.749131813377623e-05, "loss": 1.0283, "step": 50725 }, { "epoch": 0.13, "learning_rate": 8.749005989229452e-05, "loss": 1.0268, "step": 50730 }, { "epoch": 0.13, "learning_rate": 8.748880165081283e-05, "loss": 1.0278, "step": 50735 }, { "epoch": 0.13, "learning_rate": 8.748754340933112e-05, "loss": 1.0283, "step": 50740 }, { "epoch": 0.13, "learning_rate": 8.748628516784941e-05, "loss": 1.0259, "step": 50745 }, { "epoch": 0.13, "learning_rate": 8.74850269263677e-05, "loss": 1.0288, "step": 50750 }, { "epoch": 0.13, "learning_rate": 8.748376868488601e-05, "loss": 1.0265, "step": 50755 }, { "epoch": 0.13, "learning_rate": 8.74825104434043e-05, "loss": 1.0263, "step": 50760 }, { "epoch": 0.13, "learning_rate": 8.748125220192259e-05, "loss": 1.0295, "step": 50765 }, { "epoch": 0.13, "learning_rate": 8.747999396044088e-05, "loss": 1.0292, "step": 50770 }, { "epoch": 0.13, "learning_rate": 8.747873571895918e-05, "loss": 1.0291, "step": 50775 }, { "epoch": 0.13, "learning_rate": 8.747747747747748e-05, "loss": 1.0284, "step": 50780 }, { "epoch": 0.13, "learning_rate": 8.747621923599577e-05, "loss": 1.0266, "step": 50785 }, { "epoch": 0.13, "learning_rate": 8.747496099451406e-05, "loss": 1.0278, "step": 50790 }, { "epoch": 0.13, "learning_rate": 8.747370275303237e-05, "loss": 1.0262, "step": 50795 }, { "epoch": 0.13, "learning_rate": 8.747244451155066e-05, "loss": 1.0288, "step": 50800 }, { "epoch": 0.13, "learning_rate": 8.747118627006897e-05, "loss": 1.0257, "step": 50805 }, { "epoch": 0.13, "learning_rate": 8.746992802858726e-05, "loss": 1.027, "step": 50810 }, { "epoch": 0.13, "learning_rate": 8.746866978710555e-05, "loss": 1.0282, "step": 50815 }, { "epoch": 0.13, "learning_rate": 8.746741154562384e-05, "loss": 1.0269, "step": 50820 }, { "epoch": 0.13, "learning_rate": 8.746615330414215e-05, "loss": 1.0268, "step": 50825 }, { "epoch": 0.13, "learning_rate": 8.746489506266044e-05, "loss": 1.0307, "step": 50830 }, { "epoch": 0.13, "learning_rate": 8.746363682117873e-05, "loss": 1.0284, "step": 50835 }, { "epoch": 0.13, "learning_rate": 8.746237857969702e-05, "loss": 1.0286, "step": 50840 }, { "epoch": 0.13, "learning_rate": 8.746112033821531e-05, "loss": 1.0276, "step": 50845 }, { "epoch": 0.13, "learning_rate": 8.745986209673362e-05, "loss": 1.0286, "step": 50850 }, { "epoch": 0.13, "learning_rate": 8.745860385525191e-05, "loss": 1.0274, "step": 50855 }, { "epoch": 0.13, "learning_rate": 8.74573456137702e-05, "loss": 1.0274, "step": 50860 }, { "epoch": 0.13, "learning_rate": 8.745608737228849e-05, "loss": 1.0272, "step": 50865 }, { "epoch": 0.13, "learning_rate": 8.74548291308068e-05, "loss": 1.0281, "step": 50870 }, { "epoch": 0.13, "learning_rate": 8.745357088932509e-05, "loss": 1.029, "step": 50875 }, { "epoch": 0.13, "learning_rate": 8.745231264784338e-05, "loss": 1.0292, "step": 50880 }, { "epoch": 0.13, "learning_rate": 8.745105440636167e-05, "loss": 1.0284, "step": 50885 }, { "epoch": 0.13, "learning_rate": 8.744979616487998e-05, "loss": 1.0301, "step": 50890 }, { "epoch": 0.13, "learning_rate": 8.744853792339827e-05, "loss": 1.0312, "step": 50895 }, { "epoch": 0.13, "learning_rate": 8.744727968191656e-05, "loss": 1.0295, "step": 50900 }, { "epoch": 0.13, "learning_rate": 8.744602144043485e-05, "loss": 1.0274, "step": 50905 }, { "epoch": 0.13, "learning_rate": 8.744476319895314e-05, "loss": 1.0278, "step": 50910 }, { "epoch": 0.13, "learning_rate": 8.744350495747145e-05, "loss": 1.0284, "step": 50915 }, { "epoch": 0.13, "learning_rate": 8.744224671598974e-05, "loss": 1.0273, "step": 50920 }, { "epoch": 0.13, "learning_rate": 8.744098847450803e-05, "loss": 1.0271, "step": 50925 }, { "epoch": 0.13, "learning_rate": 8.743973023302632e-05, "loss": 1.0281, "step": 50930 }, { "epoch": 0.13, "learning_rate": 8.743847199154463e-05, "loss": 1.026, "step": 50935 }, { "epoch": 0.13, "learning_rate": 8.743721375006292e-05, "loss": 1.0277, "step": 50940 }, { "epoch": 0.13, "learning_rate": 8.743595550858121e-05, "loss": 1.0301, "step": 50945 }, { "epoch": 0.13, "learning_rate": 8.74346972670995e-05, "loss": 1.0288, "step": 50950 }, { "epoch": 0.13, "learning_rate": 8.743343902561781e-05, "loss": 1.0266, "step": 50955 }, { "epoch": 0.13, "learning_rate": 8.74321807841361e-05, "loss": 1.0265, "step": 50960 }, { "epoch": 0.13, "learning_rate": 8.743092254265439e-05, "loss": 1.0261, "step": 50965 }, { "epoch": 0.13, "learning_rate": 8.742966430117268e-05, "loss": 1.0271, "step": 50970 }, { "epoch": 0.13, "learning_rate": 8.742840605969098e-05, "loss": 1.0261, "step": 50975 }, { "epoch": 0.13, "learning_rate": 8.742714781820928e-05, "loss": 1.0271, "step": 50980 }, { "epoch": 0.13, "learning_rate": 8.742588957672757e-05, "loss": 1.0288, "step": 50985 }, { "epoch": 0.13, "learning_rate": 8.742463133524586e-05, "loss": 1.0286, "step": 50990 }, { "epoch": 0.13, "learning_rate": 8.742337309376416e-05, "loss": 1.0277, "step": 50995 }, { "epoch": 0.13, "learning_rate": 8.742211485228246e-05, "loss": 1.0272, "step": 51000 }, { "epoch": 0.13, "learning_rate": 8.742085661080075e-05, "loss": 1.0285, "step": 51005 }, { "epoch": 0.13, "learning_rate": 8.741959836931904e-05, "loss": 1.0283, "step": 51010 }, { "epoch": 0.13, "learning_rate": 8.741834012783734e-05, "loss": 1.0288, "step": 51015 }, { "epoch": 0.13, "learning_rate": 8.741708188635564e-05, "loss": 1.0296, "step": 51020 }, { "epoch": 0.13, "learning_rate": 8.741582364487393e-05, "loss": 1.0291, "step": 51025 }, { "epoch": 0.13, "learning_rate": 8.741456540339222e-05, "loss": 1.0262, "step": 51030 }, { "epoch": 0.13, "learning_rate": 8.741330716191052e-05, "loss": 1.0321, "step": 51035 }, { "epoch": 0.13, "learning_rate": 8.74120489204288e-05, "loss": 1.0259, "step": 51040 }, { "epoch": 0.13, "learning_rate": 8.741079067894711e-05, "loss": 1.0286, "step": 51045 }, { "epoch": 0.13, "learning_rate": 8.74095324374654e-05, "loss": 1.0266, "step": 51050 }, { "epoch": 0.13, "learning_rate": 8.74082741959837e-05, "loss": 1.0274, "step": 51055 }, { "epoch": 0.13, "learning_rate": 8.740701595450199e-05, "loss": 1.0276, "step": 51060 }, { "epoch": 0.13, "learning_rate": 8.740575771302029e-05, "loss": 1.0279, "step": 51065 }, { "epoch": 0.13, "learning_rate": 8.740449947153858e-05, "loss": 1.0263, "step": 51070 }, { "epoch": 0.13, "learning_rate": 8.740324123005688e-05, "loss": 1.0286, "step": 51075 }, { "epoch": 0.13, "learning_rate": 8.740198298857517e-05, "loss": 1.0278, "step": 51080 }, { "epoch": 0.13, "learning_rate": 8.740072474709347e-05, "loss": 1.0282, "step": 51085 }, { "epoch": 0.13, "learning_rate": 8.739946650561176e-05, "loss": 1.0278, "step": 51090 }, { "epoch": 0.13, "learning_rate": 8.739820826413005e-05, "loss": 1.0285, "step": 51095 }, { "epoch": 0.13, "learning_rate": 8.739695002264835e-05, "loss": 1.0278, "step": 51100 }, { "epoch": 0.13, "learning_rate": 8.739569178116664e-05, "loss": 1.0259, "step": 51105 }, { "epoch": 0.13, "learning_rate": 8.739443353968494e-05, "loss": 1.0313, "step": 51110 }, { "epoch": 0.13, "learning_rate": 8.739317529820323e-05, "loss": 1.0279, "step": 51115 }, { "epoch": 0.13, "learning_rate": 8.739191705672153e-05, "loss": 1.0321, "step": 51120 }, { "epoch": 0.13, "learning_rate": 8.739065881523982e-05, "loss": 1.0294, "step": 51125 }, { "epoch": 0.13, "learning_rate": 8.738940057375812e-05, "loss": 1.0292, "step": 51130 }, { "epoch": 0.13, "learning_rate": 8.738814233227641e-05, "loss": 1.0286, "step": 51135 }, { "epoch": 0.13, "learning_rate": 8.73868840907947e-05, "loss": 1.0286, "step": 51140 }, { "epoch": 0.13, "learning_rate": 8.7385625849313e-05, "loss": 1.0275, "step": 51145 }, { "epoch": 0.13, "learning_rate": 8.73843676078313e-05, "loss": 1.0268, "step": 51150 }, { "epoch": 0.13, "learning_rate": 8.73831093663496e-05, "loss": 1.0257, "step": 51155 }, { "epoch": 0.13, "learning_rate": 8.738185112486789e-05, "loss": 1.0263, "step": 51160 }, { "epoch": 0.13, "learning_rate": 8.738059288338618e-05, "loss": 1.0281, "step": 51165 }, { "epoch": 0.13, "learning_rate": 8.737933464190447e-05, "loss": 1.0274, "step": 51170 }, { "epoch": 0.13, "learning_rate": 8.737807640042277e-05, "loss": 1.0399, "step": 51175 }, { "epoch": 0.13, "learning_rate": 8.737681815894107e-05, "loss": 1.0297, "step": 51180 }, { "epoch": 0.13, "learning_rate": 8.737555991745936e-05, "loss": 1.0291, "step": 51185 }, { "epoch": 0.13, "learning_rate": 8.737430167597765e-05, "loss": 1.0267, "step": 51190 }, { "epoch": 0.13, "learning_rate": 8.737304343449595e-05, "loss": 1.0261, "step": 51195 }, { "epoch": 0.13, "learning_rate": 8.737178519301425e-05, "loss": 1.0268, "step": 51200 }, { "epoch": 0.13, "learning_rate": 8.737052695153254e-05, "loss": 1.0288, "step": 51205 }, { "epoch": 0.13, "learning_rate": 8.736926871005083e-05, "loss": 1.0259, "step": 51210 }, { "epoch": 0.13, "learning_rate": 8.736801046856913e-05, "loss": 1.0287, "step": 51215 }, { "epoch": 0.13, "learning_rate": 8.736675222708743e-05, "loss": 1.0241, "step": 51220 }, { "epoch": 0.13, "learning_rate": 8.736549398560572e-05, "loss": 1.0275, "step": 51225 }, { "epoch": 0.13, "learning_rate": 8.736423574412401e-05, "loss": 1.0255, "step": 51230 }, { "epoch": 0.13, "learning_rate": 8.73629775026423e-05, "loss": 1.0306, "step": 51235 }, { "epoch": 0.13, "learning_rate": 8.73617192611606e-05, "loss": 1.0279, "step": 51240 }, { "epoch": 0.13, "learning_rate": 8.73604610196789e-05, "loss": 1.0258, "step": 51245 }, { "epoch": 0.13, "learning_rate": 8.735920277819719e-05, "loss": 1.029, "step": 51250 }, { "epoch": 0.13, "learning_rate": 8.735794453671548e-05, "loss": 1.0273, "step": 51255 }, { "epoch": 0.13, "learning_rate": 8.735668629523379e-05, "loss": 1.0531, "step": 51260 }, { "epoch": 0.13, "learning_rate": 8.735542805375208e-05, "loss": 1.0297, "step": 51265 }, { "epoch": 0.13, "learning_rate": 8.735416981227037e-05, "loss": 1.0283, "step": 51270 }, { "epoch": 0.13, "learning_rate": 8.735291157078866e-05, "loss": 1.0278, "step": 51275 }, { "epoch": 0.13, "learning_rate": 8.735165332930697e-05, "loss": 1.0485, "step": 51280 }, { "epoch": 0.13, "learning_rate": 8.735039508782526e-05, "loss": 1.027, "step": 51285 }, { "epoch": 0.13, "learning_rate": 8.734913684634356e-05, "loss": 1.0295, "step": 51290 }, { "epoch": 0.13, "learning_rate": 8.734787860486185e-05, "loss": 1.0265, "step": 51295 }, { "epoch": 0.13, "learning_rate": 8.734662036338015e-05, "loss": 1.0285, "step": 51300 }, { "epoch": 0.13, "learning_rate": 8.734536212189844e-05, "loss": 1.028, "step": 51305 }, { "epoch": 0.13, "learning_rate": 8.734410388041674e-05, "loss": 1.0275, "step": 51310 }, { "epoch": 0.13, "learning_rate": 8.734284563893503e-05, "loss": 1.0262, "step": 51315 }, { "epoch": 0.13, "learning_rate": 8.734158739745333e-05, "loss": 1.0279, "step": 51320 }, { "epoch": 0.13, "learning_rate": 8.734032915597162e-05, "loss": 1.027, "step": 51325 }, { "epoch": 0.13, "learning_rate": 8.733907091448992e-05, "loss": 1.0294, "step": 51330 }, { "epoch": 0.13, "learning_rate": 8.733781267300821e-05, "loss": 1.0267, "step": 51335 }, { "epoch": 0.13, "learning_rate": 8.73365544315265e-05, "loss": 1.0268, "step": 51340 }, { "epoch": 0.13, "learning_rate": 8.73352961900448e-05, "loss": 1.0266, "step": 51345 }, { "epoch": 0.13, "learning_rate": 8.73340379485631e-05, "loss": 1.0285, "step": 51350 }, { "epoch": 0.13, "learning_rate": 8.73327797070814e-05, "loss": 1.0291, "step": 51355 }, { "epoch": 0.13, "learning_rate": 8.733152146559969e-05, "loss": 1.0262, "step": 51360 }, { "epoch": 0.13, "learning_rate": 8.733026322411798e-05, "loss": 1.0271, "step": 51365 }, { "epoch": 0.13, "learning_rate": 8.732900498263627e-05, "loss": 1.0262, "step": 51370 }, { "epoch": 0.13, "learning_rate": 8.732774674115457e-05, "loss": 1.0269, "step": 51375 }, { "epoch": 0.13, "learning_rate": 8.732648849967287e-05, "loss": 1.028, "step": 51380 }, { "epoch": 0.13, "learning_rate": 8.732523025819116e-05, "loss": 1.0295, "step": 51385 }, { "epoch": 0.13, "learning_rate": 8.732397201670945e-05, "loss": 1.0284, "step": 51390 }, { "epoch": 0.13, "learning_rate": 8.732271377522775e-05, "loss": 1.0279, "step": 51395 }, { "epoch": 0.13, "learning_rate": 8.732145553374605e-05, "loss": 1.0291, "step": 51400 }, { "epoch": 0.13, "learning_rate": 8.732019729226434e-05, "loss": 1.0295, "step": 51405 }, { "epoch": 0.13, "learning_rate": 8.731893905078263e-05, "loss": 1.0276, "step": 51410 }, { "epoch": 0.13, "learning_rate": 8.731768080930093e-05, "loss": 1.0271, "step": 51415 }, { "epoch": 0.13, "learning_rate": 8.731642256781923e-05, "loss": 1.0285, "step": 51420 }, { "epoch": 0.13, "learning_rate": 8.731516432633752e-05, "loss": 1.0294, "step": 51425 }, { "epoch": 0.13, "learning_rate": 8.731390608485581e-05, "loss": 1.0282, "step": 51430 }, { "epoch": 0.13, "learning_rate": 8.73126478433741e-05, "loss": 1.0297, "step": 51435 }, { "epoch": 0.13, "learning_rate": 8.731164125018874e-05, "loss": 1.0256, "step": 51440 }, { "epoch": 0.13, "learning_rate": 8.731038300870703e-05, "loss": 1.0295, "step": 51445 }, { "epoch": 0.13, "learning_rate": 8.730912476722532e-05, "loss": 1.0274, "step": 51450 }, { "epoch": 0.13, "learning_rate": 8.730786652574362e-05, "loss": 1.0282, "step": 51455 }, { "epoch": 0.13, "learning_rate": 8.730660828426192e-05, "loss": 1.0279, "step": 51460 }, { "epoch": 0.13, "learning_rate": 8.730535004278021e-05, "loss": 1.0281, "step": 51465 }, { "epoch": 0.13, "learning_rate": 8.73040918012985e-05, "loss": 1.048, "step": 51470 }, { "epoch": 0.13, "learning_rate": 8.73028335598168e-05, "loss": 1.0423, "step": 51475 }, { "epoch": 0.13, "learning_rate": 8.73015753183351e-05, "loss": 1.0272, "step": 51480 }, { "epoch": 0.13, "learning_rate": 8.730031707685339e-05, "loss": 1.0301, "step": 51485 }, { "epoch": 0.13, "learning_rate": 8.729905883537168e-05, "loss": 1.0282, "step": 51490 }, { "epoch": 0.13, "learning_rate": 8.729780059388998e-05, "loss": 1.0277, "step": 51495 }, { "epoch": 0.13, "learning_rate": 8.729654235240829e-05, "loss": 1.0273, "step": 51500 }, { "epoch": 0.13, "learning_rate": 8.729528411092658e-05, "loss": 1.0282, "step": 51505 }, { "epoch": 0.13, "learning_rate": 8.729402586944487e-05, "loss": 1.0277, "step": 51510 }, { "epoch": 0.13, "learning_rate": 8.729276762796316e-05, "loss": 1.0296, "step": 51515 }, { "epoch": 0.13, "learning_rate": 8.729150938648146e-05, "loss": 1.0277, "step": 51520 }, { "epoch": 0.13, "learning_rate": 8.729025114499976e-05, "loss": 1.0298, "step": 51525 }, { "epoch": 0.13, "learning_rate": 8.728899290351805e-05, "loss": 1.0274, "step": 51530 }, { "epoch": 0.13, "learning_rate": 8.728773466203634e-05, "loss": 1.0285, "step": 51535 }, { "epoch": 0.13, "learning_rate": 8.728647642055464e-05, "loss": 1.024, "step": 51540 }, { "epoch": 0.13, "learning_rate": 8.728521817907294e-05, "loss": 1.0286, "step": 51545 }, { "epoch": 0.13, "learning_rate": 8.728395993759123e-05, "loss": 1.0282, "step": 51550 }, { "epoch": 0.13, "learning_rate": 8.728270169610952e-05, "loss": 1.0277, "step": 51555 }, { "epoch": 0.13, "learning_rate": 8.728144345462782e-05, "loss": 1.0268, "step": 51560 }, { "epoch": 0.13, "learning_rate": 8.728018521314612e-05, "loss": 1.0268, "step": 51565 }, { "epoch": 0.13, "learning_rate": 8.727892697166441e-05, "loss": 1.0283, "step": 51570 }, { "epoch": 0.13, "learning_rate": 8.72776687301827e-05, "loss": 1.0292, "step": 51575 }, { "epoch": 0.13, "learning_rate": 8.7276410488701e-05, "loss": 1.0293, "step": 51580 }, { "epoch": 0.13, "learning_rate": 8.727515224721929e-05, "loss": 1.0267, "step": 51585 }, { "epoch": 0.13, "learning_rate": 8.727389400573759e-05, "loss": 1.0277, "step": 51590 }, { "epoch": 0.13, "learning_rate": 8.727263576425588e-05, "loss": 1.0282, "step": 51595 }, { "epoch": 0.13, "learning_rate": 8.727137752277418e-05, "loss": 1.0278, "step": 51600 }, { "epoch": 0.13, "learning_rate": 8.727011928129247e-05, "loss": 1.0265, "step": 51605 }, { "epoch": 0.13, "learning_rate": 8.726886103981077e-05, "loss": 1.0296, "step": 51610 }, { "epoch": 0.13, "learning_rate": 8.726760279832906e-05, "loss": 1.0281, "step": 51615 }, { "epoch": 0.13, "learning_rate": 8.726634455684736e-05, "loss": 1.0274, "step": 51620 }, { "epoch": 0.13, "learning_rate": 8.726508631536565e-05, "loss": 1.029, "step": 51625 }, { "epoch": 0.13, "learning_rate": 8.726382807388395e-05, "loss": 1.0282, "step": 51630 }, { "epoch": 0.13, "learning_rate": 8.726256983240224e-05, "loss": 1.0279, "step": 51635 }, { "epoch": 0.13, "learning_rate": 8.726131159092054e-05, "loss": 1.0292, "step": 51640 }, { "epoch": 0.13, "learning_rate": 8.726005334943883e-05, "loss": 1.0297, "step": 51645 }, { "epoch": 0.13, "learning_rate": 8.725879510795712e-05, "loss": 1.027, "step": 51650 }, { "epoch": 0.13, "learning_rate": 8.725753686647542e-05, "loss": 1.0279, "step": 51655 }, { "epoch": 0.13, "learning_rate": 8.725627862499372e-05, "loss": 1.0495, "step": 51660 }, { "epoch": 0.13, "learning_rate": 8.725502038351201e-05, "loss": 1.0284, "step": 51665 }, { "epoch": 0.13, "learning_rate": 8.72537621420303e-05, "loss": 1.0265, "step": 51670 }, { "epoch": 0.13, "learning_rate": 8.72525039005486e-05, "loss": 1.0274, "step": 51675 }, { "epoch": 0.13, "learning_rate": 8.72512456590669e-05, "loss": 1.0277, "step": 51680 }, { "epoch": 0.13, "learning_rate": 8.724998741758519e-05, "loss": 1.0286, "step": 51685 }, { "epoch": 0.13, "learning_rate": 8.724872917610348e-05, "loss": 1.0253, "step": 51690 }, { "epoch": 0.13, "learning_rate": 8.724747093462178e-05, "loss": 1.0284, "step": 51695 }, { "epoch": 0.13, "learning_rate": 8.724621269314008e-05, "loss": 1.0293, "step": 51700 }, { "epoch": 0.13, "learning_rate": 8.724495445165837e-05, "loss": 1.0284, "step": 51705 }, { "epoch": 0.13, "learning_rate": 8.724369621017666e-05, "loss": 1.0282, "step": 51710 }, { "epoch": 0.13, "learning_rate": 8.724243796869495e-05, "loss": 1.0256, "step": 51715 }, { "epoch": 0.13, "learning_rate": 8.724117972721326e-05, "loss": 1.0268, "step": 51720 }, { "epoch": 0.13, "learning_rate": 8.723992148573155e-05, "loss": 1.0269, "step": 51725 }, { "epoch": 0.13, "learning_rate": 8.723866324424984e-05, "loss": 1.0279, "step": 51730 }, { "epoch": 0.13, "learning_rate": 8.723740500276813e-05, "loss": 1.029, "step": 51735 }, { "epoch": 0.13, "learning_rate": 8.723614676128644e-05, "loss": 1.0266, "step": 51740 }, { "epoch": 0.13, "learning_rate": 8.723488851980473e-05, "loss": 1.0266, "step": 51745 }, { "epoch": 0.13, "learning_rate": 8.723363027832302e-05, "loss": 1.029, "step": 51750 }, { "epoch": 0.13, "learning_rate": 8.723237203684131e-05, "loss": 1.0278, "step": 51755 }, { "epoch": 0.13, "learning_rate": 8.723111379535962e-05, "loss": 1.0283, "step": 51760 }, { "epoch": 0.13, "learning_rate": 8.722985555387791e-05, "loss": 1.0263, "step": 51765 }, { "epoch": 0.13, "learning_rate": 8.72285973123962e-05, "loss": 1.0283, "step": 51770 }, { "epoch": 0.13, "learning_rate": 8.722733907091449e-05, "loss": 1.0275, "step": 51775 }, { "epoch": 0.13, "learning_rate": 8.722608082943278e-05, "loss": 1.0277, "step": 51780 }, { "epoch": 0.13, "learning_rate": 8.722482258795109e-05, "loss": 1.0249, "step": 51785 }, { "epoch": 0.13, "learning_rate": 8.722356434646938e-05, "loss": 1.0475, "step": 51790 }, { "epoch": 0.13, "learning_rate": 8.722230610498767e-05, "loss": 1.0298, "step": 51795 }, { "epoch": 0.13, "learning_rate": 8.722104786350596e-05, "loss": 1.0282, "step": 51800 }, { "epoch": 0.13, "learning_rate": 8.721978962202427e-05, "loss": 1.0298, "step": 51805 }, { "epoch": 0.13, "learning_rate": 8.721853138054256e-05, "loss": 1.0237, "step": 51810 }, { "epoch": 0.13, "learning_rate": 8.721727313906085e-05, "loss": 1.0274, "step": 51815 }, { "epoch": 0.13, "learning_rate": 8.721601489757914e-05, "loss": 1.0284, "step": 51820 }, { "epoch": 0.13, "learning_rate": 8.721475665609745e-05, "loss": 1.0248, "step": 51825 }, { "epoch": 0.13, "learning_rate": 8.721349841461574e-05, "loss": 1.0283, "step": 51830 }, { "epoch": 0.13, "learning_rate": 8.721224017313403e-05, "loss": 1.0289, "step": 51835 }, { "epoch": 0.13, "learning_rate": 8.721098193165232e-05, "loss": 1.0294, "step": 51840 }, { "epoch": 0.13, "learning_rate": 8.720972369017061e-05, "loss": 1.0268, "step": 51845 }, { "epoch": 0.13, "learning_rate": 8.720846544868892e-05, "loss": 1.0303, "step": 51850 }, { "epoch": 0.13, "learning_rate": 8.720720720720721e-05, "loss": 1.0261, "step": 51855 }, { "epoch": 0.13, "learning_rate": 8.72059489657255e-05, "loss": 1.0291, "step": 51860 }, { "epoch": 0.13, "learning_rate": 8.72046907242438e-05, "loss": 1.028, "step": 51865 }, { "epoch": 0.13, "learning_rate": 8.72034324827621e-05, "loss": 1.0279, "step": 51870 }, { "epoch": 0.13, "learning_rate": 8.720217424128039e-05, "loss": 1.0294, "step": 51875 }, { "epoch": 0.13, "learning_rate": 8.720091599979868e-05, "loss": 1.0287, "step": 51880 }, { "epoch": 0.13, "learning_rate": 8.719965775831697e-05, "loss": 1.0268, "step": 51885 }, { "epoch": 0.13, "learning_rate": 8.719839951683528e-05, "loss": 1.0265, "step": 51890 }, { "epoch": 0.13, "learning_rate": 8.719714127535357e-05, "loss": 1.0259, "step": 51895 }, { "epoch": 0.13, "learning_rate": 8.719588303387186e-05, "loss": 1.0261, "step": 51900 }, { "epoch": 0.13, "learning_rate": 8.719462479239015e-05, "loss": 1.0504, "step": 51905 }, { "epoch": 0.13, "learning_rate": 8.719336655090844e-05, "loss": 1.026, "step": 51910 }, { "epoch": 0.13, "learning_rate": 8.719210830942675e-05, "loss": 1.0288, "step": 51915 }, { "epoch": 0.13, "learning_rate": 8.719085006794504e-05, "loss": 1.0277, "step": 51920 }, { "epoch": 0.13, "learning_rate": 8.718959182646333e-05, "loss": 1.0271, "step": 51925 }, { "epoch": 0.13, "learning_rate": 8.718833358498162e-05, "loss": 1.0265, "step": 51930 }, { "epoch": 0.13, "learning_rate": 8.718707534349993e-05, "loss": 1.0285, "step": 51935 }, { "epoch": 0.13, "learning_rate": 8.718581710201822e-05, "loss": 1.0273, "step": 51940 }, { "epoch": 0.13, "learning_rate": 8.718455886053651e-05, "loss": 1.0266, "step": 51945 }, { "epoch": 0.13, "learning_rate": 8.71833006190548e-05, "loss": 1.0262, "step": 51950 }, { "epoch": 0.13, "learning_rate": 8.718204237757311e-05, "loss": 1.0251, "step": 51955 }, { "epoch": 0.13, "learning_rate": 8.71807841360914e-05, "loss": 1.0283, "step": 51960 }, { "epoch": 0.13, "learning_rate": 8.717952589460969e-05, "loss": 1.0272, "step": 51965 }, { "epoch": 0.13, "learning_rate": 8.717826765312798e-05, "loss": 1.0273, "step": 51970 }, { "epoch": 0.13, "learning_rate": 8.717700941164628e-05, "loss": 1.0293, "step": 51975 }, { "epoch": 0.13, "learning_rate": 8.717575117016458e-05, "loss": 1.0276, "step": 51980 }, { "epoch": 0.13, "learning_rate": 8.717449292868287e-05, "loss": 1.0293, "step": 51985 }, { "epoch": 0.13, "learning_rate": 8.717323468720118e-05, "loss": 1.0304, "step": 51990 }, { "epoch": 0.13, "learning_rate": 8.717197644571947e-05, "loss": 1.0279, "step": 51995 }, { "epoch": 0.13, "learning_rate": 8.717071820423776e-05, "loss": 1.0288, "step": 52000 }, { "epoch": 0.13, "learning_rate": 8.716945996275607e-05, "loss": 1.0286, "step": 52005 }, { "epoch": 0.13, "learning_rate": 8.716820172127436e-05, "loss": 1.0274, "step": 52010 }, { "epoch": 0.13, "learning_rate": 8.716694347979265e-05, "loss": 1.0267, "step": 52015 }, { "epoch": 0.13, "learning_rate": 8.716568523831094e-05, "loss": 1.028, "step": 52020 }, { "epoch": 0.13, "learning_rate": 8.716442699682925e-05, "loss": 1.0286, "step": 52025 }, { "epoch": 0.13, "learning_rate": 8.716316875534754e-05, "loss": 1.0256, "step": 52030 }, { "epoch": 0.13, "learning_rate": 8.716191051386583e-05, "loss": 1.0283, "step": 52035 }, { "epoch": 0.13, "learning_rate": 8.716065227238412e-05, "loss": 1.0276, "step": 52040 }, { "epoch": 0.13, "learning_rate": 8.715939403090241e-05, "loss": 1.0271, "step": 52045 }, { "epoch": 0.13, "learning_rate": 8.715813578942072e-05, "loss": 1.0295, "step": 52050 }, { "epoch": 0.13, "learning_rate": 8.715687754793901e-05, "loss": 1.0277, "step": 52055 }, { "epoch": 0.13, "learning_rate": 8.71556193064573e-05, "loss": 1.0303, "step": 52060 }, { "epoch": 0.13, "learning_rate": 8.715436106497559e-05, "loss": 1.0276, "step": 52065 }, { "epoch": 0.13, "learning_rate": 8.71531028234939e-05, "loss": 1.0296, "step": 52070 }, { "epoch": 0.13, "learning_rate": 8.715184458201219e-05, "loss": 1.0289, "step": 52075 }, { "epoch": 0.13, "learning_rate": 8.715058634053048e-05, "loss": 1.0305, "step": 52080 }, { "epoch": 0.13, "learning_rate": 8.714932809904877e-05, "loss": 1.0266, "step": 52085 }, { "epoch": 0.13, "learning_rate": 8.714806985756708e-05, "loss": 1.0276, "step": 52090 }, { "epoch": 0.13, "learning_rate": 8.714681161608537e-05, "loss": 1.0287, "step": 52095 }, { "epoch": 0.13, "learning_rate": 8.714555337460366e-05, "loss": 1.0264, "step": 52100 }, { "epoch": 0.13, "learning_rate": 8.714429513312195e-05, "loss": 1.0275, "step": 52105 }, { "epoch": 0.13, "learning_rate": 8.714303689164024e-05, "loss": 1.0272, "step": 52110 }, { "epoch": 0.13, "learning_rate": 8.714177865015855e-05, "loss": 1.0281, "step": 52115 }, { "epoch": 0.13, "learning_rate": 8.714052040867684e-05, "loss": 1.0286, "step": 52120 }, { "epoch": 0.13, "learning_rate": 8.713926216719513e-05, "loss": 1.0295, "step": 52125 }, { "epoch": 0.13, "learning_rate": 8.713800392571342e-05, "loss": 1.0269, "step": 52130 }, { "epoch": 0.13, "learning_rate": 8.713674568423173e-05, "loss": 1.027, "step": 52135 }, { "epoch": 0.13, "learning_rate": 8.713548744275002e-05, "loss": 1.0285, "step": 52140 }, { "epoch": 0.13, "learning_rate": 8.713422920126831e-05, "loss": 1.0273, "step": 52145 }, { "epoch": 0.13, "learning_rate": 8.71329709597866e-05, "loss": 1.0263, "step": 52150 }, { "epoch": 0.13, "learning_rate": 8.713171271830491e-05, "loss": 1.0306, "step": 52155 }, { "epoch": 0.13, "learning_rate": 8.71304544768232e-05, "loss": 1.0291, "step": 52160 }, { "epoch": 0.13, "learning_rate": 8.712919623534149e-05, "loss": 1.0284, "step": 52165 }, { "epoch": 0.13, "learning_rate": 8.712793799385978e-05, "loss": 1.0274, "step": 52170 }, { "epoch": 0.13, "learning_rate": 8.712667975237808e-05, "loss": 1.0292, "step": 52175 }, { "epoch": 0.13, "learning_rate": 8.712542151089638e-05, "loss": 1.0486, "step": 52180 }, { "epoch": 0.13, "learning_rate": 8.712416326941467e-05, "loss": 1.0269, "step": 52185 }, { "epoch": 0.13, "learning_rate": 8.712290502793296e-05, "loss": 1.029, "step": 52190 }, { "epoch": 0.13, "learning_rate": 8.712164678645126e-05, "loss": 1.0278, "step": 52195 }, { "epoch": 0.13, "learning_rate": 8.712038854496956e-05, "loss": 1.0283, "step": 52200 }, { "epoch": 0.13, "learning_rate": 8.711913030348785e-05, "loss": 1.0294, "step": 52205 }, { "epoch": 0.13, "learning_rate": 8.711787206200614e-05, "loss": 1.0278, "step": 52210 }, { "epoch": 0.13, "learning_rate": 8.711661382052444e-05, "loss": 1.0266, "step": 52215 }, { "epoch": 0.13, "learning_rate": 8.711535557904274e-05, "loss": 1.0276, "step": 52220 }, { "epoch": 0.13, "learning_rate": 8.711409733756103e-05, "loss": 1.0271, "step": 52225 }, { "epoch": 0.13, "learning_rate": 8.711283909607932e-05, "loss": 1.0248, "step": 52230 }, { "epoch": 0.13, "learning_rate": 8.711158085459762e-05, "loss": 1.0269, "step": 52235 }, { "epoch": 0.13, "learning_rate": 8.711032261311591e-05, "loss": 1.0287, "step": 52240 }, { "epoch": 0.13, "learning_rate": 8.710906437163421e-05, "loss": 1.0282, "step": 52245 }, { "epoch": 0.13, "learning_rate": 8.71078061301525e-05, "loss": 1.0298, "step": 52250 }, { "epoch": 0.13, "learning_rate": 8.71065478886708e-05, "loss": 1.0273, "step": 52255 }, { "epoch": 0.13, "learning_rate": 8.710528964718909e-05, "loss": 1.0272, "step": 52260 }, { "epoch": 0.13, "learning_rate": 8.710403140570739e-05, "loss": 1.0291, "step": 52265 }, { "epoch": 0.13, "learning_rate": 8.710277316422568e-05, "loss": 1.028, "step": 52270 }, { "epoch": 0.13, "learning_rate": 8.710151492274398e-05, "loss": 1.0279, "step": 52275 }, { "epoch": 0.13, "learning_rate": 8.710025668126227e-05, "loss": 1.0288, "step": 52280 }, { "epoch": 0.13, "learning_rate": 8.709899843978057e-05, "loss": 1.0286, "step": 52285 }, { "epoch": 0.13, "learning_rate": 8.709774019829886e-05, "loss": 1.0276, "step": 52290 }, { "epoch": 0.13, "learning_rate": 8.709648195681716e-05, "loss": 1.028, "step": 52295 }, { "epoch": 0.13, "learning_rate": 8.709522371533545e-05, "loss": 1.0269, "step": 52300 }, { "epoch": 0.13, "learning_rate": 8.709396547385374e-05, "loss": 1.0268, "step": 52305 }, { "epoch": 0.13, "learning_rate": 8.709270723237204e-05, "loss": 1.0301, "step": 52310 }, { "epoch": 0.13, "learning_rate": 8.709144899089034e-05, "loss": 1.026, "step": 52315 }, { "epoch": 0.13, "learning_rate": 8.709019074940863e-05, "loss": 1.0291, "step": 52320 }, { "epoch": 0.13, "learning_rate": 8.708893250792692e-05, "loss": 1.0289, "step": 52325 }, { "epoch": 0.13, "learning_rate": 8.708767426644522e-05, "loss": 1.0268, "step": 52330 }, { "epoch": 0.13, "learning_rate": 8.708641602496352e-05, "loss": 1.0526, "step": 52335 }, { "epoch": 0.13, "learning_rate": 8.70851577834818e-05, "loss": 1.0261, "step": 52340 }, { "epoch": 0.13, "learning_rate": 8.70838995420001e-05, "loss": 1.0261, "step": 52345 }, { "epoch": 0.13, "learning_rate": 8.70826413005184e-05, "loss": 1.0314, "step": 52350 }, { "epoch": 0.13, "learning_rate": 8.70813830590367e-05, "loss": 1.0278, "step": 52355 }, { "epoch": 0.13, "learning_rate": 8.708012481755499e-05, "loss": 1.0275, "step": 52360 }, { "epoch": 0.13, "learning_rate": 8.707886657607328e-05, "loss": 1.0274, "step": 52365 }, { "epoch": 0.13, "learning_rate": 8.707760833459157e-05, "loss": 1.0272, "step": 52370 }, { "epoch": 0.13, "learning_rate": 8.707635009310988e-05, "loss": 1.0256, "step": 52375 }, { "epoch": 0.13, "learning_rate": 8.707509185162817e-05, "loss": 1.0246, "step": 52380 }, { "epoch": 0.13, "learning_rate": 8.707383361014646e-05, "loss": 1.0308, "step": 52385 }, { "epoch": 0.13, "learning_rate": 8.707257536866475e-05, "loss": 1.0249, "step": 52390 }, { "epoch": 0.13, "learning_rate": 8.707131712718305e-05, "loss": 1.0285, "step": 52395 }, { "epoch": 0.13, "learning_rate": 8.707005888570135e-05, "loss": 1.029, "step": 52400 }, { "epoch": 0.13, "learning_rate": 8.706880064421964e-05, "loss": 1.028, "step": 52405 }, { "epoch": 0.13, "learning_rate": 8.706754240273793e-05, "loss": 1.0285, "step": 52410 }, { "epoch": 0.13, "learning_rate": 8.706628416125623e-05, "loss": 1.0263, "step": 52415 }, { "epoch": 0.13, "learning_rate": 8.706502591977453e-05, "loss": 1.0439, "step": 52420 }, { "epoch": 0.13, "learning_rate": 8.706376767829282e-05, "loss": 1.0264, "step": 52425 }, { "epoch": 0.13, "learning_rate": 8.706250943681111e-05, "loss": 1.0265, "step": 52430 }, { "epoch": 0.13, "learning_rate": 8.70612511953294e-05, "loss": 1.0479, "step": 52435 }, { "epoch": 0.13, "learning_rate": 8.70599929538477e-05, "loss": 1.0268, "step": 52440 }, { "epoch": 0.13, "learning_rate": 8.7058734712366e-05, "loss": 1.0308, "step": 52445 }, { "epoch": 0.13, "learning_rate": 8.705747647088429e-05, "loss": 1.0259, "step": 52450 }, { "epoch": 0.13, "learning_rate": 8.705621822940258e-05, "loss": 1.0311, "step": 52455 }, { "epoch": 0.13, "learning_rate": 8.705495998792089e-05, "loss": 1.0281, "step": 52460 }, { "epoch": 0.13, "learning_rate": 8.705370174643918e-05, "loss": 1.0286, "step": 52465 }, { "epoch": 0.13, "learning_rate": 8.705244350495747e-05, "loss": 1.0269, "step": 52470 }, { "epoch": 0.13, "learning_rate": 8.705118526347576e-05, "loss": 1.0281, "step": 52475 }, { "epoch": 0.13, "learning_rate": 8.704992702199407e-05, "loss": 1.0282, "step": 52480 }, { "epoch": 0.13, "learning_rate": 8.704866878051236e-05, "loss": 1.0265, "step": 52485 }, { "epoch": 0.13, "learning_rate": 8.704741053903066e-05, "loss": 1.025, "step": 52490 }, { "epoch": 0.13, "learning_rate": 8.704615229754895e-05, "loss": 1.0276, "step": 52495 }, { "epoch": 0.13, "learning_rate": 8.704489405606725e-05, "loss": 1.0283, "step": 52500 }, { "epoch": 0.13, "learning_rate": 8.704363581458554e-05, "loss": 1.0275, "step": 52505 }, { "epoch": 0.13, "learning_rate": 8.704237757310384e-05, "loss": 1.0293, "step": 52510 }, { "epoch": 0.13, "learning_rate": 8.704111933162213e-05, "loss": 1.0266, "step": 52515 }, { "epoch": 0.13, "learning_rate": 8.703986109014043e-05, "loss": 1.0272, "step": 52520 }, { "epoch": 0.13, "learning_rate": 8.703860284865872e-05, "loss": 1.0254, "step": 52525 }, { "epoch": 0.13, "learning_rate": 8.703734460717702e-05, "loss": 1.0289, "step": 52530 }, { "epoch": 0.13, "learning_rate": 8.703608636569531e-05, "loss": 1.0268, "step": 52535 }, { "epoch": 0.13, "learning_rate": 8.70348281242136e-05, "loss": 1.0279, "step": 52540 }, { "epoch": 0.13, "learning_rate": 8.70335698827319e-05, "loss": 1.0276, "step": 52545 }, { "epoch": 0.13, "learning_rate": 8.70323116412502e-05, "loss": 1.0267, "step": 52550 }, { "epoch": 0.13, "learning_rate": 8.70310533997685e-05, "loss": 1.0249, "step": 52555 }, { "epoch": 0.13, "learning_rate": 8.702979515828679e-05, "loss": 1.0294, "step": 52560 }, { "epoch": 0.13, "learning_rate": 8.702853691680508e-05, "loss": 1.0297, "step": 52565 }, { "epoch": 0.13, "learning_rate": 8.702727867532337e-05, "loss": 1.0284, "step": 52570 }, { "epoch": 0.13, "learning_rate": 8.702602043384167e-05, "loss": 1.0252, "step": 52575 }, { "epoch": 0.13, "learning_rate": 8.702476219235997e-05, "loss": 1.0275, "step": 52580 }, { "epoch": 0.13, "learning_rate": 8.702350395087826e-05, "loss": 1.029, "step": 52585 }, { "epoch": 0.13, "learning_rate": 8.702224570939655e-05, "loss": 1.027, "step": 52590 }, { "epoch": 0.13, "learning_rate": 8.702098746791485e-05, "loss": 1.0273, "step": 52595 }, { "epoch": 0.13, "learning_rate": 8.701972922643315e-05, "loss": 1.0255, "step": 52600 }, { "epoch": 0.13, "learning_rate": 8.701847098495144e-05, "loss": 1.0282, "step": 52605 }, { "epoch": 0.13, "learning_rate": 8.701721274346973e-05, "loss": 1.0291, "step": 52610 }, { "epoch": 0.13, "learning_rate": 8.701595450198803e-05, "loss": 1.0258, "step": 52615 }, { "epoch": 0.13, "learning_rate": 8.701469626050633e-05, "loss": 1.0272, "step": 52620 }, { "epoch": 0.13, "learning_rate": 8.701343801902462e-05, "loss": 1.0251, "step": 52625 }, { "epoch": 0.13, "learning_rate": 8.701217977754291e-05, "loss": 1.0285, "step": 52630 }, { "epoch": 0.13, "learning_rate": 8.70109215360612e-05, "loss": 1.0249, "step": 52635 }, { "epoch": 0.13, "learning_rate": 8.70096632945795e-05, "loss": 1.0303, "step": 52640 }, { "epoch": 0.13, "learning_rate": 8.70084050530978e-05, "loss": 1.0275, "step": 52645 }, { "epoch": 0.13, "learning_rate": 8.700714681161609e-05, "loss": 1.0252, "step": 52650 }, { "epoch": 0.13, "learning_rate": 8.700588857013438e-05, "loss": 1.0282, "step": 52655 }, { "epoch": 0.13, "learning_rate": 8.700463032865269e-05, "loss": 1.0273, "step": 52660 }, { "epoch": 0.13, "learning_rate": 8.700337208717098e-05, "loss": 1.0282, "step": 52665 }, { "epoch": 0.13, "learning_rate": 8.700211384568927e-05, "loss": 1.0492, "step": 52670 }, { "epoch": 0.13, "learning_rate": 8.700085560420756e-05, "loss": 1.0277, "step": 52675 }, { "epoch": 0.13, "learning_rate": 8.699959736272587e-05, "loss": 1.0273, "step": 52680 }, { "epoch": 0.13, "learning_rate": 8.699833912124416e-05, "loss": 1.0282, "step": 52685 }, { "epoch": 0.13, "learning_rate": 8.699708087976245e-05, "loss": 1.0271, "step": 52690 }, { "epoch": 0.13, "learning_rate": 8.699582263828074e-05, "loss": 1.0271, "step": 52695 }, { "epoch": 0.13, "learning_rate": 8.699456439679903e-05, "loss": 1.0267, "step": 52700 }, { "epoch": 0.13, "learning_rate": 8.699330615531734e-05, "loss": 1.0272, "step": 52705 }, { "epoch": 0.13, "learning_rate": 8.699204791383563e-05, "loss": 1.0267, "step": 52710 }, { "epoch": 0.13, "learning_rate": 8.699078967235392e-05, "loss": 1.0294, "step": 52715 }, { "epoch": 0.13, "learning_rate": 8.698953143087221e-05, "loss": 1.0278, "step": 52720 }, { "epoch": 0.13, "learning_rate": 8.698827318939052e-05, "loss": 1.0279, "step": 52725 }, { "epoch": 0.13, "learning_rate": 8.698701494790881e-05, "loss": 1.0266, "step": 52730 }, { "epoch": 0.13, "learning_rate": 8.69857567064271e-05, "loss": 1.0282, "step": 52735 }, { "epoch": 0.13, "learning_rate": 8.698449846494539e-05, "loss": 1.0271, "step": 52740 }, { "epoch": 0.13, "learning_rate": 8.69832402234637e-05, "loss": 1.0277, "step": 52745 }, { "epoch": 0.13, "learning_rate": 8.698198198198199e-05, "loss": 1.0474, "step": 52750 }, { "epoch": 0.13, "learning_rate": 8.698072374050028e-05, "loss": 1.0257, "step": 52755 }, { "epoch": 0.13, "learning_rate": 8.697946549901857e-05, "loss": 1.0284, "step": 52760 }, { "epoch": 0.13, "learning_rate": 8.697820725753686e-05, "loss": 1.0264, "step": 52765 }, { "epoch": 0.13, "learning_rate": 8.697694901605517e-05, "loss": 1.0264, "step": 52770 }, { "epoch": 0.13, "learning_rate": 8.697569077457346e-05, "loss": 1.0276, "step": 52775 }, { "epoch": 0.13, "learning_rate": 8.697443253309175e-05, "loss": 1.0275, "step": 52780 }, { "epoch": 0.13, "learning_rate": 8.697317429161004e-05, "loss": 1.0286, "step": 52785 }, { "epoch": 0.13, "learning_rate": 8.697191605012835e-05, "loss": 1.0274, "step": 52790 }, { "epoch": 0.13, "learning_rate": 8.697065780864664e-05, "loss": 1.0286, "step": 52795 }, { "epoch": 0.13, "learning_rate": 8.696939956716493e-05, "loss": 1.0259, "step": 52800 }, { "epoch": 0.13, "learning_rate": 8.696814132568322e-05, "loss": 1.0289, "step": 52805 }, { "epoch": 0.13, "learning_rate": 8.696688308420153e-05, "loss": 1.0263, "step": 52810 }, { "epoch": 0.13, "learning_rate": 8.696562484271982e-05, "loss": 1.0265, "step": 52815 }, { "epoch": 0.13, "learning_rate": 8.696436660123811e-05, "loss": 1.0271, "step": 52820 }, { "epoch": 0.13, "learning_rate": 8.69631083597564e-05, "loss": 1.0265, "step": 52825 }, { "epoch": 0.13, "learning_rate": 8.69618501182747e-05, "loss": 1.0259, "step": 52830 }, { "epoch": 0.13, "learning_rate": 8.6960591876793e-05, "loss": 1.0255, "step": 52835 }, { "epoch": 0.13, "learning_rate": 8.695933363531129e-05, "loss": 1.0266, "step": 52840 }, { "epoch": 0.13, "learning_rate": 8.695807539382958e-05, "loss": 1.0273, "step": 52845 }, { "epoch": 0.13, "learning_rate": 8.695681715234787e-05, "loss": 1.0254, "step": 52850 }, { "epoch": 0.13, "learning_rate": 8.695555891086618e-05, "loss": 1.0281, "step": 52855 }, { "epoch": 0.13, "learning_rate": 8.695430066938447e-05, "loss": 1.028, "step": 52860 }, { "epoch": 0.13, "learning_rate": 8.695304242790276e-05, "loss": 1.0268, "step": 52865 }, { "epoch": 0.13, "learning_rate": 8.695178418642105e-05, "loss": 1.029, "step": 52870 }, { "epoch": 0.13, "learning_rate": 8.695052594493936e-05, "loss": 1.0269, "step": 52875 }, { "epoch": 0.13, "learning_rate": 8.694926770345765e-05, "loss": 1.0267, "step": 52880 }, { "epoch": 0.13, "learning_rate": 8.694800946197594e-05, "loss": 1.0265, "step": 52885 }, { "epoch": 0.13, "learning_rate": 8.694675122049423e-05, "loss": 1.0275, "step": 52890 }, { "epoch": 0.13, "learning_rate": 8.694549297901253e-05, "loss": 1.0294, "step": 52895 }, { "epoch": 0.13, "learning_rate": 8.694423473753083e-05, "loss": 1.0288, "step": 52900 }, { "epoch": 0.13, "learning_rate": 8.694297649604912e-05, "loss": 1.0279, "step": 52905 }, { "epoch": 0.13, "learning_rate": 8.694171825456741e-05, "loss": 1.0278, "step": 52910 }, { "epoch": 0.13, "learning_rate": 8.69404600130857e-05, "loss": 1.0303, "step": 52915 }, { "epoch": 0.13, "learning_rate": 8.693920177160401e-05, "loss": 1.025, "step": 52920 }, { "epoch": 0.13, "learning_rate": 8.69379435301223e-05, "loss": 1.0291, "step": 52925 }, { "epoch": 0.13, "learning_rate": 8.69366852886406e-05, "loss": 1.0287, "step": 52930 }, { "epoch": 0.13, "learning_rate": 8.693542704715889e-05, "loss": 1.0277, "step": 52935 }, { "epoch": 0.13, "learning_rate": 8.693416880567719e-05, "loss": 1.0267, "step": 52940 }, { "epoch": 0.13, "learning_rate": 8.693291056419548e-05, "loss": 1.0264, "step": 52945 }, { "epoch": 0.13, "learning_rate": 8.693165232271377e-05, "loss": 1.028, "step": 52950 }, { "epoch": 0.13, "learning_rate": 8.693039408123207e-05, "loss": 1.0457, "step": 52955 }, { "epoch": 0.13, "learning_rate": 8.692913583975036e-05, "loss": 1.0261, "step": 52960 }, { "epoch": 0.13, "learning_rate": 8.692787759826866e-05, "loss": 1.0263, "step": 52965 }, { "epoch": 0.13, "learning_rate": 8.692661935678695e-05, "loss": 1.0292, "step": 52970 }, { "epoch": 0.13, "learning_rate": 8.692536111530525e-05, "loss": 1.029, "step": 52975 }, { "epoch": 0.13, "learning_rate": 8.692410287382354e-05, "loss": 1.026, "step": 52980 }, { "epoch": 0.13, "learning_rate": 8.692284463234184e-05, "loss": 1.0274, "step": 52985 }, { "epoch": 0.13, "learning_rate": 8.692158639086015e-05, "loss": 1.0278, "step": 52990 }, { "epoch": 0.13, "learning_rate": 8.692032814937844e-05, "loss": 1.0254, "step": 52995 }, { "epoch": 0.13, "learning_rate": 8.691906990789673e-05, "loss": 1.0496, "step": 53000 }, { "epoch": 0.13, "learning_rate": 8.691781166641502e-05, "loss": 1.0285, "step": 53005 }, { "epoch": 0.13, "learning_rate": 8.691655342493333e-05, "loss": 1.0277, "step": 53010 }, { "epoch": 0.13, "learning_rate": 8.691529518345162e-05, "loss": 1.0282, "step": 53015 }, { "epoch": 0.13, "learning_rate": 8.691403694196991e-05, "loss": 1.0276, "step": 53020 }, { "epoch": 0.13, "learning_rate": 8.69127787004882e-05, "loss": 1.0271, "step": 53025 }, { "epoch": 0.13, "learning_rate": 8.691152045900651e-05, "loss": 1.029, "step": 53030 }, { "epoch": 0.13, "learning_rate": 8.69102622175248e-05, "loss": 1.0275, "step": 53035 }, { "epoch": 0.13, "learning_rate": 8.690900397604309e-05, "loss": 1.0301, "step": 53040 }, { "epoch": 0.13, "learning_rate": 8.690774573456138e-05, "loss": 1.0289, "step": 53045 }, { "epoch": 0.13, "learning_rate": 8.690648749307967e-05, "loss": 1.0259, "step": 53050 }, { "epoch": 0.13, "learning_rate": 8.690522925159798e-05, "loss": 1.0275, "step": 53055 }, { "epoch": 0.13, "learning_rate": 8.690397101011627e-05, "loss": 1.0257, "step": 53060 }, { "epoch": 0.13, "learning_rate": 8.690271276863456e-05, "loss": 1.0258, "step": 53065 }, { "epoch": 0.13, "learning_rate": 8.690145452715285e-05, "loss": 1.0258, "step": 53070 }, { "epoch": 0.13, "learning_rate": 8.690019628567116e-05, "loss": 1.0245, "step": 53075 }, { "epoch": 0.13, "learning_rate": 8.689893804418945e-05, "loss": 1.0269, "step": 53080 }, { "epoch": 0.13, "learning_rate": 8.689767980270774e-05, "loss": 1.0283, "step": 53085 }, { "epoch": 0.13, "learning_rate": 8.689642156122603e-05, "loss": 1.0274, "step": 53090 }, { "epoch": 0.13, "learning_rate": 8.689516331974434e-05, "loss": 1.0279, "step": 53095 }, { "epoch": 0.13, "learning_rate": 8.689390507826263e-05, "loss": 1.0274, "step": 53100 }, { "epoch": 0.13, "learning_rate": 8.689264683678092e-05, "loss": 1.0271, "step": 53105 }, { "epoch": 0.13, "learning_rate": 8.689138859529921e-05, "loss": 1.0272, "step": 53110 }, { "epoch": 0.13, "learning_rate": 8.68901303538175e-05, "loss": 1.0282, "step": 53115 }, { "epoch": 0.13, "learning_rate": 8.688887211233581e-05, "loss": 1.0298, "step": 53120 }, { "epoch": 0.13, "learning_rate": 8.68876138708541e-05, "loss": 1.0518, "step": 53125 }, { "epoch": 0.13, "learning_rate": 8.68863556293724e-05, "loss": 1.0266, "step": 53130 }, { "epoch": 0.13, "learning_rate": 8.688509738789069e-05, "loss": 1.0252, "step": 53135 }, { "epoch": 0.13, "learning_rate": 8.688383914640899e-05, "loss": 1.0275, "step": 53140 }, { "epoch": 0.13, "learning_rate": 8.688258090492728e-05, "loss": 1.0301, "step": 53145 }, { "epoch": 0.13, "learning_rate": 8.688132266344557e-05, "loss": 1.0264, "step": 53150 }, { "epoch": 0.13, "learning_rate": 8.688006442196387e-05, "loss": 1.0292, "step": 53155 }, { "epoch": 0.13, "learning_rate": 8.687880618048216e-05, "loss": 1.027, "step": 53160 }, { "epoch": 0.13, "learning_rate": 8.687754793900046e-05, "loss": 1.0258, "step": 53165 }, { "epoch": 0.13, "learning_rate": 8.687628969751875e-05, "loss": 1.0258, "step": 53170 }, { "epoch": 0.13, "learning_rate": 8.687503145603705e-05, "loss": 1.0499, "step": 53175 }, { "epoch": 0.13, "learning_rate": 8.687377321455534e-05, "loss": 1.0268, "step": 53180 }, { "epoch": 0.13, "learning_rate": 8.687251497307364e-05, "loss": 1.0251, "step": 53185 }, { "epoch": 0.13, "learning_rate": 8.687125673159193e-05, "loss": 1.0272, "step": 53190 }, { "epoch": 0.13, "learning_rate": 8.686999849011023e-05, "loss": 1.0296, "step": 53195 }, { "epoch": 0.13, "learning_rate": 8.686874024862852e-05, "loss": 1.0246, "step": 53200 }, { "epoch": 0.13, "learning_rate": 8.686748200714682e-05, "loss": 1.0266, "step": 53205 }, { "epoch": 0.13, "learning_rate": 8.686622376566511e-05, "loss": 1.027, "step": 53210 }, { "epoch": 0.13, "learning_rate": 8.68649655241834e-05, "loss": 1.0483, "step": 53215 }, { "epoch": 0.13, "learning_rate": 8.68637072827017e-05, "loss": 1.0286, "step": 53220 }, { "epoch": 0.13, "learning_rate": 8.686244904121999e-05, "loss": 1.027, "step": 53225 }, { "epoch": 0.13, "learning_rate": 8.68611907997383e-05, "loss": 1.0269, "step": 53230 }, { "epoch": 0.13, "learning_rate": 8.685993255825659e-05, "loss": 1.0258, "step": 53235 }, { "epoch": 0.13, "learning_rate": 8.685867431677488e-05, "loss": 1.0291, "step": 53240 }, { "epoch": 0.13, "learning_rate": 8.685741607529317e-05, "loss": 1.0273, "step": 53245 }, { "epoch": 0.13, "learning_rate": 8.685615783381147e-05, "loss": 1.0258, "step": 53250 }, { "epoch": 0.13, "learning_rate": 8.685489959232977e-05, "loss": 1.027, "step": 53255 }, { "epoch": 0.13, "learning_rate": 8.685364135084806e-05, "loss": 1.0268, "step": 53260 }, { "epoch": 0.13, "learning_rate": 8.685238310936635e-05, "loss": 1.0287, "step": 53265 }, { "epoch": 0.13, "learning_rate": 8.685112486788465e-05, "loss": 1.0264, "step": 53270 }, { "epoch": 0.13, "learning_rate": 8.684986662640295e-05, "loss": 1.0269, "step": 53275 }, { "epoch": 0.13, "learning_rate": 8.684860838492124e-05, "loss": 1.0266, "step": 53280 }, { "epoch": 0.13, "learning_rate": 8.684735014343953e-05, "loss": 1.0279, "step": 53285 }, { "epoch": 0.13, "learning_rate": 8.684609190195782e-05, "loss": 1.0255, "step": 53290 }, { "epoch": 0.13, "learning_rate": 8.684483366047613e-05, "loss": 1.0256, "step": 53295 }, { "epoch": 0.13, "learning_rate": 8.684357541899442e-05, "loss": 1.0282, "step": 53300 }, { "epoch": 0.13, "learning_rate": 8.684231717751271e-05, "loss": 1.0274, "step": 53305 }, { "epoch": 0.13, "learning_rate": 8.6841058936031e-05, "loss": 1.0281, "step": 53310 }, { "epoch": 0.13, "learning_rate": 8.68398006945493e-05, "loss": 1.0292, "step": 53315 }, { "epoch": 0.13, "learning_rate": 8.68385424530676e-05, "loss": 1.0268, "step": 53320 }, { "epoch": 0.13, "learning_rate": 8.683728421158589e-05, "loss": 1.0274, "step": 53325 }, { "epoch": 0.13, "learning_rate": 8.683602597010418e-05, "loss": 1.0267, "step": 53330 }, { "epoch": 0.13, "learning_rate": 8.683476772862249e-05, "loss": 1.0266, "step": 53335 }, { "epoch": 0.13, "learning_rate": 8.683350948714078e-05, "loss": 1.026, "step": 53340 }, { "epoch": 0.13, "learning_rate": 8.683225124565907e-05, "loss": 1.0271, "step": 53345 }, { "epoch": 0.13, "learning_rate": 8.683099300417736e-05, "loss": 1.0266, "step": 53350 }, { "epoch": 0.13, "learning_rate": 8.682973476269565e-05, "loss": 1.0261, "step": 53355 }, { "epoch": 0.13, "learning_rate": 8.682847652121396e-05, "loss": 1.0281, "step": 53360 }, { "epoch": 0.13, "learning_rate": 8.682721827973225e-05, "loss": 1.0278, "step": 53365 }, { "epoch": 0.13, "learning_rate": 8.682596003825054e-05, "loss": 1.0266, "step": 53370 }, { "epoch": 0.13, "learning_rate": 8.682470179676883e-05, "loss": 1.0269, "step": 53375 }, { "epoch": 0.13, "learning_rate": 8.682344355528714e-05, "loss": 1.027, "step": 53380 }, { "epoch": 0.13, "learning_rate": 8.682218531380543e-05, "loss": 1.0287, "step": 53385 }, { "epoch": 0.13, "learning_rate": 8.682092707232372e-05, "loss": 1.0282, "step": 53390 }, { "epoch": 0.13, "learning_rate": 8.681966883084201e-05, "loss": 1.0277, "step": 53395 }, { "epoch": 0.13, "learning_rate": 8.681841058936032e-05, "loss": 1.0278, "step": 53400 }, { "epoch": 0.13, "learning_rate": 8.681715234787861e-05, "loss": 1.0287, "step": 53405 }, { "epoch": 0.13, "learning_rate": 8.68158941063969e-05, "loss": 1.0283, "step": 53410 }, { "epoch": 0.13, "learning_rate": 8.681463586491519e-05, "loss": 1.0275, "step": 53415 }, { "epoch": 0.13, "learning_rate": 8.681337762343348e-05, "loss": 1.0287, "step": 53420 }, { "epoch": 0.13, "learning_rate": 8.681211938195179e-05, "loss": 1.0271, "step": 53425 }, { "epoch": 0.13, "learning_rate": 8.681086114047008e-05, "loss": 1.0261, "step": 53430 }, { "epoch": 0.13, "learning_rate": 8.680960289898837e-05, "loss": 1.0255, "step": 53435 }, { "epoch": 0.13, "learning_rate": 8.680834465750666e-05, "loss": 1.0279, "step": 53440 }, { "epoch": 0.13, "learning_rate": 8.680708641602497e-05, "loss": 1.0276, "step": 53445 }, { "epoch": 0.13, "learning_rate": 8.680582817454326e-05, "loss": 1.0279, "step": 53450 }, { "epoch": 0.13, "learning_rate": 8.680456993306155e-05, "loss": 1.0294, "step": 53455 }, { "epoch": 0.13, "learning_rate": 8.680331169157984e-05, "loss": 1.0255, "step": 53460 }, { "epoch": 0.13, "learning_rate": 8.680205345009815e-05, "loss": 1.0281, "step": 53465 }, { "epoch": 0.13, "learning_rate": 8.680079520861644e-05, "loss": 1.0293, "step": 53470 }, { "epoch": 0.13, "learning_rate": 8.679953696713473e-05, "loss": 1.0275, "step": 53475 }, { "epoch": 0.13, "learning_rate": 8.679827872565302e-05, "loss": 1.0273, "step": 53480 }, { "epoch": 0.13, "learning_rate": 8.679702048417131e-05, "loss": 1.0283, "step": 53485 }, { "epoch": 0.13, "learning_rate": 8.679576224268963e-05, "loss": 1.0268, "step": 53490 }, { "epoch": 0.13, "learning_rate": 8.679450400120792e-05, "loss": 1.0253, "step": 53495 }, { "epoch": 0.13, "learning_rate": 8.679324575972622e-05, "loss": 1.0252, "step": 53500 }, { "epoch": 0.13, "learning_rate": 8.679198751824451e-05, "loss": 1.0288, "step": 53505 }, { "epoch": 0.13, "learning_rate": 8.67907292767628e-05, "loss": 1.0246, "step": 53510 }, { "epoch": 0.13, "learning_rate": 8.67894710352811e-05, "loss": 1.0288, "step": 53515 }, { "epoch": 0.13, "learning_rate": 8.67882127937994e-05, "loss": 1.0266, "step": 53520 }, { "epoch": 0.13, "learning_rate": 8.678695455231769e-05, "loss": 1.0263, "step": 53525 }, { "epoch": 0.13, "learning_rate": 8.678569631083598e-05, "loss": 1.0269, "step": 53530 }, { "epoch": 0.13, "learning_rate": 8.678443806935428e-05, "loss": 1.026, "step": 53535 }, { "epoch": 0.13, "learning_rate": 8.678317982787258e-05, "loss": 1.0295, "step": 53540 }, { "epoch": 0.13, "learning_rate": 8.678192158639087e-05, "loss": 1.0273, "step": 53545 }, { "epoch": 0.13, "learning_rate": 8.678066334490916e-05, "loss": 1.0258, "step": 53550 }, { "epoch": 0.13, "learning_rate": 8.677940510342746e-05, "loss": 1.0293, "step": 53555 }, { "epoch": 0.13, "learning_rate": 8.677814686194576e-05, "loss": 1.0278, "step": 53560 }, { "epoch": 0.13, "learning_rate": 8.677688862046405e-05, "loss": 1.0261, "step": 53565 }, { "epoch": 0.13, "learning_rate": 8.677563037898234e-05, "loss": 1.0274, "step": 53570 }, { "epoch": 0.13, "learning_rate": 8.677437213750063e-05, "loss": 1.028, "step": 53575 }, { "epoch": 0.13, "learning_rate": 8.677311389601894e-05, "loss": 1.0275, "step": 53580 }, { "epoch": 0.13, "learning_rate": 8.677185565453723e-05, "loss": 1.0295, "step": 53585 }, { "epoch": 0.13, "learning_rate": 8.677059741305552e-05, "loss": 1.0293, "step": 53590 }, { "epoch": 0.13, "learning_rate": 8.676933917157381e-05, "loss": 1.0284, "step": 53595 }, { "epoch": 0.13, "learning_rate": 8.676808093009212e-05, "loss": 1.0278, "step": 53600 }, { "epoch": 0.13, "learning_rate": 8.676682268861041e-05, "loss": 1.028, "step": 53605 }, { "epoch": 0.13, "learning_rate": 8.67655644471287e-05, "loss": 1.0316, "step": 53610 }, { "epoch": 0.13, "learning_rate": 8.676430620564699e-05, "loss": 1.0282, "step": 53615 }, { "epoch": 0.13, "learning_rate": 8.67630479641653e-05, "loss": 1.0245, "step": 53620 }, { "epoch": 0.13, "learning_rate": 8.676178972268359e-05, "loss": 1.0301, "step": 53625 }, { "epoch": 0.13, "learning_rate": 8.676053148120188e-05, "loss": 1.0275, "step": 53630 }, { "epoch": 0.13, "learning_rate": 8.675927323972017e-05, "loss": 1.0284, "step": 53635 }, { "epoch": 0.13, "learning_rate": 8.675801499823846e-05, "loss": 1.0272, "step": 53640 }, { "epoch": 0.13, "learning_rate": 8.675675675675677e-05, "loss": 1.026, "step": 53645 }, { "epoch": 0.13, "learning_rate": 8.675549851527506e-05, "loss": 1.0276, "step": 53650 }, { "epoch": 0.13, "learning_rate": 8.675424027379335e-05, "loss": 1.0277, "step": 53655 }, { "epoch": 0.13, "learning_rate": 8.675298203231164e-05, "loss": 1.0281, "step": 53660 }, { "epoch": 0.13, "learning_rate": 8.675172379082995e-05, "loss": 1.0273, "step": 53665 }, { "epoch": 0.13, "learning_rate": 8.675046554934824e-05, "loss": 1.0274, "step": 53670 }, { "epoch": 0.13, "learning_rate": 8.674920730786653e-05, "loss": 1.0289, "step": 53675 }, { "epoch": 0.13, "learning_rate": 8.674794906638482e-05, "loss": 1.0254, "step": 53680 }, { "epoch": 0.13, "learning_rate": 8.674669082490313e-05, "loss": 1.0294, "step": 53685 }, { "epoch": 0.13, "learning_rate": 8.674543258342142e-05, "loss": 1.027, "step": 53690 }, { "epoch": 0.13, "learning_rate": 8.674417434193971e-05, "loss": 1.0282, "step": 53695 }, { "epoch": 0.13, "learning_rate": 8.6742916100458e-05, "loss": 1.0264, "step": 53700 }, { "epoch": 0.13, "learning_rate": 8.67416578589763e-05, "loss": 1.027, "step": 53705 }, { "epoch": 0.13, "learning_rate": 8.67403996174946e-05, "loss": 1.0292, "step": 53710 }, { "epoch": 0.13, "learning_rate": 8.673914137601289e-05, "loss": 1.026, "step": 53715 }, { "epoch": 0.13, "learning_rate": 8.673788313453118e-05, "loss": 1.0264, "step": 53720 }, { "epoch": 0.13, "learning_rate": 8.673662489304947e-05, "loss": 1.0274, "step": 53725 }, { "epoch": 0.13, "learning_rate": 8.673536665156778e-05, "loss": 1.0249, "step": 53730 }, { "epoch": 0.13, "learning_rate": 8.673410841008607e-05, "loss": 1.0265, "step": 53735 }, { "epoch": 0.13, "learning_rate": 8.673285016860436e-05, "loss": 1.0279, "step": 53740 }, { "epoch": 0.13, "learning_rate": 8.673159192712265e-05, "loss": 1.0286, "step": 53745 }, { "epoch": 0.13, "learning_rate": 8.673033368564096e-05, "loss": 1.0287, "step": 53750 }, { "epoch": 0.13, "learning_rate": 8.672907544415925e-05, "loss": 1.0274, "step": 53755 }, { "epoch": 0.13, "learning_rate": 8.672781720267754e-05, "loss": 1.0262, "step": 53760 }, { "epoch": 0.13, "learning_rate": 8.672655896119583e-05, "loss": 1.0277, "step": 53765 }, { "epoch": 0.13, "learning_rate": 8.672530071971413e-05, "loss": 1.0268, "step": 53770 }, { "epoch": 0.13, "learning_rate": 8.672404247823243e-05, "loss": 1.0278, "step": 53775 }, { "epoch": 0.13, "learning_rate": 8.672278423675072e-05, "loss": 1.026, "step": 53780 }, { "epoch": 0.14, "learning_rate": 8.672152599526901e-05, "loss": 1.026, "step": 53785 }, { "epoch": 0.14, "learning_rate": 8.67202677537873e-05, "loss": 1.0259, "step": 53790 }, { "epoch": 0.14, "learning_rate": 8.671900951230561e-05, "loss": 1.0274, "step": 53795 }, { "epoch": 0.14, "learning_rate": 8.67177512708239e-05, "loss": 1.0293, "step": 53800 }, { "epoch": 0.14, "learning_rate": 8.67164930293422e-05, "loss": 1.0265, "step": 53805 }, { "epoch": 0.14, "learning_rate": 8.671523478786049e-05, "loss": 1.0262, "step": 53810 }, { "epoch": 0.14, "learning_rate": 8.671397654637879e-05, "loss": 1.0273, "step": 53815 }, { "epoch": 0.14, "learning_rate": 8.671271830489708e-05, "loss": 1.0297, "step": 53820 }, { "epoch": 0.14, "learning_rate": 8.671146006341537e-05, "loss": 1.0278, "step": 53825 }, { "epoch": 0.14, "learning_rate": 8.671020182193366e-05, "loss": 1.0265, "step": 53830 }, { "epoch": 0.14, "learning_rate": 8.670894358045196e-05, "loss": 1.0257, "step": 53835 }, { "epoch": 0.14, "learning_rate": 8.670768533897026e-05, "loss": 1.0271, "step": 53840 }, { "epoch": 0.14, "learning_rate": 8.670642709748855e-05, "loss": 1.028, "step": 53845 }, { "epoch": 0.14, "learning_rate": 8.670516885600684e-05, "loss": 1.0509, "step": 53850 }, { "epoch": 0.14, "learning_rate": 8.670391061452514e-05, "loss": 1.0269, "step": 53855 }, { "epoch": 0.14, "learning_rate": 8.670265237304344e-05, "loss": 1.0263, "step": 53860 }, { "epoch": 0.14, "learning_rate": 8.670139413156173e-05, "loss": 1.028, "step": 53865 }, { "epoch": 0.14, "learning_rate": 8.670013589008002e-05, "loss": 1.0292, "step": 53870 }, { "epoch": 0.14, "learning_rate": 8.669887764859832e-05, "loss": 1.0262, "step": 53875 }, { "epoch": 0.14, "learning_rate": 8.669761940711661e-05, "loss": 1.0279, "step": 53880 }, { "epoch": 0.14, "learning_rate": 8.669636116563491e-05, "loss": 1.0266, "step": 53885 }, { "epoch": 0.14, "learning_rate": 8.66951029241532e-05, "loss": 1.0245, "step": 53890 }, { "epoch": 0.14, "learning_rate": 8.66938446826715e-05, "loss": 1.0272, "step": 53895 }, { "epoch": 0.14, "learning_rate": 8.669258644118979e-05, "loss": 1.0273, "step": 53900 }, { "epoch": 0.14, "learning_rate": 8.669132819970809e-05, "loss": 1.0289, "step": 53905 }, { "epoch": 0.14, "learning_rate": 8.669006995822638e-05, "loss": 1.0268, "step": 53910 }, { "epoch": 0.14, "learning_rate": 8.668881171674468e-05, "loss": 1.0258, "step": 53915 }, { "epoch": 0.14, "learning_rate": 8.668755347526297e-05, "loss": 1.0278, "step": 53920 }, { "epoch": 0.14, "learning_rate": 8.668629523378127e-05, "loss": 1.0288, "step": 53925 }, { "epoch": 0.14, "learning_rate": 8.668503699229956e-05, "loss": 1.0273, "step": 53930 }, { "epoch": 0.14, "learning_rate": 8.668377875081786e-05, "loss": 1.0285, "step": 53935 }, { "epoch": 0.14, "learning_rate": 8.668252050933615e-05, "loss": 1.0273, "step": 53940 }, { "epoch": 0.14, "learning_rate": 8.668126226785444e-05, "loss": 1.0263, "step": 53945 }, { "epoch": 0.14, "learning_rate": 8.668000402637274e-05, "loss": 1.0286, "step": 53950 }, { "epoch": 0.14, "learning_rate": 8.667874578489104e-05, "loss": 1.0264, "step": 53955 }, { "epoch": 0.14, "learning_rate": 8.667748754340933e-05, "loss": 1.0279, "step": 53960 }, { "epoch": 0.14, "learning_rate": 8.667622930192762e-05, "loss": 1.0285, "step": 53965 }, { "epoch": 0.14, "learning_rate": 8.667497106044592e-05, "loss": 1.0286, "step": 53970 }, { "epoch": 0.14, "learning_rate": 8.667371281896422e-05, "loss": 1.0261, "step": 53975 }, { "epoch": 0.14, "learning_rate": 8.667245457748251e-05, "loss": 1.0478, "step": 53980 }, { "epoch": 0.14, "learning_rate": 8.66711963360008e-05, "loss": 1.0253, "step": 53985 }, { "epoch": 0.14, "learning_rate": 8.66699380945191e-05, "loss": 1.0264, "step": 53990 }, { "epoch": 0.14, "learning_rate": 8.666867985303741e-05, "loss": 1.0258, "step": 53995 }, { "epoch": 0.14, "learning_rate": 8.66674216115557e-05, "loss": 1.0275, "step": 54000 }, { "epoch": 0.14, "learning_rate": 8.666616337007399e-05, "loss": 1.0302, "step": 54005 }, { "epoch": 0.14, "learning_rate": 8.666490512859228e-05, "loss": 1.0255, "step": 54010 }, { "epoch": 0.14, "learning_rate": 8.666364688711059e-05, "loss": 1.0257, "step": 54015 }, { "epoch": 0.14, "learning_rate": 8.666238864562888e-05, "loss": 1.0273, "step": 54020 }, { "epoch": 0.14, "learning_rate": 8.666113040414717e-05, "loss": 1.0279, "step": 54025 }, { "epoch": 0.14, "learning_rate": 8.665987216266546e-05, "loss": 1.027, "step": 54030 }, { "epoch": 0.14, "learning_rate": 8.665861392118376e-05, "loss": 1.0515, "step": 54035 }, { "epoch": 0.14, "learning_rate": 8.665735567970206e-05, "loss": 1.0269, "step": 54040 }, { "epoch": 0.14, "learning_rate": 8.665609743822035e-05, "loss": 1.0507, "step": 54045 }, { "epoch": 0.14, "learning_rate": 8.665483919673864e-05, "loss": 1.0279, "step": 54050 }, { "epoch": 0.14, "learning_rate": 8.665358095525694e-05, "loss": 1.0253, "step": 54055 }, { "epoch": 0.14, "learning_rate": 8.665232271377524e-05, "loss": 1.0285, "step": 54060 }, { "epoch": 0.14, "learning_rate": 8.665106447229353e-05, "loss": 1.0258, "step": 54065 }, { "epoch": 0.14, "learning_rate": 8.664980623081182e-05, "loss": 1.0235, "step": 54070 }, { "epoch": 0.14, "learning_rate": 8.664854798933012e-05, "loss": 1.0277, "step": 54075 }, { "epoch": 0.14, "learning_rate": 8.664728974784842e-05, "loss": 1.0495, "step": 54080 }, { "epoch": 0.14, "learning_rate": 8.664603150636671e-05, "loss": 1.0271, "step": 54085 }, { "epoch": 0.14, "learning_rate": 8.6644773264885e-05, "loss": 1.0295, "step": 54090 }, { "epoch": 0.14, "learning_rate": 8.66435150234033e-05, "loss": 1.0497, "step": 54095 }, { "epoch": 0.14, "learning_rate": 8.664225678192159e-05, "loss": 1.0286, "step": 54100 }, { "epoch": 0.14, "learning_rate": 8.664099854043989e-05, "loss": 1.0299, "step": 54105 }, { "epoch": 0.14, "learning_rate": 8.663974029895818e-05, "loss": 1.0288, "step": 54110 }, { "epoch": 0.14, "learning_rate": 8.663848205747648e-05, "loss": 1.0434, "step": 54115 }, { "epoch": 0.14, "learning_rate": 8.663722381599477e-05, "loss": 1.0259, "step": 54120 }, { "epoch": 0.14, "learning_rate": 8.663596557451307e-05, "loss": 1.0259, "step": 54125 }, { "epoch": 0.14, "learning_rate": 8.663470733303136e-05, "loss": 1.0302, "step": 54130 }, { "epoch": 0.14, "learning_rate": 8.663344909154966e-05, "loss": 1.0292, "step": 54135 }, { "epoch": 0.14, "learning_rate": 8.663219085006795e-05, "loss": 1.0264, "step": 54140 }, { "epoch": 0.14, "learning_rate": 8.663093260858625e-05, "loss": 1.0291, "step": 54145 }, { "epoch": 0.14, "learning_rate": 8.662967436710454e-05, "loss": 1.0297, "step": 54150 }, { "epoch": 0.14, "learning_rate": 8.662841612562284e-05, "loss": 1.0265, "step": 54155 }, { "epoch": 0.14, "learning_rate": 8.662715788414113e-05, "loss": 1.0271, "step": 54160 }, { "epoch": 0.14, "learning_rate": 8.662589964265942e-05, "loss": 1.0265, "step": 54165 }, { "epoch": 0.14, "learning_rate": 8.662464140117772e-05, "loss": 1.0262, "step": 54170 }, { "epoch": 0.14, "learning_rate": 8.662338315969602e-05, "loss": 1.0262, "step": 54175 }, { "epoch": 0.14, "learning_rate": 8.662212491821431e-05, "loss": 1.0268, "step": 54180 }, { "epoch": 0.14, "learning_rate": 8.66208666767326e-05, "loss": 1.0288, "step": 54185 }, { "epoch": 0.14, "learning_rate": 8.66196084352509e-05, "loss": 1.0282, "step": 54190 }, { "epoch": 0.14, "learning_rate": 8.66183501937692e-05, "loss": 1.0281, "step": 54195 }, { "epoch": 0.14, "learning_rate": 8.661709195228749e-05, "loss": 1.028, "step": 54200 }, { "epoch": 0.14, "learning_rate": 8.661583371080578e-05, "loss": 1.0264, "step": 54205 }, { "epoch": 0.14, "learning_rate": 8.661457546932408e-05, "loss": 1.028, "step": 54210 }, { "epoch": 0.14, "learning_rate": 8.661331722784238e-05, "loss": 1.0283, "step": 54215 }, { "epoch": 0.14, "learning_rate": 8.661205898636067e-05, "loss": 1.0281, "step": 54220 }, { "epoch": 0.14, "learning_rate": 8.661080074487896e-05, "loss": 1.0272, "step": 54225 }, { "epoch": 0.14, "learning_rate": 8.660954250339725e-05, "loss": 1.0285, "step": 54230 }, { "epoch": 0.14, "learning_rate": 8.660828426191556e-05, "loss": 1.0258, "step": 54235 }, { "epoch": 0.14, "learning_rate": 8.660702602043385e-05, "loss": 1.0289, "step": 54240 }, { "epoch": 0.14, "learning_rate": 8.660576777895214e-05, "loss": 1.0289, "step": 54245 }, { "epoch": 0.14, "learning_rate": 8.660450953747043e-05, "loss": 1.0272, "step": 54250 }, { "epoch": 0.14, "learning_rate": 8.660325129598874e-05, "loss": 1.0275, "step": 54255 }, { "epoch": 0.14, "learning_rate": 8.660199305450703e-05, "loss": 1.0275, "step": 54260 }, { "epoch": 0.14, "learning_rate": 8.660073481302532e-05, "loss": 1.0257, "step": 54265 }, { "epoch": 0.14, "learning_rate": 8.659947657154361e-05, "loss": 1.0245, "step": 54270 }, { "epoch": 0.14, "learning_rate": 8.659821833006192e-05, "loss": 1.0288, "step": 54275 }, { "epoch": 0.14, "learning_rate": 8.659696008858021e-05, "loss": 1.0272, "step": 54280 }, { "epoch": 0.14, "learning_rate": 8.65957018470985e-05, "loss": 1.0271, "step": 54285 }, { "epoch": 0.14, "learning_rate": 8.659444360561679e-05, "loss": 1.0279, "step": 54290 }, { "epoch": 0.14, "learning_rate": 8.659318536413508e-05, "loss": 1.0253, "step": 54295 }, { "epoch": 0.14, "learning_rate": 8.659192712265339e-05, "loss": 1.0271, "step": 54300 }, { "epoch": 0.14, "learning_rate": 8.659066888117168e-05, "loss": 1.0288, "step": 54305 }, { "epoch": 0.14, "learning_rate": 8.658941063968997e-05, "loss": 1.0286, "step": 54310 }, { "epoch": 0.14, "learning_rate": 8.658815239820826e-05, "loss": 1.0263, "step": 54315 }, { "epoch": 0.14, "learning_rate": 8.658689415672657e-05, "loss": 1.0276, "step": 54320 }, { "epoch": 0.14, "learning_rate": 8.658563591524486e-05, "loss": 1.0276, "step": 54325 }, { "epoch": 0.14, "learning_rate": 8.658437767376315e-05, "loss": 1.0291, "step": 54330 }, { "epoch": 0.14, "learning_rate": 8.658311943228144e-05, "loss": 1.0265, "step": 54335 }, { "epoch": 0.14, "learning_rate": 8.658186119079975e-05, "loss": 1.0284, "step": 54340 }, { "epoch": 0.14, "learning_rate": 8.658060294931804e-05, "loss": 1.0273, "step": 54345 }, { "epoch": 0.14, "learning_rate": 8.657934470783633e-05, "loss": 1.028, "step": 54350 }, { "epoch": 0.14, "learning_rate": 8.657808646635462e-05, "loss": 1.026, "step": 54355 }, { "epoch": 0.14, "learning_rate": 8.657682822487291e-05, "loss": 1.0284, "step": 54360 }, { "epoch": 0.14, "learning_rate": 8.657556998339122e-05, "loss": 1.0279, "step": 54365 }, { "epoch": 0.14, "learning_rate": 8.657431174190951e-05, "loss": 1.0248, "step": 54370 }, { "epoch": 0.14, "learning_rate": 8.65730535004278e-05, "loss": 1.0275, "step": 54375 }, { "epoch": 0.14, "learning_rate": 8.657179525894609e-05, "loss": 1.0288, "step": 54380 }, { "epoch": 0.14, "learning_rate": 8.65705370174644e-05, "loss": 1.0264, "step": 54385 }, { "epoch": 0.14, "learning_rate": 8.656927877598269e-05, "loss": 1.0258, "step": 54390 }, { "epoch": 0.14, "learning_rate": 8.656802053450098e-05, "loss": 1.0284, "step": 54395 }, { "epoch": 0.14, "learning_rate": 8.656676229301927e-05, "loss": 1.0274, "step": 54400 }, { "epoch": 0.14, "learning_rate": 8.656550405153758e-05, "loss": 1.026, "step": 54405 }, { "epoch": 0.14, "learning_rate": 8.656424581005587e-05, "loss": 1.0242, "step": 54410 }, { "epoch": 0.14, "learning_rate": 8.656298756857416e-05, "loss": 1.0245, "step": 54415 }, { "epoch": 0.14, "learning_rate": 8.656172932709245e-05, "loss": 1.0291, "step": 54420 }, { "epoch": 0.14, "learning_rate": 8.656047108561074e-05, "loss": 1.0279, "step": 54425 }, { "epoch": 0.14, "learning_rate": 8.655921284412905e-05, "loss": 1.0286, "step": 54430 }, { "epoch": 0.14, "learning_rate": 8.655795460264734e-05, "loss": 1.0293, "step": 54435 }, { "epoch": 0.14, "learning_rate": 8.655669636116563e-05, "loss": 1.0279, "step": 54440 }, { "epoch": 0.14, "learning_rate": 8.655543811968392e-05, "loss": 1.0293, "step": 54445 }, { "epoch": 0.14, "learning_rate": 8.655417987820223e-05, "loss": 1.0273, "step": 54450 }, { "epoch": 0.14, "learning_rate": 8.655292163672052e-05, "loss": 1.0271, "step": 54455 }, { "epoch": 0.14, "learning_rate": 8.655166339523881e-05, "loss": 1.0282, "step": 54460 }, { "epoch": 0.14, "learning_rate": 8.65504051537571e-05, "loss": 1.0241, "step": 54465 }, { "epoch": 0.14, "learning_rate": 8.654914691227541e-05, "loss": 1.0264, "step": 54470 }, { "epoch": 0.14, "learning_rate": 8.65478886707937e-05, "loss": 1.0257, "step": 54475 }, { "epoch": 0.14, "learning_rate": 8.654663042931199e-05, "loss": 1.0276, "step": 54480 }, { "epoch": 0.14, "learning_rate": 8.654537218783028e-05, "loss": 1.0277, "step": 54485 }, { "epoch": 0.14, "learning_rate": 8.654411394634859e-05, "loss": 1.0269, "step": 54490 }, { "epoch": 0.14, "learning_rate": 8.654285570486688e-05, "loss": 1.028, "step": 54495 }, { "epoch": 0.14, "learning_rate": 8.654159746338519e-05, "loss": 1.0261, "step": 54500 }, { "epoch": 0.14, "learning_rate": 8.654033922190348e-05, "loss": 1.0259, "step": 54505 }, { "epoch": 0.14, "learning_rate": 8.653908098042177e-05, "loss": 1.0257, "step": 54510 }, { "epoch": 0.14, "learning_rate": 8.653782273894006e-05, "loss": 1.0259, "step": 54515 }, { "epoch": 0.14, "learning_rate": 8.653656449745837e-05, "loss": 1.0294, "step": 54520 }, { "epoch": 0.14, "learning_rate": 8.653530625597666e-05, "loss": 1.026, "step": 54525 }, { "epoch": 0.14, "learning_rate": 8.653404801449495e-05, "loss": 1.0266, "step": 54530 }, { "epoch": 0.14, "learning_rate": 8.653278977301324e-05, "loss": 1.0288, "step": 54535 }, { "epoch": 0.14, "learning_rate": 8.653153153153155e-05, "loss": 1.0274, "step": 54540 }, { "epoch": 0.14, "learning_rate": 8.653027329004984e-05, "loss": 1.0247, "step": 54545 }, { "epoch": 0.14, "learning_rate": 8.652901504856813e-05, "loss": 1.0243, "step": 54550 }, { "epoch": 0.14, "learning_rate": 8.652775680708642e-05, "loss": 1.0266, "step": 54555 }, { "epoch": 0.14, "learning_rate": 8.652649856560471e-05, "loss": 1.0291, "step": 54560 }, { "epoch": 0.14, "learning_rate": 8.652524032412302e-05, "loss": 1.027, "step": 54565 }, { "epoch": 0.14, "learning_rate": 8.652398208264131e-05, "loss": 1.0283, "step": 54570 }, { "epoch": 0.14, "learning_rate": 8.65227238411596e-05, "loss": 1.0296, "step": 54575 }, { "epoch": 0.14, "learning_rate": 8.652146559967789e-05, "loss": 1.0263, "step": 54580 }, { "epoch": 0.14, "learning_rate": 8.65202073581962e-05, "loss": 1.0259, "step": 54585 }, { "epoch": 0.14, "learning_rate": 8.651894911671449e-05, "loss": 1.0288, "step": 54590 }, { "epoch": 0.14, "learning_rate": 8.651769087523278e-05, "loss": 1.0261, "step": 54595 }, { "epoch": 0.14, "learning_rate": 8.651643263375107e-05, "loss": 1.0246, "step": 54600 }, { "epoch": 0.14, "learning_rate": 8.651517439226938e-05, "loss": 1.0276, "step": 54605 }, { "epoch": 0.14, "learning_rate": 8.651391615078767e-05, "loss": 1.0287, "step": 54610 }, { "epoch": 0.14, "learning_rate": 8.651265790930596e-05, "loss": 1.028, "step": 54615 }, { "epoch": 0.14, "learning_rate": 8.651139966782425e-05, "loss": 1.0262, "step": 54620 }, { "epoch": 0.14, "learning_rate": 8.651014142634254e-05, "loss": 1.0247, "step": 54625 }, { "epoch": 0.14, "learning_rate": 8.650888318486085e-05, "loss": 1.0264, "step": 54630 }, { "epoch": 0.14, "learning_rate": 8.650762494337914e-05, "loss": 1.0288, "step": 54635 }, { "epoch": 0.14, "learning_rate": 8.650636670189743e-05, "loss": 1.0238, "step": 54640 }, { "epoch": 0.14, "learning_rate": 8.650510846041572e-05, "loss": 1.0286, "step": 54645 }, { "epoch": 0.14, "learning_rate": 8.650385021893403e-05, "loss": 1.0271, "step": 54650 }, { "epoch": 0.14, "learning_rate": 8.650259197745232e-05, "loss": 1.0281, "step": 54655 }, { "epoch": 0.14, "learning_rate": 8.650133373597061e-05, "loss": 1.0278, "step": 54660 }, { "epoch": 0.14, "learning_rate": 8.65000754944889e-05, "loss": 1.0267, "step": 54665 }, { "epoch": 0.14, "learning_rate": 8.649881725300721e-05, "loss": 1.0286, "step": 54670 }, { "epoch": 0.14, "learning_rate": 8.64975590115255e-05, "loss": 1.0255, "step": 54675 }, { "epoch": 0.14, "learning_rate": 8.649630077004379e-05, "loss": 1.0266, "step": 54680 }, { "epoch": 0.14, "learning_rate": 8.649504252856208e-05, "loss": 1.0283, "step": 54685 }, { "epoch": 0.14, "learning_rate": 8.649378428708038e-05, "loss": 1.0274, "step": 54690 }, { "epoch": 0.14, "learning_rate": 8.649252604559868e-05, "loss": 1.0263, "step": 54695 }, { "epoch": 0.14, "learning_rate": 8.649126780411697e-05, "loss": 1.0297, "step": 54700 }, { "epoch": 0.14, "learning_rate": 8.649000956263526e-05, "loss": 1.0271, "step": 54705 }, { "epoch": 0.14, "learning_rate": 8.648875132115356e-05, "loss": 1.0265, "step": 54710 }, { "epoch": 0.14, "learning_rate": 8.648749307967186e-05, "loss": 1.0279, "step": 54715 }, { "epoch": 0.14, "learning_rate": 8.648623483819015e-05, "loss": 1.026, "step": 54720 }, { "epoch": 0.14, "learning_rate": 8.648497659670844e-05, "loss": 1.0281, "step": 54725 }, { "epoch": 0.14, "learning_rate": 8.648371835522674e-05, "loss": 1.027, "step": 54730 }, { "epoch": 0.14, "learning_rate": 8.648246011374504e-05, "loss": 1.0273, "step": 54735 }, { "epoch": 0.14, "learning_rate": 8.648120187226333e-05, "loss": 1.03, "step": 54740 }, { "epoch": 0.14, "learning_rate": 8.647994363078162e-05, "loss": 1.0267, "step": 54745 }, { "epoch": 0.14, "learning_rate": 8.647868538929992e-05, "loss": 1.0276, "step": 54750 }, { "epoch": 0.14, "learning_rate": 8.647742714781821e-05, "loss": 1.0286, "step": 54755 }, { "epoch": 0.14, "learning_rate": 8.647616890633651e-05, "loss": 1.0269, "step": 54760 }, { "epoch": 0.14, "learning_rate": 8.64749106648548e-05, "loss": 1.0276, "step": 54765 }, { "epoch": 0.14, "learning_rate": 8.64736524233731e-05, "loss": 1.027, "step": 54770 }, { "epoch": 0.14, "learning_rate": 8.647239418189139e-05, "loss": 1.027, "step": 54775 }, { "epoch": 0.14, "learning_rate": 8.647113594040969e-05, "loss": 1.0253, "step": 54780 }, { "epoch": 0.14, "learning_rate": 8.646987769892798e-05, "loss": 1.0257, "step": 54785 }, { "epoch": 0.14, "learning_rate": 8.646861945744628e-05, "loss": 1.0273, "step": 54790 }, { "epoch": 0.14, "learning_rate": 8.646736121596457e-05, "loss": 1.0265, "step": 54795 }, { "epoch": 0.14, "learning_rate": 8.646610297448287e-05, "loss": 1.0281, "step": 54800 }, { "epoch": 0.14, "learning_rate": 8.646484473300116e-05, "loss": 1.0281, "step": 54805 }, { "epoch": 0.14, "learning_rate": 8.646358649151946e-05, "loss": 1.0271, "step": 54810 }, { "epoch": 0.14, "learning_rate": 8.646232825003775e-05, "loss": 1.0321, "step": 54815 }, { "epoch": 0.14, "learning_rate": 8.646107000855604e-05, "loss": 1.026, "step": 54820 }, { "epoch": 0.14, "learning_rate": 8.645981176707434e-05, "loss": 1.0277, "step": 54825 }, { "epoch": 0.14, "learning_rate": 8.645855352559264e-05, "loss": 1.0268, "step": 54830 }, { "epoch": 0.14, "learning_rate": 8.645729528411093e-05, "loss": 1.0254, "step": 54835 }, { "epoch": 0.14, "learning_rate": 8.645603704262922e-05, "loss": 1.0265, "step": 54840 }, { "epoch": 0.14, "learning_rate": 8.645477880114752e-05, "loss": 1.0265, "step": 54845 }, { "epoch": 0.14, "learning_rate": 8.645352055966581e-05, "loss": 1.0459, "step": 54850 }, { "epoch": 0.14, "learning_rate": 8.64522623181841e-05, "loss": 1.0281, "step": 54855 }, { "epoch": 0.14, "learning_rate": 8.64510040767024e-05, "loss": 1.0298, "step": 54860 }, { "epoch": 0.14, "learning_rate": 8.64497458352207e-05, "loss": 1.0259, "step": 54865 }, { "epoch": 0.14, "learning_rate": 8.6448487593739e-05, "loss": 1.0263, "step": 54870 }, { "epoch": 0.14, "learning_rate": 8.644722935225729e-05, "loss": 1.027, "step": 54875 }, { "epoch": 0.14, "learning_rate": 8.644597111077558e-05, "loss": 1.0279, "step": 54880 }, { "epoch": 0.14, "learning_rate": 8.644471286929387e-05, "loss": 1.048, "step": 54885 }, { "epoch": 0.14, "learning_rate": 8.644345462781217e-05, "loss": 1.0277, "step": 54890 }, { "epoch": 0.14, "learning_rate": 8.644219638633047e-05, "loss": 1.0277, "step": 54895 }, { "epoch": 0.14, "learning_rate": 8.644093814484876e-05, "loss": 1.0284, "step": 54900 }, { "epoch": 0.14, "learning_rate": 8.643967990336705e-05, "loss": 1.0296, "step": 54905 }, { "epoch": 0.14, "learning_rate": 8.643842166188535e-05, "loss": 1.0273, "step": 54910 }, { "epoch": 0.14, "learning_rate": 8.643716342040365e-05, "loss": 1.027, "step": 54915 }, { "epoch": 0.14, "learning_rate": 8.643615682721828e-05, "loss": 1.0266, "step": 54920 }, { "epoch": 0.14, "learning_rate": 8.643489858573657e-05, "loss": 1.0284, "step": 54925 }, { "epoch": 0.14, "learning_rate": 8.643364034425488e-05, "loss": 1.0252, "step": 54930 }, { "epoch": 0.14, "learning_rate": 8.643238210277317e-05, "loss": 1.0281, "step": 54935 }, { "epoch": 0.14, "learning_rate": 8.643112386129146e-05, "loss": 1.0258, "step": 54940 }, { "epoch": 0.14, "learning_rate": 8.642986561980975e-05, "loss": 1.0299, "step": 54945 }, { "epoch": 0.14, "learning_rate": 8.642860737832806e-05, "loss": 1.0304, "step": 54950 }, { "epoch": 0.14, "learning_rate": 8.642734913684635e-05, "loss": 1.0293, "step": 54955 }, { "epoch": 0.14, "learning_rate": 8.642609089536464e-05, "loss": 1.0266, "step": 54960 }, { "epoch": 0.14, "learning_rate": 8.642483265388293e-05, "loss": 1.0258, "step": 54965 }, { "epoch": 0.14, "learning_rate": 8.642357441240123e-05, "loss": 1.0261, "step": 54970 }, { "epoch": 0.14, "learning_rate": 8.642231617091953e-05, "loss": 1.0282, "step": 54975 }, { "epoch": 0.14, "learning_rate": 8.642105792943782e-05, "loss": 1.0259, "step": 54980 }, { "epoch": 0.14, "learning_rate": 8.641979968795611e-05, "loss": 1.0283, "step": 54985 }, { "epoch": 0.14, "learning_rate": 8.64185414464744e-05, "loss": 1.0267, "step": 54990 }, { "epoch": 0.14, "learning_rate": 8.641728320499271e-05, "loss": 1.0287, "step": 54995 }, { "epoch": 0.14, "learning_rate": 8.6416024963511e-05, "loss": 1.0271, "step": 55000 }, { "epoch": 0.14, "learning_rate": 8.64147667220293e-05, "loss": 1.0274, "step": 55005 }, { "epoch": 0.14, "learning_rate": 8.641350848054759e-05, "loss": 1.0282, "step": 55010 }, { "epoch": 0.14, "learning_rate": 8.641225023906589e-05, "loss": 1.0268, "step": 55015 }, { "epoch": 0.14, "learning_rate": 8.641099199758418e-05, "loss": 1.0305, "step": 55020 }, { "epoch": 0.14, "learning_rate": 8.640973375610247e-05, "loss": 1.0266, "step": 55025 }, { "epoch": 0.14, "learning_rate": 8.640847551462077e-05, "loss": 1.0274, "step": 55030 }, { "epoch": 0.14, "learning_rate": 8.640721727313906e-05, "loss": 1.0266, "step": 55035 }, { "epoch": 0.14, "learning_rate": 8.640595903165736e-05, "loss": 1.0282, "step": 55040 }, { "epoch": 0.14, "learning_rate": 8.640470079017565e-05, "loss": 1.0273, "step": 55045 }, { "epoch": 0.14, "learning_rate": 8.640344254869395e-05, "loss": 1.0293, "step": 55050 }, { "epoch": 0.14, "learning_rate": 8.640218430721224e-05, "loss": 1.0276, "step": 55055 }, { "epoch": 0.14, "learning_rate": 8.640092606573054e-05, "loss": 1.03, "step": 55060 }, { "epoch": 0.14, "learning_rate": 8.639966782424883e-05, "loss": 1.0278, "step": 55065 }, { "epoch": 0.14, "learning_rate": 8.639840958276713e-05, "loss": 1.0273, "step": 55070 }, { "epoch": 0.14, "learning_rate": 8.639715134128542e-05, "loss": 1.0263, "step": 55075 }, { "epoch": 0.14, "learning_rate": 8.639589309980372e-05, "loss": 1.0281, "step": 55080 }, { "epoch": 0.14, "learning_rate": 8.639463485832201e-05, "loss": 1.0287, "step": 55085 }, { "epoch": 0.14, "learning_rate": 8.63933766168403e-05, "loss": 1.0278, "step": 55090 }, { "epoch": 0.14, "learning_rate": 8.63921183753586e-05, "loss": 1.0255, "step": 55095 }, { "epoch": 0.14, "learning_rate": 8.639086013387689e-05, "loss": 1.0284, "step": 55100 }, { "epoch": 0.14, "learning_rate": 8.63896018923952e-05, "loss": 1.0272, "step": 55105 }, { "epoch": 0.14, "learning_rate": 8.638834365091349e-05, "loss": 1.0281, "step": 55110 }, { "epoch": 0.14, "learning_rate": 8.638708540943178e-05, "loss": 1.0274, "step": 55115 }, { "epoch": 0.14, "learning_rate": 8.638582716795007e-05, "loss": 1.0274, "step": 55120 }, { "epoch": 0.14, "learning_rate": 8.638456892646837e-05, "loss": 1.0249, "step": 55125 }, { "epoch": 0.14, "learning_rate": 8.638331068498666e-05, "loss": 1.0286, "step": 55130 }, { "epoch": 0.14, "learning_rate": 8.638205244350496e-05, "loss": 1.0274, "step": 55135 }, { "epoch": 0.14, "learning_rate": 8.638079420202325e-05, "loss": 1.0263, "step": 55140 }, { "epoch": 0.14, "learning_rate": 8.637953596054155e-05, "loss": 1.0263, "step": 55145 }, { "epoch": 0.14, "learning_rate": 8.637827771905984e-05, "loss": 1.0278, "step": 55150 }, { "epoch": 0.14, "learning_rate": 8.637701947757814e-05, "loss": 1.0294, "step": 55155 }, { "epoch": 0.14, "learning_rate": 8.637576123609643e-05, "loss": 1.0251, "step": 55160 }, { "epoch": 0.14, "learning_rate": 8.637450299461472e-05, "loss": 1.0253, "step": 55165 }, { "epoch": 0.14, "learning_rate": 8.637324475313302e-05, "loss": 1.0274, "step": 55170 }, { "epoch": 0.14, "learning_rate": 8.637198651165132e-05, "loss": 1.0267, "step": 55175 }, { "epoch": 0.14, "learning_rate": 8.637072827016961e-05, "loss": 1.0287, "step": 55180 }, { "epoch": 0.14, "learning_rate": 8.63694700286879e-05, "loss": 1.0269, "step": 55185 }, { "epoch": 0.14, "learning_rate": 8.63682117872062e-05, "loss": 1.0493, "step": 55190 }, { "epoch": 0.14, "learning_rate": 8.636695354572451e-05, "loss": 1.0244, "step": 55195 }, { "epoch": 0.14, "learning_rate": 8.63656953042428e-05, "loss": 1.0254, "step": 55200 }, { "epoch": 0.14, "learning_rate": 8.636443706276109e-05, "loss": 1.0276, "step": 55205 }, { "epoch": 0.14, "learning_rate": 8.636317882127938e-05, "loss": 1.0266, "step": 55210 }, { "epoch": 0.14, "learning_rate": 8.636192057979769e-05, "loss": 1.025, "step": 55215 }, { "epoch": 0.14, "learning_rate": 8.636066233831598e-05, "loss": 1.0289, "step": 55220 }, { "epoch": 0.14, "learning_rate": 8.635940409683427e-05, "loss": 1.0267, "step": 55225 }, { "epoch": 0.14, "learning_rate": 8.635814585535256e-05, "loss": 1.0269, "step": 55230 }, { "epoch": 0.14, "learning_rate": 8.635688761387086e-05, "loss": 1.0295, "step": 55235 }, { "epoch": 0.14, "learning_rate": 8.635562937238916e-05, "loss": 1.0276, "step": 55240 }, { "epoch": 0.14, "learning_rate": 8.635437113090745e-05, "loss": 1.0266, "step": 55245 }, { "epoch": 0.14, "learning_rate": 8.635311288942574e-05, "loss": 1.0257, "step": 55250 }, { "epoch": 0.14, "learning_rate": 8.635185464794404e-05, "loss": 1.03, "step": 55255 }, { "epoch": 0.14, "learning_rate": 8.635059640646234e-05, "loss": 1.0282, "step": 55260 }, { "epoch": 0.14, "learning_rate": 8.634933816498063e-05, "loss": 1.0275, "step": 55265 }, { "epoch": 0.14, "learning_rate": 8.634807992349892e-05, "loss": 1.0281, "step": 55270 }, { "epoch": 0.14, "learning_rate": 8.634682168201722e-05, "loss": 1.0278, "step": 55275 }, { "epoch": 0.14, "learning_rate": 8.634556344053552e-05, "loss": 1.027, "step": 55280 }, { "epoch": 0.14, "learning_rate": 8.634430519905381e-05, "loss": 1.0283, "step": 55285 }, { "epoch": 0.14, "learning_rate": 8.63430469575721e-05, "loss": 1.0276, "step": 55290 }, { "epoch": 0.14, "learning_rate": 8.63417887160904e-05, "loss": 1.0277, "step": 55295 }, { "epoch": 0.14, "learning_rate": 8.634053047460869e-05, "loss": 1.0278, "step": 55300 }, { "epoch": 0.14, "learning_rate": 8.633927223312699e-05, "loss": 1.0313, "step": 55305 }, { "epoch": 0.14, "learning_rate": 8.633801399164528e-05, "loss": 1.0282, "step": 55310 }, { "epoch": 0.14, "learning_rate": 8.633675575016358e-05, "loss": 1.0258, "step": 55315 }, { "epoch": 0.14, "learning_rate": 8.633549750868187e-05, "loss": 1.0334, "step": 55320 }, { "epoch": 0.14, "learning_rate": 8.633423926720017e-05, "loss": 1.0272, "step": 55325 }, { "epoch": 0.14, "learning_rate": 8.633298102571846e-05, "loss": 1.0238, "step": 55330 }, { "epoch": 0.14, "learning_rate": 8.633172278423676e-05, "loss": 1.029, "step": 55335 }, { "epoch": 0.14, "learning_rate": 8.633046454275505e-05, "loss": 1.029, "step": 55340 }, { "epoch": 0.14, "learning_rate": 8.632920630127335e-05, "loss": 1.0288, "step": 55345 }, { "epoch": 0.14, "learning_rate": 8.632794805979164e-05, "loss": 1.0258, "step": 55350 }, { "epoch": 0.14, "learning_rate": 8.632668981830994e-05, "loss": 1.028, "step": 55355 }, { "epoch": 0.14, "learning_rate": 8.632543157682823e-05, "loss": 1.0267, "step": 55360 }, { "epoch": 0.14, "learning_rate": 8.632417333534652e-05, "loss": 1.0261, "step": 55365 }, { "epoch": 0.14, "learning_rate": 8.632291509386482e-05, "loss": 1.0268, "step": 55370 }, { "epoch": 0.14, "learning_rate": 8.632165685238312e-05, "loss": 1.0282, "step": 55375 }, { "epoch": 0.14, "learning_rate": 8.632039861090141e-05, "loss": 1.0278, "step": 55380 }, { "epoch": 0.14, "learning_rate": 8.63191403694197e-05, "loss": 1.0267, "step": 55385 }, { "epoch": 0.14, "learning_rate": 8.6317882127938e-05, "loss": 1.0274, "step": 55390 }, { "epoch": 0.14, "learning_rate": 8.63166238864563e-05, "loss": 1.0527, "step": 55395 }, { "epoch": 0.14, "learning_rate": 8.631536564497459e-05, "loss": 1.0267, "step": 55400 }, { "epoch": 0.14, "learning_rate": 8.631410740349288e-05, "loss": 1.0265, "step": 55405 }, { "epoch": 0.14, "learning_rate": 8.631284916201118e-05, "loss": 1.0266, "step": 55410 }, { "epoch": 0.14, "learning_rate": 8.631159092052948e-05, "loss": 1.0271, "step": 55415 }, { "epoch": 0.14, "learning_rate": 8.631033267904777e-05, "loss": 1.0291, "step": 55420 }, { "epoch": 0.14, "learning_rate": 8.630907443756606e-05, "loss": 1.0296, "step": 55425 }, { "epoch": 0.14, "learning_rate": 8.630781619608435e-05, "loss": 1.0291, "step": 55430 }, { "epoch": 0.14, "learning_rate": 8.630655795460266e-05, "loss": 1.0251, "step": 55435 }, { "epoch": 0.14, "learning_rate": 8.630529971312095e-05, "loss": 1.026, "step": 55440 }, { "epoch": 0.14, "learning_rate": 8.630404147163924e-05, "loss": 1.0282, "step": 55445 }, { "epoch": 0.14, "learning_rate": 8.630278323015753e-05, "loss": 1.0252, "step": 55450 }, { "epoch": 0.14, "learning_rate": 8.630152498867584e-05, "loss": 1.027, "step": 55455 }, { "epoch": 0.14, "learning_rate": 8.630026674719413e-05, "loss": 1.0268, "step": 55460 }, { "epoch": 0.14, "learning_rate": 8.629900850571242e-05, "loss": 1.0281, "step": 55465 }, { "epoch": 0.14, "learning_rate": 8.629775026423071e-05, "loss": 1.0476, "step": 55470 }, { "epoch": 0.14, "learning_rate": 8.629649202274902e-05, "loss": 1.0271, "step": 55475 }, { "epoch": 0.14, "learning_rate": 8.629523378126731e-05, "loss": 1.027, "step": 55480 }, { "epoch": 0.14, "learning_rate": 8.62939755397856e-05, "loss": 1.0287, "step": 55485 }, { "epoch": 0.14, "learning_rate": 8.629271729830389e-05, "loss": 1.0265, "step": 55490 }, { "epoch": 0.14, "learning_rate": 8.629145905682218e-05, "loss": 1.0246, "step": 55495 }, { "epoch": 0.14, "learning_rate": 8.629020081534049e-05, "loss": 1.0266, "step": 55500 }, { "epoch": 0.14, "learning_rate": 8.628894257385878e-05, "loss": 1.0271, "step": 55505 }, { "epoch": 0.14, "learning_rate": 8.628768433237707e-05, "loss": 1.0292, "step": 55510 }, { "epoch": 0.14, "learning_rate": 8.628642609089536e-05, "loss": 1.0273, "step": 55515 }, { "epoch": 0.14, "learning_rate": 8.628516784941367e-05, "loss": 1.0289, "step": 55520 }, { "epoch": 0.14, "learning_rate": 8.628390960793196e-05, "loss": 1.0274, "step": 55525 }, { "epoch": 0.14, "learning_rate": 8.628265136645025e-05, "loss": 1.0273, "step": 55530 }, { "epoch": 0.14, "learning_rate": 8.628139312496854e-05, "loss": 1.028, "step": 55535 }, { "epoch": 0.14, "learning_rate": 8.628013488348685e-05, "loss": 1.027, "step": 55540 }, { "epoch": 0.14, "learning_rate": 8.627887664200514e-05, "loss": 1.0271, "step": 55545 }, { "epoch": 0.14, "learning_rate": 8.627761840052343e-05, "loss": 1.027, "step": 55550 }, { "epoch": 0.14, "learning_rate": 8.627636015904172e-05, "loss": 1.0276, "step": 55555 }, { "epoch": 0.14, "learning_rate": 8.627510191756001e-05, "loss": 1.0266, "step": 55560 }, { "epoch": 0.14, "learning_rate": 8.627384367607832e-05, "loss": 1.0271, "step": 55565 }, { "epoch": 0.14, "learning_rate": 8.627258543459661e-05, "loss": 1.0293, "step": 55570 }, { "epoch": 0.14, "learning_rate": 8.62713271931149e-05, "loss": 1.0279, "step": 55575 }, { "epoch": 0.14, "learning_rate": 8.62700689516332e-05, "loss": 1.0252, "step": 55580 }, { "epoch": 0.14, "learning_rate": 8.62688107101515e-05, "loss": 1.0244, "step": 55585 }, { "epoch": 0.14, "learning_rate": 8.626755246866979e-05, "loss": 1.0265, "step": 55590 }, { "epoch": 0.14, "learning_rate": 8.626629422718808e-05, "loss": 1.0278, "step": 55595 }, { "epoch": 0.14, "learning_rate": 8.626503598570637e-05, "loss": 1.0272, "step": 55600 }, { "epoch": 0.14, "learning_rate": 8.626377774422468e-05, "loss": 1.0264, "step": 55605 }, { "epoch": 0.14, "learning_rate": 8.626251950274297e-05, "loss": 1.0279, "step": 55610 }, { "epoch": 0.14, "learning_rate": 8.626126126126126e-05, "loss": 1.027, "step": 55615 }, { "epoch": 0.14, "learning_rate": 8.626000301977955e-05, "loss": 1.0271, "step": 55620 }, { "epoch": 0.14, "learning_rate": 8.625874477829784e-05, "loss": 1.0287, "step": 55625 }, { "epoch": 0.14, "learning_rate": 8.625748653681615e-05, "loss": 1.0275, "step": 55630 }, { "epoch": 0.14, "learning_rate": 8.625622829533444e-05, "loss": 1.0289, "step": 55635 }, { "epoch": 0.14, "learning_rate": 8.625497005385273e-05, "loss": 1.0291, "step": 55640 }, { "epoch": 0.14, "learning_rate": 8.625371181237102e-05, "loss": 1.0261, "step": 55645 }, { "epoch": 0.14, "learning_rate": 8.625245357088933e-05, "loss": 1.0262, "step": 55650 }, { "epoch": 0.14, "learning_rate": 8.625119532940762e-05, "loss": 1.0279, "step": 55655 }, { "epoch": 0.14, "learning_rate": 8.624993708792591e-05, "loss": 1.0273, "step": 55660 }, { "epoch": 0.14, "learning_rate": 8.62486788464442e-05, "loss": 1.0268, "step": 55665 }, { "epoch": 0.14, "learning_rate": 8.624742060496251e-05, "loss": 1.0269, "step": 55670 }, { "epoch": 0.14, "learning_rate": 8.62461623634808e-05, "loss": 1.0284, "step": 55675 }, { "epoch": 0.14, "learning_rate": 8.624490412199909e-05, "loss": 1.0278, "step": 55680 }, { "epoch": 0.14, "learning_rate": 8.62436458805174e-05, "loss": 1.0285, "step": 55685 }, { "epoch": 0.14, "learning_rate": 8.624238763903569e-05, "loss": 1.0274, "step": 55690 }, { "epoch": 0.14, "learning_rate": 8.624112939755398e-05, "loss": 1.0256, "step": 55695 }, { "epoch": 0.14, "learning_rate": 8.623987115607229e-05, "loss": 1.0252, "step": 55700 }, { "epoch": 0.14, "learning_rate": 8.623861291459058e-05, "loss": 1.0283, "step": 55705 }, { "epoch": 0.14, "learning_rate": 8.623735467310887e-05, "loss": 1.0256, "step": 55710 }, { "epoch": 0.14, "learning_rate": 8.623609643162716e-05, "loss": 1.0278, "step": 55715 }, { "epoch": 0.14, "learning_rate": 8.623483819014547e-05, "loss": 1.0273, "step": 55720 }, { "epoch": 0.14, "learning_rate": 8.623357994866376e-05, "loss": 1.0293, "step": 55725 }, { "epoch": 0.14, "learning_rate": 8.623232170718205e-05, "loss": 1.0295, "step": 55730 }, { "epoch": 0.14, "learning_rate": 8.623106346570034e-05, "loss": 1.0253, "step": 55735 }, { "epoch": 0.14, "learning_rate": 8.622980522421865e-05, "loss": 1.0286, "step": 55740 }, { "epoch": 0.14, "learning_rate": 8.622854698273694e-05, "loss": 1.0276, "step": 55745 }, { "epoch": 0.14, "learning_rate": 8.622728874125523e-05, "loss": 1.0283, "step": 55750 }, { "epoch": 0.14, "learning_rate": 8.622603049977352e-05, "loss": 1.0265, "step": 55755 }, { "epoch": 0.14, "learning_rate": 8.622477225829181e-05, "loss": 1.0289, "step": 55760 }, { "epoch": 0.14, "learning_rate": 8.622351401681012e-05, "loss": 1.0281, "step": 55765 }, { "epoch": 0.14, "learning_rate": 8.622225577532841e-05, "loss": 1.0264, "step": 55770 }, { "epoch": 0.14, "learning_rate": 8.62209975338467e-05, "loss": 1.0244, "step": 55775 }, { "epoch": 0.14, "learning_rate": 8.621973929236499e-05, "loss": 1.029, "step": 55780 }, { "epoch": 0.14, "learning_rate": 8.62184810508833e-05, "loss": 1.028, "step": 55785 }, { "epoch": 0.14, "learning_rate": 8.621722280940159e-05, "loss": 1.0491, "step": 55790 }, { "epoch": 0.14, "learning_rate": 8.621596456791988e-05, "loss": 1.0264, "step": 55795 }, { "epoch": 0.14, "learning_rate": 8.621470632643817e-05, "loss": 1.0256, "step": 55800 }, { "epoch": 0.14, "learning_rate": 8.621344808495648e-05, "loss": 1.0271, "step": 55805 }, { "epoch": 0.14, "learning_rate": 8.621218984347477e-05, "loss": 1.0269, "step": 55810 }, { "epoch": 0.14, "learning_rate": 8.621093160199306e-05, "loss": 1.0277, "step": 55815 }, { "epoch": 0.14, "learning_rate": 8.620967336051135e-05, "loss": 1.0271, "step": 55820 }, { "epoch": 0.14, "learning_rate": 8.620841511902964e-05, "loss": 1.0283, "step": 55825 }, { "epoch": 0.14, "learning_rate": 8.620715687754795e-05, "loss": 1.0286, "step": 55830 }, { "epoch": 0.14, "learning_rate": 8.620589863606624e-05, "loss": 1.0265, "step": 55835 }, { "epoch": 0.14, "learning_rate": 8.620464039458453e-05, "loss": 1.0267, "step": 55840 }, { "epoch": 0.14, "learning_rate": 8.620338215310282e-05, "loss": 1.0278, "step": 55845 }, { "epoch": 0.14, "learning_rate": 8.620212391162113e-05, "loss": 1.0253, "step": 55850 }, { "epoch": 0.14, "learning_rate": 8.620086567013942e-05, "loss": 1.0241, "step": 55855 }, { "epoch": 0.14, "learning_rate": 8.619960742865771e-05, "loss": 1.0279, "step": 55860 }, { "epoch": 0.14, "learning_rate": 8.6198349187176e-05, "loss": 1.025, "step": 55865 }, { "epoch": 0.14, "learning_rate": 8.619709094569431e-05, "loss": 1.0265, "step": 55870 }, { "epoch": 0.14, "learning_rate": 8.61958327042126e-05, "loss": 1.0277, "step": 55875 }, { "epoch": 0.14, "learning_rate": 8.619457446273089e-05, "loss": 1.0264, "step": 55880 }, { "epoch": 0.14, "learning_rate": 8.619331622124918e-05, "loss": 1.0261, "step": 55885 }, { "epoch": 0.14, "learning_rate": 8.619205797976748e-05, "loss": 1.0288, "step": 55890 }, { "epoch": 0.14, "learning_rate": 8.619079973828578e-05, "loss": 1.0275, "step": 55895 }, { "epoch": 0.14, "learning_rate": 8.618954149680407e-05, "loss": 1.0244, "step": 55900 }, { "epoch": 0.14, "learning_rate": 8.618828325532236e-05, "loss": 1.0286, "step": 55905 }, { "epoch": 0.14, "learning_rate": 8.618702501384066e-05, "loss": 1.0272, "step": 55910 }, { "epoch": 0.14, "learning_rate": 8.618576677235896e-05, "loss": 1.029, "step": 55915 }, { "epoch": 0.14, "learning_rate": 8.618450853087725e-05, "loss": 1.0269, "step": 55920 }, { "epoch": 0.14, "learning_rate": 8.618325028939554e-05, "loss": 1.0279, "step": 55925 }, { "epoch": 0.14, "learning_rate": 8.618199204791384e-05, "loss": 1.0254, "step": 55930 }, { "epoch": 0.14, "learning_rate": 8.618073380643214e-05, "loss": 1.0273, "step": 55935 }, { "epoch": 0.14, "learning_rate": 8.617947556495043e-05, "loss": 1.0267, "step": 55940 }, { "epoch": 0.14, "learning_rate": 8.617821732346872e-05, "loss": 1.0281, "step": 55945 }, { "epoch": 0.14, "learning_rate": 8.617695908198702e-05, "loss": 1.0284, "step": 55950 }, { "epoch": 0.14, "learning_rate": 8.617570084050531e-05, "loss": 1.0295, "step": 55955 }, { "epoch": 0.14, "learning_rate": 8.617444259902361e-05, "loss": 1.0251, "step": 55960 }, { "epoch": 0.14, "learning_rate": 8.61731843575419e-05, "loss": 1.0265, "step": 55965 }, { "epoch": 0.14, "learning_rate": 8.61719261160602e-05, "loss": 1.0278, "step": 55970 }, { "epoch": 0.14, "learning_rate": 8.617066787457849e-05, "loss": 1.0266, "step": 55975 }, { "epoch": 0.14, "learning_rate": 8.616940963309679e-05, "loss": 1.0262, "step": 55980 }, { "epoch": 0.14, "learning_rate": 8.616815139161508e-05, "loss": 1.0267, "step": 55985 }, { "epoch": 0.14, "learning_rate": 8.616689315013338e-05, "loss": 1.0275, "step": 55990 }, { "epoch": 0.14, "learning_rate": 8.616563490865167e-05, "loss": 1.0268, "step": 55995 }, { "epoch": 0.14, "learning_rate": 8.616437666716997e-05, "loss": 1.0271, "step": 56000 }, { "epoch": 0.14, "learning_rate": 8.616311842568826e-05, "loss": 1.0262, "step": 56005 }, { "epoch": 0.14, "learning_rate": 8.616186018420656e-05, "loss": 1.0266, "step": 56010 }, { "epoch": 0.14, "learning_rate": 8.616060194272485e-05, "loss": 1.0299, "step": 56015 }, { "epoch": 0.14, "learning_rate": 8.615934370124314e-05, "loss": 1.0277, "step": 56020 }, { "epoch": 0.14, "learning_rate": 8.615808545976144e-05, "loss": 1.0287, "step": 56025 }, { "epoch": 0.14, "learning_rate": 8.615682721827974e-05, "loss": 1.0267, "step": 56030 }, { "epoch": 0.14, "learning_rate": 8.615556897679803e-05, "loss": 1.0268, "step": 56035 }, { "epoch": 0.14, "learning_rate": 8.615431073531632e-05, "loss": 1.026, "step": 56040 }, { "epoch": 0.14, "learning_rate": 8.615305249383462e-05, "loss": 1.0286, "step": 56045 }, { "epoch": 0.14, "learning_rate": 8.615179425235292e-05, "loss": 1.0278, "step": 56050 }, { "epoch": 0.14, "learning_rate": 8.615053601087121e-05, "loss": 1.0256, "step": 56055 }, { "epoch": 0.14, "learning_rate": 8.61492777693895e-05, "loss": 1.029, "step": 56060 }, { "epoch": 0.14, "learning_rate": 8.61480195279078e-05, "loss": 1.0258, "step": 56065 }, { "epoch": 0.14, "learning_rate": 8.61467612864261e-05, "loss": 1.0267, "step": 56070 }, { "epoch": 0.14, "learning_rate": 8.614550304494439e-05, "loss": 1.0266, "step": 56075 }, { "epoch": 0.14, "learning_rate": 8.614424480346268e-05, "loss": 1.0283, "step": 56080 }, { "epoch": 0.14, "learning_rate": 8.614298656198097e-05, "loss": 1.029, "step": 56085 }, { "epoch": 0.14, "learning_rate": 8.614172832049928e-05, "loss": 1.0269, "step": 56090 }, { "epoch": 0.14, "learning_rate": 8.614047007901757e-05, "loss": 1.0281, "step": 56095 }, { "epoch": 0.14, "learning_rate": 8.613921183753586e-05, "loss": 1.0268, "step": 56100 }, { "epoch": 0.14, "learning_rate": 8.613795359605415e-05, "loss": 1.0248, "step": 56105 }, { "epoch": 0.14, "learning_rate": 8.613669535457246e-05, "loss": 1.0254, "step": 56110 }, { "epoch": 0.14, "learning_rate": 8.613543711309075e-05, "loss": 1.0274, "step": 56115 }, { "epoch": 0.14, "learning_rate": 8.613417887160904e-05, "loss": 1.0289, "step": 56120 }, { "epoch": 0.14, "learning_rate": 8.613292063012733e-05, "loss": 1.0281, "step": 56125 }, { "epoch": 0.14, "learning_rate": 8.613166238864563e-05, "loss": 1.0283, "step": 56130 }, { "epoch": 0.14, "learning_rate": 8.613040414716393e-05, "loss": 1.0292, "step": 56135 }, { "epoch": 0.14, "learning_rate": 8.612914590568222e-05, "loss": 1.0285, "step": 56140 }, { "epoch": 0.14, "learning_rate": 8.612788766420051e-05, "loss": 1.0278, "step": 56145 }, { "epoch": 0.14, "learning_rate": 8.61266294227188e-05, "loss": 1.0292, "step": 56150 }, { "epoch": 0.14, "learning_rate": 8.61253711812371e-05, "loss": 1.0275, "step": 56155 }, { "epoch": 0.14, "learning_rate": 8.61241129397554e-05, "loss": 1.029, "step": 56160 }, { "epoch": 0.14, "learning_rate": 8.612285469827369e-05, "loss": 1.0256, "step": 56165 }, { "epoch": 0.14, "learning_rate": 8.612159645679198e-05, "loss": 1.0268, "step": 56170 }, { "epoch": 0.14, "learning_rate": 8.612033821531029e-05, "loss": 1.0259, "step": 56175 }, { "epoch": 0.14, "learning_rate": 8.611907997382858e-05, "loss": 1.0252, "step": 56180 }, { "epoch": 0.14, "learning_rate": 8.611782173234688e-05, "loss": 1.0267, "step": 56185 }, { "epoch": 0.14, "learning_rate": 8.611656349086517e-05, "loss": 1.0264, "step": 56190 }, { "epoch": 0.14, "learning_rate": 8.611530524938347e-05, "loss": 1.0266, "step": 56195 }, { "epoch": 0.14, "learning_rate": 8.611404700790177e-05, "loss": 1.0297, "step": 56200 }, { "epoch": 0.14, "learning_rate": 8.611278876642006e-05, "loss": 1.0257, "step": 56205 }, { "epoch": 0.14, "learning_rate": 8.611153052493835e-05, "loss": 1.0255, "step": 56210 }, { "epoch": 0.14, "learning_rate": 8.611027228345665e-05, "loss": 1.0269, "step": 56215 }, { "epoch": 0.14, "learning_rate": 8.610901404197494e-05, "loss": 1.0254, "step": 56220 }, { "epoch": 0.14, "learning_rate": 8.610775580049324e-05, "loss": 1.0265, "step": 56225 }, { "epoch": 0.14, "learning_rate": 8.610649755901153e-05, "loss": 1.0292, "step": 56230 }, { "epoch": 0.14, "learning_rate": 8.610523931752983e-05, "loss": 1.0257, "step": 56235 }, { "epoch": 0.14, "learning_rate": 8.610398107604812e-05, "loss": 1.0281, "step": 56240 }, { "epoch": 0.14, "learning_rate": 8.610272283456642e-05, "loss": 1.0283, "step": 56245 }, { "epoch": 0.14, "learning_rate": 8.610146459308471e-05, "loss": 1.0287, "step": 56250 }, { "epoch": 0.14, "learning_rate": 8.6100206351603e-05, "loss": 1.0263, "step": 56255 }, { "epoch": 0.14, "learning_rate": 8.60989481101213e-05, "loss": 1.0354, "step": 56260 }, { "epoch": 0.14, "learning_rate": 8.60976898686396e-05, "loss": 1.0259, "step": 56265 }, { "epoch": 0.14, "learning_rate": 8.60964316271579e-05, "loss": 1.0247, "step": 56270 }, { "epoch": 0.14, "learning_rate": 8.609517338567619e-05, "loss": 1.0262, "step": 56275 }, { "epoch": 0.14, "learning_rate": 8.609391514419448e-05, "loss": 1.028, "step": 56280 }, { "epoch": 0.14, "learning_rate": 8.609265690271277e-05, "loss": 1.027, "step": 56285 }, { "epoch": 0.14, "learning_rate": 8.609139866123107e-05, "loss": 1.0365, "step": 56290 }, { "epoch": 0.14, "learning_rate": 8.609014041974937e-05, "loss": 1.0275, "step": 56295 }, { "epoch": 0.14, "learning_rate": 8.608888217826766e-05, "loss": 1.0273, "step": 56300 }, { "epoch": 0.14, "learning_rate": 8.608762393678595e-05, "loss": 1.0265, "step": 56305 }, { "epoch": 0.14, "learning_rate": 8.608636569530425e-05, "loss": 1.0283, "step": 56310 }, { "epoch": 0.14, "learning_rate": 8.608510745382255e-05, "loss": 1.0285, "step": 56315 }, { "epoch": 0.14, "learning_rate": 8.608384921234084e-05, "loss": 1.0266, "step": 56320 }, { "epoch": 0.14, "learning_rate": 8.608259097085913e-05, "loss": 1.0272, "step": 56325 }, { "epoch": 0.14, "learning_rate": 8.608133272937743e-05, "loss": 1.0302, "step": 56330 }, { "epoch": 0.14, "learning_rate": 8.608007448789573e-05, "loss": 1.0278, "step": 56335 }, { "epoch": 0.14, "learning_rate": 8.607881624641402e-05, "loss": 1.024, "step": 56340 }, { "epoch": 0.14, "learning_rate": 8.607755800493231e-05, "loss": 1.0259, "step": 56345 }, { "epoch": 0.14, "learning_rate": 8.60762997634506e-05, "loss": 1.0414, "step": 56350 }, { "epoch": 0.14, "learning_rate": 8.60750415219689e-05, "loss": 1.0297, "step": 56355 }, { "epoch": 0.14, "learning_rate": 8.60737832804872e-05, "loss": 1.0257, "step": 56360 }, { "epoch": 0.14, "learning_rate": 8.607252503900549e-05, "loss": 1.0469, "step": 56365 }, { "epoch": 0.14, "learning_rate": 8.607126679752378e-05, "loss": 1.0287, "step": 56370 }, { "epoch": 0.14, "learning_rate": 8.607000855604209e-05, "loss": 1.0259, "step": 56375 }, { "epoch": 0.14, "learning_rate": 8.606875031456038e-05, "loss": 1.0284, "step": 56380 }, { "epoch": 0.14, "learning_rate": 8.606749207307867e-05, "loss": 1.0286, "step": 56385 }, { "epoch": 0.14, "learning_rate": 8.606623383159696e-05, "loss": 1.0263, "step": 56390 }, { "epoch": 0.14, "learning_rate": 8.606497559011527e-05, "loss": 1.0283, "step": 56395 }, { "epoch": 0.14, "learning_rate": 8.606371734863356e-05, "loss": 1.029, "step": 56400 }, { "epoch": 0.14, "learning_rate": 8.606245910715185e-05, "loss": 1.0287, "step": 56405 }, { "epoch": 0.14, "learning_rate": 8.606120086567014e-05, "loss": 1.0273, "step": 56410 }, { "epoch": 0.14, "learning_rate": 8.605994262418843e-05, "loss": 1.0269, "step": 56415 }, { "epoch": 0.14, "learning_rate": 8.605868438270674e-05, "loss": 1.0272, "step": 56420 }, { "epoch": 0.14, "learning_rate": 8.605742614122503e-05, "loss": 1.026, "step": 56425 }, { "epoch": 0.14, "learning_rate": 8.605616789974332e-05, "loss": 1.026, "step": 56430 }, { "epoch": 0.14, "learning_rate": 8.605490965826161e-05, "loss": 1.0276, "step": 56435 }, { "epoch": 0.14, "learning_rate": 8.605365141677992e-05, "loss": 1.0288, "step": 56440 }, { "epoch": 0.14, "learning_rate": 8.605239317529821e-05, "loss": 1.0265, "step": 56445 }, { "epoch": 0.14, "learning_rate": 8.60511349338165e-05, "loss": 1.0281, "step": 56450 }, { "epoch": 0.14, "learning_rate": 8.604987669233479e-05, "loss": 1.0475, "step": 56455 }, { "epoch": 0.14, "learning_rate": 8.60486184508531e-05, "loss": 1.0261, "step": 56460 }, { "epoch": 0.14, "learning_rate": 8.604736020937139e-05, "loss": 1.026, "step": 56465 }, { "epoch": 0.14, "learning_rate": 8.604610196788968e-05, "loss": 1.0276, "step": 56470 }, { "epoch": 0.14, "learning_rate": 8.604484372640797e-05, "loss": 1.025, "step": 56475 }, { "epoch": 0.14, "learning_rate": 8.604358548492626e-05, "loss": 1.0281, "step": 56480 }, { "epoch": 0.14, "learning_rate": 8.604232724344457e-05, "loss": 1.0289, "step": 56485 }, { "epoch": 0.14, "learning_rate": 8.604106900196286e-05, "loss": 1.028, "step": 56490 }, { "epoch": 0.14, "learning_rate": 8.603981076048115e-05, "loss": 1.026, "step": 56495 }, { "epoch": 0.14, "learning_rate": 8.603855251899944e-05, "loss": 1.0248, "step": 56500 }, { "epoch": 0.14, "learning_rate": 8.603729427751775e-05, "loss": 1.0256, "step": 56505 }, { "epoch": 0.14, "learning_rate": 8.603603603603604e-05, "loss": 1.0268, "step": 56510 }, { "epoch": 0.14, "learning_rate": 8.603477779455433e-05, "loss": 1.0279, "step": 56515 }, { "epoch": 0.14, "learning_rate": 8.603351955307262e-05, "loss": 1.0288, "step": 56520 }, { "epoch": 0.14, "learning_rate": 8.603226131159093e-05, "loss": 1.0256, "step": 56525 }, { "epoch": 0.14, "learning_rate": 8.603100307010922e-05, "loss": 1.0247, "step": 56530 }, { "epoch": 0.14, "learning_rate": 8.602974482862751e-05, "loss": 1.0272, "step": 56535 }, { "epoch": 0.14, "learning_rate": 8.60284865871458e-05, "loss": 1.0263, "step": 56540 }, { "epoch": 0.14, "learning_rate": 8.60272283456641e-05, "loss": 1.0273, "step": 56545 }, { "epoch": 0.14, "learning_rate": 8.60259701041824e-05, "loss": 1.0261, "step": 56550 }, { "epoch": 0.14, "learning_rate": 8.602471186270069e-05, "loss": 1.0264, "step": 56555 }, { "epoch": 0.14, "learning_rate": 8.602345362121898e-05, "loss": 1.0295, "step": 56560 }, { "epoch": 0.14, "learning_rate": 8.602219537973728e-05, "loss": 1.0243, "step": 56565 }, { "epoch": 0.14, "learning_rate": 8.602093713825558e-05, "loss": 1.0286, "step": 56570 }, { "epoch": 0.14, "learning_rate": 8.601967889677387e-05, "loss": 1.0263, "step": 56575 }, { "epoch": 0.14, "learning_rate": 8.601842065529216e-05, "loss": 1.0593, "step": 56580 }, { "epoch": 0.14, "learning_rate": 8.601716241381045e-05, "loss": 1.0259, "step": 56585 }, { "epoch": 0.14, "learning_rate": 8.601590417232876e-05, "loss": 1.0259, "step": 56590 }, { "epoch": 0.14, "learning_rate": 8.601464593084705e-05, "loss": 1.0406, "step": 56595 }, { "epoch": 0.14, "learning_rate": 8.601338768936534e-05, "loss": 1.0282, "step": 56600 }, { "epoch": 0.14, "learning_rate": 8.601212944788363e-05, "loss": 1.0294, "step": 56605 }, { "epoch": 0.14, "learning_rate": 8.601087120640193e-05, "loss": 1.0269, "step": 56610 }, { "epoch": 0.14, "learning_rate": 8.600961296492023e-05, "loss": 1.0277, "step": 56615 }, { "epoch": 0.14, "learning_rate": 8.600835472343852e-05, "loss": 1.0274, "step": 56620 }, { "epoch": 0.14, "learning_rate": 8.600709648195681e-05, "loss": 1.0299, "step": 56625 }, { "epoch": 0.14, "learning_rate": 8.60058382404751e-05, "loss": 1.0245, "step": 56630 }, { "epoch": 0.14, "learning_rate": 8.600457999899341e-05, "loss": 1.0256, "step": 56635 }, { "epoch": 0.14, "learning_rate": 8.60033217575117e-05, "loss": 1.0292, "step": 56640 }, { "epoch": 0.14, "learning_rate": 8.600206351603e-05, "loss": 1.0289, "step": 56645 }, { "epoch": 0.14, "learning_rate": 8.600080527454829e-05, "loss": 1.0269, "step": 56650 }, { "epoch": 0.14, "learning_rate": 8.599954703306659e-05, "loss": 1.0468, "step": 56655 }, { "epoch": 0.14, "learning_rate": 8.599828879158488e-05, "loss": 1.0282, "step": 56660 }, { "epoch": 0.14, "learning_rate": 8.599703055010317e-05, "loss": 1.0287, "step": 56665 }, { "epoch": 0.14, "learning_rate": 8.599577230862147e-05, "loss": 1.0271, "step": 56670 }, { "epoch": 0.14, "learning_rate": 8.599451406713976e-05, "loss": 1.0265, "step": 56675 }, { "epoch": 0.14, "learning_rate": 8.599325582565806e-05, "loss": 1.028, "step": 56680 }, { "epoch": 0.14, "learning_rate": 8.599199758417637e-05, "loss": 1.0278, "step": 56685 }, { "epoch": 0.14, "learning_rate": 8.599073934269466e-05, "loss": 1.0263, "step": 56690 }, { "epoch": 0.14, "learning_rate": 8.598948110121295e-05, "loss": 1.0279, "step": 56695 }, { "epoch": 0.14, "learning_rate": 8.598822285973124e-05, "loss": 1.0265, "step": 56700 }, { "epoch": 0.14, "learning_rate": 8.598696461824955e-05, "loss": 1.0247, "step": 56705 }, { "epoch": 0.14, "learning_rate": 8.598570637676784e-05, "loss": 1.0292, "step": 56710 }, { "epoch": 0.14, "learning_rate": 8.598444813528613e-05, "loss": 1.0271, "step": 56715 }, { "epoch": 0.14, "learning_rate": 8.598318989380442e-05, "loss": 1.0271, "step": 56720 }, { "epoch": 0.14, "learning_rate": 8.598193165232273e-05, "loss": 1.0272, "step": 56725 }, { "epoch": 0.14, "learning_rate": 8.598067341084102e-05, "loss": 1.0286, "step": 56730 }, { "epoch": 0.14, "learning_rate": 8.597941516935931e-05, "loss": 1.025, "step": 56735 }, { "epoch": 0.14, "learning_rate": 8.59781569278776e-05, "loss": 1.0264, "step": 56740 }, { "epoch": 0.14, "learning_rate": 8.597689868639591e-05, "loss": 1.025, "step": 56745 }, { "epoch": 0.14, "learning_rate": 8.59756404449142e-05, "loss": 1.0261, "step": 56750 }, { "epoch": 0.14, "learning_rate": 8.597438220343249e-05, "loss": 1.0272, "step": 56755 }, { "epoch": 0.14, "learning_rate": 8.597312396195078e-05, "loss": 1.0257, "step": 56760 }, { "epoch": 0.14, "learning_rate": 8.597186572046907e-05, "loss": 1.0251, "step": 56765 }, { "epoch": 0.14, "learning_rate": 8.597060747898738e-05, "loss": 1.0272, "step": 56770 }, { "epoch": 0.14, "learning_rate": 8.596934923750567e-05, "loss": 1.0274, "step": 56775 }, { "epoch": 0.14, "learning_rate": 8.596809099602396e-05, "loss": 1.0263, "step": 56780 }, { "epoch": 0.14, "learning_rate": 8.596683275454225e-05, "loss": 1.028, "step": 56785 }, { "epoch": 0.14, "learning_rate": 8.596557451306056e-05, "loss": 1.0268, "step": 56790 }, { "epoch": 0.14, "learning_rate": 8.596431627157885e-05, "loss": 1.0277, "step": 56795 }, { "epoch": 0.14, "learning_rate": 8.596305803009714e-05, "loss": 1.0271, "step": 56800 }, { "epoch": 0.14, "learning_rate": 8.596179978861543e-05, "loss": 1.0271, "step": 56805 }, { "epoch": 0.14, "learning_rate": 8.596054154713374e-05, "loss": 1.0254, "step": 56810 }, { "epoch": 0.14, "learning_rate": 8.595928330565203e-05, "loss": 1.0271, "step": 56815 }, { "epoch": 0.14, "learning_rate": 8.595802506417032e-05, "loss": 1.0266, "step": 56820 }, { "epoch": 0.14, "learning_rate": 8.595676682268861e-05, "loss": 1.0259, "step": 56825 }, { "epoch": 0.14, "learning_rate": 8.59555085812069e-05, "loss": 1.0275, "step": 56830 }, { "epoch": 0.14, "learning_rate": 8.595425033972521e-05, "loss": 1.0265, "step": 56835 }, { "epoch": 0.14, "learning_rate": 8.59529920982435e-05, "loss": 1.0227, "step": 56840 }, { "epoch": 0.14, "learning_rate": 8.59517338567618e-05, "loss": 1.0285, "step": 56845 }, { "epoch": 0.14, "learning_rate": 8.595047561528009e-05, "loss": 1.0284, "step": 56850 }, { "epoch": 0.14, "learning_rate": 8.594921737379839e-05, "loss": 1.0275, "step": 56855 }, { "epoch": 0.14, "learning_rate": 8.594795913231668e-05, "loss": 1.0286, "step": 56860 }, { "epoch": 0.14, "learning_rate": 8.594670089083497e-05, "loss": 1.0273, "step": 56865 }, { "epoch": 0.14, "learning_rate": 8.594544264935327e-05, "loss": 1.0306, "step": 56870 }, { "epoch": 0.14, "learning_rate": 8.594418440787156e-05, "loss": 1.029, "step": 56875 }, { "epoch": 0.14, "learning_rate": 8.594292616638986e-05, "loss": 1.0277, "step": 56880 }, { "epoch": 0.14, "learning_rate": 8.594166792490815e-05, "loss": 1.0265, "step": 56885 }, { "epoch": 0.14, "learning_rate": 8.594040968342645e-05, "loss": 1.0256, "step": 56890 }, { "epoch": 0.14, "learning_rate": 8.593915144194474e-05, "loss": 1.0243, "step": 56895 }, { "epoch": 0.14, "learning_rate": 8.593789320046304e-05, "loss": 1.0262, "step": 56900 }, { "epoch": 0.14, "learning_rate": 8.593663495898133e-05, "loss": 1.0284, "step": 56905 }, { "epoch": 0.14, "learning_rate": 8.593537671749963e-05, "loss": 1.0294, "step": 56910 }, { "epoch": 0.14, "learning_rate": 8.593411847601792e-05, "loss": 1.0272, "step": 56915 }, { "epoch": 0.14, "learning_rate": 8.593286023453622e-05, "loss": 1.0275, "step": 56920 }, { "epoch": 0.14, "learning_rate": 8.593160199305451e-05, "loss": 1.0271, "step": 56925 }, { "epoch": 0.14, "learning_rate": 8.59303437515728e-05, "loss": 1.0277, "step": 56930 }, { "epoch": 0.14, "learning_rate": 8.59290855100911e-05, "loss": 1.0259, "step": 56935 }, { "epoch": 0.14, "learning_rate": 8.592782726860939e-05, "loss": 1.0266, "step": 56940 }, { "epoch": 0.14, "learning_rate": 8.59265690271277e-05, "loss": 1.0266, "step": 56945 }, { "epoch": 0.14, "learning_rate": 8.592531078564599e-05, "loss": 1.0267, "step": 56950 }, { "epoch": 0.14, "learning_rate": 8.592405254416428e-05, "loss": 1.0269, "step": 56955 }, { "epoch": 0.14, "learning_rate": 8.592279430268257e-05, "loss": 1.0289, "step": 56960 }, { "epoch": 0.14, "learning_rate": 8.592153606120087e-05, "loss": 1.0275, "step": 56965 }, { "epoch": 0.14, "learning_rate": 8.592027781971917e-05, "loss": 1.0276, "step": 56970 }, { "epoch": 0.14, "learning_rate": 8.591901957823746e-05, "loss": 1.0272, "step": 56975 }, { "epoch": 0.14, "learning_rate": 8.591776133675575e-05, "loss": 1.027, "step": 56980 }, { "epoch": 0.14, "learning_rate": 8.591650309527405e-05, "loss": 1.0252, "step": 56985 }, { "epoch": 0.14, "learning_rate": 8.591524485379235e-05, "loss": 1.0502, "step": 56990 }, { "epoch": 0.14, "learning_rate": 8.591398661231064e-05, "loss": 1.0281, "step": 56995 }, { "epoch": 0.14, "learning_rate": 8.591272837082893e-05, "loss": 1.0257, "step": 57000 }, { "epoch": 0.14, "learning_rate": 8.591147012934722e-05, "loss": 1.0254, "step": 57005 }, { "epoch": 0.14, "learning_rate": 8.591021188786553e-05, "loss": 1.0262, "step": 57010 }, { "epoch": 0.14, "learning_rate": 8.590895364638382e-05, "loss": 1.0271, "step": 57015 }, { "epoch": 0.14, "learning_rate": 8.590769540490211e-05, "loss": 1.0278, "step": 57020 }, { "epoch": 0.14, "learning_rate": 8.59064371634204e-05, "loss": 1.0289, "step": 57025 }, { "epoch": 0.14, "learning_rate": 8.59051789219387e-05, "loss": 1.027, "step": 57030 }, { "epoch": 0.14, "learning_rate": 8.5903920680457e-05, "loss": 1.0266, "step": 57035 }, { "epoch": 0.14, "learning_rate": 8.590266243897529e-05, "loss": 1.0293, "step": 57040 }, { "epoch": 0.14, "learning_rate": 8.590140419749358e-05, "loss": 1.0273, "step": 57045 }, { "epoch": 0.14, "learning_rate": 8.590014595601189e-05, "loss": 1.0259, "step": 57050 }, { "epoch": 0.14, "learning_rate": 8.589888771453018e-05, "loss": 1.0269, "step": 57055 }, { "epoch": 0.14, "learning_rate": 8.589762947304847e-05, "loss": 1.0244, "step": 57060 }, { "epoch": 0.14, "learning_rate": 8.589637123156676e-05, "loss": 1.024, "step": 57065 }, { "epoch": 0.14, "learning_rate": 8.589511299008505e-05, "loss": 1.0295, "step": 57070 }, { "epoch": 0.14, "learning_rate": 8.589385474860336e-05, "loss": 1.0275, "step": 57075 }, { "epoch": 0.14, "learning_rate": 8.589259650712165e-05, "loss": 1.0249, "step": 57080 }, { "epoch": 0.14, "learning_rate": 8.589133826563994e-05, "loss": 1.0257, "step": 57085 }, { "epoch": 0.14, "learning_rate": 8.589008002415823e-05, "loss": 1.0257, "step": 57090 }, { "epoch": 0.14, "learning_rate": 8.588882178267654e-05, "loss": 1.0254, "step": 57095 }, { "epoch": 0.14, "learning_rate": 8.588756354119483e-05, "loss": 1.0257, "step": 57100 }, { "epoch": 0.14, "learning_rate": 8.588630529971312e-05, "loss": 1.0261, "step": 57105 }, { "epoch": 0.14, "learning_rate": 8.588504705823141e-05, "loss": 1.0263, "step": 57110 }, { "epoch": 0.14, "learning_rate": 8.588378881674972e-05, "loss": 1.0272, "step": 57115 }, { "epoch": 0.14, "learning_rate": 8.588253057526801e-05, "loss": 1.0264, "step": 57120 }, { "epoch": 0.14, "learning_rate": 8.58812723337863e-05, "loss": 1.0451, "step": 57125 }, { "epoch": 0.14, "learning_rate": 8.588001409230459e-05, "loss": 1.0276, "step": 57130 }, { "epoch": 0.14, "learning_rate": 8.587875585082288e-05, "loss": 1.0277, "step": 57135 }, { "epoch": 0.14, "learning_rate": 8.587749760934119e-05, "loss": 1.0261, "step": 57140 }, { "epoch": 0.14, "learning_rate": 8.587623936785948e-05, "loss": 1.0271, "step": 57145 }, { "epoch": 0.14, "learning_rate": 8.587498112637777e-05, "loss": 1.0281, "step": 57150 }, { "epoch": 0.14, "learning_rate": 8.587372288489606e-05, "loss": 1.0275, "step": 57155 }, { "epoch": 0.14, "learning_rate": 8.587246464341437e-05, "loss": 1.0278, "step": 57160 }, { "epoch": 0.14, "learning_rate": 8.587120640193266e-05, "loss": 1.0291, "step": 57165 }, { "epoch": 0.14, "learning_rate": 8.586994816045095e-05, "loss": 1.0276, "step": 57170 }, { "epoch": 0.14, "learning_rate": 8.586868991896924e-05, "loss": 1.0269, "step": 57175 }, { "epoch": 0.14, "learning_rate": 8.586743167748755e-05, "loss": 1.0281, "step": 57180 }, { "epoch": 0.14, "learning_rate": 8.586617343600585e-05, "loss": 1.0272, "step": 57185 }, { "epoch": 0.14, "learning_rate": 8.586491519452414e-05, "loss": 1.0282, "step": 57190 }, { "epoch": 0.14, "learning_rate": 8.586365695304244e-05, "loss": 1.027, "step": 57195 }, { "epoch": 0.14, "learning_rate": 8.586239871156073e-05, "loss": 1.027, "step": 57200 }, { "epoch": 0.14, "learning_rate": 8.586114047007903e-05, "loss": 1.0278, "step": 57205 }, { "epoch": 0.14, "learning_rate": 8.585988222859732e-05, "loss": 1.027, "step": 57210 }, { "epoch": 0.14, "learning_rate": 8.585862398711562e-05, "loss": 1.0293, "step": 57215 }, { "epoch": 0.14, "learning_rate": 8.585736574563391e-05, "loss": 1.0286, "step": 57220 }, { "epoch": 0.14, "learning_rate": 8.58561075041522e-05, "loss": 1.0278, "step": 57225 }, { "epoch": 0.14, "learning_rate": 8.58548492626705e-05, "loss": 1.0272, "step": 57230 }, { "epoch": 0.14, "learning_rate": 8.58535910211888e-05, "loss": 1.0282, "step": 57235 }, { "epoch": 0.14, "learning_rate": 8.585233277970709e-05, "loss": 1.0262, "step": 57240 }, { "epoch": 0.14, "learning_rate": 8.585107453822538e-05, "loss": 1.0266, "step": 57245 }, { "epoch": 0.14, "learning_rate": 8.584981629674368e-05, "loss": 1.0264, "step": 57250 }, { "epoch": 0.14, "learning_rate": 8.584855805526198e-05, "loss": 1.0269, "step": 57255 }, { "epoch": 0.14, "learning_rate": 8.584729981378027e-05, "loss": 1.027, "step": 57260 }, { "epoch": 0.14, "learning_rate": 8.584604157229856e-05, "loss": 1.0294, "step": 57265 }, { "epoch": 0.14, "learning_rate": 8.584478333081686e-05, "loss": 1.0256, "step": 57270 }, { "epoch": 0.14, "learning_rate": 8.584352508933516e-05, "loss": 1.0292, "step": 57275 }, { "epoch": 0.14, "learning_rate": 8.584226684785345e-05, "loss": 1.0295, "step": 57280 }, { "epoch": 0.14, "learning_rate": 8.584100860637174e-05, "loss": 1.0258, "step": 57285 }, { "epoch": 0.14, "learning_rate": 8.583975036489003e-05, "loss": 1.0271, "step": 57290 }, { "epoch": 0.14, "learning_rate": 8.583849212340834e-05, "loss": 1.0274, "step": 57295 }, { "epoch": 0.14, "learning_rate": 8.583723388192663e-05, "loss": 1.0274, "step": 57300 }, { "epoch": 0.14, "learning_rate": 8.583597564044492e-05, "loss": 1.027, "step": 57305 }, { "epoch": 0.14, "learning_rate": 8.583471739896321e-05, "loss": 1.0259, "step": 57310 }, { "epoch": 0.14, "learning_rate": 8.583345915748152e-05, "loss": 1.0269, "step": 57315 }, { "epoch": 0.14, "learning_rate": 8.583220091599981e-05, "loss": 1.0289, "step": 57320 }, { "epoch": 0.14, "learning_rate": 8.58309426745181e-05, "loss": 1.0267, "step": 57325 }, { "epoch": 0.14, "learning_rate": 8.582968443303639e-05, "loss": 1.0328, "step": 57330 }, { "epoch": 0.14, "learning_rate": 8.58284261915547e-05, "loss": 1.0263, "step": 57335 }, { "epoch": 0.14, "learning_rate": 8.582716795007299e-05, "loss": 1.0282, "step": 57340 }, { "epoch": 0.14, "learning_rate": 8.582590970859128e-05, "loss": 1.0269, "step": 57345 }, { "epoch": 0.14, "learning_rate": 8.582465146710957e-05, "loss": 1.0238, "step": 57350 }, { "epoch": 0.14, "learning_rate": 8.582339322562786e-05, "loss": 1.0289, "step": 57355 }, { "epoch": 0.14, "learning_rate": 8.582213498414617e-05, "loss": 1.0255, "step": 57360 }, { "epoch": 0.14, "learning_rate": 8.582087674266446e-05, "loss": 1.0268, "step": 57365 }, { "epoch": 0.14, "learning_rate": 8.581961850118275e-05, "loss": 1.0274, "step": 57370 }, { "epoch": 0.14, "learning_rate": 8.581836025970104e-05, "loss": 1.0268, "step": 57375 }, { "epoch": 0.14, "learning_rate": 8.581710201821935e-05, "loss": 1.0271, "step": 57380 }, { "epoch": 0.14, "learning_rate": 8.581584377673764e-05, "loss": 1.0292, "step": 57385 }, { "epoch": 0.14, "learning_rate": 8.581458553525593e-05, "loss": 1.0249, "step": 57390 }, { "epoch": 0.14, "learning_rate": 8.581332729377422e-05, "loss": 1.0276, "step": 57395 }, { "epoch": 0.14, "learning_rate": 8.581206905229253e-05, "loss": 1.0262, "step": 57400 }, { "epoch": 0.14, "learning_rate": 8.581081081081082e-05, "loss": 1.0246, "step": 57405 }, { "epoch": 0.14, "learning_rate": 8.580955256932911e-05, "loss": 1.026, "step": 57410 }, { "epoch": 0.14, "learning_rate": 8.58082943278474e-05, "loss": 1.0269, "step": 57415 }, { "epoch": 0.14, "learning_rate": 8.58070360863657e-05, "loss": 1.0281, "step": 57420 }, { "epoch": 0.14, "learning_rate": 8.5805777844884e-05, "loss": 1.0267, "step": 57425 }, { "epoch": 0.14, "learning_rate": 8.580451960340229e-05, "loss": 1.027, "step": 57430 }, { "epoch": 0.14, "learning_rate": 8.580326136192058e-05, "loss": 1.0271, "step": 57435 }, { "epoch": 0.14, "learning_rate": 8.580200312043887e-05, "loss": 1.0264, "step": 57440 }, { "epoch": 0.14, "learning_rate": 8.580074487895718e-05, "loss": 1.0261, "step": 57445 }, { "epoch": 0.14, "learning_rate": 8.579948663747547e-05, "loss": 1.0286, "step": 57450 }, { "epoch": 0.14, "learning_rate": 8.579822839599376e-05, "loss": 1.0492, "step": 57455 }, { "epoch": 0.14, "learning_rate": 8.579697015451205e-05, "loss": 1.0257, "step": 57460 }, { "epoch": 0.14, "learning_rate": 8.579571191303036e-05, "loss": 1.0268, "step": 57465 }, { "epoch": 0.14, "learning_rate": 8.579445367154865e-05, "loss": 1.029, "step": 57470 }, { "epoch": 0.14, "learning_rate": 8.579319543006694e-05, "loss": 1.0266, "step": 57475 }, { "epoch": 0.14, "learning_rate": 8.579193718858523e-05, "loss": 1.0285, "step": 57480 }, { "epoch": 0.14, "learning_rate": 8.579067894710353e-05, "loss": 1.0283, "step": 57485 }, { "epoch": 0.14, "learning_rate": 8.578942070562183e-05, "loss": 1.0275, "step": 57490 }, { "epoch": 0.14, "learning_rate": 8.578816246414012e-05, "loss": 1.0264, "step": 57495 }, { "epoch": 0.14, "learning_rate": 8.578690422265841e-05, "loss": 1.0273, "step": 57500 }, { "epoch": 0.14, "learning_rate": 8.57856459811767e-05, "loss": 1.0263, "step": 57505 }, { "epoch": 0.14, "learning_rate": 8.578438773969501e-05, "loss": 1.028, "step": 57510 }, { "epoch": 0.14, "learning_rate": 8.57831294982133e-05, "loss": 1.0266, "step": 57515 }, { "epoch": 0.14, "learning_rate": 8.57818712567316e-05, "loss": 1.0254, "step": 57520 }, { "epoch": 0.14, "learning_rate": 8.578061301524989e-05, "loss": 1.0278, "step": 57525 }, { "epoch": 0.14, "learning_rate": 8.577935477376819e-05, "loss": 1.0285, "step": 57530 }, { "epoch": 0.14, "learning_rate": 8.577809653228648e-05, "loss": 1.0285, "step": 57535 }, { "epoch": 0.14, "learning_rate": 8.577683829080477e-05, "loss": 1.0259, "step": 57540 }, { "epoch": 0.14, "learning_rate": 8.577558004932307e-05, "loss": 1.0266, "step": 57545 }, { "epoch": 0.14, "learning_rate": 8.577432180784136e-05, "loss": 1.0267, "step": 57550 }, { "epoch": 0.14, "learning_rate": 8.577306356635966e-05, "loss": 1.0281, "step": 57555 }, { "epoch": 0.14, "learning_rate": 8.577180532487795e-05, "loss": 1.0299, "step": 57560 }, { "epoch": 0.14, "learning_rate": 8.577054708339625e-05, "loss": 1.0256, "step": 57565 }, { "epoch": 0.14, "learning_rate": 8.576928884191454e-05, "loss": 1.0282, "step": 57570 }, { "epoch": 0.14, "learning_rate": 8.576803060043284e-05, "loss": 1.028, "step": 57575 }, { "epoch": 0.14, "learning_rate": 8.576677235895113e-05, "loss": 1.0262, "step": 57580 }, { "epoch": 0.14, "learning_rate": 8.576551411746942e-05, "loss": 1.0268, "step": 57585 }, { "epoch": 0.14, "learning_rate": 8.576425587598772e-05, "loss": 1.0252, "step": 57590 }, { "epoch": 0.14, "learning_rate": 8.576299763450601e-05, "loss": 1.0288, "step": 57595 }, { "epoch": 0.14, "learning_rate": 8.576173939302431e-05, "loss": 1.026, "step": 57600 }, { "epoch": 0.14, "learning_rate": 8.57604811515426e-05, "loss": 1.0271, "step": 57605 }, { "epoch": 0.14, "learning_rate": 8.57592229100609e-05, "loss": 1.0259, "step": 57610 }, { "epoch": 0.14, "learning_rate": 8.575796466857919e-05, "loss": 1.0269, "step": 57615 }, { "epoch": 0.14, "learning_rate": 8.57567064270975e-05, "loss": 1.0282, "step": 57620 }, { "epoch": 0.14, "learning_rate": 8.575544818561578e-05, "loss": 1.026, "step": 57625 }, { "epoch": 0.14, "learning_rate": 8.575418994413408e-05, "loss": 1.0297, "step": 57630 }, { "epoch": 0.14, "learning_rate": 8.575293170265237e-05, "loss": 1.0275, "step": 57635 }, { "epoch": 0.14, "learning_rate": 8.575167346117067e-05, "loss": 1.0282, "step": 57640 }, { "epoch": 0.14, "learning_rate": 8.575041521968896e-05, "loss": 1.0284, "step": 57645 }, { "epoch": 0.14, "learning_rate": 8.574915697820726e-05, "loss": 1.0256, "step": 57650 }, { "epoch": 0.14, "learning_rate": 8.574789873672555e-05, "loss": 1.0292, "step": 57655 }, { "epoch": 0.14, "learning_rate": 8.574664049524384e-05, "loss": 1.027, "step": 57660 }, { "epoch": 0.14, "learning_rate": 8.574538225376214e-05, "loss": 1.0243, "step": 57665 }, { "epoch": 0.14, "learning_rate": 8.574412401228044e-05, "loss": 1.025, "step": 57670 }, { "epoch": 0.14, "learning_rate": 8.574286577079873e-05, "loss": 1.0258, "step": 57675 }, { "epoch": 0.14, "learning_rate": 8.574160752931702e-05, "loss": 1.0273, "step": 57680 }, { "epoch": 0.14, "learning_rate": 8.574034928783532e-05, "loss": 1.0264, "step": 57685 }, { "epoch": 0.14, "learning_rate": 8.573909104635363e-05, "loss": 1.0275, "step": 57690 }, { "epoch": 0.14, "learning_rate": 8.573783280487192e-05, "loss": 1.0255, "step": 57695 }, { "epoch": 0.14, "learning_rate": 8.573657456339021e-05, "loss": 1.0268, "step": 57700 }, { "epoch": 0.14, "learning_rate": 8.57353163219085e-05, "loss": 1.0263, "step": 57705 }, { "epoch": 0.14, "learning_rate": 8.573405808042681e-05, "loss": 1.028, "step": 57710 }, { "epoch": 0.14, "learning_rate": 8.57327998389451e-05, "loss": 1.028, "step": 57715 }, { "epoch": 0.14, "learning_rate": 8.573154159746339e-05, "loss": 1.0343, "step": 57720 }, { "epoch": 0.14, "learning_rate": 8.573028335598168e-05, "loss": 1.0272, "step": 57725 }, { "epoch": 0.14, "learning_rate": 8.572902511449999e-05, "loss": 1.0279, "step": 57730 }, { "epoch": 0.14, "learning_rate": 8.572776687301828e-05, "loss": 1.0257, "step": 57735 }, { "epoch": 0.14, "learning_rate": 8.572650863153657e-05, "loss": 1.0271, "step": 57740 }, { "epoch": 0.14, "learning_rate": 8.572525039005486e-05, "loss": 1.0287, "step": 57745 }, { "epoch": 0.14, "learning_rate": 8.572399214857316e-05, "loss": 1.0284, "step": 57750 }, { "epoch": 0.14, "learning_rate": 8.572273390709146e-05, "loss": 1.0269, "step": 57755 }, { "epoch": 0.14, "learning_rate": 8.572147566560975e-05, "loss": 1.0254, "step": 57760 }, { "epoch": 0.14, "learning_rate": 8.572021742412804e-05, "loss": 1.0266, "step": 57765 }, { "epoch": 0.15, "learning_rate": 8.571895918264634e-05, "loss": 1.0266, "step": 57770 }, { "epoch": 0.15, "learning_rate": 8.571770094116464e-05, "loss": 1.0267, "step": 57775 }, { "epoch": 0.15, "learning_rate": 8.571644269968293e-05, "loss": 1.0259, "step": 57780 }, { "epoch": 0.15, "learning_rate": 8.571518445820122e-05, "loss": 1.0244, "step": 57785 }, { "epoch": 0.15, "learning_rate": 8.571392621671952e-05, "loss": 1.0271, "step": 57790 }, { "epoch": 0.15, "learning_rate": 8.571266797523782e-05, "loss": 1.0264, "step": 57795 }, { "epoch": 0.15, "learning_rate": 8.571140973375611e-05, "loss": 1.0285, "step": 57800 }, { "epoch": 0.15, "learning_rate": 8.57101514922744e-05, "loss": 1.0273, "step": 57805 }, { "epoch": 0.15, "learning_rate": 8.57088932507927e-05, "loss": 1.029, "step": 57810 }, { "epoch": 0.15, "learning_rate": 8.570763500931099e-05, "loss": 1.0263, "step": 57815 }, { "epoch": 0.15, "learning_rate": 8.570637676782929e-05, "loss": 1.0278, "step": 57820 }, { "epoch": 0.15, "learning_rate": 8.570511852634758e-05, "loss": 1.0253, "step": 57825 }, { "epoch": 0.15, "learning_rate": 8.570386028486588e-05, "loss": 1.0272, "step": 57830 }, { "epoch": 0.15, "learning_rate": 8.570260204338417e-05, "loss": 1.026, "step": 57835 }, { "epoch": 0.15, "learning_rate": 8.570134380190247e-05, "loss": 1.0273, "step": 57840 }, { "epoch": 0.15, "learning_rate": 8.570008556042076e-05, "loss": 1.0256, "step": 57845 }, { "epoch": 0.15, "learning_rate": 8.569882731893906e-05, "loss": 1.0273, "step": 57850 }, { "epoch": 0.15, "learning_rate": 8.569756907745735e-05, "loss": 1.0277, "step": 57855 }, { "epoch": 0.15, "learning_rate": 8.569631083597565e-05, "loss": 1.0271, "step": 57860 }, { "epoch": 0.15, "learning_rate": 8.569505259449394e-05, "loss": 1.0285, "step": 57865 }, { "epoch": 0.15, "learning_rate": 8.569379435301224e-05, "loss": 1.0271, "step": 57870 }, { "epoch": 0.15, "learning_rate": 8.569253611153053e-05, "loss": 1.0265, "step": 57875 }, { "epoch": 0.15, "learning_rate": 8.569127787004882e-05, "loss": 1.0266, "step": 57880 }, { "epoch": 0.15, "learning_rate": 8.569001962856712e-05, "loss": 1.0279, "step": 57885 }, { "epoch": 0.15, "learning_rate": 8.568876138708542e-05, "loss": 1.0298, "step": 57890 }, { "epoch": 0.15, "learning_rate": 8.568750314560371e-05, "loss": 1.0281, "step": 57895 }, { "epoch": 0.15, "learning_rate": 8.5686244904122e-05, "loss": 1.0276, "step": 57900 }, { "epoch": 0.15, "learning_rate": 8.56849866626403e-05, "loss": 1.0275, "step": 57905 }, { "epoch": 0.15, "learning_rate": 8.56837284211586e-05, "loss": 1.026, "step": 57910 }, { "epoch": 0.15, "learning_rate": 8.568247017967689e-05, "loss": 1.0268, "step": 57915 }, { "epoch": 0.15, "learning_rate": 8.568121193819518e-05, "loss": 1.027, "step": 57920 }, { "epoch": 0.15, "learning_rate": 8.567995369671348e-05, "loss": 1.0261, "step": 57925 }, { "epoch": 0.15, "learning_rate": 8.567869545523178e-05, "loss": 1.0269, "step": 57930 }, { "epoch": 0.15, "learning_rate": 8.567743721375007e-05, "loss": 1.0276, "step": 57935 }, { "epoch": 0.15, "learning_rate": 8.567617897226836e-05, "loss": 1.0248, "step": 57940 }, { "epoch": 0.15, "learning_rate": 8.567492073078665e-05, "loss": 1.0281, "step": 57945 }, { "epoch": 0.15, "learning_rate": 8.567366248930496e-05, "loss": 1.0289, "step": 57950 }, { "epoch": 0.15, "learning_rate": 8.567240424782325e-05, "loss": 1.0265, "step": 57955 }, { "epoch": 0.15, "learning_rate": 8.567114600634154e-05, "loss": 1.0261, "step": 57960 }, { "epoch": 0.15, "learning_rate": 8.566988776485983e-05, "loss": 1.0257, "step": 57965 }, { "epoch": 0.15, "learning_rate": 8.566862952337814e-05, "loss": 1.0253, "step": 57970 }, { "epoch": 0.15, "learning_rate": 8.566737128189643e-05, "loss": 1.0275, "step": 57975 }, { "epoch": 0.15, "learning_rate": 8.566611304041472e-05, "loss": 1.0275, "step": 57980 }, { "epoch": 0.15, "learning_rate": 8.566485479893301e-05, "loss": 1.0253, "step": 57985 }, { "epoch": 0.15, "learning_rate": 8.566359655745132e-05, "loss": 1.0299, "step": 57990 }, { "epoch": 0.15, "learning_rate": 8.566233831596961e-05, "loss": 1.0267, "step": 57995 }, { "epoch": 0.15, "learning_rate": 8.56610800744879e-05, "loss": 1.0272, "step": 58000 }, { "epoch": 0.15, "learning_rate": 8.565982183300619e-05, "loss": 1.0255, "step": 58005 }, { "epoch": 0.15, "learning_rate": 8.565856359152448e-05, "loss": 1.0259, "step": 58010 }, { "epoch": 0.15, "learning_rate": 8.565730535004279e-05, "loss": 1.0283, "step": 58015 }, { "epoch": 0.15, "learning_rate": 8.565604710856108e-05, "loss": 1.0282, "step": 58020 }, { "epoch": 0.15, "learning_rate": 8.565478886707937e-05, "loss": 1.0274, "step": 58025 }, { "epoch": 0.15, "learning_rate": 8.565353062559766e-05, "loss": 1.0287, "step": 58030 }, { "epoch": 0.15, "learning_rate": 8.565227238411597e-05, "loss": 1.0283, "step": 58035 }, { "epoch": 0.15, "learning_rate": 8.565101414263426e-05, "loss": 1.0277, "step": 58040 }, { "epoch": 0.15, "learning_rate": 8.564975590115255e-05, "loss": 1.0274, "step": 58045 }, { "epoch": 0.15, "learning_rate": 8.564849765967084e-05, "loss": 1.0261, "step": 58050 }, { "epoch": 0.15, "learning_rate": 8.564723941818915e-05, "loss": 1.0273, "step": 58055 }, { "epoch": 0.15, "learning_rate": 8.564598117670744e-05, "loss": 1.0277, "step": 58060 }, { "epoch": 0.15, "learning_rate": 8.564472293522573e-05, "loss": 1.0476, "step": 58065 }, { "epoch": 0.15, "learning_rate": 8.564346469374402e-05, "loss": 1.027, "step": 58070 }, { "epoch": 0.15, "learning_rate": 8.564220645226231e-05, "loss": 1.0277, "step": 58075 }, { "epoch": 0.15, "learning_rate": 8.564094821078062e-05, "loss": 1.0294, "step": 58080 }, { "epoch": 0.15, "learning_rate": 8.563968996929891e-05, "loss": 1.0285, "step": 58085 }, { "epoch": 0.15, "learning_rate": 8.56384317278172e-05, "loss": 1.0304, "step": 58090 }, { "epoch": 0.15, "learning_rate": 8.563717348633549e-05, "loss": 1.028, "step": 58095 }, { "epoch": 0.15, "learning_rate": 8.56359152448538e-05, "loss": 1.0257, "step": 58100 }, { "epoch": 0.15, "learning_rate": 8.563465700337209e-05, "loss": 1.0263, "step": 58105 }, { "epoch": 0.15, "learning_rate": 8.563339876189038e-05, "loss": 1.0296, "step": 58110 }, { "epoch": 0.15, "learning_rate": 8.563214052040867e-05, "loss": 1.0269, "step": 58115 }, { "epoch": 0.15, "learning_rate": 8.563088227892698e-05, "loss": 1.0275, "step": 58120 }, { "epoch": 0.15, "learning_rate": 8.562962403744527e-05, "loss": 1.0269, "step": 58125 }, { "epoch": 0.15, "learning_rate": 8.562836579596356e-05, "loss": 1.0279, "step": 58130 }, { "epoch": 0.15, "learning_rate": 8.562710755448185e-05, "loss": 1.0271, "step": 58135 }, { "epoch": 0.15, "learning_rate": 8.562584931300014e-05, "loss": 1.0281, "step": 58140 }, { "epoch": 0.15, "learning_rate": 8.562459107151845e-05, "loss": 1.0269, "step": 58145 }, { "epoch": 0.15, "learning_rate": 8.562333283003674e-05, "loss": 1.0286, "step": 58150 }, { "epoch": 0.15, "learning_rate": 8.562207458855503e-05, "loss": 1.0264, "step": 58155 }, { "epoch": 0.15, "learning_rate": 8.562081634707332e-05, "loss": 1.0256, "step": 58160 }, { "epoch": 0.15, "learning_rate": 8.561955810559163e-05, "loss": 1.0285, "step": 58165 }, { "epoch": 0.15, "learning_rate": 8.561829986410992e-05, "loss": 1.0289, "step": 58170 }, { "epoch": 0.15, "learning_rate": 8.561704162262821e-05, "loss": 1.0269, "step": 58175 }, { "epoch": 0.15, "learning_rate": 8.56157833811465e-05, "loss": 1.0262, "step": 58180 }, { "epoch": 0.15, "learning_rate": 8.561452513966481e-05, "loss": 1.029, "step": 58185 }, { "epoch": 0.15, "learning_rate": 8.561326689818311e-05, "loss": 1.0278, "step": 58190 }, { "epoch": 0.15, "learning_rate": 8.56120086567014e-05, "loss": 1.0264, "step": 58195 }, { "epoch": 0.15, "learning_rate": 8.56107504152197e-05, "loss": 1.0279, "step": 58200 }, { "epoch": 0.15, "learning_rate": 8.560949217373799e-05, "loss": 1.0257, "step": 58205 }, { "epoch": 0.15, "learning_rate": 8.560823393225628e-05, "loss": 1.0272, "step": 58210 }, { "epoch": 0.15, "learning_rate": 8.560697569077459e-05, "loss": 1.0289, "step": 58215 }, { "epoch": 0.15, "learning_rate": 8.560571744929288e-05, "loss": 1.0279, "step": 58220 }, { "epoch": 0.15, "learning_rate": 8.560445920781117e-05, "loss": 1.0305, "step": 58225 }, { "epoch": 0.15, "learning_rate": 8.560320096632946e-05, "loss": 1.0275, "step": 58230 }, { "epoch": 0.15, "learning_rate": 8.560194272484777e-05, "loss": 1.027, "step": 58235 }, { "epoch": 0.15, "learning_rate": 8.560068448336606e-05, "loss": 1.0271, "step": 58240 }, { "epoch": 0.15, "learning_rate": 8.559942624188435e-05, "loss": 1.0484, "step": 58245 }, { "epoch": 0.15, "learning_rate": 8.559816800040264e-05, "loss": 1.0281, "step": 58250 }, { "epoch": 0.15, "learning_rate": 8.559690975892095e-05, "loss": 1.0249, "step": 58255 }, { "epoch": 0.15, "learning_rate": 8.559565151743924e-05, "loss": 1.0516, "step": 58260 }, { "epoch": 0.15, "learning_rate": 8.559439327595753e-05, "loss": 1.0252, "step": 58265 }, { "epoch": 0.15, "learning_rate": 8.559313503447582e-05, "loss": 1.0268, "step": 58270 }, { "epoch": 0.15, "learning_rate": 8.559187679299411e-05, "loss": 1.0259, "step": 58275 }, { "epoch": 0.15, "learning_rate": 8.559061855151242e-05, "loss": 1.0272, "step": 58280 }, { "epoch": 0.15, "learning_rate": 8.558936031003071e-05, "loss": 1.03, "step": 58285 }, { "epoch": 0.15, "learning_rate": 8.5588102068549e-05, "loss": 1.0278, "step": 58290 }, { "epoch": 0.15, "learning_rate": 8.558684382706729e-05, "loss": 1.0265, "step": 58295 }, { "epoch": 0.15, "learning_rate": 8.55855855855856e-05, "loss": 1.0262, "step": 58300 }, { "epoch": 0.15, "learning_rate": 8.558432734410389e-05, "loss": 1.0296, "step": 58305 }, { "epoch": 0.15, "learning_rate": 8.558306910262218e-05, "loss": 1.0287, "step": 58310 }, { "epoch": 0.15, "learning_rate": 8.558181086114047e-05, "loss": 1.0271, "step": 58315 }, { "epoch": 0.15, "learning_rate": 8.558055261965878e-05, "loss": 1.027, "step": 58320 }, { "epoch": 0.15, "learning_rate": 8.557929437817707e-05, "loss": 1.0263, "step": 58325 }, { "epoch": 0.15, "learning_rate": 8.557803613669536e-05, "loss": 1.0279, "step": 58330 }, { "epoch": 0.15, "learning_rate": 8.557677789521365e-05, "loss": 1.0293, "step": 58335 }, { "epoch": 0.15, "learning_rate": 8.557551965373194e-05, "loss": 1.027, "step": 58340 }, { "epoch": 0.15, "learning_rate": 8.557426141225025e-05, "loss": 1.0266, "step": 58345 }, { "epoch": 0.15, "learning_rate": 8.557300317076854e-05, "loss": 1.0249, "step": 58350 }, { "epoch": 0.15, "learning_rate": 8.557174492928683e-05, "loss": 1.026, "step": 58355 }, { "epoch": 0.15, "learning_rate": 8.557048668780512e-05, "loss": 1.0265, "step": 58360 }, { "epoch": 0.15, "learning_rate": 8.556922844632343e-05, "loss": 1.0274, "step": 58365 }, { "epoch": 0.15, "learning_rate": 8.556797020484172e-05, "loss": 1.0246, "step": 58370 }, { "epoch": 0.15, "learning_rate": 8.556671196336001e-05, "loss": 1.0257, "step": 58375 }, { "epoch": 0.15, "learning_rate": 8.55654537218783e-05, "loss": 1.0272, "step": 58380 }, { "epoch": 0.15, "learning_rate": 8.556419548039661e-05, "loss": 1.0259, "step": 58385 }, { "epoch": 0.15, "learning_rate": 8.55629372389149e-05, "loss": 1.0266, "step": 58390 }, { "epoch": 0.15, "learning_rate": 8.556167899743319e-05, "loss": 1.0238, "step": 58395 }, { "epoch": 0.15, "learning_rate": 8.556042075595148e-05, "loss": 1.0281, "step": 58400 }, { "epoch": 0.15, "learning_rate": 8.555916251446978e-05, "loss": 1.0253, "step": 58405 }, { "epoch": 0.15, "learning_rate": 8.555790427298808e-05, "loss": 1.0272, "step": 58410 }, { "epoch": 0.15, "learning_rate": 8.555664603150637e-05, "loss": 1.0278, "step": 58415 }, { "epoch": 0.15, "learning_rate": 8.555538779002466e-05, "loss": 1.0257, "step": 58420 }, { "epoch": 0.15, "learning_rate": 8.555412954854296e-05, "loss": 1.0267, "step": 58425 }, { "epoch": 0.15, "learning_rate": 8.555287130706126e-05, "loss": 1.0265, "step": 58430 }, { "epoch": 0.15, "learning_rate": 8.555161306557955e-05, "loss": 1.0268, "step": 58435 }, { "epoch": 0.15, "learning_rate": 8.555035482409784e-05, "loss": 1.0275, "step": 58440 }, { "epoch": 0.15, "learning_rate": 8.554909658261614e-05, "loss": 1.0278, "step": 58445 }, { "epoch": 0.15, "learning_rate": 8.554783834113444e-05, "loss": 1.0264, "step": 58450 }, { "epoch": 0.15, "learning_rate": 8.554658009965273e-05, "loss": 1.0248, "step": 58455 }, { "epoch": 0.15, "learning_rate": 8.554532185817102e-05, "loss": 1.0277, "step": 58460 }, { "epoch": 0.15, "learning_rate": 8.554406361668932e-05, "loss": 1.03, "step": 58465 }, { "epoch": 0.15, "learning_rate": 8.554280537520761e-05, "loss": 1.0272, "step": 58470 }, { "epoch": 0.15, "learning_rate": 8.554154713372591e-05, "loss": 1.0281, "step": 58475 }, { "epoch": 0.15, "learning_rate": 8.55402888922442e-05, "loss": 1.0251, "step": 58480 }, { "epoch": 0.15, "learning_rate": 8.55390306507625e-05, "loss": 1.0248, "step": 58485 }, { "epoch": 0.15, "learning_rate": 8.553777240928079e-05, "loss": 1.0265, "step": 58490 }, { "epoch": 0.15, "learning_rate": 8.553651416779909e-05, "loss": 1.0272, "step": 58495 }, { "epoch": 0.15, "learning_rate": 8.553525592631738e-05, "loss": 1.0249, "step": 58500 }, { "epoch": 0.15, "learning_rate": 8.553399768483568e-05, "loss": 1.028, "step": 58505 }, { "epoch": 0.15, "learning_rate": 8.553273944335397e-05, "loss": 1.0273, "step": 58510 }, { "epoch": 0.15, "learning_rate": 8.553148120187227e-05, "loss": 1.0251, "step": 58515 }, { "epoch": 0.15, "learning_rate": 8.553022296039056e-05, "loss": 1.0265, "step": 58520 }, { "epoch": 0.15, "learning_rate": 8.552896471890886e-05, "loss": 1.0303, "step": 58525 }, { "epoch": 0.15, "learning_rate": 8.552770647742715e-05, "loss": 1.0254, "step": 58530 }, { "epoch": 0.15, "learning_rate": 8.552644823594544e-05, "loss": 1.0288, "step": 58535 }, { "epoch": 0.15, "learning_rate": 8.552518999446374e-05, "loss": 1.0267, "step": 58540 }, { "epoch": 0.15, "learning_rate": 8.552393175298204e-05, "loss": 1.0246, "step": 58545 }, { "epoch": 0.15, "learning_rate": 8.552267351150033e-05, "loss": 1.0273, "step": 58550 }, { "epoch": 0.15, "learning_rate": 8.552141527001862e-05, "loss": 1.0283, "step": 58555 }, { "epoch": 0.15, "learning_rate": 8.552015702853692e-05, "loss": 1.0278, "step": 58560 }, { "epoch": 0.15, "learning_rate": 8.551889878705522e-05, "loss": 1.028, "step": 58565 }, { "epoch": 0.15, "learning_rate": 8.55176405455735e-05, "loss": 1.0268, "step": 58570 }, { "epoch": 0.15, "learning_rate": 8.55163823040918e-05, "loss": 1.0284, "step": 58575 }, { "epoch": 0.15, "learning_rate": 8.55151240626101e-05, "loss": 1.0278, "step": 58580 }, { "epoch": 0.15, "learning_rate": 8.55138658211284e-05, "loss": 1.0253, "step": 58585 }, { "epoch": 0.15, "learning_rate": 8.551260757964669e-05, "loss": 1.0278, "step": 58590 }, { "epoch": 0.15, "learning_rate": 8.551134933816498e-05, "loss": 1.0259, "step": 58595 }, { "epoch": 0.15, "learning_rate": 8.551009109668327e-05, "loss": 1.028, "step": 58600 }, { "epoch": 0.15, "learning_rate": 8.550883285520157e-05, "loss": 1.0251, "step": 58605 }, { "epoch": 0.15, "learning_rate": 8.550757461371987e-05, "loss": 1.0265, "step": 58610 }, { "epoch": 0.15, "learning_rate": 8.550631637223816e-05, "loss": 1.0272, "step": 58615 }, { "epoch": 0.15, "learning_rate": 8.550505813075645e-05, "loss": 1.0273, "step": 58620 }, { "epoch": 0.15, "learning_rate": 8.550379988927475e-05, "loss": 1.0278, "step": 58625 }, { "epoch": 0.15, "learning_rate": 8.550254164779305e-05, "loss": 1.025, "step": 58630 }, { "epoch": 0.15, "learning_rate": 8.550128340631134e-05, "loss": 1.0269, "step": 58635 }, { "epoch": 0.15, "learning_rate": 8.550002516482963e-05, "loss": 1.0267, "step": 58640 }, { "epoch": 0.15, "learning_rate": 8.549876692334793e-05, "loss": 1.027, "step": 58645 }, { "epoch": 0.15, "learning_rate": 8.549750868186623e-05, "loss": 1.0285, "step": 58650 }, { "epoch": 0.15, "learning_rate": 8.549625044038452e-05, "loss": 1.0264, "step": 58655 }, { "epoch": 0.15, "learning_rate": 8.549499219890281e-05, "loss": 1.0269, "step": 58660 }, { "epoch": 0.15, "learning_rate": 8.54937339574211e-05, "loss": 1.0269, "step": 58665 }, { "epoch": 0.15, "learning_rate": 8.54924757159394e-05, "loss": 1.0238, "step": 58670 }, { "epoch": 0.15, "learning_rate": 8.54912174744577e-05, "loss": 1.026, "step": 58675 }, { "epoch": 0.15, "learning_rate": 8.548995923297599e-05, "loss": 1.0271, "step": 58680 }, { "epoch": 0.15, "learning_rate": 8.54887009914943e-05, "loss": 1.0258, "step": 58685 }, { "epoch": 0.15, "learning_rate": 8.548744275001259e-05, "loss": 1.0283, "step": 58690 }, { "epoch": 0.15, "learning_rate": 8.548618450853089e-05, "loss": 1.0266, "step": 58695 }, { "epoch": 0.15, "learning_rate": 8.548492626704918e-05, "loss": 1.0282, "step": 58700 }, { "epoch": 0.15, "learning_rate": 8.548366802556747e-05, "loss": 1.0241, "step": 58705 }, { "epoch": 0.15, "learning_rate": 8.548240978408577e-05, "loss": 1.0263, "step": 58710 }, { "epoch": 0.15, "learning_rate": 8.548115154260407e-05, "loss": 1.0274, "step": 58715 }, { "epoch": 0.15, "learning_rate": 8.547989330112236e-05, "loss": 1.0285, "step": 58720 }, { "epoch": 0.15, "learning_rate": 8.547863505964065e-05, "loss": 1.0244, "step": 58725 }, { "epoch": 0.15, "learning_rate": 8.547737681815895e-05, "loss": 1.0275, "step": 58730 }, { "epoch": 0.15, "learning_rate": 8.547611857667724e-05, "loss": 1.0261, "step": 58735 }, { "epoch": 0.15, "learning_rate": 8.547486033519554e-05, "loss": 1.0263, "step": 58740 }, { "epoch": 0.15, "learning_rate": 8.547360209371383e-05, "loss": 1.0271, "step": 58745 }, { "epoch": 0.15, "learning_rate": 8.547234385223213e-05, "loss": 1.0263, "step": 58750 }, { "epoch": 0.15, "learning_rate": 8.547108561075042e-05, "loss": 1.0263, "step": 58755 }, { "epoch": 0.15, "learning_rate": 8.546982736926872e-05, "loss": 1.0272, "step": 58760 }, { "epoch": 0.15, "learning_rate": 8.546856912778701e-05, "loss": 1.027, "step": 58765 }, { "epoch": 0.15, "learning_rate": 8.54673108863053e-05, "loss": 1.0278, "step": 58770 }, { "epoch": 0.15, "learning_rate": 8.54660526448236e-05, "loss": 1.0246, "step": 58775 }, { "epoch": 0.15, "learning_rate": 8.54647944033419e-05, "loss": 1.0283, "step": 58780 }, { "epoch": 0.15, "learning_rate": 8.54635361618602e-05, "loss": 1.0242, "step": 58785 }, { "epoch": 0.15, "learning_rate": 8.546227792037849e-05, "loss": 1.0282, "step": 58790 }, { "epoch": 0.15, "learning_rate": 8.546101967889678e-05, "loss": 1.025, "step": 58795 }, { "epoch": 0.15, "learning_rate": 8.545976143741507e-05, "loss": 1.0271, "step": 58800 }, { "epoch": 0.15, "learning_rate": 8.545850319593337e-05, "loss": 1.0274, "step": 58805 }, { "epoch": 0.15, "learning_rate": 8.545724495445167e-05, "loss": 1.027, "step": 58810 }, { "epoch": 0.15, "learning_rate": 8.545598671296996e-05, "loss": 1.0261, "step": 58815 }, { "epoch": 0.15, "learning_rate": 8.545472847148825e-05, "loss": 1.0255, "step": 58820 }, { "epoch": 0.15, "learning_rate": 8.545347023000655e-05, "loss": 1.0271, "step": 58825 }, { "epoch": 0.15, "learning_rate": 8.545221198852485e-05, "loss": 1.0263, "step": 58830 }, { "epoch": 0.15, "learning_rate": 8.545095374704314e-05, "loss": 1.0286, "step": 58835 }, { "epoch": 0.15, "learning_rate": 8.544969550556143e-05, "loss": 1.0276, "step": 58840 }, { "epoch": 0.15, "learning_rate": 8.544843726407973e-05, "loss": 1.0278, "step": 58845 }, { "epoch": 0.15, "learning_rate": 8.544717902259803e-05, "loss": 1.0286, "step": 58850 }, { "epoch": 0.15, "learning_rate": 8.544592078111632e-05, "loss": 1.0278, "step": 58855 }, { "epoch": 0.15, "learning_rate": 8.544466253963461e-05, "loss": 1.0278, "step": 58860 }, { "epoch": 0.15, "learning_rate": 8.54434042981529e-05, "loss": 1.0256, "step": 58865 }, { "epoch": 0.15, "learning_rate": 8.54421460566712e-05, "loss": 1.0264, "step": 58870 }, { "epoch": 0.15, "learning_rate": 8.54408878151895e-05, "loss": 1.0263, "step": 58875 }, { "epoch": 0.15, "learning_rate": 8.543962957370779e-05, "loss": 1.0285, "step": 58880 }, { "epoch": 0.15, "learning_rate": 8.543837133222608e-05, "loss": 1.0249, "step": 58885 }, { "epoch": 0.15, "learning_rate": 8.543711309074439e-05, "loss": 1.0277, "step": 58890 }, { "epoch": 0.15, "learning_rate": 8.543585484926268e-05, "loss": 1.0261, "step": 58895 }, { "epoch": 0.15, "learning_rate": 8.543459660778097e-05, "loss": 1.0273, "step": 58900 }, { "epoch": 0.15, "learning_rate": 8.543333836629926e-05, "loss": 1.0254, "step": 58905 }, { "epoch": 0.15, "learning_rate": 8.543208012481757e-05, "loss": 1.0277, "step": 58910 }, { "epoch": 0.15, "learning_rate": 8.543082188333586e-05, "loss": 1.0268, "step": 58915 }, { "epoch": 0.15, "learning_rate": 8.542956364185415e-05, "loss": 1.0278, "step": 58920 }, { "epoch": 0.15, "learning_rate": 8.542830540037244e-05, "loss": 1.0268, "step": 58925 }, { "epoch": 0.15, "learning_rate": 8.542704715889073e-05, "loss": 1.0267, "step": 58930 }, { "epoch": 0.15, "learning_rate": 8.542578891740904e-05, "loss": 1.0259, "step": 58935 }, { "epoch": 0.15, "learning_rate": 8.542453067592733e-05, "loss": 1.0269, "step": 58940 }, { "epoch": 0.15, "learning_rate": 8.542327243444562e-05, "loss": 1.0283, "step": 58945 }, { "epoch": 0.15, "learning_rate": 8.542201419296391e-05, "loss": 1.0255, "step": 58950 }, { "epoch": 0.15, "learning_rate": 8.542075595148222e-05, "loss": 1.0272, "step": 58955 }, { "epoch": 0.15, "learning_rate": 8.541949771000051e-05, "loss": 1.0265, "step": 58960 }, { "epoch": 0.15, "learning_rate": 8.54182394685188e-05, "loss": 1.0467, "step": 58965 }, { "epoch": 0.15, "learning_rate": 8.541698122703709e-05, "loss": 1.0269, "step": 58970 }, { "epoch": 0.15, "learning_rate": 8.54157229855554e-05, "loss": 1.0297, "step": 58975 }, { "epoch": 0.15, "learning_rate": 8.541446474407369e-05, "loss": 1.0259, "step": 58980 }, { "epoch": 0.15, "learning_rate": 8.541320650259198e-05, "loss": 1.0251, "step": 58985 }, { "epoch": 0.15, "learning_rate": 8.541194826111027e-05, "loss": 1.027, "step": 58990 }, { "epoch": 0.15, "learning_rate": 8.541069001962856e-05, "loss": 1.0273, "step": 58995 }, { "epoch": 0.15, "learning_rate": 8.540943177814687e-05, "loss": 1.0285, "step": 59000 }, { "epoch": 0.15, "learning_rate": 8.540817353666516e-05, "loss": 1.0246, "step": 59005 }, { "epoch": 0.15, "learning_rate": 8.540691529518345e-05, "loss": 1.0266, "step": 59010 }, { "epoch": 0.15, "learning_rate": 8.540565705370174e-05, "loss": 1.0265, "step": 59015 }, { "epoch": 0.15, "learning_rate": 8.540439881222005e-05, "loss": 1.0263, "step": 59020 }, { "epoch": 0.15, "learning_rate": 8.540314057073834e-05, "loss": 1.0236, "step": 59025 }, { "epoch": 0.15, "learning_rate": 8.540188232925663e-05, "loss": 1.0247, "step": 59030 }, { "epoch": 0.15, "learning_rate": 8.540062408777492e-05, "loss": 1.0252, "step": 59035 }, { "epoch": 0.15, "learning_rate": 8.539936584629323e-05, "loss": 1.025, "step": 59040 }, { "epoch": 0.15, "learning_rate": 8.539810760481152e-05, "loss": 1.0241, "step": 59045 }, { "epoch": 0.15, "learning_rate": 8.539684936332981e-05, "loss": 1.0276, "step": 59050 }, { "epoch": 0.15, "learning_rate": 8.53955911218481e-05, "loss": 1.0523, "step": 59055 }, { "epoch": 0.15, "learning_rate": 8.53943328803664e-05, "loss": 1.0265, "step": 59060 }, { "epoch": 0.15, "learning_rate": 8.53930746388847e-05, "loss": 1.0285, "step": 59065 }, { "epoch": 0.15, "learning_rate": 8.539181639740299e-05, "loss": 1.0241, "step": 59070 }, { "epoch": 0.15, "learning_rate": 8.539055815592128e-05, "loss": 1.0257, "step": 59075 }, { "epoch": 0.15, "learning_rate": 8.538929991443957e-05, "loss": 1.0244, "step": 59080 }, { "epoch": 0.15, "learning_rate": 8.538804167295788e-05, "loss": 1.0282, "step": 59085 }, { "epoch": 0.15, "learning_rate": 8.538678343147617e-05, "loss": 1.0275, "step": 59090 }, { "epoch": 0.15, "learning_rate": 8.538552518999446e-05, "loss": 1.0267, "step": 59095 }, { "epoch": 0.15, "learning_rate": 8.538426694851275e-05, "loss": 1.0273, "step": 59100 }, { "epoch": 0.15, "learning_rate": 8.538300870703106e-05, "loss": 1.0284, "step": 59105 }, { "epoch": 0.15, "learning_rate": 8.538175046554935e-05, "loss": 1.0293, "step": 59110 }, { "epoch": 0.15, "learning_rate": 8.538049222406764e-05, "loss": 1.0253, "step": 59115 }, { "epoch": 0.15, "learning_rate": 8.537923398258593e-05, "loss": 1.0293, "step": 59120 }, { "epoch": 0.15, "learning_rate": 8.537797574110423e-05, "loss": 1.0267, "step": 59125 }, { "epoch": 0.15, "learning_rate": 8.537671749962253e-05, "loss": 1.025, "step": 59130 }, { "epoch": 0.15, "learning_rate": 8.537545925814082e-05, "loss": 1.0274, "step": 59135 }, { "epoch": 0.15, "learning_rate": 8.537420101665911e-05, "loss": 1.0279, "step": 59140 }, { "epoch": 0.15, "learning_rate": 8.53729427751774e-05, "loss": 1.0263, "step": 59145 }, { "epoch": 0.15, "learning_rate": 8.537168453369571e-05, "loss": 1.0246, "step": 59150 }, { "epoch": 0.15, "learning_rate": 8.5370426292214e-05, "loss": 1.0261, "step": 59155 }, { "epoch": 0.15, "learning_rate": 8.53691680507323e-05, "loss": 1.0417, "step": 59160 }, { "epoch": 0.15, "learning_rate": 8.536790980925059e-05, "loss": 1.0249, "step": 59165 }, { "epoch": 0.15, "learning_rate": 8.536665156776889e-05, "loss": 1.0275, "step": 59170 }, { "epoch": 0.15, "learning_rate": 8.536539332628718e-05, "loss": 1.0276, "step": 59175 }, { "epoch": 0.15, "learning_rate": 8.536413508480549e-05, "loss": 1.0264, "step": 59180 }, { "epoch": 0.15, "learning_rate": 8.536287684332378e-05, "loss": 1.0262, "step": 59185 }, { "epoch": 0.15, "learning_rate": 8.536161860184207e-05, "loss": 1.0279, "step": 59190 }, { "epoch": 0.15, "learning_rate": 8.536036036036036e-05, "loss": 1.0262, "step": 59195 }, { "epoch": 0.15, "learning_rate": 8.535910211887867e-05, "loss": 1.0274, "step": 59200 }, { "epoch": 0.15, "learning_rate": 8.535784387739696e-05, "loss": 1.0278, "step": 59205 }, { "epoch": 0.15, "learning_rate": 8.535658563591525e-05, "loss": 1.0268, "step": 59210 }, { "epoch": 0.15, "learning_rate": 8.535532739443354e-05, "loss": 1.0262, "step": 59215 }, { "epoch": 0.15, "learning_rate": 8.535406915295185e-05, "loss": 1.0264, "step": 59220 }, { "epoch": 0.15, "learning_rate": 8.535281091147014e-05, "loss": 1.0286, "step": 59225 }, { "epoch": 0.15, "learning_rate": 8.535155266998843e-05, "loss": 1.0243, "step": 59230 }, { "epoch": 0.15, "learning_rate": 8.535029442850672e-05, "loss": 1.0289, "step": 59235 }, { "epoch": 0.15, "learning_rate": 8.534928783532136e-05, "loss": 1.0271, "step": 59240 }, { "epoch": 0.15, "learning_rate": 8.534802959383965e-05, "loss": 1.0287, "step": 59245 }, { "epoch": 0.15, "learning_rate": 8.534677135235794e-05, "loss": 1.0267, "step": 59250 }, { "epoch": 0.15, "learning_rate": 8.534551311087625e-05, "loss": 1.0265, "step": 59255 }, { "epoch": 0.15, "learning_rate": 8.534425486939454e-05, "loss": 1.0274, "step": 59260 }, { "epoch": 0.15, "learning_rate": 8.534299662791283e-05, "loss": 1.0266, "step": 59265 }, { "epoch": 0.15, "learning_rate": 8.534173838643112e-05, "loss": 1.0254, "step": 59270 }, { "epoch": 0.15, "learning_rate": 8.534048014494941e-05, "loss": 1.026, "step": 59275 }, { "epoch": 0.15, "learning_rate": 8.533922190346772e-05, "loss": 1.0283, "step": 59280 }, { "epoch": 0.15, "learning_rate": 8.533796366198601e-05, "loss": 1.0273, "step": 59285 }, { "epoch": 0.15, "learning_rate": 8.53367054205043e-05, "loss": 1.0253, "step": 59290 }, { "epoch": 0.15, "learning_rate": 8.53354471790226e-05, "loss": 1.0283, "step": 59295 }, { "epoch": 0.15, "learning_rate": 8.53341889375409e-05, "loss": 1.0256, "step": 59300 }, { "epoch": 0.15, "learning_rate": 8.533293069605919e-05, "loss": 1.0249, "step": 59305 }, { "epoch": 0.15, "learning_rate": 8.533167245457748e-05, "loss": 1.0271, "step": 59310 }, { "epoch": 0.15, "learning_rate": 8.533041421309577e-05, "loss": 1.0283, "step": 59315 }, { "epoch": 0.15, "learning_rate": 8.532915597161408e-05, "loss": 1.0255, "step": 59320 }, { "epoch": 0.15, "learning_rate": 8.532789773013237e-05, "loss": 1.0288, "step": 59325 }, { "epoch": 0.15, "learning_rate": 8.532663948865066e-05, "loss": 1.0255, "step": 59330 }, { "epoch": 0.15, "learning_rate": 8.532538124716895e-05, "loss": 1.0235, "step": 59335 }, { "epoch": 0.15, "learning_rate": 8.532412300568724e-05, "loss": 1.0265, "step": 59340 }, { "epoch": 0.15, "learning_rate": 8.532286476420555e-05, "loss": 1.0276, "step": 59345 }, { "epoch": 0.15, "learning_rate": 8.532160652272384e-05, "loss": 1.0252, "step": 59350 }, { "epoch": 0.15, "learning_rate": 8.532034828124213e-05, "loss": 1.0285, "step": 59355 }, { "epoch": 0.15, "learning_rate": 8.531909003976042e-05, "loss": 1.0233, "step": 59360 }, { "epoch": 0.15, "learning_rate": 8.531783179827873e-05, "loss": 1.0262, "step": 59365 }, { "epoch": 0.15, "learning_rate": 8.531657355679702e-05, "loss": 1.0263, "step": 59370 }, { "epoch": 0.15, "learning_rate": 8.531531531531531e-05, "loss": 1.0271, "step": 59375 }, { "epoch": 0.15, "learning_rate": 8.531405707383362e-05, "loss": 1.0252, "step": 59380 }, { "epoch": 0.15, "learning_rate": 8.531279883235191e-05, "loss": 1.0289, "step": 59385 }, { "epoch": 0.15, "learning_rate": 8.531154059087022e-05, "loss": 1.0273, "step": 59390 }, { "epoch": 0.15, "learning_rate": 8.53102823493885e-05, "loss": 1.0247, "step": 59395 }, { "epoch": 0.15, "learning_rate": 8.53090241079068e-05, "loss": 1.0271, "step": 59400 }, { "epoch": 0.15, "learning_rate": 8.530776586642509e-05, "loss": 1.0262, "step": 59405 }, { "epoch": 0.15, "learning_rate": 8.530650762494338e-05, "loss": 1.0289, "step": 59410 }, { "epoch": 0.15, "learning_rate": 8.530524938346169e-05, "loss": 1.0267, "step": 59415 }, { "epoch": 0.15, "learning_rate": 8.530399114197998e-05, "loss": 1.0264, "step": 59420 }, { "epoch": 0.15, "learning_rate": 8.530273290049827e-05, "loss": 1.0255, "step": 59425 }, { "epoch": 0.15, "learning_rate": 8.530147465901656e-05, "loss": 1.0264, "step": 59430 }, { "epoch": 0.15, "learning_rate": 8.530021641753487e-05, "loss": 1.0264, "step": 59435 }, { "epoch": 0.15, "learning_rate": 8.529895817605316e-05, "loss": 1.0285, "step": 59440 }, { "epoch": 0.15, "learning_rate": 8.529769993457145e-05, "loss": 1.0262, "step": 59445 }, { "epoch": 0.15, "learning_rate": 8.529644169308974e-05, "loss": 1.0281, "step": 59450 }, { "epoch": 0.15, "learning_rate": 8.529518345160805e-05, "loss": 1.0284, "step": 59455 }, { "epoch": 0.15, "learning_rate": 8.529392521012634e-05, "loss": 1.0281, "step": 59460 }, { "epoch": 0.15, "learning_rate": 8.529266696864463e-05, "loss": 1.0256, "step": 59465 }, { "epoch": 0.15, "learning_rate": 8.529140872716292e-05, "loss": 1.0275, "step": 59470 }, { "epoch": 0.15, "learning_rate": 8.529015048568121e-05, "loss": 1.0267, "step": 59475 }, { "epoch": 0.15, "learning_rate": 8.528889224419952e-05, "loss": 1.0251, "step": 59480 }, { "epoch": 0.15, "learning_rate": 8.528763400271781e-05, "loss": 1.0254, "step": 59485 }, { "epoch": 0.15, "learning_rate": 8.52863757612361e-05, "loss": 1.0257, "step": 59490 }, { "epoch": 0.15, "learning_rate": 8.528511751975439e-05, "loss": 1.0269, "step": 59495 }, { "epoch": 0.15, "learning_rate": 8.52838592782727e-05, "loss": 1.0265, "step": 59500 }, { "epoch": 0.15, "learning_rate": 8.528260103679099e-05, "loss": 1.0245, "step": 59505 }, { "epoch": 0.15, "learning_rate": 8.528134279530928e-05, "loss": 1.0254, "step": 59510 }, { "epoch": 0.15, "learning_rate": 8.528008455382757e-05, "loss": 1.0265, "step": 59515 }, { "epoch": 0.15, "learning_rate": 8.527882631234588e-05, "loss": 1.0262, "step": 59520 }, { "epoch": 0.15, "learning_rate": 8.527756807086417e-05, "loss": 1.0287, "step": 59525 }, { "epoch": 0.15, "learning_rate": 8.527630982938246e-05, "loss": 1.0275, "step": 59530 }, { "epoch": 0.15, "learning_rate": 8.527505158790075e-05, "loss": 1.0286, "step": 59535 }, { "epoch": 0.15, "learning_rate": 8.527379334641904e-05, "loss": 1.0286, "step": 59540 }, { "epoch": 0.15, "learning_rate": 8.527253510493735e-05, "loss": 1.0265, "step": 59545 }, { "epoch": 0.15, "learning_rate": 8.527127686345564e-05, "loss": 1.0246, "step": 59550 }, { "epoch": 0.15, "learning_rate": 8.527001862197393e-05, "loss": 1.0252, "step": 59555 }, { "epoch": 0.15, "learning_rate": 8.526876038049222e-05, "loss": 1.0269, "step": 59560 }, { "epoch": 0.15, "learning_rate": 8.526750213901053e-05, "loss": 1.0256, "step": 59565 }, { "epoch": 0.15, "learning_rate": 8.526624389752882e-05, "loss": 1.0258, "step": 59570 }, { "epoch": 0.15, "learning_rate": 8.526498565604711e-05, "loss": 1.0254, "step": 59575 }, { "epoch": 0.15, "learning_rate": 8.52637274145654e-05, "loss": 1.0261, "step": 59580 }, { "epoch": 0.15, "learning_rate": 8.526246917308371e-05, "loss": 1.0256, "step": 59585 }, { "epoch": 0.15, "learning_rate": 8.5261210931602e-05, "loss": 1.0503, "step": 59590 }, { "epoch": 0.15, "learning_rate": 8.525995269012029e-05, "loss": 1.0268, "step": 59595 }, { "epoch": 0.15, "learning_rate": 8.525869444863858e-05, "loss": 1.0269, "step": 59600 }, { "epoch": 0.15, "learning_rate": 8.525743620715688e-05, "loss": 1.0268, "step": 59605 }, { "epoch": 0.15, "learning_rate": 8.525617796567518e-05, "loss": 1.0296, "step": 59610 }, { "epoch": 0.15, "learning_rate": 8.525491972419347e-05, "loss": 1.0271, "step": 59615 }, { "epoch": 0.15, "learning_rate": 8.525366148271176e-05, "loss": 1.0266, "step": 59620 }, { "epoch": 0.15, "learning_rate": 8.525240324123006e-05, "loss": 1.0283, "step": 59625 }, { "epoch": 0.15, "learning_rate": 8.525114499974836e-05, "loss": 1.027, "step": 59630 }, { "epoch": 0.15, "learning_rate": 8.524988675826665e-05, "loss": 1.0265, "step": 59635 }, { "epoch": 0.15, "learning_rate": 8.524862851678494e-05, "loss": 1.0258, "step": 59640 }, { "epoch": 0.15, "learning_rate": 8.524737027530324e-05, "loss": 1.0269, "step": 59645 }, { "epoch": 0.15, "learning_rate": 8.524611203382154e-05, "loss": 1.0255, "step": 59650 }, { "epoch": 0.15, "learning_rate": 8.524485379233983e-05, "loss": 1.0298, "step": 59655 }, { "epoch": 0.15, "learning_rate": 8.524359555085812e-05, "loss": 1.0256, "step": 59660 }, { "epoch": 0.15, "learning_rate": 8.524233730937642e-05, "loss": 1.0255, "step": 59665 }, { "epoch": 0.15, "learning_rate": 8.524107906789471e-05, "loss": 1.0267, "step": 59670 }, { "epoch": 0.15, "learning_rate": 8.523982082641301e-05, "loss": 1.0258, "step": 59675 }, { "epoch": 0.15, "learning_rate": 8.52385625849313e-05, "loss": 1.0268, "step": 59680 }, { "epoch": 0.15, "learning_rate": 8.52373043434496e-05, "loss": 1.0269, "step": 59685 }, { "epoch": 0.15, "learning_rate": 8.523604610196789e-05, "loss": 1.0251, "step": 59690 }, { "epoch": 0.15, "learning_rate": 8.523478786048619e-05, "loss": 1.0248, "step": 59695 }, { "epoch": 0.15, "learning_rate": 8.523352961900448e-05, "loss": 1.0279, "step": 59700 }, { "epoch": 0.15, "learning_rate": 8.523227137752278e-05, "loss": 1.0295, "step": 59705 }, { "epoch": 0.15, "learning_rate": 8.523101313604107e-05, "loss": 1.0285, "step": 59710 }, { "epoch": 0.15, "learning_rate": 8.522975489455937e-05, "loss": 1.0283, "step": 59715 }, { "epoch": 0.15, "learning_rate": 8.522849665307766e-05, "loss": 1.0278, "step": 59720 }, { "epoch": 0.15, "learning_rate": 8.522723841159596e-05, "loss": 1.0284, "step": 59725 }, { "epoch": 0.15, "learning_rate": 8.522648346670694e-05, "loss": 1.026, "step": 59730 }, { "epoch": 0.15, "learning_rate": 8.522522522522523e-05, "loss": 1.0274, "step": 59735 }, { "epoch": 0.15, "learning_rate": 8.522396698374352e-05, "loss": 1.0273, "step": 59740 }, { "epoch": 0.15, "learning_rate": 8.522270874226181e-05, "loss": 1.0464, "step": 59745 }, { "epoch": 0.15, "learning_rate": 8.52214505007801e-05, "loss": 1.0426, "step": 59750 }, { "epoch": 0.15, "learning_rate": 8.522019225929841e-05, "loss": 1.025, "step": 59755 }, { "epoch": 0.15, "learning_rate": 8.52189340178167e-05, "loss": 1.028, "step": 59760 }, { "epoch": 0.15, "learning_rate": 8.521767577633499e-05, "loss": 1.0286, "step": 59765 }, { "epoch": 0.15, "learning_rate": 8.521641753485328e-05, "loss": 1.0257, "step": 59770 }, { "epoch": 0.15, "learning_rate": 8.521541094166793e-05, "loss": 1.0331, "step": 59775 }, { "epoch": 0.15, "learning_rate": 8.521415270018622e-05, "loss": 1.0262, "step": 59780 }, { "epoch": 0.15, "learning_rate": 8.521289445870451e-05, "loss": 1.0262, "step": 59785 }, { "epoch": 0.15, "learning_rate": 8.52116362172228e-05, "loss": 1.0278, "step": 59790 }, { "epoch": 0.15, "learning_rate": 8.521037797574111e-05, "loss": 1.0268, "step": 59795 }, { "epoch": 0.15, "learning_rate": 8.52091197342594e-05, "loss": 1.0263, "step": 59800 }, { "epoch": 0.15, "learning_rate": 8.52078614927777e-05, "loss": 1.0294, "step": 59805 }, { "epoch": 0.15, "learning_rate": 8.520660325129599e-05, "loss": 1.0264, "step": 59810 }, { "epoch": 0.15, "learning_rate": 8.520534500981429e-05, "loss": 1.03, "step": 59815 }, { "epoch": 0.15, "learning_rate": 8.520408676833258e-05, "loss": 1.0272, "step": 59820 }, { "epoch": 0.15, "learning_rate": 8.520282852685087e-05, "loss": 1.0278, "step": 59825 }, { "epoch": 0.15, "learning_rate": 8.520157028536917e-05, "loss": 1.0266, "step": 59830 }, { "epoch": 0.15, "learning_rate": 8.520031204388746e-05, "loss": 1.0269, "step": 59835 }, { "epoch": 0.15, "learning_rate": 8.519905380240576e-05, "loss": 1.0271, "step": 59840 }, { "epoch": 0.15, "learning_rate": 8.519779556092405e-05, "loss": 1.0249, "step": 59845 }, { "epoch": 0.15, "learning_rate": 8.519653731944235e-05, "loss": 1.0272, "step": 59850 }, { "epoch": 0.15, "learning_rate": 8.519527907796064e-05, "loss": 1.0266, "step": 59855 }, { "epoch": 0.15, "learning_rate": 8.519402083647894e-05, "loss": 1.0386, "step": 59860 }, { "epoch": 0.15, "learning_rate": 8.519276259499723e-05, "loss": 1.0251, "step": 59865 }, { "epoch": 0.15, "learning_rate": 8.519150435351553e-05, "loss": 1.026, "step": 59870 }, { "epoch": 0.15, "learning_rate": 8.519024611203382e-05, "loss": 1.026, "step": 59875 }, { "epoch": 0.15, "learning_rate": 8.518898787055212e-05, "loss": 1.0264, "step": 59880 }, { "epoch": 0.15, "learning_rate": 8.518772962907041e-05, "loss": 1.027, "step": 59885 }, { "epoch": 0.15, "learning_rate": 8.51864713875887e-05, "loss": 1.0282, "step": 59890 }, { "epoch": 0.15, "learning_rate": 8.5185213146107e-05, "loss": 1.025, "step": 59895 }, { "epoch": 0.15, "learning_rate": 8.518395490462529e-05, "loss": 1.0269, "step": 59900 }, { "epoch": 0.15, "learning_rate": 8.51826966631436e-05, "loss": 1.0241, "step": 59905 }, { "epoch": 0.15, "learning_rate": 8.518143842166189e-05, "loss": 1.0262, "step": 59910 }, { "epoch": 0.15, "learning_rate": 8.518018018018018e-05, "loss": 1.0264, "step": 59915 }, { "epoch": 0.15, "learning_rate": 8.517892193869847e-05, "loss": 1.0262, "step": 59920 }, { "epoch": 0.15, "learning_rate": 8.517766369721677e-05, "loss": 1.0285, "step": 59925 }, { "epoch": 0.15, "learning_rate": 8.517640545573507e-05, "loss": 1.0267, "step": 59930 }, { "epoch": 0.15, "learning_rate": 8.517514721425336e-05, "loss": 1.0286, "step": 59935 }, { "epoch": 0.15, "learning_rate": 8.517388897277165e-05, "loss": 1.0281, "step": 59940 }, { "epoch": 0.15, "learning_rate": 8.517263073128995e-05, "loss": 1.0269, "step": 59945 }, { "epoch": 0.15, "learning_rate": 8.517137248980825e-05, "loss": 1.0256, "step": 59950 }, { "epoch": 0.15, "learning_rate": 8.517011424832654e-05, "loss": 1.0266, "step": 59955 }, { "epoch": 0.15, "learning_rate": 8.516885600684483e-05, "loss": 1.0267, "step": 59960 }, { "epoch": 0.15, "learning_rate": 8.516759776536312e-05, "loss": 1.0256, "step": 59965 }, { "epoch": 0.15, "learning_rate": 8.516633952388143e-05, "loss": 1.0269, "step": 59970 }, { "epoch": 0.15, "learning_rate": 8.516508128239972e-05, "loss": 1.0487, "step": 59975 }, { "epoch": 0.15, "learning_rate": 8.516382304091801e-05, "loss": 1.0263, "step": 59980 }, { "epoch": 0.15, "learning_rate": 8.516256479943631e-05, "loss": 1.0245, "step": 59985 }, { "epoch": 0.15, "learning_rate": 8.51613065579546e-05, "loss": 1.0261, "step": 59990 }, { "epoch": 0.15, "learning_rate": 8.516004831647291e-05, "loss": 1.0272, "step": 59995 }, { "epoch": 0.15, "learning_rate": 8.51587900749912e-05, "loss": 1.0267, "step": 60000 }, { "epoch": 0.15, "learning_rate": 8.51575318335095e-05, "loss": 1.0252, "step": 60005 }, { "epoch": 0.15, "learning_rate": 8.515627359202779e-05, "loss": 1.0258, "step": 60010 }, { "epoch": 0.15, "learning_rate": 8.515501535054609e-05, "loss": 1.0268, "step": 60015 }, { "epoch": 0.15, "learning_rate": 8.515375710906438e-05, "loss": 1.0274, "step": 60020 }, { "epoch": 0.15, "learning_rate": 8.515249886758267e-05, "loss": 1.0263, "step": 60025 }, { "epoch": 0.15, "learning_rate": 8.515124062610097e-05, "loss": 1.0257, "step": 60030 }, { "epoch": 0.15, "learning_rate": 8.514998238461927e-05, "loss": 1.0262, "step": 60035 }, { "epoch": 0.15, "learning_rate": 8.514872414313756e-05, "loss": 1.0261, "step": 60040 }, { "epoch": 0.15, "learning_rate": 8.514746590165585e-05, "loss": 1.0258, "step": 60045 }, { "epoch": 0.15, "learning_rate": 8.514620766017415e-05, "loss": 1.0278, "step": 60050 }, { "epoch": 0.15, "learning_rate": 8.514494941869244e-05, "loss": 1.0282, "step": 60055 }, { "epoch": 0.15, "learning_rate": 8.514369117721074e-05, "loss": 1.0262, "step": 60060 }, { "epoch": 0.15, "learning_rate": 8.514243293572903e-05, "loss": 1.0276, "step": 60065 }, { "epoch": 0.15, "learning_rate": 8.514117469424733e-05, "loss": 1.0278, "step": 60070 }, { "epoch": 0.15, "learning_rate": 8.513991645276562e-05, "loss": 1.0267, "step": 60075 }, { "epoch": 0.15, "learning_rate": 8.513865821128392e-05, "loss": 1.024, "step": 60080 }, { "epoch": 0.15, "learning_rate": 8.513739996980221e-05, "loss": 1.0268, "step": 60085 }, { "epoch": 0.15, "learning_rate": 8.51361417283205e-05, "loss": 1.026, "step": 60090 }, { "epoch": 0.15, "learning_rate": 8.51348834868388e-05, "loss": 1.0265, "step": 60095 }, { "epoch": 0.15, "learning_rate": 8.51336252453571e-05, "loss": 1.0257, "step": 60100 }, { "epoch": 0.15, "learning_rate": 8.51323670038754e-05, "loss": 1.0274, "step": 60105 }, { "epoch": 0.15, "learning_rate": 8.513110876239369e-05, "loss": 1.0272, "step": 60110 }, { "epoch": 0.15, "learning_rate": 8.512985052091198e-05, "loss": 1.025, "step": 60115 }, { "epoch": 0.15, "learning_rate": 8.512859227943027e-05, "loss": 1.0248, "step": 60120 }, { "epoch": 0.15, "learning_rate": 8.512733403794857e-05, "loss": 1.027, "step": 60125 }, { "epoch": 0.15, "learning_rate": 8.512607579646687e-05, "loss": 1.0283, "step": 60130 }, { "epoch": 0.15, "learning_rate": 8.512481755498516e-05, "loss": 1.0264, "step": 60135 }, { "epoch": 0.15, "learning_rate": 8.512355931350345e-05, "loss": 1.027, "step": 60140 }, { "epoch": 0.15, "learning_rate": 8.512230107202175e-05, "loss": 1.0256, "step": 60145 }, { "epoch": 0.15, "learning_rate": 8.512104283054005e-05, "loss": 1.0252, "step": 60150 }, { "epoch": 0.15, "learning_rate": 8.511978458905834e-05, "loss": 1.0293, "step": 60155 }, { "epoch": 0.15, "learning_rate": 8.511852634757663e-05, "loss": 1.0262, "step": 60160 }, { "epoch": 0.15, "learning_rate": 8.511726810609493e-05, "loss": 1.0278, "step": 60165 }, { "epoch": 0.15, "learning_rate": 8.511600986461323e-05, "loss": 1.0264, "step": 60170 }, { "epoch": 0.15, "learning_rate": 8.511475162313152e-05, "loss": 1.0261, "step": 60175 }, { "epoch": 0.15, "learning_rate": 8.511349338164981e-05, "loss": 1.0287, "step": 60180 }, { "epoch": 0.15, "learning_rate": 8.51122351401681e-05, "loss": 1.0281, "step": 60185 }, { "epoch": 0.15, "learning_rate": 8.51109768986864e-05, "loss": 1.0263, "step": 60190 }, { "epoch": 0.15, "learning_rate": 8.51097186572047e-05, "loss": 1.0262, "step": 60195 }, { "epoch": 0.15, "learning_rate": 8.510846041572299e-05, "loss": 1.0248, "step": 60200 }, { "epoch": 0.15, "learning_rate": 8.510720217424128e-05, "loss": 1.0256, "step": 60205 }, { "epoch": 0.15, "learning_rate": 8.510594393275959e-05, "loss": 1.0269, "step": 60210 }, { "epoch": 0.15, "learning_rate": 8.510468569127788e-05, "loss": 1.026, "step": 60215 }, { "epoch": 0.15, "learning_rate": 8.510342744979617e-05, "loss": 1.0262, "step": 60220 }, { "epoch": 0.15, "learning_rate": 8.510216920831446e-05, "loss": 1.0296, "step": 60225 }, { "epoch": 0.15, "learning_rate": 8.510091096683277e-05, "loss": 1.0237, "step": 60230 }, { "epoch": 0.15, "learning_rate": 8.509965272535106e-05, "loss": 1.0263, "step": 60235 }, { "epoch": 0.15, "learning_rate": 8.509839448386935e-05, "loss": 1.0241, "step": 60240 }, { "epoch": 0.15, "learning_rate": 8.509713624238764e-05, "loss": 1.0256, "step": 60245 }, { "epoch": 0.15, "learning_rate": 8.509587800090593e-05, "loss": 1.027, "step": 60250 }, { "epoch": 0.15, "learning_rate": 8.509461975942424e-05, "loss": 1.0246, "step": 60255 }, { "epoch": 0.15, "learning_rate": 8.509336151794253e-05, "loss": 1.0264, "step": 60260 }, { "epoch": 0.15, "learning_rate": 8.509210327646082e-05, "loss": 1.0259, "step": 60265 }, { "epoch": 0.15, "learning_rate": 8.509084503497911e-05, "loss": 1.029, "step": 60270 }, { "epoch": 0.15, "learning_rate": 8.508958679349742e-05, "loss": 1.0262, "step": 60275 }, { "epoch": 0.15, "learning_rate": 8.508832855201571e-05, "loss": 1.0264, "step": 60280 }, { "epoch": 0.15, "learning_rate": 8.5087070310534e-05, "loss": 1.0283, "step": 60285 }, { "epoch": 0.15, "learning_rate": 8.508581206905229e-05, "loss": 1.0263, "step": 60290 }, { "epoch": 0.15, "learning_rate": 8.50845538275706e-05, "loss": 1.027, "step": 60295 }, { "epoch": 0.15, "learning_rate": 8.508329558608889e-05, "loss": 1.0286, "step": 60300 }, { "epoch": 0.15, "learning_rate": 8.508203734460718e-05, "loss": 1.0442, "step": 60305 }, { "epoch": 0.15, "learning_rate": 8.508077910312547e-05, "loss": 1.0265, "step": 60310 }, { "epoch": 0.15, "learning_rate": 8.507952086164376e-05, "loss": 1.0269, "step": 60315 }, { "epoch": 0.15, "learning_rate": 8.507826262016207e-05, "loss": 1.0299, "step": 60320 }, { "epoch": 0.15, "learning_rate": 8.507700437868036e-05, "loss": 1.0257, "step": 60325 }, { "epoch": 0.15, "learning_rate": 8.507574613719865e-05, "loss": 1.0251, "step": 60330 }, { "epoch": 0.15, "learning_rate": 8.507448789571694e-05, "loss": 1.0248, "step": 60335 }, { "epoch": 0.15, "learning_rate": 8.507322965423525e-05, "loss": 1.0253, "step": 60340 }, { "epoch": 0.15, "learning_rate": 8.507197141275354e-05, "loss": 1.0286, "step": 60345 }, { "epoch": 0.15, "learning_rate": 8.507071317127183e-05, "loss": 1.0514, "step": 60350 }, { "epoch": 0.15, "learning_rate": 8.506945492979012e-05, "loss": 1.0266, "step": 60355 }, { "epoch": 0.15, "learning_rate": 8.506819668830843e-05, "loss": 1.0244, "step": 60360 }, { "epoch": 0.15, "learning_rate": 8.506693844682672e-05, "loss": 1.027, "step": 60365 }, { "epoch": 0.15, "learning_rate": 8.506568020534501e-05, "loss": 1.0262, "step": 60370 }, { "epoch": 0.15, "learning_rate": 8.50644219638633e-05, "loss": 1.0255, "step": 60375 }, { "epoch": 0.15, "learning_rate": 8.50631637223816e-05, "loss": 1.0268, "step": 60380 }, { "epoch": 0.15, "learning_rate": 8.50619054808999e-05, "loss": 1.0254, "step": 60385 }, { "epoch": 0.15, "learning_rate": 8.506064723941819e-05, "loss": 1.0261, "step": 60390 }, { "epoch": 0.15, "learning_rate": 8.505938899793648e-05, "loss": 1.0261, "step": 60395 }, { "epoch": 0.15, "learning_rate": 8.505813075645477e-05, "loss": 1.0277, "step": 60400 }, { "epoch": 0.15, "learning_rate": 8.505687251497308e-05, "loss": 1.0264, "step": 60405 }, { "epoch": 0.15, "learning_rate": 8.505561427349137e-05, "loss": 1.0265, "step": 60410 }, { "epoch": 0.15, "learning_rate": 8.505435603200966e-05, "loss": 1.0256, "step": 60415 }, { "epoch": 0.15, "learning_rate": 8.505309779052795e-05, "loss": 1.0257, "step": 60420 }, { "epoch": 0.15, "learning_rate": 8.505183954904626e-05, "loss": 1.0285, "step": 60425 }, { "epoch": 0.15, "learning_rate": 8.505058130756455e-05, "loss": 1.0289, "step": 60430 }, { "epoch": 0.15, "learning_rate": 8.504932306608284e-05, "loss": 1.0267, "step": 60435 }, { "epoch": 0.15, "learning_rate": 8.504806482460113e-05, "loss": 1.0254, "step": 60440 }, { "epoch": 0.15, "learning_rate": 8.504680658311943e-05, "loss": 1.027, "step": 60445 }, { "epoch": 0.15, "learning_rate": 8.504554834163773e-05, "loss": 1.0263, "step": 60450 }, { "epoch": 0.15, "learning_rate": 8.504429010015602e-05, "loss": 1.0249, "step": 60455 }, { "epoch": 0.15, "learning_rate": 8.504303185867431e-05, "loss": 1.0269, "step": 60460 }, { "epoch": 0.15, "learning_rate": 8.50417736171926e-05, "loss": 1.0246, "step": 60465 }, { "epoch": 0.15, "learning_rate": 8.504051537571091e-05, "loss": 1.0275, "step": 60470 }, { "epoch": 0.15, "learning_rate": 8.50392571342292e-05, "loss": 1.0261, "step": 60475 }, { "epoch": 0.15, "learning_rate": 8.50379988927475e-05, "loss": 1.0276, "step": 60480 }, { "epoch": 0.15, "learning_rate": 8.50367406512658e-05, "loss": 1.0254, "step": 60485 }, { "epoch": 0.15, "learning_rate": 8.503548240978409e-05, "loss": 1.0261, "step": 60490 }, { "epoch": 0.15, "learning_rate": 8.50342241683024e-05, "loss": 1.0286, "step": 60495 }, { "epoch": 0.15, "learning_rate": 8.503296592682069e-05, "loss": 1.0281, "step": 60500 }, { "epoch": 0.15, "learning_rate": 8.503170768533898e-05, "loss": 1.0261, "step": 60505 }, { "epoch": 0.15, "learning_rate": 8.503044944385727e-05, "loss": 1.0295, "step": 60510 }, { "epoch": 0.15, "learning_rate": 8.502919120237556e-05, "loss": 1.0255, "step": 60515 }, { "epoch": 0.15, "learning_rate": 8.502793296089387e-05, "loss": 1.0283, "step": 60520 }, { "epoch": 0.15, "learning_rate": 8.502667471941216e-05, "loss": 1.029, "step": 60525 }, { "epoch": 0.15, "learning_rate": 8.502541647793045e-05, "loss": 1.0261, "step": 60530 }, { "epoch": 0.15, "learning_rate": 8.502415823644874e-05, "loss": 1.0263, "step": 60535 }, { "epoch": 0.15, "learning_rate": 8.502289999496705e-05, "loss": 1.0257, "step": 60540 }, { "epoch": 0.15, "learning_rate": 8.502164175348534e-05, "loss": 1.0253, "step": 60545 }, { "epoch": 0.15, "learning_rate": 8.502038351200363e-05, "loss": 1.0266, "step": 60550 }, { "epoch": 0.15, "learning_rate": 8.501912527052192e-05, "loss": 1.0248, "step": 60555 }, { "epoch": 0.15, "learning_rate": 8.501786702904023e-05, "loss": 1.0264, "step": 60560 }, { "epoch": 0.15, "learning_rate": 8.501660878755852e-05, "loss": 1.0254, "step": 60565 }, { "epoch": 0.15, "learning_rate": 8.501535054607681e-05, "loss": 1.0272, "step": 60570 }, { "epoch": 0.15, "learning_rate": 8.50140923045951e-05, "loss": 1.0261, "step": 60575 }, { "epoch": 0.15, "learning_rate": 8.50128340631134e-05, "loss": 1.0258, "step": 60580 }, { "epoch": 0.15, "learning_rate": 8.50115758216317e-05, "loss": 1.0272, "step": 60585 }, { "epoch": 0.15, "learning_rate": 8.501031758014999e-05, "loss": 1.0261, "step": 60590 }, { "epoch": 0.15, "learning_rate": 8.500905933866828e-05, "loss": 1.025, "step": 60595 }, { "epoch": 0.15, "learning_rate": 8.500780109718657e-05, "loss": 1.0261, "step": 60600 }, { "epoch": 0.15, "learning_rate": 8.500654285570488e-05, "loss": 1.0272, "step": 60605 }, { "epoch": 0.15, "learning_rate": 8.500528461422317e-05, "loss": 1.0245, "step": 60610 }, { "epoch": 0.15, "learning_rate": 8.500402637274146e-05, "loss": 1.0288, "step": 60615 }, { "epoch": 0.15, "learning_rate": 8.500276813125975e-05, "loss": 1.0272, "step": 60620 }, { "epoch": 0.15, "learning_rate": 8.500150988977806e-05, "loss": 1.0257, "step": 60625 }, { "epoch": 0.15, "learning_rate": 8.500025164829635e-05, "loss": 1.0278, "step": 60630 }, { "epoch": 0.15, "learning_rate": 8.499899340681464e-05, "loss": 1.0266, "step": 60635 }, { "epoch": 0.15, "learning_rate": 8.499773516533293e-05, "loss": 1.049, "step": 60640 }, { "epoch": 0.15, "learning_rate": 8.499647692385123e-05, "loss": 1.0267, "step": 60645 }, { "epoch": 0.15, "learning_rate": 8.499521868236953e-05, "loss": 1.0261, "step": 60650 }, { "epoch": 0.15, "learning_rate": 8.499396044088782e-05, "loss": 1.0262, "step": 60655 }, { "epoch": 0.15, "learning_rate": 8.499270219940611e-05, "loss": 1.028, "step": 60660 }, { "epoch": 0.15, "learning_rate": 8.49914439579244e-05, "loss": 1.0473, "step": 60665 }, { "epoch": 0.15, "learning_rate": 8.499018571644271e-05, "loss": 1.0259, "step": 60670 }, { "epoch": 0.15, "learning_rate": 8.4988927474961e-05, "loss": 1.0463, "step": 60675 }, { "epoch": 0.15, "learning_rate": 8.49876692334793e-05, "loss": 1.0255, "step": 60680 }, { "epoch": 0.15, "learning_rate": 8.498641099199759e-05, "loss": 1.0271, "step": 60685 }, { "epoch": 0.15, "learning_rate": 8.498515275051589e-05, "loss": 1.0266, "step": 60690 }, { "epoch": 0.15, "learning_rate": 8.498389450903418e-05, "loss": 1.028, "step": 60695 }, { "epoch": 0.15, "learning_rate": 8.498263626755247e-05, "loss": 1.0252, "step": 60700 }, { "epoch": 0.15, "learning_rate": 8.498137802607077e-05, "loss": 1.0265, "step": 60705 }, { "epoch": 0.15, "learning_rate": 8.498011978458906e-05, "loss": 1.0259, "step": 60710 }, { "epoch": 0.15, "learning_rate": 8.497886154310736e-05, "loss": 1.0472, "step": 60715 }, { "epoch": 0.15, "learning_rate": 8.497760330162565e-05, "loss": 1.0264, "step": 60720 }, { "epoch": 0.15, "learning_rate": 8.497634506014395e-05, "loss": 1.0255, "step": 60725 }, { "epoch": 0.15, "learning_rate": 8.497508681866224e-05, "loss": 1.0283, "step": 60730 }, { "epoch": 0.15, "learning_rate": 8.497382857718054e-05, "loss": 1.0251, "step": 60735 }, { "epoch": 0.15, "learning_rate": 8.497257033569883e-05, "loss": 1.0526, "step": 60740 }, { "epoch": 0.15, "learning_rate": 8.497131209421713e-05, "loss": 1.0243, "step": 60745 }, { "epoch": 0.15, "learning_rate": 8.497005385273542e-05, "loss": 1.0264, "step": 60750 }, { "epoch": 0.15, "learning_rate": 8.496879561125372e-05, "loss": 1.0251, "step": 60755 }, { "epoch": 0.15, "learning_rate": 8.496753736977201e-05, "loss": 1.0269, "step": 60760 }, { "epoch": 0.15, "learning_rate": 8.49662791282903e-05, "loss": 1.048, "step": 60765 }, { "epoch": 0.15, "learning_rate": 8.49650208868086e-05, "loss": 1.0249, "step": 60770 }, { "epoch": 0.15, "learning_rate": 8.496376264532689e-05, "loss": 1.0253, "step": 60775 }, { "epoch": 0.15, "learning_rate": 8.49625044038452e-05, "loss": 1.0272, "step": 60780 }, { "epoch": 0.15, "learning_rate": 8.496124616236348e-05, "loss": 1.048, "step": 60785 }, { "epoch": 0.15, "learning_rate": 8.495998792088178e-05, "loss": 1.0273, "step": 60790 }, { "epoch": 0.15, "learning_rate": 8.495872967940007e-05, "loss": 1.0267, "step": 60795 }, { "epoch": 0.15, "learning_rate": 8.495747143791837e-05, "loss": 1.0271, "step": 60800 }, { "epoch": 0.15, "learning_rate": 8.495621319643666e-05, "loss": 1.0283, "step": 60805 }, { "epoch": 0.15, "learning_rate": 8.495495495495496e-05, "loss": 1.0263, "step": 60810 }, { "epoch": 0.15, "learning_rate": 8.495369671347325e-05, "loss": 1.0267, "step": 60815 }, { "epoch": 0.15, "learning_rate": 8.495243847199155e-05, "loss": 1.0282, "step": 60820 }, { "epoch": 0.15, "learning_rate": 8.495118023050984e-05, "loss": 1.0457, "step": 60825 }, { "epoch": 0.15, "learning_rate": 8.494992198902814e-05, "loss": 1.0256, "step": 60830 }, { "epoch": 0.15, "learning_rate": 8.494866374754643e-05, "loss": 1.0252, "step": 60835 }, { "epoch": 0.15, "learning_rate": 8.494740550606472e-05, "loss": 1.026, "step": 60840 }, { "epoch": 0.15, "learning_rate": 8.494614726458302e-05, "loss": 1.0267, "step": 60845 }, { "epoch": 0.15, "learning_rate": 8.494488902310132e-05, "loss": 1.0464, "step": 60850 }, { "epoch": 0.15, "learning_rate": 8.494363078161961e-05, "loss": 1.0254, "step": 60855 }, { "epoch": 0.15, "learning_rate": 8.49423725401379e-05, "loss": 1.0241, "step": 60860 }, { "epoch": 0.15, "learning_rate": 8.49411142986562e-05, "loss": 1.0283, "step": 60865 }, { "epoch": 0.15, "learning_rate": 8.49398560571745e-05, "loss": 1.0243, "step": 60870 }, { "epoch": 0.15, "learning_rate": 8.493859781569279e-05, "loss": 1.0291, "step": 60875 }, { "epoch": 0.15, "learning_rate": 8.493733957421108e-05, "loss": 1.0278, "step": 60880 }, { "epoch": 0.15, "learning_rate": 8.493608133272938e-05, "loss": 1.0266, "step": 60885 }, { "epoch": 0.15, "learning_rate": 8.493482309124768e-05, "loss": 1.0271, "step": 60890 }, { "epoch": 0.15, "learning_rate": 8.493356484976597e-05, "loss": 1.0252, "step": 60895 }, { "epoch": 0.15, "learning_rate": 8.493230660828426e-05, "loss": 1.0275, "step": 60900 }, { "epoch": 0.15, "learning_rate": 8.493104836680255e-05, "loss": 1.0272, "step": 60905 }, { "epoch": 0.15, "learning_rate": 8.492979012532086e-05, "loss": 1.0264, "step": 60910 }, { "epoch": 0.15, "learning_rate": 8.492853188383915e-05, "loss": 1.0258, "step": 60915 }, { "epoch": 0.15, "learning_rate": 8.492727364235744e-05, "loss": 1.0261, "step": 60920 }, { "epoch": 0.15, "learning_rate": 8.492601540087573e-05, "loss": 1.026, "step": 60925 }, { "epoch": 0.15, "learning_rate": 8.492475715939404e-05, "loss": 1.0261, "step": 60930 }, { "epoch": 0.15, "learning_rate": 8.492349891791233e-05, "loss": 1.0258, "step": 60935 }, { "epoch": 0.15, "learning_rate": 8.492224067643062e-05, "loss": 1.0234, "step": 60940 }, { "epoch": 0.15, "learning_rate": 8.492098243494891e-05, "loss": 1.0263, "step": 60945 }, { "epoch": 0.15, "learning_rate": 8.491972419346722e-05, "loss": 1.0276, "step": 60950 }, { "epoch": 0.15, "learning_rate": 8.491846595198551e-05, "loss": 1.0256, "step": 60955 }, { "epoch": 0.15, "learning_rate": 8.49172077105038e-05, "loss": 1.0251, "step": 60960 }, { "epoch": 0.15, "learning_rate": 8.491594946902209e-05, "loss": 1.0285, "step": 60965 }, { "epoch": 0.15, "learning_rate": 8.491469122754038e-05, "loss": 1.0276, "step": 60970 }, { "epoch": 0.15, "learning_rate": 8.491343298605869e-05, "loss": 1.026, "step": 60975 }, { "epoch": 0.15, "learning_rate": 8.491217474457698e-05, "loss": 1.0285, "step": 60980 }, { "epoch": 0.15, "learning_rate": 8.491091650309528e-05, "loss": 1.0279, "step": 60985 }, { "epoch": 0.15, "learning_rate": 8.490965826161358e-05, "loss": 1.026, "step": 60990 }, { "epoch": 0.15, "learning_rate": 8.490840002013187e-05, "loss": 1.0264, "step": 60995 }, { "epoch": 0.15, "learning_rate": 8.490714177865017e-05, "loss": 1.0258, "step": 61000 }, { "epoch": 0.15, "learning_rate": 8.490588353716846e-05, "loss": 1.0262, "step": 61005 }, { "epoch": 0.15, "learning_rate": 8.490462529568676e-05, "loss": 1.0256, "step": 61010 }, { "epoch": 0.15, "learning_rate": 8.490336705420505e-05, "loss": 1.0286, "step": 61015 }, { "epoch": 0.15, "learning_rate": 8.490210881272335e-05, "loss": 1.0271, "step": 61020 }, { "epoch": 0.15, "learning_rate": 8.490085057124164e-05, "loss": 1.026, "step": 61025 }, { "epoch": 0.15, "learning_rate": 8.489959232975994e-05, "loss": 1.0261, "step": 61030 }, { "epoch": 0.15, "learning_rate": 8.489833408827823e-05, "loss": 1.0249, "step": 61035 }, { "epoch": 0.15, "learning_rate": 8.489707584679652e-05, "loss": 1.0262, "step": 61040 }, { "epoch": 0.15, "learning_rate": 8.489581760531482e-05, "loss": 1.0273, "step": 61045 }, { "epoch": 0.15, "learning_rate": 8.489455936383312e-05, "loss": 1.0257, "step": 61050 }, { "epoch": 0.15, "learning_rate": 8.489330112235141e-05, "loss": 1.0259, "step": 61055 }, { "epoch": 0.15, "learning_rate": 8.48920428808697e-05, "loss": 1.026, "step": 61060 }, { "epoch": 0.15, "learning_rate": 8.4890784639388e-05, "loss": 1.0267, "step": 61065 }, { "epoch": 0.15, "learning_rate": 8.48895263979063e-05, "loss": 1.0275, "step": 61070 }, { "epoch": 0.15, "learning_rate": 8.488826815642459e-05, "loss": 1.0266, "step": 61075 }, { "epoch": 0.15, "learning_rate": 8.488700991494288e-05, "loss": 1.0265, "step": 61080 }, { "epoch": 0.15, "learning_rate": 8.488575167346118e-05, "loss": 1.0269, "step": 61085 }, { "epoch": 0.15, "learning_rate": 8.488449343197948e-05, "loss": 1.0253, "step": 61090 }, { "epoch": 0.15, "learning_rate": 8.488323519049777e-05, "loss": 1.0283, "step": 61095 }, { "epoch": 0.15, "learning_rate": 8.488197694901606e-05, "loss": 1.025, "step": 61100 }, { "epoch": 0.15, "learning_rate": 8.488071870753435e-05, "loss": 1.0277, "step": 61105 }, { "epoch": 0.15, "learning_rate": 8.487946046605266e-05, "loss": 1.0268, "step": 61110 }, { "epoch": 0.15, "learning_rate": 8.487820222457095e-05, "loss": 1.0258, "step": 61115 }, { "epoch": 0.15, "learning_rate": 8.487694398308924e-05, "loss": 1.0265, "step": 61120 }, { "epoch": 0.15, "learning_rate": 8.487568574160753e-05, "loss": 1.0278, "step": 61125 }, { "epoch": 0.15, "learning_rate": 8.487442750012584e-05, "loss": 1.0253, "step": 61130 }, { "epoch": 0.15, "learning_rate": 8.487316925864413e-05, "loss": 1.0267, "step": 61135 }, { "epoch": 0.15, "learning_rate": 8.487191101716242e-05, "loss": 1.0271, "step": 61140 }, { "epoch": 0.15, "learning_rate": 8.487065277568071e-05, "loss": 1.0253, "step": 61145 }, { "epoch": 0.15, "learning_rate": 8.486939453419902e-05, "loss": 1.0265, "step": 61150 }, { "epoch": 0.15, "learning_rate": 8.486813629271731e-05, "loss": 1.0284, "step": 61155 }, { "epoch": 0.15, "learning_rate": 8.48668780512356e-05, "loss": 1.028, "step": 61160 }, { "epoch": 0.15, "learning_rate": 8.486561980975389e-05, "loss": 1.0254, "step": 61165 }, { "epoch": 0.15, "learning_rate": 8.486436156827218e-05, "loss": 1.0252, "step": 61170 }, { "epoch": 0.15, "learning_rate": 8.486310332679049e-05, "loss": 1.026, "step": 61175 }, { "epoch": 0.15, "learning_rate": 8.486184508530878e-05, "loss": 1.0275, "step": 61180 }, { "epoch": 0.15, "learning_rate": 8.486058684382707e-05, "loss": 1.0291, "step": 61185 }, { "epoch": 0.15, "learning_rate": 8.485932860234536e-05, "loss": 1.0264, "step": 61190 }, { "epoch": 0.15, "learning_rate": 8.485807036086367e-05, "loss": 1.0247, "step": 61195 }, { "epoch": 0.15, "learning_rate": 8.485681211938196e-05, "loss": 1.0248, "step": 61200 }, { "epoch": 0.15, "learning_rate": 8.485555387790025e-05, "loss": 1.0261, "step": 61205 }, { "epoch": 0.15, "learning_rate": 8.485429563641854e-05, "loss": 1.0254, "step": 61210 }, { "epoch": 0.15, "learning_rate": 8.485303739493685e-05, "loss": 1.0262, "step": 61215 }, { "epoch": 0.15, "learning_rate": 8.485177915345514e-05, "loss": 1.0258, "step": 61220 }, { "epoch": 0.15, "learning_rate": 8.485052091197343e-05, "loss": 1.0275, "step": 61225 }, { "epoch": 0.15, "learning_rate": 8.484926267049172e-05, "loss": 1.0253, "step": 61230 }, { "epoch": 0.15, "learning_rate": 8.484800442901001e-05, "loss": 1.0225, "step": 61235 }, { "epoch": 0.15, "learning_rate": 8.484674618752832e-05, "loss": 1.0241, "step": 61240 }, { "epoch": 0.15, "learning_rate": 8.484548794604661e-05, "loss": 1.0262, "step": 61245 }, { "epoch": 0.15, "learning_rate": 8.48442297045649e-05, "loss": 1.0255, "step": 61250 }, { "epoch": 0.15, "learning_rate": 8.484297146308319e-05, "loss": 1.0273, "step": 61255 }, { "epoch": 0.15, "learning_rate": 8.48417132216015e-05, "loss": 1.0257, "step": 61260 }, { "epoch": 0.15, "learning_rate": 8.484045498011979e-05, "loss": 1.0275, "step": 61265 }, { "epoch": 0.15, "learning_rate": 8.483919673863808e-05, "loss": 1.0292, "step": 61270 }, { "epoch": 0.15, "learning_rate": 8.483793849715637e-05, "loss": 1.048, "step": 61275 }, { "epoch": 0.15, "learning_rate": 8.483668025567468e-05, "loss": 1.0236, "step": 61280 }, { "epoch": 0.15, "learning_rate": 8.483542201419297e-05, "loss": 1.0505, "step": 61285 }, { "epoch": 0.15, "learning_rate": 8.483416377271126e-05, "loss": 1.0252, "step": 61290 }, { "epoch": 0.15, "learning_rate": 8.483290553122955e-05, "loss": 1.0263, "step": 61295 }, { "epoch": 0.15, "learning_rate": 8.483164728974784e-05, "loss": 1.027, "step": 61300 }, { "epoch": 0.15, "learning_rate": 8.483038904826615e-05, "loss": 1.0272, "step": 61305 }, { "epoch": 0.15, "learning_rate": 8.482913080678444e-05, "loss": 1.0271, "step": 61310 }, { "epoch": 0.15, "learning_rate": 8.482787256530273e-05, "loss": 1.0271, "step": 61315 }, { "epoch": 0.15, "learning_rate": 8.482661432382102e-05, "loss": 1.0273, "step": 61320 }, { "epoch": 0.15, "learning_rate": 8.482535608233933e-05, "loss": 1.0272, "step": 61325 }, { "epoch": 0.15, "learning_rate": 8.482409784085762e-05, "loss": 1.0286, "step": 61330 }, { "epoch": 0.15, "learning_rate": 8.482283959937591e-05, "loss": 1.0268, "step": 61335 }, { "epoch": 0.15, "learning_rate": 8.48215813578942e-05, "loss": 1.0242, "step": 61340 }, { "epoch": 0.15, "learning_rate": 8.482032311641251e-05, "loss": 1.0273, "step": 61345 }, { "epoch": 0.15, "learning_rate": 8.48190648749308e-05, "loss": 1.0262, "step": 61350 }, { "epoch": 0.15, "learning_rate": 8.481780663344909e-05, "loss": 1.0261, "step": 61355 }, { "epoch": 0.15, "learning_rate": 8.481654839196738e-05, "loss": 1.0259, "step": 61360 }, { "epoch": 0.15, "learning_rate": 8.481529015048568e-05, "loss": 1.028, "step": 61365 }, { "epoch": 0.15, "learning_rate": 8.481403190900398e-05, "loss": 1.0271, "step": 61370 }, { "epoch": 0.15, "learning_rate": 8.481277366752227e-05, "loss": 1.0269, "step": 61375 }, { "epoch": 0.15, "learning_rate": 8.481151542604056e-05, "loss": 1.0251, "step": 61380 }, { "epoch": 0.15, "learning_rate": 8.481025718455886e-05, "loss": 1.0264, "step": 61385 }, { "epoch": 0.15, "learning_rate": 8.480899894307716e-05, "loss": 1.0262, "step": 61390 }, { "epoch": 0.15, "learning_rate": 8.480774070159545e-05, "loss": 1.0275, "step": 61395 }, { "epoch": 0.15, "learning_rate": 8.480648246011374e-05, "loss": 1.027, "step": 61400 }, { "epoch": 0.15, "learning_rate": 8.480522421863204e-05, "loss": 1.0243, "step": 61405 }, { "epoch": 0.15, "learning_rate": 8.480396597715034e-05, "loss": 1.026, "step": 61410 }, { "epoch": 0.15, "learning_rate": 8.480270773566863e-05, "loss": 1.0257, "step": 61415 }, { "epoch": 0.15, "learning_rate": 8.480144949418692e-05, "loss": 1.0267, "step": 61420 }, { "epoch": 0.15, "learning_rate": 8.480019125270522e-05, "loss": 1.0259, "step": 61425 }, { "epoch": 0.15, "learning_rate": 8.479893301122351e-05, "loss": 1.028, "step": 61430 }, { "epoch": 0.15, "learning_rate": 8.479767476974181e-05, "loss": 1.027, "step": 61435 }, { "epoch": 0.15, "learning_rate": 8.47964165282601e-05, "loss": 1.026, "step": 61440 }, { "epoch": 0.15, "learning_rate": 8.47951582867784e-05, "loss": 1.0263, "step": 61445 }, { "epoch": 0.15, "learning_rate": 8.479390004529669e-05, "loss": 1.0269, "step": 61450 }, { "epoch": 0.15, "learning_rate": 8.479264180381499e-05, "loss": 1.0249, "step": 61455 }, { "epoch": 0.15, "learning_rate": 8.479138356233328e-05, "loss": 1.0288, "step": 61460 }, { "epoch": 0.15, "learning_rate": 8.479012532085158e-05, "loss": 1.0271, "step": 61465 }, { "epoch": 0.15, "learning_rate": 8.478886707936987e-05, "loss": 1.0266, "step": 61470 }, { "epoch": 0.15, "learning_rate": 8.478760883788817e-05, "loss": 1.0252, "step": 61475 }, { "epoch": 0.15, "learning_rate": 8.478635059640648e-05, "loss": 1.0253, "step": 61480 }, { "epoch": 0.15, "learning_rate": 8.478509235492477e-05, "loss": 1.0278, "step": 61485 }, { "epoch": 0.15, "learning_rate": 8.478383411344306e-05, "loss": 1.0258, "step": 61490 }, { "epoch": 0.15, "learning_rate": 8.478257587196135e-05, "loss": 1.0272, "step": 61495 }, { "epoch": 0.15, "learning_rate": 8.478131763047964e-05, "loss": 1.026, "step": 61500 }, { "epoch": 0.15, "learning_rate": 8.478005938899795e-05, "loss": 1.0249, "step": 61505 }, { "epoch": 0.15, "learning_rate": 8.477880114751624e-05, "loss": 1.0283, "step": 61510 }, { "epoch": 0.15, "learning_rate": 8.477754290603453e-05, "loss": 1.0253, "step": 61515 }, { "epoch": 0.15, "learning_rate": 8.477628466455282e-05, "loss": 1.0262, "step": 61520 }, { "epoch": 0.15, "learning_rate": 8.477502642307113e-05, "loss": 1.0246, "step": 61525 }, { "epoch": 0.15, "learning_rate": 8.477376818158942e-05, "loss": 1.0268, "step": 61530 }, { "epoch": 0.15, "learning_rate": 8.477250994010771e-05, "loss": 1.0255, "step": 61535 }, { "epoch": 0.15, "learning_rate": 8.4771251698626e-05, "loss": 1.0262, "step": 61540 }, { "epoch": 0.15, "learning_rate": 8.476999345714431e-05, "loss": 1.0259, "step": 61545 }, { "epoch": 0.15, "learning_rate": 8.47687352156626e-05, "loss": 1.0287, "step": 61550 }, { "epoch": 0.15, "learning_rate": 8.476747697418089e-05, "loss": 1.0244, "step": 61555 }, { "epoch": 0.15, "learning_rate": 8.476621873269918e-05, "loss": 1.0276, "step": 61560 }, { "epoch": 0.15, "learning_rate": 8.476496049121748e-05, "loss": 1.0279, "step": 61565 }, { "epoch": 0.15, "learning_rate": 8.476370224973578e-05, "loss": 1.0267, "step": 61570 }, { "epoch": 0.15, "learning_rate": 8.476244400825407e-05, "loss": 1.0276, "step": 61575 }, { "epoch": 0.15, "learning_rate": 8.476118576677236e-05, "loss": 1.0289, "step": 61580 }, { "epoch": 0.15, "learning_rate": 8.475992752529066e-05, "loss": 1.0269, "step": 61585 }, { "epoch": 0.15, "learning_rate": 8.475866928380896e-05, "loss": 1.0247, "step": 61590 }, { "epoch": 0.15, "learning_rate": 8.475741104232725e-05, "loss": 1.0259, "step": 61595 }, { "epoch": 0.15, "learning_rate": 8.475615280084554e-05, "loss": 1.0264, "step": 61600 }, { "epoch": 0.15, "learning_rate": 8.475489455936384e-05, "loss": 1.0254, "step": 61605 }, { "epoch": 0.15, "learning_rate": 8.475363631788214e-05, "loss": 1.0262, "step": 61610 }, { "epoch": 0.15, "learning_rate": 8.475237807640043e-05, "loss": 1.0249, "step": 61615 }, { "epoch": 0.15, "learning_rate": 8.475111983491872e-05, "loss": 1.0266, "step": 61620 }, { "epoch": 0.15, "learning_rate": 8.474986159343702e-05, "loss": 1.0253, "step": 61625 }, { "epoch": 0.15, "learning_rate": 8.474860335195531e-05, "loss": 1.0238, "step": 61630 }, { "epoch": 0.15, "learning_rate": 8.474734511047361e-05, "loss": 1.0258, "step": 61635 }, { "epoch": 0.15, "learning_rate": 8.47460868689919e-05, "loss": 1.0265, "step": 61640 }, { "epoch": 0.15, "learning_rate": 8.47448286275102e-05, "loss": 1.0277, "step": 61645 }, { "epoch": 0.15, "learning_rate": 8.474357038602849e-05, "loss": 1.0283, "step": 61650 }, { "epoch": 0.15, "learning_rate": 8.474231214454679e-05, "loss": 1.0259, "step": 61655 }, { "epoch": 0.15, "learning_rate": 8.474105390306508e-05, "loss": 1.0266, "step": 61660 }, { "epoch": 0.15, "learning_rate": 8.473979566158338e-05, "loss": 1.0276, "step": 61665 }, { "epoch": 0.15, "learning_rate": 8.473853742010167e-05, "loss": 1.0264, "step": 61670 }, { "epoch": 0.15, "learning_rate": 8.473727917861997e-05, "loss": 1.0271, "step": 61675 }, { "epoch": 0.15, "learning_rate": 8.473602093713826e-05, "loss": 1.0259, "step": 61680 }, { "epoch": 0.15, "learning_rate": 8.473476269565656e-05, "loss": 1.0261, "step": 61685 }, { "epoch": 0.15, "learning_rate": 8.473350445417485e-05, "loss": 1.0254, "step": 61690 }, { "epoch": 0.15, "learning_rate": 8.473224621269314e-05, "loss": 1.0262, "step": 61695 }, { "epoch": 0.15, "learning_rate": 8.473098797121144e-05, "loss": 1.0252, "step": 61700 }, { "epoch": 0.15, "learning_rate": 8.472972972972974e-05, "loss": 1.0267, "step": 61705 }, { "epoch": 0.15, "learning_rate": 8.472847148824803e-05, "loss": 1.0258, "step": 61710 }, { "epoch": 0.15, "learning_rate": 8.472721324676632e-05, "loss": 1.027, "step": 61715 }, { "epoch": 0.15, "learning_rate": 8.472595500528462e-05, "loss": 1.0288, "step": 61720 }, { "epoch": 0.15, "learning_rate": 8.472469676380292e-05, "loss": 1.0258, "step": 61725 }, { "epoch": 0.15, "learning_rate": 8.47234385223212e-05, "loss": 1.0265, "step": 61730 }, { "epoch": 0.15, "learning_rate": 8.47221802808395e-05, "loss": 1.0243, "step": 61735 }, { "epoch": 0.15, "learning_rate": 8.47209220393578e-05, "loss": 1.0275, "step": 61740 }, { "epoch": 0.15, "learning_rate": 8.47196637978761e-05, "loss": 1.0272, "step": 61745 }, { "epoch": 0.16, "learning_rate": 8.471840555639439e-05, "loss": 1.0266, "step": 61750 }, { "epoch": 0.16, "learning_rate": 8.471714731491268e-05, "loss": 1.0266, "step": 61755 }, { "epoch": 0.16, "learning_rate": 8.471588907343097e-05, "loss": 1.0264, "step": 61760 }, { "epoch": 0.16, "learning_rate": 8.471463083194927e-05, "loss": 1.0239, "step": 61765 }, { "epoch": 0.16, "learning_rate": 8.471337259046757e-05, "loss": 1.0248, "step": 61770 }, { "epoch": 0.16, "learning_rate": 8.471211434898586e-05, "loss": 1.028, "step": 61775 }, { "epoch": 0.16, "learning_rate": 8.471085610750415e-05, "loss": 1.0684, "step": 61780 }, { "epoch": 0.16, "learning_rate": 8.470959786602245e-05, "loss": 1.0283, "step": 61785 }, { "epoch": 0.16, "learning_rate": 8.470833962454075e-05, "loss": 1.0286, "step": 61790 }, { "epoch": 0.16, "learning_rate": 8.470708138305904e-05, "loss": 1.0286, "step": 61795 }, { "epoch": 0.16, "learning_rate": 8.470582314157733e-05, "loss": 1.0259, "step": 61800 }, { "epoch": 0.16, "learning_rate": 8.470456490009563e-05, "loss": 1.0257, "step": 61805 }, { "epoch": 0.16, "learning_rate": 8.470330665861393e-05, "loss": 1.0274, "step": 61810 }, { "epoch": 0.16, "learning_rate": 8.470204841713222e-05, "loss": 1.0272, "step": 61815 }, { "epoch": 0.16, "learning_rate": 8.470079017565051e-05, "loss": 1.0241, "step": 61820 }, { "epoch": 0.16, "learning_rate": 8.46995319341688e-05, "loss": 1.0242, "step": 61825 }, { "epoch": 0.16, "learning_rate": 8.46982736926871e-05, "loss": 1.0282, "step": 61830 }, { "epoch": 0.16, "learning_rate": 8.46970154512054e-05, "loss": 1.0276, "step": 61835 }, { "epoch": 0.16, "learning_rate": 8.469575720972369e-05, "loss": 1.0268, "step": 61840 }, { "epoch": 0.16, "learning_rate": 8.469449896824198e-05, "loss": 1.0236, "step": 61845 }, { "epoch": 0.16, "learning_rate": 8.469324072676029e-05, "loss": 1.026, "step": 61850 }, { "epoch": 0.16, "learning_rate": 8.469198248527858e-05, "loss": 1.0268, "step": 61855 }, { "epoch": 0.16, "learning_rate": 8.469072424379687e-05, "loss": 1.0254, "step": 61860 }, { "epoch": 0.16, "learning_rate": 8.468946600231516e-05, "loss": 1.0248, "step": 61865 }, { "epoch": 0.16, "learning_rate": 8.468820776083347e-05, "loss": 1.0249, "step": 61870 }, { "epoch": 0.16, "learning_rate": 8.468694951935176e-05, "loss": 1.0269, "step": 61875 }, { "epoch": 0.16, "learning_rate": 8.468569127787005e-05, "loss": 1.0271, "step": 61880 }, { "epoch": 0.16, "learning_rate": 8.468443303638834e-05, "loss": 1.0282, "step": 61885 }, { "epoch": 0.16, "learning_rate": 8.468317479490663e-05, "loss": 1.0261, "step": 61890 }, { "epoch": 0.16, "learning_rate": 8.468191655342494e-05, "loss": 1.0255, "step": 61895 }, { "epoch": 0.16, "learning_rate": 8.468065831194323e-05, "loss": 1.0283, "step": 61900 }, { "epoch": 0.16, "learning_rate": 8.467940007046152e-05, "loss": 1.0272, "step": 61905 }, { "epoch": 0.16, "learning_rate": 8.467814182897981e-05, "loss": 1.0277, "step": 61910 }, { "epoch": 0.16, "learning_rate": 8.467688358749812e-05, "loss": 1.0272, "step": 61915 }, { "epoch": 0.16, "learning_rate": 8.467562534601641e-05, "loss": 1.0283, "step": 61920 }, { "epoch": 0.16, "learning_rate": 8.46743671045347e-05, "loss": 1.0464, "step": 61925 }, { "epoch": 0.16, "learning_rate": 8.467310886305299e-05, "loss": 1.0265, "step": 61930 }, { "epoch": 0.16, "learning_rate": 8.46718506215713e-05, "loss": 1.0265, "step": 61935 }, { "epoch": 0.16, "learning_rate": 8.467059238008959e-05, "loss": 1.0282, "step": 61940 }, { "epoch": 0.16, "learning_rate": 8.466933413860788e-05, "loss": 1.0258, "step": 61945 }, { "epoch": 0.16, "learning_rate": 8.466807589712617e-05, "loss": 1.0274, "step": 61950 }, { "epoch": 0.16, "learning_rate": 8.466681765564446e-05, "loss": 1.0255, "step": 61955 }, { "epoch": 0.16, "learning_rate": 8.466555941416277e-05, "loss": 1.0272, "step": 61960 }, { "epoch": 0.16, "learning_rate": 8.466430117268106e-05, "loss": 1.0247, "step": 61965 }, { "epoch": 0.16, "learning_rate": 8.466304293119935e-05, "loss": 1.0246, "step": 61970 }, { "epoch": 0.16, "learning_rate": 8.466178468971764e-05, "loss": 1.0277, "step": 61975 }, { "epoch": 0.16, "learning_rate": 8.466052644823595e-05, "loss": 1.0271, "step": 61980 }, { "epoch": 0.16, "learning_rate": 8.465926820675425e-05, "loss": 1.0299, "step": 61985 }, { "epoch": 0.16, "learning_rate": 8.465800996527255e-05, "loss": 1.027, "step": 61990 }, { "epoch": 0.16, "learning_rate": 8.465675172379084e-05, "loss": 1.0245, "step": 61995 }, { "epoch": 0.16, "learning_rate": 8.465549348230913e-05, "loss": 1.0266, "step": 62000 }, { "epoch": 0.16, "learning_rate": 8.465423524082743e-05, "loss": 1.0261, "step": 62005 }, { "epoch": 0.16, "learning_rate": 8.465297699934573e-05, "loss": 1.0247, "step": 62010 }, { "epoch": 0.16, "learning_rate": 8.465171875786402e-05, "loss": 1.0261, "step": 62015 }, { "epoch": 0.16, "learning_rate": 8.465046051638231e-05, "loss": 1.0262, "step": 62020 }, { "epoch": 0.16, "learning_rate": 8.464920227490061e-05, "loss": 1.0264, "step": 62025 }, { "epoch": 0.16, "learning_rate": 8.46479440334189e-05, "loss": 1.025, "step": 62030 }, { "epoch": 0.16, "learning_rate": 8.46466857919372e-05, "loss": 1.0278, "step": 62035 }, { "epoch": 0.16, "learning_rate": 8.464542755045549e-05, "loss": 1.0273, "step": 62040 }, { "epoch": 0.16, "learning_rate": 8.464416930897378e-05, "loss": 1.027, "step": 62045 }, { "epoch": 0.16, "learning_rate": 8.464291106749209e-05, "loss": 1.0271, "step": 62050 }, { "epoch": 0.16, "learning_rate": 8.464165282601038e-05, "loss": 1.0256, "step": 62055 }, { "epoch": 0.16, "learning_rate": 8.464039458452867e-05, "loss": 1.0267, "step": 62060 }, { "epoch": 0.16, "learning_rate": 8.463913634304696e-05, "loss": 1.0261, "step": 62065 }, { "epoch": 0.16, "learning_rate": 8.463787810156527e-05, "loss": 1.028, "step": 62070 }, { "epoch": 0.16, "learning_rate": 8.463661986008356e-05, "loss": 1.0273, "step": 62075 }, { "epoch": 0.16, "learning_rate": 8.463536161860185e-05, "loss": 1.0268, "step": 62080 }, { "epoch": 0.16, "learning_rate": 8.463410337712014e-05, "loss": 1.0271, "step": 62085 }, { "epoch": 0.16, "learning_rate": 8.463284513563845e-05, "loss": 1.0278, "step": 62090 }, { "epoch": 0.16, "learning_rate": 8.463158689415674e-05, "loss": 1.0244, "step": 62095 }, { "epoch": 0.16, "learning_rate": 8.463032865267503e-05, "loss": 1.0277, "step": 62100 }, { "epoch": 0.16, "learning_rate": 8.462907041119332e-05, "loss": 1.0262, "step": 62105 }, { "epoch": 0.16, "learning_rate": 8.462781216971161e-05, "loss": 1.0259, "step": 62110 }, { "epoch": 0.16, "learning_rate": 8.462655392822992e-05, "loss": 1.0274, "step": 62115 }, { "epoch": 0.16, "learning_rate": 8.462529568674821e-05, "loss": 1.0264, "step": 62120 }, { "epoch": 0.16, "learning_rate": 8.46240374452665e-05, "loss": 1.0279, "step": 62125 }, { "epoch": 0.16, "learning_rate": 8.462277920378479e-05, "loss": 1.0257, "step": 62130 }, { "epoch": 0.16, "learning_rate": 8.46215209623031e-05, "loss": 1.0248, "step": 62135 }, { "epoch": 0.16, "learning_rate": 8.462026272082139e-05, "loss": 1.0267, "step": 62140 }, { "epoch": 0.16, "learning_rate": 8.461900447933968e-05, "loss": 1.0269, "step": 62145 }, { "epoch": 0.16, "learning_rate": 8.461774623785797e-05, "loss": 1.0262, "step": 62150 }, { "epoch": 0.16, "learning_rate": 8.461648799637626e-05, "loss": 1.027, "step": 62155 }, { "epoch": 0.16, "learning_rate": 8.461522975489457e-05, "loss": 1.0269, "step": 62160 }, { "epoch": 0.16, "learning_rate": 8.461397151341286e-05, "loss": 1.0241, "step": 62165 }, { "epoch": 0.16, "learning_rate": 8.461271327193115e-05, "loss": 1.0264, "step": 62170 }, { "epoch": 0.16, "learning_rate": 8.461145503044944e-05, "loss": 1.0269, "step": 62175 }, { "epoch": 0.16, "learning_rate": 8.461019678896775e-05, "loss": 1.0245, "step": 62180 }, { "epoch": 0.16, "learning_rate": 8.460893854748604e-05, "loss": 1.0277, "step": 62185 }, { "epoch": 0.16, "learning_rate": 8.460768030600433e-05, "loss": 1.0252, "step": 62190 }, { "epoch": 0.16, "learning_rate": 8.460642206452262e-05, "loss": 1.0274, "step": 62195 }, { "epoch": 0.16, "learning_rate": 8.460516382304093e-05, "loss": 1.0265, "step": 62200 }, { "epoch": 0.16, "learning_rate": 8.460390558155922e-05, "loss": 1.0276, "step": 62205 }, { "epoch": 0.16, "learning_rate": 8.460264734007751e-05, "loss": 1.025, "step": 62210 }, { "epoch": 0.16, "learning_rate": 8.46013890985958e-05, "loss": 1.0266, "step": 62215 }, { "epoch": 0.16, "learning_rate": 8.46001308571141e-05, "loss": 1.0227, "step": 62220 }, { "epoch": 0.16, "learning_rate": 8.45988726156324e-05, "loss": 1.0252, "step": 62225 }, { "epoch": 0.16, "learning_rate": 8.459761437415069e-05, "loss": 1.0241, "step": 62230 }, { "epoch": 0.16, "learning_rate": 8.459635613266898e-05, "loss": 1.0266, "step": 62235 }, { "epoch": 0.16, "learning_rate": 8.459509789118727e-05, "loss": 1.0273, "step": 62240 }, { "epoch": 0.16, "learning_rate": 8.459383964970558e-05, "loss": 1.0259, "step": 62245 }, { "epoch": 0.16, "learning_rate": 8.459258140822387e-05, "loss": 1.0267, "step": 62250 }, { "epoch": 0.16, "learning_rate": 8.459132316674216e-05, "loss": 1.0269, "step": 62255 }, { "epoch": 0.16, "learning_rate": 8.459006492526045e-05, "loss": 1.0265, "step": 62260 }, { "epoch": 0.16, "learning_rate": 8.458880668377876e-05, "loss": 1.0261, "step": 62265 }, { "epoch": 0.16, "learning_rate": 8.458754844229705e-05, "loss": 1.0257, "step": 62270 }, { "epoch": 0.16, "learning_rate": 8.458629020081534e-05, "loss": 1.0285, "step": 62275 }, { "epoch": 0.16, "learning_rate": 8.458503195933363e-05, "loss": 1.0241, "step": 62280 }, { "epoch": 0.16, "learning_rate": 8.458377371785193e-05, "loss": 1.0289, "step": 62285 }, { "epoch": 0.16, "learning_rate": 8.458251547637023e-05, "loss": 1.0272, "step": 62290 }, { "epoch": 0.16, "learning_rate": 8.458125723488852e-05, "loss": 1.027, "step": 62295 }, { "epoch": 0.16, "learning_rate": 8.457999899340681e-05, "loss": 1.0274, "step": 62300 }, { "epoch": 0.16, "learning_rate": 8.45787407519251e-05, "loss": 1.0279, "step": 62305 }, { "epoch": 0.16, "learning_rate": 8.457748251044341e-05, "loss": 1.0282, "step": 62310 }, { "epoch": 0.16, "learning_rate": 8.45762242689617e-05, "loss": 1.0261, "step": 62315 }, { "epoch": 0.16, "learning_rate": 8.457496602748e-05, "loss": 1.0254, "step": 62320 }, { "epoch": 0.16, "learning_rate": 8.457370778599829e-05, "loss": 1.0281, "step": 62325 }, { "epoch": 0.16, "learning_rate": 8.457244954451659e-05, "loss": 1.0273, "step": 62330 }, { "epoch": 0.16, "learning_rate": 8.457119130303488e-05, "loss": 1.0254, "step": 62335 }, { "epoch": 0.16, "learning_rate": 8.456993306155317e-05, "loss": 1.0271, "step": 62340 }, { "epoch": 0.16, "learning_rate": 8.456867482007147e-05, "loss": 1.0254, "step": 62345 }, { "epoch": 0.16, "learning_rate": 8.456741657858976e-05, "loss": 1.026, "step": 62350 }, { "epoch": 0.16, "learning_rate": 8.456615833710806e-05, "loss": 1.0264, "step": 62355 }, { "epoch": 0.16, "learning_rate": 8.456490009562635e-05, "loss": 1.0265, "step": 62360 }, { "epoch": 0.16, "learning_rate": 8.456364185414465e-05, "loss": 1.026, "step": 62365 }, { "epoch": 0.16, "learning_rate": 8.456238361266294e-05, "loss": 1.0273, "step": 62370 }, { "epoch": 0.16, "learning_rate": 8.456112537118124e-05, "loss": 1.0293, "step": 62375 }, { "epoch": 0.16, "learning_rate": 8.455986712969953e-05, "loss": 1.0275, "step": 62380 }, { "epoch": 0.16, "learning_rate": 8.455860888821783e-05, "loss": 1.0283, "step": 62385 }, { "epoch": 0.16, "learning_rate": 8.455735064673612e-05, "loss": 1.0268, "step": 62390 }, { "epoch": 0.16, "learning_rate": 8.455609240525442e-05, "loss": 1.0283, "step": 62395 }, { "epoch": 0.16, "learning_rate": 8.455483416377271e-05, "loss": 1.0253, "step": 62400 }, { "epoch": 0.16, "learning_rate": 8.4553575922291e-05, "loss": 1.0272, "step": 62405 }, { "epoch": 0.16, "learning_rate": 8.45523176808093e-05, "loss": 1.0512, "step": 62410 }, { "epoch": 0.16, "learning_rate": 8.455105943932759e-05, "loss": 1.0259, "step": 62415 }, { "epoch": 0.16, "learning_rate": 8.45498011978459e-05, "loss": 1.0247, "step": 62420 }, { "epoch": 0.16, "learning_rate": 8.454854295636419e-05, "loss": 1.0425, "step": 62425 }, { "epoch": 0.16, "learning_rate": 8.454728471488248e-05, "loss": 1.0284, "step": 62430 }, { "epoch": 0.16, "learning_rate": 8.454602647340077e-05, "loss": 1.028, "step": 62435 }, { "epoch": 0.16, "learning_rate": 8.454476823191907e-05, "loss": 1.026, "step": 62440 }, { "epoch": 0.16, "learning_rate": 8.454350999043737e-05, "loss": 1.0278, "step": 62445 }, { "epoch": 0.16, "learning_rate": 8.454225174895566e-05, "loss": 1.028, "step": 62450 }, { "epoch": 0.16, "learning_rate": 8.454099350747395e-05, "loss": 1.0274, "step": 62455 }, { "epoch": 0.16, "learning_rate": 8.453973526599225e-05, "loss": 1.0274, "step": 62460 }, { "epoch": 0.16, "learning_rate": 8.453847702451055e-05, "loss": 1.0271, "step": 62465 }, { "epoch": 0.16, "learning_rate": 8.453721878302884e-05, "loss": 1.0266, "step": 62470 }, { "epoch": 0.16, "learning_rate": 8.453596054154713e-05, "loss": 1.0244, "step": 62475 }, { "epoch": 0.16, "learning_rate": 8.453470230006543e-05, "loss": 1.0255, "step": 62480 }, { "epoch": 0.16, "learning_rate": 8.453344405858374e-05, "loss": 1.0274, "step": 62485 }, { "epoch": 0.16, "learning_rate": 8.453218581710203e-05, "loss": 1.026, "step": 62490 }, { "epoch": 0.16, "learning_rate": 8.453092757562032e-05, "loss": 1.0285, "step": 62495 }, { "epoch": 0.16, "learning_rate": 8.452966933413861e-05, "loss": 1.0285, "step": 62500 }, { "epoch": 0.16, "learning_rate": 8.45284110926569e-05, "loss": 1.0275, "step": 62505 }, { "epoch": 0.16, "learning_rate": 8.452715285117521e-05, "loss": 1.0252, "step": 62510 }, { "epoch": 0.16, "learning_rate": 8.45258946096935e-05, "loss": 1.0284, "step": 62515 }, { "epoch": 0.16, "learning_rate": 8.45246363682118e-05, "loss": 1.0262, "step": 62520 }, { "epoch": 0.16, "learning_rate": 8.452337812673009e-05, "loss": 1.0251, "step": 62525 }, { "epoch": 0.16, "learning_rate": 8.452211988524839e-05, "loss": 1.0251, "step": 62530 }, { "epoch": 0.16, "learning_rate": 8.452086164376668e-05, "loss": 1.0269, "step": 62535 }, { "epoch": 0.16, "learning_rate": 8.451960340228497e-05, "loss": 1.0271, "step": 62540 }, { "epoch": 0.16, "learning_rate": 8.451834516080327e-05, "loss": 1.0259, "step": 62545 }, { "epoch": 0.16, "learning_rate": 8.451708691932157e-05, "loss": 1.0256, "step": 62550 }, { "epoch": 0.16, "learning_rate": 8.451582867783986e-05, "loss": 1.0262, "step": 62555 }, { "epoch": 0.16, "learning_rate": 8.451457043635815e-05, "loss": 1.0243, "step": 62560 }, { "epoch": 0.16, "learning_rate": 8.451331219487645e-05, "loss": 1.0285, "step": 62565 }, { "epoch": 0.16, "learning_rate": 8.451205395339474e-05, "loss": 1.0268, "step": 62570 }, { "epoch": 0.16, "learning_rate": 8.451079571191304e-05, "loss": 1.0256, "step": 62575 }, { "epoch": 0.16, "learning_rate": 8.450953747043133e-05, "loss": 1.0262, "step": 62580 }, { "epoch": 0.16, "learning_rate": 8.450827922894963e-05, "loss": 1.0245, "step": 62585 }, { "epoch": 0.16, "learning_rate": 8.450702098746792e-05, "loss": 1.0271, "step": 62590 }, { "epoch": 0.16, "learning_rate": 8.450576274598622e-05, "loss": 1.0283, "step": 62595 }, { "epoch": 0.16, "learning_rate": 8.450450450450451e-05, "loss": 1.0259, "step": 62600 }, { "epoch": 0.16, "learning_rate": 8.45032462630228e-05, "loss": 1.0279, "step": 62605 }, { "epoch": 0.16, "learning_rate": 8.45019880215411e-05, "loss": 1.0255, "step": 62610 }, { "epoch": 0.16, "learning_rate": 8.45007297800594e-05, "loss": 1.0259, "step": 62615 }, { "epoch": 0.16, "learning_rate": 8.44994715385777e-05, "loss": 1.0244, "step": 62620 }, { "epoch": 0.16, "learning_rate": 8.449821329709599e-05, "loss": 1.0279, "step": 62625 }, { "epoch": 0.16, "learning_rate": 8.449695505561428e-05, "loss": 1.0275, "step": 62630 }, { "epoch": 0.16, "learning_rate": 8.449569681413257e-05, "loss": 1.0247, "step": 62635 }, { "epoch": 0.16, "learning_rate": 8.449443857265087e-05, "loss": 1.025, "step": 62640 }, { "epoch": 0.16, "learning_rate": 8.449318033116917e-05, "loss": 1.0249, "step": 62645 }, { "epoch": 0.16, "learning_rate": 8.449192208968746e-05, "loss": 1.0254, "step": 62650 }, { "epoch": 0.16, "learning_rate": 8.449066384820575e-05, "loss": 1.0265, "step": 62655 }, { "epoch": 0.16, "learning_rate": 8.448940560672405e-05, "loss": 1.0275, "step": 62660 }, { "epoch": 0.16, "learning_rate": 8.448814736524235e-05, "loss": 1.0248, "step": 62665 }, { "epoch": 0.16, "learning_rate": 8.448688912376064e-05, "loss": 1.0271, "step": 62670 }, { "epoch": 0.16, "learning_rate": 8.448563088227893e-05, "loss": 1.0291, "step": 62675 }, { "epoch": 0.16, "learning_rate": 8.448437264079723e-05, "loss": 1.0264, "step": 62680 }, { "epoch": 0.16, "learning_rate": 8.448311439931553e-05, "loss": 1.0267, "step": 62685 }, { "epoch": 0.16, "learning_rate": 8.448185615783382e-05, "loss": 1.0263, "step": 62690 }, { "epoch": 0.16, "learning_rate": 8.448059791635211e-05, "loss": 1.0258, "step": 62695 }, { "epoch": 0.16, "learning_rate": 8.44793396748704e-05, "loss": 1.0268, "step": 62700 }, { "epoch": 0.16, "learning_rate": 8.44780814333887e-05, "loss": 1.0251, "step": 62705 }, { "epoch": 0.16, "learning_rate": 8.4476823191907e-05, "loss": 1.0276, "step": 62710 }, { "epoch": 0.16, "learning_rate": 8.447556495042529e-05, "loss": 1.0249, "step": 62715 }, { "epoch": 0.16, "learning_rate": 8.447430670894358e-05, "loss": 1.027, "step": 62720 }, { "epoch": 0.16, "learning_rate": 8.447304846746189e-05, "loss": 1.0241, "step": 62725 }, { "epoch": 0.16, "learning_rate": 8.447179022598018e-05, "loss": 1.024, "step": 62730 }, { "epoch": 0.16, "learning_rate": 8.447053198449847e-05, "loss": 1.0265, "step": 62735 }, { "epoch": 0.16, "learning_rate": 8.446927374301676e-05, "loss": 1.0262, "step": 62740 }, { "epoch": 0.16, "learning_rate": 8.446801550153507e-05, "loss": 1.0275, "step": 62745 }, { "epoch": 0.16, "learning_rate": 8.446675726005336e-05, "loss": 1.0253, "step": 62750 }, { "epoch": 0.16, "learning_rate": 8.446549901857165e-05, "loss": 1.0276, "step": 62755 }, { "epoch": 0.16, "learning_rate": 8.446424077708994e-05, "loss": 1.0259, "step": 62760 }, { "epoch": 0.16, "learning_rate": 8.446298253560823e-05, "loss": 1.0262, "step": 62765 }, { "epoch": 0.16, "learning_rate": 8.446172429412654e-05, "loss": 1.0244, "step": 62770 }, { "epoch": 0.16, "learning_rate": 8.446046605264483e-05, "loss": 1.0262, "step": 62775 }, { "epoch": 0.16, "learning_rate": 8.445920781116312e-05, "loss": 1.0278, "step": 62780 }, { "epoch": 0.16, "learning_rate": 8.445794956968141e-05, "loss": 1.024, "step": 62785 }, { "epoch": 0.16, "learning_rate": 8.445669132819972e-05, "loss": 1.0249, "step": 62790 }, { "epoch": 0.16, "learning_rate": 8.445543308671801e-05, "loss": 1.0267, "step": 62795 }, { "epoch": 0.16, "learning_rate": 8.44541748452363e-05, "loss": 1.0246, "step": 62800 }, { "epoch": 0.16, "learning_rate": 8.445291660375459e-05, "loss": 1.029, "step": 62805 }, { "epoch": 0.16, "learning_rate": 8.445165836227288e-05, "loss": 1.0264, "step": 62810 }, { "epoch": 0.16, "learning_rate": 8.445040012079119e-05, "loss": 1.0246, "step": 62815 }, { "epoch": 0.16, "learning_rate": 8.444914187930948e-05, "loss": 1.0259, "step": 62820 }, { "epoch": 0.16, "learning_rate": 8.444788363782777e-05, "loss": 1.0262, "step": 62825 }, { "epoch": 0.16, "learning_rate": 8.444662539634606e-05, "loss": 1.0268, "step": 62830 }, { "epoch": 0.16, "learning_rate": 8.444536715486437e-05, "loss": 1.0257, "step": 62835 }, { "epoch": 0.16, "learning_rate": 8.444410891338266e-05, "loss": 1.0277, "step": 62840 }, { "epoch": 0.16, "learning_rate": 8.444285067190095e-05, "loss": 1.0255, "step": 62845 }, { "epoch": 0.16, "learning_rate": 8.444159243041924e-05, "loss": 1.023, "step": 62850 }, { "epoch": 0.16, "learning_rate": 8.444033418893755e-05, "loss": 1.044, "step": 62855 }, { "epoch": 0.16, "learning_rate": 8.443907594745584e-05, "loss": 1.0264, "step": 62860 }, { "epoch": 0.16, "learning_rate": 8.443781770597413e-05, "loss": 1.0272, "step": 62865 }, { "epoch": 0.16, "learning_rate": 8.443655946449242e-05, "loss": 1.027, "step": 62870 }, { "epoch": 0.16, "learning_rate": 8.443530122301071e-05, "loss": 1.0263, "step": 62875 }, { "epoch": 0.16, "learning_rate": 8.443404298152902e-05, "loss": 1.0257, "step": 62880 }, { "epoch": 0.16, "learning_rate": 8.443278474004731e-05, "loss": 1.0251, "step": 62885 }, { "epoch": 0.16, "learning_rate": 8.44315264985656e-05, "loss": 1.0253, "step": 62890 }, { "epoch": 0.16, "learning_rate": 8.44302682570839e-05, "loss": 1.0255, "step": 62895 }, { "epoch": 0.16, "learning_rate": 8.44290100156022e-05, "loss": 1.0254, "step": 62900 }, { "epoch": 0.16, "learning_rate": 8.442775177412049e-05, "loss": 1.0272, "step": 62905 }, { "epoch": 0.16, "learning_rate": 8.442649353263878e-05, "loss": 1.026, "step": 62910 }, { "epoch": 0.16, "learning_rate": 8.442523529115707e-05, "loss": 1.0281, "step": 62915 }, { "epoch": 0.16, "learning_rate": 8.442397704967538e-05, "loss": 1.0251, "step": 62920 }, { "epoch": 0.16, "learning_rate": 8.442271880819367e-05, "loss": 1.0278, "step": 62925 }, { "epoch": 0.16, "learning_rate": 8.442146056671196e-05, "loss": 1.0278, "step": 62930 }, { "epoch": 0.16, "learning_rate": 8.442020232523025e-05, "loss": 1.0257, "step": 62935 }, { "epoch": 0.16, "learning_rate": 8.441894408374855e-05, "loss": 1.0264, "step": 62940 }, { "epoch": 0.16, "learning_rate": 8.441768584226685e-05, "loss": 1.0261, "step": 62945 }, { "epoch": 0.16, "learning_rate": 8.441642760078514e-05, "loss": 1.0258, "step": 62950 }, { "epoch": 0.16, "learning_rate": 8.441516935930343e-05, "loss": 1.0276, "step": 62955 }, { "epoch": 0.16, "learning_rate": 8.441391111782173e-05, "loss": 1.0279, "step": 62960 }, { "epoch": 0.16, "learning_rate": 8.441265287634003e-05, "loss": 1.0263, "step": 62965 }, { "epoch": 0.16, "learning_rate": 8.441139463485832e-05, "loss": 1.0254, "step": 62970 }, { "epoch": 0.16, "learning_rate": 8.441013639337661e-05, "loss": 1.0475, "step": 62975 }, { "epoch": 0.16, "learning_rate": 8.440887815189492e-05, "loss": 1.0267, "step": 62980 }, { "epoch": 0.16, "learning_rate": 8.440761991041321e-05, "loss": 1.0273, "step": 62985 }, { "epoch": 0.16, "learning_rate": 8.440636166893152e-05, "loss": 1.0266, "step": 62990 }, { "epoch": 0.16, "learning_rate": 8.440510342744981e-05, "loss": 1.0259, "step": 62995 }, { "epoch": 0.16, "learning_rate": 8.44038451859681e-05, "loss": 1.0265, "step": 63000 }, { "epoch": 0.16, "learning_rate": 8.440258694448639e-05, "loss": 1.0275, "step": 63005 }, { "epoch": 0.16, "learning_rate": 8.44013287030047e-05, "loss": 1.0251, "step": 63010 }, { "epoch": 0.16, "learning_rate": 8.440007046152299e-05, "loss": 1.0261, "step": 63015 }, { "epoch": 0.16, "learning_rate": 8.439881222004128e-05, "loss": 1.0259, "step": 63020 }, { "epoch": 0.16, "learning_rate": 8.439755397855957e-05, "loss": 1.0275, "step": 63025 }, { "epoch": 0.16, "learning_rate": 8.439629573707786e-05, "loss": 1.0251, "step": 63030 }, { "epoch": 0.16, "learning_rate": 8.439503749559617e-05, "loss": 1.0271, "step": 63035 }, { "epoch": 0.16, "learning_rate": 8.439377925411446e-05, "loss": 1.028, "step": 63040 }, { "epoch": 0.16, "learning_rate": 8.439252101263275e-05, "loss": 1.049, "step": 63045 }, { "epoch": 0.16, "learning_rate": 8.439126277115104e-05, "loss": 1.0265, "step": 63050 }, { "epoch": 0.16, "learning_rate": 8.439000452966935e-05, "loss": 1.0243, "step": 63055 }, { "epoch": 0.16, "learning_rate": 8.438874628818764e-05, "loss": 1.0251, "step": 63060 }, { "epoch": 0.16, "learning_rate": 8.438748804670593e-05, "loss": 1.026, "step": 63065 }, { "epoch": 0.16, "learning_rate": 8.438622980522422e-05, "loss": 1.0244, "step": 63070 }, { "epoch": 0.16, "learning_rate": 8.438497156374253e-05, "loss": 1.026, "step": 63075 }, { "epoch": 0.16, "learning_rate": 8.438371332226082e-05, "loss": 1.0255, "step": 63080 }, { "epoch": 0.16, "learning_rate": 8.438245508077911e-05, "loss": 1.0267, "step": 63085 }, { "epoch": 0.16, "learning_rate": 8.43811968392974e-05, "loss": 1.0282, "step": 63090 }, { "epoch": 0.16, "learning_rate": 8.43799385978157e-05, "loss": 1.024, "step": 63095 }, { "epoch": 0.16, "learning_rate": 8.4378680356334e-05, "loss": 1.0252, "step": 63100 }, { "epoch": 0.16, "learning_rate": 8.437742211485229e-05, "loss": 1.0253, "step": 63105 }, { "epoch": 0.16, "learning_rate": 8.437616387337058e-05, "loss": 1.028, "step": 63110 }, { "epoch": 0.16, "learning_rate": 8.437490563188887e-05, "loss": 1.0262, "step": 63115 }, { "epoch": 0.16, "learning_rate": 8.437364739040718e-05, "loss": 1.026, "step": 63120 }, { "epoch": 0.16, "learning_rate": 8.437238914892547e-05, "loss": 1.0246, "step": 63125 }, { "epoch": 0.16, "learning_rate": 8.437113090744376e-05, "loss": 1.0245, "step": 63130 }, { "epoch": 0.16, "learning_rate": 8.436987266596205e-05, "loss": 1.0459, "step": 63135 }, { "epoch": 0.16, "learning_rate": 8.436861442448036e-05, "loss": 1.0229, "step": 63140 }, { "epoch": 0.16, "learning_rate": 8.436735618299865e-05, "loss": 1.0264, "step": 63145 }, { "epoch": 0.16, "learning_rate": 8.436609794151694e-05, "loss": 1.0261, "step": 63150 }, { "epoch": 0.16, "learning_rate": 8.436483970003523e-05, "loss": 1.0283, "step": 63155 }, { "epoch": 0.16, "learning_rate": 8.436358145855353e-05, "loss": 1.0252, "step": 63160 }, { "epoch": 0.16, "learning_rate": 8.436232321707183e-05, "loss": 1.0239, "step": 63165 }, { "epoch": 0.16, "learning_rate": 8.436106497559012e-05, "loss": 1.0272, "step": 63170 }, { "epoch": 0.16, "learning_rate": 8.435980673410841e-05, "loss": 1.0257, "step": 63175 }, { "epoch": 0.16, "learning_rate": 8.43585484926267e-05, "loss": 1.0263, "step": 63180 }, { "epoch": 0.16, "learning_rate": 8.435729025114501e-05, "loss": 1.0277, "step": 63185 }, { "epoch": 0.16, "learning_rate": 8.43560320096633e-05, "loss": 1.0282, "step": 63190 }, { "epoch": 0.16, "learning_rate": 8.43547737681816e-05, "loss": 1.0261, "step": 63195 }, { "epoch": 0.16, "learning_rate": 8.435351552669988e-05, "loss": 1.0273, "step": 63200 }, { "epoch": 0.16, "learning_rate": 8.435225728521819e-05, "loss": 1.0266, "step": 63205 }, { "epoch": 0.16, "learning_rate": 8.435099904373648e-05, "loss": 1.0265, "step": 63210 }, { "epoch": 0.16, "learning_rate": 8.434974080225477e-05, "loss": 1.0266, "step": 63215 }, { "epoch": 0.16, "learning_rate": 8.434848256077306e-05, "loss": 1.0259, "step": 63220 }, { "epoch": 0.16, "learning_rate": 8.434722431929136e-05, "loss": 1.0254, "step": 63225 }, { "epoch": 0.16, "learning_rate": 8.434596607780966e-05, "loss": 1.027, "step": 63230 }, { "epoch": 0.16, "learning_rate": 8.434470783632795e-05, "loss": 1.024, "step": 63235 }, { "epoch": 0.16, "learning_rate": 8.434344959484624e-05, "loss": 1.0272, "step": 63240 }, { "epoch": 0.16, "learning_rate": 8.434219135336454e-05, "loss": 1.028, "step": 63245 }, { "epoch": 0.16, "learning_rate": 8.434093311188284e-05, "loss": 1.026, "step": 63250 }, { "epoch": 0.16, "learning_rate": 8.433967487040113e-05, "loss": 1.0276, "step": 63255 }, { "epoch": 0.16, "learning_rate": 8.433841662891942e-05, "loss": 1.0249, "step": 63260 }, { "epoch": 0.16, "learning_rate": 8.433715838743772e-05, "loss": 1.0245, "step": 63265 }, { "epoch": 0.16, "learning_rate": 8.433590014595602e-05, "loss": 1.0263, "step": 63270 }, { "epoch": 0.16, "learning_rate": 8.433464190447431e-05, "loss": 1.0262, "step": 63275 }, { "epoch": 0.16, "learning_rate": 8.43333836629926e-05, "loss": 1.0259, "step": 63280 }, { "epoch": 0.16, "learning_rate": 8.43321254215109e-05, "loss": 1.0247, "step": 63285 }, { "epoch": 0.16, "learning_rate": 8.433086718002919e-05, "loss": 1.0244, "step": 63290 }, { "epoch": 0.16, "learning_rate": 8.432960893854749e-05, "loss": 1.0262, "step": 63295 }, { "epoch": 0.16, "learning_rate": 8.432835069706578e-05, "loss": 1.0275, "step": 63300 }, { "epoch": 0.16, "learning_rate": 8.432709245558408e-05, "loss": 1.0263, "step": 63305 }, { "epoch": 0.16, "learning_rate": 8.432583421410237e-05, "loss": 1.0262, "step": 63310 }, { "epoch": 0.16, "learning_rate": 8.432457597262067e-05, "loss": 1.026, "step": 63315 }, { "epoch": 0.16, "learning_rate": 8.432331773113896e-05, "loss": 1.0261, "step": 63320 }, { "epoch": 0.16, "learning_rate": 8.432205948965726e-05, "loss": 1.0267, "step": 63325 }, { "epoch": 0.16, "learning_rate": 8.432080124817555e-05, "loss": 1.028, "step": 63330 }, { "epoch": 0.16, "learning_rate": 8.431954300669385e-05, "loss": 1.0257, "step": 63335 }, { "epoch": 0.16, "learning_rate": 8.431828476521214e-05, "loss": 1.0269, "step": 63340 }, { "epoch": 0.16, "learning_rate": 8.431702652373044e-05, "loss": 1.0262, "step": 63345 }, { "epoch": 0.16, "learning_rate": 8.431576828224873e-05, "loss": 1.0266, "step": 63350 }, { "epoch": 0.16, "learning_rate": 8.431451004076702e-05, "loss": 1.027, "step": 63355 }, { "epoch": 0.16, "learning_rate": 8.431325179928532e-05, "loss": 1.0258, "step": 63360 }, { "epoch": 0.16, "learning_rate": 8.431199355780362e-05, "loss": 1.0288, "step": 63365 }, { "epoch": 0.16, "learning_rate": 8.431073531632191e-05, "loss": 1.0247, "step": 63370 }, { "epoch": 0.16, "learning_rate": 8.43094770748402e-05, "loss": 1.0238, "step": 63375 }, { "epoch": 0.16, "learning_rate": 8.43082188333585e-05, "loss": 1.0282, "step": 63380 }, { "epoch": 0.16, "learning_rate": 8.43069605918768e-05, "loss": 1.0251, "step": 63385 }, { "epoch": 0.16, "learning_rate": 8.430570235039509e-05, "loss": 1.0244, "step": 63390 }, { "epoch": 0.16, "learning_rate": 8.430444410891338e-05, "loss": 1.0271, "step": 63395 }, { "epoch": 0.16, "learning_rate": 8.430318586743168e-05, "loss": 1.0251, "step": 63400 }, { "epoch": 0.16, "learning_rate": 8.430192762594998e-05, "loss": 1.0278, "step": 63405 }, { "epoch": 0.16, "learning_rate": 8.430066938446827e-05, "loss": 1.0274, "step": 63410 }, { "epoch": 0.16, "learning_rate": 8.429941114298656e-05, "loss": 1.0239, "step": 63415 }, { "epoch": 0.16, "learning_rate": 8.429815290150485e-05, "loss": 1.0273, "step": 63420 }, { "epoch": 0.16, "learning_rate": 8.429689466002316e-05, "loss": 1.026, "step": 63425 }, { "epoch": 0.16, "learning_rate": 8.429563641854145e-05, "loss": 1.0275, "step": 63430 }, { "epoch": 0.16, "learning_rate": 8.429437817705974e-05, "loss": 1.026, "step": 63435 }, { "epoch": 0.16, "learning_rate": 8.429311993557803e-05, "loss": 1.0239, "step": 63440 }, { "epoch": 0.16, "learning_rate": 8.429186169409634e-05, "loss": 1.0263, "step": 63445 }, { "epoch": 0.16, "learning_rate": 8.429060345261463e-05, "loss": 1.0254, "step": 63450 }, { "epoch": 0.16, "learning_rate": 8.428934521113292e-05, "loss": 1.0238, "step": 63455 }, { "epoch": 0.16, "learning_rate": 8.428808696965121e-05, "loss": 1.0283, "step": 63460 }, { "epoch": 0.16, "learning_rate": 8.42868287281695e-05, "loss": 1.0245, "step": 63465 }, { "epoch": 0.16, "learning_rate": 8.428557048668781e-05, "loss": 1.0463, "step": 63470 }, { "epoch": 0.16, "learning_rate": 8.42843122452061e-05, "loss": 1.0254, "step": 63475 }, { "epoch": 0.16, "learning_rate": 8.42830540037244e-05, "loss": 1.0285, "step": 63480 }, { "epoch": 0.16, "learning_rate": 8.42817957622427e-05, "loss": 1.026, "step": 63485 }, { "epoch": 0.16, "learning_rate": 8.428053752076099e-05, "loss": 1.0237, "step": 63490 }, { "epoch": 0.16, "learning_rate": 8.427927927927929e-05, "loss": 1.0275, "step": 63495 }, { "epoch": 0.16, "learning_rate": 8.427802103779758e-05, "loss": 1.0284, "step": 63500 }, { "epoch": 0.16, "learning_rate": 8.427676279631588e-05, "loss": 1.027, "step": 63505 }, { "epoch": 0.16, "learning_rate": 8.427550455483417e-05, "loss": 1.0388, "step": 63510 }, { "epoch": 0.16, "learning_rate": 8.427424631335247e-05, "loss": 1.0252, "step": 63515 }, { "epoch": 0.16, "learning_rate": 8.427298807187076e-05, "loss": 1.0282, "step": 63520 }, { "epoch": 0.16, "learning_rate": 8.427172983038906e-05, "loss": 1.0255, "step": 63525 }, { "epoch": 0.16, "learning_rate": 8.427047158890735e-05, "loss": 1.0274, "step": 63530 }, { "epoch": 0.16, "learning_rate": 8.426921334742565e-05, "loss": 1.0247, "step": 63535 }, { "epoch": 0.16, "learning_rate": 8.426795510594394e-05, "loss": 1.0258, "step": 63540 }, { "epoch": 0.16, "learning_rate": 8.426669686446224e-05, "loss": 1.0267, "step": 63545 }, { "epoch": 0.16, "learning_rate": 8.426543862298053e-05, "loss": 1.0256, "step": 63550 }, { "epoch": 0.16, "learning_rate": 8.426418038149882e-05, "loss": 1.0272, "step": 63555 }, { "epoch": 0.16, "learning_rate": 8.426292214001712e-05, "loss": 1.0273, "step": 63560 }, { "epoch": 0.16, "learning_rate": 8.426166389853542e-05, "loss": 1.0265, "step": 63565 }, { "epoch": 0.16, "learning_rate": 8.426040565705371e-05, "loss": 1.028, "step": 63570 }, { "epoch": 0.16, "learning_rate": 8.4259147415572e-05, "loss": 1.0264, "step": 63575 }, { "epoch": 0.16, "learning_rate": 8.42578891740903e-05, "loss": 1.0247, "step": 63580 }, { "epoch": 0.16, "learning_rate": 8.42566309326086e-05, "loss": 1.047, "step": 63585 }, { "epoch": 0.16, "learning_rate": 8.425537269112689e-05, "loss": 1.026, "step": 63590 }, { "epoch": 0.16, "learning_rate": 8.425411444964518e-05, "loss": 1.0269, "step": 63595 }, { "epoch": 0.16, "learning_rate": 8.425285620816348e-05, "loss": 1.027, "step": 63600 }, { "epoch": 0.16, "learning_rate": 8.425159796668178e-05, "loss": 1.026, "step": 63605 }, { "epoch": 0.16, "learning_rate": 8.425033972520007e-05, "loss": 1.0279, "step": 63610 }, { "epoch": 0.16, "learning_rate": 8.424908148371836e-05, "loss": 1.0263, "step": 63615 }, { "epoch": 0.16, "learning_rate": 8.424782324223665e-05, "loss": 1.0235, "step": 63620 }, { "epoch": 0.16, "learning_rate": 8.424656500075496e-05, "loss": 1.026, "step": 63625 }, { "epoch": 0.16, "learning_rate": 8.424530675927325e-05, "loss": 1.0262, "step": 63630 }, { "epoch": 0.16, "learning_rate": 8.424404851779154e-05, "loss": 1.0266, "step": 63635 }, { "epoch": 0.16, "learning_rate": 8.424279027630983e-05, "loss": 1.0279, "step": 63640 }, { "epoch": 0.16, "learning_rate": 8.424153203482814e-05, "loss": 1.0277, "step": 63645 }, { "epoch": 0.16, "learning_rate": 8.424027379334643e-05, "loss": 1.0251, "step": 63650 }, { "epoch": 0.16, "learning_rate": 8.423901555186472e-05, "loss": 1.027, "step": 63655 }, { "epoch": 0.16, "learning_rate": 8.423775731038301e-05, "loss": 1.0254, "step": 63660 }, { "epoch": 0.16, "learning_rate": 8.423649906890132e-05, "loss": 1.0253, "step": 63665 }, { "epoch": 0.16, "learning_rate": 8.423524082741961e-05, "loss": 1.0277, "step": 63670 }, { "epoch": 0.16, "learning_rate": 8.42339825859379e-05, "loss": 1.0258, "step": 63675 }, { "epoch": 0.16, "learning_rate": 8.423272434445619e-05, "loss": 1.0268, "step": 63680 }, { "epoch": 0.16, "learning_rate": 8.423146610297448e-05, "loss": 1.0284, "step": 63685 }, { "epoch": 0.16, "learning_rate": 8.423020786149279e-05, "loss": 1.0261, "step": 63690 }, { "epoch": 0.16, "learning_rate": 8.422894962001108e-05, "loss": 1.027, "step": 63695 }, { "epoch": 0.16, "learning_rate": 8.422769137852937e-05, "loss": 1.0274, "step": 63700 }, { "epoch": 0.16, "learning_rate": 8.422643313704766e-05, "loss": 1.0282, "step": 63705 }, { "epoch": 0.16, "learning_rate": 8.422517489556597e-05, "loss": 1.0279, "step": 63710 }, { "epoch": 0.16, "learning_rate": 8.422391665408426e-05, "loss": 1.0247, "step": 63715 }, { "epoch": 0.16, "learning_rate": 8.422265841260255e-05, "loss": 1.0258, "step": 63720 }, { "epoch": 0.16, "learning_rate": 8.422140017112084e-05, "loss": 1.0256, "step": 63725 }, { "epoch": 0.16, "learning_rate": 8.422014192963915e-05, "loss": 1.0261, "step": 63730 }, { "epoch": 0.16, "learning_rate": 8.421888368815744e-05, "loss": 1.0257, "step": 63735 }, { "epoch": 0.16, "learning_rate": 8.421762544667573e-05, "loss": 1.0494, "step": 63740 }, { "epoch": 0.16, "learning_rate": 8.421636720519402e-05, "loss": 1.0273, "step": 63745 }, { "epoch": 0.16, "learning_rate": 8.421510896371231e-05, "loss": 1.024, "step": 63750 }, { "epoch": 0.16, "learning_rate": 8.421385072223062e-05, "loss": 1.024, "step": 63755 }, { "epoch": 0.16, "learning_rate": 8.421259248074891e-05, "loss": 1.0264, "step": 63760 }, { "epoch": 0.16, "learning_rate": 8.42113342392672e-05, "loss": 1.026, "step": 63765 }, { "epoch": 0.16, "learning_rate": 8.421007599778549e-05, "loss": 1.0245, "step": 63770 }, { "epoch": 0.16, "learning_rate": 8.42088177563038e-05, "loss": 1.0255, "step": 63775 }, { "epoch": 0.16, "learning_rate": 8.420755951482209e-05, "loss": 1.0248, "step": 63780 }, { "epoch": 0.16, "learning_rate": 8.420630127334038e-05, "loss": 1.025, "step": 63785 }, { "epoch": 0.16, "learning_rate": 8.420504303185867e-05, "loss": 1.0281, "step": 63790 }, { "epoch": 0.16, "learning_rate": 8.420378479037698e-05, "loss": 1.0264, "step": 63795 }, { "epoch": 0.16, "learning_rate": 8.420252654889527e-05, "loss": 1.0269, "step": 63800 }, { "epoch": 0.16, "learning_rate": 8.420126830741356e-05, "loss": 1.0246, "step": 63805 }, { "epoch": 0.16, "learning_rate": 8.420001006593185e-05, "loss": 1.0261, "step": 63810 }, { "epoch": 0.16, "learning_rate": 8.419875182445014e-05, "loss": 1.0266, "step": 63815 }, { "epoch": 0.16, "learning_rate": 8.419749358296845e-05, "loss": 1.0273, "step": 63820 }, { "epoch": 0.16, "learning_rate": 8.419623534148674e-05, "loss": 1.0293, "step": 63825 }, { "epoch": 0.16, "learning_rate": 8.419497710000503e-05, "loss": 1.025, "step": 63830 }, { "epoch": 0.16, "learning_rate": 8.419371885852332e-05, "loss": 1.0257, "step": 63835 }, { "epoch": 0.16, "learning_rate": 8.419246061704163e-05, "loss": 1.0242, "step": 63840 }, { "epoch": 0.16, "learning_rate": 8.419120237555992e-05, "loss": 1.0262, "step": 63845 }, { "epoch": 0.16, "learning_rate": 8.418994413407821e-05, "loss": 1.0256, "step": 63850 }, { "epoch": 0.16, "learning_rate": 8.41886858925965e-05, "loss": 1.0265, "step": 63855 }, { "epoch": 0.16, "learning_rate": 8.418742765111481e-05, "loss": 1.0278, "step": 63860 }, { "epoch": 0.16, "learning_rate": 8.41861694096331e-05, "loss": 1.0232, "step": 63865 }, { "epoch": 0.16, "learning_rate": 8.418491116815139e-05, "loss": 1.0262, "step": 63870 }, { "epoch": 0.16, "learning_rate": 8.418365292666968e-05, "loss": 1.0259, "step": 63875 }, { "epoch": 0.16, "learning_rate": 8.418239468518798e-05, "loss": 1.0244, "step": 63880 }, { "epoch": 0.16, "learning_rate": 8.418113644370628e-05, "loss": 1.0252, "step": 63885 }, { "epoch": 0.16, "learning_rate": 8.417987820222457e-05, "loss": 1.0285, "step": 63890 }, { "epoch": 0.16, "learning_rate": 8.417861996074286e-05, "loss": 1.0277, "step": 63895 }, { "epoch": 0.16, "learning_rate": 8.417736171926116e-05, "loss": 1.0242, "step": 63900 }, { "epoch": 0.16, "learning_rate": 8.417610347777946e-05, "loss": 1.0265, "step": 63905 }, { "epoch": 0.16, "learning_rate": 8.417484523629775e-05, "loss": 1.0268, "step": 63910 }, { "epoch": 0.16, "learning_rate": 8.417358699481604e-05, "loss": 1.0255, "step": 63915 }, { "epoch": 0.16, "learning_rate": 8.417232875333434e-05, "loss": 1.0281, "step": 63920 }, { "epoch": 0.16, "learning_rate": 8.417107051185264e-05, "loss": 1.0246, "step": 63925 }, { "epoch": 0.16, "learning_rate": 8.416981227037093e-05, "loss": 1.0267, "step": 63930 }, { "epoch": 0.16, "learning_rate": 8.416855402888922e-05, "loss": 1.026, "step": 63935 }, { "epoch": 0.16, "learning_rate": 8.416729578740752e-05, "loss": 1.0261, "step": 63940 }, { "epoch": 0.16, "learning_rate": 8.416603754592581e-05, "loss": 1.0246, "step": 63945 }, { "epoch": 0.16, "learning_rate": 8.416477930444411e-05, "loss": 1.0279, "step": 63950 }, { "epoch": 0.16, "learning_rate": 8.41635210629624e-05, "loss": 1.0257, "step": 63955 }, { "epoch": 0.16, "learning_rate": 8.41622628214807e-05, "loss": 1.0264, "step": 63960 }, { "epoch": 0.16, "learning_rate": 8.416100457999899e-05, "loss": 1.0247, "step": 63965 }, { "epoch": 0.16, "learning_rate": 8.415974633851729e-05, "loss": 1.0283, "step": 63970 }, { "epoch": 0.16, "learning_rate": 8.415848809703558e-05, "loss": 1.0254, "step": 63975 }, { "epoch": 0.16, "learning_rate": 8.415722985555389e-05, "loss": 1.0246, "step": 63980 }, { "epoch": 0.16, "learning_rate": 8.415597161407218e-05, "loss": 1.0269, "step": 63985 }, { "epoch": 0.16, "learning_rate": 8.415471337259047e-05, "loss": 1.0267, "step": 63990 }, { "epoch": 0.16, "learning_rate": 8.415345513110878e-05, "loss": 1.0265, "step": 63995 }, { "epoch": 0.16, "learning_rate": 8.415219688962707e-05, "loss": 1.0272, "step": 64000 }, { "epoch": 0.16, "learning_rate": 8.415093864814536e-05, "loss": 1.0272, "step": 64005 }, { "epoch": 0.16, "learning_rate": 8.414968040666365e-05, "loss": 1.0437, "step": 64010 }, { "epoch": 0.16, "learning_rate": 8.414842216518194e-05, "loss": 1.0257, "step": 64015 }, { "epoch": 0.16, "learning_rate": 8.414716392370025e-05, "loss": 1.0262, "step": 64020 }, { "epoch": 0.16, "learning_rate": 8.414590568221854e-05, "loss": 1.0278, "step": 64025 }, { "epoch": 0.16, "learning_rate": 8.414464744073683e-05, "loss": 1.0253, "step": 64030 }, { "epoch": 0.16, "learning_rate": 8.414338919925512e-05, "loss": 1.0261, "step": 64035 }, { "epoch": 0.16, "learning_rate": 8.414213095777343e-05, "loss": 1.0242, "step": 64040 }, { "epoch": 0.16, "learning_rate": 8.414087271629172e-05, "loss": 1.0268, "step": 64045 }, { "epoch": 0.16, "learning_rate": 8.413961447481001e-05, "loss": 1.0267, "step": 64050 }, { "epoch": 0.16, "learning_rate": 8.41383562333283e-05, "loss": 1.0259, "step": 64055 }, { "epoch": 0.16, "learning_rate": 8.413709799184661e-05, "loss": 1.025, "step": 64060 }, { "epoch": 0.16, "learning_rate": 8.41358397503649e-05, "loss": 1.0244, "step": 64065 }, { "epoch": 0.16, "learning_rate": 8.413458150888319e-05, "loss": 1.0282, "step": 64070 }, { "epoch": 0.16, "learning_rate": 8.413332326740148e-05, "loss": 1.0239, "step": 64075 }, { "epoch": 0.16, "learning_rate": 8.413206502591978e-05, "loss": 1.0265, "step": 64080 }, { "epoch": 0.16, "learning_rate": 8.413080678443808e-05, "loss": 1.0273, "step": 64085 }, { "epoch": 0.16, "learning_rate": 8.412954854295637e-05, "loss": 1.0268, "step": 64090 }, { "epoch": 0.16, "learning_rate": 8.412829030147466e-05, "loss": 1.0276, "step": 64095 }, { "epoch": 0.16, "learning_rate": 8.412703205999296e-05, "loss": 1.0271, "step": 64100 }, { "epoch": 0.16, "learning_rate": 8.412577381851126e-05, "loss": 1.0255, "step": 64105 }, { "epoch": 0.16, "learning_rate": 8.412451557702955e-05, "loss": 1.0274, "step": 64110 }, { "epoch": 0.16, "learning_rate": 8.412325733554784e-05, "loss": 1.0261, "step": 64115 }, { "epoch": 0.16, "learning_rate": 8.412199909406614e-05, "loss": 1.0251, "step": 64120 }, { "epoch": 0.16, "learning_rate": 8.412074085258444e-05, "loss": 1.0285, "step": 64125 }, { "epoch": 0.16, "learning_rate": 8.411948261110273e-05, "loss": 1.0253, "step": 64130 }, { "epoch": 0.16, "learning_rate": 8.411822436962102e-05, "loss": 1.0255, "step": 64135 }, { "epoch": 0.16, "learning_rate": 8.411696612813932e-05, "loss": 1.0257, "step": 64140 }, { "epoch": 0.16, "learning_rate": 8.41157078866576e-05, "loss": 1.0264, "step": 64145 }, { "epoch": 0.16, "learning_rate": 8.411444964517591e-05, "loss": 1.0249, "step": 64150 }, { "epoch": 0.16, "learning_rate": 8.41131914036942e-05, "loss": 1.0257, "step": 64155 }, { "epoch": 0.16, "learning_rate": 8.41119331622125e-05, "loss": 1.027, "step": 64160 }, { "epoch": 0.16, "learning_rate": 8.411067492073079e-05, "loss": 1.0247, "step": 64165 }, { "epoch": 0.16, "learning_rate": 8.410941667924909e-05, "loss": 1.0469, "step": 64170 }, { "epoch": 0.16, "learning_rate": 8.410815843776738e-05, "loss": 1.0269, "step": 64175 }, { "epoch": 0.16, "learning_rate": 8.410690019628568e-05, "loss": 1.0265, "step": 64180 }, { "epoch": 0.16, "learning_rate": 8.410564195480397e-05, "loss": 1.0272, "step": 64185 }, { "epoch": 0.16, "learning_rate": 8.410438371332227e-05, "loss": 1.0271, "step": 64190 }, { "epoch": 0.16, "learning_rate": 8.410312547184056e-05, "loss": 1.0261, "step": 64195 }, { "epoch": 0.16, "learning_rate": 8.410186723035886e-05, "loss": 1.027, "step": 64200 }, { "epoch": 0.16, "learning_rate": 8.410060898887715e-05, "loss": 1.0261, "step": 64205 }, { "epoch": 0.16, "learning_rate": 8.409935074739544e-05, "loss": 1.0271, "step": 64210 }, { "epoch": 0.16, "learning_rate": 8.409809250591374e-05, "loss": 1.0281, "step": 64215 }, { "epoch": 0.16, "learning_rate": 8.409683426443203e-05, "loss": 1.0278, "step": 64220 }, { "epoch": 0.16, "learning_rate": 8.409557602295033e-05, "loss": 1.0258, "step": 64225 }, { "epoch": 0.16, "learning_rate": 8.409431778146862e-05, "loss": 1.0263, "step": 64230 }, { "epoch": 0.16, "learning_rate": 8.409305953998692e-05, "loss": 1.0433, "step": 64235 }, { "epoch": 0.16, "learning_rate": 8.409180129850521e-05, "loss": 1.0262, "step": 64240 }, { "epoch": 0.16, "learning_rate": 8.40905430570235e-05, "loss": 1.0269, "step": 64245 }, { "epoch": 0.16, "learning_rate": 8.40892848155418e-05, "loss": 1.0269, "step": 64250 }, { "epoch": 0.16, "learning_rate": 8.40880265740601e-05, "loss": 1.0247, "step": 64255 }, { "epoch": 0.16, "learning_rate": 8.40867683325784e-05, "loss": 1.0272, "step": 64260 }, { "epoch": 0.16, "learning_rate": 8.408551009109669e-05, "loss": 1.0255, "step": 64265 }, { "epoch": 0.16, "learning_rate": 8.408425184961498e-05, "loss": 1.0244, "step": 64270 }, { "epoch": 0.16, "learning_rate": 8.408299360813327e-05, "loss": 1.0246, "step": 64275 }, { "epoch": 0.16, "learning_rate": 8.408173536665157e-05, "loss": 1.0266, "step": 64280 }, { "epoch": 0.16, "learning_rate": 8.408047712516987e-05, "loss": 1.026, "step": 64285 }, { "epoch": 0.16, "learning_rate": 8.407921888368816e-05, "loss": 1.0675, "step": 64290 }, { "epoch": 0.16, "learning_rate": 8.407796064220645e-05, "loss": 1.0255, "step": 64295 }, { "epoch": 0.16, "learning_rate": 8.407670240072475e-05, "loss": 1.0244, "step": 64300 }, { "epoch": 0.16, "learning_rate": 8.407544415924305e-05, "loss": 1.0246, "step": 64305 }, { "epoch": 0.16, "learning_rate": 8.407418591776134e-05, "loss": 1.0249, "step": 64310 }, { "epoch": 0.16, "learning_rate": 8.407292767627963e-05, "loss": 1.0267, "step": 64315 }, { "epoch": 0.16, "learning_rate": 8.407166943479793e-05, "loss": 1.0252, "step": 64320 }, { "epoch": 0.16, "learning_rate": 8.407041119331623e-05, "loss": 1.0244, "step": 64325 }, { "epoch": 0.16, "learning_rate": 8.406915295183452e-05, "loss": 1.0255, "step": 64330 }, { "epoch": 0.16, "learning_rate": 8.406789471035281e-05, "loss": 1.0292, "step": 64335 }, { "epoch": 0.16, "learning_rate": 8.40666364688711e-05, "loss": 1.0256, "step": 64340 }, { "epoch": 0.16, "learning_rate": 8.40653782273894e-05, "loss": 1.0275, "step": 64345 }, { "epoch": 0.16, "learning_rate": 8.40641199859077e-05, "loss": 1.0247, "step": 64350 }, { "epoch": 0.16, "learning_rate": 8.406286174442599e-05, "loss": 1.0264, "step": 64355 }, { "epoch": 0.16, "learning_rate": 8.406160350294428e-05, "loss": 1.027, "step": 64360 }, { "epoch": 0.16, "learning_rate": 8.406034526146259e-05, "loss": 1.0235, "step": 64365 }, { "epoch": 0.16, "learning_rate": 8.405908701998088e-05, "loss": 1.0261, "step": 64370 }, { "epoch": 0.16, "learning_rate": 8.405782877849917e-05, "loss": 1.0261, "step": 64375 }, { "epoch": 0.16, "learning_rate": 8.405657053701746e-05, "loss": 1.0246, "step": 64380 }, { "epoch": 0.16, "learning_rate": 8.405531229553577e-05, "loss": 1.0258, "step": 64385 }, { "epoch": 0.16, "learning_rate": 8.405405405405406e-05, "loss": 1.0249, "step": 64390 }, { "epoch": 0.16, "learning_rate": 8.405279581257235e-05, "loss": 1.0266, "step": 64395 }, { "epoch": 0.16, "learning_rate": 8.405153757109064e-05, "loss": 1.0288, "step": 64400 }, { "epoch": 0.16, "learning_rate": 8.405027932960893e-05, "loss": 1.0261, "step": 64405 }, { "epoch": 0.16, "learning_rate": 8.404902108812724e-05, "loss": 1.0282, "step": 64410 }, { "epoch": 0.16, "learning_rate": 8.404776284664553e-05, "loss": 1.0249, "step": 64415 }, { "epoch": 0.16, "learning_rate": 8.404650460516382e-05, "loss": 1.0263, "step": 64420 }, { "epoch": 0.16, "learning_rate": 8.404524636368211e-05, "loss": 1.0253, "step": 64425 }, { "epoch": 0.16, "learning_rate": 8.404398812220042e-05, "loss": 1.026, "step": 64430 }, { "epoch": 0.16, "learning_rate": 8.404272988071871e-05, "loss": 1.0264, "step": 64435 }, { "epoch": 0.16, "learning_rate": 8.4041471639237e-05, "loss": 1.0255, "step": 64440 }, { "epoch": 0.16, "learning_rate": 8.404021339775529e-05, "loss": 1.0474, "step": 64445 }, { "epoch": 0.16, "learning_rate": 8.40389551562736e-05, "loss": 1.0283, "step": 64450 }, { "epoch": 0.16, "learning_rate": 8.403769691479189e-05, "loss": 1.0268, "step": 64455 }, { "epoch": 0.16, "learning_rate": 8.403643867331018e-05, "loss": 1.0273, "step": 64460 }, { "epoch": 0.16, "learning_rate": 8.403518043182847e-05, "loss": 1.0258, "step": 64465 }, { "epoch": 0.16, "learning_rate": 8.403392219034676e-05, "loss": 1.0264, "step": 64470 }, { "epoch": 0.16, "learning_rate": 8.403266394886507e-05, "loss": 1.0255, "step": 64475 }, { "epoch": 0.16, "learning_rate": 8.403140570738337e-05, "loss": 1.0285, "step": 64480 }, { "epoch": 0.16, "learning_rate": 8.403014746590167e-05, "loss": 1.0252, "step": 64485 }, { "epoch": 0.16, "learning_rate": 8.402888922441996e-05, "loss": 1.0266, "step": 64490 }, { "epoch": 0.16, "learning_rate": 8.402763098293825e-05, "loss": 1.0281, "step": 64495 }, { "epoch": 0.16, "learning_rate": 8.402637274145655e-05, "loss": 1.0253, "step": 64500 }, { "epoch": 0.16, "learning_rate": 8.402511449997485e-05, "loss": 1.0282, "step": 64505 }, { "epoch": 0.16, "learning_rate": 8.402385625849314e-05, "loss": 1.0275, "step": 64510 }, { "epoch": 0.16, "learning_rate": 8.402259801701143e-05, "loss": 1.0266, "step": 64515 }, { "epoch": 0.16, "learning_rate": 8.402133977552973e-05, "loss": 1.0282, "step": 64520 }, { "epoch": 0.16, "learning_rate": 8.402008153404803e-05, "loss": 1.0248, "step": 64525 }, { "epoch": 0.16, "learning_rate": 8.401882329256632e-05, "loss": 1.0245, "step": 64530 }, { "epoch": 0.16, "learning_rate": 8.401756505108461e-05, "loss": 1.025, "step": 64535 }, { "epoch": 0.16, "learning_rate": 8.40163068096029e-05, "loss": 1.0261, "step": 64540 }, { "epoch": 0.16, "learning_rate": 8.40150485681212e-05, "loss": 1.0256, "step": 64545 }, { "epoch": 0.16, "learning_rate": 8.40137903266395e-05, "loss": 1.0254, "step": 64550 }, { "epoch": 0.16, "learning_rate": 8.401253208515779e-05, "loss": 1.0254, "step": 64555 }, { "epoch": 0.16, "learning_rate": 8.401127384367608e-05, "loss": 1.0262, "step": 64560 }, { "epoch": 0.16, "learning_rate": 8.401001560219439e-05, "loss": 1.0291, "step": 64565 }, { "epoch": 0.16, "learning_rate": 8.400875736071268e-05, "loss": 1.0261, "step": 64570 }, { "epoch": 0.16, "learning_rate": 8.400749911923097e-05, "loss": 1.0285, "step": 64575 }, { "epoch": 0.16, "learning_rate": 8.400624087774926e-05, "loss": 1.0281, "step": 64580 }, { "epoch": 0.16, "learning_rate": 8.400498263626757e-05, "loss": 1.0259, "step": 64585 }, { "epoch": 0.16, "learning_rate": 8.400372439478586e-05, "loss": 1.0271, "step": 64590 }, { "epoch": 0.16, "learning_rate": 8.400246615330415e-05, "loss": 1.0243, "step": 64595 }, { "epoch": 0.16, "learning_rate": 8.400120791182244e-05, "loss": 1.0262, "step": 64600 }, { "epoch": 0.16, "learning_rate": 8.399994967034073e-05, "loss": 1.0257, "step": 64605 }, { "epoch": 0.16, "learning_rate": 8.399869142885904e-05, "loss": 1.025, "step": 64610 }, { "epoch": 0.16, "learning_rate": 8.399743318737733e-05, "loss": 1.0251, "step": 64615 }, { "epoch": 0.16, "learning_rate": 8.399617494589562e-05, "loss": 1.0271, "step": 64620 }, { "epoch": 0.16, "learning_rate": 8.399491670441391e-05, "loss": 1.0265, "step": 64625 }, { "epoch": 0.16, "learning_rate": 8.399365846293222e-05, "loss": 1.0249, "step": 64630 }, { "epoch": 0.16, "learning_rate": 8.399240022145051e-05, "loss": 1.0271, "step": 64635 }, { "epoch": 0.16, "learning_rate": 8.39911419799688e-05, "loss": 1.0269, "step": 64640 }, { "epoch": 0.16, "learning_rate": 8.398988373848709e-05, "loss": 1.0283, "step": 64645 }, { "epoch": 0.16, "learning_rate": 8.39886254970054e-05, "loss": 1.024, "step": 64650 }, { "epoch": 0.16, "learning_rate": 8.398736725552369e-05, "loss": 1.0254, "step": 64655 }, { "epoch": 0.16, "learning_rate": 8.398610901404198e-05, "loss": 1.0264, "step": 64660 }, { "epoch": 0.16, "learning_rate": 8.398485077256027e-05, "loss": 1.0273, "step": 64665 }, { "epoch": 0.16, "learning_rate": 8.398359253107856e-05, "loss": 1.0262, "step": 64670 }, { "epoch": 0.16, "learning_rate": 8.398233428959687e-05, "loss": 1.0253, "step": 64675 }, { "epoch": 0.16, "learning_rate": 8.398107604811516e-05, "loss": 1.0237, "step": 64680 }, { "epoch": 0.16, "learning_rate": 8.397981780663345e-05, "loss": 1.027, "step": 64685 }, { "epoch": 0.16, "learning_rate": 8.397855956515174e-05, "loss": 1.0245, "step": 64690 }, { "epoch": 0.16, "learning_rate": 8.397730132367005e-05, "loss": 1.0258, "step": 64695 }, { "epoch": 0.16, "learning_rate": 8.397604308218834e-05, "loss": 1.024, "step": 64700 }, { "epoch": 0.16, "learning_rate": 8.397478484070663e-05, "loss": 1.026, "step": 64705 }, { "epoch": 0.16, "learning_rate": 8.397352659922492e-05, "loss": 1.0278, "step": 64710 }, { "epoch": 0.16, "learning_rate": 8.397226835774323e-05, "loss": 1.0241, "step": 64715 }, { "epoch": 0.16, "learning_rate": 8.397101011626152e-05, "loss": 1.0257, "step": 64720 }, { "epoch": 0.16, "learning_rate": 8.396975187477981e-05, "loss": 1.0274, "step": 64725 }, { "epoch": 0.16, "learning_rate": 8.39684936332981e-05, "loss": 1.0243, "step": 64730 }, { "epoch": 0.16, "learning_rate": 8.39672353918164e-05, "loss": 1.0267, "step": 64735 }, { "epoch": 0.16, "learning_rate": 8.39659771503347e-05, "loss": 1.0263, "step": 64740 }, { "epoch": 0.16, "learning_rate": 8.396471890885299e-05, "loss": 1.0254, "step": 64745 }, { "epoch": 0.16, "learning_rate": 8.396346066737128e-05, "loss": 1.0246, "step": 64750 }, { "epoch": 0.16, "learning_rate": 8.396220242588957e-05, "loss": 1.0259, "step": 64755 }, { "epoch": 0.16, "learning_rate": 8.396094418440788e-05, "loss": 1.0275, "step": 64760 }, { "epoch": 0.16, "learning_rate": 8.395968594292617e-05, "loss": 1.0251, "step": 64765 }, { "epoch": 0.16, "learning_rate": 8.395842770144446e-05, "loss": 1.0261, "step": 64770 }, { "epoch": 0.16, "learning_rate": 8.395716945996275e-05, "loss": 1.0259, "step": 64775 }, { "epoch": 0.16, "learning_rate": 8.395591121848106e-05, "loss": 1.0294, "step": 64780 }, { "epoch": 0.16, "learning_rate": 8.395465297699935e-05, "loss": 1.0277, "step": 64785 }, { "epoch": 0.16, "learning_rate": 8.395339473551764e-05, "loss": 1.0248, "step": 64790 }, { "epoch": 0.16, "learning_rate": 8.395213649403593e-05, "loss": 1.0266, "step": 64795 }, { "epoch": 0.16, "learning_rate": 8.395087825255423e-05, "loss": 1.0273, "step": 64800 }, { "epoch": 0.16, "learning_rate": 8.394962001107253e-05, "loss": 1.0287, "step": 64805 }, { "epoch": 0.16, "learning_rate": 8.394836176959082e-05, "loss": 1.0264, "step": 64810 }, { "epoch": 0.16, "learning_rate": 8.394710352810911e-05, "loss": 1.0261, "step": 64815 }, { "epoch": 0.16, "learning_rate": 8.39458452866274e-05, "loss": 1.0249, "step": 64820 }, { "epoch": 0.16, "learning_rate": 8.394458704514571e-05, "loss": 1.0244, "step": 64825 }, { "epoch": 0.16, "learning_rate": 8.3943328803664e-05, "loss": 1.026, "step": 64830 }, { "epoch": 0.16, "learning_rate": 8.39420705621823e-05, "loss": 1.0258, "step": 64835 }, { "epoch": 0.16, "learning_rate": 8.394081232070059e-05, "loss": 1.0261, "step": 64840 }, { "epoch": 0.16, "learning_rate": 8.393955407921889e-05, "loss": 1.0253, "step": 64845 }, { "epoch": 0.16, "learning_rate": 8.393829583773718e-05, "loss": 1.0249, "step": 64850 }, { "epoch": 0.16, "learning_rate": 8.393703759625547e-05, "loss": 1.0236, "step": 64855 }, { "epoch": 0.16, "learning_rate": 8.393577935477377e-05, "loss": 1.0259, "step": 64860 }, { "epoch": 0.16, "learning_rate": 8.393452111329206e-05, "loss": 1.0269, "step": 64865 }, { "epoch": 0.16, "learning_rate": 8.393326287181036e-05, "loss": 1.0259, "step": 64870 }, { "epoch": 0.16, "learning_rate": 8.393200463032865e-05, "loss": 1.0261, "step": 64875 }, { "epoch": 0.16, "learning_rate": 8.393074638884695e-05, "loss": 1.0256, "step": 64880 }, { "epoch": 0.16, "learning_rate": 8.392948814736524e-05, "loss": 1.0267, "step": 64885 }, { "epoch": 0.16, "learning_rate": 8.392822990588354e-05, "loss": 1.0265, "step": 64890 }, { "epoch": 0.16, "learning_rate": 8.392697166440183e-05, "loss": 1.0274, "step": 64895 }, { "epoch": 0.16, "learning_rate": 8.392571342292013e-05, "loss": 1.0278, "step": 64900 }, { "epoch": 0.16, "learning_rate": 8.392445518143842e-05, "loss": 1.0236, "step": 64905 }, { "epoch": 0.16, "learning_rate": 8.392319693995672e-05, "loss": 1.0263, "step": 64910 }, { "epoch": 0.16, "learning_rate": 8.392193869847501e-05, "loss": 1.0265, "step": 64915 }, { "epoch": 0.16, "learning_rate": 8.39206804569933e-05, "loss": 1.0249, "step": 64920 }, { "epoch": 0.16, "learning_rate": 8.39194222155116e-05, "loss": 1.0284, "step": 64925 }, { "epoch": 0.16, "learning_rate": 8.391816397402989e-05, "loss": 1.0256, "step": 64930 }, { "epoch": 0.16, "learning_rate": 8.39169057325482e-05, "loss": 1.0231, "step": 64935 }, { "epoch": 0.16, "learning_rate": 8.391564749106649e-05, "loss": 1.0278, "step": 64940 }, { "epoch": 0.16, "learning_rate": 8.391438924958478e-05, "loss": 1.0265, "step": 64945 }, { "epoch": 0.16, "learning_rate": 8.391313100810307e-05, "loss": 1.0264, "step": 64950 }, { "epoch": 0.16, "learning_rate": 8.391187276662137e-05, "loss": 1.0254, "step": 64955 }, { "epoch": 0.16, "learning_rate": 8.391061452513967e-05, "loss": 1.0251, "step": 64960 }, { "epoch": 0.16, "learning_rate": 8.390935628365796e-05, "loss": 1.0251, "step": 64965 }, { "epoch": 0.16, "learning_rate": 8.390809804217625e-05, "loss": 1.0259, "step": 64970 }, { "epoch": 0.16, "learning_rate": 8.390683980069455e-05, "loss": 1.0271, "step": 64975 }, { "epoch": 0.16, "learning_rate": 8.390558155921286e-05, "loss": 1.0506, "step": 64980 }, { "epoch": 0.16, "learning_rate": 8.390432331773115e-05, "loss": 1.0275, "step": 64985 }, { "epoch": 0.16, "learning_rate": 8.390306507624944e-05, "loss": 1.0271, "step": 64990 }, { "epoch": 0.16, "learning_rate": 8.390180683476773e-05, "loss": 1.0223, "step": 64995 }, { "epoch": 0.16, "learning_rate": 8.390054859328604e-05, "loss": 1.0243, "step": 65000 }, { "epoch": 0.16, "learning_rate": 8.389929035180433e-05, "loss": 1.0269, "step": 65005 }, { "epoch": 0.16, "learning_rate": 8.389803211032262e-05, "loss": 1.0274, "step": 65010 }, { "epoch": 0.16, "learning_rate": 8.389677386884091e-05, "loss": 1.0267, "step": 65015 }, { "epoch": 0.16, "learning_rate": 8.38955156273592e-05, "loss": 1.0261, "step": 65020 }, { "epoch": 0.16, "learning_rate": 8.389425738587751e-05, "loss": 1.0256, "step": 65025 }, { "epoch": 0.16, "learning_rate": 8.38929991443958e-05, "loss": 1.0252, "step": 65030 }, { "epoch": 0.16, "learning_rate": 8.38917409029141e-05, "loss": 1.0466, "step": 65035 }, { "epoch": 0.16, "learning_rate": 8.389048266143239e-05, "loss": 1.0292, "step": 65040 }, { "epoch": 0.16, "learning_rate": 8.388922441995069e-05, "loss": 1.0274, "step": 65045 }, { "epoch": 0.16, "learning_rate": 8.388796617846898e-05, "loss": 1.0253, "step": 65050 }, { "epoch": 0.16, "learning_rate": 8.388670793698727e-05, "loss": 1.026, "step": 65055 }, { "epoch": 0.16, "learning_rate": 8.388544969550557e-05, "loss": 1.0266, "step": 65060 }, { "epoch": 0.16, "learning_rate": 8.388419145402387e-05, "loss": 1.0469, "step": 65065 }, { "epoch": 0.16, "learning_rate": 8.388293321254216e-05, "loss": 1.0251, "step": 65070 }, { "epoch": 0.16, "learning_rate": 8.388167497106045e-05, "loss": 1.0244, "step": 65075 }, { "epoch": 0.16, "learning_rate": 8.388041672957875e-05, "loss": 1.0248, "step": 65080 }, { "epoch": 0.16, "learning_rate": 8.387915848809704e-05, "loss": 1.0271, "step": 65085 }, { "epoch": 0.16, "learning_rate": 8.387790024661534e-05, "loss": 1.0262, "step": 65090 }, { "epoch": 0.16, "learning_rate": 8.387664200513363e-05, "loss": 1.0254, "step": 65095 }, { "epoch": 0.16, "learning_rate": 8.387538376365193e-05, "loss": 1.0282, "step": 65100 }, { "epoch": 0.16, "learning_rate": 8.387412552217022e-05, "loss": 1.0256, "step": 65105 }, { "epoch": 0.16, "learning_rate": 8.387286728068852e-05, "loss": 1.0284, "step": 65110 }, { "epoch": 0.16, "learning_rate": 8.387160903920681e-05, "loss": 1.0256, "step": 65115 }, { "epoch": 0.16, "learning_rate": 8.38703507977251e-05, "loss": 1.0258, "step": 65120 }, { "epoch": 0.16, "learning_rate": 8.38690925562434e-05, "loss": 1.0269, "step": 65125 }, { "epoch": 0.16, "learning_rate": 8.386783431476169e-05, "loss": 1.0254, "step": 65130 }, { "epoch": 0.16, "learning_rate": 8.386657607328e-05, "loss": 1.025, "step": 65135 }, { "epoch": 0.16, "learning_rate": 8.386531783179829e-05, "loss": 1.0256, "step": 65140 }, { "epoch": 0.16, "learning_rate": 8.386405959031658e-05, "loss": 1.0248, "step": 65145 }, { "epoch": 0.16, "learning_rate": 8.386280134883487e-05, "loss": 1.0281, "step": 65150 }, { "epoch": 0.16, "learning_rate": 8.386154310735317e-05, "loss": 1.0247, "step": 65155 }, { "epoch": 0.16, "learning_rate": 8.386028486587147e-05, "loss": 1.0316, "step": 65160 }, { "epoch": 0.16, "learning_rate": 8.385902662438976e-05, "loss": 1.0259, "step": 65165 }, { "epoch": 0.16, "learning_rate": 8.385776838290805e-05, "loss": 1.0264, "step": 65170 }, { "epoch": 0.16, "learning_rate": 8.385651014142635e-05, "loss": 1.0259, "step": 65175 }, { "epoch": 0.16, "learning_rate": 8.385525189994465e-05, "loss": 1.0468, "step": 65180 }, { "epoch": 0.16, "learning_rate": 8.385399365846294e-05, "loss": 1.0255, "step": 65185 }, { "epoch": 0.16, "learning_rate": 8.385273541698123e-05, "loss": 1.0259, "step": 65190 }, { "epoch": 0.16, "learning_rate": 8.385147717549952e-05, "loss": 1.0264, "step": 65195 }, { "epoch": 0.16, "learning_rate": 8.385021893401783e-05, "loss": 1.0265, "step": 65200 }, { "epoch": 0.16, "learning_rate": 8.384896069253612e-05, "loss": 1.0258, "step": 65205 }, { "epoch": 0.16, "learning_rate": 8.384770245105441e-05, "loss": 1.0248, "step": 65210 }, { "epoch": 0.16, "learning_rate": 8.38464442095727e-05, "loss": 1.0276, "step": 65215 }, { "epoch": 0.16, "learning_rate": 8.3845185968091e-05, "loss": 1.028, "step": 65220 }, { "epoch": 0.16, "learning_rate": 8.38439277266093e-05, "loss": 1.0255, "step": 65225 }, { "epoch": 0.16, "learning_rate": 8.384266948512759e-05, "loss": 1.0282, "step": 65230 }, { "epoch": 0.16, "learning_rate": 8.384141124364588e-05, "loss": 1.0262, "step": 65235 }, { "epoch": 0.16, "learning_rate": 8.384015300216418e-05, "loss": 1.0254, "step": 65240 }, { "epoch": 0.16, "learning_rate": 8.383889476068248e-05, "loss": 1.0258, "step": 65245 }, { "epoch": 0.16, "learning_rate": 8.383763651920077e-05, "loss": 1.0271, "step": 65250 }, { "epoch": 0.16, "learning_rate": 8.383637827771906e-05, "loss": 1.0263, "step": 65255 }, { "epoch": 0.16, "learning_rate": 8.383512003623735e-05, "loss": 1.0248, "step": 65260 }, { "epoch": 0.16, "learning_rate": 8.383386179475566e-05, "loss": 1.0251, "step": 65265 }, { "epoch": 0.16, "learning_rate": 8.383260355327395e-05, "loss": 1.0286, "step": 65270 }, { "epoch": 0.16, "learning_rate": 8.383134531179224e-05, "loss": 1.0274, "step": 65275 }, { "epoch": 0.16, "learning_rate": 8.383008707031053e-05, "loss": 1.0241, "step": 65280 }, { "epoch": 0.16, "learning_rate": 8.382882882882884e-05, "loss": 1.0254, "step": 65285 }, { "epoch": 0.16, "learning_rate": 8.382757058734713e-05, "loss": 1.0262, "step": 65290 }, { "epoch": 0.16, "learning_rate": 8.382631234586542e-05, "loss": 1.0247, "step": 65295 }, { "epoch": 0.16, "learning_rate": 8.382505410438371e-05, "loss": 1.0244, "step": 65300 }, { "epoch": 0.16, "learning_rate": 8.382379586290202e-05, "loss": 1.028, "step": 65305 }, { "epoch": 0.16, "learning_rate": 8.382253762142031e-05, "loss": 1.0272, "step": 65310 }, { "epoch": 0.16, "learning_rate": 8.38212793799386e-05, "loss": 1.0292, "step": 65315 }, { "epoch": 0.16, "learning_rate": 8.382002113845689e-05, "loss": 1.0262, "step": 65320 }, { "epoch": 0.16, "learning_rate": 8.381876289697518e-05, "loss": 1.027, "step": 65325 }, { "epoch": 0.16, "learning_rate": 8.381750465549349e-05, "loss": 1.0254, "step": 65330 }, { "epoch": 0.16, "learning_rate": 8.381624641401178e-05, "loss": 1.0276, "step": 65335 }, { "epoch": 0.16, "learning_rate": 8.381498817253007e-05, "loss": 1.0243, "step": 65340 }, { "epoch": 0.16, "learning_rate": 8.381372993104836e-05, "loss": 1.0228, "step": 65345 }, { "epoch": 0.16, "learning_rate": 8.381247168956667e-05, "loss": 1.0267, "step": 65350 }, { "epoch": 0.16, "learning_rate": 8.381121344808496e-05, "loss": 1.0263, "step": 65355 }, { "epoch": 0.16, "learning_rate": 8.380995520660325e-05, "loss": 1.0259, "step": 65360 }, { "epoch": 0.16, "learning_rate": 8.380869696512154e-05, "loss": 1.0267, "step": 65365 }, { "epoch": 0.16, "learning_rate": 8.380743872363985e-05, "loss": 1.0252, "step": 65370 }, { "epoch": 0.16, "learning_rate": 8.380618048215814e-05, "loss": 1.0231, "step": 65375 }, { "epoch": 0.16, "learning_rate": 8.380492224067643e-05, "loss": 1.0218, "step": 65380 }, { "epoch": 0.16, "learning_rate": 8.380366399919472e-05, "loss": 1.0274, "step": 65385 }, { "epoch": 0.16, "learning_rate": 8.380240575771301e-05, "loss": 1.0279, "step": 65390 }, { "epoch": 0.16, "learning_rate": 8.380114751623132e-05, "loss": 1.027, "step": 65395 }, { "epoch": 0.16, "learning_rate": 8.379988927474961e-05, "loss": 1.0259, "step": 65400 }, { "epoch": 0.16, "learning_rate": 8.37986310332679e-05, "loss": 1.0248, "step": 65405 }, { "epoch": 0.16, "learning_rate": 8.37973727917862e-05, "loss": 1.0248, "step": 65410 }, { "epoch": 0.16, "learning_rate": 8.37961145503045e-05, "loss": 1.0284, "step": 65415 }, { "epoch": 0.16, "learning_rate": 8.379485630882279e-05, "loss": 1.0255, "step": 65420 }, { "epoch": 0.16, "learning_rate": 8.379359806734108e-05, "loss": 1.0259, "step": 65425 }, { "epoch": 0.16, "learning_rate": 8.379233982585937e-05, "loss": 1.0235, "step": 65430 }, { "epoch": 0.16, "learning_rate": 8.379108158437768e-05, "loss": 1.0268, "step": 65435 }, { "epoch": 0.16, "learning_rate": 8.378982334289597e-05, "loss": 1.0279, "step": 65440 }, { "epoch": 0.16, "learning_rate": 8.378856510141426e-05, "loss": 1.0262, "step": 65445 }, { "epoch": 0.16, "learning_rate": 8.378730685993255e-05, "loss": 1.0268, "step": 65450 }, { "epoch": 0.16, "learning_rate": 8.378604861845085e-05, "loss": 1.0258, "step": 65455 }, { "epoch": 0.16, "learning_rate": 8.378479037696915e-05, "loss": 1.0253, "step": 65460 }, { "epoch": 0.16, "learning_rate": 8.378353213548744e-05, "loss": 1.0274, "step": 65465 }, { "epoch": 0.16, "learning_rate": 8.378227389400573e-05, "loss": 1.0258, "step": 65470 }, { "epoch": 0.16, "learning_rate": 8.378101565252404e-05, "loss": 1.0284, "step": 65475 }, { "epoch": 0.16, "learning_rate": 8.377975741104233e-05, "loss": 1.0269, "step": 65480 }, { "epoch": 0.16, "learning_rate": 8.377849916956064e-05, "loss": 1.0263, "step": 65485 }, { "epoch": 0.16, "learning_rate": 8.377724092807893e-05, "loss": 1.0241, "step": 65490 }, { "epoch": 0.16, "learning_rate": 8.377598268659722e-05, "loss": 1.0258, "step": 65495 }, { "epoch": 0.16, "learning_rate": 8.377472444511551e-05, "loss": 1.0253, "step": 65500 }, { "epoch": 0.16, "learning_rate": 8.377346620363382e-05, "loss": 1.0259, "step": 65505 }, { "epoch": 0.16, "learning_rate": 8.377220796215211e-05, "loss": 1.0262, "step": 65510 }, { "epoch": 0.16, "learning_rate": 8.37709497206704e-05, "loss": 1.0254, "step": 65515 }, { "epoch": 0.16, "learning_rate": 8.376969147918869e-05, "loss": 1.026, "step": 65520 }, { "epoch": 0.16, "learning_rate": 8.3768433237707e-05, "loss": 1.0266, "step": 65525 }, { "epoch": 0.16, "learning_rate": 8.376717499622529e-05, "loss": 1.0271, "step": 65530 }, { "epoch": 0.16, "learning_rate": 8.376591675474358e-05, "loss": 1.026, "step": 65535 }, { "epoch": 0.16, "learning_rate": 8.376465851326187e-05, "loss": 1.0266, "step": 65540 }, { "epoch": 0.16, "learning_rate": 8.376340027178016e-05, "loss": 1.0267, "step": 65545 }, { "epoch": 0.16, "learning_rate": 8.376214203029847e-05, "loss": 1.0278, "step": 65550 }, { "epoch": 0.16, "learning_rate": 8.376088378881676e-05, "loss": 1.0269, "step": 65555 }, { "epoch": 0.16, "learning_rate": 8.375962554733505e-05, "loss": 1.0259, "step": 65560 }, { "epoch": 0.16, "learning_rate": 8.375836730585334e-05, "loss": 1.0244, "step": 65565 }, { "epoch": 0.16, "learning_rate": 8.375710906437165e-05, "loss": 1.0252, "step": 65570 }, { "epoch": 0.16, "learning_rate": 8.375585082288994e-05, "loss": 1.0239, "step": 65575 }, { "epoch": 0.16, "learning_rate": 8.375459258140823e-05, "loss": 1.0281, "step": 65580 }, { "epoch": 0.16, "learning_rate": 8.375333433992652e-05, "loss": 1.0256, "step": 65585 }, { "epoch": 0.16, "learning_rate": 8.375207609844483e-05, "loss": 1.0255, "step": 65590 }, { "epoch": 0.16, "learning_rate": 8.375081785696312e-05, "loss": 1.0274, "step": 65595 }, { "epoch": 0.16, "learning_rate": 8.374955961548141e-05, "loss": 1.0286, "step": 65600 }, { "epoch": 0.16, "learning_rate": 8.37483013739997e-05, "loss": 1.0291, "step": 65605 }, { "epoch": 0.16, "learning_rate": 8.3747043132518e-05, "loss": 1.0252, "step": 65610 }, { "epoch": 0.16, "learning_rate": 8.37457848910363e-05, "loss": 1.0265, "step": 65615 }, { "epoch": 0.16, "learning_rate": 8.374452664955459e-05, "loss": 1.0271, "step": 65620 }, { "epoch": 0.16, "learning_rate": 8.374326840807288e-05, "loss": 1.0258, "step": 65625 }, { "epoch": 0.16, "learning_rate": 8.374201016659117e-05, "loss": 1.025, "step": 65630 }, { "epoch": 0.16, "learning_rate": 8.374075192510948e-05, "loss": 1.026, "step": 65635 }, { "epoch": 0.16, "learning_rate": 8.373949368362777e-05, "loss": 1.025, "step": 65640 }, { "epoch": 0.16, "learning_rate": 8.373823544214606e-05, "loss": 1.0254, "step": 65645 }, { "epoch": 0.16, "learning_rate": 8.373697720066435e-05, "loss": 1.0253, "step": 65650 }, { "epoch": 0.16, "learning_rate": 8.373571895918266e-05, "loss": 1.026, "step": 65655 }, { "epoch": 0.16, "learning_rate": 8.373446071770095e-05, "loss": 1.0243, "step": 65660 }, { "epoch": 0.16, "learning_rate": 8.373320247621924e-05, "loss": 1.0252, "step": 65665 }, { "epoch": 0.16, "learning_rate": 8.373194423473753e-05, "loss": 1.0258, "step": 65670 }, { "epoch": 0.16, "learning_rate": 8.373068599325582e-05, "loss": 1.0258, "step": 65675 }, { "epoch": 0.16, "learning_rate": 8.372942775177413e-05, "loss": 1.024, "step": 65680 }, { "epoch": 0.16, "learning_rate": 8.372816951029242e-05, "loss": 1.0251, "step": 65685 }, { "epoch": 0.16, "learning_rate": 8.372691126881071e-05, "loss": 1.0255, "step": 65690 }, { "epoch": 0.16, "learning_rate": 8.3725653027329e-05, "loss": 1.0261, "step": 65695 }, { "epoch": 0.16, "learning_rate": 8.372439478584731e-05, "loss": 1.0261, "step": 65700 }, { "epoch": 0.16, "learning_rate": 8.37231365443656e-05, "loss": 1.0283, "step": 65705 }, { "epoch": 0.16, "learning_rate": 8.372187830288389e-05, "loss": 1.0246, "step": 65710 }, { "epoch": 0.16, "learning_rate": 8.372062006140218e-05, "loss": 1.0256, "step": 65715 }, { "epoch": 0.16, "learning_rate": 8.371936181992049e-05, "loss": 1.0261, "step": 65720 }, { "epoch": 0.16, "learning_rate": 8.371810357843878e-05, "loss": 1.0235, "step": 65725 }, { "epoch": 0.16, "learning_rate": 8.371684533695707e-05, "loss": 1.0245, "step": 65730 }, { "epoch": 0.17, "learning_rate": 8.371558709547536e-05, "loss": 1.0266, "step": 65735 }, { "epoch": 0.17, "learning_rate": 8.371432885399366e-05, "loss": 1.0265, "step": 65740 }, { "epoch": 0.17, "learning_rate": 8.371307061251196e-05, "loss": 1.0255, "step": 65745 }, { "epoch": 0.17, "learning_rate": 8.371181237103025e-05, "loss": 1.0268, "step": 65750 }, { "epoch": 0.17, "learning_rate": 8.371055412954854e-05, "loss": 1.0263, "step": 65755 }, { "epoch": 0.17, "learning_rate": 8.370929588806684e-05, "loss": 1.027, "step": 65760 }, { "epoch": 0.17, "learning_rate": 8.370803764658514e-05, "loss": 1.0247, "step": 65765 }, { "epoch": 0.17, "learning_rate": 8.370677940510343e-05, "loss": 1.0264, "step": 65770 }, { "epoch": 0.17, "learning_rate": 8.370552116362172e-05, "loss": 1.0258, "step": 65775 }, { "epoch": 0.17, "learning_rate": 8.370426292214002e-05, "loss": 1.027, "step": 65780 }, { "epoch": 0.17, "learning_rate": 8.370300468065831e-05, "loss": 1.025, "step": 65785 }, { "epoch": 0.17, "learning_rate": 8.370174643917661e-05, "loss": 1.0255, "step": 65790 }, { "epoch": 0.17, "learning_rate": 8.37004881976949e-05, "loss": 1.0239, "step": 65795 }, { "epoch": 0.17, "learning_rate": 8.36992299562132e-05, "loss": 1.0258, "step": 65800 }, { "epoch": 0.17, "learning_rate": 8.369797171473149e-05, "loss": 1.0269, "step": 65805 }, { "epoch": 0.17, "learning_rate": 8.369671347324979e-05, "loss": 1.0257, "step": 65810 }, { "epoch": 0.17, "learning_rate": 8.369545523176808e-05, "loss": 1.0307, "step": 65815 }, { "epoch": 0.17, "learning_rate": 8.369419699028638e-05, "loss": 1.0255, "step": 65820 }, { "epoch": 0.17, "learning_rate": 8.369293874880467e-05, "loss": 1.0277, "step": 65825 }, { "epoch": 0.17, "learning_rate": 8.369168050732297e-05, "loss": 1.0248, "step": 65830 }, { "epoch": 0.17, "learning_rate": 8.369042226584126e-05, "loss": 1.0243, "step": 65835 }, { "epoch": 0.17, "learning_rate": 8.368916402435956e-05, "loss": 1.0279, "step": 65840 }, { "epoch": 0.17, "learning_rate": 8.368790578287785e-05, "loss": 1.0269, "step": 65845 }, { "epoch": 0.17, "learning_rate": 8.368664754139614e-05, "loss": 1.0274, "step": 65850 }, { "epoch": 0.17, "learning_rate": 8.368538929991444e-05, "loss": 1.0289, "step": 65855 }, { "epoch": 0.17, "learning_rate": 8.368413105843274e-05, "loss": 1.0275, "step": 65860 }, { "epoch": 0.17, "learning_rate": 8.368287281695103e-05, "loss": 1.0262, "step": 65865 }, { "epoch": 0.17, "learning_rate": 8.368161457546932e-05, "loss": 1.0266, "step": 65870 }, { "epoch": 0.17, "learning_rate": 8.368035633398762e-05, "loss": 1.0277, "step": 65875 }, { "epoch": 0.17, "learning_rate": 8.367909809250592e-05, "loss": 1.0278, "step": 65880 }, { "epoch": 0.17, "learning_rate": 8.367783985102421e-05, "loss": 1.0262, "step": 65885 }, { "epoch": 0.17, "learning_rate": 8.36765816095425e-05, "loss": 1.0251, "step": 65890 }, { "epoch": 0.17, "learning_rate": 8.36753233680608e-05, "loss": 1.027, "step": 65895 }, { "epoch": 0.17, "learning_rate": 8.36740651265791e-05, "loss": 1.0263, "step": 65900 }, { "epoch": 0.17, "learning_rate": 8.367280688509739e-05, "loss": 1.0268, "step": 65905 }, { "epoch": 0.17, "learning_rate": 8.367154864361568e-05, "loss": 1.0274, "step": 65910 }, { "epoch": 0.17, "learning_rate": 8.367029040213397e-05, "loss": 1.0251, "step": 65915 }, { "epoch": 0.17, "learning_rate": 8.366903216065228e-05, "loss": 1.0253, "step": 65920 }, { "epoch": 0.17, "learning_rate": 8.366777391917057e-05, "loss": 1.0256, "step": 65925 }, { "epoch": 0.17, "learning_rate": 8.366651567768886e-05, "loss": 1.0259, "step": 65930 }, { "epoch": 0.17, "learning_rate": 8.366525743620715e-05, "loss": 1.0274, "step": 65935 }, { "epoch": 0.17, "learning_rate": 8.366399919472546e-05, "loss": 1.0239, "step": 65940 }, { "epoch": 0.17, "learning_rate": 8.366274095324375e-05, "loss": 1.0263, "step": 65945 }, { "epoch": 0.17, "learning_rate": 8.366148271176204e-05, "loss": 1.0259, "step": 65950 }, { "epoch": 0.17, "learning_rate": 8.366022447028033e-05, "loss": 1.0275, "step": 65955 }, { "epoch": 0.17, "learning_rate": 8.365896622879864e-05, "loss": 1.0255, "step": 65960 }, { "epoch": 0.17, "learning_rate": 8.365770798731693e-05, "loss": 1.0274, "step": 65965 }, { "epoch": 0.17, "learning_rate": 8.365644974583522e-05, "loss": 1.0265, "step": 65970 }, { "epoch": 0.17, "learning_rate": 8.365519150435352e-05, "loss": 1.0259, "step": 65975 }, { "epoch": 0.17, "learning_rate": 8.365393326287182e-05, "loss": 1.0255, "step": 65980 }, { "epoch": 0.17, "learning_rate": 8.365267502139012e-05, "loss": 1.0249, "step": 65985 }, { "epoch": 0.17, "learning_rate": 8.365141677990841e-05, "loss": 1.0251, "step": 65990 }, { "epoch": 0.17, "learning_rate": 8.36501585384267e-05, "loss": 1.0472, "step": 65995 }, { "epoch": 0.17, "learning_rate": 8.3648900296945e-05, "loss": 1.0268, "step": 66000 }, { "epoch": 0.17, "learning_rate": 8.364764205546329e-05, "loss": 1.027, "step": 66005 }, { "epoch": 0.17, "learning_rate": 8.364638381398159e-05, "loss": 1.0255, "step": 66010 }, { "epoch": 0.17, "learning_rate": 8.364512557249988e-05, "loss": 1.0244, "step": 66015 }, { "epoch": 0.17, "learning_rate": 8.364386733101818e-05, "loss": 1.0257, "step": 66020 }, { "epoch": 0.17, "learning_rate": 8.364260908953647e-05, "loss": 1.0258, "step": 66025 }, { "epoch": 0.17, "learning_rate": 8.364135084805477e-05, "loss": 1.0259, "step": 66030 }, { "epoch": 0.17, "learning_rate": 8.364009260657306e-05, "loss": 1.0241, "step": 66035 }, { "epoch": 0.17, "learning_rate": 8.363883436509136e-05, "loss": 1.0246, "step": 66040 }, { "epoch": 0.17, "learning_rate": 8.363757612360965e-05, "loss": 1.0251, "step": 66045 }, { "epoch": 0.17, "learning_rate": 8.363631788212795e-05, "loss": 1.0242, "step": 66050 }, { "epoch": 0.17, "learning_rate": 8.363505964064624e-05, "loss": 1.0275, "step": 66055 }, { "epoch": 0.17, "learning_rate": 8.363380139916454e-05, "loss": 1.027, "step": 66060 }, { "epoch": 0.17, "learning_rate": 8.363254315768283e-05, "loss": 1.0277, "step": 66065 }, { "epoch": 0.17, "learning_rate": 8.363128491620112e-05, "loss": 1.0247, "step": 66070 }, { "epoch": 0.17, "learning_rate": 8.363002667471942e-05, "loss": 1.0242, "step": 66075 }, { "epoch": 0.17, "learning_rate": 8.362876843323772e-05, "loss": 1.0225, "step": 66080 }, { "epoch": 0.17, "learning_rate": 8.362751019175601e-05, "loss": 1.0247, "step": 66085 }, { "epoch": 0.17, "learning_rate": 8.36262519502743e-05, "loss": 1.0268, "step": 66090 }, { "epoch": 0.17, "learning_rate": 8.36249937087926e-05, "loss": 1.0285, "step": 66095 }, { "epoch": 0.17, "learning_rate": 8.36237354673109e-05, "loss": 1.025, "step": 66100 }, { "epoch": 0.17, "learning_rate": 8.362247722582919e-05, "loss": 1.0239, "step": 66105 }, { "epoch": 0.17, "learning_rate": 8.362121898434748e-05, "loss": 1.0249, "step": 66110 }, { "epoch": 0.17, "learning_rate": 8.361996074286578e-05, "loss": 1.0257, "step": 66115 }, { "epoch": 0.17, "learning_rate": 8.361870250138408e-05, "loss": 1.0261, "step": 66120 }, { "epoch": 0.17, "learning_rate": 8.361744425990237e-05, "loss": 1.0272, "step": 66125 }, { "epoch": 0.17, "learning_rate": 8.361618601842066e-05, "loss": 1.0262, "step": 66130 }, { "epoch": 0.17, "learning_rate": 8.361492777693895e-05, "loss": 1.0279, "step": 66135 }, { "epoch": 0.17, "learning_rate": 8.361366953545726e-05, "loss": 1.027, "step": 66140 }, { "epoch": 0.17, "learning_rate": 8.361241129397555e-05, "loss": 1.0262, "step": 66145 }, { "epoch": 0.17, "learning_rate": 8.361115305249384e-05, "loss": 1.0264, "step": 66150 }, { "epoch": 0.17, "learning_rate": 8.360989481101213e-05, "loss": 1.0258, "step": 66155 }, { "epoch": 0.17, "learning_rate": 8.360863656953044e-05, "loss": 1.0254, "step": 66160 }, { "epoch": 0.17, "learning_rate": 8.360737832804873e-05, "loss": 1.0248, "step": 66165 }, { "epoch": 0.17, "learning_rate": 8.360612008656702e-05, "loss": 1.0256, "step": 66170 }, { "epoch": 0.17, "learning_rate": 8.360486184508531e-05, "loss": 1.0247, "step": 66175 }, { "epoch": 0.17, "learning_rate": 8.360360360360362e-05, "loss": 1.0247, "step": 66180 }, { "epoch": 0.17, "learning_rate": 8.36023453621219e-05, "loss": 1.0262, "step": 66185 }, { "epoch": 0.17, "learning_rate": 8.36010871206402e-05, "loss": 1.0261, "step": 66190 }, { "epoch": 0.17, "learning_rate": 8.359982887915849e-05, "loss": 1.0265, "step": 66195 }, { "epoch": 0.17, "learning_rate": 8.359857063767678e-05, "loss": 1.0245, "step": 66200 }, { "epoch": 0.17, "learning_rate": 8.359731239619509e-05, "loss": 1.0248, "step": 66205 }, { "epoch": 0.17, "learning_rate": 8.359605415471338e-05, "loss": 1.027, "step": 66210 }, { "epoch": 0.17, "learning_rate": 8.359479591323167e-05, "loss": 1.0263, "step": 66215 }, { "epoch": 0.17, "learning_rate": 8.359353767174996e-05, "loss": 1.0253, "step": 66220 }, { "epoch": 0.17, "learning_rate": 8.359227943026827e-05, "loss": 1.0288, "step": 66225 }, { "epoch": 0.17, "learning_rate": 8.359102118878656e-05, "loss": 1.0261, "step": 66230 }, { "epoch": 0.17, "learning_rate": 8.358976294730485e-05, "loss": 1.0265, "step": 66235 }, { "epoch": 0.17, "learning_rate": 8.358850470582314e-05, "loss": 1.0268, "step": 66240 }, { "epoch": 0.17, "learning_rate": 8.358724646434145e-05, "loss": 1.0276, "step": 66245 }, { "epoch": 0.17, "learning_rate": 8.358598822285974e-05, "loss": 1.0241, "step": 66250 }, { "epoch": 0.17, "learning_rate": 8.358472998137803e-05, "loss": 1.0249, "step": 66255 }, { "epoch": 0.17, "learning_rate": 8.358347173989632e-05, "loss": 1.0448, "step": 66260 }, { "epoch": 0.17, "learning_rate": 8.358221349841461e-05, "loss": 1.0249, "step": 66265 }, { "epoch": 0.17, "learning_rate": 8.358095525693292e-05, "loss": 1.0248, "step": 66270 }, { "epoch": 0.17, "learning_rate": 8.357969701545121e-05, "loss": 1.0247, "step": 66275 }, { "epoch": 0.17, "learning_rate": 8.35784387739695e-05, "loss": 1.0263, "step": 66280 }, { "epoch": 0.17, "learning_rate": 8.357718053248779e-05, "loss": 1.0237, "step": 66285 }, { "epoch": 0.17, "learning_rate": 8.35759222910061e-05, "loss": 1.0227, "step": 66290 }, { "epoch": 0.17, "learning_rate": 8.357466404952439e-05, "loss": 1.0273, "step": 66295 }, { "epoch": 0.17, "learning_rate": 8.357340580804268e-05, "loss": 1.0264, "step": 66300 }, { "epoch": 0.17, "learning_rate": 8.357214756656097e-05, "loss": 1.0251, "step": 66305 }, { "epoch": 0.17, "learning_rate": 8.357088932507928e-05, "loss": 1.0275, "step": 66310 }, { "epoch": 0.17, "learning_rate": 8.356963108359757e-05, "loss": 1.025, "step": 66315 }, { "epoch": 0.17, "learning_rate": 8.356837284211586e-05, "loss": 1.0261, "step": 66320 }, { "epoch": 0.17, "learning_rate": 8.356711460063415e-05, "loss": 1.027, "step": 66325 }, { "epoch": 0.17, "learning_rate": 8.356585635915244e-05, "loss": 1.025, "step": 66330 }, { "epoch": 0.17, "learning_rate": 8.356459811767075e-05, "loss": 1.0253, "step": 66335 }, { "epoch": 0.17, "learning_rate": 8.356333987618904e-05, "loss": 1.0255, "step": 66340 }, { "epoch": 0.17, "learning_rate": 8.356208163470733e-05, "loss": 1.0271, "step": 66345 }, { "epoch": 0.17, "learning_rate": 8.356082339322562e-05, "loss": 1.0264, "step": 66350 }, { "epoch": 0.17, "learning_rate": 8.355956515174393e-05, "loss": 1.0257, "step": 66355 }, { "epoch": 0.17, "learning_rate": 8.355830691026222e-05, "loss": 1.0283, "step": 66360 }, { "epoch": 0.17, "learning_rate": 8.355704866878051e-05, "loss": 1.0263, "step": 66365 }, { "epoch": 0.17, "learning_rate": 8.35557904272988e-05, "loss": 1.0266, "step": 66370 }, { "epoch": 0.17, "learning_rate": 8.355453218581711e-05, "loss": 1.045, "step": 66375 }, { "epoch": 0.17, "learning_rate": 8.35532739443354e-05, "loss": 1.0272, "step": 66380 }, { "epoch": 0.17, "learning_rate": 8.355201570285369e-05, "loss": 1.0462, "step": 66385 }, { "epoch": 0.17, "learning_rate": 8.355075746137198e-05, "loss": 1.0248, "step": 66390 }, { "epoch": 0.17, "learning_rate": 8.354949921989028e-05, "loss": 1.0276, "step": 66395 }, { "epoch": 0.17, "learning_rate": 8.354824097840858e-05, "loss": 1.0253, "step": 66400 }, { "epoch": 0.17, "learning_rate": 8.354698273692687e-05, "loss": 1.0476, "step": 66405 }, { "epoch": 0.17, "learning_rate": 8.354572449544516e-05, "loss": 1.0272, "step": 66410 }, { "epoch": 0.17, "learning_rate": 8.354446625396346e-05, "loss": 1.0245, "step": 66415 }, { "epoch": 0.17, "learning_rate": 8.354320801248176e-05, "loss": 1.0243, "step": 66420 }, { "epoch": 0.17, "learning_rate": 8.354194977100005e-05, "loss": 1.0266, "step": 66425 }, { "epoch": 0.17, "learning_rate": 8.354069152951834e-05, "loss": 1.0257, "step": 66430 }, { "epoch": 0.17, "learning_rate": 8.353943328803664e-05, "loss": 1.0251, "step": 66435 }, { "epoch": 0.17, "learning_rate": 8.353817504655493e-05, "loss": 1.0264, "step": 66440 }, { "epoch": 0.17, "learning_rate": 8.353691680507323e-05, "loss": 1.0253, "step": 66445 }, { "epoch": 0.17, "learning_rate": 8.353565856359152e-05, "loss": 1.0256, "step": 66450 }, { "epoch": 0.17, "learning_rate": 8.353440032210982e-05, "loss": 1.0277, "step": 66455 }, { "epoch": 0.17, "learning_rate": 8.353314208062811e-05, "loss": 1.0239, "step": 66460 }, { "epoch": 0.17, "learning_rate": 8.353188383914641e-05, "loss": 1.0252, "step": 66465 }, { "epoch": 0.17, "learning_rate": 8.35306255976647e-05, "loss": 1.0236, "step": 66470 }, { "epoch": 0.17, "learning_rate": 8.352936735618301e-05, "loss": 1.0261, "step": 66475 }, { "epoch": 0.17, "learning_rate": 8.35281091147013e-05, "loss": 1.0253, "step": 66480 }, { "epoch": 0.17, "learning_rate": 8.352685087321959e-05, "loss": 1.0263, "step": 66485 }, { "epoch": 0.17, "learning_rate": 8.35255926317379e-05, "loss": 1.0246, "step": 66490 }, { "epoch": 0.17, "learning_rate": 8.352433439025619e-05, "loss": 1.0267, "step": 66495 }, { "epoch": 0.17, "learning_rate": 8.352307614877448e-05, "loss": 1.0258, "step": 66500 }, { "epoch": 0.17, "learning_rate": 8.352181790729277e-05, "loss": 1.0251, "step": 66505 }, { "epoch": 0.17, "learning_rate": 8.352055966581108e-05, "loss": 1.0272, "step": 66510 }, { "epoch": 0.17, "learning_rate": 8.351930142432937e-05, "loss": 1.0238, "step": 66515 }, { "epoch": 0.17, "learning_rate": 8.351804318284766e-05, "loss": 1.0259, "step": 66520 }, { "epoch": 0.17, "learning_rate": 8.351678494136595e-05, "loss": 1.0274, "step": 66525 }, { "epoch": 0.17, "learning_rate": 8.351552669988424e-05, "loss": 1.0279, "step": 66530 }, { "epoch": 0.17, "learning_rate": 8.351426845840255e-05, "loss": 1.0277, "step": 66535 }, { "epoch": 0.17, "learning_rate": 8.351301021692084e-05, "loss": 1.0287, "step": 66540 }, { "epoch": 0.17, "learning_rate": 8.351175197543913e-05, "loss": 1.0232, "step": 66545 }, { "epoch": 0.17, "learning_rate": 8.351049373395742e-05, "loss": 1.0271, "step": 66550 }, { "epoch": 0.17, "learning_rate": 8.350923549247573e-05, "loss": 1.0248, "step": 66555 }, { "epoch": 0.17, "learning_rate": 8.350797725099402e-05, "loss": 1.0255, "step": 66560 }, { "epoch": 0.17, "learning_rate": 8.350671900951231e-05, "loss": 1.0238, "step": 66565 }, { "epoch": 0.17, "learning_rate": 8.35054607680306e-05, "loss": 1.0258, "step": 66570 }, { "epoch": 0.17, "learning_rate": 8.350420252654891e-05, "loss": 1.0263, "step": 66575 }, { "epoch": 0.17, "learning_rate": 8.350369922995622e-05, "loss": 1.0481, "step": 66580 }, { "epoch": 0.17, "learning_rate": 8.350244098847451e-05, "loss": 1.0265, "step": 66585 }, { "epoch": 0.17, "learning_rate": 8.350118274699282e-05, "loss": 1.0242, "step": 66590 }, { "epoch": 0.17, "learning_rate": 8.349992450551111e-05, "loss": 1.0266, "step": 66595 }, { "epoch": 0.17, "learning_rate": 8.34986662640294e-05, "loss": 1.0261, "step": 66600 }, { "epoch": 0.17, "learning_rate": 8.349740802254769e-05, "loss": 1.0245, "step": 66605 }, { "epoch": 0.17, "learning_rate": 8.3496149781066e-05, "loss": 1.0265, "step": 66610 }, { "epoch": 0.17, "learning_rate": 8.349489153958429e-05, "loss": 1.0264, "step": 66615 }, { "epoch": 0.17, "learning_rate": 8.349363329810258e-05, "loss": 1.0261, "step": 66620 }, { "epoch": 0.17, "learning_rate": 8.349237505662087e-05, "loss": 1.0245, "step": 66625 }, { "epoch": 0.17, "learning_rate": 8.349111681513916e-05, "loss": 1.0283, "step": 66630 }, { "epoch": 0.17, "learning_rate": 8.348985857365747e-05, "loss": 1.0273, "step": 66635 }, { "epoch": 0.17, "learning_rate": 8.348860033217576e-05, "loss": 1.0247, "step": 66640 }, { "epoch": 0.17, "learning_rate": 8.348734209069405e-05, "loss": 1.0252, "step": 66645 }, { "epoch": 0.17, "learning_rate": 8.348608384921234e-05, "loss": 1.0269, "step": 66650 }, { "epoch": 0.17, "learning_rate": 8.348482560773065e-05, "loss": 1.0475, "step": 66655 }, { "epoch": 0.17, "learning_rate": 8.348356736624894e-05, "loss": 1.0254, "step": 66660 }, { "epoch": 0.17, "learning_rate": 8.348230912476723e-05, "loss": 1.0268, "step": 66665 }, { "epoch": 0.17, "learning_rate": 8.348105088328552e-05, "loss": 1.0283, "step": 66670 }, { "epoch": 0.17, "learning_rate": 8.347979264180383e-05, "loss": 1.024, "step": 66675 }, { "epoch": 0.17, "learning_rate": 8.347853440032212e-05, "loss": 1.0253, "step": 66680 }, { "epoch": 0.17, "learning_rate": 8.347727615884041e-05, "loss": 1.0279, "step": 66685 }, { "epoch": 0.17, "learning_rate": 8.34760179173587e-05, "loss": 1.0276, "step": 66690 }, { "epoch": 0.17, "learning_rate": 8.3474759675877e-05, "loss": 1.0275, "step": 66695 }, { "epoch": 0.17, "learning_rate": 8.34735014343953e-05, "loss": 1.0247, "step": 66700 }, { "epoch": 0.17, "learning_rate": 8.347224319291359e-05, "loss": 1.0457, "step": 66705 }, { "epoch": 0.17, "learning_rate": 8.347098495143188e-05, "loss": 1.0266, "step": 66710 }, { "epoch": 0.17, "learning_rate": 8.346972670995017e-05, "loss": 1.0249, "step": 66715 }, { "epoch": 0.17, "learning_rate": 8.346846846846848e-05, "loss": 1.0253, "step": 66720 }, { "epoch": 0.17, "learning_rate": 8.346721022698677e-05, "loss": 1.0266, "step": 66725 }, { "epoch": 0.17, "learning_rate": 8.346595198550506e-05, "loss": 1.0268, "step": 66730 }, { "epoch": 0.17, "learning_rate": 8.346469374402335e-05, "loss": 1.0231, "step": 66735 }, { "epoch": 0.17, "learning_rate": 8.346343550254166e-05, "loss": 1.0268, "step": 66740 }, { "epoch": 0.17, "learning_rate": 8.346217726105995e-05, "loss": 1.0258, "step": 66745 }, { "epoch": 0.17, "learning_rate": 8.346091901957824e-05, "loss": 1.0284, "step": 66750 }, { "epoch": 0.17, "learning_rate": 8.345966077809653e-05, "loss": 1.0247, "step": 66755 }, { "epoch": 0.17, "learning_rate": 8.345840253661483e-05, "loss": 1.0253, "step": 66760 }, { "epoch": 0.17, "learning_rate": 8.345714429513313e-05, "loss": 1.0254, "step": 66765 }, { "epoch": 0.17, "learning_rate": 8.345588605365142e-05, "loss": 1.028, "step": 66770 }, { "epoch": 0.17, "learning_rate": 8.345462781216971e-05, "loss": 1.0234, "step": 66775 }, { "epoch": 0.17, "learning_rate": 8.3453369570688e-05, "loss": 1.0266, "step": 66780 }, { "epoch": 0.17, "learning_rate": 8.345211132920631e-05, "loss": 1.0258, "step": 66785 }, { "epoch": 0.17, "learning_rate": 8.34508530877246e-05, "loss": 1.0244, "step": 66790 }, { "epoch": 0.17, "learning_rate": 8.34495948462429e-05, "loss": 1.0252, "step": 66795 }, { "epoch": 0.17, "learning_rate": 8.344833660476119e-05, "loss": 1.0268, "step": 66800 }, { "epoch": 0.17, "learning_rate": 8.344707836327949e-05, "loss": 1.0257, "step": 66805 }, { "epoch": 0.17, "learning_rate": 8.344582012179778e-05, "loss": 1.0267, "step": 66810 }, { "epoch": 0.17, "learning_rate": 8.344456188031607e-05, "loss": 1.0256, "step": 66815 }, { "epoch": 0.17, "learning_rate": 8.344330363883437e-05, "loss": 1.0286, "step": 66820 }, { "epoch": 0.17, "learning_rate": 8.344204539735266e-05, "loss": 1.026, "step": 66825 }, { "epoch": 0.17, "learning_rate": 8.344078715587096e-05, "loss": 1.0258, "step": 66830 }, { "epoch": 0.17, "learning_rate": 8.343952891438925e-05, "loss": 1.0249, "step": 66835 }, { "epoch": 0.17, "learning_rate": 8.343827067290755e-05, "loss": 1.0238, "step": 66840 }, { "epoch": 0.17, "learning_rate": 8.343701243142584e-05, "loss": 1.0249, "step": 66845 }, { "epoch": 0.17, "learning_rate": 8.343575418994414e-05, "loss": 1.0277, "step": 66850 }, { "epoch": 0.17, "learning_rate": 8.343449594846243e-05, "loss": 1.0262, "step": 66855 }, { "epoch": 0.17, "learning_rate": 8.343323770698073e-05, "loss": 1.0254, "step": 66860 }, { "epoch": 0.17, "learning_rate": 8.343197946549902e-05, "loss": 1.0258, "step": 66865 }, { "epoch": 0.17, "learning_rate": 8.343072122401732e-05, "loss": 1.0247, "step": 66870 }, { "epoch": 0.17, "learning_rate": 8.342946298253561e-05, "loss": 1.0265, "step": 66875 }, { "epoch": 0.17, "learning_rate": 8.34282047410539e-05, "loss": 1.0254, "step": 66880 }, { "epoch": 0.17, "learning_rate": 8.34269464995722e-05, "loss": 1.0271, "step": 66885 }, { "epoch": 0.17, "learning_rate": 8.342568825809049e-05, "loss": 1.0397, "step": 66890 }, { "epoch": 0.17, "learning_rate": 8.34244300166088e-05, "loss": 1.0257, "step": 66895 }, { "epoch": 0.17, "learning_rate": 8.342317177512709e-05, "loss": 1.0271, "step": 66900 }, { "epoch": 0.17, "learning_rate": 8.342191353364538e-05, "loss": 1.0244, "step": 66905 }, { "epoch": 0.17, "learning_rate": 8.342065529216367e-05, "loss": 1.0254, "step": 66910 }, { "epoch": 0.17, "learning_rate": 8.341939705068197e-05, "loss": 1.0263, "step": 66915 }, { "epoch": 0.17, "learning_rate": 8.341813880920027e-05, "loss": 1.0255, "step": 66920 }, { "epoch": 0.17, "learning_rate": 8.341688056771856e-05, "loss": 1.0266, "step": 66925 }, { "epoch": 0.17, "learning_rate": 8.341562232623685e-05, "loss": 1.0269, "step": 66930 }, { "epoch": 0.17, "learning_rate": 8.341436408475515e-05, "loss": 1.0264, "step": 66935 }, { "epoch": 0.17, "learning_rate": 8.341310584327345e-05, "loss": 1.0259, "step": 66940 }, { "epoch": 0.17, "learning_rate": 8.341184760179174e-05, "loss": 1.0401, "step": 66945 }, { "epoch": 0.17, "learning_rate": 8.341058936031003e-05, "loss": 1.0269, "step": 66950 }, { "epoch": 0.17, "learning_rate": 8.340933111882832e-05, "loss": 1.0265, "step": 66955 }, { "epoch": 0.17, "learning_rate": 8.340807287734663e-05, "loss": 1.0268, "step": 66960 }, { "epoch": 0.17, "learning_rate": 8.340681463586492e-05, "loss": 1.0259, "step": 66965 }, { "epoch": 0.17, "learning_rate": 8.340555639438321e-05, "loss": 1.0253, "step": 66970 }, { "epoch": 0.17, "learning_rate": 8.34042981529015e-05, "loss": 1.0262, "step": 66975 }, { "epoch": 0.17, "learning_rate": 8.34030399114198e-05, "loss": 1.0259, "step": 66980 }, { "epoch": 0.17, "learning_rate": 8.34017816699381e-05, "loss": 1.0272, "step": 66985 }, { "epoch": 0.17, "learning_rate": 8.340052342845639e-05, "loss": 1.029, "step": 66990 }, { "epoch": 0.17, "learning_rate": 8.339926518697468e-05, "loss": 1.0246, "step": 66995 }, { "epoch": 0.17, "learning_rate": 8.339800694549299e-05, "loss": 1.0259, "step": 67000 }, { "epoch": 0.17, "learning_rate": 8.339674870401128e-05, "loss": 1.0247, "step": 67005 }, { "epoch": 0.17, "learning_rate": 8.339549046252957e-05, "loss": 1.0271, "step": 67010 }, { "epoch": 0.17, "learning_rate": 8.339423222104786e-05, "loss": 1.0239, "step": 67015 }, { "epoch": 0.17, "learning_rate": 8.339297397956615e-05, "loss": 1.027, "step": 67020 }, { "epoch": 0.17, "learning_rate": 8.339171573808446e-05, "loss": 1.0257, "step": 67025 }, { "epoch": 0.17, "learning_rate": 8.339045749660275e-05, "loss": 1.0285, "step": 67030 }, { "epoch": 0.17, "learning_rate": 8.338919925512104e-05, "loss": 1.0263, "step": 67035 }, { "epoch": 0.17, "learning_rate": 8.338794101363933e-05, "loss": 1.0274, "step": 67040 }, { "epoch": 0.17, "learning_rate": 8.338668277215764e-05, "loss": 1.0241, "step": 67045 }, { "epoch": 0.17, "learning_rate": 8.338542453067593e-05, "loss": 1.025, "step": 67050 }, { "epoch": 0.17, "learning_rate": 8.338416628919422e-05, "loss": 1.0245, "step": 67055 }, { "epoch": 0.17, "learning_rate": 8.338290804771251e-05, "loss": 1.024, "step": 67060 }, { "epoch": 0.17, "learning_rate": 8.338164980623082e-05, "loss": 1.0278, "step": 67065 }, { "epoch": 0.17, "learning_rate": 8.338039156474911e-05, "loss": 1.0246, "step": 67070 }, { "epoch": 0.17, "learning_rate": 8.337913332326741e-05, "loss": 1.0259, "step": 67075 }, { "epoch": 0.17, "learning_rate": 8.33778750817857e-05, "loss": 1.027, "step": 67080 }, { "epoch": 0.17, "learning_rate": 8.3376616840304e-05, "loss": 1.0256, "step": 67085 }, { "epoch": 0.17, "learning_rate": 8.33753585988223e-05, "loss": 1.0261, "step": 67090 }, { "epoch": 0.17, "learning_rate": 8.33741003573406e-05, "loss": 1.0266, "step": 67095 }, { "epoch": 0.17, "learning_rate": 8.337284211585888e-05, "loss": 1.0276, "step": 67100 }, { "epoch": 0.17, "learning_rate": 8.337158387437718e-05, "loss": 1.0261, "step": 67105 }, { "epoch": 0.17, "learning_rate": 8.337032563289547e-05, "loss": 1.027, "step": 67110 }, { "epoch": 0.17, "learning_rate": 8.336906739141377e-05, "loss": 1.0233, "step": 67115 }, { "epoch": 0.17, "learning_rate": 8.336780914993206e-05, "loss": 1.0263, "step": 67120 }, { "epoch": 0.17, "learning_rate": 8.336655090845036e-05, "loss": 1.0256, "step": 67125 }, { "epoch": 0.17, "learning_rate": 8.336529266696865e-05, "loss": 1.0264, "step": 67130 }, { "epoch": 0.17, "learning_rate": 8.336403442548695e-05, "loss": 1.0234, "step": 67135 }, { "epoch": 0.17, "learning_rate": 8.336277618400524e-05, "loss": 1.0254, "step": 67140 }, { "epoch": 0.17, "learning_rate": 8.336151794252354e-05, "loss": 1.0239, "step": 67145 }, { "epoch": 0.17, "learning_rate": 8.336025970104183e-05, "loss": 1.0258, "step": 67150 }, { "epoch": 0.17, "learning_rate": 8.335900145956013e-05, "loss": 1.0262, "step": 67155 }, { "epoch": 0.17, "learning_rate": 8.335774321807842e-05, "loss": 1.0258, "step": 67160 }, { "epoch": 0.17, "learning_rate": 8.335648497659672e-05, "loss": 1.0253, "step": 67165 }, { "epoch": 0.17, "learning_rate": 8.335522673511501e-05, "loss": 1.0266, "step": 67170 }, { "epoch": 0.17, "learning_rate": 8.33539684936333e-05, "loss": 1.0242, "step": 67175 }, { "epoch": 0.17, "learning_rate": 8.33527102521516e-05, "loss": 1.0243, "step": 67180 }, { "epoch": 0.17, "learning_rate": 8.33514520106699e-05, "loss": 1.0275, "step": 67185 }, { "epoch": 0.17, "learning_rate": 8.335019376918819e-05, "loss": 1.0259, "step": 67190 }, { "epoch": 0.17, "learning_rate": 8.334893552770648e-05, "loss": 1.0263, "step": 67195 }, { "epoch": 0.17, "learning_rate": 8.334767728622478e-05, "loss": 1.0265, "step": 67200 }, { "epoch": 0.17, "learning_rate": 8.334641904474308e-05, "loss": 1.0267, "step": 67205 }, { "epoch": 0.17, "learning_rate": 8.334516080326137e-05, "loss": 1.0247, "step": 67210 }, { "epoch": 0.17, "learning_rate": 8.334390256177966e-05, "loss": 1.0261, "step": 67215 }, { "epoch": 0.17, "learning_rate": 8.334264432029796e-05, "loss": 1.0242, "step": 67220 }, { "epoch": 0.17, "learning_rate": 8.334138607881626e-05, "loss": 1.0236, "step": 67225 }, { "epoch": 0.17, "learning_rate": 8.334012783733455e-05, "loss": 1.0252, "step": 67230 }, { "epoch": 0.17, "learning_rate": 8.333886959585284e-05, "loss": 1.0252, "step": 67235 }, { "epoch": 0.17, "learning_rate": 8.333761135437113e-05, "loss": 1.0251, "step": 67240 }, { "epoch": 0.17, "learning_rate": 8.333635311288944e-05, "loss": 1.0272, "step": 67245 }, { "epoch": 0.17, "learning_rate": 8.333509487140773e-05, "loss": 1.0242, "step": 67250 }, { "epoch": 0.17, "learning_rate": 8.333383662992602e-05, "loss": 1.0451, "step": 67255 }, { "epoch": 0.17, "learning_rate": 8.333257838844431e-05, "loss": 1.0261, "step": 67260 }, { "epoch": 0.17, "learning_rate": 8.333132014696262e-05, "loss": 1.0255, "step": 67265 }, { "epoch": 0.17, "learning_rate": 8.333006190548091e-05, "loss": 1.0284, "step": 67270 }, { "epoch": 0.17, "learning_rate": 8.33288036639992e-05, "loss": 1.0276, "step": 67275 }, { "epoch": 0.17, "learning_rate": 8.332754542251749e-05, "loss": 1.025, "step": 67280 }, { "epoch": 0.17, "learning_rate": 8.332628718103578e-05, "loss": 1.0263, "step": 67285 }, { "epoch": 0.17, "learning_rate": 8.332502893955409e-05, "loss": 1.0259, "step": 67290 }, { "epoch": 0.17, "learning_rate": 8.332377069807238e-05, "loss": 1.0244, "step": 67295 }, { "epoch": 0.17, "learning_rate": 8.332251245659067e-05, "loss": 1.0246, "step": 67300 }, { "epoch": 0.17, "learning_rate": 8.332125421510896e-05, "loss": 1.0256, "step": 67305 }, { "epoch": 0.17, "learning_rate": 8.331999597362727e-05, "loss": 1.0265, "step": 67310 }, { "epoch": 0.17, "learning_rate": 8.331873773214556e-05, "loss": 1.0244, "step": 67315 }, { "epoch": 0.17, "learning_rate": 8.331747949066385e-05, "loss": 1.0269, "step": 67320 }, { "epoch": 0.17, "learning_rate": 8.331622124918214e-05, "loss": 1.0267, "step": 67325 }, { "epoch": 0.17, "learning_rate": 8.331496300770045e-05, "loss": 1.0227, "step": 67330 }, { "epoch": 0.17, "learning_rate": 8.331370476621874e-05, "loss": 1.0271, "step": 67335 }, { "epoch": 0.17, "learning_rate": 8.331244652473703e-05, "loss": 1.0254, "step": 67340 }, { "epoch": 0.17, "learning_rate": 8.331118828325532e-05, "loss": 1.0267, "step": 67345 }, { "epoch": 0.17, "learning_rate": 8.330993004177361e-05, "loss": 1.0262, "step": 67350 }, { "epoch": 0.17, "learning_rate": 8.330867180029192e-05, "loss": 1.0247, "step": 67355 }, { "epoch": 0.17, "learning_rate": 8.330741355881021e-05, "loss": 1.0273, "step": 67360 }, { "epoch": 0.17, "learning_rate": 8.33061553173285e-05, "loss": 1.0251, "step": 67365 }, { "epoch": 0.17, "learning_rate": 8.33048970758468e-05, "loss": 1.0261, "step": 67370 }, { "epoch": 0.17, "learning_rate": 8.33036388343651e-05, "loss": 1.0276, "step": 67375 }, { "epoch": 0.17, "learning_rate": 8.330238059288339e-05, "loss": 1.0253, "step": 67380 }, { "epoch": 0.17, "learning_rate": 8.330112235140168e-05, "loss": 1.0266, "step": 67385 }, { "epoch": 0.17, "learning_rate": 8.329986410991997e-05, "loss": 1.0248, "step": 67390 }, { "epoch": 0.17, "learning_rate": 8.329860586843828e-05, "loss": 1.0255, "step": 67395 }, { "epoch": 0.17, "learning_rate": 8.329734762695657e-05, "loss": 1.0257, "step": 67400 }, { "epoch": 0.17, "learning_rate": 8.329608938547486e-05, "loss": 1.0267, "step": 67405 }, { "epoch": 0.17, "learning_rate": 8.329483114399315e-05, "loss": 1.0244, "step": 67410 }, { "epoch": 0.17, "learning_rate": 8.329357290251145e-05, "loss": 1.0278, "step": 67415 }, { "epoch": 0.17, "learning_rate": 8.329231466102975e-05, "loss": 1.0242, "step": 67420 }, { "epoch": 0.17, "learning_rate": 8.329105641954804e-05, "loss": 1.0262, "step": 67425 }, { "epoch": 0.17, "learning_rate": 8.328979817806633e-05, "loss": 1.0245, "step": 67430 }, { "epoch": 0.17, "learning_rate": 8.328853993658463e-05, "loss": 1.0262, "step": 67435 }, { "epoch": 0.17, "learning_rate": 8.328728169510293e-05, "loss": 1.0252, "step": 67440 }, { "epoch": 0.17, "learning_rate": 8.328602345362122e-05, "loss": 1.0224, "step": 67445 }, { "epoch": 0.17, "learning_rate": 8.328476521213951e-05, "loss": 1.0237, "step": 67450 }, { "epoch": 0.17, "learning_rate": 8.32835069706578e-05, "loss": 1.0269, "step": 67455 }, { "epoch": 0.17, "learning_rate": 8.328224872917611e-05, "loss": 1.0249, "step": 67460 }, { "epoch": 0.17, "learning_rate": 8.32809904876944e-05, "loss": 1.0253, "step": 67465 }, { "epoch": 0.17, "learning_rate": 8.32797322462127e-05, "loss": 1.0277, "step": 67470 }, { "epoch": 0.17, "learning_rate": 8.327847400473099e-05, "loss": 1.0255, "step": 67475 }, { "epoch": 0.17, "learning_rate": 8.327721576324928e-05, "loss": 1.0236, "step": 67480 }, { "epoch": 0.17, "learning_rate": 8.327595752176758e-05, "loss": 1.0277, "step": 67485 }, { "epoch": 0.17, "learning_rate": 8.327469928028587e-05, "loss": 1.0241, "step": 67490 }, { "epoch": 0.17, "learning_rate": 8.327344103880417e-05, "loss": 1.027, "step": 67495 }, { "epoch": 0.17, "learning_rate": 8.327218279732246e-05, "loss": 1.0266, "step": 67500 }, { "epoch": 0.17, "learning_rate": 8.327092455584076e-05, "loss": 1.0261, "step": 67505 }, { "epoch": 0.17, "learning_rate": 8.326966631435905e-05, "loss": 1.0254, "step": 67510 }, { "epoch": 0.17, "learning_rate": 8.326840807287735e-05, "loss": 1.0257, "step": 67515 }, { "epoch": 0.17, "learning_rate": 8.326714983139564e-05, "loss": 1.0256, "step": 67520 }, { "epoch": 0.17, "learning_rate": 8.326589158991394e-05, "loss": 1.024, "step": 67525 }, { "epoch": 0.17, "learning_rate": 8.326463334843223e-05, "loss": 1.0259, "step": 67530 }, { "epoch": 0.17, "learning_rate": 8.326337510695052e-05, "loss": 1.0272, "step": 67535 }, { "epoch": 0.17, "learning_rate": 8.326211686546882e-05, "loss": 1.0283, "step": 67540 }, { "epoch": 0.17, "learning_rate": 8.326085862398711e-05, "loss": 1.0251, "step": 67545 }, { "epoch": 0.17, "learning_rate": 8.325960038250541e-05, "loss": 1.0261, "step": 67550 }, { "epoch": 0.17, "learning_rate": 8.32583421410237e-05, "loss": 1.0233, "step": 67555 }, { "epoch": 0.17, "learning_rate": 8.3257083899542e-05, "loss": 1.0263, "step": 67560 }, { "epoch": 0.17, "learning_rate": 8.325582565806029e-05, "loss": 1.0245, "step": 67565 }, { "epoch": 0.17, "learning_rate": 8.32545674165786e-05, "loss": 1.0265, "step": 67570 }, { "epoch": 0.17, "learning_rate": 8.32533091750969e-05, "loss": 1.0273, "step": 67575 }, { "epoch": 0.17, "learning_rate": 8.325205093361519e-05, "loss": 1.0257, "step": 67580 }, { "epoch": 0.17, "learning_rate": 8.325079269213348e-05, "loss": 1.0485, "step": 67585 }, { "epoch": 0.17, "learning_rate": 8.324953445065177e-05, "loss": 1.0261, "step": 67590 }, { "epoch": 0.17, "learning_rate": 8.324827620917008e-05, "loss": 1.0251, "step": 67595 }, { "epoch": 0.17, "learning_rate": 8.324701796768837e-05, "loss": 1.025, "step": 67600 }, { "epoch": 0.17, "learning_rate": 8.324575972620666e-05, "loss": 1.0235, "step": 67605 }, { "epoch": 0.17, "learning_rate": 8.324450148472495e-05, "loss": 1.0247, "step": 67610 }, { "epoch": 0.17, "learning_rate": 8.324324324324326e-05, "loss": 1.0246, "step": 67615 }, { "epoch": 0.17, "learning_rate": 8.324198500176155e-05, "loss": 1.0284, "step": 67620 }, { "epoch": 0.17, "learning_rate": 8.324072676027984e-05, "loss": 1.0271, "step": 67625 }, { "epoch": 0.17, "learning_rate": 8.323946851879813e-05, "loss": 1.0256, "step": 67630 }, { "epoch": 0.17, "learning_rate": 8.323821027731642e-05, "loss": 1.025, "step": 67635 }, { "epoch": 0.17, "learning_rate": 8.323695203583473e-05, "loss": 1.0472, "step": 67640 }, { "epoch": 0.17, "learning_rate": 8.323569379435302e-05, "loss": 1.0251, "step": 67645 }, { "epoch": 0.17, "learning_rate": 8.323443555287131e-05, "loss": 1.0243, "step": 67650 }, { "epoch": 0.17, "learning_rate": 8.32331773113896e-05, "loss": 1.0265, "step": 67655 }, { "epoch": 0.17, "learning_rate": 8.323191906990791e-05, "loss": 1.0272, "step": 67660 }, { "epoch": 0.17, "learning_rate": 8.32306608284262e-05, "loss": 1.0242, "step": 67665 }, { "epoch": 0.17, "learning_rate": 8.322940258694449e-05, "loss": 1.0248, "step": 67670 }, { "epoch": 0.17, "learning_rate": 8.322814434546278e-05, "loss": 1.0278, "step": 67675 }, { "epoch": 0.17, "learning_rate": 8.322688610398109e-05, "loss": 1.0427, "step": 67680 }, { "epoch": 0.17, "learning_rate": 8.322562786249938e-05, "loss": 1.0263, "step": 67685 }, { "epoch": 0.17, "learning_rate": 8.322436962101767e-05, "loss": 1.0263, "step": 67690 }, { "epoch": 0.17, "learning_rate": 8.322311137953596e-05, "loss": 1.0257, "step": 67695 }, { "epoch": 0.17, "learning_rate": 8.322185313805426e-05, "loss": 1.0255, "step": 67700 }, { "epoch": 0.17, "learning_rate": 8.322059489657256e-05, "loss": 1.025, "step": 67705 }, { "epoch": 0.17, "learning_rate": 8.321933665509085e-05, "loss": 1.0251, "step": 67710 }, { "epoch": 0.17, "learning_rate": 8.321807841360914e-05, "loss": 1.0278, "step": 67715 }, { "epoch": 0.17, "learning_rate": 8.321682017212744e-05, "loss": 1.0258, "step": 67720 }, { "epoch": 0.17, "learning_rate": 8.321556193064574e-05, "loss": 1.026, "step": 67725 }, { "epoch": 0.17, "learning_rate": 8.321430368916403e-05, "loss": 1.023, "step": 67730 }, { "epoch": 0.17, "learning_rate": 8.321304544768232e-05, "loss": 1.0267, "step": 67735 }, { "epoch": 0.17, "learning_rate": 8.321178720620062e-05, "loss": 1.026, "step": 67740 }, { "epoch": 0.17, "learning_rate": 8.321052896471892e-05, "loss": 1.0268, "step": 67745 }, { "epoch": 0.17, "learning_rate": 8.320927072323721e-05, "loss": 1.0288, "step": 67750 }, { "epoch": 0.17, "learning_rate": 8.32080124817555e-05, "loss": 1.0244, "step": 67755 }, { "epoch": 0.17, "learning_rate": 8.32067542402738e-05, "loss": 1.0235, "step": 67760 }, { "epoch": 0.17, "learning_rate": 8.320549599879209e-05, "loss": 1.0276, "step": 67765 }, { "epoch": 0.17, "learning_rate": 8.320423775731039e-05, "loss": 1.0277, "step": 67770 }, { "epoch": 0.17, "learning_rate": 8.320297951582868e-05, "loss": 1.0251, "step": 67775 }, { "epoch": 0.17, "learning_rate": 8.320172127434698e-05, "loss": 1.0228, "step": 67780 }, { "epoch": 0.17, "learning_rate": 8.320046303286527e-05, "loss": 1.0271, "step": 67785 }, { "epoch": 0.17, "learning_rate": 8.319920479138357e-05, "loss": 1.0266, "step": 67790 }, { "epoch": 0.17, "learning_rate": 8.319794654990186e-05, "loss": 1.0239, "step": 67795 }, { "epoch": 0.17, "learning_rate": 8.319668830842016e-05, "loss": 1.0254, "step": 67800 }, { "epoch": 0.17, "learning_rate": 8.319543006693845e-05, "loss": 1.0245, "step": 67805 }, { "epoch": 0.17, "learning_rate": 8.319417182545675e-05, "loss": 1.027, "step": 67810 }, { "epoch": 0.17, "learning_rate": 8.319291358397504e-05, "loss": 1.0249, "step": 67815 }, { "epoch": 0.17, "learning_rate": 8.319165534249334e-05, "loss": 1.0261, "step": 67820 }, { "epoch": 0.17, "learning_rate": 8.319039710101163e-05, "loss": 1.0248, "step": 67825 }, { "epoch": 0.17, "learning_rate": 8.318913885952992e-05, "loss": 1.026, "step": 67830 }, { "epoch": 0.17, "learning_rate": 8.318788061804822e-05, "loss": 1.0276, "step": 67835 }, { "epoch": 0.17, "learning_rate": 8.318662237656652e-05, "loss": 1.0268, "step": 67840 }, { "epoch": 0.17, "learning_rate": 8.318536413508481e-05, "loss": 1.0257, "step": 67845 }, { "epoch": 0.17, "learning_rate": 8.31841058936031e-05, "loss": 1.024, "step": 67850 }, { "epoch": 0.17, "learning_rate": 8.31828476521214e-05, "loss": 1.0261, "step": 67855 }, { "epoch": 0.17, "learning_rate": 8.31815894106397e-05, "loss": 1.0234, "step": 67860 }, { "epoch": 0.17, "learning_rate": 8.318033116915799e-05, "loss": 1.0245, "step": 67865 }, { "epoch": 0.17, "learning_rate": 8.317907292767628e-05, "loss": 1.0254, "step": 67870 }, { "epoch": 0.17, "learning_rate": 8.317781468619458e-05, "loss": 1.0262, "step": 67875 }, { "epoch": 0.17, "learning_rate": 8.317655644471288e-05, "loss": 1.0261, "step": 67880 }, { "epoch": 0.17, "learning_rate": 8.317529820323117e-05, "loss": 1.027, "step": 67885 }, { "epoch": 0.17, "learning_rate": 8.317403996174946e-05, "loss": 1.0286, "step": 67890 }, { "epoch": 0.17, "learning_rate": 8.317278172026775e-05, "loss": 1.0254, "step": 67895 }, { "epoch": 0.17, "learning_rate": 8.317152347878606e-05, "loss": 1.0238, "step": 67900 }, { "epoch": 0.17, "learning_rate": 8.317026523730435e-05, "loss": 1.0268, "step": 67905 }, { "epoch": 0.17, "learning_rate": 8.316900699582264e-05, "loss": 1.0253, "step": 67910 }, { "epoch": 0.17, "learning_rate": 8.316774875434093e-05, "loss": 1.0265, "step": 67915 }, { "epoch": 0.17, "learning_rate": 8.316649051285924e-05, "loss": 1.0269, "step": 67920 }, { "epoch": 0.17, "learning_rate": 8.316523227137753e-05, "loss": 1.0261, "step": 67925 }, { "epoch": 0.17, "learning_rate": 8.316397402989582e-05, "loss": 1.0257, "step": 67930 }, { "epoch": 0.17, "learning_rate": 8.316271578841411e-05, "loss": 1.0253, "step": 67935 }, { "epoch": 0.17, "learning_rate": 8.316145754693242e-05, "loss": 1.0248, "step": 67940 }, { "epoch": 0.17, "learning_rate": 8.316019930545071e-05, "loss": 1.0249, "step": 67945 }, { "epoch": 0.17, "learning_rate": 8.3158941063969e-05, "loss": 1.0235, "step": 67950 }, { "epoch": 0.17, "learning_rate": 8.315768282248729e-05, "loss": 1.0257, "step": 67955 }, { "epoch": 0.17, "learning_rate": 8.315642458100558e-05, "loss": 1.0253, "step": 67960 }, { "epoch": 0.17, "learning_rate": 8.315516633952389e-05, "loss": 1.0251, "step": 67965 }, { "epoch": 0.17, "learning_rate": 8.315390809804218e-05, "loss": 1.0256, "step": 67970 }, { "epoch": 0.17, "learning_rate": 8.315264985656047e-05, "loss": 1.0251, "step": 67975 }, { "epoch": 0.17, "learning_rate": 8.315139161507876e-05, "loss": 1.0499, "step": 67980 }, { "epoch": 0.17, "learning_rate": 8.315013337359707e-05, "loss": 1.0257, "step": 67985 }, { "epoch": 0.17, "learning_rate": 8.314887513211536e-05, "loss": 1.024, "step": 67990 }, { "epoch": 0.17, "learning_rate": 8.314761689063365e-05, "loss": 1.0431, "step": 67995 }, { "epoch": 0.17, "learning_rate": 8.314635864915194e-05, "loss": 1.025, "step": 68000 }, { "epoch": 0.17, "learning_rate": 8.314510040767023e-05, "loss": 1.0262, "step": 68005 }, { "epoch": 0.17, "learning_rate": 8.314384216618854e-05, "loss": 1.0249, "step": 68010 }, { "epoch": 0.17, "learning_rate": 8.314258392470683e-05, "loss": 1.0261, "step": 68015 }, { "epoch": 0.17, "learning_rate": 8.314132568322512e-05, "loss": 1.025, "step": 68020 }, { "epoch": 0.17, "learning_rate": 8.314006744174341e-05, "loss": 1.0251, "step": 68025 }, { "epoch": 0.17, "learning_rate": 8.313880920026172e-05, "loss": 1.0264, "step": 68030 }, { "epoch": 0.17, "learning_rate": 8.313755095878001e-05, "loss": 1.0247, "step": 68035 }, { "epoch": 0.17, "learning_rate": 8.31362927172983e-05, "loss": 1.0258, "step": 68040 }, { "epoch": 0.17, "learning_rate": 8.313503447581659e-05, "loss": 1.025, "step": 68045 }, { "epoch": 0.17, "learning_rate": 8.31337762343349e-05, "loss": 1.0251, "step": 68050 }, { "epoch": 0.17, "learning_rate": 8.313251799285319e-05, "loss": 1.0668, "step": 68055 }, { "epoch": 0.17, "learning_rate": 8.313125975137148e-05, "loss": 1.0276, "step": 68060 }, { "epoch": 0.17, "learning_rate": 8.313000150988977e-05, "loss": 1.0251, "step": 68065 }, { "epoch": 0.17, "learning_rate": 8.312874326840806e-05, "loss": 1.0282, "step": 68070 }, { "epoch": 0.17, "learning_rate": 8.312748502692638e-05, "loss": 1.0242, "step": 68075 }, { "epoch": 0.17, "learning_rate": 8.312622678544468e-05, "loss": 1.0275, "step": 68080 }, { "epoch": 0.17, "learning_rate": 8.312496854396297e-05, "loss": 1.0265, "step": 68085 }, { "epoch": 0.17, "learning_rate": 8.312371030248126e-05, "loss": 1.0254, "step": 68090 }, { "epoch": 0.17, "learning_rate": 8.312245206099955e-05, "loss": 1.0282, "step": 68095 }, { "epoch": 0.17, "learning_rate": 8.312119381951785e-05, "loss": 1.0293, "step": 68100 }, { "epoch": 0.17, "learning_rate": 8.311993557803615e-05, "loss": 1.0258, "step": 68105 }, { "epoch": 0.17, "learning_rate": 8.311867733655444e-05, "loss": 1.0248, "step": 68110 }, { "epoch": 0.17, "learning_rate": 8.311741909507273e-05, "loss": 1.0422, "step": 68115 }, { "epoch": 0.17, "learning_rate": 8.311616085359103e-05, "loss": 1.0265, "step": 68120 }, { "epoch": 0.17, "learning_rate": 8.311490261210933e-05, "loss": 1.0245, "step": 68125 }, { "epoch": 0.17, "learning_rate": 8.311364437062762e-05, "loss": 1.025, "step": 68130 }, { "epoch": 0.17, "learning_rate": 8.311238612914591e-05, "loss": 1.0272, "step": 68135 }, { "epoch": 0.17, "learning_rate": 8.311112788766421e-05, "loss": 1.0264, "step": 68140 }, { "epoch": 0.17, "learning_rate": 8.31098696461825e-05, "loss": 1.0441, "step": 68145 }, { "epoch": 0.17, "learning_rate": 8.31086114047008e-05, "loss": 1.0244, "step": 68150 }, { "epoch": 0.17, "learning_rate": 8.310735316321909e-05, "loss": 1.0259, "step": 68155 }, { "epoch": 0.17, "learning_rate": 8.310609492173738e-05, "loss": 1.0263, "step": 68160 }, { "epoch": 0.17, "learning_rate": 8.310483668025569e-05, "loss": 1.0246, "step": 68165 }, { "epoch": 0.17, "learning_rate": 8.310357843877398e-05, "loss": 1.0223, "step": 68170 }, { "epoch": 0.17, "learning_rate": 8.310232019729227e-05, "loss": 1.0253, "step": 68175 }, { "epoch": 0.17, "learning_rate": 8.310106195581056e-05, "loss": 1.0277, "step": 68180 }, { "epoch": 0.17, "learning_rate": 8.309980371432887e-05, "loss": 1.0261, "step": 68185 }, { "epoch": 0.17, "learning_rate": 8.309854547284716e-05, "loss": 1.0273, "step": 68190 }, { "epoch": 0.17, "learning_rate": 8.309728723136545e-05, "loss": 1.0498, "step": 68195 }, { "epoch": 0.17, "learning_rate": 8.309602898988374e-05, "loss": 1.0253, "step": 68200 }, { "epoch": 0.17, "learning_rate": 8.309477074840205e-05, "loss": 1.0248, "step": 68205 }, { "epoch": 0.17, "learning_rate": 8.309351250692034e-05, "loss": 1.0248, "step": 68210 }, { "epoch": 0.17, "learning_rate": 8.309225426543863e-05, "loss": 1.0265, "step": 68215 }, { "epoch": 0.17, "learning_rate": 8.309099602395692e-05, "loss": 1.0276, "step": 68220 }, { "epoch": 0.17, "learning_rate": 8.308973778247521e-05, "loss": 1.0253, "step": 68225 }, { "epoch": 0.17, "learning_rate": 8.308847954099352e-05, "loss": 1.0274, "step": 68230 }, { "epoch": 0.17, "learning_rate": 8.308722129951181e-05, "loss": 1.0249, "step": 68235 }, { "epoch": 0.17, "learning_rate": 8.30859630580301e-05, "loss": 1.0253, "step": 68240 }, { "epoch": 0.17, "learning_rate": 8.308470481654839e-05, "loss": 1.0285, "step": 68245 }, { "epoch": 0.17, "learning_rate": 8.30834465750667e-05, "loss": 1.0257, "step": 68250 }, { "epoch": 0.17, "learning_rate": 8.308218833358499e-05, "loss": 1.0254, "step": 68255 }, { "epoch": 0.17, "learning_rate": 8.308093009210328e-05, "loss": 1.0264, "step": 68260 }, { "epoch": 0.17, "learning_rate": 8.307967185062157e-05, "loss": 1.0253, "step": 68265 }, { "epoch": 0.17, "learning_rate": 8.307841360913988e-05, "loss": 1.0266, "step": 68270 }, { "epoch": 0.17, "learning_rate": 8.307715536765817e-05, "loss": 1.024, "step": 68275 }, { "epoch": 0.17, "learning_rate": 8.307589712617646e-05, "loss": 1.0272, "step": 68280 }, { "epoch": 0.17, "learning_rate": 8.307463888469475e-05, "loss": 1.0273, "step": 68285 }, { "epoch": 0.17, "learning_rate": 8.307338064321304e-05, "loss": 1.0281, "step": 68290 }, { "epoch": 0.17, "learning_rate": 8.307212240173135e-05, "loss": 1.0253, "step": 68295 }, { "epoch": 0.17, "learning_rate": 8.307086416024964e-05, "loss": 1.0263, "step": 68300 }, { "epoch": 0.17, "learning_rate": 8.306960591876793e-05, "loss": 1.0253, "step": 68305 }, { "epoch": 0.17, "learning_rate": 8.306834767728622e-05, "loss": 1.0232, "step": 68310 }, { "epoch": 0.17, "learning_rate": 8.306708943580453e-05, "loss": 1.0493, "step": 68315 }, { "epoch": 0.17, "learning_rate": 8.306583119432282e-05, "loss": 1.0238, "step": 68320 }, { "epoch": 0.17, "learning_rate": 8.306457295284111e-05, "loss": 1.0272, "step": 68325 }, { "epoch": 0.17, "learning_rate": 8.30633147113594e-05, "loss": 1.0249, "step": 68330 }, { "epoch": 0.17, "learning_rate": 8.306205646987771e-05, "loss": 1.0263, "step": 68335 }, { "epoch": 0.17, "learning_rate": 8.3060798228396e-05, "loss": 1.0256, "step": 68340 }, { "epoch": 0.17, "learning_rate": 8.305953998691429e-05, "loss": 1.0249, "step": 68345 }, { "epoch": 0.17, "learning_rate": 8.305828174543258e-05, "loss": 1.0249, "step": 68350 }, { "epoch": 0.17, "learning_rate": 8.305702350395088e-05, "loss": 1.0451, "step": 68355 }, { "epoch": 0.17, "learning_rate": 8.305576526246918e-05, "loss": 1.0247, "step": 68360 }, { "epoch": 0.17, "learning_rate": 8.305450702098747e-05, "loss": 1.0257, "step": 68365 }, { "epoch": 0.17, "learning_rate": 8.305324877950576e-05, "loss": 1.0253, "step": 68370 }, { "epoch": 0.17, "learning_rate": 8.305199053802406e-05, "loss": 1.0255, "step": 68375 }, { "epoch": 0.17, "learning_rate": 8.305073229654236e-05, "loss": 1.0274, "step": 68380 }, { "epoch": 0.17, "learning_rate": 8.304947405506065e-05, "loss": 1.0266, "step": 68385 }, { "epoch": 0.17, "learning_rate": 8.304821581357894e-05, "loss": 1.0264, "step": 68390 }, { "epoch": 0.17, "learning_rate": 8.304695757209724e-05, "loss": 1.0261, "step": 68395 }, { "epoch": 0.17, "learning_rate": 8.304569933061554e-05, "loss": 1.0249, "step": 68400 }, { "epoch": 0.17, "learning_rate": 8.304444108913383e-05, "loss": 1.0267, "step": 68405 }, { "epoch": 0.17, "learning_rate": 8.304318284765212e-05, "loss": 1.0239, "step": 68410 }, { "epoch": 0.17, "learning_rate": 8.304192460617042e-05, "loss": 1.0274, "step": 68415 }, { "epoch": 0.17, "learning_rate": 8.304066636468871e-05, "loss": 1.025, "step": 68420 }, { "epoch": 0.17, "learning_rate": 8.303940812320701e-05, "loss": 1.0271, "step": 68425 }, { "epoch": 0.17, "learning_rate": 8.30381498817253e-05, "loss": 1.0262, "step": 68430 }, { "epoch": 0.17, "learning_rate": 8.30368916402436e-05, "loss": 1.0261, "step": 68435 }, { "epoch": 0.17, "learning_rate": 8.303563339876189e-05, "loss": 1.0244, "step": 68440 }, { "epoch": 0.17, "learning_rate": 8.303437515728019e-05, "loss": 1.0258, "step": 68445 }, { "epoch": 0.17, "learning_rate": 8.303311691579848e-05, "loss": 1.026, "step": 68450 }, { "epoch": 0.17, "learning_rate": 8.303185867431678e-05, "loss": 1.0255, "step": 68455 }, { "epoch": 0.17, "learning_rate": 8.303060043283507e-05, "loss": 1.026, "step": 68460 }, { "epoch": 0.17, "learning_rate": 8.302934219135337e-05, "loss": 1.0249, "step": 68465 }, { "epoch": 0.17, "learning_rate": 8.302808394987166e-05, "loss": 1.0211, "step": 68470 }, { "epoch": 0.17, "learning_rate": 8.302682570838996e-05, "loss": 1.0267, "step": 68475 }, { "epoch": 0.17, "learning_rate": 8.302556746690825e-05, "loss": 1.0244, "step": 68480 }, { "epoch": 0.17, "learning_rate": 8.302430922542654e-05, "loss": 1.0262, "step": 68485 }, { "epoch": 0.17, "learning_rate": 8.302305098394484e-05, "loss": 1.0239, "step": 68490 }, { "epoch": 0.17, "learning_rate": 8.302179274246314e-05, "loss": 1.0253, "step": 68495 }, { "epoch": 0.17, "learning_rate": 8.302053450098143e-05, "loss": 1.0264, "step": 68500 }, { "epoch": 0.17, "learning_rate": 8.301927625949972e-05, "loss": 1.0245, "step": 68505 }, { "epoch": 0.17, "learning_rate": 8.301801801801802e-05, "loss": 1.0237, "step": 68510 }, { "epoch": 0.17, "learning_rate": 8.301675977653632e-05, "loss": 1.0245, "step": 68515 }, { "epoch": 0.17, "learning_rate": 8.30155015350546e-05, "loss": 1.0244, "step": 68520 }, { "epoch": 0.17, "learning_rate": 8.30142432935729e-05, "loss": 1.0247, "step": 68525 }, { "epoch": 0.17, "learning_rate": 8.30129850520912e-05, "loss": 1.0258, "step": 68530 }, { "epoch": 0.17, "learning_rate": 8.30117268106095e-05, "loss": 1.0266, "step": 68535 }, { "epoch": 0.17, "learning_rate": 8.301046856912779e-05, "loss": 1.0248, "step": 68540 }, { "epoch": 0.17, "learning_rate": 8.300921032764608e-05, "loss": 1.0254, "step": 68545 }, { "epoch": 0.17, "learning_rate": 8.300795208616437e-05, "loss": 1.0284, "step": 68550 }, { "epoch": 0.17, "learning_rate": 8.300669384468267e-05, "loss": 1.0246, "step": 68555 }, { "epoch": 0.17, "learning_rate": 8.300543560320097e-05, "loss": 1.0277, "step": 68560 }, { "epoch": 0.17, "learning_rate": 8.300417736171926e-05, "loss": 1.0253, "step": 68565 }, { "epoch": 0.17, "learning_rate": 8.300291912023755e-05, "loss": 1.0278, "step": 68570 }, { "epoch": 0.17, "learning_rate": 8.300166087875585e-05, "loss": 1.0242, "step": 68575 }, { "epoch": 0.17, "learning_rate": 8.300040263727416e-05, "loss": 1.027, "step": 68580 }, { "epoch": 0.17, "learning_rate": 8.299914439579245e-05, "loss": 1.027, "step": 68585 }, { "epoch": 0.17, "learning_rate": 8.299788615431074e-05, "loss": 1.0245, "step": 68590 }, { "epoch": 0.17, "learning_rate": 8.299662791282903e-05, "loss": 1.0245, "step": 68595 }, { "epoch": 0.17, "learning_rate": 8.299536967134734e-05, "loss": 1.0248, "step": 68600 }, { "epoch": 0.17, "learning_rate": 8.299411142986563e-05, "loss": 1.025, "step": 68605 }, { "epoch": 0.17, "learning_rate": 8.299285318838392e-05, "loss": 1.0249, "step": 68610 }, { "epoch": 0.17, "learning_rate": 8.299159494690221e-05, "loss": 1.0269, "step": 68615 }, { "epoch": 0.17, "learning_rate": 8.29903367054205e-05, "loss": 1.0255, "step": 68620 }, { "epoch": 0.17, "learning_rate": 8.298907846393881e-05, "loss": 1.0253, "step": 68625 }, { "epoch": 0.17, "learning_rate": 8.29878202224571e-05, "loss": 1.0253, "step": 68630 }, { "epoch": 0.17, "learning_rate": 8.29865619809754e-05, "loss": 1.0242, "step": 68635 }, { "epoch": 0.17, "learning_rate": 8.298530373949369e-05, "loss": 1.0255, "step": 68640 }, { "epoch": 0.17, "learning_rate": 8.298404549801199e-05, "loss": 1.0255, "step": 68645 }, { "epoch": 0.17, "learning_rate": 8.298278725653028e-05, "loss": 1.0251, "step": 68650 }, { "epoch": 0.17, "learning_rate": 8.298152901504857e-05, "loss": 1.0249, "step": 68655 }, { "epoch": 0.17, "learning_rate": 8.298027077356687e-05, "loss": 1.0262, "step": 68660 }, { "epoch": 0.17, "learning_rate": 8.297901253208517e-05, "loss": 1.0246, "step": 68665 }, { "epoch": 0.17, "learning_rate": 8.297775429060346e-05, "loss": 1.0266, "step": 68670 }, { "epoch": 0.17, "learning_rate": 8.297649604912175e-05, "loss": 1.0245, "step": 68675 }, { "epoch": 0.17, "learning_rate": 8.297523780764005e-05, "loss": 1.0256, "step": 68680 }, { "epoch": 0.17, "learning_rate": 8.297397956615834e-05, "loss": 1.0258, "step": 68685 }, { "epoch": 0.17, "learning_rate": 8.297272132467664e-05, "loss": 1.0243, "step": 68690 }, { "epoch": 0.17, "learning_rate": 8.297146308319493e-05, "loss": 1.0502, "step": 68695 }, { "epoch": 0.17, "learning_rate": 8.297020484171323e-05, "loss": 1.0236, "step": 68700 }, { "epoch": 0.17, "learning_rate": 8.296894660023152e-05, "loss": 1.0229, "step": 68705 }, { "epoch": 0.17, "learning_rate": 8.296768835874982e-05, "loss": 1.0255, "step": 68710 }, { "epoch": 0.17, "learning_rate": 8.296643011726811e-05, "loss": 1.0248, "step": 68715 }, { "epoch": 0.17, "learning_rate": 8.29651718757864e-05, "loss": 1.0269, "step": 68720 }, { "epoch": 0.17, "learning_rate": 8.29639136343047e-05, "loss": 1.026, "step": 68725 }, { "epoch": 0.17, "learning_rate": 8.2962655392823e-05, "loss": 1.0266, "step": 68730 }, { "epoch": 0.17, "learning_rate": 8.29613971513413e-05, "loss": 1.0269, "step": 68735 }, { "epoch": 0.17, "learning_rate": 8.296013890985959e-05, "loss": 1.0277, "step": 68740 }, { "epoch": 0.17, "learning_rate": 8.295888066837788e-05, "loss": 1.0236, "step": 68745 }, { "epoch": 0.17, "learning_rate": 8.295762242689617e-05, "loss": 1.0262, "step": 68750 }, { "epoch": 0.17, "learning_rate": 8.295636418541447e-05, "loss": 1.0255, "step": 68755 }, { "epoch": 0.17, "learning_rate": 8.295510594393277e-05, "loss": 1.0263, "step": 68760 }, { "epoch": 0.17, "learning_rate": 8.295384770245106e-05, "loss": 1.0257, "step": 68765 }, { "epoch": 0.17, "learning_rate": 8.295258946096935e-05, "loss": 1.025, "step": 68770 }, { "epoch": 0.17, "learning_rate": 8.295133121948765e-05, "loss": 1.0265, "step": 68775 }, { "epoch": 0.17, "learning_rate": 8.295007297800595e-05, "loss": 1.0517, "step": 68780 }, { "epoch": 0.17, "learning_rate": 8.294881473652424e-05, "loss": 1.0252, "step": 68785 }, { "epoch": 0.17, "learning_rate": 8.294755649504253e-05, "loss": 1.0263, "step": 68790 }, { "epoch": 0.17, "learning_rate": 8.294629825356083e-05, "loss": 1.0255, "step": 68795 }, { "epoch": 0.17, "learning_rate": 8.294504001207913e-05, "loss": 1.0269, "step": 68800 }, { "epoch": 0.17, "learning_rate": 8.294378177059742e-05, "loss": 1.0263, "step": 68805 }, { "epoch": 0.17, "learning_rate": 8.294252352911571e-05, "loss": 1.0243, "step": 68810 }, { "epoch": 0.17, "learning_rate": 8.2941265287634e-05, "loss": 1.0251, "step": 68815 }, { "epoch": 0.17, "learning_rate": 8.29400070461523e-05, "loss": 1.0268, "step": 68820 }, { "epoch": 0.17, "learning_rate": 8.29387488046706e-05, "loss": 1.0262, "step": 68825 }, { "epoch": 0.17, "learning_rate": 8.293749056318889e-05, "loss": 1.0242, "step": 68830 }, { "epoch": 0.17, "learning_rate": 8.293623232170718e-05, "loss": 1.0237, "step": 68835 }, { "epoch": 0.17, "learning_rate": 8.293497408022549e-05, "loss": 1.0261, "step": 68840 }, { "epoch": 0.17, "learning_rate": 8.293371583874378e-05, "loss": 1.026, "step": 68845 }, { "epoch": 0.17, "learning_rate": 8.293245759726207e-05, "loss": 1.0262, "step": 68850 }, { "epoch": 0.17, "learning_rate": 8.293119935578036e-05, "loss": 1.0257, "step": 68855 }, { "epoch": 0.17, "learning_rate": 8.292994111429867e-05, "loss": 1.0245, "step": 68860 }, { "epoch": 0.17, "learning_rate": 8.292868287281696e-05, "loss": 1.0269, "step": 68865 }, { "epoch": 0.17, "learning_rate": 8.292742463133525e-05, "loss": 1.0249, "step": 68870 }, { "epoch": 0.17, "learning_rate": 8.292616638985354e-05, "loss": 1.0268, "step": 68875 }, { "epoch": 0.17, "learning_rate": 8.292490814837183e-05, "loss": 1.0251, "step": 68880 }, { "epoch": 0.17, "learning_rate": 8.292364990689014e-05, "loss": 1.0254, "step": 68885 }, { "epoch": 0.17, "learning_rate": 8.292239166540843e-05, "loss": 1.0252, "step": 68890 }, { "epoch": 0.17, "learning_rate": 8.292113342392672e-05, "loss": 1.0504, "step": 68895 }, { "epoch": 0.17, "learning_rate": 8.291987518244501e-05, "loss": 1.0268, "step": 68900 }, { "epoch": 0.17, "learning_rate": 8.291861694096332e-05, "loss": 1.0273, "step": 68905 }, { "epoch": 0.17, "learning_rate": 8.291735869948161e-05, "loss": 1.0253, "step": 68910 }, { "epoch": 0.17, "learning_rate": 8.29161004579999e-05, "loss": 1.0245, "step": 68915 }, { "epoch": 0.17, "learning_rate": 8.291484221651819e-05, "loss": 1.0259, "step": 68920 }, { "epoch": 0.17, "learning_rate": 8.29135839750365e-05, "loss": 1.026, "step": 68925 }, { "epoch": 0.17, "learning_rate": 8.291232573355479e-05, "loss": 1.0256, "step": 68930 }, { "epoch": 0.17, "learning_rate": 8.291106749207308e-05, "loss": 1.0251, "step": 68935 }, { "epoch": 0.17, "learning_rate": 8.290980925059137e-05, "loss": 1.0272, "step": 68940 }, { "epoch": 0.17, "learning_rate": 8.290855100910966e-05, "loss": 1.0261, "step": 68945 }, { "epoch": 0.17, "learning_rate": 8.290729276762797e-05, "loss": 1.0267, "step": 68950 }, { "epoch": 0.17, "learning_rate": 8.290603452614626e-05, "loss": 1.0265, "step": 68955 }, { "epoch": 0.17, "learning_rate": 8.290477628466455e-05, "loss": 1.0474, "step": 68960 }, { "epoch": 0.17, "learning_rate": 8.290351804318284e-05, "loss": 1.0243, "step": 68965 }, { "epoch": 0.17, "learning_rate": 8.290225980170115e-05, "loss": 1.0222, "step": 68970 }, { "epoch": 0.17, "learning_rate": 8.290100156021944e-05, "loss": 1.0258, "step": 68975 }, { "epoch": 0.17, "learning_rate": 8.289974331873773e-05, "loss": 1.0246, "step": 68980 }, { "epoch": 0.17, "learning_rate": 8.289848507725602e-05, "loss": 1.0257, "step": 68985 }, { "epoch": 0.17, "learning_rate": 8.289722683577433e-05, "loss": 1.0244, "step": 68990 }, { "epoch": 0.17, "learning_rate": 8.289596859429262e-05, "loss": 1.0264, "step": 68995 }, { "epoch": 0.17, "learning_rate": 8.289471035281091e-05, "loss": 1.0248, "step": 69000 }, { "epoch": 0.17, "learning_rate": 8.28934521113292e-05, "loss": 1.0241, "step": 69005 }, { "epoch": 0.17, "learning_rate": 8.28921938698475e-05, "loss": 1.0262, "step": 69010 }, { "epoch": 0.17, "learning_rate": 8.28909356283658e-05, "loss": 1.0279, "step": 69015 }, { "epoch": 0.17, "learning_rate": 8.288967738688409e-05, "loss": 1.0234, "step": 69020 }, { "epoch": 0.17, "learning_rate": 8.288841914540238e-05, "loss": 1.025, "step": 69025 }, { "epoch": 0.17, "learning_rate": 8.288716090392067e-05, "loss": 1.0266, "step": 69030 }, { "epoch": 0.17, "learning_rate": 8.288590266243898e-05, "loss": 1.0253, "step": 69035 }, { "epoch": 0.17, "learning_rate": 8.288464442095727e-05, "loss": 1.0259, "step": 69040 }, { "epoch": 0.17, "learning_rate": 8.288338617947556e-05, "loss": 1.0448, "step": 69045 }, { "epoch": 0.17, "learning_rate": 8.288212793799385e-05, "loss": 1.0241, "step": 69050 }, { "epoch": 0.17, "learning_rate": 8.288086969651216e-05, "loss": 1.0272, "step": 69055 }, { "epoch": 0.17, "learning_rate": 8.287961145503045e-05, "loss": 1.0257, "step": 69060 }, { "epoch": 0.17, "learning_rate": 8.287835321354874e-05, "loss": 1.0254, "step": 69065 }, { "epoch": 0.17, "learning_rate": 8.287709497206703e-05, "loss": 1.025, "step": 69070 }, { "epoch": 0.17, "learning_rate": 8.287583673058534e-05, "loss": 1.0257, "step": 69075 }, { "epoch": 0.17, "learning_rate": 8.287457848910363e-05, "loss": 1.0249, "step": 69080 }, { "epoch": 0.17, "learning_rate": 8.287332024762194e-05, "loss": 1.0261, "step": 69085 }, { "epoch": 0.17, "learning_rate": 8.287206200614023e-05, "loss": 1.0276, "step": 69090 }, { "epoch": 0.17, "learning_rate": 8.287080376465852e-05, "loss": 1.0283, "step": 69095 }, { "epoch": 0.17, "learning_rate": 8.286954552317681e-05, "loss": 1.0272, "step": 69100 }, { "epoch": 0.17, "learning_rate": 8.286828728169512e-05, "loss": 1.0274, "step": 69105 }, { "epoch": 0.17, "learning_rate": 8.286702904021341e-05, "loss": 1.027, "step": 69110 }, { "epoch": 0.17, "learning_rate": 8.28657707987317e-05, "loss": 1.0264, "step": 69115 }, { "epoch": 0.17, "learning_rate": 8.286451255724999e-05, "loss": 1.0263, "step": 69120 }, { "epoch": 0.17, "learning_rate": 8.28632543157683e-05, "loss": 1.0268, "step": 69125 }, { "epoch": 0.17, "learning_rate": 8.286199607428659e-05, "loss": 1.0241, "step": 69130 }, { "epoch": 0.17, "learning_rate": 8.286073783280488e-05, "loss": 1.0257, "step": 69135 }, { "epoch": 0.17, "learning_rate": 8.285947959132317e-05, "loss": 1.0256, "step": 69140 }, { "epoch": 0.17, "learning_rate": 8.285822134984146e-05, "loss": 1.0243, "step": 69145 }, { "epoch": 0.17, "learning_rate": 8.285696310835977e-05, "loss": 1.0262, "step": 69150 }, { "epoch": 0.17, "learning_rate": 8.285570486687806e-05, "loss": 1.0445, "step": 69155 }, { "epoch": 0.17, "learning_rate": 8.285444662539635e-05, "loss": 1.0263, "step": 69160 }, { "epoch": 0.17, "learning_rate": 8.285318838391464e-05, "loss": 1.0247, "step": 69165 }, { "epoch": 0.17, "learning_rate": 8.285193014243295e-05, "loss": 1.0253, "step": 69170 }, { "epoch": 0.17, "learning_rate": 8.285067190095124e-05, "loss": 1.0264, "step": 69175 }, { "epoch": 0.17, "learning_rate": 8.284941365946953e-05, "loss": 1.0245, "step": 69180 }, { "epoch": 0.17, "learning_rate": 8.284815541798782e-05, "loss": 1.0246, "step": 69185 }, { "epoch": 0.17, "learning_rate": 8.284689717650613e-05, "loss": 1.0274, "step": 69190 }, { "epoch": 0.17, "learning_rate": 8.284563893502442e-05, "loss": 1.0271, "step": 69195 }, { "epoch": 0.17, "learning_rate": 8.284438069354271e-05, "loss": 1.0267, "step": 69200 }, { "epoch": 0.17, "learning_rate": 8.2843122452061e-05, "loss": 1.0247, "step": 69205 }, { "epoch": 0.17, "learning_rate": 8.28418642105793e-05, "loss": 1.025, "step": 69210 }, { "epoch": 0.17, "learning_rate": 8.28406059690976e-05, "loss": 1.0271, "step": 69215 }, { "epoch": 0.17, "learning_rate": 8.283934772761589e-05, "loss": 1.0275, "step": 69220 }, { "epoch": 0.17, "learning_rate": 8.283808948613418e-05, "loss": 1.026, "step": 69225 }, { "epoch": 0.17, "learning_rate": 8.283683124465247e-05, "loss": 1.0263, "step": 69230 }, { "epoch": 0.17, "learning_rate": 8.283557300317078e-05, "loss": 1.0506, "step": 69235 }, { "epoch": 0.17, "learning_rate": 8.283431476168907e-05, "loss": 1.028, "step": 69240 }, { "epoch": 0.17, "learning_rate": 8.283305652020736e-05, "loss": 1.0244, "step": 69245 }, { "epoch": 0.17, "learning_rate": 8.283179827872565e-05, "loss": 1.0263, "step": 69250 }, { "epoch": 0.17, "learning_rate": 8.283054003724396e-05, "loss": 1.0232, "step": 69255 }, { "epoch": 0.17, "learning_rate": 8.282928179576225e-05, "loss": 1.0259, "step": 69260 }, { "epoch": 0.17, "learning_rate": 8.282802355428054e-05, "loss": 1.0497, "step": 69265 }, { "epoch": 0.17, "learning_rate": 8.282676531279883e-05, "loss": 1.0256, "step": 69270 }, { "epoch": 0.17, "learning_rate": 8.282550707131713e-05, "loss": 1.0261, "step": 69275 }, { "epoch": 0.17, "learning_rate": 8.282424882983543e-05, "loss": 1.0257, "step": 69280 }, { "epoch": 0.17, "learning_rate": 8.282299058835372e-05, "loss": 1.0263, "step": 69285 }, { "epoch": 0.17, "learning_rate": 8.282173234687201e-05, "loss": 1.0252, "step": 69290 }, { "epoch": 0.17, "learning_rate": 8.28204741053903e-05, "loss": 1.0256, "step": 69295 }, { "epoch": 0.17, "learning_rate": 8.281921586390861e-05, "loss": 1.0462, "step": 69300 }, { "epoch": 0.17, "learning_rate": 8.28179576224269e-05, "loss": 1.0267, "step": 69305 }, { "epoch": 0.17, "learning_rate": 8.28166993809452e-05, "loss": 1.0262, "step": 69310 }, { "epoch": 0.17, "learning_rate": 8.281544113946349e-05, "loss": 1.0273, "step": 69315 }, { "epoch": 0.17, "learning_rate": 8.281418289798179e-05, "loss": 1.0271, "step": 69320 }, { "epoch": 0.17, "learning_rate": 8.281292465650008e-05, "loss": 1.0258, "step": 69325 }, { "epoch": 0.17, "learning_rate": 8.281166641501837e-05, "loss": 1.026, "step": 69330 }, { "epoch": 0.17, "learning_rate": 8.281040817353667e-05, "loss": 1.0255, "step": 69335 }, { "epoch": 0.17, "learning_rate": 8.280914993205496e-05, "loss": 1.0282, "step": 69340 }, { "epoch": 0.17, "learning_rate": 8.280789169057326e-05, "loss": 1.0268, "step": 69345 }, { "epoch": 0.17, "learning_rate": 8.280663344909155e-05, "loss": 1.0231, "step": 69350 }, { "epoch": 0.17, "learning_rate": 8.280537520760985e-05, "loss": 1.0255, "step": 69355 }, { "epoch": 0.17, "learning_rate": 8.280411696612814e-05, "loss": 1.0249, "step": 69360 }, { "epoch": 0.17, "learning_rate": 8.280285872464644e-05, "loss": 1.0244, "step": 69365 }, { "epoch": 0.17, "learning_rate": 8.280160048316473e-05, "loss": 1.0249, "step": 69370 }, { "epoch": 0.17, "learning_rate": 8.280034224168303e-05, "loss": 1.0249, "step": 69375 }, { "epoch": 0.17, "learning_rate": 8.279908400020132e-05, "loss": 1.0233, "step": 69380 }, { "epoch": 0.17, "learning_rate": 8.279782575871962e-05, "loss": 1.0278, "step": 69385 }, { "epoch": 0.17, "learning_rate": 8.279656751723791e-05, "loss": 1.0255, "step": 69390 }, { "epoch": 0.17, "learning_rate": 8.27953092757562e-05, "loss": 1.0249, "step": 69395 }, { "epoch": 0.17, "learning_rate": 8.27940510342745e-05, "loss": 1.0263, "step": 69400 }, { "epoch": 0.17, "learning_rate": 8.279279279279279e-05, "loss": 1.0262, "step": 69405 }, { "epoch": 0.17, "learning_rate": 8.27915345513111e-05, "loss": 1.0275, "step": 69410 }, { "epoch": 0.17, "learning_rate": 8.279027630982939e-05, "loss": 1.0264, "step": 69415 }, { "epoch": 0.17, "learning_rate": 8.278901806834768e-05, "loss": 1.0257, "step": 69420 }, { "epoch": 0.17, "learning_rate": 8.278775982686597e-05, "loss": 1.0276, "step": 69425 }, { "epoch": 0.17, "learning_rate": 8.278650158538427e-05, "loss": 1.0249, "step": 69430 }, { "epoch": 0.17, "learning_rate": 8.278524334390257e-05, "loss": 1.0242, "step": 69435 }, { "epoch": 0.17, "learning_rate": 8.278398510242086e-05, "loss": 1.0256, "step": 69440 }, { "epoch": 0.17, "learning_rate": 8.278272686093915e-05, "loss": 1.0246, "step": 69445 }, { "epoch": 0.17, "learning_rate": 8.278146861945745e-05, "loss": 1.0273, "step": 69450 }, { "epoch": 0.17, "learning_rate": 8.278021037797575e-05, "loss": 1.0251, "step": 69455 }, { "epoch": 0.17, "learning_rate": 8.277895213649404e-05, "loss": 1.0247, "step": 69460 }, { "epoch": 0.17, "learning_rate": 8.277769389501233e-05, "loss": 1.0253, "step": 69465 }, { "epoch": 0.17, "learning_rate": 8.277643565353062e-05, "loss": 1.027, "step": 69470 }, { "epoch": 0.17, "learning_rate": 8.277517741204893e-05, "loss": 1.0253, "step": 69475 }, { "epoch": 0.17, "learning_rate": 8.277391917056722e-05, "loss": 1.0263, "step": 69480 }, { "epoch": 0.17, "learning_rate": 8.277266092908551e-05, "loss": 1.0272, "step": 69485 }, { "epoch": 0.17, "learning_rate": 8.27714026876038e-05, "loss": 1.0246, "step": 69490 }, { "epoch": 0.17, "learning_rate": 8.27701444461221e-05, "loss": 1.024, "step": 69495 }, { "epoch": 0.17, "learning_rate": 8.27688862046404e-05, "loss": 1.0233, "step": 69500 }, { "epoch": 0.17, "learning_rate": 8.276762796315869e-05, "loss": 1.0253, "step": 69505 }, { "epoch": 0.17, "learning_rate": 8.276636972167698e-05, "loss": 1.0264, "step": 69510 }, { "epoch": 0.17, "learning_rate": 8.276511148019529e-05, "loss": 1.0265, "step": 69515 }, { "epoch": 0.17, "learning_rate": 8.276385323871358e-05, "loss": 1.0266, "step": 69520 }, { "epoch": 0.17, "learning_rate": 8.276259499723187e-05, "loss": 1.0244, "step": 69525 }, { "epoch": 0.17, "learning_rate": 8.276133675575016e-05, "loss": 1.025, "step": 69530 }, { "epoch": 0.17, "learning_rate": 8.276007851426845e-05, "loss": 1.0262, "step": 69535 }, { "epoch": 0.17, "learning_rate": 8.275882027278676e-05, "loss": 1.0257, "step": 69540 }, { "epoch": 0.17, "learning_rate": 8.275756203130505e-05, "loss": 1.0252, "step": 69545 }, { "epoch": 0.17, "learning_rate": 8.275630378982334e-05, "loss": 1.0256, "step": 69550 }, { "epoch": 0.17, "learning_rate": 8.275504554834163e-05, "loss": 1.028, "step": 69555 }, { "epoch": 0.17, "learning_rate": 8.275378730685994e-05, "loss": 1.0265, "step": 69560 }, { "epoch": 0.17, "learning_rate": 8.275252906537823e-05, "loss": 1.0259, "step": 69565 }, { "epoch": 0.17, "learning_rate": 8.275127082389652e-05, "loss": 1.0244, "step": 69570 }, { "epoch": 0.17, "learning_rate": 8.275001258241482e-05, "loss": 1.0259, "step": 69575 }, { "epoch": 0.17, "learning_rate": 8.274875434093312e-05, "loss": 1.0251, "step": 69580 }, { "epoch": 0.17, "learning_rate": 8.274749609945142e-05, "loss": 1.0258, "step": 69585 }, { "epoch": 0.17, "learning_rate": 8.274623785796971e-05, "loss": 1.0264, "step": 69590 }, { "epoch": 0.17, "learning_rate": 8.2744979616488e-05, "loss": 1.0274, "step": 69595 }, { "epoch": 0.17, "learning_rate": 8.27437213750063e-05, "loss": 1.0259, "step": 69600 }, { "epoch": 0.17, "learning_rate": 8.274246313352459e-05, "loss": 1.0258, "step": 69605 }, { "epoch": 0.17, "learning_rate": 8.274120489204289e-05, "loss": 1.0262, "step": 69610 }, { "epoch": 0.17, "learning_rate": 8.273994665056118e-05, "loss": 1.0262, "step": 69615 }, { "epoch": 0.17, "learning_rate": 8.273868840907948e-05, "loss": 1.0254, "step": 69620 }, { "epoch": 0.17, "learning_rate": 8.273743016759777e-05, "loss": 1.0247, "step": 69625 }, { "epoch": 0.17, "learning_rate": 8.273617192611607e-05, "loss": 1.0284, "step": 69630 }, { "epoch": 0.17, "learning_rate": 8.273491368463436e-05, "loss": 1.0273, "step": 69635 }, { "epoch": 0.17, "learning_rate": 8.273365544315266e-05, "loss": 1.0253, "step": 69640 }, { "epoch": 0.17, "learning_rate": 8.273239720167095e-05, "loss": 1.0256, "step": 69645 }, { "epoch": 0.17, "learning_rate": 8.273113896018925e-05, "loss": 1.0255, "step": 69650 }, { "epoch": 0.17, "learning_rate": 8.272988071870754e-05, "loss": 1.0254, "step": 69655 }, { "epoch": 0.17, "learning_rate": 8.272862247722584e-05, "loss": 1.0278, "step": 69660 }, { "epoch": 0.17, "learning_rate": 8.272736423574413e-05, "loss": 1.0254, "step": 69665 }, { "epoch": 0.17, "learning_rate": 8.272610599426242e-05, "loss": 1.0255, "step": 69670 }, { "epoch": 0.17, "learning_rate": 8.272484775278072e-05, "loss": 1.0273, "step": 69675 }, { "epoch": 0.17, "learning_rate": 8.272358951129902e-05, "loss": 1.0243, "step": 69680 }, { "epoch": 0.17, "learning_rate": 8.272233126981731e-05, "loss": 1.0239, "step": 69685 }, { "epoch": 0.17, "learning_rate": 8.27210730283356e-05, "loss": 1.0263, "step": 69690 }, { "epoch": 0.17, "learning_rate": 8.27198147868539e-05, "loss": 1.0261, "step": 69695 }, { "epoch": 0.17, "learning_rate": 8.27185565453722e-05, "loss": 1.0251, "step": 69700 }, { "epoch": 0.17, "learning_rate": 8.271729830389049e-05, "loss": 1.0282, "step": 69705 }, { "epoch": 0.17, "learning_rate": 8.271604006240878e-05, "loss": 1.0412, "step": 69710 }, { "epoch": 0.17, "learning_rate": 8.271478182092708e-05, "loss": 1.0273, "step": 69715 }, { "epoch": 0.18, "learning_rate": 8.271352357944538e-05, "loss": 1.0267, "step": 69720 }, { "epoch": 0.18, "learning_rate": 8.271226533796367e-05, "loss": 1.0255, "step": 69725 }, { "epoch": 0.18, "learning_rate": 8.271100709648196e-05, "loss": 1.0252, "step": 69730 }, { "epoch": 0.18, "learning_rate": 8.270974885500025e-05, "loss": 1.0245, "step": 69735 }, { "epoch": 0.18, "learning_rate": 8.270849061351856e-05, "loss": 1.0277, "step": 69740 }, { "epoch": 0.18, "learning_rate": 8.270723237203685e-05, "loss": 1.0252, "step": 69745 }, { "epoch": 0.18, "learning_rate": 8.270597413055514e-05, "loss": 1.0256, "step": 69750 }, { "epoch": 0.18, "learning_rate": 8.270471588907343e-05, "loss": 1.0254, "step": 69755 }, { "epoch": 0.18, "learning_rate": 8.270345764759174e-05, "loss": 1.0236, "step": 69760 }, { "epoch": 0.18, "learning_rate": 8.270219940611003e-05, "loss": 1.0253, "step": 69765 }, { "epoch": 0.18, "learning_rate": 8.270094116462832e-05, "loss": 1.0239, "step": 69770 }, { "epoch": 0.18, "learning_rate": 8.269968292314661e-05, "loss": 1.0262, "step": 69775 }, { "epoch": 0.18, "learning_rate": 8.269842468166492e-05, "loss": 1.0274, "step": 69780 }, { "epoch": 0.18, "learning_rate": 8.269716644018321e-05, "loss": 1.0267, "step": 69785 }, { "epoch": 0.18, "learning_rate": 8.26959081987015e-05, "loss": 1.0253, "step": 69790 }, { "epoch": 0.18, "learning_rate": 8.269464995721979e-05, "loss": 1.0214, "step": 69795 }, { "epoch": 0.18, "learning_rate": 8.269339171573808e-05, "loss": 1.0262, "step": 69800 }, { "epoch": 0.18, "learning_rate": 8.269213347425639e-05, "loss": 1.0248, "step": 69805 }, { "epoch": 0.18, "learning_rate": 8.269087523277468e-05, "loss": 1.0265, "step": 69810 }, { "epoch": 0.18, "learning_rate": 8.268961699129297e-05, "loss": 1.0258, "step": 69815 }, { "epoch": 0.18, "learning_rate": 8.268835874981126e-05, "loss": 1.0257, "step": 69820 }, { "epoch": 0.18, "learning_rate": 8.268710050832957e-05, "loss": 1.0232, "step": 69825 }, { "epoch": 0.18, "learning_rate": 8.268584226684786e-05, "loss": 1.027, "step": 69830 }, { "epoch": 0.18, "learning_rate": 8.268458402536615e-05, "loss": 1.0251, "step": 69835 }, { "epoch": 0.18, "learning_rate": 8.268332578388444e-05, "loss": 1.025, "step": 69840 }, { "epoch": 0.18, "learning_rate": 8.268206754240275e-05, "loss": 1.0246, "step": 69845 }, { "epoch": 0.18, "learning_rate": 8.268080930092104e-05, "loss": 1.0246, "step": 69850 }, { "epoch": 0.18, "learning_rate": 8.267955105943933e-05, "loss": 1.0252, "step": 69855 }, { "epoch": 0.18, "learning_rate": 8.267829281795762e-05, "loss": 1.0259, "step": 69860 }, { "epoch": 0.18, "learning_rate": 8.267703457647591e-05, "loss": 1.0261, "step": 69865 }, { "epoch": 0.18, "learning_rate": 8.267577633499422e-05, "loss": 1.0249, "step": 69870 }, { "epoch": 0.18, "learning_rate": 8.267451809351251e-05, "loss": 1.0257, "step": 69875 }, { "epoch": 0.18, "learning_rate": 8.26732598520308e-05, "loss": 1.034, "step": 69880 }, { "epoch": 0.18, "learning_rate": 8.26720016105491e-05, "loss": 1.0257, "step": 69885 }, { "epoch": 0.18, "learning_rate": 8.26707433690674e-05, "loss": 1.0267, "step": 69890 }, { "epoch": 0.18, "learning_rate": 8.266948512758569e-05, "loss": 1.0267, "step": 69895 }, { "epoch": 0.18, "learning_rate": 8.266822688610398e-05, "loss": 1.0263, "step": 69900 }, { "epoch": 0.18, "learning_rate": 8.266696864462227e-05, "loss": 1.0235, "step": 69905 }, { "epoch": 0.18, "learning_rate": 8.266571040314058e-05, "loss": 1.0251, "step": 69910 }, { "epoch": 0.18, "learning_rate": 8.266445216165887e-05, "loss": 1.0243, "step": 69915 }, { "epoch": 0.18, "learning_rate": 8.266319392017716e-05, "loss": 1.0236, "step": 69920 }, { "epoch": 0.18, "learning_rate": 8.266193567869545e-05, "loss": 1.0261, "step": 69925 }, { "epoch": 0.18, "learning_rate": 8.266067743721375e-05, "loss": 1.0226, "step": 69930 }, { "epoch": 0.18, "learning_rate": 8.265941919573205e-05, "loss": 1.0282, "step": 69935 }, { "epoch": 0.18, "learning_rate": 8.265816095425034e-05, "loss": 1.0259, "step": 69940 }, { "epoch": 0.18, "learning_rate": 8.265690271276863e-05, "loss": 1.0284, "step": 69945 }, { "epoch": 0.18, "learning_rate": 8.265564447128693e-05, "loss": 1.0254, "step": 69950 }, { "epoch": 0.18, "learning_rate": 8.265438622980523e-05, "loss": 1.0245, "step": 69955 }, { "epoch": 0.18, "learning_rate": 8.265312798832352e-05, "loss": 1.0469, "step": 69960 }, { "epoch": 0.18, "learning_rate": 8.265186974684181e-05, "loss": 1.0272, "step": 69965 }, { "epoch": 0.18, "learning_rate": 8.26506115053601e-05, "loss": 1.0264, "step": 69970 }, { "epoch": 0.18, "learning_rate": 8.264935326387841e-05, "loss": 1.0257, "step": 69975 }, { "epoch": 0.18, "learning_rate": 8.26480950223967e-05, "loss": 1.0248, "step": 69980 }, { "epoch": 0.18, "learning_rate": 8.2646836780915e-05, "loss": 1.0246, "step": 69985 }, { "epoch": 0.18, "learning_rate": 8.264557853943328e-05, "loss": 1.0267, "step": 69990 }, { "epoch": 0.18, "learning_rate": 8.264432029795158e-05, "loss": 1.0277, "step": 69995 }, { "epoch": 0.18, "learning_rate": 8.264306205646988e-05, "loss": 1.0316, "step": 70000 }, { "epoch": 0.18, "learning_rate": 8.264180381498817e-05, "loss": 1.0268, "step": 70005 }, { "epoch": 0.18, "learning_rate": 8.264054557350646e-05, "loss": 1.0248, "step": 70010 }, { "epoch": 0.18, "learning_rate": 8.263928733202476e-05, "loss": 1.0269, "step": 70015 }, { "epoch": 0.18, "learning_rate": 8.263802909054306e-05, "loss": 1.0482, "step": 70020 }, { "epoch": 0.18, "learning_rate": 8.263677084906135e-05, "loss": 1.0259, "step": 70025 }, { "epoch": 0.18, "learning_rate": 8.263551260757964e-05, "loss": 1.025, "step": 70030 }, { "epoch": 0.18, "learning_rate": 8.263425436609794e-05, "loss": 1.0252, "step": 70035 }, { "epoch": 0.18, "learning_rate": 8.263299612461624e-05, "loss": 1.0241, "step": 70040 }, { "epoch": 0.18, "learning_rate": 8.263173788313453e-05, "loss": 1.0239, "step": 70045 }, { "epoch": 0.18, "learning_rate": 8.263047964165282e-05, "loss": 1.0248, "step": 70050 }, { "epoch": 0.18, "learning_rate": 8.262922140017112e-05, "loss": 1.0262, "step": 70055 }, { "epoch": 0.18, "learning_rate": 8.262796315868941e-05, "loss": 1.0236, "step": 70060 }, { "epoch": 0.18, "learning_rate": 8.262670491720771e-05, "loss": 1.035, "step": 70065 }, { "epoch": 0.18, "learning_rate": 8.262544667572602e-05, "loss": 1.026, "step": 70070 }, { "epoch": 0.18, "learning_rate": 8.262418843424431e-05, "loss": 1.026, "step": 70075 }, { "epoch": 0.18, "learning_rate": 8.26229301927626e-05, "loss": 1.0267, "step": 70080 }, { "epoch": 0.18, "learning_rate": 8.262167195128089e-05, "loss": 1.0266, "step": 70085 }, { "epoch": 0.18, "learning_rate": 8.26204137097992e-05, "loss": 1.0242, "step": 70090 }, { "epoch": 0.18, "learning_rate": 8.261915546831749e-05, "loss": 1.0242, "step": 70095 }, { "epoch": 0.18, "learning_rate": 8.261789722683578e-05, "loss": 1.0272, "step": 70100 }, { "epoch": 0.18, "learning_rate": 8.261663898535407e-05, "loss": 1.0259, "step": 70105 }, { "epoch": 0.18, "learning_rate": 8.261538074387238e-05, "loss": 1.0262, "step": 70110 }, { "epoch": 0.18, "learning_rate": 8.261412250239067e-05, "loss": 1.0248, "step": 70115 }, { "epoch": 0.18, "learning_rate": 8.261286426090896e-05, "loss": 1.0246, "step": 70120 }, { "epoch": 0.18, "learning_rate": 8.261160601942725e-05, "loss": 1.0262, "step": 70125 }, { "epoch": 0.18, "learning_rate": 8.261034777794556e-05, "loss": 1.0253, "step": 70130 }, { "epoch": 0.18, "learning_rate": 8.260908953646385e-05, "loss": 1.0259, "step": 70135 }, { "epoch": 0.18, "learning_rate": 8.260783129498214e-05, "loss": 1.0254, "step": 70140 }, { "epoch": 0.18, "learning_rate": 8.260657305350043e-05, "loss": 1.0266, "step": 70145 }, { "epoch": 0.18, "learning_rate": 8.260531481201872e-05, "loss": 1.0256, "step": 70150 }, { "epoch": 0.18, "learning_rate": 8.260405657053703e-05, "loss": 1.0246, "step": 70155 }, { "epoch": 0.18, "learning_rate": 8.260279832905532e-05, "loss": 1.0254, "step": 70160 }, { "epoch": 0.18, "learning_rate": 8.260154008757361e-05, "loss": 1.0247, "step": 70165 }, { "epoch": 0.18, "learning_rate": 8.26002818460919e-05, "loss": 1.0249, "step": 70170 }, { "epoch": 0.18, "learning_rate": 8.259902360461021e-05, "loss": 1.0253, "step": 70175 }, { "epoch": 0.18, "learning_rate": 8.25977653631285e-05, "loss": 1.0264, "step": 70180 }, { "epoch": 0.18, "learning_rate": 8.259650712164679e-05, "loss": 1.024, "step": 70185 }, { "epoch": 0.18, "learning_rate": 8.259524888016508e-05, "loss": 1.0246, "step": 70190 }, { "epoch": 0.18, "learning_rate": 8.259399063868339e-05, "loss": 1.0254, "step": 70195 }, { "epoch": 0.18, "learning_rate": 8.259273239720168e-05, "loss": 1.0249, "step": 70200 }, { "epoch": 0.18, "learning_rate": 8.259147415571997e-05, "loss": 1.025, "step": 70205 }, { "epoch": 0.18, "learning_rate": 8.259021591423826e-05, "loss": 1.0264, "step": 70210 }, { "epoch": 0.18, "learning_rate": 8.258895767275656e-05, "loss": 1.0274, "step": 70215 }, { "epoch": 0.18, "learning_rate": 8.258769943127486e-05, "loss": 1.0277, "step": 70220 }, { "epoch": 0.18, "learning_rate": 8.258644118979315e-05, "loss": 1.0264, "step": 70225 }, { "epoch": 0.18, "learning_rate": 8.258518294831144e-05, "loss": 1.025, "step": 70230 }, { "epoch": 0.18, "learning_rate": 8.258392470682974e-05, "loss": 1.0271, "step": 70235 }, { "epoch": 0.18, "learning_rate": 8.258266646534804e-05, "loss": 1.0264, "step": 70240 }, { "epoch": 0.18, "learning_rate": 8.258140822386633e-05, "loss": 1.0281, "step": 70245 }, { "epoch": 0.18, "learning_rate": 8.258014998238462e-05, "loss": 1.025, "step": 70250 }, { "epoch": 0.18, "learning_rate": 8.257889174090292e-05, "loss": 1.0262, "step": 70255 }, { "epoch": 0.18, "learning_rate": 8.257763349942121e-05, "loss": 1.0261, "step": 70260 }, { "epoch": 0.18, "learning_rate": 8.257637525793951e-05, "loss": 1.0271, "step": 70265 }, { "epoch": 0.18, "learning_rate": 8.25751170164578e-05, "loss": 1.0273, "step": 70270 }, { "epoch": 0.18, "learning_rate": 8.25738587749761e-05, "loss": 1.0252, "step": 70275 }, { "epoch": 0.18, "learning_rate": 8.257260053349439e-05, "loss": 1.0274, "step": 70280 }, { "epoch": 0.18, "learning_rate": 8.257134229201269e-05, "loss": 1.0248, "step": 70285 }, { "epoch": 0.18, "learning_rate": 8.257008405053098e-05, "loss": 1.0246, "step": 70290 }, { "epoch": 0.18, "learning_rate": 8.256882580904928e-05, "loss": 1.0245, "step": 70295 }, { "epoch": 0.18, "learning_rate": 8.256756756756757e-05, "loss": 1.0275, "step": 70300 }, { "epoch": 0.18, "learning_rate": 8.256630932608587e-05, "loss": 1.0257, "step": 70305 }, { "epoch": 0.18, "learning_rate": 8.256505108460416e-05, "loss": 1.0261, "step": 70310 }, { "epoch": 0.18, "learning_rate": 8.256379284312246e-05, "loss": 1.0264, "step": 70315 }, { "epoch": 0.18, "learning_rate": 8.256253460164075e-05, "loss": 1.0265, "step": 70320 }, { "epoch": 0.18, "learning_rate": 8.256127636015904e-05, "loss": 1.0247, "step": 70325 }, { "epoch": 0.18, "learning_rate": 8.256001811867734e-05, "loss": 1.0261, "step": 70330 }, { "epoch": 0.18, "learning_rate": 8.255875987719564e-05, "loss": 1.0263, "step": 70335 }, { "epoch": 0.18, "learning_rate": 8.255750163571393e-05, "loss": 1.026, "step": 70340 }, { "epoch": 0.18, "learning_rate": 8.255624339423222e-05, "loss": 1.0247, "step": 70345 }, { "epoch": 0.18, "learning_rate": 8.255498515275052e-05, "loss": 1.0235, "step": 70350 }, { "epoch": 0.18, "learning_rate": 8.255372691126882e-05, "loss": 1.0255, "step": 70355 }, { "epoch": 0.18, "learning_rate": 8.255246866978711e-05, "loss": 1.026, "step": 70360 }, { "epoch": 0.18, "learning_rate": 8.25512104283054e-05, "loss": 1.0294, "step": 70365 }, { "epoch": 0.18, "learning_rate": 8.25499521868237e-05, "loss": 1.0276, "step": 70370 }, { "epoch": 0.18, "learning_rate": 8.2548693945342e-05, "loss": 1.0242, "step": 70375 }, { "epoch": 0.18, "learning_rate": 8.254743570386029e-05, "loss": 1.0239, "step": 70380 }, { "epoch": 0.18, "learning_rate": 8.254617746237858e-05, "loss": 1.0278, "step": 70385 }, { "epoch": 0.18, "learning_rate": 8.254491922089687e-05, "loss": 1.0257, "step": 70390 }, { "epoch": 0.18, "learning_rate": 8.254366097941518e-05, "loss": 1.0261, "step": 70395 }, { "epoch": 0.18, "learning_rate": 8.254240273793347e-05, "loss": 1.0222, "step": 70400 }, { "epoch": 0.18, "learning_rate": 8.254114449645176e-05, "loss": 1.0221, "step": 70405 }, { "epoch": 0.18, "learning_rate": 8.253988625497005e-05, "loss": 1.0252, "step": 70410 }, { "epoch": 0.18, "learning_rate": 8.253862801348836e-05, "loss": 1.0226, "step": 70415 }, { "epoch": 0.18, "learning_rate": 8.253736977200665e-05, "loss": 1.0279, "step": 70420 }, { "epoch": 0.18, "learning_rate": 8.253611153052494e-05, "loss": 1.025, "step": 70425 }, { "epoch": 0.18, "learning_rate": 8.253485328904323e-05, "loss": 1.0255, "step": 70430 }, { "epoch": 0.18, "learning_rate": 8.253359504756154e-05, "loss": 1.0266, "step": 70435 }, { "epoch": 0.18, "learning_rate": 8.253233680607983e-05, "loss": 1.0233, "step": 70440 }, { "epoch": 0.18, "learning_rate": 8.253107856459812e-05, "loss": 1.0262, "step": 70445 }, { "epoch": 0.18, "learning_rate": 8.252982032311641e-05, "loss": 1.0244, "step": 70450 }, { "epoch": 0.18, "learning_rate": 8.25285620816347e-05, "loss": 1.0252, "step": 70455 }, { "epoch": 0.18, "learning_rate": 8.252730384015301e-05, "loss": 1.0262, "step": 70460 }, { "epoch": 0.18, "learning_rate": 8.25260455986713e-05, "loss": 1.0261, "step": 70465 }, { "epoch": 0.18, "learning_rate": 8.252478735718959e-05, "loss": 1.0253, "step": 70470 }, { "epoch": 0.18, "learning_rate": 8.252352911570788e-05, "loss": 1.0255, "step": 70475 }, { "epoch": 0.18, "learning_rate": 8.252227087422619e-05, "loss": 1.0256, "step": 70480 }, { "epoch": 0.18, "learning_rate": 8.252101263274448e-05, "loss": 1.0257, "step": 70485 }, { "epoch": 0.18, "learning_rate": 8.251975439126277e-05, "loss": 1.0245, "step": 70490 }, { "epoch": 0.18, "learning_rate": 8.251849614978106e-05, "loss": 1.0254, "step": 70495 }, { "epoch": 0.18, "learning_rate": 8.251723790829937e-05, "loss": 1.0251, "step": 70500 }, { "epoch": 0.18, "learning_rate": 8.251597966681766e-05, "loss": 1.0273, "step": 70505 }, { "epoch": 0.18, "learning_rate": 8.251472142533595e-05, "loss": 1.0266, "step": 70510 }, { "epoch": 0.18, "learning_rate": 8.251346318385424e-05, "loss": 1.0255, "step": 70515 }, { "epoch": 0.18, "learning_rate": 8.251220494237253e-05, "loss": 1.0265, "step": 70520 }, { "epoch": 0.18, "learning_rate": 8.251094670089084e-05, "loss": 1.0274, "step": 70525 }, { "epoch": 0.18, "learning_rate": 8.250968845940913e-05, "loss": 1.0255, "step": 70530 }, { "epoch": 0.18, "learning_rate": 8.250843021792742e-05, "loss": 1.025, "step": 70535 }, { "epoch": 0.18, "learning_rate": 8.250717197644571e-05, "loss": 1.0242, "step": 70540 }, { "epoch": 0.18, "learning_rate": 8.250591373496402e-05, "loss": 1.0263, "step": 70545 }, { "epoch": 0.18, "learning_rate": 8.250465549348231e-05, "loss": 1.0253, "step": 70550 }, { "epoch": 0.18, "learning_rate": 8.25033972520006e-05, "loss": 1.0258, "step": 70555 }, { "epoch": 0.18, "learning_rate": 8.250213901051889e-05, "loss": 1.0234, "step": 70560 }, { "epoch": 0.18, "learning_rate": 8.25008807690372e-05, "loss": 1.0239, "step": 70565 }, { "epoch": 0.18, "learning_rate": 8.24996225275555e-05, "loss": 1.0243, "step": 70570 }, { "epoch": 0.18, "learning_rate": 8.24983642860738e-05, "loss": 1.0275, "step": 70575 }, { "epoch": 0.18, "learning_rate": 8.249710604459209e-05, "loss": 1.0267, "step": 70580 }, { "epoch": 0.18, "learning_rate": 8.249584780311038e-05, "loss": 1.0231, "step": 70585 }, { "epoch": 0.18, "learning_rate": 8.249458956162868e-05, "loss": 1.0231, "step": 70590 }, { "epoch": 0.18, "learning_rate": 8.249333132014697e-05, "loss": 1.0257, "step": 70595 }, { "epoch": 0.18, "learning_rate": 8.249207307866527e-05, "loss": 1.0235, "step": 70600 }, { "epoch": 0.18, "learning_rate": 8.249081483718356e-05, "loss": 1.0244, "step": 70605 }, { "epoch": 0.18, "learning_rate": 8.248955659570185e-05, "loss": 1.0275, "step": 70610 }, { "epoch": 0.18, "learning_rate": 8.248829835422015e-05, "loss": 1.0256, "step": 70615 }, { "epoch": 0.18, "learning_rate": 8.248704011273845e-05, "loss": 1.0256, "step": 70620 }, { "epoch": 0.18, "learning_rate": 8.248578187125674e-05, "loss": 1.0266, "step": 70625 }, { "epoch": 0.18, "learning_rate": 8.248452362977503e-05, "loss": 1.0232, "step": 70630 }, { "epoch": 0.18, "learning_rate": 8.248326538829333e-05, "loss": 1.0257, "step": 70635 }, { "epoch": 0.18, "learning_rate": 8.248200714681163e-05, "loss": 1.024, "step": 70640 }, { "epoch": 0.18, "learning_rate": 8.248074890532992e-05, "loss": 1.0254, "step": 70645 }, { "epoch": 0.18, "learning_rate": 8.247949066384821e-05, "loss": 1.0261, "step": 70650 }, { "epoch": 0.18, "learning_rate": 8.247823242236651e-05, "loss": 1.025, "step": 70655 }, { "epoch": 0.18, "learning_rate": 8.24769741808848e-05, "loss": 1.0252, "step": 70660 }, { "epoch": 0.18, "learning_rate": 8.24757159394031e-05, "loss": 1.0255, "step": 70665 }, { "epoch": 0.18, "learning_rate": 8.247445769792139e-05, "loss": 1.0248, "step": 70670 }, { "epoch": 0.18, "learning_rate": 8.247319945643968e-05, "loss": 1.0272, "step": 70675 }, { "epoch": 0.18, "learning_rate": 8.247194121495799e-05, "loss": 1.0238, "step": 70680 }, { "epoch": 0.18, "learning_rate": 8.247068297347628e-05, "loss": 1.0284, "step": 70685 }, { "epoch": 0.18, "learning_rate": 8.246942473199457e-05, "loss": 1.0271, "step": 70690 }, { "epoch": 0.18, "learning_rate": 8.246816649051286e-05, "loss": 1.0296, "step": 70695 }, { "epoch": 0.18, "learning_rate": 8.246690824903117e-05, "loss": 1.0261, "step": 70700 }, { "epoch": 0.18, "learning_rate": 8.246565000754946e-05, "loss": 1.0297, "step": 70705 }, { "epoch": 0.18, "learning_rate": 8.246439176606775e-05, "loss": 1.0265, "step": 70710 }, { "epoch": 0.18, "learning_rate": 8.246313352458604e-05, "loss": 1.0247, "step": 70715 }, { "epoch": 0.18, "learning_rate": 8.246187528310435e-05, "loss": 1.0271, "step": 70720 }, { "epoch": 0.18, "learning_rate": 8.246061704162264e-05, "loss": 1.0268, "step": 70725 }, { "epoch": 0.18, "learning_rate": 8.245935880014093e-05, "loss": 1.0248, "step": 70730 }, { "epoch": 0.18, "learning_rate": 8.245810055865922e-05, "loss": 1.0276, "step": 70735 }, { "epoch": 0.18, "learning_rate": 8.245684231717751e-05, "loss": 1.025, "step": 70740 }, { "epoch": 0.18, "learning_rate": 8.245558407569582e-05, "loss": 1.026, "step": 70745 }, { "epoch": 0.18, "learning_rate": 8.245432583421411e-05, "loss": 1.0251, "step": 70750 }, { "epoch": 0.18, "learning_rate": 8.24530675927324e-05, "loss": 1.025, "step": 70755 }, { "epoch": 0.18, "learning_rate": 8.245180935125069e-05, "loss": 1.0258, "step": 70760 }, { "epoch": 0.18, "learning_rate": 8.2450551109769e-05, "loss": 1.025, "step": 70765 }, { "epoch": 0.18, "learning_rate": 8.244929286828729e-05, "loss": 1.025, "step": 70770 }, { "epoch": 0.18, "learning_rate": 8.244803462680558e-05, "loss": 1.0245, "step": 70775 }, { "epoch": 0.18, "learning_rate": 8.244677638532387e-05, "loss": 1.0244, "step": 70780 }, { "epoch": 0.18, "learning_rate": 8.244551814384218e-05, "loss": 1.0247, "step": 70785 }, { "epoch": 0.18, "learning_rate": 8.244425990236047e-05, "loss": 1.024, "step": 70790 }, { "epoch": 0.18, "learning_rate": 8.244300166087876e-05, "loss": 1.0278, "step": 70795 }, { "epoch": 0.18, "learning_rate": 8.244174341939705e-05, "loss": 1.0264, "step": 70800 }, { "epoch": 0.18, "learning_rate": 8.244048517791534e-05, "loss": 1.027, "step": 70805 }, { "epoch": 0.18, "learning_rate": 8.243922693643365e-05, "loss": 1.0268, "step": 70810 }, { "epoch": 0.18, "learning_rate": 8.243796869495194e-05, "loss": 1.0261, "step": 70815 }, { "epoch": 0.18, "learning_rate": 8.243671045347023e-05, "loss": 1.0261, "step": 70820 }, { "epoch": 0.18, "learning_rate": 8.243545221198852e-05, "loss": 1.0247, "step": 70825 }, { "epoch": 0.18, "learning_rate": 8.243419397050683e-05, "loss": 1.0246, "step": 70830 }, { "epoch": 0.18, "learning_rate": 8.243293572902512e-05, "loss": 1.0417, "step": 70835 }, { "epoch": 0.18, "learning_rate": 8.243167748754341e-05, "loss": 1.0246, "step": 70840 }, { "epoch": 0.18, "learning_rate": 8.24304192460617e-05, "loss": 1.0234, "step": 70845 }, { "epoch": 0.18, "learning_rate": 8.242916100458001e-05, "loss": 1.0247, "step": 70850 }, { "epoch": 0.18, "learning_rate": 8.24279027630983e-05, "loss": 1.0262, "step": 70855 }, { "epoch": 0.18, "learning_rate": 8.242664452161659e-05, "loss": 1.0243, "step": 70860 }, { "epoch": 0.18, "learning_rate": 8.242538628013488e-05, "loss": 1.0278, "step": 70865 }, { "epoch": 0.18, "learning_rate": 8.242412803865318e-05, "loss": 1.0271, "step": 70870 }, { "epoch": 0.18, "learning_rate": 8.242286979717148e-05, "loss": 1.0256, "step": 70875 }, { "epoch": 0.18, "learning_rate": 8.242161155568977e-05, "loss": 1.0248, "step": 70880 }, { "epoch": 0.18, "learning_rate": 8.242035331420806e-05, "loss": 1.0261, "step": 70885 }, { "epoch": 0.18, "learning_rate": 8.241909507272636e-05, "loss": 1.0263, "step": 70890 }, { "epoch": 0.18, "learning_rate": 8.241783683124466e-05, "loss": 1.026, "step": 70895 }, { "epoch": 0.18, "learning_rate": 8.241657858976295e-05, "loss": 1.0255, "step": 70900 }, { "epoch": 0.18, "learning_rate": 8.241532034828124e-05, "loss": 1.024, "step": 70905 }, { "epoch": 0.18, "learning_rate": 8.241406210679954e-05, "loss": 1.0255, "step": 70910 }, { "epoch": 0.18, "learning_rate": 8.241280386531784e-05, "loss": 1.0267, "step": 70915 }, { "epoch": 0.18, "learning_rate": 8.241154562383613e-05, "loss": 1.0276, "step": 70920 }, { "epoch": 0.18, "learning_rate": 8.241028738235442e-05, "loss": 1.0243, "step": 70925 }, { "epoch": 0.18, "learning_rate": 8.240902914087272e-05, "loss": 1.0246, "step": 70930 }, { "epoch": 0.18, "learning_rate": 8.2407770899391e-05, "loss": 1.0511, "step": 70935 }, { "epoch": 0.18, "learning_rate": 8.240651265790931e-05, "loss": 1.0253, "step": 70940 }, { "epoch": 0.18, "learning_rate": 8.24052544164276e-05, "loss": 1.0279, "step": 70945 }, { "epoch": 0.18, "learning_rate": 8.24039961749459e-05, "loss": 1.0268, "step": 70950 }, { "epoch": 0.18, "learning_rate": 8.240273793346419e-05, "loss": 1.0255, "step": 70955 }, { "epoch": 0.18, "learning_rate": 8.240147969198249e-05, "loss": 1.0252, "step": 70960 }, { "epoch": 0.18, "learning_rate": 8.240022145050078e-05, "loss": 1.0499, "step": 70965 }, { "epoch": 0.18, "learning_rate": 8.239896320901908e-05, "loss": 1.0241, "step": 70970 }, { "epoch": 0.18, "learning_rate": 8.239770496753737e-05, "loss": 1.0262, "step": 70975 }, { "epoch": 0.18, "learning_rate": 8.239644672605566e-05, "loss": 1.0261, "step": 70980 }, { "epoch": 0.18, "learning_rate": 8.239518848457396e-05, "loss": 1.0249, "step": 70985 }, { "epoch": 0.18, "learning_rate": 8.239393024309225e-05, "loss": 1.0246, "step": 70990 }, { "epoch": 0.18, "learning_rate": 8.239267200161055e-05, "loss": 1.0267, "step": 70995 }, { "epoch": 0.18, "learning_rate": 8.239141376012884e-05, "loss": 1.0249, "step": 71000 }, { "epoch": 0.18, "learning_rate": 8.239015551864714e-05, "loss": 1.0255, "step": 71005 }, { "epoch": 0.18, "learning_rate": 8.238889727716543e-05, "loss": 1.0247, "step": 71010 }, { "epoch": 0.18, "learning_rate": 8.238763903568373e-05, "loss": 1.0275, "step": 71015 }, { "epoch": 0.18, "learning_rate": 8.238638079420202e-05, "loss": 1.049, "step": 71020 }, { "epoch": 0.18, "learning_rate": 8.238512255272032e-05, "loss": 1.0259, "step": 71025 }, { "epoch": 0.18, "learning_rate": 8.238386431123861e-05, "loss": 1.0263, "step": 71030 }, { "epoch": 0.18, "learning_rate": 8.23826060697569e-05, "loss": 1.0251, "step": 71035 }, { "epoch": 0.18, "learning_rate": 8.23813478282752e-05, "loss": 1.0246, "step": 71040 }, { "epoch": 0.18, "learning_rate": 8.238008958679349e-05, "loss": 1.0245, "step": 71045 }, { "epoch": 0.18, "learning_rate": 8.23788313453118e-05, "loss": 1.028, "step": 71050 }, { "epoch": 0.18, "learning_rate": 8.237757310383009e-05, "loss": 1.023, "step": 71055 }, { "epoch": 0.18, "learning_rate": 8.237631486234838e-05, "loss": 1.0269, "step": 71060 }, { "epoch": 0.18, "learning_rate": 8.237505662086667e-05, "loss": 1.0235, "step": 71065 }, { "epoch": 0.18, "learning_rate": 8.237379837938497e-05, "loss": 1.0233, "step": 71070 }, { "epoch": 0.18, "learning_rate": 8.237254013790328e-05, "loss": 1.0267, "step": 71075 }, { "epoch": 0.18, "learning_rate": 8.237128189642157e-05, "loss": 1.0243, "step": 71080 }, { "epoch": 0.18, "learning_rate": 8.237002365493986e-05, "loss": 1.0255, "step": 71085 }, { "epoch": 0.18, "learning_rate": 8.236876541345815e-05, "loss": 1.0251, "step": 71090 }, { "epoch": 0.18, "learning_rate": 8.236750717197646e-05, "loss": 1.0233, "step": 71095 }, { "epoch": 0.18, "learning_rate": 8.236624893049475e-05, "loss": 1.0453, "step": 71100 }, { "epoch": 0.18, "learning_rate": 8.236499068901304e-05, "loss": 1.0257, "step": 71105 }, { "epoch": 0.18, "learning_rate": 8.236373244753133e-05, "loss": 1.0248, "step": 71110 }, { "epoch": 0.18, "learning_rate": 8.236247420604964e-05, "loss": 1.0232, "step": 71115 }, { "epoch": 0.18, "learning_rate": 8.236121596456793e-05, "loss": 1.0241, "step": 71120 }, { "epoch": 0.18, "learning_rate": 8.235995772308622e-05, "loss": 1.0244, "step": 71125 }, { "epoch": 0.18, "learning_rate": 8.235869948160451e-05, "loss": 1.0246, "step": 71130 }, { "epoch": 0.18, "learning_rate": 8.23574412401228e-05, "loss": 1.0261, "step": 71135 }, { "epoch": 0.18, "learning_rate": 8.235618299864111e-05, "loss": 1.0284, "step": 71140 }, { "epoch": 0.18, "learning_rate": 8.23549247571594e-05, "loss": 1.0244, "step": 71145 }, { "epoch": 0.18, "learning_rate": 8.23536665156777e-05, "loss": 1.0288, "step": 71150 }, { "epoch": 0.18, "learning_rate": 8.235240827419599e-05, "loss": 1.0251, "step": 71155 }, { "epoch": 0.18, "learning_rate": 8.235115003271429e-05, "loss": 1.0258, "step": 71160 }, { "epoch": 0.18, "learning_rate": 8.234989179123258e-05, "loss": 1.028, "step": 71165 }, { "epoch": 0.18, "learning_rate": 8.234863354975087e-05, "loss": 1.026, "step": 71170 }, { "epoch": 0.18, "learning_rate": 8.234737530826917e-05, "loss": 1.0257, "step": 71175 }, { "epoch": 0.18, "learning_rate": 8.234611706678747e-05, "loss": 1.0245, "step": 71180 }, { "epoch": 0.18, "learning_rate": 8.234485882530576e-05, "loss": 1.0261, "step": 71185 }, { "epoch": 0.18, "learning_rate": 8.234360058382405e-05, "loss": 1.0479, "step": 71190 }, { "epoch": 0.18, "learning_rate": 8.234234234234235e-05, "loss": 1.0256, "step": 71195 }, { "epoch": 0.18, "learning_rate": 8.234108410086064e-05, "loss": 1.0272, "step": 71200 }, { "epoch": 0.18, "learning_rate": 8.233982585937894e-05, "loss": 1.0262, "step": 71205 }, { "epoch": 0.18, "learning_rate": 8.233856761789723e-05, "loss": 1.0248, "step": 71210 }, { "epoch": 0.18, "learning_rate": 8.233730937641553e-05, "loss": 1.0243, "step": 71215 }, { "epoch": 0.18, "learning_rate": 8.233605113493382e-05, "loss": 1.0248, "step": 71220 }, { "epoch": 0.18, "learning_rate": 8.233479289345212e-05, "loss": 1.0255, "step": 71225 }, { "epoch": 0.18, "learning_rate": 8.233353465197041e-05, "loss": 1.0245, "step": 71230 }, { "epoch": 0.18, "learning_rate": 8.23322764104887e-05, "loss": 1.0253, "step": 71235 }, { "epoch": 0.18, "learning_rate": 8.2331018169007e-05, "loss": 1.0266, "step": 71240 }, { "epoch": 0.18, "learning_rate": 8.23297599275253e-05, "loss": 1.0228, "step": 71245 }, { "epoch": 0.18, "learning_rate": 8.23285016860436e-05, "loss": 1.0243, "step": 71250 }, { "epoch": 0.18, "learning_rate": 8.232724344456189e-05, "loss": 1.0231, "step": 71255 }, { "epoch": 0.18, "learning_rate": 8.232598520308018e-05, "loss": 1.0246, "step": 71260 }, { "epoch": 0.18, "learning_rate": 8.232472696159847e-05, "loss": 1.0273, "step": 71265 }, { "epoch": 0.18, "learning_rate": 8.232346872011677e-05, "loss": 1.0248, "step": 71270 }, { "epoch": 0.18, "learning_rate": 8.232221047863507e-05, "loss": 1.0242, "step": 71275 }, { "epoch": 0.18, "learning_rate": 8.232095223715336e-05, "loss": 1.0263, "step": 71280 }, { "epoch": 0.18, "learning_rate": 8.231969399567165e-05, "loss": 1.0246, "step": 71285 }, { "epoch": 0.18, "learning_rate": 8.231843575418995e-05, "loss": 1.0264, "step": 71290 }, { "epoch": 0.18, "learning_rate": 8.231717751270825e-05, "loss": 1.0243, "step": 71295 }, { "epoch": 0.18, "learning_rate": 8.231591927122654e-05, "loss": 1.0265, "step": 71300 }, { "epoch": 0.18, "learning_rate": 8.231466102974483e-05, "loss": 1.026, "step": 71305 }, { "epoch": 0.18, "learning_rate": 8.231340278826313e-05, "loss": 1.0255, "step": 71310 }, { "epoch": 0.18, "learning_rate": 8.231214454678143e-05, "loss": 1.0278, "step": 71315 }, { "epoch": 0.18, "learning_rate": 8.231088630529972e-05, "loss": 1.0262, "step": 71320 }, { "epoch": 0.18, "learning_rate": 8.230962806381801e-05, "loss": 1.0241, "step": 71325 }, { "epoch": 0.18, "learning_rate": 8.23083698223363e-05, "loss": 1.0246, "step": 71330 }, { "epoch": 0.18, "learning_rate": 8.23071115808546e-05, "loss": 1.0246, "step": 71335 }, { "epoch": 0.18, "learning_rate": 8.23058533393729e-05, "loss": 1.0248, "step": 71340 }, { "epoch": 0.18, "learning_rate": 8.230459509789119e-05, "loss": 1.0258, "step": 71345 }, { "epoch": 0.18, "learning_rate": 8.230333685640948e-05, "loss": 1.0249, "step": 71350 }, { "epoch": 0.18, "learning_rate": 8.230207861492779e-05, "loss": 1.026, "step": 71355 }, { "epoch": 0.18, "learning_rate": 8.230082037344608e-05, "loss": 1.0248, "step": 71360 }, { "epoch": 0.18, "learning_rate": 8.229956213196437e-05, "loss": 1.0265, "step": 71365 }, { "epoch": 0.18, "learning_rate": 8.229830389048266e-05, "loss": 1.0285, "step": 71370 }, { "epoch": 0.18, "learning_rate": 8.229704564900097e-05, "loss": 1.0488, "step": 71375 }, { "epoch": 0.18, "learning_rate": 8.229578740751926e-05, "loss": 1.0241, "step": 71380 }, { "epoch": 0.18, "learning_rate": 8.229452916603755e-05, "loss": 1.0255, "step": 71385 }, { "epoch": 0.18, "learning_rate": 8.229327092455584e-05, "loss": 1.0259, "step": 71390 }, { "epoch": 0.18, "learning_rate": 8.229201268307413e-05, "loss": 1.0243, "step": 71395 }, { "epoch": 0.18, "learning_rate": 8.229075444159244e-05, "loss": 1.0283, "step": 71400 }, { "epoch": 0.18, "learning_rate": 8.228949620011073e-05, "loss": 1.0253, "step": 71405 }, { "epoch": 0.18, "learning_rate": 8.228823795862902e-05, "loss": 1.0261, "step": 71410 }, { "epoch": 0.18, "learning_rate": 8.228697971714731e-05, "loss": 1.0248, "step": 71415 }, { "epoch": 0.18, "learning_rate": 8.228572147566562e-05, "loss": 1.0268, "step": 71420 }, { "epoch": 0.18, "learning_rate": 8.228446323418391e-05, "loss": 1.0264, "step": 71425 }, { "epoch": 0.18, "learning_rate": 8.22832049927022e-05, "loss": 1.0237, "step": 71430 }, { "epoch": 0.18, "learning_rate": 8.228194675122049e-05, "loss": 1.0249, "step": 71435 }, { "epoch": 0.18, "learning_rate": 8.22806885097388e-05, "loss": 1.0233, "step": 71440 }, { "epoch": 0.18, "learning_rate": 8.227943026825709e-05, "loss": 1.0254, "step": 71445 }, { "epoch": 0.18, "learning_rate": 8.227817202677538e-05, "loss": 1.026, "step": 71450 }, { "epoch": 0.18, "learning_rate": 8.227691378529367e-05, "loss": 1.0258, "step": 71455 }, { "epoch": 0.18, "learning_rate": 8.227565554381196e-05, "loss": 1.0265, "step": 71460 }, { "epoch": 0.18, "learning_rate": 8.227439730233027e-05, "loss": 1.0256, "step": 71465 }, { "epoch": 0.18, "learning_rate": 8.227313906084856e-05, "loss": 1.0255, "step": 71470 }, { "epoch": 0.18, "learning_rate": 8.227188081936685e-05, "loss": 1.0431, "step": 71475 }, { "epoch": 0.18, "learning_rate": 8.227062257788514e-05, "loss": 1.0248, "step": 71480 }, { "epoch": 0.18, "learning_rate": 8.226936433640345e-05, "loss": 1.0226, "step": 71485 }, { "epoch": 0.18, "learning_rate": 8.226810609492174e-05, "loss": 1.0251, "step": 71490 }, { "epoch": 0.18, "learning_rate": 8.226684785344003e-05, "loss": 1.0241, "step": 71495 }, { "epoch": 0.18, "learning_rate": 8.226558961195832e-05, "loss": 1.0259, "step": 71500 }, { "epoch": 0.18, "learning_rate": 8.226433137047663e-05, "loss": 1.0266, "step": 71505 }, { "epoch": 0.18, "learning_rate": 8.226307312899492e-05, "loss": 1.0242, "step": 71510 }, { "epoch": 0.18, "learning_rate": 8.226181488751321e-05, "loss": 1.0422, "step": 71515 }, { "epoch": 0.18, "learning_rate": 8.22605566460315e-05, "loss": 1.0262, "step": 71520 }, { "epoch": 0.18, "learning_rate": 8.22592984045498e-05, "loss": 1.022, "step": 71525 }, { "epoch": 0.18, "learning_rate": 8.22580401630681e-05, "loss": 1.0233, "step": 71530 }, { "epoch": 0.18, "learning_rate": 8.225678192158639e-05, "loss": 1.0239, "step": 71535 }, { "epoch": 0.18, "learning_rate": 8.225552368010468e-05, "loss": 1.0273, "step": 71540 }, { "epoch": 0.18, "learning_rate": 8.225426543862297e-05, "loss": 1.0239, "step": 71545 }, { "epoch": 0.18, "learning_rate": 8.225300719714128e-05, "loss": 1.0262, "step": 71550 }, { "epoch": 0.18, "learning_rate": 8.225174895565957e-05, "loss": 1.0284, "step": 71555 }, { "epoch": 0.18, "learning_rate": 8.225049071417786e-05, "loss": 1.0257, "step": 71560 }, { "epoch": 0.18, "learning_rate": 8.224923247269615e-05, "loss": 1.0229, "step": 71565 }, { "epoch": 0.18, "learning_rate": 8.224797423121446e-05, "loss": 1.0255, "step": 71570 }, { "epoch": 0.18, "learning_rate": 8.224671598973276e-05, "loss": 1.0266, "step": 71575 }, { "epoch": 0.18, "learning_rate": 8.224545774825106e-05, "loss": 1.0258, "step": 71580 }, { "epoch": 0.18, "learning_rate": 8.224419950676935e-05, "loss": 1.0246, "step": 71585 }, { "epoch": 0.18, "learning_rate": 8.224294126528764e-05, "loss": 1.0281, "step": 71590 }, { "epoch": 0.18, "learning_rate": 8.224168302380593e-05, "loss": 1.0257, "step": 71595 }, { "epoch": 0.18, "learning_rate": 8.224042478232424e-05, "loss": 1.0251, "step": 71600 }, { "epoch": 0.18, "learning_rate": 8.223916654084253e-05, "loss": 1.0265, "step": 71605 }, { "epoch": 0.18, "learning_rate": 8.223790829936082e-05, "loss": 1.027, "step": 71610 }, { "epoch": 0.18, "learning_rate": 8.223665005787911e-05, "loss": 1.0262, "step": 71615 }, { "epoch": 0.18, "learning_rate": 8.223539181639742e-05, "loss": 1.0258, "step": 71620 }, { "epoch": 0.18, "learning_rate": 8.223413357491571e-05, "loss": 1.0264, "step": 71625 }, { "epoch": 0.18, "learning_rate": 8.2232875333434e-05, "loss": 1.0255, "step": 71630 }, { "epoch": 0.18, "learning_rate": 8.223161709195229e-05, "loss": 1.0227, "step": 71635 }, { "epoch": 0.18, "learning_rate": 8.22303588504706e-05, "loss": 1.0273, "step": 71640 }, { "epoch": 0.18, "learning_rate": 8.222910060898889e-05, "loss": 1.0274, "step": 71645 }, { "epoch": 0.18, "learning_rate": 8.222784236750718e-05, "loss": 1.0268, "step": 71650 }, { "epoch": 0.18, "learning_rate": 8.222658412602547e-05, "loss": 1.0257, "step": 71655 }, { "epoch": 0.18, "learning_rate": 8.222532588454376e-05, "loss": 1.0247, "step": 71660 }, { "epoch": 0.18, "learning_rate": 8.222406764306207e-05, "loss": 1.0243, "step": 71665 }, { "epoch": 0.18, "learning_rate": 8.222280940158036e-05, "loss": 1.0261, "step": 71670 }, { "epoch": 0.18, "learning_rate": 8.222155116009865e-05, "loss": 1.0235, "step": 71675 }, { "epoch": 0.18, "learning_rate": 8.222029291861694e-05, "loss": 1.0479, "step": 71680 }, { "epoch": 0.18, "learning_rate": 8.221903467713525e-05, "loss": 1.0273, "step": 71685 }, { "epoch": 0.18, "learning_rate": 8.221777643565354e-05, "loss": 1.0226, "step": 71690 }, { "epoch": 0.18, "learning_rate": 8.221651819417183e-05, "loss": 1.0257, "step": 71695 }, { "epoch": 0.18, "learning_rate": 8.221525995269012e-05, "loss": 1.0239, "step": 71700 }, { "epoch": 0.18, "learning_rate": 8.221400171120843e-05, "loss": 1.0243, "step": 71705 }, { "epoch": 0.18, "learning_rate": 8.221274346972672e-05, "loss": 1.0247, "step": 71710 }, { "epoch": 0.18, "learning_rate": 8.221148522824501e-05, "loss": 1.0219, "step": 71715 }, { "epoch": 0.18, "learning_rate": 8.22102269867633e-05, "loss": 1.0245, "step": 71720 }, { "epoch": 0.18, "learning_rate": 8.22089687452816e-05, "loss": 1.0267, "step": 71725 }, { "epoch": 0.18, "learning_rate": 8.22077105037999e-05, "loss": 1.0271, "step": 71730 }, { "epoch": 0.18, "learning_rate": 8.220645226231819e-05, "loss": 1.0255, "step": 71735 }, { "epoch": 0.18, "learning_rate": 8.220519402083648e-05, "loss": 1.0234, "step": 71740 }, { "epoch": 0.18, "learning_rate": 8.220393577935477e-05, "loss": 1.0237, "step": 71745 }, { "epoch": 0.18, "learning_rate": 8.220267753787308e-05, "loss": 1.0274, "step": 71750 }, { "epoch": 0.18, "learning_rate": 8.220141929639137e-05, "loss": 1.0242, "step": 71755 }, { "epoch": 0.18, "learning_rate": 8.220016105490966e-05, "loss": 1.0247, "step": 71760 }, { "epoch": 0.18, "learning_rate": 8.219890281342795e-05, "loss": 1.0247, "step": 71765 }, { "epoch": 0.18, "learning_rate": 8.219764457194626e-05, "loss": 1.025, "step": 71770 }, { "epoch": 0.18, "learning_rate": 8.219638633046455e-05, "loss": 1.0272, "step": 71775 }, { "epoch": 0.18, "learning_rate": 8.219512808898284e-05, "loss": 1.0229, "step": 71780 }, { "epoch": 0.18, "learning_rate": 8.219386984750113e-05, "loss": 1.0281, "step": 71785 }, { "epoch": 0.18, "learning_rate": 8.219261160601943e-05, "loss": 1.0279, "step": 71790 }, { "epoch": 0.18, "learning_rate": 8.219135336453773e-05, "loss": 1.0232, "step": 71795 }, { "epoch": 0.18, "learning_rate": 8.219009512305602e-05, "loss": 1.0246, "step": 71800 }, { "epoch": 0.18, "learning_rate": 8.218883688157431e-05, "loss": 1.0264, "step": 71805 }, { "epoch": 0.18, "learning_rate": 8.21875786400926e-05, "loss": 1.0234, "step": 71810 }, { "epoch": 0.18, "learning_rate": 8.218632039861091e-05, "loss": 1.0262, "step": 71815 }, { "epoch": 0.18, "learning_rate": 8.21850621571292e-05, "loss": 1.0238, "step": 71820 }, { "epoch": 0.18, "learning_rate": 8.21838039156475e-05, "loss": 1.0283, "step": 71825 }, { "epoch": 0.18, "learning_rate": 8.218254567416579e-05, "loss": 1.024, "step": 71830 }, { "epoch": 0.18, "learning_rate": 8.218128743268409e-05, "loss": 1.0258, "step": 71835 }, { "epoch": 0.18, "learning_rate": 8.218002919120238e-05, "loss": 1.0258, "step": 71840 }, { "epoch": 0.18, "learning_rate": 8.217877094972067e-05, "loss": 1.0243, "step": 71845 }, { "epoch": 0.18, "learning_rate": 8.217751270823897e-05, "loss": 1.0271, "step": 71850 }, { "epoch": 0.18, "learning_rate": 8.217625446675726e-05, "loss": 1.0243, "step": 71855 }, { "epoch": 0.18, "learning_rate": 8.217499622527556e-05, "loss": 1.0256, "step": 71860 }, { "epoch": 0.18, "learning_rate": 8.217373798379385e-05, "loss": 1.0247, "step": 71865 }, { "epoch": 0.18, "learning_rate": 8.217247974231215e-05, "loss": 1.0258, "step": 71870 }, { "epoch": 0.18, "learning_rate": 8.217122150083044e-05, "loss": 1.0244, "step": 71875 }, { "epoch": 0.18, "learning_rate": 8.216996325934874e-05, "loss": 1.0244, "step": 71880 }, { "epoch": 0.18, "learning_rate": 8.216870501786703e-05, "loss": 1.0251, "step": 71885 }, { "epoch": 0.18, "learning_rate": 8.216744677638533e-05, "loss": 1.0253, "step": 71890 }, { "epoch": 0.18, "learning_rate": 8.216618853490362e-05, "loss": 1.0255, "step": 71895 }, { "epoch": 0.18, "learning_rate": 8.216493029342192e-05, "loss": 1.023, "step": 71900 }, { "epoch": 0.18, "learning_rate": 8.216367205194021e-05, "loss": 1.0255, "step": 71905 }, { "epoch": 0.18, "learning_rate": 8.21624138104585e-05, "loss": 1.0264, "step": 71910 }, { "epoch": 0.18, "learning_rate": 8.21611555689768e-05, "loss": 1.0267, "step": 71915 }, { "epoch": 0.18, "learning_rate": 8.215989732749509e-05, "loss": 1.0258, "step": 71920 }, { "epoch": 0.18, "learning_rate": 8.21586390860134e-05, "loss": 1.0267, "step": 71925 }, { "epoch": 0.18, "learning_rate": 8.215738084453169e-05, "loss": 1.0265, "step": 71930 }, { "epoch": 0.18, "learning_rate": 8.215612260304998e-05, "loss": 1.0225, "step": 71935 }, { "epoch": 0.18, "learning_rate": 8.215486436156827e-05, "loss": 1.0243, "step": 71940 }, { "epoch": 0.18, "learning_rate": 8.215360612008657e-05, "loss": 1.0267, "step": 71945 }, { "epoch": 0.18, "learning_rate": 8.215234787860487e-05, "loss": 1.0454, "step": 71950 }, { "epoch": 0.18, "learning_rate": 8.215108963712316e-05, "loss": 1.0263, "step": 71955 }, { "epoch": 0.18, "learning_rate": 8.214983139564145e-05, "loss": 1.0254, "step": 71960 }, { "epoch": 0.18, "learning_rate": 8.214857315415975e-05, "loss": 1.0235, "step": 71965 }, { "epoch": 0.18, "learning_rate": 8.214731491267805e-05, "loss": 1.0264, "step": 71970 }, { "epoch": 0.18, "learning_rate": 8.214605667119634e-05, "loss": 1.0267, "step": 71975 }, { "epoch": 0.18, "learning_rate": 8.214479842971463e-05, "loss": 1.0254, "step": 71980 }, { "epoch": 0.18, "learning_rate": 8.214354018823292e-05, "loss": 1.0236, "step": 71985 }, { "epoch": 0.18, "learning_rate": 8.214228194675122e-05, "loss": 1.0259, "step": 71990 }, { "epoch": 0.18, "learning_rate": 8.214102370526952e-05, "loss": 1.025, "step": 71995 }, { "epoch": 0.18, "learning_rate": 8.213976546378781e-05, "loss": 1.0272, "step": 72000 }, { "epoch": 0.18, "learning_rate": 8.21385072223061e-05, "loss": 1.0276, "step": 72005 }, { "epoch": 0.18, "learning_rate": 8.21372489808244e-05, "loss": 1.0239, "step": 72010 }, { "epoch": 0.18, "learning_rate": 8.21359907393427e-05, "loss": 1.0248, "step": 72015 }, { "epoch": 0.18, "learning_rate": 8.213473249786099e-05, "loss": 1.0265, "step": 72020 }, { "epoch": 0.18, "learning_rate": 8.213347425637928e-05, "loss": 1.0257, "step": 72025 }, { "epoch": 0.18, "learning_rate": 8.213221601489758e-05, "loss": 1.0255, "step": 72030 }, { "epoch": 0.18, "learning_rate": 8.213095777341588e-05, "loss": 1.0236, "step": 72035 }, { "epoch": 0.18, "learning_rate": 8.212969953193417e-05, "loss": 1.0265, "step": 72040 }, { "epoch": 0.18, "learning_rate": 8.212844129045246e-05, "loss": 1.0267, "step": 72045 }, { "epoch": 0.18, "learning_rate": 8.212718304897075e-05, "loss": 1.0224, "step": 72050 }, { "epoch": 0.18, "learning_rate": 8.212592480748906e-05, "loss": 1.0278, "step": 72055 }, { "epoch": 0.18, "learning_rate": 8.212466656600735e-05, "loss": 1.0509, "step": 72060 }, { "epoch": 0.18, "learning_rate": 8.212340832452564e-05, "loss": 1.0277, "step": 72065 }, { "epoch": 0.18, "learning_rate": 8.212215008304394e-05, "loss": 1.0258, "step": 72070 }, { "epoch": 0.18, "learning_rate": 8.212089184156224e-05, "loss": 1.0266, "step": 72075 }, { "epoch": 0.18, "learning_rate": 8.211963360008054e-05, "loss": 1.0231, "step": 72080 }, { "epoch": 0.18, "learning_rate": 8.211837535859883e-05, "loss": 1.0247, "step": 72085 }, { "epoch": 0.18, "learning_rate": 8.211711711711712e-05, "loss": 1.0249, "step": 72090 }, { "epoch": 0.18, "learning_rate": 8.211585887563542e-05, "loss": 1.0258, "step": 72095 }, { "epoch": 0.18, "learning_rate": 8.211460063415372e-05, "loss": 1.0271, "step": 72100 }, { "epoch": 0.18, "learning_rate": 8.211334239267201e-05, "loss": 1.0254, "step": 72105 }, { "epoch": 0.18, "learning_rate": 8.21120841511903e-05, "loss": 1.0287, "step": 72110 }, { "epoch": 0.18, "learning_rate": 8.21108259097086e-05, "loss": 1.0261, "step": 72115 }, { "epoch": 0.18, "learning_rate": 8.210956766822689e-05, "loss": 1.0229, "step": 72120 }, { "epoch": 0.18, "learning_rate": 8.210830942674519e-05, "loss": 1.0243, "step": 72125 }, { "epoch": 0.18, "learning_rate": 8.210705118526348e-05, "loss": 1.0241, "step": 72130 }, { "epoch": 0.18, "learning_rate": 8.210579294378178e-05, "loss": 1.0253, "step": 72135 }, { "epoch": 0.18, "learning_rate": 8.210453470230007e-05, "loss": 1.0269, "step": 72140 }, { "epoch": 0.18, "learning_rate": 8.210327646081837e-05, "loss": 1.0251, "step": 72145 }, { "epoch": 0.18, "learning_rate": 8.210201821933666e-05, "loss": 1.0262, "step": 72150 }, { "epoch": 0.18, "learning_rate": 8.210075997785496e-05, "loss": 1.0263, "step": 72155 }, { "epoch": 0.18, "learning_rate": 8.209950173637325e-05, "loss": 1.0247, "step": 72160 }, { "epoch": 0.18, "learning_rate": 8.209824349489155e-05, "loss": 1.0274, "step": 72165 }, { "epoch": 0.18, "learning_rate": 8.209698525340984e-05, "loss": 1.0249, "step": 72170 }, { "epoch": 0.18, "learning_rate": 8.209572701192814e-05, "loss": 1.0273, "step": 72175 }, { "epoch": 0.18, "learning_rate": 8.209446877044643e-05, "loss": 1.0253, "step": 72180 }, { "epoch": 0.18, "learning_rate": 8.209321052896472e-05, "loss": 1.0257, "step": 72185 }, { "epoch": 0.18, "learning_rate": 8.209195228748302e-05, "loss": 1.0232, "step": 72190 }, { "epoch": 0.18, "learning_rate": 8.209069404600132e-05, "loss": 1.0251, "step": 72195 }, { "epoch": 0.18, "learning_rate": 8.208943580451961e-05, "loss": 1.0235, "step": 72200 }, { "epoch": 0.18, "learning_rate": 8.20881775630379e-05, "loss": 1.0272, "step": 72205 }, { "epoch": 0.18, "learning_rate": 8.20869193215562e-05, "loss": 1.0306, "step": 72210 }, { "epoch": 0.18, "learning_rate": 8.20856610800745e-05, "loss": 1.0253, "step": 72215 }, { "epoch": 0.18, "learning_rate": 8.208440283859279e-05, "loss": 1.0234, "step": 72220 }, { "epoch": 0.18, "learning_rate": 8.208314459711108e-05, "loss": 1.0267, "step": 72225 }, { "epoch": 0.18, "learning_rate": 8.208188635562938e-05, "loss": 1.0228, "step": 72230 }, { "epoch": 0.18, "learning_rate": 8.208062811414768e-05, "loss": 1.0254, "step": 72235 }, { "epoch": 0.18, "learning_rate": 8.207936987266597e-05, "loss": 1.0264, "step": 72240 }, { "epoch": 0.18, "learning_rate": 8.207811163118426e-05, "loss": 1.0234, "step": 72245 }, { "epoch": 0.18, "learning_rate": 8.207685338970255e-05, "loss": 1.0249, "step": 72250 }, { "epoch": 0.18, "learning_rate": 8.207559514822086e-05, "loss": 1.0269, "step": 72255 }, { "epoch": 0.18, "learning_rate": 8.207433690673915e-05, "loss": 1.0286, "step": 72260 }, { "epoch": 0.18, "learning_rate": 8.207307866525744e-05, "loss": 1.0261, "step": 72265 }, { "epoch": 0.18, "learning_rate": 8.207182042377573e-05, "loss": 1.0238, "step": 72270 }, { "epoch": 0.18, "learning_rate": 8.207056218229404e-05, "loss": 1.0246, "step": 72275 }, { "epoch": 0.18, "learning_rate": 8.206930394081233e-05, "loss": 1.025, "step": 72280 }, { "epoch": 0.18, "learning_rate": 8.206804569933062e-05, "loss": 1.0259, "step": 72285 }, { "epoch": 0.18, "learning_rate": 8.206678745784891e-05, "loss": 1.0258, "step": 72290 }, { "epoch": 0.18, "learning_rate": 8.206552921636722e-05, "loss": 1.0261, "step": 72295 }, { "epoch": 0.18, "learning_rate": 8.206427097488551e-05, "loss": 1.0237, "step": 72300 }, { "epoch": 0.18, "learning_rate": 8.20630127334038e-05, "loss": 1.0256, "step": 72305 }, { "epoch": 0.18, "learning_rate": 8.206175449192209e-05, "loss": 1.024, "step": 72310 }, { "epoch": 0.18, "learning_rate": 8.206049625044038e-05, "loss": 1.0258, "step": 72315 }, { "epoch": 0.18, "learning_rate": 8.205923800895869e-05, "loss": 1.025, "step": 72320 }, { "epoch": 0.18, "learning_rate": 8.205797976747698e-05, "loss": 1.0247, "step": 72325 }, { "epoch": 0.18, "learning_rate": 8.205672152599527e-05, "loss": 1.0263, "step": 72330 }, { "epoch": 0.18, "learning_rate": 8.205546328451356e-05, "loss": 1.0256, "step": 72335 }, { "epoch": 0.18, "learning_rate": 8.205420504303187e-05, "loss": 1.026, "step": 72340 }, { "epoch": 0.18, "learning_rate": 8.205294680155016e-05, "loss": 1.0238, "step": 72345 }, { "epoch": 0.18, "learning_rate": 8.205168856006845e-05, "loss": 1.0268, "step": 72350 }, { "epoch": 0.18, "learning_rate": 8.205043031858674e-05, "loss": 1.0253, "step": 72355 }, { "epoch": 0.18, "learning_rate": 8.204917207710505e-05, "loss": 1.024, "step": 72360 }, { "epoch": 0.18, "learning_rate": 8.204791383562334e-05, "loss": 1.0238, "step": 72365 }, { "epoch": 0.18, "learning_rate": 8.204665559414163e-05, "loss": 1.0272, "step": 72370 }, { "epoch": 0.18, "learning_rate": 8.204539735265992e-05, "loss": 1.0281, "step": 72375 }, { "epoch": 0.18, "learning_rate": 8.204413911117821e-05, "loss": 1.0243, "step": 72380 }, { "epoch": 0.18, "learning_rate": 8.204288086969652e-05, "loss": 1.0251, "step": 72385 }, { "epoch": 0.18, "learning_rate": 8.204162262821481e-05, "loss": 1.0249, "step": 72390 }, { "epoch": 0.18, "learning_rate": 8.20403643867331e-05, "loss": 1.0239, "step": 72395 }, { "epoch": 0.18, "learning_rate": 8.20391061452514e-05, "loss": 1.024, "step": 72400 }, { "epoch": 0.18, "learning_rate": 8.20378479037697e-05, "loss": 1.025, "step": 72405 }, { "epoch": 0.18, "learning_rate": 8.203658966228799e-05, "loss": 1.0253, "step": 72410 }, { "epoch": 0.18, "learning_rate": 8.203533142080628e-05, "loss": 1.0238, "step": 72415 }, { "epoch": 0.18, "learning_rate": 8.203407317932457e-05, "loss": 1.0231, "step": 72420 }, { "epoch": 0.18, "learning_rate": 8.203281493784288e-05, "loss": 1.0242, "step": 72425 }, { "epoch": 0.18, "learning_rate": 8.203155669636117e-05, "loss": 1.028, "step": 72430 }, { "epoch": 0.18, "learning_rate": 8.203029845487946e-05, "loss": 1.025, "step": 72435 }, { "epoch": 0.18, "learning_rate": 8.202904021339775e-05, "loss": 1.0242, "step": 72440 }, { "epoch": 0.18, "learning_rate": 8.202778197191604e-05, "loss": 1.0252, "step": 72445 }, { "epoch": 0.18, "learning_rate": 8.202652373043435e-05, "loss": 1.0265, "step": 72450 }, { "epoch": 0.18, "learning_rate": 8.202526548895264e-05, "loss": 1.0259, "step": 72455 }, { "epoch": 0.18, "learning_rate": 8.202400724747093e-05, "loss": 1.0249, "step": 72460 }, { "epoch": 0.18, "learning_rate": 8.202274900598922e-05, "loss": 1.0271, "step": 72465 }, { "epoch": 0.18, "learning_rate": 8.202149076450753e-05, "loss": 1.0274, "step": 72470 }, { "epoch": 0.18, "learning_rate": 8.202023252302582e-05, "loss": 1.0239, "step": 72475 }, { "epoch": 0.18, "learning_rate": 8.201897428154411e-05, "loss": 1.0232, "step": 72480 }, { "epoch": 0.18, "learning_rate": 8.20177160400624e-05, "loss": 1.0251, "step": 72485 }, { "epoch": 0.18, "learning_rate": 8.201645779858071e-05, "loss": 1.0257, "step": 72490 }, { "epoch": 0.18, "learning_rate": 8.2015199557099e-05, "loss": 1.0261, "step": 72495 }, { "epoch": 0.18, "learning_rate": 8.201394131561729e-05, "loss": 1.0258, "step": 72500 }, { "epoch": 0.18, "learning_rate": 8.201268307413558e-05, "loss": 1.0281, "step": 72505 }, { "epoch": 0.18, "learning_rate": 8.201142483265388e-05, "loss": 1.0242, "step": 72510 }, { "epoch": 0.18, "learning_rate": 8.201016659117218e-05, "loss": 1.0276, "step": 72515 }, { "epoch": 0.18, "learning_rate": 8.200890834969047e-05, "loss": 1.0242, "step": 72520 }, { "epoch": 0.18, "learning_rate": 8.200765010820876e-05, "loss": 1.024, "step": 72525 }, { "epoch": 0.18, "learning_rate": 8.200639186672706e-05, "loss": 1.0266, "step": 72530 }, { "epoch": 0.18, "learning_rate": 8.200513362524536e-05, "loss": 1.0257, "step": 72535 }, { "epoch": 0.18, "learning_rate": 8.200387538376365e-05, "loss": 1.0462, "step": 72540 }, { "epoch": 0.18, "learning_rate": 8.200261714228194e-05, "loss": 1.0256, "step": 72545 }, { "epoch": 0.18, "learning_rate": 8.200135890080024e-05, "loss": 1.0268, "step": 72550 }, { "epoch": 0.18, "learning_rate": 8.200010065931854e-05, "loss": 1.0261, "step": 72555 }, { "epoch": 0.18, "learning_rate": 8.199884241783683e-05, "loss": 1.027, "step": 72560 }, { "epoch": 0.18, "learning_rate": 8.199758417635512e-05, "loss": 1.0247, "step": 72565 }, { "epoch": 0.18, "learning_rate": 8.199632593487343e-05, "loss": 1.0246, "step": 72570 }, { "epoch": 0.18, "learning_rate": 8.199506769339172e-05, "loss": 1.0256, "step": 72575 }, { "epoch": 0.18, "learning_rate": 8.199380945191001e-05, "loss": 1.0269, "step": 72580 }, { "epoch": 0.18, "learning_rate": 8.199255121042832e-05, "loss": 1.0256, "step": 72585 }, { "epoch": 0.18, "learning_rate": 8.199129296894661e-05, "loss": 1.0264, "step": 72590 }, { "epoch": 0.18, "learning_rate": 8.19900347274649e-05, "loss": 1.0242, "step": 72595 }, { "epoch": 0.18, "learning_rate": 8.198877648598319e-05, "loss": 1.0261, "step": 72600 }, { "epoch": 0.18, "learning_rate": 8.19875182445015e-05, "loss": 1.0237, "step": 72605 }, { "epoch": 0.18, "learning_rate": 8.198626000301979e-05, "loss": 1.0244, "step": 72610 }, { "epoch": 0.18, "learning_rate": 8.198500176153808e-05, "loss": 1.0267, "step": 72615 }, { "epoch": 0.18, "learning_rate": 8.198374352005637e-05, "loss": 1.0244, "step": 72620 }, { "epoch": 0.18, "learning_rate": 8.198248527857468e-05, "loss": 1.0267, "step": 72625 }, { "epoch": 0.18, "learning_rate": 8.198122703709297e-05, "loss": 1.0262, "step": 72630 }, { "epoch": 0.18, "learning_rate": 8.197996879561126e-05, "loss": 1.0261, "step": 72635 }, { "epoch": 0.18, "learning_rate": 8.197871055412955e-05, "loss": 1.0252, "step": 72640 }, { "epoch": 0.18, "learning_rate": 8.197745231264784e-05, "loss": 1.0244, "step": 72645 }, { "epoch": 0.18, "learning_rate": 8.197619407116615e-05, "loss": 1.0252, "step": 72650 }, { "epoch": 0.18, "learning_rate": 8.197493582968444e-05, "loss": 1.0301, "step": 72655 }, { "epoch": 0.18, "learning_rate": 8.197367758820273e-05, "loss": 1.0259, "step": 72660 }, { "epoch": 0.18, "learning_rate": 8.197241934672102e-05, "loss": 1.0252, "step": 72665 }, { "epoch": 0.18, "learning_rate": 8.197116110523933e-05, "loss": 1.0237, "step": 72670 }, { "epoch": 0.18, "learning_rate": 8.196990286375762e-05, "loss": 1.0223, "step": 72675 }, { "epoch": 0.18, "learning_rate": 8.196864462227591e-05, "loss": 1.0254, "step": 72680 }, { "epoch": 0.18, "learning_rate": 8.19673863807942e-05, "loss": 1.0273, "step": 72685 }, { "epoch": 0.18, "learning_rate": 8.196612813931251e-05, "loss": 1.0268, "step": 72690 }, { "epoch": 0.18, "learning_rate": 8.19648698978308e-05, "loss": 1.0247, "step": 72695 }, { "epoch": 0.18, "learning_rate": 8.196361165634909e-05, "loss": 1.0263, "step": 72700 }, { "epoch": 0.18, "learning_rate": 8.196235341486738e-05, "loss": 1.0257, "step": 72705 }, { "epoch": 0.18, "learning_rate": 8.196109517338568e-05, "loss": 1.026, "step": 72710 }, { "epoch": 0.18, "learning_rate": 8.195983693190398e-05, "loss": 1.0253, "step": 72715 }, { "epoch": 0.18, "learning_rate": 8.195857869042227e-05, "loss": 1.027, "step": 72720 }, { "epoch": 0.18, "learning_rate": 8.195732044894056e-05, "loss": 1.0276, "step": 72725 }, { "epoch": 0.18, "learning_rate": 8.195606220745886e-05, "loss": 1.0242, "step": 72730 }, { "epoch": 0.18, "learning_rate": 8.195480396597716e-05, "loss": 1.0236, "step": 72735 }, { "epoch": 0.18, "learning_rate": 8.195354572449545e-05, "loss": 1.0233, "step": 72740 }, { "epoch": 0.18, "learning_rate": 8.195228748301374e-05, "loss": 1.0255, "step": 72745 }, { "epoch": 0.18, "learning_rate": 8.195102924153204e-05, "loss": 1.0253, "step": 72750 }, { "epoch": 0.18, "learning_rate": 8.194977100005034e-05, "loss": 1.0274, "step": 72755 }, { "epoch": 0.18, "learning_rate": 8.194851275856863e-05, "loss": 1.0245, "step": 72760 }, { "epoch": 0.18, "learning_rate": 8.194725451708692e-05, "loss": 1.0255, "step": 72765 }, { "epoch": 0.18, "learning_rate": 8.194599627560522e-05, "loss": 1.0232, "step": 72770 }, { "epoch": 0.18, "learning_rate": 8.194473803412351e-05, "loss": 1.0264, "step": 72775 }, { "epoch": 0.18, "learning_rate": 8.194347979264181e-05, "loss": 1.0257, "step": 72780 }, { "epoch": 0.18, "learning_rate": 8.19422215511601e-05, "loss": 1.0247, "step": 72785 }, { "epoch": 0.18, "learning_rate": 8.19409633096784e-05, "loss": 1.0267, "step": 72790 }, { "epoch": 0.18, "learning_rate": 8.193970506819669e-05, "loss": 1.0238, "step": 72795 }, { "epoch": 0.18, "learning_rate": 8.193844682671499e-05, "loss": 1.0261, "step": 72800 }, { "epoch": 0.18, "learning_rate": 8.193718858523328e-05, "loss": 1.0237, "step": 72805 }, { "epoch": 0.18, "learning_rate": 8.193593034375158e-05, "loss": 1.0244, "step": 72810 }, { "epoch": 0.18, "learning_rate": 8.193467210226987e-05, "loss": 1.0265, "step": 72815 }, { "epoch": 0.18, "learning_rate": 8.193341386078817e-05, "loss": 1.025, "step": 72820 }, { "epoch": 0.18, "learning_rate": 8.193215561930646e-05, "loss": 1.0258, "step": 72825 }, { "epoch": 0.18, "learning_rate": 8.193089737782476e-05, "loss": 1.0255, "step": 72830 }, { "epoch": 0.18, "learning_rate": 8.192963913634305e-05, "loss": 1.0259, "step": 72835 }, { "epoch": 0.18, "learning_rate": 8.192838089486134e-05, "loss": 1.0254, "step": 72840 }, { "epoch": 0.18, "learning_rate": 8.192712265337964e-05, "loss": 1.0238, "step": 72845 }, { "epoch": 0.18, "learning_rate": 8.192586441189794e-05, "loss": 1.0222, "step": 72850 }, { "epoch": 0.18, "learning_rate": 8.192460617041623e-05, "loss": 1.0267, "step": 72855 }, { "epoch": 0.18, "learning_rate": 8.192334792893452e-05, "loss": 1.026, "step": 72860 }, { "epoch": 0.18, "learning_rate": 8.192208968745282e-05, "loss": 1.0262, "step": 72865 }, { "epoch": 0.18, "learning_rate": 8.192083144597112e-05, "loss": 1.0266, "step": 72870 }, { "epoch": 0.18, "learning_rate": 8.191957320448941e-05, "loss": 1.023, "step": 72875 }, { "epoch": 0.18, "learning_rate": 8.19183149630077e-05, "loss": 1.0275, "step": 72880 }, { "epoch": 0.18, "learning_rate": 8.1917056721526e-05, "loss": 1.0268, "step": 72885 }, { "epoch": 0.18, "learning_rate": 8.19157984800443e-05, "loss": 1.0264, "step": 72890 }, { "epoch": 0.18, "learning_rate": 8.191454023856259e-05, "loss": 1.025, "step": 72895 }, { "epoch": 0.18, "learning_rate": 8.191328199708088e-05, "loss": 1.024, "step": 72900 }, { "epoch": 0.18, "learning_rate": 8.191202375559917e-05, "loss": 1.0245, "step": 72905 }, { "epoch": 0.18, "learning_rate": 8.191076551411748e-05, "loss": 1.0235, "step": 72910 }, { "epoch": 0.18, "learning_rate": 8.190950727263577e-05, "loss": 1.0503, "step": 72915 }, { "epoch": 0.18, "learning_rate": 8.190824903115406e-05, "loss": 1.0258, "step": 72920 }, { "epoch": 0.18, "learning_rate": 8.190699078967235e-05, "loss": 1.0251, "step": 72925 }, { "epoch": 0.18, "learning_rate": 8.190573254819066e-05, "loss": 1.0266, "step": 72930 }, { "epoch": 0.18, "learning_rate": 8.190447430670895e-05, "loss": 1.0261, "step": 72935 }, { "epoch": 0.18, "learning_rate": 8.190321606522724e-05, "loss": 1.0258, "step": 72940 }, { "epoch": 0.18, "learning_rate": 8.190195782374553e-05, "loss": 1.0232, "step": 72945 }, { "epoch": 0.18, "learning_rate": 8.190069958226384e-05, "loss": 1.0248, "step": 72950 }, { "epoch": 0.18, "learning_rate": 8.189944134078213e-05, "loss": 1.0262, "step": 72955 }, { "epoch": 0.18, "learning_rate": 8.189818309930042e-05, "loss": 1.027, "step": 72960 }, { "epoch": 0.18, "learning_rate": 8.189692485781871e-05, "loss": 1.0254, "step": 72965 }, { "epoch": 0.18, "learning_rate": 8.1895666616337e-05, "loss": 1.0247, "step": 72970 }, { "epoch": 0.18, "learning_rate": 8.18944083748553e-05, "loss": 1.0234, "step": 72975 }, { "epoch": 0.18, "learning_rate": 8.18931501333736e-05, "loss": 1.0257, "step": 72980 }, { "epoch": 0.18, "learning_rate": 8.189189189189189e-05, "loss": 1.0261, "step": 72985 }, { "epoch": 0.18, "learning_rate": 8.189063365041018e-05, "loss": 1.0241, "step": 72990 }, { "epoch": 0.18, "learning_rate": 8.188937540892849e-05, "loss": 1.0249, "step": 72995 }, { "epoch": 0.18, "learning_rate": 8.188811716744678e-05, "loss": 1.0254, "step": 73000 }, { "epoch": 0.18, "learning_rate": 8.188685892596507e-05, "loss": 1.0247, "step": 73005 }, { "epoch": 0.18, "learning_rate": 8.188560068448336e-05, "loss": 1.0251, "step": 73010 }, { "epoch": 0.18, "learning_rate": 8.188434244300167e-05, "loss": 1.0238, "step": 73015 }, { "epoch": 0.18, "learning_rate": 8.188308420151996e-05, "loss": 1.0237, "step": 73020 }, { "epoch": 0.18, "learning_rate": 8.188182596003825e-05, "loss": 1.0245, "step": 73025 }, { "epoch": 0.18, "learning_rate": 8.188056771855654e-05, "loss": 1.0237, "step": 73030 }, { "epoch": 0.18, "learning_rate": 8.187930947707483e-05, "loss": 1.0251, "step": 73035 }, { "epoch": 0.18, "learning_rate": 8.187805123559314e-05, "loss": 1.0251, "step": 73040 }, { "epoch": 0.18, "learning_rate": 8.187679299411143e-05, "loss": 1.0241, "step": 73045 }, { "epoch": 0.18, "learning_rate": 8.187553475262972e-05, "loss": 1.0274, "step": 73050 }, { "epoch": 0.18, "learning_rate": 8.187427651114801e-05, "loss": 1.0256, "step": 73055 }, { "epoch": 0.18, "learning_rate": 8.187301826966632e-05, "loss": 1.0255, "step": 73060 }, { "epoch": 0.18, "learning_rate": 8.187176002818462e-05, "loss": 1.024, "step": 73065 }, { "epoch": 0.18, "learning_rate": 8.187050178670291e-05, "loss": 1.0257, "step": 73070 }, { "epoch": 0.18, "learning_rate": 8.18692435452212e-05, "loss": 1.0258, "step": 73075 }, { "epoch": 0.18, "learning_rate": 8.18679853037395e-05, "loss": 1.0251, "step": 73080 }, { "epoch": 0.18, "learning_rate": 8.18667270622578e-05, "loss": 1.0248, "step": 73085 }, { "epoch": 0.18, "learning_rate": 8.18654688207761e-05, "loss": 1.0475, "step": 73090 }, { "epoch": 0.18, "learning_rate": 8.186421057929439e-05, "loss": 1.0248, "step": 73095 }, { "epoch": 0.18, "learning_rate": 8.186295233781268e-05, "loss": 1.0266, "step": 73100 }, { "epoch": 0.18, "learning_rate": 8.186169409633098e-05, "loss": 1.0248, "step": 73105 }, { "epoch": 0.18, "learning_rate": 8.186043585484927e-05, "loss": 1.0238, "step": 73110 }, { "epoch": 0.18, "learning_rate": 8.185917761336757e-05, "loss": 1.0248, "step": 73115 }, { "epoch": 0.18, "learning_rate": 8.185791937188586e-05, "loss": 1.0268, "step": 73120 }, { "epoch": 0.18, "learning_rate": 8.185666113040415e-05, "loss": 1.025, "step": 73125 }, { "epoch": 0.18, "learning_rate": 8.185540288892245e-05, "loss": 1.0241, "step": 73130 }, { "epoch": 0.18, "learning_rate": 8.185414464744075e-05, "loss": 1.0245, "step": 73135 }, { "epoch": 0.18, "learning_rate": 8.185288640595904e-05, "loss": 1.0257, "step": 73140 }, { "epoch": 0.18, "learning_rate": 8.185162816447733e-05, "loss": 1.0258, "step": 73145 }, { "epoch": 0.18, "learning_rate": 8.185036992299563e-05, "loss": 1.0269, "step": 73150 }, { "epoch": 0.18, "learning_rate": 8.184911168151393e-05, "loss": 1.0251, "step": 73155 }, { "epoch": 0.18, "learning_rate": 8.184785344003222e-05, "loss": 1.0259, "step": 73160 }, { "epoch": 0.18, "learning_rate": 8.184659519855051e-05, "loss": 1.0272, "step": 73165 }, { "epoch": 0.18, "learning_rate": 8.184533695706881e-05, "loss": 1.0236, "step": 73170 }, { "epoch": 0.18, "learning_rate": 8.18440787155871e-05, "loss": 1.0255, "step": 73175 }, { "epoch": 0.18, "learning_rate": 8.18428204741054e-05, "loss": 1.0264, "step": 73180 }, { "epoch": 0.18, "learning_rate": 8.184156223262369e-05, "loss": 1.0243, "step": 73185 }, { "epoch": 0.18, "learning_rate": 8.184030399114198e-05, "loss": 1.0243, "step": 73190 }, { "epoch": 0.18, "learning_rate": 8.183904574966029e-05, "loss": 1.0279, "step": 73195 }, { "epoch": 0.18, "learning_rate": 8.183778750817858e-05, "loss": 1.0268, "step": 73200 }, { "epoch": 0.18, "learning_rate": 8.183652926669687e-05, "loss": 1.0252, "step": 73205 }, { "epoch": 0.18, "learning_rate": 8.183527102521516e-05, "loss": 1.0246, "step": 73210 }, { "epoch": 0.18, "learning_rate": 8.183401278373347e-05, "loss": 1.0244, "step": 73215 }, { "epoch": 0.18, "learning_rate": 8.183275454225176e-05, "loss": 1.0252, "step": 73220 }, { "epoch": 0.18, "learning_rate": 8.183149630077005e-05, "loss": 1.0268, "step": 73225 }, { "epoch": 0.18, "learning_rate": 8.183023805928834e-05, "loss": 1.0266, "step": 73230 }, { "epoch": 0.18, "learning_rate": 8.182897981780665e-05, "loss": 1.0257, "step": 73235 }, { "epoch": 0.18, "learning_rate": 8.182772157632494e-05, "loss": 1.0265, "step": 73240 }, { "epoch": 0.18, "learning_rate": 8.182646333484323e-05, "loss": 1.0251, "step": 73245 }, { "epoch": 0.18, "learning_rate": 8.182520509336152e-05, "loss": 1.0271, "step": 73250 }, { "epoch": 0.18, "learning_rate": 8.182394685187981e-05, "loss": 1.0261, "step": 73255 }, { "epoch": 0.18, "learning_rate": 8.182268861039812e-05, "loss": 1.0254, "step": 73260 }, { "epoch": 0.18, "learning_rate": 8.182143036891641e-05, "loss": 1.0253, "step": 73265 }, { "epoch": 0.18, "learning_rate": 8.18201721274347e-05, "loss": 1.0246, "step": 73270 }, { "epoch": 0.18, "learning_rate": 8.181891388595299e-05, "loss": 1.026, "step": 73275 }, { "epoch": 0.18, "learning_rate": 8.18176556444713e-05, "loss": 1.0274, "step": 73280 }, { "epoch": 0.18, "learning_rate": 8.181639740298959e-05, "loss": 1.0256, "step": 73285 }, { "epoch": 0.18, "learning_rate": 8.181513916150788e-05, "loss": 1.0257, "step": 73290 }, { "epoch": 0.18, "learning_rate": 8.181388092002617e-05, "loss": 1.0248, "step": 73295 }, { "epoch": 0.18, "learning_rate": 8.181262267854446e-05, "loss": 1.0238, "step": 73300 }, { "epoch": 0.18, "learning_rate": 8.181136443706277e-05, "loss": 1.0254, "step": 73305 }, { "epoch": 0.18, "learning_rate": 8.181010619558106e-05, "loss": 1.0282, "step": 73310 }, { "epoch": 0.18, "learning_rate": 8.180884795409935e-05, "loss": 1.0262, "step": 73315 }, { "epoch": 0.18, "learning_rate": 8.180758971261764e-05, "loss": 1.0264, "step": 73320 }, { "epoch": 0.18, "learning_rate": 8.180633147113595e-05, "loss": 1.0241, "step": 73325 }, { "epoch": 0.18, "learning_rate": 8.180507322965424e-05, "loss": 1.0242, "step": 73330 }, { "epoch": 0.18, "learning_rate": 8.180381498817253e-05, "loss": 1.026, "step": 73335 }, { "epoch": 0.18, "learning_rate": 8.180255674669082e-05, "loss": 1.0277, "step": 73340 }, { "epoch": 0.18, "learning_rate": 8.180129850520913e-05, "loss": 1.0244, "step": 73345 }, { "epoch": 0.18, "learning_rate": 8.180004026372742e-05, "loss": 1.0258, "step": 73350 }, { "epoch": 0.18, "learning_rate": 8.179878202224571e-05, "loss": 1.0412, "step": 73355 }, { "epoch": 0.18, "learning_rate": 8.1797523780764e-05, "loss": 1.0254, "step": 73360 }, { "epoch": 0.18, "learning_rate": 8.17962655392823e-05, "loss": 1.0266, "step": 73365 }, { "epoch": 0.18, "learning_rate": 8.17950072978006e-05, "loss": 1.0253, "step": 73370 }, { "epoch": 0.18, "learning_rate": 8.179374905631889e-05, "loss": 1.0241, "step": 73375 }, { "epoch": 0.18, "learning_rate": 8.179249081483718e-05, "loss": 1.0249, "step": 73380 }, { "epoch": 0.18, "learning_rate": 8.179123257335548e-05, "loss": 1.0261, "step": 73385 }, { "epoch": 0.18, "learning_rate": 8.178997433187378e-05, "loss": 1.0256, "step": 73390 }, { "epoch": 0.18, "learning_rate": 8.178871609039207e-05, "loss": 1.0237, "step": 73395 }, { "epoch": 0.18, "learning_rate": 8.178745784891036e-05, "loss": 1.0277, "step": 73400 }, { "epoch": 0.18, "learning_rate": 8.178619960742866e-05, "loss": 1.0242, "step": 73405 }, { "epoch": 0.18, "learning_rate": 8.178494136594696e-05, "loss": 1.0258, "step": 73410 }, { "epoch": 0.18, "learning_rate": 8.178368312446525e-05, "loss": 1.0261, "step": 73415 }, { "epoch": 0.18, "learning_rate": 8.178242488298354e-05, "loss": 1.0248, "step": 73420 }, { "epoch": 0.18, "learning_rate": 8.178116664150184e-05, "loss": 1.0259, "step": 73425 }, { "epoch": 0.18, "learning_rate": 8.177990840002013e-05, "loss": 1.0261, "step": 73430 }, { "epoch": 0.18, "learning_rate": 8.177865015853843e-05, "loss": 1.0241, "step": 73435 }, { "epoch": 0.18, "learning_rate": 8.177739191705672e-05, "loss": 1.0236, "step": 73440 }, { "epoch": 0.18, "learning_rate": 8.177613367557501e-05, "loss": 1.0293, "step": 73445 }, { "epoch": 0.18, "learning_rate": 8.17748754340933e-05, "loss": 1.0259, "step": 73450 }, { "epoch": 0.18, "learning_rate": 8.177361719261161e-05, "loss": 1.025, "step": 73455 }, { "epoch": 0.18, "learning_rate": 8.17723589511299e-05, "loss": 1.0243, "step": 73460 }, { "epoch": 0.18, "learning_rate": 8.17711007096482e-05, "loss": 1.0257, "step": 73465 }, { "epoch": 0.18, "learning_rate": 8.176984246816649e-05, "loss": 1.0269, "step": 73470 }, { "epoch": 0.18, "learning_rate": 8.176858422668479e-05, "loss": 1.0272, "step": 73475 }, { "epoch": 0.18, "learning_rate": 8.176732598520308e-05, "loss": 1.0244, "step": 73480 }, { "epoch": 0.18, "learning_rate": 8.176606774372137e-05, "loss": 1.0256, "step": 73485 }, { "epoch": 0.18, "learning_rate": 8.176480950223967e-05, "loss": 1.0245, "step": 73490 }, { "epoch": 0.18, "learning_rate": 8.176355126075796e-05, "loss": 1.0257, "step": 73495 }, { "epoch": 0.18, "learning_rate": 8.176229301927626e-05, "loss": 1.0285, "step": 73500 }, { "epoch": 0.18, "learning_rate": 8.176103477779455e-05, "loss": 1.026, "step": 73505 }, { "epoch": 0.18, "learning_rate": 8.175977653631285e-05, "loss": 1.0242, "step": 73510 }, { "epoch": 0.18, "learning_rate": 8.175851829483114e-05, "loss": 1.0264, "step": 73515 }, { "epoch": 0.18, "learning_rate": 8.175726005334944e-05, "loss": 1.0242, "step": 73520 }, { "epoch": 0.18, "learning_rate": 8.175600181186773e-05, "loss": 1.0253, "step": 73525 }, { "epoch": 0.18, "learning_rate": 8.175474357038603e-05, "loss": 1.0245, "step": 73530 }, { "epoch": 0.18, "learning_rate": 8.175348532890432e-05, "loss": 1.0238, "step": 73535 }, { "epoch": 0.18, "learning_rate": 8.175222708742262e-05, "loss": 1.026, "step": 73540 }, { "epoch": 0.18, "learning_rate": 8.175096884594091e-05, "loss": 1.0251, "step": 73545 }, { "epoch": 0.18, "learning_rate": 8.17497106044592e-05, "loss": 1.0256, "step": 73550 }, { "epoch": 0.18, "learning_rate": 8.17484523629775e-05, "loss": 1.0266, "step": 73555 }, { "epoch": 0.18, "learning_rate": 8.174719412149579e-05, "loss": 1.027, "step": 73560 }, { "epoch": 0.18, "learning_rate": 8.174593588001411e-05, "loss": 1.0268, "step": 73565 }, { "epoch": 0.18, "learning_rate": 8.17446776385324e-05, "loss": 1.0247, "step": 73570 }, { "epoch": 0.18, "learning_rate": 8.174341939705069e-05, "loss": 1.026, "step": 73575 }, { "epoch": 0.18, "learning_rate": 8.174216115556898e-05, "loss": 1.0256, "step": 73580 }, { "epoch": 0.18, "learning_rate": 8.174090291408727e-05, "loss": 1.0271, "step": 73585 }, { "epoch": 0.18, "learning_rate": 8.173964467260558e-05, "loss": 1.026, "step": 73590 }, { "epoch": 0.18, "learning_rate": 8.173838643112387e-05, "loss": 1.0273, "step": 73595 }, { "epoch": 0.18, "learning_rate": 8.173712818964216e-05, "loss": 1.0246, "step": 73600 }, { "epoch": 0.18, "learning_rate": 8.173586994816045e-05, "loss": 1.0266, "step": 73605 }, { "epoch": 0.18, "learning_rate": 8.173461170667876e-05, "loss": 1.0272, "step": 73610 }, { "epoch": 0.18, "learning_rate": 8.173335346519705e-05, "loss": 1.0285, "step": 73615 }, { "epoch": 0.18, "learning_rate": 8.173209522371534e-05, "loss": 1.0245, "step": 73620 }, { "epoch": 0.18, "learning_rate": 8.173083698223363e-05, "loss": 1.0252, "step": 73625 }, { "epoch": 0.18, "learning_rate": 8.172957874075194e-05, "loss": 1.0243, "step": 73630 }, { "epoch": 0.18, "learning_rate": 8.172832049927023e-05, "loss": 1.0235, "step": 73635 }, { "epoch": 0.18, "learning_rate": 8.172706225778852e-05, "loss": 1.0251, "step": 73640 }, { "epoch": 0.18, "learning_rate": 8.172580401630681e-05, "loss": 1.0228, "step": 73645 }, { "epoch": 0.18, "learning_rate": 8.17245457748251e-05, "loss": 1.026, "step": 73650 }, { "epoch": 0.18, "learning_rate": 8.172328753334341e-05, "loss": 1.0234, "step": 73655 }, { "epoch": 0.18, "learning_rate": 8.17220292918617e-05, "loss": 1.0263, "step": 73660 }, { "epoch": 0.18, "learning_rate": 8.172077105038e-05, "loss": 1.0256, "step": 73665 }, { "epoch": 0.18, "learning_rate": 8.171951280889829e-05, "loss": 1.0263, "step": 73670 }, { "epoch": 0.18, "learning_rate": 8.171825456741659e-05, "loss": 1.0267, "step": 73675 }, { "epoch": 0.18, "learning_rate": 8.171699632593488e-05, "loss": 1.0231, "step": 73680 }, { "epoch": 0.18, "learning_rate": 8.171573808445317e-05, "loss": 1.024, "step": 73685 }, { "epoch": 0.18, "learning_rate": 8.171447984297147e-05, "loss": 1.0259, "step": 73690 }, { "epoch": 0.18, "learning_rate": 8.171322160148977e-05, "loss": 1.0249, "step": 73695 }, { "epoch": 0.18, "learning_rate": 8.171196336000806e-05, "loss": 1.0252, "step": 73700 }, { "epoch": 0.19, "learning_rate": 8.171070511852635e-05, "loss": 1.0232, "step": 73705 }, { "epoch": 0.19, "learning_rate": 8.170944687704465e-05, "loss": 1.0242, "step": 73710 }, { "epoch": 0.19, "learning_rate": 8.170818863556294e-05, "loss": 1.0256, "step": 73715 }, { "epoch": 0.19, "learning_rate": 8.170693039408124e-05, "loss": 1.0255, "step": 73720 }, { "epoch": 0.19, "learning_rate": 8.170567215259953e-05, "loss": 1.0249, "step": 73725 }, { "epoch": 0.19, "learning_rate": 8.170441391111783e-05, "loss": 1.0292, "step": 73730 }, { "epoch": 0.19, "learning_rate": 8.170315566963612e-05, "loss": 1.0259, "step": 73735 }, { "epoch": 0.19, "learning_rate": 8.170189742815442e-05, "loss": 1.0235, "step": 73740 }, { "epoch": 0.19, "learning_rate": 8.170063918667271e-05, "loss": 1.0239, "step": 73745 }, { "epoch": 0.19, "learning_rate": 8.1699380945191e-05, "loss": 1.0251, "step": 73750 }, { "epoch": 0.19, "learning_rate": 8.16981227037093e-05, "loss": 1.0258, "step": 73755 }, { "epoch": 0.19, "learning_rate": 8.16968644622276e-05, "loss": 1.0271, "step": 73760 }, { "epoch": 0.19, "learning_rate": 8.16956062207459e-05, "loss": 1.0256, "step": 73765 }, { "epoch": 0.19, "learning_rate": 8.169434797926419e-05, "loss": 1.0244, "step": 73770 }, { "epoch": 0.19, "learning_rate": 8.169308973778248e-05, "loss": 1.0245, "step": 73775 }, { "epoch": 0.19, "learning_rate": 8.169183149630077e-05, "loss": 1.0498, "step": 73780 }, { "epoch": 0.19, "learning_rate": 8.169057325481907e-05, "loss": 1.0246, "step": 73785 }, { "epoch": 0.19, "learning_rate": 8.168931501333737e-05, "loss": 1.0241, "step": 73790 }, { "epoch": 0.19, "learning_rate": 8.168805677185566e-05, "loss": 1.0259, "step": 73795 }, { "epoch": 0.19, "learning_rate": 8.168679853037395e-05, "loss": 1.025, "step": 73800 }, { "epoch": 0.19, "learning_rate": 8.168554028889225e-05, "loss": 1.0269, "step": 73805 }, { "epoch": 0.19, "learning_rate": 8.168428204741055e-05, "loss": 1.0261, "step": 73810 }, { "epoch": 0.19, "learning_rate": 8.168302380592884e-05, "loss": 1.026, "step": 73815 }, { "epoch": 0.19, "learning_rate": 8.168176556444713e-05, "loss": 1.0249, "step": 73820 }, { "epoch": 0.19, "learning_rate": 8.168050732296543e-05, "loss": 1.0267, "step": 73825 }, { "epoch": 0.19, "learning_rate": 8.167924908148373e-05, "loss": 1.0254, "step": 73830 }, { "epoch": 0.19, "learning_rate": 8.167799084000202e-05, "loss": 1.0259, "step": 73835 }, { "epoch": 0.19, "learning_rate": 8.167673259852031e-05, "loss": 1.0276, "step": 73840 }, { "epoch": 0.19, "learning_rate": 8.16754743570386e-05, "loss": 1.0235, "step": 73845 }, { "epoch": 0.19, "learning_rate": 8.16742161155569e-05, "loss": 1.0252, "step": 73850 }, { "epoch": 0.19, "learning_rate": 8.16729578740752e-05, "loss": 1.0241, "step": 73855 }, { "epoch": 0.19, "learning_rate": 8.167169963259349e-05, "loss": 1.0249, "step": 73860 }, { "epoch": 0.19, "learning_rate": 8.167044139111178e-05, "loss": 1.0277, "step": 73865 }, { "epoch": 0.19, "learning_rate": 8.166918314963009e-05, "loss": 1.0261, "step": 73870 }, { "epoch": 0.19, "learning_rate": 8.166792490814838e-05, "loss": 1.0254, "step": 73875 }, { "epoch": 0.19, "learning_rate": 8.166666666666667e-05, "loss": 1.0241, "step": 73880 }, { "epoch": 0.19, "learning_rate": 8.166540842518496e-05, "loss": 1.0261, "step": 73885 }, { "epoch": 0.19, "learning_rate": 8.166415018370327e-05, "loss": 1.0239, "step": 73890 }, { "epoch": 0.19, "learning_rate": 8.166289194222156e-05, "loss": 1.0457, "step": 73895 }, { "epoch": 0.19, "learning_rate": 8.166163370073985e-05, "loss": 1.0221, "step": 73900 }, { "epoch": 0.19, "learning_rate": 8.166037545925814e-05, "loss": 1.0274, "step": 73905 }, { "epoch": 0.19, "learning_rate": 8.165911721777643e-05, "loss": 1.0246, "step": 73910 }, { "epoch": 0.19, "learning_rate": 8.165785897629474e-05, "loss": 1.0288, "step": 73915 }, { "epoch": 0.19, "learning_rate": 8.165660073481303e-05, "loss": 1.0264, "step": 73920 }, { "epoch": 0.19, "learning_rate": 8.165534249333132e-05, "loss": 1.0253, "step": 73925 }, { "epoch": 0.19, "learning_rate": 8.165408425184961e-05, "loss": 1.0232, "step": 73930 }, { "epoch": 0.19, "learning_rate": 8.165282601036792e-05, "loss": 1.0249, "step": 73935 }, { "epoch": 0.19, "learning_rate": 8.165156776888621e-05, "loss": 1.0249, "step": 73940 }, { "epoch": 0.19, "learning_rate": 8.16503095274045e-05, "loss": 1.0262, "step": 73945 }, { "epoch": 0.19, "learning_rate": 8.164905128592279e-05, "loss": 1.0243, "step": 73950 }, { "epoch": 0.19, "learning_rate": 8.164779304444108e-05, "loss": 1.0266, "step": 73955 }, { "epoch": 0.19, "learning_rate": 8.164653480295939e-05, "loss": 1.0488, "step": 73960 }, { "epoch": 0.19, "learning_rate": 8.164527656147768e-05, "loss": 1.0249, "step": 73965 }, { "epoch": 0.19, "learning_rate": 8.164401831999597e-05, "loss": 1.0263, "step": 73970 }, { "epoch": 0.19, "learning_rate": 8.164276007851426e-05, "loss": 1.0246, "step": 73975 }, { "epoch": 0.19, "learning_rate": 8.164150183703257e-05, "loss": 1.0228, "step": 73980 }, { "epoch": 0.19, "learning_rate": 8.164024359555086e-05, "loss": 1.0259, "step": 73985 }, { "epoch": 0.19, "learning_rate": 8.163898535406915e-05, "loss": 1.0254, "step": 73990 }, { "epoch": 0.19, "learning_rate": 8.163772711258744e-05, "loss": 1.0243, "step": 73995 }, { "epoch": 0.19, "learning_rate": 8.163646887110575e-05, "loss": 1.0242, "step": 74000 }, { "epoch": 0.19, "learning_rate": 8.163521062962404e-05, "loss": 1.0257, "step": 74005 }, { "epoch": 0.19, "learning_rate": 8.163395238814233e-05, "loss": 1.0259, "step": 74010 }, { "epoch": 0.19, "learning_rate": 8.163269414666062e-05, "loss": 1.0259, "step": 74015 }, { "epoch": 0.19, "learning_rate": 8.163143590517891e-05, "loss": 1.0257, "step": 74020 }, { "epoch": 0.19, "learning_rate": 8.163017766369722e-05, "loss": 1.0265, "step": 74025 }, { "epoch": 0.19, "learning_rate": 8.162891942221551e-05, "loss": 1.023, "step": 74030 }, { "epoch": 0.19, "learning_rate": 8.16276611807338e-05, "loss": 1.0245, "step": 74035 }, { "epoch": 0.19, "learning_rate": 8.16264029392521e-05, "loss": 1.0273, "step": 74040 }, { "epoch": 0.19, "learning_rate": 8.16251446977704e-05, "loss": 1.025, "step": 74045 }, { "epoch": 0.19, "learning_rate": 8.162388645628869e-05, "loss": 1.0252, "step": 74050 }, { "epoch": 0.19, "learning_rate": 8.162262821480698e-05, "loss": 1.0235, "step": 74055 }, { "epoch": 0.19, "learning_rate": 8.162136997332527e-05, "loss": 1.0227, "step": 74060 }, { "epoch": 0.19, "learning_rate": 8.162011173184358e-05, "loss": 1.0264, "step": 74065 }, { "epoch": 0.19, "learning_rate": 8.161885349036188e-05, "loss": 1.0228, "step": 74070 }, { "epoch": 0.19, "learning_rate": 8.161759524888018e-05, "loss": 1.0234, "step": 74075 }, { "epoch": 0.19, "learning_rate": 8.161633700739847e-05, "loss": 1.027, "step": 74080 }, { "epoch": 0.19, "learning_rate": 8.161507876591676e-05, "loss": 1.0256, "step": 74085 }, { "epoch": 0.19, "learning_rate": 8.161382052443506e-05, "loss": 1.0237, "step": 74090 }, { "epoch": 0.19, "learning_rate": 8.161256228295336e-05, "loss": 1.0243, "step": 74095 }, { "epoch": 0.19, "learning_rate": 8.161130404147165e-05, "loss": 1.0227, "step": 74100 }, { "epoch": 0.19, "learning_rate": 8.161004579998994e-05, "loss": 1.026, "step": 74105 }, { "epoch": 0.19, "learning_rate": 8.160878755850823e-05, "loss": 1.0251, "step": 74110 }, { "epoch": 0.19, "learning_rate": 8.160752931702654e-05, "loss": 1.0277, "step": 74115 }, { "epoch": 0.19, "learning_rate": 8.160627107554483e-05, "loss": 1.0239, "step": 74120 }, { "epoch": 0.19, "learning_rate": 8.160501283406312e-05, "loss": 1.028, "step": 74125 }, { "epoch": 0.19, "learning_rate": 8.160375459258141e-05, "loss": 1.0267, "step": 74130 }, { "epoch": 0.19, "learning_rate": 8.160249635109972e-05, "loss": 1.0286, "step": 74135 }, { "epoch": 0.19, "learning_rate": 8.160123810961801e-05, "loss": 1.0239, "step": 74140 }, { "epoch": 0.19, "learning_rate": 8.15999798681363e-05, "loss": 1.023, "step": 74145 }, { "epoch": 0.19, "learning_rate": 8.159872162665459e-05, "loss": 1.0244, "step": 74150 }, { "epoch": 0.19, "learning_rate": 8.15974633851729e-05, "loss": 1.0242, "step": 74155 }, { "epoch": 0.19, "learning_rate": 8.159620514369119e-05, "loss": 1.0253, "step": 74160 }, { "epoch": 0.19, "learning_rate": 8.159494690220948e-05, "loss": 1.0261, "step": 74165 }, { "epoch": 0.19, "learning_rate": 8.159368866072777e-05, "loss": 1.0463, "step": 74170 }, { "epoch": 0.19, "learning_rate": 8.159243041924606e-05, "loss": 1.0472, "step": 74175 }, { "epoch": 0.19, "learning_rate": 8.159117217776437e-05, "loss": 1.026, "step": 74180 }, { "epoch": 0.19, "learning_rate": 8.158991393628266e-05, "loss": 1.0274, "step": 74185 }, { "epoch": 0.19, "learning_rate": 8.158865569480095e-05, "loss": 1.0263, "step": 74190 }, { "epoch": 0.19, "learning_rate": 8.158739745331924e-05, "loss": 1.0263, "step": 74195 }, { "epoch": 0.19, "learning_rate": 8.158613921183755e-05, "loss": 1.0245, "step": 74200 }, { "epoch": 0.19, "learning_rate": 8.158488097035584e-05, "loss": 1.0277, "step": 74205 }, { "epoch": 0.19, "learning_rate": 8.158362272887413e-05, "loss": 1.0261, "step": 74210 }, { "epoch": 0.19, "learning_rate": 8.158236448739242e-05, "loss": 1.0241, "step": 74215 }, { "epoch": 0.19, "learning_rate": 8.158110624591073e-05, "loss": 1.0264, "step": 74220 }, { "epoch": 0.19, "learning_rate": 8.157984800442902e-05, "loss": 1.023, "step": 74225 }, { "epoch": 0.19, "learning_rate": 8.157858976294731e-05, "loss": 1.0252, "step": 74230 }, { "epoch": 0.19, "learning_rate": 8.15773315214656e-05, "loss": 1.026, "step": 74235 }, { "epoch": 0.19, "learning_rate": 8.15760732799839e-05, "loss": 1.0249, "step": 74240 }, { "epoch": 0.19, "learning_rate": 8.15748150385022e-05, "loss": 1.0254, "step": 74245 }, { "epoch": 0.19, "learning_rate": 8.157355679702049e-05, "loss": 1.0241, "step": 74250 }, { "epoch": 0.19, "learning_rate": 8.157229855553878e-05, "loss": 1.0232, "step": 74255 }, { "epoch": 0.19, "learning_rate": 8.157104031405707e-05, "loss": 1.0269, "step": 74260 }, { "epoch": 0.19, "learning_rate": 8.156978207257538e-05, "loss": 1.0249, "step": 74265 }, { "epoch": 0.19, "learning_rate": 8.156852383109367e-05, "loss": 1.0245, "step": 74270 }, { "epoch": 0.19, "learning_rate": 8.156726558961196e-05, "loss": 1.0243, "step": 74275 }, { "epoch": 0.19, "learning_rate": 8.156600734813025e-05, "loss": 1.0263, "step": 74280 }, { "epoch": 0.19, "learning_rate": 8.156474910664856e-05, "loss": 1.0244, "step": 74285 }, { "epoch": 0.19, "learning_rate": 8.156349086516685e-05, "loss": 1.0242, "step": 74290 }, { "epoch": 0.19, "learning_rate": 8.156223262368514e-05, "loss": 1.0246, "step": 74295 }, { "epoch": 0.19, "learning_rate": 8.156097438220343e-05, "loss": 1.0231, "step": 74300 }, { "epoch": 0.19, "learning_rate": 8.155971614072173e-05, "loss": 1.0273, "step": 74305 }, { "epoch": 0.19, "learning_rate": 8.155845789924003e-05, "loss": 1.0264, "step": 74310 }, { "epoch": 0.19, "learning_rate": 8.155719965775832e-05, "loss": 1.0246, "step": 74315 }, { "epoch": 0.19, "learning_rate": 8.155594141627661e-05, "loss": 1.0237, "step": 74320 }, { "epoch": 0.19, "learning_rate": 8.15546831747949e-05, "loss": 1.0258, "step": 74325 }, { "epoch": 0.19, "learning_rate": 8.155342493331321e-05, "loss": 1.0286, "step": 74330 }, { "epoch": 0.19, "learning_rate": 8.15521666918315e-05, "loss": 1.0261, "step": 74335 }, { "epoch": 0.19, "learning_rate": 8.15509084503498e-05, "loss": 1.0269, "step": 74340 }, { "epoch": 0.19, "learning_rate": 8.154965020886809e-05, "loss": 1.0258, "step": 74345 }, { "epoch": 0.19, "learning_rate": 8.154839196738639e-05, "loss": 1.0262, "step": 74350 }, { "epoch": 0.19, "learning_rate": 8.154713372590468e-05, "loss": 1.0262, "step": 74355 }, { "epoch": 0.19, "learning_rate": 8.154587548442297e-05, "loss": 1.0248, "step": 74360 }, { "epoch": 0.19, "learning_rate": 8.154461724294127e-05, "loss": 1.0249, "step": 74365 }, { "epoch": 0.19, "learning_rate": 8.154335900145956e-05, "loss": 1.0247, "step": 74370 }, { "epoch": 0.19, "learning_rate": 8.154210075997786e-05, "loss": 1.0223, "step": 74375 }, { "epoch": 0.19, "learning_rate": 8.154084251849615e-05, "loss": 1.0268, "step": 74380 }, { "epoch": 0.19, "learning_rate": 8.153958427701445e-05, "loss": 1.0278, "step": 74385 }, { "epoch": 0.19, "learning_rate": 8.153832603553274e-05, "loss": 1.0233, "step": 74390 }, { "epoch": 0.19, "learning_rate": 8.153706779405104e-05, "loss": 1.0234, "step": 74395 }, { "epoch": 0.19, "learning_rate": 8.153580955256933e-05, "loss": 1.0234, "step": 74400 }, { "epoch": 0.19, "learning_rate": 8.153455131108763e-05, "loss": 1.0274, "step": 74405 }, { "epoch": 0.19, "learning_rate": 8.153329306960592e-05, "loss": 1.0242, "step": 74410 }, { "epoch": 0.19, "learning_rate": 8.153203482812422e-05, "loss": 1.0249, "step": 74415 }, { "epoch": 0.19, "learning_rate": 8.153077658664251e-05, "loss": 1.026, "step": 74420 }, { "epoch": 0.19, "learning_rate": 8.15295183451608e-05, "loss": 1.0259, "step": 74425 }, { "epoch": 0.19, "learning_rate": 8.15282601036791e-05, "loss": 1.0253, "step": 74430 }, { "epoch": 0.19, "learning_rate": 8.152700186219739e-05, "loss": 1.0249, "step": 74435 }, { "epoch": 0.19, "learning_rate": 8.15257436207157e-05, "loss": 1.0269, "step": 74440 }, { "epoch": 0.19, "learning_rate": 8.152448537923398e-05, "loss": 1.0232, "step": 74445 }, { "epoch": 0.19, "learning_rate": 8.152322713775228e-05, "loss": 1.0228, "step": 74450 }, { "epoch": 0.19, "learning_rate": 8.152196889627057e-05, "loss": 1.0228, "step": 74455 }, { "epoch": 0.19, "learning_rate": 8.152071065478887e-05, "loss": 1.0258, "step": 74460 }, { "epoch": 0.19, "learning_rate": 8.151945241330716e-05, "loss": 1.0247, "step": 74465 }, { "epoch": 0.19, "learning_rate": 8.151819417182546e-05, "loss": 1.0249, "step": 74470 }, { "epoch": 0.19, "learning_rate": 8.151693593034375e-05, "loss": 1.0229, "step": 74475 }, { "epoch": 0.19, "learning_rate": 8.151567768886205e-05, "loss": 1.0244, "step": 74480 }, { "epoch": 0.19, "learning_rate": 8.151441944738034e-05, "loss": 1.0255, "step": 74485 }, { "epoch": 0.19, "learning_rate": 8.151316120589864e-05, "loss": 1.0248, "step": 74490 }, { "epoch": 0.19, "learning_rate": 8.151190296441693e-05, "loss": 1.0238, "step": 74495 }, { "epoch": 0.19, "learning_rate": 8.151064472293522e-05, "loss": 1.0265, "step": 74500 }, { "epoch": 0.19, "learning_rate": 8.150938648145352e-05, "loss": 1.0253, "step": 74505 }, { "epoch": 0.19, "learning_rate": 8.150812823997182e-05, "loss": 1.0241, "step": 74510 }, { "epoch": 0.19, "learning_rate": 8.150686999849011e-05, "loss": 1.0267, "step": 74515 }, { "epoch": 0.19, "learning_rate": 8.15056117570084e-05, "loss": 1.0269, "step": 74520 }, { "epoch": 0.19, "learning_rate": 8.15043535155267e-05, "loss": 1.0244, "step": 74525 }, { "epoch": 0.19, "learning_rate": 8.1503095274045e-05, "loss": 1.0263, "step": 74530 }, { "epoch": 0.19, "learning_rate": 8.150183703256329e-05, "loss": 1.0251, "step": 74535 }, { "epoch": 0.19, "learning_rate": 8.150057879108158e-05, "loss": 1.0236, "step": 74540 }, { "epoch": 0.19, "learning_rate": 8.149932054959988e-05, "loss": 1.0246, "step": 74545 }, { "epoch": 0.19, "learning_rate": 8.149806230811818e-05, "loss": 1.026, "step": 74550 }, { "epoch": 0.19, "learning_rate": 8.149680406663647e-05, "loss": 1.0237, "step": 74555 }, { "epoch": 0.19, "learning_rate": 8.149554582515476e-05, "loss": 1.0221, "step": 74560 }, { "epoch": 0.19, "learning_rate": 8.149428758367306e-05, "loss": 1.0273, "step": 74565 }, { "epoch": 0.19, "learning_rate": 8.149302934219136e-05, "loss": 1.0227, "step": 74570 }, { "epoch": 0.19, "learning_rate": 8.149177110070966e-05, "loss": 1.0256, "step": 74575 }, { "epoch": 0.19, "learning_rate": 8.149051285922795e-05, "loss": 1.0244, "step": 74580 }, { "epoch": 0.19, "learning_rate": 8.148925461774624e-05, "loss": 1.0246, "step": 74585 }, { "epoch": 0.19, "learning_rate": 8.148799637626454e-05, "loss": 1.0248, "step": 74590 }, { "epoch": 0.19, "learning_rate": 8.148673813478284e-05, "loss": 1.0268, "step": 74595 }, { "epoch": 0.19, "learning_rate": 8.148547989330113e-05, "loss": 1.0285, "step": 74600 }, { "epoch": 0.19, "learning_rate": 8.148422165181942e-05, "loss": 1.0268, "step": 74605 }, { "epoch": 0.19, "learning_rate": 8.148296341033772e-05, "loss": 1.0255, "step": 74610 }, { "epoch": 0.19, "learning_rate": 8.148170516885602e-05, "loss": 1.0245, "step": 74615 }, { "epoch": 0.19, "learning_rate": 8.148044692737431e-05, "loss": 1.0453, "step": 74620 }, { "epoch": 0.19, "learning_rate": 8.14791886858926e-05, "loss": 1.0272, "step": 74625 }, { "epoch": 0.19, "learning_rate": 8.14779304444109e-05, "loss": 1.0264, "step": 74630 }, { "epoch": 0.19, "learning_rate": 8.147667220292919e-05, "loss": 1.0257, "step": 74635 }, { "epoch": 0.19, "learning_rate": 8.147541396144749e-05, "loss": 1.0263, "step": 74640 }, { "epoch": 0.19, "learning_rate": 8.147415571996578e-05, "loss": 1.0244, "step": 74645 }, { "epoch": 0.19, "learning_rate": 8.147289747848408e-05, "loss": 1.0246, "step": 74650 }, { "epoch": 0.19, "learning_rate": 8.147163923700237e-05, "loss": 1.0253, "step": 74655 }, { "epoch": 0.19, "learning_rate": 8.147038099552067e-05, "loss": 1.0267, "step": 74660 }, { "epoch": 0.19, "learning_rate": 8.146912275403896e-05, "loss": 1.0242, "step": 74665 }, { "epoch": 0.19, "learning_rate": 8.146786451255726e-05, "loss": 1.0253, "step": 74670 }, { "epoch": 0.19, "learning_rate": 8.146660627107555e-05, "loss": 1.0235, "step": 74675 }, { "epoch": 0.19, "learning_rate": 8.146534802959385e-05, "loss": 1.0244, "step": 74680 }, { "epoch": 0.19, "learning_rate": 8.146408978811214e-05, "loss": 1.0253, "step": 74685 }, { "epoch": 0.19, "learning_rate": 8.146283154663044e-05, "loss": 1.0248, "step": 74690 }, { "epoch": 0.19, "learning_rate": 8.146157330514873e-05, "loss": 1.0243, "step": 74695 }, { "epoch": 0.19, "learning_rate": 8.146031506366702e-05, "loss": 1.0243, "step": 74700 }, { "epoch": 0.19, "learning_rate": 8.145905682218532e-05, "loss": 1.0283, "step": 74705 }, { "epoch": 0.19, "learning_rate": 8.145779858070362e-05, "loss": 1.0264, "step": 74710 }, { "epoch": 0.19, "learning_rate": 8.145654033922191e-05, "loss": 1.0246, "step": 74715 }, { "epoch": 0.19, "learning_rate": 8.14552820977402e-05, "loss": 1.0254, "step": 74720 }, { "epoch": 0.19, "learning_rate": 8.14540238562585e-05, "loss": 1.0252, "step": 74725 }, { "epoch": 0.19, "learning_rate": 8.14527656147768e-05, "loss": 1.0248, "step": 74730 }, { "epoch": 0.19, "learning_rate": 8.145150737329509e-05, "loss": 1.0245, "step": 74735 }, { "epoch": 0.19, "learning_rate": 8.145024913181338e-05, "loss": 1.0257, "step": 74740 }, { "epoch": 0.19, "learning_rate": 8.144899089033168e-05, "loss": 1.0249, "step": 74745 }, { "epoch": 0.19, "learning_rate": 8.144773264884998e-05, "loss": 1.0266, "step": 74750 }, { "epoch": 0.19, "learning_rate": 8.144647440736827e-05, "loss": 1.0247, "step": 74755 }, { "epoch": 0.19, "learning_rate": 8.144521616588656e-05, "loss": 1.0243, "step": 74760 }, { "epoch": 0.19, "learning_rate": 8.144395792440485e-05, "loss": 1.0235, "step": 74765 }, { "epoch": 0.19, "learning_rate": 8.144269968292316e-05, "loss": 1.0278, "step": 74770 }, { "epoch": 0.19, "learning_rate": 8.144144144144145e-05, "loss": 1.0236, "step": 74775 }, { "epoch": 0.19, "learning_rate": 8.144018319995974e-05, "loss": 1.0252, "step": 74780 }, { "epoch": 0.19, "learning_rate": 8.143892495847803e-05, "loss": 1.0232, "step": 74785 }, { "epoch": 0.19, "learning_rate": 8.143766671699634e-05, "loss": 1.026, "step": 74790 }, { "epoch": 0.19, "learning_rate": 8.143640847551463e-05, "loss": 1.0251, "step": 74795 }, { "epoch": 0.19, "learning_rate": 8.143515023403292e-05, "loss": 1.0239, "step": 74800 }, { "epoch": 0.19, "learning_rate": 8.143389199255121e-05, "loss": 1.0248, "step": 74805 }, { "epoch": 0.19, "learning_rate": 8.143263375106952e-05, "loss": 1.0214, "step": 74810 }, { "epoch": 0.19, "learning_rate": 8.143137550958781e-05, "loss": 1.025, "step": 74815 }, { "epoch": 0.19, "learning_rate": 8.14301172681061e-05, "loss": 1.0251, "step": 74820 }, { "epoch": 0.19, "learning_rate": 8.142885902662439e-05, "loss": 1.0264, "step": 74825 }, { "epoch": 0.19, "learning_rate": 8.142760078514268e-05, "loss": 1.0241, "step": 74830 }, { "epoch": 0.19, "learning_rate": 8.142634254366099e-05, "loss": 1.0257, "step": 74835 }, { "epoch": 0.19, "learning_rate": 8.142508430217928e-05, "loss": 1.0268, "step": 74840 }, { "epoch": 0.19, "learning_rate": 8.142382606069757e-05, "loss": 1.0236, "step": 74845 }, { "epoch": 0.19, "learning_rate": 8.142256781921586e-05, "loss": 1.0256, "step": 74850 }, { "epoch": 0.19, "learning_rate": 8.142130957773417e-05, "loss": 1.0242, "step": 74855 }, { "epoch": 0.19, "learning_rate": 8.142005133625246e-05, "loss": 1.0216, "step": 74860 }, { "epoch": 0.19, "learning_rate": 8.141879309477075e-05, "loss": 1.0257, "step": 74865 }, { "epoch": 0.19, "learning_rate": 8.141753485328904e-05, "loss": 1.045, "step": 74870 }, { "epoch": 0.19, "learning_rate": 8.141627661180735e-05, "loss": 1.0242, "step": 74875 }, { "epoch": 0.19, "learning_rate": 8.141501837032564e-05, "loss": 1.0281, "step": 74880 }, { "epoch": 0.19, "learning_rate": 8.141376012884393e-05, "loss": 1.0256, "step": 74885 }, { "epoch": 0.19, "learning_rate": 8.141250188736222e-05, "loss": 1.0261, "step": 74890 }, { "epoch": 0.19, "learning_rate": 8.141124364588051e-05, "loss": 1.0219, "step": 74895 }, { "epoch": 0.19, "learning_rate": 8.140998540439882e-05, "loss": 1.0247, "step": 74900 }, { "epoch": 0.19, "learning_rate": 8.140897881121345e-05, "loss": 1.0235, "step": 74905 }, { "epoch": 0.19, "learning_rate": 8.140772056973175e-05, "loss": 1.0283, "step": 74910 }, { "epoch": 0.19, "learning_rate": 8.140646232825004e-05, "loss": 1.0259, "step": 74915 }, { "epoch": 0.19, "learning_rate": 8.140520408676834e-05, "loss": 1.0244, "step": 74920 }, { "epoch": 0.19, "learning_rate": 8.140394584528663e-05, "loss": 1.0244, "step": 74925 }, { "epoch": 0.19, "learning_rate": 8.140268760380493e-05, "loss": 1.0236, "step": 74930 }, { "epoch": 0.19, "learning_rate": 8.140142936232322e-05, "loss": 1.0237, "step": 74935 }, { "epoch": 0.19, "learning_rate": 8.140017112084152e-05, "loss": 1.0237, "step": 74940 }, { "epoch": 0.19, "learning_rate": 8.139891287935981e-05, "loss": 1.0254, "step": 74945 }, { "epoch": 0.19, "learning_rate": 8.13976546378781e-05, "loss": 1.0216, "step": 74950 }, { "epoch": 0.19, "learning_rate": 8.13963963963964e-05, "loss": 1.0258, "step": 74955 }, { "epoch": 0.19, "learning_rate": 8.13951381549147e-05, "loss": 1.0246, "step": 74960 }, { "epoch": 0.19, "learning_rate": 8.1393879913433e-05, "loss": 1.0254, "step": 74965 }, { "epoch": 0.19, "learning_rate": 8.139262167195129e-05, "loss": 1.0234, "step": 74970 }, { "epoch": 0.19, "learning_rate": 8.139136343046958e-05, "loss": 1.0259, "step": 74975 }, { "epoch": 0.19, "learning_rate": 8.139010518898787e-05, "loss": 1.0264, "step": 74980 }, { "epoch": 0.19, "learning_rate": 8.138884694750617e-05, "loss": 1.0261, "step": 74985 }, { "epoch": 0.19, "learning_rate": 8.138758870602447e-05, "loss": 1.0238, "step": 74990 }, { "epoch": 0.19, "learning_rate": 8.138633046454276e-05, "loss": 1.0254, "step": 74995 }, { "epoch": 0.19, "learning_rate": 8.138507222306105e-05, "loss": 1.0258, "step": 75000 }, { "epoch": 0.19, "learning_rate": 8.138381398157935e-05, "loss": 1.0255, "step": 75005 }, { "epoch": 0.19, "learning_rate": 8.138255574009765e-05, "loss": 1.0252, "step": 75010 }, { "epoch": 0.19, "learning_rate": 8.138129749861594e-05, "loss": 1.0486, "step": 75015 }, { "epoch": 0.19, "learning_rate": 8.138003925713423e-05, "loss": 1.0264, "step": 75020 }, { "epoch": 0.19, "learning_rate": 8.137878101565253e-05, "loss": 1.0248, "step": 75025 }, { "epoch": 0.19, "learning_rate": 8.137752277417083e-05, "loss": 1.051, "step": 75030 }, { "epoch": 0.19, "learning_rate": 8.137626453268912e-05, "loss": 1.0251, "step": 75035 }, { "epoch": 0.19, "learning_rate": 8.137500629120741e-05, "loss": 1.0266, "step": 75040 }, { "epoch": 0.19, "learning_rate": 8.13737480497257e-05, "loss": 1.0245, "step": 75045 }, { "epoch": 0.19, "learning_rate": 8.1372489808244e-05, "loss": 1.0231, "step": 75050 }, { "epoch": 0.19, "learning_rate": 8.13712315667623e-05, "loss": 1.0259, "step": 75055 }, { "epoch": 0.19, "learning_rate": 8.136997332528059e-05, "loss": 1.0262, "step": 75060 }, { "epoch": 0.19, "learning_rate": 8.136871508379888e-05, "loss": 1.0271, "step": 75065 }, { "epoch": 0.19, "learning_rate": 8.136745684231719e-05, "loss": 1.0228, "step": 75070 }, { "epoch": 0.19, "learning_rate": 8.136619860083548e-05, "loss": 1.0243, "step": 75075 }, { "epoch": 0.19, "learning_rate": 8.136494035935377e-05, "loss": 1.0242, "step": 75080 }, { "epoch": 0.19, "learning_rate": 8.136368211787206e-05, "loss": 1.0251, "step": 75085 }, { "epoch": 0.19, "learning_rate": 8.136242387639037e-05, "loss": 1.0266, "step": 75090 }, { "epoch": 0.19, "learning_rate": 8.136116563490866e-05, "loss": 1.0242, "step": 75095 }, { "epoch": 0.19, "learning_rate": 8.135990739342695e-05, "loss": 1.0239, "step": 75100 }, { "epoch": 0.19, "learning_rate": 8.135864915194524e-05, "loss": 1.0249, "step": 75105 }, { "epoch": 0.19, "learning_rate": 8.135739091046353e-05, "loss": 1.0248, "step": 75110 }, { "epoch": 0.19, "learning_rate": 8.135613266898184e-05, "loss": 1.0251, "step": 75115 }, { "epoch": 0.19, "learning_rate": 8.135487442750013e-05, "loss": 1.026, "step": 75120 }, { "epoch": 0.19, "learning_rate": 8.135361618601842e-05, "loss": 1.0261, "step": 75125 }, { "epoch": 0.19, "learning_rate": 8.135235794453671e-05, "loss": 1.025, "step": 75130 }, { "epoch": 0.19, "learning_rate": 8.135109970305502e-05, "loss": 1.025, "step": 75135 }, { "epoch": 0.19, "learning_rate": 8.134984146157331e-05, "loss": 1.0272, "step": 75140 }, { "epoch": 0.19, "learning_rate": 8.13485832200916e-05, "loss": 1.0257, "step": 75145 }, { "epoch": 0.19, "learning_rate": 8.134732497860989e-05, "loss": 1.0245, "step": 75150 }, { "epoch": 0.19, "learning_rate": 8.13460667371282e-05, "loss": 1.0226, "step": 75155 }, { "epoch": 0.19, "learning_rate": 8.134480849564649e-05, "loss": 1.0234, "step": 75160 }, { "epoch": 0.19, "learning_rate": 8.134355025416478e-05, "loss": 1.0271, "step": 75165 }, { "epoch": 0.19, "learning_rate": 8.134229201268307e-05, "loss": 1.0261, "step": 75170 }, { "epoch": 0.19, "learning_rate": 8.134103377120136e-05, "loss": 1.043, "step": 75175 }, { "epoch": 0.19, "learning_rate": 8.133977552971967e-05, "loss": 1.0275, "step": 75180 }, { "epoch": 0.19, "learning_rate": 8.133851728823796e-05, "loss": 1.0266, "step": 75185 }, { "epoch": 0.19, "learning_rate": 8.133725904675625e-05, "loss": 1.0235, "step": 75190 }, { "epoch": 0.19, "learning_rate": 8.133600080527454e-05, "loss": 1.0267, "step": 75195 }, { "epoch": 0.19, "learning_rate": 8.133474256379285e-05, "loss": 1.0242, "step": 75200 }, { "epoch": 0.19, "learning_rate": 8.133373597060748e-05, "loss": 1.0259, "step": 75205 }, { "epoch": 0.19, "learning_rate": 8.133247772912578e-05, "loss": 1.0253, "step": 75210 }, { "epoch": 0.19, "learning_rate": 8.133121948764407e-05, "loss": 1.0258, "step": 75215 }, { "epoch": 0.19, "learning_rate": 8.132996124616237e-05, "loss": 1.0263, "step": 75220 }, { "epoch": 0.19, "learning_rate": 8.132870300468066e-05, "loss": 1.0242, "step": 75225 }, { "epoch": 0.19, "learning_rate": 8.132744476319896e-05, "loss": 1.0258, "step": 75230 }, { "epoch": 0.19, "learning_rate": 8.132618652171725e-05, "loss": 1.0265, "step": 75235 }, { "epoch": 0.19, "learning_rate": 8.132492828023555e-05, "loss": 1.0255, "step": 75240 }, { "epoch": 0.19, "learning_rate": 8.132367003875384e-05, "loss": 1.0257, "step": 75245 }, { "epoch": 0.19, "learning_rate": 8.132241179727214e-05, "loss": 1.026, "step": 75250 }, { "epoch": 0.19, "learning_rate": 8.132115355579043e-05, "loss": 1.0258, "step": 75255 }, { "epoch": 0.19, "learning_rate": 8.131989531430872e-05, "loss": 1.0245, "step": 75260 }, { "epoch": 0.19, "learning_rate": 8.131863707282702e-05, "loss": 1.0245, "step": 75265 }, { "epoch": 0.19, "learning_rate": 8.131737883134532e-05, "loss": 1.0231, "step": 75270 }, { "epoch": 0.19, "learning_rate": 8.131612058986361e-05, "loss": 1.0252, "step": 75275 }, { "epoch": 0.19, "learning_rate": 8.13148623483819e-05, "loss": 1.0248, "step": 75280 }, { "epoch": 0.19, "learning_rate": 8.13136041069002e-05, "loss": 1.0482, "step": 75285 }, { "epoch": 0.19, "learning_rate": 8.13123458654185e-05, "loss": 1.0248, "step": 75290 }, { "epoch": 0.19, "learning_rate": 8.131108762393679e-05, "loss": 1.0269, "step": 75295 }, { "epoch": 0.19, "learning_rate": 8.130982938245508e-05, "loss": 1.0244, "step": 75300 }, { "epoch": 0.19, "learning_rate": 8.130857114097338e-05, "loss": 1.0253, "step": 75305 }, { "epoch": 0.19, "learning_rate": 8.130731289949168e-05, "loss": 1.0257, "step": 75310 }, { "epoch": 0.19, "learning_rate": 8.130605465800997e-05, "loss": 1.0238, "step": 75315 }, { "epoch": 0.19, "learning_rate": 8.130479641652826e-05, "loss": 1.0234, "step": 75320 }, { "epoch": 0.19, "learning_rate": 8.130353817504655e-05, "loss": 1.024, "step": 75325 }, { "epoch": 0.19, "learning_rate": 8.130227993356486e-05, "loss": 1.0278, "step": 75330 }, { "epoch": 0.19, "learning_rate": 8.130102169208315e-05, "loss": 1.0284, "step": 75335 }, { "epoch": 0.19, "learning_rate": 8.129976345060144e-05, "loss": 1.0253, "step": 75340 }, { "epoch": 0.19, "learning_rate": 8.129850520911973e-05, "loss": 1.0248, "step": 75345 }, { "epoch": 0.19, "learning_rate": 8.129724696763804e-05, "loss": 1.0246, "step": 75350 }, { "epoch": 0.19, "learning_rate": 8.129598872615633e-05, "loss": 1.0212, "step": 75355 }, { "epoch": 0.19, "learning_rate": 8.129473048467462e-05, "loss": 1.0243, "step": 75360 }, { "epoch": 0.19, "learning_rate": 8.129347224319291e-05, "loss": 1.0233, "step": 75365 }, { "epoch": 0.19, "learning_rate": 8.129221400171122e-05, "loss": 1.025, "step": 75370 }, { "epoch": 0.19, "learning_rate": 8.129095576022951e-05, "loss": 1.0245, "step": 75375 }, { "epoch": 0.19, "learning_rate": 8.12896975187478e-05, "loss": 1.0256, "step": 75380 }, { "epoch": 0.19, "learning_rate": 8.128843927726609e-05, "loss": 1.0252, "step": 75385 }, { "epoch": 0.19, "learning_rate": 8.128718103578438e-05, "loss": 1.0226, "step": 75390 }, { "epoch": 0.19, "learning_rate": 8.128592279430269e-05, "loss": 1.0254, "step": 75395 }, { "epoch": 0.19, "learning_rate": 8.128466455282098e-05, "loss": 1.0258, "step": 75400 }, { "epoch": 0.19, "learning_rate": 8.128340631133927e-05, "loss": 1.0258, "step": 75405 }, { "epoch": 0.19, "learning_rate": 8.128214806985756e-05, "loss": 1.0247, "step": 75410 }, { "epoch": 0.19, "learning_rate": 8.128088982837587e-05, "loss": 1.0252, "step": 75415 }, { "epoch": 0.19, "learning_rate": 8.127963158689416e-05, "loss": 1.0243, "step": 75420 }, { "epoch": 0.19, "learning_rate": 8.127837334541245e-05, "loss": 1.0262, "step": 75425 }, { "epoch": 0.19, "learning_rate": 8.127711510393074e-05, "loss": 1.0229, "step": 75430 }, { "epoch": 0.19, "learning_rate": 8.127585686244903e-05, "loss": 1.0233, "step": 75435 }, { "epoch": 0.19, "learning_rate": 8.127459862096734e-05, "loss": 1.025, "step": 75440 }, { "epoch": 0.19, "learning_rate": 8.127334037948563e-05, "loss": 1.0256, "step": 75445 }, { "epoch": 0.19, "learning_rate": 8.127208213800392e-05, "loss": 1.0242, "step": 75450 }, { "epoch": 0.19, "learning_rate": 8.127082389652221e-05, "loss": 1.0239, "step": 75455 }, { "epoch": 0.19, "learning_rate": 8.126956565504052e-05, "loss": 1.0261, "step": 75460 }, { "epoch": 0.19, "learning_rate": 8.126830741355882e-05, "loss": 1.026, "step": 75465 }, { "epoch": 0.19, "learning_rate": 8.126704917207712e-05, "loss": 1.0241, "step": 75470 }, { "epoch": 0.19, "learning_rate": 8.126579093059541e-05, "loss": 1.0256, "step": 75475 }, { "epoch": 0.19, "learning_rate": 8.12645326891137e-05, "loss": 1.0263, "step": 75480 }, { "epoch": 0.19, "learning_rate": 8.1263274447632e-05, "loss": 1.0232, "step": 75485 }, { "epoch": 0.19, "learning_rate": 8.12620162061503e-05, "loss": 1.0248, "step": 75490 }, { "epoch": 0.19, "learning_rate": 8.126075796466859e-05, "loss": 1.025, "step": 75495 }, { "epoch": 0.19, "learning_rate": 8.125949972318688e-05, "loss": 1.0286, "step": 75500 }, { "epoch": 0.19, "learning_rate": 8.125824148170518e-05, "loss": 1.027, "step": 75505 }, { "epoch": 0.19, "learning_rate": 8.125698324022348e-05, "loss": 1.0244, "step": 75510 }, { "epoch": 0.19, "learning_rate": 8.125572499874177e-05, "loss": 1.0227, "step": 75515 }, { "epoch": 0.19, "learning_rate": 8.125446675726006e-05, "loss": 1.0232, "step": 75520 }, { "epoch": 0.19, "learning_rate": 8.125320851577835e-05, "loss": 1.0246, "step": 75525 }, { "epoch": 0.19, "learning_rate": 8.125195027429666e-05, "loss": 1.0223, "step": 75530 }, { "epoch": 0.19, "learning_rate": 8.125069203281495e-05, "loss": 1.0264, "step": 75535 }, { "epoch": 0.19, "learning_rate": 8.124943379133324e-05, "loss": 1.0235, "step": 75540 }, { "epoch": 0.19, "learning_rate": 8.124817554985153e-05, "loss": 1.0264, "step": 75545 }, { "epoch": 0.19, "learning_rate": 8.124691730836984e-05, "loss": 1.0249, "step": 75550 }, { "epoch": 0.19, "learning_rate": 8.124565906688813e-05, "loss": 1.0241, "step": 75555 }, { "epoch": 0.19, "learning_rate": 8.124440082540642e-05, "loss": 1.0258, "step": 75560 }, { "epoch": 0.19, "learning_rate": 8.124314258392471e-05, "loss": 1.024, "step": 75565 }, { "epoch": 0.19, "learning_rate": 8.124188434244302e-05, "loss": 1.0242, "step": 75570 }, { "epoch": 0.19, "learning_rate": 8.12406261009613e-05, "loss": 1.0258, "step": 75575 }, { "epoch": 0.19, "learning_rate": 8.12393678594796e-05, "loss": 1.0233, "step": 75580 }, { "epoch": 0.19, "learning_rate": 8.123810961799789e-05, "loss": 1.0246, "step": 75585 }, { "epoch": 0.19, "learning_rate": 8.123685137651618e-05, "loss": 1.027, "step": 75590 }, { "epoch": 0.19, "learning_rate": 8.123559313503449e-05, "loss": 1.0233, "step": 75595 }, { "epoch": 0.19, "learning_rate": 8.123433489355278e-05, "loss": 1.0234, "step": 75600 }, { "epoch": 0.19, "learning_rate": 8.123307665207107e-05, "loss": 1.0246, "step": 75605 }, { "epoch": 0.19, "learning_rate": 8.123181841058936e-05, "loss": 1.0235, "step": 75610 }, { "epoch": 0.19, "learning_rate": 8.123056016910767e-05, "loss": 1.0257, "step": 75615 }, { "epoch": 0.19, "learning_rate": 8.122930192762596e-05, "loss": 1.0257, "step": 75620 }, { "epoch": 0.19, "learning_rate": 8.122804368614425e-05, "loss": 1.0256, "step": 75625 }, { "epoch": 0.19, "learning_rate": 8.122678544466254e-05, "loss": 1.024, "step": 75630 }, { "epoch": 0.19, "learning_rate": 8.122552720318085e-05, "loss": 1.026, "step": 75635 }, { "epoch": 0.19, "learning_rate": 8.122426896169914e-05, "loss": 1.0252, "step": 75640 }, { "epoch": 0.19, "learning_rate": 8.122301072021743e-05, "loss": 1.0267, "step": 75645 }, { "epoch": 0.19, "learning_rate": 8.122175247873572e-05, "loss": 1.026, "step": 75650 }, { "epoch": 0.19, "learning_rate": 8.122049423725401e-05, "loss": 1.0264, "step": 75655 }, { "epoch": 0.19, "learning_rate": 8.121923599577232e-05, "loss": 1.0274, "step": 75660 }, { "epoch": 0.19, "learning_rate": 8.121797775429061e-05, "loss": 1.0251, "step": 75665 }, { "epoch": 0.19, "learning_rate": 8.12167195128089e-05, "loss": 1.0252, "step": 75670 }, { "epoch": 0.19, "learning_rate": 8.121546127132719e-05, "loss": 1.0258, "step": 75675 }, { "epoch": 0.19, "learning_rate": 8.12142030298455e-05, "loss": 1.0235, "step": 75680 }, { "epoch": 0.19, "learning_rate": 8.121294478836379e-05, "loss": 1.0244, "step": 75685 }, { "epoch": 0.19, "learning_rate": 8.121168654688208e-05, "loss": 1.0267, "step": 75690 }, { "epoch": 0.19, "learning_rate": 8.121042830540037e-05, "loss": 1.0247, "step": 75695 }, { "epoch": 0.19, "learning_rate": 8.120917006391868e-05, "loss": 1.0238, "step": 75700 }, { "epoch": 0.19, "learning_rate": 8.120791182243697e-05, "loss": 1.0246, "step": 75705 }, { "epoch": 0.19, "learning_rate": 8.120665358095526e-05, "loss": 1.0248, "step": 75710 }, { "epoch": 0.19, "learning_rate": 8.120539533947355e-05, "loss": 1.025, "step": 75715 }, { "epoch": 0.19, "learning_rate": 8.120413709799184e-05, "loss": 1.0244, "step": 75720 }, { "epoch": 0.19, "learning_rate": 8.120287885651015e-05, "loss": 1.0234, "step": 75725 }, { "epoch": 0.19, "learning_rate": 8.120162061502844e-05, "loss": 1.0239, "step": 75730 }, { "epoch": 0.19, "learning_rate": 8.120036237354673e-05, "loss": 1.0235, "step": 75735 }, { "epoch": 0.19, "learning_rate": 8.119910413206502e-05, "loss": 1.0245, "step": 75740 }, { "epoch": 0.19, "learning_rate": 8.119784589058333e-05, "loss": 1.0246, "step": 75745 }, { "epoch": 0.19, "learning_rate": 8.119658764910162e-05, "loss": 1.0244, "step": 75750 }, { "epoch": 0.19, "learning_rate": 8.119532940761991e-05, "loss": 1.0251, "step": 75755 }, { "epoch": 0.19, "learning_rate": 8.11940711661382e-05, "loss": 1.0264, "step": 75760 }, { "epoch": 0.19, "learning_rate": 8.119281292465651e-05, "loss": 1.0242, "step": 75765 }, { "epoch": 0.19, "learning_rate": 8.11915546831748e-05, "loss": 1.0236, "step": 75770 }, { "epoch": 0.19, "learning_rate": 8.119029644169309e-05, "loss": 1.0244, "step": 75775 }, { "epoch": 0.19, "learning_rate": 8.118903820021138e-05, "loss": 1.0282, "step": 75780 }, { "epoch": 0.19, "learning_rate": 8.118777995872968e-05, "loss": 1.0256, "step": 75785 }, { "epoch": 0.19, "learning_rate": 8.118652171724798e-05, "loss": 1.0253, "step": 75790 }, { "epoch": 0.19, "learning_rate": 8.118526347576627e-05, "loss": 1.024, "step": 75795 }, { "epoch": 0.19, "learning_rate": 8.118400523428456e-05, "loss": 1.0274, "step": 75800 }, { "epoch": 0.19, "learning_rate": 8.118274699280286e-05, "loss": 1.0256, "step": 75805 }, { "epoch": 0.19, "learning_rate": 8.118148875132116e-05, "loss": 1.0249, "step": 75810 }, { "epoch": 0.19, "learning_rate": 8.118023050983945e-05, "loss": 1.0484, "step": 75815 }, { "epoch": 0.19, "learning_rate": 8.117897226835774e-05, "loss": 1.0217, "step": 75820 }, { "epoch": 0.19, "learning_rate": 8.117771402687604e-05, "loss": 1.0247, "step": 75825 }, { "epoch": 0.19, "learning_rate": 8.117645578539434e-05, "loss": 1.0244, "step": 75830 }, { "epoch": 0.19, "learning_rate": 8.117519754391263e-05, "loss": 1.0227, "step": 75835 }, { "epoch": 0.19, "learning_rate": 8.117393930243092e-05, "loss": 1.025, "step": 75840 }, { "epoch": 0.19, "learning_rate": 8.117268106094922e-05, "loss": 1.0245, "step": 75845 }, { "epoch": 0.19, "learning_rate": 8.117142281946751e-05, "loss": 1.0253, "step": 75850 }, { "epoch": 0.19, "learning_rate": 8.117016457798581e-05, "loss": 1.0251, "step": 75855 }, { "epoch": 0.19, "learning_rate": 8.11689063365041e-05, "loss": 1.0254, "step": 75860 }, { "epoch": 0.19, "learning_rate": 8.11676480950224e-05, "loss": 1.0263, "step": 75865 }, { "epoch": 0.19, "learning_rate": 8.116638985354069e-05, "loss": 1.0261, "step": 75870 }, { "epoch": 0.19, "learning_rate": 8.116513161205899e-05, "loss": 1.0265, "step": 75875 }, { "epoch": 0.19, "learning_rate": 8.116387337057728e-05, "loss": 1.0228, "step": 75880 }, { "epoch": 0.19, "learning_rate": 8.116261512909558e-05, "loss": 1.0259, "step": 75885 }, { "epoch": 0.19, "learning_rate": 8.116135688761387e-05, "loss": 1.0248, "step": 75890 }, { "epoch": 0.19, "learning_rate": 8.116009864613217e-05, "loss": 1.0247, "step": 75895 }, { "epoch": 0.19, "learning_rate": 8.115884040465046e-05, "loss": 1.0258, "step": 75900 }, { "epoch": 0.19, "learning_rate": 8.115758216316876e-05, "loss": 1.0257, "step": 75905 }, { "epoch": 0.19, "learning_rate": 8.115632392168705e-05, "loss": 1.0228, "step": 75910 }, { "epoch": 0.19, "learning_rate": 8.115506568020534e-05, "loss": 1.0241, "step": 75915 }, { "epoch": 0.19, "learning_rate": 8.115380743872364e-05, "loss": 1.026, "step": 75920 }, { "epoch": 0.19, "learning_rate": 8.115254919724194e-05, "loss": 1.0233, "step": 75925 }, { "epoch": 0.19, "learning_rate": 8.115129095576023e-05, "loss": 1.0232, "step": 75930 }, { "epoch": 0.19, "learning_rate": 8.115003271427852e-05, "loss": 1.0281, "step": 75935 }, { "epoch": 0.19, "learning_rate": 8.114877447279682e-05, "loss": 1.0246, "step": 75940 }, { "epoch": 0.19, "learning_rate": 8.114751623131512e-05, "loss": 1.0248, "step": 75945 }, { "epoch": 0.19, "learning_rate": 8.114625798983341e-05, "loss": 1.0261, "step": 75950 }, { "epoch": 0.19, "learning_rate": 8.11449997483517e-05, "loss": 1.0249, "step": 75955 }, { "epoch": 0.19, "learning_rate": 8.114374150687e-05, "loss": 1.0237, "step": 75960 }, { "epoch": 0.19, "learning_rate": 8.114248326538831e-05, "loss": 1.0245, "step": 75965 }, { "epoch": 0.19, "learning_rate": 8.11412250239066e-05, "loss": 1.0457, "step": 75970 }, { "epoch": 0.19, "learning_rate": 8.113996678242489e-05, "loss": 1.0255, "step": 75975 }, { "epoch": 0.19, "learning_rate": 8.113870854094318e-05, "loss": 1.023, "step": 75980 }, { "epoch": 0.19, "learning_rate": 8.113745029946148e-05, "loss": 1.0256, "step": 75985 }, { "epoch": 0.19, "learning_rate": 8.113619205797978e-05, "loss": 1.0242, "step": 75990 }, { "epoch": 0.19, "learning_rate": 8.113493381649807e-05, "loss": 1.0251, "step": 75995 }, { "epoch": 0.19, "learning_rate": 8.113367557501636e-05, "loss": 1.0256, "step": 76000 }, { "epoch": 0.19, "learning_rate": 8.113241733353466e-05, "loss": 1.0399, "step": 76005 }, { "epoch": 0.19, "learning_rate": 8.113115909205296e-05, "loss": 1.0267, "step": 76010 }, { "epoch": 0.19, "learning_rate": 8.112990085057125e-05, "loss": 1.0258, "step": 76015 }, { "epoch": 0.19, "learning_rate": 8.112864260908954e-05, "loss": 1.0243, "step": 76020 }, { "epoch": 0.19, "learning_rate": 8.112738436760783e-05, "loss": 1.025, "step": 76025 }, { "epoch": 0.19, "learning_rate": 8.112612612612614e-05, "loss": 1.0279, "step": 76030 }, { "epoch": 0.19, "learning_rate": 8.112486788464443e-05, "loss": 1.0282, "step": 76035 }, { "epoch": 0.19, "learning_rate": 8.112360964316272e-05, "loss": 1.0284, "step": 76040 }, { "epoch": 0.19, "learning_rate": 8.112235140168101e-05, "loss": 1.0265, "step": 76045 }, { "epoch": 0.19, "learning_rate": 8.11210931601993e-05, "loss": 1.0263, "step": 76050 }, { "epoch": 0.19, "learning_rate": 8.111983491871761e-05, "loss": 1.0241, "step": 76055 }, { "epoch": 0.19, "learning_rate": 8.11185766772359e-05, "loss": 1.0266, "step": 76060 }, { "epoch": 0.19, "learning_rate": 8.11173184357542e-05, "loss": 1.0231, "step": 76065 }, { "epoch": 0.19, "learning_rate": 8.111606019427249e-05, "loss": 1.0254, "step": 76070 }, { "epoch": 0.19, "learning_rate": 8.111480195279079e-05, "loss": 1.0249, "step": 76075 }, { "epoch": 0.19, "learning_rate": 8.111354371130908e-05, "loss": 1.0251, "step": 76080 }, { "epoch": 0.19, "learning_rate": 8.111228546982737e-05, "loss": 1.0251, "step": 76085 }, { "epoch": 0.19, "learning_rate": 8.111102722834567e-05, "loss": 1.0243, "step": 76090 }, { "epoch": 0.19, "learning_rate": 8.110976898686397e-05, "loss": 1.0261, "step": 76095 }, { "epoch": 0.19, "learning_rate": 8.110851074538226e-05, "loss": 1.0262, "step": 76100 }, { "epoch": 0.19, "learning_rate": 8.110725250390055e-05, "loss": 1.0242, "step": 76105 }, { "epoch": 0.19, "learning_rate": 8.110599426241885e-05, "loss": 1.0263, "step": 76110 }, { "epoch": 0.19, "learning_rate": 8.110473602093714e-05, "loss": 1.0259, "step": 76115 }, { "epoch": 0.19, "learning_rate": 8.110347777945544e-05, "loss": 1.0256, "step": 76120 }, { "epoch": 0.19, "learning_rate": 8.110221953797373e-05, "loss": 1.0258, "step": 76125 }, { "epoch": 0.19, "learning_rate": 8.110096129649203e-05, "loss": 1.0263, "step": 76130 }, { "epoch": 0.19, "learning_rate": 8.109970305501032e-05, "loss": 1.0266, "step": 76135 }, { "epoch": 0.19, "learning_rate": 8.109844481352862e-05, "loss": 1.0256, "step": 76140 }, { "epoch": 0.19, "learning_rate": 8.109718657204691e-05, "loss": 1.0265, "step": 76145 }, { "epoch": 0.19, "learning_rate": 8.10959283305652e-05, "loss": 1.0231, "step": 76150 }, { "epoch": 0.19, "learning_rate": 8.10946700890835e-05, "loss": 1.0241, "step": 76155 }, { "epoch": 0.19, "learning_rate": 8.10934118476018e-05, "loss": 1.0253, "step": 76160 }, { "epoch": 0.19, "learning_rate": 8.10921536061201e-05, "loss": 1.0266, "step": 76165 }, { "epoch": 0.19, "learning_rate": 8.109089536463839e-05, "loss": 1.027, "step": 76170 }, { "epoch": 0.19, "learning_rate": 8.108963712315668e-05, "loss": 1.0257, "step": 76175 }, { "epoch": 0.19, "learning_rate": 8.108837888167497e-05, "loss": 1.025, "step": 76180 }, { "epoch": 0.19, "learning_rate": 8.108712064019327e-05, "loss": 1.0273, "step": 76185 }, { "epoch": 0.19, "learning_rate": 8.108586239871157e-05, "loss": 1.0276, "step": 76190 }, { "epoch": 0.19, "learning_rate": 8.108460415722986e-05, "loss": 1.0233, "step": 76195 }, { "epoch": 0.19, "learning_rate": 8.108334591574815e-05, "loss": 1.0252, "step": 76200 }, { "epoch": 0.19, "learning_rate": 8.108208767426645e-05, "loss": 1.0261, "step": 76205 }, { "epoch": 0.19, "learning_rate": 8.108082943278475e-05, "loss": 1.0243, "step": 76210 }, { "epoch": 0.19, "learning_rate": 8.107957119130304e-05, "loss": 1.0412, "step": 76215 }, { "epoch": 0.19, "learning_rate": 8.107831294982133e-05, "loss": 1.0243, "step": 76220 }, { "epoch": 0.19, "learning_rate": 8.107705470833963e-05, "loss": 1.0244, "step": 76225 }, { "epoch": 0.19, "learning_rate": 8.107579646685793e-05, "loss": 1.0272, "step": 76230 }, { "epoch": 0.19, "learning_rate": 8.107453822537622e-05, "loss": 1.0239, "step": 76235 }, { "epoch": 0.19, "learning_rate": 8.107327998389451e-05, "loss": 1.0247, "step": 76240 }, { "epoch": 0.19, "learning_rate": 8.10720217424128e-05, "loss": 1.0239, "step": 76245 }, { "epoch": 0.19, "learning_rate": 8.10707635009311e-05, "loss": 1.0252, "step": 76250 }, { "epoch": 0.19, "learning_rate": 8.10695052594494e-05, "loss": 1.0244, "step": 76255 }, { "epoch": 0.19, "learning_rate": 8.106824701796769e-05, "loss": 1.0237, "step": 76260 }, { "epoch": 0.19, "learning_rate": 8.106698877648598e-05, "loss": 1.0253, "step": 76265 }, { "epoch": 0.19, "learning_rate": 8.106573053500429e-05, "loss": 1.0239, "step": 76270 }, { "epoch": 0.19, "learning_rate": 8.106447229352258e-05, "loss": 1.0252, "step": 76275 }, { "epoch": 0.19, "learning_rate": 8.106321405204087e-05, "loss": 1.0248, "step": 76280 }, { "epoch": 0.19, "learning_rate": 8.106195581055916e-05, "loss": 1.0257, "step": 76285 }, { "epoch": 0.19, "learning_rate": 8.106069756907747e-05, "loss": 1.0244, "step": 76290 }, { "epoch": 0.19, "learning_rate": 8.105943932759576e-05, "loss": 1.0254, "step": 76295 }, { "epoch": 0.19, "learning_rate": 8.105818108611405e-05, "loss": 1.0255, "step": 76300 }, { "epoch": 0.19, "learning_rate": 8.105692284463234e-05, "loss": 1.0254, "step": 76305 }, { "epoch": 0.19, "learning_rate": 8.105566460315063e-05, "loss": 1.0251, "step": 76310 }, { "epoch": 0.19, "learning_rate": 8.105440636166894e-05, "loss": 1.023, "step": 76315 }, { "epoch": 0.19, "learning_rate": 8.105314812018723e-05, "loss": 1.021, "step": 76320 }, { "epoch": 0.19, "learning_rate": 8.105188987870552e-05, "loss": 1.025, "step": 76325 }, { "epoch": 0.19, "learning_rate": 8.105063163722381e-05, "loss": 1.0258, "step": 76330 }, { "epoch": 0.19, "learning_rate": 8.104937339574212e-05, "loss": 1.0231, "step": 76335 }, { "epoch": 0.19, "learning_rate": 8.104811515426041e-05, "loss": 1.0449, "step": 76340 }, { "epoch": 0.19, "learning_rate": 8.10468569127787e-05, "loss": 1.0258, "step": 76345 }, { "epoch": 0.19, "learning_rate": 8.104559867129699e-05, "loss": 1.0248, "step": 76350 }, { "epoch": 0.19, "learning_rate": 8.10443404298153e-05, "loss": 1.0247, "step": 76355 }, { "epoch": 0.19, "learning_rate": 8.104308218833359e-05, "loss": 1.0255, "step": 76360 }, { "epoch": 0.19, "learning_rate": 8.104182394685188e-05, "loss": 1.0249, "step": 76365 }, { "epoch": 0.19, "learning_rate": 8.104056570537017e-05, "loss": 1.0469, "step": 76370 }, { "epoch": 0.19, "learning_rate": 8.103930746388846e-05, "loss": 1.0244, "step": 76375 }, { "epoch": 0.19, "learning_rate": 8.103804922240677e-05, "loss": 1.0247, "step": 76380 }, { "epoch": 0.19, "learning_rate": 8.103679098092506e-05, "loss": 1.0261, "step": 76385 }, { "epoch": 0.19, "learning_rate": 8.103553273944335e-05, "loss": 1.0251, "step": 76390 }, { "epoch": 0.19, "learning_rate": 8.103427449796164e-05, "loss": 1.0268, "step": 76395 }, { "epoch": 0.19, "learning_rate": 8.103301625647995e-05, "loss": 1.0234, "step": 76400 }, { "epoch": 0.19, "learning_rate": 8.103175801499824e-05, "loss": 1.0257, "step": 76405 }, { "epoch": 0.19, "learning_rate": 8.103049977351653e-05, "loss": 1.0257, "step": 76410 }, { "epoch": 0.19, "learning_rate": 8.102924153203482e-05, "loss": 1.0253, "step": 76415 }, { "epoch": 0.19, "learning_rate": 8.102798329055313e-05, "loss": 1.0238, "step": 76420 }, { "epoch": 0.19, "learning_rate": 8.102672504907142e-05, "loss": 1.0266, "step": 76425 }, { "epoch": 0.19, "learning_rate": 8.102546680758971e-05, "loss": 1.0247, "step": 76430 }, { "epoch": 0.19, "learning_rate": 8.1024208566108e-05, "loss": 1.0232, "step": 76435 }, { "epoch": 0.19, "learning_rate": 8.10229503246263e-05, "loss": 1.025, "step": 76440 }, { "epoch": 0.19, "learning_rate": 8.10216920831446e-05, "loss": 1.0265, "step": 76445 }, { "epoch": 0.19, "learning_rate": 8.102043384166289e-05, "loss": 1.0274, "step": 76450 }, { "epoch": 0.19, "learning_rate": 8.101917560018118e-05, "loss": 1.0252, "step": 76455 }, { "epoch": 0.19, "learning_rate": 8.101791735869948e-05, "loss": 1.0249, "step": 76460 }, { "epoch": 0.19, "learning_rate": 8.101665911721778e-05, "loss": 1.026, "step": 76465 }, { "epoch": 0.19, "learning_rate": 8.101540087573609e-05, "loss": 1.0254, "step": 76470 }, { "epoch": 0.19, "learning_rate": 8.101414263425438e-05, "loss": 1.0233, "step": 76475 }, { "epoch": 0.19, "learning_rate": 8.101288439277267e-05, "loss": 1.0241, "step": 76480 }, { "epoch": 0.19, "learning_rate": 8.101162615129096e-05, "loss": 1.0234, "step": 76485 }, { "epoch": 0.19, "learning_rate": 8.101036790980927e-05, "loss": 1.0289, "step": 76490 }, { "epoch": 0.19, "learning_rate": 8.100910966832756e-05, "loss": 1.0252, "step": 76495 }, { "epoch": 0.19, "learning_rate": 8.100785142684585e-05, "loss": 1.026, "step": 76500 }, { "epoch": 0.19, "learning_rate": 8.100659318536414e-05, "loss": 1.0257, "step": 76505 }, { "epoch": 0.19, "learning_rate": 8.100533494388243e-05, "loss": 1.0246, "step": 76510 }, { "epoch": 0.19, "learning_rate": 8.100407670240074e-05, "loss": 1.0267, "step": 76515 }, { "epoch": 0.19, "learning_rate": 8.100281846091903e-05, "loss": 1.025, "step": 76520 }, { "epoch": 0.19, "learning_rate": 8.100156021943732e-05, "loss": 1.026, "step": 76525 }, { "epoch": 0.19, "learning_rate": 8.100030197795561e-05, "loss": 1.0249, "step": 76530 }, { "epoch": 0.19, "learning_rate": 8.099904373647392e-05, "loss": 1.025, "step": 76535 }, { "epoch": 0.19, "learning_rate": 8.099778549499221e-05, "loss": 1.0256, "step": 76540 }, { "epoch": 0.19, "learning_rate": 8.09965272535105e-05, "loss": 1.0245, "step": 76545 }, { "epoch": 0.19, "learning_rate": 8.099526901202879e-05, "loss": 1.0515, "step": 76550 }, { "epoch": 0.19, "learning_rate": 8.09940107705471e-05, "loss": 1.0237, "step": 76555 }, { "epoch": 0.19, "learning_rate": 8.099275252906539e-05, "loss": 1.0262, "step": 76560 }, { "epoch": 0.19, "learning_rate": 8.099149428758368e-05, "loss": 1.0251, "step": 76565 }, { "epoch": 0.19, "learning_rate": 8.099023604610197e-05, "loss": 1.0278, "step": 76570 }, { "epoch": 0.19, "learning_rate": 8.098897780462026e-05, "loss": 1.0264, "step": 76575 }, { "epoch": 0.19, "learning_rate": 8.098771956313857e-05, "loss": 1.0465, "step": 76580 }, { "epoch": 0.19, "learning_rate": 8.098646132165686e-05, "loss": 1.025, "step": 76585 }, { "epoch": 0.19, "learning_rate": 8.098520308017515e-05, "loss": 1.0248, "step": 76590 }, { "epoch": 0.19, "learning_rate": 8.098394483869344e-05, "loss": 1.0251, "step": 76595 }, { "epoch": 0.19, "learning_rate": 8.098268659721175e-05, "loss": 1.0239, "step": 76600 }, { "epoch": 0.19, "learning_rate": 8.098142835573004e-05, "loss": 1.0267, "step": 76605 }, { "epoch": 0.19, "learning_rate": 8.098017011424833e-05, "loss": 1.0267, "step": 76610 }, { "epoch": 0.19, "learning_rate": 8.097891187276662e-05, "loss": 1.0254, "step": 76615 }, { "epoch": 0.19, "learning_rate": 8.097765363128493e-05, "loss": 1.0242, "step": 76620 }, { "epoch": 0.19, "learning_rate": 8.097639538980322e-05, "loss": 1.0271, "step": 76625 }, { "epoch": 0.19, "learning_rate": 8.097513714832151e-05, "loss": 1.0281, "step": 76630 }, { "epoch": 0.19, "learning_rate": 8.09738789068398e-05, "loss": 1.0256, "step": 76635 }, { "epoch": 0.19, "learning_rate": 8.09726206653581e-05, "loss": 1.0257, "step": 76640 }, { "epoch": 0.19, "learning_rate": 8.09713624238764e-05, "loss": 1.0243, "step": 76645 }, { "epoch": 0.19, "learning_rate": 8.097010418239469e-05, "loss": 1.0464, "step": 76650 }, { "epoch": 0.19, "learning_rate": 8.096884594091298e-05, "loss": 1.0248, "step": 76655 }, { "epoch": 0.19, "learning_rate": 8.096758769943127e-05, "loss": 1.0239, "step": 76660 }, { "epoch": 0.19, "learning_rate": 8.096632945794958e-05, "loss": 1.0257, "step": 76665 }, { "epoch": 0.19, "learning_rate": 8.096507121646787e-05, "loss": 1.0253, "step": 76670 }, { "epoch": 0.19, "learning_rate": 8.096381297498616e-05, "loss": 1.0238, "step": 76675 }, { "epoch": 0.19, "learning_rate": 8.096255473350445e-05, "loss": 1.024, "step": 76680 }, { "epoch": 0.19, "learning_rate": 8.096129649202276e-05, "loss": 1.0448, "step": 76685 }, { "epoch": 0.19, "learning_rate": 8.096003825054105e-05, "loss": 1.0244, "step": 76690 }, { "epoch": 0.19, "learning_rate": 8.095878000905934e-05, "loss": 1.0247, "step": 76695 }, { "epoch": 0.19, "learning_rate": 8.095752176757763e-05, "loss": 1.0252, "step": 76700 }, { "epoch": 0.19, "learning_rate": 8.095626352609593e-05, "loss": 1.0256, "step": 76705 }, { "epoch": 0.19, "learning_rate": 8.095500528461423e-05, "loss": 1.0277, "step": 76710 }, { "epoch": 0.19, "learning_rate": 8.095374704313252e-05, "loss": 1.0227, "step": 76715 }, { "epoch": 0.19, "learning_rate": 8.095248880165081e-05, "loss": 1.0246, "step": 76720 }, { "epoch": 0.19, "learning_rate": 8.09512305601691e-05, "loss": 1.0253, "step": 76725 }, { "epoch": 0.19, "learning_rate": 8.094997231868741e-05, "loss": 1.0252, "step": 76730 }, { "epoch": 0.19, "learning_rate": 8.09487140772057e-05, "loss": 1.0264, "step": 76735 }, { "epoch": 0.19, "learning_rate": 8.0947455835724e-05, "loss": 1.0255, "step": 76740 }, { "epoch": 0.19, "learning_rate": 8.094619759424229e-05, "loss": 1.0233, "step": 76745 }, { "epoch": 0.19, "learning_rate": 8.094493935276059e-05, "loss": 1.0263, "step": 76750 }, { "epoch": 0.19, "learning_rate": 8.094368111127888e-05, "loss": 1.0224, "step": 76755 }, { "epoch": 0.19, "learning_rate": 8.094242286979717e-05, "loss": 1.024, "step": 76760 }, { "epoch": 0.19, "learning_rate": 8.094116462831547e-05, "loss": 1.0241, "step": 76765 }, { "epoch": 0.19, "learning_rate": 8.093990638683376e-05, "loss": 1.0266, "step": 76770 }, { "epoch": 0.19, "learning_rate": 8.093864814535206e-05, "loss": 1.024, "step": 76775 }, { "epoch": 0.19, "learning_rate": 8.093738990387035e-05, "loss": 1.0269, "step": 76780 }, { "epoch": 0.19, "learning_rate": 8.093613166238865e-05, "loss": 1.0226, "step": 76785 }, { "epoch": 0.19, "learning_rate": 8.093487342090694e-05, "loss": 1.0263, "step": 76790 }, { "epoch": 0.19, "learning_rate": 8.093361517942524e-05, "loss": 1.0255, "step": 76795 }, { "epoch": 0.19, "learning_rate": 8.093235693794353e-05, "loss": 1.027, "step": 76800 }, { "epoch": 0.19, "learning_rate": 8.093109869646183e-05, "loss": 1.0233, "step": 76805 }, { "epoch": 0.19, "learning_rate": 8.092984045498012e-05, "loss": 1.0254, "step": 76810 }, { "epoch": 0.19, "learning_rate": 8.092858221349842e-05, "loss": 1.0248, "step": 76815 }, { "epoch": 0.19, "learning_rate": 8.092732397201671e-05, "loss": 1.0254, "step": 76820 }, { "epoch": 0.19, "learning_rate": 8.0926065730535e-05, "loss": 1.0244, "step": 76825 }, { "epoch": 0.19, "learning_rate": 8.09248074890533e-05, "loss": 1.0267, "step": 76830 }, { "epoch": 0.19, "learning_rate": 8.092354924757159e-05, "loss": 1.0262, "step": 76835 }, { "epoch": 0.19, "learning_rate": 8.09222910060899e-05, "loss": 1.0249, "step": 76840 }, { "epoch": 0.19, "learning_rate": 8.092103276460819e-05, "loss": 1.0235, "step": 76845 }, { "epoch": 0.19, "learning_rate": 8.091977452312648e-05, "loss": 1.0248, "step": 76850 }, { "epoch": 0.19, "learning_rate": 8.091851628164477e-05, "loss": 1.0266, "step": 76855 }, { "epoch": 0.19, "learning_rate": 8.091725804016307e-05, "loss": 1.0231, "step": 76860 }, { "epoch": 0.19, "learning_rate": 8.091599979868137e-05, "loss": 1.0246, "step": 76865 }, { "epoch": 0.19, "learning_rate": 8.091474155719966e-05, "loss": 1.0266, "step": 76870 }, { "epoch": 0.19, "learning_rate": 8.091348331571795e-05, "loss": 1.0261, "step": 76875 }, { "epoch": 0.19, "learning_rate": 8.091222507423625e-05, "loss": 1.0259, "step": 76880 }, { "epoch": 0.19, "learning_rate": 8.091096683275455e-05, "loss": 1.0253, "step": 76885 }, { "epoch": 0.19, "learning_rate": 8.090970859127284e-05, "loss": 1.0243, "step": 76890 }, { "epoch": 0.19, "learning_rate": 8.090845034979113e-05, "loss": 1.0249, "step": 76895 }, { "epoch": 0.19, "learning_rate": 8.090719210830942e-05, "loss": 1.0261, "step": 76900 }, { "epoch": 0.19, "learning_rate": 8.090593386682773e-05, "loss": 1.0453, "step": 76905 }, { "epoch": 0.19, "learning_rate": 8.090467562534602e-05, "loss": 1.0241, "step": 76910 }, { "epoch": 0.19, "learning_rate": 8.090341738386431e-05, "loss": 1.0268, "step": 76915 }, { "epoch": 0.19, "learning_rate": 8.09021591423826e-05, "loss": 1.025, "step": 76920 }, { "epoch": 0.19, "learning_rate": 8.09009009009009e-05, "loss": 1.0251, "step": 76925 }, { "epoch": 0.19, "learning_rate": 8.08996426594192e-05, "loss": 1.0257, "step": 76930 }, { "epoch": 0.19, "learning_rate": 8.089838441793749e-05, "loss": 1.0271, "step": 76935 }, { "epoch": 0.19, "learning_rate": 8.089712617645578e-05, "loss": 1.027, "step": 76940 }, { "epoch": 0.19, "learning_rate": 8.089586793497409e-05, "loss": 1.0273, "step": 76945 }, { "epoch": 0.19, "learning_rate": 8.089460969349238e-05, "loss": 1.0243, "step": 76950 }, { "epoch": 0.19, "learning_rate": 8.089335145201067e-05, "loss": 1.024, "step": 76955 }, { "epoch": 0.19, "learning_rate": 8.089209321052896e-05, "loss": 1.0247, "step": 76960 }, { "epoch": 0.19, "learning_rate": 8.089083496904727e-05, "loss": 1.0246, "step": 76965 }, { "epoch": 0.19, "learning_rate": 8.088957672756557e-05, "loss": 1.0237, "step": 76970 }, { "epoch": 0.19, "learning_rate": 8.088831848608386e-05, "loss": 1.0246, "step": 76975 }, { "epoch": 0.19, "learning_rate": 8.088706024460215e-05, "loss": 1.0239, "step": 76980 }, { "epoch": 0.19, "learning_rate": 8.088580200312045e-05, "loss": 1.0263, "step": 76985 }, { "epoch": 0.19, "learning_rate": 8.088454376163874e-05, "loss": 1.0254, "step": 76990 }, { "epoch": 0.19, "learning_rate": 8.088328552015704e-05, "loss": 1.0258, "step": 76995 }, { "epoch": 0.19, "learning_rate": 8.088202727867533e-05, "loss": 1.0258, "step": 77000 }, { "epoch": 0.19, "learning_rate": 8.088076903719363e-05, "loss": 1.0254, "step": 77005 }, { "epoch": 0.19, "learning_rate": 8.087951079571192e-05, "loss": 1.0254, "step": 77010 }, { "epoch": 0.19, "learning_rate": 8.087825255423022e-05, "loss": 1.025, "step": 77015 }, { "epoch": 0.19, "learning_rate": 8.087699431274851e-05, "loss": 1.0245, "step": 77020 }, { "epoch": 0.19, "learning_rate": 8.08757360712668e-05, "loss": 1.0244, "step": 77025 }, { "epoch": 0.19, "learning_rate": 8.08744778297851e-05, "loss": 1.0252, "step": 77030 }, { "epoch": 0.19, "learning_rate": 8.087321958830339e-05, "loss": 1.0285, "step": 77035 }, { "epoch": 0.19, "learning_rate": 8.08719613468217e-05, "loss": 1.0245, "step": 77040 }, { "epoch": 0.19, "learning_rate": 8.087070310533998e-05, "loss": 1.0258, "step": 77045 }, { "epoch": 0.19, "learning_rate": 8.086944486385828e-05, "loss": 1.0258, "step": 77050 }, { "epoch": 0.19, "learning_rate": 8.086818662237657e-05, "loss": 1.0236, "step": 77055 }, { "epoch": 0.19, "learning_rate": 8.086692838089487e-05, "loss": 1.0249, "step": 77060 }, { "epoch": 0.19, "learning_rate": 8.086567013941316e-05, "loss": 1.0249, "step": 77065 }, { "epoch": 0.19, "learning_rate": 8.086441189793146e-05, "loss": 1.0257, "step": 77070 }, { "epoch": 0.19, "learning_rate": 8.086315365644975e-05, "loss": 1.0235, "step": 77075 }, { "epoch": 0.19, "learning_rate": 8.086189541496805e-05, "loss": 1.0256, "step": 77080 }, { "epoch": 0.19, "learning_rate": 8.086063717348634e-05, "loss": 1.023, "step": 77085 }, { "epoch": 0.19, "learning_rate": 8.085937893200464e-05, "loss": 1.023, "step": 77090 }, { "epoch": 0.19, "learning_rate": 8.085812069052293e-05, "loss": 1.0256, "step": 77095 }, { "epoch": 0.19, "learning_rate": 8.085686244904122e-05, "loss": 1.0236, "step": 77100 }, { "epoch": 0.19, "learning_rate": 8.085560420755952e-05, "loss": 1.0242, "step": 77105 }, { "epoch": 0.19, "learning_rate": 8.085434596607782e-05, "loss": 1.0237, "step": 77110 }, { "epoch": 0.19, "learning_rate": 8.085308772459611e-05, "loss": 1.0222, "step": 77115 }, { "epoch": 0.19, "learning_rate": 8.08518294831144e-05, "loss": 1.0255, "step": 77120 }, { "epoch": 0.19, "learning_rate": 8.08505712416327e-05, "loss": 1.0237, "step": 77125 }, { "epoch": 0.19, "learning_rate": 8.0849313000151e-05, "loss": 1.0243, "step": 77130 }, { "epoch": 0.19, "learning_rate": 8.084805475866929e-05, "loss": 1.0247, "step": 77135 }, { "epoch": 0.19, "learning_rate": 8.084679651718758e-05, "loss": 1.0248, "step": 77140 }, { "epoch": 0.19, "learning_rate": 8.084553827570588e-05, "loss": 1.026, "step": 77145 }, { "epoch": 0.19, "learning_rate": 8.084428003422418e-05, "loss": 1.023, "step": 77150 }, { "epoch": 0.19, "learning_rate": 8.084302179274247e-05, "loss": 1.0244, "step": 77155 }, { "epoch": 0.19, "learning_rate": 8.084176355126076e-05, "loss": 1.0249, "step": 77160 }, { "epoch": 0.19, "learning_rate": 8.084050530977905e-05, "loss": 1.0257, "step": 77165 }, { "epoch": 0.19, "learning_rate": 8.083924706829736e-05, "loss": 1.024, "step": 77170 }, { "epoch": 0.19, "learning_rate": 8.083798882681565e-05, "loss": 1.0259, "step": 77175 }, { "epoch": 0.19, "learning_rate": 8.083673058533394e-05, "loss": 1.0269, "step": 77180 }, { "epoch": 0.19, "learning_rate": 8.083547234385223e-05, "loss": 1.0268, "step": 77185 }, { "epoch": 0.19, "learning_rate": 8.083421410237054e-05, "loss": 1.0248, "step": 77190 }, { "epoch": 0.19, "learning_rate": 8.083295586088883e-05, "loss": 1.0251, "step": 77195 }, { "epoch": 0.19, "learning_rate": 8.083169761940712e-05, "loss": 1.0243, "step": 77200 }, { "epoch": 0.19, "learning_rate": 8.083043937792541e-05, "loss": 1.0252, "step": 77205 }, { "epoch": 0.19, "learning_rate": 8.082918113644372e-05, "loss": 1.0246, "step": 77210 }, { "epoch": 0.19, "learning_rate": 8.082792289496201e-05, "loss": 1.0237, "step": 77215 }, { "epoch": 0.19, "learning_rate": 8.08266646534803e-05, "loss": 1.0254, "step": 77220 }, { "epoch": 0.19, "learning_rate": 8.082540641199859e-05, "loss": 1.0254, "step": 77225 }, { "epoch": 0.19, "learning_rate": 8.082414817051688e-05, "loss": 1.0252, "step": 77230 }, { "epoch": 0.19, "learning_rate": 8.082288992903519e-05, "loss": 1.0239, "step": 77235 }, { "epoch": 0.19, "learning_rate": 8.082163168755348e-05, "loss": 1.0233, "step": 77240 }, { "epoch": 0.19, "learning_rate": 8.082062509436812e-05, "loss": 1.0253, "step": 77245 }, { "epoch": 0.19, "learning_rate": 8.081936685288641e-05, "loss": 1.0252, "step": 77250 }, { "epoch": 0.19, "learning_rate": 8.081810861140471e-05, "loss": 1.0265, "step": 77255 }, { "epoch": 0.19, "learning_rate": 8.0816850369923e-05, "loss": 1.0254, "step": 77260 }, { "epoch": 0.19, "learning_rate": 8.08155921284413e-05, "loss": 1.0263, "step": 77265 }, { "epoch": 0.19, "learning_rate": 8.081433388695959e-05, "loss": 1.0243, "step": 77270 }, { "epoch": 0.19, "learning_rate": 8.081307564547789e-05, "loss": 1.0222, "step": 77275 }, { "epoch": 0.19, "learning_rate": 8.081181740399618e-05, "loss": 1.0246, "step": 77280 }, { "epoch": 0.19, "learning_rate": 8.081055916251448e-05, "loss": 1.025, "step": 77285 }, { "epoch": 0.19, "learning_rate": 8.080930092103277e-05, "loss": 1.0277, "step": 77290 }, { "epoch": 0.19, "learning_rate": 8.080804267955107e-05, "loss": 1.0251, "step": 77295 }, { "epoch": 0.19, "learning_rate": 8.080678443806936e-05, "loss": 1.0248, "step": 77300 }, { "epoch": 0.19, "learning_rate": 8.080552619658766e-05, "loss": 1.0254, "step": 77305 }, { "epoch": 0.19, "learning_rate": 8.080426795510595e-05, "loss": 1.0233, "step": 77310 }, { "epoch": 0.19, "learning_rate": 8.080300971362424e-05, "loss": 1.0254, "step": 77315 }, { "epoch": 0.19, "learning_rate": 8.080175147214254e-05, "loss": 1.0245, "step": 77320 }, { "epoch": 0.19, "learning_rate": 8.080049323066083e-05, "loss": 1.025, "step": 77325 }, { "epoch": 0.19, "learning_rate": 8.079923498917913e-05, "loss": 1.0235, "step": 77330 }, { "epoch": 0.19, "learning_rate": 8.079797674769742e-05, "loss": 1.0267, "step": 77335 }, { "epoch": 0.19, "learning_rate": 8.079671850621572e-05, "loss": 1.025, "step": 77340 }, { "epoch": 0.19, "learning_rate": 8.079546026473401e-05, "loss": 1.0477, "step": 77345 }, { "epoch": 0.19, "learning_rate": 8.07942020232523e-05, "loss": 1.0256, "step": 77350 }, { "epoch": 0.19, "learning_rate": 8.07929437817706e-05, "loss": 1.0248, "step": 77355 }, { "epoch": 0.19, "learning_rate": 8.07916855402889e-05, "loss": 1.025, "step": 77360 }, { "epoch": 0.19, "learning_rate": 8.07904272988072e-05, "loss": 1.0264, "step": 77365 }, { "epoch": 0.19, "learning_rate": 8.078916905732549e-05, "loss": 1.0249, "step": 77370 }, { "epoch": 0.19, "learning_rate": 8.078791081584378e-05, "loss": 1.0231, "step": 77375 }, { "epoch": 0.19, "learning_rate": 8.078665257436207e-05, "loss": 1.0234, "step": 77380 }, { "epoch": 0.19, "learning_rate": 8.078539433288037e-05, "loss": 1.0231, "step": 77385 }, { "epoch": 0.19, "learning_rate": 8.078413609139867e-05, "loss": 1.0258, "step": 77390 }, { "epoch": 0.19, "learning_rate": 8.078287784991696e-05, "loss": 1.0253, "step": 77395 }, { "epoch": 0.19, "learning_rate": 8.078161960843525e-05, "loss": 1.0233, "step": 77400 }, { "epoch": 0.19, "learning_rate": 8.078036136695355e-05, "loss": 1.0274, "step": 77405 }, { "epoch": 0.19, "learning_rate": 8.077910312547185e-05, "loss": 1.0232, "step": 77410 }, { "epoch": 0.19, "learning_rate": 8.077784488399014e-05, "loss": 1.0262, "step": 77415 }, { "epoch": 0.19, "learning_rate": 8.077658664250843e-05, "loss": 1.0239, "step": 77420 }, { "epoch": 0.19, "learning_rate": 8.077532840102673e-05, "loss": 1.0281, "step": 77425 }, { "epoch": 0.19, "learning_rate": 8.077407015954503e-05, "loss": 1.027, "step": 77430 }, { "epoch": 0.19, "learning_rate": 8.077281191806332e-05, "loss": 1.0258, "step": 77435 }, { "epoch": 0.19, "learning_rate": 8.077155367658161e-05, "loss": 1.0261, "step": 77440 }, { "epoch": 0.19, "learning_rate": 8.07702954350999e-05, "loss": 1.0237, "step": 77445 }, { "epoch": 0.19, "learning_rate": 8.07690371936182e-05, "loss": 1.0258, "step": 77450 }, { "epoch": 0.19, "learning_rate": 8.07677789521365e-05, "loss": 1.0244, "step": 77455 }, { "epoch": 0.19, "learning_rate": 8.076652071065479e-05, "loss": 1.0245, "step": 77460 }, { "epoch": 0.19, "learning_rate": 8.076526246917308e-05, "loss": 1.0248, "step": 77465 }, { "epoch": 0.19, "learning_rate": 8.076400422769139e-05, "loss": 1.025, "step": 77470 }, { "epoch": 0.19, "learning_rate": 8.076274598620968e-05, "loss": 1.026, "step": 77475 }, { "epoch": 0.19, "learning_rate": 8.076148774472797e-05, "loss": 1.0237, "step": 77480 }, { "epoch": 0.19, "learning_rate": 8.076022950324626e-05, "loss": 1.0241, "step": 77485 }, { "epoch": 0.19, "learning_rate": 8.075897126176457e-05, "loss": 1.049, "step": 77490 }, { "epoch": 0.19, "learning_rate": 8.075771302028286e-05, "loss": 1.023, "step": 77495 }, { "epoch": 0.19, "learning_rate": 8.075645477880115e-05, "loss": 1.025, "step": 77500 }, { "epoch": 0.19, "learning_rate": 8.075519653731944e-05, "loss": 1.0259, "step": 77505 }, { "epoch": 0.19, "learning_rate": 8.075393829583773e-05, "loss": 1.0266, "step": 77510 }, { "epoch": 0.19, "learning_rate": 8.075268005435604e-05, "loss": 1.0264, "step": 77515 }, { "epoch": 0.19, "learning_rate": 8.075142181287433e-05, "loss": 1.0269, "step": 77520 }, { "epoch": 0.19, "learning_rate": 8.075016357139262e-05, "loss": 1.0242, "step": 77525 }, { "epoch": 0.19, "learning_rate": 8.074890532991091e-05, "loss": 1.0228, "step": 77530 }, { "epoch": 0.19, "learning_rate": 8.074764708842922e-05, "loss": 1.0237, "step": 77535 }, { "epoch": 0.19, "learning_rate": 8.074638884694751e-05, "loss": 1.0247, "step": 77540 }, { "epoch": 0.19, "learning_rate": 8.07451306054658e-05, "loss": 1.0249, "step": 77545 }, { "epoch": 0.19, "learning_rate": 8.074387236398409e-05, "loss": 1.0275, "step": 77550 }, { "epoch": 0.19, "learning_rate": 8.07426141225024e-05, "loss": 1.026, "step": 77555 }, { "epoch": 0.19, "learning_rate": 8.074135588102069e-05, "loss": 1.0209, "step": 77560 }, { "epoch": 0.19, "learning_rate": 8.074009763953898e-05, "loss": 1.0244, "step": 77565 }, { "epoch": 0.19, "learning_rate": 8.073883939805727e-05, "loss": 1.0236, "step": 77570 }, { "epoch": 0.19, "learning_rate": 8.073758115657556e-05, "loss": 1.0243, "step": 77575 }, { "epoch": 0.19, "learning_rate": 8.073632291509387e-05, "loss": 1.0247, "step": 77580 }, { "epoch": 0.19, "learning_rate": 8.073506467361216e-05, "loss": 1.0236, "step": 77585 }, { "epoch": 0.19, "learning_rate": 8.073380643213045e-05, "loss": 1.0261, "step": 77590 }, { "epoch": 0.19, "learning_rate": 8.073254819064874e-05, "loss": 1.0247, "step": 77595 }, { "epoch": 0.19, "learning_rate": 8.073128994916705e-05, "loss": 1.0242, "step": 77600 }, { "epoch": 0.19, "learning_rate": 8.073003170768534e-05, "loss": 1.0245, "step": 77605 }, { "epoch": 0.19, "learning_rate": 8.072877346620363e-05, "loss": 1.026, "step": 77610 }, { "epoch": 0.19, "learning_rate": 8.072751522472192e-05, "loss": 1.0235, "step": 77615 }, { "epoch": 0.19, "learning_rate": 8.072625698324023e-05, "loss": 1.0283, "step": 77620 }, { "epoch": 0.19, "learning_rate": 8.072499874175852e-05, "loss": 1.0244, "step": 77625 }, { "epoch": 0.19, "learning_rate": 8.072374050027681e-05, "loss": 1.0415, "step": 77630 }, { "epoch": 0.19, "learning_rate": 8.07224822587951e-05, "loss": 1.026, "step": 77635 }, { "epoch": 0.19, "learning_rate": 8.07212240173134e-05, "loss": 1.0266, "step": 77640 }, { "epoch": 0.19, "learning_rate": 8.07199657758317e-05, "loss": 1.0245, "step": 77645 }, { "epoch": 0.19, "learning_rate": 8.071870753434999e-05, "loss": 1.0239, "step": 77650 }, { "epoch": 0.19, "learning_rate": 8.071744929286828e-05, "loss": 1.0229, "step": 77655 }, { "epoch": 0.19, "learning_rate": 8.071619105138658e-05, "loss": 1.0266, "step": 77660 }, { "epoch": 0.19, "learning_rate": 8.071493280990488e-05, "loss": 1.0251, "step": 77665 }, { "epoch": 0.19, "learning_rate": 8.071367456842319e-05, "loss": 1.0257, "step": 77670 }, { "epoch": 0.19, "learning_rate": 8.071241632694148e-05, "loss": 1.0246, "step": 77675 }, { "epoch": 0.19, "learning_rate": 8.071115808545977e-05, "loss": 1.0258, "step": 77680 }, { "epoch": 0.2, "learning_rate": 8.070989984397806e-05, "loss": 1.0261, "step": 77685 }, { "epoch": 0.2, "learning_rate": 8.070864160249637e-05, "loss": 1.0254, "step": 77690 }, { "epoch": 0.2, "learning_rate": 8.070738336101466e-05, "loss": 1.0225, "step": 77695 }, { "epoch": 0.2, "learning_rate": 8.070612511953295e-05, "loss": 1.0247, "step": 77700 }, { "epoch": 0.2, "learning_rate": 8.070486687805124e-05, "loss": 1.0244, "step": 77705 }, { "epoch": 0.2, "learning_rate": 8.070360863656955e-05, "loss": 1.0266, "step": 77710 }, { "epoch": 0.2, "learning_rate": 8.070235039508784e-05, "loss": 1.023, "step": 77715 }, { "epoch": 0.2, "learning_rate": 8.070109215360613e-05, "loss": 1.0246, "step": 77720 }, { "epoch": 0.2, "learning_rate": 8.069983391212442e-05, "loss": 1.0235, "step": 77725 }, { "epoch": 0.2, "learning_rate": 8.069857567064271e-05, "loss": 1.0255, "step": 77730 }, { "epoch": 0.2, "learning_rate": 8.069731742916102e-05, "loss": 1.024, "step": 77735 }, { "epoch": 0.2, "learning_rate": 8.069605918767931e-05, "loss": 1.0266, "step": 77740 }, { "epoch": 0.2, "learning_rate": 8.06948009461976e-05, "loss": 1.0213, "step": 77745 }, { "epoch": 0.2, "learning_rate": 8.069354270471589e-05, "loss": 1.0243, "step": 77750 }, { "epoch": 0.2, "learning_rate": 8.06922844632342e-05, "loss": 1.0253, "step": 77755 }, { "epoch": 0.2, "learning_rate": 8.069102622175249e-05, "loss": 1.0254, "step": 77760 }, { "epoch": 0.2, "learning_rate": 8.068976798027078e-05, "loss": 1.0251, "step": 77765 }, { "epoch": 0.2, "learning_rate": 8.068850973878907e-05, "loss": 1.0262, "step": 77770 }, { "epoch": 0.2, "learning_rate": 8.068725149730736e-05, "loss": 1.0246, "step": 77775 }, { "epoch": 0.2, "learning_rate": 8.068599325582567e-05, "loss": 1.027, "step": 77780 }, { "epoch": 0.2, "learning_rate": 8.068473501434396e-05, "loss": 1.0237, "step": 77785 }, { "epoch": 0.2, "learning_rate": 8.068347677286225e-05, "loss": 1.028, "step": 77790 }, { "epoch": 0.2, "learning_rate": 8.068221853138054e-05, "loss": 1.0246, "step": 77795 }, { "epoch": 0.2, "learning_rate": 8.068096028989885e-05, "loss": 1.0238, "step": 77800 }, { "epoch": 0.2, "learning_rate": 8.067970204841714e-05, "loss": 1.0268, "step": 77805 }, { "epoch": 0.2, "learning_rate": 8.067844380693543e-05, "loss": 1.0258, "step": 77810 }, { "epoch": 0.2, "learning_rate": 8.067718556545372e-05, "loss": 1.0264, "step": 77815 }, { "epoch": 0.2, "learning_rate": 8.067617897226836e-05, "loss": 1.0259, "step": 77820 }, { "epoch": 0.2, "learning_rate": 8.0675172379083e-05, "loss": 1.0259, "step": 77825 }, { "epoch": 0.2, "learning_rate": 8.067391413760129e-05, "loss": 1.0271, "step": 77830 }, { "epoch": 0.2, "learning_rate": 8.067265589611959e-05, "loss": 1.0247, "step": 77835 }, { "epoch": 0.2, "learning_rate": 8.067139765463788e-05, "loss": 1.023, "step": 77840 }, { "epoch": 0.2, "learning_rate": 8.067013941315618e-05, "loss": 1.0256, "step": 77845 }, { "epoch": 0.2, "learning_rate": 8.066888117167447e-05, "loss": 1.0258, "step": 77850 }, { "epoch": 0.2, "learning_rate": 8.066787457848912e-05, "loss": 1.0232, "step": 77855 }, { "epoch": 0.2, "learning_rate": 8.066661633700741e-05, "loss": 1.0252, "step": 77860 }, { "epoch": 0.2, "learning_rate": 8.06653580955257e-05, "loss": 1.0263, "step": 77865 }, { "epoch": 0.2, "learning_rate": 8.066409985404399e-05, "loss": 1.0286, "step": 77870 }, { "epoch": 0.2, "learning_rate": 8.06628416125623e-05, "loss": 1.0245, "step": 77875 }, { "epoch": 0.2, "learning_rate": 8.066158337108059e-05, "loss": 1.0235, "step": 77880 }, { "epoch": 0.2, "learning_rate": 8.066032512959888e-05, "loss": 1.0238, "step": 77885 }, { "epoch": 0.2, "learning_rate": 8.065906688811717e-05, "loss": 1.0261, "step": 77890 }, { "epoch": 0.2, "learning_rate": 8.065780864663546e-05, "loss": 1.0258, "step": 77895 }, { "epoch": 0.2, "learning_rate": 8.065655040515377e-05, "loss": 1.0232, "step": 77900 }, { "epoch": 0.2, "learning_rate": 8.065529216367206e-05, "loss": 1.0246, "step": 77905 }, { "epoch": 0.2, "learning_rate": 8.065403392219035e-05, "loss": 1.0257, "step": 77910 }, { "epoch": 0.2, "learning_rate": 8.065277568070864e-05, "loss": 1.0237, "step": 77915 }, { "epoch": 0.2, "learning_rate": 8.065151743922695e-05, "loss": 1.0248, "step": 77920 }, { "epoch": 0.2, "learning_rate": 8.065025919774524e-05, "loss": 1.0253, "step": 77925 }, { "epoch": 0.2, "learning_rate": 8.064900095626353e-05, "loss": 1.0256, "step": 77930 }, { "epoch": 0.2, "learning_rate": 8.064774271478182e-05, "loss": 1.0238, "step": 77935 }, { "epoch": 0.2, "learning_rate": 8.064648447330013e-05, "loss": 1.0273, "step": 77940 }, { "epoch": 0.2, "learning_rate": 8.064522623181842e-05, "loss": 1.0235, "step": 77945 }, { "epoch": 0.2, "learning_rate": 8.064396799033671e-05, "loss": 1.0237, "step": 77950 }, { "epoch": 0.2, "learning_rate": 8.0642709748855e-05, "loss": 1.0252, "step": 77955 }, { "epoch": 0.2, "learning_rate": 8.06414515073733e-05, "loss": 1.0491, "step": 77960 }, { "epoch": 0.2, "learning_rate": 8.06401932658916e-05, "loss": 1.0267, "step": 77965 }, { "epoch": 0.2, "learning_rate": 8.063893502440989e-05, "loss": 1.0235, "step": 77970 }, { "epoch": 0.2, "learning_rate": 8.063767678292818e-05, "loss": 1.0243, "step": 77975 }, { "epoch": 0.2, "learning_rate": 8.063641854144647e-05, "loss": 1.0251, "step": 77980 }, { "epoch": 0.2, "learning_rate": 8.063516029996478e-05, "loss": 1.0254, "step": 77985 }, { "epoch": 0.2, "learning_rate": 8.063390205848307e-05, "loss": 1.0276, "step": 77990 }, { "epoch": 0.2, "learning_rate": 8.063264381700136e-05, "loss": 1.0255, "step": 77995 }, { "epoch": 0.2, "learning_rate": 8.063138557551965e-05, "loss": 1.0253, "step": 78000 }, { "epoch": 0.2, "learning_rate": 8.063012733403796e-05, "loss": 1.025, "step": 78005 }, { "epoch": 0.2, "learning_rate": 8.062886909255625e-05, "loss": 1.0238, "step": 78010 }, { "epoch": 0.2, "learning_rate": 8.062761085107454e-05, "loss": 1.0235, "step": 78015 }, { "epoch": 0.2, "learning_rate": 8.062635260959283e-05, "loss": 1.0245, "step": 78020 }, { "epoch": 0.2, "learning_rate": 8.062509436811113e-05, "loss": 1.0236, "step": 78025 }, { "epoch": 0.2, "learning_rate": 8.062383612662943e-05, "loss": 1.0246, "step": 78030 }, { "epoch": 0.2, "learning_rate": 8.062257788514772e-05, "loss": 1.0269, "step": 78035 }, { "epoch": 0.2, "learning_rate": 8.062131964366601e-05, "loss": 1.0242, "step": 78040 }, { "epoch": 0.2, "learning_rate": 8.06200614021843e-05, "loss": 1.0255, "step": 78045 }, { "epoch": 0.2, "learning_rate": 8.061880316070261e-05, "loss": 1.0238, "step": 78050 }, { "epoch": 0.2, "learning_rate": 8.06175449192209e-05, "loss": 1.0248, "step": 78055 }, { "epoch": 0.2, "learning_rate": 8.06162866777392e-05, "loss": 1.0264, "step": 78060 }, { "epoch": 0.2, "learning_rate": 8.061502843625749e-05, "loss": 1.0239, "step": 78065 }, { "epoch": 0.2, "learning_rate": 8.061377019477579e-05, "loss": 1.0262, "step": 78070 }, { "epoch": 0.2, "learning_rate": 8.061251195329408e-05, "loss": 1.0254, "step": 78075 }, { "epoch": 0.2, "learning_rate": 8.061125371181237e-05, "loss": 1.025, "step": 78080 }, { "epoch": 0.2, "learning_rate": 8.060999547033067e-05, "loss": 1.0255, "step": 78085 }, { "epoch": 0.2, "learning_rate": 8.060873722884896e-05, "loss": 1.0227, "step": 78090 }, { "epoch": 0.2, "learning_rate": 8.060747898736726e-05, "loss": 1.0243, "step": 78095 }, { "epoch": 0.2, "learning_rate": 8.060622074588555e-05, "loss": 1.0241, "step": 78100 }, { "epoch": 0.2, "learning_rate": 8.060496250440385e-05, "loss": 1.0254, "step": 78105 }, { "epoch": 0.2, "learning_rate": 8.060370426292214e-05, "loss": 1.0269, "step": 78110 }, { "epoch": 0.2, "learning_rate": 8.060244602144044e-05, "loss": 1.0249, "step": 78115 }, { "epoch": 0.2, "learning_rate": 8.060118777995873e-05, "loss": 1.0245, "step": 78120 }, { "epoch": 0.2, "learning_rate": 8.059992953847703e-05, "loss": 1.0243, "step": 78125 }, { "epoch": 0.2, "learning_rate": 8.059867129699532e-05, "loss": 1.0253, "step": 78130 }, { "epoch": 0.2, "learning_rate": 8.059741305551362e-05, "loss": 1.0242, "step": 78135 }, { "epoch": 0.2, "learning_rate": 8.059615481403191e-05, "loss": 1.025, "step": 78140 }, { "epoch": 0.2, "learning_rate": 8.05948965725502e-05, "loss": 1.0251, "step": 78145 }, { "epoch": 0.2, "learning_rate": 8.05936383310685e-05, "loss": 1.0249, "step": 78150 }, { "epoch": 0.2, "learning_rate": 8.059238008958679e-05, "loss": 1.0248, "step": 78155 }, { "epoch": 0.2, "learning_rate": 8.05911218481051e-05, "loss": 1.0244, "step": 78160 }, { "epoch": 0.2, "learning_rate": 8.058986360662339e-05, "loss": 1.0265, "step": 78165 }, { "epoch": 0.2, "learning_rate": 8.058860536514168e-05, "loss": 1.024, "step": 78170 }, { "epoch": 0.2, "learning_rate": 8.058734712365997e-05, "loss": 1.0258, "step": 78175 }, { "epoch": 0.2, "learning_rate": 8.058608888217827e-05, "loss": 1.0264, "step": 78180 }, { "epoch": 0.2, "learning_rate": 8.058483064069657e-05, "loss": 1.0244, "step": 78185 }, { "epoch": 0.2, "learning_rate": 8.058357239921486e-05, "loss": 1.0245, "step": 78190 }, { "epoch": 0.2, "learning_rate": 8.058231415773315e-05, "loss": 1.0241, "step": 78195 }, { "epoch": 0.2, "learning_rate": 8.058105591625145e-05, "loss": 1.0246, "step": 78200 }, { "epoch": 0.2, "learning_rate": 8.057979767476974e-05, "loss": 1.0235, "step": 78205 }, { "epoch": 0.2, "learning_rate": 8.057853943328804e-05, "loss": 1.0233, "step": 78210 }, { "epoch": 0.2, "learning_rate": 8.057728119180633e-05, "loss": 1.0242, "step": 78215 }, { "epoch": 0.2, "learning_rate": 8.057602295032462e-05, "loss": 1.027, "step": 78220 }, { "epoch": 0.2, "learning_rate": 8.057476470884292e-05, "loss": 1.024, "step": 78225 }, { "epoch": 0.2, "learning_rate": 8.057350646736122e-05, "loss": 1.0251, "step": 78230 }, { "epoch": 0.2, "learning_rate": 8.057224822587951e-05, "loss": 1.0465, "step": 78235 }, { "epoch": 0.2, "learning_rate": 8.05709899843978e-05, "loss": 1.0217, "step": 78240 }, { "epoch": 0.2, "learning_rate": 8.05697317429161e-05, "loss": 1.0242, "step": 78245 }, { "epoch": 0.2, "learning_rate": 8.05684735014344e-05, "loss": 1.0262, "step": 78250 }, { "epoch": 0.2, "learning_rate": 8.056721525995269e-05, "loss": 1.0263, "step": 78255 }, { "epoch": 0.2, "learning_rate": 8.056595701847098e-05, "loss": 1.0235, "step": 78260 }, { "epoch": 0.2, "learning_rate": 8.056469877698928e-05, "loss": 1.0253, "step": 78265 }, { "epoch": 0.2, "learning_rate": 8.056344053550759e-05, "loss": 1.0258, "step": 78270 }, { "epoch": 0.2, "learning_rate": 8.056218229402588e-05, "loss": 1.0237, "step": 78275 }, { "epoch": 0.2, "learning_rate": 8.056092405254417e-05, "loss": 1.0235, "step": 78280 }, { "epoch": 0.2, "learning_rate": 8.055966581106246e-05, "loss": 1.0233, "step": 78285 }, { "epoch": 0.2, "learning_rate": 8.055840756958076e-05, "loss": 1.0253, "step": 78290 }, { "epoch": 0.2, "learning_rate": 8.055714932809906e-05, "loss": 1.0452, "step": 78295 }, { "epoch": 0.2, "learning_rate": 8.055589108661735e-05, "loss": 1.0342, "step": 78300 }, { "epoch": 0.2, "learning_rate": 8.055463284513564e-05, "loss": 1.0255, "step": 78305 }, { "epoch": 0.2, "learning_rate": 8.055337460365394e-05, "loss": 1.0248, "step": 78310 }, { "epoch": 0.2, "learning_rate": 8.055211636217224e-05, "loss": 1.0244, "step": 78315 }, { "epoch": 0.2, "learning_rate": 8.055085812069053e-05, "loss": 1.0252, "step": 78320 }, { "epoch": 0.2, "learning_rate": 8.054959987920882e-05, "loss": 1.0269, "step": 78325 }, { "epoch": 0.2, "learning_rate": 8.054834163772712e-05, "loss": 1.0227, "step": 78330 }, { "epoch": 0.2, "learning_rate": 8.054708339624542e-05, "loss": 1.0254, "step": 78335 }, { "epoch": 0.2, "learning_rate": 8.054582515476371e-05, "loss": 1.0248, "step": 78340 }, { "epoch": 0.2, "learning_rate": 8.0544566913282e-05, "loss": 1.0258, "step": 78345 }, { "epoch": 0.2, "learning_rate": 8.05433086718003e-05, "loss": 1.0262, "step": 78350 }, { "epoch": 0.2, "learning_rate": 8.054205043031859e-05, "loss": 1.0238, "step": 78355 }, { "epoch": 0.2, "learning_rate": 8.054079218883689e-05, "loss": 1.0439, "step": 78360 }, { "epoch": 0.2, "learning_rate": 8.053953394735518e-05, "loss": 1.0267, "step": 78365 }, { "epoch": 0.2, "learning_rate": 8.053827570587348e-05, "loss": 1.0246, "step": 78370 }, { "epoch": 0.2, "learning_rate": 8.053701746439177e-05, "loss": 1.0245, "step": 78375 }, { "epoch": 0.2, "learning_rate": 8.053575922291007e-05, "loss": 1.0243, "step": 78380 }, { "epoch": 0.2, "learning_rate": 8.053450098142836e-05, "loss": 1.0241, "step": 78385 }, { "epoch": 0.2, "learning_rate": 8.053324273994666e-05, "loss": 1.0252, "step": 78390 }, { "epoch": 0.2, "learning_rate": 8.053198449846495e-05, "loss": 1.0255, "step": 78395 }, { "epoch": 0.2, "learning_rate": 8.053072625698325e-05, "loss": 1.0243, "step": 78400 }, { "epoch": 0.2, "learning_rate": 8.052946801550154e-05, "loss": 1.0255, "step": 78405 }, { "epoch": 0.2, "learning_rate": 8.052820977401984e-05, "loss": 1.0261, "step": 78410 }, { "epoch": 0.2, "learning_rate": 8.052695153253813e-05, "loss": 1.0266, "step": 78415 }, { "epoch": 0.2, "learning_rate": 8.052569329105642e-05, "loss": 1.0233, "step": 78420 }, { "epoch": 0.2, "learning_rate": 8.052443504957472e-05, "loss": 1.0256, "step": 78425 }, { "epoch": 0.2, "learning_rate": 8.052317680809302e-05, "loss": 1.0257, "step": 78430 }, { "epoch": 0.2, "learning_rate": 8.052191856661131e-05, "loss": 1.0262, "step": 78435 }, { "epoch": 0.2, "learning_rate": 8.05206603251296e-05, "loss": 1.0239, "step": 78440 }, { "epoch": 0.2, "learning_rate": 8.05194020836479e-05, "loss": 1.0253, "step": 78445 }, { "epoch": 0.2, "learning_rate": 8.05181438421662e-05, "loss": 1.0249, "step": 78450 }, { "epoch": 0.2, "learning_rate": 8.051688560068449e-05, "loss": 1.0271, "step": 78455 }, { "epoch": 0.2, "learning_rate": 8.051562735920278e-05, "loss": 1.0224, "step": 78460 }, { "epoch": 0.2, "learning_rate": 8.051436911772108e-05, "loss": 1.024, "step": 78465 }, { "epoch": 0.2, "learning_rate": 8.051311087623938e-05, "loss": 1.0238, "step": 78470 }, { "epoch": 0.2, "learning_rate": 8.051185263475767e-05, "loss": 1.0222, "step": 78475 }, { "epoch": 0.2, "learning_rate": 8.051059439327596e-05, "loss": 1.0238, "step": 78480 }, { "epoch": 0.2, "learning_rate": 8.050933615179425e-05, "loss": 1.0252, "step": 78485 }, { "epoch": 0.2, "learning_rate": 8.050807791031256e-05, "loss": 1.0219, "step": 78490 }, { "epoch": 0.2, "learning_rate": 8.050681966883085e-05, "loss": 1.0248, "step": 78495 }, { "epoch": 0.2, "learning_rate": 8.050556142734914e-05, "loss": 1.0263, "step": 78500 }, { "epoch": 0.2, "learning_rate": 8.050430318586743e-05, "loss": 1.0254, "step": 78505 }, { "epoch": 0.2, "learning_rate": 8.050304494438574e-05, "loss": 1.0261, "step": 78510 }, { "epoch": 0.2, "learning_rate": 8.050178670290403e-05, "loss": 1.0242, "step": 78515 }, { "epoch": 0.2, "learning_rate": 8.050052846142232e-05, "loss": 1.0256, "step": 78520 }, { "epoch": 0.2, "learning_rate": 8.049927021994061e-05, "loss": 1.0249, "step": 78525 }, { "epoch": 0.2, "learning_rate": 8.049801197845892e-05, "loss": 1.0261, "step": 78530 }, { "epoch": 0.2, "learning_rate": 8.049675373697721e-05, "loss": 1.0251, "step": 78535 }, { "epoch": 0.2, "learning_rate": 8.04954954954955e-05, "loss": 1.0244, "step": 78540 }, { "epoch": 0.2, "learning_rate": 8.049423725401379e-05, "loss": 1.0251, "step": 78545 }, { "epoch": 0.2, "learning_rate": 8.049297901253208e-05, "loss": 1.0246, "step": 78550 }, { "epoch": 0.2, "learning_rate": 8.049172077105039e-05, "loss": 1.0377, "step": 78555 }, { "epoch": 0.2, "learning_rate": 8.049046252956868e-05, "loss": 1.0259, "step": 78560 }, { "epoch": 0.2, "learning_rate": 8.048920428808697e-05, "loss": 1.0253, "step": 78565 }, { "epoch": 0.2, "learning_rate": 8.048794604660526e-05, "loss": 1.0234, "step": 78570 }, { "epoch": 0.2, "learning_rate": 8.048668780512357e-05, "loss": 1.026, "step": 78575 }, { "epoch": 0.2, "learning_rate": 8.048542956364186e-05, "loss": 1.0251, "step": 78580 }, { "epoch": 0.2, "learning_rate": 8.048417132216015e-05, "loss": 1.0243, "step": 78585 }, { "epoch": 0.2, "learning_rate": 8.048291308067844e-05, "loss": 1.0229, "step": 78590 }, { "epoch": 0.2, "learning_rate": 8.048165483919675e-05, "loss": 1.0231, "step": 78595 }, { "epoch": 0.2, "learning_rate": 8.048039659771504e-05, "loss": 1.0247, "step": 78600 }, { "epoch": 0.2, "learning_rate": 8.047913835623333e-05, "loss": 1.0245, "step": 78605 }, { "epoch": 0.2, "learning_rate": 8.047788011475162e-05, "loss": 1.0272, "step": 78610 }, { "epoch": 0.2, "learning_rate": 8.047662187326991e-05, "loss": 1.0258, "step": 78615 }, { "epoch": 0.2, "learning_rate": 8.047536363178822e-05, "loss": 1.0273, "step": 78620 }, { "epoch": 0.2, "learning_rate": 8.047410539030651e-05, "loss": 1.0251, "step": 78625 }, { "epoch": 0.2, "learning_rate": 8.04728471488248e-05, "loss": 1.0249, "step": 78630 }, { "epoch": 0.2, "learning_rate": 8.04715889073431e-05, "loss": 1.0238, "step": 78635 }, { "epoch": 0.2, "learning_rate": 8.04703306658614e-05, "loss": 1.0243, "step": 78640 }, { "epoch": 0.2, "learning_rate": 8.046907242437969e-05, "loss": 1.0254, "step": 78645 }, { "epoch": 0.2, "learning_rate": 8.046781418289798e-05, "loss": 1.024, "step": 78650 }, { "epoch": 0.2, "learning_rate": 8.046655594141627e-05, "loss": 1.0251, "step": 78655 }, { "epoch": 0.2, "learning_rate": 8.046529769993458e-05, "loss": 1.0253, "step": 78660 }, { "epoch": 0.2, "learning_rate": 8.046403945845287e-05, "loss": 1.0256, "step": 78665 }, { "epoch": 0.2, "learning_rate": 8.046278121697116e-05, "loss": 1.0457, "step": 78670 }, { "epoch": 0.2, "learning_rate": 8.046152297548945e-05, "loss": 1.0235, "step": 78675 }, { "epoch": 0.2, "learning_rate": 8.046026473400774e-05, "loss": 1.0253, "step": 78680 }, { "epoch": 0.2, "learning_rate": 8.045900649252605e-05, "loss": 1.0267, "step": 78685 }, { "epoch": 0.2, "learning_rate": 8.045774825104434e-05, "loss": 1.0247, "step": 78690 }, { "epoch": 0.2, "learning_rate": 8.045649000956263e-05, "loss": 1.025, "step": 78695 }, { "epoch": 0.2, "learning_rate": 8.045523176808092e-05, "loss": 1.0248, "step": 78700 }, { "epoch": 0.2, "learning_rate": 8.045397352659923e-05, "loss": 1.0264, "step": 78705 }, { "epoch": 0.2, "learning_rate": 8.045271528511752e-05, "loss": 1.0242, "step": 78710 }, { "epoch": 0.2, "learning_rate": 8.045145704363581e-05, "loss": 1.0229, "step": 78715 }, { "epoch": 0.2, "learning_rate": 8.04501988021541e-05, "loss": 1.0267, "step": 78720 }, { "epoch": 0.2, "learning_rate": 8.044894056067241e-05, "loss": 1.0241, "step": 78725 }, { "epoch": 0.2, "learning_rate": 8.04476823191907e-05, "loss": 1.0251, "step": 78730 }, { "epoch": 0.2, "learning_rate": 8.044642407770899e-05, "loss": 1.0253, "step": 78735 }, { "epoch": 0.2, "learning_rate": 8.044516583622728e-05, "loss": 1.0257, "step": 78740 }, { "epoch": 0.2, "learning_rate": 8.044390759474558e-05, "loss": 1.024, "step": 78745 }, { "epoch": 0.2, "learning_rate": 8.044264935326388e-05, "loss": 1.0267, "step": 78750 }, { "epoch": 0.2, "learning_rate": 8.044139111178217e-05, "loss": 1.024, "step": 78755 }, { "epoch": 0.2, "learning_rate": 8.044013287030046e-05, "loss": 1.0233, "step": 78760 }, { "epoch": 0.2, "learning_rate": 8.043887462881877e-05, "loss": 1.0272, "step": 78765 }, { "epoch": 0.2, "learning_rate": 8.043761638733706e-05, "loss": 1.0273, "step": 78770 }, { "epoch": 0.2, "learning_rate": 8.043635814585537e-05, "loss": 1.0251, "step": 78775 }, { "epoch": 0.2, "learning_rate": 8.043509990437366e-05, "loss": 1.0246, "step": 78780 }, { "epoch": 0.2, "learning_rate": 8.043384166289195e-05, "loss": 1.023, "step": 78785 }, { "epoch": 0.2, "learning_rate": 8.043258342141024e-05, "loss": 1.0229, "step": 78790 }, { "epoch": 0.2, "learning_rate": 8.043132517992855e-05, "loss": 1.0251, "step": 78795 }, { "epoch": 0.2, "learning_rate": 8.043006693844684e-05, "loss": 1.0425, "step": 78800 }, { "epoch": 0.2, "learning_rate": 8.042880869696513e-05, "loss": 1.0252, "step": 78805 }, { "epoch": 0.2, "learning_rate": 8.042755045548342e-05, "loss": 1.0256, "step": 78810 }, { "epoch": 0.2, "learning_rate": 8.042629221400171e-05, "loss": 1.0276, "step": 78815 }, { "epoch": 0.2, "learning_rate": 8.042503397252002e-05, "loss": 1.0258, "step": 78820 }, { "epoch": 0.2, "learning_rate": 8.042377573103831e-05, "loss": 1.0273, "step": 78825 }, { "epoch": 0.2, "learning_rate": 8.04225174895566e-05, "loss": 1.0255, "step": 78830 }, { "epoch": 0.2, "learning_rate": 8.042125924807489e-05, "loss": 1.0223, "step": 78835 }, { "epoch": 0.2, "learning_rate": 8.04200010065932e-05, "loss": 1.0242, "step": 78840 }, { "epoch": 0.2, "learning_rate": 8.041874276511149e-05, "loss": 1.0458, "step": 78845 }, { "epoch": 0.2, "learning_rate": 8.041748452362978e-05, "loss": 1.0253, "step": 78850 }, { "epoch": 0.2, "learning_rate": 8.041622628214807e-05, "loss": 1.0235, "step": 78855 }, { "epoch": 0.2, "learning_rate": 8.041496804066638e-05, "loss": 1.0253, "step": 78860 }, { "epoch": 0.2, "learning_rate": 8.041370979918467e-05, "loss": 1.0277, "step": 78865 }, { "epoch": 0.2, "learning_rate": 8.041245155770296e-05, "loss": 1.0234, "step": 78870 }, { "epoch": 0.2, "learning_rate": 8.041119331622125e-05, "loss": 1.0255, "step": 78875 }, { "epoch": 0.2, "learning_rate": 8.040993507473954e-05, "loss": 1.0273, "step": 78880 }, { "epoch": 0.2, "learning_rate": 8.040867683325785e-05, "loss": 1.0255, "step": 78885 }, { "epoch": 0.2, "learning_rate": 8.040741859177614e-05, "loss": 1.0249, "step": 78890 }, { "epoch": 0.2, "learning_rate": 8.040616035029443e-05, "loss": 1.0264, "step": 78895 }, { "epoch": 0.2, "learning_rate": 8.040490210881272e-05, "loss": 1.0247, "step": 78900 }, { "epoch": 0.2, "learning_rate": 8.040364386733103e-05, "loss": 1.0258, "step": 78905 }, { "epoch": 0.2, "learning_rate": 8.040238562584932e-05, "loss": 1.0249, "step": 78910 }, { "epoch": 0.2, "learning_rate": 8.040112738436761e-05, "loss": 1.0246, "step": 78915 }, { "epoch": 0.2, "learning_rate": 8.03998691428859e-05, "loss": 1.0249, "step": 78920 }, { "epoch": 0.2, "learning_rate": 8.039861090140421e-05, "loss": 1.0235, "step": 78925 }, { "epoch": 0.2, "learning_rate": 8.03973526599225e-05, "loss": 1.0257, "step": 78930 }, { "epoch": 0.2, "learning_rate": 8.039609441844079e-05, "loss": 1.0244, "step": 78935 }, { "epoch": 0.2, "learning_rate": 8.039483617695908e-05, "loss": 1.0264, "step": 78940 }, { "epoch": 0.2, "learning_rate": 8.039357793547738e-05, "loss": 1.0234, "step": 78945 }, { "epoch": 0.2, "learning_rate": 8.039231969399568e-05, "loss": 1.0244, "step": 78950 }, { "epoch": 0.2, "learning_rate": 8.039106145251397e-05, "loss": 1.0218, "step": 78955 }, { "epoch": 0.2, "learning_rate": 8.038980321103226e-05, "loss": 1.0265, "step": 78960 }, { "epoch": 0.2, "learning_rate": 8.038854496955056e-05, "loss": 1.0233, "step": 78965 }, { "epoch": 0.2, "learning_rate": 8.038728672806886e-05, "loss": 1.0249, "step": 78970 }, { "epoch": 0.2, "learning_rate": 8.038602848658715e-05, "loss": 1.0258, "step": 78975 }, { "epoch": 0.2, "learning_rate": 8.038477024510544e-05, "loss": 1.0237, "step": 78980 }, { "epoch": 0.2, "learning_rate": 8.038351200362374e-05, "loss": 1.0256, "step": 78985 }, { "epoch": 0.2, "learning_rate": 8.038225376214204e-05, "loss": 1.0236, "step": 78990 }, { "epoch": 0.2, "learning_rate": 8.038099552066033e-05, "loss": 1.0435, "step": 78995 }, { "epoch": 0.2, "learning_rate": 8.037973727917862e-05, "loss": 1.0234, "step": 79000 }, { "epoch": 0.2, "learning_rate": 8.037847903769692e-05, "loss": 1.025, "step": 79005 }, { "epoch": 0.2, "learning_rate": 8.037722079621521e-05, "loss": 1.0251, "step": 79010 }, { "epoch": 0.2, "learning_rate": 8.037596255473351e-05, "loss": 1.0252, "step": 79015 }, { "epoch": 0.2, "learning_rate": 8.03747043132518e-05, "loss": 1.0266, "step": 79020 }, { "epoch": 0.2, "learning_rate": 8.03734460717701e-05, "loss": 1.0269, "step": 79025 }, { "epoch": 0.2, "learning_rate": 8.037218783028839e-05, "loss": 1.0248, "step": 79030 }, { "epoch": 0.2, "learning_rate": 8.037092958880669e-05, "loss": 1.0208, "step": 79035 }, { "epoch": 0.2, "learning_rate": 8.036967134732498e-05, "loss": 1.0246, "step": 79040 }, { "epoch": 0.2, "learning_rate": 8.036841310584328e-05, "loss": 1.027, "step": 79045 }, { "epoch": 0.2, "learning_rate": 8.036715486436157e-05, "loss": 1.0272, "step": 79050 }, { "epoch": 0.2, "learning_rate": 8.036589662287987e-05, "loss": 1.0258, "step": 79055 }, { "epoch": 0.2, "learning_rate": 8.036463838139816e-05, "loss": 1.0428, "step": 79060 }, { "epoch": 0.2, "learning_rate": 8.036338013991646e-05, "loss": 1.0229, "step": 79065 }, { "epoch": 0.2, "learning_rate": 8.036212189843475e-05, "loss": 1.024, "step": 79070 }, { "epoch": 0.2, "learning_rate": 8.036086365695304e-05, "loss": 1.027, "step": 79075 }, { "epoch": 0.2, "learning_rate": 8.035960541547134e-05, "loss": 1.0244, "step": 79080 }, { "epoch": 0.2, "learning_rate": 8.035834717398964e-05, "loss": 1.0222, "step": 79085 }, { "epoch": 0.2, "learning_rate": 8.035708893250793e-05, "loss": 1.0251, "step": 79090 }, { "epoch": 0.2, "learning_rate": 8.035583069102622e-05, "loss": 1.022, "step": 79095 }, { "epoch": 0.2, "learning_rate": 8.035457244954452e-05, "loss": 1.0235, "step": 79100 }, { "epoch": 0.2, "learning_rate": 8.035331420806282e-05, "loss": 1.0234, "step": 79105 }, { "epoch": 0.2, "learning_rate": 8.035205596658111e-05, "loss": 1.0236, "step": 79110 }, { "epoch": 0.2, "learning_rate": 8.03507977250994e-05, "loss": 1.0244, "step": 79115 }, { "epoch": 0.2, "learning_rate": 8.03495394836177e-05, "loss": 1.0244, "step": 79120 }, { "epoch": 0.2, "learning_rate": 8.0348281242136e-05, "loss": 1.0234, "step": 79125 }, { "epoch": 0.2, "learning_rate": 8.034702300065429e-05, "loss": 1.0245, "step": 79130 }, { "epoch": 0.2, "learning_rate": 8.034576475917258e-05, "loss": 1.0241, "step": 79135 }, { "epoch": 0.2, "learning_rate": 8.034450651769087e-05, "loss": 1.024, "step": 79140 }, { "epoch": 0.2, "learning_rate": 8.034324827620918e-05, "loss": 1.0242, "step": 79145 }, { "epoch": 0.2, "learning_rate": 8.034199003472747e-05, "loss": 1.0255, "step": 79150 }, { "epoch": 0.2, "learning_rate": 8.034073179324576e-05, "loss": 1.0236, "step": 79155 }, { "epoch": 0.2, "learning_rate": 8.033947355176405e-05, "loss": 1.0232, "step": 79160 }, { "epoch": 0.2, "learning_rate": 8.033821531028236e-05, "loss": 1.0254, "step": 79165 }, { "epoch": 0.2, "learning_rate": 8.033695706880065e-05, "loss": 1.0252, "step": 79170 }, { "epoch": 0.2, "learning_rate": 8.033569882731894e-05, "loss": 1.0249, "step": 79175 }, { "epoch": 0.2, "learning_rate": 8.033444058583723e-05, "loss": 1.0263, "step": 79180 }, { "epoch": 0.2, "learning_rate": 8.033318234435554e-05, "loss": 1.0283, "step": 79185 }, { "epoch": 0.2, "learning_rate": 8.033192410287383e-05, "loss": 1.023, "step": 79190 }, { "epoch": 0.2, "learning_rate": 8.033066586139212e-05, "loss": 1.0266, "step": 79195 }, { "epoch": 0.2, "learning_rate": 8.032940761991041e-05, "loss": 1.0257, "step": 79200 }, { "epoch": 0.2, "learning_rate": 8.03281493784287e-05, "loss": 1.0252, "step": 79205 }, { "epoch": 0.2, "learning_rate": 8.0326891136947e-05, "loss": 1.0265, "step": 79210 }, { "epoch": 0.2, "learning_rate": 8.03256328954653e-05, "loss": 1.0252, "step": 79215 }, { "epoch": 0.2, "learning_rate": 8.032437465398359e-05, "loss": 1.0253, "step": 79220 }, { "epoch": 0.2, "learning_rate": 8.032311641250188e-05, "loss": 1.0235, "step": 79225 }, { "epoch": 0.2, "learning_rate": 8.032185817102019e-05, "loss": 1.0222, "step": 79230 }, { "epoch": 0.2, "learning_rate": 8.032059992953848e-05, "loss": 1.0264, "step": 79235 }, { "epoch": 0.2, "learning_rate": 8.031934168805677e-05, "loss": 1.0227, "step": 79240 }, { "epoch": 0.2, "learning_rate": 8.031808344657506e-05, "loss": 1.0401, "step": 79245 }, { "epoch": 0.2, "learning_rate": 8.031682520509337e-05, "loss": 1.0253, "step": 79250 }, { "epoch": 0.2, "learning_rate": 8.031556696361166e-05, "loss": 1.0244, "step": 79255 }, { "epoch": 0.2, "learning_rate": 8.031430872212995e-05, "loss": 1.0453, "step": 79260 }, { "epoch": 0.2, "learning_rate": 8.031305048064825e-05, "loss": 1.026, "step": 79265 }, { "epoch": 0.2, "learning_rate": 8.031179223916655e-05, "loss": 1.025, "step": 79270 }, { "epoch": 0.2, "learning_rate": 8.031053399768484e-05, "loss": 1.0239, "step": 79275 }, { "epoch": 0.2, "learning_rate": 8.030927575620314e-05, "loss": 1.0261, "step": 79280 }, { "epoch": 0.2, "learning_rate": 8.030801751472143e-05, "loss": 1.024, "step": 79285 }, { "epoch": 0.2, "learning_rate": 8.030675927323973e-05, "loss": 1.0271, "step": 79290 }, { "epoch": 0.2, "learning_rate": 8.030550103175802e-05, "loss": 1.0265, "step": 79295 }, { "epoch": 0.2, "learning_rate": 8.030424279027632e-05, "loss": 1.0262, "step": 79300 }, { "epoch": 0.2, "learning_rate": 8.030298454879461e-05, "loss": 1.0243, "step": 79305 }, { "epoch": 0.2, "learning_rate": 8.03017263073129e-05, "loss": 1.0251, "step": 79310 }, { "epoch": 0.2, "learning_rate": 8.03004680658312e-05, "loss": 1.0228, "step": 79315 }, { "epoch": 0.2, "learning_rate": 8.02992098243495e-05, "loss": 1.0237, "step": 79320 }, { "epoch": 0.2, "learning_rate": 8.02979515828678e-05, "loss": 1.0254, "step": 79325 }, { "epoch": 0.2, "learning_rate": 8.029669334138609e-05, "loss": 1.0265, "step": 79330 }, { "epoch": 0.2, "learning_rate": 8.029543509990438e-05, "loss": 1.0223, "step": 79335 }, { "epoch": 0.2, "learning_rate": 8.029417685842267e-05, "loss": 1.0228, "step": 79340 }, { "epoch": 0.2, "learning_rate": 8.029291861694097e-05, "loss": 1.0248, "step": 79345 }, { "epoch": 0.2, "learning_rate": 8.029166037545927e-05, "loss": 1.0253, "step": 79350 }, { "epoch": 0.2, "learning_rate": 8.029040213397756e-05, "loss": 1.0246, "step": 79355 }, { "epoch": 0.2, "learning_rate": 8.028914389249585e-05, "loss": 1.0246, "step": 79360 }, { "epoch": 0.2, "learning_rate": 8.028788565101415e-05, "loss": 1.0225, "step": 79365 }, { "epoch": 0.2, "learning_rate": 8.028662740953245e-05, "loss": 1.023, "step": 79370 }, { "epoch": 0.2, "learning_rate": 8.028536916805074e-05, "loss": 1.026, "step": 79375 }, { "epoch": 0.2, "learning_rate": 8.028411092656903e-05, "loss": 1.0258, "step": 79380 }, { "epoch": 0.2, "learning_rate": 8.028285268508733e-05, "loss": 1.0245, "step": 79385 }, { "epoch": 0.2, "learning_rate": 8.028159444360563e-05, "loss": 1.0252, "step": 79390 }, { "epoch": 0.2, "learning_rate": 8.028033620212392e-05, "loss": 1.025, "step": 79395 }, { "epoch": 0.2, "learning_rate": 8.027907796064221e-05, "loss": 1.0247, "step": 79400 }, { "epoch": 0.2, "learning_rate": 8.02778197191605e-05, "loss": 1.0249, "step": 79405 }, { "epoch": 0.2, "learning_rate": 8.02765614776788e-05, "loss": 1.0487, "step": 79410 }, { "epoch": 0.2, "learning_rate": 8.02753032361971e-05, "loss": 1.0256, "step": 79415 }, { "epoch": 0.2, "learning_rate": 8.027404499471539e-05, "loss": 1.0252, "step": 79420 }, { "epoch": 0.2, "learning_rate": 8.027278675323368e-05, "loss": 1.0249, "step": 79425 }, { "epoch": 0.2, "learning_rate": 8.027152851175199e-05, "loss": 1.0239, "step": 79430 }, { "epoch": 0.2, "learning_rate": 8.027027027027028e-05, "loss": 1.0256, "step": 79435 }, { "epoch": 0.2, "learning_rate": 8.026901202878857e-05, "loss": 1.0246, "step": 79440 }, { "epoch": 0.2, "learning_rate": 8.026775378730686e-05, "loss": 1.0237, "step": 79445 }, { "epoch": 0.2, "learning_rate": 8.026649554582517e-05, "loss": 1.0238, "step": 79450 }, { "epoch": 0.2, "learning_rate": 8.026523730434346e-05, "loss": 1.0252, "step": 79455 }, { "epoch": 0.2, "learning_rate": 8.026397906286175e-05, "loss": 1.0266, "step": 79460 }, { "epoch": 0.2, "learning_rate": 8.026272082138004e-05, "loss": 1.023, "step": 79465 }, { "epoch": 0.2, "learning_rate": 8.026146257989833e-05, "loss": 1.0259, "step": 79470 }, { "epoch": 0.2, "learning_rate": 8.026020433841664e-05, "loss": 1.0262, "step": 79475 }, { "epoch": 0.2, "learning_rate": 8.025894609693493e-05, "loss": 1.0271, "step": 79480 }, { "epoch": 0.2, "learning_rate": 8.025768785545322e-05, "loss": 1.0268, "step": 79485 }, { "epoch": 0.2, "learning_rate": 8.025642961397151e-05, "loss": 1.0246, "step": 79490 }, { "epoch": 0.2, "learning_rate": 8.025517137248982e-05, "loss": 1.0369, "step": 79495 }, { "epoch": 0.2, "learning_rate": 8.025391313100811e-05, "loss": 1.0257, "step": 79500 }, { "epoch": 0.2, "learning_rate": 8.02526548895264e-05, "loss": 1.0257, "step": 79505 }, { "epoch": 0.2, "learning_rate": 8.025139664804469e-05, "loss": 1.0242, "step": 79510 }, { "epoch": 0.2, "learning_rate": 8.0250138406563e-05, "loss": 1.025, "step": 79515 }, { "epoch": 0.2, "learning_rate": 8.024888016508129e-05, "loss": 1.0236, "step": 79520 }, { "epoch": 0.2, "learning_rate": 8.024762192359958e-05, "loss": 1.0262, "step": 79525 }, { "epoch": 0.2, "learning_rate": 8.024636368211787e-05, "loss": 1.0229, "step": 79530 }, { "epoch": 0.2, "learning_rate": 8.024510544063616e-05, "loss": 1.0238, "step": 79535 }, { "epoch": 0.2, "learning_rate": 8.024384719915447e-05, "loss": 1.024, "step": 79540 }, { "epoch": 0.2, "learning_rate": 8.024258895767276e-05, "loss": 1.0281, "step": 79545 }, { "epoch": 0.2, "learning_rate": 8.024133071619105e-05, "loss": 1.0259, "step": 79550 }, { "epoch": 0.2, "learning_rate": 8.024007247470934e-05, "loss": 1.0268, "step": 79555 }, { "epoch": 0.2, "learning_rate": 8.023881423322765e-05, "loss": 1.0254, "step": 79560 }, { "epoch": 0.2, "learning_rate": 8.023755599174594e-05, "loss": 1.0257, "step": 79565 }, { "epoch": 0.2, "learning_rate": 8.023629775026423e-05, "loss": 1.0254, "step": 79570 }, { "epoch": 0.2, "learning_rate": 8.023503950878252e-05, "loss": 1.0249, "step": 79575 }, { "epoch": 0.2, "learning_rate": 8.023378126730083e-05, "loss": 1.0251, "step": 79580 }, { "epoch": 0.2, "learning_rate": 8.023252302581912e-05, "loss": 1.0223, "step": 79585 }, { "epoch": 0.2, "learning_rate": 8.023126478433741e-05, "loss": 1.0247, "step": 79590 }, { "epoch": 0.2, "learning_rate": 8.02300065428557e-05, "loss": 1.0239, "step": 79595 }, { "epoch": 0.2, "learning_rate": 8.0228748301374e-05, "loss": 1.0269, "step": 79600 }, { "epoch": 0.2, "learning_rate": 8.02274900598923e-05, "loss": 1.0258, "step": 79605 }, { "epoch": 0.2, "learning_rate": 8.022623181841059e-05, "loss": 1.025, "step": 79610 }, { "epoch": 0.2, "learning_rate": 8.022497357692888e-05, "loss": 1.0255, "step": 79615 }, { "epoch": 0.2, "learning_rate": 8.022371533544718e-05, "loss": 1.0243, "step": 79620 }, { "epoch": 0.2, "learning_rate": 8.022245709396548e-05, "loss": 1.0247, "step": 79625 }, { "epoch": 0.2, "learning_rate": 8.022119885248377e-05, "loss": 1.025, "step": 79630 }, { "epoch": 0.2, "learning_rate": 8.021994061100206e-05, "loss": 1.0254, "step": 79635 }, { "epoch": 0.2, "learning_rate": 8.021868236952035e-05, "loss": 1.0266, "step": 79640 }, { "epoch": 0.2, "learning_rate": 8.021742412803866e-05, "loss": 1.0262, "step": 79645 }, { "epoch": 0.2, "learning_rate": 8.021616588655695e-05, "loss": 1.0261, "step": 79650 }, { "epoch": 0.2, "learning_rate": 8.021490764507524e-05, "loss": 1.025, "step": 79655 }, { "epoch": 0.2, "learning_rate": 8.021364940359353e-05, "loss": 1.0268, "step": 79660 }, { "epoch": 0.2, "learning_rate": 8.021239116211183e-05, "loss": 1.0258, "step": 79665 }, { "epoch": 0.2, "learning_rate": 8.021113292063013e-05, "loss": 1.0251, "step": 79670 }, { "epoch": 0.2, "learning_rate": 8.020987467914842e-05, "loss": 1.0248, "step": 79675 }, { "epoch": 0.2, "learning_rate": 8.020861643766671e-05, "loss": 1.0269, "step": 79680 }, { "epoch": 0.2, "learning_rate": 8.0207358196185e-05, "loss": 1.0245, "step": 79685 }, { "epoch": 0.2, "learning_rate": 8.020609995470331e-05, "loss": 1.0436, "step": 79690 }, { "epoch": 0.2, "learning_rate": 8.02048417132216e-05, "loss": 1.0265, "step": 79695 }, { "epoch": 0.2, "learning_rate": 8.02035834717399e-05, "loss": 1.0235, "step": 79700 }, { "epoch": 0.2, "learning_rate": 8.020232523025819e-05, "loss": 1.0258, "step": 79705 }, { "epoch": 0.2, "learning_rate": 8.020106698877649e-05, "loss": 1.0442, "step": 79710 }, { "epoch": 0.2, "learning_rate": 8.019980874729478e-05, "loss": 1.0242, "step": 79715 }, { "epoch": 0.2, "learning_rate": 8.019855050581307e-05, "loss": 1.0256, "step": 79720 }, { "epoch": 0.2, "learning_rate": 8.019729226433137e-05, "loss": 1.0245, "step": 79725 }, { "epoch": 0.2, "learning_rate": 8.019603402284966e-05, "loss": 1.0244, "step": 79730 }, { "epoch": 0.2, "learning_rate": 8.019477578136796e-05, "loss": 1.0247, "step": 79735 }, { "epoch": 0.2, "learning_rate": 8.019351753988625e-05, "loss": 1.0226, "step": 79740 }, { "epoch": 0.2, "learning_rate": 8.019225929840455e-05, "loss": 1.027, "step": 79745 }, { "epoch": 0.2, "learning_rate": 8.019100105692284e-05, "loss": 1.0255, "step": 79750 }, { "epoch": 0.2, "learning_rate": 8.018974281544114e-05, "loss": 1.0252, "step": 79755 }, { "epoch": 0.2, "learning_rate": 8.018848457395943e-05, "loss": 1.0248, "step": 79760 }, { "epoch": 0.2, "learning_rate": 8.018722633247774e-05, "loss": 1.025, "step": 79765 }, { "epoch": 0.2, "learning_rate": 8.018596809099603e-05, "loss": 1.0268, "step": 79770 }, { "epoch": 0.2, "learning_rate": 8.018470984951432e-05, "loss": 1.0238, "step": 79775 }, { "epoch": 0.2, "learning_rate": 8.018345160803263e-05, "loss": 1.0247, "step": 79780 }, { "epoch": 0.2, "learning_rate": 8.018219336655092e-05, "loss": 1.0222, "step": 79785 }, { "epoch": 0.2, "learning_rate": 8.018093512506921e-05, "loss": 1.0233, "step": 79790 }, { "epoch": 0.2, "learning_rate": 8.01796768835875e-05, "loss": 1.0243, "step": 79795 }, { "epoch": 0.2, "learning_rate": 8.017841864210581e-05, "loss": 1.0271, "step": 79800 }, { "epoch": 0.2, "learning_rate": 8.01771604006241e-05, "loss": 1.0248, "step": 79805 }, { "epoch": 0.2, "learning_rate": 8.017590215914239e-05, "loss": 1.0246, "step": 79810 }, { "epoch": 0.2, "learning_rate": 8.017464391766068e-05, "loss": 1.0244, "step": 79815 }, { "epoch": 0.2, "learning_rate": 8.017338567617897e-05, "loss": 1.0256, "step": 79820 }, { "epoch": 0.2, "learning_rate": 8.017212743469728e-05, "loss": 1.0269, "step": 79825 }, { "epoch": 0.2, "learning_rate": 8.017086919321557e-05, "loss": 1.0237, "step": 79830 }, { "epoch": 0.2, "learning_rate": 8.016961095173386e-05, "loss": 1.025, "step": 79835 }, { "epoch": 0.2, "learning_rate": 8.016835271025215e-05, "loss": 1.0279, "step": 79840 }, { "epoch": 0.2, "learning_rate": 8.016709446877046e-05, "loss": 1.0252, "step": 79845 }, { "epoch": 0.2, "learning_rate": 8.016583622728875e-05, "loss": 1.0271, "step": 79850 }, { "epoch": 0.2, "learning_rate": 8.016457798580704e-05, "loss": 1.0268, "step": 79855 }, { "epoch": 0.2, "learning_rate": 8.016331974432533e-05, "loss": 1.0254, "step": 79860 }, { "epoch": 0.2, "learning_rate": 8.016206150284364e-05, "loss": 1.0234, "step": 79865 }, { "epoch": 0.2, "learning_rate": 8.016080326136193e-05, "loss": 1.0241, "step": 79870 }, { "epoch": 0.2, "learning_rate": 8.015954501988022e-05, "loss": 1.0221, "step": 79875 }, { "epoch": 0.2, "learning_rate": 8.015828677839851e-05, "loss": 1.024, "step": 79880 }, { "epoch": 0.2, "learning_rate": 8.01570285369168e-05, "loss": 1.0245, "step": 79885 }, { "epoch": 0.2, "learning_rate": 8.015577029543511e-05, "loss": 1.0262, "step": 79890 }, { "epoch": 0.2, "learning_rate": 8.01545120539534e-05, "loss": 1.0233, "step": 79895 }, { "epoch": 0.2, "learning_rate": 8.01532538124717e-05, "loss": 1.0246, "step": 79900 }, { "epoch": 0.2, "learning_rate": 8.015199557098999e-05, "loss": 1.0235, "step": 79905 }, { "epoch": 0.2, "learning_rate": 8.015073732950829e-05, "loss": 1.0244, "step": 79910 }, { "epoch": 0.2, "learning_rate": 8.014947908802658e-05, "loss": 1.0248, "step": 79915 }, { "epoch": 0.2, "learning_rate": 8.014822084654487e-05, "loss": 1.0264, "step": 79920 }, { "epoch": 0.2, "learning_rate": 8.014696260506317e-05, "loss": 1.026, "step": 79925 }, { "epoch": 0.2, "learning_rate": 8.014570436358147e-05, "loss": 1.0228, "step": 79930 }, { "epoch": 0.2, "learning_rate": 8.014444612209976e-05, "loss": 1.0246, "step": 79935 }, { "epoch": 0.2, "learning_rate": 8.014318788061805e-05, "loss": 1.0254, "step": 79940 }, { "epoch": 0.2, "learning_rate": 8.014192963913635e-05, "loss": 1.0272, "step": 79945 }, { "epoch": 0.2, "learning_rate": 8.014067139765464e-05, "loss": 1.0249, "step": 79950 }, { "epoch": 0.2, "learning_rate": 8.013941315617294e-05, "loss": 1.022, "step": 79955 }, { "epoch": 0.2, "learning_rate": 8.013815491469123e-05, "loss": 1.0242, "step": 79960 }, { "epoch": 0.2, "learning_rate": 8.013689667320953e-05, "loss": 1.0261, "step": 79965 }, { "epoch": 0.2, "learning_rate": 8.013563843172782e-05, "loss": 1.0229, "step": 79970 }, { "epoch": 0.2, "learning_rate": 8.013438019024612e-05, "loss": 1.0233, "step": 79975 }, { "epoch": 0.2, "learning_rate": 8.013312194876441e-05, "loss": 1.0247, "step": 79980 }, { "epoch": 0.2, "learning_rate": 8.01318637072827e-05, "loss": 1.0261, "step": 79985 }, { "epoch": 0.2, "learning_rate": 8.0130605465801e-05, "loss": 1.0259, "step": 79990 }, { "epoch": 0.2, "learning_rate": 8.012934722431929e-05, "loss": 1.0411, "step": 79995 }, { "epoch": 0.2, "learning_rate": 8.01280889828376e-05, "loss": 1.0244, "step": 80000 }, { "epoch": 0.2, "learning_rate": 8.012683074135589e-05, "loss": 1.0263, "step": 80005 }, { "epoch": 0.2, "learning_rate": 8.012557249987418e-05, "loss": 1.025, "step": 80010 }, { "epoch": 0.2, "learning_rate": 8.012431425839247e-05, "loss": 1.0252, "step": 80015 }, { "epoch": 0.2, "learning_rate": 8.012305601691077e-05, "loss": 1.0249, "step": 80020 }, { "epoch": 0.2, "learning_rate": 8.012179777542907e-05, "loss": 1.0249, "step": 80025 }, { "epoch": 0.2, "learning_rate": 8.012053953394736e-05, "loss": 1.0441, "step": 80030 }, { "epoch": 0.2, "learning_rate": 8.011928129246565e-05, "loss": 1.0251, "step": 80035 }, { "epoch": 0.2, "learning_rate": 8.011802305098395e-05, "loss": 1.0237, "step": 80040 }, { "epoch": 0.2, "learning_rate": 8.011676480950225e-05, "loss": 1.0239, "step": 80045 }, { "epoch": 0.2, "learning_rate": 8.011550656802054e-05, "loss": 1.023, "step": 80050 }, { "epoch": 0.2, "learning_rate": 8.011424832653883e-05, "loss": 1.0221, "step": 80055 }, { "epoch": 0.2, "learning_rate": 8.011299008505712e-05, "loss": 1.0245, "step": 80060 }, { "epoch": 0.2, "learning_rate": 8.011173184357543e-05, "loss": 1.0238, "step": 80065 }, { "epoch": 0.2, "learning_rate": 8.011047360209372e-05, "loss": 1.0262, "step": 80070 }, { "epoch": 0.2, "learning_rate": 8.010921536061201e-05, "loss": 1.0231, "step": 80075 }, { "epoch": 0.2, "learning_rate": 8.01079571191303e-05, "loss": 1.0252, "step": 80080 }, { "epoch": 0.2, "learning_rate": 8.01066988776486e-05, "loss": 1.0257, "step": 80085 }, { "epoch": 0.2, "learning_rate": 8.01054406361669e-05, "loss": 1.026, "step": 80090 }, { "epoch": 0.2, "learning_rate": 8.010418239468519e-05, "loss": 1.0236, "step": 80095 }, { "epoch": 0.2, "learning_rate": 8.010292415320348e-05, "loss": 1.0264, "step": 80100 }, { "epoch": 0.2, "learning_rate": 8.010166591172179e-05, "loss": 1.0249, "step": 80105 }, { "epoch": 0.2, "learning_rate": 8.010040767024008e-05, "loss": 1.0245, "step": 80110 }, { "epoch": 0.2, "learning_rate": 8.009914942875837e-05, "loss": 1.0231, "step": 80115 }, { "epoch": 0.2, "learning_rate": 8.009789118727666e-05, "loss": 1.0265, "step": 80120 }, { "epoch": 0.2, "learning_rate": 8.009663294579495e-05, "loss": 1.0238, "step": 80125 }, { "epoch": 0.2, "learning_rate": 8.009537470431326e-05, "loss": 1.0267, "step": 80130 }, { "epoch": 0.2, "learning_rate": 8.009411646283155e-05, "loss": 1.026, "step": 80135 }, { "epoch": 0.2, "learning_rate": 8.009285822134984e-05, "loss": 1.0249, "step": 80140 }, { "epoch": 0.2, "learning_rate": 8.009159997986813e-05, "loss": 1.0263, "step": 80145 }, { "epoch": 0.2, "learning_rate": 8.009034173838644e-05, "loss": 1.022, "step": 80150 }, { "epoch": 0.2, "learning_rate": 8.008908349690473e-05, "loss": 1.0265, "step": 80155 }, { "epoch": 0.2, "learning_rate": 8.008782525542302e-05, "loss": 1.0249, "step": 80160 }, { "epoch": 0.2, "learning_rate": 8.008656701394131e-05, "loss": 1.0229, "step": 80165 }, { "epoch": 0.2, "learning_rate": 8.008530877245962e-05, "loss": 1.026, "step": 80170 }, { "epoch": 0.2, "learning_rate": 8.008405053097791e-05, "loss": 1.0259, "step": 80175 }, { "epoch": 0.2, "learning_rate": 8.00827922894962e-05, "loss": 1.0227, "step": 80180 }, { "epoch": 0.2, "learning_rate": 8.008153404801449e-05, "loss": 1.0246, "step": 80185 }, { "epoch": 0.2, "learning_rate": 8.008027580653278e-05, "loss": 1.0242, "step": 80190 }, { "epoch": 0.2, "learning_rate": 8.007901756505109e-05, "loss": 1.0249, "step": 80195 }, { "epoch": 0.2, "learning_rate": 8.007775932356938e-05, "loss": 1.0212, "step": 80200 }, { "epoch": 0.2, "learning_rate": 8.007650108208767e-05, "loss": 1.0228, "step": 80205 }, { "epoch": 0.2, "learning_rate": 8.007524284060596e-05, "loss": 1.0237, "step": 80210 }, { "epoch": 0.2, "learning_rate": 8.007398459912427e-05, "loss": 1.0248, "step": 80215 }, { "epoch": 0.2, "learning_rate": 8.007272635764256e-05, "loss": 1.0249, "step": 80220 }, { "epoch": 0.2, "learning_rate": 8.007146811616085e-05, "loss": 1.0242, "step": 80225 }, { "epoch": 0.2, "learning_rate": 8.007020987467914e-05, "loss": 1.0262, "step": 80230 }, { "epoch": 0.2, "learning_rate": 8.006895163319745e-05, "loss": 1.025, "step": 80235 }, { "epoch": 0.2, "learning_rate": 8.006769339171574e-05, "loss": 1.023, "step": 80240 }, { "epoch": 0.2, "learning_rate": 8.006643515023403e-05, "loss": 1.0256, "step": 80245 }, { "epoch": 0.2, "learning_rate": 8.006517690875232e-05, "loss": 1.0241, "step": 80250 }, { "epoch": 0.2, "learning_rate": 8.006391866727061e-05, "loss": 1.0441, "step": 80255 }, { "epoch": 0.2, "learning_rate": 8.006266042578893e-05, "loss": 1.0224, "step": 80260 }, { "epoch": 0.2, "learning_rate": 8.006140218430722e-05, "loss": 1.0216, "step": 80265 }, { "epoch": 0.2, "learning_rate": 8.006014394282552e-05, "loss": 1.0252, "step": 80270 }, { "epoch": 0.2, "learning_rate": 8.005888570134381e-05, "loss": 1.0238, "step": 80275 }, { "epoch": 0.2, "learning_rate": 8.00576274598621e-05, "loss": 1.0237, "step": 80280 }, { "epoch": 0.2, "learning_rate": 8.00563692183804e-05, "loss": 1.0253, "step": 80285 }, { "epoch": 0.2, "learning_rate": 8.00551109768987e-05, "loss": 1.0402, "step": 80290 }, { "epoch": 0.2, "learning_rate": 8.005385273541699e-05, "loss": 1.0243, "step": 80295 }, { "epoch": 0.2, "learning_rate": 8.005259449393528e-05, "loss": 1.0225, "step": 80300 }, { "epoch": 0.2, "learning_rate": 8.005133625245358e-05, "loss": 1.0236, "step": 80305 }, { "epoch": 0.2, "learning_rate": 8.005007801097188e-05, "loss": 1.0473, "step": 80310 }, { "epoch": 0.2, "learning_rate": 8.004881976949017e-05, "loss": 1.0221, "step": 80315 }, { "epoch": 0.2, "learning_rate": 8.004756152800846e-05, "loss": 1.0238, "step": 80320 }, { "epoch": 0.2, "learning_rate": 8.004630328652676e-05, "loss": 1.0237, "step": 80325 }, { "epoch": 0.2, "learning_rate": 8.004504504504506e-05, "loss": 1.0263, "step": 80330 }, { "epoch": 0.2, "learning_rate": 8.004378680356335e-05, "loss": 1.0247, "step": 80335 }, { "epoch": 0.2, "learning_rate": 8.004252856208164e-05, "loss": 1.0235, "step": 80340 }, { "epoch": 0.2, "learning_rate": 8.004127032059993e-05, "loss": 1.0268, "step": 80345 }, { "epoch": 0.2, "learning_rate": 8.004001207911824e-05, "loss": 1.0241, "step": 80350 }, { "epoch": 0.2, "learning_rate": 8.003875383763653e-05, "loss": 1.025, "step": 80355 }, { "epoch": 0.2, "learning_rate": 8.003749559615482e-05, "loss": 1.025, "step": 80360 }, { "epoch": 0.2, "learning_rate": 8.003623735467311e-05, "loss": 1.0223, "step": 80365 }, { "epoch": 0.2, "learning_rate": 8.003497911319142e-05, "loss": 1.0247, "step": 80370 }, { "epoch": 0.2, "learning_rate": 8.003372087170971e-05, "loss": 1.0273, "step": 80375 }, { "epoch": 0.2, "learning_rate": 8.0032462630228e-05, "loss": 1.023, "step": 80380 }, { "epoch": 0.2, "learning_rate": 8.003120438874629e-05, "loss": 1.0223, "step": 80385 }, { "epoch": 0.2, "learning_rate": 8.00299461472646e-05, "loss": 1.0244, "step": 80390 }, { "epoch": 0.2, "learning_rate": 8.002868790578289e-05, "loss": 1.023, "step": 80395 }, { "epoch": 0.2, "learning_rate": 8.002742966430118e-05, "loss": 1.0241, "step": 80400 }, { "epoch": 0.2, "learning_rate": 8.002617142281947e-05, "loss": 1.026, "step": 80405 }, { "epoch": 0.2, "learning_rate": 8.002491318133776e-05, "loss": 1.0237, "step": 80410 }, { "epoch": 0.2, "learning_rate": 8.002365493985607e-05, "loss": 1.0235, "step": 80415 }, { "epoch": 0.2, "learning_rate": 8.002239669837436e-05, "loss": 1.0261, "step": 80420 }, { "epoch": 0.2, "learning_rate": 8.002113845689265e-05, "loss": 1.0268, "step": 80425 }, { "epoch": 0.2, "learning_rate": 8.001988021541094e-05, "loss": 1.0267, "step": 80430 }, { "epoch": 0.2, "learning_rate": 8.001862197392925e-05, "loss": 1.0226, "step": 80435 }, { "epoch": 0.2, "learning_rate": 8.001736373244754e-05, "loss": 1.0261, "step": 80440 }, { "epoch": 0.2, "learning_rate": 8.001610549096583e-05, "loss": 1.0241, "step": 80445 }, { "epoch": 0.2, "learning_rate": 8.001484724948412e-05, "loss": 1.0234, "step": 80450 }, { "epoch": 0.2, "learning_rate": 8.001358900800243e-05, "loss": 1.0246, "step": 80455 }, { "epoch": 0.2, "learning_rate": 8.001233076652072e-05, "loss": 1.0238, "step": 80460 }, { "epoch": 0.2, "learning_rate": 8.001107252503901e-05, "loss": 1.0264, "step": 80465 }, { "epoch": 0.2, "learning_rate": 8.00098142835573e-05, "loss": 1.0229, "step": 80470 }, { "epoch": 0.2, "learning_rate": 8.00085560420756e-05, "loss": 1.0228, "step": 80475 }, { "epoch": 0.2, "learning_rate": 8.00072978005939e-05, "loss": 1.0251, "step": 80480 }, { "epoch": 0.2, "learning_rate": 8.000603955911219e-05, "loss": 1.0256, "step": 80485 }, { "epoch": 0.2, "learning_rate": 8.000478131763048e-05, "loss": 1.0238, "step": 80490 }, { "epoch": 0.2, "learning_rate": 8.000352307614877e-05, "loss": 1.0239, "step": 80495 }, { "epoch": 0.2, "learning_rate": 8.000226483466708e-05, "loss": 1.025, "step": 80500 }, { "epoch": 0.2, "learning_rate": 8.000100659318537e-05, "loss": 1.0263, "step": 80505 }, { "epoch": 0.2, "learning_rate": 7.999974835170366e-05, "loss": 1.0247, "step": 80510 }, { "epoch": 0.2, "learning_rate": 7.999849011022195e-05, "loss": 1.0265, "step": 80515 }, { "epoch": 0.2, "learning_rate": 7.999723186874026e-05, "loss": 1.0238, "step": 80520 }, { "epoch": 0.2, "learning_rate": 7.999597362725855e-05, "loss": 1.0251, "step": 80525 }, { "epoch": 0.2, "learning_rate": 7.999471538577684e-05, "loss": 1.0245, "step": 80530 }, { "epoch": 0.2, "learning_rate": 7.999345714429513e-05, "loss": 1.0258, "step": 80535 }, { "epoch": 0.2, "learning_rate": 7.999219890281343e-05, "loss": 1.0238, "step": 80540 }, { "epoch": 0.2, "learning_rate": 7.999094066133173e-05, "loss": 1.0232, "step": 80545 }, { "epoch": 0.2, "learning_rate": 7.998968241985002e-05, "loss": 1.0225, "step": 80550 }, { "epoch": 0.2, "learning_rate": 7.998842417836831e-05, "loss": 1.0241, "step": 80555 }, { "epoch": 0.2, "learning_rate": 7.99871659368866e-05, "loss": 1.0237, "step": 80560 }, { "epoch": 0.2, "learning_rate": 7.998590769540491e-05, "loss": 1.0243, "step": 80565 }, { "epoch": 0.2, "learning_rate": 7.99846494539232e-05, "loss": 1.0253, "step": 80570 }, { "epoch": 0.2, "learning_rate": 7.99833912124415e-05, "loss": 1.0251, "step": 80575 }, { "epoch": 0.2, "learning_rate": 7.998213297095979e-05, "loss": 1.0246, "step": 80580 }, { "epoch": 0.2, "learning_rate": 7.998087472947809e-05, "loss": 1.024, "step": 80585 }, { "epoch": 0.2, "learning_rate": 7.997961648799638e-05, "loss": 1.023, "step": 80590 }, { "epoch": 0.2, "learning_rate": 7.997835824651467e-05, "loss": 1.0264, "step": 80595 }, { "epoch": 0.2, "learning_rate": 7.997710000503297e-05, "loss": 1.0243, "step": 80600 }, { "epoch": 0.2, "learning_rate": 7.997584176355126e-05, "loss": 1.0235, "step": 80605 }, { "epoch": 0.2, "learning_rate": 7.997458352206956e-05, "loss": 1.0211, "step": 80610 }, { "epoch": 0.2, "learning_rate": 7.997332528058785e-05, "loss": 1.0267, "step": 80615 }, { "epoch": 0.2, "learning_rate": 7.997206703910615e-05, "loss": 1.0254, "step": 80620 }, { "epoch": 0.2, "learning_rate": 7.997080879762444e-05, "loss": 1.0247, "step": 80625 }, { "epoch": 0.2, "learning_rate": 7.996955055614274e-05, "loss": 1.0255, "step": 80630 }, { "epoch": 0.2, "learning_rate": 7.996829231466103e-05, "loss": 1.0274, "step": 80635 }, { "epoch": 0.2, "learning_rate": 7.996703407317933e-05, "loss": 1.0238, "step": 80640 }, { "epoch": 0.2, "learning_rate": 7.996577583169762e-05, "loss": 1.0261, "step": 80645 }, { "epoch": 0.2, "learning_rate": 7.996451759021591e-05, "loss": 1.0267, "step": 80650 }, { "epoch": 0.2, "learning_rate": 7.996325934873421e-05, "loss": 1.0248, "step": 80655 }, { "epoch": 0.2, "learning_rate": 7.99620011072525e-05, "loss": 1.0233, "step": 80660 }, { "epoch": 0.2, "learning_rate": 7.99607428657708e-05, "loss": 1.0245, "step": 80665 }, { "epoch": 0.2, "learning_rate": 7.995948462428909e-05, "loss": 1.0238, "step": 80670 }, { "epoch": 0.2, "learning_rate": 7.99582263828074e-05, "loss": 1.0259, "step": 80675 }, { "epoch": 0.2, "learning_rate": 7.995696814132568e-05, "loss": 1.0224, "step": 80680 }, { "epoch": 0.2, "learning_rate": 7.995570989984398e-05, "loss": 1.0251, "step": 80685 }, { "epoch": 0.2, "learning_rate": 7.995445165836227e-05, "loss": 1.0232, "step": 80690 }, { "epoch": 0.2, "learning_rate": 7.995319341688057e-05, "loss": 1.0239, "step": 80695 }, { "epoch": 0.2, "learning_rate": 7.995193517539886e-05, "loss": 1.0259, "step": 80700 }, { "epoch": 0.2, "learning_rate": 7.995067693391716e-05, "loss": 1.0221, "step": 80705 }, { "epoch": 0.2, "learning_rate": 7.994941869243545e-05, "loss": 1.026, "step": 80710 }, { "epoch": 0.2, "learning_rate": 7.994816045095374e-05, "loss": 1.0241, "step": 80715 }, { "epoch": 0.2, "learning_rate": 7.994690220947204e-05, "loss": 1.0252, "step": 80720 }, { "epoch": 0.2, "learning_rate": 7.994564396799034e-05, "loss": 1.0243, "step": 80725 }, { "epoch": 0.2, "learning_rate": 7.994438572650863e-05, "loss": 1.0247, "step": 80730 }, { "epoch": 0.2, "learning_rate": 7.994312748502692e-05, "loss": 1.0224, "step": 80735 }, { "epoch": 0.2, "learning_rate": 7.994186924354522e-05, "loss": 1.025, "step": 80740 }, { "epoch": 0.2, "learning_rate": 7.994061100206352e-05, "loss": 1.0253, "step": 80745 }, { "epoch": 0.2, "learning_rate": 7.993935276058181e-05, "loss": 1.0232, "step": 80750 }, { "epoch": 0.2, "learning_rate": 7.99380945191001e-05, "loss": 1.0251, "step": 80755 }, { "epoch": 0.2, "learning_rate": 7.99368362776184e-05, "loss": 1.0251, "step": 80760 }, { "epoch": 0.2, "learning_rate": 7.993557803613671e-05, "loss": 1.0248, "step": 80765 }, { "epoch": 0.2, "learning_rate": 7.9934319794655e-05, "loss": 1.0265, "step": 80770 }, { "epoch": 0.2, "learning_rate": 7.993306155317329e-05, "loss": 1.0244, "step": 80775 }, { "epoch": 0.2, "learning_rate": 7.993180331169158e-05, "loss": 1.024, "step": 80780 }, { "epoch": 0.2, "learning_rate": 7.993054507020989e-05, "loss": 1.024, "step": 80785 }, { "epoch": 0.2, "learning_rate": 7.992928682872818e-05, "loss": 1.0261, "step": 80790 }, { "epoch": 0.2, "learning_rate": 7.992802858724647e-05, "loss": 1.0243, "step": 80795 }, { "epoch": 0.2, "learning_rate": 7.992677034576476e-05, "loss": 1.0254, "step": 80800 }, { "epoch": 0.2, "learning_rate": 7.992551210428306e-05, "loss": 1.0475, "step": 80805 }, { "epoch": 0.2, "learning_rate": 7.992425386280136e-05, "loss": 1.0228, "step": 80810 }, { "epoch": 0.2, "learning_rate": 7.992299562131965e-05, "loss": 1.0251, "step": 80815 }, { "epoch": 0.2, "learning_rate": 7.992173737983794e-05, "loss": 1.0246, "step": 80820 }, { "epoch": 0.2, "learning_rate": 7.992047913835624e-05, "loss": 1.0254, "step": 80825 }, { "epoch": 0.2, "learning_rate": 7.991922089687454e-05, "loss": 1.0232, "step": 80830 }, { "epoch": 0.2, "learning_rate": 7.991796265539283e-05, "loss": 1.0234, "step": 80835 }, { "epoch": 0.2, "learning_rate": 7.991670441391112e-05, "loss": 1.0244, "step": 80840 }, { "epoch": 0.2, "learning_rate": 7.991544617242942e-05, "loss": 1.0245, "step": 80845 }, { "epoch": 0.2, "learning_rate": 7.991418793094772e-05, "loss": 1.0265, "step": 80850 }, { "epoch": 0.2, "learning_rate": 7.991292968946601e-05, "loss": 1.0226, "step": 80855 }, { "epoch": 0.2, "learning_rate": 7.99116714479843e-05, "loss": 1.0254, "step": 80860 }, { "epoch": 0.2, "learning_rate": 7.99104132065026e-05, "loss": 1.024, "step": 80865 }, { "epoch": 0.2, "learning_rate": 7.990915496502089e-05, "loss": 1.0261, "step": 80870 }, { "epoch": 0.2, "learning_rate": 7.990789672353919e-05, "loss": 1.0229, "step": 80875 }, { "epoch": 0.2, "learning_rate": 7.990663848205748e-05, "loss": 1.0254, "step": 80880 }, { "epoch": 0.2, "learning_rate": 7.990538024057578e-05, "loss": 1.0232, "step": 80885 }, { "epoch": 0.2, "learning_rate": 7.990412199909407e-05, "loss": 1.0262, "step": 80890 }, { "epoch": 0.2, "learning_rate": 7.990286375761237e-05, "loss": 1.0271, "step": 80895 }, { "epoch": 0.2, "learning_rate": 7.990160551613066e-05, "loss": 1.0228, "step": 80900 }, { "epoch": 0.2, "learning_rate": 7.990034727464896e-05, "loss": 1.0248, "step": 80905 }, { "epoch": 0.2, "learning_rate": 7.989908903316725e-05, "loss": 1.0227, "step": 80910 }, { "epoch": 0.2, "learning_rate": 7.989783079168555e-05, "loss": 1.024, "step": 80915 }, { "epoch": 0.2, "learning_rate": 7.989657255020384e-05, "loss": 1.0244, "step": 80920 }, { "epoch": 0.2, "learning_rate": 7.989531430872214e-05, "loss": 1.0252, "step": 80925 }, { "epoch": 0.2, "learning_rate": 7.989405606724043e-05, "loss": 1.0259, "step": 80930 }, { "epoch": 0.2, "learning_rate": 7.989279782575872e-05, "loss": 1.0263, "step": 80935 }, { "epoch": 0.2, "learning_rate": 7.989153958427702e-05, "loss": 1.0242, "step": 80940 }, { "epoch": 0.2, "learning_rate": 7.989028134279532e-05, "loss": 1.0215, "step": 80945 }, { "epoch": 0.2, "learning_rate": 7.988902310131361e-05, "loss": 1.0239, "step": 80950 }, { "epoch": 0.2, "learning_rate": 7.98877648598319e-05, "loss": 1.0256, "step": 80955 }, { "epoch": 0.2, "learning_rate": 7.98865066183502e-05, "loss": 1.0276, "step": 80960 }, { "epoch": 0.2, "learning_rate": 7.98852483768685e-05, "loss": 1.0248, "step": 80965 }, { "epoch": 0.2, "learning_rate": 7.988399013538679e-05, "loss": 1.0236, "step": 80970 }, { "epoch": 0.2, "learning_rate": 7.988273189390508e-05, "loss": 1.0257, "step": 80975 }, { "epoch": 0.2, "learning_rate": 7.988147365242338e-05, "loss": 1.0262, "step": 80980 }, { "epoch": 0.2, "learning_rate": 7.988021541094168e-05, "loss": 1.0261, "step": 80985 }, { "epoch": 0.2, "learning_rate": 7.987895716945997e-05, "loss": 1.0243, "step": 80990 }, { "epoch": 0.2, "learning_rate": 7.987769892797826e-05, "loss": 1.025, "step": 80995 }, { "epoch": 0.2, "learning_rate": 7.987644068649655e-05, "loss": 1.0244, "step": 81000 }, { "epoch": 0.2, "learning_rate": 7.987518244501486e-05, "loss": 1.0238, "step": 81005 }, { "epoch": 0.2, "learning_rate": 7.987392420353315e-05, "loss": 1.0232, "step": 81010 }, { "epoch": 0.2, "learning_rate": 7.987266596205144e-05, "loss": 1.0235, "step": 81015 }, { "epoch": 0.2, "learning_rate": 7.987140772056973e-05, "loss": 1.0234, "step": 81020 }, { "epoch": 0.2, "learning_rate": 7.987014947908804e-05, "loss": 1.0228, "step": 81025 }, { "epoch": 0.2, "learning_rate": 7.986889123760633e-05, "loss": 1.0266, "step": 81030 }, { "epoch": 0.2, "learning_rate": 7.986763299612462e-05, "loss": 1.025, "step": 81035 }, { "epoch": 0.2, "learning_rate": 7.986637475464291e-05, "loss": 1.0237, "step": 81040 }, { "epoch": 0.2, "learning_rate": 7.986511651316122e-05, "loss": 1.0247, "step": 81045 }, { "epoch": 0.2, "learning_rate": 7.986385827167951e-05, "loss": 1.0231, "step": 81050 }, { "epoch": 0.2, "learning_rate": 7.98626000301978e-05, "loss": 1.025, "step": 81055 }, { "epoch": 0.2, "learning_rate": 7.986134178871609e-05, "loss": 1.022, "step": 81060 }, { "epoch": 0.2, "learning_rate": 7.986008354723438e-05, "loss": 1.0424, "step": 81065 }, { "epoch": 0.2, "learning_rate": 7.985882530575269e-05, "loss": 1.0248, "step": 81070 }, { "epoch": 0.2, "learning_rate": 7.985756706427098e-05, "loss": 1.0254, "step": 81075 }, { "epoch": 0.2, "learning_rate": 7.985630882278927e-05, "loss": 1.0232, "step": 81080 }, { "epoch": 0.2, "learning_rate": 7.985505058130756e-05, "loss": 1.024, "step": 81085 }, { "epoch": 0.2, "learning_rate": 7.985379233982587e-05, "loss": 1.0255, "step": 81090 }, { "epoch": 0.2, "learning_rate": 7.985253409834416e-05, "loss": 1.0253, "step": 81095 }, { "epoch": 0.2, "learning_rate": 7.985127585686245e-05, "loss": 1.0254, "step": 81100 }, { "epoch": 0.2, "learning_rate": 7.985001761538074e-05, "loss": 1.0243, "step": 81105 }, { "epoch": 0.2, "learning_rate": 7.984875937389905e-05, "loss": 1.0244, "step": 81110 }, { "epoch": 0.2, "learning_rate": 7.984750113241734e-05, "loss": 1.0247, "step": 81115 }, { "epoch": 0.2, "learning_rate": 7.984624289093563e-05, "loss": 1.0267, "step": 81120 }, { "epoch": 0.2, "learning_rate": 7.984498464945392e-05, "loss": 1.026, "step": 81125 }, { "epoch": 0.2, "learning_rate": 7.984372640797221e-05, "loss": 1.0271, "step": 81130 }, { "epoch": 0.2, "learning_rate": 7.984246816649052e-05, "loss": 1.0225, "step": 81135 }, { "epoch": 0.2, "learning_rate": 7.984120992500881e-05, "loss": 1.0239, "step": 81140 }, { "epoch": 0.2, "learning_rate": 7.98399516835271e-05, "loss": 1.0226, "step": 81145 }, { "epoch": 0.2, "learning_rate": 7.983869344204539e-05, "loss": 1.0246, "step": 81150 }, { "epoch": 0.2, "learning_rate": 7.98374352005637e-05, "loss": 1.0233, "step": 81155 }, { "epoch": 0.2, "learning_rate": 7.983617695908199e-05, "loss": 1.025, "step": 81160 }, { "epoch": 0.2, "learning_rate": 7.983491871760028e-05, "loss": 1.0255, "step": 81165 }, { "epoch": 0.2, "learning_rate": 7.983366047611857e-05, "loss": 1.0251, "step": 81170 }, { "epoch": 0.2, "learning_rate": 7.983240223463688e-05, "loss": 1.0428, "step": 81175 }, { "epoch": 0.2, "learning_rate": 7.983114399315517e-05, "loss": 1.0251, "step": 81180 }, { "epoch": 0.2, "learning_rate": 7.982988575167346e-05, "loss": 1.0239, "step": 81185 }, { "epoch": 0.2, "learning_rate": 7.982862751019175e-05, "loss": 1.0238, "step": 81190 }, { "epoch": 0.2, "learning_rate": 7.982736926871004e-05, "loss": 1.0244, "step": 81195 }, { "epoch": 0.2, "learning_rate": 7.982611102722835e-05, "loss": 1.0245, "step": 81200 }, { "epoch": 0.2, "learning_rate": 7.982485278574664e-05, "loss": 1.0232, "step": 81205 }, { "epoch": 0.2, "learning_rate": 7.982359454426493e-05, "loss": 1.0252, "step": 81210 }, { "epoch": 0.2, "learning_rate": 7.982233630278322e-05, "loss": 1.023, "step": 81215 }, { "epoch": 0.2, "learning_rate": 7.982107806130153e-05, "loss": 1.0246, "step": 81220 }, { "epoch": 0.2, "learning_rate": 7.981981981981982e-05, "loss": 1.0237, "step": 81225 }, { "epoch": 0.2, "learning_rate": 7.981856157833811e-05, "loss": 1.028, "step": 81230 }, { "epoch": 0.2, "learning_rate": 7.98173033368564e-05, "loss": 1.0257, "step": 81235 }, { "epoch": 0.2, "learning_rate": 7.981604509537471e-05, "loss": 1.0261, "step": 81240 }, { "epoch": 0.2, "learning_rate": 7.9814786853893e-05, "loss": 1.024, "step": 81245 }, { "epoch": 0.2, "learning_rate": 7.981352861241129e-05, "loss": 1.0254, "step": 81250 }, { "epoch": 0.2, "learning_rate": 7.981227037092958e-05, "loss": 1.0246, "step": 81255 }, { "epoch": 0.2, "learning_rate": 7.981101212944789e-05, "loss": 1.0235, "step": 81260 }, { "epoch": 0.2, "learning_rate": 7.980975388796618e-05, "loss": 1.0251, "step": 81265 }, { "epoch": 0.2, "learning_rate": 7.980849564648449e-05, "loss": 1.0242, "step": 81270 }, { "epoch": 0.2, "learning_rate": 7.980723740500278e-05, "loss": 1.0238, "step": 81275 }, { "epoch": 0.2, "learning_rate": 7.980597916352107e-05, "loss": 1.0236, "step": 81280 }, { "epoch": 0.2, "learning_rate": 7.980472092203936e-05, "loss": 1.0269, "step": 81285 }, { "epoch": 0.2, "learning_rate": 7.980346268055767e-05, "loss": 1.0233, "step": 81290 }, { "epoch": 0.2, "learning_rate": 7.980220443907596e-05, "loss": 1.0254, "step": 81295 }, { "epoch": 0.2, "learning_rate": 7.980094619759425e-05, "loss": 1.0255, "step": 81300 }, { "epoch": 0.2, "learning_rate": 7.979968795611254e-05, "loss": 1.0256, "step": 81305 }, { "epoch": 0.2, "learning_rate": 7.979842971463085e-05, "loss": 1.0241, "step": 81310 }, { "epoch": 0.2, "learning_rate": 7.979717147314914e-05, "loss": 1.0255, "step": 81315 }, { "epoch": 0.2, "learning_rate": 7.979591323166743e-05, "loss": 1.0248, "step": 81320 }, { "epoch": 0.2, "learning_rate": 7.979465499018572e-05, "loss": 1.0261, "step": 81325 }, { "epoch": 0.2, "learning_rate": 7.979339674870401e-05, "loss": 1.024, "step": 81330 }, { "epoch": 0.2, "learning_rate": 7.979213850722232e-05, "loss": 1.0254, "step": 81335 }, { "epoch": 0.2, "learning_rate": 7.979088026574061e-05, "loss": 1.0248, "step": 81340 }, { "epoch": 0.2, "learning_rate": 7.97896220242589e-05, "loss": 1.0237, "step": 81345 }, { "epoch": 0.2, "learning_rate": 7.978836378277719e-05, "loss": 1.0233, "step": 81350 }, { "epoch": 0.2, "learning_rate": 7.97871055412955e-05, "loss": 1.0257, "step": 81355 }, { "epoch": 0.2, "learning_rate": 7.978584729981379e-05, "loss": 1.0241, "step": 81360 }, { "epoch": 0.2, "learning_rate": 7.978458905833208e-05, "loss": 1.022, "step": 81365 }, { "epoch": 0.2, "learning_rate": 7.978333081685037e-05, "loss": 1.0245, "step": 81370 }, { "epoch": 0.2, "learning_rate": 7.978207257536868e-05, "loss": 1.0223, "step": 81375 }, { "epoch": 0.2, "learning_rate": 7.978081433388697e-05, "loss": 1.0255, "step": 81380 }, { "epoch": 0.2, "learning_rate": 7.977955609240526e-05, "loss": 1.0234, "step": 81385 }, { "epoch": 0.2, "learning_rate": 7.977829785092355e-05, "loss": 1.0222, "step": 81390 }, { "epoch": 0.2, "learning_rate": 7.977703960944184e-05, "loss": 1.0252, "step": 81395 }, { "epoch": 0.2, "learning_rate": 7.977578136796015e-05, "loss": 1.0247, "step": 81400 }, { "epoch": 0.2, "learning_rate": 7.977452312647844e-05, "loss": 1.0243, "step": 81405 }, { "epoch": 0.2, "learning_rate": 7.977326488499673e-05, "loss": 1.023, "step": 81410 }, { "epoch": 0.2, "learning_rate": 7.977200664351502e-05, "loss": 1.0249, "step": 81415 }, { "epoch": 0.2, "learning_rate": 7.977074840203333e-05, "loss": 1.0256, "step": 81420 }, { "epoch": 0.2, "learning_rate": 7.976949016055162e-05, "loss": 1.0246, "step": 81425 }, { "epoch": 0.2, "learning_rate": 7.976823191906991e-05, "loss": 1.0239, "step": 81430 }, { "epoch": 0.2, "learning_rate": 7.97669736775882e-05, "loss": 1.0266, "step": 81435 }, { "epoch": 0.2, "learning_rate": 7.976571543610651e-05, "loss": 1.0243, "step": 81440 }, { "epoch": 0.2, "learning_rate": 7.97644571946248e-05, "loss": 1.0404, "step": 81445 }, { "epoch": 0.2, "learning_rate": 7.976319895314309e-05, "loss": 1.0252, "step": 81450 }, { "epoch": 0.2, "learning_rate": 7.976194071166138e-05, "loss": 1.0219, "step": 81455 }, { "epoch": 0.2, "learning_rate": 7.976068247017968e-05, "loss": 1.0255, "step": 81460 }, { "epoch": 0.2, "learning_rate": 7.975942422869798e-05, "loss": 1.0235, "step": 81465 }, { "epoch": 0.2, "learning_rate": 7.975816598721627e-05, "loss": 1.0473, "step": 81470 }, { "epoch": 0.2, "learning_rate": 7.975690774573456e-05, "loss": 1.0238, "step": 81475 }, { "epoch": 0.2, "learning_rate": 7.975564950425286e-05, "loss": 1.0248, "step": 81480 }, { "epoch": 0.2, "learning_rate": 7.975439126277116e-05, "loss": 1.0257, "step": 81485 }, { "epoch": 0.2, "learning_rate": 7.975313302128945e-05, "loss": 1.026, "step": 81490 }, { "epoch": 0.2, "learning_rate": 7.975187477980774e-05, "loss": 1.0256, "step": 81495 }, { "epoch": 0.2, "learning_rate": 7.975061653832604e-05, "loss": 1.0236, "step": 81500 }, { "epoch": 0.2, "learning_rate": 7.974935829684434e-05, "loss": 1.0253, "step": 81505 }, { "epoch": 0.2, "learning_rate": 7.974810005536263e-05, "loss": 1.0234, "step": 81510 }, { "epoch": 0.2, "learning_rate": 7.974684181388092e-05, "loss": 1.0255, "step": 81515 }, { "epoch": 0.2, "learning_rate": 7.974558357239922e-05, "loss": 1.0257, "step": 81520 }, { "epoch": 0.2, "learning_rate": 7.974432533091751e-05, "loss": 1.025, "step": 81525 }, { "epoch": 0.2, "learning_rate": 7.974306708943581e-05, "loss": 1.0237, "step": 81530 }, { "epoch": 0.2, "learning_rate": 7.97418088479541e-05, "loss": 1.0251, "step": 81535 }, { "epoch": 0.2, "learning_rate": 7.97405506064724e-05, "loss": 1.0267, "step": 81540 }, { "epoch": 0.2, "learning_rate": 7.973929236499069e-05, "loss": 1.0255, "step": 81545 }, { "epoch": 0.2, "learning_rate": 7.973803412350899e-05, "loss": 1.0258, "step": 81550 }, { "epoch": 0.2, "learning_rate": 7.973677588202728e-05, "loss": 1.0242, "step": 81555 }, { "epoch": 0.2, "learning_rate": 7.973551764054558e-05, "loss": 1.0226, "step": 81560 }, { "epoch": 0.2, "learning_rate": 7.973425939906387e-05, "loss": 1.0237, "step": 81565 }, { "epoch": 0.2, "learning_rate": 7.973300115758217e-05, "loss": 1.0232, "step": 81570 }, { "epoch": 0.2, "learning_rate": 7.973174291610046e-05, "loss": 1.0251, "step": 81575 }, { "epoch": 0.2, "learning_rate": 7.973048467461876e-05, "loss": 1.0249, "step": 81580 }, { "epoch": 0.2, "learning_rate": 7.972922643313705e-05, "loss": 1.0272, "step": 81585 }, { "epoch": 0.2, "learning_rate": 7.972796819165534e-05, "loss": 1.0237, "step": 81590 }, { "epoch": 0.2, "learning_rate": 7.972670995017364e-05, "loss": 1.0249, "step": 81595 }, { "epoch": 0.2, "learning_rate": 7.972545170869194e-05, "loss": 1.025, "step": 81600 }, { "epoch": 0.2, "learning_rate": 7.972419346721023e-05, "loss": 1.0263, "step": 81605 }, { "epoch": 0.2, "learning_rate": 7.972293522572852e-05, "loss": 1.0251, "step": 81610 }, { "epoch": 0.2, "learning_rate": 7.972167698424682e-05, "loss": 1.024, "step": 81615 }, { "epoch": 0.2, "learning_rate": 7.972041874276512e-05, "loss": 1.0231, "step": 81620 }, { "epoch": 0.2, "learning_rate": 7.97191605012834e-05, "loss": 1.0229, "step": 81625 }, { "epoch": 0.2, "learning_rate": 7.97179022598017e-05, "loss": 1.0245, "step": 81630 }, { "epoch": 0.2, "learning_rate": 7.971664401832e-05, "loss": 1.0256, "step": 81635 }, { "epoch": 0.2, "learning_rate": 7.97153857768383e-05, "loss": 1.0447, "step": 81640 }, { "epoch": 0.2, "learning_rate": 7.971412753535659e-05, "loss": 1.0249, "step": 81645 }, { "epoch": 0.2, "learning_rate": 7.971286929387488e-05, "loss": 1.026, "step": 81650 }, { "epoch": 0.2, "learning_rate": 7.971161105239317e-05, "loss": 1.0236, "step": 81655 }, { "epoch": 0.2, "learning_rate": 7.971035281091147e-05, "loss": 1.0252, "step": 81660 }, { "epoch": 0.2, "learning_rate": 7.970909456942977e-05, "loss": 1.023, "step": 81665 }, { "epoch": 0.21, "learning_rate": 7.970783632794806e-05, "loss": 1.0234, "step": 81670 }, { "epoch": 0.21, "learning_rate": 7.970657808646635e-05, "loss": 1.0238, "step": 81675 }, { "epoch": 0.21, "learning_rate": 7.970531984498465e-05, "loss": 1.0247, "step": 81680 }, { "epoch": 0.21, "learning_rate": 7.970406160350295e-05, "loss": 1.0227, "step": 81685 }, { "epoch": 0.21, "learning_rate": 7.970280336202124e-05, "loss": 1.0238, "step": 81690 }, { "epoch": 0.21, "learning_rate": 7.970154512053953e-05, "loss": 1.023, "step": 81695 }, { "epoch": 0.21, "learning_rate": 7.970028687905783e-05, "loss": 1.0248, "step": 81700 }, { "epoch": 0.21, "learning_rate": 7.969902863757613e-05, "loss": 1.045, "step": 81705 }, { "epoch": 0.21, "learning_rate": 7.969777039609442e-05, "loss": 1.0237, "step": 81710 }, { "epoch": 0.21, "learning_rate": 7.969651215461271e-05, "loss": 1.0262, "step": 81715 }, { "epoch": 0.21, "learning_rate": 7.9695253913131e-05, "loss": 1.0247, "step": 81720 }, { "epoch": 0.21, "learning_rate": 7.96939956716493e-05, "loss": 1.0246, "step": 81725 }, { "epoch": 0.21, "learning_rate": 7.96927374301676e-05, "loss": 1.0256, "step": 81730 }, { "epoch": 0.21, "learning_rate": 7.969147918868589e-05, "loss": 1.0258, "step": 81735 }, { "epoch": 0.21, "learning_rate": 7.969022094720418e-05, "loss": 1.0252, "step": 81740 }, { "epoch": 0.21, "learning_rate": 7.968896270572249e-05, "loss": 1.025, "step": 81745 }, { "epoch": 0.21, "learning_rate": 7.968770446424078e-05, "loss": 1.0235, "step": 81750 }, { "epoch": 0.21, "learning_rate": 7.968644622275907e-05, "loss": 1.026, "step": 81755 }, { "epoch": 0.21, "learning_rate": 7.968518798127737e-05, "loss": 1.0228, "step": 81760 }, { "epoch": 0.21, "learning_rate": 7.968392973979567e-05, "loss": 1.0239, "step": 81765 }, { "epoch": 0.21, "learning_rate": 7.968267149831397e-05, "loss": 1.0231, "step": 81770 }, { "epoch": 0.21, "learning_rate": 7.968141325683226e-05, "loss": 1.0238, "step": 81775 }, { "epoch": 0.21, "learning_rate": 7.968015501535055e-05, "loss": 1.0244, "step": 81780 }, { "epoch": 0.21, "learning_rate": 7.967889677386885e-05, "loss": 1.0228, "step": 81785 }, { "epoch": 0.21, "learning_rate": 7.967763853238714e-05, "loss": 1.0231, "step": 81790 }, { "epoch": 0.21, "learning_rate": 7.967638029090544e-05, "loss": 1.0248, "step": 81795 }, { "epoch": 0.21, "learning_rate": 7.967512204942373e-05, "loss": 1.025, "step": 81800 }, { "epoch": 0.21, "learning_rate": 7.967386380794203e-05, "loss": 1.0237, "step": 81805 }, { "epoch": 0.21, "learning_rate": 7.967260556646032e-05, "loss": 1.0247, "step": 81810 }, { "epoch": 0.21, "learning_rate": 7.967134732497862e-05, "loss": 1.0455, "step": 81815 }, { "epoch": 0.21, "learning_rate": 7.967008908349691e-05, "loss": 1.0251, "step": 81820 }, { "epoch": 0.21, "learning_rate": 7.96688308420152e-05, "loss": 1.0264, "step": 81825 }, { "epoch": 0.21, "learning_rate": 7.96675726005335e-05, "loss": 1.0231, "step": 81830 }, { "epoch": 0.21, "learning_rate": 7.96663143590518e-05, "loss": 1.023, "step": 81835 }, { "epoch": 0.21, "learning_rate": 7.96650561175701e-05, "loss": 1.0237, "step": 81840 }, { "epoch": 0.21, "learning_rate": 7.966379787608839e-05, "loss": 1.0269, "step": 81845 }, { "epoch": 0.21, "learning_rate": 7.966253963460668e-05, "loss": 1.0258, "step": 81850 }, { "epoch": 0.21, "learning_rate": 7.966128139312497e-05, "loss": 1.0266, "step": 81855 }, { "epoch": 0.21, "learning_rate": 7.966002315164327e-05, "loss": 1.0242, "step": 81860 }, { "epoch": 0.21, "learning_rate": 7.965876491016157e-05, "loss": 1.0244, "step": 81865 }, { "epoch": 0.21, "learning_rate": 7.965750666867986e-05, "loss": 1.045, "step": 81870 }, { "epoch": 0.21, "learning_rate": 7.965624842719815e-05, "loss": 1.0244, "step": 81875 }, { "epoch": 0.21, "learning_rate": 7.965499018571645e-05, "loss": 1.0233, "step": 81880 }, { "epoch": 0.21, "learning_rate": 7.965373194423475e-05, "loss": 1.0241, "step": 81885 }, { "epoch": 0.21, "learning_rate": 7.965247370275304e-05, "loss": 1.0253, "step": 81890 }, { "epoch": 0.21, "learning_rate": 7.965121546127133e-05, "loss": 1.0244, "step": 81895 }, { "epoch": 0.21, "learning_rate": 7.964995721978963e-05, "loss": 1.024, "step": 81900 }, { "epoch": 0.21, "learning_rate": 7.964869897830793e-05, "loss": 1.0237, "step": 81905 }, { "epoch": 0.21, "learning_rate": 7.964744073682622e-05, "loss": 1.0434, "step": 81910 }, { "epoch": 0.21, "learning_rate": 7.964618249534451e-05, "loss": 1.0236, "step": 81915 }, { "epoch": 0.21, "learning_rate": 7.96449242538628e-05, "loss": 1.0237, "step": 81920 }, { "epoch": 0.21, "learning_rate": 7.96436660123811e-05, "loss": 1.0236, "step": 81925 }, { "epoch": 0.21, "learning_rate": 7.96424077708994e-05, "loss": 1.0252, "step": 81930 }, { "epoch": 0.21, "learning_rate": 7.964114952941769e-05, "loss": 1.0246, "step": 81935 }, { "epoch": 0.21, "learning_rate": 7.963989128793598e-05, "loss": 1.0245, "step": 81940 }, { "epoch": 0.21, "learning_rate": 7.963863304645429e-05, "loss": 1.0261, "step": 81945 }, { "epoch": 0.21, "learning_rate": 7.963737480497258e-05, "loss": 1.024, "step": 81950 }, { "epoch": 0.21, "learning_rate": 7.963611656349087e-05, "loss": 1.0274, "step": 81955 }, { "epoch": 0.21, "learning_rate": 7.963485832200916e-05, "loss": 1.0267, "step": 81960 }, { "epoch": 0.21, "learning_rate": 7.963360008052747e-05, "loss": 1.0272, "step": 81965 }, { "epoch": 0.21, "learning_rate": 7.963234183904576e-05, "loss": 1.0245, "step": 81970 }, { "epoch": 0.21, "learning_rate": 7.963108359756405e-05, "loss": 1.0254, "step": 81975 }, { "epoch": 0.21, "learning_rate": 7.962982535608234e-05, "loss": 1.0208, "step": 81980 }, { "epoch": 0.21, "learning_rate": 7.962856711460063e-05, "loss": 1.0251, "step": 81985 }, { "epoch": 0.21, "learning_rate": 7.962730887311894e-05, "loss": 1.022, "step": 81990 }, { "epoch": 0.21, "learning_rate": 7.962605063163723e-05, "loss": 1.0236, "step": 81995 }, { "epoch": 0.21, "learning_rate": 7.962479239015552e-05, "loss": 1.0247, "step": 82000 }, { "epoch": 0.21, "learning_rate": 7.962353414867381e-05, "loss": 1.0259, "step": 82005 }, { "epoch": 0.21, "learning_rate": 7.962227590719212e-05, "loss": 1.0356, "step": 82010 }, { "epoch": 0.21, "learning_rate": 7.962101766571041e-05, "loss": 1.0268, "step": 82015 }, { "epoch": 0.21, "learning_rate": 7.96197594242287e-05, "loss": 1.0258, "step": 82020 }, { "epoch": 0.21, "learning_rate": 7.961850118274699e-05, "loss": 1.0257, "step": 82025 }, { "epoch": 0.21, "learning_rate": 7.96172429412653e-05, "loss": 1.0263, "step": 82030 }, { "epoch": 0.21, "learning_rate": 7.961598469978359e-05, "loss": 1.0255, "step": 82035 }, { "epoch": 0.21, "learning_rate": 7.961472645830188e-05, "loss": 1.026, "step": 82040 }, { "epoch": 0.21, "learning_rate": 7.961346821682017e-05, "loss": 1.0255, "step": 82045 }, { "epoch": 0.21, "learning_rate": 7.961220997533846e-05, "loss": 1.0266, "step": 82050 }, { "epoch": 0.21, "learning_rate": 7.961095173385677e-05, "loss": 1.0264, "step": 82055 }, { "epoch": 0.21, "learning_rate": 7.960969349237506e-05, "loss": 1.0245, "step": 82060 }, { "epoch": 0.21, "learning_rate": 7.960843525089335e-05, "loss": 1.023, "step": 82065 }, { "epoch": 0.21, "learning_rate": 7.960717700941164e-05, "loss": 1.0242, "step": 82070 }, { "epoch": 0.21, "learning_rate": 7.960591876792995e-05, "loss": 1.0248, "step": 82075 }, { "epoch": 0.21, "learning_rate": 7.960466052644824e-05, "loss": 1.0265, "step": 82080 }, { "epoch": 0.21, "learning_rate": 7.960340228496653e-05, "loss": 1.0253, "step": 82085 }, { "epoch": 0.21, "learning_rate": 7.960214404348482e-05, "loss": 1.021, "step": 82090 }, { "epoch": 0.21, "learning_rate": 7.960088580200313e-05, "loss": 1.0231, "step": 82095 }, { "epoch": 0.21, "learning_rate": 7.959962756052142e-05, "loss": 1.0243, "step": 82100 }, { "epoch": 0.21, "learning_rate": 7.959836931903971e-05, "loss": 1.0262, "step": 82105 }, { "epoch": 0.21, "learning_rate": 7.9597111077558e-05, "loss": 1.025, "step": 82110 }, { "epoch": 0.21, "learning_rate": 7.95958528360763e-05, "loss": 1.0242, "step": 82115 }, { "epoch": 0.21, "learning_rate": 7.95945945945946e-05, "loss": 1.0239, "step": 82120 }, { "epoch": 0.21, "learning_rate": 7.959333635311289e-05, "loss": 1.0234, "step": 82125 }, { "epoch": 0.21, "learning_rate": 7.959207811163118e-05, "loss": 1.0249, "step": 82130 }, { "epoch": 0.21, "learning_rate": 7.959081987014947e-05, "loss": 1.025, "step": 82135 }, { "epoch": 0.21, "learning_rate": 7.958956162866778e-05, "loss": 1.021, "step": 82140 }, { "epoch": 0.21, "learning_rate": 7.958830338718607e-05, "loss": 1.0264, "step": 82145 }, { "epoch": 0.21, "learning_rate": 7.958704514570436e-05, "loss": 1.026, "step": 82150 }, { "epoch": 0.21, "learning_rate": 7.958578690422265e-05, "loss": 1.0232, "step": 82155 }, { "epoch": 0.21, "learning_rate": 7.958452866274096e-05, "loss": 1.0261, "step": 82160 }, { "epoch": 0.21, "learning_rate": 7.958327042125925e-05, "loss": 1.026, "step": 82165 }, { "epoch": 0.21, "learning_rate": 7.958201217977754e-05, "loss": 1.0238, "step": 82170 }, { "epoch": 0.21, "learning_rate": 7.958075393829583e-05, "loss": 1.0263, "step": 82175 }, { "epoch": 0.21, "learning_rate": 7.957949569681413e-05, "loss": 1.0241, "step": 82180 }, { "epoch": 0.21, "learning_rate": 7.957823745533243e-05, "loss": 1.0226, "step": 82185 }, { "epoch": 0.21, "learning_rate": 7.957697921385072e-05, "loss": 1.0255, "step": 82190 }, { "epoch": 0.21, "learning_rate": 7.957572097236901e-05, "loss": 1.0244, "step": 82195 }, { "epoch": 0.21, "learning_rate": 7.95744627308873e-05, "loss": 1.0254, "step": 82200 }, { "epoch": 0.21, "learning_rate": 7.957320448940561e-05, "loss": 1.0216, "step": 82205 }, { "epoch": 0.21, "learning_rate": 7.95719462479239e-05, "loss": 1.0234, "step": 82210 }, { "epoch": 0.21, "learning_rate": 7.95706880064422e-05, "loss": 1.0259, "step": 82215 }, { "epoch": 0.21, "learning_rate": 7.956942976496049e-05, "loss": 1.0264, "step": 82220 }, { "epoch": 0.21, "learning_rate": 7.956817152347879e-05, "loss": 1.0231, "step": 82225 }, { "epoch": 0.21, "learning_rate": 7.956691328199708e-05, "loss": 1.025, "step": 82230 }, { "epoch": 0.21, "learning_rate": 7.956565504051537e-05, "loss": 1.0233, "step": 82235 }, { "epoch": 0.21, "learning_rate": 7.956439679903367e-05, "loss": 1.0242, "step": 82240 }, { "epoch": 0.21, "learning_rate": 7.956313855755196e-05, "loss": 1.0251, "step": 82245 }, { "epoch": 0.21, "learning_rate": 7.956188031607026e-05, "loss": 1.0236, "step": 82250 }, { "epoch": 0.21, "learning_rate": 7.956062207458855e-05, "loss": 1.0243, "step": 82255 }, { "epoch": 0.21, "learning_rate": 7.955936383310686e-05, "loss": 1.027, "step": 82260 }, { "epoch": 0.21, "learning_rate": 7.955810559162515e-05, "loss": 1.0242, "step": 82265 }, { "epoch": 0.21, "learning_rate": 7.955684735014344e-05, "loss": 1.0251, "step": 82270 }, { "epoch": 0.21, "learning_rate": 7.955558910866175e-05, "loss": 1.0239, "step": 82275 }, { "epoch": 0.21, "learning_rate": 7.955433086718004e-05, "loss": 1.0239, "step": 82280 }, { "epoch": 0.21, "learning_rate": 7.955307262569833e-05, "loss": 1.0242, "step": 82285 }, { "epoch": 0.21, "learning_rate": 7.955181438421662e-05, "loss": 1.0274, "step": 82290 }, { "epoch": 0.21, "learning_rate": 7.955055614273493e-05, "loss": 1.0234, "step": 82295 }, { "epoch": 0.21, "learning_rate": 7.954929790125322e-05, "loss": 1.0271, "step": 82300 }, { "epoch": 0.21, "learning_rate": 7.954803965977151e-05, "loss": 1.0259, "step": 82305 }, { "epoch": 0.21, "learning_rate": 7.95467814182898e-05, "loss": 1.0251, "step": 82310 }, { "epoch": 0.21, "learning_rate": 7.95455231768081e-05, "loss": 1.024, "step": 82315 }, { "epoch": 0.21, "learning_rate": 7.95442649353264e-05, "loss": 1.0234, "step": 82320 }, { "epoch": 0.21, "learning_rate": 7.954300669384469e-05, "loss": 1.0251, "step": 82325 }, { "epoch": 0.21, "learning_rate": 7.954174845236298e-05, "loss": 1.0245, "step": 82330 }, { "epoch": 0.21, "learning_rate": 7.954049021088127e-05, "loss": 1.0245, "step": 82335 }, { "epoch": 0.21, "learning_rate": 7.953923196939958e-05, "loss": 1.0242, "step": 82340 }, { "epoch": 0.21, "learning_rate": 7.953797372791787e-05, "loss": 1.0255, "step": 82345 }, { "epoch": 0.21, "learning_rate": 7.953671548643616e-05, "loss": 1.0232, "step": 82350 }, { "epoch": 0.21, "learning_rate": 7.953545724495445e-05, "loss": 1.0242, "step": 82355 }, { "epoch": 0.21, "learning_rate": 7.953419900347276e-05, "loss": 1.0272, "step": 82360 }, { "epoch": 0.21, "learning_rate": 7.953294076199105e-05, "loss": 1.0236, "step": 82365 }, { "epoch": 0.21, "learning_rate": 7.953168252050934e-05, "loss": 1.0239, "step": 82370 }, { "epoch": 0.21, "learning_rate": 7.953042427902763e-05, "loss": 1.0238, "step": 82375 }, { "epoch": 0.21, "learning_rate": 7.952916603754593e-05, "loss": 1.0244, "step": 82380 }, { "epoch": 0.21, "learning_rate": 7.952790779606423e-05, "loss": 1.0233, "step": 82385 }, { "epoch": 0.21, "learning_rate": 7.952664955458252e-05, "loss": 1.0245, "step": 82390 }, { "epoch": 0.21, "learning_rate": 7.952539131310081e-05, "loss": 1.0252, "step": 82395 }, { "epoch": 0.21, "learning_rate": 7.95241330716191e-05, "loss": 1.0249, "step": 82400 }, { "epoch": 0.21, "learning_rate": 7.952287483013741e-05, "loss": 1.0254, "step": 82405 }, { "epoch": 0.21, "learning_rate": 7.95216165886557e-05, "loss": 1.0241, "step": 82410 }, { "epoch": 0.21, "learning_rate": 7.9520358347174e-05, "loss": 1.0224, "step": 82415 }, { "epoch": 0.21, "learning_rate": 7.951910010569229e-05, "loss": 1.0248, "step": 82420 }, { "epoch": 0.21, "learning_rate": 7.951784186421059e-05, "loss": 1.0224, "step": 82425 }, { "epoch": 0.21, "learning_rate": 7.951658362272888e-05, "loss": 1.0237, "step": 82430 }, { "epoch": 0.21, "learning_rate": 7.951532538124717e-05, "loss": 1.024, "step": 82435 }, { "epoch": 0.21, "learning_rate": 7.951406713976547e-05, "loss": 1.0241, "step": 82440 }, { "epoch": 0.21, "learning_rate": 7.951280889828376e-05, "loss": 1.024, "step": 82445 }, { "epoch": 0.21, "learning_rate": 7.951155065680206e-05, "loss": 1.0241, "step": 82450 }, { "epoch": 0.21, "learning_rate": 7.951029241532035e-05, "loss": 1.025, "step": 82455 }, { "epoch": 0.21, "learning_rate": 7.950903417383865e-05, "loss": 1.023, "step": 82460 }, { "epoch": 0.21, "learning_rate": 7.950777593235694e-05, "loss": 1.0257, "step": 82465 }, { "epoch": 0.21, "learning_rate": 7.950651769087524e-05, "loss": 1.0245, "step": 82470 }, { "epoch": 0.21, "learning_rate": 7.950525944939353e-05, "loss": 1.025, "step": 82475 }, { "epoch": 0.21, "learning_rate": 7.950400120791183e-05, "loss": 1.0242, "step": 82480 }, { "epoch": 0.21, "learning_rate": 7.950274296643012e-05, "loss": 1.0239, "step": 82485 }, { "epoch": 0.21, "learning_rate": 7.950148472494842e-05, "loss": 1.0211, "step": 82490 }, { "epoch": 0.21, "learning_rate": 7.950022648346671e-05, "loss": 1.0248, "step": 82495 }, { "epoch": 0.21, "learning_rate": 7.9498968241985e-05, "loss": 1.043, "step": 82500 }, { "epoch": 0.21, "learning_rate": 7.94977100005033e-05, "loss": 1.025, "step": 82505 }, { "epoch": 0.21, "learning_rate": 7.949645175902159e-05, "loss": 1.0244, "step": 82510 }, { "epoch": 0.21, "learning_rate": 7.94951935175399e-05, "loss": 1.0238, "step": 82515 }, { "epoch": 0.21, "learning_rate": 7.949393527605819e-05, "loss": 1.0255, "step": 82520 }, { "epoch": 0.21, "learning_rate": 7.949267703457648e-05, "loss": 1.025, "step": 82525 }, { "epoch": 0.21, "learning_rate": 7.949141879309477e-05, "loss": 1.0239, "step": 82530 }, { "epoch": 0.21, "learning_rate": 7.949016055161307e-05, "loss": 1.0251, "step": 82535 }, { "epoch": 0.21, "learning_rate": 7.948890231013137e-05, "loss": 1.0248, "step": 82540 }, { "epoch": 0.21, "learning_rate": 7.948764406864966e-05, "loss": 1.0253, "step": 82545 }, { "epoch": 0.21, "learning_rate": 7.948638582716795e-05, "loss": 1.0254, "step": 82550 }, { "epoch": 0.21, "learning_rate": 7.948512758568625e-05, "loss": 1.0265, "step": 82555 }, { "epoch": 0.21, "learning_rate": 7.948386934420455e-05, "loss": 1.0247, "step": 82560 }, { "epoch": 0.21, "learning_rate": 7.948261110272284e-05, "loss": 1.0242, "step": 82565 }, { "epoch": 0.21, "learning_rate": 7.948135286124113e-05, "loss": 1.0241, "step": 82570 }, { "epoch": 0.21, "learning_rate": 7.948009461975942e-05, "loss": 1.0234, "step": 82575 }, { "epoch": 0.21, "learning_rate": 7.947883637827773e-05, "loss": 1.0252, "step": 82580 }, { "epoch": 0.21, "learning_rate": 7.947757813679602e-05, "loss": 1.0251, "step": 82585 }, { "epoch": 0.21, "learning_rate": 7.947631989531431e-05, "loss": 1.0244, "step": 82590 }, { "epoch": 0.21, "learning_rate": 7.94750616538326e-05, "loss": 1.0226, "step": 82595 }, { "epoch": 0.21, "learning_rate": 7.94738034123509e-05, "loss": 1.0248, "step": 82600 }, { "epoch": 0.21, "learning_rate": 7.94725451708692e-05, "loss": 1.023, "step": 82605 }, { "epoch": 0.21, "learning_rate": 7.947128692938749e-05, "loss": 1.0233, "step": 82610 }, { "epoch": 0.21, "learning_rate": 7.947002868790578e-05, "loss": 1.0235, "step": 82615 }, { "epoch": 0.21, "learning_rate": 7.946877044642409e-05, "loss": 1.0239, "step": 82620 }, { "epoch": 0.21, "learning_rate": 7.946751220494238e-05, "loss": 1.025, "step": 82625 }, { "epoch": 0.21, "learning_rate": 7.946625396346067e-05, "loss": 1.0235, "step": 82630 }, { "epoch": 0.21, "learning_rate": 7.946499572197896e-05, "loss": 1.0229, "step": 82635 }, { "epoch": 0.21, "learning_rate": 7.946373748049725e-05, "loss": 1.0268, "step": 82640 }, { "epoch": 0.21, "learning_rate": 7.946247923901556e-05, "loss": 1.026, "step": 82645 }, { "epoch": 0.21, "learning_rate": 7.946122099753385e-05, "loss": 1.0264, "step": 82650 }, { "epoch": 0.21, "learning_rate": 7.945996275605214e-05, "loss": 1.0231, "step": 82655 }, { "epoch": 0.21, "learning_rate": 7.945870451457043e-05, "loss": 1.0262, "step": 82660 }, { "epoch": 0.21, "learning_rate": 7.945744627308874e-05, "loss": 1.0265, "step": 82665 }, { "epoch": 0.21, "learning_rate": 7.945618803160703e-05, "loss": 1.0235, "step": 82670 }, { "epoch": 0.21, "learning_rate": 7.945492979012532e-05, "loss": 1.023, "step": 82675 }, { "epoch": 0.21, "learning_rate": 7.945367154864361e-05, "loss": 1.0256, "step": 82680 }, { "epoch": 0.21, "learning_rate": 7.945241330716192e-05, "loss": 1.0246, "step": 82685 }, { "epoch": 0.21, "learning_rate": 7.945115506568021e-05, "loss": 1.0234, "step": 82690 }, { "epoch": 0.21, "learning_rate": 7.94498968241985e-05, "loss": 1.0243, "step": 82695 }, { "epoch": 0.21, "learning_rate": 7.944863858271679e-05, "loss": 1.026, "step": 82700 }, { "epoch": 0.21, "learning_rate": 7.944738034123508e-05, "loss": 1.0237, "step": 82705 }, { "epoch": 0.21, "learning_rate": 7.944612209975339e-05, "loss": 1.0242, "step": 82710 }, { "epoch": 0.21, "learning_rate": 7.944486385827168e-05, "loss": 1.0257, "step": 82715 }, { "epoch": 0.21, "learning_rate": 7.944360561678997e-05, "loss": 1.0264, "step": 82720 }, { "epoch": 0.21, "learning_rate": 7.944234737530826e-05, "loss": 1.0264, "step": 82725 }, { "epoch": 0.21, "learning_rate": 7.944108913382657e-05, "loss": 1.0217, "step": 82730 }, { "epoch": 0.21, "learning_rate": 7.943983089234486e-05, "loss": 1.0259, "step": 82735 }, { "epoch": 0.21, "learning_rate": 7.943857265086315e-05, "loss": 1.0483, "step": 82740 }, { "epoch": 0.21, "learning_rate": 7.943731440938144e-05, "loss": 1.0232, "step": 82745 }, { "epoch": 0.21, "learning_rate": 7.943605616789975e-05, "loss": 1.0247, "step": 82750 }, { "epoch": 0.21, "learning_rate": 7.943479792641804e-05, "loss": 1.0237, "step": 82755 }, { "epoch": 0.21, "learning_rate": 7.943353968493634e-05, "loss": 1.0233, "step": 82760 }, { "epoch": 0.21, "learning_rate": 7.943228144345464e-05, "loss": 1.0255, "step": 82765 }, { "epoch": 0.21, "learning_rate": 7.943102320197293e-05, "loss": 1.0243, "step": 82770 }, { "epoch": 0.21, "learning_rate": 7.942976496049123e-05, "loss": 1.0244, "step": 82775 }, { "epoch": 0.21, "learning_rate": 7.942850671900952e-05, "loss": 1.0227, "step": 82780 }, { "epoch": 0.21, "learning_rate": 7.942724847752782e-05, "loss": 1.0255, "step": 82785 }, { "epoch": 0.21, "learning_rate": 7.942599023604611e-05, "loss": 1.0236, "step": 82790 }, { "epoch": 0.21, "learning_rate": 7.94247319945644e-05, "loss": 1.0226, "step": 82795 }, { "epoch": 0.21, "learning_rate": 7.94234737530827e-05, "loss": 1.0241, "step": 82800 }, { "epoch": 0.21, "learning_rate": 7.9422215511601e-05, "loss": 1.0236, "step": 82805 }, { "epoch": 0.21, "learning_rate": 7.942095727011929e-05, "loss": 1.024, "step": 82810 }, { "epoch": 0.21, "learning_rate": 7.941969902863758e-05, "loss": 1.0258, "step": 82815 }, { "epoch": 0.21, "learning_rate": 7.941844078715588e-05, "loss": 1.0228, "step": 82820 }, { "epoch": 0.21, "learning_rate": 7.941718254567418e-05, "loss": 1.0255, "step": 82825 }, { "epoch": 0.21, "learning_rate": 7.941592430419247e-05, "loss": 1.0243, "step": 82830 }, { "epoch": 0.21, "learning_rate": 7.941466606271076e-05, "loss": 1.0214, "step": 82835 }, { "epoch": 0.21, "learning_rate": 7.941340782122906e-05, "loss": 1.0247, "step": 82840 }, { "epoch": 0.21, "learning_rate": 7.941214957974736e-05, "loss": 1.0257, "step": 82845 }, { "epoch": 0.21, "learning_rate": 7.941089133826565e-05, "loss": 1.0229, "step": 82850 }, { "epoch": 0.21, "learning_rate": 7.940963309678394e-05, "loss": 1.0248, "step": 82855 }, { "epoch": 0.21, "learning_rate": 7.940837485530223e-05, "loss": 1.0239, "step": 82860 }, { "epoch": 0.21, "learning_rate": 7.940711661382054e-05, "loss": 1.0229, "step": 82865 }, { "epoch": 0.21, "learning_rate": 7.940585837233883e-05, "loss": 1.0246, "step": 82870 }, { "epoch": 0.21, "learning_rate": 7.940460013085712e-05, "loss": 1.0224, "step": 82875 }, { "epoch": 0.21, "learning_rate": 7.940334188937541e-05, "loss": 1.0258, "step": 82880 }, { "epoch": 0.21, "learning_rate": 7.940208364789372e-05, "loss": 1.0256, "step": 82885 }, { "epoch": 0.21, "learning_rate": 7.940082540641201e-05, "loss": 1.0242, "step": 82890 }, { "epoch": 0.21, "learning_rate": 7.93995671649303e-05, "loss": 1.0262, "step": 82895 }, { "epoch": 0.21, "learning_rate": 7.939830892344859e-05, "loss": 1.0249, "step": 82900 }, { "epoch": 0.21, "learning_rate": 7.93970506819669e-05, "loss": 1.0254, "step": 82905 }, { "epoch": 0.21, "learning_rate": 7.939579244048519e-05, "loss": 1.0241, "step": 82910 }, { "epoch": 0.21, "learning_rate": 7.939453419900348e-05, "loss": 1.0248, "step": 82915 }, { "epoch": 0.21, "learning_rate": 7.939327595752177e-05, "loss": 1.0244, "step": 82920 }, { "epoch": 0.21, "learning_rate": 7.939201771604006e-05, "loss": 1.0266, "step": 82925 }, { "epoch": 0.21, "learning_rate": 7.939075947455837e-05, "loss": 1.0258, "step": 82930 }, { "epoch": 0.21, "learning_rate": 7.938950123307666e-05, "loss": 1.0266, "step": 82935 }, { "epoch": 0.21, "learning_rate": 7.938824299159495e-05, "loss": 1.0234, "step": 82940 }, { "epoch": 0.21, "learning_rate": 7.938698475011324e-05, "loss": 1.0264, "step": 82945 }, { "epoch": 0.21, "learning_rate": 7.938572650863155e-05, "loss": 1.0231, "step": 82950 }, { "epoch": 0.21, "learning_rate": 7.938446826714984e-05, "loss": 1.0254, "step": 82955 }, { "epoch": 0.21, "learning_rate": 7.938321002566813e-05, "loss": 1.0241, "step": 82960 }, { "epoch": 0.21, "learning_rate": 7.938195178418642e-05, "loss": 1.0262, "step": 82965 }, { "epoch": 0.21, "learning_rate": 7.938069354270471e-05, "loss": 1.0256, "step": 82970 }, { "epoch": 0.21, "learning_rate": 7.937943530122302e-05, "loss": 1.0254, "step": 82975 }, { "epoch": 0.21, "learning_rate": 7.937817705974131e-05, "loss": 1.0226, "step": 82980 }, { "epoch": 0.21, "learning_rate": 7.93769188182596e-05, "loss": 1.024, "step": 82985 }, { "epoch": 0.21, "learning_rate": 7.93756605767779e-05, "loss": 1.0242, "step": 82990 }, { "epoch": 0.21, "learning_rate": 7.93744023352962e-05, "loss": 1.0244, "step": 82995 }, { "epoch": 0.21, "learning_rate": 7.937314409381449e-05, "loss": 1.0265, "step": 83000 }, { "epoch": 0.21, "learning_rate": 7.937188585233278e-05, "loss": 1.0237, "step": 83005 }, { "epoch": 0.21, "learning_rate": 7.937062761085107e-05, "loss": 1.0262, "step": 83010 }, { "epoch": 0.21, "learning_rate": 7.936936936936938e-05, "loss": 1.024, "step": 83015 }, { "epoch": 0.21, "learning_rate": 7.936811112788767e-05, "loss": 1.0272, "step": 83020 }, { "epoch": 0.21, "learning_rate": 7.936685288640596e-05, "loss": 1.0241, "step": 83025 }, { "epoch": 0.21, "learning_rate": 7.936559464492425e-05, "loss": 1.0244, "step": 83030 }, { "epoch": 0.21, "learning_rate": 7.936433640344255e-05, "loss": 1.0248, "step": 83035 }, { "epoch": 0.21, "learning_rate": 7.936307816196085e-05, "loss": 1.0252, "step": 83040 }, { "epoch": 0.21, "learning_rate": 7.936181992047914e-05, "loss": 1.0244, "step": 83045 }, { "epoch": 0.21, "learning_rate": 7.936056167899743e-05, "loss": 1.0249, "step": 83050 }, { "epoch": 0.21, "learning_rate": 7.935930343751573e-05, "loss": 1.0244, "step": 83055 }, { "epoch": 0.21, "learning_rate": 7.935804519603403e-05, "loss": 1.0247, "step": 83060 }, { "epoch": 0.21, "learning_rate": 7.935678695455232e-05, "loss": 1.0246, "step": 83065 }, { "epoch": 0.21, "learning_rate": 7.935552871307061e-05, "loss": 1.0236, "step": 83070 }, { "epoch": 0.21, "learning_rate": 7.93542704715889e-05, "loss": 1.0244, "step": 83075 }, { "epoch": 0.21, "learning_rate": 7.935301223010721e-05, "loss": 1.0243, "step": 83080 }, { "epoch": 0.21, "learning_rate": 7.93517539886255e-05, "loss": 1.0264, "step": 83085 }, { "epoch": 0.21, "learning_rate": 7.93504957471438e-05, "loss": 1.0224, "step": 83090 }, { "epoch": 0.21, "learning_rate": 7.934923750566208e-05, "loss": 1.0253, "step": 83095 }, { "epoch": 0.21, "learning_rate": 7.934797926418038e-05, "loss": 1.0228, "step": 83100 }, { "epoch": 0.21, "learning_rate": 7.934672102269868e-05, "loss": 1.0235, "step": 83105 }, { "epoch": 0.21, "learning_rate": 7.934546278121697e-05, "loss": 1.0245, "step": 83110 }, { "epoch": 0.21, "learning_rate": 7.934420453973526e-05, "loss": 1.0269, "step": 83115 }, { "epoch": 0.21, "learning_rate": 7.934294629825356e-05, "loss": 1.0262, "step": 83120 }, { "epoch": 0.21, "learning_rate": 7.934168805677186e-05, "loss": 1.0242, "step": 83125 }, { "epoch": 0.21, "learning_rate": 7.934042981529015e-05, "loss": 1.0246, "step": 83130 }, { "epoch": 0.21, "learning_rate": 7.933917157380844e-05, "loss": 1.0257, "step": 83135 }, { "epoch": 0.21, "learning_rate": 7.933791333232674e-05, "loss": 1.0246, "step": 83140 }, { "epoch": 0.21, "learning_rate": 7.933665509084504e-05, "loss": 1.0248, "step": 83145 }, { "epoch": 0.21, "learning_rate": 7.933539684936333e-05, "loss": 1.0217, "step": 83150 }, { "epoch": 0.21, "learning_rate": 7.933413860788162e-05, "loss": 1.0233, "step": 83155 }, { "epoch": 0.21, "learning_rate": 7.933288036639992e-05, "loss": 1.0245, "step": 83160 }, { "epoch": 0.21, "learning_rate": 7.933162212491821e-05, "loss": 1.0243, "step": 83165 }, { "epoch": 0.21, "learning_rate": 7.933036388343651e-05, "loss": 1.0249, "step": 83170 }, { "epoch": 0.21, "learning_rate": 7.93291056419548e-05, "loss": 1.0244, "step": 83175 }, { "epoch": 0.21, "learning_rate": 7.93278474004731e-05, "loss": 1.0235, "step": 83180 }, { "epoch": 0.21, "learning_rate": 7.932658915899139e-05, "loss": 1.025, "step": 83185 }, { "epoch": 0.21, "learning_rate": 7.932533091750969e-05, "loss": 1.0243, "step": 83190 }, { "epoch": 0.21, "learning_rate": 7.932407267602798e-05, "loss": 1.0233, "step": 83195 }, { "epoch": 0.21, "learning_rate": 7.932281443454628e-05, "loss": 1.0251, "step": 83200 }, { "epoch": 0.21, "learning_rate": 7.932155619306457e-05, "loss": 1.0239, "step": 83205 }, { "epoch": 0.21, "learning_rate": 7.932029795158287e-05, "loss": 1.0236, "step": 83210 }, { "epoch": 0.21, "learning_rate": 7.931903971010116e-05, "loss": 1.0244, "step": 83215 }, { "epoch": 0.21, "learning_rate": 7.931778146861946e-05, "loss": 1.0252, "step": 83220 }, { "epoch": 0.21, "learning_rate": 7.931652322713775e-05, "loss": 1.0265, "step": 83225 }, { "epoch": 0.21, "learning_rate": 7.931526498565604e-05, "loss": 1.0271, "step": 83230 }, { "epoch": 0.21, "learning_rate": 7.931400674417434e-05, "loss": 1.0234, "step": 83235 }, { "epoch": 0.21, "learning_rate": 7.931274850269264e-05, "loss": 1.0254, "step": 83240 }, { "epoch": 0.21, "learning_rate": 7.931149026121093e-05, "loss": 1.0241, "step": 83245 }, { "epoch": 0.21, "learning_rate": 7.931023201972922e-05, "loss": 1.0256, "step": 83250 }, { "epoch": 0.21, "learning_rate": 7.930897377824752e-05, "loss": 1.025, "step": 83255 }, { "epoch": 0.21, "learning_rate": 7.930771553676583e-05, "loss": 1.0244, "step": 83260 }, { "epoch": 0.21, "learning_rate": 7.930645729528412e-05, "loss": 1.0232, "step": 83265 }, { "epoch": 0.21, "learning_rate": 7.930519905380241e-05, "loss": 1.0238, "step": 83270 }, { "epoch": 0.21, "learning_rate": 7.93039408123207e-05, "loss": 1.0411, "step": 83275 }, { "epoch": 0.21, "learning_rate": 7.930268257083901e-05, "loss": 1.0233, "step": 83280 }, { "epoch": 0.21, "learning_rate": 7.93014243293573e-05, "loss": 1.023, "step": 83285 }, { "epoch": 0.21, "learning_rate": 7.930016608787559e-05, "loss": 1.0226, "step": 83290 }, { "epoch": 0.21, "learning_rate": 7.929890784639388e-05, "loss": 1.0246, "step": 83295 }, { "epoch": 0.21, "learning_rate": 7.929764960491219e-05, "loss": 1.0241, "step": 83300 }, { "epoch": 0.21, "learning_rate": 7.929639136343048e-05, "loss": 1.0226, "step": 83305 }, { "epoch": 0.21, "learning_rate": 7.929513312194877e-05, "loss": 1.0247, "step": 83310 }, { "epoch": 0.21, "learning_rate": 7.929387488046706e-05, "loss": 1.0247, "step": 83315 }, { "epoch": 0.21, "learning_rate": 7.929261663898536e-05, "loss": 1.0249, "step": 83320 }, { "epoch": 0.21, "learning_rate": 7.929135839750366e-05, "loss": 1.0241, "step": 83325 }, { "epoch": 0.21, "learning_rate": 7.929010015602195e-05, "loss": 1.0238, "step": 83330 }, { "epoch": 0.21, "learning_rate": 7.928884191454024e-05, "loss": 1.0269, "step": 83335 }, { "epoch": 0.21, "learning_rate": 7.928758367305854e-05, "loss": 1.0245, "step": 83340 }, { "epoch": 0.21, "learning_rate": 7.928632543157684e-05, "loss": 1.0255, "step": 83345 }, { "epoch": 0.21, "learning_rate": 7.928506719009513e-05, "loss": 1.026, "step": 83350 }, { "epoch": 0.21, "learning_rate": 7.928380894861342e-05, "loss": 1.0247, "step": 83355 }, { "epoch": 0.21, "learning_rate": 7.928255070713172e-05, "loss": 1.0251, "step": 83360 }, { "epoch": 0.21, "learning_rate": 7.928129246565002e-05, "loss": 1.0218, "step": 83365 }, { "epoch": 0.21, "learning_rate": 7.928003422416831e-05, "loss": 1.0257, "step": 83370 }, { "epoch": 0.21, "learning_rate": 7.92787759826866e-05, "loss": 1.023, "step": 83375 }, { "epoch": 0.21, "learning_rate": 7.92775177412049e-05, "loss": 1.024, "step": 83380 }, { "epoch": 0.21, "learning_rate": 7.927625949972319e-05, "loss": 1.0272, "step": 83385 }, { "epoch": 0.21, "learning_rate": 7.927500125824149e-05, "loss": 1.0248, "step": 83390 }, { "epoch": 0.21, "learning_rate": 7.927374301675978e-05, "loss": 1.0253, "step": 83395 }, { "epoch": 0.21, "learning_rate": 7.927248477527808e-05, "loss": 1.0231, "step": 83400 }, { "epoch": 0.21, "learning_rate": 7.927122653379637e-05, "loss": 1.0235, "step": 83405 }, { "epoch": 0.21, "learning_rate": 7.926996829231467e-05, "loss": 1.0252, "step": 83410 }, { "epoch": 0.21, "learning_rate": 7.926871005083296e-05, "loss": 1.0233, "step": 83415 }, { "epoch": 0.21, "learning_rate": 7.926745180935126e-05, "loss": 1.0246, "step": 83420 }, { "epoch": 0.21, "learning_rate": 7.926619356786955e-05, "loss": 1.0233, "step": 83425 }, { "epoch": 0.21, "learning_rate": 7.926493532638785e-05, "loss": 1.0257, "step": 83430 }, { "epoch": 0.21, "learning_rate": 7.926367708490614e-05, "loss": 1.0237, "step": 83435 }, { "epoch": 0.21, "learning_rate": 7.926241884342444e-05, "loss": 1.0263, "step": 83440 }, { "epoch": 0.21, "learning_rate": 7.926116060194273e-05, "loss": 1.0269, "step": 83445 }, { "epoch": 0.21, "learning_rate": 7.925990236046102e-05, "loss": 1.0257, "step": 83450 }, { "epoch": 0.21, "learning_rate": 7.925864411897932e-05, "loss": 1.0232, "step": 83455 }, { "epoch": 0.21, "learning_rate": 7.925738587749762e-05, "loss": 1.025, "step": 83460 }, { "epoch": 0.21, "learning_rate": 7.925612763601591e-05, "loss": 1.0222, "step": 83465 }, { "epoch": 0.21, "learning_rate": 7.92548693945342e-05, "loss": 1.0237, "step": 83470 }, { "epoch": 0.21, "learning_rate": 7.92536111530525e-05, "loss": 1.0259, "step": 83475 }, { "epoch": 0.21, "learning_rate": 7.92523529115708e-05, "loss": 1.0227, "step": 83480 }, { "epoch": 0.21, "learning_rate": 7.925109467008909e-05, "loss": 1.026, "step": 83485 }, { "epoch": 0.21, "learning_rate": 7.924983642860738e-05, "loss": 1.0239, "step": 83490 }, { "epoch": 0.21, "learning_rate": 7.924857818712568e-05, "loss": 1.0243, "step": 83495 }, { "epoch": 0.21, "learning_rate": 7.924731994564398e-05, "loss": 1.0249, "step": 83500 }, { "epoch": 0.21, "learning_rate": 7.924606170416227e-05, "loss": 1.0242, "step": 83505 }, { "epoch": 0.21, "learning_rate": 7.924480346268056e-05, "loss": 1.0251, "step": 83510 }, { "epoch": 0.21, "learning_rate": 7.924354522119885e-05, "loss": 1.0242, "step": 83515 }, { "epoch": 0.21, "learning_rate": 7.924228697971716e-05, "loss": 1.0246, "step": 83520 }, { "epoch": 0.21, "learning_rate": 7.924102873823545e-05, "loss": 1.0249, "step": 83525 }, { "epoch": 0.21, "learning_rate": 7.923977049675374e-05, "loss": 1.0219, "step": 83530 }, { "epoch": 0.21, "learning_rate": 7.923851225527203e-05, "loss": 1.023, "step": 83535 }, { "epoch": 0.21, "learning_rate": 7.923725401379034e-05, "loss": 1.0239, "step": 83540 }, { "epoch": 0.21, "learning_rate": 7.923599577230863e-05, "loss": 1.0261, "step": 83545 }, { "epoch": 0.21, "learning_rate": 7.923473753082692e-05, "loss": 1.0256, "step": 83550 }, { "epoch": 0.21, "learning_rate": 7.923347928934521e-05, "loss": 1.0246, "step": 83555 }, { "epoch": 0.21, "learning_rate": 7.923222104786352e-05, "loss": 1.0223, "step": 83560 }, { "epoch": 0.21, "learning_rate": 7.923096280638181e-05, "loss": 1.0229, "step": 83565 }, { "epoch": 0.21, "learning_rate": 7.92297045649001e-05, "loss": 1.0215, "step": 83570 }, { "epoch": 0.21, "learning_rate": 7.922844632341839e-05, "loss": 1.0227, "step": 83575 }, { "epoch": 0.21, "learning_rate": 7.922718808193668e-05, "loss": 1.0244, "step": 83580 }, { "epoch": 0.21, "learning_rate": 7.922592984045499e-05, "loss": 1.0252, "step": 83585 }, { "epoch": 0.21, "learning_rate": 7.922467159897328e-05, "loss": 1.0264, "step": 83590 }, { "epoch": 0.21, "learning_rate": 7.922341335749157e-05, "loss": 1.0461, "step": 83595 }, { "epoch": 0.21, "learning_rate": 7.922215511600986e-05, "loss": 1.0503, "step": 83600 }, { "epoch": 0.21, "learning_rate": 7.922089687452817e-05, "loss": 1.022, "step": 83605 }, { "epoch": 0.21, "learning_rate": 7.921963863304646e-05, "loss": 1.0232, "step": 83610 }, { "epoch": 0.21, "learning_rate": 7.921838039156475e-05, "loss": 1.0216, "step": 83615 }, { "epoch": 0.21, "learning_rate": 7.921712215008304e-05, "loss": 1.0262, "step": 83620 }, { "epoch": 0.21, "learning_rate": 7.921586390860133e-05, "loss": 1.0262, "step": 83625 }, { "epoch": 0.21, "learning_rate": 7.921460566711964e-05, "loss": 1.0231, "step": 83630 }, { "epoch": 0.21, "learning_rate": 7.921334742563793e-05, "loss": 1.0252, "step": 83635 }, { "epoch": 0.21, "learning_rate": 7.921208918415622e-05, "loss": 1.0236, "step": 83640 }, { "epoch": 0.21, "learning_rate": 7.921083094267451e-05, "loss": 1.0251, "step": 83645 }, { "epoch": 0.21, "learning_rate": 7.920957270119282e-05, "loss": 1.0236, "step": 83650 }, { "epoch": 0.21, "learning_rate": 7.920831445971111e-05, "loss": 1.0268, "step": 83655 }, { "epoch": 0.21, "learning_rate": 7.92070562182294e-05, "loss": 1.0247, "step": 83660 }, { "epoch": 0.21, "learning_rate": 7.920579797674769e-05, "loss": 1.0249, "step": 83665 }, { "epoch": 0.21, "learning_rate": 7.9204539735266e-05, "loss": 1.0232, "step": 83670 }, { "epoch": 0.21, "learning_rate": 7.920328149378429e-05, "loss": 1.0279, "step": 83675 }, { "epoch": 0.21, "learning_rate": 7.920202325230258e-05, "loss": 1.0256, "step": 83680 }, { "epoch": 0.21, "learning_rate": 7.920076501082087e-05, "loss": 1.0261, "step": 83685 }, { "epoch": 0.21, "learning_rate": 7.919950676933916e-05, "loss": 1.0242, "step": 83690 }, { "epoch": 0.21, "learning_rate": 7.919824852785747e-05, "loss": 1.023, "step": 83695 }, { "epoch": 0.21, "learning_rate": 7.919699028637576e-05, "loss": 1.023, "step": 83700 }, { "epoch": 0.21, "learning_rate": 7.919573204489405e-05, "loss": 1.0263, "step": 83705 }, { "epoch": 0.21, "learning_rate": 7.919447380341234e-05, "loss": 1.0246, "step": 83710 }, { "epoch": 0.21, "learning_rate": 7.919321556193065e-05, "loss": 1.0238, "step": 83715 }, { "epoch": 0.21, "learning_rate": 7.919195732044894e-05, "loss": 1.0258, "step": 83720 }, { "epoch": 0.21, "learning_rate": 7.919069907896723e-05, "loss": 1.0268, "step": 83725 }, { "epoch": 0.21, "learning_rate": 7.918944083748552e-05, "loss": 1.0249, "step": 83730 }, { "epoch": 0.21, "learning_rate": 7.918818259600383e-05, "loss": 1.023, "step": 83735 }, { "epoch": 0.21, "learning_rate": 7.918692435452212e-05, "loss": 1.0442, "step": 83740 }, { "epoch": 0.21, "learning_rate": 7.918566611304041e-05, "loss": 1.0243, "step": 83745 }, { "epoch": 0.21, "learning_rate": 7.91844078715587e-05, "loss": 1.0237, "step": 83750 }, { "epoch": 0.21, "learning_rate": 7.918314963007701e-05, "loss": 1.0226, "step": 83755 }, { "epoch": 0.21, "learning_rate": 7.918189138859531e-05, "loss": 1.024, "step": 83760 }, { "epoch": 0.21, "learning_rate": 7.91806331471136e-05, "loss": 1.0238, "step": 83765 }, { "epoch": 0.21, "learning_rate": 7.91793749056319e-05, "loss": 1.0265, "step": 83770 }, { "epoch": 0.21, "learning_rate": 7.917811666415019e-05, "loss": 1.0235, "step": 83775 }, { "epoch": 0.21, "learning_rate": 7.917685842266848e-05, "loss": 1.0232, "step": 83780 }, { "epoch": 0.21, "learning_rate": 7.917560018118679e-05, "loss": 1.0227, "step": 83785 }, { "epoch": 0.21, "learning_rate": 7.917434193970508e-05, "loss": 1.0247, "step": 83790 }, { "epoch": 0.21, "learning_rate": 7.917308369822337e-05, "loss": 1.0239, "step": 83795 }, { "epoch": 0.21, "learning_rate": 7.917182545674166e-05, "loss": 1.0247, "step": 83800 }, { "epoch": 0.21, "learning_rate": 7.917056721525997e-05, "loss": 1.0233, "step": 83805 }, { "epoch": 0.21, "learning_rate": 7.916930897377826e-05, "loss": 1.0231, "step": 83810 }, { "epoch": 0.21, "learning_rate": 7.916805073229655e-05, "loss": 1.0251, "step": 83815 }, { "epoch": 0.21, "learning_rate": 7.916679249081484e-05, "loss": 1.0233, "step": 83820 }, { "epoch": 0.21, "learning_rate": 7.916553424933315e-05, "loss": 1.0235, "step": 83825 }, { "epoch": 0.21, "learning_rate": 7.916427600785144e-05, "loss": 1.0252, "step": 83830 }, { "epoch": 0.21, "learning_rate": 7.916301776636973e-05, "loss": 1.0249, "step": 83835 }, { "epoch": 0.21, "learning_rate": 7.916175952488802e-05, "loss": 1.0236, "step": 83840 }, { "epoch": 0.21, "learning_rate": 7.916050128340631e-05, "loss": 1.0252, "step": 83845 }, { "epoch": 0.21, "learning_rate": 7.915924304192462e-05, "loss": 1.0247, "step": 83850 }, { "epoch": 0.21, "learning_rate": 7.915798480044291e-05, "loss": 1.0246, "step": 83855 }, { "epoch": 0.21, "learning_rate": 7.91567265589612e-05, "loss": 1.0231, "step": 83860 }, { "epoch": 0.21, "learning_rate": 7.915546831747949e-05, "loss": 1.0242, "step": 83865 }, { "epoch": 0.21, "learning_rate": 7.91542100759978e-05, "loss": 1.025, "step": 83870 }, { "epoch": 0.21, "learning_rate": 7.915295183451609e-05, "loss": 1.0264, "step": 83875 }, { "epoch": 0.21, "learning_rate": 7.915169359303438e-05, "loss": 1.0238, "step": 83880 }, { "epoch": 0.21, "learning_rate": 7.915043535155267e-05, "loss": 1.0234, "step": 83885 }, { "epoch": 0.21, "learning_rate": 7.914917711007098e-05, "loss": 1.0272, "step": 83890 }, { "epoch": 0.21, "learning_rate": 7.914791886858927e-05, "loss": 1.0238, "step": 83895 }, { "epoch": 0.21, "learning_rate": 7.914666062710756e-05, "loss": 1.0259, "step": 83900 }, { "epoch": 0.21, "learning_rate": 7.914540238562585e-05, "loss": 1.0228, "step": 83905 }, { "epoch": 0.21, "learning_rate": 7.914414414414414e-05, "loss": 1.0233, "step": 83910 }, { "epoch": 0.21, "learning_rate": 7.914288590266245e-05, "loss": 1.0246, "step": 83915 }, { "epoch": 0.21, "learning_rate": 7.914162766118074e-05, "loss": 1.0444, "step": 83920 }, { "epoch": 0.21, "learning_rate": 7.914036941969903e-05, "loss": 1.0231, "step": 83925 }, { "epoch": 0.21, "learning_rate": 7.913911117821732e-05, "loss": 1.0239, "step": 83930 }, { "epoch": 0.21, "learning_rate": 7.913785293673563e-05, "loss": 1.026, "step": 83935 }, { "epoch": 0.21, "learning_rate": 7.913659469525392e-05, "loss": 1.0236, "step": 83940 }, { "epoch": 0.21, "learning_rate": 7.913533645377221e-05, "loss": 1.0241, "step": 83945 }, { "epoch": 0.21, "learning_rate": 7.91340782122905e-05, "loss": 1.0462, "step": 83950 }, { "epoch": 0.21, "learning_rate": 7.913281997080881e-05, "loss": 1.0237, "step": 83955 }, { "epoch": 0.21, "learning_rate": 7.91315617293271e-05, "loss": 1.026, "step": 83960 }, { "epoch": 0.21, "learning_rate": 7.913030348784539e-05, "loss": 1.0267, "step": 83965 }, { "epoch": 0.21, "learning_rate": 7.912904524636368e-05, "loss": 1.0462, "step": 83970 }, { "epoch": 0.21, "learning_rate": 7.912778700488198e-05, "loss": 1.0242, "step": 83975 }, { "epoch": 0.21, "learning_rate": 7.912652876340028e-05, "loss": 1.024, "step": 83980 }, { "epoch": 0.21, "learning_rate": 7.912527052191857e-05, "loss": 1.0235, "step": 83985 }, { "epoch": 0.21, "learning_rate": 7.912401228043686e-05, "loss": 1.0241, "step": 83990 }, { "epoch": 0.21, "learning_rate": 7.912275403895516e-05, "loss": 1.0462, "step": 83995 }, { "epoch": 0.21, "learning_rate": 7.912149579747346e-05, "loss": 1.0231, "step": 84000 }, { "epoch": 0.21, "learning_rate": 7.912023755599175e-05, "loss": 1.0267, "step": 84005 }, { "epoch": 0.21, "learning_rate": 7.911897931451004e-05, "loss": 1.024, "step": 84010 }, { "epoch": 0.21, "learning_rate": 7.911772107302834e-05, "loss": 1.0226, "step": 84015 }, { "epoch": 0.21, "learning_rate": 7.911646283154664e-05, "loss": 1.0247, "step": 84020 }, { "epoch": 0.21, "learning_rate": 7.911520459006493e-05, "loss": 1.0231, "step": 84025 }, { "epoch": 0.21, "learning_rate": 7.911394634858322e-05, "loss": 1.0258, "step": 84030 }, { "epoch": 0.21, "learning_rate": 7.911268810710152e-05, "loss": 1.0231, "step": 84035 }, { "epoch": 0.21, "learning_rate": 7.91114298656198e-05, "loss": 1.0244, "step": 84040 }, { "epoch": 0.21, "learning_rate": 7.911017162413811e-05, "loss": 1.0231, "step": 84045 }, { "epoch": 0.21, "learning_rate": 7.91089133826564e-05, "loss": 1.025, "step": 84050 }, { "epoch": 0.21, "learning_rate": 7.91076551411747e-05, "loss": 1.0258, "step": 84055 }, { "epoch": 0.21, "learning_rate": 7.910639689969299e-05, "loss": 1.0235, "step": 84060 }, { "epoch": 0.21, "learning_rate": 7.910513865821129e-05, "loss": 1.0232, "step": 84065 }, { "epoch": 0.21, "learning_rate": 7.910388041672958e-05, "loss": 1.025, "step": 84070 }, { "epoch": 0.21, "learning_rate": 7.910262217524788e-05, "loss": 1.0242, "step": 84075 }, { "epoch": 0.21, "learning_rate": 7.910136393376617e-05, "loss": 1.0239, "step": 84080 }, { "epoch": 0.21, "learning_rate": 7.910010569228447e-05, "loss": 1.0253, "step": 84085 }, { "epoch": 0.21, "learning_rate": 7.909884745080276e-05, "loss": 1.0243, "step": 84090 }, { "epoch": 0.21, "learning_rate": 7.909758920932106e-05, "loss": 1.0237, "step": 84095 }, { "epoch": 0.21, "learning_rate": 7.909633096783935e-05, "loss": 1.0251, "step": 84100 }, { "epoch": 0.21, "learning_rate": 7.909507272635764e-05, "loss": 1.0263, "step": 84105 }, { "epoch": 0.21, "learning_rate": 7.909381448487594e-05, "loss": 1.024, "step": 84110 }, { "epoch": 0.21, "learning_rate": 7.909255624339423e-05, "loss": 1.0229, "step": 84115 }, { "epoch": 0.21, "learning_rate": 7.909129800191253e-05, "loss": 1.0225, "step": 84120 }, { "epoch": 0.21, "learning_rate": 7.909003976043082e-05, "loss": 1.0239, "step": 84125 }, { "epoch": 0.21, "learning_rate": 7.908878151894912e-05, "loss": 1.0242, "step": 84130 }, { "epoch": 0.21, "learning_rate": 7.908752327746741e-05, "loss": 1.0219, "step": 84135 }, { "epoch": 0.21, "learning_rate": 7.90862650359857e-05, "loss": 1.023, "step": 84140 }, { "epoch": 0.21, "learning_rate": 7.9085006794504e-05, "loss": 1.0252, "step": 84145 }, { "epoch": 0.21, "learning_rate": 7.90837485530223e-05, "loss": 1.0231, "step": 84150 }, { "epoch": 0.21, "learning_rate": 7.90824903115406e-05, "loss": 1.0247, "step": 84155 }, { "epoch": 0.21, "learning_rate": 7.908123207005889e-05, "loss": 1.0243, "step": 84160 }, { "epoch": 0.21, "learning_rate": 7.907997382857718e-05, "loss": 1.0252, "step": 84165 }, { "epoch": 0.21, "learning_rate": 7.907871558709547e-05, "loss": 1.0254, "step": 84170 }, { "epoch": 0.21, "learning_rate": 7.907745734561377e-05, "loss": 1.0282, "step": 84175 }, { "epoch": 0.21, "learning_rate": 7.907619910413207e-05, "loss": 1.0257, "step": 84180 }, { "epoch": 0.21, "learning_rate": 7.907494086265036e-05, "loss": 1.036, "step": 84185 }, { "epoch": 0.21, "learning_rate": 7.907368262116865e-05, "loss": 1.0266, "step": 84190 }, { "epoch": 0.21, "learning_rate": 7.907242437968695e-05, "loss": 1.0238, "step": 84195 }, { "epoch": 0.21, "learning_rate": 7.907116613820525e-05, "loss": 1.0236, "step": 84200 }, { "epoch": 0.21, "learning_rate": 7.906990789672354e-05, "loss": 1.0263, "step": 84205 }, { "epoch": 0.21, "learning_rate": 7.906864965524183e-05, "loss": 1.0265, "step": 84210 }, { "epoch": 0.21, "learning_rate": 7.906739141376013e-05, "loss": 1.0253, "step": 84215 }, { "epoch": 0.21, "learning_rate": 7.906613317227843e-05, "loss": 1.0227, "step": 84220 }, { "epoch": 0.21, "learning_rate": 7.906487493079672e-05, "loss": 1.0241, "step": 84225 }, { "epoch": 0.21, "learning_rate": 7.906361668931501e-05, "loss": 1.026, "step": 84230 }, { "epoch": 0.21, "learning_rate": 7.90623584478333e-05, "loss": 1.0239, "step": 84235 }, { "epoch": 0.21, "learning_rate": 7.90611002063516e-05, "loss": 1.0252, "step": 84240 }, { "epoch": 0.21, "learning_rate": 7.90598419648699e-05, "loss": 1.041, "step": 84245 }, { "epoch": 0.21, "learning_rate": 7.905858372338819e-05, "loss": 1.0245, "step": 84250 }, { "epoch": 0.21, "learning_rate": 7.90573254819065e-05, "loss": 1.0221, "step": 84255 }, { "epoch": 0.21, "learning_rate": 7.905606724042479e-05, "loss": 1.0228, "step": 84260 }, { "epoch": 0.21, "learning_rate": 7.905480899894309e-05, "loss": 1.0243, "step": 84265 }, { "epoch": 0.21, "learning_rate": 7.905355075746138e-05, "loss": 1.0248, "step": 84270 }, { "epoch": 0.21, "learning_rate": 7.905229251597967e-05, "loss": 1.024, "step": 84275 }, { "epoch": 0.21, "learning_rate": 7.905103427449797e-05, "loss": 1.0261, "step": 84280 }, { "epoch": 0.21, "learning_rate": 7.904977603301627e-05, "loss": 1.0279, "step": 84285 }, { "epoch": 0.21, "learning_rate": 7.904851779153456e-05, "loss": 1.0266, "step": 84290 }, { "epoch": 0.21, "learning_rate": 7.904725955005285e-05, "loss": 1.0243, "step": 84295 }, { "epoch": 0.21, "learning_rate": 7.904600130857115e-05, "loss": 1.0249, "step": 84300 }, { "epoch": 0.21, "learning_rate": 7.904474306708944e-05, "loss": 1.025, "step": 84305 }, { "epoch": 0.21, "learning_rate": 7.904348482560774e-05, "loss": 1.0228, "step": 84310 }, { "epoch": 0.21, "learning_rate": 7.904222658412603e-05, "loss": 1.0233, "step": 84315 }, { "epoch": 0.21, "learning_rate": 7.904096834264433e-05, "loss": 1.0217, "step": 84320 }, { "epoch": 0.21, "learning_rate": 7.903971010116262e-05, "loss": 1.0211, "step": 84325 }, { "epoch": 0.21, "learning_rate": 7.903845185968092e-05, "loss": 1.0247, "step": 84330 }, { "epoch": 0.21, "learning_rate": 7.903719361819921e-05, "loss": 1.0233, "step": 84335 }, { "epoch": 0.21, "learning_rate": 7.90359353767175e-05, "loss": 1.0257, "step": 84340 }, { "epoch": 0.21, "learning_rate": 7.90346771352358e-05, "loss": 1.026, "step": 84345 }, { "epoch": 0.21, "learning_rate": 7.90334188937541e-05, "loss": 1.0252, "step": 84350 }, { "epoch": 0.21, "learning_rate": 7.90321606522724e-05, "loss": 1.0234, "step": 84355 }, { "epoch": 0.21, "learning_rate": 7.903090241079069e-05, "loss": 1.0243, "step": 84360 }, { "epoch": 0.21, "learning_rate": 7.902964416930898e-05, "loss": 1.0247, "step": 84365 }, { "epoch": 0.21, "learning_rate": 7.902838592782727e-05, "loss": 1.0229, "step": 84370 }, { "epoch": 0.21, "learning_rate": 7.902712768634557e-05, "loss": 1.024, "step": 84375 }, { "epoch": 0.21, "learning_rate": 7.902586944486387e-05, "loss": 1.0253, "step": 84380 }, { "epoch": 0.21, "learning_rate": 7.902461120338216e-05, "loss": 1.0262, "step": 84385 }, { "epoch": 0.21, "learning_rate": 7.902335296190045e-05, "loss": 1.0245, "step": 84390 }, { "epoch": 0.21, "learning_rate": 7.902209472041875e-05, "loss": 1.0233, "step": 84395 }, { "epoch": 0.21, "learning_rate": 7.902083647893705e-05, "loss": 1.0345, "step": 84400 }, { "epoch": 0.21, "learning_rate": 7.901957823745534e-05, "loss": 1.0255, "step": 84405 }, { "epoch": 0.21, "learning_rate": 7.901831999597363e-05, "loss": 1.0236, "step": 84410 }, { "epoch": 0.21, "learning_rate": 7.901706175449193e-05, "loss": 1.0249, "step": 84415 }, { "epoch": 0.21, "learning_rate": 7.901580351301023e-05, "loss": 1.0237, "step": 84420 }, { "epoch": 0.21, "learning_rate": 7.901454527152852e-05, "loss": 1.0227, "step": 84425 }, { "epoch": 0.21, "learning_rate": 7.901328703004681e-05, "loss": 1.0248, "step": 84430 }, { "epoch": 0.21, "learning_rate": 7.90120287885651e-05, "loss": 1.0224, "step": 84435 }, { "epoch": 0.21, "learning_rate": 7.90107705470834e-05, "loss": 1.0246, "step": 84440 }, { "epoch": 0.21, "learning_rate": 7.90095123056017e-05, "loss": 1.0244, "step": 84445 }, { "epoch": 0.21, "learning_rate": 7.900825406411999e-05, "loss": 1.0229, "step": 84450 }, { "epoch": 0.21, "learning_rate": 7.900699582263828e-05, "loss": 1.0248, "step": 84455 }, { "epoch": 0.21, "learning_rate": 7.900573758115659e-05, "loss": 1.0267, "step": 84460 }, { "epoch": 0.21, "learning_rate": 7.900447933967488e-05, "loss": 1.0228, "step": 84465 }, { "epoch": 0.21, "learning_rate": 7.900322109819317e-05, "loss": 1.0238, "step": 84470 }, { "epoch": 0.21, "learning_rate": 7.900196285671146e-05, "loss": 1.0241, "step": 84475 }, { "epoch": 0.21, "learning_rate": 7.900070461522977e-05, "loss": 1.0265, "step": 84480 }, { "epoch": 0.21, "learning_rate": 7.899944637374806e-05, "loss": 1.0244, "step": 84485 }, { "epoch": 0.21, "learning_rate": 7.899818813226635e-05, "loss": 1.0252, "step": 84490 }, { "epoch": 0.21, "learning_rate": 7.899692989078464e-05, "loss": 1.0238, "step": 84495 }, { "epoch": 0.21, "learning_rate": 7.899567164930293e-05, "loss": 1.0224, "step": 84500 }, { "epoch": 0.21, "learning_rate": 7.899441340782124e-05, "loss": 1.0243, "step": 84505 }, { "epoch": 0.21, "learning_rate": 7.899315516633953e-05, "loss": 1.0236, "step": 84510 }, { "epoch": 0.21, "learning_rate": 7.899189692485782e-05, "loss": 1.0246, "step": 84515 }, { "epoch": 0.21, "learning_rate": 7.899063868337611e-05, "loss": 1.0233, "step": 84520 }, { "epoch": 0.21, "learning_rate": 7.898938044189442e-05, "loss": 1.0327, "step": 84525 }, { "epoch": 0.21, "learning_rate": 7.898812220041271e-05, "loss": 1.0235, "step": 84530 }, { "epoch": 0.21, "learning_rate": 7.8986863958931e-05, "loss": 1.026, "step": 84535 }, { "epoch": 0.21, "learning_rate": 7.898560571744929e-05, "loss": 1.0262, "step": 84540 }, { "epoch": 0.21, "learning_rate": 7.89843474759676e-05, "loss": 1.0233, "step": 84545 }, { "epoch": 0.21, "learning_rate": 7.898308923448589e-05, "loss": 1.0241, "step": 84550 }, { "epoch": 0.21, "learning_rate": 7.898183099300418e-05, "loss": 1.023, "step": 84555 }, { "epoch": 0.21, "learning_rate": 7.898057275152247e-05, "loss": 1.0237, "step": 84560 }, { "epoch": 0.21, "learning_rate": 7.897931451004076e-05, "loss": 1.0258, "step": 84565 }, { "epoch": 0.21, "learning_rate": 7.897805626855907e-05, "loss": 1.0224, "step": 84570 }, { "epoch": 0.21, "learning_rate": 7.897679802707736e-05, "loss": 1.023, "step": 84575 }, { "epoch": 0.21, "learning_rate": 7.897553978559565e-05, "loss": 1.0224, "step": 84580 }, { "epoch": 0.21, "learning_rate": 7.897428154411394e-05, "loss": 1.0226, "step": 84585 }, { "epoch": 0.21, "learning_rate": 7.897302330263225e-05, "loss": 1.0237, "step": 84590 }, { "epoch": 0.21, "learning_rate": 7.897176506115054e-05, "loss": 1.0247, "step": 84595 }, { "epoch": 0.21, "learning_rate": 7.897050681966883e-05, "loss": 1.0206, "step": 84600 }, { "epoch": 0.21, "learning_rate": 7.896924857818712e-05, "loss": 1.0272, "step": 84605 }, { "epoch": 0.21, "learning_rate": 7.896799033670543e-05, "loss": 1.0227, "step": 84610 }, { "epoch": 0.21, "learning_rate": 7.896673209522372e-05, "loss": 1.0249, "step": 84615 }, { "epoch": 0.21, "learning_rate": 7.896547385374201e-05, "loss": 1.0221, "step": 84620 }, { "epoch": 0.21, "learning_rate": 7.89642156122603e-05, "loss": 1.0235, "step": 84625 }, { "epoch": 0.21, "learning_rate": 7.89629573707786e-05, "loss": 1.0249, "step": 84630 }, { "epoch": 0.21, "learning_rate": 7.89616991292969e-05, "loss": 1.0247, "step": 84635 }, { "epoch": 0.21, "learning_rate": 7.896044088781519e-05, "loss": 1.0244, "step": 84640 }, { "epoch": 0.21, "learning_rate": 7.895918264633348e-05, "loss": 1.0257, "step": 84645 }, { "epoch": 0.21, "learning_rate": 7.895792440485177e-05, "loss": 1.0433, "step": 84650 }, { "epoch": 0.21, "learning_rate": 7.895666616337008e-05, "loss": 1.0237, "step": 84655 }, { "epoch": 0.21, "learning_rate": 7.895540792188837e-05, "loss": 1.0245, "step": 84660 }, { "epoch": 0.21, "learning_rate": 7.895414968040666e-05, "loss": 1.0248, "step": 84665 }, { "epoch": 0.21, "learning_rate": 7.895289143892495e-05, "loss": 1.0255, "step": 84670 }, { "epoch": 0.21, "learning_rate": 7.895163319744326e-05, "loss": 1.0248, "step": 84675 }, { "epoch": 0.21, "learning_rate": 7.895037495596155e-05, "loss": 1.0205, "step": 84680 }, { "epoch": 0.21, "learning_rate": 7.894911671447984e-05, "loss": 1.033, "step": 84685 }, { "epoch": 0.21, "learning_rate": 7.894785847299813e-05, "loss": 1.0257, "step": 84690 }, { "epoch": 0.21, "learning_rate": 7.894660023151643e-05, "loss": 1.0233, "step": 84695 }, { "epoch": 0.21, "learning_rate": 7.894534199003473e-05, "loss": 1.0242, "step": 84700 }, { "epoch": 0.21, "learning_rate": 7.894408374855302e-05, "loss": 1.0245, "step": 84705 }, { "epoch": 0.21, "learning_rate": 7.894282550707131e-05, "loss": 1.0257, "step": 84710 }, { "epoch": 0.21, "learning_rate": 7.89415672655896e-05, "loss": 1.0236, "step": 84715 }, { "epoch": 0.21, "learning_rate": 7.894030902410791e-05, "loss": 1.0254, "step": 84720 }, { "epoch": 0.21, "learning_rate": 7.89390507826262e-05, "loss": 1.0265, "step": 84725 }, { "epoch": 0.21, "learning_rate": 7.89377925411445e-05, "loss": 1.0251, "step": 84730 }, { "epoch": 0.21, "learning_rate": 7.893653429966279e-05, "loss": 1.0209, "step": 84735 }, { "epoch": 0.21, "learning_rate": 7.893527605818109e-05, "loss": 1.0231, "step": 84740 }, { "epoch": 0.21, "learning_rate": 7.893401781669938e-05, "loss": 1.0229, "step": 84745 }, { "epoch": 0.21, "learning_rate": 7.893275957521767e-05, "loss": 1.025, "step": 84750 }, { "epoch": 0.21, "learning_rate": 7.893150133373598e-05, "loss": 1.0242, "step": 84755 }, { "epoch": 0.21, "learning_rate": 7.893024309225427e-05, "loss": 1.0244, "step": 84760 }, { "epoch": 0.21, "learning_rate": 7.892898485077256e-05, "loss": 1.0261, "step": 84765 }, { "epoch": 0.21, "learning_rate": 7.892772660929087e-05, "loss": 1.0267, "step": 84770 }, { "epoch": 0.21, "learning_rate": 7.892646836780916e-05, "loss": 1.0265, "step": 84775 }, { "epoch": 0.21, "learning_rate": 7.892521012632745e-05, "loss": 1.0256, "step": 84780 }, { "epoch": 0.21, "learning_rate": 7.892395188484574e-05, "loss": 1.0223, "step": 84785 }, { "epoch": 0.21, "learning_rate": 7.892269364336405e-05, "loss": 1.0245, "step": 84790 }, { "epoch": 0.21, "learning_rate": 7.892143540188234e-05, "loss": 1.0236, "step": 84795 }, { "epoch": 0.21, "learning_rate": 7.892017716040063e-05, "loss": 1.025, "step": 84800 }, { "epoch": 0.21, "learning_rate": 7.891891891891892e-05, "loss": 1.0256, "step": 84805 }, { "epoch": 0.21, "learning_rate": 7.891766067743723e-05, "loss": 1.0346, "step": 84810 }, { "epoch": 0.21, "learning_rate": 7.891640243595552e-05, "loss": 1.0255, "step": 84815 }, { "epoch": 0.21, "learning_rate": 7.891514419447381e-05, "loss": 1.0227, "step": 84820 }, { "epoch": 0.21, "learning_rate": 7.89138859529921e-05, "loss": 1.0234, "step": 84825 }, { "epoch": 0.21, "learning_rate": 7.89126277115104e-05, "loss": 1.0249, "step": 84830 }, { "epoch": 0.21, "learning_rate": 7.89113694700287e-05, "loss": 1.0242, "step": 84835 }, { "epoch": 0.21, "learning_rate": 7.891011122854699e-05, "loss": 1.0239, "step": 84840 }, { "epoch": 0.21, "learning_rate": 7.890885298706528e-05, "loss": 1.0221, "step": 84845 }, { "epoch": 0.21, "learning_rate": 7.890759474558357e-05, "loss": 1.0238, "step": 84850 }, { "epoch": 0.21, "learning_rate": 7.890633650410188e-05, "loss": 1.0262, "step": 84855 }, { "epoch": 0.21, "learning_rate": 7.890507826262017e-05, "loss": 1.0261, "step": 84860 }, { "epoch": 0.21, "learning_rate": 7.890382002113846e-05, "loss": 1.0223, "step": 84865 }, { "epoch": 0.21, "learning_rate": 7.890256177965675e-05, "loss": 1.0238, "step": 84870 }, { "epoch": 0.21, "learning_rate": 7.890130353817506e-05, "loss": 1.0237, "step": 84875 }, { "epoch": 0.21, "learning_rate": 7.890004529669335e-05, "loss": 1.0263, "step": 84880 }, { "epoch": 0.21, "learning_rate": 7.889878705521164e-05, "loss": 1.0249, "step": 84885 }, { "epoch": 0.21, "learning_rate": 7.889752881372993e-05, "loss": 1.0221, "step": 84890 }, { "epoch": 0.21, "learning_rate": 7.889627057224823e-05, "loss": 1.0241, "step": 84895 }, { "epoch": 0.21, "learning_rate": 7.889501233076653e-05, "loss": 1.0249, "step": 84900 }, { "epoch": 0.21, "learning_rate": 7.889375408928482e-05, "loss": 1.0233, "step": 84905 }, { "epoch": 0.21, "learning_rate": 7.889249584780311e-05, "loss": 1.0216, "step": 84910 }, { "epoch": 0.21, "learning_rate": 7.88912376063214e-05, "loss": 1.025, "step": 84915 }, { "epoch": 0.21, "learning_rate": 7.888997936483971e-05, "loss": 1.0238, "step": 84920 }, { "epoch": 0.21, "learning_rate": 7.8888721123358e-05, "loss": 1.0254, "step": 84925 }, { "epoch": 0.21, "learning_rate": 7.88874628818763e-05, "loss": 1.0263, "step": 84930 }, { "epoch": 0.21, "learning_rate": 7.888620464039459e-05, "loss": 1.0236, "step": 84935 }, { "epoch": 0.21, "learning_rate": 7.888494639891289e-05, "loss": 1.024, "step": 84940 }, { "epoch": 0.21, "learning_rate": 7.888368815743118e-05, "loss": 1.0248, "step": 84945 }, { "epoch": 0.21, "learning_rate": 7.888242991594947e-05, "loss": 1.0252, "step": 84950 }, { "epoch": 0.21, "learning_rate": 7.888117167446777e-05, "loss": 1.0241, "step": 84955 }, { "epoch": 0.21, "learning_rate": 7.887991343298606e-05, "loss": 1.0241, "step": 84960 }, { "epoch": 0.21, "learning_rate": 7.887865519150436e-05, "loss": 1.0234, "step": 84965 }, { "epoch": 0.21, "learning_rate": 7.887739695002265e-05, "loss": 1.0253, "step": 84970 }, { "epoch": 0.21, "learning_rate": 7.887613870854095e-05, "loss": 1.0237, "step": 84975 }, { "epoch": 0.21, "learning_rate": 7.887488046705924e-05, "loss": 1.0225, "step": 84980 }, { "epoch": 0.21, "learning_rate": 7.887362222557754e-05, "loss": 1.0246, "step": 84985 }, { "epoch": 0.21, "learning_rate": 7.887236398409583e-05, "loss": 1.0235, "step": 84990 }, { "epoch": 0.21, "learning_rate": 7.887110574261413e-05, "loss": 1.0437, "step": 84995 }, { "epoch": 0.21, "learning_rate": 7.886984750113242e-05, "loss": 1.0254, "step": 85000 }, { "epoch": 0.21, "learning_rate": 7.886858925965072e-05, "loss": 1.0245, "step": 85005 }, { "epoch": 0.21, "learning_rate": 7.886733101816901e-05, "loss": 1.0237, "step": 85010 }, { "epoch": 0.21, "learning_rate": 7.88660727766873e-05, "loss": 1.0243, "step": 85015 }, { "epoch": 0.21, "learning_rate": 7.88648145352056e-05, "loss": 1.0237, "step": 85020 }, { "epoch": 0.21, "learning_rate": 7.886355629372389e-05, "loss": 1.0243, "step": 85025 }, { "epoch": 0.21, "learning_rate": 7.88622980522422e-05, "loss": 1.0255, "step": 85030 }, { "epoch": 0.21, "learning_rate": 7.886103981076049e-05, "loss": 1.0219, "step": 85035 }, { "epoch": 0.21, "learning_rate": 7.885978156927878e-05, "loss": 1.0227, "step": 85040 }, { "epoch": 0.21, "learning_rate": 7.885852332779707e-05, "loss": 1.024, "step": 85045 }, { "epoch": 0.21, "learning_rate": 7.885726508631537e-05, "loss": 1.0231, "step": 85050 }, { "epoch": 0.21, "learning_rate": 7.885600684483367e-05, "loss": 1.024, "step": 85055 }, { "epoch": 0.21, "learning_rate": 7.885474860335196e-05, "loss": 1.0246, "step": 85060 }, { "epoch": 0.21, "learning_rate": 7.885349036187025e-05, "loss": 1.0239, "step": 85065 }, { "epoch": 0.21, "learning_rate": 7.885223212038855e-05, "loss": 1.0249, "step": 85070 }, { "epoch": 0.21, "learning_rate": 7.885097387890685e-05, "loss": 1.0459, "step": 85075 }, { "epoch": 0.21, "learning_rate": 7.884971563742514e-05, "loss": 1.0234, "step": 85080 }, { "epoch": 0.21, "learning_rate": 7.884845739594343e-05, "loss": 1.0235, "step": 85085 }, { "epoch": 0.21, "learning_rate": 7.884719915446172e-05, "loss": 1.026, "step": 85090 }, { "epoch": 0.21, "learning_rate": 7.884594091298003e-05, "loss": 1.0254, "step": 85095 }, { "epoch": 0.21, "learning_rate": 7.884468267149832e-05, "loss": 1.0243, "step": 85100 }, { "epoch": 0.21, "learning_rate": 7.884342443001661e-05, "loss": 1.0243, "step": 85105 }, { "epoch": 0.21, "learning_rate": 7.88421661885349e-05, "loss": 1.0492, "step": 85110 }, { "epoch": 0.21, "learning_rate": 7.88409079470532e-05, "loss": 1.0256, "step": 85115 }, { "epoch": 0.21, "learning_rate": 7.88396497055715e-05, "loss": 1.0224, "step": 85120 }, { "epoch": 0.21, "learning_rate": 7.883839146408979e-05, "loss": 1.0236, "step": 85125 }, { "epoch": 0.21, "learning_rate": 7.883713322260808e-05, "loss": 1.0225, "step": 85130 }, { "epoch": 0.21, "learning_rate": 7.883587498112638e-05, "loss": 1.0235, "step": 85135 }, { "epoch": 0.21, "learning_rate": 7.883461673964468e-05, "loss": 1.0254, "step": 85140 }, { "epoch": 0.21, "learning_rate": 7.883335849816297e-05, "loss": 1.0232, "step": 85145 }, { "epoch": 0.21, "learning_rate": 7.883210025668126e-05, "loss": 1.021, "step": 85150 }, { "epoch": 0.21, "learning_rate": 7.883084201519955e-05, "loss": 1.0235, "step": 85155 }, { "epoch": 0.21, "learning_rate": 7.882958377371786e-05, "loss": 1.0255, "step": 85160 }, { "epoch": 0.21, "learning_rate": 7.882832553223615e-05, "loss": 1.0241, "step": 85165 }, { "epoch": 0.21, "learning_rate": 7.882706729075444e-05, "loss": 1.0238, "step": 85170 }, { "epoch": 0.21, "learning_rate": 7.882580904927273e-05, "loss": 1.0258, "step": 85175 }, { "epoch": 0.21, "learning_rate": 7.882455080779104e-05, "loss": 1.0264, "step": 85180 }, { "epoch": 0.21, "learning_rate": 7.882329256630933e-05, "loss": 1.0238, "step": 85185 }, { "epoch": 0.21, "learning_rate": 7.882203432482762e-05, "loss": 1.0237, "step": 85190 }, { "epoch": 0.21, "learning_rate": 7.882077608334591e-05, "loss": 1.0211, "step": 85195 }, { "epoch": 0.21, "learning_rate": 7.881951784186422e-05, "loss": 1.0225, "step": 85200 }, { "epoch": 0.21, "learning_rate": 7.881825960038251e-05, "loss": 1.0206, "step": 85205 }, { "epoch": 0.21, "learning_rate": 7.88170013589008e-05, "loss": 1.0239, "step": 85210 }, { "epoch": 0.21, "learning_rate": 7.881574311741909e-05, "loss": 1.0249, "step": 85215 }, { "epoch": 0.21, "learning_rate": 7.881448487593738e-05, "loss": 1.027, "step": 85220 }, { "epoch": 0.21, "learning_rate": 7.881322663445569e-05, "loss": 1.0238, "step": 85225 }, { "epoch": 0.21, "learning_rate": 7.881196839297398e-05, "loss": 1.0245, "step": 85230 }, { "epoch": 0.21, "learning_rate": 7.881071015149227e-05, "loss": 1.0303, "step": 85235 }, { "epoch": 0.21, "learning_rate": 7.880945191001056e-05, "loss": 1.0253, "step": 85240 }, { "epoch": 0.21, "learning_rate": 7.880819366852887e-05, "loss": 1.0258, "step": 85245 }, { "epoch": 0.21, "learning_rate": 7.880693542704716e-05, "loss": 1.0244, "step": 85250 }, { "epoch": 0.21, "learning_rate": 7.880567718556546e-05, "loss": 1.0214, "step": 85255 }, { "epoch": 0.21, "learning_rate": 7.880441894408376e-05, "loss": 1.0239, "step": 85260 }, { "epoch": 0.21, "learning_rate": 7.880316070260205e-05, "loss": 1.0246, "step": 85265 }, { "epoch": 0.21, "learning_rate": 7.880190246112035e-05, "loss": 1.0225, "step": 85270 }, { "epoch": 0.21, "learning_rate": 7.880064421963864e-05, "loss": 1.0226, "step": 85275 }, { "epoch": 0.21, "learning_rate": 7.879938597815694e-05, "loss": 1.0263, "step": 85280 }, { "epoch": 0.21, "learning_rate": 7.879812773667523e-05, "loss": 1.0231, "step": 85285 }, { "epoch": 0.21, "learning_rate": 7.879686949519352e-05, "loss": 1.0239, "step": 85290 }, { "epoch": 0.21, "learning_rate": 7.879561125371182e-05, "loss": 1.0239, "step": 85295 }, { "epoch": 0.21, "learning_rate": 7.879435301223012e-05, "loss": 1.025, "step": 85300 }, { "epoch": 0.21, "learning_rate": 7.879309477074841e-05, "loss": 1.0261, "step": 85305 }, { "epoch": 0.21, "learning_rate": 7.87918365292667e-05, "loss": 1.0252, "step": 85310 }, { "epoch": 0.21, "learning_rate": 7.8790578287785e-05, "loss": 1.0253, "step": 85315 }, { "epoch": 0.21, "learning_rate": 7.87893200463033e-05, "loss": 1.024, "step": 85320 }, { "epoch": 0.21, "learning_rate": 7.878806180482159e-05, "loss": 1.0251, "step": 85325 }, { "epoch": 0.21, "learning_rate": 7.878680356333988e-05, "loss": 1.0473, "step": 85330 }, { "epoch": 0.21, "learning_rate": 7.878554532185818e-05, "loss": 1.0238, "step": 85335 }, { "epoch": 0.21, "learning_rate": 7.878428708037648e-05, "loss": 1.026, "step": 85340 }, { "epoch": 0.21, "learning_rate": 7.878302883889477e-05, "loss": 1.0245, "step": 85345 }, { "epoch": 0.21, "learning_rate": 7.878177059741306e-05, "loss": 1.0236, "step": 85350 }, { "epoch": 0.21, "learning_rate": 7.878051235593135e-05, "loss": 1.0227, "step": 85355 }, { "epoch": 0.21, "learning_rate": 7.877925411444966e-05, "loss": 1.0247, "step": 85360 }, { "epoch": 0.21, "learning_rate": 7.877799587296795e-05, "loss": 1.0239, "step": 85365 }, { "epoch": 0.21, "learning_rate": 7.877673763148624e-05, "loss": 1.0257, "step": 85370 }, { "epoch": 0.21, "learning_rate": 7.877547939000453e-05, "loss": 1.0256, "step": 85375 }, { "epoch": 0.21, "learning_rate": 7.877422114852284e-05, "loss": 1.024, "step": 85380 }, { "epoch": 0.21, "learning_rate": 7.877296290704113e-05, "loss": 1.023, "step": 85385 }, { "epoch": 0.21, "learning_rate": 7.877170466555942e-05, "loss": 1.0238, "step": 85390 }, { "epoch": 0.21, "learning_rate": 7.877044642407771e-05, "loss": 1.0365, "step": 85395 }, { "epoch": 0.21, "learning_rate": 7.876918818259602e-05, "loss": 1.0262, "step": 85400 }, { "epoch": 0.21, "learning_rate": 7.876792994111431e-05, "loss": 1.023, "step": 85405 }, { "epoch": 0.21, "learning_rate": 7.87666716996326e-05, "loss": 1.0248, "step": 85410 }, { "epoch": 0.21, "learning_rate": 7.876541345815089e-05, "loss": 1.0246, "step": 85415 }, { "epoch": 0.21, "learning_rate": 7.876415521666918e-05, "loss": 1.0288, "step": 85420 }, { "epoch": 0.21, "learning_rate": 7.876289697518749e-05, "loss": 1.0228, "step": 85425 }, { "epoch": 0.21, "learning_rate": 7.876163873370578e-05, "loss": 1.0247, "step": 85430 }, { "epoch": 0.21, "learning_rate": 7.876038049222407e-05, "loss": 1.0212, "step": 85435 }, { "epoch": 0.21, "learning_rate": 7.875912225074236e-05, "loss": 1.0261, "step": 85440 }, { "epoch": 0.21, "learning_rate": 7.875786400926067e-05, "loss": 1.0244, "step": 85445 }, { "epoch": 0.21, "learning_rate": 7.875660576777896e-05, "loss": 1.0198, "step": 85450 }, { "epoch": 0.21, "learning_rate": 7.875534752629725e-05, "loss": 1.0243, "step": 85455 }, { "epoch": 0.21, "learning_rate": 7.875408928481554e-05, "loss": 1.0233, "step": 85460 }, { "epoch": 0.21, "learning_rate": 7.875283104333385e-05, "loss": 1.0246, "step": 85465 }, { "epoch": 0.21, "learning_rate": 7.875157280185214e-05, "loss": 1.0256, "step": 85470 }, { "epoch": 0.21, "learning_rate": 7.875031456037043e-05, "loss": 1.0241, "step": 85475 }, { "epoch": 0.21, "learning_rate": 7.874905631888872e-05, "loss": 1.0215, "step": 85480 }, { "epoch": 0.21, "learning_rate": 7.874779807740701e-05, "loss": 1.048, "step": 85485 }, { "epoch": 0.21, "learning_rate": 7.874653983592532e-05, "loss": 1.0245, "step": 85490 }, { "epoch": 0.21, "learning_rate": 7.874528159444361e-05, "loss": 1.0234, "step": 85495 }, { "epoch": 0.21, "learning_rate": 7.87440233529619e-05, "loss": 1.0243, "step": 85500 }, { "epoch": 0.21, "learning_rate": 7.87427651114802e-05, "loss": 1.0258, "step": 85505 }, { "epoch": 0.21, "learning_rate": 7.87415068699985e-05, "loss": 1.0239, "step": 85510 }, { "epoch": 0.21, "learning_rate": 7.874024862851679e-05, "loss": 1.0223, "step": 85515 }, { "epoch": 0.21, "learning_rate": 7.873899038703508e-05, "loss": 1.0262, "step": 85520 }, { "epoch": 0.21, "learning_rate": 7.873773214555337e-05, "loss": 1.0274, "step": 85525 }, { "epoch": 0.21, "learning_rate": 7.873647390407168e-05, "loss": 1.0373, "step": 85530 }, { "epoch": 0.21, "learning_rate": 7.873521566258997e-05, "loss": 1.024, "step": 85535 }, { "epoch": 0.21, "learning_rate": 7.873395742110826e-05, "loss": 1.0244, "step": 85540 }, { "epoch": 0.21, "learning_rate": 7.873269917962655e-05, "loss": 1.025, "step": 85545 }, { "epoch": 0.21, "learning_rate": 7.873144093814484e-05, "loss": 1.0243, "step": 85550 }, { "epoch": 0.21, "learning_rate": 7.873018269666315e-05, "loss": 1.0263, "step": 85555 }, { "epoch": 0.21, "learning_rate": 7.872892445518144e-05, "loss": 1.0231, "step": 85560 }, { "epoch": 0.21, "learning_rate": 7.872766621369973e-05, "loss": 1.0224, "step": 85565 }, { "epoch": 0.21, "learning_rate": 7.872640797221802e-05, "loss": 1.0255, "step": 85570 }, { "epoch": 0.21, "learning_rate": 7.872514973073633e-05, "loss": 1.0232, "step": 85575 }, { "epoch": 0.21, "learning_rate": 7.872389148925462e-05, "loss": 1.0234, "step": 85580 }, { "epoch": 0.21, "learning_rate": 7.872263324777291e-05, "loss": 1.0248, "step": 85585 }, { "epoch": 0.21, "learning_rate": 7.87213750062912e-05, "loss": 1.0481, "step": 85590 }, { "epoch": 0.21, "learning_rate": 7.872011676480951e-05, "loss": 1.0234, "step": 85595 }, { "epoch": 0.21, "learning_rate": 7.87188585233278e-05, "loss": 1.0217, "step": 85600 }, { "epoch": 0.21, "learning_rate": 7.871760028184609e-05, "loss": 1.0251, "step": 85605 }, { "epoch": 0.21, "learning_rate": 7.871634204036438e-05, "loss": 1.0244, "step": 85610 }, { "epoch": 0.21, "learning_rate": 7.871508379888268e-05, "loss": 1.0236, "step": 85615 }, { "epoch": 0.21, "learning_rate": 7.871382555740098e-05, "loss": 1.0236, "step": 85620 }, { "epoch": 0.21, "learning_rate": 7.871256731591927e-05, "loss": 1.0266, "step": 85625 }, { "epoch": 0.21, "learning_rate": 7.871130907443756e-05, "loss": 1.0254, "step": 85630 }, { "epoch": 0.21, "learning_rate": 7.871005083295586e-05, "loss": 1.026, "step": 85635 }, { "epoch": 0.21, "learning_rate": 7.870879259147416e-05, "loss": 1.0246, "step": 85640 }, { "epoch": 0.21, "learning_rate": 7.870753434999245e-05, "loss": 1.0229, "step": 85645 }, { "epoch": 0.21, "learning_rate": 7.870627610851074e-05, "loss": 1.0232, "step": 85650 }, { "epoch": 0.22, "learning_rate": 7.870501786702904e-05, "loss": 1.0244, "step": 85655 }, { "epoch": 0.22, "learning_rate": 7.870375962554734e-05, "loss": 1.0224, "step": 85660 }, { "epoch": 0.22, "learning_rate": 7.870250138406563e-05, "loss": 1.023, "step": 85665 }, { "epoch": 0.22, "learning_rate": 7.870124314258392e-05, "loss": 1.0223, "step": 85670 }, { "epoch": 0.22, "learning_rate": 7.869998490110222e-05, "loss": 1.0235, "step": 85675 }, { "epoch": 0.22, "learning_rate": 7.869872665962051e-05, "loss": 1.0227, "step": 85680 }, { "epoch": 0.22, "learning_rate": 7.869746841813881e-05, "loss": 1.0244, "step": 85685 }, { "epoch": 0.22, "learning_rate": 7.86962101766571e-05, "loss": 1.0251, "step": 85690 }, { "epoch": 0.22, "learning_rate": 7.86949519351754e-05, "loss": 1.0255, "step": 85695 }, { "epoch": 0.22, "learning_rate": 7.869369369369369e-05, "loss": 1.025, "step": 85700 }, { "epoch": 0.22, "learning_rate": 7.869243545221199e-05, "loss": 1.0248, "step": 85705 }, { "epoch": 0.22, "learning_rate": 7.869117721073028e-05, "loss": 1.0239, "step": 85710 }, { "epoch": 0.22, "learning_rate": 7.868991896924858e-05, "loss": 1.0224, "step": 85715 }, { "epoch": 0.22, "learning_rate": 7.868866072776687e-05, "loss": 1.0246, "step": 85720 }, { "epoch": 0.22, "learning_rate": 7.868740248628517e-05, "loss": 1.0248, "step": 85725 }, { "epoch": 0.22, "learning_rate": 7.868614424480346e-05, "loss": 1.0256, "step": 85730 }, { "epoch": 0.22, "learning_rate": 7.868488600332176e-05, "loss": 1.0246, "step": 85735 }, { "epoch": 0.22, "learning_rate": 7.868362776184005e-05, "loss": 1.0259, "step": 85740 }, { "epoch": 0.22, "learning_rate": 7.868236952035834e-05, "loss": 1.0231, "step": 85745 }, { "epoch": 0.22, "learning_rate": 7.868111127887664e-05, "loss": 1.0248, "step": 85750 }, { "epoch": 0.22, "learning_rate": 7.867985303739495e-05, "loss": 1.0227, "step": 85755 }, { "epoch": 0.22, "learning_rate": 7.867859479591324e-05, "loss": 1.0232, "step": 85760 }, { "epoch": 0.22, "learning_rate": 7.867733655443153e-05, "loss": 1.0238, "step": 85765 }, { "epoch": 0.22, "learning_rate": 7.867607831294982e-05, "loss": 1.0234, "step": 85770 }, { "epoch": 0.22, "learning_rate": 7.867482007146813e-05, "loss": 1.0255, "step": 85775 }, { "epoch": 0.22, "learning_rate": 7.867356182998642e-05, "loss": 1.0252, "step": 85780 }, { "epoch": 0.22, "learning_rate": 7.867230358850471e-05, "loss": 1.0236, "step": 85785 }, { "epoch": 0.22, "learning_rate": 7.8671045347023e-05, "loss": 1.025, "step": 85790 }, { "epoch": 0.22, "learning_rate": 7.866978710554131e-05, "loss": 1.0236, "step": 85795 }, { "epoch": 0.22, "learning_rate": 7.86685288640596e-05, "loss": 1.0233, "step": 85800 }, { "epoch": 0.22, "learning_rate": 7.866727062257789e-05, "loss": 1.0262, "step": 85805 }, { "epoch": 0.22, "learning_rate": 7.866601238109618e-05, "loss": 1.0242, "step": 85810 }, { "epoch": 0.22, "learning_rate": 7.866475413961449e-05, "loss": 1.0249, "step": 85815 }, { "epoch": 0.22, "learning_rate": 7.866349589813278e-05, "loss": 1.0221, "step": 85820 }, { "epoch": 0.22, "learning_rate": 7.866223765665107e-05, "loss": 1.0243, "step": 85825 }, { "epoch": 0.22, "learning_rate": 7.866097941516936e-05, "loss": 1.0218, "step": 85830 }, { "epoch": 0.22, "learning_rate": 7.865972117368766e-05, "loss": 1.0261, "step": 85835 }, { "epoch": 0.22, "learning_rate": 7.865846293220596e-05, "loss": 1.025, "step": 85840 }, { "epoch": 0.22, "learning_rate": 7.865720469072425e-05, "loss": 1.0249, "step": 85845 }, { "epoch": 0.22, "learning_rate": 7.865594644924254e-05, "loss": 1.0477, "step": 85850 }, { "epoch": 0.22, "learning_rate": 7.865468820776084e-05, "loss": 1.0251, "step": 85855 }, { "epoch": 0.22, "learning_rate": 7.865342996627914e-05, "loss": 1.0233, "step": 85860 }, { "epoch": 0.22, "learning_rate": 7.865217172479743e-05, "loss": 1.0263, "step": 85865 }, { "epoch": 0.22, "learning_rate": 7.865091348331572e-05, "loss": 1.024, "step": 85870 }, { "epoch": 0.22, "learning_rate": 7.864965524183402e-05, "loss": 1.0238, "step": 85875 }, { "epoch": 0.22, "learning_rate": 7.864839700035232e-05, "loss": 1.0235, "step": 85880 }, { "epoch": 0.22, "learning_rate": 7.864713875887061e-05, "loss": 1.0248, "step": 85885 }, { "epoch": 0.22, "learning_rate": 7.86458805173889e-05, "loss": 1.0235, "step": 85890 }, { "epoch": 0.22, "learning_rate": 7.86446222759072e-05, "loss": 1.0227, "step": 85895 }, { "epoch": 0.22, "learning_rate": 7.864336403442549e-05, "loss": 1.0226, "step": 85900 }, { "epoch": 0.22, "learning_rate": 7.864210579294379e-05, "loss": 1.0247, "step": 85905 }, { "epoch": 0.22, "learning_rate": 7.864084755146208e-05, "loss": 1.023, "step": 85910 }, { "epoch": 0.22, "learning_rate": 7.863958930998038e-05, "loss": 1.0246, "step": 85915 }, { "epoch": 0.22, "learning_rate": 7.863833106849867e-05, "loss": 1.0234, "step": 85920 }, { "epoch": 0.22, "learning_rate": 7.863707282701697e-05, "loss": 1.0262, "step": 85925 }, { "epoch": 0.22, "learning_rate": 7.863581458553526e-05, "loss": 1.0241, "step": 85930 }, { "epoch": 0.22, "learning_rate": 7.863455634405356e-05, "loss": 1.0246, "step": 85935 }, { "epoch": 0.22, "learning_rate": 7.863329810257185e-05, "loss": 1.0276, "step": 85940 }, { "epoch": 0.22, "learning_rate": 7.863203986109014e-05, "loss": 1.0244, "step": 85945 }, { "epoch": 0.22, "learning_rate": 7.863078161960844e-05, "loss": 1.0252, "step": 85950 }, { "epoch": 0.22, "learning_rate": 7.862952337812674e-05, "loss": 1.021, "step": 85955 }, { "epoch": 0.22, "learning_rate": 7.862826513664503e-05, "loss": 1.024, "step": 85960 }, { "epoch": 0.22, "learning_rate": 7.862700689516332e-05, "loss": 1.0231, "step": 85965 }, { "epoch": 0.22, "learning_rate": 7.862574865368162e-05, "loss": 1.0248, "step": 85970 }, { "epoch": 0.22, "learning_rate": 7.862449041219992e-05, "loss": 1.0245, "step": 85975 }, { "epoch": 0.22, "learning_rate": 7.862323217071821e-05, "loss": 1.0254, "step": 85980 }, { "epoch": 0.22, "learning_rate": 7.86219739292365e-05, "loss": 1.0256, "step": 85985 }, { "epoch": 0.22, "learning_rate": 7.86207156877548e-05, "loss": 1.0235, "step": 85990 }, { "epoch": 0.22, "learning_rate": 7.86194574462731e-05, "loss": 1.024, "step": 85995 }, { "epoch": 0.22, "learning_rate": 7.861819920479139e-05, "loss": 1.026, "step": 86000 }, { "epoch": 0.22, "learning_rate": 7.861694096330968e-05, "loss": 1.0252, "step": 86005 }, { "epoch": 0.22, "learning_rate": 7.861568272182797e-05, "loss": 1.0247, "step": 86010 }, { "epoch": 0.22, "learning_rate": 7.861442448034628e-05, "loss": 1.0214, "step": 86015 }, { "epoch": 0.22, "learning_rate": 7.861316623886457e-05, "loss": 1.0241, "step": 86020 }, { "epoch": 0.22, "learning_rate": 7.861190799738286e-05, "loss": 1.0255, "step": 86025 }, { "epoch": 0.22, "learning_rate": 7.861064975590115e-05, "loss": 1.025, "step": 86030 }, { "epoch": 0.22, "learning_rate": 7.860939151441946e-05, "loss": 1.0221, "step": 86035 }, { "epoch": 0.22, "learning_rate": 7.860813327293775e-05, "loss": 1.023, "step": 86040 }, { "epoch": 0.22, "learning_rate": 7.860687503145604e-05, "loss": 1.0245, "step": 86045 }, { "epoch": 0.22, "learning_rate": 7.860561678997433e-05, "loss": 1.0247, "step": 86050 }, { "epoch": 0.22, "learning_rate": 7.860435854849264e-05, "loss": 1.0227, "step": 86055 }, { "epoch": 0.22, "learning_rate": 7.860310030701093e-05, "loss": 1.0249, "step": 86060 }, { "epoch": 0.22, "learning_rate": 7.860184206552922e-05, "loss": 1.0245, "step": 86065 }, { "epoch": 0.22, "learning_rate": 7.860058382404751e-05, "loss": 1.0248, "step": 86070 }, { "epoch": 0.22, "learning_rate": 7.85993255825658e-05, "loss": 1.025, "step": 86075 }, { "epoch": 0.22, "learning_rate": 7.85980673410841e-05, "loss": 1.0255, "step": 86080 }, { "epoch": 0.22, "learning_rate": 7.85968090996024e-05, "loss": 1.0227, "step": 86085 }, { "epoch": 0.22, "learning_rate": 7.859555085812069e-05, "loss": 1.0235, "step": 86090 }, { "epoch": 0.22, "learning_rate": 7.859429261663898e-05, "loss": 1.0247, "step": 86095 }, { "epoch": 0.22, "learning_rate": 7.859303437515729e-05, "loss": 1.024, "step": 86100 }, { "epoch": 0.22, "learning_rate": 7.859177613367558e-05, "loss": 1.0244, "step": 86105 }, { "epoch": 0.22, "learning_rate": 7.859051789219387e-05, "loss": 1.0237, "step": 86110 }, { "epoch": 0.22, "learning_rate": 7.858925965071216e-05, "loss": 1.0256, "step": 86115 }, { "epoch": 0.22, "learning_rate": 7.858800140923047e-05, "loss": 1.0494, "step": 86120 }, { "epoch": 0.22, "learning_rate": 7.858674316774876e-05, "loss": 1.0244, "step": 86125 }, { "epoch": 0.22, "learning_rate": 7.858548492626705e-05, "loss": 1.0237, "step": 86130 }, { "epoch": 0.22, "learning_rate": 7.858422668478534e-05, "loss": 1.0244, "step": 86135 }, { "epoch": 0.22, "learning_rate": 7.858296844330363e-05, "loss": 1.0234, "step": 86140 }, { "epoch": 0.22, "learning_rate": 7.858171020182194e-05, "loss": 1.0234, "step": 86145 }, { "epoch": 0.22, "learning_rate": 7.858045196034023e-05, "loss": 1.0269, "step": 86150 }, { "epoch": 0.22, "learning_rate": 7.857919371885852e-05, "loss": 1.0258, "step": 86155 }, { "epoch": 0.22, "learning_rate": 7.857793547737681e-05, "loss": 1.0247, "step": 86160 }, { "epoch": 0.22, "learning_rate": 7.857667723589512e-05, "loss": 1.0465, "step": 86165 }, { "epoch": 0.22, "learning_rate": 7.857541899441341e-05, "loss": 1.0245, "step": 86170 }, { "epoch": 0.22, "learning_rate": 7.85741607529317e-05, "loss": 1.0236, "step": 86175 }, { "epoch": 0.22, "learning_rate": 7.857290251144999e-05, "loss": 1.0248, "step": 86180 }, { "epoch": 0.22, "learning_rate": 7.85716442699683e-05, "loss": 1.0248, "step": 86185 }, { "epoch": 0.22, "learning_rate": 7.857038602848659e-05, "loss": 1.022, "step": 86190 }, { "epoch": 0.22, "learning_rate": 7.856912778700488e-05, "loss": 1.0224, "step": 86195 }, { "epoch": 0.22, "learning_rate": 7.856786954552317e-05, "loss": 1.0254, "step": 86200 }, { "epoch": 0.22, "learning_rate": 7.856661130404146e-05, "loss": 1.0255, "step": 86205 }, { "epoch": 0.22, "learning_rate": 7.856535306255977e-05, "loss": 1.026, "step": 86210 }, { "epoch": 0.22, "learning_rate": 7.856409482107806e-05, "loss": 1.0221, "step": 86215 }, { "epoch": 0.22, "learning_rate": 7.856283657959635e-05, "loss": 1.0244, "step": 86220 }, { "epoch": 0.22, "learning_rate": 7.856157833811464e-05, "loss": 1.0237, "step": 86225 }, { "epoch": 0.22, "learning_rate": 7.856032009663295e-05, "loss": 1.0237, "step": 86230 }, { "epoch": 0.22, "learning_rate": 7.855906185515124e-05, "loss": 1.0229, "step": 86235 }, { "epoch": 0.22, "learning_rate": 7.855780361366953e-05, "loss": 1.0252, "step": 86240 }, { "epoch": 0.22, "learning_rate": 7.855654537218782e-05, "loss": 1.0224, "step": 86245 }, { "epoch": 0.22, "learning_rate": 7.855528713070613e-05, "loss": 1.0259, "step": 86250 }, { "epoch": 0.22, "learning_rate": 7.855402888922443e-05, "loss": 1.0263, "step": 86255 }, { "epoch": 0.22, "learning_rate": 7.855277064774273e-05, "loss": 1.0235, "step": 86260 }, { "epoch": 0.22, "learning_rate": 7.855151240626102e-05, "loss": 1.0233, "step": 86265 }, { "epoch": 0.22, "learning_rate": 7.855025416477931e-05, "loss": 1.0237, "step": 86270 }, { "epoch": 0.22, "learning_rate": 7.854899592329761e-05, "loss": 1.0251, "step": 86275 }, { "epoch": 0.22, "learning_rate": 7.85477376818159e-05, "loss": 1.0251, "step": 86280 }, { "epoch": 0.22, "learning_rate": 7.85464794403342e-05, "loss": 1.0241, "step": 86285 }, { "epoch": 0.22, "learning_rate": 7.854522119885249e-05, "loss": 1.0262, "step": 86290 }, { "epoch": 0.22, "learning_rate": 7.854396295737078e-05, "loss": 1.0452, "step": 86295 }, { "epoch": 0.22, "learning_rate": 7.854270471588909e-05, "loss": 1.0256, "step": 86300 }, { "epoch": 0.22, "learning_rate": 7.854144647440738e-05, "loss": 1.0256, "step": 86305 }, { "epoch": 0.22, "learning_rate": 7.854018823292567e-05, "loss": 1.0231, "step": 86310 }, { "epoch": 0.22, "learning_rate": 7.853892999144396e-05, "loss": 1.0251, "step": 86315 }, { "epoch": 0.22, "learning_rate": 7.853767174996227e-05, "loss": 1.0252, "step": 86320 }, { "epoch": 0.22, "learning_rate": 7.853641350848056e-05, "loss": 1.0255, "step": 86325 }, { "epoch": 0.22, "learning_rate": 7.853515526699885e-05, "loss": 1.0246, "step": 86330 }, { "epoch": 0.22, "learning_rate": 7.853389702551714e-05, "loss": 1.0227, "step": 86335 }, { "epoch": 0.22, "learning_rate": 7.853263878403545e-05, "loss": 1.0227, "step": 86340 }, { "epoch": 0.22, "learning_rate": 7.853138054255374e-05, "loss": 1.0255, "step": 86345 }, { "epoch": 0.22, "learning_rate": 7.853012230107203e-05, "loss": 1.0226, "step": 86350 }, { "epoch": 0.22, "learning_rate": 7.852886405959032e-05, "loss": 1.0242, "step": 86355 }, { "epoch": 0.22, "learning_rate": 7.852760581810861e-05, "loss": 1.0259, "step": 86360 }, { "epoch": 0.22, "learning_rate": 7.852634757662692e-05, "loss": 1.023, "step": 86365 }, { "epoch": 0.22, "learning_rate": 7.852508933514521e-05, "loss": 1.0258, "step": 86370 }, { "epoch": 0.22, "learning_rate": 7.85238310936635e-05, "loss": 1.025, "step": 86375 }, { "epoch": 0.22, "learning_rate": 7.852257285218179e-05, "loss": 1.0237, "step": 86380 }, { "epoch": 0.22, "learning_rate": 7.85213146107001e-05, "loss": 1.0251, "step": 86385 }, { "epoch": 0.22, "learning_rate": 7.852005636921839e-05, "loss": 1.0244, "step": 86390 }, { "epoch": 0.22, "learning_rate": 7.851879812773668e-05, "loss": 1.027, "step": 86395 }, { "epoch": 0.22, "learning_rate": 7.851753988625497e-05, "loss": 1.0241, "step": 86400 }, { "epoch": 0.22, "learning_rate": 7.851628164477328e-05, "loss": 1.0258, "step": 86405 }, { "epoch": 0.22, "learning_rate": 7.851502340329157e-05, "loss": 1.0238, "step": 86410 }, { "epoch": 0.22, "learning_rate": 7.851376516180986e-05, "loss": 1.0259, "step": 86415 }, { "epoch": 0.22, "learning_rate": 7.851250692032815e-05, "loss": 1.0219, "step": 86420 }, { "epoch": 0.22, "learning_rate": 7.851124867884644e-05, "loss": 1.0484, "step": 86425 }, { "epoch": 0.22, "learning_rate": 7.850999043736475e-05, "loss": 1.0231, "step": 86430 }, { "epoch": 0.22, "learning_rate": 7.850873219588304e-05, "loss": 1.0243, "step": 86435 }, { "epoch": 0.22, "learning_rate": 7.850747395440133e-05, "loss": 1.0228, "step": 86440 }, { "epoch": 0.22, "learning_rate": 7.850621571291962e-05, "loss": 1.0236, "step": 86445 }, { "epoch": 0.22, "learning_rate": 7.850495747143793e-05, "loss": 1.0248, "step": 86450 }, { "epoch": 0.22, "learning_rate": 7.850369922995622e-05, "loss": 1.0259, "step": 86455 }, { "epoch": 0.22, "learning_rate": 7.850244098847451e-05, "loss": 1.0235, "step": 86460 }, { "epoch": 0.22, "learning_rate": 7.85011827469928e-05, "loss": 1.0275, "step": 86465 }, { "epoch": 0.22, "learning_rate": 7.849992450551111e-05, "loss": 1.0247, "step": 86470 }, { "epoch": 0.22, "learning_rate": 7.84986662640294e-05, "loss": 1.0252, "step": 86475 }, { "epoch": 0.22, "learning_rate": 7.849740802254769e-05, "loss": 1.0238, "step": 86480 }, { "epoch": 0.22, "learning_rate": 7.849614978106598e-05, "loss": 1.0226, "step": 86485 }, { "epoch": 0.22, "learning_rate": 7.849489153958428e-05, "loss": 1.0236, "step": 86490 }, { "epoch": 0.22, "learning_rate": 7.849363329810258e-05, "loss": 1.0227, "step": 86495 }, { "epoch": 0.22, "learning_rate": 7.849237505662087e-05, "loss": 1.0244, "step": 86500 }, { "epoch": 0.22, "learning_rate": 7.849111681513916e-05, "loss": 1.022, "step": 86505 }, { "epoch": 0.22, "learning_rate": 7.848985857365746e-05, "loss": 1.0218, "step": 86510 }, { "epoch": 0.22, "learning_rate": 7.848860033217576e-05, "loss": 1.0263, "step": 86515 }, { "epoch": 0.22, "learning_rate": 7.848734209069405e-05, "loss": 1.0287, "step": 86520 }, { "epoch": 0.22, "learning_rate": 7.848608384921234e-05, "loss": 1.0251, "step": 86525 }, { "epoch": 0.22, "learning_rate": 7.848482560773064e-05, "loss": 1.0259, "step": 86530 }, { "epoch": 0.22, "learning_rate": 7.848356736624894e-05, "loss": 1.0234, "step": 86535 }, { "epoch": 0.22, "learning_rate": 7.848230912476723e-05, "loss": 1.0254, "step": 86540 }, { "epoch": 0.22, "learning_rate": 7.848105088328552e-05, "loss": 1.0232, "step": 86545 }, { "epoch": 0.22, "learning_rate": 7.847979264180381e-05, "loss": 1.0229, "step": 86550 }, { "epoch": 0.22, "learning_rate": 7.84785344003221e-05, "loss": 1.0236, "step": 86555 }, { "epoch": 0.22, "learning_rate": 7.847727615884041e-05, "loss": 1.0239, "step": 86560 }, { "epoch": 0.22, "learning_rate": 7.84760179173587e-05, "loss": 1.0245, "step": 86565 }, { "epoch": 0.22, "learning_rate": 7.8474759675877e-05, "loss": 1.0255, "step": 86570 }, { "epoch": 0.22, "learning_rate": 7.847350143439529e-05, "loss": 1.0236, "step": 86575 }, { "epoch": 0.22, "learning_rate": 7.847224319291359e-05, "loss": 1.0429, "step": 86580 }, { "epoch": 0.22, "learning_rate": 7.847098495143188e-05, "loss": 1.0265, "step": 86585 }, { "epoch": 0.22, "learning_rate": 7.846972670995017e-05, "loss": 1.0433, "step": 86590 }, { "epoch": 0.22, "learning_rate": 7.846846846846847e-05, "loss": 1.0229, "step": 86595 }, { "epoch": 0.22, "learning_rate": 7.846721022698676e-05, "loss": 1.0248, "step": 86600 }, { "epoch": 0.22, "learning_rate": 7.846595198550506e-05, "loss": 1.0251, "step": 86605 }, { "epoch": 0.22, "learning_rate": 7.846469374402335e-05, "loss": 1.0231, "step": 86610 }, { "epoch": 0.22, "learning_rate": 7.846343550254165e-05, "loss": 1.0248, "step": 86615 }, { "epoch": 0.22, "learning_rate": 7.846217726105994e-05, "loss": 1.0212, "step": 86620 }, { "epoch": 0.22, "learning_rate": 7.846091901957824e-05, "loss": 1.0226, "step": 86625 }, { "epoch": 0.22, "learning_rate": 7.845966077809653e-05, "loss": 1.0234, "step": 86630 }, { "epoch": 0.22, "learning_rate": 7.845840253661483e-05, "loss": 1.0228, "step": 86635 }, { "epoch": 0.22, "learning_rate": 7.845714429513312e-05, "loss": 1.0226, "step": 86640 }, { "epoch": 0.22, "learning_rate": 7.845588605365142e-05, "loss": 1.0255, "step": 86645 }, { "epoch": 0.22, "learning_rate": 7.845462781216971e-05, "loss": 1.0234, "step": 86650 }, { "epoch": 0.22, "learning_rate": 7.8453369570688e-05, "loss": 1.0249, "step": 86655 }, { "epoch": 0.22, "learning_rate": 7.84521113292063e-05, "loss": 1.0241, "step": 86660 }, { "epoch": 0.22, "learning_rate": 7.845085308772459e-05, "loss": 1.0211, "step": 86665 }, { "epoch": 0.22, "learning_rate": 7.84495948462429e-05, "loss": 1.0236, "step": 86670 }, { "epoch": 0.22, "learning_rate": 7.844833660476119e-05, "loss": 1.0226, "step": 86675 }, { "epoch": 0.22, "learning_rate": 7.844707836327948e-05, "loss": 1.0245, "step": 86680 }, { "epoch": 0.22, "learning_rate": 7.844582012179777e-05, "loss": 1.0357, "step": 86685 }, { "epoch": 0.22, "learning_rate": 7.844456188031607e-05, "loss": 1.0262, "step": 86690 }, { "epoch": 0.22, "learning_rate": 7.844330363883437e-05, "loss": 1.0248, "step": 86695 }, { "epoch": 0.22, "learning_rate": 7.844204539735266e-05, "loss": 1.0245, "step": 86700 }, { "epoch": 0.22, "learning_rate": 7.844078715587095e-05, "loss": 1.0254, "step": 86705 }, { "epoch": 0.22, "learning_rate": 7.843952891438925e-05, "loss": 1.0229, "step": 86710 }, { "epoch": 0.22, "learning_rate": 7.843827067290755e-05, "loss": 1.0253, "step": 86715 }, { "epoch": 0.22, "learning_rate": 7.843701243142584e-05, "loss": 1.0246, "step": 86720 }, { "epoch": 0.22, "learning_rate": 7.843575418994413e-05, "loss": 1.0232, "step": 86725 }, { "epoch": 0.22, "learning_rate": 7.843449594846242e-05, "loss": 1.0442, "step": 86730 }, { "epoch": 0.22, "learning_rate": 7.843323770698073e-05, "loss": 1.0238, "step": 86735 }, { "epoch": 0.22, "learning_rate": 7.843197946549902e-05, "loss": 1.0259, "step": 86740 }, { "epoch": 0.22, "learning_rate": 7.843072122401731e-05, "loss": 1.0247, "step": 86745 }, { "epoch": 0.22, "learning_rate": 7.842946298253561e-05, "loss": 1.0236, "step": 86750 }, { "epoch": 0.22, "learning_rate": 7.84282047410539e-05, "loss": 1.0224, "step": 86755 }, { "epoch": 0.22, "learning_rate": 7.842694649957221e-05, "loss": 1.0218, "step": 86760 }, { "epoch": 0.22, "learning_rate": 7.84256882580905e-05, "loss": 1.024, "step": 86765 }, { "epoch": 0.22, "learning_rate": 7.84244300166088e-05, "loss": 1.0243, "step": 86770 }, { "epoch": 0.22, "learning_rate": 7.842317177512709e-05, "loss": 1.0225, "step": 86775 }, { "epoch": 0.22, "learning_rate": 7.842191353364539e-05, "loss": 1.0249, "step": 86780 }, { "epoch": 0.22, "learning_rate": 7.842065529216368e-05, "loss": 1.0247, "step": 86785 }, { "epoch": 0.22, "learning_rate": 7.841939705068197e-05, "loss": 1.024, "step": 86790 }, { "epoch": 0.22, "learning_rate": 7.841813880920027e-05, "loss": 1.027, "step": 86795 }, { "epoch": 0.22, "learning_rate": 7.841688056771857e-05, "loss": 1.0209, "step": 86800 }, { "epoch": 0.22, "learning_rate": 7.841562232623686e-05, "loss": 1.0267, "step": 86805 }, { "epoch": 0.22, "learning_rate": 7.841436408475515e-05, "loss": 1.0247, "step": 86810 }, { "epoch": 0.22, "learning_rate": 7.841310584327345e-05, "loss": 1.0239, "step": 86815 }, { "epoch": 0.22, "learning_rate": 7.841184760179174e-05, "loss": 1.0217, "step": 86820 }, { "epoch": 0.22, "learning_rate": 7.841058936031004e-05, "loss": 1.0441, "step": 86825 }, { "epoch": 0.22, "learning_rate": 7.840933111882833e-05, "loss": 1.0238, "step": 86830 }, { "epoch": 0.22, "learning_rate": 7.840807287734663e-05, "loss": 1.0255, "step": 86835 }, { "epoch": 0.22, "learning_rate": 7.840681463586492e-05, "loss": 1.0256, "step": 86840 }, { "epoch": 0.22, "learning_rate": 7.840555639438322e-05, "loss": 1.0252, "step": 86845 }, { "epoch": 0.22, "learning_rate": 7.840429815290151e-05, "loss": 1.0233, "step": 86850 }, { "epoch": 0.22, "learning_rate": 7.84030399114198e-05, "loss": 1.0252, "step": 86855 }, { "epoch": 0.22, "learning_rate": 7.84017816699381e-05, "loss": 1.0245, "step": 86860 }, { "epoch": 0.22, "learning_rate": 7.84005234284564e-05, "loss": 1.0256, "step": 86865 }, { "epoch": 0.22, "learning_rate": 7.83992651869747e-05, "loss": 1.0231, "step": 86870 }, { "epoch": 0.22, "learning_rate": 7.839800694549299e-05, "loss": 1.0238, "step": 86875 }, { "epoch": 0.22, "learning_rate": 7.839674870401128e-05, "loss": 1.0232, "step": 86880 }, { "epoch": 0.22, "learning_rate": 7.839549046252957e-05, "loss": 1.0236, "step": 86885 }, { "epoch": 0.22, "learning_rate": 7.839423222104787e-05, "loss": 1.0261, "step": 86890 }, { "epoch": 0.22, "learning_rate": 7.839297397956617e-05, "loss": 1.0223, "step": 86895 }, { "epoch": 0.22, "learning_rate": 7.839171573808446e-05, "loss": 1.0228, "step": 86900 }, { "epoch": 0.22, "learning_rate": 7.839045749660275e-05, "loss": 1.0226, "step": 86905 }, { "epoch": 0.22, "learning_rate": 7.838919925512105e-05, "loss": 1.0257, "step": 86910 }, { "epoch": 0.22, "learning_rate": 7.838794101363935e-05, "loss": 1.0246, "step": 86915 }, { "epoch": 0.22, "learning_rate": 7.838668277215764e-05, "loss": 1.0266, "step": 86920 }, { "epoch": 0.22, "learning_rate": 7.838542453067593e-05, "loss": 1.0261, "step": 86925 }, { "epoch": 0.22, "learning_rate": 7.838416628919423e-05, "loss": 1.0233, "step": 86930 }, { "epoch": 0.22, "learning_rate": 7.838290804771253e-05, "loss": 1.0235, "step": 86935 }, { "epoch": 0.22, "learning_rate": 7.838164980623082e-05, "loss": 1.0253, "step": 86940 }, { "epoch": 0.22, "learning_rate": 7.838039156474911e-05, "loss": 1.0237, "step": 86945 }, { "epoch": 0.22, "learning_rate": 7.83791333232674e-05, "loss": 1.0226, "step": 86950 }, { "epoch": 0.22, "learning_rate": 7.83778750817857e-05, "loss": 1.0243, "step": 86955 }, { "epoch": 0.22, "learning_rate": 7.8376616840304e-05, "loss": 1.0242, "step": 86960 }, { "epoch": 0.22, "learning_rate": 7.837535859882229e-05, "loss": 1.0232, "step": 86965 }, { "epoch": 0.22, "learning_rate": 7.837410035734058e-05, "loss": 1.0217, "step": 86970 }, { "epoch": 0.22, "learning_rate": 7.837284211585889e-05, "loss": 1.0245, "step": 86975 }, { "epoch": 0.22, "learning_rate": 7.837158387437718e-05, "loss": 1.0235, "step": 86980 }, { "epoch": 0.22, "learning_rate": 7.837032563289547e-05, "loss": 1.0251, "step": 86985 }, { "epoch": 0.22, "learning_rate": 7.836906739141376e-05, "loss": 1.0235, "step": 86990 }, { "epoch": 0.22, "learning_rate": 7.836780914993207e-05, "loss": 1.025, "step": 86995 }, { "epoch": 0.22, "learning_rate": 7.836655090845036e-05, "loss": 1.0246, "step": 87000 }, { "epoch": 0.22, "learning_rate": 7.836529266696865e-05, "loss": 1.0252, "step": 87005 }, { "epoch": 0.22, "learning_rate": 7.836403442548694e-05, "loss": 1.0249, "step": 87010 }, { "epoch": 0.22, "learning_rate": 7.836277618400523e-05, "loss": 1.0252, "step": 87015 }, { "epoch": 0.22, "learning_rate": 7.836151794252354e-05, "loss": 1.0238, "step": 87020 }, { "epoch": 0.22, "learning_rate": 7.836025970104183e-05, "loss": 1.0459, "step": 87025 }, { "epoch": 0.22, "learning_rate": 7.835900145956012e-05, "loss": 1.0254, "step": 87030 }, { "epoch": 0.22, "learning_rate": 7.835774321807841e-05, "loss": 1.0241, "step": 87035 }, { "epoch": 0.22, "learning_rate": 7.835648497659672e-05, "loss": 1.0238, "step": 87040 }, { "epoch": 0.22, "learning_rate": 7.835522673511501e-05, "loss": 1.0245, "step": 87045 }, { "epoch": 0.22, "learning_rate": 7.83539684936333e-05, "loss": 1.0257, "step": 87050 }, { "epoch": 0.22, "learning_rate": 7.835271025215159e-05, "loss": 1.0225, "step": 87055 }, { "epoch": 0.22, "learning_rate": 7.83514520106699e-05, "loss": 1.0251, "step": 87060 }, { "epoch": 0.22, "learning_rate": 7.835019376918819e-05, "loss": 1.0215, "step": 87065 }, { "epoch": 0.22, "learning_rate": 7.834893552770648e-05, "loss": 1.0248, "step": 87070 }, { "epoch": 0.22, "learning_rate": 7.834767728622477e-05, "loss": 1.023, "step": 87075 }, { "epoch": 0.22, "learning_rate": 7.834641904474306e-05, "loss": 1.0242, "step": 87080 }, { "epoch": 0.22, "learning_rate": 7.834516080326137e-05, "loss": 1.024, "step": 87085 }, { "epoch": 0.22, "learning_rate": 7.834390256177966e-05, "loss": 1.0241, "step": 87090 }, { "epoch": 0.22, "learning_rate": 7.834264432029795e-05, "loss": 1.0256, "step": 87095 }, { "epoch": 0.22, "learning_rate": 7.834138607881624e-05, "loss": 1.0238, "step": 87100 }, { "epoch": 0.22, "learning_rate": 7.834012783733455e-05, "loss": 1.0254, "step": 87105 }, { "epoch": 0.22, "learning_rate": 7.833886959585284e-05, "loss": 1.0253, "step": 87110 }, { "epoch": 0.22, "learning_rate": 7.833761135437113e-05, "loss": 1.0224, "step": 87115 }, { "epoch": 0.22, "learning_rate": 7.833635311288942e-05, "loss": 1.0243, "step": 87120 }, { "epoch": 0.22, "learning_rate": 7.833509487140773e-05, "loss": 1.0225, "step": 87125 }, { "epoch": 0.22, "learning_rate": 7.833383662992602e-05, "loss": 1.0232, "step": 87130 }, { "epoch": 0.22, "learning_rate": 7.833257838844431e-05, "loss": 1.0237, "step": 87135 }, { "epoch": 0.22, "learning_rate": 7.83313201469626e-05, "loss": 1.0228, "step": 87140 }, { "epoch": 0.22, "learning_rate": 7.83300619054809e-05, "loss": 1.0232, "step": 87145 }, { "epoch": 0.22, "learning_rate": 7.83288036639992e-05, "loss": 1.0251, "step": 87150 }, { "epoch": 0.22, "learning_rate": 7.832754542251749e-05, "loss": 1.0235, "step": 87155 }, { "epoch": 0.22, "learning_rate": 7.832628718103578e-05, "loss": 1.0229, "step": 87160 }, { "epoch": 0.22, "learning_rate": 7.832502893955407e-05, "loss": 1.0234, "step": 87165 }, { "epoch": 0.22, "learning_rate": 7.832377069807238e-05, "loss": 1.024, "step": 87170 }, { "epoch": 0.22, "learning_rate": 7.832251245659067e-05, "loss": 1.0196, "step": 87175 }, { "epoch": 0.22, "learning_rate": 7.832125421510896e-05, "loss": 1.0249, "step": 87180 }, { "epoch": 0.22, "learning_rate": 7.831999597362725e-05, "loss": 1.0228, "step": 87185 }, { "epoch": 0.22, "learning_rate": 7.831873773214556e-05, "loss": 1.0263, "step": 87190 }, { "epoch": 0.22, "learning_rate": 7.831747949066385e-05, "loss": 1.0238, "step": 87195 }, { "epoch": 0.22, "learning_rate": 7.831622124918214e-05, "loss": 1.0251, "step": 87200 }, { "epoch": 0.22, "learning_rate": 7.831496300770043e-05, "loss": 1.0248, "step": 87205 }, { "epoch": 0.22, "learning_rate": 7.831370476621873e-05, "loss": 1.0256, "step": 87210 }, { "epoch": 0.22, "learning_rate": 7.831244652473703e-05, "loss": 1.025, "step": 87215 }, { "epoch": 0.22, "learning_rate": 7.831118828325532e-05, "loss": 1.023, "step": 87220 }, { "epoch": 0.22, "learning_rate": 7.830993004177361e-05, "loss": 1.0244, "step": 87225 }, { "epoch": 0.22, "learning_rate": 7.83086718002919e-05, "loss": 1.0255, "step": 87230 }, { "epoch": 0.22, "learning_rate": 7.830741355881021e-05, "loss": 1.0254, "step": 87235 }, { "epoch": 0.22, "learning_rate": 7.83061553173285e-05, "loss": 1.0234, "step": 87240 }, { "epoch": 0.22, "learning_rate": 7.83048970758468e-05, "loss": 1.0256, "step": 87245 }, { "epoch": 0.22, "learning_rate": 7.83036388343651e-05, "loss": 1.0229, "step": 87250 }, { "epoch": 0.22, "learning_rate": 7.830238059288339e-05, "loss": 1.0248, "step": 87255 }, { "epoch": 0.22, "learning_rate": 7.83011223514017e-05, "loss": 1.0466, "step": 87260 }, { "epoch": 0.22, "learning_rate": 7.829986410991999e-05, "loss": 1.0221, "step": 87265 }, { "epoch": 0.22, "learning_rate": 7.829860586843828e-05, "loss": 1.0232, "step": 87270 }, { "epoch": 0.22, "learning_rate": 7.829734762695657e-05, "loss": 1.0245, "step": 87275 }, { "epoch": 0.22, "learning_rate": 7.829608938547486e-05, "loss": 1.0249, "step": 87280 }, { "epoch": 0.22, "learning_rate": 7.829483114399317e-05, "loss": 1.0246, "step": 87285 }, { "epoch": 0.22, "learning_rate": 7.829357290251146e-05, "loss": 1.0248, "step": 87290 }, { "epoch": 0.22, "learning_rate": 7.829231466102975e-05, "loss": 1.0232, "step": 87295 }, { "epoch": 0.22, "learning_rate": 7.829105641954804e-05, "loss": 1.0228, "step": 87300 }, { "epoch": 0.22, "learning_rate": 7.828979817806635e-05, "loss": 1.0268, "step": 87305 }, { "epoch": 0.22, "learning_rate": 7.828853993658464e-05, "loss": 1.0232, "step": 87310 }, { "epoch": 0.22, "learning_rate": 7.828728169510293e-05, "loss": 1.0251, "step": 87315 }, { "epoch": 0.22, "learning_rate": 7.828602345362122e-05, "loss": 1.0238, "step": 87320 }, { "epoch": 0.22, "learning_rate": 7.828476521213953e-05, "loss": 1.0253, "step": 87325 }, { "epoch": 0.22, "learning_rate": 7.828350697065782e-05, "loss": 1.0237, "step": 87330 }, { "epoch": 0.22, "learning_rate": 7.828224872917611e-05, "loss": 1.0241, "step": 87335 }, { "epoch": 0.22, "learning_rate": 7.82809904876944e-05, "loss": 1.027, "step": 87340 }, { "epoch": 0.22, "learning_rate": 7.82797322462127e-05, "loss": 1.0216, "step": 87345 }, { "epoch": 0.22, "learning_rate": 7.8278474004731e-05, "loss": 1.0235, "step": 87350 }, { "epoch": 0.22, "learning_rate": 7.827721576324929e-05, "loss": 1.0254, "step": 87355 }, { "epoch": 0.22, "learning_rate": 7.827595752176758e-05, "loss": 1.0241, "step": 87360 }, { "epoch": 0.22, "learning_rate": 7.827469928028587e-05, "loss": 1.0269, "step": 87365 }, { "epoch": 0.22, "learning_rate": 7.827344103880418e-05, "loss": 1.0229, "step": 87370 }, { "epoch": 0.22, "learning_rate": 7.827218279732247e-05, "loss": 1.026, "step": 87375 }, { "epoch": 0.22, "learning_rate": 7.827092455584076e-05, "loss": 1.0235, "step": 87380 }, { "epoch": 0.22, "learning_rate": 7.826966631435905e-05, "loss": 1.0264, "step": 87385 }, { "epoch": 0.22, "learning_rate": 7.826840807287736e-05, "loss": 1.0237, "step": 87390 }, { "epoch": 0.22, "learning_rate": 7.826714983139565e-05, "loss": 1.0254, "step": 87395 }, { "epoch": 0.22, "learning_rate": 7.826589158991394e-05, "loss": 1.0255, "step": 87400 }, { "epoch": 0.22, "learning_rate": 7.826463334843223e-05, "loss": 1.0258, "step": 87405 }, { "epoch": 0.22, "learning_rate": 7.826337510695053e-05, "loss": 1.0213, "step": 87410 }, { "epoch": 0.22, "learning_rate": 7.826211686546883e-05, "loss": 1.0245, "step": 87415 }, { "epoch": 0.22, "learning_rate": 7.826085862398712e-05, "loss": 1.0244, "step": 87420 }, { "epoch": 0.22, "learning_rate": 7.825960038250541e-05, "loss": 1.0249, "step": 87425 }, { "epoch": 0.22, "learning_rate": 7.82583421410237e-05, "loss": 1.0219, "step": 87430 }, { "epoch": 0.22, "learning_rate": 7.825708389954201e-05, "loss": 1.025, "step": 87435 }, { "epoch": 0.22, "learning_rate": 7.82558256580603e-05, "loss": 1.0231, "step": 87440 }, { "epoch": 0.22, "learning_rate": 7.82545674165786e-05, "loss": 1.0228, "step": 87445 }, { "epoch": 0.22, "learning_rate": 7.825330917509689e-05, "loss": 1.0265, "step": 87450 }, { "epoch": 0.22, "learning_rate": 7.825205093361519e-05, "loss": 1.0203, "step": 87455 }, { "epoch": 0.22, "learning_rate": 7.825079269213348e-05, "loss": 1.0497, "step": 87460 }, { "epoch": 0.22, "learning_rate": 7.824953445065177e-05, "loss": 1.0231, "step": 87465 }, { "epoch": 0.22, "learning_rate": 7.824827620917007e-05, "loss": 1.0233, "step": 87470 }, { "epoch": 0.22, "learning_rate": 7.824726961598471e-05, "loss": 1.0438, "step": 87475 }, { "epoch": 0.22, "learning_rate": 7.8246011374503e-05, "loss": 1.0242, "step": 87480 }, { "epoch": 0.22, "learning_rate": 7.82447531330213e-05, "loss": 1.0246, "step": 87485 }, { "epoch": 0.22, "learning_rate": 7.824349489153959e-05, "loss": 1.0228, "step": 87490 }, { "epoch": 0.22, "learning_rate": 7.824223665005788e-05, "loss": 1.0241, "step": 87495 }, { "epoch": 0.22, "learning_rate": 7.824097840857619e-05, "loss": 1.0468, "step": 87500 }, { "epoch": 0.22, "learning_rate": 7.823972016709448e-05, "loss": 1.0219, "step": 87505 }, { "epoch": 0.22, "learning_rate": 7.823846192561277e-05, "loss": 1.0255, "step": 87510 }, { "epoch": 0.22, "learning_rate": 7.823720368413106e-05, "loss": 1.0254, "step": 87515 }, { "epoch": 0.22, "learning_rate": 7.823594544264937e-05, "loss": 1.0236, "step": 87520 }, { "epoch": 0.22, "learning_rate": 7.823468720116766e-05, "loss": 1.0243, "step": 87525 }, { "epoch": 0.22, "learning_rate": 7.823342895968595e-05, "loss": 1.0222, "step": 87530 }, { "epoch": 0.22, "learning_rate": 7.823217071820424e-05, "loss": 1.0239, "step": 87535 }, { "epoch": 0.22, "learning_rate": 7.823091247672255e-05, "loss": 1.0228, "step": 87540 }, { "epoch": 0.22, "learning_rate": 7.822965423524084e-05, "loss": 1.0243, "step": 87545 }, { "epoch": 0.22, "learning_rate": 7.822839599375913e-05, "loss": 1.0214, "step": 87550 }, { "epoch": 0.22, "learning_rate": 7.822713775227742e-05, "loss": 1.0237, "step": 87555 }, { "epoch": 0.22, "learning_rate": 7.822587951079571e-05, "loss": 1.022, "step": 87560 }, { "epoch": 0.22, "learning_rate": 7.822462126931402e-05, "loss": 1.0243, "step": 87565 }, { "epoch": 0.22, "learning_rate": 7.822336302783231e-05, "loss": 1.0249, "step": 87570 }, { "epoch": 0.22, "learning_rate": 7.82221047863506e-05, "loss": 1.0239, "step": 87575 }, { "epoch": 0.22, "learning_rate": 7.822084654486889e-05, "loss": 1.0239, "step": 87580 }, { "epoch": 0.22, "learning_rate": 7.82195883033872e-05, "loss": 1.025, "step": 87585 }, { "epoch": 0.22, "learning_rate": 7.821833006190549e-05, "loss": 1.0258, "step": 87590 }, { "epoch": 0.22, "learning_rate": 7.821707182042378e-05, "loss": 1.0246, "step": 87595 }, { "epoch": 0.22, "learning_rate": 7.821581357894207e-05, "loss": 1.0249, "step": 87600 }, { "epoch": 0.22, "learning_rate": 7.821455533746038e-05, "loss": 1.0222, "step": 87605 }, { "epoch": 0.22, "learning_rate": 7.821329709597867e-05, "loss": 1.0229, "step": 87610 }, { "epoch": 0.22, "learning_rate": 7.821203885449696e-05, "loss": 1.0245, "step": 87615 }, { "epoch": 0.22, "learning_rate": 7.821078061301525e-05, "loss": 1.025, "step": 87620 }, { "epoch": 0.22, "learning_rate": 7.820952237153354e-05, "loss": 1.0223, "step": 87625 }, { "epoch": 0.22, "learning_rate": 7.820826413005185e-05, "loss": 1.022, "step": 87630 }, { "epoch": 0.22, "learning_rate": 7.820700588857014e-05, "loss": 1.0219, "step": 87635 }, { "epoch": 0.22, "learning_rate": 7.820574764708843e-05, "loss": 1.0219, "step": 87640 }, { "epoch": 0.22, "learning_rate": 7.820448940560672e-05, "loss": 1.0253, "step": 87645 }, { "epoch": 0.22, "learning_rate": 7.820323116412503e-05, "loss": 1.0263, "step": 87650 }, { "epoch": 0.22, "learning_rate": 7.820197292264332e-05, "loss": 1.0257, "step": 87655 }, { "epoch": 0.22, "learning_rate": 7.820071468116161e-05, "loss": 1.0231, "step": 87660 }, { "epoch": 0.22, "learning_rate": 7.81994564396799e-05, "loss": 1.0238, "step": 87665 }, { "epoch": 0.22, "learning_rate": 7.819819819819821e-05, "loss": 1.0246, "step": 87670 }, { "epoch": 0.22, "learning_rate": 7.81969399567165e-05, "loss": 1.0247, "step": 87675 }, { "epoch": 0.22, "learning_rate": 7.819568171523479e-05, "loss": 1.0228, "step": 87680 }, { "epoch": 0.22, "learning_rate": 7.819442347375308e-05, "loss": 1.0228, "step": 87685 }, { "epoch": 0.22, "learning_rate": 7.819316523227138e-05, "loss": 1.0231, "step": 87690 }, { "epoch": 0.22, "learning_rate": 7.819190699078968e-05, "loss": 1.036, "step": 87695 }, { "epoch": 0.22, "learning_rate": 7.819064874930797e-05, "loss": 1.0259, "step": 87700 }, { "epoch": 0.22, "learning_rate": 7.818939050782626e-05, "loss": 1.0244, "step": 87705 }, { "epoch": 0.22, "learning_rate": 7.818813226634456e-05, "loss": 1.0223, "step": 87710 }, { "epoch": 0.22, "learning_rate": 7.818687402486286e-05, "loss": 1.0234, "step": 87715 }, { "epoch": 0.22, "learning_rate": 7.818561578338115e-05, "loss": 1.0245, "step": 87720 }, { "epoch": 0.22, "learning_rate": 7.818435754189944e-05, "loss": 1.0228, "step": 87725 }, { "epoch": 0.22, "learning_rate": 7.818309930041774e-05, "loss": 1.0243, "step": 87730 }, { "epoch": 0.22, "learning_rate": 7.818184105893604e-05, "loss": 1.025, "step": 87735 }, { "epoch": 0.22, "learning_rate": 7.818058281745433e-05, "loss": 1.0237, "step": 87740 }, { "epoch": 0.22, "learning_rate": 7.817932457597262e-05, "loss": 1.0258, "step": 87745 }, { "epoch": 0.22, "learning_rate": 7.817806633449092e-05, "loss": 1.0236, "step": 87750 }, { "epoch": 0.22, "learning_rate": 7.817680809300921e-05, "loss": 1.0241, "step": 87755 }, { "epoch": 0.22, "learning_rate": 7.817554985152751e-05, "loss": 1.0234, "step": 87760 }, { "epoch": 0.22, "learning_rate": 7.81742916100458e-05, "loss": 1.0244, "step": 87765 }, { "epoch": 0.22, "learning_rate": 7.81730333685641e-05, "loss": 1.0216, "step": 87770 }, { "epoch": 0.22, "learning_rate": 7.817177512708239e-05, "loss": 1.0241, "step": 87775 }, { "epoch": 0.22, "learning_rate": 7.817051688560069e-05, "loss": 1.0238, "step": 87780 }, { "epoch": 0.22, "learning_rate": 7.816925864411898e-05, "loss": 1.0253, "step": 87785 }, { "epoch": 0.22, "learning_rate": 7.816800040263728e-05, "loss": 1.0261, "step": 87790 }, { "epoch": 0.22, "learning_rate": 7.816674216115557e-05, "loss": 1.0243, "step": 87795 }, { "epoch": 0.22, "learning_rate": 7.816548391967387e-05, "loss": 1.0404, "step": 87800 }, { "epoch": 0.22, "learning_rate": 7.816422567819216e-05, "loss": 1.0265, "step": 87805 }, { "epoch": 0.22, "learning_rate": 7.816296743671046e-05, "loss": 1.024, "step": 87810 }, { "epoch": 0.22, "learning_rate": 7.816170919522875e-05, "loss": 1.0236, "step": 87815 }, { "epoch": 0.22, "learning_rate": 7.816045095374704e-05, "loss": 1.0249, "step": 87820 }, { "epoch": 0.22, "learning_rate": 7.815919271226534e-05, "loss": 1.0245, "step": 87825 }, { "epoch": 0.22, "learning_rate": 7.815793447078364e-05, "loss": 1.0221, "step": 87830 }, { "epoch": 0.22, "learning_rate": 7.815667622930193e-05, "loss": 1.0256, "step": 87835 }, { "epoch": 0.22, "learning_rate": 7.815541798782022e-05, "loss": 1.0222, "step": 87840 }, { "epoch": 0.22, "learning_rate": 7.815415974633852e-05, "loss": 1.0251, "step": 87845 }, { "epoch": 0.22, "learning_rate": 7.815290150485681e-05, "loss": 1.0242, "step": 87850 }, { "epoch": 0.22, "learning_rate": 7.81516432633751e-05, "loss": 1.0238, "step": 87855 }, { "epoch": 0.22, "learning_rate": 7.81503850218934e-05, "loss": 1.0228, "step": 87860 }, { "epoch": 0.22, "learning_rate": 7.81491267804117e-05, "loss": 1.0225, "step": 87865 }, { "epoch": 0.22, "learning_rate": 7.814786853893e-05, "loss": 1.0248, "step": 87870 }, { "epoch": 0.22, "learning_rate": 7.814661029744829e-05, "loss": 1.0231, "step": 87875 }, { "epoch": 0.22, "learning_rate": 7.814535205596658e-05, "loss": 1.0241, "step": 87880 }, { "epoch": 0.22, "learning_rate": 7.814409381448487e-05, "loss": 1.0253, "step": 87885 }, { "epoch": 0.22, "learning_rate": 7.814283557300317e-05, "loss": 1.0238, "step": 87890 }, { "epoch": 0.22, "learning_rate": 7.814157733152147e-05, "loss": 1.0226, "step": 87895 }, { "epoch": 0.22, "learning_rate": 7.814031909003976e-05, "loss": 1.0229, "step": 87900 }, { "epoch": 0.22, "learning_rate": 7.813906084855805e-05, "loss": 1.0247, "step": 87905 }, { "epoch": 0.22, "learning_rate": 7.813780260707635e-05, "loss": 1.0231, "step": 87910 }, { "epoch": 0.22, "learning_rate": 7.813654436559465e-05, "loss": 1.0242, "step": 87915 }, { "epoch": 0.22, "learning_rate": 7.813528612411294e-05, "loss": 1.0253, "step": 87920 }, { "epoch": 0.22, "learning_rate": 7.813402788263123e-05, "loss": 1.024, "step": 87925 }, { "epoch": 0.22, "learning_rate": 7.813276964114953e-05, "loss": 1.0239, "step": 87930 }, { "epoch": 0.22, "learning_rate": 7.813151139966783e-05, "loss": 1.0252, "step": 87935 }, { "epoch": 0.22, "learning_rate": 7.813025315818612e-05, "loss": 1.0231, "step": 87940 }, { "epoch": 0.22, "learning_rate": 7.812899491670441e-05, "loss": 1.0216, "step": 87945 }, { "epoch": 0.22, "learning_rate": 7.812773667522271e-05, "loss": 1.0235, "step": 87950 }, { "epoch": 0.22, "learning_rate": 7.8126478433741e-05, "loss": 1.0244, "step": 87955 }, { "epoch": 0.22, "learning_rate": 7.812522019225931e-05, "loss": 1.0236, "step": 87960 }, { "epoch": 0.22, "learning_rate": 7.81239619507776e-05, "loss": 1.0253, "step": 87965 }, { "epoch": 0.22, "learning_rate": 7.81227037092959e-05, "loss": 1.0242, "step": 87970 }, { "epoch": 0.22, "learning_rate": 7.812144546781419e-05, "loss": 1.0242, "step": 87975 }, { "epoch": 0.22, "learning_rate": 7.812018722633249e-05, "loss": 1.0248, "step": 87980 }, { "epoch": 0.22, "learning_rate": 7.811892898485078e-05, "loss": 1.0225, "step": 87985 }, { "epoch": 0.22, "learning_rate": 7.811767074336907e-05, "loss": 1.0215, "step": 87990 }, { "epoch": 0.22, "learning_rate": 7.811641250188737e-05, "loss": 1.0241, "step": 87995 }, { "epoch": 0.22, "learning_rate": 7.811515426040567e-05, "loss": 1.0227, "step": 88000 }, { "epoch": 0.22, "learning_rate": 7.811389601892396e-05, "loss": 1.0234, "step": 88005 }, { "epoch": 0.22, "learning_rate": 7.811263777744225e-05, "loss": 1.0234, "step": 88010 }, { "epoch": 0.22, "learning_rate": 7.811137953596055e-05, "loss": 1.0231, "step": 88015 }, { "epoch": 0.22, "learning_rate": 7.811012129447884e-05, "loss": 1.0227, "step": 88020 }, { "epoch": 0.22, "learning_rate": 7.810886305299714e-05, "loss": 1.0236, "step": 88025 }, { "epoch": 0.22, "learning_rate": 7.810760481151543e-05, "loss": 1.0237, "step": 88030 }, { "epoch": 0.22, "learning_rate": 7.810634657003373e-05, "loss": 1.0234, "step": 88035 }, { "epoch": 0.22, "learning_rate": 7.810508832855202e-05, "loss": 1.0245, "step": 88040 }, { "epoch": 0.22, "learning_rate": 7.810383008707032e-05, "loss": 1.026, "step": 88045 }, { "epoch": 0.22, "learning_rate": 7.810257184558861e-05, "loss": 1.0433, "step": 88050 }, { "epoch": 0.22, "learning_rate": 7.81013136041069e-05, "loss": 1.023, "step": 88055 }, { "epoch": 0.22, "learning_rate": 7.81000553626252e-05, "loss": 1.0228, "step": 88060 }, { "epoch": 0.22, "learning_rate": 7.80987971211435e-05, "loss": 1.0241, "step": 88065 }, { "epoch": 0.22, "learning_rate": 7.80975388796618e-05, "loss": 1.0252, "step": 88070 }, { "epoch": 0.22, "learning_rate": 7.809628063818009e-05, "loss": 1.0248, "step": 88075 }, { "epoch": 0.22, "learning_rate": 7.809502239669838e-05, "loss": 1.0236, "step": 88080 }, { "epoch": 0.22, "learning_rate": 7.809376415521667e-05, "loss": 1.0477, "step": 88085 }, { "epoch": 0.22, "learning_rate": 7.809250591373497e-05, "loss": 1.0216, "step": 88090 }, { "epoch": 0.22, "learning_rate": 7.809124767225327e-05, "loss": 1.0225, "step": 88095 }, { "epoch": 0.22, "learning_rate": 7.808998943077156e-05, "loss": 1.0245, "step": 88100 }, { "epoch": 0.22, "learning_rate": 7.808873118928985e-05, "loss": 1.0256, "step": 88105 }, { "epoch": 0.22, "learning_rate": 7.808747294780815e-05, "loss": 1.0252, "step": 88110 }, { "epoch": 0.22, "learning_rate": 7.808621470632645e-05, "loss": 1.0243, "step": 88115 }, { "epoch": 0.22, "learning_rate": 7.808495646484474e-05, "loss": 1.0215, "step": 88120 }, { "epoch": 0.22, "learning_rate": 7.808369822336303e-05, "loss": 1.0248, "step": 88125 }, { "epoch": 0.22, "learning_rate": 7.808243998188133e-05, "loss": 1.0226, "step": 88130 }, { "epoch": 0.22, "learning_rate": 7.808118174039963e-05, "loss": 1.0229, "step": 88135 }, { "epoch": 0.22, "learning_rate": 7.807992349891792e-05, "loss": 1.0484, "step": 88140 }, { "epoch": 0.22, "learning_rate": 7.807866525743621e-05, "loss": 1.0245, "step": 88145 }, { "epoch": 0.22, "learning_rate": 7.80774070159545e-05, "loss": 1.0219, "step": 88150 }, { "epoch": 0.22, "learning_rate": 7.80761487744728e-05, "loss": 1.0234, "step": 88155 }, { "epoch": 0.22, "learning_rate": 7.80748905329911e-05, "loss": 1.0241, "step": 88160 }, { "epoch": 0.22, "learning_rate": 7.807363229150939e-05, "loss": 1.0231, "step": 88165 }, { "epoch": 0.22, "learning_rate": 7.807237405002768e-05, "loss": 1.0247, "step": 88170 }, { "epoch": 0.22, "learning_rate": 7.807111580854599e-05, "loss": 1.0252, "step": 88175 }, { "epoch": 0.22, "learning_rate": 7.806985756706428e-05, "loss": 1.0466, "step": 88180 }, { "epoch": 0.22, "learning_rate": 7.806859932558257e-05, "loss": 1.0216, "step": 88185 }, { "epoch": 0.22, "learning_rate": 7.806734108410086e-05, "loss": 1.0271, "step": 88190 }, { "epoch": 0.22, "learning_rate": 7.806608284261917e-05, "loss": 1.0236, "step": 88195 }, { "epoch": 0.22, "learning_rate": 7.806482460113746e-05, "loss": 1.025, "step": 88200 }, { "epoch": 0.22, "learning_rate": 7.806356635965575e-05, "loss": 1.0239, "step": 88205 }, { "epoch": 0.22, "learning_rate": 7.806230811817404e-05, "loss": 1.0245, "step": 88210 }, { "epoch": 0.22, "learning_rate": 7.806104987669233e-05, "loss": 1.0254, "step": 88215 }, { "epoch": 0.22, "learning_rate": 7.805979163521064e-05, "loss": 1.0229, "step": 88220 }, { "epoch": 0.22, "learning_rate": 7.805853339372893e-05, "loss": 1.0238, "step": 88225 }, { "epoch": 0.22, "learning_rate": 7.805727515224722e-05, "loss": 1.0234, "step": 88230 }, { "epoch": 0.22, "learning_rate": 7.805601691076551e-05, "loss": 1.0242, "step": 88235 }, { "epoch": 0.22, "learning_rate": 7.805475866928382e-05, "loss": 1.0243, "step": 88240 }, { "epoch": 0.22, "learning_rate": 7.805350042780211e-05, "loss": 1.0235, "step": 88245 }, { "epoch": 0.22, "learning_rate": 7.80522421863204e-05, "loss": 1.0236, "step": 88250 }, { "epoch": 0.22, "learning_rate": 7.805098394483869e-05, "loss": 1.0237, "step": 88255 }, { "epoch": 0.22, "learning_rate": 7.8049725703357e-05, "loss": 1.0229, "step": 88260 }, { "epoch": 0.22, "learning_rate": 7.804846746187529e-05, "loss": 1.0221, "step": 88265 }, { "epoch": 0.22, "learning_rate": 7.804720922039358e-05, "loss": 1.0243, "step": 88270 }, { "epoch": 0.22, "learning_rate": 7.804595097891187e-05, "loss": 1.0243, "step": 88275 }, { "epoch": 0.22, "learning_rate": 7.804469273743016e-05, "loss": 1.0235, "step": 88280 }, { "epoch": 0.22, "learning_rate": 7.804343449594847e-05, "loss": 1.0242, "step": 88285 }, { "epoch": 0.22, "learning_rate": 7.804217625446676e-05, "loss": 1.0233, "step": 88290 }, { "epoch": 0.22, "learning_rate": 7.804091801298505e-05, "loss": 1.024, "step": 88295 }, { "epoch": 0.22, "learning_rate": 7.803965977150334e-05, "loss": 1.0247, "step": 88300 }, { "epoch": 0.22, "learning_rate": 7.803840153002165e-05, "loss": 1.0263, "step": 88305 }, { "epoch": 0.22, "learning_rate": 7.803714328853994e-05, "loss": 1.0244, "step": 88310 }, { "epoch": 0.22, "learning_rate": 7.803588504705823e-05, "loss": 1.0392, "step": 88315 }, { "epoch": 0.22, "learning_rate": 7.803462680557652e-05, "loss": 1.0257, "step": 88320 }, { "epoch": 0.22, "learning_rate": 7.803336856409483e-05, "loss": 1.0262, "step": 88325 }, { "epoch": 0.22, "learning_rate": 7.803211032261312e-05, "loss": 1.0257, "step": 88330 }, { "epoch": 0.22, "learning_rate": 7.803085208113141e-05, "loss": 1.0223, "step": 88335 }, { "epoch": 0.22, "learning_rate": 7.80295938396497e-05, "loss": 1.0226, "step": 88340 }, { "epoch": 0.22, "learning_rate": 7.8028335598168e-05, "loss": 1.0229, "step": 88345 }, { "epoch": 0.22, "learning_rate": 7.80270773566863e-05, "loss": 1.0477, "step": 88350 }, { "epoch": 0.22, "learning_rate": 7.802581911520459e-05, "loss": 1.0261, "step": 88355 }, { "epoch": 0.22, "learning_rate": 7.802456087372288e-05, "loss": 1.0254, "step": 88360 }, { "epoch": 0.22, "learning_rate": 7.802330263224117e-05, "loss": 1.0348, "step": 88365 }, { "epoch": 0.22, "learning_rate": 7.802204439075948e-05, "loss": 1.0247, "step": 88370 }, { "epoch": 0.22, "learning_rate": 7.802078614927777e-05, "loss": 1.0262, "step": 88375 }, { "epoch": 0.22, "learning_rate": 7.801952790779606e-05, "loss": 1.0239, "step": 88380 }, { "epoch": 0.22, "learning_rate": 7.801826966631435e-05, "loss": 1.0232, "step": 88385 }, { "epoch": 0.22, "learning_rate": 7.801701142483266e-05, "loss": 1.0241, "step": 88390 }, { "epoch": 0.22, "learning_rate": 7.801575318335095e-05, "loss": 1.0257, "step": 88395 }, { "epoch": 0.22, "learning_rate": 7.801449494186924e-05, "loss": 1.024, "step": 88400 }, { "epoch": 0.22, "learning_rate": 7.801323670038753e-05, "loss": 1.0234, "step": 88405 }, { "epoch": 0.22, "learning_rate": 7.801197845890583e-05, "loss": 1.0234, "step": 88410 }, { "epoch": 0.22, "learning_rate": 7.801072021742413e-05, "loss": 1.0224, "step": 88415 }, { "epoch": 0.22, "learning_rate": 7.800946197594242e-05, "loss": 1.0255, "step": 88420 }, { "epoch": 0.22, "learning_rate": 7.800820373446071e-05, "loss": 1.0232, "step": 88425 }, { "epoch": 0.22, "learning_rate": 7.8006945492979e-05, "loss": 1.0269, "step": 88430 }, { "epoch": 0.22, "learning_rate": 7.800568725149731e-05, "loss": 1.025, "step": 88435 }, { "epoch": 0.22, "learning_rate": 7.80044290100156e-05, "loss": 1.025, "step": 88440 }, { "epoch": 0.22, "learning_rate": 7.80031707685339e-05, "loss": 1.0247, "step": 88445 }, { "epoch": 0.22, "learning_rate": 7.80019125270522e-05, "loss": 1.0237, "step": 88450 }, { "epoch": 0.22, "learning_rate": 7.800065428557049e-05, "loss": 1.0233, "step": 88455 }, { "epoch": 0.22, "learning_rate": 7.79993960440888e-05, "loss": 1.0252, "step": 88460 }, { "epoch": 0.22, "learning_rate": 7.799813780260709e-05, "loss": 1.0252, "step": 88465 }, { "epoch": 0.22, "learning_rate": 7.799687956112538e-05, "loss": 1.0259, "step": 88470 }, { "epoch": 0.22, "learning_rate": 7.799562131964367e-05, "loss": 1.0234, "step": 88475 }, { "epoch": 0.22, "learning_rate": 7.799436307816196e-05, "loss": 1.0255, "step": 88480 }, { "epoch": 0.22, "learning_rate": 7.799310483668027e-05, "loss": 1.0226, "step": 88485 }, { "epoch": 0.22, "learning_rate": 7.799184659519856e-05, "loss": 1.024, "step": 88490 }, { "epoch": 0.22, "learning_rate": 7.799058835371685e-05, "loss": 1.0235, "step": 88495 }, { "epoch": 0.22, "learning_rate": 7.798933011223514e-05, "loss": 1.0257, "step": 88500 }, { "epoch": 0.22, "learning_rate": 7.798807187075345e-05, "loss": 1.0228, "step": 88505 }, { "epoch": 0.22, "learning_rate": 7.798681362927174e-05, "loss": 1.0214, "step": 88510 }, { "epoch": 0.22, "learning_rate": 7.798555538779003e-05, "loss": 1.0264, "step": 88515 }, { "epoch": 0.22, "learning_rate": 7.798429714630832e-05, "loss": 1.0244, "step": 88520 }, { "epoch": 0.22, "learning_rate": 7.798303890482663e-05, "loss": 1.026, "step": 88525 }, { "epoch": 0.22, "learning_rate": 7.798178066334492e-05, "loss": 1.0269, "step": 88530 }, { "epoch": 0.22, "learning_rate": 7.798052242186321e-05, "loss": 1.0221, "step": 88535 }, { "epoch": 0.22, "learning_rate": 7.79792641803815e-05, "loss": 1.0248, "step": 88540 }, { "epoch": 0.22, "learning_rate": 7.79780059388998e-05, "loss": 1.0253, "step": 88545 }, { "epoch": 0.22, "learning_rate": 7.79767476974181e-05, "loss": 1.0248, "step": 88550 }, { "epoch": 0.22, "learning_rate": 7.797548945593639e-05, "loss": 1.0233, "step": 88555 }, { "epoch": 0.22, "learning_rate": 7.797423121445468e-05, "loss": 1.0236, "step": 88560 }, { "epoch": 0.22, "learning_rate": 7.797297297297297e-05, "loss": 1.0454, "step": 88565 }, { "epoch": 0.22, "learning_rate": 7.797171473149128e-05, "loss": 1.0482, "step": 88570 }, { "epoch": 0.22, "learning_rate": 7.797045649000957e-05, "loss": 1.0487, "step": 88575 }, { "epoch": 0.22, "learning_rate": 7.796919824852786e-05, "loss": 1.025, "step": 88580 }, { "epoch": 0.22, "learning_rate": 7.796794000704615e-05, "loss": 1.0242, "step": 88585 }, { "epoch": 0.22, "learning_rate": 7.796668176556446e-05, "loss": 1.0251, "step": 88590 }, { "epoch": 0.22, "learning_rate": 7.796542352408275e-05, "loss": 1.0241, "step": 88595 }, { "epoch": 0.22, "learning_rate": 7.796416528260104e-05, "loss": 1.0228, "step": 88600 }, { "epoch": 0.22, "learning_rate": 7.796290704111933e-05, "loss": 1.0236, "step": 88605 }, { "epoch": 0.22, "learning_rate": 7.796164879963763e-05, "loss": 1.0263, "step": 88610 }, { "epoch": 0.22, "learning_rate": 7.796039055815593e-05, "loss": 1.0238, "step": 88615 }, { "epoch": 0.22, "learning_rate": 7.795913231667422e-05, "loss": 1.0242, "step": 88620 }, { "epoch": 0.22, "learning_rate": 7.795787407519251e-05, "loss": 1.0212, "step": 88625 }, { "epoch": 0.22, "learning_rate": 7.79566158337108e-05, "loss": 1.0241, "step": 88630 }, { "epoch": 0.22, "learning_rate": 7.795535759222911e-05, "loss": 1.0257, "step": 88635 }, { "epoch": 0.22, "learning_rate": 7.79540993507474e-05, "loss": 1.0247, "step": 88640 }, { "epoch": 0.22, "learning_rate": 7.79528411092657e-05, "loss": 1.0251, "step": 88645 }, { "epoch": 0.22, "learning_rate": 7.795158286778399e-05, "loss": 1.0452, "step": 88650 }, { "epoch": 0.22, "learning_rate": 7.795032462630229e-05, "loss": 1.023, "step": 88655 }, { "epoch": 0.22, "learning_rate": 7.794906638482058e-05, "loss": 1.0247, "step": 88660 }, { "epoch": 0.22, "learning_rate": 7.794780814333887e-05, "loss": 1.0251, "step": 88665 }, { "epoch": 0.22, "learning_rate": 7.794654990185717e-05, "loss": 1.0218, "step": 88670 }, { "epoch": 0.22, "learning_rate": 7.794529166037546e-05, "loss": 1.0253, "step": 88675 }, { "epoch": 0.22, "learning_rate": 7.794403341889376e-05, "loss": 1.0243, "step": 88680 }, { "epoch": 0.22, "learning_rate": 7.794277517741205e-05, "loss": 1.0254, "step": 88685 }, { "epoch": 0.22, "learning_rate": 7.794151693593035e-05, "loss": 1.0244, "step": 88690 }, { "epoch": 0.22, "learning_rate": 7.794025869444864e-05, "loss": 1.0249, "step": 88695 }, { "epoch": 0.22, "learning_rate": 7.793900045296694e-05, "loss": 1.0252, "step": 88700 }, { "epoch": 0.22, "learning_rate": 7.793774221148523e-05, "loss": 1.0263, "step": 88705 }, { "epoch": 0.22, "learning_rate": 7.793648397000353e-05, "loss": 1.0249, "step": 88710 }, { "epoch": 0.22, "learning_rate": 7.793522572852182e-05, "loss": 1.024, "step": 88715 }, { "epoch": 0.22, "learning_rate": 7.793396748704012e-05, "loss": 1.0238, "step": 88720 }, { "epoch": 0.22, "learning_rate": 7.793270924555841e-05, "loss": 1.0238, "step": 88725 }, { "epoch": 0.22, "learning_rate": 7.79314510040767e-05, "loss": 1.0251, "step": 88730 }, { "epoch": 0.22, "learning_rate": 7.7930192762595e-05, "loss": 1.0239, "step": 88735 }, { "epoch": 0.22, "learning_rate": 7.792893452111329e-05, "loss": 1.0254, "step": 88740 }, { "epoch": 0.22, "learning_rate": 7.79276762796316e-05, "loss": 1.0211, "step": 88745 }, { "epoch": 0.22, "learning_rate": 7.792641803814989e-05, "loss": 1.0226, "step": 88750 }, { "epoch": 0.22, "learning_rate": 7.792515979666818e-05, "loss": 1.0267, "step": 88755 }, { "epoch": 0.22, "learning_rate": 7.792390155518647e-05, "loss": 1.0244, "step": 88760 }, { "epoch": 0.22, "learning_rate": 7.792264331370477e-05, "loss": 1.0252, "step": 88765 }, { "epoch": 0.22, "learning_rate": 7.792138507222307e-05, "loss": 1.0231, "step": 88770 }, { "epoch": 0.22, "learning_rate": 7.792012683074136e-05, "loss": 1.0266, "step": 88775 }, { "epoch": 0.22, "learning_rate": 7.791886858925965e-05, "loss": 1.0249, "step": 88780 }, { "epoch": 0.22, "learning_rate": 7.791761034777795e-05, "loss": 1.0236, "step": 88785 }, { "epoch": 0.22, "learning_rate": 7.791635210629625e-05, "loss": 1.0239, "step": 88790 }, { "epoch": 0.22, "learning_rate": 7.791509386481454e-05, "loss": 1.0263, "step": 88795 }, { "epoch": 0.22, "learning_rate": 7.791383562333283e-05, "loss": 1.0253, "step": 88800 }, { "epoch": 0.22, "learning_rate": 7.791257738185112e-05, "loss": 1.0243, "step": 88805 }, { "epoch": 0.22, "learning_rate": 7.791131914036943e-05, "loss": 1.0253, "step": 88810 }, { "epoch": 0.22, "learning_rate": 7.791006089888772e-05, "loss": 1.0249, "step": 88815 }, { "epoch": 0.22, "learning_rate": 7.790880265740601e-05, "loss": 1.0265, "step": 88820 }, { "epoch": 0.22, "learning_rate": 7.79075444159243e-05, "loss": 1.0233, "step": 88825 }, { "epoch": 0.22, "learning_rate": 7.79062861744426e-05, "loss": 1.0238, "step": 88830 }, { "epoch": 0.22, "learning_rate": 7.79050279329609e-05, "loss": 1.0255, "step": 88835 }, { "epoch": 0.22, "learning_rate": 7.790376969147919e-05, "loss": 1.0244, "step": 88840 }, { "epoch": 0.22, "learning_rate": 7.790251144999748e-05, "loss": 1.0228, "step": 88845 }, { "epoch": 0.22, "learning_rate": 7.790125320851578e-05, "loss": 1.023, "step": 88850 }, { "epoch": 0.22, "learning_rate": 7.789999496703408e-05, "loss": 1.0239, "step": 88855 }, { "epoch": 0.22, "learning_rate": 7.789873672555237e-05, "loss": 1.0212, "step": 88860 }, { "epoch": 0.22, "learning_rate": 7.789747848407066e-05, "loss": 1.0251, "step": 88865 }, { "epoch": 0.22, "learning_rate": 7.789622024258895e-05, "loss": 1.0238, "step": 88870 }, { "epoch": 0.22, "learning_rate": 7.789496200110726e-05, "loss": 1.0251, "step": 88875 }, { "epoch": 0.22, "learning_rate": 7.789370375962555e-05, "loss": 1.0222, "step": 88880 }, { "epoch": 0.22, "learning_rate": 7.789244551814384e-05, "loss": 1.0271, "step": 88885 }, { "epoch": 0.22, "learning_rate": 7.789118727666213e-05, "loss": 1.0248, "step": 88890 }, { "epoch": 0.22, "learning_rate": 7.788992903518044e-05, "loss": 1.0225, "step": 88895 }, { "epoch": 0.22, "learning_rate": 7.788867079369873e-05, "loss": 1.0249, "step": 88900 }, { "epoch": 0.22, "learning_rate": 7.788741255221702e-05, "loss": 1.025, "step": 88905 }, { "epoch": 0.22, "learning_rate": 7.788615431073531e-05, "loss": 1.027, "step": 88910 }, { "epoch": 0.22, "learning_rate": 7.788489606925362e-05, "loss": 1.0237, "step": 88915 }, { "epoch": 0.22, "learning_rate": 7.788363782777191e-05, "loss": 1.022, "step": 88920 }, { "epoch": 0.22, "learning_rate": 7.78823795862902e-05, "loss": 1.0233, "step": 88925 }, { "epoch": 0.22, "learning_rate": 7.788112134480849e-05, "loss": 1.0241, "step": 88930 }, { "epoch": 0.22, "learning_rate": 7.787986310332678e-05, "loss": 1.0255, "step": 88935 }, { "epoch": 0.22, "learning_rate": 7.787860486184509e-05, "loss": 1.0258, "step": 88940 }, { "epoch": 0.22, "learning_rate": 7.787734662036338e-05, "loss": 1.0228, "step": 88945 }, { "epoch": 0.22, "learning_rate": 7.787608837888168e-05, "loss": 1.0253, "step": 88950 }, { "epoch": 0.22, "learning_rate": 7.787483013739998e-05, "loss": 1.0246, "step": 88955 }, { "epoch": 0.22, "learning_rate": 7.787357189591827e-05, "loss": 1.0241, "step": 88960 }, { "epoch": 0.22, "learning_rate": 7.787231365443657e-05, "loss": 1.0234, "step": 88965 }, { "epoch": 0.22, "learning_rate": 7.787105541295486e-05, "loss": 1.0242, "step": 88970 }, { "epoch": 0.22, "learning_rate": 7.786979717147316e-05, "loss": 1.0431, "step": 88975 }, { "epoch": 0.22, "learning_rate": 7.786853892999145e-05, "loss": 1.0249, "step": 88980 }, { "epoch": 0.22, "learning_rate": 7.786728068850975e-05, "loss": 1.023, "step": 88985 }, { "epoch": 0.22, "learning_rate": 7.786602244702804e-05, "loss": 1.0408, "step": 88990 }, { "epoch": 0.22, "learning_rate": 7.786476420554634e-05, "loss": 1.0258, "step": 88995 }, { "epoch": 0.22, "learning_rate": 7.786350596406463e-05, "loss": 1.0243, "step": 89000 }, { "epoch": 0.22, "learning_rate": 7.786224772258292e-05, "loss": 1.0238, "step": 89005 }, { "epoch": 0.22, "learning_rate": 7.786098948110122e-05, "loss": 1.0256, "step": 89010 }, { "epoch": 0.22, "learning_rate": 7.785973123961952e-05, "loss": 1.0242, "step": 89015 }, { "epoch": 0.22, "learning_rate": 7.785847299813781e-05, "loss": 1.0228, "step": 89020 }, { "epoch": 0.22, "learning_rate": 7.78572147566561e-05, "loss": 1.0205, "step": 89025 }, { "epoch": 0.22, "learning_rate": 7.78559565151744e-05, "loss": 1.0236, "step": 89030 }, { "epoch": 0.22, "learning_rate": 7.78546982736927e-05, "loss": 1.0251, "step": 89035 }, { "epoch": 0.22, "learning_rate": 7.785344003221099e-05, "loss": 1.0464, "step": 89040 }, { "epoch": 0.22, "learning_rate": 7.785218179072928e-05, "loss": 1.0246, "step": 89045 }, { "epoch": 0.22, "learning_rate": 7.785092354924758e-05, "loss": 1.0239, "step": 89050 }, { "epoch": 0.22, "learning_rate": 7.784966530776588e-05, "loss": 1.0236, "step": 89055 }, { "epoch": 0.22, "learning_rate": 7.784840706628417e-05, "loss": 1.0233, "step": 89060 }, { "epoch": 0.22, "learning_rate": 7.784714882480246e-05, "loss": 1.025, "step": 89065 }, { "epoch": 0.22, "learning_rate": 7.784589058332075e-05, "loss": 1.0244, "step": 89070 }, { "epoch": 0.22, "learning_rate": 7.784463234183906e-05, "loss": 1.023, "step": 89075 }, { "epoch": 0.22, "learning_rate": 7.784337410035735e-05, "loss": 1.0227, "step": 89080 }, { "epoch": 0.22, "learning_rate": 7.784211585887564e-05, "loss": 1.0239, "step": 89085 }, { "epoch": 0.22, "learning_rate": 7.784085761739393e-05, "loss": 1.0231, "step": 89090 }, { "epoch": 0.22, "learning_rate": 7.783959937591224e-05, "loss": 1.0228, "step": 89095 }, { "epoch": 0.22, "learning_rate": 7.783834113443053e-05, "loss": 1.0236, "step": 89100 }, { "epoch": 0.22, "learning_rate": 7.783708289294882e-05, "loss": 1.0226, "step": 89105 }, { "epoch": 0.22, "learning_rate": 7.783582465146711e-05, "loss": 1.0215, "step": 89110 }, { "epoch": 0.22, "learning_rate": 7.783456640998542e-05, "loss": 1.0235, "step": 89115 }, { "epoch": 0.22, "learning_rate": 7.783330816850371e-05, "loss": 1.0229, "step": 89120 }, { "epoch": 0.22, "learning_rate": 7.7832049927022e-05, "loss": 1.0245, "step": 89125 }, { "epoch": 0.22, "learning_rate": 7.783079168554029e-05, "loss": 1.0241, "step": 89130 }, { "epoch": 0.22, "learning_rate": 7.782953344405858e-05, "loss": 1.025, "step": 89135 }, { "epoch": 0.22, "learning_rate": 7.782827520257689e-05, "loss": 1.0235, "step": 89140 }, { "epoch": 0.22, "learning_rate": 7.782701696109518e-05, "loss": 1.0233, "step": 89145 }, { "epoch": 0.22, "learning_rate": 7.782575871961347e-05, "loss": 1.0232, "step": 89150 }, { "epoch": 0.22, "learning_rate": 7.782450047813176e-05, "loss": 1.0258, "step": 89155 }, { "epoch": 0.22, "learning_rate": 7.782324223665007e-05, "loss": 1.0235, "step": 89160 }, { "epoch": 0.22, "learning_rate": 7.782198399516836e-05, "loss": 1.0251, "step": 89165 }, { "epoch": 0.22, "learning_rate": 7.782072575368665e-05, "loss": 1.026, "step": 89170 }, { "epoch": 0.22, "learning_rate": 7.781946751220494e-05, "loss": 1.0222, "step": 89175 }, { "epoch": 0.22, "learning_rate": 7.781820927072325e-05, "loss": 1.0226, "step": 89180 }, { "epoch": 0.22, "learning_rate": 7.781695102924154e-05, "loss": 1.0239, "step": 89185 }, { "epoch": 0.22, "learning_rate": 7.781569278775983e-05, "loss": 1.0245, "step": 89190 }, { "epoch": 0.22, "learning_rate": 7.781443454627812e-05, "loss": 1.0235, "step": 89195 }, { "epoch": 0.22, "learning_rate": 7.781317630479641e-05, "loss": 1.0245, "step": 89200 }, { "epoch": 0.22, "learning_rate": 7.781191806331472e-05, "loss": 1.024, "step": 89205 }, { "epoch": 0.22, "learning_rate": 7.781065982183301e-05, "loss": 1.0247, "step": 89210 }, { "epoch": 0.22, "learning_rate": 7.78094015803513e-05, "loss": 1.0246, "step": 89215 }, { "epoch": 0.22, "learning_rate": 7.78081433388696e-05, "loss": 1.0232, "step": 89220 }, { "epoch": 0.22, "learning_rate": 7.78068850973879e-05, "loss": 1.0237, "step": 89225 }, { "epoch": 0.22, "learning_rate": 7.780562685590619e-05, "loss": 1.0253, "step": 89230 }, { "epoch": 0.22, "learning_rate": 7.780436861442448e-05, "loss": 1.0239, "step": 89235 }, { "epoch": 0.22, "learning_rate": 7.780311037294277e-05, "loss": 1.0875, "step": 89240 }, { "epoch": 0.22, "learning_rate": 7.780185213146108e-05, "loss": 1.0236, "step": 89245 }, { "epoch": 0.22, "learning_rate": 7.780059388997937e-05, "loss": 1.0234, "step": 89250 }, { "epoch": 0.22, "learning_rate": 7.779933564849766e-05, "loss": 1.0239, "step": 89255 }, { "epoch": 0.22, "learning_rate": 7.779807740701595e-05, "loss": 1.0251, "step": 89260 }, { "epoch": 0.22, "learning_rate": 7.779681916553425e-05, "loss": 1.0221, "step": 89265 }, { "epoch": 0.22, "learning_rate": 7.779556092405255e-05, "loss": 1.0222, "step": 89270 }, { "epoch": 0.22, "learning_rate": 7.779430268257084e-05, "loss": 1.0232, "step": 89275 }, { "epoch": 0.22, "learning_rate": 7.779304444108913e-05, "loss": 1.0249, "step": 89280 }, { "epoch": 0.22, "learning_rate": 7.779178619960743e-05, "loss": 1.0244, "step": 89285 }, { "epoch": 0.22, "learning_rate": 7.779052795812573e-05, "loss": 1.0252, "step": 89290 }, { "epoch": 0.22, "learning_rate": 7.778926971664402e-05, "loss": 1.0241, "step": 89295 }, { "epoch": 0.22, "learning_rate": 7.778801147516231e-05, "loss": 1.0233, "step": 89300 }, { "epoch": 0.22, "learning_rate": 7.77867532336806e-05, "loss": 1.0256, "step": 89305 }, { "epoch": 0.22, "learning_rate": 7.778549499219891e-05, "loss": 1.0202, "step": 89310 }, { "epoch": 0.22, "learning_rate": 7.77842367507172e-05, "loss": 1.0225, "step": 89315 }, { "epoch": 0.22, "learning_rate": 7.77829785092355e-05, "loss": 1.0238, "step": 89320 }, { "epoch": 0.22, "learning_rate": 7.778172026775378e-05, "loss": 1.0259, "step": 89325 }, { "epoch": 0.22, "learning_rate": 7.778046202627208e-05, "loss": 1.0239, "step": 89330 }, { "epoch": 0.22, "learning_rate": 7.777920378479038e-05, "loss": 1.0249, "step": 89335 }, { "epoch": 0.22, "learning_rate": 7.777794554330867e-05, "loss": 1.0267, "step": 89340 }, { "epoch": 0.22, "learning_rate": 7.777668730182696e-05, "loss": 1.023, "step": 89345 }, { "epoch": 0.22, "learning_rate": 7.777542906034526e-05, "loss": 1.0208, "step": 89350 }, { "epoch": 0.22, "learning_rate": 7.777417081886356e-05, "loss": 1.0269, "step": 89355 }, { "epoch": 0.22, "learning_rate": 7.777291257738185e-05, "loss": 1.0249, "step": 89360 }, { "epoch": 0.22, "learning_rate": 7.777165433590014e-05, "loss": 1.0231, "step": 89365 }, { "epoch": 0.22, "learning_rate": 7.777039609441844e-05, "loss": 1.0232, "step": 89370 }, { "epoch": 0.22, "learning_rate": 7.776913785293674e-05, "loss": 1.0236, "step": 89375 }, { "epoch": 0.22, "learning_rate": 7.776787961145503e-05, "loss": 1.0248, "step": 89380 }, { "epoch": 0.22, "learning_rate": 7.776662136997332e-05, "loss": 1.0693, "step": 89385 }, { "epoch": 0.22, "learning_rate": 7.776536312849162e-05, "loss": 1.025, "step": 89390 }, { "epoch": 0.22, "learning_rate": 7.776410488700991e-05, "loss": 1.0229, "step": 89395 }, { "epoch": 0.22, "learning_rate": 7.776284664552821e-05, "loss": 1.0213, "step": 89400 }, { "epoch": 0.22, "learning_rate": 7.77615884040465e-05, "loss": 1.0253, "step": 89405 }, { "epoch": 0.22, "learning_rate": 7.77603301625648e-05, "loss": 1.0319, "step": 89410 }, { "epoch": 0.22, "learning_rate": 7.775907192108309e-05, "loss": 1.0234, "step": 89415 }, { "epoch": 0.22, "learning_rate": 7.775781367960139e-05, "loss": 1.0244, "step": 89420 }, { "epoch": 0.22, "learning_rate": 7.775655543811968e-05, "loss": 1.0268, "step": 89425 }, { "epoch": 0.22, "learning_rate": 7.775529719663798e-05, "loss": 1.0449, "step": 89430 }, { "epoch": 0.22, "learning_rate": 7.775403895515627e-05, "loss": 1.0487, "step": 89435 }, { "epoch": 0.22, "learning_rate": 7.775278071367457e-05, "loss": 1.0239, "step": 89440 }, { "epoch": 0.22, "learning_rate": 7.775152247219286e-05, "loss": 1.021, "step": 89445 }, { "epoch": 0.22, "learning_rate": 7.775026423071117e-05, "loss": 1.0224, "step": 89450 }, { "epoch": 0.22, "learning_rate": 7.774900598922946e-05, "loss": 1.0243, "step": 89455 }, { "epoch": 0.22, "learning_rate": 7.774774774774775e-05, "loss": 1.0257, "step": 89460 }, { "epoch": 0.22, "learning_rate": 7.774648950626606e-05, "loss": 1.0444, "step": 89465 }, { "epoch": 0.22, "learning_rate": 7.774523126478435e-05, "loss": 1.0246, "step": 89470 }, { "epoch": 0.22, "learning_rate": 7.774397302330264e-05, "loss": 1.024, "step": 89475 }, { "epoch": 0.22, "learning_rate": 7.774271478182093e-05, "loss": 1.0241, "step": 89480 }, { "epoch": 0.22, "learning_rate": 7.774145654033922e-05, "loss": 1.0241, "step": 89485 }, { "epoch": 0.22, "learning_rate": 7.774019829885753e-05, "loss": 1.023, "step": 89490 }, { "epoch": 0.22, "learning_rate": 7.773894005737582e-05, "loss": 1.0457, "step": 89495 }, { "epoch": 0.22, "learning_rate": 7.773768181589411e-05, "loss": 1.0246, "step": 89500 }, { "epoch": 0.22, "learning_rate": 7.77364235744124e-05, "loss": 1.025, "step": 89505 }, { "epoch": 0.22, "learning_rate": 7.773516533293071e-05, "loss": 1.0219, "step": 89510 }, { "epoch": 0.22, "learning_rate": 7.7733907091449e-05, "loss": 1.0228, "step": 89515 }, { "epoch": 0.22, "learning_rate": 7.773264884996729e-05, "loss": 1.0223, "step": 89520 }, { "epoch": 0.22, "learning_rate": 7.773139060848558e-05, "loss": 1.0245, "step": 89525 }, { "epoch": 0.22, "learning_rate": 7.773013236700389e-05, "loss": 1.0219, "step": 89530 }, { "epoch": 0.22, "learning_rate": 7.772887412552218e-05, "loss": 1.0224, "step": 89535 }, { "epoch": 0.22, "learning_rate": 7.772761588404047e-05, "loss": 1.0241, "step": 89540 }, { "epoch": 0.22, "learning_rate": 7.772635764255876e-05, "loss": 1.0223, "step": 89545 }, { "epoch": 0.22, "learning_rate": 7.772509940107706e-05, "loss": 1.0239, "step": 89550 }, { "epoch": 0.22, "learning_rate": 7.772384115959536e-05, "loss": 1.0247, "step": 89555 }, { "epoch": 0.22, "learning_rate": 7.772258291811365e-05, "loss": 1.025, "step": 89560 }, { "epoch": 0.22, "learning_rate": 7.772132467663194e-05, "loss": 1.0469, "step": 89565 }, { "epoch": 0.22, "learning_rate": 7.772006643515024e-05, "loss": 1.0237, "step": 89570 }, { "epoch": 0.22, "learning_rate": 7.771880819366854e-05, "loss": 1.0269, "step": 89575 }, { "epoch": 0.22, "learning_rate": 7.771754995218683e-05, "loss": 1.0252, "step": 89580 }, { "epoch": 0.22, "learning_rate": 7.771629171070512e-05, "loss": 1.0227, "step": 89585 }, { "epoch": 0.22, "learning_rate": 7.771503346922342e-05, "loss": 1.0234, "step": 89590 }, { "epoch": 0.22, "learning_rate": 7.771377522774172e-05, "loss": 1.0231, "step": 89595 }, { "epoch": 0.22, "learning_rate": 7.771251698626001e-05, "loss": 1.0246, "step": 89600 }, { "epoch": 0.22, "learning_rate": 7.77112587447783e-05, "loss": 1.0261, "step": 89605 }, { "epoch": 0.22, "learning_rate": 7.77100005032966e-05, "loss": 1.0246, "step": 89610 }, { "epoch": 0.22, "learning_rate": 7.770874226181489e-05, "loss": 1.023, "step": 89615 }, { "epoch": 0.22, "learning_rate": 7.770748402033319e-05, "loss": 1.0241, "step": 89620 }, { "epoch": 0.22, "learning_rate": 7.770622577885148e-05, "loss": 1.023, "step": 89625 }, { "epoch": 0.22, "learning_rate": 7.770496753736978e-05, "loss": 1.046, "step": 89630 }, { "epoch": 0.22, "learning_rate": 7.770370929588807e-05, "loss": 1.022, "step": 89635 }, { "epoch": 0.23, "learning_rate": 7.770245105440637e-05, "loss": 1.0258, "step": 89640 }, { "epoch": 0.23, "learning_rate": 7.770119281292466e-05, "loss": 1.0235, "step": 89645 }, { "epoch": 0.23, "learning_rate": 7.769993457144296e-05, "loss": 1.0225, "step": 89650 }, { "epoch": 0.23, "learning_rate": 7.769867632996125e-05, "loss": 1.0234, "step": 89655 }, { "epoch": 0.23, "learning_rate": 7.769741808847954e-05, "loss": 1.0239, "step": 89660 }, { "epoch": 0.23, "learning_rate": 7.769615984699784e-05, "loss": 1.0234, "step": 89665 }, { "epoch": 0.23, "learning_rate": 7.769490160551614e-05, "loss": 1.0229, "step": 89670 }, { "epoch": 0.23, "learning_rate": 7.769364336403443e-05, "loss": 1.0242, "step": 89675 }, { "epoch": 0.23, "learning_rate": 7.769238512255272e-05, "loss": 1.0252, "step": 89680 }, { "epoch": 0.23, "learning_rate": 7.769112688107102e-05, "loss": 1.0223, "step": 89685 }, { "epoch": 0.23, "learning_rate": 7.768986863958932e-05, "loss": 1.0237, "step": 89690 }, { "epoch": 0.23, "learning_rate": 7.768861039810761e-05, "loss": 1.0262, "step": 89695 }, { "epoch": 0.23, "learning_rate": 7.76873521566259e-05, "loss": 1.0229, "step": 89700 }, { "epoch": 0.23, "learning_rate": 7.76860939151442e-05, "loss": 1.0269, "step": 89705 }, { "epoch": 0.23, "learning_rate": 7.76848356736625e-05, "loss": 1.0251, "step": 89710 }, { "epoch": 0.23, "learning_rate": 7.768357743218079e-05, "loss": 1.0243, "step": 89715 }, { "epoch": 0.23, "learning_rate": 7.768231919069908e-05, "loss": 1.0236, "step": 89720 }, { "epoch": 0.23, "learning_rate": 7.768106094921737e-05, "loss": 1.0204, "step": 89725 }, { "epoch": 0.23, "learning_rate": 7.767980270773568e-05, "loss": 1.0219, "step": 89730 }, { "epoch": 0.23, "learning_rate": 7.767854446625397e-05, "loss": 1.0428, "step": 89735 }, { "epoch": 0.23, "learning_rate": 7.767728622477226e-05, "loss": 1.0255, "step": 89740 }, { "epoch": 0.23, "learning_rate": 7.767602798329055e-05, "loss": 1.0245, "step": 89745 }, { "epoch": 0.23, "learning_rate": 7.767476974180886e-05, "loss": 1.0211, "step": 89750 }, { "epoch": 0.23, "learning_rate": 7.767351150032715e-05, "loss": 1.0242, "step": 89755 }, { "epoch": 0.23, "learning_rate": 7.767225325884544e-05, "loss": 1.0233, "step": 89760 }, { "epoch": 0.23, "learning_rate": 7.767099501736373e-05, "loss": 1.0229, "step": 89765 }, { "epoch": 0.23, "learning_rate": 7.766973677588204e-05, "loss": 1.0229, "step": 89770 }, { "epoch": 0.23, "learning_rate": 7.766847853440033e-05, "loss": 1.0257, "step": 89775 }, { "epoch": 0.23, "learning_rate": 7.766722029291862e-05, "loss": 1.0226, "step": 89780 }, { "epoch": 0.23, "learning_rate": 7.766596205143691e-05, "loss": 1.0254, "step": 89785 }, { "epoch": 0.23, "learning_rate": 7.76647038099552e-05, "loss": 1.0244, "step": 89790 }, { "epoch": 0.23, "learning_rate": 7.76634455684735e-05, "loss": 1.0248, "step": 89795 }, { "epoch": 0.23, "learning_rate": 7.76621873269918e-05, "loss": 1.0267, "step": 89800 }, { "epoch": 0.23, "learning_rate": 7.766092908551009e-05, "loss": 1.0217, "step": 89805 }, { "epoch": 0.23, "learning_rate": 7.765967084402838e-05, "loss": 1.0241, "step": 89810 }, { "epoch": 0.23, "learning_rate": 7.765841260254669e-05, "loss": 1.0238, "step": 89815 }, { "epoch": 0.23, "learning_rate": 7.765715436106498e-05, "loss": 1.0241, "step": 89820 }, { "epoch": 0.23, "learning_rate": 7.765589611958327e-05, "loss": 1.0243, "step": 89825 }, { "epoch": 0.23, "learning_rate": 7.765463787810156e-05, "loss": 1.0231, "step": 89830 }, { "epoch": 0.23, "learning_rate": 7.765337963661987e-05, "loss": 1.0259, "step": 89835 }, { "epoch": 0.23, "learning_rate": 7.765212139513816e-05, "loss": 1.0234, "step": 89840 }, { "epoch": 0.23, "learning_rate": 7.765086315365645e-05, "loss": 1.0229, "step": 89845 }, { "epoch": 0.23, "learning_rate": 7.764960491217474e-05, "loss": 1.0229, "step": 89850 }, { "epoch": 0.23, "learning_rate": 7.764834667069303e-05, "loss": 1.023, "step": 89855 }, { "epoch": 0.23, "learning_rate": 7.764708842921134e-05, "loss": 1.0216, "step": 89860 }, { "epoch": 0.23, "learning_rate": 7.764583018772963e-05, "loss": 1.0258, "step": 89865 }, { "epoch": 0.23, "learning_rate": 7.764457194624792e-05, "loss": 1.0243, "step": 89870 }, { "epoch": 0.23, "learning_rate": 7.764331370476621e-05, "loss": 1.0252, "step": 89875 }, { "epoch": 0.23, "learning_rate": 7.764205546328452e-05, "loss": 1.0242, "step": 89880 }, { "epoch": 0.23, "learning_rate": 7.764079722180281e-05, "loss": 1.0225, "step": 89885 }, { "epoch": 0.23, "learning_rate": 7.76395389803211e-05, "loss": 1.0246, "step": 89890 }, { "epoch": 0.23, "learning_rate": 7.763828073883939e-05, "loss": 1.0271, "step": 89895 }, { "epoch": 0.23, "learning_rate": 7.76370224973577e-05, "loss": 1.0228, "step": 89900 }, { "epoch": 0.23, "learning_rate": 7.763576425587599e-05, "loss": 1.0218, "step": 89905 }, { "epoch": 0.23, "learning_rate": 7.763450601439428e-05, "loss": 1.0243, "step": 89910 }, { "epoch": 0.23, "learning_rate": 7.763324777291257e-05, "loss": 1.0224, "step": 89915 }, { "epoch": 0.23, "learning_rate": 7.763198953143086e-05, "loss": 1.0224, "step": 89920 }, { "epoch": 0.23, "learning_rate": 7.763073128994917e-05, "loss": 1.0255, "step": 89925 }, { "epoch": 0.23, "learning_rate": 7.762947304846746e-05, "loss": 1.0232, "step": 89930 }, { "epoch": 0.23, "learning_rate": 7.762821480698575e-05, "loss": 1.022, "step": 89935 }, { "epoch": 0.23, "learning_rate": 7.762695656550404e-05, "loss": 1.0249, "step": 89940 }, { "epoch": 0.23, "learning_rate": 7.762569832402235e-05, "loss": 1.0224, "step": 89945 }, { "epoch": 0.23, "learning_rate": 7.762444008254065e-05, "loss": 1.0227, "step": 89950 }, { "epoch": 0.23, "learning_rate": 7.762318184105895e-05, "loss": 1.0255, "step": 89955 }, { "epoch": 0.23, "learning_rate": 7.762192359957724e-05, "loss": 1.0252, "step": 89960 }, { "epoch": 0.23, "learning_rate": 7.762066535809553e-05, "loss": 1.0263, "step": 89965 }, { "epoch": 0.23, "learning_rate": 7.761940711661383e-05, "loss": 1.0449, "step": 89970 }, { "epoch": 0.23, "learning_rate": 7.761814887513213e-05, "loss": 1.0235, "step": 89975 }, { "epoch": 0.23, "learning_rate": 7.761689063365042e-05, "loss": 1.0227, "step": 89980 }, { "epoch": 0.23, "learning_rate": 7.761563239216871e-05, "loss": 1.0242, "step": 89985 }, { "epoch": 0.23, "learning_rate": 7.761437415068701e-05, "loss": 1.0233, "step": 89990 }, { "epoch": 0.23, "learning_rate": 7.76131159092053e-05, "loss": 1.0256, "step": 89995 }, { "epoch": 0.23, "learning_rate": 7.76118576677236e-05, "loss": 1.0232, "step": 90000 }, { "epoch": 0.23, "learning_rate": 7.761059942624189e-05, "loss": 1.025, "step": 90005 }, { "epoch": 0.23, "learning_rate": 7.760934118476018e-05, "loss": 1.0237, "step": 90010 }, { "epoch": 0.23, "learning_rate": 7.760808294327849e-05, "loss": 1.0208, "step": 90015 }, { "epoch": 0.23, "learning_rate": 7.760682470179678e-05, "loss": 1.0248, "step": 90020 }, { "epoch": 0.23, "learning_rate": 7.760556646031507e-05, "loss": 1.0237, "step": 90025 }, { "epoch": 0.23, "learning_rate": 7.760430821883336e-05, "loss": 1.0218, "step": 90030 }, { "epoch": 0.23, "learning_rate": 7.760304997735167e-05, "loss": 1.0263, "step": 90035 }, { "epoch": 0.23, "learning_rate": 7.760179173586996e-05, "loss": 1.0284, "step": 90040 }, { "epoch": 0.23, "learning_rate": 7.760053349438825e-05, "loss": 1.0232, "step": 90045 }, { "epoch": 0.23, "learning_rate": 7.759927525290654e-05, "loss": 1.0231, "step": 90050 }, { "epoch": 0.23, "learning_rate": 7.759801701142485e-05, "loss": 1.0247, "step": 90055 }, { "epoch": 0.23, "learning_rate": 7.759675876994314e-05, "loss": 1.0237, "step": 90060 }, { "epoch": 0.23, "learning_rate": 7.759550052846143e-05, "loss": 1.0251, "step": 90065 }, { "epoch": 0.23, "learning_rate": 7.759424228697972e-05, "loss": 1.0243, "step": 90070 }, { "epoch": 0.23, "learning_rate": 7.759298404549801e-05, "loss": 1.0243, "step": 90075 }, { "epoch": 0.23, "learning_rate": 7.759172580401632e-05, "loss": 1.0249, "step": 90080 }, { "epoch": 0.23, "learning_rate": 7.759046756253461e-05, "loss": 1.027, "step": 90085 }, { "epoch": 0.23, "learning_rate": 7.75892093210529e-05, "loss": 1.0223, "step": 90090 }, { "epoch": 0.23, "learning_rate": 7.758795107957119e-05, "loss": 1.0243, "step": 90095 }, { "epoch": 0.23, "learning_rate": 7.75866928380895e-05, "loss": 1.0417, "step": 90100 }, { "epoch": 0.23, "learning_rate": 7.758543459660779e-05, "loss": 1.0244, "step": 90105 }, { "epoch": 0.23, "learning_rate": 7.758417635512608e-05, "loss": 1.0244, "step": 90110 }, { "epoch": 0.23, "learning_rate": 7.758291811364437e-05, "loss": 1.0225, "step": 90115 }, { "epoch": 0.23, "learning_rate": 7.758165987216268e-05, "loss": 1.0237, "step": 90120 }, { "epoch": 0.23, "learning_rate": 7.758040163068097e-05, "loss": 1.0251, "step": 90125 }, { "epoch": 0.23, "learning_rate": 7.757914338919926e-05, "loss": 1.024, "step": 90130 }, { "epoch": 0.23, "learning_rate": 7.757788514771755e-05, "loss": 1.0242, "step": 90135 }, { "epoch": 0.23, "learning_rate": 7.757662690623584e-05, "loss": 1.0229, "step": 90140 }, { "epoch": 0.23, "learning_rate": 7.757536866475415e-05, "loss": 1.0239, "step": 90145 }, { "epoch": 0.23, "learning_rate": 7.757411042327244e-05, "loss": 1.0243, "step": 90150 }, { "epoch": 0.23, "learning_rate": 7.757285218179073e-05, "loss": 1.0248, "step": 90155 }, { "epoch": 0.23, "learning_rate": 7.757159394030902e-05, "loss": 1.0235, "step": 90160 }, { "epoch": 0.23, "learning_rate": 7.757033569882733e-05, "loss": 1.024, "step": 90165 }, { "epoch": 0.23, "learning_rate": 7.756907745734562e-05, "loss": 1.0249, "step": 90170 }, { "epoch": 0.23, "learning_rate": 7.756781921586391e-05, "loss": 1.0252, "step": 90175 }, { "epoch": 0.23, "learning_rate": 7.75665609743822e-05, "loss": 1.0217, "step": 90180 }, { "epoch": 0.23, "learning_rate": 7.756530273290051e-05, "loss": 1.0227, "step": 90185 }, { "epoch": 0.23, "learning_rate": 7.75640444914188e-05, "loss": 1.0225, "step": 90190 }, { "epoch": 0.23, "learning_rate": 7.756278624993709e-05, "loss": 1.0243, "step": 90195 }, { "epoch": 0.23, "learning_rate": 7.756152800845538e-05, "loss": 1.0234, "step": 90200 }, { "epoch": 0.23, "learning_rate": 7.756026976697368e-05, "loss": 1.0232, "step": 90205 }, { "epoch": 0.23, "learning_rate": 7.755901152549198e-05, "loss": 1.0251, "step": 90210 }, { "epoch": 0.23, "learning_rate": 7.755775328401027e-05, "loss": 1.0237, "step": 90215 }, { "epoch": 0.23, "learning_rate": 7.755649504252856e-05, "loss": 1.0242, "step": 90220 }, { "epoch": 0.23, "learning_rate": 7.755523680104686e-05, "loss": 1.0241, "step": 90225 }, { "epoch": 0.23, "learning_rate": 7.755397855956516e-05, "loss": 1.0252, "step": 90230 }, { "epoch": 0.23, "learning_rate": 7.755272031808345e-05, "loss": 1.0241, "step": 90235 }, { "epoch": 0.23, "learning_rate": 7.755146207660174e-05, "loss": 1.0218, "step": 90240 }, { "epoch": 0.23, "learning_rate": 7.755020383512004e-05, "loss": 1.0276, "step": 90245 }, { "epoch": 0.23, "learning_rate": 7.754894559363834e-05, "loss": 1.0236, "step": 90250 }, { "epoch": 0.23, "learning_rate": 7.754768735215663e-05, "loss": 1.0257, "step": 90255 }, { "epoch": 0.23, "learning_rate": 7.754642911067492e-05, "loss": 1.0241, "step": 90260 }, { "epoch": 0.23, "learning_rate": 7.754517086919322e-05, "loss": 1.0242, "step": 90265 }, { "epoch": 0.23, "learning_rate": 7.75439126277115e-05, "loss": 1.0236, "step": 90270 }, { "epoch": 0.23, "learning_rate": 7.754265438622981e-05, "loss": 1.0236, "step": 90275 }, { "epoch": 0.23, "learning_rate": 7.75413961447481e-05, "loss": 1.0263, "step": 90280 }, { "epoch": 0.23, "learning_rate": 7.75401379032664e-05, "loss": 1.0476, "step": 90285 }, { "epoch": 0.23, "learning_rate": 7.753887966178469e-05, "loss": 1.0232, "step": 90290 }, { "epoch": 0.23, "learning_rate": 7.753762142030299e-05, "loss": 1.0226, "step": 90295 }, { "epoch": 0.23, "learning_rate": 7.753636317882128e-05, "loss": 1.0232, "step": 90300 }, { "epoch": 0.23, "learning_rate": 7.753510493733957e-05, "loss": 1.0251, "step": 90305 }, { "epoch": 0.23, "learning_rate": 7.753384669585787e-05, "loss": 1.0222, "step": 90310 }, { "epoch": 0.23, "learning_rate": 7.753258845437616e-05, "loss": 1.0245, "step": 90315 }, { "epoch": 0.23, "learning_rate": 7.753133021289446e-05, "loss": 1.0217, "step": 90320 }, { "epoch": 0.23, "learning_rate": 7.753007197141275e-05, "loss": 1.0239, "step": 90325 }, { "epoch": 0.23, "learning_rate": 7.752881372993105e-05, "loss": 1.0221, "step": 90330 }, { "epoch": 0.23, "learning_rate": 7.752755548844934e-05, "loss": 1.0275, "step": 90335 }, { "epoch": 0.23, "learning_rate": 7.752629724696764e-05, "loss": 1.024, "step": 90340 }, { "epoch": 0.23, "learning_rate": 7.752503900548593e-05, "loss": 1.0402, "step": 90345 }, { "epoch": 0.23, "learning_rate": 7.752378076400423e-05, "loss": 1.0239, "step": 90350 }, { "epoch": 0.23, "learning_rate": 7.752252252252252e-05, "loss": 1.0226, "step": 90355 }, { "epoch": 0.23, "learning_rate": 7.752126428104082e-05, "loss": 1.0241, "step": 90360 }, { "epoch": 0.23, "learning_rate": 7.752000603955911e-05, "loss": 1.0226, "step": 90365 }, { "epoch": 0.23, "learning_rate": 7.75187477980774e-05, "loss": 1.0232, "step": 90370 }, { "epoch": 0.23, "learning_rate": 7.75174895565957e-05, "loss": 1.0226, "step": 90375 }, { "epoch": 0.23, "learning_rate": 7.751623131511399e-05, "loss": 1.0257, "step": 90380 }, { "epoch": 0.23, "learning_rate": 7.75149730736323e-05, "loss": 1.0237, "step": 90385 }, { "epoch": 0.23, "learning_rate": 7.751371483215059e-05, "loss": 1.0255, "step": 90390 }, { "epoch": 0.23, "learning_rate": 7.751245659066888e-05, "loss": 1.023, "step": 90395 }, { "epoch": 0.23, "learning_rate": 7.751119834918717e-05, "loss": 1.026, "step": 90400 }, { "epoch": 0.23, "learning_rate": 7.750994010770547e-05, "loss": 1.0221, "step": 90405 }, { "epoch": 0.23, "learning_rate": 7.750868186622377e-05, "loss": 1.0247, "step": 90410 }, { "epoch": 0.23, "learning_rate": 7.750742362474206e-05, "loss": 1.025, "step": 90415 }, { "epoch": 0.23, "learning_rate": 7.750616538326035e-05, "loss": 1.0232, "step": 90420 }, { "epoch": 0.23, "learning_rate": 7.750490714177865e-05, "loss": 1.0252, "step": 90425 }, { "epoch": 0.23, "learning_rate": 7.750364890029695e-05, "loss": 1.0234, "step": 90430 }, { "epoch": 0.23, "learning_rate": 7.750239065881524e-05, "loss": 1.0218, "step": 90435 }, { "epoch": 0.23, "learning_rate": 7.750113241733353e-05, "loss": 1.0234, "step": 90440 }, { "epoch": 0.23, "learning_rate": 7.749987417585183e-05, "loss": 1.0235, "step": 90445 }, { "epoch": 0.23, "learning_rate": 7.749861593437014e-05, "loss": 1.0246, "step": 90450 }, { "epoch": 0.23, "learning_rate": 7.749735769288843e-05, "loss": 1.0238, "step": 90455 }, { "epoch": 0.23, "learning_rate": 7.749609945140672e-05, "loss": 1.0188, "step": 90460 }, { "epoch": 0.23, "learning_rate": 7.749484120992501e-05, "loss": 1.0238, "step": 90465 }, { "epoch": 0.23, "learning_rate": 7.74935829684433e-05, "loss": 1.0219, "step": 90470 }, { "epoch": 0.23, "learning_rate": 7.749232472696161e-05, "loss": 1.024, "step": 90475 }, { "epoch": 0.23, "learning_rate": 7.74910664854799e-05, "loss": 1.0239, "step": 90480 }, { "epoch": 0.23, "learning_rate": 7.74898082439982e-05, "loss": 1.0236, "step": 90485 }, { "epoch": 0.23, "learning_rate": 7.748855000251649e-05, "loss": 1.0445, "step": 90490 }, { "epoch": 0.23, "learning_rate": 7.748729176103479e-05, "loss": 1.0231, "step": 90495 }, { "epoch": 0.23, "learning_rate": 7.748603351955308e-05, "loss": 1.0248, "step": 90500 }, { "epoch": 0.23, "learning_rate": 7.748477527807137e-05, "loss": 1.0222, "step": 90505 }, { "epoch": 0.23, "learning_rate": 7.748351703658967e-05, "loss": 1.0229, "step": 90510 }, { "epoch": 0.23, "learning_rate": 7.748225879510797e-05, "loss": 1.0235, "step": 90515 }, { "epoch": 0.23, "learning_rate": 7.748100055362626e-05, "loss": 1.0235, "step": 90520 }, { "epoch": 0.23, "learning_rate": 7.747974231214455e-05, "loss": 1.0247, "step": 90525 }, { "epoch": 0.23, "learning_rate": 7.747848407066285e-05, "loss": 1.0262, "step": 90530 }, { "epoch": 0.23, "learning_rate": 7.747722582918114e-05, "loss": 1.0237, "step": 90535 }, { "epoch": 0.23, "learning_rate": 7.747596758769944e-05, "loss": 1.0267, "step": 90540 }, { "epoch": 0.23, "learning_rate": 7.747470934621773e-05, "loss": 1.0286, "step": 90545 }, { "epoch": 0.23, "learning_rate": 7.747345110473603e-05, "loss": 1.0249, "step": 90550 }, { "epoch": 0.23, "learning_rate": 7.747219286325432e-05, "loss": 1.0231, "step": 90555 }, { "epoch": 0.23, "learning_rate": 7.747093462177262e-05, "loss": 1.0227, "step": 90560 }, { "epoch": 0.23, "learning_rate": 7.746967638029091e-05, "loss": 1.0224, "step": 90565 }, { "epoch": 0.23, "learning_rate": 7.74684181388092e-05, "loss": 1.0266, "step": 90570 }, { "epoch": 0.23, "learning_rate": 7.74671598973275e-05, "loss": 1.0234, "step": 90575 }, { "epoch": 0.23, "learning_rate": 7.74659016558458e-05, "loss": 1.0223, "step": 90580 }, { "epoch": 0.23, "learning_rate": 7.74646434143641e-05, "loss": 1.0216, "step": 90585 }, { "epoch": 0.23, "learning_rate": 7.746338517288239e-05, "loss": 1.0275, "step": 90590 }, { "epoch": 0.23, "learning_rate": 7.746212693140068e-05, "loss": 1.0234, "step": 90595 }, { "epoch": 0.23, "learning_rate": 7.746086868991897e-05, "loss": 1.0246, "step": 90600 }, { "epoch": 0.23, "learning_rate": 7.745961044843727e-05, "loss": 1.0239, "step": 90605 }, { "epoch": 0.23, "learning_rate": 7.745835220695557e-05, "loss": 1.0242, "step": 90610 }, { "epoch": 0.23, "learning_rate": 7.745709396547386e-05, "loss": 1.0209, "step": 90615 }, { "epoch": 0.23, "learning_rate": 7.745583572399215e-05, "loss": 1.024, "step": 90620 }, { "epoch": 0.23, "learning_rate": 7.745457748251045e-05, "loss": 1.0248, "step": 90625 }, { "epoch": 0.23, "learning_rate": 7.745331924102875e-05, "loss": 1.0235, "step": 90630 }, { "epoch": 0.23, "learning_rate": 7.745206099954704e-05, "loss": 1.0214, "step": 90635 }, { "epoch": 0.23, "learning_rate": 7.745080275806533e-05, "loss": 1.0434, "step": 90640 }, { "epoch": 0.23, "learning_rate": 7.744954451658363e-05, "loss": 1.0255, "step": 90645 }, { "epoch": 0.23, "learning_rate": 7.744828627510193e-05, "loss": 1.0233, "step": 90650 }, { "epoch": 0.23, "learning_rate": 7.744702803362022e-05, "loss": 1.0238, "step": 90655 }, { "epoch": 0.23, "learning_rate": 7.744576979213851e-05, "loss": 1.0254, "step": 90660 }, { "epoch": 0.23, "learning_rate": 7.74445115506568e-05, "loss": 1.0236, "step": 90665 }, { "epoch": 0.23, "learning_rate": 7.74432533091751e-05, "loss": 1.0235, "step": 90670 }, { "epoch": 0.23, "learning_rate": 7.74419950676934e-05, "loss": 1.0222, "step": 90675 }, { "epoch": 0.23, "learning_rate": 7.744073682621169e-05, "loss": 1.0247, "step": 90680 }, { "epoch": 0.23, "learning_rate": 7.743947858472998e-05, "loss": 1.0237, "step": 90685 }, { "epoch": 0.23, "learning_rate": 7.743822034324829e-05, "loss": 1.024, "step": 90690 }, { "epoch": 0.23, "learning_rate": 7.743696210176658e-05, "loss": 1.0219, "step": 90695 }, { "epoch": 0.23, "learning_rate": 7.743570386028487e-05, "loss": 1.0249, "step": 90700 }, { "epoch": 0.23, "learning_rate": 7.743444561880316e-05, "loss": 1.0219, "step": 90705 }, { "epoch": 0.23, "learning_rate": 7.743318737732147e-05, "loss": 1.0235, "step": 90710 }, { "epoch": 0.23, "learning_rate": 7.743192913583976e-05, "loss": 1.0229, "step": 90715 }, { "epoch": 0.23, "learning_rate": 7.743067089435805e-05, "loss": 1.0478, "step": 90720 }, { "epoch": 0.23, "learning_rate": 7.742941265287634e-05, "loss": 1.0239, "step": 90725 }, { "epoch": 0.23, "learning_rate": 7.742815441139463e-05, "loss": 1.0226, "step": 90730 }, { "epoch": 0.23, "learning_rate": 7.742689616991294e-05, "loss": 1.0232, "step": 90735 }, { "epoch": 0.23, "learning_rate": 7.742563792843123e-05, "loss": 1.0271, "step": 90740 }, { "epoch": 0.23, "learning_rate": 7.742437968694952e-05, "loss": 1.0223, "step": 90745 }, { "epoch": 0.23, "learning_rate": 7.742312144546781e-05, "loss": 1.0243, "step": 90750 }, { "epoch": 0.23, "learning_rate": 7.742186320398612e-05, "loss": 1.0228, "step": 90755 }, { "epoch": 0.23, "learning_rate": 7.742060496250441e-05, "loss": 1.023, "step": 90760 }, { "epoch": 0.23, "learning_rate": 7.74193467210227e-05, "loss": 1.0239, "step": 90765 }, { "epoch": 0.23, "learning_rate": 7.741808847954099e-05, "loss": 1.0232, "step": 90770 }, { "epoch": 0.23, "learning_rate": 7.74168302380593e-05, "loss": 1.0227, "step": 90775 }, { "epoch": 0.23, "learning_rate": 7.741557199657759e-05, "loss": 1.0261, "step": 90780 }, { "epoch": 0.23, "learning_rate": 7.741431375509588e-05, "loss": 1.025, "step": 90785 }, { "epoch": 0.23, "learning_rate": 7.741305551361417e-05, "loss": 1.0252, "step": 90790 }, { "epoch": 0.23, "learning_rate": 7.741179727213246e-05, "loss": 1.0222, "step": 90795 }, { "epoch": 0.23, "learning_rate": 7.741053903065077e-05, "loss": 1.0247, "step": 90800 }, { "epoch": 0.23, "learning_rate": 7.740928078916906e-05, "loss": 1.0255, "step": 90805 }, { "epoch": 0.23, "learning_rate": 7.740802254768735e-05, "loss": 1.0253, "step": 90810 }, { "epoch": 0.23, "learning_rate": 7.740676430620564e-05, "loss": 1.0318, "step": 90815 }, { "epoch": 0.23, "learning_rate": 7.740550606472395e-05, "loss": 1.0241, "step": 90820 }, { "epoch": 0.23, "learning_rate": 7.740424782324224e-05, "loss": 1.0235, "step": 90825 }, { "epoch": 0.23, "learning_rate": 7.740298958176053e-05, "loss": 1.0258, "step": 90830 }, { "epoch": 0.23, "learning_rate": 7.740173134027882e-05, "loss": 1.0238, "step": 90835 }, { "epoch": 0.23, "learning_rate": 7.740047309879713e-05, "loss": 1.0271, "step": 90840 }, { "epoch": 0.23, "learning_rate": 7.739921485731542e-05, "loss": 1.0258, "step": 90845 }, { "epoch": 0.23, "learning_rate": 7.739795661583371e-05, "loss": 1.0263, "step": 90850 }, { "epoch": 0.23, "learning_rate": 7.7396698374352e-05, "loss": 1.0231, "step": 90855 }, { "epoch": 0.23, "learning_rate": 7.73954401328703e-05, "loss": 1.0226, "step": 90860 }, { "epoch": 0.23, "learning_rate": 7.73941818913886e-05, "loss": 1.0243, "step": 90865 }, { "epoch": 0.23, "learning_rate": 7.739292364990689e-05, "loss": 1.0465, "step": 90870 }, { "epoch": 0.23, "learning_rate": 7.739166540842518e-05, "loss": 1.0238, "step": 90875 }, { "epoch": 0.23, "learning_rate": 7.739040716694347e-05, "loss": 1.025, "step": 90880 }, { "epoch": 0.23, "learning_rate": 7.738914892546178e-05, "loss": 1.0223, "step": 90885 }, { "epoch": 0.23, "learning_rate": 7.738789068398007e-05, "loss": 1.0254, "step": 90890 }, { "epoch": 0.23, "learning_rate": 7.738663244249836e-05, "loss": 1.0227, "step": 90895 }, { "epoch": 0.23, "learning_rate": 7.738537420101665e-05, "loss": 1.0225, "step": 90900 }, { "epoch": 0.23, "learning_rate": 7.738411595953496e-05, "loss": 1.025, "step": 90905 }, { "epoch": 0.23, "learning_rate": 7.738285771805325e-05, "loss": 1.0231, "step": 90910 }, { "epoch": 0.23, "learning_rate": 7.738159947657154e-05, "loss": 1.0229, "step": 90915 }, { "epoch": 0.23, "learning_rate": 7.738034123508983e-05, "loss": 1.0241, "step": 90920 }, { "epoch": 0.23, "learning_rate": 7.737908299360813e-05, "loss": 1.023, "step": 90925 }, { "epoch": 0.23, "learning_rate": 7.737782475212643e-05, "loss": 1.0224, "step": 90930 }, { "epoch": 0.23, "learning_rate": 7.737656651064472e-05, "loss": 1.0236, "step": 90935 }, { "epoch": 0.23, "learning_rate": 7.737530826916301e-05, "loss": 1.0242, "step": 90940 }, { "epoch": 0.23, "learning_rate": 7.737405002768132e-05, "loss": 1.0241, "step": 90945 }, { "epoch": 0.23, "learning_rate": 7.737279178619961e-05, "loss": 1.0247, "step": 90950 }, { "epoch": 0.23, "learning_rate": 7.737153354471792e-05, "loss": 1.0204, "step": 90955 }, { "epoch": 0.23, "learning_rate": 7.737027530323621e-05, "loss": 1.0237, "step": 90960 }, { "epoch": 0.23, "learning_rate": 7.73690170617545e-05, "loss": 1.0229, "step": 90965 }, { "epoch": 0.23, "learning_rate": 7.736775882027279e-05, "loss": 1.0233, "step": 90970 }, { "epoch": 0.23, "learning_rate": 7.73665005787911e-05, "loss": 1.0259, "step": 90975 }, { "epoch": 0.23, "learning_rate": 7.736524233730939e-05, "loss": 1.0233, "step": 90980 }, { "epoch": 0.23, "learning_rate": 7.736398409582768e-05, "loss": 1.0247, "step": 90985 }, { "epoch": 0.23, "learning_rate": 7.736272585434597e-05, "loss": 1.0231, "step": 90990 }, { "epoch": 0.23, "learning_rate": 7.736146761286426e-05, "loss": 1.0253, "step": 90995 }, { "epoch": 0.23, "learning_rate": 7.736020937138257e-05, "loss": 1.0225, "step": 91000 }, { "epoch": 0.23, "learning_rate": 7.735895112990086e-05, "loss": 1.0259, "step": 91005 }, { "epoch": 0.23, "learning_rate": 7.735769288841915e-05, "loss": 1.0242, "step": 91010 }, { "epoch": 0.23, "learning_rate": 7.735643464693744e-05, "loss": 1.0249, "step": 91015 }, { "epoch": 0.23, "learning_rate": 7.735517640545575e-05, "loss": 1.0235, "step": 91020 }, { "epoch": 0.23, "learning_rate": 7.735391816397404e-05, "loss": 1.0216, "step": 91025 }, { "epoch": 0.23, "learning_rate": 7.735265992249233e-05, "loss": 1.0221, "step": 91030 }, { "epoch": 0.23, "learning_rate": 7.735140168101062e-05, "loss": 1.0248, "step": 91035 }, { "epoch": 0.23, "learning_rate": 7.735014343952893e-05, "loss": 1.0221, "step": 91040 }, { "epoch": 0.23, "learning_rate": 7.734888519804722e-05, "loss": 1.0444, "step": 91045 }, { "epoch": 0.23, "learning_rate": 7.734762695656551e-05, "loss": 1.0239, "step": 91050 }, { "epoch": 0.23, "learning_rate": 7.73463687150838e-05, "loss": 1.0256, "step": 91055 }, { "epoch": 0.23, "learning_rate": 7.73451104736021e-05, "loss": 1.0237, "step": 91060 }, { "epoch": 0.23, "learning_rate": 7.73438522321204e-05, "loss": 1.0231, "step": 91065 }, { "epoch": 0.23, "learning_rate": 7.734259399063869e-05, "loss": 1.0217, "step": 91070 }, { "epoch": 0.23, "learning_rate": 7.734133574915698e-05, "loss": 1.023, "step": 91075 }, { "epoch": 0.23, "learning_rate": 7.734007750767527e-05, "loss": 1.0246, "step": 91080 }, { "epoch": 0.23, "learning_rate": 7.733881926619358e-05, "loss": 1.0239, "step": 91085 }, { "epoch": 0.23, "learning_rate": 7.733756102471187e-05, "loss": 1.0248, "step": 91090 }, { "epoch": 0.23, "learning_rate": 7.733630278323016e-05, "loss": 1.0239, "step": 91095 }, { "epoch": 0.23, "learning_rate": 7.733504454174845e-05, "loss": 1.0241, "step": 91100 }, { "epoch": 0.23, "learning_rate": 7.733378630026676e-05, "loss": 1.0229, "step": 91105 }, { "epoch": 0.23, "learning_rate": 7.733252805878505e-05, "loss": 1.0241, "step": 91110 }, { "epoch": 0.23, "learning_rate": 7.733126981730334e-05, "loss": 1.0242, "step": 91115 }, { "epoch": 0.23, "learning_rate": 7.733001157582163e-05, "loss": 1.0261, "step": 91120 }, { "epoch": 0.23, "learning_rate": 7.732875333433993e-05, "loss": 1.0212, "step": 91125 }, { "epoch": 0.23, "learning_rate": 7.732749509285823e-05, "loss": 1.0476, "step": 91130 }, { "epoch": 0.23, "learning_rate": 7.732623685137652e-05, "loss": 1.0223, "step": 91135 }, { "epoch": 0.23, "learning_rate": 7.732497860989481e-05, "loss": 1.0233, "step": 91140 }, { "epoch": 0.23, "learning_rate": 7.73237203684131e-05, "loss": 1.0243, "step": 91145 }, { "epoch": 0.23, "learning_rate": 7.732246212693141e-05, "loss": 1.0248, "step": 91150 }, { "epoch": 0.23, "learning_rate": 7.73212038854497e-05, "loss": 1.0234, "step": 91155 }, { "epoch": 0.23, "learning_rate": 7.7319945643968e-05, "loss": 1.0223, "step": 91160 }, { "epoch": 0.23, "learning_rate": 7.731868740248629e-05, "loss": 1.024, "step": 91165 }, { "epoch": 0.23, "learning_rate": 7.731742916100459e-05, "loss": 1.0242, "step": 91170 }, { "epoch": 0.23, "learning_rate": 7.731617091952288e-05, "loss": 1.0432, "step": 91175 }, { "epoch": 0.23, "learning_rate": 7.731491267804117e-05, "loss": 1.0245, "step": 91180 }, { "epoch": 0.23, "learning_rate": 7.731365443655947e-05, "loss": 1.0247, "step": 91185 }, { "epoch": 0.23, "learning_rate": 7.731239619507776e-05, "loss": 1.0261, "step": 91190 }, { "epoch": 0.23, "learning_rate": 7.731113795359606e-05, "loss": 1.0234, "step": 91195 }, { "epoch": 0.23, "learning_rate": 7.730987971211435e-05, "loss": 1.0269, "step": 91200 }, { "epoch": 0.23, "learning_rate": 7.730862147063265e-05, "loss": 1.0243, "step": 91205 }, { "epoch": 0.23, "learning_rate": 7.730736322915094e-05, "loss": 1.024, "step": 91210 }, { "epoch": 0.23, "learning_rate": 7.730610498766924e-05, "loss": 1.0233, "step": 91215 }, { "epoch": 0.23, "learning_rate": 7.730484674618753e-05, "loss": 1.0234, "step": 91220 }, { "epoch": 0.23, "learning_rate": 7.730358850470583e-05, "loss": 1.0235, "step": 91225 }, { "epoch": 0.23, "learning_rate": 7.730233026322412e-05, "loss": 1.0242, "step": 91230 }, { "epoch": 0.23, "learning_rate": 7.730107202174242e-05, "loss": 1.0243, "step": 91235 }, { "epoch": 0.23, "learning_rate": 7.729981378026071e-05, "loss": 1.0245, "step": 91240 }, { "epoch": 0.23, "learning_rate": 7.7298555538779e-05, "loss": 1.0231, "step": 91245 }, { "epoch": 0.23, "learning_rate": 7.72972972972973e-05, "loss": 1.0224, "step": 91250 }, { "epoch": 0.23, "learning_rate": 7.729603905581559e-05, "loss": 1.0243, "step": 91255 }, { "epoch": 0.23, "learning_rate": 7.72947808143339e-05, "loss": 1.0242, "step": 91260 }, { "epoch": 0.23, "learning_rate": 7.729352257285219e-05, "loss": 1.0236, "step": 91265 }, { "epoch": 0.23, "learning_rate": 7.729226433137048e-05, "loss": 1.0233, "step": 91270 }, { "epoch": 0.23, "learning_rate": 7.729100608988877e-05, "loss": 1.0242, "step": 91275 }, { "epoch": 0.23, "learning_rate": 7.728974784840707e-05, "loss": 1.0241, "step": 91280 }, { "epoch": 0.23, "learning_rate": 7.728848960692537e-05, "loss": 1.0235, "step": 91285 }, { "epoch": 0.23, "learning_rate": 7.728723136544366e-05, "loss": 1.0242, "step": 91290 }, { "epoch": 0.23, "learning_rate": 7.728597312396195e-05, "loss": 1.0254, "step": 91295 }, { "epoch": 0.23, "learning_rate": 7.728471488248025e-05, "loss": 1.0243, "step": 91300 }, { "epoch": 0.23, "learning_rate": 7.728345664099854e-05, "loss": 1.0238, "step": 91305 }, { "epoch": 0.23, "learning_rate": 7.728219839951684e-05, "loss": 1.0244, "step": 91310 }, { "epoch": 0.23, "learning_rate": 7.728094015803513e-05, "loss": 1.0242, "step": 91315 }, { "epoch": 0.23, "learning_rate": 7.727968191655342e-05, "loss": 1.026, "step": 91320 }, { "epoch": 0.23, "learning_rate": 7.727842367507172e-05, "loss": 1.0226, "step": 91325 }, { "epoch": 0.23, "learning_rate": 7.727716543359002e-05, "loss": 1.0235, "step": 91330 }, { "epoch": 0.23, "learning_rate": 7.727590719210831e-05, "loss": 1.0238, "step": 91335 }, { "epoch": 0.23, "learning_rate": 7.72746489506266e-05, "loss": 1.0235, "step": 91340 }, { "epoch": 0.23, "learning_rate": 7.72733907091449e-05, "loss": 1.0231, "step": 91345 }, { "epoch": 0.23, "learning_rate": 7.72721324676632e-05, "loss": 1.0251, "step": 91350 }, { "epoch": 0.23, "learning_rate": 7.727087422618149e-05, "loss": 1.0241, "step": 91355 }, { "epoch": 0.23, "learning_rate": 7.726961598469978e-05, "loss": 1.0236, "step": 91360 }, { "epoch": 0.23, "learning_rate": 7.726835774321808e-05, "loss": 1.026, "step": 91365 }, { "epoch": 0.23, "learning_rate": 7.726709950173638e-05, "loss": 1.0249, "step": 91370 }, { "epoch": 0.23, "learning_rate": 7.726584126025467e-05, "loss": 1.0233, "step": 91375 }, { "epoch": 0.23, "learning_rate": 7.726458301877296e-05, "loss": 1.0239, "step": 91380 }, { "epoch": 0.23, "learning_rate": 7.726332477729125e-05, "loss": 1.0246, "step": 91385 }, { "epoch": 0.23, "learning_rate": 7.726206653580956e-05, "loss": 1.0215, "step": 91390 }, { "epoch": 0.23, "learning_rate": 7.726080829432785e-05, "loss": 1.025, "step": 91395 }, { "epoch": 0.23, "learning_rate": 7.725955005284614e-05, "loss": 1.0235, "step": 91400 }, { "epoch": 0.23, "learning_rate": 7.725829181136443e-05, "loss": 1.0217, "step": 91405 }, { "epoch": 0.23, "learning_rate": 7.725703356988274e-05, "loss": 1.0224, "step": 91410 }, { "epoch": 0.23, "learning_rate": 7.725577532840103e-05, "loss": 1.0253, "step": 91415 }, { "epoch": 0.23, "learning_rate": 7.725451708691932e-05, "loss": 1.0242, "step": 91420 }, { "epoch": 0.23, "learning_rate": 7.725325884543761e-05, "loss": 1.0224, "step": 91425 }, { "epoch": 0.23, "learning_rate": 7.725200060395592e-05, "loss": 1.0249, "step": 91430 }, { "epoch": 0.23, "learning_rate": 7.725074236247421e-05, "loss": 1.0265, "step": 91435 }, { "epoch": 0.23, "learning_rate": 7.72494841209925e-05, "loss": 1.0244, "step": 91440 }, { "epoch": 0.23, "learning_rate": 7.72482258795108e-05, "loss": 1.0244, "step": 91445 }, { "epoch": 0.23, "learning_rate": 7.72469676380291e-05, "loss": 1.0254, "step": 91450 }, { "epoch": 0.23, "learning_rate": 7.724570939654739e-05, "loss": 1.0258, "step": 91455 }, { "epoch": 0.23, "learning_rate": 7.724445115506569e-05, "loss": 1.0232, "step": 91460 }, { "epoch": 0.23, "learning_rate": 7.724319291358398e-05, "loss": 1.0237, "step": 91465 }, { "epoch": 0.23, "learning_rate": 7.724193467210228e-05, "loss": 1.0233, "step": 91470 }, { "epoch": 0.23, "learning_rate": 7.724067643062057e-05, "loss": 1.0223, "step": 91475 }, { "epoch": 0.23, "learning_rate": 7.723941818913887e-05, "loss": 1.0247, "step": 91480 }, { "epoch": 0.23, "learning_rate": 7.723815994765716e-05, "loss": 1.0208, "step": 91485 }, { "epoch": 0.23, "learning_rate": 7.723690170617546e-05, "loss": 1.0458, "step": 91490 }, { "epoch": 0.23, "learning_rate": 7.723564346469375e-05, "loss": 1.0216, "step": 91495 }, { "epoch": 0.23, "learning_rate": 7.723438522321205e-05, "loss": 1.0463, "step": 91500 }, { "epoch": 0.23, "learning_rate": 7.723312698173034e-05, "loss": 1.0244, "step": 91505 }, { "epoch": 0.23, "learning_rate": 7.723186874024864e-05, "loss": 1.0224, "step": 91510 }, { "epoch": 0.23, "learning_rate": 7.723061049876693e-05, "loss": 1.0226, "step": 91515 }, { "epoch": 0.23, "learning_rate": 7.722935225728522e-05, "loss": 1.0259, "step": 91520 }, { "epoch": 0.23, "learning_rate": 7.722809401580352e-05, "loss": 1.025, "step": 91525 }, { "epoch": 0.23, "learning_rate": 7.722683577432182e-05, "loss": 1.0237, "step": 91530 }, { "epoch": 0.23, "learning_rate": 7.722557753284011e-05, "loss": 1.0207, "step": 91535 }, { "epoch": 0.23, "learning_rate": 7.72243192913584e-05, "loss": 1.0242, "step": 91540 }, { "epoch": 0.23, "learning_rate": 7.72230610498767e-05, "loss": 1.0256, "step": 91545 }, { "epoch": 0.23, "learning_rate": 7.7221802808395e-05, "loss": 1.0221, "step": 91550 }, { "epoch": 0.23, "learning_rate": 7.722054456691329e-05, "loss": 1.0232, "step": 91555 }, { "epoch": 0.23, "learning_rate": 7.721928632543158e-05, "loss": 1.0239, "step": 91560 }, { "epoch": 0.23, "learning_rate": 7.721802808394988e-05, "loss": 1.0236, "step": 91565 }, { "epoch": 0.23, "learning_rate": 7.721676984246818e-05, "loss": 1.0233, "step": 91570 }, { "epoch": 0.23, "learning_rate": 7.721551160098647e-05, "loss": 1.0248, "step": 91575 }, { "epoch": 0.23, "learning_rate": 7.721425335950476e-05, "loss": 1.0215, "step": 91580 }, { "epoch": 0.23, "learning_rate": 7.721299511802305e-05, "loss": 1.0231, "step": 91585 }, { "epoch": 0.23, "learning_rate": 7.721173687654136e-05, "loss": 1.0232, "step": 91590 }, { "epoch": 0.23, "learning_rate": 7.721047863505965e-05, "loss": 1.0244, "step": 91595 }, { "epoch": 0.23, "learning_rate": 7.720922039357794e-05, "loss": 1.0237, "step": 91600 }, { "epoch": 0.23, "learning_rate": 7.720796215209623e-05, "loss": 1.0248, "step": 91605 }, { "epoch": 0.23, "learning_rate": 7.720670391061454e-05, "loss": 1.0247, "step": 91610 }, { "epoch": 0.23, "learning_rate": 7.720544566913283e-05, "loss": 1.0267, "step": 91615 }, { "epoch": 0.23, "learning_rate": 7.720418742765112e-05, "loss": 1.0248, "step": 91620 }, { "epoch": 0.23, "learning_rate": 7.720292918616941e-05, "loss": 1.0262, "step": 91625 }, { "epoch": 0.23, "learning_rate": 7.720167094468772e-05, "loss": 1.0255, "step": 91630 }, { "epoch": 0.23, "learning_rate": 7.720041270320601e-05, "loss": 1.0243, "step": 91635 }, { "epoch": 0.23, "learning_rate": 7.71991544617243e-05, "loss": 1.0217, "step": 91640 }, { "epoch": 0.23, "learning_rate": 7.719789622024259e-05, "loss": 1.0221, "step": 91645 }, { "epoch": 0.23, "learning_rate": 7.719663797876088e-05, "loss": 1.0229, "step": 91650 }, { "epoch": 0.23, "learning_rate": 7.719537973727919e-05, "loss": 1.0242, "step": 91655 }, { "epoch": 0.23, "learning_rate": 7.719412149579748e-05, "loss": 1.0249, "step": 91660 }, { "epoch": 0.23, "learning_rate": 7.719286325431577e-05, "loss": 1.0209, "step": 91665 }, { "epoch": 0.23, "learning_rate": 7.719160501283406e-05, "loss": 1.0256, "step": 91670 }, { "epoch": 0.23, "learning_rate": 7.719034677135237e-05, "loss": 1.0224, "step": 91675 }, { "epoch": 0.23, "learning_rate": 7.718908852987066e-05, "loss": 1.0231, "step": 91680 }, { "epoch": 0.23, "learning_rate": 7.718783028838895e-05, "loss": 1.0241, "step": 91685 }, { "epoch": 0.23, "learning_rate": 7.718657204690724e-05, "loss": 1.0235, "step": 91690 }, { "epoch": 0.23, "learning_rate": 7.718531380542555e-05, "loss": 1.0255, "step": 91695 }, { "epoch": 0.23, "learning_rate": 7.718405556394384e-05, "loss": 1.0249, "step": 91700 }, { "epoch": 0.23, "learning_rate": 7.718279732246213e-05, "loss": 1.023, "step": 91705 }, { "epoch": 0.23, "learning_rate": 7.718153908098042e-05, "loss": 1.0283, "step": 91710 }, { "epoch": 0.23, "learning_rate": 7.718028083949871e-05, "loss": 1.0246, "step": 91715 }, { "epoch": 0.23, "learning_rate": 7.717902259801702e-05, "loss": 1.0224, "step": 91720 }, { "epoch": 0.23, "learning_rate": 7.717776435653531e-05, "loss": 1.0467, "step": 91725 }, { "epoch": 0.23, "learning_rate": 7.71765061150536e-05, "loss": 1.0241, "step": 91730 }, { "epoch": 0.23, "learning_rate": 7.71752478735719e-05, "loss": 1.024, "step": 91735 }, { "epoch": 0.23, "learning_rate": 7.71739896320902e-05, "loss": 1.0243, "step": 91740 }, { "epoch": 0.23, "learning_rate": 7.717273139060849e-05, "loss": 1.0252, "step": 91745 }, { "epoch": 0.23, "learning_rate": 7.717147314912678e-05, "loss": 1.0231, "step": 91750 }, { "epoch": 0.23, "learning_rate": 7.717021490764507e-05, "loss": 1.0221, "step": 91755 }, { "epoch": 0.23, "learning_rate": 7.716895666616338e-05, "loss": 1.0248, "step": 91760 }, { "epoch": 0.23, "learning_rate": 7.716769842468167e-05, "loss": 1.0235, "step": 91765 }, { "epoch": 0.23, "learning_rate": 7.716644018319996e-05, "loss": 1.0243, "step": 91770 }, { "epoch": 0.23, "learning_rate": 7.716518194171825e-05, "loss": 1.025, "step": 91775 }, { "epoch": 0.23, "learning_rate": 7.716392370023654e-05, "loss": 1.0256, "step": 91780 }, { "epoch": 0.23, "learning_rate": 7.716266545875485e-05, "loss": 1.0266, "step": 91785 }, { "epoch": 0.23, "learning_rate": 7.716140721727314e-05, "loss": 1.0434, "step": 91790 }, { "epoch": 0.23, "learning_rate": 7.716014897579143e-05, "loss": 1.0228, "step": 91795 }, { "epoch": 0.23, "learning_rate": 7.715889073430972e-05, "loss": 1.0261, "step": 91800 }, { "epoch": 0.23, "learning_rate": 7.715763249282803e-05, "loss": 1.0255, "step": 91805 }, { "epoch": 0.23, "learning_rate": 7.715637425134632e-05, "loss": 1.0254, "step": 91810 }, { "epoch": 0.23, "learning_rate": 7.715511600986461e-05, "loss": 1.024, "step": 91815 }, { "epoch": 0.23, "learning_rate": 7.71538577683829e-05, "loss": 1.0239, "step": 91820 }, { "epoch": 0.23, "learning_rate": 7.715259952690121e-05, "loss": 1.023, "step": 91825 }, { "epoch": 0.23, "learning_rate": 7.71513412854195e-05, "loss": 1.023, "step": 91830 }, { "epoch": 0.23, "learning_rate": 7.715008304393779e-05, "loss": 1.0221, "step": 91835 }, { "epoch": 0.23, "learning_rate": 7.714882480245608e-05, "loss": 1.0235, "step": 91840 }, { "epoch": 0.23, "learning_rate": 7.714756656097438e-05, "loss": 1.0253, "step": 91845 }, { "epoch": 0.23, "learning_rate": 7.714630831949268e-05, "loss": 1.0258, "step": 91850 }, { "epoch": 0.23, "learning_rate": 7.714505007801097e-05, "loss": 1.0229, "step": 91855 }, { "epoch": 0.23, "learning_rate": 7.714379183652926e-05, "loss": 1.0217, "step": 91860 }, { "epoch": 0.23, "learning_rate": 7.714253359504756e-05, "loss": 1.0212, "step": 91865 }, { "epoch": 0.23, "learning_rate": 7.714127535356586e-05, "loss": 1.025, "step": 91870 }, { "epoch": 0.23, "learning_rate": 7.714001711208415e-05, "loss": 1.0252, "step": 91875 }, { "epoch": 0.23, "learning_rate": 7.713875887060244e-05, "loss": 1.0231, "step": 91880 }, { "epoch": 0.23, "learning_rate": 7.713750062912074e-05, "loss": 1.024, "step": 91885 }, { "epoch": 0.23, "learning_rate": 7.713624238763904e-05, "loss": 1.0246, "step": 91890 }, { "epoch": 0.23, "learning_rate": 7.713498414615733e-05, "loss": 1.0234, "step": 91895 }, { "epoch": 0.23, "learning_rate": 7.713372590467562e-05, "loss": 1.0264, "step": 91900 }, { "epoch": 0.23, "learning_rate": 7.713246766319392e-05, "loss": 1.0225, "step": 91905 }, { "epoch": 0.23, "learning_rate": 7.713120942171221e-05, "loss": 1.0239, "step": 91910 }, { "epoch": 0.23, "learning_rate": 7.712995118023051e-05, "loss": 1.0225, "step": 91915 }, { "epoch": 0.23, "learning_rate": 7.71286929387488e-05, "loss": 1.0241, "step": 91920 }, { "epoch": 0.23, "learning_rate": 7.71274346972671e-05, "loss": 1.0264, "step": 91925 }, { "epoch": 0.23, "learning_rate": 7.712617645578539e-05, "loss": 1.0225, "step": 91930 }, { "epoch": 0.23, "learning_rate": 7.712491821430369e-05, "loss": 1.0494, "step": 91935 }, { "epoch": 0.23, "learning_rate": 7.712365997282198e-05, "loss": 1.0224, "step": 91940 }, { "epoch": 0.23, "learning_rate": 7.712240173134029e-05, "loss": 1.023, "step": 91945 }, { "epoch": 0.23, "learning_rate": 7.712114348985858e-05, "loss": 1.0243, "step": 91950 }, { "epoch": 0.23, "learning_rate": 7.711988524837687e-05, "loss": 1.0237, "step": 91955 }, { "epoch": 0.23, "learning_rate": 7.711862700689518e-05, "loss": 1.0429, "step": 91960 }, { "epoch": 0.23, "learning_rate": 7.711736876541347e-05, "loss": 1.0234, "step": 91965 }, { "epoch": 0.23, "learning_rate": 7.711611052393176e-05, "loss": 1.0247, "step": 91970 }, { "epoch": 0.23, "learning_rate": 7.711485228245005e-05, "loss": 1.0232, "step": 91975 }, { "epoch": 0.23, "learning_rate": 7.711359404096834e-05, "loss": 1.0214, "step": 91980 }, { "epoch": 0.23, "learning_rate": 7.711233579948665e-05, "loss": 1.025, "step": 91985 }, { "epoch": 0.23, "learning_rate": 7.711107755800494e-05, "loss": 1.0242, "step": 91990 }, { "epoch": 0.23, "learning_rate": 7.710981931652323e-05, "loss": 1.024, "step": 91995 }, { "epoch": 0.23, "learning_rate": 7.710856107504152e-05, "loss": 1.0234, "step": 92000 }, { "epoch": 0.23, "learning_rate": 7.710730283355983e-05, "loss": 1.0241, "step": 92005 }, { "epoch": 0.23, "learning_rate": 7.710604459207812e-05, "loss": 1.0226, "step": 92010 }, { "epoch": 0.23, "learning_rate": 7.710478635059641e-05, "loss": 1.0231, "step": 92015 }, { "epoch": 0.23, "learning_rate": 7.71035281091147e-05, "loss": 1.0234, "step": 92020 }, { "epoch": 0.23, "learning_rate": 7.710226986763301e-05, "loss": 1.0262, "step": 92025 }, { "epoch": 0.23, "learning_rate": 7.71010116261513e-05, "loss": 1.0243, "step": 92030 }, { "epoch": 0.23, "learning_rate": 7.709975338466959e-05, "loss": 1.0247, "step": 92035 }, { "epoch": 0.23, "learning_rate": 7.709849514318788e-05, "loss": 1.0224, "step": 92040 }, { "epoch": 0.23, "learning_rate": 7.709723690170618e-05, "loss": 1.0206, "step": 92045 }, { "epoch": 0.23, "learning_rate": 7.709597866022448e-05, "loss": 1.0228, "step": 92050 }, { "epoch": 0.23, "learning_rate": 7.709472041874277e-05, "loss": 1.024, "step": 92055 }, { "epoch": 0.23, "learning_rate": 7.709346217726106e-05, "loss": 1.0244, "step": 92060 }, { "epoch": 0.23, "learning_rate": 7.709220393577936e-05, "loss": 1.0232, "step": 92065 }, { "epoch": 0.23, "learning_rate": 7.709094569429766e-05, "loss": 1.0231, "step": 92070 }, { "epoch": 0.23, "learning_rate": 7.708968745281595e-05, "loss": 1.0239, "step": 92075 }, { "epoch": 0.23, "learning_rate": 7.708842921133424e-05, "loss": 1.0228, "step": 92080 }, { "epoch": 0.23, "learning_rate": 7.708717096985254e-05, "loss": 1.0242, "step": 92085 }, { "epoch": 0.23, "learning_rate": 7.708591272837084e-05, "loss": 1.0248, "step": 92090 }, { "epoch": 0.23, "learning_rate": 7.708465448688913e-05, "loss": 1.0229, "step": 92095 }, { "epoch": 0.23, "learning_rate": 7.708339624540742e-05, "loss": 1.0221, "step": 92100 }, { "epoch": 0.23, "learning_rate": 7.708213800392572e-05, "loss": 1.025, "step": 92105 }, { "epoch": 0.23, "learning_rate": 7.708087976244401e-05, "loss": 1.0224, "step": 92110 }, { "epoch": 0.23, "learning_rate": 7.707962152096231e-05, "loss": 1.0229, "step": 92115 }, { "epoch": 0.23, "learning_rate": 7.70783632794806e-05, "loss": 1.0248, "step": 92120 }, { "epoch": 0.23, "learning_rate": 7.70771050379989e-05, "loss": 1.0236, "step": 92125 }, { "epoch": 0.23, "learning_rate": 7.707584679651719e-05, "loss": 1.0233, "step": 92130 }, { "epoch": 0.23, "learning_rate": 7.707458855503549e-05, "loss": 1.0229, "step": 92135 }, { "epoch": 0.23, "learning_rate": 7.707333031355378e-05, "loss": 1.0239, "step": 92140 }, { "epoch": 0.23, "learning_rate": 7.707207207207208e-05, "loss": 1.0209, "step": 92145 }, { "epoch": 0.23, "learning_rate": 7.707081383059037e-05, "loss": 1.0255, "step": 92150 }, { "epoch": 0.23, "learning_rate": 7.706955558910867e-05, "loss": 1.0238, "step": 92155 }, { "epoch": 0.23, "learning_rate": 7.706829734762696e-05, "loss": 1.0247, "step": 92160 }, { "epoch": 0.23, "learning_rate": 7.706703910614526e-05, "loss": 1.0231, "step": 92165 }, { "epoch": 0.23, "learning_rate": 7.706578086466355e-05, "loss": 1.0245, "step": 92170 }, { "epoch": 0.23, "learning_rate": 7.706452262318184e-05, "loss": 1.0228, "step": 92175 }, { "epoch": 0.23, "learning_rate": 7.706326438170014e-05, "loss": 1.0251, "step": 92180 }, { "epoch": 0.23, "learning_rate": 7.706200614021844e-05, "loss": 1.0259, "step": 92185 }, { "epoch": 0.23, "learning_rate": 7.706074789873673e-05, "loss": 1.0258, "step": 92190 }, { "epoch": 0.23, "learning_rate": 7.705948965725502e-05, "loss": 1.0215, "step": 92195 }, { "epoch": 0.23, "learning_rate": 7.705823141577332e-05, "loss": 1.024, "step": 92200 }, { "epoch": 0.23, "learning_rate": 7.705697317429162e-05, "loss": 1.021, "step": 92205 }, { "epoch": 0.23, "learning_rate": 7.705571493280991e-05, "loss": 1.0218, "step": 92210 }, { "epoch": 0.23, "learning_rate": 7.70544566913282e-05, "loss": 1.0227, "step": 92215 }, { "epoch": 0.23, "learning_rate": 7.70531984498465e-05, "loss": 1.026, "step": 92220 }, { "epoch": 0.23, "learning_rate": 7.70519402083648e-05, "loss": 1.0217, "step": 92225 }, { "epoch": 0.23, "learning_rate": 7.705068196688309e-05, "loss": 1.0233, "step": 92230 }, { "epoch": 0.23, "learning_rate": 7.704967537369772e-05, "loss": 1.0257, "step": 92235 }, { "epoch": 0.23, "learning_rate": 7.704841713221603e-05, "loss": 1.0238, "step": 92240 }, { "epoch": 0.23, "learning_rate": 7.704715889073432e-05, "loss": 1.0226, "step": 92245 }, { "epoch": 0.23, "learning_rate": 7.704590064925261e-05, "loss": 1.024, "step": 92250 }, { "epoch": 0.23, "learning_rate": 7.70446424077709e-05, "loss": 1.0246, "step": 92255 }, { "epoch": 0.23, "learning_rate": 7.70433841662892e-05, "loss": 1.024, "step": 92260 }, { "epoch": 0.23, "learning_rate": 7.70421259248075e-05, "loss": 1.0219, "step": 92265 }, { "epoch": 0.23, "learning_rate": 7.704086768332579e-05, "loss": 1.0223, "step": 92270 }, { "epoch": 0.23, "learning_rate": 7.703960944184408e-05, "loss": 1.0238, "step": 92275 }, { "epoch": 0.23, "learning_rate": 7.703835120036237e-05, "loss": 1.0231, "step": 92280 }, { "epoch": 0.23, "learning_rate": 7.703709295888068e-05, "loss": 1.0233, "step": 92285 }, { "epoch": 0.23, "learning_rate": 7.703583471739897e-05, "loss": 1.0234, "step": 92290 }, { "epoch": 0.23, "learning_rate": 7.703457647591726e-05, "loss": 1.0235, "step": 92295 }, { "epoch": 0.23, "learning_rate": 7.703331823443555e-05, "loss": 1.0244, "step": 92300 }, { "epoch": 0.23, "learning_rate": 7.703205999295386e-05, "loss": 1.0244, "step": 92305 }, { "epoch": 0.23, "learning_rate": 7.703080175147215e-05, "loss": 1.0232, "step": 92310 }, { "epoch": 0.23, "learning_rate": 7.702954350999044e-05, "loss": 1.0227, "step": 92315 }, { "epoch": 0.23, "learning_rate": 7.702828526850873e-05, "loss": 1.0256, "step": 92320 }, { "epoch": 0.23, "learning_rate": 7.702702702702703e-05, "loss": 1.0253, "step": 92325 }, { "epoch": 0.23, "learning_rate": 7.702576878554533e-05, "loss": 1.0231, "step": 92330 }, { "epoch": 0.23, "learning_rate": 7.702451054406362e-05, "loss": 1.0253, "step": 92335 }, { "epoch": 0.23, "learning_rate": 7.702325230258191e-05, "loss": 1.027, "step": 92340 }, { "epoch": 0.23, "learning_rate": 7.70219940611002e-05, "loss": 1.0242, "step": 92345 }, { "epoch": 0.23, "learning_rate": 7.702073581961851e-05, "loss": 1.0245, "step": 92350 }, { "epoch": 0.23, "learning_rate": 7.70194775781368e-05, "loss": 1.0239, "step": 92355 }, { "epoch": 0.23, "learning_rate": 7.70182193366551e-05, "loss": 1.0245, "step": 92360 }, { "epoch": 0.23, "learning_rate": 7.701696109517339e-05, "loss": 1.0222, "step": 92365 }, { "epoch": 0.23, "learning_rate": 7.701570285369169e-05, "loss": 1.0469, "step": 92370 }, { "epoch": 0.23, "learning_rate": 7.701444461220998e-05, "loss": 1.0268, "step": 92375 }, { "epoch": 0.23, "learning_rate": 7.701318637072827e-05, "loss": 1.0248, "step": 92380 }, { "epoch": 0.23, "learning_rate": 7.701192812924657e-05, "loss": 1.024, "step": 92385 }, { "epoch": 0.23, "learning_rate": 7.701066988776486e-05, "loss": 1.0234, "step": 92390 }, { "epoch": 0.23, "learning_rate": 7.700941164628316e-05, "loss": 1.0225, "step": 92395 }, { "epoch": 0.23, "learning_rate": 7.700815340480145e-05, "loss": 1.0221, "step": 92400 }, { "epoch": 0.23, "learning_rate": 7.700689516331975e-05, "loss": 1.0243, "step": 92405 }, { "epoch": 0.23, "learning_rate": 7.700563692183804e-05, "loss": 1.024, "step": 92410 }, { "epoch": 0.23, "learning_rate": 7.700437868035634e-05, "loss": 1.0233, "step": 92415 }, { "epoch": 0.23, "learning_rate": 7.700312043887463e-05, "loss": 1.0463, "step": 92420 }, { "epoch": 0.23, "learning_rate": 7.700186219739293e-05, "loss": 1.0245, "step": 92425 }, { "epoch": 0.23, "learning_rate": 7.700060395591122e-05, "loss": 1.024, "step": 92430 }, { "epoch": 0.23, "learning_rate": 7.699934571442952e-05, "loss": 1.0263, "step": 92435 }, { "epoch": 0.23, "learning_rate": 7.699808747294781e-05, "loss": 1.0239, "step": 92440 }, { "epoch": 0.23, "learning_rate": 7.69968292314661e-05, "loss": 1.0232, "step": 92445 }, { "epoch": 0.23, "learning_rate": 7.69955709899844e-05, "loss": 1.0275, "step": 92450 }, { "epoch": 0.23, "learning_rate": 7.699431274850269e-05, "loss": 1.0218, "step": 92455 }, { "epoch": 0.23, "learning_rate": 7.6993054507021e-05, "loss": 1.0239, "step": 92460 }, { "epoch": 0.23, "learning_rate": 7.699179626553929e-05, "loss": 1.0249, "step": 92465 }, { "epoch": 0.23, "learning_rate": 7.699053802405758e-05, "loss": 1.0218, "step": 92470 }, { "epoch": 0.23, "learning_rate": 7.698927978257587e-05, "loss": 1.0253, "step": 92475 }, { "epoch": 0.23, "learning_rate": 7.698802154109417e-05, "loss": 1.024, "step": 92480 }, { "epoch": 0.23, "learning_rate": 7.698676329961247e-05, "loss": 1.0228, "step": 92485 }, { "epoch": 0.23, "learning_rate": 7.698550505813076e-05, "loss": 1.0259, "step": 92490 }, { "epoch": 0.23, "learning_rate": 7.698424681664905e-05, "loss": 1.0241, "step": 92495 }, { "epoch": 0.23, "learning_rate": 7.698298857516735e-05, "loss": 1.0236, "step": 92500 }, { "epoch": 0.23, "learning_rate": 7.698173033368565e-05, "loss": 1.0227, "step": 92505 }, { "epoch": 0.23, "learning_rate": 7.698047209220394e-05, "loss": 1.0237, "step": 92510 }, { "epoch": 0.23, "learning_rate": 7.697921385072223e-05, "loss": 1.0248, "step": 92515 }, { "epoch": 0.23, "learning_rate": 7.697795560924052e-05, "loss": 1.024, "step": 92520 }, { "epoch": 0.23, "learning_rate": 7.697669736775883e-05, "loss": 1.0256, "step": 92525 }, { "epoch": 0.23, "learning_rate": 7.697543912627712e-05, "loss": 1.0229, "step": 92530 }, { "epoch": 0.23, "learning_rate": 7.697418088479541e-05, "loss": 1.0229, "step": 92535 }, { "epoch": 0.23, "learning_rate": 7.69729226433137e-05, "loss": 1.0325, "step": 92540 }, { "epoch": 0.23, "learning_rate": 7.6971664401832e-05, "loss": 1.0225, "step": 92545 }, { "epoch": 0.23, "learning_rate": 7.69704061603503e-05, "loss": 1.0241, "step": 92550 }, { "epoch": 0.23, "learning_rate": 7.696914791886859e-05, "loss": 1.0239, "step": 92555 }, { "epoch": 0.23, "learning_rate": 7.696788967738688e-05, "loss": 1.0222, "step": 92560 }, { "epoch": 0.23, "learning_rate": 7.696663143590519e-05, "loss": 1.0227, "step": 92565 }, { "epoch": 0.23, "learning_rate": 7.696537319442348e-05, "loss": 1.0253, "step": 92570 }, { "epoch": 0.23, "learning_rate": 7.696411495294177e-05, "loss": 1.023, "step": 92575 }, { "epoch": 0.23, "learning_rate": 7.696285671146006e-05, "loss": 1.0246, "step": 92580 }, { "epoch": 0.23, "learning_rate": 7.696159846997835e-05, "loss": 1.0256, "step": 92585 }, { "epoch": 0.23, "learning_rate": 7.696034022849666e-05, "loss": 1.0403, "step": 92590 }, { "epoch": 0.23, "learning_rate": 7.695908198701495e-05, "loss": 1.0222, "step": 92595 }, { "epoch": 0.23, "learning_rate": 7.695782374553324e-05, "loss": 1.0261, "step": 92600 }, { "epoch": 0.23, "learning_rate": 7.695656550405153e-05, "loss": 1.0448, "step": 92605 }, { "epoch": 0.23, "learning_rate": 7.695530726256984e-05, "loss": 1.0233, "step": 92610 }, { "epoch": 0.23, "learning_rate": 7.695404902108813e-05, "loss": 1.023, "step": 92615 }, { "epoch": 0.23, "learning_rate": 7.695279077960642e-05, "loss": 1.0238, "step": 92620 }, { "epoch": 0.23, "learning_rate": 7.695153253812471e-05, "loss": 1.0252, "step": 92625 }, { "epoch": 0.23, "learning_rate": 7.695027429664302e-05, "loss": 1.0232, "step": 92630 }, { "epoch": 0.23, "learning_rate": 7.694901605516131e-05, "loss": 1.046, "step": 92635 }, { "epoch": 0.23, "learning_rate": 7.69477578136796e-05, "loss": 1.0216, "step": 92640 }, { "epoch": 0.23, "learning_rate": 7.69464995721979e-05, "loss": 1.0238, "step": 92645 }, { "epoch": 0.23, "learning_rate": 7.69452413307162e-05, "loss": 1.0231, "step": 92650 }, { "epoch": 0.23, "learning_rate": 7.69439830892345e-05, "loss": 1.025, "step": 92655 }, { "epoch": 0.23, "learning_rate": 7.694272484775279e-05, "loss": 1.0257, "step": 92660 }, { "epoch": 0.23, "learning_rate": 7.694146660627108e-05, "loss": 1.0236, "step": 92665 }, { "epoch": 0.23, "learning_rate": 7.694020836478938e-05, "loss": 1.0235, "step": 92670 }, { "epoch": 0.23, "learning_rate": 7.693895012330767e-05, "loss": 1.0251, "step": 92675 }, { "epoch": 0.23, "learning_rate": 7.693769188182597e-05, "loss": 1.0243, "step": 92680 }, { "epoch": 0.23, "learning_rate": 7.693643364034426e-05, "loss": 1.023, "step": 92685 }, { "epoch": 0.23, "learning_rate": 7.693517539886256e-05, "loss": 1.025, "step": 92690 }, { "epoch": 0.23, "learning_rate": 7.693391715738085e-05, "loss": 1.0248, "step": 92695 }, { "epoch": 0.23, "learning_rate": 7.693265891589915e-05, "loss": 1.0217, "step": 92700 }, { "epoch": 0.23, "learning_rate": 7.693140067441744e-05, "loss": 1.0246, "step": 92705 }, { "epoch": 0.23, "learning_rate": 7.693014243293574e-05, "loss": 1.0218, "step": 92710 }, { "epoch": 0.23, "learning_rate": 7.692888419145403e-05, "loss": 1.0216, "step": 92715 }, { "epoch": 0.23, "learning_rate": 7.692762594997232e-05, "loss": 1.0228, "step": 92720 }, { "epoch": 0.23, "learning_rate": 7.692636770849062e-05, "loss": 1.0246, "step": 92725 }, { "epoch": 0.23, "learning_rate": 7.692510946700892e-05, "loss": 1.0232, "step": 92730 }, { "epoch": 0.23, "learning_rate": 7.692385122552721e-05, "loss": 1.0235, "step": 92735 }, { "epoch": 0.23, "learning_rate": 7.69225929840455e-05, "loss": 1.0217, "step": 92740 }, { "epoch": 0.23, "learning_rate": 7.69213347425638e-05, "loss": 1.0246, "step": 92745 }, { "epoch": 0.23, "learning_rate": 7.69200765010821e-05, "loss": 1.0243, "step": 92750 }, { "epoch": 0.23, "learning_rate": 7.691881825960039e-05, "loss": 1.0249, "step": 92755 }, { "epoch": 0.23, "learning_rate": 7.691756001811868e-05, "loss": 1.0248, "step": 92760 }, { "epoch": 0.23, "learning_rate": 7.691630177663698e-05, "loss": 1.0243, "step": 92765 }, { "epoch": 0.23, "learning_rate": 7.691504353515528e-05, "loss": 1.0247, "step": 92770 }, { "epoch": 0.23, "learning_rate": 7.691378529367357e-05, "loss": 1.0243, "step": 92775 }, { "epoch": 0.23, "learning_rate": 7.691252705219186e-05, "loss": 1.0456, "step": 92780 }, { "epoch": 0.23, "learning_rate": 7.691126881071015e-05, "loss": 1.0249, "step": 92785 }, { "epoch": 0.23, "learning_rate": 7.691001056922846e-05, "loss": 1.023, "step": 92790 }, { "epoch": 0.23, "learning_rate": 7.690875232774675e-05, "loss": 1.0232, "step": 92795 }, { "epoch": 0.23, "learning_rate": 7.690749408626504e-05, "loss": 1.0258, "step": 92800 }, { "epoch": 0.23, "learning_rate": 7.690623584478333e-05, "loss": 1.0226, "step": 92805 }, { "epoch": 0.23, "learning_rate": 7.690497760330164e-05, "loss": 1.0226, "step": 92810 }, { "epoch": 0.23, "learning_rate": 7.690371936181993e-05, "loss": 1.0215, "step": 92815 }, { "epoch": 0.23, "learning_rate": 7.690246112033822e-05, "loss": 1.0253, "step": 92820 }, { "epoch": 0.23, "learning_rate": 7.690120287885651e-05, "loss": 1.0265, "step": 92825 }, { "epoch": 0.23, "learning_rate": 7.689994463737482e-05, "loss": 1.0233, "step": 92830 }, { "epoch": 0.23, "learning_rate": 7.689868639589311e-05, "loss": 1.0257, "step": 92835 }, { "epoch": 0.23, "learning_rate": 7.68974281544114e-05, "loss": 1.0227, "step": 92840 }, { "epoch": 0.23, "learning_rate": 7.689616991292969e-05, "loss": 1.022, "step": 92845 }, { "epoch": 0.23, "learning_rate": 7.689491167144798e-05, "loss": 1.0254, "step": 92850 }, { "epoch": 0.23, "learning_rate": 7.689365342996629e-05, "loss": 1.0235, "step": 92855 }, { "epoch": 0.23, "learning_rate": 7.689239518848458e-05, "loss": 1.0224, "step": 92860 }, { "epoch": 0.23, "learning_rate": 7.689113694700287e-05, "loss": 1.0229, "step": 92865 }, { "epoch": 0.23, "learning_rate": 7.688987870552116e-05, "loss": 1.0232, "step": 92870 }, { "epoch": 0.23, "learning_rate": 7.688862046403947e-05, "loss": 1.041, "step": 92875 }, { "epoch": 0.23, "learning_rate": 7.688736222255776e-05, "loss": 1.0235, "step": 92880 }, { "epoch": 0.23, "learning_rate": 7.688610398107605e-05, "loss": 1.0214, "step": 92885 }, { "epoch": 0.23, "learning_rate": 7.688484573959434e-05, "loss": 1.0246, "step": 92890 }, { "epoch": 0.23, "learning_rate": 7.688358749811265e-05, "loss": 1.0228, "step": 92895 }, { "epoch": 0.23, "learning_rate": 7.688232925663094e-05, "loss": 1.0259, "step": 92900 }, { "epoch": 0.23, "learning_rate": 7.688107101514923e-05, "loss": 1.0233, "step": 92905 }, { "epoch": 0.23, "learning_rate": 7.687981277366752e-05, "loss": 1.0242, "step": 92910 }, { "epoch": 0.23, "learning_rate": 7.687855453218581e-05, "loss": 1.0245, "step": 92915 }, { "epoch": 0.23, "learning_rate": 7.687729629070412e-05, "loss": 1.0237, "step": 92920 }, { "epoch": 0.23, "learning_rate": 7.687603804922241e-05, "loss": 1.024, "step": 92925 }, { "epoch": 0.23, "learning_rate": 7.68747798077407e-05, "loss": 1.0251, "step": 92930 }, { "epoch": 0.23, "learning_rate": 7.6873521566259e-05, "loss": 1.0241, "step": 92935 }, { "epoch": 0.23, "learning_rate": 7.68722633247773e-05, "loss": 1.023, "step": 92940 }, { "epoch": 0.23, "learning_rate": 7.687100508329559e-05, "loss": 1.0245, "step": 92945 }, { "epoch": 0.23, "learning_rate": 7.686974684181388e-05, "loss": 1.0231, "step": 92950 }, { "epoch": 0.23, "learning_rate": 7.686848860033217e-05, "loss": 1.0249, "step": 92955 }, { "epoch": 0.23, "learning_rate": 7.686723035885048e-05, "loss": 1.0242, "step": 92960 }, { "epoch": 0.23, "learning_rate": 7.686597211736877e-05, "loss": 1.0242, "step": 92965 }, { "epoch": 0.23, "learning_rate": 7.686471387588706e-05, "loss": 1.0226, "step": 92970 }, { "epoch": 0.23, "learning_rate": 7.686345563440535e-05, "loss": 1.0217, "step": 92975 }, { "epoch": 0.23, "learning_rate": 7.686219739292365e-05, "loss": 1.0227, "step": 92980 }, { "epoch": 0.23, "learning_rate": 7.686093915144195e-05, "loss": 1.021, "step": 92985 }, { "epoch": 0.23, "learning_rate": 7.685968090996024e-05, "loss": 1.0237, "step": 92990 }, { "epoch": 0.23, "learning_rate": 7.685842266847853e-05, "loss": 1.0247, "step": 92995 }, { "epoch": 0.23, "learning_rate": 7.685716442699683e-05, "loss": 1.0212, "step": 93000 }, { "epoch": 0.23, "learning_rate": 7.685590618551513e-05, "loss": 1.0247, "step": 93005 }, { "epoch": 0.23, "learning_rate": 7.685464794403342e-05, "loss": 1.0243, "step": 93010 }, { "epoch": 0.23, "learning_rate": 7.685338970255171e-05, "loss": 1.0243, "step": 93015 }, { "epoch": 0.23, "learning_rate": 7.685213146107e-05, "loss": 1.0257, "step": 93020 }, { "epoch": 0.23, "learning_rate": 7.685087321958831e-05, "loss": 1.0245, "step": 93025 }, { "epoch": 0.23, "learning_rate": 7.68496149781066e-05, "loss": 1.0218, "step": 93030 }, { "epoch": 0.23, "learning_rate": 7.68483567366249e-05, "loss": 1.0233, "step": 93035 }, { "epoch": 0.23, "learning_rate": 7.684709849514318e-05, "loss": 1.0221, "step": 93040 }, { "epoch": 0.23, "learning_rate": 7.684609190195783e-05, "loss": 1.0228, "step": 93045 }, { "epoch": 0.23, "learning_rate": 7.684483366047613e-05, "loss": 1.0217, "step": 93050 }, { "epoch": 0.23, "learning_rate": 7.684357541899442e-05, "loss": 1.0259, "step": 93055 }, { "epoch": 0.23, "learning_rate": 7.684231717751271e-05, "loss": 1.0247, "step": 93060 }, { "epoch": 0.23, "learning_rate": 7.6841058936031e-05, "loss": 1.0226, "step": 93065 }, { "epoch": 0.23, "learning_rate": 7.68398006945493e-05, "loss": 1.0244, "step": 93070 }, { "epoch": 0.23, "learning_rate": 7.68385424530676e-05, "loss": 1.0485, "step": 93075 }, { "epoch": 0.23, "learning_rate": 7.683728421158589e-05, "loss": 1.024, "step": 93080 }, { "epoch": 0.23, "learning_rate": 7.683602597010418e-05, "loss": 1.0225, "step": 93085 }, { "epoch": 0.23, "learning_rate": 7.683476772862249e-05, "loss": 1.0214, "step": 93090 }, { "epoch": 0.23, "learning_rate": 7.683350948714078e-05, "loss": 1.0244, "step": 93095 }, { "epoch": 0.23, "learning_rate": 7.683225124565907e-05, "loss": 1.0233, "step": 93100 }, { "epoch": 0.23, "learning_rate": 7.683099300417736e-05, "loss": 1.023, "step": 93105 }, { "epoch": 0.23, "learning_rate": 7.682973476269567e-05, "loss": 1.0247, "step": 93110 }, { "epoch": 0.23, "learning_rate": 7.682847652121396e-05, "loss": 1.023, "step": 93115 }, { "epoch": 0.23, "learning_rate": 7.682721827973225e-05, "loss": 1.0219, "step": 93120 }, { "epoch": 0.23, "learning_rate": 7.682596003825054e-05, "loss": 1.0226, "step": 93125 }, { "epoch": 0.23, "learning_rate": 7.682470179676883e-05, "loss": 1.0222, "step": 93130 }, { "epoch": 0.23, "learning_rate": 7.682344355528714e-05, "loss": 1.0255, "step": 93135 }, { "epoch": 0.23, "learning_rate": 7.682218531380543e-05, "loss": 1.0245, "step": 93140 }, { "epoch": 0.23, "learning_rate": 7.682092707232372e-05, "loss": 1.023, "step": 93145 }, { "epoch": 0.23, "learning_rate": 7.681966883084201e-05, "loss": 1.0264, "step": 93150 }, { "epoch": 0.23, "learning_rate": 7.681841058936032e-05, "loss": 1.0239, "step": 93155 }, { "epoch": 0.23, "learning_rate": 7.681715234787861e-05, "loss": 1.0221, "step": 93160 }, { "epoch": 0.23, "learning_rate": 7.68158941063969e-05, "loss": 1.0251, "step": 93165 }, { "epoch": 0.23, "learning_rate": 7.681463586491519e-05, "loss": 1.0217, "step": 93170 }, { "epoch": 0.23, "learning_rate": 7.68133776234335e-05, "loss": 1.0228, "step": 93175 }, { "epoch": 0.23, "learning_rate": 7.681211938195179e-05, "loss": 1.0228, "step": 93180 }, { "epoch": 0.23, "learning_rate": 7.681086114047008e-05, "loss": 1.0215, "step": 93185 }, { "epoch": 0.23, "learning_rate": 7.680960289898837e-05, "loss": 1.0233, "step": 93190 }, { "epoch": 0.23, "learning_rate": 7.680834465750666e-05, "loss": 1.0235, "step": 93195 }, { "epoch": 0.23, "learning_rate": 7.680708641602497e-05, "loss": 1.0263, "step": 93200 }, { "epoch": 0.23, "learning_rate": 7.680582817454326e-05, "loss": 1.0256, "step": 93205 }, { "epoch": 0.23, "learning_rate": 7.680456993306155e-05, "loss": 1.0217, "step": 93210 }, { "epoch": 0.23, "learning_rate": 7.680331169157984e-05, "loss": 1.0236, "step": 93215 }, { "epoch": 0.23, "learning_rate": 7.680205345009815e-05, "loss": 1.0232, "step": 93220 }, { "epoch": 0.23, "learning_rate": 7.680079520861644e-05, "loss": 1.024, "step": 93225 }, { "epoch": 0.23, "learning_rate": 7.679953696713473e-05, "loss": 1.0415, "step": 93230 }, { "epoch": 0.23, "learning_rate": 7.679827872565302e-05, "loss": 1.0229, "step": 93235 }, { "epoch": 0.23, "learning_rate": 7.679702048417133e-05, "loss": 1.0249, "step": 93240 }, { "epoch": 0.23, "learning_rate": 7.679576224268962e-05, "loss": 1.0243, "step": 93245 }, { "epoch": 0.23, "learning_rate": 7.679450400120791e-05, "loss": 1.0254, "step": 93250 }, { "epoch": 0.23, "learning_rate": 7.67932457597262e-05, "loss": 1.0222, "step": 93255 }, { "epoch": 0.23, "learning_rate": 7.67919875182445e-05, "loss": 1.0233, "step": 93260 }, { "epoch": 0.23, "learning_rate": 7.67907292767628e-05, "loss": 1.0223, "step": 93265 }, { "epoch": 0.23, "learning_rate": 7.678947103528109e-05, "loss": 1.0207, "step": 93270 }, { "epoch": 0.23, "learning_rate": 7.678821279379938e-05, "loss": 1.0415, "step": 93275 }, { "epoch": 0.23, "learning_rate": 7.678695455231768e-05, "loss": 1.0232, "step": 93280 }, { "epoch": 0.23, "learning_rate": 7.678569631083598e-05, "loss": 1.0229, "step": 93285 }, { "epoch": 0.23, "learning_rate": 7.678443806935427e-05, "loss": 1.0235, "step": 93290 }, { "epoch": 0.23, "learning_rate": 7.678317982787256e-05, "loss": 1.0241, "step": 93295 }, { "epoch": 0.23, "learning_rate": 7.678192158639086e-05, "loss": 1.0276, "step": 93300 }, { "epoch": 0.23, "learning_rate": 7.678066334490916e-05, "loss": 1.0243, "step": 93305 }, { "epoch": 0.23, "learning_rate": 7.677940510342745e-05, "loss": 1.0223, "step": 93310 }, { "epoch": 0.23, "learning_rate": 7.677814686194574e-05, "loss": 1.0242, "step": 93315 }, { "epoch": 0.23, "learning_rate": 7.677688862046404e-05, "loss": 1.0216, "step": 93320 }, { "epoch": 0.23, "learning_rate": 7.677563037898233e-05, "loss": 1.0244, "step": 93325 }, { "epoch": 0.23, "learning_rate": 7.677437213750063e-05, "loss": 1.023, "step": 93330 }, { "epoch": 0.23, "learning_rate": 7.677311389601892e-05, "loss": 1.0234, "step": 93335 }, { "epoch": 0.23, "learning_rate": 7.677185565453721e-05, "loss": 1.0207, "step": 93340 }, { "epoch": 0.23, "learning_rate": 7.677059741305552e-05, "loss": 1.0235, "step": 93345 }, { "epoch": 0.23, "learning_rate": 7.676933917157381e-05, "loss": 1.043, "step": 93350 }, { "epoch": 0.23, "learning_rate": 7.676808093009212e-05, "loss": 1.025, "step": 93355 }, { "epoch": 0.23, "learning_rate": 7.676682268861041e-05, "loss": 1.0241, "step": 93360 }, { "epoch": 0.23, "learning_rate": 7.67655644471287e-05, "loss": 1.0249, "step": 93365 }, { "epoch": 0.23, "learning_rate": 7.676430620564699e-05, "loss": 1.0242, "step": 93370 }, { "epoch": 0.23, "learning_rate": 7.67630479641653e-05, "loss": 1.0233, "step": 93375 }, { "epoch": 0.23, "learning_rate": 7.676178972268359e-05, "loss": 1.0237, "step": 93380 }, { "epoch": 0.23, "learning_rate": 7.676053148120188e-05, "loss": 1.0248, "step": 93385 }, { "epoch": 0.23, "learning_rate": 7.675927323972017e-05, "loss": 1.0216, "step": 93390 }, { "epoch": 0.23, "learning_rate": 7.675801499823848e-05, "loss": 1.0263, "step": 93395 }, { "epoch": 0.23, "learning_rate": 7.675675675675677e-05, "loss": 1.0236, "step": 93400 }, { "epoch": 0.23, "learning_rate": 7.675549851527506e-05, "loss": 1.0237, "step": 93405 }, { "epoch": 0.23, "learning_rate": 7.675424027379335e-05, "loss": 1.0237, "step": 93410 }, { "epoch": 0.23, "learning_rate": 7.675298203231164e-05, "loss": 1.021, "step": 93415 }, { "epoch": 0.23, "learning_rate": 7.675172379082995e-05, "loss": 1.0231, "step": 93420 }, { "epoch": 0.23, "learning_rate": 7.675046554934824e-05, "loss": 1.0213, "step": 93425 }, { "epoch": 0.23, "learning_rate": 7.674920730786653e-05, "loss": 1.0235, "step": 93430 }, { "epoch": 0.23, "learning_rate": 7.674794906638482e-05, "loss": 1.0235, "step": 93435 }, { "epoch": 0.23, "learning_rate": 7.674669082490313e-05, "loss": 1.0246, "step": 93440 }, { "epoch": 0.23, "learning_rate": 7.674543258342142e-05, "loss": 1.0244, "step": 93445 }, { "epoch": 0.23, "learning_rate": 7.674417434193971e-05, "loss": 1.0249, "step": 93450 }, { "epoch": 0.23, "learning_rate": 7.6742916100458e-05, "loss": 1.0246, "step": 93455 }, { "epoch": 0.23, "learning_rate": 7.67416578589763e-05, "loss": 1.0228, "step": 93460 }, { "epoch": 0.23, "learning_rate": 7.67403996174946e-05, "loss": 1.022, "step": 93465 }, { "epoch": 0.23, "learning_rate": 7.673914137601289e-05, "loss": 1.0238, "step": 93470 }, { "epoch": 0.23, "learning_rate": 7.673788313453118e-05, "loss": 1.024, "step": 93475 }, { "epoch": 0.23, "learning_rate": 7.673662489304947e-05, "loss": 1.0221, "step": 93480 }, { "epoch": 0.23, "learning_rate": 7.673536665156778e-05, "loss": 1.0231, "step": 93485 }, { "epoch": 0.23, "learning_rate": 7.673410841008607e-05, "loss": 1.0212, "step": 93490 }, { "epoch": 0.23, "learning_rate": 7.673285016860436e-05, "loss": 1.0232, "step": 93495 }, { "epoch": 0.23, "learning_rate": 7.673159192712265e-05, "loss": 1.0229, "step": 93500 }, { "epoch": 0.23, "learning_rate": 7.673033368564096e-05, "loss": 1.024, "step": 93505 }, { "epoch": 0.23, "learning_rate": 7.672907544415925e-05, "loss": 1.0235, "step": 93510 }, { "epoch": 0.23, "learning_rate": 7.672781720267754e-05, "loss": 1.0234, "step": 93515 }, { "epoch": 0.23, "learning_rate": 7.672655896119583e-05, "loss": 1.0256, "step": 93520 }, { "epoch": 0.23, "learning_rate": 7.672530071971413e-05, "loss": 1.0211, "step": 93525 }, { "epoch": 0.23, "learning_rate": 7.672404247823243e-05, "loss": 1.0256, "step": 93530 }, { "epoch": 0.23, "learning_rate": 7.672278423675072e-05, "loss": 1.023, "step": 93535 }, { "epoch": 0.23, "learning_rate": 7.672152599526901e-05, "loss": 1.0224, "step": 93540 }, { "epoch": 0.23, "learning_rate": 7.67202677537873e-05, "loss": 1.0236, "step": 93545 }, { "epoch": 0.23, "learning_rate": 7.671900951230561e-05, "loss": 1.023, "step": 93550 }, { "epoch": 0.23, "learning_rate": 7.67177512708239e-05, "loss": 1.0225, "step": 93555 }, { "epoch": 0.23, "learning_rate": 7.67164930293422e-05, "loss": 1.0255, "step": 93560 }, { "epoch": 0.23, "learning_rate": 7.671523478786049e-05, "loss": 1.0246, "step": 93565 }, { "epoch": 0.23, "learning_rate": 7.671397654637879e-05, "loss": 1.0231, "step": 93570 }, { "epoch": 0.23, "learning_rate": 7.671271830489708e-05, "loss": 1.0236, "step": 93575 }, { "epoch": 0.23, "learning_rate": 7.671146006341537e-05, "loss": 1.0245, "step": 93580 }, { "epoch": 0.23, "learning_rate": 7.671020182193367e-05, "loss": 1.0223, "step": 93585 }, { "epoch": 0.23, "learning_rate": 7.670894358045196e-05, "loss": 1.0204, "step": 93590 }, { "epoch": 0.23, "learning_rate": 7.670768533897026e-05, "loss": 1.0223, "step": 93595 }, { "epoch": 0.23, "learning_rate": 7.670642709748855e-05, "loss": 1.0232, "step": 93600 }, { "epoch": 0.23, "learning_rate": 7.670516885600685e-05, "loss": 1.023, "step": 93605 }, { "epoch": 0.23, "learning_rate": 7.670391061452514e-05, "loss": 1.0224, "step": 93610 }, { "epoch": 0.23, "learning_rate": 7.670265237304344e-05, "loss": 1.0215, "step": 93615 }, { "epoch": 0.24, "learning_rate": 7.670139413156173e-05, "loss": 1.0231, "step": 93620 }, { "epoch": 0.24, "learning_rate": 7.670013589008003e-05, "loss": 1.0237, "step": 93625 }, { "epoch": 0.24, "learning_rate": 7.669887764859832e-05, "loss": 1.024, "step": 93630 }, { "epoch": 0.24, "learning_rate": 7.669761940711662e-05, "loss": 1.0243, "step": 93635 }, { "epoch": 0.24, "learning_rate": 7.669636116563491e-05, "loss": 1.0229, "step": 93640 }, { "epoch": 0.24, "learning_rate": 7.66951029241532e-05, "loss": 1.0226, "step": 93645 }, { "epoch": 0.24, "learning_rate": 7.66938446826715e-05, "loss": 1.0239, "step": 93650 }, { "epoch": 0.24, "learning_rate": 7.669258644118979e-05, "loss": 1.0238, "step": 93655 }, { "epoch": 0.24, "learning_rate": 7.66913281997081e-05, "loss": 1.0235, "step": 93660 }, { "epoch": 0.24, "learning_rate": 7.669006995822639e-05, "loss": 1.0245, "step": 93665 }, { "epoch": 0.24, "learning_rate": 7.668881171674468e-05, "loss": 1.0224, "step": 93670 }, { "epoch": 0.24, "learning_rate": 7.668755347526297e-05, "loss": 1.0222, "step": 93675 }, { "epoch": 0.24, "learning_rate": 7.668629523378127e-05, "loss": 1.022, "step": 93680 }, { "epoch": 0.24, "learning_rate": 7.668503699229957e-05, "loss": 1.0252, "step": 93685 }, { "epoch": 0.24, "learning_rate": 7.668377875081786e-05, "loss": 1.024, "step": 93690 }, { "epoch": 0.24, "learning_rate": 7.668252050933615e-05, "loss": 1.0385, "step": 93695 }, { "epoch": 0.24, "learning_rate": 7.668126226785445e-05, "loss": 1.0221, "step": 93700 }, { "epoch": 0.24, "learning_rate": 7.668000402637275e-05, "loss": 1.0237, "step": 93705 }, { "epoch": 0.24, "learning_rate": 7.667874578489104e-05, "loss": 1.0224, "step": 93710 }, { "epoch": 0.24, "learning_rate": 7.667748754340933e-05, "loss": 1.0237, "step": 93715 }, { "epoch": 0.24, "learning_rate": 7.667622930192762e-05, "loss": 1.0185, "step": 93720 }, { "epoch": 0.24, "learning_rate": 7.667497106044593e-05, "loss": 1.0232, "step": 93725 }, { "epoch": 0.24, "learning_rate": 7.667371281896422e-05, "loss": 1.0254, "step": 93730 }, { "epoch": 0.24, "learning_rate": 7.667245457748251e-05, "loss": 1.0246, "step": 93735 }, { "epoch": 0.24, "learning_rate": 7.66711963360008e-05, "loss": 1.0227, "step": 93740 }, { "epoch": 0.24, "learning_rate": 7.66699380945191e-05, "loss": 1.0214, "step": 93745 }, { "epoch": 0.24, "learning_rate": 7.66686798530374e-05, "loss": 1.0227, "step": 93750 }, { "epoch": 0.24, "learning_rate": 7.666742161155569e-05, "loss": 1.0234, "step": 93755 }, { "epoch": 0.24, "learning_rate": 7.666616337007398e-05, "loss": 1.024, "step": 93760 }, { "epoch": 0.24, "learning_rate": 7.666490512859229e-05, "loss": 1.0254, "step": 93765 }, { "epoch": 0.24, "learning_rate": 7.666364688711058e-05, "loss": 1.0243, "step": 93770 }, { "epoch": 0.24, "learning_rate": 7.666238864562887e-05, "loss": 1.0235, "step": 93775 }, { "epoch": 0.24, "learning_rate": 7.666113040414716e-05, "loss": 1.0248, "step": 93780 }, { "epoch": 0.24, "learning_rate": 7.665987216266545e-05, "loss": 1.0223, "step": 93785 }, { "epoch": 0.24, "learning_rate": 7.665861392118376e-05, "loss": 1.0256, "step": 93790 }, { "epoch": 0.24, "learning_rate": 7.665735567970205e-05, "loss": 1.0235, "step": 93795 }, { "epoch": 0.24, "learning_rate": 7.665609743822034e-05, "loss": 1.0228, "step": 93800 }, { "epoch": 0.24, "learning_rate": 7.665483919673863e-05, "loss": 1.0225, "step": 93805 }, { "epoch": 0.24, "learning_rate": 7.665358095525694e-05, "loss": 1.0223, "step": 93810 }, { "epoch": 0.24, "learning_rate": 7.665232271377523e-05, "loss": 1.0238, "step": 93815 }, { "epoch": 0.24, "learning_rate": 7.665106447229352e-05, "loss": 1.0251, "step": 93820 }, { "epoch": 0.24, "learning_rate": 7.664980623081181e-05, "loss": 1.0265, "step": 93825 }, { "epoch": 0.24, "learning_rate": 7.664854798933012e-05, "loss": 1.0236, "step": 93830 }, { "epoch": 0.24, "learning_rate": 7.664728974784841e-05, "loss": 1.0223, "step": 93835 }, { "epoch": 0.24, "learning_rate": 7.66460315063667e-05, "loss": 1.0224, "step": 93840 }, { "epoch": 0.24, "learning_rate": 7.6644773264885e-05, "loss": 1.044, "step": 93845 }, { "epoch": 0.24, "learning_rate": 7.66435150234033e-05, "loss": 1.0237, "step": 93850 }, { "epoch": 0.24, "learning_rate": 7.66422567819216e-05, "loss": 1.0253, "step": 93855 }, { "epoch": 0.24, "learning_rate": 7.66409985404399e-05, "loss": 1.0271, "step": 93860 }, { "epoch": 0.24, "learning_rate": 7.663974029895819e-05, "loss": 1.0241, "step": 93865 }, { "epoch": 0.24, "learning_rate": 7.663848205747648e-05, "loss": 1.0249, "step": 93870 }, { "epoch": 0.24, "learning_rate": 7.663722381599477e-05, "loss": 1.022, "step": 93875 }, { "epoch": 0.24, "learning_rate": 7.663596557451307e-05, "loss": 1.0244, "step": 93880 }, { "epoch": 0.24, "learning_rate": 7.663470733303137e-05, "loss": 1.0223, "step": 93885 }, { "epoch": 0.24, "learning_rate": 7.663344909154966e-05, "loss": 1.0244, "step": 93890 }, { "epoch": 0.24, "learning_rate": 7.663219085006795e-05, "loss": 1.0233, "step": 93895 }, { "epoch": 0.24, "learning_rate": 7.663093260858625e-05, "loss": 1.041, "step": 93900 }, { "epoch": 0.24, "learning_rate": 7.662967436710454e-05, "loss": 1.0254, "step": 93905 }, { "epoch": 0.24, "learning_rate": 7.662841612562284e-05, "loss": 1.0228, "step": 93910 }, { "epoch": 0.24, "learning_rate": 7.662715788414113e-05, "loss": 1.023, "step": 93915 }, { "epoch": 0.24, "learning_rate": 7.662589964265943e-05, "loss": 1.0227, "step": 93920 }, { "epoch": 0.24, "learning_rate": 7.662464140117772e-05, "loss": 1.0249, "step": 93925 }, { "epoch": 0.24, "learning_rate": 7.662338315969602e-05, "loss": 1.024, "step": 93930 }, { "epoch": 0.24, "learning_rate": 7.662212491821431e-05, "loss": 1.0238, "step": 93935 }, { "epoch": 0.24, "learning_rate": 7.66208666767326e-05, "loss": 1.045, "step": 93940 }, { "epoch": 0.24, "learning_rate": 7.66196084352509e-05, "loss": 1.024, "step": 93945 }, { "epoch": 0.24, "learning_rate": 7.66183501937692e-05, "loss": 1.0214, "step": 93950 }, { "epoch": 0.24, "learning_rate": 7.661709195228749e-05, "loss": 1.0247, "step": 93955 }, { "epoch": 0.24, "learning_rate": 7.661583371080578e-05, "loss": 1.0246, "step": 93960 }, { "epoch": 0.24, "learning_rate": 7.661457546932408e-05, "loss": 1.0221, "step": 93965 }, { "epoch": 0.24, "learning_rate": 7.661331722784238e-05, "loss": 1.0216, "step": 93970 }, { "epoch": 0.24, "learning_rate": 7.661205898636067e-05, "loss": 1.0231, "step": 93975 }, { "epoch": 0.24, "learning_rate": 7.661080074487896e-05, "loss": 1.0221, "step": 93980 }, { "epoch": 0.24, "learning_rate": 7.660954250339726e-05, "loss": 1.0233, "step": 93985 }, { "epoch": 0.24, "learning_rate": 7.660828426191556e-05, "loss": 1.0455, "step": 93990 }, { "epoch": 0.24, "learning_rate": 7.660702602043385e-05, "loss": 1.0243, "step": 93995 }, { "epoch": 0.24, "learning_rate": 7.660576777895214e-05, "loss": 1.0223, "step": 94000 }, { "epoch": 0.24, "learning_rate": 7.660450953747043e-05, "loss": 1.0216, "step": 94005 }, { "epoch": 0.24, "learning_rate": 7.660325129598874e-05, "loss": 1.0218, "step": 94010 }, { "epoch": 0.24, "learning_rate": 7.660199305450703e-05, "loss": 1.0233, "step": 94015 }, { "epoch": 0.24, "learning_rate": 7.660073481302532e-05, "loss": 1.025, "step": 94020 }, { "epoch": 0.24, "learning_rate": 7.659947657154361e-05, "loss": 1.0257, "step": 94025 }, { "epoch": 0.24, "learning_rate": 7.659821833006192e-05, "loss": 1.0217, "step": 94030 }, { "epoch": 0.24, "learning_rate": 7.659696008858021e-05, "loss": 1.026, "step": 94035 }, { "epoch": 0.24, "learning_rate": 7.65957018470985e-05, "loss": 1.0234, "step": 94040 }, { "epoch": 0.24, "learning_rate": 7.659444360561679e-05, "loss": 1.024, "step": 94045 }, { "epoch": 0.24, "learning_rate": 7.65931853641351e-05, "loss": 1.021, "step": 94050 }, { "epoch": 0.24, "learning_rate": 7.659192712265339e-05, "loss": 1.0261, "step": 94055 }, { "epoch": 0.24, "learning_rate": 7.659066888117168e-05, "loss": 1.0229, "step": 94060 }, { "epoch": 0.24, "learning_rate": 7.658941063968997e-05, "loss": 1.0217, "step": 94065 }, { "epoch": 0.24, "learning_rate": 7.658815239820826e-05, "loss": 1.0235, "step": 94070 }, { "epoch": 0.24, "learning_rate": 7.658689415672657e-05, "loss": 1.0251, "step": 94075 }, { "epoch": 0.24, "learning_rate": 7.658563591524486e-05, "loss": 1.0215, "step": 94080 }, { "epoch": 0.24, "learning_rate": 7.658437767376315e-05, "loss": 1.0235, "step": 94085 }, { "epoch": 0.24, "learning_rate": 7.658311943228144e-05, "loss": 1.0257, "step": 94090 }, { "epoch": 0.24, "learning_rate": 7.658186119079975e-05, "loss": 1.026, "step": 94095 }, { "epoch": 0.24, "learning_rate": 7.658060294931804e-05, "loss": 1.0203, "step": 94100 }, { "epoch": 0.24, "learning_rate": 7.657934470783633e-05, "loss": 1.0229, "step": 94105 }, { "epoch": 0.24, "learning_rate": 7.657808646635462e-05, "loss": 1.0226, "step": 94110 }, { "epoch": 0.24, "learning_rate": 7.657682822487291e-05, "loss": 1.0254, "step": 94115 }, { "epoch": 0.24, "learning_rate": 7.657556998339122e-05, "loss": 1.0245, "step": 94120 }, { "epoch": 0.24, "learning_rate": 7.657431174190951e-05, "loss": 1.0223, "step": 94125 }, { "epoch": 0.24, "learning_rate": 7.65730535004278e-05, "loss": 1.0246, "step": 94130 }, { "epoch": 0.24, "learning_rate": 7.65717952589461e-05, "loss": 1.0253, "step": 94135 }, { "epoch": 0.24, "learning_rate": 7.65705370174644e-05, "loss": 1.0241, "step": 94140 }, { "epoch": 0.24, "learning_rate": 7.656927877598269e-05, "loss": 1.021, "step": 94145 }, { "epoch": 0.24, "learning_rate": 7.656802053450098e-05, "loss": 1.0239, "step": 94150 }, { "epoch": 0.24, "learning_rate": 7.656676229301927e-05, "loss": 1.023, "step": 94155 }, { "epoch": 0.24, "learning_rate": 7.656550405153758e-05, "loss": 1.0222, "step": 94160 }, { "epoch": 0.24, "learning_rate": 7.656424581005587e-05, "loss": 1.0226, "step": 94165 }, { "epoch": 0.24, "learning_rate": 7.656298756857416e-05, "loss": 1.0251, "step": 94170 }, { "epoch": 0.24, "learning_rate": 7.656172932709245e-05, "loss": 1.0227, "step": 94175 }, { "epoch": 0.24, "learning_rate": 7.656047108561075e-05, "loss": 1.0209, "step": 94180 }, { "epoch": 0.24, "learning_rate": 7.655921284412905e-05, "loss": 1.0255, "step": 94185 }, { "epoch": 0.24, "learning_rate": 7.655795460264734e-05, "loss": 1.0244, "step": 94190 }, { "epoch": 0.24, "learning_rate": 7.655669636116563e-05, "loss": 1.0238, "step": 94195 }, { "epoch": 0.24, "learning_rate": 7.655543811968393e-05, "loss": 1.0243, "step": 94200 }, { "epoch": 0.24, "learning_rate": 7.655417987820223e-05, "loss": 1.0246, "step": 94205 }, { "epoch": 0.24, "learning_rate": 7.655292163672052e-05, "loss": 1.0233, "step": 94210 }, { "epoch": 0.24, "learning_rate": 7.655166339523881e-05, "loss": 1.0255, "step": 94215 }, { "epoch": 0.24, "learning_rate": 7.65504051537571e-05, "loss": 1.0242, "step": 94220 }, { "epoch": 0.24, "learning_rate": 7.654914691227541e-05, "loss": 1.0247, "step": 94225 }, { "epoch": 0.24, "learning_rate": 7.65478886707937e-05, "loss": 1.0263, "step": 94230 }, { "epoch": 0.24, "learning_rate": 7.6546630429312e-05, "loss": 1.0234, "step": 94235 }, { "epoch": 0.24, "learning_rate": 7.654537218783029e-05, "loss": 1.0232, "step": 94240 }, { "epoch": 0.24, "learning_rate": 7.654411394634858e-05, "loss": 1.0247, "step": 94245 }, { "epoch": 0.24, "learning_rate": 7.654285570486688e-05, "loss": 1.0211, "step": 94250 }, { "epoch": 0.24, "learning_rate": 7.654159746338517e-05, "loss": 1.0241, "step": 94255 }, { "epoch": 0.24, "learning_rate": 7.654033922190347e-05, "loss": 1.0222, "step": 94260 }, { "epoch": 0.24, "learning_rate": 7.653908098042176e-05, "loss": 1.0244, "step": 94265 }, { "epoch": 0.24, "learning_rate": 7.653782273894006e-05, "loss": 1.0252, "step": 94270 }, { "epoch": 0.24, "learning_rate": 7.653656449745835e-05, "loss": 1.0235, "step": 94275 }, { "epoch": 0.24, "learning_rate": 7.653530625597665e-05, "loss": 1.0214, "step": 94280 }, { "epoch": 0.24, "learning_rate": 7.653404801449494e-05, "loss": 1.0238, "step": 94285 }, { "epoch": 0.24, "learning_rate": 7.653278977301324e-05, "loss": 1.0209, "step": 94290 }, { "epoch": 0.24, "learning_rate": 7.653153153153153e-05, "loss": 1.0242, "step": 94295 }, { "epoch": 0.24, "learning_rate": 7.653027329004983e-05, "loss": 1.0215, "step": 94300 }, { "epoch": 0.24, "learning_rate": 7.652901504856812e-05, "loss": 1.0242, "step": 94305 }, { "epoch": 0.24, "learning_rate": 7.652775680708641e-05, "loss": 1.0224, "step": 94310 }, { "epoch": 0.24, "learning_rate": 7.652649856560471e-05, "loss": 1.024, "step": 94315 }, { "epoch": 0.24, "learning_rate": 7.6525240324123e-05, "loss": 1.0253, "step": 94320 }, { "epoch": 0.24, "learning_rate": 7.65239820826413e-05, "loss": 1.024, "step": 94325 }, { "epoch": 0.24, "learning_rate": 7.652272384115959e-05, "loss": 1.0224, "step": 94330 }, { "epoch": 0.24, "learning_rate": 7.65214655996779e-05, "loss": 1.0448, "step": 94335 }, { "epoch": 0.24, "learning_rate": 7.652020735819618e-05, "loss": 1.0235, "step": 94340 }, { "epoch": 0.24, "learning_rate": 7.651894911671449e-05, "loss": 1.0255, "step": 94345 }, { "epoch": 0.24, "learning_rate": 7.651769087523278e-05, "loss": 1.0221, "step": 94350 }, { "epoch": 0.24, "learning_rate": 7.651643263375107e-05, "loss": 1.0457, "step": 94355 }, { "epoch": 0.24, "learning_rate": 7.651517439226938e-05, "loss": 1.0201, "step": 94360 }, { "epoch": 0.24, "learning_rate": 7.651391615078767e-05, "loss": 1.0231, "step": 94365 }, { "epoch": 0.24, "learning_rate": 7.651265790930596e-05, "loss": 1.0253, "step": 94370 }, { "epoch": 0.24, "learning_rate": 7.651139966782425e-05, "loss": 1.0257, "step": 94375 }, { "epoch": 0.24, "learning_rate": 7.651014142634256e-05, "loss": 1.0222, "step": 94380 }, { "epoch": 0.24, "learning_rate": 7.650888318486085e-05, "loss": 1.0227, "step": 94385 }, { "epoch": 0.24, "learning_rate": 7.650762494337914e-05, "loss": 1.0224, "step": 94390 }, { "epoch": 0.24, "learning_rate": 7.650636670189743e-05, "loss": 1.0233, "step": 94395 }, { "epoch": 0.24, "learning_rate": 7.650510846041572e-05, "loss": 1.0235, "step": 94400 }, { "epoch": 0.24, "learning_rate": 7.650385021893403e-05, "loss": 1.0242, "step": 94405 }, { "epoch": 0.24, "learning_rate": 7.650259197745232e-05, "loss": 1.024, "step": 94410 }, { "epoch": 0.24, "learning_rate": 7.650133373597061e-05, "loss": 1.0226, "step": 94415 }, { "epoch": 0.24, "learning_rate": 7.65000754944889e-05, "loss": 1.0257, "step": 94420 }, { "epoch": 0.24, "learning_rate": 7.649881725300721e-05, "loss": 1.0231, "step": 94425 }, { "epoch": 0.24, "learning_rate": 7.64975590115255e-05, "loss": 1.024, "step": 94430 }, { "epoch": 0.24, "learning_rate": 7.649630077004379e-05, "loss": 1.0224, "step": 94435 }, { "epoch": 0.24, "learning_rate": 7.649504252856208e-05, "loss": 1.0244, "step": 94440 }, { "epoch": 0.24, "learning_rate": 7.649378428708039e-05, "loss": 1.0242, "step": 94445 }, { "epoch": 0.24, "learning_rate": 7.649252604559868e-05, "loss": 1.0249, "step": 94450 }, { "epoch": 0.24, "learning_rate": 7.649126780411697e-05, "loss": 1.0225, "step": 94455 }, { "epoch": 0.24, "learning_rate": 7.649000956263526e-05, "loss": 1.0247, "step": 94460 }, { "epoch": 0.24, "learning_rate": 7.648875132115356e-05, "loss": 1.0233, "step": 94465 }, { "epoch": 0.24, "learning_rate": 7.648749307967186e-05, "loss": 1.0268, "step": 94470 }, { "epoch": 0.24, "learning_rate": 7.648623483819015e-05, "loss": 1.022, "step": 94475 }, { "epoch": 0.24, "learning_rate": 7.648497659670844e-05, "loss": 1.0215, "step": 94480 }, { "epoch": 0.24, "learning_rate": 7.648371835522674e-05, "loss": 1.0217, "step": 94485 }, { "epoch": 0.24, "learning_rate": 7.648246011374504e-05, "loss": 1.0234, "step": 94490 }, { "epoch": 0.24, "learning_rate": 7.648120187226333e-05, "loss": 1.0235, "step": 94495 }, { "epoch": 0.24, "learning_rate": 7.647994363078162e-05, "loss": 1.0421, "step": 94500 }, { "epoch": 0.24, "learning_rate": 7.647868538929992e-05, "loss": 1.0226, "step": 94505 }, { "epoch": 0.24, "learning_rate": 7.647742714781822e-05, "loss": 1.0214, "step": 94510 }, { "epoch": 0.24, "learning_rate": 7.647616890633651e-05, "loss": 1.0236, "step": 94515 }, { "epoch": 0.24, "learning_rate": 7.64749106648548e-05, "loss": 1.0369, "step": 94520 }, { "epoch": 0.24, "learning_rate": 7.64736524233731e-05, "loss": 1.0213, "step": 94525 }, { "epoch": 0.24, "learning_rate": 7.647239418189139e-05, "loss": 1.0253, "step": 94530 }, { "epoch": 0.24, "learning_rate": 7.647113594040969e-05, "loss": 1.023, "step": 94535 }, { "epoch": 0.24, "learning_rate": 7.646987769892798e-05, "loss": 1.0259, "step": 94540 }, { "epoch": 0.24, "learning_rate": 7.646861945744628e-05, "loss": 1.0216, "step": 94545 }, { "epoch": 0.24, "learning_rate": 7.646736121596457e-05, "loss": 1.0239, "step": 94550 }, { "epoch": 0.24, "learning_rate": 7.646610297448287e-05, "loss": 1.025, "step": 94555 }, { "epoch": 0.24, "learning_rate": 7.646484473300116e-05, "loss": 1.0253, "step": 94560 }, { "epoch": 0.24, "learning_rate": 7.646358649151946e-05, "loss": 1.0246, "step": 94565 }, { "epoch": 0.24, "learning_rate": 7.646232825003775e-05, "loss": 1.038, "step": 94570 }, { "epoch": 0.24, "learning_rate": 7.646107000855605e-05, "loss": 1.0245, "step": 94575 }, { "epoch": 0.24, "learning_rate": 7.645981176707434e-05, "loss": 1.0222, "step": 94580 }, { "epoch": 0.24, "learning_rate": 7.645855352559264e-05, "loss": 1.0253, "step": 94585 }, { "epoch": 0.24, "learning_rate": 7.645729528411093e-05, "loss": 1.0225, "step": 94590 }, { "epoch": 0.24, "learning_rate": 7.645603704262922e-05, "loss": 1.0233, "step": 94595 }, { "epoch": 0.24, "learning_rate": 7.645477880114752e-05, "loss": 1.0237, "step": 94600 }, { "epoch": 0.24, "learning_rate": 7.645352055966582e-05, "loss": 1.0233, "step": 94605 }, { "epoch": 0.24, "learning_rate": 7.645226231818411e-05, "loss": 1.0237, "step": 94610 }, { "epoch": 0.24, "learning_rate": 7.64510040767024e-05, "loss": 1.0227, "step": 94615 }, { "epoch": 0.24, "learning_rate": 7.64497458352207e-05, "loss": 1.0239, "step": 94620 }, { "epoch": 0.24, "learning_rate": 7.6448487593739e-05, "loss": 1.024, "step": 94625 }, { "epoch": 0.24, "learning_rate": 7.644722935225729e-05, "loss": 1.0231, "step": 94630 }, { "epoch": 0.24, "learning_rate": 7.644597111077558e-05, "loss": 1.0225, "step": 94635 }, { "epoch": 0.24, "learning_rate": 7.644471286929388e-05, "loss": 1.0215, "step": 94640 }, { "epoch": 0.24, "learning_rate": 7.644345462781218e-05, "loss": 1.0236, "step": 94645 }, { "epoch": 0.24, "learning_rate": 7.644219638633047e-05, "loss": 1.0249, "step": 94650 }, { "epoch": 0.24, "learning_rate": 7.644093814484876e-05, "loss": 1.0234, "step": 94655 }, { "epoch": 0.24, "learning_rate": 7.643967990336705e-05, "loss": 1.02, "step": 94660 }, { "epoch": 0.24, "learning_rate": 7.643842166188536e-05, "loss": 1.0258, "step": 94665 }, { "epoch": 0.24, "learning_rate": 7.643716342040365e-05, "loss": 1.0255, "step": 94670 }, { "epoch": 0.24, "learning_rate": 7.643590517892194e-05, "loss": 1.0239, "step": 94675 }, { "epoch": 0.24, "learning_rate": 7.643464693744023e-05, "loss": 1.0467, "step": 94680 }, { "epoch": 0.24, "learning_rate": 7.643338869595854e-05, "loss": 1.0243, "step": 94685 }, { "epoch": 0.24, "learning_rate": 7.643213045447683e-05, "loss": 1.0244, "step": 94690 }, { "epoch": 0.24, "learning_rate": 7.643087221299512e-05, "loss": 1.0239, "step": 94695 }, { "epoch": 0.24, "learning_rate": 7.642961397151341e-05, "loss": 1.0242, "step": 94700 }, { "epoch": 0.24, "learning_rate": 7.642835573003172e-05, "loss": 1.0242, "step": 94705 }, { "epoch": 0.24, "learning_rate": 7.642709748855001e-05, "loss": 1.0229, "step": 94710 }, { "epoch": 0.24, "learning_rate": 7.64258392470683e-05, "loss": 1.0209, "step": 94715 }, { "epoch": 0.24, "learning_rate": 7.642458100558659e-05, "loss": 1.0218, "step": 94720 }, { "epoch": 0.24, "learning_rate": 7.642332276410488e-05, "loss": 1.0248, "step": 94725 }, { "epoch": 0.24, "learning_rate": 7.642206452262319e-05, "loss": 1.0226, "step": 94730 }, { "epoch": 0.24, "learning_rate": 7.642080628114148e-05, "loss": 1.0245, "step": 94735 }, { "epoch": 0.24, "learning_rate": 7.641954803965977e-05, "loss": 1.0238, "step": 94740 }, { "epoch": 0.24, "learning_rate": 7.641828979817806e-05, "loss": 1.0222, "step": 94745 }, { "epoch": 0.24, "learning_rate": 7.641703155669637e-05, "loss": 1.0229, "step": 94750 }, { "epoch": 0.24, "learning_rate": 7.641577331521466e-05, "loss": 1.0236, "step": 94755 }, { "epoch": 0.24, "learning_rate": 7.641451507373295e-05, "loss": 1.0253, "step": 94760 }, { "epoch": 0.24, "learning_rate": 7.641325683225124e-05, "loss": 1.0231, "step": 94765 }, { "epoch": 0.24, "learning_rate": 7.641199859076953e-05, "loss": 1.0239, "step": 94770 }, { "epoch": 0.24, "learning_rate": 7.641074034928784e-05, "loss": 1.0218, "step": 94775 }, { "epoch": 0.24, "learning_rate": 7.640948210780613e-05, "loss": 1.0225, "step": 94780 }, { "epoch": 0.24, "learning_rate": 7.640822386632442e-05, "loss": 1.0232, "step": 94785 }, { "epoch": 0.24, "learning_rate": 7.640696562484271e-05, "loss": 1.0238, "step": 94790 }, { "epoch": 0.24, "learning_rate": 7.640570738336102e-05, "loss": 1.023, "step": 94795 }, { "epoch": 0.24, "learning_rate": 7.640444914187931e-05, "loss": 1.0236, "step": 94800 }, { "epoch": 0.24, "learning_rate": 7.64031909003976e-05, "loss": 1.0249, "step": 94805 }, { "epoch": 0.24, "learning_rate": 7.64019326589159e-05, "loss": 1.0205, "step": 94810 }, { "epoch": 0.24, "learning_rate": 7.64006744174342e-05, "loss": 1.0263, "step": 94815 }, { "epoch": 0.24, "learning_rate": 7.639941617595249e-05, "loss": 1.0241, "step": 94820 }, { "epoch": 0.24, "learning_rate": 7.639815793447078e-05, "loss": 1.0214, "step": 94825 }, { "epoch": 0.24, "learning_rate": 7.639689969298907e-05, "loss": 1.0228, "step": 94830 }, { "epoch": 0.24, "learning_rate": 7.639564145150736e-05, "loss": 1.0246, "step": 94835 }, { "epoch": 0.24, "learning_rate": 7.639438321002568e-05, "loss": 1.0248, "step": 94840 }, { "epoch": 0.24, "learning_rate": 7.639312496854398e-05, "loss": 1.0233, "step": 94845 }, { "epoch": 0.24, "learning_rate": 7.639186672706227e-05, "loss": 1.0233, "step": 94850 }, { "epoch": 0.24, "learning_rate": 7.639060848558056e-05, "loss": 1.0234, "step": 94855 }, { "epoch": 0.24, "learning_rate": 7.638935024409885e-05, "loss": 1.023, "step": 94860 }, { "epoch": 0.24, "learning_rate": 7.638809200261716e-05, "loss": 1.0251, "step": 94865 }, { "epoch": 0.24, "learning_rate": 7.638683376113545e-05, "loss": 1.0437, "step": 94870 }, { "epoch": 0.24, "learning_rate": 7.638557551965374e-05, "loss": 1.0214, "step": 94875 }, { "epoch": 0.24, "learning_rate": 7.638431727817203e-05, "loss": 1.041, "step": 94880 }, { "epoch": 0.24, "learning_rate": 7.638305903669034e-05, "loss": 1.0243, "step": 94885 }, { "epoch": 0.24, "learning_rate": 7.638180079520863e-05, "loss": 1.0244, "step": 94890 }, { "epoch": 0.24, "learning_rate": 7.638054255372692e-05, "loss": 1.0224, "step": 94895 }, { "epoch": 0.24, "learning_rate": 7.637928431224521e-05, "loss": 1.0233, "step": 94900 }, { "epoch": 0.24, "learning_rate": 7.637802607076351e-05, "loss": 1.0201, "step": 94905 }, { "epoch": 0.24, "learning_rate": 7.63767678292818e-05, "loss": 1.0232, "step": 94910 }, { "epoch": 0.24, "learning_rate": 7.63755095878001e-05, "loss": 1.0385, "step": 94915 }, { "epoch": 0.24, "learning_rate": 7.637425134631839e-05, "loss": 1.0246, "step": 94920 }, { "epoch": 0.24, "learning_rate": 7.637299310483668e-05, "loss": 1.0268, "step": 94925 }, { "epoch": 0.24, "learning_rate": 7.637173486335499e-05, "loss": 1.0243, "step": 94930 }, { "epoch": 0.24, "learning_rate": 7.637047662187328e-05, "loss": 1.0227, "step": 94935 }, { "epoch": 0.24, "learning_rate": 7.636921838039157e-05, "loss": 1.0244, "step": 94940 }, { "epoch": 0.24, "learning_rate": 7.636796013890986e-05, "loss": 1.0225, "step": 94945 }, { "epoch": 0.24, "learning_rate": 7.636670189742817e-05, "loss": 1.0229, "step": 94950 }, { "epoch": 0.24, "learning_rate": 7.636544365594646e-05, "loss": 1.0246, "step": 94955 }, { "epoch": 0.24, "learning_rate": 7.636418541446475e-05, "loss": 1.0236, "step": 94960 }, { "epoch": 0.24, "learning_rate": 7.636292717298304e-05, "loss": 1.0231, "step": 94965 }, { "epoch": 0.24, "learning_rate": 7.636166893150135e-05, "loss": 1.0207, "step": 94970 }, { "epoch": 0.24, "learning_rate": 7.636041069001964e-05, "loss": 1.0224, "step": 94975 }, { "epoch": 0.24, "learning_rate": 7.635915244853793e-05, "loss": 1.0228, "step": 94980 }, { "epoch": 0.24, "learning_rate": 7.635789420705622e-05, "loss": 1.0256, "step": 94985 }, { "epoch": 0.24, "learning_rate": 7.635663596557451e-05, "loss": 1.0248, "step": 94990 }, { "epoch": 0.24, "learning_rate": 7.635537772409282e-05, "loss": 1.0203, "step": 94995 }, { "epoch": 0.24, "learning_rate": 7.635411948261111e-05, "loss": 1.0224, "step": 95000 }, { "epoch": 0.24, "learning_rate": 7.63528612411294e-05, "loss": 1.0217, "step": 95005 }, { "epoch": 0.24, "learning_rate": 7.635160299964769e-05, "loss": 1.0223, "step": 95010 }, { "epoch": 0.24, "learning_rate": 7.6350344758166e-05, "loss": 1.0239, "step": 95015 }, { "epoch": 0.24, "learning_rate": 7.634908651668429e-05, "loss": 1.0247, "step": 95020 }, { "epoch": 0.24, "learning_rate": 7.634782827520258e-05, "loss": 1.0225, "step": 95025 }, { "epoch": 0.24, "learning_rate": 7.634657003372087e-05, "loss": 1.0232, "step": 95030 }, { "epoch": 0.24, "learning_rate": 7.634531179223918e-05, "loss": 1.0256, "step": 95035 }, { "epoch": 0.24, "learning_rate": 7.634405355075747e-05, "loss": 1.0233, "step": 95040 }, { "epoch": 0.24, "learning_rate": 7.634279530927576e-05, "loss": 1.025, "step": 95045 }, { "epoch": 0.24, "learning_rate": 7.634153706779405e-05, "loss": 1.0243, "step": 95050 }, { "epoch": 0.24, "learning_rate": 7.634027882631234e-05, "loss": 1.0229, "step": 95055 }, { "epoch": 0.24, "learning_rate": 7.633902058483065e-05, "loss": 1.0426, "step": 95060 }, { "epoch": 0.24, "learning_rate": 7.633776234334894e-05, "loss": 1.0235, "step": 95065 }, { "epoch": 0.24, "learning_rate": 7.633650410186723e-05, "loss": 1.023, "step": 95070 }, { "epoch": 0.24, "learning_rate": 7.633524586038552e-05, "loss": 1.0224, "step": 95075 }, { "epoch": 0.24, "learning_rate": 7.633398761890383e-05, "loss": 1.0373, "step": 95080 }, { "epoch": 0.24, "learning_rate": 7.633272937742212e-05, "loss": 1.027, "step": 95085 }, { "epoch": 0.24, "learning_rate": 7.633147113594041e-05, "loss": 1.025, "step": 95090 }, { "epoch": 0.24, "learning_rate": 7.63302128944587e-05, "loss": 1.0207, "step": 95095 }, { "epoch": 0.24, "learning_rate": 7.632895465297701e-05, "loss": 1.0235, "step": 95100 }, { "epoch": 0.24, "learning_rate": 7.63276964114953e-05, "loss": 1.026, "step": 95105 }, { "epoch": 0.24, "learning_rate": 7.632643817001359e-05, "loss": 1.0255, "step": 95110 }, { "epoch": 0.24, "learning_rate": 7.632517992853188e-05, "loss": 1.025, "step": 95115 }, { "epoch": 0.24, "learning_rate": 7.632392168705018e-05, "loss": 1.021, "step": 95120 }, { "epoch": 0.24, "learning_rate": 7.632266344556848e-05, "loss": 1.0228, "step": 95125 }, { "epoch": 0.24, "learning_rate": 7.632140520408677e-05, "loss": 1.0243, "step": 95130 }, { "epoch": 0.24, "learning_rate": 7.632014696260506e-05, "loss": 1.024, "step": 95135 }, { "epoch": 0.24, "learning_rate": 7.631888872112336e-05, "loss": 1.0255, "step": 95140 }, { "epoch": 0.24, "learning_rate": 7.631763047964166e-05, "loss": 1.023, "step": 95145 }, { "epoch": 0.24, "learning_rate": 7.631637223815995e-05, "loss": 1.0236, "step": 95150 }, { "epoch": 0.24, "learning_rate": 7.631511399667824e-05, "loss": 1.0231, "step": 95155 }, { "epoch": 0.24, "learning_rate": 7.631385575519654e-05, "loss": 1.0239, "step": 95160 }, { "epoch": 0.24, "learning_rate": 7.631259751371484e-05, "loss": 1.0235, "step": 95165 }, { "epoch": 0.24, "learning_rate": 7.631133927223313e-05, "loss": 1.0223, "step": 95170 }, { "epoch": 0.24, "learning_rate": 7.631008103075142e-05, "loss": 1.0241, "step": 95175 }, { "epoch": 0.24, "learning_rate": 7.630882278926972e-05, "loss": 1.025, "step": 95180 }, { "epoch": 0.24, "learning_rate": 7.630756454778801e-05, "loss": 1.0228, "step": 95185 }, { "epoch": 0.24, "learning_rate": 7.630630630630631e-05, "loss": 1.0255, "step": 95190 }, { "epoch": 0.24, "learning_rate": 7.63050480648246e-05, "loss": 1.0231, "step": 95195 }, { "epoch": 0.24, "learning_rate": 7.63037898233429e-05, "loss": 1.0216, "step": 95200 }, { "epoch": 0.24, "learning_rate": 7.630253158186119e-05, "loss": 1.0217, "step": 95205 }, { "epoch": 0.24, "learning_rate": 7.630127334037949e-05, "loss": 1.0231, "step": 95210 }, { "epoch": 0.24, "learning_rate": 7.630001509889778e-05, "loss": 1.0229, "step": 95215 }, { "epoch": 0.24, "learning_rate": 7.629875685741608e-05, "loss": 1.0239, "step": 95220 }, { "epoch": 0.24, "learning_rate": 7.629749861593437e-05, "loss": 1.0251, "step": 95225 }, { "epoch": 0.24, "learning_rate": 7.629624037445267e-05, "loss": 1.0212, "step": 95230 }, { "epoch": 0.24, "learning_rate": 7.629498213297096e-05, "loss": 1.023, "step": 95235 }, { "epoch": 0.24, "learning_rate": 7.629372389148926e-05, "loss": 1.0244, "step": 95240 }, { "epoch": 0.24, "learning_rate": 7.629246565000755e-05, "loss": 1.0253, "step": 95245 }, { "epoch": 0.24, "learning_rate": 7.629120740852584e-05, "loss": 1.0239, "step": 95250 }, { "epoch": 0.24, "learning_rate": 7.628994916704414e-05, "loss": 1.0224, "step": 95255 }, { "epoch": 0.24, "learning_rate": 7.628869092556244e-05, "loss": 1.0258, "step": 95260 }, { "epoch": 0.24, "learning_rate": 7.628743268408073e-05, "loss": 1.0258, "step": 95265 }, { "epoch": 0.24, "learning_rate": 7.628617444259902e-05, "loss": 1.0227, "step": 95270 }, { "epoch": 0.24, "learning_rate": 7.628491620111732e-05, "loss": 1.0231, "step": 95275 }, { "epoch": 0.24, "learning_rate": 7.628365795963562e-05, "loss": 1.0234, "step": 95280 }, { "epoch": 0.24, "learning_rate": 7.62823997181539e-05, "loss": 1.025, "step": 95285 }, { "epoch": 0.24, "learning_rate": 7.62811414766722e-05, "loss": 1.0232, "step": 95290 }, { "epoch": 0.24, "learning_rate": 7.62798832351905e-05, "loss": 1.0225, "step": 95295 }, { "epoch": 0.24, "learning_rate": 7.62786249937088e-05, "loss": 1.0225, "step": 95300 }, { "epoch": 0.24, "learning_rate": 7.627736675222709e-05, "loss": 1.0261, "step": 95305 }, { "epoch": 0.24, "learning_rate": 7.627610851074538e-05, "loss": 1.0248, "step": 95310 }, { "epoch": 0.24, "learning_rate": 7.627485026926367e-05, "loss": 1.0233, "step": 95315 }, { "epoch": 0.24, "learning_rate": 7.627359202778198e-05, "loss": 1.0224, "step": 95320 }, { "epoch": 0.24, "learning_rate": 7.627233378630027e-05, "loss": 1.0244, "step": 95325 }, { "epoch": 0.24, "learning_rate": 7.627107554481856e-05, "loss": 1.0225, "step": 95330 }, { "epoch": 0.24, "learning_rate": 7.626981730333685e-05, "loss": 1.0219, "step": 95335 }, { "epoch": 0.24, "learning_rate": 7.626855906185515e-05, "loss": 1.0243, "step": 95340 }, { "epoch": 0.24, "learning_rate": 7.626730082037346e-05, "loss": 1.0255, "step": 95345 }, { "epoch": 0.24, "learning_rate": 7.626604257889175e-05, "loss": 1.0236, "step": 95350 }, { "epoch": 0.24, "learning_rate": 7.626478433741004e-05, "loss": 1.0225, "step": 95355 }, { "epoch": 0.24, "learning_rate": 7.626352609592833e-05, "loss": 1.0233, "step": 95360 }, { "epoch": 0.24, "learning_rate": 7.626226785444664e-05, "loss": 1.0234, "step": 95365 }, { "epoch": 0.24, "learning_rate": 7.626100961296493e-05, "loss": 1.0215, "step": 95370 }, { "epoch": 0.24, "learning_rate": 7.625975137148322e-05, "loss": 1.0223, "step": 95375 }, { "epoch": 0.24, "learning_rate": 7.625849313000151e-05, "loss": 1.0218, "step": 95380 }, { "epoch": 0.24, "learning_rate": 7.62572348885198e-05, "loss": 1.0225, "step": 95385 }, { "epoch": 0.24, "learning_rate": 7.625597664703811e-05, "loss": 1.0214, "step": 95390 }, { "epoch": 0.24, "learning_rate": 7.62547184055564e-05, "loss": 1.0233, "step": 95395 }, { "epoch": 0.24, "learning_rate": 7.62534601640747e-05, "loss": 1.0258, "step": 95400 }, { "epoch": 0.24, "learning_rate": 7.625220192259299e-05, "loss": 1.0453, "step": 95405 }, { "epoch": 0.24, "learning_rate": 7.625094368111129e-05, "loss": 1.0417, "step": 95410 }, { "epoch": 0.24, "learning_rate": 7.624968543962958e-05, "loss": 1.0222, "step": 95415 }, { "epoch": 0.24, "learning_rate": 7.624842719814787e-05, "loss": 1.0226, "step": 95420 }, { "epoch": 0.24, "learning_rate": 7.624716895666617e-05, "loss": 1.0221, "step": 95425 }, { "epoch": 0.24, "learning_rate": 7.624591071518447e-05, "loss": 1.0226, "step": 95430 }, { "epoch": 0.24, "learning_rate": 7.624465247370276e-05, "loss": 1.0227, "step": 95435 }, { "epoch": 0.24, "learning_rate": 7.624339423222105e-05, "loss": 1.0225, "step": 95440 }, { "epoch": 0.24, "learning_rate": 7.624213599073935e-05, "loss": 1.0234, "step": 95445 }, { "epoch": 0.24, "learning_rate": 7.624087774925764e-05, "loss": 1.024, "step": 95450 }, { "epoch": 0.24, "learning_rate": 7.623961950777594e-05, "loss": 1.0459, "step": 95455 }, { "epoch": 0.24, "learning_rate": 7.623836126629423e-05, "loss": 1.0207, "step": 95460 }, { "epoch": 0.24, "learning_rate": 7.623710302481253e-05, "loss": 1.021, "step": 95465 }, { "epoch": 0.24, "learning_rate": 7.623584478333082e-05, "loss": 1.0234, "step": 95470 }, { "epoch": 0.24, "learning_rate": 7.623458654184912e-05, "loss": 1.0232, "step": 95475 }, { "epoch": 0.24, "learning_rate": 7.623332830036741e-05, "loss": 1.0229, "step": 95480 }, { "epoch": 0.24, "learning_rate": 7.62320700588857e-05, "loss": 1.0234, "step": 95485 }, { "epoch": 0.24, "learning_rate": 7.6230811817404e-05, "loss": 1.0246, "step": 95490 }, { "epoch": 0.24, "learning_rate": 7.62295535759223e-05, "loss": 1.0256, "step": 95495 }, { "epoch": 0.24, "learning_rate": 7.62282953344406e-05, "loss": 1.0235, "step": 95500 }, { "epoch": 0.24, "learning_rate": 7.622703709295889e-05, "loss": 1.0241, "step": 95505 }, { "epoch": 0.24, "learning_rate": 7.622577885147718e-05, "loss": 1.0223, "step": 95510 }, { "epoch": 0.24, "learning_rate": 7.622452060999547e-05, "loss": 1.0242, "step": 95515 }, { "epoch": 0.24, "learning_rate": 7.622326236851377e-05, "loss": 1.0226, "step": 95520 }, { "epoch": 0.24, "learning_rate": 7.622200412703207e-05, "loss": 1.0446, "step": 95525 }, { "epoch": 0.24, "learning_rate": 7.622074588555036e-05, "loss": 1.0218, "step": 95530 }, { "epoch": 0.24, "learning_rate": 7.621948764406865e-05, "loss": 1.022, "step": 95535 }, { "epoch": 0.24, "learning_rate": 7.621822940258695e-05, "loss": 1.0243, "step": 95540 }, { "epoch": 0.24, "learning_rate": 7.621697116110525e-05, "loss": 1.0361, "step": 95545 }, { "epoch": 0.24, "learning_rate": 7.621571291962354e-05, "loss": 1.022, "step": 95550 }, { "epoch": 0.24, "learning_rate": 7.621445467814183e-05, "loss": 1.0217, "step": 95555 }, { "epoch": 0.24, "learning_rate": 7.621319643666013e-05, "loss": 1.0272, "step": 95560 }, { "epoch": 0.24, "learning_rate": 7.621193819517843e-05, "loss": 1.022, "step": 95565 }, { "epoch": 0.24, "learning_rate": 7.621067995369672e-05, "loss": 1.0235, "step": 95570 }, { "epoch": 0.24, "learning_rate": 7.620942171221501e-05, "loss": 1.0248, "step": 95575 }, { "epoch": 0.24, "learning_rate": 7.62081634707333e-05, "loss": 1.0231, "step": 95580 }, { "epoch": 0.24, "learning_rate": 7.62069052292516e-05, "loss": 1.0224, "step": 95585 }, { "epoch": 0.24, "learning_rate": 7.62056469877699e-05, "loss": 1.0263, "step": 95590 }, { "epoch": 0.24, "learning_rate": 7.620438874628819e-05, "loss": 1.0258, "step": 95595 }, { "epoch": 0.24, "learning_rate": 7.620313050480648e-05, "loss": 1.022, "step": 95600 }, { "epoch": 0.24, "learning_rate": 7.620187226332479e-05, "loss": 1.0242, "step": 95605 }, { "epoch": 0.24, "learning_rate": 7.620061402184308e-05, "loss": 1.0235, "step": 95610 }, { "epoch": 0.24, "learning_rate": 7.619935578036137e-05, "loss": 1.0232, "step": 95615 }, { "epoch": 0.24, "learning_rate": 7.619809753887966e-05, "loss": 1.0416, "step": 95620 }, { "epoch": 0.24, "learning_rate": 7.619683929739797e-05, "loss": 1.0245, "step": 95625 }, { "epoch": 0.24, "learning_rate": 7.619558105591626e-05, "loss": 1.024, "step": 95630 }, { "epoch": 0.24, "learning_rate": 7.619432281443455e-05, "loss": 1.0238, "step": 95635 }, { "epoch": 0.24, "learning_rate": 7.619306457295284e-05, "loss": 1.0228, "step": 95640 }, { "epoch": 0.24, "learning_rate": 7.619180633147113e-05, "loss": 1.0246, "step": 95645 }, { "epoch": 0.24, "learning_rate": 7.619054808998944e-05, "loss": 1.0232, "step": 95650 }, { "epoch": 0.24, "learning_rate": 7.618928984850773e-05, "loss": 1.0226, "step": 95655 }, { "epoch": 0.24, "learning_rate": 7.618803160702602e-05, "loss": 1.024, "step": 95660 }, { "epoch": 0.24, "learning_rate": 7.618677336554431e-05, "loss": 1.0225, "step": 95665 }, { "epoch": 0.24, "learning_rate": 7.618551512406262e-05, "loss": 1.0219, "step": 95670 }, { "epoch": 0.24, "learning_rate": 7.618425688258091e-05, "loss": 1.0225, "step": 95675 }, { "epoch": 0.24, "learning_rate": 7.61829986410992e-05, "loss": 1.0221, "step": 95680 }, { "epoch": 0.24, "learning_rate": 7.618174039961749e-05, "loss": 1.0238, "step": 95685 }, { "epoch": 0.24, "learning_rate": 7.61804821581358e-05, "loss": 1.0226, "step": 95690 }, { "epoch": 0.24, "learning_rate": 7.617922391665409e-05, "loss": 1.0229, "step": 95695 }, { "epoch": 0.24, "learning_rate": 7.617796567517238e-05, "loss": 1.0249, "step": 95700 }, { "epoch": 0.24, "learning_rate": 7.617670743369067e-05, "loss": 1.0224, "step": 95705 }, { "epoch": 0.24, "learning_rate": 7.617544919220896e-05, "loss": 1.0273, "step": 95710 }, { "epoch": 0.24, "learning_rate": 7.617419095072727e-05, "loss": 1.0245, "step": 95715 }, { "epoch": 0.24, "learning_rate": 7.617293270924556e-05, "loss": 1.0229, "step": 95720 }, { "epoch": 0.24, "learning_rate": 7.617167446776385e-05, "loss": 1.025, "step": 95725 }, { "epoch": 0.24, "learning_rate": 7.617041622628214e-05, "loss": 1.0202, "step": 95730 }, { "epoch": 0.24, "learning_rate": 7.616915798480045e-05, "loss": 1.0221, "step": 95735 }, { "epoch": 0.24, "learning_rate": 7.616789974331874e-05, "loss": 1.0251, "step": 95740 }, { "epoch": 0.24, "learning_rate": 7.616664150183703e-05, "loss": 1.024, "step": 95745 }, { "epoch": 0.24, "learning_rate": 7.616538326035532e-05, "loss": 1.0248, "step": 95750 }, { "epoch": 0.24, "learning_rate": 7.616412501887363e-05, "loss": 1.0272, "step": 95755 }, { "epoch": 0.24, "learning_rate": 7.616286677739192e-05, "loss": 1.0245, "step": 95760 }, { "epoch": 0.24, "learning_rate": 7.616160853591021e-05, "loss": 1.0237, "step": 95765 }, { "epoch": 0.24, "learning_rate": 7.61603502944285e-05, "loss": 1.0233, "step": 95770 }, { "epoch": 0.24, "learning_rate": 7.61590920529468e-05, "loss": 1.0234, "step": 95775 }, { "epoch": 0.24, "learning_rate": 7.61578338114651e-05, "loss": 1.0231, "step": 95780 }, { "epoch": 0.24, "learning_rate": 7.615657556998339e-05, "loss": 1.0239, "step": 95785 }, { "epoch": 0.24, "learning_rate": 7.615531732850168e-05, "loss": 1.0226, "step": 95790 }, { "epoch": 0.24, "learning_rate": 7.615405908701997e-05, "loss": 1.0237, "step": 95795 }, { "epoch": 0.24, "learning_rate": 7.615280084553828e-05, "loss": 1.022, "step": 95800 }, { "epoch": 0.24, "learning_rate": 7.615154260405657e-05, "loss": 1.0228, "step": 95805 }, { "epoch": 0.24, "learning_rate": 7.615028436257486e-05, "loss": 1.0247, "step": 95810 }, { "epoch": 0.24, "learning_rate": 7.614902612109315e-05, "loss": 1.0247, "step": 95815 }, { "epoch": 0.24, "learning_rate": 7.614776787961146e-05, "loss": 1.0223, "step": 95820 }, { "epoch": 0.24, "learning_rate": 7.614650963812975e-05, "loss": 1.0224, "step": 95825 }, { "epoch": 0.24, "learning_rate": 7.614525139664804e-05, "loss": 1.024, "step": 95830 }, { "epoch": 0.24, "learning_rate": 7.614399315516633e-05, "loss": 1.0256, "step": 95835 }, { "epoch": 0.24, "learning_rate": 7.614273491368464e-05, "loss": 1.0233, "step": 95840 }, { "epoch": 0.24, "learning_rate": 7.614147667220293e-05, "loss": 1.0237, "step": 95845 }, { "epoch": 0.24, "learning_rate": 7.614021843072124e-05, "loss": 1.0251, "step": 95850 }, { "epoch": 0.24, "learning_rate": 7.613896018923953e-05, "loss": 1.0457, "step": 95855 }, { "epoch": 0.24, "learning_rate": 7.613770194775782e-05, "loss": 1.0243, "step": 95860 }, { "epoch": 0.24, "learning_rate": 7.613644370627611e-05, "loss": 1.0238, "step": 95865 }, { "epoch": 0.24, "learning_rate": 7.613518546479442e-05, "loss": 1.0218, "step": 95870 }, { "epoch": 0.24, "learning_rate": 7.613417887160904e-05, "loss": 1.0244, "step": 95875 }, { "epoch": 0.24, "learning_rate": 7.613292063012733e-05, "loss": 1.0214, "step": 95880 }, { "epoch": 0.24, "learning_rate": 7.613166238864564e-05, "loss": 1.0253, "step": 95885 }, { "epoch": 0.24, "learning_rate": 7.613040414716393e-05, "loss": 1.0207, "step": 95890 }, { "epoch": 0.24, "learning_rate": 7.612914590568222e-05, "loss": 1.0235, "step": 95895 }, { "epoch": 0.24, "learning_rate": 7.612788766420051e-05, "loss": 1.0263, "step": 95900 }, { "epoch": 0.24, "learning_rate": 7.612662942271882e-05, "loss": 1.0262, "step": 95905 }, { "epoch": 0.24, "learning_rate": 7.612537118123711e-05, "loss": 1.0233, "step": 95910 }, { "epoch": 0.24, "learning_rate": 7.61241129397554e-05, "loss": 1.0205, "step": 95915 }, { "epoch": 0.24, "learning_rate": 7.612285469827369e-05, "loss": 1.0446, "step": 95920 }, { "epoch": 0.24, "learning_rate": 7.612159645679198e-05, "loss": 1.0244, "step": 95925 }, { "epoch": 0.24, "learning_rate": 7.612033821531029e-05, "loss": 1.0222, "step": 95930 }, { "epoch": 0.24, "learning_rate": 7.611907997382858e-05, "loss": 1.0227, "step": 95935 }, { "epoch": 0.24, "learning_rate": 7.611782173234687e-05, "loss": 1.0232, "step": 95940 }, { "epoch": 0.24, "learning_rate": 7.611656349086516e-05, "loss": 1.0241, "step": 95945 }, { "epoch": 0.24, "learning_rate": 7.611530524938347e-05, "loss": 1.023, "step": 95950 }, { "epoch": 0.24, "learning_rate": 7.611404700790176e-05, "loss": 1.0249, "step": 95955 }, { "epoch": 0.24, "learning_rate": 7.611278876642005e-05, "loss": 1.0227, "step": 95960 }, { "epoch": 0.24, "learning_rate": 7.611153052493834e-05, "loss": 1.0238, "step": 95965 }, { "epoch": 0.24, "learning_rate": 7.611027228345665e-05, "loss": 1.0244, "step": 95970 }, { "epoch": 0.24, "learning_rate": 7.610901404197494e-05, "loss": 1.0258, "step": 95975 }, { "epoch": 0.24, "learning_rate": 7.610775580049323e-05, "loss": 1.0225, "step": 95980 }, { "epoch": 0.24, "learning_rate": 7.610649755901152e-05, "loss": 1.0241, "step": 95985 }, { "epoch": 0.24, "learning_rate": 7.610523931752981e-05, "loss": 1.0254, "step": 95990 }, { "epoch": 0.24, "learning_rate": 7.610398107604812e-05, "loss": 1.0248, "step": 95995 }, { "epoch": 0.24, "learning_rate": 7.610272283456641e-05, "loss": 1.025, "step": 96000 }, { "epoch": 0.24, "learning_rate": 7.61014645930847e-05, "loss": 1.0234, "step": 96005 }, { "epoch": 0.24, "learning_rate": 7.6100206351603e-05, "loss": 1.0242, "step": 96010 }, { "epoch": 0.24, "learning_rate": 7.60989481101213e-05, "loss": 1.0207, "step": 96015 }, { "epoch": 0.24, "learning_rate": 7.609768986863959e-05, "loss": 1.0249, "step": 96020 }, { "epoch": 0.24, "learning_rate": 7.609643162715788e-05, "loss": 1.0405, "step": 96025 }, { "epoch": 0.24, "learning_rate": 7.609517338567617e-05, "loss": 1.0243, "step": 96030 }, { "epoch": 0.24, "learning_rate": 7.609391514419448e-05, "loss": 1.0214, "step": 96035 }, { "epoch": 0.24, "learning_rate": 7.609265690271278e-05, "loss": 1.0221, "step": 96040 }, { "epoch": 0.24, "learning_rate": 7.609139866123108e-05, "loss": 1.0214, "step": 96045 }, { "epoch": 0.24, "learning_rate": 7.609014041974937e-05, "loss": 1.0222, "step": 96050 }, { "epoch": 0.24, "learning_rate": 7.608888217826766e-05, "loss": 1.023, "step": 96055 }, { "epoch": 0.24, "learning_rate": 7.608762393678595e-05, "loss": 1.0234, "step": 96060 }, { "epoch": 0.24, "learning_rate": 7.608636569530426e-05, "loss": 1.0249, "step": 96065 }, { "epoch": 0.24, "learning_rate": 7.608510745382255e-05, "loss": 1.0234, "step": 96070 }, { "epoch": 0.24, "learning_rate": 7.608384921234084e-05, "loss": 1.0234, "step": 96075 }, { "epoch": 0.24, "learning_rate": 7.608259097085913e-05, "loss": 1.0244, "step": 96080 }, { "epoch": 0.24, "learning_rate": 7.608133272937744e-05, "loss": 1.0218, "step": 96085 }, { "epoch": 0.24, "learning_rate": 7.608007448789573e-05, "loss": 1.0223, "step": 96090 }, { "epoch": 0.24, "learning_rate": 7.607881624641402e-05, "loss": 1.0204, "step": 96095 }, { "epoch": 0.24, "learning_rate": 7.607755800493231e-05, "loss": 1.0216, "step": 96100 }, { "epoch": 0.24, "learning_rate": 7.607629976345062e-05, "loss": 1.0256, "step": 96105 }, { "epoch": 0.24, "learning_rate": 7.607504152196891e-05, "loss": 1.0218, "step": 96110 }, { "epoch": 0.24, "learning_rate": 7.60737832804872e-05, "loss": 1.0211, "step": 96115 }, { "epoch": 0.24, "learning_rate": 7.607252503900549e-05, "loss": 1.0255, "step": 96120 }, { "epoch": 0.24, "learning_rate": 7.607126679752378e-05, "loss": 1.0231, "step": 96125 }, { "epoch": 0.24, "learning_rate": 7.607000855604209e-05, "loss": 1.0237, "step": 96130 }, { "epoch": 0.24, "learning_rate": 7.606875031456038e-05, "loss": 1.0253, "step": 96135 }, { "epoch": 0.24, "learning_rate": 7.606749207307867e-05, "loss": 1.0247, "step": 96140 }, { "epoch": 0.24, "learning_rate": 7.606623383159696e-05, "loss": 1.0231, "step": 96145 }, { "epoch": 0.24, "learning_rate": 7.606497559011527e-05, "loss": 1.0231, "step": 96150 }, { "epoch": 0.24, "learning_rate": 7.606371734863356e-05, "loss": 1.0234, "step": 96155 }, { "epoch": 0.24, "learning_rate": 7.606245910715185e-05, "loss": 1.0216, "step": 96160 }, { "epoch": 0.24, "learning_rate": 7.606120086567014e-05, "loss": 1.0218, "step": 96165 }, { "epoch": 0.24, "learning_rate": 7.605994262418845e-05, "loss": 1.0233, "step": 96170 }, { "epoch": 0.24, "learning_rate": 7.605868438270674e-05, "loss": 1.025, "step": 96175 }, { "epoch": 0.24, "learning_rate": 7.605742614122503e-05, "loss": 1.0227, "step": 96180 }, { "epoch": 0.24, "learning_rate": 7.605616789974332e-05, "loss": 1.023, "step": 96185 }, { "epoch": 0.24, "learning_rate": 7.605490965826161e-05, "loss": 1.0253, "step": 96190 }, { "epoch": 0.24, "learning_rate": 7.605365141677992e-05, "loss": 1.0237, "step": 96195 }, { "epoch": 0.24, "learning_rate": 7.605239317529821e-05, "loss": 1.0241, "step": 96200 }, { "epoch": 0.24, "learning_rate": 7.60511349338165e-05, "loss": 1.0236, "step": 96205 }, { "epoch": 0.24, "learning_rate": 7.604987669233479e-05, "loss": 1.0221, "step": 96210 }, { "epoch": 0.24, "learning_rate": 7.60486184508531e-05, "loss": 1.0221, "step": 96215 }, { "epoch": 0.24, "learning_rate": 7.604736020937139e-05, "loss": 1.0227, "step": 96220 }, { "epoch": 0.24, "learning_rate": 7.604610196788968e-05, "loss": 1.0247, "step": 96225 }, { "epoch": 0.24, "learning_rate": 7.604484372640797e-05, "loss": 1.023, "step": 96230 }, { "epoch": 0.24, "learning_rate": 7.604358548492628e-05, "loss": 1.0218, "step": 96235 }, { "epoch": 0.24, "learning_rate": 7.604232724344457e-05, "loss": 1.0236, "step": 96240 }, { "epoch": 0.24, "learning_rate": 7.604106900196286e-05, "loss": 1.0237, "step": 96245 }, { "epoch": 0.24, "learning_rate": 7.603981076048115e-05, "loss": 1.0222, "step": 96250 }, { "epoch": 0.24, "learning_rate": 7.603855251899944e-05, "loss": 1.0211, "step": 96255 }, { "epoch": 0.24, "learning_rate": 7.603729427751775e-05, "loss": 1.0242, "step": 96260 }, { "epoch": 0.24, "learning_rate": 7.603603603603604e-05, "loss": 1.0229, "step": 96265 }, { "epoch": 0.24, "learning_rate": 7.603477779455433e-05, "loss": 1.0232, "step": 96270 }, { "epoch": 0.24, "learning_rate": 7.603351955307262e-05, "loss": 1.0212, "step": 96275 }, { "epoch": 0.24, "learning_rate": 7.603226131159093e-05, "loss": 1.0223, "step": 96280 }, { "epoch": 0.24, "learning_rate": 7.603100307010922e-05, "loss": 1.0221, "step": 96285 }, { "epoch": 0.24, "learning_rate": 7.602974482862751e-05, "loss": 1.0241, "step": 96290 }, { "epoch": 0.24, "learning_rate": 7.60284865871458e-05, "loss": 1.0209, "step": 96295 }, { "epoch": 0.24, "learning_rate": 7.602722834566411e-05, "loss": 1.0204, "step": 96300 }, { "epoch": 0.24, "learning_rate": 7.60259701041824e-05, "loss": 1.0254, "step": 96305 }, { "epoch": 0.24, "learning_rate": 7.602471186270069e-05, "loss": 1.0232, "step": 96310 }, { "epoch": 0.24, "learning_rate": 7.602345362121898e-05, "loss": 1.0246, "step": 96315 }, { "epoch": 0.24, "learning_rate": 7.602219537973728e-05, "loss": 1.0238, "step": 96320 }, { "epoch": 0.24, "learning_rate": 7.602093713825558e-05, "loss": 1.0233, "step": 96325 }, { "epoch": 0.24, "learning_rate": 7.601967889677387e-05, "loss": 1.0219, "step": 96330 }, { "epoch": 0.24, "learning_rate": 7.601842065529216e-05, "loss": 1.0246, "step": 96335 }, { "epoch": 0.24, "learning_rate": 7.601716241381046e-05, "loss": 1.0238, "step": 96340 }, { "epoch": 0.24, "learning_rate": 7.601590417232876e-05, "loss": 1.0228, "step": 96345 }, { "epoch": 0.24, "learning_rate": 7.601464593084705e-05, "loss": 1.0249, "step": 96350 }, { "epoch": 0.24, "learning_rate": 7.601338768936534e-05, "loss": 1.0449, "step": 96355 }, { "epoch": 0.24, "learning_rate": 7.601212944788364e-05, "loss": 1.0252, "step": 96360 }, { "epoch": 0.24, "learning_rate": 7.601087120640194e-05, "loss": 1.0216, "step": 96365 }, { "epoch": 0.24, "learning_rate": 7.600961296492023e-05, "loss": 1.0477, "step": 96370 }, { "epoch": 0.24, "learning_rate": 7.600835472343852e-05, "loss": 1.0249, "step": 96375 }, { "epoch": 0.24, "learning_rate": 7.600709648195682e-05, "loss": 1.0248, "step": 96380 }, { "epoch": 0.24, "learning_rate": 7.600583824047511e-05, "loss": 1.0259, "step": 96385 }, { "epoch": 0.24, "learning_rate": 7.600457999899341e-05, "loss": 1.0252, "step": 96390 }, { "epoch": 0.24, "learning_rate": 7.60033217575117e-05, "loss": 1.0227, "step": 96395 }, { "epoch": 0.24, "learning_rate": 7.600206351603e-05, "loss": 1.0233, "step": 96400 }, { "epoch": 0.24, "learning_rate": 7.600080527454829e-05, "loss": 1.0248, "step": 96405 }, { "epoch": 0.24, "learning_rate": 7.599954703306659e-05, "loss": 1.0228, "step": 96410 }, { "epoch": 0.24, "learning_rate": 7.599828879158488e-05, "loss": 1.0245, "step": 96415 }, { "epoch": 0.24, "learning_rate": 7.599703055010318e-05, "loss": 1.0462, "step": 96420 }, { "epoch": 0.24, "learning_rate": 7.599577230862147e-05, "loss": 1.0226, "step": 96425 }, { "epoch": 0.24, "learning_rate": 7.599451406713977e-05, "loss": 1.0215, "step": 96430 }, { "epoch": 0.24, "learning_rate": 7.599325582565806e-05, "loss": 1.0222, "step": 96435 }, { "epoch": 0.24, "learning_rate": 7.599199758417636e-05, "loss": 1.0207, "step": 96440 }, { "epoch": 0.24, "learning_rate": 7.599073934269465e-05, "loss": 1.0232, "step": 96445 }, { "epoch": 0.24, "learning_rate": 7.598948110121294e-05, "loss": 1.0224, "step": 96450 }, { "epoch": 0.24, "learning_rate": 7.598822285973124e-05, "loss": 1.0254, "step": 96455 }, { "epoch": 0.24, "learning_rate": 7.598696461824954e-05, "loss": 1.0238, "step": 96460 }, { "epoch": 0.24, "learning_rate": 7.598570637676783e-05, "loss": 1.0235, "step": 96465 }, { "epoch": 0.24, "learning_rate": 7.598444813528612e-05, "loss": 1.0223, "step": 96470 }, { "epoch": 0.24, "learning_rate": 7.598318989380442e-05, "loss": 1.0247, "step": 96475 }, { "epoch": 0.24, "learning_rate": 7.598193165232272e-05, "loss": 1.0237, "step": 96480 }, { "epoch": 0.24, "learning_rate": 7.598067341084101e-05, "loss": 1.0237, "step": 96485 }, { "epoch": 0.24, "learning_rate": 7.59794151693593e-05, "loss": 1.0252, "step": 96490 }, { "epoch": 0.24, "learning_rate": 7.59781569278776e-05, "loss": 1.0234, "step": 96495 }, { "epoch": 0.24, "learning_rate": 7.59768986863959e-05, "loss": 1.0248, "step": 96500 }, { "epoch": 0.24, "learning_rate": 7.597564044491419e-05, "loss": 1.0231, "step": 96505 }, { "epoch": 0.24, "learning_rate": 7.597438220343248e-05, "loss": 1.0475, "step": 96510 }, { "epoch": 0.24, "learning_rate": 7.597312396195077e-05, "loss": 1.0226, "step": 96515 }, { "epoch": 0.24, "learning_rate": 7.597186572046908e-05, "loss": 1.0202, "step": 96520 }, { "epoch": 0.24, "learning_rate": 7.597060747898737e-05, "loss": 1.0223, "step": 96525 }, { "epoch": 0.24, "learning_rate": 7.596934923750566e-05, "loss": 1.0247, "step": 96530 }, { "epoch": 0.24, "learning_rate": 7.596809099602395e-05, "loss": 1.0227, "step": 96535 }, { "epoch": 0.24, "learning_rate": 7.596683275454226e-05, "loss": 1.0214, "step": 96540 }, { "epoch": 0.24, "learning_rate": 7.596557451306056e-05, "loss": 1.0231, "step": 96545 }, { "epoch": 0.24, "learning_rate": 7.596431627157885e-05, "loss": 1.0449, "step": 96550 }, { "epoch": 0.24, "learning_rate": 7.596305803009714e-05, "loss": 1.0233, "step": 96555 }, { "epoch": 0.24, "learning_rate": 7.596179978861544e-05, "loss": 1.0224, "step": 96560 }, { "epoch": 0.24, "learning_rate": 7.596054154713374e-05, "loss": 1.0216, "step": 96565 }, { "epoch": 0.24, "learning_rate": 7.595928330565203e-05, "loss": 1.0238, "step": 96570 }, { "epoch": 0.24, "learning_rate": 7.595802506417032e-05, "loss": 1.0242, "step": 96575 }, { "epoch": 0.24, "learning_rate": 7.595676682268862e-05, "loss": 1.0233, "step": 96580 }, { "epoch": 0.24, "learning_rate": 7.59555085812069e-05, "loss": 1.0409, "step": 96585 }, { "epoch": 0.24, "learning_rate": 7.595425033972521e-05, "loss": 1.0193, "step": 96590 }, { "epoch": 0.24, "learning_rate": 7.59529920982435e-05, "loss": 1.0225, "step": 96595 }, { "epoch": 0.24, "learning_rate": 7.59517338567618e-05, "loss": 1.0251, "step": 96600 }, { "epoch": 0.24, "learning_rate": 7.595047561528009e-05, "loss": 1.0232, "step": 96605 }, { "epoch": 0.24, "learning_rate": 7.594921737379839e-05, "loss": 1.0236, "step": 96610 }, { "epoch": 0.24, "learning_rate": 7.594795913231668e-05, "loss": 1.0247, "step": 96615 }, { "epoch": 0.24, "learning_rate": 7.594670089083498e-05, "loss": 1.0238, "step": 96620 }, { "epoch": 0.24, "learning_rate": 7.594544264935327e-05, "loss": 1.0221, "step": 96625 }, { "epoch": 0.24, "learning_rate": 7.594418440787157e-05, "loss": 1.0268, "step": 96630 }, { "epoch": 0.24, "learning_rate": 7.594292616638986e-05, "loss": 1.0226, "step": 96635 }, { "epoch": 0.24, "learning_rate": 7.594166792490815e-05, "loss": 1.0224, "step": 96640 }, { "epoch": 0.24, "learning_rate": 7.594040968342645e-05, "loss": 1.0244, "step": 96645 }, { "epoch": 0.24, "learning_rate": 7.593915144194474e-05, "loss": 1.0235, "step": 96650 }, { "epoch": 0.24, "learning_rate": 7.593789320046304e-05, "loss": 1.0224, "step": 96655 }, { "epoch": 0.24, "learning_rate": 7.593663495898133e-05, "loss": 1.0211, "step": 96660 }, { "epoch": 0.24, "learning_rate": 7.593537671749963e-05, "loss": 1.0243, "step": 96665 }, { "epoch": 0.24, "learning_rate": 7.593411847601792e-05, "loss": 1.0239, "step": 96670 }, { "epoch": 0.24, "learning_rate": 7.593286023453622e-05, "loss": 1.0245, "step": 96675 }, { "epoch": 0.24, "learning_rate": 7.593160199305451e-05, "loss": 1.0235, "step": 96680 }, { "epoch": 0.24, "learning_rate": 7.59303437515728e-05, "loss": 1.0227, "step": 96685 }, { "epoch": 0.24, "learning_rate": 7.59290855100911e-05, "loss": 1.0232, "step": 96690 }, { "epoch": 0.24, "learning_rate": 7.59278272686094e-05, "loss": 1.0205, "step": 96695 }, { "epoch": 0.24, "learning_rate": 7.59265690271277e-05, "loss": 1.0254, "step": 96700 }, { "epoch": 0.24, "learning_rate": 7.592531078564599e-05, "loss": 1.0249, "step": 96705 }, { "epoch": 0.24, "learning_rate": 7.592405254416428e-05, "loss": 1.0503, "step": 96710 }, { "epoch": 0.24, "learning_rate": 7.592279430268257e-05, "loss": 1.0242, "step": 96715 }, { "epoch": 0.24, "learning_rate": 7.592153606120087e-05, "loss": 1.0223, "step": 96720 }, { "epoch": 0.24, "learning_rate": 7.592027781971917e-05, "loss": 1.024, "step": 96725 }, { "epoch": 0.24, "learning_rate": 7.591901957823746e-05, "loss": 1.024, "step": 96730 }, { "epoch": 0.24, "learning_rate": 7.591776133675575e-05, "loss": 1.0215, "step": 96735 }, { "epoch": 0.24, "learning_rate": 7.591650309527405e-05, "loss": 1.0237, "step": 96740 }, { "epoch": 0.24, "learning_rate": 7.591524485379235e-05, "loss": 1.023, "step": 96745 }, { "epoch": 0.24, "learning_rate": 7.591398661231064e-05, "loss": 1.021, "step": 96750 }, { "epoch": 0.24, "learning_rate": 7.591272837082893e-05, "loss": 1.0225, "step": 96755 }, { "epoch": 0.24, "learning_rate": 7.591147012934723e-05, "loss": 1.025, "step": 96760 }, { "epoch": 0.24, "learning_rate": 7.591021188786553e-05, "loss": 1.0233, "step": 96765 }, { "epoch": 0.24, "learning_rate": 7.590895364638382e-05, "loss": 1.0222, "step": 96770 }, { "epoch": 0.24, "learning_rate": 7.590769540490211e-05, "loss": 1.0234, "step": 96775 }, { "epoch": 0.24, "learning_rate": 7.59064371634204e-05, "loss": 1.0248, "step": 96780 }, { "epoch": 0.24, "learning_rate": 7.59051789219387e-05, "loss": 1.0222, "step": 96785 }, { "epoch": 0.24, "learning_rate": 7.5903920680457e-05, "loss": 1.0228, "step": 96790 }, { "epoch": 0.24, "learning_rate": 7.590266243897529e-05, "loss": 1.0214, "step": 96795 }, { "epoch": 0.24, "learning_rate": 7.590140419749358e-05, "loss": 1.0217, "step": 96800 }, { "epoch": 0.24, "learning_rate": 7.590014595601189e-05, "loss": 1.0231, "step": 96805 }, { "epoch": 0.24, "learning_rate": 7.589888771453018e-05, "loss": 1.0233, "step": 96810 }, { "epoch": 0.24, "learning_rate": 7.589762947304847e-05, "loss": 1.024, "step": 96815 }, { "epoch": 0.24, "learning_rate": 7.589637123156676e-05, "loss": 1.0232, "step": 96820 }, { "epoch": 0.24, "learning_rate": 7.589511299008507e-05, "loss": 1.0229, "step": 96825 }, { "epoch": 0.24, "learning_rate": 7.589385474860336e-05, "loss": 1.023, "step": 96830 }, { "epoch": 0.24, "learning_rate": 7.589259650712165e-05, "loss": 1.0231, "step": 96835 }, { "epoch": 0.24, "learning_rate": 7.589133826563994e-05, "loss": 1.0239, "step": 96840 }, { "epoch": 0.24, "learning_rate": 7.589008002415823e-05, "loss": 1.0229, "step": 96845 }, { "epoch": 0.24, "learning_rate": 7.588882178267654e-05, "loss": 1.0219, "step": 96850 }, { "epoch": 0.24, "learning_rate": 7.588756354119483e-05, "loss": 1.0229, "step": 96855 }, { "epoch": 0.24, "learning_rate": 7.588630529971312e-05, "loss": 1.0242, "step": 96860 }, { "epoch": 0.24, "learning_rate": 7.588504705823141e-05, "loss": 1.023, "step": 96865 }, { "epoch": 0.24, "learning_rate": 7.588378881674972e-05, "loss": 1.0213, "step": 96870 }, { "epoch": 0.24, "learning_rate": 7.588253057526801e-05, "loss": 1.0214, "step": 96875 }, { "epoch": 0.24, "learning_rate": 7.58812723337863e-05, "loss": 1.0258, "step": 96880 }, { "epoch": 0.24, "learning_rate": 7.588001409230459e-05, "loss": 1.0228, "step": 96885 }, { "epoch": 0.24, "learning_rate": 7.58787558508229e-05, "loss": 1.0219, "step": 96890 }, { "epoch": 0.24, "learning_rate": 7.587749760934119e-05, "loss": 1.0211, "step": 96895 }, { "epoch": 0.24, "learning_rate": 7.587623936785948e-05, "loss": 1.0219, "step": 96900 }, { "epoch": 0.24, "learning_rate": 7.587498112637777e-05, "loss": 1.0244, "step": 96905 }, { "epoch": 0.24, "learning_rate": 7.587372288489606e-05, "loss": 1.0242, "step": 96910 }, { "epoch": 0.24, "learning_rate": 7.587246464341437e-05, "loss": 1.0241, "step": 96915 }, { "epoch": 0.24, "learning_rate": 7.587120640193266e-05, "loss": 1.0234, "step": 96920 }, { "epoch": 0.24, "learning_rate": 7.586994816045095e-05, "loss": 1.0457, "step": 96925 }, { "epoch": 0.24, "learning_rate": 7.586868991896924e-05, "loss": 1.0239, "step": 96930 }, { "epoch": 0.24, "learning_rate": 7.586743167748755e-05, "loss": 1.0228, "step": 96935 }, { "epoch": 0.24, "learning_rate": 7.586617343600584e-05, "loss": 1.0247, "step": 96940 }, { "epoch": 0.24, "learning_rate": 7.586491519452413e-05, "loss": 1.0213, "step": 96945 }, { "epoch": 0.24, "learning_rate": 7.586365695304242e-05, "loss": 1.0256, "step": 96950 }, { "epoch": 0.24, "learning_rate": 7.586239871156073e-05, "loss": 1.0253, "step": 96955 }, { "epoch": 0.24, "learning_rate": 7.586114047007902e-05, "loss": 1.0196, "step": 96960 }, { "epoch": 0.24, "learning_rate": 7.585988222859731e-05, "loss": 1.024, "step": 96965 }, { "epoch": 0.24, "learning_rate": 7.58586239871156e-05, "loss": 1.0221, "step": 96970 }, { "epoch": 0.24, "learning_rate": 7.58573657456339e-05, "loss": 1.0231, "step": 96975 }, { "epoch": 0.24, "learning_rate": 7.58561075041522e-05, "loss": 1.0255, "step": 96980 }, { "epoch": 0.24, "learning_rate": 7.585484926267049e-05, "loss": 1.0275, "step": 96985 }, { "epoch": 0.24, "learning_rate": 7.585359102118878e-05, "loss": 1.021, "step": 96990 }, { "epoch": 0.24, "learning_rate": 7.585233277970708e-05, "loss": 1.0223, "step": 96995 }, { "epoch": 0.24, "learning_rate": 7.585107453822538e-05, "loss": 1.025, "step": 97000 }, { "epoch": 0.24, "learning_rate": 7.584981629674367e-05, "loss": 1.0228, "step": 97005 }, { "epoch": 0.24, "learning_rate": 7.584855805526196e-05, "loss": 1.0232, "step": 97010 }, { "epoch": 0.24, "learning_rate": 7.584729981378026e-05, "loss": 1.0242, "step": 97015 }, { "epoch": 0.24, "learning_rate": 7.584604157229856e-05, "loss": 1.025, "step": 97020 }, { "epoch": 0.24, "learning_rate": 7.584478333081685e-05, "loss": 1.0239, "step": 97025 }, { "epoch": 0.24, "learning_rate": 7.584352508933514e-05, "loss": 1.0233, "step": 97030 }, { "epoch": 0.24, "learning_rate": 7.584226684785344e-05, "loss": 1.0235, "step": 97035 }, { "epoch": 0.24, "learning_rate": 7.584100860637174e-05, "loss": 1.0249, "step": 97040 }, { "epoch": 0.24, "learning_rate": 7.583975036489005e-05, "loss": 1.0225, "step": 97045 }, { "epoch": 0.24, "learning_rate": 7.583849212340834e-05, "loss": 1.0225, "step": 97050 }, { "epoch": 0.24, "learning_rate": 7.583723388192663e-05, "loss": 1.0222, "step": 97055 }, { "epoch": 0.24, "learning_rate": 7.583597564044492e-05, "loss": 1.0252, "step": 97060 }, { "epoch": 0.24, "learning_rate": 7.583471739896321e-05, "loss": 1.0231, "step": 97065 }, { "epoch": 0.24, "learning_rate": 7.583345915748152e-05, "loss": 1.0202, "step": 97070 }, { "epoch": 0.24, "learning_rate": 7.583220091599981e-05, "loss": 1.0215, "step": 97075 }, { "epoch": 0.24, "learning_rate": 7.58309426745181e-05, "loss": 1.0232, "step": 97080 }, { "epoch": 0.24, "learning_rate": 7.582968443303639e-05, "loss": 1.0242, "step": 97085 }, { "epoch": 0.24, "learning_rate": 7.58284261915547e-05, "loss": 1.0237, "step": 97090 }, { "epoch": 0.24, "learning_rate": 7.582716795007299e-05, "loss": 1.0216, "step": 97095 }, { "epoch": 0.24, "learning_rate": 7.582590970859128e-05, "loss": 1.0232, "step": 97100 }, { "epoch": 0.24, "learning_rate": 7.582465146710957e-05, "loss": 1.0253, "step": 97105 }, { "epoch": 0.24, "learning_rate": 7.582339322562788e-05, "loss": 1.0249, "step": 97110 }, { "epoch": 0.24, "learning_rate": 7.582213498414617e-05, "loss": 1.0223, "step": 97115 }, { "epoch": 0.24, "learning_rate": 7.582087674266446e-05, "loss": 1.0238, "step": 97120 }, { "epoch": 0.24, "learning_rate": 7.581961850118275e-05, "loss": 1.0219, "step": 97125 }, { "epoch": 0.24, "learning_rate": 7.581836025970104e-05, "loss": 1.0228, "step": 97130 }, { "epoch": 0.24, "learning_rate": 7.581710201821935e-05, "loss": 1.0211, "step": 97135 }, { "epoch": 0.24, "learning_rate": 7.581584377673764e-05, "loss": 1.0218, "step": 97140 }, { "epoch": 0.24, "learning_rate": 7.581458553525593e-05, "loss": 1.0243, "step": 97145 }, { "epoch": 0.24, "learning_rate": 7.581332729377422e-05, "loss": 1.0227, "step": 97150 }, { "epoch": 0.24, "learning_rate": 7.581206905229253e-05, "loss": 1.0231, "step": 97155 }, { "epoch": 0.24, "learning_rate": 7.581081081081082e-05, "loss": 1.0226, "step": 97160 }, { "epoch": 0.24, "learning_rate": 7.580955256932911e-05, "loss": 1.0223, "step": 97165 }, { "epoch": 0.24, "learning_rate": 7.58082943278474e-05, "loss": 1.0229, "step": 97170 }, { "epoch": 0.24, "learning_rate": 7.58070360863657e-05, "loss": 1.0243, "step": 97175 }, { "epoch": 0.24, "learning_rate": 7.5805777844884e-05, "loss": 1.0222, "step": 97180 }, { "epoch": 0.24, "learning_rate": 7.580451960340229e-05, "loss": 1.0251, "step": 97185 }, { "epoch": 0.24, "learning_rate": 7.580326136192058e-05, "loss": 1.0391, "step": 97190 }, { "epoch": 0.24, "learning_rate": 7.580200312043887e-05, "loss": 1.0254, "step": 97195 }, { "epoch": 0.24, "learning_rate": 7.580074487895718e-05, "loss": 1.0239, "step": 97200 }, { "epoch": 0.24, "learning_rate": 7.579948663747547e-05, "loss": 1.0225, "step": 97205 }, { "epoch": 0.24, "learning_rate": 7.579822839599376e-05, "loss": 1.0244, "step": 97210 }, { "epoch": 0.24, "learning_rate": 7.579697015451205e-05, "loss": 1.023, "step": 97215 }, { "epoch": 0.24, "learning_rate": 7.579571191303036e-05, "loss": 1.0255, "step": 97220 }, { "epoch": 0.24, "learning_rate": 7.579445367154865e-05, "loss": 1.025, "step": 97225 }, { "epoch": 0.24, "learning_rate": 7.579319543006694e-05, "loss": 1.0227, "step": 97230 }, { "epoch": 0.24, "learning_rate": 7.579193718858523e-05, "loss": 1.0222, "step": 97235 }, { "epoch": 0.24, "learning_rate": 7.579067894710353e-05, "loss": 1.0231, "step": 97240 }, { "epoch": 0.24, "learning_rate": 7.578942070562183e-05, "loss": 1.0239, "step": 97245 }, { "epoch": 0.24, "learning_rate": 7.578816246414012e-05, "loss": 1.0224, "step": 97250 }, { "epoch": 0.24, "learning_rate": 7.578690422265841e-05, "loss": 1.0206, "step": 97255 }, { "epoch": 0.24, "learning_rate": 7.57856459811767e-05, "loss": 1.0236, "step": 97260 }, { "epoch": 0.24, "learning_rate": 7.578438773969501e-05, "loss": 1.0437, "step": 97265 }, { "epoch": 0.24, "learning_rate": 7.57831294982133e-05, "loss": 1.0242, "step": 97270 }, { "epoch": 0.24, "learning_rate": 7.57818712567316e-05, "loss": 1.0241, "step": 97275 }, { "epoch": 0.24, "learning_rate": 7.578061301524989e-05, "loss": 1.0254, "step": 97280 }, { "epoch": 0.24, "learning_rate": 7.577935477376819e-05, "loss": 1.0241, "step": 97285 }, { "epoch": 0.24, "learning_rate": 7.577809653228648e-05, "loss": 1.0241, "step": 97290 }, { "epoch": 0.24, "learning_rate": 7.577683829080477e-05, "loss": 1.0232, "step": 97295 }, { "epoch": 0.24, "learning_rate": 7.577558004932307e-05, "loss": 1.0444, "step": 97300 }, { "epoch": 0.24, "learning_rate": 7.577432180784136e-05, "loss": 1.0215, "step": 97305 }, { "epoch": 0.24, "learning_rate": 7.577306356635966e-05, "loss": 1.0231, "step": 97310 }, { "epoch": 0.24, "learning_rate": 7.577180532487795e-05, "loss": 1.0221, "step": 97315 }, { "epoch": 0.24, "learning_rate": 7.577054708339625e-05, "loss": 1.0232, "step": 97320 }, { "epoch": 0.24, "learning_rate": 7.576928884191454e-05, "loss": 1.0237, "step": 97325 }, { "epoch": 0.24, "learning_rate": 7.576803060043284e-05, "loss": 1.024, "step": 97330 }, { "epoch": 0.24, "learning_rate": 7.576677235895113e-05, "loss": 1.0236, "step": 97335 }, { "epoch": 0.24, "learning_rate": 7.576551411746943e-05, "loss": 1.0234, "step": 97340 }, { "epoch": 0.24, "learning_rate": 7.576425587598772e-05, "loss": 1.0242, "step": 97345 }, { "epoch": 0.24, "learning_rate": 7.576299763450602e-05, "loss": 1.0233, "step": 97350 }, { "epoch": 0.24, "learning_rate": 7.576173939302431e-05, "loss": 1.0214, "step": 97355 }, { "epoch": 0.24, "learning_rate": 7.57604811515426e-05, "loss": 1.0236, "step": 97360 }, { "epoch": 0.24, "learning_rate": 7.57592229100609e-05, "loss": 1.0239, "step": 97365 }, { "epoch": 0.24, "learning_rate": 7.575796466857919e-05, "loss": 1.0232, "step": 97370 }, { "epoch": 0.24, "learning_rate": 7.57567064270975e-05, "loss": 1.0226, "step": 97375 }, { "epoch": 0.24, "learning_rate": 7.575544818561579e-05, "loss": 1.026, "step": 97380 }, { "epoch": 0.24, "learning_rate": 7.575418994413408e-05, "loss": 1.0216, "step": 97385 }, { "epoch": 0.24, "learning_rate": 7.575293170265237e-05, "loss": 1.0226, "step": 97390 }, { "epoch": 0.24, "learning_rate": 7.575167346117067e-05, "loss": 1.0238, "step": 97395 }, { "epoch": 0.24, "learning_rate": 7.575041521968897e-05, "loss": 1.0244, "step": 97400 }, { "epoch": 0.24, "learning_rate": 7.574915697820726e-05, "loss": 1.0206, "step": 97405 }, { "epoch": 0.24, "learning_rate": 7.574789873672555e-05, "loss": 1.0214, "step": 97410 }, { "epoch": 0.24, "learning_rate": 7.574664049524385e-05, "loss": 1.0239, "step": 97415 }, { "epoch": 0.24, "learning_rate": 7.574538225376215e-05, "loss": 1.0239, "step": 97420 }, { "epoch": 0.24, "learning_rate": 7.574412401228044e-05, "loss": 1.0248, "step": 97425 }, { "epoch": 0.24, "learning_rate": 7.574286577079873e-05, "loss": 1.0227, "step": 97430 }, { "epoch": 0.24, "learning_rate": 7.574160752931702e-05, "loss": 1.0242, "step": 97435 }, { "epoch": 0.24, "learning_rate": 7.574034928783533e-05, "loss": 1.024, "step": 97440 }, { "epoch": 0.24, "learning_rate": 7.573909104635362e-05, "loss": 1.0239, "step": 97445 }, { "epoch": 0.24, "learning_rate": 7.573783280487191e-05, "loss": 1.0251, "step": 97450 }, { "epoch": 0.24, "learning_rate": 7.57365745633902e-05, "loss": 1.0245, "step": 97455 }, { "epoch": 0.24, "learning_rate": 7.57353163219085e-05, "loss": 1.0256, "step": 97460 }, { "epoch": 0.24, "learning_rate": 7.57340580804268e-05, "loss": 1.0245, "step": 97465 }, { "epoch": 0.24, "learning_rate": 7.573279983894509e-05, "loss": 1.0234, "step": 97470 }, { "epoch": 0.24, "learning_rate": 7.573154159746338e-05, "loss": 1.0229, "step": 97475 }, { "epoch": 0.24, "learning_rate": 7.573028335598169e-05, "loss": 1.0229, "step": 97480 }, { "epoch": 0.24, "learning_rate": 7.572902511449998e-05, "loss": 1.0213, "step": 97485 }, { "epoch": 0.24, "learning_rate": 7.572776687301827e-05, "loss": 1.0237, "step": 97490 }, { "epoch": 0.24, "learning_rate": 7.572650863153656e-05, "loss": 1.022, "step": 97495 }, { "epoch": 0.24, "learning_rate": 7.572525039005485e-05, "loss": 1.0256, "step": 97500 }, { "epoch": 0.24, "learning_rate": 7.572399214857316e-05, "loss": 1.0247, "step": 97505 }, { "epoch": 0.24, "learning_rate": 7.572273390709145e-05, "loss": 1.0233, "step": 97510 }, { "epoch": 0.24, "learning_rate": 7.572147566560974e-05, "loss": 1.0245, "step": 97515 }, { "epoch": 0.24, "learning_rate": 7.572021742412803e-05, "loss": 1.0236, "step": 97520 }, { "epoch": 0.24, "learning_rate": 7.571895918264634e-05, "loss": 1.024, "step": 97525 }, { "epoch": 0.24, "learning_rate": 7.571770094116463e-05, "loss": 1.0225, "step": 97530 }, { "epoch": 0.24, "learning_rate": 7.571644269968292e-05, "loss": 1.0235, "step": 97535 }, { "epoch": 0.24, "learning_rate": 7.571518445820123e-05, "loss": 1.0214, "step": 97540 }, { "epoch": 0.24, "learning_rate": 7.571392621671952e-05, "loss": 1.021, "step": 97545 }, { "epoch": 0.24, "learning_rate": 7.571266797523782e-05, "loss": 1.0226, "step": 97550 }, { "epoch": 0.24, "learning_rate": 7.571140973375611e-05, "loss": 1.0252, "step": 97555 }, { "epoch": 0.24, "learning_rate": 7.57101514922744e-05, "loss": 1.0231, "step": 97560 }, { "epoch": 0.24, "learning_rate": 7.57088932507927e-05, "loss": 1.0249, "step": 97565 }, { "epoch": 0.24, "learning_rate": 7.5707635009311e-05, "loss": 1.0241, "step": 97570 }, { "epoch": 0.24, "learning_rate": 7.57063767678293e-05, "loss": 1.0251, "step": 97575 }, { "epoch": 0.24, "learning_rate": 7.570511852634759e-05, "loss": 1.0233, "step": 97580 }, { "epoch": 0.24, "learning_rate": 7.570386028486588e-05, "loss": 1.0211, "step": 97585 }, { "epoch": 0.24, "learning_rate": 7.570260204338417e-05, "loss": 1.024, "step": 97590 }, { "epoch": 0.24, "learning_rate": 7.570134380190247e-05, "loss": 1.0249, "step": 97595 }, { "epoch": 0.24, "learning_rate": 7.570008556042077e-05, "loss": 1.0212, "step": 97600 }, { "epoch": 0.25, "learning_rate": 7.569882731893906e-05, "loss": 1.0241, "step": 97605 }, { "epoch": 0.25, "learning_rate": 7.569756907745735e-05, "loss": 1.0245, "step": 97610 }, { "epoch": 0.25, "learning_rate": 7.569631083597565e-05, "loss": 1.0244, "step": 97615 }, { "epoch": 0.25, "learning_rate": 7.569505259449395e-05, "loss": 1.0246, "step": 97620 }, { "epoch": 0.25, "learning_rate": 7.569379435301224e-05, "loss": 1.0211, "step": 97625 }, { "epoch": 0.25, "learning_rate": 7.569253611153053e-05, "loss": 1.0234, "step": 97630 }, { "epoch": 0.25, "learning_rate": 7.569127787004883e-05, "loss": 1.0242, "step": 97635 }, { "epoch": 0.25, "learning_rate": 7.569001962856712e-05, "loss": 1.0237, "step": 97640 }, { "epoch": 0.25, "learning_rate": 7.568876138708542e-05, "loss": 1.0242, "step": 97645 }, { "epoch": 0.25, "learning_rate": 7.568750314560371e-05, "loss": 1.0232, "step": 97650 }, { "epoch": 0.25, "learning_rate": 7.5686244904122e-05, "loss": 1.0236, "step": 97655 }, { "epoch": 0.25, "learning_rate": 7.56849866626403e-05, "loss": 1.0213, "step": 97660 }, { "epoch": 0.25, "learning_rate": 7.56837284211586e-05, "loss": 1.0215, "step": 97665 }, { "epoch": 0.25, "learning_rate": 7.568247017967689e-05, "loss": 1.0219, "step": 97670 }, { "epoch": 0.25, "learning_rate": 7.568121193819518e-05, "loss": 1.0231, "step": 97675 }, { "epoch": 0.25, "learning_rate": 7.567995369671348e-05, "loss": 1.0233, "step": 97680 }, { "epoch": 0.25, "learning_rate": 7.567869545523178e-05, "loss": 1.0273, "step": 97685 }, { "epoch": 0.25, "learning_rate": 7.567743721375007e-05, "loss": 1.0229, "step": 97690 }, { "epoch": 0.25, "learning_rate": 7.567617897226836e-05, "loss": 1.0255, "step": 97695 }, { "epoch": 0.25, "learning_rate": 7.567492073078666e-05, "loss": 1.0199, "step": 97700 }, { "epoch": 0.25, "learning_rate": 7.567366248930496e-05, "loss": 1.0217, "step": 97705 }, { "epoch": 0.25, "learning_rate": 7.567240424782325e-05, "loss": 1.0246, "step": 97710 }, { "epoch": 0.25, "learning_rate": 7.567114600634154e-05, "loss": 1.0222, "step": 97715 }, { "epoch": 0.25, "learning_rate": 7.566988776485983e-05, "loss": 1.0247, "step": 97720 }, { "epoch": 0.25, "learning_rate": 7.566862952337814e-05, "loss": 1.0342, "step": 97725 }, { "epoch": 0.25, "learning_rate": 7.566737128189643e-05, "loss": 1.0242, "step": 97730 }, { "epoch": 0.25, "learning_rate": 7.566611304041472e-05, "loss": 1.021, "step": 97735 }, { "epoch": 0.25, "learning_rate": 7.566485479893301e-05, "loss": 1.0226, "step": 97740 }, { "epoch": 0.25, "learning_rate": 7.566359655745132e-05, "loss": 1.0251, "step": 97745 }, { "epoch": 0.25, "learning_rate": 7.566233831596961e-05, "loss": 1.0212, "step": 97750 }, { "epoch": 0.25, "learning_rate": 7.56610800744879e-05, "loss": 1.0218, "step": 97755 }, { "epoch": 0.25, "learning_rate": 7.565982183300619e-05, "loss": 1.0227, "step": 97760 }, { "epoch": 0.25, "learning_rate": 7.56585635915245e-05, "loss": 1.0232, "step": 97765 }, { "epoch": 0.25, "learning_rate": 7.565730535004279e-05, "loss": 1.0231, "step": 97770 }, { "epoch": 0.25, "learning_rate": 7.565604710856108e-05, "loss": 1.0222, "step": 97775 }, { "epoch": 0.25, "learning_rate": 7.565478886707937e-05, "loss": 1.0219, "step": 97780 }, { "epoch": 0.25, "learning_rate": 7.565353062559766e-05, "loss": 1.0225, "step": 97785 }, { "epoch": 0.25, "learning_rate": 7.565227238411597e-05, "loss": 1.0218, "step": 97790 }, { "epoch": 0.25, "learning_rate": 7.565101414263426e-05, "loss": 1.0467, "step": 97795 }, { "epoch": 0.25, "learning_rate": 7.564975590115255e-05, "loss": 1.0239, "step": 97800 }, { "epoch": 0.25, "learning_rate": 7.564849765967084e-05, "loss": 1.024, "step": 97805 }, { "epoch": 0.25, "learning_rate": 7.564723941818915e-05, "loss": 1.0227, "step": 97810 }, { "epoch": 0.25, "learning_rate": 7.564598117670744e-05, "loss": 1.0243, "step": 97815 }, { "epoch": 0.25, "learning_rate": 7.564472293522573e-05, "loss": 1.0441, "step": 97820 }, { "epoch": 0.25, "learning_rate": 7.564346469374402e-05, "loss": 1.0232, "step": 97825 }, { "epoch": 0.25, "learning_rate": 7.564220645226231e-05, "loss": 1.0265, "step": 97830 }, { "epoch": 0.25, "learning_rate": 7.564094821078062e-05, "loss": 1.0241, "step": 97835 }, { "epoch": 0.25, "learning_rate": 7.563968996929891e-05, "loss": 1.0228, "step": 97840 }, { "epoch": 0.25, "learning_rate": 7.56384317278172e-05, "loss": 1.0227, "step": 97845 }, { "epoch": 0.25, "learning_rate": 7.56371734863355e-05, "loss": 1.0237, "step": 97850 }, { "epoch": 0.25, "learning_rate": 7.56359152448538e-05, "loss": 1.022, "step": 97855 }, { "epoch": 0.25, "learning_rate": 7.563465700337209e-05, "loss": 1.0237, "step": 97860 }, { "epoch": 0.25, "learning_rate": 7.563339876189038e-05, "loss": 1.022, "step": 97865 }, { "epoch": 0.25, "learning_rate": 7.563214052040867e-05, "loss": 1.0243, "step": 97870 }, { "epoch": 0.25, "learning_rate": 7.563088227892698e-05, "loss": 1.0223, "step": 97875 }, { "epoch": 0.25, "learning_rate": 7.562962403744527e-05, "loss": 1.0236, "step": 97880 }, { "epoch": 0.25, "learning_rate": 7.562836579596356e-05, "loss": 1.0248, "step": 97885 }, { "epoch": 0.25, "learning_rate": 7.562710755448185e-05, "loss": 1.0233, "step": 97890 }, { "epoch": 0.25, "learning_rate": 7.562584931300015e-05, "loss": 1.0209, "step": 97895 }, { "epoch": 0.25, "learning_rate": 7.562459107151845e-05, "loss": 1.0239, "step": 97900 }, { "epoch": 0.25, "learning_rate": 7.562333283003674e-05, "loss": 1.0242, "step": 97905 }, { "epoch": 0.25, "learning_rate": 7.562207458855503e-05, "loss": 1.0232, "step": 97910 }, { "epoch": 0.25, "learning_rate": 7.562081634707333e-05, "loss": 1.0251, "step": 97915 }, { "epoch": 0.25, "learning_rate": 7.561955810559163e-05, "loss": 1.0218, "step": 97920 }, { "epoch": 0.25, "learning_rate": 7.561829986410992e-05, "loss": 1.0241, "step": 97925 }, { "epoch": 0.25, "learning_rate": 7.561704162262821e-05, "loss": 1.0224, "step": 97930 }, { "epoch": 0.25, "learning_rate": 7.56157833811465e-05, "loss": 1.0251, "step": 97935 }, { "epoch": 0.25, "learning_rate": 7.561452513966481e-05, "loss": 1.0247, "step": 97940 }, { "epoch": 0.25, "learning_rate": 7.56132668981831e-05, "loss": 1.0238, "step": 97945 }, { "epoch": 0.25, "learning_rate": 7.56120086567014e-05, "loss": 1.0223, "step": 97950 }, { "epoch": 0.25, "learning_rate": 7.561075041521969e-05, "loss": 1.022, "step": 97955 }, { "epoch": 0.25, "learning_rate": 7.560949217373798e-05, "loss": 1.027, "step": 97960 }, { "epoch": 0.25, "learning_rate": 7.560823393225628e-05, "loss": 1.0223, "step": 97965 }, { "epoch": 0.25, "learning_rate": 7.560697569077457e-05, "loss": 1.0226, "step": 97970 }, { "epoch": 0.25, "learning_rate": 7.560571744929287e-05, "loss": 1.0226, "step": 97975 }, { "epoch": 0.25, "learning_rate": 7.560445920781116e-05, "loss": 1.0235, "step": 97980 }, { "epoch": 0.25, "learning_rate": 7.560320096632946e-05, "loss": 1.02, "step": 97985 }, { "epoch": 0.25, "learning_rate": 7.560194272484775e-05, "loss": 1.0233, "step": 97990 }, { "epoch": 0.25, "learning_rate": 7.560068448336605e-05, "loss": 1.0228, "step": 97995 }, { "epoch": 0.25, "learning_rate": 7.559942624188434e-05, "loss": 1.0238, "step": 98000 }, { "epoch": 0.25, "learning_rate": 7.559816800040264e-05, "loss": 1.0241, "step": 98005 }, { "epoch": 0.25, "learning_rate": 7.559690975892093e-05, "loss": 1.0222, "step": 98010 }, { "epoch": 0.25, "learning_rate": 7.559565151743923e-05, "loss": 1.0239, "step": 98015 }, { "epoch": 0.25, "learning_rate": 7.559439327595752e-05, "loss": 1.0234, "step": 98020 }, { "epoch": 0.25, "learning_rate": 7.559313503447581e-05, "loss": 1.0217, "step": 98025 }, { "epoch": 0.25, "learning_rate": 7.559187679299411e-05, "loss": 1.0235, "step": 98030 }, { "epoch": 0.25, "learning_rate": 7.55906185515124e-05, "loss": 1.0238, "step": 98035 }, { "epoch": 0.25, "learning_rate": 7.558936031003071e-05, "loss": 1.0255, "step": 98040 }, { "epoch": 0.25, "learning_rate": 7.5588102068549e-05, "loss": 1.025, "step": 98045 }, { "epoch": 0.25, "learning_rate": 7.55868438270673e-05, "loss": 1.0221, "step": 98050 }, { "epoch": 0.25, "learning_rate": 7.55855855855856e-05, "loss": 1.0232, "step": 98055 }, { "epoch": 0.25, "learning_rate": 7.558432734410389e-05, "loss": 1.0232, "step": 98060 }, { "epoch": 0.25, "learning_rate": 7.558306910262218e-05, "loss": 1.0202, "step": 98065 }, { "epoch": 0.25, "learning_rate": 7.558181086114047e-05, "loss": 1.0219, "step": 98070 }, { "epoch": 0.25, "learning_rate": 7.558055261965878e-05, "loss": 1.023, "step": 98075 }, { "epoch": 0.25, "learning_rate": 7.557929437817707e-05, "loss": 1.0244, "step": 98080 }, { "epoch": 0.25, "learning_rate": 7.557803613669536e-05, "loss": 1.0263, "step": 98085 }, { "epoch": 0.25, "learning_rate": 7.557677789521365e-05, "loss": 1.0211, "step": 98090 }, { "epoch": 0.25, "learning_rate": 7.557551965373196e-05, "loss": 1.0213, "step": 98095 }, { "epoch": 0.25, "learning_rate": 7.557426141225025e-05, "loss": 1.0236, "step": 98100 }, { "epoch": 0.25, "learning_rate": 7.557300317076854e-05, "loss": 1.0217, "step": 98105 }, { "epoch": 0.25, "learning_rate": 7.557174492928683e-05, "loss": 1.0207, "step": 98110 }, { "epoch": 0.25, "learning_rate": 7.557048668780512e-05, "loss": 1.0197, "step": 98115 }, { "epoch": 0.25, "learning_rate": 7.556922844632343e-05, "loss": 1.024, "step": 98120 }, { "epoch": 0.25, "learning_rate": 7.556797020484172e-05, "loss": 1.0233, "step": 98125 }, { "epoch": 0.25, "learning_rate": 7.556671196336001e-05, "loss": 1.0228, "step": 98130 }, { "epoch": 0.25, "learning_rate": 7.55654537218783e-05, "loss": 1.0267, "step": 98135 }, { "epoch": 0.25, "learning_rate": 7.556419548039661e-05, "loss": 1.0231, "step": 98140 }, { "epoch": 0.25, "learning_rate": 7.55629372389149e-05, "loss": 1.024, "step": 98145 }, { "epoch": 0.25, "learning_rate": 7.556167899743319e-05, "loss": 1.022, "step": 98150 }, { "epoch": 0.25, "learning_rate": 7.556042075595148e-05, "loss": 1.0235, "step": 98155 }, { "epoch": 0.25, "learning_rate": 7.555916251446979e-05, "loss": 1.0442, "step": 98160 }, { "epoch": 0.25, "learning_rate": 7.555790427298808e-05, "loss": 1.0238, "step": 98165 }, { "epoch": 0.25, "learning_rate": 7.555664603150637e-05, "loss": 1.0226, "step": 98170 }, { "epoch": 0.25, "learning_rate": 7.555538779002466e-05, "loss": 1.023, "step": 98175 }, { "epoch": 0.25, "learning_rate": 7.555412954854296e-05, "loss": 1.02, "step": 98180 }, { "epoch": 0.25, "learning_rate": 7.555287130706126e-05, "loss": 1.0242, "step": 98185 }, { "epoch": 0.25, "learning_rate": 7.555161306557955e-05, "loss": 1.0219, "step": 98190 }, { "epoch": 0.25, "learning_rate": 7.555035482409784e-05, "loss": 1.0221, "step": 98195 }, { "epoch": 0.25, "learning_rate": 7.554909658261614e-05, "loss": 1.0232, "step": 98200 }, { "epoch": 0.25, "learning_rate": 7.554783834113444e-05, "loss": 1.0216, "step": 98205 }, { "epoch": 0.25, "learning_rate": 7.554658009965273e-05, "loss": 1.0247, "step": 98210 }, { "epoch": 0.25, "learning_rate": 7.554532185817102e-05, "loss": 1.0226, "step": 98215 }, { "epoch": 0.25, "learning_rate": 7.554406361668932e-05, "loss": 1.0239, "step": 98220 }, { "epoch": 0.25, "learning_rate": 7.554280537520762e-05, "loss": 1.0231, "step": 98225 }, { "epoch": 0.25, "learning_rate": 7.554154713372591e-05, "loss": 1.0222, "step": 98230 }, { "epoch": 0.25, "learning_rate": 7.55402888922442e-05, "loss": 1.0243, "step": 98235 }, { "epoch": 0.25, "learning_rate": 7.55390306507625e-05, "loss": 1.023, "step": 98240 }, { "epoch": 0.25, "learning_rate": 7.553777240928079e-05, "loss": 1.0246, "step": 98245 }, { "epoch": 0.25, "learning_rate": 7.553651416779909e-05, "loss": 1.0219, "step": 98250 }, { "epoch": 0.25, "learning_rate": 7.553525592631738e-05, "loss": 1.039, "step": 98255 }, { "epoch": 0.25, "learning_rate": 7.553399768483568e-05, "loss": 1.0242, "step": 98260 }, { "epoch": 0.25, "learning_rate": 7.553273944335397e-05, "loss": 1.0242, "step": 98265 }, { "epoch": 0.25, "learning_rate": 7.553148120187227e-05, "loss": 1.0216, "step": 98270 }, { "epoch": 0.25, "learning_rate": 7.553022296039056e-05, "loss": 1.0242, "step": 98275 }, { "epoch": 0.25, "learning_rate": 7.552896471890886e-05, "loss": 1.0238, "step": 98280 }, { "epoch": 0.25, "learning_rate": 7.552770647742715e-05, "loss": 1.0218, "step": 98285 }, { "epoch": 0.25, "learning_rate": 7.552644823594545e-05, "loss": 1.0232, "step": 98290 }, { "epoch": 0.25, "learning_rate": 7.552518999446374e-05, "loss": 1.0221, "step": 98295 }, { "epoch": 0.25, "learning_rate": 7.552393175298204e-05, "loss": 1.0235, "step": 98300 }, { "epoch": 0.25, "learning_rate": 7.552267351150033e-05, "loss": 1.0234, "step": 98305 }, { "epoch": 0.25, "learning_rate": 7.552141527001862e-05, "loss": 1.0221, "step": 98310 }, { "epoch": 0.25, "learning_rate": 7.552015702853692e-05, "loss": 1.0228, "step": 98315 }, { "epoch": 0.25, "learning_rate": 7.551889878705522e-05, "loss": 1.024, "step": 98320 }, { "epoch": 0.25, "learning_rate": 7.551764054557351e-05, "loss": 1.0454, "step": 98325 }, { "epoch": 0.25, "learning_rate": 7.55163823040918e-05, "loss": 1.0223, "step": 98330 }, { "epoch": 0.25, "learning_rate": 7.55151240626101e-05, "loss": 1.0235, "step": 98335 }, { "epoch": 0.25, "learning_rate": 7.55138658211284e-05, "loss": 1.0248, "step": 98340 }, { "epoch": 0.25, "learning_rate": 7.551260757964669e-05, "loss": 1.0224, "step": 98345 }, { "epoch": 0.25, "learning_rate": 7.551134933816498e-05, "loss": 1.0231, "step": 98350 }, { "epoch": 0.25, "learning_rate": 7.551009109668328e-05, "loss": 1.0229, "step": 98355 }, { "epoch": 0.25, "learning_rate": 7.550883285520158e-05, "loss": 1.0237, "step": 98360 }, { "epoch": 0.25, "learning_rate": 7.550757461371987e-05, "loss": 1.0217, "step": 98365 }, { "epoch": 0.25, "learning_rate": 7.550631637223816e-05, "loss": 1.0234, "step": 98370 }, { "epoch": 0.25, "learning_rate": 7.550505813075645e-05, "loss": 1.0234, "step": 98375 }, { "epoch": 0.25, "learning_rate": 7.550379988927476e-05, "loss": 1.0246, "step": 98380 }, { "epoch": 0.25, "learning_rate": 7.550254164779305e-05, "loss": 1.0222, "step": 98385 }, { "epoch": 0.25, "learning_rate": 7.550128340631134e-05, "loss": 1.0213, "step": 98390 }, { "epoch": 0.25, "learning_rate": 7.550002516482963e-05, "loss": 1.023, "step": 98395 }, { "epoch": 0.25, "learning_rate": 7.549876692334794e-05, "loss": 1.0406, "step": 98400 }, { "epoch": 0.25, "learning_rate": 7.549750868186623e-05, "loss": 1.0237, "step": 98405 }, { "epoch": 0.25, "learning_rate": 7.549625044038452e-05, "loss": 1.0223, "step": 98410 }, { "epoch": 0.25, "learning_rate": 7.549499219890281e-05, "loss": 1.0231, "step": 98415 }, { "epoch": 0.25, "learning_rate": 7.549373395742112e-05, "loss": 1.0223, "step": 98420 }, { "epoch": 0.25, "learning_rate": 7.549247571593941e-05, "loss": 1.0226, "step": 98425 }, { "epoch": 0.25, "learning_rate": 7.54912174744577e-05, "loss": 1.0232, "step": 98430 }, { "epoch": 0.25, "learning_rate": 7.548995923297599e-05, "loss": 1.0238, "step": 98435 }, { "epoch": 0.25, "learning_rate": 7.548870099149428e-05, "loss": 1.023, "step": 98440 }, { "epoch": 0.25, "learning_rate": 7.548744275001259e-05, "loss": 1.0216, "step": 98445 }, { "epoch": 0.25, "learning_rate": 7.548618450853088e-05, "loss": 1.0226, "step": 98450 }, { "epoch": 0.25, "learning_rate": 7.548492626704917e-05, "loss": 1.0224, "step": 98455 }, { "epoch": 0.25, "learning_rate": 7.548366802556746e-05, "loss": 1.0227, "step": 98460 }, { "epoch": 0.25, "learning_rate": 7.548240978408577e-05, "loss": 1.0242, "step": 98465 }, { "epoch": 0.25, "learning_rate": 7.548115154260406e-05, "loss": 1.0251, "step": 98470 }, { "epoch": 0.25, "learning_rate": 7.547989330112235e-05, "loss": 1.0211, "step": 98475 }, { "epoch": 0.25, "learning_rate": 7.547863505964064e-05, "loss": 1.0222, "step": 98480 }, { "epoch": 0.25, "learning_rate": 7.547737681815893e-05, "loss": 1.0199, "step": 98485 }, { "epoch": 0.25, "learning_rate": 7.547611857667724e-05, "loss": 1.0248, "step": 98490 }, { "epoch": 0.25, "learning_rate": 7.547486033519553e-05, "loss": 1.0231, "step": 98495 }, { "epoch": 0.25, "learning_rate": 7.547360209371382e-05, "loss": 1.024, "step": 98500 }, { "epoch": 0.25, "learning_rate": 7.547234385223211e-05, "loss": 1.021, "step": 98505 }, { "epoch": 0.25, "learning_rate": 7.547108561075042e-05, "loss": 1.026, "step": 98510 }, { "epoch": 0.25, "learning_rate": 7.546982736926871e-05, "loss": 1.0203, "step": 98515 }, { "epoch": 0.25, "learning_rate": 7.5468569127787e-05, "loss": 1.0227, "step": 98520 }, { "epoch": 0.25, "learning_rate": 7.54673108863053e-05, "loss": 1.0245, "step": 98525 }, { "epoch": 0.25, "learning_rate": 7.54660526448236e-05, "loss": 1.0231, "step": 98530 }, { "epoch": 0.25, "learning_rate": 7.54647944033419e-05, "loss": 1.0233, "step": 98535 }, { "epoch": 0.25, "learning_rate": 7.54635361618602e-05, "loss": 1.0204, "step": 98540 }, { "epoch": 0.25, "learning_rate": 7.546227792037849e-05, "loss": 1.0228, "step": 98545 }, { "epoch": 0.25, "learning_rate": 7.546101967889678e-05, "loss": 1.0221, "step": 98550 }, { "epoch": 0.25, "learning_rate": 7.545976143741508e-05, "loss": 1.0224, "step": 98555 }, { "epoch": 0.25, "learning_rate": 7.545850319593338e-05, "loss": 1.0234, "step": 98560 }, { "epoch": 0.25, "learning_rate": 7.545724495445167e-05, "loss": 1.0219, "step": 98565 }, { "epoch": 0.25, "learning_rate": 7.545598671296996e-05, "loss": 1.0243, "step": 98570 }, { "epoch": 0.25, "learning_rate": 7.545472847148825e-05, "loss": 1.0204, "step": 98575 }, { "epoch": 0.25, "learning_rate": 7.545347023000656e-05, "loss": 1.025, "step": 98580 }, { "epoch": 0.25, "learning_rate": 7.545221198852485e-05, "loss": 1.0223, "step": 98585 }, { "epoch": 0.25, "learning_rate": 7.545095374704314e-05, "loss": 1.0234, "step": 98590 }, { "epoch": 0.25, "learning_rate": 7.544969550556143e-05, "loss": 1.0242, "step": 98595 }, { "epoch": 0.25, "learning_rate": 7.544843726407974e-05, "loss": 1.0237, "step": 98600 }, { "epoch": 0.25, "learning_rate": 7.544717902259803e-05, "loss": 1.0239, "step": 98605 }, { "epoch": 0.25, "learning_rate": 7.544592078111632e-05, "loss": 1.0234, "step": 98610 }, { "epoch": 0.25, "learning_rate": 7.544466253963461e-05, "loss": 1.0226, "step": 98615 }, { "epoch": 0.25, "learning_rate": 7.544340429815292e-05, "loss": 1.0212, "step": 98620 }, { "epoch": 0.25, "learning_rate": 7.54421460566712e-05, "loss": 1.0246, "step": 98625 }, { "epoch": 0.25, "learning_rate": 7.54408878151895e-05, "loss": 1.0238, "step": 98630 }, { "epoch": 0.25, "learning_rate": 7.543962957370779e-05, "loss": 1.0217, "step": 98635 }, { "epoch": 0.25, "learning_rate": 7.543837133222608e-05, "loss": 1.0225, "step": 98640 }, { "epoch": 0.25, "learning_rate": 7.543711309074439e-05, "loss": 1.0244, "step": 98645 }, { "epoch": 0.25, "learning_rate": 7.543585484926268e-05, "loss": 1.023, "step": 98650 }, { "epoch": 0.25, "learning_rate": 7.543459660778097e-05, "loss": 1.0212, "step": 98655 }, { "epoch": 0.25, "learning_rate": 7.543333836629926e-05, "loss": 1.0248, "step": 98660 }, { "epoch": 0.25, "learning_rate": 7.543208012481757e-05, "loss": 1.0235, "step": 98665 }, { "epoch": 0.25, "learning_rate": 7.543082188333586e-05, "loss": 1.0245, "step": 98670 }, { "epoch": 0.25, "learning_rate": 7.542956364185415e-05, "loss": 1.0214, "step": 98675 }, { "epoch": 0.25, "learning_rate": 7.542830540037244e-05, "loss": 1.0231, "step": 98680 }, { "epoch": 0.25, "learning_rate": 7.542704715889075e-05, "loss": 1.0233, "step": 98685 }, { "epoch": 0.25, "learning_rate": 7.542578891740904e-05, "loss": 1.0215, "step": 98690 }, { "epoch": 0.25, "learning_rate": 7.542453067592733e-05, "loss": 1.018, "step": 98695 }, { "epoch": 0.25, "learning_rate": 7.542327243444562e-05, "loss": 1.0228, "step": 98700 }, { "epoch": 0.25, "learning_rate": 7.542201419296391e-05, "loss": 1.0232, "step": 98705 }, { "epoch": 0.25, "learning_rate": 7.542075595148222e-05, "loss": 1.0216, "step": 98710 }, { "epoch": 0.25, "learning_rate": 7.541949771000051e-05, "loss": 1.0428, "step": 98715 }, { "epoch": 0.25, "learning_rate": 7.541849111681513e-05, "loss": 1.0199, "step": 98720 }, { "epoch": 0.25, "learning_rate": 7.541723287533344e-05, "loss": 1.025, "step": 98725 }, { "epoch": 0.25, "learning_rate": 7.541597463385173e-05, "loss": 1.0229, "step": 98730 }, { "epoch": 0.25, "learning_rate": 7.541471639237002e-05, "loss": 1.0226, "step": 98735 }, { "epoch": 0.25, "learning_rate": 7.541345815088833e-05, "loss": 1.0233, "step": 98740 }, { "epoch": 0.25, "learning_rate": 7.541219990940662e-05, "loss": 1.0234, "step": 98745 }, { "epoch": 0.25, "learning_rate": 7.541094166792492e-05, "loss": 1.0211, "step": 98750 }, { "epoch": 0.25, "learning_rate": 7.540968342644321e-05, "loss": 1.0207, "step": 98755 }, { "epoch": 0.25, "learning_rate": 7.54084251849615e-05, "loss": 1.0219, "step": 98760 }, { "epoch": 0.25, "learning_rate": 7.54071669434798e-05, "loss": 1.022, "step": 98765 }, { "epoch": 0.25, "learning_rate": 7.54059087019981e-05, "loss": 1.0245, "step": 98770 }, { "epoch": 0.25, "learning_rate": 7.54046504605164e-05, "loss": 1.0214, "step": 98775 }, { "epoch": 0.25, "learning_rate": 7.540339221903469e-05, "loss": 1.0239, "step": 98780 }, { "epoch": 0.25, "learning_rate": 7.540213397755298e-05, "loss": 1.0239, "step": 98785 }, { "epoch": 0.25, "learning_rate": 7.540087573607127e-05, "loss": 1.0244, "step": 98790 }, { "epoch": 0.25, "learning_rate": 7.539961749458957e-05, "loss": 1.0238, "step": 98795 }, { "epoch": 0.25, "learning_rate": 7.539835925310787e-05, "loss": 1.0233, "step": 98800 }, { "epoch": 0.25, "learning_rate": 7.539710101162616e-05, "loss": 1.0219, "step": 98805 }, { "epoch": 0.25, "learning_rate": 7.539584277014445e-05, "loss": 1.0252, "step": 98810 }, { "epoch": 0.25, "learning_rate": 7.539458452866275e-05, "loss": 1.0225, "step": 98815 }, { "epoch": 0.25, "learning_rate": 7.539332628718105e-05, "loss": 1.0226, "step": 98820 }, { "epoch": 0.25, "learning_rate": 7.539206804569934e-05, "loss": 1.0217, "step": 98825 }, { "epoch": 0.25, "learning_rate": 7.539080980421763e-05, "loss": 1.046, "step": 98830 }, { "epoch": 0.25, "learning_rate": 7.538955156273593e-05, "loss": 1.0242, "step": 98835 }, { "epoch": 0.25, "learning_rate": 7.538829332125423e-05, "loss": 1.0232, "step": 98840 }, { "epoch": 0.25, "learning_rate": 7.538703507977252e-05, "loss": 1.0241, "step": 98845 }, { "epoch": 0.25, "learning_rate": 7.538577683829081e-05, "loss": 1.0214, "step": 98850 }, { "epoch": 0.25, "learning_rate": 7.53845185968091e-05, "loss": 1.0231, "step": 98855 }, { "epoch": 0.25, "learning_rate": 7.53832603553274e-05, "loss": 1.025, "step": 98860 }, { "epoch": 0.25, "learning_rate": 7.53820021138457e-05, "loss": 1.0224, "step": 98865 }, { "epoch": 0.25, "learning_rate": 7.538074387236399e-05, "loss": 1.0205, "step": 98870 }, { "epoch": 0.25, "learning_rate": 7.537948563088228e-05, "loss": 1.0203, "step": 98875 }, { "epoch": 0.25, "learning_rate": 7.537822738940059e-05, "loss": 1.0241, "step": 98880 }, { "epoch": 0.25, "learning_rate": 7.537696914791888e-05, "loss": 1.025, "step": 98885 }, { "epoch": 0.25, "learning_rate": 7.537571090643717e-05, "loss": 1.024, "step": 98890 }, { "epoch": 0.25, "learning_rate": 7.537445266495546e-05, "loss": 1.0225, "step": 98895 }, { "epoch": 0.25, "learning_rate": 7.537319442347377e-05, "loss": 1.0232, "step": 98900 }, { "epoch": 0.25, "learning_rate": 7.537193618199206e-05, "loss": 1.0233, "step": 98905 }, { "epoch": 0.25, "learning_rate": 7.537067794051035e-05, "loss": 1.0217, "step": 98910 }, { "epoch": 0.25, "learning_rate": 7.536941969902864e-05, "loss": 1.0248, "step": 98915 }, { "epoch": 0.25, "learning_rate": 7.536816145754693e-05, "loss": 1.0239, "step": 98920 }, { "epoch": 0.25, "learning_rate": 7.536690321606524e-05, "loss": 1.0232, "step": 98925 }, { "epoch": 0.25, "learning_rate": 7.536564497458353e-05, "loss": 1.0265, "step": 98930 }, { "epoch": 0.25, "learning_rate": 7.536438673310182e-05, "loss": 1.0241, "step": 98935 }, { "epoch": 0.25, "learning_rate": 7.536312849162011e-05, "loss": 1.024, "step": 98940 }, { "epoch": 0.25, "learning_rate": 7.536187025013842e-05, "loss": 1.022, "step": 98945 }, { "epoch": 0.25, "learning_rate": 7.536061200865671e-05, "loss": 1.0222, "step": 98950 }, { "epoch": 0.25, "learning_rate": 7.5359353767175e-05, "loss": 1.0238, "step": 98955 }, { "epoch": 0.25, "learning_rate": 7.535809552569329e-05, "loss": 1.0238, "step": 98960 }, { "epoch": 0.25, "learning_rate": 7.53568372842116e-05, "loss": 1.0217, "step": 98965 }, { "epoch": 0.25, "learning_rate": 7.535557904272989e-05, "loss": 1.0215, "step": 98970 }, { "epoch": 0.25, "learning_rate": 7.535432080124818e-05, "loss": 1.0227, "step": 98975 }, { "epoch": 0.25, "learning_rate": 7.535306255976647e-05, "loss": 1.0254, "step": 98980 }, { "epoch": 0.25, "learning_rate": 7.535180431828476e-05, "loss": 1.0399, "step": 98985 }, { "epoch": 0.25, "learning_rate": 7.535054607680307e-05, "loss": 1.0238, "step": 98990 }, { "epoch": 0.25, "learning_rate": 7.534928783532136e-05, "loss": 1.0246, "step": 98995 }, { "epoch": 0.25, "learning_rate": 7.534802959383965e-05, "loss": 1.0238, "step": 99000 }, { "epoch": 0.25, "learning_rate": 7.534677135235794e-05, "loss": 1.0235, "step": 99005 }, { "epoch": 0.25, "learning_rate": 7.534551311087625e-05, "loss": 1.0236, "step": 99010 }, { "epoch": 0.25, "learning_rate": 7.534425486939454e-05, "loss": 1.0234, "step": 99015 }, { "epoch": 0.25, "learning_rate": 7.534299662791283e-05, "loss": 1.02, "step": 99020 }, { "epoch": 0.25, "learning_rate": 7.534173838643112e-05, "loss": 1.0237, "step": 99025 }, { "epoch": 0.25, "learning_rate": 7.534048014494943e-05, "loss": 1.0219, "step": 99030 }, { "epoch": 0.25, "learning_rate": 7.533922190346772e-05, "loss": 1.0244, "step": 99035 }, { "epoch": 0.25, "learning_rate": 7.533796366198601e-05, "loss": 1.0258, "step": 99040 }, { "epoch": 0.25, "learning_rate": 7.53367054205043e-05, "loss": 1.0246, "step": 99045 }, { "epoch": 0.25, "learning_rate": 7.53354471790226e-05, "loss": 1.0214, "step": 99050 }, { "epoch": 0.25, "learning_rate": 7.53341889375409e-05, "loss": 1.0254, "step": 99055 }, { "epoch": 0.25, "learning_rate": 7.533293069605919e-05, "loss": 1.0209, "step": 99060 }, { "epoch": 0.25, "learning_rate": 7.533167245457748e-05, "loss": 1.0226, "step": 99065 }, { "epoch": 0.25, "learning_rate": 7.533041421309577e-05, "loss": 1.0227, "step": 99070 }, { "epoch": 0.25, "learning_rate": 7.532915597161408e-05, "loss": 1.0233, "step": 99075 }, { "epoch": 0.25, "learning_rate": 7.532789773013237e-05, "loss": 1.0232, "step": 99080 }, { "epoch": 0.25, "learning_rate": 7.532663948865066e-05, "loss": 1.0242, "step": 99085 }, { "epoch": 0.25, "learning_rate": 7.532538124716895e-05, "loss": 1.0227, "step": 99090 }, { "epoch": 0.25, "learning_rate": 7.532412300568726e-05, "loss": 1.0217, "step": 99095 }, { "epoch": 0.25, "learning_rate": 7.532286476420555e-05, "loss": 1.0235, "step": 99100 }, { "epoch": 0.25, "learning_rate": 7.532160652272384e-05, "loss": 1.0217, "step": 99105 }, { "epoch": 0.25, "learning_rate": 7.532034828124213e-05, "loss": 1.023, "step": 99110 }, { "epoch": 0.25, "learning_rate": 7.531909003976043e-05, "loss": 1.0225, "step": 99115 }, { "epoch": 0.25, "learning_rate": 7.531783179827873e-05, "loss": 1.0247, "step": 99120 }, { "epoch": 0.25, "learning_rate": 7.531657355679702e-05, "loss": 1.0238, "step": 99125 }, { "epoch": 0.25, "learning_rate": 7.531531531531531e-05, "loss": 1.0216, "step": 99130 }, { "epoch": 0.25, "learning_rate": 7.53140570738336e-05, "loss": 1.0228, "step": 99135 }, { "epoch": 0.25, "learning_rate": 7.531279883235191e-05, "loss": 1.0251, "step": 99140 }, { "epoch": 0.25, "learning_rate": 7.53115405908702e-05, "loss": 1.0248, "step": 99145 }, { "epoch": 0.25, "learning_rate": 7.53102823493885e-05, "loss": 1.0254, "step": 99150 }, { "epoch": 0.25, "learning_rate": 7.530902410790679e-05, "loss": 1.0232, "step": 99155 }, { "epoch": 0.25, "learning_rate": 7.530776586642509e-05, "loss": 1.0218, "step": 99160 }, { "epoch": 0.25, "learning_rate": 7.530650762494338e-05, "loss": 1.0223, "step": 99165 }, { "epoch": 0.25, "learning_rate": 7.530524938346167e-05, "loss": 1.0237, "step": 99170 }, { "epoch": 0.25, "learning_rate": 7.530399114197997e-05, "loss": 1.0246, "step": 99175 }, { "epoch": 0.25, "learning_rate": 7.530273290049826e-05, "loss": 1.0227, "step": 99180 }, { "epoch": 0.25, "learning_rate": 7.530147465901656e-05, "loss": 1.0228, "step": 99185 }, { "epoch": 0.25, "learning_rate": 7.530021641753485e-05, "loss": 1.0224, "step": 99190 }, { "epoch": 0.25, "learning_rate": 7.529895817605315e-05, "loss": 1.0256, "step": 99195 }, { "epoch": 0.25, "learning_rate": 7.529769993457144e-05, "loss": 1.0244, "step": 99200 }, { "epoch": 0.25, "learning_rate": 7.529644169308974e-05, "loss": 1.0222, "step": 99205 }, { "epoch": 0.25, "learning_rate": 7.529518345160803e-05, "loss": 1.0246, "step": 99210 }, { "epoch": 0.25, "learning_rate": 7.529392521012633e-05, "loss": 1.0248, "step": 99215 }, { "epoch": 0.25, "learning_rate": 7.529266696864462e-05, "loss": 1.0217, "step": 99220 }, { "epoch": 0.25, "learning_rate": 7.529140872716291e-05, "loss": 1.026, "step": 99225 }, { "epoch": 0.25, "learning_rate": 7.529015048568121e-05, "loss": 1.0223, "step": 99230 }, { "epoch": 0.25, "learning_rate": 7.528889224419952e-05, "loss": 1.0223, "step": 99235 }, { "epoch": 0.25, "learning_rate": 7.528763400271781e-05, "loss": 1.0232, "step": 99240 }, { "epoch": 0.25, "learning_rate": 7.52863757612361e-05, "loss": 1.0219, "step": 99245 }, { "epoch": 0.25, "learning_rate": 7.52851175197544e-05, "loss": 1.0246, "step": 99250 }, { "epoch": 0.25, "learning_rate": 7.52838592782727e-05, "loss": 1.0245, "step": 99255 }, { "epoch": 0.25, "learning_rate": 7.528260103679099e-05, "loss": 1.0224, "step": 99260 }, { "epoch": 0.25, "learning_rate": 7.528134279530928e-05, "loss": 1.0226, "step": 99265 }, { "epoch": 0.25, "learning_rate": 7.528008455382757e-05, "loss": 1.0232, "step": 99270 }, { "epoch": 0.25, "learning_rate": 7.527882631234588e-05, "loss": 1.0231, "step": 99275 }, { "epoch": 0.25, "learning_rate": 7.527756807086417e-05, "loss": 1.0231, "step": 99280 }, { "epoch": 0.25, "learning_rate": 7.527630982938246e-05, "loss": 1.0204, "step": 99285 }, { "epoch": 0.25, "learning_rate": 7.527505158790075e-05, "loss": 1.0234, "step": 99290 }, { "epoch": 0.25, "learning_rate": 7.527379334641906e-05, "loss": 1.0229, "step": 99295 }, { "epoch": 0.25, "learning_rate": 7.527253510493735e-05, "loss": 1.0247, "step": 99300 }, { "epoch": 0.25, "learning_rate": 7.527127686345564e-05, "loss": 1.0231, "step": 99305 }, { "epoch": 0.25, "learning_rate": 7.527001862197393e-05, "loss": 1.0242, "step": 99310 }, { "epoch": 0.25, "learning_rate": 7.526876038049223e-05, "loss": 1.0423, "step": 99315 }, { "epoch": 0.25, "learning_rate": 7.526750213901053e-05, "loss": 1.0232, "step": 99320 }, { "epoch": 0.25, "learning_rate": 7.526624389752882e-05, "loss": 1.0221, "step": 99325 }, { "epoch": 0.25, "learning_rate": 7.526498565604711e-05, "loss": 1.0236, "step": 99330 }, { "epoch": 0.25, "learning_rate": 7.52637274145654e-05, "loss": 1.0229, "step": 99335 }, { "epoch": 0.25, "learning_rate": 7.526246917308371e-05, "loss": 1.0257, "step": 99340 }, { "epoch": 0.25, "learning_rate": 7.5261210931602e-05, "loss": 1.0233, "step": 99345 }, { "epoch": 0.25, "learning_rate": 7.52599526901203e-05, "loss": 1.0224, "step": 99350 }, { "epoch": 0.25, "learning_rate": 7.525869444863859e-05, "loss": 1.0258, "step": 99355 }, { "epoch": 0.25, "learning_rate": 7.525743620715689e-05, "loss": 1.0208, "step": 99360 }, { "epoch": 0.25, "learning_rate": 7.525617796567518e-05, "loss": 1.0221, "step": 99365 }, { "epoch": 0.25, "learning_rate": 7.525491972419347e-05, "loss": 1.0235, "step": 99370 }, { "epoch": 0.25, "learning_rate": 7.525366148271176e-05, "loss": 1.0229, "step": 99375 }, { "epoch": 0.25, "learning_rate": 7.525240324123006e-05, "loss": 1.0239, "step": 99380 }, { "epoch": 0.25, "learning_rate": 7.525114499974836e-05, "loss": 1.0225, "step": 99385 }, { "epoch": 0.25, "learning_rate": 7.524988675826665e-05, "loss": 1.0234, "step": 99390 }, { "epoch": 0.25, "learning_rate": 7.524862851678494e-05, "loss": 1.0246, "step": 99395 }, { "epoch": 0.25, "learning_rate": 7.524737027530324e-05, "loss": 1.0238, "step": 99400 }, { "epoch": 0.25, "learning_rate": 7.524611203382154e-05, "loss": 1.0225, "step": 99405 }, { "epoch": 0.25, "learning_rate": 7.524485379233983e-05, "loss": 1.0236, "step": 99410 }, { "epoch": 0.25, "learning_rate": 7.524359555085812e-05, "loss": 1.0225, "step": 99415 }, { "epoch": 0.25, "learning_rate": 7.524233730937642e-05, "loss": 1.0245, "step": 99420 }, { "epoch": 0.25, "learning_rate": 7.524107906789472e-05, "loss": 1.0227, "step": 99425 }, { "epoch": 0.25, "learning_rate": 7.523982082641301e-05, "loss": 1.0229, "step": 99430 }, { "epoch": 0.25, "learning_rate": 7.52385625849313e-05, "loss": 1.0236, "step": 99435 }, { "epoch": 0.25, "learning_rate": 7.52373043434496e-05, "loss": 1.0216, "step": 99440 }, { "epoch": 0.25, "learning_rate": 7.523604610196789e-05, "loss": 1.0209, "step": 99445 }, { "epoch": 0.25, "learning_rate": 7.523478786048619e-05, "loss": 1.0413, "step": 99450 }, { "epoch": 0.25, "learning_rate": 7.523352961900448e-05, "loss": 1.021, "step": 99455 }, { "epoch": 0.25, "learning_rate": 7.523227137752278e-05, "loss": 1.023, "step": 99460 }, { "epoch": 0.25, "learning_rate": 7.523101313604107e-05, "loss": 1.0233, "step": 99465 }, { "epoch": 0.25, "learning_rate": 7.522975489455937e-05, "loss": 1.0236, "step": 99470 }, { "epoch": 0.25, "learning_rate": 7.522849665307766e-05, "loss": 1.0221, "step": 99475 }, { "epoch": 0.25, "learning_rate": 7.522723841159596e-05, "loss": 1.0224, "step": 99480 }, { "epoch": 0.25, "learning_rate": 7.522598017011425e-05, "loss": 1.0231, "step": 99485 }, { "epoch": 0.25, "learning_rate": 7.522472192863255e-05, "loss": 1.0254, "step": 99490 }, { "epoch": 0.25, "learning_rate": 7.522346368715084e-05, "loss": 1.0235, "step": 99495 }, { "epoch": 0.25, "learning_rate": 7.522220544566914e-05, "loss": 1.0232, "step": 99500 }, { "epoch": 0.25, "learning_rate": 7.522094720418743e-05, "loss": 1.0214, "step": 99505 }, { "epoch": 0.25, "learning_rate": 7.521968896270572e-05, "loss": 1.0236, "step": 99510 }, { "epoch": 0.25, "learning_rate": 7.521843072122402e-05, "loss": 1.0229, "step": 99515 }, { "epoch": 0.25, "learning_rate": 7.521717247974232e-05, "loss": 1.0472, "step": 99520 }, { "epoch": 0.25, "learning_rate": 7.521591423826061e-05, "loss": 1.0229, "step": 99525 }, { "epoch": 0.25, "learning_rate": 7.52146559967789e-05, "loss": 1.0259, "step": 99530 }, { "epoch": 0.25, "learning_rate": 7.52133977552972e-05, "loss": 1.0225, "step": 99535 }, { "epoch": 0.25, "learning_rate": 7.52121395138155e-05, "loss": 1.0216, "step": 99540 }, { "epoch": 0.25, "learning_rate": 7.521088127233379e-05, "loss": 1.0249, "step": 99545 }, { "epoch": 0.25, "learning_rate": 7.520962303085208e-05, "loss": 1.0243, "step": 99550 }, { "epoch": 0.25, "learning_rate": 7.520836478937038e-05, "loss": 1.0217, "step": 99555 }, { "epoch": 0.25, "learning_rate": 7.520710654788868e-05, "loss": 1.0241, "step": 99560 }, { "epoch": 0.25, "learning_rate": 7.520584830640697e-05, "loss": 1.0248, "step": 99565 }, { "epoch": 0.25, "learning_rate": 7.520459006492526e-05, "loss": 1.0225, "step": 99570 }, { "epoch": 0.25, "learning_rate": 7.520333182344355e-05, "loss": 1.0239, "step": 99575 }, { "epoch": 0.25, "learning_rate": 7.520207358196186e-05, "loss": 1.0208, "step": 99580 }, { "epoch": 0.25, "learning_rate": 7.520081534048015e-05, "loss": 1.0232, "step": 99585 }, { "epoch": 0.25, "learning_rate": 7.519955709899844e-05, "loss": 1.0227, "step": 99590 }, { "epoch": 0.25, "learning_rate": 7.519829885751673e-05, "loss": 1.0237, "step": 99595 }, { "epoch": 0.25, "learning_rate": 7.519704061603504e-05, "loss": 1.0226, "step": 99600 }, { "epoch": 0.25, "learning_rate": 7.519578237455333e-05, "loss": 1.0226, "step": 99605 }, { "epoch": 0.25, "learning_rate": 7.519452413307162e-05, "loss": 1.0264, "step": 99610 }, { "epoch": 0.25, "learning_rate": 7.519326589158991e-05, "loss": 1.0231, "step": 99615 }, { "epoch": 0.25, "learning_rate": 7.519200765010822e-05, "loss": 1.0256, "step": 99620 }, { "epoch": 0.25, "learning_rate": 7.519074940862651e-05, "loss": 1.027, "step": 99625 }, { "epoch": 0.25, "learning_rate": 7.51894911671448e-05, "loss": 1.0215, "step": 99630 }, { "epoch": 0.25, "learning_rate": 7.518823292566309e-05, "loss": 1.0214, "step": 99635 }, { "epoch": 0.25, "learning_rate": 7.518697468418138e-05, "loss": 1.0252, "step": 99640 }, { "epoch": 0.25, "learning_rate": 7.518571644269969e-05, "loss": 1.0235, "step": 99645 }, { "epoch": 0.25, "learning_rate": 7.518445820121798e-05, "loss": 1.0235, "step": 99650 }, { "epoch": 0.25, "learning_rate": 7.518319995973627e-05, "loss": 1.0224, "step": 99655 }, { "epoch": 0.25, "learning_rate": 7.518194171825456e-05, "loss": 1.0214, "step": 99660 }, { "epoch": 0.25, "learning_rate": 7.518068347677287e-05, "loss": 1.0254, "step": 99665 }, { "epoch": 0.25, "learning_rate": 7.517942523529116e-05, "loss": 1.0227, "step": 99670 }, { "epoch": 0.25, "learning_rate": 7.517816699380945e-05, "loss": 1.0235, "step": 99675 }, { "epoch": 0.25, "learning_rate": 7.517690875232774e-05, "loss": 1.0235, "step": 99680 }, { "epoch": 0.25, "learning_rate": 7.517565051084605e-05, "loss": 1.0239, "step": 99685 }, { "epoch": 0.25, "learning_rate": 7.517439226936434e-05, "loss": 1.0224, "step": 99690 }, { "epoch": 0.25, "learning_rate": 7.517313402788263e-05, "loss": 1.0245, "step": 99695 }, { "epoch": 0.25, "learning_rate": 7.517187578640092e-05, "loss": 1.0218, "step": 99700 }, { "epoch": 0.25, "learning_rate": 7.517061754491921e-05, "loss": 1.0238, "step": 99705 }, { "epoch": 0.25, "learning_rate": 7.516935930343752e-05, "loss": 1.0236, "step": 99710 }, { "epoch": 0.25, "learning_rate": 7.516810106195581e-05, "loss": 1.0236, "step": 99715 }, { "epoch": 0.25, "learning_rate": 7.51668428204741e-05, "loss": 1.023, "step": 99720 }, { "epoch": 0.25, "learning_rate": 7.51655845789924e-05, "loss": 1.0236, "step": 99725 }, { "epoch": 0.25, "learning_rate": 7.51643263375107e-05, "loss": 1.0228, "step": 99730 }, { "epoch": 0.25, "learning_rate": 7.5163068096029e-05, "loss": 1.0247, "step": 99735 }, { "epoch": 0.25, "learning_rate": 7.51618098545473e-05, "loss": 1.0223, "step": 99740 }, { "epoch": 0.25, "learning_rate": 7.516055161306559e-05, "loss": 1.0241, "step": 99745 }, { "epoch": 0.25, "learning_rate": 7.515929337158388e-05, "loss": 1.0224, "step": 99750 }, { "epoch": 0.25, "learning_rate": 7.515803513010218e-05, "loss": 1.0265, "step": 99755 }, { "epoch": 0.25, "learning_rate": 7.515677688862048e-05, "loss": 1.0247, "step": 99760 }, { "epoch": 0.25, "learning_rate": 7.515551864713877e-05, "loss": 1.0241, "step": 99765 }, { "epoch": 0.25, "learning_rate": 7.515426040565706e-05, "loss": 1.0238, "step": 99770 }, { "epoch": 0.25, "learning_rate": 7.515300216417535e-05, "loss": 1.0433, "step": 99775 }, { "epoch": 0.25, "learning_rate": 7.515174392269366e-05, "loss": 1.0236, "step": 99780 }, { "epoch": 0.25, "learning_rate": 7.515048568121195e-05, "loss": 1.0235, "step": 99785 }, { "epoch": 0.25, "learning_rate": 7.514922743973024e-05, "loss": 1.0218, "step": 99790 }, { "epoch": 0.25, "learning_rate": 7.514796919824853e-05, "loss": 1.0233, "step": 99795 }, { "epoch": 0.25, "learning_rate": 7.514671095676684e-05, "loss": 1.0234, "step": 99800 }, { "epoch": 0.25, "learning_rate": 7.514545271528513e-05, "loss": 1.0213, "step": 99805 }, { "epoch": 0.25, "learning_rate": 7.514419447380342e-05, "loss": 1.0239, "step": 99810 }, { "epoch": 0.25, "learning_rate": 7.514293623232171e-05, "loss": 1.0222, "step": 99815 }, { "epoch": 0.25, "learning_rate": 7.514167799084002e-05, "loss": 1.0216, "step": 99820 }, { "epoch": 0.25, "learning_rate": 7.514041974935831e-05, "loss": 1.0208, "step": 99825 }, { "epoch": 0.25, "learning_rate": 7.51391615078766e-05, "loss": 1.0257, "step": 99830 }, { "epoch": 0.25, "learning_rate": 7.513790326639489e-05, "loss": 1.0234, "step": 99835 }, { "epoch": 0.25, "learning_rate": 7.513664502491318e-05, "loss": 1.0236, "step": 99840 }, { "epoch": 0.25, "learning_rate": 7.513538678343149e-05, "loss": 1.0244, "step": 99845 }, { "epoch": 0.25, "learning_rate": 7.513412854194978e-05, "loss": 1.0231, "step": 99850 }, { "epoch": 0.25, "learning_rate": 7.513287030046807e-05, "loss": 1.023, "step": 99855 }, { "epoch": 0.25, "learning_rate": 7.513161205898636e-05, "loss": 1.0235, "step": 99860 }, { "epoch": 0.25, "learning_rate": 7.513035381750467e-05, "loss": 1.0233, "step": 99865 }, { "epoch": 0.25, "learning_rate": 7.512909557602296e-05, "loss": 1.0236, "step": 99870 }, { "epoch": 0.25, "learning_rate": 7.512783733454125e-05, "loss": 1.0234, "step": 99875 }, { "epoch": 0.25, "learning_rate": 7.512657909305954e-05, "loss": 1.0212, "step": 99880 }, { "epoch": 0.25, "learning_rate": 7.512532085157785e-05, "loss": 1.0247, "step": 99885 }, { "epoch": 0.25, "learning_rate": 7.512406261009614e-05, "loss": 1.0233, "step": 99890 }, { "epoch": 0.25, "learning_rate": 7.512280436861443e-05, "loss": 1.024, "step": 99895 }, { "epoch": 0.25, "learning_rate": 7.512154612713272e-05, "loss": 1.0202, "step": 99900 }, { "epoch": 0.25, "learning_rate": 7.512028788565101e-05, "loss": 1.0235, "step": 99905 }, { "epoch": 0.25, "learning_rate": 7.511902964416932e-05, "loss": 1.0264, "step": 99910 }, { "epoch": 0.25, "learning_rate": 7.511777140268761e-05, "loss": 1.0234, "step": 99915 }, { "epoch": 0.25, "learning_rate": 7.51165131612059e-05, "loss": 1.0247, "step": 99920 }, { "epoch": 0.25, "learning_rate": 7.511525491972419e-05, "loss": 1.0227, "step": 99925 }, { "epoch": 0.25, "learning_rate": 7.51139966782425e-05, "loss": 1.0223, "step": 99930 }, { "epoch": 0.25, "learning_rate": 7.511273843676079e-05, "loss": 1.0229, "step": 99935 }, { "epoch": 0.25, "learning_rate": 7.511148019527908e-05, "loss": 1.0244, "step": 99940 }, { "epoch": 0.25, "learning_rate": 7.511022195379737e-05, "loss": 1.0232, "step": 99945 }, { "epoch": 0.25, "learning_rate": 7.510896371231568e-05, "loss": 1.0241, "step": 99950 }, { "epoch": 0.25, "learning_rate": 7.510770547083397e-05, "loss": 1.0366, "step": 99955 }, { "epoch": 0.25, "learning_rate": 7.510644722935226e-05, "loss": 1.0233, "step": 99960 }, { "epoch": 0.25, "learning_rate": 7.510518898787055e-05, "loss": 1.0231, "step": 99965 }, { "epoch": 0.25, "learning_rate": 7.510393074638884e-05, "loss": 1.0246, "step": 99970 }, { "epoch": 0.25, "learning_rate": 7.510267250490715e-05, "loss": 1.0211, "step": 99975 }, { "epoch": 0.25, "learning_rate": 7.510141426342544e-05, "loss": 1.0231, "step": 99980 }, { "epoch": 0.25, "learning_rate": 7.510015602194373e-05, "loss": 1.0234, "step": 99985 }, { "epoch": 0.25, "learning_rate": 7.509889778046202e-05, "loss": 1.0232, "step": 99990 }, { "epoch": 0.25, "learning_rate": 7.509763953898033e-05, "loss": 1.0238, "step": 99995 }, { "epoch": 0.25, "learning_rate": 7.509638129749862e-05, "loss": 1.0239, "step": 100000 }, { "epoch": 0.25, "learning_rate": 7.509512305601691e-05, "loss": 1.0227, "step": 100005 }, { "epoch": 0.25, "learning_rate": 7.50938648145352e-05, "loss": 1.022, "step": 100010 }, { "epoch": 0.25, "learning_rate": 7.509260657305351e-05, "loss": 1.025, "step": 100015 }, { "epoch": 0.25, "learning_rate": 7.50913483315718e-05, "loss": 1.023, "step": 100020 }, { "epoch": 0.25, "learning_rate": 7.509009009009009e-05, "loss": 1.0251, "step": 100025 }, { "epoch": 0.25, "learning_rate": 7.508883184860838e-05, "loss": 1.0266, "step": 100030 }, { "epoch": 0.25, "learning_rate": 7.508757360712668e-05, "loss": 1.0225, "step": 100035 }, { "epoch": 0.25, "learning_rate": 7.508631536564498e-05, "loss": 1.0211, "step": 100040 }, { "epoch": 0.25, "learning_rate": 7.508505712416327e-05, "loss": 1.0247, "step": 100045 }, { "epoch": 0.25, "learning_rate": 7.508379888268156e-05, "loss": 1.043, "step": 100050 }, { "epoch": 0.25, "learning_rate": 7.508254064119986e-05, "loss": 1.0435, "step": 100055 }, { "epoch": 0.25, "learning_rate": 7.508128239971816e-05, "loss": 1.0235, "step": 100060 }, { "epoch": 0.25, "learning_rate": 7.508002415823645e-05, "loss": 1.0249, "step": 100065 }, { "epoch": 0.25, "learning_rate": 7.507876591675474e-05, "loss": 1.0217, "step": 100070 }, { "epoch": 0.25, "learning_rate": 7.507750767527304e-05, "loss": 1.0221, "step": 100075 }, { "epoch": 0.25, "learning_rate": 7.507624943379134e-05, "loss": 1.0251, "step": 100080 }, { "epoch": 0.25, "learning_rate": 7.507499119230963e-05, "loss": 1.0219, "step": 100085 }, { "epoch": 0.25, "learning_rate": 7.507373295082792e-05, "loss": 1.0258, "step": 100090 }, { "epoch": 0.25, "learning_rate": 7.507247470934622e-05, "loss": 1.0241, "step": 100095 }, { "epoch": 0.25, "learning_rate": 7.507121646786451e-05, "loss": 1.0235, "step": 100100 }, { "epoch": 0.25, "learning_rate": 7.506995822638281e-05, "loss": 1.0235, "step": 100105 }, { "epoch": 0.25, "learning_rate": 7.50686999849011e-05, "loss": 1.0256, "step": 100110 }, { "epoch": 0.25, "learning_rate": 7.50674417434194e-05, "loss": 1.0208, "step": 100115 }, { "epoch": 0.25, "learning_rate": 7.506618350193769e-05, "loss": 1.0201, "step": 100120 }, { "epoch": 0.25, "learning_rate": 7.506492526045599e-05, "loss": 1.0258, "step": 100125 }, { "epoch": 0.25, "learning_rate": 7.506366701897428e-05, "loss": 1.0242, "step": 100130 }, { "epoch": 0.25, "learning_rate": 7.506240877749258e-05, "loss": 1.0249, "step": 100135 }, { "epoch": 0.25, "learning_rate": 7.506115053601087e-05, "loss": 1.0248, "step": 100140 }, { "epoch": 0.25, "learning_rate": 7.505989229452917e-05, "loss": 1.022, "step": 100145 }, { "epoch": 0.25, "learning_rate": 7.505863405304746e-05, "loss": 1.0253, "step": 100150 }, { "epoch": 0.25, "learning_rate": 7.505737581156576e-05, "loss": 1.0227, "step": 100155 }, { "epoch": 0.25, "learning_rate": 7.505611757008405e-05, "loss": 1.0219, "step": 100160 }, { "epoch": 0.25, "learning_rate": 7.505485932860234e-05, "loss": 1.0225, "step": 100165 }, { "epoch": 0.25, "learning_rate": 7.505360108712064e-05, "loss": 1.0223, "step": 100170 }, { "epoch": 0.25, "learning_rate": 7.505234284563894e-05, "loss": 1.0247, "step": 100175 }, { "epoch": 0.25, "learning_rate": 7.505108460415723e-05, "loss": 1.021, "step": 100180 }, { "epoch": 0.25, "learning_rate": 7.504982636267552e-05, "loss": 1.0232, "step": 100185 }, { "epoch": 0.25, "learning_rate": 7.504856812119382e-05, "loss": 1.0227, "step": 100190 }, { "epoch": 0.25, "learning_rate": 7.504730987971212e-05, "loss": 1.0395, "step": 100195 }, { "epoch": 0.25, "learning_rate": 7.504605163823041e-05, "loss": 1.0246, "step": 100200 }, { "epoch": 0.25, "learning_rate": 7.50447933967487e-05, "loss": 1.0227, "step": 100205 }, { "epoch": 0.25, "learning_rate": 7.5043535155267e-05, "loss": 1.0232, "step": 100210 }, { "epoch": 0.25, "learning_rate": 7.50422769137853e-05, "loss": 1.0215, "step": 100215 }, { "epoch": 0.25, "learning_rate": 7.504101867230359e-05, "loss": 1.0217, "step": 100220 }, { "epoch": 0.25, "learning_rate": 7.503976043082188e-05, "loss": 1.0186, "step": 100225 }, { "epoch": 0.25, "learning_rate": 7.503850218934017e-05, "loss": 1.0239, "step": 100230 }, { "epoch": 0.25, "learning_rate": 7.503724394785848e-05, "loss": 1.0213, "step": 100235 }, { "epoch": 0.25, "learning_rate": 7.503598570637678e-05, "loss": 1.0272, "step": 100240 }, { "epoch": 0.25, "learning_rate": 7.503472746489507e-05, "loss": 1.0225, "step": 100245 }, { "epoch": 0.25, "learning_rate": 7.503346922341336e-05, "loss": 1.0233, "step": 100250 }, { "epoch": 0.25, "learning_rate": 7.503221098193166e-05, "loss": 1.0227, "step": 100255 }, { "epoch": 0.25, "learning_rate": 7.503095274044996e-05, "loss": 1.024, "step": 100260 }, { "epoch": 0.25, "learning_rate": 7.502969449896825e-05, "loss": 1.0252, "step": 100265 }, { "epoch": 0.25, "learning_rate": 7.502843625748654e-05, "loss": 1.0251, "step": 100270 }, { "epoch": 0.25, "learning_rate": 7.502717801600484e-05, "loss": 1.0226, "step": 100275 }, { "epoch": 0.25, "learning_rate": 7.502591977452314e-05, "loss": 1.0248, "step": 100280 }, { "epoch": 0.25, "learning_rate": 7.502466153304143e-05, "loss": 1.0229, "step": 100285 }, { "epoch": 0.25, "learning_rate": 7.502340329155972e-05, "loss": 1.0245, "step": 100290 }, { "epoch": 0.25, "learning_rate": 7.502214505007802e-05, "loss": 1.0252, "step": 100295 }, { "epoch": 0.25, "learning_rate": 7.502088680859631e-05, "loss": 1.0227, "step": 100300 }, { "epoch": 0.25, "learning_rate": 7.501962856711461e-05, "loss": 1.0211, "step": 100305 }, { "epoch": 0.25, "learning_rate": 7.50183703256329e-05, "loss": 1.0227, "step": 100310 }, { "epoch": 0.25, "learning_rate": 7.50171120841512e-05, "loss": 1.0237, "step": 100315 }, { "epoch": 0.25, "learning_rate": 7.501585384266949e-05, "loss": 1.0224, "step": 100320 }, { "epoch": 0.25, "learning_rate": 7.501459560118779e-05, "loss": 1.0234, "step": 100325 }, { "epoch": 0.25, "learning_rate": 7.501333735970608e-05, "loss": 1.0215, "step": 100330 }, { "epoch": 0.25, "learning_rate": 7.501207911822438e-05, "loss": 1.0248, "step": 100335 }, { "epoch": 0.25, "learning_rate": 7.501082087674267e-05, "loss": 1.0241, "step": 100340 }, { "epoch": 0.25, "learning_rate": 7.500956263526097e-05, "loss": 1.0218, "step": 100345 }, { "epoch": 0.25, "learning_rate": 7.500830439377926e-05, "loss": 1.0219, "step": 100350 }, { "epoch": 0.25, "learning_rate": 7.500704615229756e-05, "loss": 1.0245, "step": 100355 }, { "epoch": 0.25, "learning_rate": 7.500578791081585e-05, "loss": 1.0228, "step": 100360 }, { "epoch": 0.25, "learning_rate": 7.500452966933414e-05, "loss": 1.0236, "step": 100365 }, { "epoch": 0.25, "learning_rate": 7.500327142785244e-05, "loss": 1.0236, "step": 100370 }, { "epoch": 0.25, "learning_rate": 7.500201318637074e-05, "loss": 1.0255, "step": 100375 }, { "epoch": 0.25, "learning_rate": 7.500075494488903e-05, "loss": 1.0233, "step": 100380 }, { "epoch": 0.25, "learning_rate": 7.499949670340732e-05, "loss": 1.0251, "step": 100385 }, { "epoch": 0.25, "learning_rate": 7.499823846192562e-05, "loss": 1.024, "step": 100390 }, { "epoch": 0.25, "learning_rate": 7.499698022044391e-05, "loss": 1.022, "step": 100395 }, { "epoch": 0.25, "learning_rate": 7.49957219789622e-05, "loss": 1.0239, "step": 100400 }, { "epoch": 0.25, "learning_rate": 7.49944637374805e-05, "loss": 1.0224, "step": 100405 }, { "epoch": 0.25, "learning_rate": 7.49932054959988e-05, "loss": 1.0252, "step": 100410 }, { "epoch": 0.25, "learning_rate": 7.49919472545171e-05, "loss": 1.0238, "step": 100415 }, { "epoch": 0.25, "learning_rate": 7.499068901303539e-05, "loss": 1.0404, "step": 100420 }, { "epoch": 0.25, "learning_rate": 7.498943077155368e-05, "loss": 1.0241, "step": 100425 }, { "epoch": 0.25, "learning_rate": 7.498817253007197e-05, "loss": 1.0215, "step": 100430 }, { "epoch": 0.25, "learning_rate": 7.498691428859027e-05, "loss": 1.021, "step": 100435 }, { "epoch": 0.25, "learning_rate": 7.498565604710857e-05, "loss": 1.0229, "step": 100440 }, { "epoch": 0.25, "learning_rate": 7.498439780562686e-05, "loss": 1.0214, "step": 100445 }, { "epoch": 0.25, "learning_rate": 7.498313956414515e-05, "loss": 1.0244, "step": 100450 }, { "epoch": 0.25, "learning_rate": 7.498188132266345e-05, "loss": 1.0208, "step": 100455 }, { "epoch": 0.25, "learning_rate": 7.498062308118175e-05, "loss": 1.0254, "step": 100460 }, { "epoch": 0.25, "learning_rate": 7.497936483970004e-05, "loss": 1.0231, "step": 100465 }, { "epoch": 0.25, "learning_rate": 7.497810659821833e-05, "loss": 1.0217, "step": 100470 }, { "epoch": 0.25, "learning_rate": 7.497684835673663e-05, "loss": 1.0219, "step": 100475 }, { "epoch": 0.25, "learning_rate": 7.497559011525493e-05, "loss": 1.0246, "step": 100480 }, { "epoch": 0.25, "learning_rate": 7.497433187377322e-05, "loss": 1.0222, "step": 100485 }, { "epoch": 0.25, "learning_rate": 7.497307363229151e-05, "loss": 1.0236, "step": 100490 }, { "epoch": 0.25, "learning_rate": 7.49718153908098e-05, "loss": 1.0249, "step": 100495 }, { "epoch": 0.25, "learning_rate": 7.49705571493281e-05, "loss": 1.0235, "step": 100500 }, { "epoch": 0.25, "learning_rate": 7.49692989078464e-05, "loss": 1.0234, "step": 100505 }, { "epoch": 0.25, "learning_rate": 7.496804066636469e-05, "loss": 1.0226, "step": 100510 }, { "epoch": 0.25, "learning_rate": 7.496678242488298e-05, "loss": 1.0257, "step": 100515 }, { "epoch": 0.25, "learning_rate": 7.496552418340129e-05, "loss": 1.0221, "step": 100520 }, { "epoch": 0.25, "learning_rate": 7.496426594191958e-05, "loss": 1.0228, "step": 100525 }, { "epoch": 0.25, "learning_rate": 7.496300770043787e-05, "loss": 1.0217, "step": 100530 }, { "epoch": 0.25, "learning_rate": 7.496174945895616e-05, "loss": 1.0247, "step": 100535 }, { "epoch": 0.25, "learning_rate": 7.496049121747447e-05, "loss": 1.0252, "step": 100540 }, { "epoch": 0.25, "learning_rate": 7.495923297599276e-05, "loss": 1.023, "step": 100545 }, { "epoch": 0.25, "learning_rate": 7.495797473451105e-05, "loss": 1.0198, "step": 100550 }, { "epoch": 0.25, "learning_rate": 7.495671649302934e-05, "loss": 1.0218, "step": 100555 }, { "epoch": 0.25, "learning_rate": 7.495545825154763e-05, "loss": 1.0231, "step": 100560 }, { "epoch": 0.25, "learning_rate": 7.495420001006594e-05, "loss": 1.0254, "step": 100565 }, { "epoch": 0.25, "learning_rate": 7.495294176858423e-05, "loss": 1.0237, "step": 100570 }, { "epoch": 0.25, "learning_rate": 7.495168352710252e-05, "loss": 1.022, "step": 100575 }, { "epoch": 0.25, "learning_rate": 7.495042528562081e-05, "loss": 1.0242, "step": 100580 }, { "epoch": 0.25, "learning_rate": 7.494916704413912e-05, "loss": 1.0234, "step": 100585 }, { "epoch": 0.25, "learning_rate": 7.494790880265741e-05, "loss": 1.0223, "step": 100590 }, { "epoch": 0.25, "learning_rate": 7.49466505611757e-05, "loss": 1.024, "step": 100595 }, { "epoch": 0.25, "learning_rate": 7.494539231969399e-05, "loss": 1.0215, "step": 100600 }, { "epoch": 0.25, "learning_rate": 7.49441340782123e-05, "loss": 1.0228, "step": 100605 }, { "epoch": 0.25, "learning_rate": 7.494287583673059e-05, "loss": 1.0238, "step": 100610 }, { "epoch": 0.25, "learning_rate": 7.494161759524888e-05, "loss": 1.024, "step": 100615 }, { "epoch": 0.25, "learning_rate": 7.494035935376717e-05, "loss": 1.0205, "step": 100620 }, { "epoch": 0.25, "learning_rate": 7.493910111228546e-05, "loss": 1.0232, "step": 100625 }, { "epoch": 0.25, "learning_rate": 7.493784287080377e-05, "loss": 1.0228, "step": 100630 }, { "epoch": 0.25, "learning_rate": 7.493658462932206e-05, "loss": 1.0213, "step": 100635 }, { "epoch": 0.25, "learning_rate": 7.493532638784035e-05, "loss": 1.0217, "step": 100640 }, { "epoch": 0.25, "learning_rate": 7.493406814635864e-05, "loss": 1.0215, "step": 100645 }, { "epoch": 0.25, "learning_rate": 7.493280990487695e-05, "loss": 1.0211, "step": 100650 }, { "epoch": 0.25, "learning_rate": 7.493155166339524e-05, "loss": 1.0225, "step": 100655 }, { "epoch": 0.25, "learning_rate": 7.493029342191353e-05, "loss": 1.0406, "step": 100660 }, { "epoch": 0.25, "learning_rate": 7.492903518043182e-05, "loss": 1.0224, "step": 100665 }, { "epoch": 0.25, "learning_rate": 7.492777693895013e-05, "loss": 1.0234, "step": 100670 }, { "epoch": 0.25, "learning_rate": 7.492651869746842e-05, "loss": 1.0247, "step": 100675 }, { "epoch": 0.25, "learning_rate": 7.492526045598671e-05, "loss": 1.0204, "step": 100680 }, { "epoch": 0.25, "learning_rate": 7.4924002214505e-05, "loss": 1.0217, "step": 100685 }, { "epoch": 0.25, "learning_rate": 7.49227439730233e-05, "loss": 1.023, "step": 100690 }, { "epoch": 0.25, "learning_rate": 7.49214857315416e-05, "loss": 1.0225, "step": 100695 }, { "epoch": 0.25, "learning_rate": 7.492022749005989e-05, "loss": 1.0248, "step": 100700 }, { "epoch": 0.25, "learning_rate": 7.491896924857818e-05, "loss": 1.0223, "step": 100705 }, { "epoch": 0.25, "learning_rate": 7.491771100709648e-05, "loss": 1.0199, "step": 100710 }, { "epoch": 0.25, "learning_rate": 7.491645276561478e-05, "loss": 1.0238, "step": 100715 }, { "epoch": 0.25, "learning_rate": 7.491519452413307e-05, "loss": 1.0249, "step": 100720 }, { "epoch": 0.25, "learning_rate": 7.491393628265136e-05, "loss": 1.0229, "step": 100725 }, { "epoch": 0.25, "learning_rate": 7.491267804116966e-05, "loss": 1.0224, "step": 100730 }, { "epoch": 0.25, "learning_rate": 7.491141979968796e-05, "loss": 1.0201, "step": 100735 }, { "epoch": 0.25, "learning_rate": 7.491016155820627e-05, "loss": 1.0236, "step": 100740 }, { "epoch": 0.25, "learning_rate": 7.490890331672456e-05, "loss": 1.0237, "step": 100745 }, { "epoch": 0.25, "learning_rate": 7.490764507524285e-05, "loss": 1.0203, "step": 100750 }, { "epoch": 0.25, "learning_rate": 7.490638683376114e-05, "loss": 1.023, "step": 100755 }, { "epoch": 0.25, "learning_rate": 7.490512859227945e-05, "loss": 1.0228, "step": 100760 }, { "epoch": 0.25, "learning_rate": 7.490387035079774e-05, "loss": 1.0221, "step": 100765 }, { "epoch": 0.25, "learning_rate": 7.490261210931603e-05, "loss": 1.0245, "step": 100770 }, { "epoch": 0.25, "learning_rate": 7.490135386783432e-05, "loss": 1.024, "step": 100775 }, { "epoch": 0.25, "learning_rate": 7.490009562635261e-05, "loss": 1.0234, "step": 100780 }, { "epoch": 0.25, "learning_rate": 7.489883738487092e-05, "loss": 1.0226, "step": 100785 }, { "epoch": 0.25, "learning_rate": 7.489757914338921e-05, "loss": 1.0217, "step": 100790 }, { "epoch": 0.25, "learning_rate": 7.48963209019075e-05, "loss": 1.0241, "step": 100795 }, { "epoch": 0.25, "learning_rate": 7.489506266042579e-05, "loss": 1.0232, "step": 100800 }, { "epoch": 0.25, "learning_rate": 7.48938044189441e-05, "loss": 1.022, "step": 100805 }, { "epoch": 0.25, "learning_rate": 7.489254617746239e-05, "loss": 1.0234, "step": 100810 }, { "epoch": 0.25, "learning_rate": 7.489128793598068e-05, "loss": 1.0216, "step": 100815 }, { "epoch": 0.25, "learning_rate": 7.489002969449897e-05, "loss": 1.0229, "step": 100820 }, { "epoch": 0.25, "learning_rate": 7.488877145301728e-05, "loss": 1.0218, "step": 100825 }, { "epoch": 0.25, "learning_rate": 7.488751321153557e-05, "loss": 1.0224, "step": 100830 }, { "epoch": 0.25, "learning_rate": 7.488625497005386e-05, "loss": 1.0227, "step": 100835 }, { "epoch": 0.25, "learning_rate": 7.488499672857215e-05, "loss": 1.0228, "step": 100840 }, { "epoch": 0.25, "learning_rate": 7.488373848709044e-05, "loss": 1.0238, "step": 100845 }, { "epoch": 0.25, "learning_rate": 7.488248024560875e-05, "loss": 1.025, "step": 100850 }, { "epoch": 0.25, "learning_rate": 7.488122200412704e-05, "loss": 1.0265, "step": 100855 }, { "epoch": 0.25, "learning_rate": 7.487996376264533e-05, "loss": 1.0241, "step": 100860 }, { "epoch": 0.25, "learning_rate": 7.487870552116362e-05, "loss": 1.0223, "step": 100865 }, { "epoch": 0.25, "learning_rate": 7.487744727968193e-05, "loss": 1.0215, "step": 100870 }, { "epoch": 0.25, "learning_rate": 7.487618903820022e-05, "loss": 1.0222, "step": 100875 }, { "epoch": 0.25, "learning_rate": 7.487493079671851e-05, "loss": 1.024, "step": 100880 }, { "epoch": 0.25, "learning_rate": 7.48736725552368e-05, "loss": 1.0227, "step": 100885 }, { "epoch": 0.25, "learning_rate": 7.48724143137551e-05, "loss": 1.0224, "step": 100890 }, { "epoch": 0.25, "learning_rate": 7.48711560722734e-05, "loss": 1.0221, "step": 100895 }, { "epoch": 0.25, "learning_rate": 7.486989783079169e-05, "loss": 1.0215, "step": 100900 }, { "epoch": 0.25, "learning_rate": 7.486863958930998e-05, "loss": 1.0249, "step": 100905 }, { "epoch": 0.25, "learning_rate": 7.486738134782827e-05, "loss": 1.0241, "step": 100910 }, { "epoch": 0.25, "learning_rate": 7.486612310634658e-05, "loss": 1.023, "step": 100915 }, { "epoch": 0.25, "learning_rate": 7.486486486486487e-05, "loss": 1.0214, "step": 100920 }, { "epoch": 0.25, "learning_rate": 7.486360662338316e-05, "loss": 1.0242, "step": 100925 }, { "epoch": 0.25, "learning_rate": 7.486234838190145e-05, "loss": 1.0228, "step": 100930 }, { "epoch": 0.25, "learning_rate": 7.486109014041976e-05, "loss": 1.0224, "step": 100935 }, { "epoch": 0.25, "learning_rate": 7.485983189893805e-05, "loss": 1.0212, "step": 100940 }, { "epoch": 0.25, "learning_rate": 7.485857365745634e-05, "loss": 1.0233, "step": 100945 }, { "epoch": 0.25, "learning_rate": 7.485731541597463e-05, "loss": 1.0221, "step": 100950 }, { "epoch": 0.25, "learning_rate": 7.485605717449293e-05, "loss": 1.0234, "step": 100955 }, { "epoch": 0.25, "learning_rate": 7.485479893301123e-05, "loss": 1.0246, "step": 100960 }, { "epoch": 0.25, "learning_rate": 7.485354069152952e-05, "loss": 1.0223, "step": 100965 }, { "epoch": 0.25, "learning_rate": 7.485228245004781e-05, "loss": 1.0222, "step": 100970 }, { "epoch": 0.25, "learning_rate": 7.48510242085661e-05, "loss": 1.0238, "step": 100975 }, { "epoch": 0.25, "learning_rate": 7.484976596708441e-05, "loss": 1.022, "step": 100980 }, { "epoch": 0.25, "learning_rate": 7.48485077256027e-05, "loss": 1.0236, "step": 100985 }, { "epoch": 0.25, "learning_rate": 7.4847249484121e-05, "loss": 1.0229, "step": 100990 }, { "epoch": 0.25, "learning_rate": 7.484599124263929e-05, "loss": 1.0231, "step": 100995 }, { "epoch": 0.25, "learning_rate": 7.484473300115759e-05, "loss": 1.0231, "step": 101000 }, { "epoch": 0.25, "learning_rate": 7.484347475967588e-05, "loss": 1.0237, "step": 101005 }, { "epoch": 0.25, "learning_rate": 7.484221651819417e-05, "loss": 1.023, "step": 101010 }, { "epoch": 0.25, "learning_rate": 7.484095827671247e-05, "loss": 1.0214, "step": 101015 }, { "epoch": 0.25, "learning_rate": 7.483970003523076e-05, "loss": 1.0224, "step": 101020 }, { "epoch": 0.25, "learning_rate": 7.483844179374906e-05, "loss": 1.0212, "step": 101025 }, { "epoch": 0.25, "learning_rate": 7.483718355226735e-05, "loss": 1.0205, "step": 101030 }, { "epoch": 0.25, "learning_rate": 7.483592531078565e-05, "loss": 1.0235, "step": 101035 }, { "epoch": 0.25, "learning_rate": 7.483466706930394e-05, "loss": 1.023, "step": 101040 }, { "epoch": 0.25, "learning_rate": 7.483340882782224e-05, "loss": 1.0246, "step": 101045 }, { "epoch": 0.25, "learning_rate": 7.483215058634053e-05, "loss": 1.022, "step": 101050 }, { "epoch": 0.25, "learning_rate": 7.483089234485883e-05, "loss": 1.0226, "step": 101055 }, { "epoch": 0.25, "learning_rate": 7.482963410337712e-05, "loss": 1.0239, "step": 101060 }, { "epoch": 0.25, "learning_rate": 7.482837586189542e-05, "loss": 1.0245, "step": 101065 }, { "epoch": 0.25, "learning_rate": 7.482711762041371e-05, "loss": 1.0232, "step": 101070 }, { "epoch": 0.25, "learning_rate": 7.4825859378932e-05, "loss": 1.0239, "step": 101075 }, { "epoch": 0.25, "learning_rate": 7.48246011374503e-05, "loss": 1.0272, "step": 101080 }, { "epoch": 0.25, "learning_rate": 7.482334289596859e-05, "loss": 1.0203, "step": 101085 }, { "epoch": 0.25, "learning_rate": 7.48220846544869e-05, "loss": 1.0238, "step": 101090 }, { "epoch": 0.25, "learning_rate": 7.482082641300519e-05, "loss": 1.0235, "step": 101095 }, { "epoch": 0.25, "learning_rate": 7.481956817152348e-05, "loss": 1.0216, "step": 101100 }, { "epoch": 0.25, "learning_rate": 7.481830993004177e-05, "loss": 1.0255, "step": 101105 }, { "epoch": 0.25, "learning_rate": 7.481705168856007e-05, "loss": 1.0242, "step": 101110 }, { "epoch": 0.25, "learning_rate": 7.481579344707837e-05, "loss": 1.0212, "step": 101115 }, { "epoch": 0.25, "learning_rate": 7.481453520559666e-05, "loss": 1.0243, "step": 101120 }, { "epoch": 0.25, "learning_rate": 7.481327696411495e-05, "loss": 1.0234, "step": 101125 }, { "epoch": 0.25, "learning_rate": 7.481201872263325e-05, "loss": 1.0247, "step": 101130 }, { "epoch": 0.25, "learning_rate": 7.481076048115155e-05, "loss": 1.0222, "step": 101135 }, { "epoch": 0.25, "learning_rate": 7.480950223966984e-05, "loss": 1.0234, "step": 101140 }, { "epoch": 0.25, "learning_rate": 7.480824399818813e-05, "loss": 1.0218, "step": 101145 }, { "epoch": 0.25, "learning_rate": 7.480698575670642e-05, "loss": 1.0236, "step": 101150 }, { "epoch": 0.25, "learning_rate": 7.480572751522473e-05, "loss": 1.0229, "step": 101155 }, { "epoch": 0.25, "learning_rate": 7.480446927374302e-05, "loss": 1.0237, "step": 101160 }, { "epoch": 0.25, "learning_rate": 7.480321103226131e-05, "loss": 1.0262, "step": 101165 }, { "epoch": 0.25, "learning_rate": 7.48019527907796e-05, "loss": 1.0226, "step": 101170 }, { "epoch": 0.25, "learning_rate": 7.48006945492979e-05, "loss": 1.0219, "step": 101175 }, { "epoch": 0.25, "learning_rate": 7.47994363078162e-05, "loss": 1.0234, "step": 101180 }, { "epoch": 0.25, "learning_rate": 7.479817806633449e-05, "loss": 1.024, "step": 101185 }, { "epoch": 0.25, "learning_rate": 7.479691982485278e-05, "loss": 1.0246, "step": 101190 }, { "epoch": 0.25, "learning_rate": 7.479566158337109e-05, "loss": 1.0234, "step": 101195 }, { "epoch": 0.25, "learning_rate": 7.479440334188938e-05, "loss": 1.0208, "step": 101200 }, { "epoch": 0.25, "learning_rate": 7.479314510040767e-05, "loss": 1.0245, "step": 101205 }, { "epoch": 0.25, "learning_rate": 7.479188685892596e-05, "loss": 1.0215, "step": 101210 }, { "epoch": 0.25, "learning_rate": 7.479062861744425e-05, "loss": 1.0231, "step": 101215 }, { "epoch": 0.25, "learning_rate": 7.478937037596256e-05, "loss": 1.0221, "step": 101220 }, { "epoch": 0.25, "learning_rate": 7.478811213448085e-05, "loss": 1.0235, "step": 101225 }, { "epoch": 0.25, "learning_rate": 7.478685389299914e-05, "loss": 1.0223, "step": 101230 }, { "epoch": 0.25, "learning_rate": 7.478559565151745e-05, "loss": 1.0253, "step": 101235 }, { "epoch": 0.25, "learning_rate": 7.478433741003574e-05, "loss": 1.023, "step": 101240 }, { "epoch": 0.25, "learning_rate": 7.478307916855404e-05, "loss": 1.0219, "step": 101245 }, { "epoch": 0.25, "learning_rate": 7.478182092707233e-05, "loss": 1.0238, "step": 101250 }, { "epoch": 0.25, "learning_rate": 7.478056268559063e-05, "loss": 1.0248, "step": 101255 }, { "epoch": 0.25, "learning_rate": 7.477930444410892e-05, "loss": 1.0236, "step": 101260 }, { "epoch": 0.25, "learning_rate": 7.477804620262722e-05, "loss": 1.0252, "step": 101265 }, { "epoch": 0.25, "learning_rate": 7.477678796114551e-05, "loss": 1.0253, "step": 101270 }, { "epoch": 0.25, "learning_rate": 7.47755297196638e-05, "loss": 1.0232, "step": 101275 }, { "epoch": 0.25, "learning_rate": 7.47742714781821e-05, "loss": 1.0461, "step": 101280 }, { "epoch": 0.25, "learning_rate": 7.47730132367004e-05, "loss": 1.0223, "step": 101285 }, { "epoch": 0.25, "learning_rate": 7.47717549952187e-05, "loss": 1.0252, "step": 101290 }, { "epoch": 0.25, "learning_rate": 7.477049675373699e-05, "loss": 1.0225, "step": 101295 }, { "epoch": 0.25, "learning_rate": 7.476923851225528e-05, "loss": 1.0253, "step": 101300 }, { "epoch": 0.25, "learning_rate": 7.476798027077357e-05, "loss": 1.0222, "step": 101305 }, { "epoch": 0.25, "learning_rate": 7.476672202929187e-05, "loss": 1.0231, "step": 101310 }, { "epoch": 0.25, "learning_rate": 7.476546378781017e-05, "loss": 1.0252, "step": 101315 }, { "epoch": 0.25, "learning_rate": 7.476420554632846e-05, "loss": 1.0229, "step": 101320 }, { "epoch": 0.25, "learning_rate": 7.476294730484675e-05, "loss": 1.0237, "step": 101325 }, { "epoch": 0.25, "learning_rate": 7.476168906336505e-05, "loss": 1.0214, "step": 101330 }, { "epoch": 0.25, "learning_rate": 7.476043082188335e-05, "loss": 1.0228, "step": 101335 }, { "epoch": 0.25, "learning_rate": 7.475917258040164e-05, "loss": 1.0241, "step": 101340 }, { "epoch": 0.25, "learning_rate": 7.475791433891993e-05, "loss": 1.0212, "step": 101345 }, { "epoch": 0.25, "learning_rate": 7.475665609743823e-05, "loss": 1.0247, "step": 101350 }, { "epoch": 0.25, "learning_rate": 7.475539785595653e-05, "loss": 1.0213, "step": 101355 }, { "epoch": 0.25, "learning_rate": 7.475413961447482e-05, "loss": 1.0258, "step": 101360 }, { "epoch": 0.25, "learning_rate": 7.475288137299311e-05, "loss": 1.0232, "step": 101365 }, { "epoch": 0.25, "learning_rate": 7.47516231315114e-05, "loss": 1.0211, "step": 101370 }, { "epoch": 0.25, "learning_rate": 7.47503648900297e-05, "loss": 1.0225, "step": 101375 }, { "epoch": 0.25, "learning_rate": 7.4749106648548e-05, "loss": 1.0246, "step": 101380 }, { "epoch": 0.25, "learning_rate": 7.474784840706629e-05, "loss": 1.0223, "step": 101385 }, { "epoch": 0.25, "learning_rate": 7.474659016558458e-05, "loss": 1.02, "step": 101390 }, { "epoch": 0.25, "learning_rate": 7.474533192410288e-05, "loss": 1.0221, "step": 101395 }, { "epoch": 0.25, "learning_rate": 7.474407368262118e-05, "loss": 1.0234, "step": 101400 }, { "epoch": 0.25, "learning_rate": 7.474281544113947e-05, "loss": 1.0249, "step": 101405 }, { "epoch": 0.25, "learning_rate": 7.474155719965776e-05, "loss": 1.0246, "step": 101410 }, { "epoch": 0.25, "learning_rate": 7.474029895817606e-05, "loss": 1.0226, "step": 101415 }, { "epoch": 0.25, "learning_rate": 7.473904071669436e-05, "loss": 1.0242, "step": 101420 }, { "epoch": 0.25, "learning_rate": 7.473778247521265e-05, "loss": 1.0248, "step": 101425 }, { "epoch": 0.25, "learning_rate": 7.473652423373094e-05, "loss": 1.024, "step": 101430 }, { "epoch": 0.25, "learning_rate": 7.473526599224923e-05, "loss": 1.0218, "step": 101435 }, { "epoch": 0.25, "learning_rate": 7.473400775076754e-05, "loss": 1.0222, "step": 101440 }, { "epoch": 0.25, "learning_rate": 7.473274950928583e-05, "loss": 1.0235, "step": 101445 }, { "epoch": 0.25, "learning_rate": 7.473149126780412e-05, "loss": 1.0244, "step": 101450 }, { "epoch": 0.25, "learning_rate": 7.473023302632241e-05, "loss": 1.0238, "step": 101455 }, { "epoch": 0.25, "learning_rate": 7.472897478484072e-05, "loss": 1.0247, "step": 101460 }, { "epoch": 0.25, "learning_rate": 7.472771654335901e-05, "loss": 1.0236, "step": 101465 }, { "epoch": 0.25, "learning_rate": 7.47264583018773e-05, "loss": 1.0444, "step": 101470 }, { "epoch": 0.25, "learning_rate": 7.472520006039559e-05, "loss": 1.0229, "step": 101475 }, { "epoch": 0.25, "learning_rate": 7.47239418189139e-05, "loss": 1.0216, "step": 101480 }, { "epoch": 0.25, "learning_rate": 7.472293522572853e-05, "loss": 1.0222, "step": 101485 }, { "epoch": 0.25, "learning_rate": 7.472167698424682e-05, "loss": 1.023, "step": 101490 }, { "epoch": 0.25, "learning_rate": 7.472041874276512e-05, "loss": 1.0425, "step": 101495 }, { "epoch": 0.25, "learning_rate": 7.471916050128342e-05, "loss": 1.0238, "step": 101500 }, { "epoch": 0.25, "learning_rate": 7.471790225980171e-05, "loss": 1.0244, "step": 101505 }, { "epoch": 0.25, "learning_rate": 7.471664401832e-05, "loss": 1.022, "step": 101510 }, { "epoch": 0.25, "learning_rate": 7.47153857768383e-05, "loss": 1.024, "step": 101515 }, { "epoch": 0.25, "learning_rate": 7.471412753535659e-05, "loss": 1.025, "step": 101520 }, { "epoch": 0.25, "learning_rate": 7.471286929387489e-05, "loss": 1.0241, "step": 101525 }, { "epoch": 0.25, "learning_rate": 7.471161105239318e-05, "loss": 1.0243, "step": 101530 }, { "epoch": 0.25, "learning_rate": 7.471035281091148e-05, "loss": 1.0226, "step": 101535 }, { "epoch": 0.25, "learning_rate": 7.470909456942977e-05, "loss": 1.0255, "step": 101540 }, { "epoch": 0.25, "learning_rate": 7.470783632794807e-05, "loss": 1.0249, "step": 101545 }, { "epoch": 0.25, "learning_rate": 7.470657808646636e-05, "loss": 1.022, "step": 101550 }, { "epoch": 0.25, "learning_rate": 7.470531984498466e-05, "loss": 1.0219, "step": 101555 }, { "epoch": 0.25, "learning_rate": 7.470406160350295e-05, "loss": 1.0244, "step": 101560 }, { "epoch": 0.25, "learning_rate": 7.470280336202124e-05, "loss": 1.0214, "step": 101565 }, { "epoch": 0.25, "learning_rate": 7.470154512053954e-05, "loss": 1.0231, "step": 101570 }, { "epoch": 0.25, "learning_rate": 7.470028687905784e-05, "loss": 1.0247, "step": 101575 }, { "epoch": 0.25, "learning_rate": 7.469902863757613e-05, "loss": 1.0232, "step": 101580 }, { "epoch": 0.25, "learning_rate": 7.469777039609442e-05, "loss": 1.0222, "step": 101585 }, { "epoch": 0.26, "learning_rate": 7.469651215461272e-05, "loss": 1.0461, "step": 101590 }, { "epoch": 0.26, "learning_rate": 7.469525391313102e-05, "loss": 1.0245, "step": 101595 }, { "epoch": 0.26, "learning_rate": 7.469399567164931e-05, "loss": 1.0233, "step": 101600 }, { "epoch": 0.26, "learning_rate": 7.46927374301676e-05, "loss": 1.0237, "step": 101605 }, { "epoch": 0.26, "learning_rate": 7.46914791886859e-05, "loss": 1.0245, "step": 101610 }, { "epoch": 0.26, "learning_rate": 7.46902209472042e-05, "loss": 1.0241, "step": 101615 }, { "epoch": 0.26, "learning_rate": 7.468896270572249e-05, "loss": 1.0235, "step": 101620 }, { "epoch": 0.26, "learning_rate": 7.468770446424078e-05, "loss": 1.0212, "step": 101625 }, { "epoch": 0.26, "learning_rate": 7.468644622275907e-05, "loss": 1.0256, "step": 101630 }, { "epoch": 0.26, "learning_rate": 7.468518798127738e-05, "loss": 1.0241, "step": 101635 }, { "epoch": 0.26, "learning_rate": 7.468392973979567e-05, "loss": 1.0242, "step": 101640 }, { "epoch": 0.26, "learning_rate": 7.468267149831396e-05, "loss": 1.0255, "step": 101645 }, { "epoch": 0.26, "learning_rate": 7.468141325683225e-05, "loss": 1.0237, "step": 101650 }, { "epoch": 0.26, "learning_rate": 7.468015501535056e-05, "loss": 1.0246, "step": 101655 }, { "epoch": 0.26, "learning_rate": 7.467889677386885e-05, "loss": 1.0218, "step": 101660 }, { "epoch": 0.26, "learning_rate": 7.467763853238714e-05, "loss": 1.0226, "step": 101665 }, { "epoch": 0.26, "learning_rate": 7.467638029090543e-05, "loss": 1.0226, "step": 101670 }, { "epoch": 0.26, "learning_rate": 7.467512204942373e-05, "loss": 1.0209, "step": 101675 }, { "epoch": 0.26, "learning_rate": 7.467386380794203e-05, "loss": 1.0239, "step": 101680 }, { "epoch": 0.26, "learning_rate": 7.467260556646032e-05, "loss": 1.0218, "step": 101685 }, { "epoch": 0.26, "learning_rate": 7.467134732497861e-05, "loss": 1.0246, "step": 101690 }, { "epoch": 0.26, "learning_rate": 7.46700890834969e-05, "loss": 1.0234, "step": 101695 }, { "epoch": 0.26, "learning_rate": 7.46688308420152e-05, "loss": 1.0231, "step": 101700 }, { "epoch": 0.26, "learning_rate": 7.46675726005335e-05, "loss": 1.0213, "step": 101705 }, { "epoch": 0.26, "learning_rate": 7.466631435905179e-05, "loss": 1.0235, "step": 101710 }, { "epoch": 0.26, "learning_rate": 7.466505611757008e-05, "loss": 1.0449, "step": 101715 }, { "epoch": 0.26, "learning_rate": 7.466379787608839e-05, "loss": 1.0257, "step": 101720 }, { "epoch": 0.26, "learning_rate": 7.466253963460668e-05, "loss": 1.0239, "step": 101725 }, { "epoch": 0.26, "learning_rate": 7.466128139312497e-05, "loss": 1.0225, "step": 101730 }, { "epoch": 0.26, "learning_rate": 7.466002315164326e-05, "loss": 1.0237, "step": 101735 }, { "epoch": 0.26, "learning_rate": 7.465876491016157e-05, "loss": 1.0223, "step": 101740 }, { "epoch": 0.26, "learning_rate": 7.465750666867986e-05, "loss": 1.0211, "step": 101745 }, { "epoch": 0.26, "learning_rate": 7.465624842719815e-05, "loss": 1.0198, "step": 101750 }, { "epoch": 0.26, "learning_rate": 7.465499018571644e-05, "loss": 1.021, "step": 101755 }, { "epoch": 0.26, "learning_rate": 7.465373194423473e-05, "loss": 1.0216, "step": 101760 }, { "epoch": 0.26, "learning_rate": 7.465247370275304e-05, "loss": 1.0251, "step": 101765 }, { "epoch": 0.26, "learning_rate": 7.465121546127133e-05, "loss": 1.0235, "step": 101770 }, { "epoch": 0.26, "learning_rate": 7.464995721978962e-05, "loss": 1.0238, "step": 101775 }, { "epoch": 0.26, "learning_rate": 7.464869897830791e-05, "loss": 1.0219, "step": 101780 }, { "epoch": 0.26, "learning_rate": 7.464744073682622e-05, "loss": 1.024, "step": 101785 }, { "epoch": 0.26, "learning_rate": 7.464618249534451e-05, "loss": 1.0259, "step": 101790 }, { "epoch": 0.26, "learning_rate": 7.46449242538628e-05, "loss": 1.0219, "step": 101795 }, { "epoch": 0.26, "learning_rate": 7.464366601238109e-05, "loss": 1.0247, "step": 101800 }, { "epoch": 0.26, "learning_rate": 7.46424077708994e-05, "loss": 1.0243, "step": 101805 }, { "epoch": 0.26, "learning_rate": 7.464114952941769e-05, "loss": 1.0233, "step": 101810 }, { "epoch": 0.26, "learning_rate": 7.463989128793598e-05, "loss": 1.0249, "step": 101815 }, { "epoch": 0.26, "learning_rate": 7.463863304645427e-05, "loss": 1.0209, "step": 101820 }, { "epoch": 0.26, "learning_rate": 7.463737480497256e-05, "loss": 1.0238, "step": 101825 }, { "epoch": 0.26, "learning_rate": 7.463611656349087e-05, "loss": 1.0241, "step": 101830 }, { "epoch": 0.26, "learning_rate": 7.463485832200916e-05, "loss": 1.0241, "step": 101835 }, { "epoch": 0.26, "learning_rate": 7.463360008052745e-05, "loss": 1.0245, "step": 101840 }, { "epoch": 0.26, "learning_rate": 7.463234183904574e-05, "loss": 1.0245, "step": 101845 }, { "epoch": 0.26, "learning_rate": 7.463108359756405e-05, "loss": 1.0234, "step": 101850 }, { "epoch": 0.26, "learning_rate": 7.462982535608234e-05, "loss": 1.0195, "step": 101855 }, { "epoch": 0.26, "learning_rate": 7.462856711460063e-05, "loss": 1.0235, "step": 101860 }, { "epoch": 0.26, "learning_rate": 7.462730887311892e-05, "loss": 1.0256, "step": 101865 }, { "epoch": 0.26, "learning_rate": 7.462605063163723e-05, "loss": 1.0216, "step": 101870 }, { "epoch": 0.26, "learning_rate": 7.462479239015552e-05, "loss": 1.0232, "step": 101875 }, { "epoch": 0.26, "learning_rate": 7.462353414867381e-05, "loss": 1.0228, "step": 101880 }, { "epoch": 0.26, "learning_rate": 7.46222759071921e-05, "loss": 1.0238, "step": 101885 }, { "epoch": 0.26, "learning_rate": 7.46210176657104e-05, "loss": 1.0246, "step": 101890 }, { "epoch": 0.26, "learning_rate": 7.46197594242287e-05, "loss": 1.0233, "step": 101895 }, { "epoch": 0.26, "learning_rate": 7.461850118274699e-05, "loss": 1.0231, "step": 101900 }, { "epoch": 0.26, "learning_rate": 7.461724294126528e-05, "loss": 1.0243, "step": 101905 }, { "epoch": 0.26, "learning_rate": 7.461598469978358e-05, "loss": 1.0226, "step": 101910 }, { "epoch": 0.26, "learning_rate": 7.461472645830188e-05, "loss": 1.0247, "step": 101915 }, { "epoch": 0.26, "learning_rate": 7.461346821682017e-05, "loss": 1.0228, "step": 101920 }, { "epoch": 0.26, "learning_rate": 7.461220997533846e-05, "loss": 1.0241, "step": 101925 }, { "epoch": 0.26, "learning_rate": 7.461095173385676e-05, "loss": 1.0231, "step": 101930 }, { "epoch": 0.26, "learning_rate": 7.460969349237506e-05, "loss": 1.023, "step": 101935 }, { "epoch": 0.26, "learning_rate": 7.460843525089337e-05, "loss": 1.0245, "step": 101940 }, { "epoch": 0.26, "learning_rate": 7.460717700941166e-05, "loss": 1.022, "step": 101945 }, { "epoch": 0.26, "learning_rate": 7.460591876792995e-05, "loss": 1.0251, "step": 101950 }, { "epoch": 0.26, "learning_rate": 7.460466052644824e-05, "loss": 1.0237, "step": 101955 }, { "epoch": 0.26, "learning_rate": 7.460340228496655e-05, "loss": 1.026, "step": 101960 }, { "epoch": 0.26, "learning_rate": 7.460214404348484e-05, "loss": 1.0244, "step": 101965 }, { "epoch": 0.26, "learning_rate": 7.460088580200313e-05, "loss": 1.0246, "step": 101970 }, { "epoch": 0.26, "learning_rate": 7.459962756052142e-05, "loss": 1.0418, "step": 101975 }, { "epoch": 0.26, "learning_rate": 7.459836931903971e-05, "loss": 1.0243, "step": 101980 }, { "epoch": 0.26, "learning_rate": 7.459711107755802e-05, "loss": 1.0225, "step": 101985 }, { "epoch": 0.26, "learning_rate": 7.459585283607631e-05, "loss": 1.0237, "step": 101990 }, { "epoch": 0.26, "learning_rate": 7.45945945945946e-05, "loss": 1.0222, "step": 101995 }, { "epoch": 0.26, "learning_rate": 7.459333635311289e-05, "loss": 1.0241, "step": 102000 }, { "epoch": 0.26, "learning_rate": 7.45920781116312e-05, "loss": 1.0208, "step": 102005 }, { "epoch": 0.26, "learning_rate": 7.459081987014949e-05, "loss": 1.0221, "step": 102010 }, { "epoch": 0.26, "learning_rate": 7.458956162866778e-05, "loss": 1.0237, "step": 102015 }, { "epoch": 0.26, "learning_rate": 7.458830338718607e-05, "loss": 1.0203, "step": 102020 }, { "epoch": 0.26, "learning_rate": 7.458704514570438e-05, "loss": 1.0231, "step": 102025 }, { "epoch": 0.26, "learning_rate": 7.458578690422267e-05, "loss": 1.0244, "step": 102030 }, { "epoch": 0.26, "learning_rate": 7.458452866274096e-05, "loss": 1.0216, "step": 102035 }, { "epoch": 0.26, "learning_rate": 7.458327042125925e-05, "loss": 1.0229, "step": 102040 }, { "epoch": 0.26, "learning_rate": 7.458201217977754e-05, "loss": 1.0212, "step": 102045 }, { "epoch": 0.26, "learning_rate": 7.458075393829585e-05, "loss": 1.0245, "step": 102050 }, { "epoch": 0.26, "learning_rate": 7.457949569681414e-05, "loss": 1.0229, "step": 102055 }, { "epoch": 0.26, "learning_rate": 7.457823745533243e-05, "loss": 1.022, "step": 102060 }, { "epoch": 0.26, "learning_rate": 7.457697921385072e-05, "loss": 1.0211, "step": 102065 }, { "epoch": 0.26, "learning_rate": 7.457572097236903e-05, "loss": 1.0238, "step": 102070 }, { "epoch": 0.26, "learning_rate": 7.457446273088732e-05, "loss": 1.0231, "step": 102075 }, { "epoch": 0.26, "learning_rate": 7.457320448940561e-05, "loss": 1.021, "step": 102080 }, { "epoch": 0.26, "learning_rate": 7.45719462479239e-05, "loss": 1.0231, "step": 102085 }, { "epoch": 0.26, "learning_rate": 7.457068800644221e-05, "loss": 1.0226, "step": 102090 }, { "epoch": 0.26, "learning_rate": 7.45694297649605e-05, "loss": 1.0234, "step": 102095 }, { "epoch": 0.26, "learning_rate": 7.456817152347879e-05, "loss": 1.0233, "step": 102100 }, { "epoch": 0.26, "learning_rate": 7.456691328199708e-05, "loss": 1.0219, "step": 102105 }, { "epoch": 0.26, "learning_rate": 7.456565504051538e-05, "loss": 1.0226, "step": 102110 }, { "epoch": 0.26, "learning_rate": 7.456439679903368e-05, "loss": 1.0224, "step": 102115 }, { "epoch": 0.26, "learning_rate": 7.456313855755197e-05, "loss": 1.0209, "step": 102120 }, { "epoch": 0.26, "learning_rate": 7.456188031607026e-05, "loss": 1.0249, "step": 102125 }, { "epoch": 0.26, "learning_rate": 7.456062207458855e-05, "loss": 1.0227, "step": 102130 }, { "epoch": 0.26, "learning_rate": 7.455936383310686e-05, "loss": 1.022, "step": 102135 }, { "epoch": 0.26, "learning_rate": 7.455810559162515e-05, "loss": 1.0239, "step": 102140 }, { "epoch": 0.26, "learning_rate": 7.455684735014344e-05, "loss": 1.0243, "step": 102145 }, { "epoch": 0.26, "learning_rate": 7.455558910866173e-05, "loss": 1.0455, "step": 102150 }, { "epoch": 0.26, "learning_rate": 7.455433086718004e-05, "loss": 1.0225, "step": 102155 }, { "epoch": 0.26, "learning_rate": 7.455307262569833e-05, "loss": 1.0226, "step": 102160 }, { "epoch": 0.26, "learning_rate": 7.455181438421662e-05, "loss": 1.0215, "step": 102165 }, { "epoch": 0.26, "learning_rate": 7.455055614273491e-05, "loss": 1.0246, "step": 102170 }, { "epoch": 0.26, "learning_rate": 7.45492979012532e-05, "loss": 1.0224, "step": 102175 }, { "epoch": 0.26, "learning_rate": 7.454803965977151e-05, "loss": 1.0222, "step": 102180 }, { "epoch": 0.26, "learning_rate": 7.45467814182898e-05, "loss": 1.0239, "step": 102185 }, { "epoch": 0.26, "learning_rate": 7.45455231768081e-05, "loss": 1.0252, "step": 102190 }, { "epoch": 0.26, "learning_rate": 7.454426493532639e-05, "loss": 1.0259, "step": 102195 }, { "epoch": 0.26, "learning_rate": 7.454300669384469e-05, "loss": 1.0243, "step": 102200 }, { "epoch": 0.26, "learning_rate": 7.454174845236298e-05, "loss": 1.0226, "step": 102205 }, { "epoch": 0.26, "learning_rate": 7.454049021088127e-05, "loss": 1.024, "step": 102210 }, { "epoch": 0.26, "learning_rate": 7.453923196939957e-05, "loss": 1.0213, "step": 102215 }, { "epoch": 0.26, "learning_rate": 7.453797372791787e-05, "loss": 1.0205, "step": 102220 }, { "epoch": 0.26, "learning_rate": 7.453671548643616e-05, "loss": 1.0223, "step": 102225 }, { "epoch": 0.26, "learning_rate": 7.453545724495445e-05, "loss": 1.0242, "step": 102230 }, { "epoch": 0.26, "learning_rate": 7.453419900347275e-05, "loss": 1.0236, "step": 102235 }, { "epoch": 0.26, "learning_rate": 7.453294076199104e-05, "loss": 1.0244, "step": 102240 }, { "epoch": 0.26, "learning_rate": 7.453168252050934e-05, "loss": 1.0225, "step": 102245 }, { "epoch": 0.26, "learning_rate": 7.453042427902763e-05, "loss": 1.023, "step": 102250 }, { "epoch": 0.26, "learning_rate": 7.452916603754593e-05, "loss": 1.0221, "step": 102255 }, { "epoch": 0.26, "learning_rate": 7.452790779606422e-05, "loss": 1.0232, "step": 102260 }, { "epoch": 0.26, "learning_rate": 7.452664955458252e-05, "loss": 1.0228, "step": 102265 }, { "epoch": 0.26, "learning_rate": 7.452539131310081e-05, "loss": 1.0205, "step": 102270 }, { "epoch": 0.26, "learning_rate": 7.45241330716191e-05, "loss": 1.0233, "step": 102275 }, { "epoch": 0.26, "learning_rate": 7.45228748301374e-05, "loss": 1.021, "step": 102280 }, { "epoch": 0.26, "learning_rate": 7.452161658865569e-05, "loss": 1.0225, "step": 102285 }, { "epoch": 0.26, "learning_rate": 7.4520358347174e-05, "loss": 1.0231, "step": 102290 }, { "epoch": 0.26, "learning_rate": 7.451910010569229e-05, "loss": 1.0231, "step": 102295 }, { "epoch": 0.26, "learning_rate": 7.451784186421058e-05, "loss": 1.0211, "step": 102300 }, { "epoch": 0.26, "learning_rate": 7.451658362272887e-05, "loss": 1.0218, "step": 102305 }, { "epoch": 0.26, "learning_rate": 7.451532538124717e-05, "loss": 1.0204, "step": 102310 }, { "epoch": 0.26, "learning_rate": 7.451406713976547e-05, "loss": 1.0243, "step": 102315 }, { "epoch": 0.26, "learning_rate": 7.451280889828376e-05, "loss": 1.0227, "step": 102320 }, { "epoch": 0.26, "learning_rate": 7.451155065680205e-05, "loss": 1.0247, "step": 102325 }, { "epoch": 0.26, "learning_rate": 7.451029241532035e-05, "loss": 1.0242, "step": 102330 }, { "epoch": 0.26, "learning_rate": 7.450903417383865e-05, "loss": 1.0224, "step": 102335 }, { "epoch": 0.26, "learning_rate": 7.450777593235694e-05, "loss": 1.0227, "step": 102340 }, { "epoch": 0.26, "learning_rate": 7.450651769087523e-05, "loss": 1.0234, "step": 102345 }, { "epoch": 0.26, "learning_rate": 7.450525944939352e-05, "loss": 1.0234, "step": 102350 }, { "epoch": 0.26, "learning_rate": 7.450400120791183e-05, "loss": 1.0242, "step": 102355 }, { "epoch": 0.26, "learning_rate": 7.450274296643012e-05, "loss": 1.0208, "step": 102360 }, { "epoch": 0.26, "learning_rate": 7.450148472494841e-05, "loss": 1.024, "step": 102365 }, { "epoch": 0.26, "learning_rate": 7.45002264834667e-05, "loss": 1.0233, "step": 102370 }, { "epoch": 0.26, "learning_rate": 7.4498968241985e-05, "loss": 1.024, "step": 102375 }, { "epoch": 0.26, "learning_rate": 7.44977100005033e-05, "loss": 1.0231, "step": 102380 }, { "epoch": 0.26, "learning_rate": 7.449645175902159e-05, "loss": 1.0231, "step": 102385 }, { "epoch": 0.26, "learning_rate": 7.449519351753988e-05, "loss": 1.0222, "step": 102390 }, { "epoch": 0.26, "learning_rate": 7.449393527605819e-05, "loss": 1.0219, "step": 102395 }, { "epoch": 0.26, "learning_rate": 7.449267703457648e-05, "loss": 1.0266, "step": 102400 }, { "epoch": 0.26, "learning_rate": 7.449141879309477e-05, "loss": 1.0237, "step": 102405 }, { "epoch": 0.26, "learning_rate": 7.449016055161306e-05, "loss": 1.0222, "step": 102410 }, { "epoch": 0.26, "learning_rate": 7.448890231013135e-05, "loss": 1.0226, "step": 102415 }, { "epoch": 0.26, "learning_rate": 7.448764406864966e-05, "loss": 1.0234, "step": 102420 }, { "epoch": 0.26, "learning_rate": 7.448638582716795e-05, "loss": 1.0215, "step": 102425 }, { "epoch": 0.26, "learning_rate": 7.448512758568624e-05, "loss": 1.0219, "step": 102430 }, { "epoch": 0.26, "learning_rate": 7.448386934420455e-05, "loss": 1.0221, "step": 102435 }, { "epoch": 0.26, "learning_rate": 7.448261110272284e-05, "loss": 1.0209, "step": 102440 }, { "epoch": 0.26, "learning_rate": 7.448135286124114e-05, "loss": 1.0245, "step": 102445 }, { "epoch": 0.26, "learning_rate": 7.448009461975943e-05, "loss": 1.0238, "step": 102450 }, { "epoch": 0.26, "learning_rate": 7.447883637827773e-05, "loss": 1.0213, "step": 102455 }, { "epoch": 0.26, "learning_rate": 7.447757813679602e-05, "loss": 1.0214, "step": 102460 }, { "epoch": 0.26, "learning_rate": 7.447631989531432e-05, "loss": 1.0233, "step": 102465 }, { "epoch": 0.26, "learning_rate": 7.447506165383261e-05, "loss": 1.0244, "step": 102470 }, { "epoch": 0.26, "learning_rate": 7.44738034123509e-05, "loss": 1.0227, "step": 102475 }, { "epoch": 0.26, "learning_rate": 7.44725451708692e-05, "loss": 1.0221, "step": 102480 }, { "epoch": 0.26, "learning_rate": 7.44712869293875e-05, "loss": 1.0231, "step": 102485 }, { "epoch": 0.26, "learning_rate": 7.44700286879058e-05, "loss": 1.0213, "step": 102490 }, { "epoch": 0.26, "learning_rate": 7.446877044642409e-05, "loss": 1.023, "step": 102495 }, { "epoch": 0.26, "learning_rate": 7.446751220494238e-05, "loss": 1.0236, "step": 102500 }, { "epoch": 0.26, "learning_rate": 7.446625396346067e-05, "loss": 1.0237, "step": 102505 }, { "epoch": 0.26, "learning_rate": 7.446499572197897e-05, "loss": 1.0237, "step": 102510 }, { "epoch": 0.26, "learning_rate": 7.446373748049727e-05, "loss": 1.024, "step": 102515 }, { "epoch": 0.26, "learning_rate": 7.446247923901556e-05, "loss": 1.0222, "step": 102520 }, { "epoch": 0.26, "learning_rate": 7.446122099753385e-05, "loss": 1.022, "step": 102525 }, { "epoch": 0.26, "learning_rate": 7.445996275605215e-05, "loss": 1.0231, "step": 102530 }, { "epoch": 0.26, "learning_rate": 7.445870451457045e-05, "loss": 1.0245, "step": 102535 }, { "epoch": 0.26, "learning_rate": 7.445744627308874e-05, "loss": 1.0227, "step": 102540 }, { "epoch": 0.26, "learning_rate": 7.445618803160703e-05, "loss": 1.0207, "step": 102545 }, { "epoch": 0.26, "learning_rate": 7.445492979012533e-05, "loss": 1.0241, "step": 102550 }, { "epoch": 0.26, "learning_rate": 7.445367154864363e-05, "loss": 1.0246, "step": 102555 }, { "epoch": 0.26, "learning_rate": 7.445241330716192e-05, "loss": 1.0222, "step": 102560 }, { "epoch": 0.26, "learning_rate": 7.445115506568021e-05, "loss": 1.0247, "step": 102565 }, { "epoch": 0.26, "learning_rate": 7.44498968241985e-05, "loss": 1.0225, "step": 102570 }, { "epoch": 0.26, "learning_rate": 7.44486385827168e-05, "loss": 1.0231, "step": 102575 }, { "epoch": 0.26, "learning_rate": 7.44473803412351e-05, "loss": 1.0228, "step": 102580 }, { "epoch": 0.26, "learning_rate": 7.444612209975339e-05, "loss": 1.0241, "step": 102585 }, { "epoch": 0.26, "learning_rate": 7.444486385827168e-05, "loss": 1.0254, "step": 102590 }, { "epoch": 0.26, "learning_rate": 7.444360561678999e-05, "loss": 1.0256, "step": 102595 }, { "epoch": 0.26, "learning_rate": 7.444234737530828e-05, "loss": 1.0199, "step": 102600 }, { "epoch": 0.26, "learning_rate": 7.444108913382657e-05, "loss": 1.0234, "step": 102605 }, { "epoch": 0.26, "learning_rate": 7.443983089234486e-05, "loss": 1.0224, "step": 102610 }, { "epoch": 0.26, "learning_rate": 7.443857265086317e-05, "loss": 1.0233, "step": 102615 }, { "epoch": 0.26, "learning_rate": 7.443731440938146e-05, "loss": 1.024, "step": 102620 }, { "epoch": 0.26, "learning_rate": 7.443605616789975e-05, "loss": 1.0221, "step": 102625 }, { "epoch": 0.26, "learning_rate": 7.443479792641804e-05, "loss": 1.0223, "step": 102630 }, { "epoch": 0.26, "learning_rate": 7.443353968493633e-05, "loss": 1.0237, "step": 102635 }, { "epoch": 0.26, "learning_rate": 7.443228144345464e-05, "loss": 1.0213, "step": 102640 }, { "epoch": 0.26, "learning_rate": 7.443102320197293e-05, "loss": 1.0248, "step": 102645 }, { "epoch": 0.26, "learning_rate": 7.442976496049122e-05, "loss": 1.0221, "step": 102650 }, { "epoch": 0.26, "learning_rate": 7.442850671900951e-05, "loss": 1.0248, "step": 102655 }, { "epoch": 0.26, "learning_rate": 7.442724847752782e-05, "loss": 1.023, "step": 102660 }, { "epoch": 0.26, "learning_rate": 7.442599023604611e-05, "loss": 1.0237, "step": 102665 }, { "epoch": 0.26, "learning_rate": 7.44247319945644e-05, "loss": 1.0245, "step": 102670 }, { "epoch": 0.26, "learning_rate": 7.442347375308269e-05, "loss": 1.0234, "step": 102675 }, { "epoch": 0.26, "learning_rate": 7.4422215511601e-05, "loss": 1.0247, "step": 102680 }, { "epoch": 0.26, "learning_rate": 7.442095727011929e-05, "loss": 1.0249, "step": 102685 }, { "epoch": 0.26, "learning_rate": 7.441969902863758e-05, "loss": 1.0229, "step": 102690 }, { "epoch": 0.26, "learning_rate": 7.441844078715587e-05, "loss": 1.0215, "step": 102695 }, { "epoch": 0.26, "learning_rate": 7.441718254567416e-05, "loss": 1.0224, "step": 102700 }, { "epoch": 0.26, "learning_rate": 7.441592430419247e-05, "loss": 1.0238, "step": 102705 }, { "epoch": 0.26, "learning_rate": 7.441466606271076e-05, "loss": 1.0251, "step": 102710 }, { "epoch": 0.26, "learning_rate": 7.441340782122905e-05, "loss": 1.0235, "step": 102715 }, { "epoch": 0.26, "learning_rate": 7.441214957974734e-05, "loss": 1.0236, "step": 102720 }, { "epoch": 0.26, "learning_rate": 7.441089133826565e-05, "loss": 1.0213, "step": 102725 }, { "epoch": 0.26, "learning_rate": 7.440963309678394e-05, "loss": 1.0246, "step": 102730 }, { "epoch": 0.26, "learning_rate": 7.440837485530223e-05, "loss": 1.0227, "step": 102735 }, { "epoch": 0.26, "learning_rate": 7.440711661382052e-05, "loss": 1.0208, "step": 102740 }, { "epoch": 0.26, "learning_rate": 7.440585837233883e-05, "loss": 1.021, "step": 102745 }, { "epoch": 0.26, "learning_rate": 7.440460013085712e-05, "loss": 1.0242, "step": 102750 }, { "epoch": 0.26, "learning_rate": 7.440334188937541e-05, "loss": 1.0259, "step": 102755 }, { "epoch": 0.26, "learning_rate": 7.44020836478937e-05, "loss": 1.0215, "step": 102760 }, { "epoch": 0.26, "learning_rate": 7.4400825406412e-05, "loss": 1.0238, "step": 102765 }, { "epoch": 0.26, "learning_rate": 7.43995671649303e-05, "loss": 1.0219, "step": 102770 }, { "epoch": 0.26, "learning_rate": 7.439830892344859e-05, "loss": 1.021, "step": 102775 }, { "epoch": 0.26, "learning_rate": 7.439705068196688e-05, "loss": 1.0217, "step": 102780 }, { "epoch": 0.26, "learning_rate": 7.439579244048517e-05, "loss": 1.0224, "step": 102785 }, { "epoch": 0.26, "learning_rate": 7.439453419900348e-05, "loss": 1.0252, "step": 102790 }, { "epoch": 0.26, "learning_rate": 7.439327595752177e-05, "loss": 1.0232, "step": 102795 }, { "epoch": 0.26, "learning_rate": 7.439201771604006e-05, "loss": 1.0232, "step": 102800 }, { "epoch": 0.26, "learning_rate": 7.439075947455835e-05, "loss": 1.0209, "step": 102805 }, { "epoch": 0.26, "learning_rate": 7.438950123307666e-05, "loss": 1.0227, "step": 102810 }, { "epoch": 0.26, "learning_rate": 7.438824299159495e-05, "loss": 1.0215, "step": 102815 }, { "epoch": 0.26, "learning_rate": 7.438698475011324e-05, "loss": 1.0246, "step": 102820 }, { "epoch": 0.26, "learning_rate": 7.438572650863153e-05, "loss": 1.0247, "step": 102825 }, { "epoch": 0.26, "learning_rate": 7.438446826714983e-05, "loss": 1.0251, "step": 102830 }, { "epoch": 0.26, "learning_rate": 7.438321002566813e-05, "loss": 1.0226, "step": 102835 }, { "epoch": 0.26, "learning_rate": 7.438195178418642e-05, "loss": 1.0204, "step": 102840 }, { "epoch": 0.26, "learning_rate": 7.438069354270471e-05, "loss": 1.0224, "step": 102845 }, { "epoch": 0.26, "learning_rate": 7.4379435301223e-05, "loss": 1.0215, "step": 102850 }, { "epoch": 0.26, "learning_rate": 7.437817705974131e-05, "loss": 1.0229, "step": 102855 }, { "epoch": 0.26, "learning_rate": 7.43769188182596e-05, "loss": 1.0213, "step": 102860 }, { "epoch": 0.26, "learning_rate": 7.43756605767779e-05, "loss": 1.023, "step": 102865 }, { "epoch": 0.26, "learning_rate": 7.437440233529619e-05, "loss": 1.0254, "step": 102870 }, { "epoch": 0.26, "learning_rate": 7.437314409381449e-05, "loss": 1.022, "step": 102875 }, { "epoch": 0.26, "learning_rate": 7.437188585233278e-05, "loss": 1.024, "step": 102880 }, { "epoch": 0.26, "learning_rate": 7.437062761085107e-05, "loss": 1.0223, "step": 102885 }, { "epoch": 0.26, "learning_rate": 7.436936936936937e-05, "loss": 1.0245, "step": 102890 }, { "epoch": 0.26, "learning_rate": 7.436811112788766e-05, "loss": 1.0225, "step": 102895 }, { "epoch": 0.26, "learning_rate": 7.436685288640596e-05, "loss": 1.0421, "step": 102900 }, { "epoch": 0.26, "learning_rate": 7.436559464492425e-05, "loss": 1.0234, "step": 102905 }, { "epoch": 0.26, "learning_rate": 7.436433640344255e-05, "loss": 1.0228, "step": 102910 }, { "epoch": 0.26, "learning_rate": 7.436307816196084e-05, "loss": 1.0225, "step": 102915 }, { "epoch": 0.26, "learning_rate": 7.436181992047914e-05, "loss": 1.0231, "step": 102920 }, { "epoch": 0.26, "learning_rate": 7.436056167899743e-05, "loss": 1.0219, "step": 102925 }, { "epoch": 0.26, "learning_rate": 7.435930343751573e-05, "loss": 1.022, "step": 102930 }, { "epoch": 0.26, "learning_rate": 7.435804519603403e-05, "loss": 1.0231, "step": 102935 }, { "epoch": 0.26, "learning_rate": 7.435678695455232e-05, "loss": 1.0232, "step": 102940 }, { "epoch": 0.26, "learning_rate": 7.435552871307063e-05, "loss": 1.0253, "step": 102945 }, { "epoch": 0.26, "learning_rate": 7.435427047158892e-05, "loss": 1.0224, "step": 102950 }, { "epoch": 0.26, "learning_rate": 7.435301223010721e-05, "loss": 1.0245, "step": 102955 }, { "epoch": 0.26, "learning_rate": 7.43517539886255e-05, "loss": 1.0234, "step": 102960 }, { "epoch": 0.26, "learning_rate": 7.43504957471438e-05, "loss": 1.0228, "step": 102965 }, { "epoch": 0.26, "learning_rate": 7.43492375056621e-05, "loss": 1.0247, "step": 102970 }, { "epoch": 0.26, "learning_rate": 7.434797926418039e-05, "loss": 1.0213, "step": 102975 }, { "epoch": 0.26, "learning_rate": 7.434672102269868e-05, "loss": 1.0276, "step": 102980 }, { "epoch": 0.26, "learning_rate": 7.434546278121697e-05, "loss": 1.0241, "step": 102985 }, { "epoch": 0.26, "learning_rate": 7.434420453973528e-05, "loss": 1.0225, "step": 102990 }, { "epoch": 0.26, "learning_rate": 7.434294629825357e-05, "loss": 1.0223, "step": 102995 }, { "epoch": 0.26, "learning_rate": 7.434168805677186e-05, "loss": 1.0217, "step": 103000 }, { "epoch": 0.26, "learning_rate": 7.434042981529015e-05, "loss": 1.0234, "step": 103005 }, { "epoch": 0.26, "learning_rate": 7.433917157380846e-05, "loss": 1.0218, "step": 103010 }, { "epoch": 0.26, "learning_rate": 7.433791333232675e-05, "loss": 1.0252, "step": 103015 }, { "epoch": 0.26, "learning_rate": 7.433665509084504e-05, "loss": 1.0235, "step": 103020 }, { "epoch": 0.26, "learning_rate": 7.433539684936333e-05, "loss": 1.0231, "step": 103025 }, { "epoch": 0.26, "learning_rate": 7.433413860788163e-05, "loss": 1.024, "step": 103030 }, { "epoch": 0.26, "learning_rate": 7.433288036639993e-05, "loss": 1.0225, "step": 103035 }, { "epoch": 0.26, "learning_rate": 7.433162212491822e-05, "loss": 1.0225, "step": 103040 }, { "epoch": 0.26, "learning_rate": 7.433036388343651e-05, "loss": 1.0244, "step": 103045 }, { "epoch": 0.26, "learning_rate": 7.43291056419548e-05, "loss": 1.0234, "step": 103050 }, { "epoch": 0.26, "learning_rate": 7.432784740047311e-05, "loss": 1.0264, "step": 103055 }, { "epoch": 0.26, "learning_rate": 7.43265891589914e-05, "loss": 1.0218, "step": 103060 }, { "epoch": 0.26, "learning_rate": 7.43253309175097e-05, "loss": 1.0248, "step": 103065 }, { "epoch": 0.26, "learning_rate": 7.432407267602799e-05, "loss": 1.0208, "step": 103070 }, { "epoch": 0.26, "learning_rate": 7.432281443454629e-05, "loss": 1.0221, "step": 103075 }, { "epoch": 0.26, "learning_rate": 7.432155619306458e-05, "loss": 1.0202, "step": 103080 }, { "epoch": 0.26, "learning_rate": 7.432029795158287e-05, "loss": 1.0231, "step": 103085 }, { "epoch": 0.26, "learning_rate": 7.431903971010117e-05, "loss": 1.0226, "step": 103090 }, { "epoch": 0.26, "learning_rate": 7.431778146861946e-05, "loss": 1.0245, "step": 103095 }, { "epoch": 0.26, "learning_rate": 7.431652322713776e-05, "loss": 1.0239, "step": 103100 }, { "epoch": 0.26, "learning_rate": 7.431526498565605e-05, "loss": 1.0212, "step": 103105 }, { "epoch": 0.26, "learning_rate": 7.431400674417435e-05, "loss": 1.0232, "step": 103110 }, { "epoch": 0.26, "learning_rate": 7.431274850269264e-05, "loss": 1.0228, "step": 103115 }, { "epoch": 0.26, "learning_rate": 7.431149026121094e-05, "loss": 1.0212, "step": 103120 }, { "epoch": 0.26, "learning_rate": 7.431023201972923e-05, "loss": 1.0223, "step": 103125 }, { "epoch": 0.26, "learning_rate": 7.430897377824752e-05, "loss": 1.0236, "step": 103130 }, { "epoch": 0.26, "learning_rate": 7.430771553676582e-05, "loss": 1.0226, "step": 103135 }, { "epoch": 0.26, "learning_rate": 7.430645729528412e-05, "loss": 1.0225, "step": 103140 }, { "epoch": 0.26, "learning_rate": 7.430519905380241e-05, "loss": 1.0231, "step": 103145 }, { "epoch": 0.26, "learning_rate": 7.43039408123207e-05, "loss": 1.023, "step": 103150 }, { "epoch": 0.26, "learning_rate": 7.4302682570839e-05, "loss": 1.0253, "step": 103155 }, { "epoch": 0.26, "learning_rate": 7.430142432935729e-05, "loss": 1.0214, "step": 103160 }, { "epoch": 0.26, "learning_rate": 7.43001660878756e-05, "loss": 1.0217, "step": 103165 }, { "epoch": 0.26, "learning_rate": 7.429890784639388e-05, "loss": 1.0212, "step": 103170 }, { "epoch": 0.26, "learning_rate": 7.429764960491218e-05, "loss": 1.0212, "step": 103175 }, { "epoch": 0.26, "learning_rate": 7.429639136343047e-05, "loss": 1.0213, "step": 103180 }, { "epoch": 0.26, "learning_rate": 7.429513312194877e-05, "loss": 1.0221, "step": 103185 }, { "epoch": 0.26, "learning_rate": 7.429387488046706e-05, "loss": 1.0237, "step": 103190 }, { "epoch": 0.26, "learning_rate": 7.429261663898536e-05, "loss": 1.0221, "step": 103195 }, { "epoch": 0.26, "learning_rate": 7.429135839750365e-05, "loss": 1.023, "step": 103200 }, { "epoch": 0.26, "learning_rate": 7.429010015602195e-05, "loss": 1.022, "step": 103205 }, { "epoch": 0.26, "learning_rate": 7.428884191454024e-05, "loss": 1.0491, "step": 103210 }, { "epoch": 0.26, "learning_rate": 7.428758367305854e-05, "loss": 1.0234, "step": 103215 }, { "epoch": 0.26, "learning_rate": 7.428632543157683e-05, "loss": 1.0223, "step": 103220 }, { "epoch": 0.26, "learning_rate": 7.428506719009512e-05, "loss": 1.0229, "step": 103225 }, { "epoch": 0.26, "learning_rate": 7.428380894861342e-05, "loss": 1.0228, "step": 103230 }, { "epoch": 0.26, "learning_rate": 7.428255070713172e-05, "loss": 1.024, "step": 103235 }, { "epoch": 0.26, "learning_rate": 7.428129246565001e-05, "loss": 1.0223, "step": 103240 }, { "epoch": 0.26, "learning_rate": 7.42800342241683e-05, "loss": 1.0223, "step": 103245 }, { "epoch": 0.26, "learning_rate": 7.42787759826866e-05, "loss": 1.0229, "step": 103250 }, { "epoch": 0.26, "learning_rate": 7.42775177412049e-05, "loss": 1.0232, "step": 103255 }, { "epoch": 0.26, "learning_rate": 7.427625949972319e-05, "loss": 1.0264, "step": 103260 }, { "epoch": 0.26, "learning_rate": 7.427500125824148e-05, "loss": 1.0242, "step": 103265 }, { "epoch": 0.26, "learning_rate": 7.427374301675978e-05, "loss": 1.0232, "step": 103270 }, { "epoch": 0.26, "learning_rate": 7.427248477527808e-05, "loss": 1.0215, "step": 103275 }, { "epoch": 0.26, "learning_rate": 7.427122653379637e-05, "loss": 1.0218, "step": 103280 }, { "epoch": 0.26, "learning_rate": 7.426996829231466e-05, "loss": 1.0216, "step": 103285 }, { "epoch": 0.26, "learning_rate": 7.426871005083295e-05, "loss": 1.0232, "step": 103290 }, { "epoch": 0.26, "learning_rate": 7.426745180935126e-05, "loss": 1.0221, "step": 103295 }, { "epoch": 0.26, "learning_rate": 7.426619356786955e-05, "loss": 1.0228, "step": 103300 }, { "epoch": 0.26, "learning_rate": 7.426493532638784e-05, "loss": 1.0241, "step": 103305 }, { "epoch": 0.26, "learning_rate": 7.426367708490613e-05, "loss": 1.0231, "step": 103310 }, { "epoch": 0.26, "learning_rate": 7.426241884342444e-05, "loss": 1.0203, "step": 103315 }, { "epoch": 0.26, "learning_rate": 7.426116060194273e-05, "loss": 1.0235, "step": 103320 }, { "epoch": 0.26, "learning_rate": 7.425990236046102e-05, "loss": 1.0225, "step": 103325 }, { "epoch": 0.26, "learning_rate": 7.425864411897931e-05, "loss": 1.021, "step": 103330 }, { "epoch": 0.26, "learning_rate": 7.425738587749762e-05, "loss": 1.0213, "step": 103335 }, { "epoch": 0.26, "learning_rate": 7.425612763601591e-05, "loss": 1.0212, "step": 103340 }, { "epoch": 0.26, "learning_rate": 7.42548693945342e-05, "loss": 1.0247, "step": 103345 }, { "epoch": 0.26, "learning_rate": 7.425361115305249e-05, "loss": 1.0209, "step": 103350 }, { "epoch": 0.26, "learning_rate": 7.425235291157078e-05, "loss": 1.0207, "step": 103355 }, { "epoch": 0.26, "learning_rate": 7.425109467008909e-05, "loss": 1.0217, "step": 103360 }, { "epoch": 0.26, "learning_rate": 7.424983642860738e-05, "loss": 1.0219, "step": 103365 }, { "epoch": 0.26, "learning_rate": 7.424857818712567e-05, "loss": 1.0217, "step": 103370 }, { "epoch": 0.26, "learning_rate": 7.424731994564396e-05, "loss": 1.0236, "step": 103375 }, { "epoch": 0.26, "learning_rate": 7.424606170416227e-05, "loss": 1.0229, "step": 103380 }, { "epoch": 0.26, "learning_rate": 7.424480346268056e-05, "loss": 1.0222, "step": 103385 }, { "epoch": 0.26, "learning_rate": 7.424354522119885e-05, "loss": 1.0226, "step": 103390 }, { "epoch": 0.26, "learning_rate": 7.424228697971714e-05, "loss": 1.0246, "step": 103395 }, { "epoch": 0.26, "learning_rate": 7.424102873823545e-05, "loss": 1.0215, "step": 103400 }, { "epoch": 0.26, "learning_rate": 7.423977049675374e-05, "loss": 1.0218, "step": 103405 }, { "epoch": 0.26, "learning_rate": 7.423851225527203e-05, "loss": 1.0213, "step": 103410 }, { "epoch": 0.26, "learning_rate": 7.423725401379032e-05, "loss": 1.0251, "step": 103415 }, { "epoch": 0.26, "learning_rate": 7.423599577230861e-05, "loss": 1.0223, "step": 103420 }, { "epoch": 0.26, "learning_rate": 7.423473753082692e-05, "loss": 1.0213, "step": 103425 }, { "epoch": 0.26, "learning_rate": 7.423347928934522e-05, "loss": 1.0228, "step": 103430 }, { "epoch": 0.26, "learning_rate": 7.423222104786352e-05, "loss": 1.0229, "step": 103435 }, { "epoch": 0.26, "learning_rate": 7.423096280638181e-05, "loss": 1.0232, "step": 103440 }, { "epoch": 0.26, "learning_rate": 7.42297045649001e-05, "loss": 1.0247, "step": 103445 }, { "epoch": 0.26, "learning_rate": 7.42284463234184e-05, "loss": 1.0245, "step": 103450 }, { "epoch": 0.26, "learning_rate": 7.42271880819367e-05, "loss": 1.0231, "step": 103455 }, { "epoch": 0.26, "learning_rate": 7.422592984045499e-05, "loss": 1.0236, "step": 103460 }, { "epoch": 0.26, "learning_rate": 7.422467159897328e-05, "loss": 1.0242, "step": 103465 }, { "epoch": 0.26, "learning_rate": 7.422341335749158e-05, "loss": 1.0203, "step": 103470 }, { "epoch": 0.26, "learning_rate": 7.422215511600988e-05, "loss": 1.0262, "step": 103475 }, { "epoch": 0.26, "learning_rate": 7.422089687452817e-05, "loss": 1.0227, "step": 103480 }, { "epoch": 0.26, "learning_rate": 7.421963863304646e-05, "loss": 1.0215, "step": 103485 }, { "epoch": 0.26, "learning_rate": 7.421838039156475e-05, "loss": 1.0233, "step": 103490 }, { "epoch": 0.26, "learning_rate": 7.421712215008306e-05, "loss": 1.0239, "step": 103495 }, { "epoch": 0.26, "learning_rate": 7.421586390860135e-05, "loss": 1.0246, "step": 103500 }, { "epoch": 0.26, "learning_rate": 7.421460566711964e-05, "loss": 1.022, "step": 103505 }, { "epoch": 0.26, "learning_rate": 7.421334742563793e-05, "loss": 1.0212, "step": 103510 }, { "epoch": 0.26, "learning_rate": 7.421208918415624e-05, "loss": 1.0206, "step": 103515 }, { "epoch": 0.26, "learning_rate": 7.421083094267453e-05, "loss": 1.0252, "step": 103520 }, { "epoch": 0.26, "learning_rate": 7.420957270119282e-05, "loss": 1.0239, "step": 103525 }, { "epoch": 0.26, "learning_rate": 7.420831445971111e-05, "loss": 1.0245, "step": 103530 }, { "epoch": 0.26, "learning_rate": 7.420705621822942e-05, "loss": 1.025, "step": 103535 }, { "epoch": 0.26, "learning_rate": 7.420579797674771e-05, "loss": 1.0483, "step": 103540 }, { "epoch": 0.26, "learning_rate": 7.4204539735266e-05, "loss": 1.024, "step": 103545 }, { "epoch": 0.26, "learning_rate": 7.420328149378429e-05, "loss": 1.0248, "step": 103550 }, { "epoch": 0.26, "learning_rate": 7.420202325230258e-05, "loss": 1.0217, "step": 103555 }, { "epoch": 0.26, "learning_rate": 7.420076501082089e-05, "loss": 1.0227, "step": 103560 }, { "epoch": 0.26, "learning_rate": 7.419950676933918e-05, "loss": 1.0244, "step": 103565 }, { "epoch": 0.26, "learning_rate": 7.419824852785747e-05, "loss": 1.0237, "step": 103570 }, { "epoch": 0.26, "learning_rate": 7.419699028637576e-05, "loss": 1.0225, "step": 103575 }, { "epoch": 0.26, "learning_rate": 7.419573204489407e-05, "loss": 1.0225, "step": 103580 }, { "epoch": 0.26, "learning_rate": 7.419447380341236e-05, "loss": 1.0215, "step": 103585 }, { "epoch": 0.26, "learning_rate": 7.419321556193065e-05, "loss": 1.0259, "step": 103590 }, { "epoch": 0.26, "learning_rate": 7.419195732044894e-05, "loss": 1.0225, "step": 103595 }, { "epoch": 0.26, "learning_rate": 7.419069907896725e-05, "loss": 1.0236, "step": 103600 }, { "epoch": 0.26, "learning_rate": 7.418944083748554e-05, "loss": 1.0239, "step": 103605 }, { "epoch": 0.26, "learning_rate": 7.418818259600383e-05, "loss": 1.0206, "step": 103610 }, { "epoch": 0.26, "learning_rate": 7.418692435452212e-05, "loss": 1.0225, "step": 103615 }, { "epoch": 0.26, "learning_rate": 7.418566611304041e-05, "loss": 1.0226, "step": 103620 }, { "epoch": 0.26, "learning_rate": 7.418440787155872e-05, "loss": 1.0213, "step": 103625 }, { "epoch": 0.26, "learning_rate": 7.418314963007701e-05, "loss": 1.023, "step": 103630 }, { "epoch": 0.26, "learning_rate": 7.41818913885953e-05, "loss": 1.0235, "step": 103635 }, { "epoch": 0.26, "learning_rate": 7.418063314711359e-05, "loss": 1.039, "step": 103640 }, { "epoch": 0.26, "learning_rate": 7.41793749056319e-05, "loss": 1.0244, "step": 103645 }, { "epoch": 0.26, "learning_rate": 7.417811666415019e-05, "loss": 1.0223, "step": 103650 }, { "epoch": 0.26, "learning_rate": 7.417685842266848e-05, "loss": 1.0226, "step": 103655 }, { "epoch": 0.26, "learning_rate": 7.417560018118677e-05, "loss": 1.0206, "step": 103660 }, { "epoch": 0.26, "learning_rate": 7.417434193970508e-05, "loss": 1.0263, "step": 103665 }, { "epoch": 0.26, "learning_rate": 7.417308369822337e-05, "loss": 1.0245, "step": 103670 }, { "epoch": 0.26, "learning_rate": 7.417182545674166e-05, "loss": 1.0249, "step": 103675 }, { "epoch": 0.26, "learning_rate": 7.417056721525995e-05, "loss": 1.0223, "step": 103680 }, { "epoch": 0.26, "learning_rate": 7.416930897377824e-05, "loss": 1.022, "step": 103685 }, { "epoch": 0.26, "learning_rate": 7.416805073229655e-05, "loss": 1.0217, "step": 103690 }, { "epoch": 0.26, "learning_rate": 7.416679249081484e-05, "loss": 1.0236, "step": 103695 }, { "epoch": 0.26, "learning_rate": 7.416553424933313e-05, "loss": 1.0235, "step": 103700 }, { "epoch": 0.26, "learning_rate": 7.416427600785142e-05, "loss": 1.0236, "step": 103705 }, { "epoch": 0.26, "learning_rate": 7.416301776636973e-05, "loss": 1.0381, "step": 103710 }, { "epoch": 0.26, "learning_rate": 7.416175952488802e-05, "loss": 1.0249, "step": 103715 }, { "epoch": 0.26, "learning_rate": 7.416050128340631e-05, "loss": 1.0244, "step": 103720 }, { "epoch": 0.26, "learning_rate": 7.41592430419246e-05, "loss": 1.0236, "step": 103725 }, { "epoch": 0.26, "learning_rate": 7.415798480044291e-05, "loss": 1.0207, "step": 103730 }, { "epoch": 0.26, "learning_rate": 7.41567265589612e-05, "loss": 1.0241, "step": 103735 }, { "epoch": 0.26, "learning_rate": 7.415546831747949e-05, "loss": 1.0486, "step": 103740 }, { "epoch": 0.26, "learning_rate": 7.415421007599778e-05, "loss": 1.0234, "step": 103745 }, { "epoch": 0.26, "learning_rate": 7.415295183451608e-05, "loss": 1.0219, "step": 103750 }, { "epoch": 0.26, "learning_rate": 7.415169359303438e-05, "loss": 1.021, "step": 103755 }, { "epoch": 0.26, "learning_rate": 7.415043535155267e-05, "loss": 1.0225, "step": 103760 }, { "epoch": 0.26, "learning_rate": 7.414917711007096e-05, "loss": 1.0238, "step": 103765 }, { "epoch": 0.26, "learning_rate": 7.414791886858926e-05, "loss": 1.0247, "step": 103770 }, { "epoch": 0.26, "learning_rate": 7.414666062710756e-05, "loss": 1.0217, "step": 103775 }, { "epoch": 0.26, "learning_rate": 7.414540238562585e-05, "loss": 1.0213, "step": 103780 }, { "epoch": 0.26, "learning_rate": 7.414414414414414e-05, "loss": 1.0238, "step": 103785 }, { "epoch": 0.26, "learning_rate": 7.414288590266244e-05, "loss": 1.0234, "step": 103790 }, { "epoch": 0.26, "learning_rate": 7.414162766118074e-05, "loss": 1.0205, "step": 103795 }, { "epoch": 0.26, "learning_rate": 7.414036941969903e-05, "loss": 1.0226, "step": 103800 }, { "epoch": 0.26, "learning_rate": 7.413911117821732e-05, "loss": 1.0246, "step": 103805 }, { "epoch": 0.26, "learning_rate": 7.413785293673562e-05, "loss": 1.022, "step": 103810 }, { "epoch": 0.26, "learning_rate": 7.413659469525391e-05, "loss": 1.0228, "step": 103815 }, { "epoch": 0.26, "learning_rate": 7.413533645377221e-05, "loss": 1.0234, "step": 103820 }, { "epoch": 0.26, "learning_rate": 7.41340782122905e-05, "loss": 1.0205, "step": 103825 }, { "epoch": 0.26, "learning_rate": 7.41328199708088e-05, "loss": 1.0216, "step": 103830 }, { "epoch": 0.26, "learning_rate": 7.413156172932709e-05, "loss": 1.0238, "step": 103835 }, { "epoch": 0.26, "learning_rate": 7.413030348784539e-05, "loss": 1.0242, "step": 103840 }, { "epoch": 0.26, "learning_rate": 7.412904524636368e-05, "loss": 1.0222, "step": 103845 }, { "epoch": 0.26, "learning_rate": 7.412778700488198e-05, "loss": 1.0225, "step": 103850 }, { "epoch": 0.26, "learning_rate": 7.412652876340027e-05, "loss": 1.0235, "step": 103855 }, { "epoch": 0.26, "learning_rate": 7.412527052191857e-05, "loss": 1.0239, "step": 103860 }, { "epoch": 0.26, "learning_rate": 7.412401228043686e-05, "loss": 1.0221, "step": 103865 }, { "epoch": 0.26, "learning_rate": 7.412275403895516e-05, "loss": 1.0256, "step": 103870 }, { "epoch": 0.26, "learning_rate": 7.412149579747345e-05, "loss": 1.0217, "step": 103875 }, { "epoch": 0.26, "learning_rate": 7.412023755599174e-05, "loss": 1.0207, "step": 103880 }, { "epoch": 0.26, "learning_rate": 7.411897931451004e-05, "loss": 1.0223, "step": 103885 }, { "epoch": 0.26, "learning_rate": 7.411772107302834e-05, "loss": 1.0226, "step": 103890 }, { "epoch": 0.26, "learning_rate": 7.411646283154663e-05, "loss": 1.0244, "step": 103895 }, { "epoch": 0.26, "learning_rate": 7.411520459006492e-05, "loss": 1.0208, "step": 103900 }, { "epoch": 0.26, "learning_rate": 7.411394634858322e-05, "loss": 1.024, "step": 103905 }, { "epoch": 0.26, "learning_rate": 7.411268810710152e-05, "loss": 1.0219, "step": 103910 }, { "epoch": 0.26, "learning_rate": 7.411142986561981e-05, "loss": 1.0209, "step": 103915 }, { "epoch": 0.26, "learning_rate": 7.41101716241381e-05, "loss": 1.0219, "step": 103920 }, { "epoch": 0.26, "learning_rate": 7.41089133826564e-05, "loss": 1.0229, "step": 103925 }, { "epoch": 0.26, "learning_rate": 7.410765514117471e-05, "loss": 1.0244, "step": 103930 }, { "epoch": 0.26, "learning_rate": 7.4106396899693e-05, "loss": 1.023, "step": 103935 }, { "epoch": 0.26, "learning_rate": 7.410513865821129e-05, "loss": 1.0256, "step": 103940 }, { "epoch": 0.26, "learning_rate": 7.410388041672958e-05, "loss": 1.0235, "step": 103945 }, { "epoch": 0.26, "learning_rate": 7.410262217524788e-05, "loss": 1.0228, "step": 103950 }, { "epoch": 0.26, "learning_rate": 7.410136393376618e-05, "loss": 1.0253, "step": 103955 }, { "epoch": 0.26, "learning_rate": 7.410010569228447e-05, "loss": 1.0216, "step": 103960 }, { "epoch": 0.26, "learning_rate": 7.409884745080276e-05, "loss": 1.0218, "step": 103965 }, { "epoch": 0.26, "learning_rate": 7.409758920932106e-05, "loss": 1.0231, "step": 103970 }, { "epoch": 0.26, "learning_rate": 7.409633096783936e-05, "loss": 1.0222, "step": 103975 }, { "epoch": 0.26, "learning_rate": 7.409507272635765e-05, "loss": 1.0223, "step": 103980 }, { "epoch": 0.26, "learning_rate": 7.409381448487594e-05, "loss": 1.0228, "step": 103985 }, { "epoch": 0.26, "learning_rate": 7.409255624339424e-05, "loss": 1.0223, "step": 103990 }, { "epoch": 0.26, "learning_rate": 7.409129800191254e-05, "loss": 1.0246, "step": 103995 }, { "epoch": 0.26, "learning_rate": 7.409003976043083e-05, "loss": 1.0242, "step": 104000 }, { "epoch": 0.26, "learning_rate": 7.408878151894912e-05, "loss": 1.023, "step": 104005 }, { "epoch": 0.26, "learning_rate": 7.408752327746742e-05, "loss": 1.0453, "step": 104010 }, { "epoch": 0.26, "learning_rate": 7.408626503598571e-05, "loss": 1.0214, "step": 104015 }, { "epoch": 0.26, "learning_rate": 7.408500679450401e-05, "loss": 1.0232, "step": 104020 }, { "epoch": 0.26, "learning_rate": 7.40837485530223e-05, "loss": 1.0232, "step": 104025 }, { "epoch": 0.26, "learning_rate": 7.40824903115406e-05, "loss": 1.0222, "step": 104030 }, { "epoch": 0.26, "learning_rate": 7.408123207005889e-05, "loss": 1.0233, "step": 104035 }, { "epoch": 0.26, "learning_rate": 7.407997382857719e-05, "loss": 1.0216, "step": 104040 }, { "epoch": 0.26, "learning_rate": 7.407871558709548e-05, "loss": 1.0227, "step": 104045 }, { "epoch": 0.26, "learning_rate": 7.407745734561378e-05, "loss": 1.0218, "step": 104050 }, { "epoch": 0.26, "learning_rate": 7.407619910413207e-05, "loss": 1.0216, "step": 104055 }, { "epoch": 0.26, "learning_rate": 7.407494086265037e-05, "loss": 1.0251, "step": 104060 }, { "epoch": 0.26, "learning_rate": 7.407368262116866e-05, "loss": 1.0218, "step": 104065 }, { "epoch": 0.26, "learning_rate": 7.407242437968696e-05, "loss": 1.0444, "step": 104070 }, { "epoch": 0.26, "learning_rate": 7.407116613820525e-05, "loss": 1.0199, "step": 104075 }, { "epoch": 0.26, "learning_rate": 7.406990789672354e-05, "loss": 1.0227, "step": 104080 }, { "epoch": 0.26, "learning_rate": 7.406864965524184e-05, "loss": 1.0229, "step": 104085 }, { "epoch": 0.26, "learning_rate": 7.406739141376014e-05, "loss": 1.0267, "step": 104090 }, { "epoch": 0.26, "learning_rate": 7.406613317227843e-05, "loss": 1.0223, "step": 104095 }, { "epoch": 0.26, "learning_rate": 7.406487493079672e-05, "loss": 1.0238, "step": 104100 }, { "epoch": 0.26, "learning_rate": 7.406361668931502e-05, "loss": 1.023, "step": 104105 }, { "epoch": 0.26, "learning_rate": 7.406235844783332e-05, "loss": 1.0228, "step": 104110 }, { "epoch": 0.26, "learning_rate": 7.40611002063516e-05, "loss": 1.0238, "step": 104115 }, { "epoch": 0.26, "learning_rate": 7.40598419648699e-05, "loss": 1.0225, "step": 104120 }, { "epoch": 0.26, "learning_rate": 7.40585837233882e-05, "loss": 1.0246, "step": 104125 }, { "epoch": 0.26, "learning_rate": 7.40573254819065e-05, "loss": 1.0212, "step": 104130 }, { "epoch": 0.26, "learning_rate": 7.405606724042479e-05, "loss": 1.0203, "step": 104135 }, { "epoch": 0.26, "learning_rate": 7.405480899894308e-05, "loss": 1.0203, "step": 104140 }, { "epoch": 0.26, "learning_rate": 7.405355075746137e-05, "loss": 1.0241, "step": 104145 }, { "epoch": 0.26, "learning_rate": 7.405229251597967e-05, "loss": 1.0225, "step": 104150 }, { "epoch": 0.26, "learning_rate": 7.405103427449797e-05, "loss": 1.0242, "step": 104155 }, { "epoch": 0.26, "learning_rate": 7.404977603301626e-05, "loss": 1.0222, "step": 104160 }, { "epoch": 0.26, "learning_rate": 7.404851779153455e-05, "loss": 1.0227, "step": 104165 }, { "epoch": 0.26, "learning_rate": 7.404725955005285e-05, "loss": 1.0229, "step": 104170 }, { "epoch": 0.26, "learning_rate": 7.404600130857115e-05, "loss": 1.0236, "step": 104175 }, { "epoch": 0.26, "learning_rate": 7.404474306708944e-05, "loss": 1.0221, "step": 104180 }, { "epoch": 0.26, "learning_rate": 7.404348482560773e-05, "loss": 1.0218, "step": 104185 }, { "epoch": 0.26, "learning_rate": 7.404222658412603e-05, "loss": 1.0218, "step": 104190 }, { "epoch": 0.26, "learning_rate": 7.404096834264433e-05, "loss": 1.0216, "step": 104195 }, { "epoch": 0.26, "learning_rate": 7.403971010116262e-05, "loss": 1.0357, "step": 104200 }, { "epoch": 0.26, "learning_rate": 7.403845185968091e-05, "loss": 1.0244, "step": 104205 }, { "epoch": 0.26, "learning_rate": 7.40371936181992e-05, "loss": 1.0242, "step": 104210 }, { "epoch": 0.26, "learning_rate": 7.40359353767175e-05, "loss": 1.0244, "step": 104215 }, { "epoch": 0.26, "learning_rate": 7.40346771352358e-05, "loss": 1.0225, "step": 104220 }, { "epoch": 0.26, "learning_rate": 7.403341889375409e-05, "loss": 1.0248, "step": 104225 }, { "epoch": 0.26, "learning_rate": 7.403216065227238e-05, "loss": 1.0207, "step": 104230 }, { "epoch": 0.26, "learning_rate": 7.403090241079069e-05, "loss": 1.0242, "step": 104235 }, { "epoch": 0.26, "learning_rate": 7.402964416930898e-05, "loss": 1.022, "step": 104240 }, { "epoch": 0.26, "learning_rate": 7.402838592782727e-05, "loss": 1.024, "step": 104245 }, { "epoch": 0.26, "learning_rate": 7.402712768634556e-05, "loss": 1.0235, "step": 104250 }, { "epoch": 0.26, "learning_rate": 7.402586944486387e-05, "loss": 1.0207, "step": 104255 }, { "epoch": 0.26, "learning_rate": 7.402461120338216e-05, "loss": 1.0245, "step": 104260 }, { "epoch": 0.26, "learning_rate": 7.402335296190045e-05, "loss": 1.0234, "step": 104265 }, { "epoch": 0.26, "learning_rate": 7.402209472041874e-05, "loss": 1.0255, "step": 104270 }, { "epoch": 0.26, "learning_rate": 7.402083647893703e-05, "loss": 1.0224, "step": 104275 }, { "epoch": 0.26, "learning_rate": 7.401957823745534e-05, "loss": 1.0246, "step": 104280 }, { "epoch": 0.26, "learning_rate": 7.401831999597363e-05, "loss": 1.022, "step": 104285 }, { "epoch": 0.26, "learning_rate": 7.401706175449192e-05, "loss": 1.0251, "step": 104290 }, { "epoch": 0.26, "learning_rate": 7.401580351301021e-05, "loss": 1.0213, "step": 104295 }, { "epoch": 0.26, "learning_rate": 7.401454527152852e-05, "loss": 1.0214, "step": 104300 }, { "epoch": 0.26, "learning_rate": 7.401328703004681e-05, "loss": 1.0221, "step": 104305 }, { "epoch": 0.26, "learning_rate": 7.40120287885651e-05, "loss": 1.023, "step": 104310 }, { "epoch": 0.26, "learning_rate": 7.401077054708339e-05, "loss": 1.0249, "step": 104315 }, { "epoch": 0.26, "learning_rate": 7.40095123056017e-05, "loss": 1.0247, "step": 104320 }, { "epoch": 0.26, "learning_rate": 7.400825406411999e-05, "loss": 1.0238, "step": 104325 }, { "epoch": 0.26, "learning_rate": 7.400699582263828e-05, "loss": 1.0258, "step": 104330 }, { "epoch": 0.26, "learning_rate": 7.400573758115657e-05, "loss": 1.0244, "step": 104335 }, { "epoch": 0.26, "learning_rate": 7.400447933967486e-05, "loss": 1.0226, "step": 104340 }, { "epoch": 0.26, "learning_rate": 7.400322109819317e-05, "loss": 1.0191, "step": 104345 }, { "epoch": 0.26, "learning_rate": 7.400196285671146e-05, "loss": 1.024, "step": 104350 }, { "epoch": 0.26, "learning_rate": 7.400070461522975e-05, "loss": 1.0236, "step": 104355 }, { "epoch": 0.26, "learning_rate": 7.399944637374804e-05, "loss": 1.0235, "step": 104360 }, { "epoch": 0.26, "learning_rate": 7.399818813226635e-05, "loss": 1.0229, "step": 104365 }, { "epoch": 0.26, "learning_rate": 7.399692989078464e-05, "loss": 1.0234, "step": 104370 }, { "epoch": 0.26, "learning_rate": 7.399567164930293e-05, "loss": 1.0215, "step": 104375 }, { "epoch": 0.26, "learning_rate": 7.399441340782122e-05, "loss": 1.023, "step": 104380 }, { "epoch": 0.26, "learning_rate": 7.399315516633953e-05, "loss": 1.0222, "step": 104385 }, { "epoch": 0.26, "learning_rate": 7.399189692485782e-05, "loss": 1.0238, "step": 104390 }, { "epoch": 0.26, "learning_rate": 7.399063868337611e-05, "loss": 1.0227, "step": 104395 }, { "epoch": 0.26, "learning_rate": 7.39893804418944e-05, "loss": 1.0223, "step": 104400 }, { "epoch": 0.26, "learning_rate": 7.39881222004127e-05, "loss": 1.0208, "step": 104405 }, { "epoch": 0.26, "learning_rate": 7.3986863958931e-05, "loss": 1.0229, "step": 104410 }, { "epoch": 0.26, "learning_rate": 7.398560571744929e-05, "loss": 1.0225, "step": 104415 }, { "epoch": 0.26, "learning_rate": 7.398434747596758e-05, "loss": 1.0446, "step": 104420 }, { "epoch": 0.26, "learning_rate": 7.398308923448588e-05, "loss": 1.0457, "step": 104425 }, { "epoch": 0.26, "learning_rate": 7.398183099300418e-05, "loss": 1.0237, "step": 104430 }, { "epoch": 0.26, "learning_rate": 7.398057275152249e-05, "loss": 1.0232, "step": 104435 }, { "epoch": 0.26, "learning_rate": 7.397931451004078e-05, "loss": 1.0229, "step": 104440 }, { "epoch": 0.26, "learning_rate": 7.397805626855907e-05, "loss": 1.0222, "step": 104445 }, { "epoch": 0.26, "learning_rate": 7.397679802707736e-05, "loss": 1.0243, "step": 104450 }, { "epoch": 0.26, "learning_rate": 7.397553978559567e-05, "loss": 1.0228, "step": 104455 }, { "epoch": 0.26, "learning_rate": 7.397428154411396e-05, "loss": 1.023, "step": 104460 }, { "epoch": 0.26, "learning_rate": 7.397302330263225e-05, "loss": 1.0223, "step": 104465 }, { "epoch": 0.26, "learning_rate": 7.397176506115054e-05, "loss": 1.0229, "step": 104470 }, { "epoch": 0.26, "learning_rate": 7.397050681966885e-05, "loss": 1.023, "step": 104475 }, { "epoch": 0.26, "learning_rate": 7.396924857818714e-05, "loss": 1.024, "step": 104480 }, { "epoch": 0.26, "learning_rate": 7.396799033670543e-05, "loss": 1.0234, "step": 104485 }, { "epoch": 0.26, "learning_rate": 7.396673209522372e-05, "loss": 1.0214, "step": 104490 }, { "epoch": 0.26, "learning_rate": 7.396547385374201e-05, "loss": 1.0257, "step": 104495 }, { "epoch": 0.26, "learning_rate": 7.396421561226032e-05, "loss": 1.0238, "step": 104500 }, { "epoch": 0.26, "learning_rate": 7.396295737077861e-05, "loss": 1.0475, "step": 104505 }, { "epoch": 0.26, "learning_rate": 7.39616991292969e-05, "loss": 1.0219, "step": 104510 }, { "epoch": 0.26, "learning_rate": 7.396044088781519e-05, "loss": 1.0207, "step": 104515 }, { "epoch": 0.26, "learning_rate": 7.39591826463335e-05, "loss": 1.0231, "step": 104520 }, { "epoch": 0.26, "learning_rate": 7.395792440485179e-05, "loss": 1.0223, "step": 104525 }, { "epoch": 0.26, "learning_rate": 7.395666616337008e-05, "loss": 1.0213, "step": 104530 }, { "epoch": 0.26, "learning_rate": 7.395540792188837e-05, "loss": 1.0241, "step": 104535 }, { "epoch": 0.26, "learning_rate": 7.395414968040666e-05, "loss": 1.0218, "step": 104540 }, { "epoch": 0.26, "learning_rate": 7.395289143892497e-05, "loss": 1.0226, "step": 104545 }, { "epoch": 0.26, "learning_rate": 7.395163319744326e-05, "loss": 1.0201, "step": 104550 }, { "epoch": 0.26, "learning_rate": 7.395037495596155e-05, "loss": 1.0221, "step": 104555 }, { "epoch": 0.26, "learning_rate": 7.394911671447984e-05, "loss": 1.0225, "step": 104560 }, { "epoch": 0.26, "learning_rate": 7.394785847299815e-05, "loss": 1.0233, "step": 104565 }, { "epoch": 0.26, "learning_rate": 7.394660023151644e-05, "loss": 1.022, "step": 104570 }, { "epoch": 0.26, "learning_rate": 7.394534199003473e-05, "loss": 1.0227, "step": 104575 }, { "epoch": 0.26, "learning_rate": 7.394408374855302e-05, "loss": 1.0242, "step": 104580 }, { "epoch": 0.26, "learning_rate": 7.394282550707133e-05, "loss": 1.0232, "step": 104585 }, { "epoch": 0.26, "learning_rate": 7.394156726558962e-05, "loss": 1.0234, "step": 104590 }, { "epoch": 0.26, "learning_rate": 7.394030902410791e-05, "loss": 1.025, "step": 104595 }, { "epoch": 0.26, "learning_rate": 7.39390507826262e-05, "loss": 1.0247, "step": 104600 }, { "epoch": 0.26, "learning_rate": 7.39377925411445e-05, "loss": 1.0217, "step": 104605 }, { "epoch": 0.26, "learning_rate": 7.39365342996628e-05, "loss": 1.0241, "step": 104610 }, { "epoch": 0.26, "learning_rate": 7.393527605818109e-05, "loss": 1.0244, "step": 104615 }, { "epoch": 0.26, "learning_rate": 7.393401781669938e-05, "loss": 1.0225, "step": 104620 }, { "epoch": 0.26, "learning_rate": 7.393275957521767e-05, "loss": 1.0224, "step": 104625 }, { "epoch": 0.26, "learning_rate": 7.393150133373598e-05, "loss": 1.0233, "step": 104630 }, { "epoch": 0.26, "learning_rate": 7.393024309225427e-05, "loss": 1.0245, "step": 104635 }, { "epoch": 0.26, "learning_rate": 7.392898485077256e-05, "loss": 1.0235, "step": 104640 }, { "epoch": 0.26, "learning_rate": 7.392772660929085e-05, "loss": 1.021, "step": 104645 }, { "epoch": 0.26, "learning_rate": 7.392646836780916e-05, "loss": 1.0214, "step": 104650 }, { "epoch": 0.26, "learning_rate": 7.392521012632745e-05, "loss": 1.0235, "step": 104655 }, { "epoch": 0.26, "learning_rate": 7.392395188484574e-05, "loss": 1.0222, "step": 104660 }, { "epoch": 0.26, "learning_rate": 7.392269364336403e-05, "loss": 1.0206, "step": 104665 }, { "epoch": 0.26, "learning_rate": 7.392143540188233e-05, "loss": 1.0221, "step": 104670 }, { "epoch": 0.26, "learning_rate": 7.392017716040063e-05, "loss": 1.0233, "step": 104675 }, { "epoch": 0.26, "learning_rate": 7.391891891891892e-05, "loss": 1.0235, "step": 104680 }, { "epoch": 0.26, "learning_rate": 7.391766067743721e-05, "loss": 1.0228, "step": 104685 }, { "epoch": 0.26, "learning_rate": 7.39164024359555e-05, "loss": 1.0216, "step": 104690 }, { "epoch": 0.26, "learning_rate": 7.391514419447381e-05, "loss": 1.0238, "step": 104695 }, { "epoch": 0.26, "learning_rate": 7.39138859529921e-05, "loss": 1.0235, "step": 104700 }, { "epoch": 0.26, "learning_rate": 7.39126277115104e-05, "loss": 1.0199, "step": 104705 }, { "epoch": 0.26, "learning_rate": 7.391136947002869e-05, "loss": 1.022, "step": 104710 }, { "epoch": 0.26, "learning_rate": 7.391011122854699e-05, "loss": 1.0207, "step": 104715 }, { "epoch": 0.26, "learning_rate": 7.390885298706528e-05, "loss": 1.0237, "step": 104720 }, { "epoch": 0.26, "learning_rate": 7.390759474558357e-05, "loss": 1.0235, "step": 104725 }, { "epoch": 0.26, "learning_rate": 7.390633650410187e-05, "loss": 1.0208, "step": 104730 }, { "epoch": 0.26, "learning_rate": 7.390507826262016e-05, "loss": 1.0235, "step": 104735 }, { "epoch": 0.26, "learning_rate": 7.390382002113846e-05, "loss": 1.0239, "step": 104740 }, { "epoch": 0.26, "learning_rate": 7.390256177965675e-05, "loss": 1.0224, "step": 104745 }, { "epoch": 0.26, "learning_rate": 7.390130353817505e-05, "loss": 1.0247, "step": 104750 }, { "epoch": 0.26, "learning_rate": 7.390004529669334e-05, "loss": 1.0243, "step": 104755 }, { "epoch": 0.26, "learning_rate": 7.389878705521164e-05, "loss": 1.0226, "step": 104760 }, { "epoch": 0.26, "learning_rate": 7.389752881372993e-05, "loss": 1.0227, "step": 104765 }, { "epoch": 0.26, "learning_rate": 7.389627057224823e-05, "loss": 1.0248, "step": 104770 }, { "epoch": 0.26, "learning_rate": 7.389501233076652e-05, "loss": 1.0216, "step": 104775 }, { "epoch": 0.26, "learning_rate": 7.389375408928482e-05, "loss": 1.0219, "step": 104780 }, { "epoch": 0.26, "learning_rate": 7.389249584780311e-05, "loss": 1.0223, "step": 104785 }, { "epoch": 0.26, "learning_rate": 7.38912376063214e-05, "loss": 1.0219, "step": 104790 }, { "epoch": 0.26, "learning_rate": 7.38899793648397e-05, "loss": 1.0221, "step": 104795 }, { "epoch": 0.26, "learning_rate": 7.388872112335799e-05, "loss": 1.0247, "step": 104800 }, { "epoch": 0.26, "learning_rate": 7.38874628818763e-05, "loss": 1.0233, "step": 104805 }, { "epoch": 0.26, "learning_rate": 7.388620464039459e-05, "loss": 1.0245, "step": 104810 }, { "epoch": 0.26, "learning_rate": 7.388494639891288e-05, "loss": 1.0226, "step": 104815 }, { "epoch": 0.26, "learning_rate": 7.388368815743117e-05, "loss": 1.0236, "step": 104820 }, { "epoch": 0.26, "learning_rate": 7.388242991594947e-05, "loss": 1.0203, "step": 104825 }, { "epoch": 0.26, "learning_rate": 7.388117167446777e-05, "loss": 1.022, "step": 104830 }, { "epoch": 0.26, "learning_rate": 7.387991343298606e-05, "loss": 1.0247, "step": 104835 }, { "epoch": 0.26, "learning_rate": 7.387865519150435e-05, "loss": 1.0228, "step": 104840 }, { "epoch": 0.26, "learning_rate": 7.387739695002265e-05, "loss": 1.0229, "step": 104845 }, { "epoch": 0.26, "learning_rate": 7.387613870854095e-05, "loss": 1.0228, "step": 104850 }, { "epoch": 0.26, "learning_rate": 7.387488046705924e-05, "loss": 1.0473, "step": 104855 }, { "epoch": 0.26, "learning_rate": 7.387362222557753e-05, "loss": 1.0236, "step": 104860 }, { "epoch": 0.26, "learning_rate": 7.387236398409582e-05, "loss": 1.02, "step": 104865 }, { "epoch": 0.26, "learning_rate": 7.387110574261413e-05, "loss": 1.0216, "step": 104870 }, { "epoch": 0.26, "learning_rate": 7.386984750113242e-05, "loss": 1.0222, "step": 104875 }, { "epoch": 0.26, "learning_rate": 7.386858925965071e-05, "loss": 1.023, "step": 104880 }, { "epoch": 0.26, "learning_rate": 7.3867331018169e-05, "loss": 1.0213, "step": 104885 }, { "epoch": 0.26, "learning_rate": 7.38660727766873e-05, "loss": 1.0226, "step": 104890 }, { "epoch": 0.26, "learning_rate": 7.38648145352056e-05, "loss": 1.0225, "step": 104895 }, { "epoch": 0.26, "learning_rate": 7.386355629372389e-05, "loss": 1.022, "step": 104900 }, { "epoch": 0.26, "learning_rate": 7.386229805224218e-05, "loss": 1.0248, "step": 104905 }, { "epoch": 0.26, "learning_rate": 7.386103981076049e-05, "loss": 1.0252, "step": 104910 }, { "epoch": 0.26, "learning_rate": 7.385978156927878e-05, "loss": 1.0234, "step": 104915 }, { "epoch": 0.26, "learning_rate": 7.385852332779707e-05, "loss": 1.0219, "step": 104920 }, { "epoch": 0.26, "learning_rate": 7.385726508631536e-05, "loss": 1.0226, "step": 104925 }, { "epoch": 0.26, "learning_rate": 7.385600684483367e-05, "loss": 1.022, "step": 104930 }, { "epoch": 0.26, "learning_rate": 7.385474860335197e-05, "loss": 1.0222, "step": 104935 }, { "epoch": 0.26, "learning_rate": 7.385349036187026e-05, "loss": 1.0235, "step": 104940 }, { "epoch": 0.26, "learning_rate": 7.385223212038855e-05, "loss": 1.0218, "step": 104945 }, { "epoch": 0.26, "learning_rate": 7.385097387890685e-05, "loss": 1.0211, "step": 104950 }, { "epoch": 0.26, "learning_rate": 7.384971563742514e-05, "loss": 1.0207, "step": 104955 }, { "epoch": 0.26, "learning_rate": 7.384845739594344e-05, "loss": 1.0229, "step": 104960 }, { "epoch": 0.26, "learning_rate": 7.384719915446173e-05, "loss": 1.0211, "step": 104965 }, { "epoch": 0.26, "learning_rate": 7.384594091298003e-05, "loss": 1.0211, "step": 104970 }, { "epoch": 0.26, "learning_rate": 7.384468267149832e-05, "loss": 1.0216, "step": 104975 }, { "epoch": 0.26, "learning_rate": 7.384342443001662e-05, "loss": 1.0241, "step": 104980 }, { "epoch": 0.26, "learning_rate": 7.384216618853491e-05, "loss": 1.0219, "step": 104985 }, { "epoch": 0.26, "learning_rate": 7.38409079470532e-05, "loss": 1.0227, "step": 104990 }, { "epoch": 0.26, "learning_rate": 7.38396497055715e-05, "loss": 1.023, "step": 104995 }, { "epoch": 0.26, "learning_rate": 7.38383914640898e-05, "loss": 1.0209, "step": 105000 }, { "epoch": 0.26, "learning_rate": 7.38371332226081e-05, "loss": 1.0225, "step": 105005 }, { "epoch": 0.26, "learning_rate": 7.383587498112639e-05, "loss": 1.0244, "step": 105010 }, { "epoch": 0.26, "learning_rate": 7.383461673964468e-05, "loss": 1.0222, "step": 105015 }, { "epoch": 0.26, "learning_rate": 7.383335849816297e-05, "loss": 1.0219, "step": 105020 }, { "epoch": 0.26, "learning_rate": 7.383210025668127e-05, "loss": 1.024, "step": 105025 }, { "epoch": 0.26, "learning_rate": 7.383084201519957e-05, "loss": 1.0232, "step": 105030 }, { "epoch": 0.26, "learning_rate": 7.382958377371786e-05, "loss": 1.0241, "step": 105035 }, { "epoch": 0.26, "learning_rate": 7.382832553223615e-05, "loss": 1.0223, "step": 105040 }, { "epoch": 0.26, "learning_rate": 7.382706729075445e-05, "loss": 1.0348, "step": 105045 }, { "epoch": 0.26, "learning_rate": 7.382580904927275e-05, "loss": 1.0244, "step": 105050 }, { "epoch": 0.26, "learning_rate": 7.382455080779104e-05, "loss": 1.0242, "step": 105055 }, { "epoch": 0.26, "learning_rate": 7.382329256630933e-05, "loss": 1.0234, "step": 105060 }, { "epoch": 0.26, "learning_rate": 7.382203432482763e-05, "loss": 1.0232, "step": 105065 }, { "epoch": 0.26, "learning_rate": 7.382077608334593e-05, "loss": 1.023, "step": 105070 }, { "epoch": 0.26, "learning_rate": 7.381951784186422e-05, "loss": 1.0223, "step": 105075 }, { "epoch": 0.26, "learning_rate": 7.381825960038251e-05, "loss": 1.0246, "step": 105080 }, { "epoch": 0.26, "learning_rate": 7.38170013589008e-05, "loss": 1.0243, "step": 105085 }, { "epoch": 0.26, "learning_rate": 7.38157431174191e-05, "loss": 1.0253, "step": 105090 }, { "epoch": 0.26, "learning_rate": 7.38144848759374e-05, "loss": 1.0232, "step": 105095 }, { "epoch": 0.26, "learning_rate": 7.381347828275202e-05, "loss": 1.0222, "step": 105100 }, { "epoch": 0.26, "learning_rate": 7.381247168956667e-05, "loss": 1.0214, "step": 105105 }, { "epoch": 0.26, "learning_rate": 7.381121344808496e-05, "loss": 1.0233, "step": 105110 }, { "epoch": 0.26, "learning_rate": 7.380995520660325e-05, "loss": 1.0488, "step": 105115 }, { "epoch": 0.26, "learning_rate": 7.380869696512154e-05, "loss": 1.0253, "step": 105120 }, { "epoch": 0.26, "learning_rate": 7.380743872363985e-05, "loss": 1.0234, "step": 105125 }, { "epoch": 0.26, "learning_rate": 7.380618048215814e-05, "loss": 1.0238, "step": 105130 }, { "epoch": 0.26, "learning_rate": 7.380492224067643e-05, "loss": 1.0224, "step": 105135 }, { "epoch": 0.26, "learning_rate": 7.380366399919472e-05, "loss": 1.0234, "step": 105140 }, { "epoch": 0.26, "learning_rate": 7.380240575771303e-05, "loss": 1.0221, "step": 105145 }, { "epoch": 0.26, "learning_rate": 7.380114751623132e-05, "loss": 1.0236, "step": 105150 }, { "epoch": 0.26, "learning_rate": 7.379988927474961e-05, "loss": 1.0431, "step": 105155 }, { "epoch": 0.26, "learning_rate": 7.37986310332679e-05, "loss": 1.0233, "step": 105160 }, { "epoch": 0.26, "learning_rate": 7.37973727917862e-05, "loss": 1.024, "step": 105165 }, { "epoch": 0.26, "learning_rate": 7.37961145503045e-05, "loss": 1.0236, "step": 105170 }, { "epoch": 0.26, "learning_rate": 7.379485630882279e-05, "loss": 1.0224, "step": 105175 }, { "epoch": 0.26, "learning_rate": 7.379359806734108e-05, "loss": 1.0209, "step": 105180 }, { "epoch": 0.26, "learning_rate": 7.379233982585937e-05, "loss": 1.0239, "step": 105185 }, { "epoch": 0.26, "learning_rate": 7.379108158437768e-05, "loss": 1.0212, "step": 105190 }, { "epoch": 0.26, "learning_rate": 7.378982334289597e-05, "loss": 1.0212, "step": 105195 }, { "epoch": 0.26, "learning_rate": 7.378856510141426e-05, "loss": 1.0234, "step": 105200 }, { "epoch": 0.26, "learning_rate": 7.378730685993255e-05, "loss": 1.0223, "step": 105205 }, { "epoch": 0.26, "learning_rate": 7.378604861845086e-05, "loss": 1.0233, "step": 105210 }, { "epoch": 0.26, "learning_rate": 7.378479037696915e-05, "loss": 1.0409, "step": 105215 }, { "epoch": 0.26, "learning_rate": 7.378353213548744e-05, "loss": 1.0201, "step": 105220 }, { "epoch": 0.26, "learning_rate": 7.378227389400573e-05, "loss": 1.0228, "step": 105225 }, { "epoch": 0.26, "learning_rate": 7.378101565252403e-05, "loss": 1.0235, "step": 105230 }, { "epoch": 0.26, "learning_rate": 7.377975741104233e-05, "loss": 1.0233, "step": 105235 }, { "epoch": 0.26, "learning_rate": 7.377849916956062e-05, "loss": 1.0226, "step": 105240 }, { "epoch": 0.26, "learning_rate": 7.377724092807891e-05, "loss": 1.0247, "step": 105245 }, { "epoch": 0.26, "learning_rate": 7.37759826865972e-05, "loss": 1.0228, "step": 105250 }, { "epoch": 0.26, "learning_rate": 7.377472444511551e-05, "loss": 1.0257, "step": 105255 }, { "epoch": 0.26, "learning_rate": 7.37734662036338e-05, "loss": 1.0187, "step": 105260 }, { "epoch": 0.26, "learning_rate": 7.37722079621521e-05, "loss": 1.0219, "step": 105265 }, { "epoch": 0.26, "learning_rate": 7.377094972067039e-05, "loss": 1.021, "step": 105270 }, { "epoch": 0.26, "learning_rate": 7.376969147918869e-05, "loss": 1.0215, "step": 105275 }, { "epoch": 0.26, "learning_rate": 7.376843323770698e-05, "loss": 1.0227, "step": 105280 }, { "epoch": 0.26, "learning_rate": 7.376717499622527e-05, "loss": 1.0248, "step": 105285 }, { "epoch": 0.26, "learning_rate": 7.376591675474357e-05, "loss": 1.0243, "step": 105290 }, { "epoch": 0.26, "learning_rate": 7.376465851326186e-05, "loss": 1.0242, "step": 105295 }, { "epoch": 0.26, "learning_rate": 7.376340027178016e-05, "loss": 1.025, "step": 105300 }, { "epoch": 0.26, "learning_rate": 7.376214203029845e-05, "loss": 1.0256, "step": 105305 }, { "epoch": 0.26, "learning_rate": 7.376088378881675e-05, "loss": 1.0247, "step": 105310 }, { "epoch": 0.26, "learning_rate": 7.375962554733504e-05, "loss": 1.0232, "step": 105315 }, { "epoch": 0.26, "learning_rate": 7.375836730585334e-05, "loss": 1.0227, "step": 105320 }, { "epoch": 0.26, "learning_rate": 7.375710906437163e-05, "loss": 1.0246, "step": 105325 }, { "epoch": 0.26, "learning_rate": 7.375585082288994e-05, "loss": 1.0227, "step": 105330 }, { "epoch": 0.26, "learning_rate": 7.375459258140823e-05, "loss": 1.0217, "step": 105335 }, { "epoch": 0.26, "learning_rate": 7.375333433992652e-05, "loss": 1.0237, "step": 105340 }, { "epoch": 0.26, "learning_rate": 7.375207609844483e-05, "loss": 1.0239, "step": 105345 }, { "epoch": 0.26, "learning_rate": 7.375081785696312e-05, "loss": 1.0223, "step": 105350 }, { "epoch": 0.26, "learning_rate": 7.374955961548141e-05, "loss": 1.0228, "step": 105355 }, { "epoch": 0.26, "learning_rate": 7.37483013739997e-05, "loss": 1.0225, "step": 105360 }, { "epoch": 0.26, "learning_rate": 7.3747043132518e-05, "loss": 1.0243, "step": 105365 }, { "epoch": 0.26, "learning_rate": 7.37457848910363e-05, "loss": 1.0227, "step": 105370 }, { "epoch": 0.26, "learning_rate": 7.374452664955459e-05, "loss": 1.0217, "step": 105375 }, { "epoch": 0.26, "learning_rate": 7.374326840807288e-05, "loss": 1.0212, "step": 105380 }, { "epoch": 0.26, "learning_rate": 7.374201016659117e-05, "loss": 1.0254, "step": 105385 }, { "epoch": 0.26, "learning_rate": 7.374075192510948e-05, "loss": 1.0232, "step": 105390 }, { "epoch": 0.26, "learning_rate": 7.373949368362777e-05, "loss": 1.0216, "step": 105395 }, { "epoch": 0.26, "learning_rate": 7.373823544214606e-05, "loss": 1.0205, "step": 105400 }, { "epoch": 0.26, "learning_rate": 7.373697720066435e-05, "loss": 1.0396, "step": 105405 }, { "epoch": 0.26, "learning_rate": 7.373571895918266e-05, "loss": 1.0239, "step": 105410 }, { "epoch": 0.26, "learning_rate": 7.373446071770095e-05, "loss": 1.0255, "step": 105415 }, { "epoch": 0.26, "learning_rate": 7.373320247621924e-05, "loss": 1.0226, "step": 105420 }, { "epoch": 0.26, "learning_rate": 7.373194423473753e-05, "loss": 1.0203, "step": 105425 }, { "epoch": 0.26, "learning_rate": 7.373068599325583e-05, "loss": 1.0229, "step": 105430 }, { "epoch": 0.26, "learning_rate": 7.372942775177413e-05, "loss": 1.0211, "step": 105435 }, { "epoch": 0.26, "learning_rate": 7.372816951029242e-05, "loss": 1.0227, "step": 105440 }, { "epoch": 0.26, "learning_rate": 7.372691126881071e-05, "loss": 1.0222, "step": 105445 }, { "epoch": 0.26, "learning_rate": 7.3725653027329e-05, "loss": 1.0249, "step": 105450 }, { "epoch": 0.26, "learning_rate": 7.372439478584731e-05, "loss": 1.0237, "step": 105455 }, { "epoch": 0.26, "learning_rate": 7.37231365443656e-05, "loss": 1.0225, "step": 105460 }, { "epoch": 0.26, "learning_rate": 7.37218783028839e-05, "loss": 1.0248, "step": 105465 }, { "epoch": 0.26, "learning_rate": 7.372062006140219e-05, "loss": 1.0227, "step": 105470 }, { "epoch": 0.26, "learning_rate": 7.371936181992049e-05, "loss": 1.022, "step": 105475 }, { "epoch": 0.26, "learning_rate": 7.371810357843878e-05, "loss": 1.0239, "step": 105480 }, { "epoch": 0.26, "learning_rate": 7.371684533695707e-05, "loss": 1.0239, "step": 105485 }, { "epoch": 0.26, "learning_rate": 7.371558709547537e-05, "loss": 1.0461, "step": 105490 }, { "epoch": 0.26, "learning_rate": 7.371432885399366e-05, "loss": 1.0215, "step": 105495 }, { "epoch": 0.26, "learning_rate": 7.371307061251196e-05, "loss": 1.0219, "step": 105500 }, { "epoch": 0.26, "learning_rate": 7.371181237103025e-05, "loss": 1.0221, "step": 105505 }, { "epoch": 0.26, "learning_rate": 7.371055412954855e-05, "loss": 1.0229, "step": 105510 }, { "epoch": 0.26, "learning_rate": 7.370929588806684e-05, "loss": 1.0234, "step": 105515 }, { "epoch": 0.26, "learning_rate": 7.370803764658514e-05, "loss": 1.0246, "step": 105520 }, { "epoch": 0.26, "learning_rate": 7.370677940510343e-05, "loss": 1.0233, "step": 105525 }, { "epoch": 0.26, "learning_rate": 7.370552116362173e-05, "loss": 1.0223, "step": 105530 }, { "epoch": 0.26, "learning_rate": 7.370426292214002e-05, "loss": 1.0216, "step": 105535 }, { "epoch": 0.26, "learning_rate": 7.370300468065832e-05, "loss": 1.0459, "step": 105540 }, { "epoch": 0.26, "learning_rate": 7.370174643917661e-05, "loss": 1.0217, "step": 105545 }, { "epoch": 0.26, "learning_rate": 7.37004881976949e-05, "loss": 1.0236, "step": 105550 }, { "epoch": 0.26, "learning_rate": 7.36992299562132e-05, "loss": 1.0415, "step": 105555 }, { "epoch": 0.26, "learning_rate": 7.369797171473149e-05, "loss": 1.0241, "step": 105560 }, { "epoch": 0.26, "learning_rate": 7.36967134732498e-05, "loss": 1.0231, "step": 105565 }, { "epoch": 0.26, "learning_rate": 7.369545523176809e-05, "loss": 1.0237, "step": 105570 }, { "epoch": 0.27, "learning_rate": 7.369419699028638e-05, "loss": 1.0214, "step": 105575 }, { "epoch": 0.27, "learning_rate": 7.369293874880467e-05, "loss": 1.021, "step": 105580 }, { "epoch": 0.27, "learning_rate": 7.369168050732297e-05, "loss": 1.0231, "step": 105585 }, { "epoch": 0.27, "learning_rate": 7.369042226584127e-05, "loss": 1.0408, "step": 105590 }, { "epoch": 0.27, "learning_rate": 7.368916402435956e-05, "loss": 1.0247, "step": 105595 }, { "epoch": 0.27, "learning_rate": 7.368790578287785e-05, "loss": 1.0213, "step": 105600 }, { "epoch": 0.27, "learning_rate": 7.368664754139615e-05, "loss": 1.0212, "step": 105605 }, { "epoch": 0.27, "learning_rate": 7.368538929991445e-05, "loss": 1.022, "step": 105610 }, { "epoch": 0.27, "learning_rate": 7.368413105843274e-05, "loss": 1.0448, "step": 105615 }, { "epoch": 0.27, "learning_rate": 7.368287281695103e-05, "loss": 1.0219, "step": 105620 }, { "epoch": 0.27, "learning_rate": 7.368161457546932e-05, "loss": 1.0228, "step": 105625 }, { "epoch": 0.27, "learning_rate": 7.368035633398763e-05, "loss": 1.0233, "step": 105630 }, { "epoch": 0.27, "learning_rate": 7.367909809250592e-05, "loss": 1.0233, "step": 105635 }, { "epoch": 0.27, "learning_rate": 7.367783985102421e-05, "loss": 1.0226, "step": 105640 }, { "epoch": 0.27, "learning_rate": 7.36765816095425e-05, "loss": 1.023, "step": 105645 }, { "epoch": 0.27, "learning_rate": 7.36753233680608e-05, "loss": 1.0412, "step": 105650 }, { "epoch": 0.27, "learning_rate": 7.36740651265791e-05, "loss": 1.0235, "step": 105655 }, { "epoch": 0.27, "learning_rate": 7.367280688509739e-05, "loss": 1.0221, "step": 105660 }, { "epoch": 0.27, "learning_rate": 7.367154864361568e-05, "loss": 1.0226, "step": 105665 }, { "epoch": 0.27, "learning_rate": 7.367029040213399e-05, "loss": 1.0233, "step": 105670 }, { "epoch": 0.27, "learning_rate": 7.366903216065228e-05, "loss": 1.0236, "step": 105675 }, { "epoch": 0.27, "learning_rate": 7.366777391917057e-05, "loss": 1.0241, "step": 105680 }, { "epoch": 0.27, "learning_rate": 7.366651567768886e-05, "loss": 1.0238, "step": 105685 }, { "epoch": 0.27, "learning_rate": 7.366525743620715e-05, "loss": 1.023, "step": 105690 }, { "epoch": 0.27, "learning_rate": 7.366399919472546e-05, "loss": 1.0251, "step": 105695 }, { "epoch": 0.27, "learning_rate": 7.366274095324375e-05, "loss": 1.0249, "step": 105700 }, { "epoch": 0.27, "learning_rate": 7.366148271176204e-05, "loss": 1.0228, "step": 105705 }, { "epoch": 0.27, "learning_rate": 7.366022447028033e-05, "loss": 1.0231, "step": 105710 }, { "epoch": 0.27, "learning_rate": 7.365896622879864e-05, "loss": 1.021, "step": 105715 }, { "epoch": 0.27, "learning_rate": 7.365770798731693e-05, "loss": 1.0233, "step": 105720 }, { "epoch": 0.27, "learning_rate": 7.365644974583522e-05, "loss": 1.0241, "step": 105725 }, { "epoch": 0.27, "learning_rate": 7.365519150435351e-05, "loss": 1.0232, "step": 105730 }, { "epoch": 0.27, "learning_rate": 7.365393326287182e-05, "loss": 1.0374, "step": 105735 }, { "epoch": 0.27, "learning_rate": 7.365267502139011e-05, "loss": 1.021, "step": 105740 }, { "epoch": 0.27, "learning_rate": 7.36514167799084e-05, "loss": 1.0223, "step": 105745 }, { "epoch": 0.27, "learning_rate": 7.365015853842669e-05, "loss": 1.024, "step": 105750 }, { "epoch": 0.27, "learning_rate": 7.364890029694498e-05, "loss": 1.0229, "step": 105755 }, { "epoch": 0.27, "learning_rate": 7.364764205546329e-05, "loss": 1.0228, "step": 105760 }, { "epoch": 0.27, "learning_rate": 7.364638381398158e-05, "loss": 1.0225, "step": 105765 }, { "epoch": 0.27, "learning_rate": 7.364512557249987e-05, "loss": 1.0242, "step": 105770 }, { "epoch": 0.27, "learning_rate": 7.364386733101816e-05, "loss": 1.0219, "step": 105775 }, { "epoch": 0.27, "learning_rate": 7.364260908953647e-05, "loss": 1.0201, "step": 105780 }, { "epoch": 0.27, "learning_rate": 7.364135084805476e-05, "loss": 1.0237, "step": 105785 }, { "epoch": 0.27, "learning_rate": 7.364009260657305e-05, "loss": 1.0223, "step": 105790 }, { "epoch": 0.27, "learning_rate": 7.363883436509134e-05, "loss": 1.0223, "step": 105795 }, { "epoch": 0.27, "learning_rate": 7.363757612360965e-05, "loss": 1.0228, "step": 105800 }, { "epoch": 0.27, "learning_rate": 7.363631788212794e-05, "loss": 1.0234, "step": 105805 }, { "epoch": 0.27, "learning_rate": 7.363505964064623e-05, "loss": 1.022, "step": 105810 }, { "epoch": 0.27, "learning_rate": 7.363380139916452e-05, "loss": 1.0235, "step": 105815 }, { "epoch": 0.27, "learning_rate": 7.363254315768281e-05, "loss": 1.022, "step": 105820 }, { "epoch": 0.27, "learning_rate": 7.363128491620112e-05, "loss": 1.0236, "step": 105825 }, { "epoch": 0.27, "learning_rate": 7.363002667471942e-05, "loss": 1.024, "step": 105830 }, { "epoch": 0.27, "learning_rate": 7.362876843323772e-05, "loss": 1.0238, "step": 105835 }, { "epoch": 0.27, "learning_rate": 7.362751019175601e-05, "loss": 1.0228, "step": 105840 }, { "epoch": 0.27, "learning_rate": 7.36262519502743e-05, "loss": 1.0241, "step": 105845 }, { "epoch": 0.27, "learning_rate": 7.36249937087926e-05, "loss": 1.0228, "step": 105850 }, { "epoch": 0.27, "learning_rate": 7.36237354673109e-05, "loss": 1.0218, "step": 105855 }, { "epoch": 0.27, "learning_rate": 7.362247722582919e-05, "loss": 1.0236, "step": 105860 }, { "epoch": 0.27, "learning_rate": 7.362121898434748e-05, "loss": 1.0217, "step": 105865 }, { "epoch": 0.27, "learning_rate": 7.361996074286578e-05, "loss": 1.0221, "step": 105870 }, { "epoch": 0.27, "learning_rate": 7.361870250138408e-05, "loss": 1.022, "step": 105875 }, { "epoch": 0.27, "learning_rate": 7.361744425990237e-05, "loss": 1.0227, "step": 105880 }, { "epoch": 0.27, "learning_rate": 7.361618601842066e-05, "loss": 1.0226, "step": 105885 }, { "epoch": 0.27, "learning_rate": 7.361492777693896e-05, "loss": 1.0215, "step": 105890 }, { "epoch": 0.27, "learning_rate": 7.361366953545726e-05, "loss": 1.0217, "step": 105895 }, { "epoch": 0.27, "learning_rate": 7.361241129397555e-05, "loss": 1.023, "step": 105900 }, { "epoch": 0.27, "learning_rate": 7.361115305249384e-05, "loss": 1.0221, "step": 105905 }, { "epoch": 0.27, "learning_rate": 7.360989481101213e-05, "loss": 1.0223, "step": 105910 }, { "epoch": 0.27, "learning_rate": 7.360863656953044e-05, "loss": 1.0216, "step": 105915 }, { "epoch": 0.27, "learning_rate": 7.360737832804873e-05, "loss": 1.024, "step": 105920 }, { "epoch": 0.27, "learning_rate": 7.360612008656702e-05, "loss": 1.0227, "step": 105925 }, { "epoch": 0.27, "learning_rate": 7.360486184508531e-05, "loss": 1.0414, "step": 105930 }, { "epoch": 0.27, "learning_rate": 7.360360360360362e-05, "loss": 1.0212, "step": 105935 }, { "epoch": 0.27, "learning_rate": 7.360234536212191e-05, "loss": 1.0228, "step": 105940 }, { "epoch": 0.27, "learning_rate": 7.36010871206402e-05, "loss": 1.0239, "step": 105945 }, { "epoch": 0.27, "learning_rate": 7.359982887915849e-05, "loss": 1.02, "step": 105950 }, { "epoch": 0.27, "learning_rate": 7.35985706376768e-05, "loss": 1.0213, "step": 105955 }, { "epoch": 0.27, "learning_rate": 7.359731239619509e-05, "loss": 1.0218, "step": 105960 }, { "epoch": 0.27, "learning_rate": 7.359605415471338e-05, "loss": 1.0227, "step": 105965 }, { "epoch": 0.27, "learning_rate": 7.359479591323167e-05, "loss": 1.022, "step": 105970 }, { "epoch": 0.27, "learning_rate": 7.359353767174996e-05, "loss": 1.0253, "step": 105975 }, { "epoch": 0.27, "learning_rate": 7.359227943026827e-05, "loss": 1.024, "step": 105980 }, { "epoch": 0.27, "learning_rate": 7.359102118878656e-05, "loss": 1.0214, "step": 105985 }, { "epoch": 0.27, "learning_rate": 7.358976294730485e-05, "loss": 1.0241, "step": 105990 }, { "epoch": 0.27, "learning_rate": 7.358850470582314e-05, "loss": 1.0248, "step": 105995 }, { "epoch": 0.27, "learning_rate": 7.358724646434145e-05, "loss": 1.0241, "step": 106000 }, { "epoch": 0.27, "learning_rate": 7.358598822285974e-05, "loss": 1.0225, "step": 106005 }, { "epoch": 0.27, "learning_rate": 7.358472998137803e-05, "loss": 1.0228, "step": 106010 }, { "epoch": 0.27, "learning_rate": 7.358347173989632e-05, "loss": 1.0413, "step": 106015 }, { "epoch": 0.27, "learning_rate": 7.358221349841461e-05, "loss": 1.0236, "step": 106020 }, { "epoch": 0.27, "learning_rate": 7.358095525693292e-05, "loss": 1.0221, "step": 106025 }, { "epoch": 0.27, "learning_rate": 7.357969701545121e-05, "loss": 1.0454, "step": 106030 }, { "epoch": 0.27, "learning_rate": 7.35784387739695e-05, "loss": 1.0244, "step": 106035 }, { "epoch": 0.27, "learning_rate": 7.35771805324878e-05, "loss": 1.024, "step": 106040 }, { "epoch": 0.27, "learning_rate": 7.35759222910061e-05, "loss": 1.0231, "step": 106045 }, { "epoch": 0.27, "learning_rate": 7.357466404952439e-05, "loss": 1.0244, "step": 106050 }, { "epoch": 0.27, "learning_rate": 7.357340580804268e-05, "loss": 1.0241, "step": 106055 }, { "epoch": 0.27, "learning_rate": 7.357214756656097e-05, "loss": 1.0225, "step": 106060 }, { "epoch": 0.27, "learning_rate": 7.357088932507928e-05, "loss": 1.0231, "step": 106065 }, { "epoch": 0.27, "learning_rate": 7.356963108359757e-05, "loss": 1.0248, "step": 106070 }, { "epoch": 0.27, "learning_rate": 7.356837284211586e-05, "loss": 1.0225, "step": 106075 }, { "epoch": 0.27, "learning_rate": 7.356711460063415e-05, "loss": 1.0249, "step": 106080 }, { "epoch": 0.27, "learning_rate": 7.356585635915245e-05, "loss": 1.0204, "step": 106085 }, { "epoch": 0.27, "learning_rate": 7.356459811767075e-05, "loss": 1.0218, "step": 106090 }, { "epoch": 0.27, "learning_rate": 7.356333987618904e-05, "loss": 1.0229, "step": 106095 }, { "epoch": 0.27, "learning_rate": 7.356208163470733e-05, "loss": 1.0264, "step": 106100 }, { "epoch": 0.27, "learning_rate": 7.356082339322563e-05, "loss": 1.0252, "step": 106105 }, { "epoch": 0.27, "learning_rate": 7.355956515174393e-05, "loss": 1.0219, "step": 106110 }, { "epoch": 0.27, "learning_rate": 7.355830691026222e-05, "loss": 1.0241, "step": 106115 }, { "epoch": 0.27, "learning_rate": 7.355704866878051e-05, "loss": 1.0249, "step": 106120 }, { "epoch": 0.27, "learning_rate": 7.35557904272988e-05, "loss": 1.0215, "step": 106125 }, { "epoch": 0.27, "learning_rate": 7.355453218581711e-05, "loss": 1.0225, "step": 106130 }, { "epoch": 0.27, "learning_rate": 7.35532739443354e-05, "loss": 1.0216, "step": 106135 }, { "epoch": 0.27, "learning_rate": 7.35520157028537e-05, "loss": 1.0259, "step": 106140 }, { "epoch": 0.27, "learning_rate": 7.355075746137199e-05, "loss": 1.0228, "step": 106145 }, { "epoch": 0.27, "learning_rate": 7.354949921989028e-05, "loss": 1.0245, "step": 106150 }, { "epoch": 0.27, "learning_rate": 7.354824097840858e-05, "loss": 1.0414, "step": 106155 }, { "epoch": 0.27, "learning_rate": 7.354698273692687e-05, "loss": 1.0226, "step": 106160 }, { "epoch": 0.27, "learning_rate": 7.354572449544516e-05, "loss": 1.0223, "step": 106165 }, { "epoch": 0.27, "learning_rate": 7.354446625396346e-05, "loss": 1.023, "step": 106170 }, { "epoch": 0.27, "learning_rate": 7.354320801248176e-05, "loss": 1.0226, "step": 106175 }, { "epoch": 0.27, "learning_rate": 7.354194977100005e-05, "loss": 1.0236, "step": 106180 }, { "epoch": 0.27, "learning_rate": 7.354069152951834e-05, "loss": 1.0382, "step": 106185 }, { "epoch": 0.27, "learning_rate": 7.353943328803664e-05, "loss": 1.023, "step": 106190 }, { "epoch": 0.27, "learning_rate": 7.353817504655494e-05, "loss": 1.0215, "step": 106195 }, { "epoch": 0.27, "learning_rate": 7.353691680507323e-05, "loss": 1.0254, "step": 106200 }, { "epoch": 0.27, "learning_rate": 7.353565856359152e-05, "loss": 1.0219, "step": 106205 }, { "epoch": 0.27, "learning_rate": 7.353440032210982e-05, "loss": 1.0212, "step": 106210 }, { "epoch": 0.27, "learning_rate": 7.353314208062811e-05, "loss": 1.022, "step": 106215 }, { "epoch": 0.27, "learning_rate": 7.353188383914641e-05, "loss": 1.0242, "step": 106220 }, { "epoch": 0.27, "learning_rate": 7.35306255976647e-05, "loss": 1.0234, "step": 106225 }, { "epoch": 0.27, "learning_rate": 7.3529367356183e-05, "loss": 1.0213, "step": 106230 }, { "epoch": 0.27, "learning_rate": 7.352810911470129e-05, "loss": 1.0229, "step": 106235 }, { "epoch": 0.27, "learning_rate": 7.352685087321959e-05, "loss": 1.0227, "step": 106240 }, { "epoch": 0.27, "learning_rate": 7.352559263173788e-05, "loss": 1.0227, "step": 106245 }, { "epoch": 0.27, "learning_rate": 7.352433439025618e-05, "loss": 1.0264, "step": 106250 }, { "epoch": 0.27, "learning_rate": 7.352307614877447e-05, "loss": 1.0223, "step": 106255 }, { "epoch": 0.27, "learning_rate": 7.352181790729277e-05, "loss": 1.0209, "step": 106260 }, { "epoch": 0.27, "learning_rate": 7.352055966581106e-05, "loss": 1.023, "step": 106265 }, { "epoch": 0.27, "learning_rate": 7.351930142432936e-05, "loss": 1.021, "step": 106270 }, { "epoch": 0.27, "learning_rate": 7.351804318284765e-05, "loss": 1.0238, "step": 106275 }, { "epoch": 0.27, "learning_rate": 7.351678494136594e-05, "loss": 1.0223, "step": 106280 }, { "epoch": 0.27, "learning_rate": 7.351552669988424e-05, "loss": 1.0215, "step": 106285 }, { "epoch": 0.27, "learning_rate": 7.351426845840254e-05, "loss": 1.0208, "step": 106290 }, { "epoch": 0.27, "learning_rate": 7.351301021692083e-05, "loss": 1.024, "step": 106295 }, { "epoch": 0.27, "learning_rate": 7.351175197543912e-05, "loss": 1.0235, "step": 106300 }, { "epoch": 0.27, "learning_rate": 7.351049373395742e-05, "loss": 1.022, "step": 106305 }, { "epoch": 0.27, "learning_rate": 7.350923549247572e-05, "loss": 1.0215, "step": 106310 }, { "epoch": 0.27, "learning_rate": 7.350797725099401e-05, "loss": 1.0228, "step": 106315 }, { "epoch": 0.27, "learning_rate": 7.35067190095123e-05, "loss": 1.0226, "step": 106320 }, { "epoch": 0.27, "learning_rate": 7.35054607680306e-05, "loss": 1.023, "step": 106325 }, { "epoch": 0.27, "learning_rate": 7.350420252654891e-05, "loss": 1.0228, "step": 106330 }, { "epoch": 0.27, "learning_rate": 7.35029442850672e-05, "loss": 1.0244, "step": 106335 }, { "epoch": 0.27, "learning_rate": 7.350168604358549e-05, "loss": 1.0233, "step": 106340 }, { "epoch": 0.27, "learning_rate": 7.350042780210378e-05, "loss": 1.0212, "step": 106345 }, { "epoch": 0.27, "learning_rate": 7.349916956062209e-05, "loss": 1.0227, "step": 106350 }, { "epoch": 0.27, "learning_rate": 7.349791131914038e-05, "loss": 1.0229, "step": 106355 }, { "epoch": 0.27, "learning_rate": 7.349665307765867e-05, "loss": 1.0233, "step": 106360 }, { "epoch": 0.27, "learning_rate": 7.349539483617696e-05, "loss": 1.0239, "step": 106365 }, { "epoch": 0.27, "learning_rate": 7.349413659469526e-05, "loss": 1.0219, "step": 106370 }, { "epoch": 0.27, "learning_rate": 7.349287835321356e-05, "loss": 1.0223, "step": 106375 }, { "epoch": 0.27, "learning_rate": 7.349162011173185e-05, "loss": 1.0209, "step": 106380 }, { "epoch": 0.27, "learning_rate": 7.349036187025014e-05, "loss": 1.022, "step": 106385 }, { "epoch": 0.27, "learning_rate": 7.348910362876844e-05, "loss": 1.0222, "step": 106390 }, { "epoch": 0.27, "learning_rate": 7.348784538728674e-05, "loss": 1.0212, "step": 106395 }, { "epoch": 0.27, "learning_rate": 7.348658714580503e-05, "loss": 1.0232, "step": 106400 }, { "epoch": 0.27, "learning_rate": 7.348532890432332e-05, "loss": 1.0226, "step": 106405 }, { "epoch": 0.27, "learning_rate": 7.348407066284162e-05, "loss": 1.022, "step": 106410 }, { "epoch": 0.27, "learning_rate": 7.348281242135992e-05, "loss": 1.0225, "step": 106415 }, { "epoch": 0.27, "learning_rate": 7.348155417987821e-05, "loss": 1.0218, "step": 106420 }, { "epoch": 0.27, "learning_rate": 7.34802959383965e-05, "loss": 1.0217, "step": 106425 }, { "epoch": 0.27, "learning_rate": 7.34790376969148e-05, "loss": 1.0219, "step": 106430 }, { "epoch": 0.27, "learning_rate": 7.347777945543309e-05, "loss": 1.0229, "step": 106435 }, { "epoch": 0.27, "learning_rate": 7.347652121395139e-05, "loss": 1.0237, "step": 106440 }, { "epoch": 0.27, "learning_rate": 7.347526297246968e-05, "loss": 1.0235, "step": 106445 }, { "epoch": 0.27, "learning_rate": 7.347400473098798e-05, "loss": 1.0243, "step": 106450 }, { "epoch": 0.27, "learning_rate": 7.347274648950627e-05, "loss": 1.023, "step": 106455 }, { "epoch": 0.27, "learning_rate": 7.347148824802457e-05, "loss": 1.0207, "step": 106460 }, { "epoch": 0.27, "learning_rate": 7.347023000654286e-05, "loss": 1.0242, "step": 106465 }, { "epoch": 0.27, "learning_rate": 7.346897176506116e-05, "loss": 1.0227, "step": 106470 }, { "epoch": 0.27, "learning_rate": 7.346771352357945e-05, "loss": 1.0218, "step": 106475 }, { "epoch": 0.27, "learning_rate": 7.346645528209775e-05, "loss": 1.0241, "step": 106480 }, { "epoch": 0.27, "learning_rate": 7.346519704061604e-05, "loss": 1.0239, "step": 106485 }, { "epoch": 0.27, "learning_rate": 7.346393879913434e-05, "loss": 1.0235, "step": 106490 }, { "epoch": 0.27, "learning_rate": 7.346268055765263e-05, "loss": 1.0241, "step": 106495 }, { "epoch": 0.27, "learning_rate": 7.346142231617092e-05, "loss": 1.0222, "step": 106500 }, { "epoch": 0.27, "learning_rate": 7.346016407468922e-05, "loss": 1.0197, "step": 106505 }, { "epoch": 0.27, "learning_rate": 7.345890583320752e-05, "loss": 1.0248, "step": 106510 }, { "epoch": 0.27, "learning_rate": 7.345764759172581e-05, "loss": 1.0239, "step": 106515 }, { "epoch": 0.27, "learning_rate": 7.34563893502441e-05, "loss": 1.0226, "step": 106520 }, { "epoch": 0.27, "learning_rate": 7.34551311087624e-05, "loss": 1.0223, "step": 106525 }, { "epoch": 0.27, "learning_rate": 7.34538728672807e-05, "loss": 1.023, "step": 106530 }, { "epoch": 0.27, "learning_rate": 7.345261462579899e-05, "loss": 1.0231, "step": 106535 }, { "epoch": 0.27, "learning_rate": 7.345135638431728e-05, "loss": 1.0229, "step": 106540 }, { "epoch": 0.27, "learning_rate": 7.345009814283558e-05, "loss": 1.0213, "step": 106545 }, { "epoch": 0.27, "learning_rate": 7.344883990135388e-05, "loss": 1.0212, "step": 106550 }, { "epoch": 0.27, "learning_rate": 7.344758165987217e-05, "loss": 1.0221, "step": 106555 }, { "epoch": 0.27, "learning_rate": 7.344632341839046e-05, "loss": 1.0235, "step": 106560 }, { "epoch": 0.27, "learning_rate": 7.344506517690875e-05, "loss": 1.0241, "step": 106565 }, { "epoch": 0.27, "learning_rate": 7.344380693542706e-05, "loss": 1.0216, "step": 106570 }, { "epoch": 0.27, "learning_rate": 7.344254869394535e-05, "loss": 1.0218, "step": 106575 }, { "epoch": 0.27, "learning_rate": 7.344129045246364e-05, "loss": 1.0227, "step": 106580 }, { "epoch": 0.27, "learning_rate": 7.344003221098193e-05, "loss": 1.0232, "step": 106585 }, { "epoch": 0.27, "learning_rate": 7.343877396950024e-05, "loss": 1.0218, "step": 106590 }, { "epoch": 0.27, "learning_rate": 7.343751572801853e-05, "loss": 1.023, "step": 106595 }, { "epoch": 0.27, "learning_rate": 7.343625748653682e-05, "loss": 1.0238, "step": 106600 }, { "epoch": 0.27, "learning_rate": 7.343499924505511e-05, "loss": 1.0225, "step": 106605 }, { "epoch": 0.27, "learning_rate": 7.343374100357342e-05, "loss": 1.0186, "step": 106610 }, { "epoch": 0.27, "learning_rate": 7.343248276209171e-05, "loss": 1.0236, "step": 106615 }, { "epoch": 0.27, "learning_rate": 7.343122452061e-05, "loss": 1.0214, "step": 106620 }, { "epoch": 0.27, "learning_rate": 7.342996627912829e-05, "loss": 1.0238, "step": 106625 }, { "epoch": 0.27, "learning_rate": 7.342870803764658e-05, "loss": 1.022, "step": 106630 }, { "epoch": 0.27, "learning_rate": 7.342744979616489e-05, "loss": 1.0239, "step": 106635 }, { "epoch": 0.27, "learning_rate": 7.342619155468318e-05, "loss": 1.0235, "step": 106640 }, { "epoch": 0.27, "learning_rate": 7.342493331320147e-05, "loss": 1.0234, "step": 106645 }, { "epoch": 0.27, "learning_rate": 7.342367507171976e-05, "loss": 1.0202, "step": 106650 }, { "epoch": 0.27, "learning_rate": 7.342241683023807e-05, "loss": 1.0441, "step": 106655 }, { "epoch": 0.27, "learning_rate": 7.342115858875636e-05, "loss": 1.0449, "step": 106660 }, { "epoch": 0.27, "learning_rate": 7.341990034727465e-05, "loss": 1.0237, "step": 106665 }, { "epoch": 0.27, "learning_rate": 7.341864210579294e-05, "loss": 1.0242, "step": 106670 }, { "epoch": 0.27, "learning_rate": 7.341738386431123e-05, "loss": 1.0242, "step": 106675 }, { "epoch": 0.27, "learning_rate": 7.341612562282954e-05, "loss": 1.0232, "step": 106680 }, { "epoch": 0.27, "learning_rate": 7.341486738134783e-05, "loss": 1.0209, "step": 106685 }, { "epoch": 0.27, "learning_rate": 7.341360913986612e-05, "loss": 1.0242, "step": 106690 }, { "epoch": 0.27, "learning_rate": 7.341235089838441e-05, "loss": 1.0236, "step": 106695 }, { "epoch": 0.27, "learning_rate": 7.341109265690272e-05, "loss": 1.0241, "step": 106700 }, { "epoch": 0.27, "learning_rate": 7.340983441542101e-05, "loss": 1.024, "step": 106705 }, { "epoch": 0.27, "learning_rate": 7.34085761739393e-05, "loss": 1.0256, "step": 106710 }, { "epoch": 0.27, "learning_rate": 7.340731793245759e-05, "loss": 1.0224, "step": 106715 }, { "epoch": 0.27, "learning_rate": 7.34060596909759e-05, "loss": 1.0242, "step": 106720 }, { "epoch": 0.27, "learning_rate": 7.340480144949419e-05, "loss": 1.0234, "step": 106725 }, { "epoch": 0.27, "learning_rate": 7.340354320801248e-05, "loss": 1.0229, "step": 106730 }, { "epoch": 0.27, "learning_rate": 7.340228496653077e-05, "loss": 1.0232, "step": 106735 }, { "epoch": 0.27, "learning_rate": 7.340102672504906e-05, "loss": 1.0226, "step": 106740 }, { "epoch": 0.27, "learning_rate": 7.339976848356737e-05, "loss": 1.0202, "step": 106745 }, { "epoch": 0.27, "learning_rate": 7.339851024208566e-05, "loss": 1.0221, "step": 106750 }, { "epoch": 0.27, "learning_rate": 7.339725200060395e-05, "loss": 1.0221, "step": 106755 }, { "epoch": 0.27, "learning_rate": 7.339599375912224e-05, "loss": 1.0213, "step": 106760 }, { "epoch": 0.27, "learning_rate": 7.339473551764055e-05, "loss": 1.0224, "step": 106765 }, { "epoch": 0.27, "learning_rate": 7.339347727615884e-05, "loss": 1.0218, "step": 106770 }, { "epoch": 0.27, "learning_rate": 7.339221903467713e-05, "loss": 1.0238, "step": 106775 }, { "epoch": 0.27, "learning_rate": 7.339096079319542e-05, "loss": 1.0241, "step": 106780 }, { "epoch": 0.27, "learning_rate": 7.338970255171373e-05, "loss": 1.0244, "step": 106785 }, { "epoch": 0.27, "learning_rate": 7.338844431023202e-05, "loss": 1.0254, "step": 106790 }, { "epoch": 0.27, "learning_rate": 7.338718606875031e-05, "loss": 1.0223, "step": 106795 }, { "epoch": 0.27, "learning_rate": 7.33859278272686e-05, "loss": 1.0226, "step": 106800 }, { "epoch": 0.27, "learning_rate": 7.33846695857869e-05, "loss": 1.0245, "step": 106805 }, { "epoch": 0.27, "learning_rate": 7.33834113443052e-05, "loss": 1.0252, "step": 106810 }, { "epoch": 0.27, "learning_rate": 7.338215310282349e-05, "loss": 1.0229, "step": 106815 }, { "epoch": 0.27, "learning_rate": 7.338089486134178e-05, "loss": 1.0251, "step": 106820 }, { "epoch": 0.27, "learning_rate": 7.337963661986008e-05, "loss": 1.0208, "step": 106825 }, { "epoch": 0.27, "learning_rate": 7.337837837837838e-05, "loss": 1.0214, "step": 106830 }, { "epoch": 0.27, "learning_rate": 7.337712013689669e-05, "loss": 1.0218, "step": 106835 }, { "epoch": 0.27, "learning_rate": 7.337586189541498e-05, "loss": 1.0224, "step": 106840 }, { "epoch": 0.27, "learning_rate": 7.337460365393327e-05, "loss": 1.0234, "step": 106845 }, { "epoch": 0.27, "learning_rate": 7.337334541245156e-05, "loss": 1.0225, "step": 106850 }, { "epoch": 0.27, "learning_rate": 7.337208717096987e-05, "loss": 1.0232, "step": 106855 }, { "epoch": 0.27, "learning_rate": 7.337082892948816e-05, "loss": 1.0231, "step": 106860 }, { "epoch": 0.27, "learning_rate": 7.336957068800645e-05, "loss": 1.0219, "step": 106865 }, { "epoch": 0.27, "learning_rate": 7.336831244652474e-05, "loss": 1.0237, "step": 106870 }, { "epoch": 0.27, "learning_rate": 7.336705420504305e-05, "loss": 1.0229, "step": 106875 }, { "epoch": 0.27, "learning_rate": 7.336579596356134e-05, "loss": 1.0237, "step": 106880 }, { "epoch": 0.27, "learning_rate": 7.336453772207963e-05, "loss": 1.0239, "step": 106885 }, { "epoch": 0.27, "learning_rate": 7.336327948059792e-05, "loss": 1.0225, "step": 106890 }, { "epoch": 0.27, "learning_rate": 7.336202123911621e-05, "loss": 1.0244, "step": 106895 }, { "epoch": 0.27, "learning_rate": 7.336076299763452e-05, "loss": 1.0343, "step": 106900 }, { "epoch": 0.27, "learning_rate": 7.335950475615281e-05, "loss": 1.0235, "step": 106905 }, { "epoch": 0.27, "learning_rate": 7.33582465146711e-05, "loss": 1.0243, "step": 106910 }, { "epoch": 0.27, "learning_rate": 7.335698827318939e-05, "loss": 1.024, "step": 106915 }, { "epoch": 0.27, "learning_rate": 7.33557300317077e-05, "loss": 1.0232, "step": 106920 }, { "epoch": 0.27, "learning_rate": 7.335447179022599e-05, "loss": 1.0236, "step": 106925 }, { "epoch": 0.27, "learning_rate": 7.335321354874428e-05, "loss": 1.0251, "step": 106930 }, { "epoch": 0.27, "learning_rate": 7.335195530726257e-05, "loss": 1.0245, "step": 106935 }, { "epoch": 0.27, "learning_rate": 7.335069706578088e-05, "loss": 1.0234, "step": 106940 }, { "epoch": 0.27, "learning_rate": 7.334943882429917e-05, "loss": 1.0221, "step": 106945 }, { "epoch": 0.27, "learning_rate": 7.334818058281746e-05, "loss": 1.0239, "step": 106950 }, { "epoch": 0.27, "learning_rate": 7.334692234133575e-05, "loss": 1.0221, "step": 106955 }, { "epoch": 0.27, "learning_rate": 7.334566409985404e-05, "loss": 1.0232, "step": 106960 }, { "epoch": 0.27, "learning_rate": 7.334440585837235e-05, "loss": 1.0217, "step": 106965 }, { "epoch": 0.27, "learning_rate": 7.334314761689064e-05, "loss": 1.0219, "step": 106970 }, { "epoch": 0.27, "learning_rate": 7.334188937540893e-05, "loss": 1.0221, "step": 106975 }, { "epoch": 0.27, "learning_rate": 7.334063113392722e-05, "loss": 1.0218, "step": 106980 }, { "epoch": 0.27, "learning_rate": 7.333937289244553e-05, "loss": 1.023, "step": 106985 }, { "epoch": 0.27, "learning_rate": 7.333811465096382e-05, "loss": 1.0215, "step": 106990 }, { "epoch": 0.27, "learning_rate": 7.333685640948211e-05, "loss": 1.0219, "step": 106995 }, { "epoch": 0.27, "learning_rate": 7.33355981680004e-05, "loss": 1.0211, "step": 107000 }, { "epoch": 0.27, "learning_rate": 7.333433992651871e-05, "loss": 1.0235, "step": 107005 }, { "epoch": 0.27, "learning_rate": 7.3333081685037e-05, "loss": 1.0222, "step": 107010 }, { "epoch": 0.27, "learning_rate": 7.333182344355529e-05, "loss": 1.0244, "step": 107015 }, { "epoch": 0.27, "learning_rate": 7.333056520207358e-05, "loss": 1.0216, "step": 107020 }, { "epoch": 0.27, "learning_rate": 7.332930696059188e-05, "loss": 1.0227, "step": 107025 }, { "epoch": 0.27, "learning_rate": 7.332804871911018e-05, "loss": 1.043, "step": 107030 }, { "epoch": 0.27, "learning_rate": 7.332679047762847e-05, "loss": 1.0209, "step": 107035 }, { "epoch": 0.27, "learning_rate": 7.332553223614676e-05, "loss": 1.0219, "step": 107040 }, { "epoch": 0.27, "learning_rate": 7.332427399466506e-05, "loss": 1.0374, "step": 107045 }, { "epoch": 0.27, "learning_rate": 7.332301575318336e-05, "loss": 1.0234, "step": 107050 }, { "epoch": 0.27, "learning_rate": 7.332175751170165e-05, "loss": 1.0231, "step": 107055 }, { "epoch": 0.27, "learning_rate": 7.332049927021994e-05, "loss": 1.0242, "step": 107060 }, { "epoch": 0.27, "learning_rate": 7.331924102873824e-05, "loss": 1.0226, "step": 107065 }, { "epoch": 0.27, "learning_rate": 7.331798278725654e-05, "loss": 1.0233, "step": 107070 }, { "epoch": 0.27, "learning_rate": 7.331672454577483e-05, "loss": 1.0225, "step": 107075 }, { "epoch": 0.27, "learning_rate": 7.331546630429312e-05, "loss": 1.0232, "step": 107080 }, { "epoch": 0.27, "learning_rate": 7.331420806281142e-05, "loss": 1.0232, "step": 107085 }, { "epoch": 0.27, "learning_rate": 7.331294982132971e-05, "loss": 1.0234, "step": 107090 }, { "epoch": 0.27, "learning_rate": 7.331169157984801e-05, "loss": 1.0224, "step": 107095 }, { "epoch": 0.27, "learning_rate": 7.33104333383663e-05, "loss": 1.0258, "step": 107100 }, { "epoch": 0.27, "learning_rate": 7.33091750968846e-05, "loss": 1.0458, "step": 107105 }, { "epoch": 0.27, "learning_rate": 7.330791685540289e-05, "loss": 1.0438, "step": 107110 }, { "epoch": 0.27, "learning_rate": 7.330665861392119e-05, "loss": 1.0224, "step": 107115 }, { "epoch": 0.27, "learning_rate": 7.330540037243948e-05, "loss": 1.0233, "step": 107120 }, { "epoch": 0.27, "learning_rate": 7.330414213095778e-05, "loss": 1.0221, "step": 107125 }, { "epoch": 0.27, "learning_rate": 7.330288388947607e-05, "loss": 1.0211, "step": 107130 }, { "epoch": 0.27, "learning_rate": 7.330162564799437e-05, "loss": 1.0201, "step": 107135 }, { "epoch": 0.27, "learning_rate": 7.330036740651266e-05, "loss": 1.0215, "step": 107140 }, { "epoch": 0.27, "learning_rate": 7.329910916503096e-05, "loss": 1.0238, "step": 107145 }, { "epoch": 0.27, "learning_rate": 7.329785092354925e-05, "loss": 1.0199, "step": 107150 }, { "epoch": 0.27, "learning_rate": 7.329659268206754e-05, "loss": 1.0238, "step": 107155 }, { "epoch": 0.27, "learning_rate": 7.329533444058584e-05, "loss": 1.0224, "step": 107160 }, { "epoch": 0.27, "learning_rate": 7.329407619910413e-05, "loss": 1.0242, "step": 107165 }, { "epoch": 0.27, "learning_rate": 7.329281795762243e-05, "loss": 1.0473, "step": 107170 }, { "epoch": 0.27, "learning_rate": 7.329155971614072e-05, "loss": 1.0221, "step": 107175 }, { "epoch": 0.27, "learning_rate": 7.329030147465902e-05, "loss": 1.0216, "step": 107180 }, { "epoch": 0.27, "learning_rate": 7.328904323317731e-05, "loss": 1.0245, "step": 107185 }, { "epoch": 0.27, "learning_rate": 7.32877849916956e-05, "loss": 1.023, "step": 107190 }, { "epoch": 0.27, "learning_rate": 7.32865267502139e-05, "loss": 1.0212, "step": 107195 }, { "epoch": 0.27, "learning_rate": 7.32852685087322e-05, "loss": 1.0232, "step": 107200 }, { "epoch": 0.27, "learning_rate": 7.32840102672505e-05, "loss": 1.0222, "step": 107205 }, { "epoch": 0.27, "learning_rate": 7.328275202576879e-05, "loss": 1.0221, "step": 107210 }, { "epoch": 0.27, "learning_rate": 7.328149378428708e-05, "loss": 1.0218, "step": 107215 }, { "epoch": 0.27, "learning_rate": 7.328023554280537e-05, "loss": 1.0239, "step": 107220 }, { "epoch": 0.27, "learning_rate": 7.327897730132367e-05, "loss": 1.022, "step": 107225 }, { "epoch": 0.27, "learning_rate": 7.327771905984197e-05, "loss": 1.021, "step": 107230 }, { "epoch": 0.27, "learning_rate": 7.327646081836026e-05, "loss": 1.0228, "step": 107235 }, { "epoch": 0.27, "learning_rate": 7.327520257687855e-05, "loss": 1.0211, "step": 107240 }, { "epoch": 0.27, "learning_rate": 7.327394433539685e-05, "loss": 1.0224, "step": 107245 }, { "epoch": 0.27, "learning_rate": 7.327268609391515e-05, "loss": 1.0214, "step": 107250 }, { "epoch": 0.27, "learning_rate": 7.327142785243344e-05, "loss": 1.0242, "step": 107255 }, { "epoch": 0.27, "learning_rate": 7.327016961095173e-05, "loss": 1.0222, "step": 107260 }, { "epoch": 0.27, "learning_rate": 7.326891136947003e-05, "loss": 1.0213, "step": 107265 }, { "epoch": 0.27, "learning_rate": 7.326765312798833e-05, "loss": 1.0227, "step": 107270 }, { "epoch": 0.27, "learning_rate": 7.326639488650662e-05, "loss": 1.0248, "step": 107275 }, { "epoch": 0.27, "learning_rate": 7.326513664502491e-05, "loss": 1.0234, "step": 107280 }, { "epoch": 0.27, "learning_rate": 7.32638784035432e-05, "loss": 1.0206, "step": 107285 }, { "epoch": 0.27, "learning_rate": 7.32626201620615e-05, "loss": 1.0462, "step": 107290 }, { "epoch": 0.27, "learning_rate": 7.32613619205798e-05, "loss": 1.0212, "step": 107295 }, { "epoch": 0.27, "learning_rate": 7.326010367909809e-05, "loss": 1.0218, "step": 107300 }, { "epoch": 0.27, "learning_rate": 7.325884543761638e-05, "loss": 1.0224, "step": 107305 }, { "epoch": 0.27, "learning_rate": 7.325758719613469e-05, "loss": 1.0218, "step": 107310 }, { "epoch": 0.27, "learning_rate": 7.325632895465298e-05, "loss": 1.0205, "step": 107315 }, { "epoch": 0.27, "learning_rate": 7.325507071317127e-05, "loss": 1.0258, "step": 107320 }, { "epoch": 0.27, "learning_rate": 7.325381247168956e-05, "loss": 1.0248, "step": 107325 }, { "epoch": 0.27, "learning_rate": 7.325255423020787e-05, "loss": 1.0232, "step": 107330 }, { "epoch": 0.27, "learning_rate": 7.325129598872617e-05, "loss": 1.024, "step": 107335 }, { "epoch": 0.27, "learning_rate": 7.325003774724446e-05, "loss": 1.0228, "step": 107340 }, { "epoch": 0.27, "learning_rate": 7.324877950576275e-05, "loss": 1.0467, "step": 107345 }, { "epoch": 0.27, "learning_rate": 7.324752126428105e-05, "loss": 1.0233, "step": 107350 }, { "epoch": 0.27, "learning_rate": 7.324626302279934e-05, "loss": 1.0244, "step": 107355 }, { "epoch": 0.27, "learning_rate": 7.324500478131764e-05, "loss": 1.0227, "step": 107360 }, { "epoch": 0.27, "learning_rate": 7.324374653983593e-05, "loss": 1.0219, "step": 107365 }, { "epoch": 0.27, "learning_rate": 7.324248829835423e-05, "loss": 1.0223, "step": 107370 }, { "epoch": 0.27, "learning_rate": 7.324123005687252e-05, "loss": 1.0231, "step": 107375 }, { "epoch": 0.27, "learning_rate": 7.323997181539082e-05, "loss": 1.0229, "step": 107380 }, { "epoch": 0.27, "learning_rate": 7.323871357390911e-05, "loss": 1.021, "step": 107385 }, { "epoch": 0.27, "learning_rate": 7.32374553324274e-05, "loss": 1.0244, "step": 107390 }, { "epoch": 0.27, "learning_rate": 7.32361970909457e-05, "loss": 1.0239, "step": 107395 }, { "epoch": 0.27, "learning_rate": 7.3234938849464e-05, "loss": 1.0239, "step": 107400 }, { "epoch": 0.27, "learning_rate": 7.32336806079823e-05, "loss": 1.0227, "step": 107405 }, { "epoch": 0.27, "learning_rate": 7.323242236650059e-05, "loss": 1.0221, "step": 107410 }, { "epoch": 0.27, "learning_rate": 7.323116412501888e-05, "loss": 1.0233, "step": 107415 }, { "epoch": 0.27, "learning_rate": 7.322990588353717e-05, "loss": 1.0216, "step": 107420 }, { "epoch": 0.27, "learning_rate": 7.322864764205547e-05, "loss": 1.021, "step": 107425 }, { "epoch": 0.27, "learning_rate": 7.322738940057377e-05, "loss": 1.0244, "step": 107430 }, { "epoch": 0.27, "learning_rate": 7.322613115909206e-05, "loss": 1.022, "step": 107435 }, { "epoch": 0.27, "learning_rate": 7.322487291761035e-05, "loss": 1.0237, "step": 107440 }, { "epoch": 0.27, "learning_rate": 7.322361467612865e-05, "loss": 1.0219, "step": 107445 }, { "epoch": 0.27, "learning_rate": 7.322235643464695e-05, "loss": 1.0229, "step": 107450 }, { "epoch": 0.27, "learning_rate": 7.322109819316524e-05, "loss": 1.024, "step": 107455 }, { "epoch": 0.27, "learning_rate": 7.321983995168353e-05, "loss": 1.0217, "step": 107460 }, { "epoch": 0.27, "learning_rate": 7.321858171020183e-05, "loss": 1.0205, "step": 107465 }, { "epoch": 0.27, "learning_rate": 7.321732346872013e-05, "loss": 1.0225, "step": 107470 }, { "epoch": 0.27, "learning_rate": 7.321606522723842e-05, "loss": 1.0238, "step": 107475 }, { "epoch": 0.27, "learning_rate": 7.321480698575671e-05, "loss": 1.0467, "step": 107480 }, { "epoch": 0.27, "learning_rate": 7.3213548744275e-05, "loss": 1.0194, "step": 107485 }, { "epoch": 0.27, "learning_rate": 7.32122905027933e-05, "loss": 1.0222, "step": 107490 }, { "epoch": 0.27, "learning_rate": 7.32110322613116e-05, "loss": 1.0252, "step": 107495 }, { "epoch": 0.27, "learning_rate": 7.320977401982989e-05, "loss": 1.0235, "step": 107500 }, { "epoch": 0.27, "learning_rate": 7.320851577834818e-05, "loss": 1.0236, "step": 107505 }, { "epoch": 0.27, "learning_rate": 7.320725753686649e-05, "loss": 1.0227, "step": 107510 }, { "epoch": 0.27, "learning_rate": 7.320599929538478e-05, "loss": 1.023, "step": 107515 }, { "epoch": 0.27, "learning_rate": 7.320474105390307e-05, "loss": 1.0218, "step": 107520 }, { "epoch": 0.27, "learning_rate": 7.320348281242136e-05, "loss": 1.0223, "step": 107525 }, { "epoch": 0.27, "learning_rate": 7.320222457093967e-05, "loss": 1.0228, "step": 107530 }, { "epoch": 0.27, "learning_rate": 7.320096632945796e-05, "loss": 1.0209, "step": 107535 }, { "epoch": 0.27, "learning_rate": 7.319970808797625e-05, "loss": 1.0223, "step": 107540 }, { "epoch": 0.27, "learning_rate": 7.319844984649454e-05, "loss": 1.0227, "step": 107545 }, { "epoch": 0.27, "learning_rate": 7.319719160501283e-05, "loss": 1.0236, "step": 107550 }, { "epoch": 0.27, "learning_rate": 7.319593336353114e-05, "loss": 1.0249, "step": 107555 }, { "epoch": 0.27, "learning_rate": 7.319467512204943e-05, "loss": 1.0232, "step": 107560 }, { "epoch": 0.27, "learning_rate": 7.319341688056772e-05, "loss": 1.0236, "step": 107565 }, { "epoch": 0.27, "learning_rate": 7.319215863908601e-05, "loss": 1.0226, "step": 107570 }, { "epoch": 0.27, "learning_rate": 7.319090039760432e-05, "loss": 1.0238, "step": 107575 }, { "epoch": 0.27, "learning_rate": 7.318964215612261e-05, "loss": 1.0422, "step": 107580 }, { "epoch": 0.27, "learning_rate": 7.31883839146409e-05, "loss": 1.0232, "step": 107585 }, { "epoch": 0.27, "learning_rate": 7.318712567315919e-05, "loss": 1.0212, "step": 107590 }, { "epoch": 0.27, "learning_rate": 7.31858674316775e-05, "loss": 1.0237, "step": 107595 }, { "epoch": 0.27, "learning_rate": 7.318460919019579e-05, "loss": 1.0233, "step": 107600 }, { "epoch": 0.27, "learning_rate": 7.318335094871408e-05, "loss": 1.0243, "step": 107605 }, { "epoch": 0.27, "learning_rate": 7.318209270723237e-05, "loss": 1.0209, "step": 107610 }, { "epoch": 0.27, "learning_rate": 7.318083446575066e-05, "loss": 1.0251, "step": 107615 }, { "epoch": 0.27, "learning_rate": 7.317957622426897e-05, "loss": 1.0236, "step": 107620 }, { "epoch": 0.27, "learning_rate": 7.317831798278726e-05, "loss": 1.022, "step": 107625 }, { "epoch": 0.27, "learning_rate": 7.317705974130555e-05, "loss": 1.0222, "step": 107630 }, { "epoch": 0.27, "learning_rate": 7.317580149982384e-05, "loss": 1.043, "step": 107635 }, { "epoch": 0.27, "learning_rate": 7.317454325834215e-05, "loss": 1.0237, "step": 107640 }, { "epoch": 0.27, "learning_rate": 7.317328501686044e-05, "loss": 1.0242, "step": 107645 }, { "epoch": 0.27, "learning_rate": 7.317202677537873e-05, "loss": 1.0232, "step": 107650 }, { "epoch": 0.27, "learning_rate": 7.317076853389702e-05, "loss": 1.0219, "step": 107655 }, { "epoch": 0.27, "learning_rate": 7.316951029241533e-05, "loss": 1.0238, "step": 107660 }, { "epoch": 0.27, "learning_rate": 7.316825205093362e-05, "loss": 1.022, "step": 107665 }, { "epoch": 0.27, "learning_rate": 7.316699380945191e-05, "loss": 1.0217, "step": 107670 }, { "epoch": 0.27, "learning_rate": 7.31657355679702e-05, "loss": 1.0205, "step": 107675 }, { "epoch": 0.27, "learning_rate": 7.31644773264885e-05, "loss": 1.0215, "step": 107680 }, { "epoch": 0.27, "learning_rate": 7.31632190850068e-05, "loss": 1.0222, "step": 107685 }, { "epoch": 0.27, "learning_rate": 7.316196084352509e-05, "loss": 1.0209, "step": 107690 }, { "epoch": 0.27, "learning_rate": 7.316070260204338e-05, "loss": 1.0215, "step": 107695 }, { "epoch": 0.27, "learning_rate": 7.315944436056167e-05, "loss": 1.0207, "step": 107700 }, { "epoch": 0.27, "learning_rate": 7.315818611907998e-05, "loss": 1.0234, "step": 107705 }, { "epoch": 0.27, "learning_rate": 7.315692787759827e-05, "loss": 1.0227, "step": 107710 }, { "epoch": 0.27, "learning_rate": 7.315566963611656e-05, "loss": 1.0234, "step": 107715 }, { "epoch": 0.27, "learning_rate": 7.315441139463485e-05, "loss": 1.0223, "step": 107720 }, { "epoch": 0.27, "learning_rate": 7.315315315315316e-05, "loss": 1.0231, "step": 107725 }, { "epoch": 0.27, "learning_rate": 7.315189491167145e-05, "loss": 1.023, "step": 107730 }, { "epoch": 0.27, "learning_rate": 7.315063667018974e-05, "loss": 1.0227, "step": 107735 }, { "epoch": 0.27, "learning_rate": 7.314937842870803e-05, "loss": 1.0216, "step": 107740 }, { "epoch": 0.27, "learning_rate": 7.314812018722633e-05, "loss": 1.0208, "step": 107745 }, { "epoch": 0.27, "learning_rate": 7.314686194574463e-05, "loss": 1.0211, "step": 107750 }, { "epoch": 0.27, "learning_rate": 7.314560370426292e-05, "loss": 1.0226, "step": 107755 }, { "epoch": 0.27, "learning_rate": 7.314434546278121e-05, "loss": 1.0229, "step": 107760 }, { "epoch": 0.27, "learning_rate": 7.31430872212995e-05, "loss": 1.0236, "step": 107765 }, { "epoch": 0.27, "learning_rate": 7.314182897981781e-05, "loss": 1.0248, "step": 107770 }, { "epoch": 0.27, "learning_rate": 7.31405707383361e-05, "loss": 1.0244, "step": 107775 }, { "epoch": 0.27, "learning_rate": 7.31393124968544e-05, "loss": 1.0212, "step": 107780 }, { "epoch": 0.27, "learning_rate": 7.313805425537269e-05, "loss": 1.0213, "step": 107785 }, { "epoch": 0.27, "learning_rate": 7.313679601389099e-05, "loss": 1.0237, "step": 107790 }, { "epoch": 0.27, "learning_rate": 7.313553777240928e-05, "loss": 1.021, "step": 107795 }, { "epoch": 0.27, "learning_rate": 7.313427953092757e-05, "loss": 1.0207, "step": 107800 }, { "epoch": 0.27, "learning_rate": 7.313302128944587e-05, "loss": 1.0446, "step": 107805 }, { "epoch": 0.27, "learning_rate": 7.313176304796416e-05, "loss": 1.0238, "step": 107810 }, { "epoch": 0.27, "learning_rate": 7.313050480648246e-05, "loss": 1.0475, "step": 107815 }, { "epoch": 0.27, "learning_rate": 7.312924656500075e-05, "loss": 1.0246, "step": 107820 }, { "epoch": 0.27, "learning_rate": 7.312798832351906e-05, "loss": 1.0235, "step": 107825 }, { "epoch": 0.27, "learning_rate": 7.312673008203735e-05, "loss": 1.0229, "step": 107830 }, { "epoch": 0.27, "learning_rate": 7.312547184055564e-05, "loss": 1.0406, "step": 107835 }, { "epoch": 0.27, "learning_rate": 7.312421359907395e-05, "loss": 1.0238, "step": 107840 }, { "epoch": 0.27, "learning_rate": 7.312295535759224e-05, "loss": 1.0218, "step": 107845 }, { "epoch": 0.27, "learning_rate": 7.312169711611053e-05, "loss": 1.0213, "step": 107850 }, { "epoch": 0.27, "learning_rate": 7.312043887462882e-05, "loss": 1.023, "step": 107855 }, { "epoch": 0.27, "learning_rate": 7.311918063314713e-05, "loss": 1.0237, "step": 107860 }, { "epoch": 0.27, "learning_rate": 7.311792239166542e-05, "loss": 1.0211, "step": 107865 }, { "epoch": 0.27, "learning_rate": 7.311666415018371e-05, "loss": 1.0219, "step": 107870 }, { "epoch": 0.27, "learning_rate": 7.3115405908702e-05, "loss": 1.022, "step": 107875 }, { "epoch": 0.27, "learning_rate": 7.31141476672203e-05, "loss": 1.0416, "step": 107880 }, { "epoch": 0.27, "learning_rate": 7.31128894257386e-05, "loss": 1.0211, "step": 107885 }, { "epoch": 0.27, "learning_rate": 7.311163118425689e-05, "loss": 1.0246, "step": 107890 }, { "epoch": 0.27, "learning_rate": 7.311037294277518e-05, "loss": 1.0208, "step": 107895 }, { "epoch": 0.27, "learning_rate": 7.310911470129347e-05, "loss": 1.0211, "step": 107900 }, { "epoch": 0.27, "learning_rate": 7.310785645981178e-05, "loss": 1.0248, "step": 107905 }, { "epoch": 0.27, "learning_rate": 7.310659821833007e-05, "loss": 1.023, "step": 107910 }, { "epoch": 0.27, "learning_rate": 7.310533997684836e-05, "loss": 1.0207, "step": 107915 }, { "epoch": 0.27, "learning_rate": 7.310408173536665e-05, "loss": 1.0211, "step": 107920 }, { "epoch": 0.27, "learning_rate": 7.310282349388496e-05, "loss": 1.0219, "step": 107925 }, { "epoch": 0.27, "learning_rate": 7.310156525240325e-05, "loss": 1.0233, "step": 107930 }, { "epoch": 0.27, "learning_rate": 7.310030701092154e-05, "loss": 1.0211, "step": 107935 }, { "epoch": 0.27, "learning_rate": 7.309904876943983e-05, "loss": 1.0232, "step": 107940 }, { "epoch": 0.27, "learning_rate": 7.309779052795813e-05, "loss": 1.0233, "step": 107945 }, { "epoch": 0.27, "learning_rate": 7.309653228647643e-05, "loss": 1.0221, "step": 107950 }, { "epoch": 0.27, "learning_rate": 7.309527404499472e-05, "loss": 1.0239, "step": 107955 }, { "epoch": 0.27, "learning_rate": 7.309401580351301e-05, "loss": 1.0209, "step": 107960 }, { "epoch": 0.27, "learning_rate": 7.30927575620313e-05, "loss": 1.0215, "step": 107965 }, { "epoch": 0.27, "learning_rate": 7.309149932054961e-05, "loss": 1.0238, "step": 107970 }, { "epoch": 0.27, "learning_rate": 7.30902410790679e-05, "loss": 1.021, "step": 107975 }, { "epoch": 0.27, "learning_rate": 7.30889828375862e-05, "loss": 1.0429, "step": 107980 }, { "epoch": 0.27, "learning_rate": 7.308772459610449e-05, "loss": 1.0235, "step": 107985 }, { "epoch": 0.27, "learning_rate": 7.308646635462279e-05, "loss": 1.0197, "step": 107990 }, { "epoch": 0.27, "learning_rate": 7.308520811314108e-05, "loss": 1.0232, "step": 107995 }, { "epoch": 0.27, "learning_rate": 7.308394987165937e-05, "loss": 1.0222, "step": 108000 }, { "epoch": 0.27, "learning_rate": 7.308269163017767e-05, "loss": 1.0269, "step": 108005 }, { "epoch": 0.27, "learning_rate": 7.308143338869596e-05, "loss": 1.0211, "step": 108010 }, { "epoch": 0.27, "learning_rate": 7.308017514721426e-05, "loss": 1.0231, "step": 108015 }, { "epoch": 0.27, "learning_rate": 7.307891690573255e-05, "loss": 1.0236, "step": 108020 }, { "epoch": 0.27, "learning_rate": 7.307765866425085e-05, "loss": 1.0216, "step": 108025 }, { "epoch": 0.27, "learning_rate": 7.307640042276914e-05, "loss": 1.0229, "step": 108030 }, { "epoch": 0.27, "learning_rate": 7.307514218128744e-05, "loss": 1.0229, "step": 108035 }, { "epoch": 0.27, "learning_rate": 7.307388393980573e-05, "loss": 1.0213, "step": 108040 }, { "epoch": 0.27, "learning_rate": 7.307262569832403e-05, "loss": 1.0409, "step": 108045 }, { "epoch": 0.27, "learning_rate": 7.307136745684232e-05, "loss": 1.0233, "step": 108050 }, { "epoch": 0.27, "learning_rate": 7.307010921536062e-05, "loss": 1.0229, "step": 108055 }, { "epoch": 0.27, "learning_rate": 7.306885097387891e-05, "loss": 1.024, "step": 108060 }, { "epoch": 0.27, "learning_rate": 7.30675927323972e-05, "loss": 1.0237, "step": 108065 }, { "epoch": 0.27, "learning_rate": 7.30663344909155e-05, "loss": 1.0233, "step": 108070 }, { "epoch": 0.27, "learning_rate": 7.306507624943379e-05, "loss": 1.0221, "step": 108075 }, { "epoch": 0.27, "learning_rate": 7.30638180079521e-05, "loss": 1.022, "step": 108080 }, { "epoch": 0.27, "learning_rate": 7.306255976647039e-05, "loss": 1.0244, "step": 108085 }, { "epoch": 0.27, "learning_rate": 7.306130152498868e-05, "loss": 1.0239, "step": 108090 }, { "epoch": 0.27, "learning_rate": 7.306004328350697e-05, "loss": 1.0232, "step": 108095 }, { "epoch": 0.27, "learning_rate": 7.305878504202527e-05, "loss": 1.0239, "step": 108100 }, { "epoch": 0.27, "learning_rate": 7.305752680054357e-05, "loss": 1.024, "step": 108105 }, { "epoch": 0.27, "learning_rate": 7.305626855906186e-05, "loss": 1.0238, "step": 108110 }, { "epoch": 0.27, "learning_rate": 7.305501031758015e-05, "loss": 1.0212, "step": 108115 }, { "epoch": 0.27, "learning_rate": 7.305375207609845e-05, "loss": 1.0232, "step": 108120 }, { "epoch": 0.27, "learning_rate": 7.305249383461675e-05, "loss": 1.0235, "step": 108125 }, { "epoch": 0.27, "learning_rate": 7.305123559313504e-05, "loss": 1.024, "step": 108130 }, { "epoch": 0.27, "learning_rate": 7.304997735165333e-05, "loss": 1.0234, "step": 108135 }, { "epoch": 0.27, "learning_rate": 7.304871911017162e-05, "loss": 1.0214, "step": 108140 }, { "epoch": 0.27, "learning_rate": 7.304746086868993e-05, "loss": 1.0228, "step": 108145 }, { "epoch": 0.27, "learning_rate": 7.304620262720822e-05, "loss": 1.0212, "step": 108150 }, { "epoch": 0.27, "learning_rate": 7.304494438572651e-05, "loss": 1.0226, "step": 108155 }, { "epoch": 0.27, "learning_rate": 7.30436861442448e-05, "loss": 1.0231, "step": 108160 }, { "epoch": 0.27, "learning_rate": 7.30424279027631e-05, "loss": 1.0246, "step": 108165 }, { "epoch": 0.27, "learning_rate": 7.30411696612814e-05, "loss": 1.0218, "step": 108170 }, { "epoch": 0.27, "learning_rate": 7.303991141979969e-05, "loss": 1.0226, "step": 108175 }, { "epoch": 0.27, "learning_rate": 7.303865317831798e-05, "loss": 1.0228, "step": 108180 }, { "epoch": 0.27, "learning_rate": 7.303739493683628e-05, "loss": 1.0226, "step": 108185 }, { "epoch": 0.27, "learning_rate": 7.303613669535458e-05, "loss": 1.0233, "step": 108190 }, { "epoch": 0.27, "learning_rate": 7.303487845387287e-05, "loss": 1.0218, "step": 108195 }, { "epoch": 0.27, "learning_rate": 7.303362021239116e-05, "loss": 1.0242, "step": 108200 }, { "epoch": 0.27, "learning_rate": 7.303236197090945e-05, "loss": 1.0237, "step": 108205 }, { "epoch": 0.27, "learning_rate": 7.303110372942776e-05, "loss": 1.0216, "step": 108210 }, { "epoch": 0.27, "learning_rate": 7.302984548794605e-05, "loss": 1.022, "step": 108215 }, { "epoch": 0.27, "learning_rate": 7.302858724646434e-05, "loss": 1.0236, "step": 108220 }, { "epoch": 0.27, "learning_rate": 7.302732900498263e-05, "loss": 1.0241, "step": 108225 }, { "epoch": 0.27, "learning_rate": 7.302607076350094e-05, "loss": 1.0409, "step": 108230 }, { "epoch": 0.27, "learning_rate": 7.302481252201923e-05, "loss": 1.0212, "step": 108235 }, { "epoch": 0.27, "learning_rate": 7.302355428053752e-05, "loss": 1.0229, "step": 108240 }, { "epoch": 0.27, "learning_rate": 7.302229603905581e-05, "loss": 1.0211, "step": 108245 }, { "epoch": 0.27, "learning_rate": 7.302103779757412e-05, "loss": 1.0207, "step": 108250 }, { "epoch": 0.27, "learning_rate": 7.301977955609241e-05, "loss": 1.0235, "step": 108255 }, { "epoch": 0.27, "learning_rate": 7.30185213146107e-05, "loss": 1.0228, "step": 108260 }, { "epoch": 0.27, "learning_rate": 7.301726307312899e-05, "loss": 1.0214, "step": 108265 }, { "epoch": 0.27, "learning_rate": 7.301600483164728e-05, "loss": 1.0221, "step": 108270 }, { "epoch": 0.27, "learning_rate": 7.301474659016559e-05, "loss": 1.0192, "step": 108275 }, { "epoch": 0.27, "learning_rate": 7.301348834868388e-05, "loss": 1.0207, "step": 108280 }, { "epoch": 0.27, "learning_rate": 7.301223010720217e-05, "loss": 1.0219, "step": 108285 }, { "epoch": 0.27, "learning_rate": 7.301097186572046e-05, "loss": 1.0215, "step": 108290 }, { "epoch": 0.27, "learning_rate": 7.300971362423877e-05, "loss": 1.0208, "step": 108295 }, { "epoch": 0.27, "learning_rate": 7.300845538275706e-05, "loss": 1.0252, "step": 108300 }, { "epoch": 0.27, "learning_rate": 7.300719714127535e-05, "loss": 1.0225, "step": 108305 }, { "epoch": 0.27, "learning_rate": 7.300593889979364e-05, "loss": 1.0211, "step": 108310 }, { "epoch": 0.27, "learning_rate": 7.300468065831195e-05, "loss": 1.0227, "step": 108315 }, { "epoch": 0.27, "learning_rate": 7.300342241683024e-05, "loss": 1.0205, "step": 108320 }, { "epoch": 0.27, "learning_rate": 7.300216417534854e-05, "loss": 1.0239, "step": 108325 }, { "epoch": 0.27, "learning_rate": 7.300090593386684e-05, "loss": 1.0219, "step": 108330 }, { "epoch": 0.27, "learning_rate": 7.299964769238513e-05, "loss": 1.0218, "step": 108335 }, { "epoch": 0.27, "learning_rate": 7.299838945090342e-05, "loss": 1.0239, "step": 108340 }, { "epoch": 0.27, "learning_rate": 7.299713120942172e-05, "loss": 1.0236, "step": 108345 }, { "epoch": 0.27, "learning_rate": 7.299587296794002e-05, "loss": 1.0233, "step": 108350 }, { "epoch": 0.27, "learning_rate": 7.299461472645831e-05, "loss": 1.0226, "step": 108355 }, { "epoch": 0.27, "learning_rate": 7.29933564849766e-05, "loss": 1.0238, "step": 108360 }, { "epoch": 0.27, "learning_rate": 7.29920982434949e-05, "loss": 1.0228, "step": 108365 }, { "epoch": 0.27, "learning_rate": 7.29908400020132e-05, "loss": 1.0234, "step": 108370 }, { "epoch": 0.27, "learning_rate": 7.298958176053149e-05, "loss": 1.0234, "step": 108375 }, { "epoch": 0.27, "learning_rate": 7.298832351904978e-05, "loss": 1.0237, "step": 108380 }, { "epoch": 0.27, "learning_rate": 7.298706527756808e-05, "loss": 1.022, "step": 108385 }, { "epoch": 0.27, "learning_rate": 7.298580703608638e-05, "loss": 1.0239, "step": 108390 }, { "epoch": 0.27, "learning_rate": 7.298454879460467e-05, "loss": 1.025, "step": 108395 }, { "epoch": 0.27, "learning_rate": 7.298329055312296e-05, "loss": 1.0219, "step": 108400 }, { "epoch": 0.27, "learning_rate": 7.298203231164125e-05, "loss": 1.0223, "step": 108405 }, { "epoch": 0.27, "learning_rate": 7.298077407015956e-05, "loss": 1.0241, "step": 108410 }, { "epoch": 0.27, "learning_rate": 7.297951582867785e-05, "loss": 1.0245, "step": 108415 }, { "epoch": 0.27, "learning_rate": 7.297825758719614e-05, "loss": 1.022, "step": 108420 }, { "epoch": 0.27, "learning_rate": 7.297699934571443e-05, "loss": 1.022, "step": 108425 }, { "epoch": 0.27, "learning_rate": 7.297574110423274e-05, "loss": 1.0227, "step": 108430 }, { "epoch": 0.27, "learning_rate": 7.297448286275103e-05, "loss": 1.0237, "step": 108435 }, { "epoch": 0.27, "learning_rate": 7.297322462126932e-05, "loss": 1.0227, "step": 108440 }, { "epoch": 0.27, "learning_rate": 7.297196637978761e-05, "loss": 1.0205, "step": 108445 }, { "epoch": 0.27, "learning_rate": 7.297070813830592e-05, "loss": 1.0222, "step": 108450 }, { "epoch": 0.27, "learning_rate": 7.296944989682421e-05, "loss": 1.0239, "step": 108455 }, { "epoch": 0.27, "learning_rate": 7.29681916553425e-05, "loss": 1.0237, "step": 108460 }, { "epoch": 0.27, "learning_rate": 7.296693341386079e-05, "loss": 1.0246, "step": 108465 }, { "epoch": 0.27, "learning_rate": 7.296567517237908e-05, "loss": 1.0232, "step": 108470 }, { "epoch": 0.27, "learning_rate": 7.296441693089739e-05, "loss": 1.0243, "step": 108475 }, { "epoch": 0.27, "learning_rate": 7.296315868941568e-05, "loss": 1.023, "step": 108480 }, { "epoch": 0.27, "learning_rate": 7.296190044793397e-05, "loss": 1.0224, "step": 108485 }, { "epoch": 0.27, "learning_rate": 7.296064220645226e-05, "loss": 1.0309, "step": 108490 }, { "epoch": 0.27, "learning_rate": 7.295938396497057e-05, "loss": 1.0207, "step": 108495 }, { "epoch": 0.27, "learning_rate": 7.295812572348886e-05, "loss": 1.0243, "step": 108500 }, { "epoch": 0.27, "learning_rate": 7.295686748200715e-05, "loss": 1.0215, "step": 108505 }, { "epoch": 0.27, "learning_rate": 7.295560924052544e-05, "loss": 1.0241, "step": 108510 }, { "epoch": 0.27, "learning_rate": 7.295435099904375e-05, "loss": 1.0227, "step": 108515 }, { "epoch": 0.27, "learning_rate": 7.295309275756204e-05, "loss": 1.0229, "step": 108520 }, { "epoch": 0.27, "learning_rate": 7.295183451608033e-05, "loss": 1.0224, "step": 108525 }, { "epoch": 0.27, "learning_rate": 7.295057627459862e-05, "loss": 1.024, "step": 108530 }, { "epoch": 0.27, "learning_rate": 7.294931803311691e-05, "loss": 1.0235, "step": 108535 }, { "epoch": 0.27, "learning_rate": 7.294805979163522e-05, "loss": 1.0242, "step": 108540 }, { "epoch": 0.27, "learning_rate": 7.294680155015351e-05, "loss": 1.023, "step": 108545 }, { "epoch": 0.27, "learning_rate": 7.29455433086718e-05, "loss": 1.0212, "step": 108550 }, { "epoch": 0.27, "learning_rate": 7.29442850671901e-05, "loss": 1.0216, "step": 108555 }, { "epoch": 0.27, "learning_rate": 7.29430268257084e-05, "loss": 1.0186, "step": 108560 }, { "epoch": 0.27, "learning_rate": 7.294176858422669e-05, "loss": 1.0227, "step": 108565 }, { "epoch": 0.27, "learning_rate": 7.294051034274498e-05, "loss": 1.0217, "step": 108570 }, { "epoch": 0.27, "learning_rate": 7.293925210126327e-05, "loss": 1.0219, "step": 108575 }, { "epoch": 0.27, "learning_rate": 7.293799385978158e-05, "loss": 1.0217, "step": 108580 }, { "epoch": 0.27, "learning_rate": 7.293673561829987e-05, "loss": 1.0235, "step": 108585 }, { "epoch": 0.27, "learning_rate": 7.293547737681816e-05, "loss": 1.0234, "step": 108590 }, { "epoch": 0.27, "learning_rate": 7.293421913533645e-05, "loss": 1.0221, "step": 108595 }, { "epoch": 0.27, "learning_rate": 7.293296089385475e-05, "loss": 1.0234, "step": 108600 }, { "epoch": 0.27, "learning_rate": 7.293170265237305e-05, "loss": 1.0227, "step": 108605 }, { "epoch": 0.27, "learning_rate": 7.293044441089134e-05, "loss": 1.0213, "step": 108610 }, { "epoch": 0.27, "learning_rate": 7.292918616940963e-05, "loss": 1.0227, "step": 108615 }, { "epoch": 0.27, "learning_rate": 7.292792792792792e-05, "loss": 1.0221, "step": 108620 }, { "epoch": 0.27, "learning_rate": 7.292666968644623e-05, "loss": 1.0215, "step": 108625 }, { "epoch": 0.27, "learning_rate": 7.292541144496452e-05, "loss": 1.023, "step": 108630 }, { "epoch": 0.27, "learning_rate": 7.292415320348281e-05, "loss": 1.0215, "step": 108635 }, { "epoch": 0.27, "learning_rate": 7.29228949620011e-05, "loss": 1.0207, "step": 108640 }, { "epoch": 0.27, "learning_rate": 7.292163672051941e-05, "loss": 1.0224, "step": 108645 }, { "epoch": 0.27, "learning_rate": 7.29203784790377e-05, "loss": 1.0229, "step": 108650 }, { "epoch": 0.27, "learning_rate": 7.2919120237556e-05, "loss": 1.0225, "step": 108655 }, { "epoch": 0.27, "learning_rate": 7.291786199607428e-05, "loss": 1.0236, "step": 108660 }, { "epoch": 0.27, "learning_rate": 7.291660375459258e-05, "loss": 1.0235, "step": 108665 }, { "epoch": 0.27, "learning_rate": 7.291534551311088e-05, "loss": 1.0217, "step": 108670 }, { "epoch": 0.27, "learning_rate": 7.291408727162917e-05, "loss": 1.023, "step": 108675 }, { "epoch": 0.27, "learning_rate": 7.291282903014746e-05, "loss": 1.0224, "step": 108680 }, { "epoch": 0.27, "learning_rate": 7.291157078866576e-05, "loss": 1.0223, "step": 108685 }, { "epoch": 0.27, "learning_rate": 7.291031254718406e-05, "loss": 1.0263, "step": 108690 }, { "epoch": 0.27, "learning_rate": 7.290905430570235e-05, "loss": 1.024, "step": 108695 }, { "epoch": 0.27, "learning_rate": 7.290779606422064e-05, "loss": 1.0207, "step": 108700 }, { "epoch": 0.27, "learning_rate": 7.290653782273894e-05, "loss": 1.0239, "step": 108705 }, { "epoch": 0.27, "learning_rate": 7.290527958125724e-05, "loss": 1.0238, "step": 108710 }, { "epoch": 0.27, "learning_rate": 7.290402133977553e-05, "loss": 1.0226, "step": 108715 }, { "epoch": 0.27, "learning_rate": 7.290276309829382e-05, "loss": 1.023, "step": 108720 }, { "epoch": 0.27, "learning_rate": 7.290150485681212e-05, "loss": 1.0227, "step": 108725 }, { "epoch": 0.27, "learning_rate": 7.290024661533041e-05, "loss": 1.0229, "step": 108730 }, { "epoch": 0.27, "learning_rate": 7.289898837384871e-05, "loss": 1.0227, "step": 108735 }, { "epoch": 0.27, "learning_rate": 7.2897730132367e-05, "loss": 1.0207, "step": 108740 }, { "epoch": 0.27, "learning_rate": 7.28964718908853e-05, "loss": 1.0228, "step": 108745 }, { "epoch": 0.27, "learning_rate": 7.289521364940359e-05, "loss": 1.0233, "step": 108750 }, { "epoch": 0.27, "learning_rate": 7.289395540792189e-05, "loss": 1.0226, "step": 108755 }, { "epoch": 0.27, "learning_rate": 7.289269716644018e-05, "loss": 1.022, "step": 108760 }, { "epoch": 0.27, "learning_rate": 7.289143892495848e-05, "loss": 1.023, "step": 108765 }, { "epoch": 0.27, "learning_rate": 7.289018068347677e-05, "loss": 1.0456, "step": 108770 }, { "epoch": 0.27, "learning_rate": 7.288892244199507e-05, "loss": 1.0224, "step": 108775 }, { "epoch": 0.27, "learning_rate": 7.288766420051336e-05, "loss": 1.0238, "step": 108780 }, { "epoch": 0.27, "learning_rate": 7.288640595903166e-05, "loss": 1.0241, "step": 108785 }, { "epoch": 0.27, "learning_rate": 7.288514771754995e-05, "loss": 1.0203, "step": 108790 }, { "epoch": 0.27, "learning_rate": 7.288388947606824e-05, "loss": 1.022, "step": 108795 }, { "epoch": 0.27, "learning_rate": 7.288263123458654e-05, "loss": 1.0227, "step": 108800 }, { "epoch": 0.27, "learning_rate": 7.288137299310484e-05, "loss": 1.0224, "step": 108805 }, { "epoch": 0.27, "learning_rate": 7.288011475162313e-05, "loss": 1.021, "step": 108810 }, { "epoch": 0.27, "learning_rate": 7.287885651014142e-05, "loss": 1.023, "step": 108815 }, { "epoch": 0.27, "learning_rate": 7.287759826865972e-05, "loss": 1.022, "step": 108820 }, { "epoch": 0.27, "learning_rate": 7.287634002717803e-05, "loss": 1.0217, "step": 108825 }, { "epoch": 0.27, "learning_rate": 7.287508178569632e-05, "loss": 1.0229, "step": 108830 }, { "epoch": 0.27, "learning_rate": 7.287382354421461e-05, "loss": 1.0229, "step": 108835 }, { "epoch": 0.27, "learning_rate": 7.28725653027329e-05, "loss": 1.022, "step": 108840 }, { "epoch": 0.27, "learning_rate": 7.287130706125121e-05, "loss": 1.0239, "step": 108845 }, { "epoch": 0.27, "learning_rate": 7.28700488197695e-05, "loss": 1.0214, "step": 108850 }, { "epoch": 0.27, "learning_rate": 7.286879057828779e-05, "loss": 1.0239, "step": 108855 }, { "epoch": 0.27, "learning_rate": 7.286753233680608e-05, "loss": 1.0214, "step": 108860 }, { "epoch": 0.27, "learning_rate": 7.286627409532439e-05, "loss": 1.0241, "step": 108865 }, { "epoch": 0.27, "learning_rate": 7.286501585384268e-05, "loss": 1.0227, "step": 108870 }, { "epoch": 0.27, "learning_rate": 7.286375761236097e-05, "loss": 1.0246, "step": 108875 }, { "epoch": 0.27, "learning_rate": 7.286249937087926e-05, "loss": 1.0231, "step": 108880 }, { "epoch": 0.27, "learning_rate": 7.286124112939756e-05, "loss": 1.0223, "step": 108885 }, { "epoch": 0.27, "learning_rate": 7.285998288791586e-05, "loss": 1.0219, "step": 108890 }, { "epoch": 0.27, "learning_rate": 7.285872464643415e-05, "loss": 1.0237, "step": 108895 }, { "epoch": 0.27, "learning_rate": 7.285746640495244e-05, "loss": 1.0229, "step": 108900 }, { "epoch": 0.27, "learning_rate": 7.285620816347074e-05, "loss": 1.0234, "step": 108905 }, { "epoch": 0.27, "learning_rate": 7.285494992198904e-05, "loss": 1.0235, "step": 108910 }, { "epoch": 0.27, "learning_rate": 7.285369168050733e-05, "loss": 1.0211, "step": 108915 }, { "epoch": 0.27, "learning_rate": 7.285243343902562e-05, "loss": 1.0232, "step": 108920 }, { "epoch": 0.27, "learning_rate": 7.285117519754392e-05, "loss": 1.0219, "step": 108925 }, { "epoch": 0.27, "learning_rate": 7.284991695606222e-05, "loss": 1.0268, "step": 108930 }, { "epoch": 0.27, "learning_rate": 7.284865871458051e-05, "loss": 1.0242, "step": 108935 }, { "epoch": 0.27, "learning_rate": 7.28474004730988e-05, "loss": 1.0212, "step": 108940 }, { "epoch": 0.27, "learning_rate": 7.28461422316171e-05, "loss": 1.022, "step": 108945 }, { "epoch": 0.27, "learning_rate": 7.284488399013539e-05, "loss": 1.0224, "step": 108950 }, { "epoch": 0.27, "learning_rate": 7.284362574865369e-05, "loss": 1.0198, "step": 108955 }, { "epoch": 0.27, "learning_rate": 7.284236750717198e-05, "loss": 1.0271, "step": 108960 }, { "epoch": 0.27, "learning_rate": 7.284110926569028e-05, "loss": 1.0253, "step": 108965 }, { "epoch": 0.27, "learning_rate": 7.283985102420857e-05, "loss": 1.0235, "step": 108970 }, { "epoch": 0.27, "learning_rate": 7.283859278272687e-05, "loss": 1.024, "step": 108975 }, { "epoch": 0.27, "learning_rate": 7.283733454124516e-05, "loss": 1.0232, "step": 108980 }, { "epoch": 0.27, "learning_rate": 7.283607629976346e-05, "loss": 1.0224, "step": 108985 }, { "epoch": 0.27, "learning_rate": 7.283481805828175e-05, "loss": 1.0237, "step": 108990 }, { "epoch": 0.27, "learning_rate": 7.283355981680004e-05, "loss": 1.0234, "step": 108995 }, { "epoch": 0.27, "learning_rate": 7.283230157531834e-05, "loss": 1.0221, "step": 109000 }, { "epoch": 0.27, "learning_rate": 7.283104333383664e-05, "loss": 1.0219, "step": 109005 }, { "epoch": 0.27, "learning_rate": 7.282978509235493e-05, "loss": 1.0236, "step": 109010 }, { "epoch": 0.27, "learning_rate": 7.282852685087322e-05, "loss": 1.0215, "step": 109015 }, { "epoch": 0.27, "learning_rate": 7.282726860939152e-05, "loss": 1.021, "step": 109020 }, { "epoch": 0.27, "learning_rate": 7.282601036790982e-05, "loss": 1.022, "step": 109025 }, { "epoch": 0.27, "learning_rate": 7.282475212642811e-05, "loss": 1.0232, "step": 109030 }, { "epoch": 0.27, "learning_rate": 7.28234938849464e-05, "loss": 1.0201, "step": 109035 }, { "epoch": 0.27, "learning_rate": 7.28222356434647e-05, "loss": 1.043, "step": 109040 }, { "epoch": 0.27, "learning_rate": 7.2820977401983e-05, "loss": 1.0407, "step": 109045 }, { "epoch": 0.27, "learning_rate": 7.281971916050129e-05, "loss": 1.0211, "step": 109050 }, { "epoch": 0.27, "learning_rate": 7.281846091901958e-05, "loss": 1.0217, "step": 109055 }, { "epoch": 0.27, "learning_rate": 7.281720267753787e-05, "loss": 1.021, "step": 109060 }, { "epoch": 0.27, "learning_rate": 7.281594443605618e-05, "loss": 1.0213, "step": 109065 }, { "epoch": 0.27, "learning_rate": 7.281468619457447e-05, "loss": 1.0235, "step": 109070 }, { "epoch": 0.27, "learning_rate": 7.281342795309276e-05, "loss": 1.024, "step": 109075 }, { "epoch": 0.27, "learning_rate": 7.281216971161105e-05, "loss": 1.0214, "step": 109080 }, { "epoch": 0.27, "learning_rate": 7.281091147012936e-05, "loss": 1.0241, "step": 109085 }, { "epoch": 0.27, "learning_rate": 7.280965322864765e-05, "loss": 1.0228, "step": 109090 }, { "epoch": 0.27, "learning_rate": 7.280839498716594e-05, "loss": 1.0216, "step": 109095 }, { "epoch": 0.27, "learning_rate": 7.280713674568423e-05, "loss": 1.023, "step": 109100 }, { "epoch": 0.27, "learning_rate": 7.280587850420254e-05, "loss": 1.0234, "step": 109105 }, { "epoch": 0.27, "learning_rate": 7.280462026272083e-05, "loss": 1.0229, "step": 109110 }, { "epoch": 0.27, "learning_rate": 7.280336202123912e-05, "loss": 1.0229, "step": 109115 }, { "epoch": 0.27, "learning_rate": 7.280210377975741e-05, "loss": 1.021, "step": 109120 }, { "epoch": 0.27, "learning_rate": 7.28008455382757e-05, "loss": 1.0222, "step": 109125 }, { "epoch": 0.27, "learning_rate": 7.2799587296794e-05, "loss": 1.0221, "step": 109130 }, { "epoch": 0.27, "learning_rate": 7.27983290553123e-05, "loss": 1.0221, "step": 109135 }, { "epoch": 0.27, "learning_rate": 7.279707081383059e-05, "loss": 1.0205, "step": 109140 }, { "epoch": 0.27, "learning_rate": 7.279581257234888e-05, "loss": 1.0221, "step": 109145 }, { "epoch": 0.27, "learning_rate": 7.279455433086719e-05, "loss": 1.0238, "step": 109150 }, { "epoch": 0.27, "learning_rate": 7.279329608938548e-05, "loss": 1.0242, "step": 109155 }, { "epoch": 0.27, "learning_rate": 7.279203784790377e-05, "loss": 1.026, "step": 109160 }, { "epoch": 0.27, "learning_rate": 7.279077960642206e-05, "loss": 1.0237, "step": 109165 }, { "epoch": 0.27, "learning_rate": 7.278952136494037e-05, "loss": 1.0204, "step": 109170 }, { "epoch": 0.27, "learning_rate": 7.278826312345866e-05, "loss": 1.0409, "step": 109175 }, { "epoch": 0.27, "learning_rate": 7.278700488197695e-05, "loss": 1.0223, "step": 109180 }, { "epoch": 0.27, "learning_rate": 7.278574664049524e-05, "loss": 1.0219, "step": 109185 }, { "epoch": 0.27, "learning_rate": 7.278448839901353e-05, "loss": 1.023, "step": 109190 }, { "epoch": 0.27, "learning_rate": 7.278323015753184e-05, "loss": 1.0217, "step": 109195 }, { "epoch": 0.27, "learning_rate": 7.278197191605013e-05, "loss": 1.0226, "step": 109200 }, { "epoch": 0.27, "learning_rate": 7.278071367456842e-05, "loss": 1.0233, "step": 109205 }, { "epoch": 0.27, "learning_rate": 7.277945543308671e-05, "loss": 1.0218, "step": 109210 }, { "epoch": 0.27, "learning_rate": 7.277819719160502e-05, "loss": 1.0228, "step": 109215 }, { "epoch": 0.27, "learning_rate": 7.277693895012331e-05, "loss": 1.0225, "step": 109220 }, { "epoch": 0.27, "learning_rate": 7.27756807086416e-05, "loss": 1.0212, "step": 109225 }, { "epoch": 0.27, "learning_rate": 7.277442246715989e-05, "loss": 1.0221, "step": 109230 }, { "epoch": 0.27, "learning_rate": 7.27731642256782e-05, "loss": 1.021, "step": 109235 }, { "epoch": 0.27, "learning_rate": 7.277190598419649e-05, "loss": 1.0242, "step": 109240 }, { "epoch": 0.27, "learning_rate": 7.277064774271478e-05, "loss": 1.022, "step": 109245 }, { "epoch": 0.27, "learning_rate": 7.276938950123307e-05, "loss": 1.0223, "step": 109250 }, { "epoch": 0.27, "learning_rate": 7.276813125975136e-05, "loss": 1.0225, "step": 109255 }, { "epoch": 0.27, "learning_rate": 7.276687301826967e-05, "loss": 1.0227, "step": 109260 }, { "epoch": 0.27, "learning_rate": 7.276561477678796e-05, "loss": 1.0226, "step": 109265 }, { "epoch": 0.27, "learning_rate": 7.276435653530625e-05, "loss": 1.021, "step": 109270 }, { "epoch": 0.27, "learning_rate": 7.276309829382454e-05, "loss": 1.0262, "step": 109275 }, { "epoch": 0.27, "learning_rate": 7.276184005234285e-05, "loss": 1.0225, "step": 109280 }, { "epoch": 0.27, "learning_rate": 7.276058181086114e-05, "loss": 1.0237, "step": 109285 }, { "epoch": 0.27, "learning_rate": 7.275932356937943e-05, "loss": 1.0238, "step": 109290 }, { "epoch": 0.27, "learning_rate": 7.275806532789772e-05, "loss": 1.0207, "step": 109295 }, { "epoch": 0.27, "learning_rate": 7.275680708641603e-05, "loss": 1.0243, "step": 109300 }, { "epoch": 0.27, "learning_rate": 7.275554884493432e-05, "loss": 1.0209, "step": 109305 }, { "epoch": 0.27, "learning_rate": 7.275429060345261e-05, "loss": 1.0225, "step": 109310 }, { "epoch": 0.27, "learning_rate": 7.27530323619709e-05, "loss": 1.0198, "step": 109315 }, { "epoch": 0.27, "learning_rate": 7.27517741204892e-05, "loss": 1.0235, "step": 109320 }, { "epoch": 0.27, "learning_rate": 7.275051587900751e-05, "loss": 1.0236, "step": 109325 }, { "epoch": 0.27, "learning_rate": 7.27492576375258e-05, "loss": 1.0238, "step": 109330 }, { "epoch": 0.27, "learning_rate": 7.27479993960441e-05, "loss": 1.0243, "step": 109335 }, { "epoch": 0.27, "learning_rate": 7.274674115456239e-05, "loss": 1.0231, "step": 109340 }, { "epoch": 0.27, "learning_rate": 7.274548291308068e-05, "loss": 1.0218, "step": 109345 }, { "epoch": 0.27, "learning_rate": 7.274422467159899e-05, "loss": 1.0226, "step": 109350 }, { "epoch": 0.27, "learning_rate": 7.274296643011728e-05, "loss": 1.0233, "step": 109355 }, { "epoch": 0.27, "learning_rate": 7.274170818863557e-05, "loss": 1.0217, "step": 109360 }, { "epoch": 0.27, "learning_rate": 7.274044994715386e-05, "loss": 1.021, "step": 109365 }, { "epoch": 0.27, "learning_rate": 7.273919170567217e-05, "loss": 1.0245, "step": 109370 }, { "epoch": 0.27, "learning_rate": 7.273793346419046e-05, "loss": 1.0228, "step": 109375 }, { "epoch": 0.27, "learning_rate": 7.273667522270875e-05, "loss": 1.0382, "step": 109380 }, { "epoch": 0.27, "learning_rate": 7.273541698122704e-05, "loss": 1.0236, "step": 109385 }, { "epoch": 0.27, "learning_rate": 7.273415873974535e-05, "loss": 1.0246, "step": 109390 }, { "epoch": 0.27, "learning_rate": 7.273290049826364e-05, "loss": 1.0234, "step": 109395 }, { "epoch": 0.27, "learning_rate": 7.273164225678193e-05, "loss": 1.0242, "step": 109400 }, { "epoch": 0.27, "learning_rate": 7.273038401530022e-05, "loss": 1.0229, "step": 109405 }, { "epoch": 0.27, "learning_rate": 7.272912577381851e-05, "loss": 1.0223, "step": 109410 }, { "epoch": 0.27, "learning_rate": 7.272786753233682e-05, "loss": 1.0218, "step": 109415 }, { "epoch": 0.27, "learning_rate": 7.272660929085511e-05, "loss": 1.0243, "step": 109420 }, { "epoch": 0.27, "learning_rate": 7.27253510493734e-05, "loss": 1.0216, "step": 109425 }, { "epoch": 0.27, "learning_rate": 7.272409280789169e-05, "loss": 1.022, "step": 109430 }, { "epoch": 0.27, "learning_rate": 7.272283456641e-05, "loss": 1.0207, "step": 109435 }, { "epoch": 0.27, "learning_rate": 7.272157632492829e-05, "loss": 1.0226, "step": 109440 }, { "epoch": 0.27, "learning_rate": 7.272031808344658e-05, "loss": 1.0201, "step": 109445 }, { "epoch": 0.27, "learning_rate": 7.271905984196487e-05, "loss": 1.0232, "step": 109450 }, { "epoch": 0.27, "learning_rate": 7.271780160048318e-05, "loss": 1.022, "step": 109455 }, { "epoch": 0.27, "learning_rate": 7.271654335900147e-05, "loss": 1.023, "step": 109460 }, { "epoch": 0.27, "learning_rate": 7.271528511751976e-05, "loss": 1.021, "step": 109465 }, { "epoch": 0.27, "learning_rate": 7.271402687603805e-05, "loss": 1.0231, "step": 109470 }, { "epoch": 0.27, "learning_rate": 7.271276863455634e-05, "loss": 1.0228, "step": 109475 }, { "epoch": 0.27, "learning_rate": 7.271151039307465e-05, "loss": 1.022, "step": 109480 }, { "epoch": 0.27, "learning_rate": 7.271025215159294e-05, "loss": 1.0234, "step": 109485 }, { "epoch": 0.27, "learning_rate": 7.270899391011123e-05, "loss": 1.0238, "step": 109490 }, { "epoch": 0.27, "learning_rate": 7.270773566862952e-05, "loss": 1.0239, "step": 109495 }, { "epoch": 0.27, "learning_rate": 7.270647742714783e-05, "loss": 1.0245, "step": 109500 }, { "epoch": 0.27, "learning_rate": 7.270521918566612e-05, "loss": 1.0237, "step": 109505 }, { "epoch": 0.27, "learning_rate": 7.270396094418441e-05, "loss": 1.024, "step": 109510 }, { "epoch": 0.27, "learning_rate": 7.27027027027027e-05, "loss": 1.0251, "step": 109515 }, { "epoch": 0.27, "learning_rate": 7.270144446122101e-05, "loss": 1.0221, "step": 109520 }, { "epoch": 0.27, "learning_rate": 7.27001862197393e-05, "loss": 1.0239, "step": 109525 }, { "epoch": 0.27, "learning_rate": 7.269892797825759e-05, "loss": 1.0232, "step": 109530 }, { "epoch": 0.27, "learning_rate": 7.269766973677588e-05, "loss": 1.0192, "step": 109535 }, { "epoch": 0.27, "learning_rate": 7.269641149529418e-05, "loss": 1.0211, "step": 109540 }, { "epoch": 0.27, "learning_rate": 7.269515325381248e-05, "loss": 1.0226, "step": 109545 }, { "epoch": 0.27, "learning_rate": 7.269389501233077e-05, "loss": 1.0231, "step": 109550 }, { "epoch": 0.28, "learning_rate": 7.269263677084906e-05, "loss": 1.0203, "step": 109555 }, { "epoch": 0.28, "learning_rate": 7.269137852936736e-05, "loss": 1.0229, "step": 109560 }, { "epoch": 0.28, "learning_rate": 7.269012028788566e-05, "loss": 1.0225, "step": 109565 }, { "epoch": 0.28, "learning_rate": 7.268886204640395e-05, "loss": 1.0231, "step": 109570 }, { "epoch": 0.28, "learning_rate": 7.268760380492224e-05, "loss": 1.0229, "step": 109575 }, { "epoch": 0.28, "learning_rate": 7.268634556344054e-05, "loss": 1.0229, "step": 109580 }, { "epoch": 0.28, "learning_rate": 7.268508732195884e-05, "loss": 1.0242, "step": 109585 }, { "epoch": 0.28, "learning_rate": 7.268382908047713e-05, "loss": 1.0425, "step": 109590 }, { "epoch": 0.28, "learning_rate": 7.268257083899542e-05, "loss": 1.0251, "step": 109595 }, { "epoch": 0.28, "learning_rate": 7.268131259751372e-05, "loss": 1.0242, "step": 109600 }, { "epoch": 0.28, "learning_rate": 7.2680054356032e-05, "loss": 1.0215, "step": 109605 }, { "epoch": 0.28, "learning_rate": 7.267879611455031e-05, "loss": 1.0206, "step": 109610 }, { "epoch": 0.28, "learning_rate": 7.26775378730686e-05, "loss": 1.0241, "step": 109615 }, { "epoch": 0.28, "learning_rate": 7.26762796315869e-05, "loss": 1.0216, "step": 109620 }, { "epoch": 0.28, "learning_rate": 7.267502139010519e-05, "loss": 1.0235, "step": 109625 }, { "epoch": 0.28, "learning_rate": 7.267376314862349e-05, "loss": 1.0203, "step": 109630 }, { "epoch": 0.28, "learning_rate": 7.267250490714178e-05, "loss": 1.0239, "step": 109635 }, { "epoch": 0.28, "learning_rate": 7.267124666566007e-05, "loss": 1.0208, "step": 109640 }, { "epoch": 0.28, "learning_rate": 7.266998842417837e-05, "loss": 1.022, "step": 109645 }, { "epoch": 0.28, "learning_rate": 7.266873018269667e-05, "loss": 1.0216, "step": 109650 }, { "epoch": 0.28, "learning_rate": 7.266747194121496e-05, "loss": 1.0247, "step": 109655 }, { "epoch": 0.28, "learning_rate": 7.266621369973325e-05, "loss": 1.0235, "step": 109660 }, { "epoch": 0.28, "learning_rate": 7.266495545825155e-05, "loss": 1.0227, "step": 109665 }, { "epoch": 0.28, "learning_rate": 7.266369721676984e-05, "loss": 1.0392, "step": 109670 }, { "epoch": 0.28, "learning_rate": 7.266243897528814e-05, "loss": 1.0238, "step": 109675 }, { "epoch": 0.28, "learning_rate": 7.266118073380643e-05, "loss": 1.0242, "step": 109680 }, { "epoch": 0.28, "learning_rate": 7.265992249232473e-05, "loss": 1.0355, "step": 109685 }, { "epoch": 0.28, "learning_rate": 7.265866425084302e-05, "loss": 1.0256, "step": 109690 }, { "epoch": 0.28, "learning_rate": 7.265740600936132e-05, "loss": 1.0213, "step": 109695 }, { "epoch": 0.28, "learning_rate": 7.265614776787961e-05, "loss": 1.022, "step": 109700 }, { "epoch": 0.28, "learning_rate": 7.26548895263979e-05, "loss": 1.0232, "step": 109705 }, { "epoch": 0.28, "learning_rate": 7.26536312849162e-05, "loss": 1.0201, "step": 109710 }, { "epoch": 0.28, "learning_rate": 7.265237304343449e-05, "loss": 1.0231, "step": 109715 }, { "epoch": 0.28, "learning_rate": 7.26511148019528e-05, "loss": 1.0224, "step": 109720 }, { "epoch": 0.28, "learning_rate": 7.265010820876743e-05, "loss": 1.0218, "step": 109725 }, { "epoch": 0.28, "learning_rate": 7.264884996728572e-05, "loss": 1.0234, "step": 109730 }, { "epoch": 0.28, "learning_rate": 7.264759172580401e-05, "loss": 1.0232, "step": 109735 }, { "epoch": 0.28, "learning_rate": 7.264633348432232e-05, "loss": 1.0229, "step": 109740 }, { "epoch": 0.28, "learning_rate": 7.264507524284061e-05, "loss": 1.0211, "step": 109745 }, { "epoch": 0.28, "learning_rate": 7.26438170013589e-05, "loss": 1.0223, "step": 109750 }, { "epoch": 0.28, "learning_rate": 7.26425587598772e-05, "loss": 1.024, "step": 109755 }, { "epoch": 0.28, "learning_rate": 7.26413005183955e-05, "loss": 1.0218, "step": 109760 }, { "epoch": 0.28, "learning_rate": 7.264004227691379e-05, "loss": 1.0219, "step": 109765 }, { "epoch": 0.28, "learning_rate": 7.263878403543208e-05, "loss": 1.0223, "step": 109770 }, { "epoch": 0.28, "learning_rate": 7.263752579395037e-05, "loss": 1.022, "step": 109775 }, { "epoch": 0.28, "learning_rate": 7.263626755246868e-05, "loss": 1.0226, "step": 109780 }, { "epoch": 0.28, "learning_rate": 7.263500931098697e-05, "loss": 1.0221, "step": 109785 }, { "epoch": 0.28, "learning_rate": 7.263375106950526e-05, "loss": 1.0234, "step": 109790 }, { "epoch": 0.28, "learning_rate": 7.263249282802355e-05, "loss": 1.0209, "step": 109795 }, { "epoch": 0.28, "learning_rate": 7.263123458654185e-05, "loss": 1.0216, "step": 109800 }, { "epoch": 0.28, "learning_rate": 7.262997634506015e-05, "loss": 1.0221, "step": 109805 }, { "epoch": 0.28, "learning_rate": 7.262871810357844e-05, "loss": 1.022, "step": 109810 }, { "epoch": 0.28, "learning_rate": 7.262745986209673e-05, "loss": 1.0213, "step": 109815 }, { "epoch": 0.28, "learning_rate": 7.262620162061503e-05, "loss": 1.0244, "step": 109820 }, { "epoch": 0.28, "learning_rate": 7.262494337913333e-05, "loss": 1.0225, "step": 109825 }, { "epoch": 0.28, "learning_rate": 7.262368513765162e-05, "loss": 1.024, "step": 109830 }, { "epoch": 0.28, "learning_rate": 7.262242689616991e-05, "loss": 1.0229, "step": 109835 }, { "epoch": 0.28, "learning_rate": 7.26211686546882e-05, "loss": 1.0239, "step": 109840 }, { "epoch": 0.28, "learning_rate": 7.261991041320651e-05, "loss": 1.022, "step": 109845 }, { "epoch": 0.28, "learning_rate": 7.26186521717248e-05, "loss": 1.023, "step": 109850 }, { "epoch": 0.28, "learning_rate": 7.26173939302431e-05, "loss": 1.0231, "step": 109855 }, { "epoch": 0.28, "learning_rate": 7.261613568876139e-05, "loss": 1.022, "step": 109860 }, { "epoch": 0.28, "learning_rate": 7.261487744727968e-05, "loss": 1.0241, "step": 109865 }, { "epoch": 0.28, "learning_rate": 7.261361920579798e-05, "loss": 1.0237, "step": 109870 }, { "epoch": 0.28, "learning_rate": 7.261236096431627e-05, "loss": 1.0226, "step": 109875 }, { "epoch": 0.28, "learning_rate": 7.261110272283457e-05, "loss": 1.0257, "step": 109880 }, { "epoch": 0.28, "learning_rate": 7.260984448135286e-05, "loss": 1.0221, "step": 109885 }, { "epoch": 0.28, "learning_rate": 7.260858623987116e-05, "loss": 1.0227, "step": 109890 }, { "epoch": 0.28, "learning_rate": 7.260732799838945e-05, "loss": 1.0232, "step": 109895 }, { "epoch": 0.28, "learning_rate": 7.260606975690774e-05, "loss": 1.0215, "step": 109900 }, { "epoch": 0.28, "learning_rate": 7.260481151542604e-05, "loss": 1.0224, "step": 109905 }, { "epoch": 0.28, "learning_rate": 7.260355327394434e-05, "loss": 1.0231, "step": 109910 }, { "epoch": 0.28, "learning_rate": 7.260229503246263e-05, "loss": 1.0226, "step": 109915 }, { "epoch": 0.28, "learning_rate": 7.260103679098092e-05, "loss": 1.024, "step": 109920 }, { "epoch": 0.28, "learning_rate": 7.259977854949922e-05, "loss": 1.0213, "step": 109925 }, { "epoch": 0.28, "learning_rate": 7.259852030801751e-05, "loss": 1.0235, "step": 109930 }, { "epoch": 0.28, "learning_rate": 7.259726206653581e-05, "loss": 1.0212, "step": 109935 }, { "epoch": 0.28, "learning_rate": 7.25960038250541e-05, "loss": 1.0218, "step": 109940 }, { "epoch": 0.28, "learning_rate": 7.25947455835724e-05, "loss": 1.023, "step": 109945 }, { "epoch": 0.28, "learning_rate": 7.259348734209069e-05, "loss": 1.0213, "step": 109950 }, { "epoch": 0.28, "learning_rate": 7.2592229100609e-05, "loss": 1.0244, "step": 109955 }, { "epoch": 0.28, "learning_rate": 7.259097085912728e-05, "loss": 1.023, "step": 109960 }, { "epoch": 0.28, "learning_rate": 7.258971261764558e-05, "loss": 1.0236, "step": 109965 }, { "epoch": 0.28, "learning_rate": 7.258845437616387e-05, "loss": 1.0235, "step": 109970 }, { "epoch": 0.28, "learning_rate": 7.258719613468217e-05, "loss": 1.0204, "step": 109975 }, { "epoch": 0.28, "learning_rate": 7.258593789320046e-05, "loss": 1.0206, "step": 109980 }, { "epoch": 0.28, "learning_rate": 7.258467965171876e-05, "loss": 1.0225, "step": 109985 }, { "epoch": 0.28, "learning_rate": 7.258342141023705e-05, "loss": 1.0236, "step": 109990 }, { "epoch": 0.28, "learning_rate": 7.258216316875534e-05, "loss": 1.0217, "step": 109995 }, { "epoch": 0.28, "learning_rate": 7.258090492727364e-05, "loss": 1.0214, "step": 110000 }, { "epoch": 0.28, "learning_rate": 7.257964668579194e-05, "loss": 1.022, "step": 110005 }, { "epoch": 0.28, "learning_rate": 7.257838844431023e-05, "loss": 1.0228, "step": 110010 }, { "epoch": 0.28, "learning_rate": 7.257713020282852e-05, "loss": 1.0223, "step": 110015 }, { "epoch": 0.28, "learning_rate": 7.257587196134682e-05, "loss": 1.0227, "step": 110020 }, { "epoch": 0.28, "learning_rate": 7.257461371986513e-05, "loss": 1.0215, "step": 110025 }, { "epoch": 0.28, "learning_rate": 7.257335547838342e-05, "loss": 1.0228, "step": 110030 }, { "epoch": 0.28, "learning_rate": 7.257209723690171e-05, "loss": 1.0217, "step": 110035 }, { "epoch": 0.28, "learning_rate": 7.257083899542e-05, "loss": 1.0217, "step": 110040 }, { "epoch": 0.28, "learning_rate": 7.256958075393831e-05, "loss": 1.0236, "step": 110045 }, { "epoch": 0.28, "learning_rate": 7.25683225124566e-05, "loss": 1.0205, "step": 110050 }, { "epoch": 0.28, "learning_rate": 7.256706427097489e-05, "loss": 1.0226, "step": 110055 }, { "epoch": 0.28, "learning_rate": 7.256580602949318e-05, "loss": 1.0212, "step": 110060 }, { "epoch": 0.28, "learning_rate": 7.256454778801149e-05, "loss": 1.0225, "step": 110065 }, { "epoch": 0.28, "learning_rate": 7.256328954652978e-05, "loss": 1.0227, "step": 110070 }, { "epoch": 0.28, "learning_rate": 7.256203130504807e-05, "loss": 1.02, "step": 110075 }, { "epoch": 0.28, "learning_rate": 7.256077306356636e-05, "loss": 1.0241, "step": 110080 }, { "epoch": 0.28, "learning_rate": 7.255951482208466e-05, "loss": 1.0249, "step": 110085 }, { "epoch": 0.28, "learning_rate": 7.255825658060296e-05, "loss": 1.0221, "step": 110090 }, { "epoch": 0.28, "learning_rate": 7.255699833912125e-05, "loss": 1.0245, "step": 110095 }, { "epoch": 0.28, "learning_rate": 7.255574009763954e-05, "loss": 1.0239, "step": 110100 }, { "epoch": 0.28, "learning_rate": 7.255448185615784e-05, "loss": 1.0231, "step": 110105 }, { "epoch": 0.28, "learning_rate": 7.255322361467614e-05, "loss": 1.0231, "step": 110110 }, { "epoch": 0.28, "learning_rate": 7.255196537319443e-05, "loss": 1.0229, "step": 110115 }, { "epoch": 0.28, "learning_rate": 7.255070713171272e-05, "loss": 1.0234, "step": 110120 }, { "epoch": 0.28, "learning_rate": 7.254944889023102e-05, "loss": 1.0213, "step": 110125 }, { "epoch": 0.28, "learning_rate": 7.254819064874932e-05, "loss": 1.0226, "step": 110130 }, { "epoch": 0.28, "learning_rate": 7.254693240726761e-05, "loss": 1.0212, "step": 110135 }, { "epoch": 0.28, "learning_rate": 7.25456741657859e-05, "loss": 1.0237, "step": 110140 }, { "epoch": 0.28, "learning_rate": 7.25444159243042e-05, "loss": 1.023, "step": 110145 }, { "epoch": 0.28, "learning_rate": 7.254315768282249e-05, "loss": 1.0212, "step": 110150 }, { "epoch": 0.28, "learning_rate": 7.254189944134079e-05, "loss": 1.0215, "step": 110155 }, { "epoch": 0.28, "learning_rate": 7.254064119985908e-05, "loss": 1.0232, "step": 110160 }, { "epoch": 0.28, "learning_rate": 7.253938295837738e-05, "loss": 1.021, "step": 110165 }, { "epoch": 0.28, "learning_rate": 7.253812471689567e-05, "loss": 1.0227, "step": 110170 }, { "epoch": 0.28, "learning_rate": 7.253686647541397e-05, "loss": 1.0218, "step": 110175 }, { "epoch": 0.28, "learning_rate": 7.253560823393226e-05, "loss": 1.0211, "step": 110180 }, { "epoch": 0.28, "learning_rate": 7.253434999245056e-05, "loss": 1.0227, "step": 110185 }, { "epoch": 0.28, "learning_rate": 7.253309175096885e-05, "loss": 1.0208, "step": 110190 }, { "epoch": 0.28, "learning_rate": 7.253183350948715e-05, "loss": 1.0417, "step": 110195 }, { "epoch": 0.28, "learning_rate": 7.253057526800544e-05, "loss": 1.0229, "step": 110200 }, { "epoch": 0.28, "learning_rate": 7.252931702652374e-05, "loss": 1.025, "step": 110205 }, { "epoch": 0.28, "learning_rate": 7.252805878504203e-05, "loss": 1.046, "step": 110210 }, { "epoch": 0.28, "learning_rate": 7.252680054356032e-05, "loss": 1.0201, "step": 110215 }, { "epoch": 0.28, "learning_rate": 7.252554230207862e-05, "loss": 1.0224, "step": 110220 }, { "epoch": 0.28, "learning_rate": 7.252428406059692e-05, "loss": 1.0221, "step": 110225 }, { "epoch": 0.28, "learning_rate": 7.252302581911521e-05, "loss": 1.0241, "step": 110230 }, { "epoch": 0.28, "learning_rate": 7.25217675776335e-05, "loss": 1.0214, "step": 110235 }, { "epoch": 0.28, "learning_rate": 7.25205093361518e-05, "loss": 1.023, "step": 110240 }, { "epoch": 0.28, "learning_rate": 7.25192510946701e-05, "loss": 1.0227, "step": 110245 }, { "epoch": 0.28, "learning_rate": 7.251799285318839e-05, "loss": 1.0232, "step": 110250 }, { "epoch": 0.28, "learning_rate": 7.251673461170668e-05, "loss": 1.0238, "step": 110255 }, { "epoch": 0.28, "learning_rate": 7.251547637022498e-05, "loss": 1.024, "step": 110260 }, { "epoch": 0.28, "learning_rate": 7.251421812874328e-05, "loss": 1.0217, "step": 110265 }, { "epoch": 0.28, "learning_rate": 7.251295988726157e-05, "loss": 1.0224, "step": 110270 }, { "epoch": 0.28, "learning_rate": 7.251170164577986e-05, "loss": 1.0248, "step": 110275 }, { "epoch": 0.28, "learning_rate": 7.251044340429815e-05, "loss": 1.0233, "step": 110280 }, { "epoch": 0.28, "learning_rate": 7.250918516281646e-05, "loss": 1.0225, "step": 110285 }, { "epoch": 0.28, "learning_rate": 7.250792692133475e-05, "loss": 1.0232, "step": 110290 }, { "epoch": 0.28, "learning_rate": 7.250666867985304e-05, "loss": 1.0232, "step": 110295 }, { "epoch": 0.28, "learning_rate": 7.250541043837133e-05, "loss": 1.022, "step": 110300 }, { "epoch": 0.28, "learning_rate": 7.250415219688964e-05, "loss": 1.0221, "step": 110305 }, { "epoch": 0.28, "learning_rate": 7.250289395540793e-05, "loss": 1.0212, "step": 110310 }, { "epoch": 0.28, "learning_rate": 7.250163571392622e-05, "loss": 1.0235, "step": 110315 }, { "epoch": 0.28, "learning_rate": 7.250037747244451e-05, "loss": 1.0232, "step": 110320 }, { "epoch": 0.28, "learning_rate": 7.249911923096282e-05, "loss": 1.023, "step": 110325 }, { "epoch": 0.28, "learning_rate": 7.249786098948111e-05, "loss": 1.0219, "step": 110330 }, { "epoch": 0.28, "learning_rate": 7.24966027479994e-05, "loss": 1.0199, "step": 110335 }, { "epoch": 0.28, "learning_rate": 7.249534450651769e-05, "loss": 1.0215, "step": 110340 }, { "epoch": 0.28, "learning_rate": 7.249408626503598e-05, "loss": 1.0226, "step": 110345 }, { "epoch": 0.28, "learning_rate": 7.249282802355429e-05, "loss": 1.0204, "step": 110350 }, { "epoch": 0.28, "learning_rate": 7.249156978207258e-05, "loss": 1.0216, "step": 110355 }, { "epoch": 0.28, "learning_rate": 7.249031154059087e-05, "loss": 1.0234, "step": 110360 }, { "epoch": 0.28, "learning_rate": 7.248905329910916e-05, "loss": 1.0246, "step": 110365 }, { "epoch": 0.28, "learning_rate": 7.248779505762747e-05, "loss": 1.0234, "step": 110370 }, { "epoch": 0.28, "learning_rate": 7.248653681614576e-05, "loss": 1.0239, "step": 110375 }, { "epoch": 0.28, "learning_rate": 7.248527857466405e-05, "loss": 1.0197, "step": 110380 }, { "epoch": 0.28, "learning_rate": 7.248402033318234e-05, "loss": 1.0226, "step": 110385 }, { "epoch": 0.28, "learning_rate": 7.248276209170063e-05, "loss": 1.021, "step": 110390 }, { "epoch": 0.28, "learning_rate": 7.248150385021894e-05, "loss": 1.0218, "step": 110395 }, { "epoch": 0.28, "learning_rate": 7.248024560873723e-05, "loss": 1.021, "step": 110400 }, { "epoch": 0.28, "learning_rate": 7.247898736725552e-05, "loss": 1.0233, "step": 110405 }, { "epoch": 0.28, "learning_rate": 7.247772912577381e-05, "loss": 1.0217, "step": 110410 }, { "epoch": 0.28, "learning_rate": 7.247647088429212e-05, "loss": 1.0208, "step": 110415 }, { "epoch": 0.28, "learning_rate": 7.247521264281041e-05, "loss": 1.0213, "step": 110420 }, { "epoch": 0.28, "learning_rate": 7.24739544013287e-05, "loss": 1.0215, "step": 110425 }, { "epoch": 0.28, "learning_rate": 7.247269615984699e-05, "loss": 1.0214, "step": 110430 }, { "epoch": 0.28, "learning_rate": 7.24714379183653e-05, "loss": 1.0189, "step": 110435 }, { "epoch": 0.28, "learning_rate": 7.247017967688359e-05, "loss": 1.0198, "step": 110440 }, { "epoch": 0.28, "learning_rate": 7.246892143540188e-05, "loss": 1.0218, "step": 110445 }, { "epoch": 0.28, "learning_rate": 7.246766319392017e-05, "loss": 1.0243, "step": 110450 }, { "epoch": 0.28, "learning_rate": 7.246640495243846e-05, "loss": 1.0241, "step": 110455 }, { "epoch": 0.28, "learning_rate": 7.246514671095677e-05, "loss": 1.024, "step": 110460 }, { "epoch": 0.28, "learning_rate": 7.246388846947506e-05, "loss": 1.0222, "step": 110465 }, { "epoch": 0.28, "learning_rate": 7.246263022799335e-05, "loss": 1.0239, "step": 110470 }, { "epoch": 0.28, "learning_rate": 7.246137198651164e-05, "loss": 1.0246, "step": 110475 }, { "epoch": 0.28, "learning_rate": 7.246011374502995e-05, "loss": 1.0209, "step": 110480 }, { "epoch": 0.28, "learning_rate": 7.245885550354824e-05, "loss": 1.0215, "step": 110485 }, { "epoch": 0.28, "learning_rate": 7.245759726206653e-05, "loss": 1.0226, "step": 110490 }, { "epoch": 0.28, "learning_rate": 7.245633902058482e-05, "loss": 1.0242, "step": 110495 }, { "epoch": 0.28, "learning_rate": 7.245508077910313e-05, "loss": 1.0208, "step": 110500 }, { "epoch": 0.28, "learning_rate": 7.245382253762142e-05, "loss": 1.0479, "step": 110505 }, { "epoch": 0.28, "learning_rate": 7.245256429613971e-05, "loss": 1.0231, "step": 110510 }, { "epoch": 0.28, "learning_rate": 7.2451306054658e-05, "loss": 1.0241, "step": 110515 }, { "epoch": 0.28, "learning_rate": 7.24500478131763e-05, "loss": 1.0235, "step": 110520 }, { "epoch": 0.28, "learning_rate": 7.244878957169461e-05, "loss": 1.0201, "step": 110525 }, { "epoch": 0.28, "learning_rate": 7.24475313302129e-05, "loss": 1.0229, "step": 110530 }, { "epoch": 0.28, "learning_rate": 7.24462730887312e-05, "loss": 1.0231, "step": 110535 }, { "epoch": 0.28, "learning_rate": 7.244501484724949e-05, "loss": 1.0223, "step": 110540 }, { "epoch": 0.28, "learning_rate": 7.244375660576778e-05, "loss": 1.0227, "step": 110545 }, { "epoch": 0.28, "learning_rate": 7.244249836428609e-05, "loss": 1.0254, "step": 110550 }, { "epoch": 0.28, "learning_rate": 7.244124012280438e-05, "loss": 1.0228, "step": 110555 }, { "epoch": 0.28, "learning_rate": 7.243998188132267e-05, "loss": 1.0242, "step": 110560 }, { "epoch": 0.28, "learning_rate": 7.243872363984096e-05, "loss": 1.0231, "step": 110565 }, { "epoch": 0.28, "learning_rate": 7.243746539835927e-05, "loss": 1.0224, "step": 110570 }, { "epoch": 0.28, "learning_rate": 7.243620715687756e-05, "loss": 1.0225, "step": 110575 }, { "epoch": 0.28, "learning_rate": 7.243494891539585e-05, "loss": 1.0234, "step": 110580 }, { "epoch": 0.28, "learning_rate": 7.243369067391414e-05, "loss": 1.0234, "step": 110585 }, { "epoch": 0.28, "learning_rate": 7.243243243243245e-05, "loss": 1.0222, "step": 110590 }, { "epoch": 0.28, "learning_rate": 7.243117419095074e-05, "loss": 1.0229, "step": 110595 }, { "epoch": 0.28, "learning_rate": 7.242991594946903e-05, "loss": 1.0203, "step": 110600 }, { "epoch": 0.28, "learning_rate": 7.242865770798732e-05, "loss": 1.0228, "step": 110605 }, { "epoch": 0.28, "learning_rate": 7.242739946650561e-05, "loss": 1.022, "step": 110610 }, { "epoch": 0.28, "learning_rate": 7.242614122502392e-05, "loss": 1.0238, "step": 110615 }, { "epoch": 0.28, "learning_rate": 7.242488298354221e-05, "loss": 1.023, "step": 110620 }, { "epoch": 0.28, "learning_rate": 7.24236247420605e-05, "loss": 1.031, "step": 110625 }, { "epoch": 0.28, "learning_rate": 7.242236650057879e-05, "loss": 1.0213, "step": 110630 }, { "epoch": 0.28, "learning_rate": 7.24211082590971e-05, "loss": 1.0218, "step": 110635 }, { "epoch": 0.28, "learning_rate": 7.241985001761539e-05, "loss": 1.0211, "step": 110640 }, { "epoch": 0.28, "learning_rate": 7.241859177613368e-05, "loss": 1.0213, "step": 110645 }, { "epoch": 0.28, "learning_rate": 7.241733353465197e-05, "loss": 1.0234, "step": 110650 }, { "epoch": 0.28, "learning_rate": 7.241607529317028e-05, "loss": 1.0227, "step": 110655 }, { "epoch": 0.28, "learning_rate": 7.241481705168857e-05, "loss": 1.0208, "step": 110660 }, { "epoch": 0.28, "learning_rate": 7.241355881020686e-05, "loss": 1.0219, "step": 110665 }, { "epoch": 0.28, "learning_rate": 7.241230056872515e-05, "loss": 1.0215, "step": 110670 }, { "epoch": 0.28, "learning_rate": 7.241104232724344e-05, "loss": 1.0221, "step": 110675 }, { "epoch": 0.28, "learning_rate": 7.240978408576175e-05, "loss": 1.0207, "step": 110680 }, { "epoch": 0.28, "learning_rate": 7.240852584428004e-05, "loss": 1.0228, "step": 110685 }, { "epoch": 0.28, "learning_rate": 7.240726760279833e-05, "loss": 1.0209, "step": 110690 }, { "epoch": 0.28, "learning_rate": 7.240600936131662e-05, "loss": 1.0226, "step": 110695 }, { "epoch": 0.28, "learning_rate": 7.240475111983493e-05, "loss": 1.0219, "step": 110700 }, { "epoch": 0.28, "learning_rate": 7.240349287835322e-05, "loss": 1.0222, "step": 110705 }, { "epoch": 0.28, "learning_rate": 7.240223463687151e-05, "loss": 1.0225, "step": 110710 }, { "epoch": 0.28, "learning_rate": 7.24009763953898e-05, "loss": 1.021, "step": 110715 }, { "epoch": 0.28, "learning_rate": 7.239971815390811e-05, "loss": 1.0221, "step": 110720 }, { "epoch": 0.28, "learning_rate": 7.23984599124264e-05, "loss": 1.0246, "step": 110725 }, { "epoch": 0.28, "learning_rate": 7.239720167094469e-05, "loss": 1.0237, "step": 110730 }, { "epoch": 0.28, "learning_rate": 7.239594342946298e-05, "loss": 1.0211, "step": 110735 }, { "epoch": 0.28, "learning_rate": 7.239468518798128e-05, "loss": 1.0232, "step": 110740 }, { "epoch": 0.28, "learning_rate": 7.239342694649958e-05, "loss": 1.0232, "step": 110745 }, { "epoch": 0.28, "learning_rate": 7.239216870501787e-05, "loss": 1.0208, "step": 110750 }, { "epoch": 0.28, "learning_rate": 7.239091046353616e-05, "loss": 1.0229, "step": 110755 }, { "epoch": 0.28, "learning_rate": 7.238965222205446e-05, "loss": 1.0248, "step": 110760 }, { "epoch": 0.28, "learning_rate": 7.238839398057276e-05, "loss": 1.0211, "step": 110765 }, { "epoch": 0.28, "learning_rate": 7.238713573909105e-05, "loss": 1.0208, "step": 110770 }, { "epoch": 0.28, "learning_rate": 7.238587749760934e-05, "loss": 1.0229, "step": 110775 }, { "epoch": 0.28, "learning_rate": 7.238461925612764e-05, "loss": 1.0251, "step": 110780 }, { "epoch": 0.28, "learning_rate": 7.238336101464594e-05, "loss": 1.0234, "step": 110785 }, { "epoch": 0.28, "learning_rate": 7.238210277316423e-05, "loss": 1.022, "step": 110790 }, { "epoch": 0.28, "learning_rate": 7.238084453168252e-05, "loss": 1.0233, "step": 110795 }, { "epoch": 0.28, "learning_rate": 7.237958629020082e-05, "loss": 1.022, "step": 110800 }, { "epoch": 0.28, "learning_rate": 7.237832804871911e-05, "loss": 1.0249, "step": 110805 }, { "epoch": 0.28, "learning_rate": 7.237706980723741e-05, "loss": 1.0205, "step": 110810 }, { "epoch": 0.28, "learning_rate": 7.23758115657557e-05, "loss": 1.0225, "step": 110815 }, { "epoch": 0.28, "learning_rate": 7.2374553324274e-05, "loss": 1.0217, "step": 110820 }, { "epoch": 0.28, "learning_rate": 7.237329508279229e-05, "loss": 1.0226, "step": 110825 }, { "epoch": 0.28, "learning_rate": 7.237203684131059e-05, "loss": 1.0235, "step": 110830 }, { "epoch": 0.28, "learning_rate": 7.237077859982888e-05, "loss": 1.0248, "step": 110835 }, { "epoch": 0.28, "learning_rate": 7.236952035834718e-05, "loss": 1.0221, "step": 110840 }, { "epoch": 0.28, "learning_rate": 7.236826211686547e-05, "loss": 1.0231, "step": 110845 }, { "epoch": 0.28, "learning_rate": 7.236700387538377e-05, "loss": 1.021, "step": 110850 }, { "epoch": 0.28, "learning_rate": 7.236574563390206e-05, "loss": 1.023, "step": 110855 }, { "epoch": 0.28, "learning_rate": 7.236448739242036e-05, "loss": 1.0228, "step": 110860 }, { "epoch": 0.28, "learning_rate": 7.236322915093865e-05, "loss": 1.0232, "step": 110865 }, { "epoch": 0.28, "learning_rate": 7.236197090945694e-05, "loss": 1.0212, "step": 110870 }, { "epoch": 0.28, "learning_rate": 7.236071266797524e-05, "loss": 1.0221, "step": 110875 }, { "epoch": 0.28, "learning_rate": 7.235945442649354e-05, "loss": 1.0226, "step": 110880 }, { "epoch": 0.28, "learning_rate": 7.235819618501183e-05, "loss": 1.025, "step": 110885 }, { "epoch": 0.28, "learning_rate": 7.235693794353012e-05, "loss": 1.0241, "step": 110890 }, { "epoch": 0.28, "learning_rate": 7.235567970204842e-05, "loss": 1.0237, "step": 110895 }, { "epoch": 0.28, "learning_rate": 7.235442146056672e-05, "loss": 1.0216, "step": 110900 }, { "epoch": 0.28, "learning_rate": 7.2353163219085e-05, "loss": 1.0217, "step": 110905 }, { "epoch": 0.28, "learning_rate": 7.23519049776033e-05, "loss": 1.0211, "step": 110910 }, { "epoch": 0.28, "learning_rate": 7.23506467361216e-05, "loss": 1.0234, "step": 110915 }, { "epoch": 0.28, "learning_rate": 7.23493884946399e-05, "loss": 1.0225, "step": 110920 }, { "epoch": 0.28, "learning_rate": 7.234813025315819e-05, "loss": 1.0239, "step": 110925 }, { "epoch": 0.28, "learning_rate": 7.234687201167648e-05, "loss": 1.0236, "step": 110930 }, { "epoch": 0.28, "learning_rate": 7.234561377019477e-05, "loss": 1.041, "step": 110935 }, { "epoch": 0.28, "learning_rate": 7.234435552871307e-05, "loss": 1.02, "step": 110940 }, { "epoch": 0.28, "learning_rate": 7.234309728723137e-05, "loss": 1.0231, "step": 110945 }, { "epoch": 0.28, "learning_rate": 7.234183904574966e-05, "loss": 1.0198, "step": 110950 }, { "epoch": 0.28, "learning_rate": 7.234058080426795e-05, "loss": 1.0248, "step": 110955 }, { "epoch": 0.28, "learning_rate": 7.233932256278625e-05, "loss": 1.0227, "step": 110960 }, { "epoch": 0.28, "learning_rate": 7.233806432130455e-05, "loss": 1.0213, "step": 110965 }, { "epoch": 0.28, "learning_rate": 7.233680607982284e-05, "loss": 1.0219, "step": 110970 }, { "epoch": 0.28, "learning_rate": 7.233554783834113e-05, "loss": 1.0241, "step": 110975 }, { "epoch": 0.28, "learning_rate": 7.233428959685943e-05, "loss": 1.0223, "step": 110980 }, { "epoch": 0.28, "learning_rate": 7.233303135537773e-05, "loss": 1.0239, "step": 110985 }, { "epoch": 0.28, "learning_rate": 7.233177311389602e-05, "loss": 1.0211, "step": 110990 }, { "epoch": 0.28, "learning_rate": 7.233051487241431e-05, "loss": 1.0213, "step": 110995 }, { "epoch": 0.28, "learning_rate": 7.23292566309326e-05, "loss": 1.0208, "step": 111000 }, { "epoch": 0.28, "learning_rate": 7.23279983894509e-05, "loss": 1.0217, "step": 111005 }, { "epoch": 0.28, "learning_rate": 7.23267401479692e-05, "loss": 1.0233, "step": 111010 }, { "epoch": 0.28, "learning_rate": 7.232548190648749e-05, "loss": 1.0234, "step": 111015 }, { "epoch": 0.28, "learning_rate": 7.232422366500578e-05, "loss": 1.0222, "step": 111020 }, { "epoch": 0.28, "learning_rate": 7.232296542352409e-05, "loss": 1.0248, "step": 111025 }, { "epoch": 0.28, "learning_rate": 7.232170718204239e-05, "loss": 1.0219, "step": 111030 }, { "epoch": 0.28, "learning_rate": 7.232044894056068e-05, "loss": 1.0233, "step": 111035 }, { "epoch": 0.28, "learning_rate": 7.231919069907897e-05, "loss": 1.0209, "step": 111040 }, { "epoch": 0.28, "learning_rate": 7.231793245759727e-05, "loss": 1.0236, "step": 111045 }, { "epoch": 0.28, "learning_rate": 7.231667421611557e-05, "loss": 1.023, "step": 111050 }, { "epoch": 0.28, "learning_rate": 7.231541597463386e-05, "loss": 1.0213, "step": 111055 }, { "epoch": 0.28, "learning_rate": 7.231415773315215e-05, "loss": 1.0222, "step": 111060 }, { "epoch": 0.28, "learning_rate": 7.231289949167045e-05, "loss": 1.0234, "step": 111065 }, { "epoch": 0.28, "learning_rate": 7.231164125018874e-05, "loss": 1.0234, "step": 111070 }, { "epoch": 0.28, "learning_rate": 7.231038300870704e-05, "loss": 1.0229, "step": 111075 }, { "epoch": 0.28, "learning_rate": 7.230912476722533e-05, "loss": 1.0226, "step": 111080 }, { "epoch": 0.28, "learning_rate": 7.230786652574363e-05, "loss": 1.0238, "step": 111085 }, { "epoch": 0.28, "learning_rate": 7.230660828426192e-05, "loss": 1.0234, "step": 111090 }, { "epoch": 0.28, "learning_rate": 7.230535004278022e-05, "loss": 1.0244, "step": 111095 }, { "epoch": 0.28, "learning_rate": 7.230409180129851e-05, "loss": 1.0221, "step": 111100 }, { "epoch": 0.28, "learning_rate": 7.23028335598168e-05, "loss": 1.0237, "step": 111105 }, { "epoch": 0.28, "learning_rate": 7.23015753183351e-05, "loss": 1.0245, "step": 111110 }, { "epoch": 0.28, "learning_rate": 7.23003170768534e-05, "loss": 1.0212, "step": 111115 }, { "epoch": 0.28, "learning_rate": 7.22990588353717e-05, "loss": 1.0228, "step": 111120 }, { "epoch": 0.28, "learning_rate": 7.229780059388999e-05, "loss": 1.0213, "step": 111125 }, { "epoch": 0.28, "learning_rate": 7.229654235240828e-05, "loss": 1.0217, "step": 111130 }, { "epoch": 0.28, "learning_rate": 7.229528411092657e-05, "loss": 1.0214, "step": 111135 }, { "epoch": 0.28, "learning_rate": 7.229402586944487e-05, "loss": 1.0242, "step": 111140 }, { "epoch": 0.28, "learning_rate": 7.229276762796317e-05, "loss": 1.0229, "step": 111145 }, { "epoch": 0.28, "learning_rate": 7.229150938648146e-05, "loss": 1.0225, "step": 111150 }, { "epoch": 0.28, "learning_rate": 7.229025114499975e-05, "loss": 1.0224, "step": 111155 }, { "epoch": 0.28, "learning_rate": 7.228899290351805e-05, "loss": 1.0268, "step": 111160 }, { "epoch": 0.28, "learning_rate": 7.228798631033268e-05, "loss": 1.0212, "step": 111165 }, { "epoch": 0.28, "learning_rate": 7.228672806885097e-05, "loss": 1.0207, "step": 111170 }, { "epoch": 0.28, "learning_rate": 7.228546982736927e-05, "loss": 1.0238, "step": 111175 }, { "epoch": 0.28, "learning_rate": 7.228421158588757e-05, "loss": 1.0229, "step": 111180 }, { "epoch": 0.28, "learning_rate": 7.228295334440586e-05, "loss": 1.023, "step": 111185 }, { "epoch": 0.28, "learning_rate": 7.228169510292415e-05, "loss": 1.0245, "step": 111190 }, { "epoch": 0.28, "learning_rate": 7.228043686144244e-05, "loss": 1.0233, "step": 111195 }, { "epoch": 0.28, "learning_rate": 7.227917861996074e-05, "loss": 1.0217, "step": 111200 }, { "epoch": 0.28, "learning_rate": 7.227792037847904e-05, "loss": 1.0241, "step": 111205 }, { "epoch": 0.28, "learning_rate": 7.227666213699733e-05, "loss": 1.0248, "step": 111210 }, { "epoch": 0.28, "learning_rate": 7.227540389551562e-05, "loss": 1.0226, "step": 111215 }, { "epoch": 0.28, "learning_rate": 7.227414565403392e-05, "loss": 1.023, "step": 111220 }, { "epoch": 0.28, "learning_rate": 7.227288741255223e-05, "loss": 1.0204, "step": 111225 }, { "epoch": 0.28, "learning_rate": 7.227162917107052e-05, "loss": 1.0221, "step": 111230 }, { "epoch": 0.28, "learning_rate": 7.227037092958881e-05, "loss": 1.0221, "step": 111235 }, { "epoch": 0.28, "learning_rate": 7.22691126881071e-05, "loss": 1.0241, "step": 111240 }, { "epoch": 0.28, "learning_rate": 7.226785444662541e-05, "loss": 1.0226, "step": 111245 }, { "epoch": 0.28, "learning_rate": 7.22665962051437e-05, "loss": 1.023, "step": 111250 }, { "epoch": 0.28, "learning_rate": 7.226533796366199e-05, "loss": 1.0217, "step": 111255 }, { "epoch": 0.28, "learning_rate": 7.226407972218028e-05, "loss": 1.0204, "step": 111260 }, { "epoch": 0.28, "learning_rate": 7.226282148069859e-05, "loss": 1.0217, "step": 111265 }, { "epoch": 0.28, "learning_rate": 7.226156323921688e-05, "loss": 1.0213, "step": 111270 }, { "epoch": 0.28, "learning_rate": 7.226030499773517e-05, "loss": 1.0225, "step": 111275 }, { "epoch": 0.28, "learning_rate": 7.225904675625346e-05, "loss": 1.0255, "step": 111280 }, { "epoch": 0.28, "learning_rate": 7.225778851477176e-05, "loss": 1.0227, "step": 111285 }, { "epoch": 0.28, "learning_rate": 7.225653027329006e-05, "loss": 1.0258, "step": 111290 }, { "epoch": 0.28, "learning_rate": 7.225527203180835e-05, "loss": 1.0227, "step": 111295 }, { "epoch": 0.28, "learning_rate": 7.225401379032664e-05, "loss": 1.0213, "step": 111300 }, { "epoch": 0.28, "learning_rate": 7.225275554884494e-05, "loss": 1.0205, "step": 111305 }, { "epoch": 0.28, "learning_rate": 7.225149730736324e-05, "loss": 1.0247, "step": 111310 }, { "epoch": 0.28, "learning_rate": 7.225023906588153e-05, "loss": 1.0207, "step": 111315 }, { "epoch": 0.28, "learning_rate": 7.224898082439982e-05, "loss": 1.0229, "step": 111320 }, { "epoch": 0.28, "learning_rate": 7.224772258291812e-05, "loss": 1.0231, "step": 111325 }, { "epoch": 0.28, "learning_rate": 7.224646434143642e-05, "loss": 1.0219, "step": 111330 }, { "epoch": 0.28, "learning_rate": 7.224520609995471e-05, "loss": 1.0219, "step": 111335 }, { "epoch": 0.28, "learning_rate": 7.2243947858473e-05, "loss": 1.0237, "step": 111340 }, { "epoch": 0.28, "learning_rate": 7.22426896169913e-05, "loss": 1.0229, "step": 111345 }, { "epoch": 0.28, "learning_rate": 7.224143137550959e-05, "loss": 1.0222, "step": 111350 }, { "epoch": 0.28, "learning_rate": 7.224017313402789e-05, "loss": 1.024, "step": 111355 }, { "epoch": 0.28, "learning_rate": 7.223891489254618e-05, "loss": 1.0219, "step": 111360 }, { "epoch": 0.28, "learning_rate": 7.223765665106448e-05, "loss": 1.024, "step": 111365 }, { "epoch": 0.28, "learning_rate": 7.223639840958277e-05, "loss": 1.0224, "step": 111370 }, { "epoch": 0.28, "learning_rate": 7.223514016810107e-05, "loss": 1.0219, "step": 111375 }, { "epoch": 0.28, "learning_rate": 7.223388192661936e-05, "loss": 1.0238, "step": 111380 }, { "epoch": 0.28, "learning_rate": 7.223262368513766e-05, "loss": 1.0202, "step": 111385 }, { "epoch": 0.28, "learning_rate": 7.223136544365595e-05, "loss": 1.0226, "step": 111390 }, { "epoch": 0.28, "learning_rate": 7.223010720217425e-05, "loss": 1.0238, "step": 111395 }, { "epoch": 0.28, "learning_rate": 7.222884896069254e-05, "loss": 1.0231, "step": 111400 }, { "epoch": 0.28, "learning_rate": 7.222759071921084e-05, "loss": 1.0232, "step": 111405 }, { "epoch": 0.28, "learning_rate": 7.222633247772913e-05, "loss": 1.0234, "step": 111410 }, { "epoch": 0.28, "learning_rate": 7.222507423624742e-05, "loss": 1.0223, "step": 111415 }, { "epoch": 0.28, "learning_rate": 7.222381599476572e-05, "loss": 1.0249, "step": 111420 }, { "epoch": 0.28, "learning_rate": 7.222255775328402e-05, "loss": 1.0247, "step": 111425 }, { "epoch": 0.28, "learning_rate": 7.222129951180231e-05, "loss": 1.0221, "step": 111430 }, { "epoch": 0.28, "learning_rate": 7.22200412703206e-05, "loss": 1.0242, "step": 111435 }, { "epoch": 0.28, "learning_rate": 7.22187830288389e-05, "loss": 1.0229, "step": 111440 }, { "epoch": 0.28, "learning_rate": 7.22175247873572e-05, "loss": 1.0235, "step": 111445 }, { "epoch": 0.28, "learning_rate": 7.221626654587549e-05, "loss": 1.0387, "step": 111450 }, { "epoch": 0.28, "learning_rate": 7.221500830439378e-05, "loss": 1.023, "step": 111455 }, { "epoch": 0.28, "learning_rate": 7.221375006291208e-05, "loss": 1.0225, "step": 111460 }, { "epoch": 0.28, "learning_rate": 7.221249182143038e-05, "loss": 1.0242, "step": 111465 }, { "epoch": 0.28, "learning_rate": 7.221123357994867e-05, "loss": 1.0244, "step": 111470 }, { "epoch": 0.28, "learning_rate": 7.220997533846696e-05, "loss": 1.0218, "step": 111475 }, { "epoch": 0.28, "learning_rate": 7.220871709698525e-05, "loss": 1.0218, "step": 111480 }, { "epoch": 0.28, "learning_rate": 7.220745885550356e-05, "loss": 1.0229, "step": 111485 }, { "epoch": 0.28, "learning_rate": 7.220620061402185e-05, "loss": 1.0213, "step": 111490 }, { "epoch": 0.28, "learning_rate": 7.220494237254014e-05, "loss": 1.0203, "step": 111495 }, { "epoch": 0.28, "learning_rate": 7.220368413105843e-05, "loss": 1.0224, "step": 111500 }, { "epoch": 0.28, "learning_rate": 7.220242588957674e-05, "loss": 1.023, "step": 111505 }, { "epoch": 0.28, "learning_rate": 7.220116764809503e-05, "loss": 1.0226, "step": 111510 }, { "epoch": 0.28, "learning_rate": 7.219990940661332e-05, "loss": 1.0437, "step": 111515 }, { "epoch": 0.28, "learning_rate": 7.219865116513161e-05, "loss": 1.0231, "step": 111520 }, { "epoch": 0.28, "learning_rate": 7.219739292364992e-05, "loss": 1.0254, "step": 111525 }, { "epoch": 0.28, "learning_rate": 7.219613468216821e-05, "loss": 1.0218, "step": 111530 }, { "epoch": 0.28, "learning_rate": 7.21948764406865e-05, "loss": 1.0243, "step": 111535 }, { "epoch": 0.28, "learning_rate": 7.219361819920479e-05, "loss": 1.0243, "step": 111540 }, { "epoch": 0.28, "learning_rate": 7.219235995772308e-05, "loss": 1.0218, "step": 111545 }, { "epoch": 0.28, "learning_rate": 7.219110171624139e-05, "loss": 1.0219, "step": 111550 }, { "epoch": 0.28, "learning_rate": 7.218984347475968e-05, "loss": 1.021, "step": 111555 }, { "epoch": 0.28, "learning_rate": 7.218858523327797e-05, "loss": 1.0217, "step": 111560 }, { "epoch": 0.28, "learning_rate": 7.218732699179626e-05, "loss": 1.0197, "step": 111565 }, { "epoch": 0.28, "learning_rate": 7.218606875031457e-05, "loss": 1.0227, "step": 111570 }, { "epoch": 0.28, "learning_rate": 7.218481050883286e-05, "loss": 1.043, "step": 111575 }, { "epoch": 0.28, "learning_rate": 7.218355226735115e-05, "loss": 1.023, "step": 111580 }, { "epoch": 0.28, "learning_rate": 7.218229402586944e-05, "loss": 1.0198, "step": 111585 }, { "epoch": 0.28, "learning_rate": 7.218103578438775e-05, "loss": 1.0243, "step": 111590 }, { "epoch": 0.28, "learning_rate": 7.217977754290604e-05, "loss": 1.0235, "step": 111595 }, { "epoch": 0.28, "learning_rate": 7.217851930142433e-05, "loss": 1.0236, "step": 111600 }, { "epoch": 0.28, "learning_rate": 7.217726105994262e-05, "loss": 1.022, "step": 111605 }, { "epoch": 0.28, "learning_rate": 7.217600281846091e-05, "loss": 1.025, "step": 111610 }, { "epoch": 0.28, "learning_rate": 7.217474457697922e-05, "loss": 1.0222, "step": 111615 }, { "epoch": 0.28, "learning_rate": 7.217348633549751e-05, "loss": 1.0233, "step": 111620 }, { "epoch": 0.28, "learning_rate": 7.21722280940158e-05, "loss": 1.0242, "step": 111625 }, { "epoch": 0.28, "learning_rate": 7.21709698525341e-05, "loss": 1.0223, "step": 111630 }, { "epoch": 0.28, "learning_rate": 7.21697116110524e-05, "loss": 1.0204, "step": 111635 }, { "epoch": 0.28, "learning_rate": 7.216845336957069e-05, "loss": 1.0209, "step": 111640 }, { "epoch": 0.28, "learning_rate": 7.216719512808898e-05, "loss": 1.0233, "step": 111645 }, { "epoch": 0.28, "learning_rate": 7.216593688660727e-05, "loss": 1.0234, "step": 111650 }, { "epoch": 0.28, "learning_rate": 7.216467864512558e-05, "loss": 1.0209, "step": 111655 }, { "epoch": 0.28, "learning_rate": 7.216342040364387e-05, "loss": 1.0204, "step": 111660 }, { "epoch": 0.28, "learning_rate": 7.216216216216216e-05, "loss": 1.0229, "step": 111665 }, { "epoch": 0.28, "learning_rate": 7.216090392068045e-05, "loss": 1.0189, "step": 111670 }, { "epoch": 0.28, "learning_rate": 7.215964567919874e-05, "loss": 1.042, "step": 111675 }, { "epoch": 0.28, "learning_rate": 7.215838743771705e-05, "loss": 1.0215, "step": 111680 }, { "epoch": 0.28, "learning_rate": 7.215712919623534e-05, "loss": 1.0204, "step": 111685 }, { "epoch": 0.28, "learning_rate": 7.215587095475363e-05, "loss": 1.0227, "step": 111690 }, { "epoch": 0.28, "learning_rate": 7.215461271327192e-05, "loss": 1.0205, "step": 111695 }, { "epoch": 0.28, "learning_rate": 7.215335447179023e-05, "loss": 1.0218, "step": 111700 }, { "epoch": 0.28, "learning_rate": 7.215209623030852e-05, "loss": 1.0219, "step": 111705 }, { "epoch": 0.28, "learning_rate": 7.215083798882681e-05, "loss": 1.0234, "step": 111710 }, { "epoch": 0.28, "learning_rate": 7.21495797473451e-05, "loss": 1.0444, "step": 111715 }, { "epoch": 0.28, "learning_rate": 7.214832150586341e-05, "loss": 1.021, "step": 111720 }, { "epoch": 0.28, "learning_rate": 7.214706326438172e-05, "loss": 1.0229, "step": 111725 }, { "epoch": 0.28, "learning_rate": 7.21458050229e-05, "loss": 1.0241, "step": 111730 }, { "epoch": 0.28, "learning_rate": 7.21445467814183e-05, "loss": 1.023, "step": 111735 }, { "epoch": 0.28, "learning_rate": 7.214328853993659e-05, "loss": 1.0251, "step": 111740 }, { "epoch": 0.28, "learning_rate": 7.214203029845488e-05, "loss": 1.0233, "step": 111745 }, { "epoch": 0.28, "learning_rate": 7.214077205697319e-05, "loss": 1.0204, "step": 111750 }, { "epoch": 0.28, "learning_rate": 7.213951381549148e-05, "loss": 1.0194, "step": 111755 }, { "epoch": 0.28, "learning_rate": 7.213825557400977e-05, "loss": 1.0232, "step": 111760 }, { "epoch": 0.28, "learning_rate": 7.213699733252806e-05, "loss": 1.0219, "step": 111765 }, { "epoch": 0.28, "learning_rate": 7.213573909104637e-05, "loss": 1.0201, "step": 111770 }, { "epoch": 0.28, "learning_rate": 7.213448084956466e-05, "loss": 1.0238, "step": 111775 }, { "epoch": 0.28, "learning_rate": 7.213322260808295e-05, "loss": 1.0213, "step": 111780 }, { "epoch": 0.28, "learning_rate": 7.213196436660124e-05, "loss": 1.0224, "step": 111785 }, { "epoch": 0.28, "learning_rate": 7.213070612511955e-05, "loss": 1.0217, "step": 111790 }, { "epoch": 0.28, "learning_rate": 7.212944788363784e-05, "loss": 1.0229, "step": 111795 }, { "epoch": 0.28, "learning_rate": 7.212818964215613e-05, "loss": 1.0246, "step": 111800 }, { "epoch": 0.28, "learning_rate": 7.212693140067442e-05, "loss": 1.0223, "step": 111805 }, { "epoch": 0.28, "learning_rate": 7.212567315919271e-05, "loss": 1.0208, "step": 111810 }, { "epoch": 0.28, "learning_rate": 7.212441491771102e-05, "loss": 1.0211, "step": 111815 }, { "epoch": 0.28, "learning_rate": 7.212315667622931e-05, "loss": 1.022, "step": 111820 }, { "epoch": 0.28, "learning_rate": 7.21218984347476e-05, "loss": 1.0222, "step": 111825 }, { "epoch": 0.28, "learning_rate": 7.212064019326589e-05, "loss": 1.0223, "step": 111830 }, { "epoch": 0.28, "learning_rate": 7.21193819517842e-05, "loss": 1.0232, "step": 111835 }, { "epoch": 0.28, "learning_rate": 7.211812371030249e-05, "loss": 1.0231, "step": 111840 }, { "epoch": 0.28, "learning_rate": 7.211686546882078e-05, "loss": 1.0226, "step": 111845 }, { "epoch": 0.28, "learning_rate": 7.211560722733907e-05, "loss": 1.0225, "step": 111850 }, { "epoch": 0.28, "learning_rate": 7.211434898585738e-05, "loss": 1.0216, "step": 111855 }, { "epoch": 0.28, "learning_rate": 7.211309074437567e-05, "loss": 1.0221, "step": 111860 }, { "epoch": 0.28, "learning_rate": 7.211183250289396e-05, "loss": 1.0197, "step": 111865 }, { "epoch": 0.28, "learning_rate": 7.211057426141225e-05, "loss": 1.0213, "step": 111870 }, { "epoch": 0.28, "learning_rate": 7.210931601993054e-05, "loss": 1.0246, "step": 111875 }, { "epoch": 0.28, "learning_rate": 7.210805777844885e-05, "loss": 1.0206, "step": 111880 }, { "epoch": 0.28, "learning_rate": 7.210679953696714e-05, "loss": 1.0228, "step": 111885 }, { "epoch": 0.28, "learning_rate": 7.210554129548543e-05, "loss": 1.0217, "step": 111890 }, { "epoch": 0.28, "learning_rate": 7.210428305400372e-05, "loss": 1.0217, "step": 111895 }, { "epoch": 0.28, "learning_rate": 7.210302481252203e-05, "loss": 1.025, "step": 111900 }, { "epoch": 0.28, "learning_rate": 7.210176657104032e-05, "loss": 1.0234, "step": 111905 }, { "epoch": 0.28, "learning_rate": 7.210050832955861e-05, "loss": 1.0427, "step": 111910 }, { "epoch": 0.28, "learning_rate": 7.20992500880769e-05, "loss": 1.0234, "step": 111915 }, { "epoch": 0.28, "learning_rate": 7.209799184659521e-05, "loss": 1.0221, "step": 111920 }, { "epoch": 0.28, "learning_rate": 7.20967336051135e-05, "loss": 1.0224, "step": 111925 }, { "epoch": 0.28, "learning_rate": 7.209547536363179e-05, "loss": 1.0224, "step": 111930 }, { "epoch": 0.28, "learning_rate": 7.209421712215008e-05, "loss": 1.0193, "step": 111935 }, { "epoch": 0.28, "learning_rate": 7.209295888066838e-05, "loss": 1.0373, "step": 111940 }, { "epoch": 0.28, "learning_rate": 7.209170063918668e-05, "loss": 1.0215, "step": 111945 }, { "epoch": 0.28, "learning_rate": 7.209044239770497e-05, "loss": 1.0253, "step": 111950 }, { "epoch": 0.28, "learning_rate": 7.208918415622326e-05, "loss": 1.0233, "step": 111955 }, { "epoch": 0.28, "learning_rate": 7.208792591474156e-05, "loss": 1.021, "step": 111960 }, { "epoch": 0.28, "learning_rate": 7.208666767325986e-05, "loss": 1.0199, "step": 111965 }, { "epoch": 0.28, "learning_rate": 7.208540943177815e-05, "loss": 1.024, "step": 111970 }, { "epoch": 0.28, "learning_rate": 7.208415119029644e-05, "loss": 1.0213, "step": 111975 }, { "epoch": 0.28, "learning_rate": 7.208289294881474e-05, "loss": 1.0221, "step": 111980 }, { "epoch": 0.28, "learning_rate": 7.208163470733304e-05, "loss": 1.0223, "step": 111985 }, { "epoch": 0.28, "learning_rate": 7.208037646585133e-05, "loss": 1.0223, "step": 111990 }, { "epoch": 0.28, "learning_rate": 7.207911822436962e-05, "loss": 1.0234, "step": 111995 }, { "epoch": 0.28, "learning_rate": 7.207785998288792e-05, "loss": 1.0224, "step": 112000 }, { "epoch": 0.28, "learning_rate": 7.207660174140621e-05, "loss": 1.0218, "step": 112005 }, { "epoch": 0.28, "learning_rate": 7.207534349992451e-05, "loss": 1.021, "step": 112010 }, { "epoch": 0.28, "learning_rate": 7.20740852584428e-05, "loss": 1.0214, "step": 112015 }, { "epoch": 0.28, "learning_rate": 7.20728270169611e-05, "loss": 1.0225, "step": 112020 }, { "epoch": 0.28, "learning_rate": 7.207156877547939e-05, "loss": 1.0435, "step": 112025 }, { "epoch": 0.28, "learning_rate": 7.207031053399769e-05, "loss": 1.0212, "step": 112030 }, { "epoch": 0.28, "learning_rate": 7.206905229251598e-05, "loss": 1.0212, "step": 112035 }, { "epoch": 0.28, "learning_rate": 7.206779405103428e-05, "loss": 1.0209, "step": 112040 }, { "epoch": 0.28, "learning_rate": 7.206653580955257e-05, "loss": 1.022, "step": 112045 }, { "epoch": 0.28, "learning_rate": 7.206527756807087e-05, "loss": 1.0219, "step": 112050 }, { "epoch": 0.28, "learning_rate": 7.206401932658916e-05, "loss": 1.0213, "step": 112055 }, { "epoch": 0.28, "learning_rate": 7.206276108510746e-05, "loss": 1.0215, "step": 112060 }, { "epoch": 0.28, "learning_rate": 7.206150284362575e-05, "loss": 1.0235, "step": 112065 }, { "epoch": 0.28, "learning_rate": 7.206024460214404e-05, "loss": 1.0221, "step": 112070 }, { "epoch": 0.28, "learning_rate": 7.205898636066234e-05, "loss": 1.0223, "step": 112075 }, { "epoch": 0.28, "learning_rate": 7.205772811918064e-05, "loss": 1.0241, "step": 112080 }, { "epoch": 0.28, "learning_rate": 7.205646987769893e-05, "loss": 1.0203, "step": 112085 }, { "epoch": 0.28, "learning_rate": 7.205521163621722e-05, "loss": 1.0221, "step": 112090 }, { "epoch": 0.28, "learning_rate": 7.205395339473552e-05, "loss": 1.0216, "step": 112095 }, { "epoch": 0.28, "learning_rate": 7.205269515325382e-05, "loss": 1.0217, "step": 112100 }, { "epoch": 0.28, "learning_rate": 7.205143691177211e-05, "loss": 1.023, "step": 112105 }, { "epoch": 0.28, "learning_rate": 7.20501786702904e-05, "loss": 1.0214, "step": 112110 }, { "epoch": 0.28, "learning_rate": 7.20489204288087e-05, "loss": 1.0245, "step": 112115 }, { "epoch": 0.28, "learning_rate": 7.2047662187327e-05, "loss": 1.0211, "step": 112120 }, { "epoch": 0.28, "learning_rate": 7.204640394584529e-05, "loss": 1.0222, "step": 112125 }, { "epoch": 0.28, "learning_rate": 7.204514570436358e-05, "loss": 1.0231, "step": 112130 }, { "epoch": 0.28, "learning_rate": 7.204388746288187e-05, "loss": 1.0225, "step": 112135 }, { "epoch": 0.28, "learning_rate": 7.204262922140018e-05, "loss": 1.0252, "step": 112140 }, { "epoch": 0.28, "learning_rate": 7.204137097991847e-05, "loss": 1.0217, "step": 112145 }, { "epoch": 0.28, "learning_rate": 7.204011273843676e-05, "loss": 1.0225, "step": 112150 }, { "epoch": 0.28, "learning_rate": 7.203885449695505e-05, "loss": 1.0195, "step": 112155 }, { "epoch": 0.28, "learning_rate": 7.203759625547336e-05, "loss": 1.0243, "step": 112160 }, { "epoch": 0.28, "learning_rate": 7.203633801399165e-05, "loss": 1.0225, "step": 112165 }, { "epoch": 0.28, "learning_rate": 7.203507977250994e-05, "loss": 1.0219, "step": 112170 }, { "epoch": 0.28, "learning_rate": 7.203382153102823e-05, "loss": 1.0237, "step": 112175 }, { "epoch": 0.28, "learning_rate": 7.203256328954654e-05, "loss": 1.0235, "step": 112180 }, { "epoch": 0.28, "learning_rate": 7.203130504806483e-05, "loss": 1.0227, "step": 112185 }, { "epoch": 0.28, "learning_rate": 7.203004680658312e-05, "loss": 1.0232, "step": 112190 }, { "epoch": 0.28, "learning_rate": 7.202878856510141e-05, "loss": 1.0206, "step": 112195 }, { "epoch": 0.28, "learning_rate": 7.20275303236197e-05, "loss": 1.0223, "step": 112200 }, { "epoch": 0.28, "learning_rate": 7.2026272082138e-05, "loss": 1.0235, "step": 112205 }, { "epoch": 0.28, "learning_rate": 7.20250138406563e-05, "loss": 1.0227, "step": 112210 }, { "epoch": 0.28, "learning_rate": 7.202375559917459e-05, "loss": 1.0244, "step": 112215 }, { "epoch": 0.28, "learning_rate": 7.20224973576929e-05, "loss": 1.0201, "step": 112220 }, { "epoch": 0.28, "learning_rate": 7.202123911621119e-05, "loss": 1.0231, "step": 112225 }, { "epoch": 0.28, "learning_rate": 7.201998087472949e-05, "loss": 1.0229, "step": 112230 }, { "epoch": 0.28, "learning_rate": 7.201872263324778e-05, "loss": 1.0235, "step": 112235 }, { "epoch": 0.28, "learning_rate": 7.201746439176607e-05, "loss": 1.023, "step": 112240 }, { "epoch": 0.28, "learning_rate": 7.201620615028437e-05, "loss": 1.0226, "step": 112245 }, { "epoch": 0.28, "learning_rate": 7.201494790880267e-05, "loss": 1.0229, "step": 112250 }, { "epoch": 0.28, "learning_rate": 7.201368966732096e-05, "loss": 1.024, "step": 112255 }, { "epoch": 0.28, "learning_rate": 7.201243142583925e-05, "loss": 1.0235, "step": 112260 }, { "epoch": 0.28, "learning_rate": 7.201117318435755e-05, "loss": 1.0211, "step": 112265 }, { "epoch": 0.28, "learning_rate": 7.200991494287584e-05, "loss": 1.0232, "step": 112270 }, { "epoch": 0.28, "learning_rate": 7.200865670139414e-05, "loss": 1.0222, "step": 112275 }, { "epoch": 0.28, "learning_rate": 7.200739845991243e-05, "loss": 1.0207, "step": 112280 }, { "epoch": 0.28, "learning_rate": 7.200614021843073e-05, "loss": 1.023, "step": 112285 }, { "epoch": 0.28, "learning_rate": 7.200488197694902e-05, "loss": 1.0211, "step": 112290 }, { "epoch": 0.28, "learning_rate": 7.200362373546732e-05, "loss": 1.0228, "step": 112295 }, { "epoch": 0.28, "learning_rate": 7.200236549398561e-05, "loss": 1.0234, "step": 112300 }, { "epoch": 0.28, "learning_rate": 7.20011072525039e-05, "loss": 1.0226, "step": 112305 }, { "epoch": 0.28, "learning_rate": 7.19998490110222e-05, "loss": 1.0203, "step": 112310 }, { "epoch": 0.28, "learning_rate": 7.19985907695405e-05, "loss": 1.021, "step": 112315 }, { "epoch": 0.28, "learning_rate": 7.19973325280588e-05, "loss": 1.023, "step": 112320 }, { "epoch": 0.28, "learning_rate": 7.199607428657709e-05, "loss": 1.0239, "step": 112325 }, { "epoch": 0.28, "learning_rate": 7.199481604509538e-05, "loss": 1.0217, "step": 112330 }, { "epoch": 0.28, "learning_rate": 7.199355780361367e-05, "loss": 1.0193, "step": 112335 }, { "epoch": 0.28, "learning_rate": 7.199229956213197e-05, "loss": 1.0205, "step": 112340 }, { "epoch": 0.28, "learning_rate": 7.199104132065027e-05, "loss": 1.0217, "step": 112345 }, { "epoch": 0.28, "learning_rate": 7.198978307916856e-05, "loss": 1.0227, "step": 112350 }, { "epoch": 0.28, "learning_rate": 7.198852483768685e-05, "loss": 1.0248, "step": 112355 }, { "epoch": 0.28, "learning_rate": 7.198726659620515e-05, "loss": 1.022, "step": 112360 }, { "epoch": 0.28, "learning_rate": 7.198600835472345e-05, "loss": 1.0249, "step": 112365 }, { "epoch": 0.28, "learning_rate": 7.198475011324174e-05, "loss": 1.0201, "step": 112370 }, { "epoch": 0.28, "learning_rate": 7.198349187176003e-05, "loss": 1.0227, "step": 112375 }, { "epoch": 0.28, "learning_rate": 7.198223363027833e-05, "loss": 1.0209, "step": 112380 }, { "epoch": 0.28, "learning_rate": 7.198097538879663e-05, "loss": 1.0226, "step": 112385 }, { "epoch": 0.28, "learning_rate": 7.197971714731492e-05, "loss": 1.024, "step": 112390 }, { "epoch": 0.28, "learning_rate": 7.197845890583321e-05, "loss": 1.0229, "step": 112395 }, { "epoch": 0.28, "learning_rate": 7.19772006643515e-05, "loss": 1.0201, "step": 112400 }, { "epoch": 0.28, "learning_rate": 7.19759424228698e-05, "loss": 1.0238, "step": 112405 }, { "epoch": 0.28, "learning_rate": 7.19746841813881e-05, "loss": 1.0208, "step": 112410 }, { "epoch": 0.28, "learning_rate": 7.197342593990639e-05, "loss": 1.0239, "step": 112415 }, { "epoch": 0.28, "learning_rate": 7.197216769842468e-05, "loss": 1.0223, "step": 112420 }, { "epoch": 0.28, "learning_rate": 7.197090945694299e-05, "loss": 1.0215, "step": 112425 }, { "epoch": 0.28, "learning_rate": 7.196965121546128e-05, "loss": 1.0222, "step": 112430 }, { "epoch": 0.28, "learning_rate": 7.196839297397957e-05, "loss": 1.0183, "step": 112435 }, { "epoch": 0.28, "learning_rate": 7.196713473249786e-05, "loss": 1.0205, "step": 112440 }, { "epoch": 0.28, "learning_rate": 7.196587649101617e-05, "loss": 1.0222, "step": 112445 }, { "epoch": 0.28, "learning_rate": 7.196461824953446e-05, "loss": 1.0227, "step": 112450 }, { "epoch": 0.28, "learning_rate": 7.196336000805275e-05, "loss": 1.0218, "step": 112455 }, { "epoch": 0.28, "learning_rate": 7.196210176657104e-05, "loss": 1.0234, "step": 112460 }, { "epoch": 0.28, "learning_rate": 7.196084352508933e-05, "loss": 1.0237, "step": 112465 }, { "epoch": 0.28, "learning_rate": 7.195958528360764e-05, "loss": 1.0495, "step": 112470 }, { "epoch": 0.28, "learning_rate": 7.195832704212593e-05, "loss": 1.0248, "step": 112475 }, { "epoch": 0.28, "learning_rate": 7.195706880064422e-05, "loss": 1.0242, "step": 112480 }, { "epoch": 0.28, "learning_rate": 7.195581055916251e-05, "loss": 1.0223, "step": 112485 }, { "epoch": 0.28, "learning_rate": 7.195455231768082e-05, "loss": 1.0246, "step": 112490 }, { "epoch": 0.28, "learning_rate": 7.195329407619911e-05, "loss": 1.0214, "step": 112495 }, { "epoch": 0.28, "learning_rate": 7.19520358347174e-05, "loss": 1.0203, "step": 112500 }, { "epoch": 0.28, "learning_rate": 7.195077759323569e-05, "loss": 1.0215, "step": 112505 }, { "epoch": 0.28, "learning_rate": 7.1949519351754e-05, "loss": 1.0231, "step": 112510 }, { "epoch": 0.28, "learning_rate": 7.194826111027229e-05, "loss": 1.0214, "step": 112515 }, { "epoch": 0.28, "learning_rate": 7.194700286879058e-05, "loss": 1.0202, "step": 112520 }, { "epoch": 0.28, "learning_rate": 7.194574462730887e-05, "loss": 1.0218, "step": 112525 }, { "epoch": 0.28, "learning_rate": 7.194448638582716e-05, "loss": 1.021, "step": 112530 }, { "epoch": 0.28, "learning_rate": 7.194322814434547e-05, "loss": 1.0211, "step": 112535 }, { "epoch": 0.28, "learning_rate": 7.194196990286376e-05, "loss": 1.0238, "step": 112540 }, { "epoch": 0.28, "learning_rate": 7.194071166138205e-05, "loss": 1.0242, "step": 112545 }, { "epoch": 0.28, "learning_rate": 7.193945341990034e-05, "loss": 1.023, "step": 112550 }, { "epoch": 0.28, "learning_rate": 7.193819517841865e-05, "loss": 1.0213, "step": 112555 }, { "epoch": 0.28, "learning_rate": 7.193693693693694e-05, "loss": 1.02, "step": 112560 }, { "epoch": 0.28, "learning_rate": 7.193567869545523e-05, "loss": 1.0232, "step": 112565 }, { "epoch": 0.28, "learning_rate": 7.193442045397352e-05, "loss": 1.0237, "step": 112570 }, { "epoch": 0.28, "learning_rate": 7.193316221249183e-05, "loss": 1.0243, "step": 112575 }, { "epoch": 0.28, "learning_rate": 7.193190397101012e-05, "loss": 1.0227, "step": 112580 }, { "epoch": 0.28, "learning_rate": 7.193064572952841e-05, "loss": 1.0233, "step": 112585 }, { "epoch": 0.28, "learning_rate": 7.19293874880467e-05, "loss": 1.0198, "step": 112590 }, { "epoch": 0.28, "learning_rate": 7.1928129246565e-05, "loss": 1.0202, "step": 112595 }, { "epoch": 0.28, "learning_rate": 7.19268710050833e-05, "loss": 1.022, "step": 112600 }, { "epoch": 0.28, "learning_rate": 7.192561276360159e-05, "loss": 1.0228, "step": 112605 }, { "epoch": 0.28, "learning_rate": 7.192435452211988e-05, "loss": 1.02, "step": 112610 }, { "epoch": 0.28, "learning_rate": 7.192309628063818e-05, "loss": 1.0219, "step": 112615 }, { "epoch": 0.28, "learning_rate": 7.192183803915648e-05, "loss": 1.0232, "step": 112620 }, { "epoch": 0.28, "learning_rate": 7.192057979767477e-05, "loss": 1.0229, "step": 112625 }, { "epoch": 0.28, "learning_rate": 7.191932155619306e-05, "loss": 1.0189, "step": 112630 }, { "epoch": 0.28, "learning_rate": 7.191806331471136e-05, "loss": 1.0219, "step": 112635 }, { "epoch": 0.28, "learning_rate": 7.191680507322966e-05, "loss": 1.0223, "step": 112640 }, { "epoch": 0.28, "learning_rate": 7.191554683174795e-05, "loss": 1.0247, "step": 112645 }, { "epoch": 0.28, "learning_rate": 7.191428859026624e-05, "loss": 1.0228, "step": 112650 }, { "epoch": 0.28, "learning_rate": 7.191303034878453e-05, "loss": 1.0223, "step": 112655 }, { "epoch": 0.28, "learning_rate": 7.191177210730283e-05, "loss": 1.0224, "step": 112660 }, { "epoch": 0.28, "learning_rate": 7.191051386582113e-05, "loss": 1.0221, "step": 112665 }, { "epoch": 0.28, "learning_rate": 7.190925562433942e-05, "loss": 1.0234, "step": 112670 }, { "epoch": 0.28, "learning_rate": 7.190799738285771e-05, "loss": 1.0196, "step": 112675 }, { "epoch": 0.28, "learning_rate": 7.1906739141376e-05, "loss": 1.0209, "step": 112680 }, { "epoch": 0.28, "learning_rate": 7.190548089989431e-05, "loss": 1.0224, "step": 112685 }, { "epoch": 0.28, "learning_rate": 7.19042226584126e-05, "loss": 1.0215, "step": 112690 }, { "epoch": 0.28, "learning_rate": 7.19029644169309e-05, "loss": 1.0213, "step": 112695 }, { "epoch": 0.28, "learning_rate": 7.190170617544919e-05, "loss": 1.0216, "step": 112700 }, { "epoch": 0.28, "learning_rate": 7.190044793396749e-05, "loss": 1.0215, "step": 112705 }, { "epoch": 0.28, "learning_rate": 7.189918969248578e-05, "loss": 1.0206, "step": 112710 }, { "epoch": 0.28, "learning_rate": 7.189793145100407e-05, "loss": 1.0226, "step": 112715 }, { "epoch": 0.28, "learning_rate": 7.189667320952238e-05, "loss": 1.0222, "step": 112720 }, { "epoch": 0.28, "learning_rate": 7.189541496804067e-05, "loss": 1.0237, "step": 112725 }, { "epoch": 0.28, "learning_rate": 7.189415672655898e-05, "loss": 1.0231, "step": 112730 }, { "epoch": 0.28, "learning_rate": 7.189289848507727e-05, "loss": 1.0217, "step": 112735 }, { "epoch": 0.28, "learning_rate": 7.189164024359556e-05, "loss": 1.0219, "step": 112740 }, { "epoch": 0.28, "learning_rate": 7.189038200211385e-05, "loss": 1.0226, "step": 112745 }, { "epoch": 0.28, "learning_rate": 7.188912376063214e-05, "loss": 1.0227, "step": 112750 }, { "epoch": 0.28, "learning_rate": 7.188786551915045e-05, "loss": 1.0225, "step": 112755 }, { "epoch": 0.28, "learning_rate": 7.188660727766874e-05, "loss": 1.022, "step": 112760 }, { "epoch": 0.28, "learning_rate": 7.188534903618703e-05, "loss": 1.0218, "step": 112765 }, { "epoch": 0.28, "learning_rate": 7.188409079470532e-05, "loss": 1.0225, "step": 112770 }, { "epoch": 0.28, "learning_rate": 7.188283255322363e-05, "loss": 1.0221, "step": 112775 }, { "epoch": 0.28, "learning_rate": 7.188157431174192e-05, "loss": 1.0238, "step": 112780 }, { "epoch": 0.28, "learning_rate": 7.188031607026021e-05, "loss": 1.0223, "step": 112785 }, { "epoch": 0.28, "learning_rate": 7.18790578287785e-05, "loss": 1.0245, "step": 112790 }, { "epoch": 0.28, "learning_rate": 7.18777995872968e-05, "loss": 1.023, "step": 112795 }, { "epoch": 0.28, "learning_rate": 7.18765413458151e-05, "loss": 1.0233, "step": 112800 }, { "epoch": 0.28, "learning_rate": 7.187528310433339e-05, "loss": 1.0231, "step": 112805 }, { "epoch": 0.28, "learning_rate": 7.187402486285168e-05, "loss": 1.0248, "step": 112810 }, { "epoch": 0.28, "learning_rate": 7.187276662136997e-05, "loss": 1.0242, "step": 112815 }, { "epoch": 0.28, "learning_rate": 7.187150837988828e-05, "loss": 1.023, "step": 112820 }, { "epoch": 0.28, "learning_rate": 7.187025013840657e-05, "loss": 1.022, "step": 112825 }, { "epoch": 0.28, "learning_rate": 7.186899189692486e-05, "loss": 1.0208, "step": 112830 }, { "epoch": 0.28, "learning_rate": 7.186773365544315e-05, "loss": 1.0193, "step": 112835 }, { "epoch": 0.28, "learning_rate": 7.186647541396146e-05, "loss": 1.0214, "step": 112840 }, { "epoch": 0.28, "learning_rate": 7.186521717247975e-05, "loss": 1.0225, "step": 112845 }, { "epoch": 0.28, "learning_rate": 7.186395893099804e-05, "loss": 1.021, "step": 112850 }, { "epoch": 0.28, "learning_rate": 7.186270068951633e-05, "loss": 1.025, "step": 112855 }, { "epoch": 0.28, "learning_rate": 7.186144244803463e-05, "loss": 1.0228, "step": 112860 }, { "epoch": 0.28, "learning_rate": 7.186018420655293e-05, "loss": 1.0226, "step": 112865 }, { "epoch": 0.28, "learning_rate": 7.185892596507122e-05, "loss": 1.0198, "step": 112870 }, { "epoch": 0.28, "learning_rate": 7.185766772358951e-05, "loss": 1.0251, "step": 112875 }, { "epoch": 0.28, "learning_rate": 7.18564094821078e-05, "loss": 1.022, "step": 112880 }, { "epoch": 0.28, "learning_rate": 7.185515124062611e-05, "loss": 1.0228, "step": 112885 }, { "epoch": 0.28, "learning_rate": 7.18538929991444e-05, "loss": 1.0224, "step": 112890 }, { "epoch": 0.28, "learning_rate": 7.18526347576627e-05, "loss": 1.0236, "step": 112895 }, { "epoch": 0.28, "learning_rate": 7.185137651618099e-05, "loss": 1.0262, "step": 112900 }, { "epoch": 0.28, "learning_rate": 7.185011827469929e-05, "loss": 1.0233, "step": 112905 }, { "epoch": 0.28, "learning_rate": 7.184886003321758e-05, "loss": 1.0225, "step": 112910 }, { "epoch": 0.28, "learning_rate": 7.184760179173587e-05, "loss": 1.0239, "step": 112915 }, { "epoch": 0.28, "learning_rate": 7.184634355025417e-05, "loss": 1.0231, "step": 112920 }, { "epoch": 0.28, "learning_rate": 7.184508530877246e-05, "loss": 1.0219, "step": 112925 }, { "epoch": 0.28, "learning_rate": 7.184382706729076e-05, "loss": 1.0215, "step": 112930 }, { "epoch": 0.28, "learning_rate": 7.184256882580905e-05, "loss": 1.025, "step": 112935 }, { "epoch": 0.28, "learning_rate": 7.184131058432735e-05, "loss": 1.0239, "step": 112940 }, { "epoch": 0.28, "learning_rate": 7.184005234284564e-05, "loss": 1.0215, "step": 112945 }, { "epoch": 0.28, "learning_rate": 7.183879410136394e-05, "loss": 1.0216, "step": 112950 }, { "epoch": 0.28, "learning_rate": 7.183753585988223e-05, "loss": 1.0211, "step": 112955 }, { "epoch": 0.28, "learning_rate": 7.183627761840053e-05, "loss": 1.025, "step": 112960 }, { "epoch": 0.28, "learning_rate": 7.183501937691882e-05, "loss": 1.0226, "step": 112965 }, { "epoch": 0.28, "learning_rate": 7.183376113543712e-05, "loss": 1.0242, "step": 112970 }, { "epoch": 0.28, "learning_rate": 7.183250289395541e-05, "loss": 1.0233, "step": 112975 }, { "epoch": 0.28, "learning_rate": 7.18312446524737e-05, "loss": 1.0221, "step": 112980 }, { "epoch": 0.28, "learning_rate": 7.1829986410992e-05, "loss": 1.0231, "step": 112985 }, { "epoch": 0.28, "learning_rate": 7.182872816951029e-05, "loss": 1.0212, "step": 112990 }, { "epoch": 0.28, "learning_rate": 7.18274699280286e-05, "loss": 1.0214, "step": 112995 }, { "epoch": 0.28, "learning_rate": 7.182621168654689e-05, "loss": 1.045, "step": 113000 }, { "epoch": 0.28, "learning_rate": 7.182495344506518e-05, "loss": 1.0213, "step": 113005 }, { "epoch": 0.28, "learning_rate": 7.182369520358347e-05, "loss": 1.0212, "step": 113010 }, { "epoch": 0.28, "learning_rate": 7.182243696210177e-05, "loss": 1.0229, "step": 113015 }, { "epoch": 0.28, "learning_rate": 7.182117872062007e-05, "loss": 1.0236, "step": 113020 }, { "epoch": 0.28, "learning_rate": 7.181992047913836e-05, "loss": 1.0224, "step": 113025 }, { "epoch": 0.28, "learning_rate": 7.181866223765665e-05, "loss": 1.0219, "step": 113030 }, { "epoch": 0.28, "learning_rate": 7.181740399617495e-05, "loss": 1.0246, "step": 113035 }, { "epoch": 0.28, "learning_rate": 7.181614575469325e-05, "loss": 1.0223, "step": 113040 }, { "epoch": 0.28, "learning_rate": 7.181488751321154e-05, "loss": 1.0213, "step": 113045 }, { "epoch": 0.28, "learning_rate": 7.181362927172983e-05, "loss": 1.0229, "step": 113050 }, { "epoch": 0.28, "learning_rate": 7.181237103024812e-05, "loss": 1.0229, "step": 113055 }, { "epoch": 0.28, "learning_rate": 7.181111278876643e-05, "loss": 1.0231, "step": 113060 }, { "epoch": 0.28, "learning_rate": 7.180985454728472e-05, "loss": 1.0217, "step": 113065 }, { "epoch": 0.28, "learning_rate": 7.180859630580301e-05, "loss": 1.0221, "step": 113070 }, { "epoch": 0.28, "learning_rate": 7.18073380643213e-05, "loss": 1.0245, "step": 113075 }, { "epoch": 0.28, "learning_rate": 7.18060798228396e-05, "loss": 1.0201, "step": 113080 }, { "epoch": 0.28, "learning_rate": 7.18048215813579e-05, "loss": 1.0223, "step": 113085 }, { "epoch": 0.28, "learning_rate": 7.180356333987619e-05, "loss": 1.0234, "step": 113090 }, { "epoch": 0.28, "learning_rate": 7.180230509839448e-05, "loss": 1.022, "step": 113095 }, { "epoch": 0.28, "learning_rate": 7.180104685691279e-05, "loss": 1.0218, "step": 113100 }, { "epoch": 0.28, "learning_rate": 7.179978861543108e-05, "loss": 1.0236, "step": 113105 }, { "epoch": 0.28, "learning_rate": 7.179853037394937e-05, "loss": 1.0209, "step": 113110 }, { "epoch": 0.28, "learning_rate": 7.179727213246766e-05, "loss": 1.022, "step": 113115 }, { "epoch": 0.28, "learning_rate": 7.179601389098595e-05, "loss": 1.0224, "step": 113120 }, { "epoch": 0.28, "learning_rate": 7.179475564950426e-05, "loss": 1.022, "step": 113125 }, { "epoch": 0.28, "learning_rate": 7.179349740802255e-05, "loss": 1.0213, "step": 113130 }, { "epoch": 0.28, "learning_rate": 7.179223916654084e-05, "loss": 1.021, "step": 113135 }, { "epoch": 0.28, "learning_rate": 7.179098092505913e-05, "loss": 1.0219, "step": 113140 }, { "epoch": 0.28, "learning_rate": 7.178972268357744e-05, "loss": 1.0216, "step": 113145 }, { "epoch": 0.28, "learning_rate": 7.178846444209573e-05, "loss": 1.0222, "step": 113150 }, { "epoch": 0.28, "learning_rate": 7.178720620061402e-05, "loss": 1.0231, "step": 113155 }, { "epoch": 0.28, "learning_rate": 7.178594795913231e-05, "loss": 1.0234, "step": 113160 }, { "epoch": 0.28, "learning_rate": 7.178468971765062e-05, "loss": 1.0202, "step": 113165 }, { "epoch": 0.28, "learning_rate": 7.178343147616891e-05, "loss": 1.0244, "step": 113170 }, { "epoch": 0.28, "learning_rate": 7.17821732346872e-05, "loss": 1.0235, "step": 113175 }, { "epoch": 0.28, "learning_rate": 7.178091499320549e-05, "loss": 1.0227, "step": 113180 }, { "epoch": 0.28, "learning_rate": 7.177965675172378e-05, "loss": 1.0219, "step": 113185 }, { "epoch": 0.28, "learning_rate": 7.177839851024209e-05, "loss": 1.0212, "step": 113190 }, { "epoch": 0.28, "learning_rate": 7.177714026876038e-05, "loss": 1.0236, "step": 113195 }, { "epoch": 0.28, "learning_rate": 7.177588202727867e-05, "loss": 1.0233, "step": 113200 }, { "epoch": 0.28, "learning_rate": 7.177462378579696e-05, "loss": 1.0223, "step": 113205 }, { "epoch": 0.28, "learning_rate": 7.177336554431527e-05, "loss": 1.021, "step": 113210 }, { "epoch": 0.28, "learning_rate": 7.177210730283356e-05, "loss": 1.0468, "step": 113215 }, { "epoch": 0.28, "learning_rate": 7.177084906135186e-05, "loss": 1.0205, "step": 113220 }, { "epoch": 0.28, "learning_rate": 7.176959081987016e-05, "loss": 1.023, "step": 113225 }, { "epoch": 0.28, "learning_rate": 7.176833257838845e-05, "loss": 1.0199, "step": 113230 }, { "epoch": 0.28, "learning_rate": 7.176707433690675e-05, "loss": 1.0226, "step": 113235 }, { "epoch": 0.28, "learning_rate": 7.176581609542504e-05, "loss": 1.0242, "step": 113240 }, { "epoch": 0.28, "learning_rate": 7.176455785394334e-05, "loss": 1.023, "step": 113245 }, { "epoch": 0.28, "learning_rate": 7.176329961246163e-05, "loss": 1.0233, "step": 113250 }, { "epoch": 0.28, "learning_rate": 7.176204137097993e-05, "loss": 1.0253, "step": 113255 }, { "epoch": 0.28, "learning_rate": 7.176078312949822e-05, "loss": 1.0196, "step": 113260 }, { "epoch": 0.28, "learning_rate": 7.175952488801652e-05, "loss": 1.0215, "step": 113265 }, { "epoch": 0.28, "learning_rate": 7.175826664653481e-05, "loss": 1.0221, "step": 113270 }, { "epoch": 0.28, "learning_rate": 7.17570084050531e-05, "loss": 1.0234, "step": 113275 }, { "epoch": 0.28, "learning_rate": 7.17557501635714e-05, "loss": 1.0224, "step": 113280 }, { "epoch": 0.28, "learning_rate": 7.17544919220897e-05, "loss": 1.0223, "step": 113285 }, { "epoch": 0.28, "learning_rate": 7.175323368060799e-05, "loss": 1.0199, "step": 113290 }, { "epoch": 0.28, "learning_rate": 7.175197543912628e-05, "loss": 1.0204, "step": 113295 }, { "epoch": 0.28, "learning_rate": 7.175071719764458e-05, "loss": 1.023, "step": 113300 }, { "epoch": 0.28, "learning_rate": 7.174945895616288e-05, "loss": 1.0214, "step": 113305 }, { "epoch": 0.28, "learning_rate": 7.174820071468117e-05, "loss": 1.0233, "step": 113310 }, { "epoch": 0.28, "learning_rate": 7.174694247319946e-05, "loss": 1.0209, "step": 113315 }, { "epoch": 0.28, "learning_rate": 7.174568423171776e-05, "loss": 1.0234, "step": 113320 }, { "epoch": 0.28, "learning_rate": 7.174442599023606e-05, "loss": 1.0208, "step": 113325 }, { "epoch": 0.28, "learning_rate": 7.174316774875435e-05, "loss": 1.0223, "step": 113330 }, { "epoch": 0.28, "learning_rate": 7.174190950727264e-05, "loss": 1.0219, "step": 113335 }, { "epoch": 0.28, "learning_rate": 7.174065126579093e-05, "loss": 1.0234, "step": 113340 }, { "epoch": 0.28, "learning_rate": 7.173939302430924e-05, "loss": 1.0232, "step": 113345 }, { "epoch": 0.28, "learning_rate": 7.173813478282753e-05, "loss": 1.0238, "step": 113350 }, { "epoch": 0.28, "learning_rate": 7.173687654134582e-05, "loss": 1.023, "step": 113355 }, { "epoch": 0.28, "learning_rate": 7.173561829986411e-05, "loss": 1.0214, "step": 113360 }, { "epoch": 0.28, "learning_rate": 7.173436005838242e-05, "loss": 1.0221, "step": 113365 }, { "epoch": 0.28, "learning_rate": 7.173310181690071e-05, "loss": 1.0195, "step": 113370 }, { "epoch": 0.28, "learning_rate": 7.1731843575419e-05, "loss": 1.0225, "step": 113375 }, { "epoch": 0.28, "learning_rate": 7.173058533393729e-05, "loss": 1.0212, "step": 113380 }, { "epoch": 0.28, "learning_rate": 7.17293270924556e-05, "loss": 1.0213, "step": 113385 }, { "epoch": 0.28, "learning_rate": 7.172806885097389e-05, "loss": 1.0251, "step": 113390 }, { "epoch": 0.28, "learning_rate": 7.172681060949218e-05, "loss": 1.0217, "step": 113395 }, { "epoch": 0.28, "learning_rate": 7.172555236801047e-05, "loss": 1.0225, "step": 113400 }, { "epoch": 0.28, "learning_rate": 7.172429412652876e-05, "loss": 1.0212, "step": 113405 }, { "epoch": 0.28, "learning_rate": 7.172303588504707e-05, "loss": 1.0199, "step": 113410 }, { "epoch": 0.28, "learning_rate": 7.172177764356536e-05, "loss": 1.0448, "step": 113415 }, { "epoch": 0.28, "learning_rate": 7.172051940208365e-05, "loss": 1.0452, "step": 113420 }, { "epoch": 0.28, "learning_rate": 7.171926116060194e-05, "loss": 1.0239, "step": 113425 }, { "epoch": 0.28, "learning_rate": 7.171800291912025e-05, "loss": 1.0221, "step": 113430 }, { "epoch": 0.28, "learning_rate": 7.171674467763854e-05, "loss": 1.0227, "step": 113435 }, { "epoch": 0.28, "learning_rate": 7.171548643615683e-05, "loss": 1.0227, "step": 113440 }, { "epoch": 0.28, "learning_rate": 7.171422819467512e-05, "loss": 1.0225, "step": 113445 }, { "epoch": 0.28, "learning_rate": 7.171296995319341e-05, "loss": 1.0238, "step": 113450 }, { "epoch": 0.28, "learning_rate": 7.171171171171172e-05, "loss": 1.0216, "step": 113455 }, { "epoch": 0.28, "learning_rate": 7.171045347023001e-05, "loss": 1.0203, "step": 113460 }, { "epoch": 0.28, "learning_rate": 7.17091952287483e-05, "loss": 1.022, "step": 113465 }, { "epoch": 0.28, "learning_rate": 7.17079369872666e-05, "loss": 1.0217, "step": 113470 }, { "epoch": 0.28, "learning_rate": 7.17066787457849e-05, "loss": 1.0228, "step": 113475 }, { "epoch": 0.28, "learning_rate": 7.170542050430319e-05, "loss": 1.0229, "step": 113480 }, { "epoch": 0.28, "learning_rate": 7.170416226282148e-05, "loss": 1.0221, "step": 113485 }, { "epoch": 0.28, "learning_rate": 7.170290402133977e-05, "loss": 1.0236, "step": 113490 }, { "epoch": 0.28, "learning_rate": 7.170164577985808e-05, "loss": 1.0222, "step": 113495 }, { "epoch": 0.28, "learning_rate": 7.170038753837637e-05, "loss": 1.0232, "step": 113500 }, { "epoch": 0.28, "learning_rate": 7.169912929689466e-05, "loss": 1.0203, "step": 113505 }, { "epoch": 0.28, "learning_rate": 7.169787105541295e-05, "loss": 1.0228, "step": 113510 }, { "epoch": 0.28, "learning_rate": 7.169661281393125e-05, "loss": 1.0192, "step": 113515 }, { "epoch": 0.28, "learning_rate": 7.169535457244955e-05, "loss": 1.0247, "step": 113520 }, { "epoch": 0.28, "learning_rate": 7.169409633096784e-05, "loss": 1.0202, "step": 113525 }, { "epoch": 0.28, "learning_rate": 7.169283808948613e-05, "loss": 1.0228, "step": 113530 }, { "epoch": 0.28, "learning_rate": 7.169157984800443e-05, "loss": 1.0223, "step": 113535 }, { "epoch": 0.29, "learning_rate": 7.169032160652273e-05, "loss": 1.0213, "step": 113540 }, { "epoch": 0.29, "learning_rate": 7.168906336504102e-05, "loss": 1.0232, "step": 113545 }, { "epoch": 0.29, "learning_rate": 7.168780512355931e-05, "loss": 1.0208, "step": 113550 }, { "epoch": 0.29, "learning_rate": 7.16865468820776e-05, "loss": 1.0215, "step": 113555 }, { "epoch": 0.29, "learning_rate": 7.168528864059591e-05, "loss": 1.0222, "step": 113560 }, { "epoch": 0.29, "learning_rate": 7.16840303991142e-05, "loss": 1.0211, "step": 113565 }, { "epoch": 0.29, "learning_rate": 7.16827721576325e-05, "loss": 1.0208, "step": 113570 }, { "epoch": 0.29, "learning_rate": 7.168151391615079e-05, "loss": 1.0217, "step": 113575 }, { "epoch": 0.29, "learning_rate": 7.168025567466908e-05, "loss": 1.0209, "step": 113580 }, { "epoch": 0.29, "learning_rate": 7.167899743318738e-05, "loss": 1.0222, "step": 113585 }, { "epoch": 0.29, "learning_rate": 7.167773919170567e-05, "loss": 1.0245, "step": 113590 }, { "epoch": 0.29, "learning_rate": 7.167648095022397e-05, "loss": 1.0216, "step": 113595 }, { "epoch": 0.29, "learning_rate": 7.167522270874226e-05, "loss": 1.0239, "step": 113600 }, { "epoch": 0.29, "learning_rate": 7.167396446726056e-05, "loss": 1.0225, "step": 113605 }, { "epoch": 0.29, "learning_rate": 7.167270622577885e-05, "loss": 1.0226, "step": 113610 }, { "epoch": 0.29, "learning_rate": 7.167144798429715e-05, "loss": 1.0247, "step": 113615 }, { "epoch": 0.29, "learning_rate": 7.167018974281544e-05, "loss": 1.0232, "step": 113620 }, { "epoch": 0.29, "learning_rate": 7.166893150133374e-05, "loss": 1.0224, "step": 113625 }, { "epoch": 0.29, "learning_rate": 7.166767325985203e-05, "loss": 1.0207, "step": 113630 }, { "epoch": 0.29, "learning_rate": 7.166641501837033e-05, "loss": 1.0226, "step": 113635 }, { "epoch": 0.29, "learning_rate": 7.166515677688862e-05, "loss": 1.0228, "step": 113640 }, { "epoch": 0.29, "learning_rate": 7.166389853540691e-05, "loss": 1.0237, "step": 113645 }, { "epoch": 0.29, "learning_rate": 7.166264029392521e-05, "loss": 1.0219, "step": 113650 }, { "epoch": 0.29, "learning_rate": 7.16613820524435e-05, "loss": 1.0247, "step": 113655 }, { "epoch": 0.29, "learning_rate": 7.16601238109618e-05, "loss": 1.0214, "step": 113660 }, { "epoch": 0.29, "learning_rate": 7.165886556948009e-05, "loss": 1.0226, "step": 113665 }, { "epoch": 0.29, "learning_rate": 7.16576073279984e-05, "loss": 1.0243, "step": 113670 }, { "epoch": 0.29, "learning_rate": 7.165634908651668e-05, "loss": 1.0222, "step": 113675 }, { "epoch": 0.29, "learning_rate": 7.165509084503498e-05, "loss": 1.0227, "step": 113680 }, { "epoch": 0.29, "learning_rate": 7.165383260355327e-05, "loss": 1.0232, "step": 113685 }, { "epoch": 0.29, "learning_rate": 7.165257436207157e-05, "loss": 1.0206, "step": 113690 }, { "epoch": 0.29, "learning_rate": 7.165131612058986e-05, "loss": 1.0221, "step": 113695 }, { "epoch": 0.29, "learning_rate": 7.165005787910816e-05, "loss": 1.0213, "step": 113700 }, { "epoch": 0.29, "learning_rate": 7.164879963762645e-05, "loss": 1.0205, "step": 113705 }, { "epoch": 0.29, "learning_rate": 7.164754139614474e-05, "loss": 1.0216, "step": 113710 }, { "epoch": 0.29, "learning_rate": 7.164628315466304e-05, "loss": 1.0236, "step": 113715 }, { "epoch": 0.29, "learning_rate": 7.164502491318135e-05, "loss": 1.0242, "step": 113720 }, { "epoch": 0.29, "learning_rate": 7.164376667169964e-05, "loss": 1.0218, "step": 113725 }, { "epoch": 0.29, "learning_rate": 7.164250843021793e-05, "loss": 1.0237, "step": 113730 }, { "epoch": 0.29, "learning_rate": 7.164125018873622e-05, "loss": 1.0221, "step": 113735 }, { "epoch": 0.29, "learning_rate": 7.163999194725453e-05, "loss": 1.0201, "step": 113740 }, { "epoch": 0.29, "learning_rate": 7.163873370577282e-05, "loss": 1.0247, "step": 113745 }, { "epoch": 0.29, "learning_rate": 7.163747546429111e-05, "loss": 1.0231, "step": 113750 }, { "epoch": 0.29, "learning_rate": 7.16362172228094e-05, "loss": 1.043, "step": 113755 }, { "epoch": 0.29, "learning_rate": 7.163495898132771e-05, "loss": 1.0214, "step": 113760 }, { "epoch": 0.29, "learning_rate": 7.1633700739846e-05, "loss": 1.0236, "step": 113765 }, { "epoch": 0.29, "learning_rate": 7.163244249836429e-05, "loss": 1.021, "step": 113770 }, { "epoch": 0.29, "learning_rate": 7.163118425688258e-05, "loss": 1.023, "step": 113775 }, { "epoch": 0.29, "learning_rate": 7.162992601540089e-05, "loss": 1.0228, "step": 113780 }, { "epoch": 0.29, "learning_rate": 7.162866777391918e-05, "loss": 1.0224, "step": 113785 }, { "epoch": 0.29, "learning_rate": 7.162740953243747e-05, "loss": 1.0205, "step": 113790 }, { "epoch": 0.29, "learning_rate": 7.162615129095576e-05, "loss": 1.0224, "step": 113795 }, { "epoch": 0.29, "learning_rate": 7.162489304947406e-05, "loss": 1.0227, "step": 113800 }, { "epoch": 0.29, "learning_rate": 7.162363480799236e-05, "loss": 1.0216, "step": 113805 }, { "epoch": 0.29, "learning_rate": 7.162237656651065e-05, "loss": 1.0212, "step": 113810 }, { "epoch": 0.29, "learning_rate": 7.162111832502894e-05, "loss": 1.0222, "step": 113815 }, { "epoch": 0.29, "learning_rate": 7.161986008354724e-05, "loss": 1.0442, "step": 113820 }, { "epoch": 0.29, "learning_rate": 7.161860184206554e-05, "loss": 1.0227, "step": 113825 }, { "epoch": 0.29, "learning_rate": 7.161734360058383e-05, "loss": 1.0204, "step": 113830 }, { "epoch": 0.29, "learning_rate": 7.161608535910212e-05, "loss": 1.0226, "step": 113835 }, { "epoch": 0.29, "learning_rate": 7.161482711762042e-05, "loss": 1.0227, "step": 113840 }, { "epoch": 0.29, "learning_rate": 7.161356887613872e-05, "loss": 1.0214, "step": 113845 }, { "epoch": 0.29, "learning_rate": 7.161231063465701e-05, "loss": 1.0209, "step": 113850 }, { "epoch": 0.29, "learning_rate": 7.16110523931753e-05, "loss": 1.0229, "step": 113855 }, { "epoch": 0.29, "learning_rate": 7.16097941516936e-05, "loss": 1.022, "step": 113860 }, { "epoch": 0.29, "learning_rate": 7.160853591021189e-05, "loss": 1.0193, "step": 113865 }, { "epoch": 0.29, "learning_rate": 7.160727766873019e-05, "loss": 1.0264, "step": 113870 }, { "epoch": 0.29, "learning_rate": 7.160601942724848e-05, "loss": 1.022, "step": 113875 }, { "epoch": 0.29, "learning_rate": 7.160476118576678e-05, "loss": 1.0214, "step": 113880 }, { "epoch": 0.29, "learning_rate": 7.160350294428507e-05, "loss": 1.0225, "step": 113885 }, { "epoch": 0.29, "learning_rate": 7.160224470280337e-05, "loss": 1.0183, "step": 113890 }, { "epoch": 0.29, "learning_rate": 7.160098646132166e-05, "loss": 1.0219, "step": 113895 }, { "epoch": 0.29, "learning_rate": 7.159972821983996e-05, "loss": 1.022, "step": 113900 }, { "epoch": 0.29, "learning_rate": 7.159846997835825e-05, "loss": 1.0223, "step": 113905 }, { "epoch": 0.29, "learning_rate": 7.159721173687655e-05, "loss": 1.0212, "step": 113910 }, { "epoch": 0.29, "learning_rate": 7.159595349539484e-05, "loss": 1.0224, "step": 113915 }, { "epoch": 0.29, "learning_rate": 7.159469525391314e-05, "loss": 1.0246, "step": 113920 }, { "epoch": 0.29, "learning_rate": 7.159343701243143e-05, "loss": 1.0219, "step": 113925 }, { "epoch": 0.29, "learning_rate": 7.159217877094972e-05, "loss": 1.0223, "step": 113930 }, { "epoch": 0.29, "learning_rate": 7.159092052946802e-05, "loss": 1.0219, "step": 113935 }, { "epoch": 0.29, "learning_rate": 7.158966228798632e-05, "loss": 1.0256, "step": 113940 }, { "epoch": 0.29, "learning_rate": 7.158840404650461e-05, "loss": 1.0224, "step": 113945 }, { "epoch": 0.29, "learning_rate": 7.15871458050229e-05, "loss": 1.023, "step": 113950 }, { "epoch": 0.29, "learning_rate": 7.15858875635412e-05, "loss": 1.0222, "step": 113955 }, { "epoch": 0.29, "learning_rate": 7.15846293220595e-05, "loss": 1.0315, "step": 113960 }, { "epoch": 0.29, "learning_rate": 7.158337108057779e-05, "loss": 1.0231, "step": 113965 }, { "epoch": 0.29, "learning_rate": 7.158211283909608e-05, "loss": 1.0214, "step": 113970 }, { "epoch": 0.29, "learning_rate": 7.158085459761438e-05, "loss": 1.024, "step": 113975 }, { "epoch": 0.29, "learning_rate": 7.157959635613268e-05, "loss": 1.0207, "step": 113980 }, { "epoch": 0.29, "learning_rate": 7.157833811465097e-05, "loss": 1.0225, "step": 113985 }, { "epoch": 0.29, "learning_rate": 7.157707987316926e-05, "loss": 1.0223, "step": 113990 }, { "epoch": 0.29, "learning_rate": 7.157582163168755e-05, "loss": 1.0227, "step": 113995 }, { "epoch": 0.29, "learning_rate": 7.157456339020586e-05, "loss": 1.0222, "step": 114000 }, { "epoch": 0.29, "learning_rate": 7.157330514872415e-05, "loss": 1.0221, "step": 114005 }, { "epoch": 0.29, "learning_rate": 7.157204690724244e-05, "loss": 1.0237, "step": 114010 }, { "epoch": 0.29, "learning_rate": 7.157078866576073e-05, "loss": 1.0225, "step": 114015 }, { "epoch": 0.29, "learning_rate": 7.156953042427904e-05, "loss": 1.0196, "step": 114020 }, { "epoch": 0.29, "learning_rate": 7.156827218279733e-05, "loss": 1.0216, "step": 114025 }, { "epoch": 0.29, "learning_rate": 7.156701394131562e-05, "loss": 1.0217, "step": 114030 }, { "epoch": 0.29, "learning_rate": 7.156575569983391e-05, "loss": 1.0215, "step": 114035 }, { "epoch": 0.29, "learning_rate": 7.156449745835222e-05, "loss": 1.023, "step": 114040 }, { "epoch": 0.29, "learning_rate": 7.156323921687051e-05, "loss": 1.0205, "step": 114045 }, { "epoch": 0.29, "learning_rate": 7.15619809753888e-05, "loss": 1.0206, "step": 114050 }, { "epoch": 0.29, "learning_rate": 7.156072273390709e-05, "loss": 1.0237, "step": 114055 }, { "epoch": 0.29, "learning_rate": 7.155946449242538e-05, "loss": 1.0237, "step": 114060 }, { "epoch": 0.29, "learning_rate": 7.155820625094369e-05, "loss": 1.0215, "step": 114065 }, { "epoch": 0.29, "learning_rate": 7.155694800946198e-05, "loss": 1.0232, "step": 114070 }, { "epoch": 0.29, "learning_rate": 7.155568976798027e-05, "loss": 1.0222, "step": 114075 }, { "epoch": 0.29, "learning_rate": 7.155443152649856e-05, "loss": 1.0223, "step": 114080 }, { "epoch": 0.29, "learning_rate": 7.155317328501687e-05, "loss": 1.022, "step": 114085 }, { "epoch": 0.29, "learning_rate": 7.155191504353516e-05, "loss": 1.0229, "step": 114090 }, { "epoch": 0.29, "learning_rate": 7.155065680205345e-05, "loss": 1.0219, "step": 114095 }, { "epoch": 0.29, "learning_rate": 7.154939856057174e-05, "loss": 1.0233, "step": 114100 }, { "epoch": 0.29, "learning_rate": 7.154814031909003e-05, "loss": 1.0243, "step": 114105 }, { "epoch": 0.29, "learning_rate": 7.154688207760834e-05, "loss": 1.0225, "step": 114110 }, { "epoch": 0.29, "learning_rate": 7.154562383612663e-05, "loss": 1.0216, "step": 114115 }, { "epoch": 0.29, "learning_rate": 7.154436559464492e-05, "loss": 1.0429, "step": 114120 }, { "epoch": 0.29, "learning_rate": 7.154310735316321e-05, "loss": 1.0233, "step": 114125 }, { "epoch": 0.29, "learning_rate": 7.154184911168152e-05, "loss": 1.0201, "step": 114130 }, { "epoch": 0.29, "learning_rate": 7.154059087019981e-05, "loss": 1.0196, "step": 114135 }, { "epoch": 0.29, "learning_rate": 7.15393326287181e-05, "loss": 1.0217, "step": 114140 }, { "epoch": 0.29, "learning_rate": 7.15380743872364e-05, "loss": 1.0195, "step": 114145 }, { "epoch": 0.29, "learning_rate": 7.15368161457547e-05, "loss": 1.0208, "step": 114150 }, { "epoch": 0.29, "learning_rate": 7.153555790427299e-05, "loss": 1.0237, "step": 114155 }, { "epoch": 0.29, "learning_rate": 7.153429966279128e-05, "loss": 1.0209, "step": 114160 }, { "epoch": 0.29, "learning_rate": 7.153304142130957e-05, "loss": 1.0218, "step": 114165 }, { "epoch": 0.29, "learning_rate": 7.153178317982786e-05, "loss": 1.0221, "step": 114170 }, { "epoch": 0.29, "learning_rate": 7.153052493834617e-05, "loss": 1.0213, "step": 114175 }, { "epoch": 0.29, "learning_rate": 7.152926669686446e-05, "loss": 1.0194, "step": 114180 }, { "epoch": 0.29, "learning_rate": 7.152800845538275e-05, "loss": 1.022, "step": 114185 }, { "epoch": 0.29, "learning_rate": 7.152675021390104e-05, "loss": 1.0229, "step": 114190 }, { "epoch": 0.29, "learning_rate": 7.152549197241935e-05, "loss": 1.0227, "step": 114195 }, { "epoch": 0.29, "learning_rate": 7.152423373093764e-05, "loss": 1.0221, "step": 114200 }, { "epoch": 0.29, "learning_rate": 7.152297548945593e-05, "loss": 1.0209, "step": 114205 }, { "epoch": 0.29, "learning_rate": 7.152171724797422e-05, "loss": 1.0217, "step": 114210 }, { "epoch": 0.29, "learning_rate": 7.152045900649253e-05, "loss": 1.0245, "step": 114215 }, { "epoch": 0.29, "learning_rate": 7.151920076501083e-05, "loss": 1.0214, "step": 114220 }, { "epoch": 0.29, "learning_rate": 7.151794252352913e-05, "loss": 1.0216, "step": 114225 }, { "epoch": 0.29, "learning_rate": 7.151668428204742e-05, "loss": 1.022, "step": 114230 }, { "epoch": 0.29, "learning_rate": 7.151542604056571e-05, "loss": 1.0227, "step": 114235 }, { "epoch": 0.29, "learning_rate": 7.151416779908401e-05, "loss": 1.0194, "step": 114240 }, { "epoch": 0.29, "learning_rate": 7.15129095576023e-05, "loss": 1.0219, "step": 114245 }, { "epoch": 0.29, "learning_rate": 7.15116513161206e-05, "loss": 1.021, "step": 114250 }, { "epoch": 0.29, "learning_rate": 7.151039307463889e-05, "loss": 1.0208, "step": 114255 }, { "epoch": 0.29, "learning_rate": 7.150913483315718e-05, "loss": 1.0234, "step": 114260 }, { "epoch": 0.29, "learning_rate": 7.150787659167549e-05, "loss": 1.022, "step": 114265 }, { "epoch": 0.29, "learning_rate": 7.150661835019378e-05, "loss": 1.0215, "step": 114270 }, { "epoch": 0.29, "learning_rate": 7.150536010871207e-05, "loss": 1.0234, "step": 114275 }, { "epoch": 0.29, "learning_rate": 7.150410186723036e-05, "loss": 1.0253, "step": 114280 }, { "epoch": 0.29, "learning_rate": 7.150284362574867e-05, "loss": 1.0235, "step": 114285 }, { "epoch": 0.29, "learning_rate": 7.150158538426696e-05, "loss": 1.0221, "step": 114290 }, { "epoch": 0.29, "learning_rate": 7.150032714278525e-05, "loss": 1.0226, "step": 114295 }, { "epoch": 0.29, "learning_rate": 7.149906890130354e-05, "loss": 1.0236, "step": 114300 }, { "epoch": 0.29, "learning_rate": 7.149781065982185e-05, "loss": 1.023, "step": 114305 }, { "epoch": 0.29, "learning_rate": 7.149655241834014e-05, "loss": 1.023, "step": 114310 }, { "epoch": 0.29, "learning_rate": 7.149529417685843e-05, "loss": 1.0211, "step": 114315 }, { "epoch": 0.29, "learning_rate": 7.149403593537672e-05, "loss": 1.0215, "step": 114320 }, { "epoch": 0.29, "learning_rate": 7.149277769389501e-05, "loss": 1.0188, "step": 114325 }, { "epoch": 0.29, "learning_rate": 7.149151945241332e-05, "loss": 1.0206, "step": 114330 }, { "epoch": 0.29, "learning_rate": 7.149026121093161e-05, "loss": 1.0203, "step": 114335 }, { "epoch": 0.29, "learning_rate": 7.14890029694499e-05, "loss": 1.0239, "step": 114340 }, { "epoch": 0.29, "learning_rate": 7.148774472796819e-05, "loss": 1.0249, "step": 114345 }, { "epoch": 0.29, "learning_rate": 7.14864864864865e-05, "loss": 1.0235, "step": 114350 }, { "epoch": 0.29, "learning_rate": 7.148522824500479e-05, "loss": 1.0213, "step": 114355 }, { "epoch": 0.29, "learning_rate": 7.148397000352308e-05, "loss": 1.0248, "step": 114360 }, { "epoch": 0.29, "learning_rate": 7.148271176204137e-05, "loss": 1.0202, "step": 114365 }, { "epoch": 0.29, "learning_rate": 7.148145352055968e-05, "loss": 1.0236, "step": 114370 }, { "epoch": 0.29, "learning_rate": 7.148019527907797e-05, "loss": 1.024, "step": 114375 }, { "epoch": 0.29, "learning_rate": 7.147893703759626e-05, "loss": 1.0232, "step": 114380 }, { "epoch": 0.29, "learning_rate": 7.147767879611455e-05, "loss": 1.0218, "step": 114385 }, { "epoch": 0.29, "learning_rate": 7.147642055463284e-05, "loss": 1.0214, "step": 114390 }, { "epoch": 0.29, "learning_rate": 7.147516231315115e-05, "loss": 1.0229, "step": 114395 }, { "epoch": 0.29, "learning_rate": 7.147390407166944e-05, "loss": 1.0231, "step": 114400 }, { "epoch": 0.29, "learning_rate": 7.147264583018773e-05, "loss": 1.0195, "step": 114405 }, { "epoch": 0.29, "learning_rate": 7.147138758870602e-05, "loss": 1.0192, "step": 114410 }, { "epoch": 0.29, "learning_rate": 7.147012934722433e-05, "loss": 1.021, "step": 114415 }, { "epoch": 0.29, "learning_rate": 7.146887110574262e-05, "loss": 1.0227, "step": 114420 }, { "epoch": 0.29, "learning_rate": 7.146761286426091e-05, "loss": 1.0202, "step": 114425 }, { "epoch": 0.29, "learning_rate": 7.14663546227792e-05, "loss": 1.0213, "step": 114430 }, { "epoch": 0.29, "learning_rate": 7.146509638129751e-05, "loss": 1.0243, "step": 114435 }, { "epoch": 0.29, "learning_rate": 7.14638381398158e-05, "loss": 1.0219, "step": 114440 }, { "epoch": 0.29, "learning_rate": 7.146257989833409e-05, "loss": 1.0229, "step": 114445 }, { "epoch": 0.29, "learning_rate": 7.146132165685238e-05, "loss": 1.0191, "step": 114450 }, { "epoch": 0.29, "learning_rate": 7.146006341537068e-05, "loss": 1.0245, "step": 114455 }, { "epoch": 0.29, "learning_rate": 7.145880517388898e-05, "loss": 1.0224, "step": 114460 }, { "epoch": 0.29, "learning_rate": 7.145754693240727e-05, "loss": 1.0221, "step": 114465 }, { "epoch": 0.29, "learning_rate": 7.145628869092556e-05, "loss": 1.0209, "step": 114470 }, { "epoch": 0.29, "learning_rate": 7.145503044944386e-05, "loss": 1.0226, "step": 114475 }, { "epoch": 0.29, "learning_rate": 7.145377220796216e-05, "loss": 1.0224, "step": 114480 }, { "epoch": 0.29, "learning_rate": 7.145251396648045e-05, "loss": 1.0232, "step": 114485 }, { "epoch": 0.29, "learning_rate": 7.145125572499874e-05, "loss": 1.0223, "step": 114490 }, { "epoch": 0.29, "learning_rate": 7.144999748351704e-05, "loss": 1.0221, "step": 114495 }, { "epoch": 0.29, "learning_rate": 7.144873924203534e-05, "loss": 1.0222, "step": 114500 }, { "epoch": 0.29, "learning_rate": 7.144748100055363e-05, "loss": 1.0218, "step": 114505 }, { "epoch": 0.29, "learning_rate": 7.144622275907192e-05, "loss": 1.0237, "step": 114510 }, { "epoch": 0.29, "learning_rate": 7.144496451759022e-05, "loss": 1.0232, "step": 114515 }, { "epoch": 0.29, "learning_rate": 7.144370627610851e-05, "loss": 1.0213, "step": 114520 }, { "epoch": 0.29, "learning_rate": 7.144244803462681e-05, "loss": 1.0239, "step": 114525 }, { "epoch": 0.29, "learning_rate": 7.14411897931451e-05, "loss": 1.0236, "step": 114530 }, { "epoch": 0.29, "learning_rate": 7.14399315516634e-05, "loss": 1.0216, "step": 114535 }, { "epoch": 0.29, "learning_rate": 7.143867331018169e-05, "loss": 1.0214, "step": 114540 }, { "epoch": 0.29, "learning_rate": 7.143741506869999e-05, "loss": 1.0209, "step": 114545 }, { "epoch": 0.29, "learning_rate": 7.143615682721828e-05, "loss": 1.0232, "step": 114550 }, { "epoch": 0.29, "learning_rate": 7.143489858573658e-05, "loss": 1.0229, "step": 114555 }, { "epoch": 0.29, "learning_rate": 7.143364034425487e-05, "loss": 1.0205, "step": 114560 }, { "epoch": 0.29, "learning_rate": 7.143238210277317e-05, "loss": 1.0193, "step": 114565 }, { "epoch": 0.29, "learning_rate": 7.143112386129146e-05, "loss": 1.0215, "step": 114570 }, { "epoch": 0.29, "learning_rate": 7.142986561980976e-05, "loss": 1.0235, "step": 114575 }, { "epoch": 0.29, "learning_rate": 7.142860737832805e-05, "loss": 1.0223, "step": 114580 }, { "epoch": 0.29, "learning_rate": 7.142734913684634e-05, "loss": 1.0201, "step": 114585 }, { "epoch": 0.29, "learning_rate": 7.142609089536464e-05, "loss": 1.0211, "step": 114590 }, { "epoch": 0.29, "learning_rate": 7.142483265388294e-05, "loss": 1.0232, "step": 114595 }, { "epoch": 0.29, "learning_rate": 7.142357441240123e-05, "loss": 1.0219, "step": 114600 }, { "epoch": 0.29, "learning_rate": 7.142231617091952e-05, "loss": 1.0207, "step": 114605 }, { "epoch": 0.29, "learning_rate": 7.142105792943782e-05, "loss": 1.0223, "step": 114610 }, { "epoch": 0.29, "learning_rate": 7.141979968795612e-05, "loss": 1.0207, "step": 114615 }, { "epoch": 0.29, "learning_rate": 7.14185414464744e-05, "loss": 1.0243, "step": 114620 }, { "epoch": 0.29, "learning_rate": 7.14172832049927e-05, "loss": 1.0219, "step": 114625 }, { "epoch": 0.29, "learning_rate": 7.1416024963511e-05, "loss": 1.0219, "step": 114630 }, { "epoch": 0.29, "learning_rate": 7.14147667220293e-05, "loss": 1.0241, "step": 114635 }, { "epoch": 0.29, "learning_rate": 7.141350848054759e-05, "loss": 1.0212, "step": 114640 }, { "epoch": 0.29, "learning_rate": 7.141225023906588e-05, "loss": 1.0224, "step": 114645 }, { "epoch": 0.29, "learning_rate": 7.141099199758417e-05, "loss": 1.022, "step": 114650 }, { "epoch": 0.29, "learning_rate": 7.140973375610247e-05, "loss": 1.0222, "step": 114655 }, { "epoch": 0.29, "learning_rate": 7.140847551462077e-05, "loss": 1.0236, "step": 114660 }, { "epoch": 0.29, "learning_rate": 7.140721727313906e-05, "loss": 1.0218, "step": 114665 }, { "epoch": 0.29, "learning_rate": 7.140595903165735e-05, "loss": 1.0206, "step": 114670 }, { "epoch": 0.29, "learning_rate": 7.140470079017565e-05, "loss": 1.0228, "step": 114675 }, { "epoch": 0.29, "learning_rate": 7.140344254869395e-05, "loss": 1.0227, "step": 114680 }, { "epoch": 0.29, "learning_rate": 7.140218430721224e-05, "loss": 1.0211, "step": 114685 }, { "epoch": 0.29, "learning_rate": 7.140092606573053e-05, "loss": 1.0219, "step": 114690 }, { "epoch": 0.29, "learning_rate": 7.139966782424883e-05, "loss": 1.0216, "step": 114695 }, { "epoch": 0.29, "learning_rate": 7.139840958276713e-05, "loss": 1.024, "step": 114700 }, { "epoch": 0.29, "learning_rate": 7.139715134128542e-05, "loss": 1.0235, "step": 114705 }, { "epoch": 0.29, "learning_rate": 7.139589309980371e-05, "loss": 1.0217, "step": 114710 }, { "epoch": 0.29, "learning_rate": 7.1394634858322e-05, "loss": 1.022, "step": 114715 }, { "epoch": 0.29, "learning_rate": 7.13933766168403e-05, "loss": 1.0223, "step": 114720 }, { "epoch": 0.29, "learning_rate": 7.139211837535861e-05, "loss": 1.0238, "step": 114725 }, { "epoch": 0.29, "learning_rate": 7.13908601338769e-05, "loss": 1.0242, "step": 114730 }, { "epoch": 0.29, "learning_rate": 7.13896018923952e-05, "loss": 1.0215, "step": 114735 }, { "epoch": 0.29, "learning_rate": 7.138834365091349e-05, "loss": 1.0242, "step": 114740 }, { "epoch": 0.29, "learning_rate": 7.138708540943179e-05, "loss": 1.0193, "step": 114745 }, { "epoch": 0.29, "learning_rate": 7.138582716795008e-05, "loss": 1.0202, "step": 114750 }, { "epoch": 0.29, "learning_rate": 7.138456892646837e-05, "loss": 1.0218, "step": 114755 }, { "epoch": 0.29, "learning_rate": 7.138331068498667e-05, "loss": 1.0213, "step": 114760 }, { "epoch": 0.29, "learning_rate": 7.138205244350497e-05, "loss": 1.0233, "step": 114765 }, { "epoch": 0.29, "learning_rate": 7.138079420202326e-05, "loss": 1.0429, "step": 114770 }, { "epoch": 0.29, "learning_rate": 7.137953596054155e-05, "loss": 1.041, "step": 114775 }, { "epoch": 0.29, "learning_rate": 7.137827771905985e-05, "loss": 1.022, "step": 114780 }, { "epoch": 0.29, "learning_rate": 7.137701947757814e-05, "loss": 1.0203, "step": 114785 }, { "epoch": 0.29, "learning_rate": 7.137576123609644e-05, "loss": 1.022, "step": 114790 }, { "epoch": 0.29, "learning_rate": 7.137450299461473e-05, "loss": 1.0237, "step": 114795 }, { "epoch": 0.29, "learning_rate": 7.137324475313303e-05, "loss": 1.0251, "step": 114800 }, { "epoch": 0.29, "learning_rate": 7.137198651165132e-05, "loss": 1.0219, "step": 114805 }, { "epoch": 0.29, "learning_rate": 7.137072827016962e-05, "loss": 1.0243, "step": 114810 }, { "epoch": 0.29, "learning_rate": 7.136947002868791e-05, "loss": 1.0226, "step": 114815 }, { "epoch": 0.29, "learning_rate": 7.13682117872062e-05, "loss": 1.0219, "step": 114820 }, { "epoch": 0.29, "learning_rate": 7.13669535457245e-05, "loss": 1.0235, "step": 114825 }, { "epoch": 0.29, "learning_rate": 7.13656953042428e-05, "loss": 1.0209, "step": 114830 }, { "epoch": 0.29, "learning_rate": 7.13644370627611e-05, "loss": 1.0218, "step": 114835 }, { "epoch": 0.29, "learning_rate": 7.136317882127939e-05, "loss": 1.0216, "step": 114840 }, { "epoch": 0.29, "learning_rate": 7.136192057979768e-05, "loss": 1.0209, "step": 114845 }, { "epoch": 0.29, "learning_rate": 7.136066233831597e-05, "loss": 1.024, "step": 114850 }, { "epoch": 0.29, "learning_rate": 7.135940409683427e-05, "loss": 1.0216, "step": 114855 }, { "epoch": 0.29, "learning_rate": 7.135814585535257e-05, "loss": 1.0244, "step": 114860 }, { "epoch": 0.29, "learning_rate": 7.135688761387086e-05, "loss": 1.0209, "step": 114865 }, { "epoch": 0.29, "learning_rate": 7.135562937238915e-05, "loss": 1.0219, "step": 114870 }, { "epoch": 0.29, "learning_rate": 7.135437113090745e-05, "loss": 1.023, "step": 114875 }, { "epoch": 0.29, "learning_rate": 7.135311288942575e-05, "loss": 1.0238, "step": 114880 }, { "epoch": 0.29, "learning_rate": 7.135185464794404e-05, "loss": 1.0245, "step": 114885 }, { "epoch": 0.29, "learning_rate": 7.135059640646233e-05, "loss": 1.0206, "step": 114890 }, { "epoch": 0.29, "learning_rate": 7.134933816498063e-05, "loss": 1.0235, "step": 114895 }, { "epoch": 0.29, "learning_rate": 7.134807992349893e-05, "loss": 1.0235, "step": 114900 }, { "epoch": 0.29, "learning_rate": 7.134682168201722e-05, "loss": 1.023, "step": 114905 }, { "epoch": 0.29, "learning_rate": 7.134556344053551e-05, "loss": 1.024, "step": 114910 }, { "epoch": 0.29, "learning_rate": 7.13443051990538e-05, "loss": 1.0227, "step": 114915 }, { "epoch": 0.29, "learning_rate": 7.13430469575721e-05, "loss": 1.0213, "step": 114920 }, { "epoch": 0.29, "learning_rate": 7.13417887160904e-05, "loss": 1.0204, "step": 114925 }, { "epoch": 0.29, "learning_rate": 7.134053047460869e-05, "loss": 1.0216, "step": 114930 }, { "epoch": 0.29, "learning_rate": 7.133927223312698e-05, "loss": 1.023, "step": 114935 }, { "epoch": 0.29, "learning_rate": 7.133801399164529e-05, "loss": 1.0227, "step": 114940 }, { "epoch": 0.29, "learning_rate": 7.133675575016358e-05, "loss": 1.0217, "step": 114945 }, { "epoch": 0.29, "learning_rate": 7.133549750868187e-05, "loss": 1.0216, "step": 114950 }, { "epoch": 0.29, "learning_rate": 7.133423926720016e-05, "loss": 1.0221, "step": 114955 }, { "epoch": 0.29, "learning_rate": 7.133298102571847e-05, "loss": 1.0244, "step": 114960 }, { "epoch": 0.29, "learning_rate": 7.133172278423676e-05, "loss": 1.0207, "step": 114965 }, { "epoch": 0.29, "learning_rate": 7.133046454275505e-05, "loss": 1.0219, "step": 114970 }, { "epoch": 0.29, "learning_rate": 7.132920630127334e-05, "loss": 1.0224, "step": 114975 }, { "epoch": 0.29, "learning_rate": 7.132794805979163e-05, "loss": 1.0226, "step": 114980 }, { "epoch": 0.29, "learning_rate": 7.132668981830994e-05, "loss": 1.023, "step": 114985 }, { "epoch": 0.29, "learning_rate": 7.132543157682823e-05, "loss": 1.0229, "step": 114990 }, { "epoch": 0.29, "learning_rate": 7.132417333534652e-05, "loss": 1.0249, "step": 114995 }, { "epoch": 0.29, "learning_rate": 7.132291509386481e-05, "loss": 1.0251, "step": 115000 }, { "epoch": 0.29, "learning_rate": 7.132165685238312e-05, "loss": 1.0237, "step": 115005 }, { "epoch": 0.29, "learning_rate": 7.132039861090141e-05, "loss": 1.0209, "step": 115010 }, { "epoch": 0.29, "learning_rate": 7.13191403694197e-05, "loss": 1.0207, "step": 115015 }, { "epoch": 0.29, "learning_rate": 7.131788212793799e-05, "loss": 1.0221, "step": 115020 }, { "epoch": 0.29, "learning_rate": 7.13166238864563e-05, "loss": 1.0197, "step": 115025 }, { "epoch": 0.29, "learning_rate": 7.131536564497459e-05, "loss": 1.0203, "step": 115030 }, { "epoch": 0.29, "learning_rate": 7.131410740349288e-05, "loss": 1.0256, "step": 115035 }, { "epoch": 0.29, "learning_rate": 7.131284916201117e-05, "loss": 1.0201, "step": 115040 }, { "epoch": 0.29, "learning_rate": 7.131159092052946e-05, "loss": 1.0203, "step": 115045 }, { "epoch": 0.29, "learning_rate": 7.131033267904777e-05, "loss": 1.0206, "step": 115050 }, { "epoch": 0.29, "learning_rate": 7.130907443756606e-05, "loss": 1.0213, "step": 115055 }, { "epoch": 0.29, "learning_rate": 7.130781619608435e-05, "loss": 1.0217, "step": 115060 }, { "epoch": 0.29, "learning_rate": 7.130655795460264e-05, "loss": 1.0197, "step": 115065 }, { "epoch": 0.29, "learning_rate": 7.130529971312095e-05, "loss": 1.0192, "step": 115070 }, { "epoch": 0.29, "learning_rate": 7.130404147163924e-05, "loss": 1.0206, "step": 115075 }, { "epoch": 0.29, "learning_rate": 7.130278323015753e-05, "loss": 1.022, "step": 115080 }, { "epoch": 0.29, "learning_rate": 7.130152498867582e-05, "loss": 1.0207, "step": 115085 }, { "epoch": 0.29, "learning_rate": 7.130026674719413e-05, "loss": 1.0194, "step": 115090 }, { "epoch": 0.29, "learning_rate": 7.129900850571242e-05, "loss": 1.0244, "step": 115095 }, { "epoch": 0.29, "learning_rate": 7.129775026423071e-05, "loss": 1.0232, "step": 115100 }, { "epoch": 0.29, "learning_rate": 7.1296492022749e-05, "loss": 1.0241, "step": 115105 }, { "epoch": 0.29, "learning_rate": 7.12952337812673e-05, "loss": 1.0244, "step": 115110 }, { "epoch": 0.29, "learning_rate": 7.12939755397856e-05, "loss": 1.024, "step": 115115 }, { "epoch": 0.29, "learning_rate": 7.129271729830389e-05, "loss": 1.0206, "step": 115120 }, { "epoch": 0.29, "learning_rate": 7.129145905682218e-05, "loss": 1.0248, "step": 115125 }, { "epoch": 0.29, "learning_rate": 7.129020081534047e-05, "loss": 1.0221, "step": 115130 }, { "epoch": 0.29, "learning_rate": 7.128894257385878e-05, "loss": 1.0225, "step": 115135 }, { "epoch": 0.29, "learning_rate": 7.128768433237707e-05, "loss": 1.0224, "step": 115140 }, { "epoch": 0.29, "learning_rate": 7.128642609089536e-05, "loss": 1.0249, "step": 115145 }, { "epoch": 0.29, "learning_rate": 7.128516784941365e-05, "loss": 1.0196, "step": 115150 }, { "epoch": 0.29, "learning_rate": 7.128390960793196e-05, "loss": 1.0214, "step": 115155 }, { "epoch": 0.29, "learning_rate": 7.128265136645025e-05, "loss": 1.023, "step": 115160 }, { "epoch": 0.29, "learning_rate": 7.128139312496854e-05, "loss": 1.0238, "step": 115165 }, { "epoch": 0.29, "learning_rate": 7.128013488348683e-05, "loss": 1.0217, "step": 115170 }, { "epoch": 0.29, "learning_rate": 7.127887664200513e-05, "loss": 1.0282, "step": 115175 }, { "epoch": 0.29, "learning_rate": 7.127761840052343e-05, "loss": 1.022, "step": 115180 }, { "epoch": 0.29, "learning_rate": 7.127636015904172e-05, "loss": 1.0221, "step": 115185 }, { "epoch": 0.29, "learning_rate": 7.127510191756001e-05, "loss": 1.0213, "step": 115190 }, { "epoch": 0.29, "learning_rate": 7.12738436760783e-05, "loss": 1.0205, "step": 115195 }, { "epoch": 0.29, "learning_rate": 7.127258543459661e-05, "loss": 1.0219, "step": 115200 }, { "epoch": 0.29, "learning_rate": 7.12713271931149e-05, "loss": 1.023, "step": 115205 }, { "epoch": 0.29, "learning_rate": 7.12700689516332e-05, "loss": 1.0218, "step": 115210 }, { "epoch": 0.29, "learning_rate": 7.12688107101515e-05, "loss": 1.0217, "step": 115215 }, { "epoch": 0.29, "learning_rate": 7.126755246866979e-05, "loss": 1.0205, "step": 115220 }, { "epoch": 0.29, "learning_rate": 7.12662942271881e-05, "loss": 1.0229, "step": 115225 }, { "epoch": 0.29, "learning_rate": 7.126503598570639e-05, "loss": 1.0225, "step": 115230 }, { "epoch": 0.29, "learning_rate": 7.126377774422468e-05, "loss": 1.0409, "step": 115235 }, { "epoch": 0.29, "learning_rate": 7.126251950274297e-05, "loss": 1.0215, "step": 115240 }, { "epoch": 0.29, "learning_rate": 7.126126126126126e-05, "loss": 1.02, "step": 115245 }, { "epoch": 0.29, "learning_rate": 7.126000301977957e-05, "loss": 1.0452, "step": 115250 }, { "epoch": 0.29, "learning_rate": 7.125874477829786e-05, "loss": 1.0203, "step": 115255 }, { "epoch": 0.29, "learning_rate": 7.125748653681615e-05, "loss": 1.024, "step": 115260 }, { "epoch": 0.29, "learning_rate": 7.125622829533444e-05, "loss": 1.0233, "step": 115265 }, { "epoch": 0.29, "learning_rate": 7.125497005385275e-05, "loss": 1.0221, "step": 115270 }, { "epoch": 0.29, "learning_rate": 7.125371181237104e-05, "loss": 1.0234, "step": 115275 }, { "epoch": 0.29, "learning_rate": 7.125245357088933e-05, "loss": 1.0234, "step": 115280 }, { "epoch": 0.29, "learning_rate": 7.125119532940762e-05, "loss": 1.0233, "step": 115285 }, { "epoch": 0.29, "learning_rate": 7.124993708792593e-05, "loss": 1.02, "step": 115290 }, { "epoch": 0.29, "learning_rate": 7.124867884644422e-05, "loss": 1.0227, "step": 115295 }, { "epoch": 0.29, "learning_rate": 7.124742060496251e-05, "loss": 1.0211, "step": 115300 }, { "epoch": 0.29, "learning_rate": 7.12461623634808e-05, "loss": 1.0184, "step": 115305 }, { "epoch": 0.29, "learning_rate": 7.12449041219991e-05, "loss": 1.0221, "step": 115310 }, { "epoch": 0.29, "learning_rate": 7.12436458805174e-05, "loss": 1.0222, "step": 115315 }, { "epoch": 0.29, "learning_rate": 7.124238763903569e-05, "loss": 1.0224, "step": 115320 }, { "epoch": 0.29, "learning_rate": 7.124112939755398e-05, "loss": 1.0236, "step": 115325 }, { "epoch": 0.29, "learning_rate": 7.123987115607227e-05, "loss": 1.0271, "step": 115330 }, { "epoch": 0.29, "learning_rate": 7.123861291459058e-05, "loss": 1.021, "step": 115335 }, { "epoch": 0.29, "learning_rate": 7.123735467310887e-05, "loss": 1.0238, "step": 115340 }, { "epoch": 0.29, "learning_rate": 7.123609643162716e-05, "loss": 1.0216, "step": 115345 }, { "epoch": 0.29, "learning_rate": 7.123483819014545e-05, "loss": 1.0207, "step": 115350 }, { "epoch": 0.29, "learning_rate": 7.123357994866376e-05, "loss": 1.0449, "step": 115355 }, { "epoch": 0.29, "learning_rate": 7.123232170718205e-05, "loss": 1.0208, "step": 115360 }, { "epoch": 0.29, "learning_rate": 7.123106346570034e-05, "loss": 1.0194, "step": 115365 }, { "epoch": 0.29, "learning_rate": 7.122980522421863e-05, "loss": 1.0232, "step": 115370 }, { "epoch": 0.29, "learning_rate": 7.122854698273693e-05, "loss": 1.0237, "step": 115375 }, { "epoch": 0.29, "learning_rate": 7.122728874125523e-05, "loss": 1.0223, "step": 115380 }, { "epoch": 0.29, "learning_rate": 7.122603049977352e-05, "loss": 1.0232, "step": 115385 }, { "epoch": 0.29, "learning_rate": 7.122477225829181e-05, "loss": 1.0218, "step": 115390 }, { "epoch": 0.29, "learning_rate": 7.12235140168101e-05, "loss": 1.0209, "step": 115395 }, { "epoch": 0.29, "learning_rate": 7.122225577532841e-05, "loss": 1.0229, "step": 115400 }, { "epoch": 0.29, "learning_rate": 7.12209975338467e-05, "loss": 1.0237, "step": 115405 }, { "epoch": 0.29, "learning_rate": 7.1219739292365e-05, "loss": 1.0203, "step": 115410 }, { "epoch": 0.29, "learning_rate": 7.121848105088329e-05, "loss": 1.0226, "step": 115415 }, { "epoch": 0.29, "learning_rate": 7.121722280940159e-05, "loss": 1.0235, "step": 115420 }, { "epoch": 0.29, "learning_rate": 7.121596456791988e-05, "loss": 1.0216, "step": 115425 }, { "epoch": 0.29, "learning_rate": 7.121470632643817e-05, "loss": 1.0208, "step": 115430 }, { "epoch": 0.29, "learning_rate": 7.121344808495647e-05, "loss": 1.022, "step": 115435 }, { "epoch": 0.29, "learning_rate": 7.121218984347476e-05, "loss": 1.0231, "step": 115440 }, { "epoch": 0.29, "learning_rate": 7.121093160199306e-05, "loss": 1.0233, "step": 115445 }, { "epoch": 0.29, "learning_rate": 7.120967336051135e-05, "loss": 1.0229, "step": 115450 }, { "epoch": 0.29, "learning_rate": 7.120841511902965e-05, "loss": 1.0229, "step": 115455 }, { "epoch": 0.29, "learning_rate": 7.120715687754794e-05, "loss": 1.022, "step": 115460 }, { "epoch": 0.29, "learning_rate": 7.120589863606624e-05, "loss": 1.021, "step": 115465 }, { "epoch": 0.29, "learning_rate": 7.120464039458453e-05, "loss": 1.0214, "step": 115470 }, { "epoch": 0.29, "learning_rate": 7.120338215310283e-05, "loss": 1.0202, "step": 115475 }, { "epoch": 0.29, "learning_rate": 7.120212391162112e-05, "loss": 1.0222, "step": 115480 }, { "epoch": 0.29, "learning_rate": 7.120086567013942e-05, "loss": 1.0207, "step": 115485 }, { "epoch": 0.29, "learning_rate": 7.119960742865771e-05, "loss": 1.0215, "step": 115490 }, { "epoch": 0.29, "learning_rate": 7.1198349187176e-05, "loss": 1.021, "step": 115495 }, { "epoch": 0.29, "learning_rate": 7.11970909456943e-05, "loss": 1.0233, "step": 115500 }, { "epoch": 0.29, "learning_rate": 7.119583270421259e-05, "loss": 1.024, "step": 115505 }, { "epoch": 0.29, "learning_rate": 7.11945744627309e-05, "loss": 1.0243, "step": 115510 }, { "epoch": 0.29, "learning_rate": 7.119331622124919e-05, "loss": 1.0468, "step": 115515 }, { "epoch": 0.29, "learning_rate": 7.119205797976748e-05, "loss": 1.0235, "step": 115520 }, { "epoch": 0.29, "learning_rate": 7.119079973828577e-05, "loss": 1.0229, "step": 115525 }, { "epoch": 0.29, "learning_rate": 7.118954149680407e-05, "loss": 1.0227, "step": 115530 }, { "epoch": 0.29, "learning_rate": 7.118828325532237e-05, "loss": 1.021, "step": 115535 }, { "epoch": 0.29, "learning_rate": 7.118702501384066e-05, "loss": 1.0206, "step": 115540 }, { "epoch": 0.29, "learning_rate": 7.118576677235895e-05, "loss": 1.0233, "step": 115545 }, { "epoch": 0.29, "learning_rate": 7.118450853087725e-05, "loss": 1.0225, "step": 115550 }, { "epoch": 0.29, "learning_rate": 7.118325028939555e-05, "loss": 1.0246, "step": 115555 }, { "epoch": 0.29, "learning_rate": 7.118199204791384e-05, "loss": 1.0228, "step": 115560 }, { "epoch": 0.29, "learning_rate": 7.118073380643213e-05, "loss": 1.0222, "step": 115565 }, { "epoch": 0.29, "learning_rate": 7.117947556495042e-05, "loss": 1.0229, "step": 115570 }, { "epoch": 0.29, "learning_rate": 7.117821732346873e-05, "loss": 1.0233, "step": 115575 }, { "epoch": 0.29, "learning_rate": 7.117695908198702e-05, "loss": 1.025, "step": 115580 }, { "epoch": 0.29, "learning_rate": 7.117570084050531e-05, "loss": 1.0204, "step": 115585 }, { "epoch": 0.29, "learning_rate": 7.11744425990236e-05, "loss": 1.023, "step": 115590 }, { "epoch": 0.29, "learning_rate": 7.11731843575419e-05, "loss": 1.0205, "step": 115595 }, { "epoch": 0.29, "learning_rate": 7.11719261160602e-05, "loss": 1.0209, "step": 115600 }, { "epoch": 0.29, "learning_rate": 7.117066787457849e-05, "loss": 1.0226, "step": 115605 }, { "epoch": 0.29, "learning_rate": 7.116940963309678e-05, "loss": 1.0216, "step": 115610 }, { "epoch": 0.29, "learning_rate": 7.116815139161509e-05, "loss": 1.0207, "step": 115615 }, { "epoch": 0.29, "learning_rate": 7.116689315013338e-05, "loss": 1.0236, "step": 115620 }, { "epoch": 0.29, "learning_rate": 7.116563490865167e-05, "loss": 1.0215, "step": 115625 }, { "epoch": 0.29, "learning_rate": 7.116437666716996e-05, "loss": 1.02, "step": 115630 }, { "epoch": 0.29, "learning_rate": 7.116311842568825e-05, "loss": 1.0215, "step": 115635 }, { "epoch": 0.29, "learning_rate": 7.116186018420656e-05, "loss": 1.0221, "step": 115640 }, { "epoch": 0.29, "learning_rate": 7.116060194272485e-05, "loss": 1.022, "step": 115645 }, { "epoch": 0.29, "learning_rate": 7.115934370124314e-05, "loss": 1.0251, "step": 115650 }, { "epoch": 0.29, "learning_rate": 7.115808545976143e-05, "loss": 1.0222, "step": 115655 }, { "epoch": 0.29, "learning_rate": 7.115682721827974e-05, "loss": 1.0222, "step": 115660 }, { "epoch": 0.29, "learning_rate": 7.115556897679803e-05, "loss": 1.0218, "step": 115665 }, { "epoch": 0.29, "learning_rate": 7.115431073531632e-05, "loss": 1.0223, "step": 115670 }, { "epoch": 0.29, "learning_rate": 7.115305249383461e-05, "loss": 1.0233, "step": 115675 }, { "epoch": 0.29, "learning_rate": 7.115179425235292e-05, "loss": 1.0232, "step": 115680 }, { "epoch": 0.29, "learning_rate": 7.115053601087121e-05, "loss": 1.0412, "step": 115685 }, { "epoch": 0.29, "learning_rate": 7.11492777693895e-05, "loss": 1.023, "step": 115690 }, { "epoch": 0.29, "learning_rate": 7.114801952790779e-05, "loss": 1.0224, "step": 115695 }, { "epoch": 0.29, "learning_rate": 7.114676128642608e-05, "loss": 1.0207, "step": 115700 }, { "epoch": 0.29, "learning_rate": 7.114550304494439e-05, "loss": 1.0221, "step": 115705 }, { "epoch": 0.29, "learning_rate": 7.114424480346268e-05, "loss": 1.0226, "step": 115710 }, { "epoch": 0.29, "learning_rate": 7.114298656198098e-05, "loss": 1.0215, "step": 115715 }, { "epoch": 0.29, "learning_rate": 7.114172832049928e-05, "loss": 1.0215, "step": 115720 }, { "epoch": 0.29, "learning_rate": 7.114047007901757e-05, "loss": 1.0198, "step": 115725 }, { "epoch": 0.29, "learning_rate": 7.113921183753587e-05, "loss": 1.0225, "step": 115730 }, { "epoch": 0.29, "learning_rate": 7.113795359605416e-05, "loss": 1.0232, "step": 115735 }, { "epoch": 0.29, "learning_rate": 7.113669535457246e-05, "loss": 1.0244, "step": 115740 }, { "epoch": 0.29, "learning_rate": 7.113543711309075e-05, "loss": 1.0228, "step": 115745 }, { "epoch": 0.29, "learning_rate": 7.113417887160905e-05, "loss": 1.0217, "step": 115750 }, { "epoch": 0.29, "learning_rate": 7.113292063012734e-05, "loss": 1.0222, "step": 115755 }, { "epoch": 0.29, "learning_rate": 7.113166238864564e-05, "loss": 1.0241, "step": 115760 }, { "epoch": 0.29, "learning_rate": 7.113040414716393e-05, "loss": 1.0204, "step": 115765 }, { "epoch": 0.29, "learning_rate": 7.112914590568222e-05, "loss": 1.0212, "step": 115770 }, { "epoch": 0.29, "learning_rate": 7.112788766420052e-05, "loss": 1.0222, "step": 115775 }, { "epoch": 0.29, "learning_rate": 7.112662942271882e-05, "loss": 1.0226, "step": 115780 }, { "epoch": 0.29, "learning_rate": 7.112537118123711e-05, "loss": 1.0219, "step": 115785 }, { "epoch": 0.29, "learning_rate": 7.11241129397554e-05, "loss": 1.0243, "step": 115790 }, { "epoch": 0.29, "learning_rate": 7.11228546982737e-05, "loss": 1.0225, "step": 115795 }, { "epoch": 0.29, "learning_rate": 7.1121596456792e-05, "loss": 1.0201, "step": 115800 }, { "epoch": 0.29, "learning_rate": 7.112033821531029e-05, "loss": 1.0236, "step": 115805 }, { "epoch": 0.29, "learning_rate": 7.111907997382858e-05, "loss": 1.0212, "step": 115810 }, { "epoch": 0.29, "learning_rate": 7.111782173234688e-05, "loss": 1.022, "step": 115815 }, { "epoch": 0.29, "learning_rate": 7.111656349086518e-05, "loss": 1.0221, "step": 115820 }, { "epoch": 0.29, "learning_rate": 7.111530524938347e-05, "loss": 1.021, "step": 115825 }, { "epoch": 0.29, "learning_rate": 7.111404700790176e-05, "loss": 1.0207, "step": 115830 }, { "epoch": 0.29, "learning_rate": 7.111278876642005e-05, "loss": 1.0206, "step": 115835 }, { "epoch": 0.29, "learning_rate": 7.111153052493836e-05, "loss": 1.023, "step": 115840 }, { "epoch": 0.29, "learning_rate": 7.111027228345665e-05, "loss": 1.0225, "step": 115845 }, { "epoch": 0.29, "learning_rate": 7.110901404197494e-05, "loss": 1.0219, "step": 115850 }, { "epoch": 0.29, "learning_rate": 7.110775580049323e-05, "loss": 1.0241, "step": 115855 }, { "epoch": 0.29, "learning_rate": 7.110649755901154e-05, "loss": 1.0243, "step": 115860 }, { "epoch": 0.29, "learning_rate": 7.110523931752983e-05, "loss": 1.0227, "step": 115865 }, { "epoch": 0.29, "learning_rate": 7.110398107604812e-05, "loss": 1.0208, "step": 115870 }, { "epoch": 0.29, "learning_rate": 7.110272283456641e-05, "loss": 1.0212, "step": 115875 }, { "epoch": 0.29, "learning_rate": 7.110146459308472e-05, "loss": 1.0242, "step": 115880 }, { "epoch": 0.29, "learning_rate": 7.110020635160301e-05, "loss": 1.0238, "step": 115885 }, { "epoch": 0.29, "learning_rate": 7.10989481101213e-05, "loss": 1.0221, "step": 115890 }, { "epoch": 0.29, "learning_rate": 7.109768986863959e-05, "loss": 1.0234, "step": 115895 }, { "epoch": 0.29, "learning_rate": 7.109643162715788e-05, "loss": 1.0218, "step": 115900 }, { "epoch": 0.29, "learning_rate": 7.109517338567619e-05, "loss": 1.02, "step": 115905 }, { "epoch": 0.29, "learning_rate": 7.109391514419448e-05, "loss": 1.0221, "step": 115910 }, { "epoch": 0.29, "learning_rate": 7.109265690271277e-05, "loss": 1.0215, "step": 115915 }, { "epoch": 0.29, "learning_rate": 7.109139866123106e-05, "loss": 1.024, "step": 115920 }, { "epoch": 0.29, "learning_rate": 7.109014041974937e-05, "loss": 1.022, "step": 115925 }, { "epoch": 0.29, "learning_rate": 7.108888217826766e-05, "loss": 1.0224, "step": 115930 }, { "epoch": 0.29, "learning_rate": 7.108762393678595e-05, "loss": 1.0216, "step": 115935 }, { "epoch": 0.29, "learning_rate": 7.108636569530424e-05, "loss": 1.0227, "step": 115940 }, { "epoch": 0.29, "learning_rate": 7.108510745382255e-05, "loss": 1.0219, "step": 115945 }, { "epoch": 0.29, "learning_rate": 7.108384921234084e-05, "loss": 1.0216, "step": 115950 }, { "epoch": 0.29, "learning_rate": 7.108259097085913e-05, "loss": 1.0322, "step": 115955 }, { "epoch": 0.29, "learning_rate": 7.108133272937742e-05, "loss": 1.0243, "step": 115960 }, { "epoch": 0.29, "learning_rate": 7.108007448789571e-05, "loss": 1.024, "step": 115965 }, { "epoch": 0.29, "learning_rate": 7.107881624641402e-05, "loss": 1.0207, "step": 115970 }, { "epoch": 0.29, "learning_rate": 7.107755800493231e-05, "loss": 1.0252, "step": 115975 }, { "epoch": 0.29, "learning_rate": 7.10762997634506e-05, "loss": 1.0226, "step": 115980 }, { "epoch": 0.29, "learning_rate": 7.10750415219689e-05, "loss": 1.0241, "step": 115985 }, { "epoch": 0.29, "learning_rate": 7.10737832804872e-05, "loss": 1.0201, "step": 115990 }, { "epoch": 0.29, "learning_rate": 7.107252503900549e-05, "loss": 1.0472, "step": 115995 }, { "epoch": 0.29, "learning_rate": 7.107126679752378e-05, "loss": 1.0205, "step": 116000 }, { "epoch": 0.29, "learning_rate": 7.107000855604207e-05, "loss": 1.0221, "step": 116005 }, { "epoch": 0.29, "learning_rate": 7.106875031456038e-05, "loss": 1.0214, "step": 116010 }, { "epoch": 0.29, "learning_rate": 7.106749207307867e-05, "loss": 1.0217, "step": 116015 }, { "epoch": 0.29, "learning_rate": 7.106623383159696e-05, "loss": 1.0234, "step": 116020 }, { "epoch": 0.29, "learning_rate": 7.106497559011525e-05, "loss": 1.0212, "step": 116025 }, { "epoch": 0.29, "learning_rate": 7.106371734863355e-05, "loss": 1.02, "step": 116030 }, { "epoch": 0.29, "learning_rate": 7.106245910715185e-05, "loss": 1.0217, "step": 116035 }, { "epoch": 0.29, "learning_rate": 7.106120086567014e-05, "loss": 1.0208, "step": 116040 }, { "epoch": 0.29, "learning_rate": 7.105994262418843e-05, "loss": 1.0225, "step": 116045 }, { "epoch": 0.29, "learning_rate": 7.105868438270673e-05, "loss": 1.02, "step": 116050 }, { "epoch": 0.29, "learning_rate": 7.105742614122503e-05, "loss": 1.0217, "step": 116055 }, { "epoch": 0.29, "learning_rate": 7.105616789974332e-05, "loss": 1.0235, "step": 116060 }, { "epoch": 0.29, "learning_rate": 7.105490965826161e-05, "loss": 1.0411, "step": 116065 }, { "epoch": 0.29, "learning_rate": 7.10536514167799e-05, "loss": 1.0218, "step": 116070 }, { "epoch": 0.29, "learning_rate": 7.105239317529821e-05, "loss": 1.0202, "step": 116075 }, { "epoch": 0.29, "learning_rate": 7.10511349338165e-05, "loss": 1.0224, "step": 116080 }, { "epoch": 0.29, "learning_rate": 7.10498766923348e-05, "loss": 1.0242, "step": 116085 }, { "epoch": 0.29, "learning_rate": 7.104861845085309e-05, "loss": 1.0252, "step": 116090 }, { "epoch": 0.29, "learning_rate": 7.104736020937138e-05, "loss": 1.022, "step": 116095 }, { "epoch": 0.29, "learning_rate": 7.104610196788968e-05, "loss": 1.023, "step": 116100 }, { "epoch": 0.29, "learning_rate": 7.104484372640797e-05, "loss": 1.0218, "step": 116105 }, { "epoch": 0.29, "learning_rate": 7.104358548492626e-05, "loss": 1.0234, "step": 116110 }, { "epoch": 0.29, "learning_rate": 7.104232724344456e-05, "loss": 1.022, "step": 116115 }, { "epoch": 0.29, "learning_rate": 7.104106900196286e-05, "loss": 1.0229, "step": 116120 }, { "epoch": 0.29, "learning_rate": 7.103981076048115e-05, "loss": 1.0214, "step": 116125 }, { "epoch": 0.29, "learning_rate": 7.103855251899944e-05, "loss": 1.0244, "step": 116130 }, { "epoch": 0.29, "learning_rate": 7.103729427751774e-05, "loss": 1.023, "step": 116135 }, { "epoch": 0.29, "learning_rate": 7.103603603603604e-05, "loss": 1.0236, "step": 116140 }, { "epoch": 0.29, "learning_rate": 7.103477779455433e-05, "loss": 1.021, "step": 116145 }, { "epoch": 0.29, "learning_rate": 7.103351955307262e-05, "loss": 1.0204, "step": 116150 }, { "epoch": 0.29, "learning_rate": 7.103226131159092e-05, "loss": 1.023, "step": 116155 }, { "epoch": 0.29, "learning_rate": 7.103100307010921e-05, "loss": 1.0212, "step": 116160 }, { "epoch": 0.29, "learning_rate": 7.102974482862751e-05, "loss": 1.0234, "step": 116165 }, { "epoch": 0.29, "learning_rate": 7.10284865871458e-05, "loss": 1.0231, "step": 116170 }, { "epoch": 0.29, "learning_rate": 7.10272283456641e-05, "loss": 1.022, "step": 116175 }, { "epoch": 0.29, "learning_rate": 7.102597010418239e-05, "loss": 1.0228, "step": 116180 }, { "epoch": 0.29, "learning_rate": 7.102471186270069e-05, "loss": 1.0217, "step": 116185 }, { "epoch": 0.29, "learning_rate": 7.102345362121898e-05, "loss": 1.0217, "step": 116190 }, { "epoch": 0.29, "learning_rate": 7.102219537973728e-05, "loss": 1.0221, "step": 116195 }, { "epoch": 0.29, "learning_rate": 7.102093713825557e-05, "loss": 1.0216, "step": 116200 }, { "epoch": 0.29, "learning_rate": 7.101967889677387e-05, "loss": 1.0453, "step": 116205 }, { "epoch": 0.29, "learning_rate": 7.101842065529216e-05, "loss": 1.0224, "step": 116210 }, { "epoch": 0.29, "learning_rate": 7.101716241381047e-05, "loss": 1.0237, "step": 116215 }, { "epoch": 0.29, "learning_rate": 7.101590417232876e-05, "loss": 1.0207, "step": 116220 }, { "epoch": 0.29, "learning_rate": 7.101464593084705e-05, "loss": 1.0234, "step": 116225 }, { "epoch": 0.29, "learning_rate": 7.101338768936536e-05, "loss": 1.0221, "step": 116230 }, { "epoch": 0.29, "learning_rate": 7.101212944788365e-05, "loss": 1.022, "step": 116235 }, { "epoch": 0.29, "learning_rate": 7.101087120640194e-05, "loss": 1.022, "step": 116240 }, { "epoch": 0.29, "learning_rate": 7.100961296492023e-05, "loss": 1.0243, "step": 116245 }, { "epoch": 0.29, "learning_rate": 7.100835472343852e-05, "loss": 1.0206, "step": 116250 }, { "epoch": 0.29, "learning_rate": 7.100709648195683e-05, "loss": 1.0233, "step": 116255 }, { "epoch": 0.29, "learning_rate": 7.100583824047512e-05, "loss": 1.022, "step": 116260 }, { "epoch": 0.29, "learning_rate": 7.100457999899341e-05, "loss": 1.0232, "step": 116265 }, { "epoch": 0.29, "learning_rate": 7.10033217575117e-05, "loss": 1.0237, "step": 116270 }, { "epoch": 0.29, "learning_rate": 7.100206351603001e-05, "loss": 1.0243, "step": 116275 }, { "epoch": 0.29, "learning_rate": 7.10008052745483e-05, "loss": 1.0228, "step": 116280 }, { "epoch": 0.29, "learning_rate": 7.099954703306659e-05, "loss": 1.0219, "step": 116285 }, { "epoch": 0.29, "learning_rate": 7.099828879158488e-05, "loss": 1.0219, "step": 116290 }, { "epoch": 0.29, "learning_rate": 7.099703055010319e-05, "loss": 1.0215, "step": 116295 }, { "epoch": 0.29, "learning_rate": 7.099577230862148e-05, "loss": 1.0214, "step": 116300 }, { "epoch": 0.29, "learning_rate": 7.099451406713977e-05, "loss": 1.023, "step": 116305 }, { "epoch": 0.29, "learning_rate": 7.099325582565806e-05, "loss": 1.0229, "step": 116310 }, { "epoch": 0.29, "learning_rate": 7.099199758417636e-05, "loss": 1.0216, "step": 116315 }, { "epoch": 0.29, "learning_rate": 7.099073934269466e-05, "loss": 1.0236, "step": 116320 }, { "epoch": 0.29, "learning_rate": 7.098948110121295e-05, "loss": 1.0239, "step": 116325 }, { "epoch": 0.29, "learning_rate": 7.098822285973124e-05, "loss": 1.022, "step": 116330 }, { "epoch": 0.29, "learning_rate": 7.098696461824954e-05, "loss": 1.0206, "step": 116335 }, { "epoch": 0.29, "learning_rate": 7.098570637676784e-05, "loss": 1.0229, "step": 116340 }, { "epoch": 0.29, "learning_rate": 7.098444813528613e-05, "loss": 1.0218, "step": 116345 }, { "epoch": 0.29, "learning_rate": 7.098318989380442e-05, "loss": 1.0208, "step": 116350 }, { "epoch": 0.29, "learning_rate": 7.098193165232272e-05, "loss": 1.022, "step": 116355 }, { "epoch": 0.29, "learning_rate": 7.098067341084102e-05, "loss": 1.0219, "step": 116360 }, { "epoch": 0.29, "learning_rate": 7.097941516935931e-05, "loss": 1.0214, "step": 116365 }, { "epoch": 0.29, "learning_rate": 7.09781569278776e-05, "loss": 1.0234, "step": 116370 }, { "epoch": 0.29, "learning_rate": 7.09768986863959e-05, "loss": 1.0216, "step": 116375 }, { "epoch": 0.29, "learning_rate": 7.097564044491419e-05, "loss": 1.0231, "step": 116380 }, { "epoch": 0.29, "learning_rate": 7.097438220343249e-05, "loss": 1.0187, "step": 116385 }, { "epoch": 0.29, "learning_rate": 7.097312396195078e-05, "loss": 1.023, "step": 116390 }, { "epoch": 0.29, "learning_rate": 7.097186572046908e-05, "loss": 1.0233, "step": 116395 }, { "epoch": 0.29, "learning_rate": 7.097060747898737e-05, "loss": 1.0221, "step": 116400 }, { "epoch": 0.29, "learning_rate": 7.096934923750567e-05, "loss": 1.0226, "step": 116405 }, { "epoch": 0.29, "learning_rate": 7.096809099602396e-05, "loss": 1.025, "step": 116410 }, { "epoch": 0.29, "learning_rate": 7.096683275454226e-05, "loss": 1.0235, "step": 116415 }, { "epoch": 0.29, "learning_rate": 7.096557451306055e-05, "loss": 1.0238, "step": 116420 }, { "epoch": 0.29, "learning_rate": 7.096431627157884e-05, "loss": 1.022, "step": 116425 }, { "epoch": 0.29, "learning_rate": 7.096305803009714e-05, "loss": 1.0229, "step": 116430 }, { "epoch": 0.29, "learning_rate": 7.096179978861544e-05, "loss": 1.0216, "step": 116435 }, { "epoch": 0.29, "learning_rate": 7.096054154713373e-05, "loss": 1.0207, "step": 116440 }, { "epoch": 0.29, "learning_rate": 7.095928330565202e-05, "loss": 1.0202, "step": 116445 }, { "epoch": 0.29, "learning_rate": 7.095802506417032e-05, "loss": 1.021, "step": 116450 }, { "epoch": 0.29, "learning_rate": 7.095676682268862e-05, "loss": 1.0223, "step": 116455 }, { "epoch": 0.29, "learning_rate": 7.095550858120691e-05, "loss": 1.019, "step": 116460 }, { "epoch": 0.29, "learning_rate": 7.09542503397252e-05, "loss": 1.0202, "step": 116465 }, { "epoch": 0.29, "learning_rate": 7.09529920982435e-05, "loss": 1.0225, "step": 116470 }, { "epoch": 0.29, "learning_rate": 7.09517338567618e-05, "loss": 1.0214, "step": 116475 }, { "epoch": 0.29, "learning_rate": 7.095047561528009e-05, "loss": 1.024, "step": 116480 }, { "epoch": 0.29, "learning_rate": 7.094921737379838e-05, "loss": 1.0205, "step": 116485 }, { "epoch": 0.29, "learning_rate": 7.094795913231667e-05, "loss": 1.0195, "step": 116490 }, { "epoch": 0.29, "learning_rate": 7.094670089083498e-05, "loss": 1.0196, "step": 116495 }, { "epoch": 0.29, "learning_rate": 7.094544264935327e-05, "loss": 1.0231, "step": 116500 }, { "epoch": 0.29, "learning_rate": 7.094418440787156e-05, "loss": 1.0215, "step": 116505 }, { "epoch": 0.29, "learning_rate": 7.094292616638985e-05, "loss": 1.0239, "step": 116510 }, { "epoch": 0.29, "learning_rate": 7.094166792490816e-05, "loss": 1.0226, "step": 116515 }, { "epoch": 0.29, "learning_rate": 7.094040968342645e-05, "loss": 1.0206, "step": 116520 }, { "epoch": 0.29, "learning_rate": 7.093915144194474e-05, "loss": 1.0224, "step": 116525 }, { "epoch": 0.29, "learning_rate": 7.093789320046303e-05, "loss": 1.0226, "step": 116530 }, { "epoch": 0.29, "learning_rate": 7.093663495898134e-05, "loss": 1.0201, "step": 116535 }, { "epoch": 0.29, "learning_rate": 7.093537671749963e-05, "loss": 1.025, "step": 116540 }, { "epoch": 0.29, "learning_rate": 7.093411847601792e-05, "loss": 1.0208, "step": 116545 }, { "epoch": 0.29, "learning_rate": 7.093286023453621e-05, "loss": 1.0228, "step": 116550 }, { "epoch": 0.29, "learning_rate": 7.09316019930545e-05, "loss": 1.02, "step": 116555 }, { "epoch": 0.29, "learning_rate": 7.093034375157281e-05, "loss": 1.0221, "step": 116560 }, { "epoch": 0.29, "learning_rate": 7.09290855100911e-05, "loss": 1.0211, "step": 116565 }, { "epoch": 0.29, "learning_rate": 7.092782726860939e-05, "loss": 1.021, "step": 116570 }, { "epoch": 0.29, "learning_rate": 7.092656902712768e-05, "loss": 1.0205, "step": 116575 }, { "epoch": 0.29, "learning_rate": 7.092531078564599e-05, "loss": 1.0233, "step": 116580 }, { "epoch": 0.29, "learning_rate": 7.092405254416428e-05, "loss": 1.0225, "step": 116585 }, { "epoch": 0.29, "learning_rate": 7.092279430268257e-05, "loss": 1.0219, "step": 116590 }, { "epoch": 0.29, "learning_rate": 7.092153606120086e-05, "loss": 1.0221, "step": 116595 }, { "epoch": 0.29, "learning_rate": 7.092027781971917e-05, "loss": 1.0217, "step": 116600 }, { "epoch": 0.29, "learning_rate": 7.091901957823746e-05, "loss": 1.0214, "step": 116605 }, { "epoch": 0.29, "learning_rate": 7.091776133675575e-05, "loss": 1.0221, "step": 116610 }, { "epoch": 0.29, "learning_rate": 7.091650309527404e-05, "loss": 1.0221, "step": 116615 }, { "epoch": 0.29, "learning_rate": 7.091524485379233e-05, "loss": 1.0219, "step": 116620 }, { "epoch": 0.29, "learning_rate": 7.091398661231064e-05, "loss": 1.0213, "step": 116625 }, { "epoch": 0.29, "learning_rate": 7.091272837082893e-05, "loss": 1.0208, "step": 116630 }, { "epoch": 0.29, "learning_rate": 7.091147012934722e-05, "loss": 1.0201, "step": 116635 }, { "epoch": 0.29, "learning_rate": 7.091021188786551e-05, "loss": 1.0202, "step": 116640 }, { "epoch": 0.29, "learning_rate": 7.090895364638382e-05, "loss": 1.0231, "step": 116645 }, { "epoch": 0.29, "learning_rate": 7.090769540490211e-05, "loss": 1.0233, "step": 116650 }, { "epoch": 0.29, "learning_rate": 7.09064371634204e-05, "loss": 1.0246, "step": 116655 }, { "epoch": 0.29, "learning_rate": 7.090517892193869e-05, "loss": 1.0216, "step": 116660 }, { "epoch": 0.29, "learning_rate": 7.0903920680457e-05, "loss": 1.022, "step": 116665 }, { "epoch": 0.29, "learning_rate": 7.090266243897529e-05, "loss": 1.0205, "step": 116670 }, { "epoch": 0.29, "learning_rate": 7.090140419749358e-05, "loss": 1.0221, "step": 116675 }, { "epoch": 0.29, "learning_rate": 7.090014595601187e-05, "loss": 1.0226, "step": 116680 }, { "epoch": 0.29, "learning_rate": 7.089888771453016e-05, "loss": 1.0222, "step": 116685 }, { "epoch": 0.29, "learning_rate": 7.089762947304847e-05, "loss": 1.0234, "step": 116690 }, { "epoch": 0.29, "learning_rate": 7.089637123156676e-05, "loss": 1.0211, "step": 116695 }, { "epoch": 0.29, "learning_rate": 7.089511299008505e-05, "loss": 1.0211, "step": 116700 }, { "epoch": 0.29, "learning_rate": 7.089385474860334e-05, "loss": 1.0207, "step": 116705 }, { "epoch": 0.29, "learning_rate": 7.089259650712165e-05, "loss": 1.0202, "step": 116710 }, { "epoch": 0.29, "learning_rate": 7.089133826563995e-05, "loss": 1.0219, "step": 116715 }, { "epoch": 0.29, "learning_rate": 7.089008002415825e-05, "loss": 1.0219, "step": 116720 }, { "epoch": 0.29, "learning_rate": 7.088882178267654e-05, "loss": 1.0204, "step": 116725 }, { "epoch": 0.29, "learning_rate": 7.088756354119483e-05, "loss": 1.0218, "step": 116730 }, { "epoch": 0.29, "learning_rate": 7.088630529971313e-05, "loss": 1.024, "step": 116735 }, { "epoch": 0.29, "learning_rate": 7.088504705823143e-05, "loss": 1.0221, "step": 116740 }, { "epoch": 0.29, "learning_rate": 7.088378881674972e-05, "loss": 1.0259, "step": 116745 }, { "epoch": 0.29, "learning_rate": 7.088253057526801e-05, "loss": 1.0227, "step": 116750 }, { "epoch": 0.29, "learning_rate": 7.088127233378631e-05, "loss": 1.0238, "step": 116755 }, { "epoch": 0.29, "learning_rate": 7.08800140923046e-05, "loss": 1.04, "step": 116760 }, { "epoch": 0.29, "learning_rate": 7.08787558508229e-05, "loss": 1.0209, "step": 116765 }, { "epoch": 0.29, "learning_rate": 7.087749760934119e-05, "loss": 1.0217, "step": 116770 }, { "epoch": 0.29, "learning_rate": 7.087623936785948e-05, "loss": 1.0249, "step": 116775 }, { "epoch": 0.29, "learning_rate": 7.087498112637779e-05, "loss": 1.0245, "step": 116780 }, { "epoch": 0.29, "learning_rate": 7.087372288489608e-05, "loss": 1.0202, "step": 116785 }, { "epoch": 0.29, "learning_rate": 7.087246464341437e-05, "loss": 1.0233, "step": 116790 }, { "epoch": 0.29, "learning_rate": 7.087120640193266e-05, "loss": 1.0203, "step": 116795 }, { "epoch": 0.29, "learning_rate": 7.086994816045097e-05, "loss": 1.0201, "step": 116800 }, { "epoch": 0.29, "learning_rate": 7.086868991896926e-05, "loss": 1.0223, "step": 116805 }, { "epoch": 0.29, "learning_rate": 7.086743167748755e-05, "loss": 1.0234, "step": 116810 }, { "epoch": 0.29, "learning_rate": 7.086617343600584e-05, "loss": 1.0221, "step": 116815 }, { "epoch": 0.29, "learning_rate": 7.086491519452415e-05, "loss": 1.0199, "step": 116820 }, { "epoch": 0.29, "learning_rate": 7.086365695304244e-05, "loss": 1.0193, "step": 116825 }, { "epoch": 0.29, "learning_rate": 7.086239871156073e-05, "loss": 1.0198, "step": 116830 }, { "epoch": 0.29, "learning_rate": 7.086114047007902e-05, "loss": 1.0238, "step": 116835 }, { "epoch": 0.29, "learning_rate": 7.085988222859731e-05, "loss": 1.0247, "step": 116840 }, { "epoch": 0.29, "learning_rate": 7.085862398711562e-05, "loss": 1.0352, "step": 116845 }, { "epoch": 0.29, "learning_rate": 7.085736574563391e-05, "loss": 1.0227, "step": 116850 }, { "epoch": 0.29, "learning_rate": 7.08561075041522e-05, "loss": 1.0225, "step": 116855 }, { "epoch": 0.29, "learning_rate": 7.085484926267049e-05, "loss": 1.0185, "step": 116860 }, { "epoch": 0.29, "learning_rate": 7.08535910211888e-05, "loss": 1.0215, "step": 116865 }, { "epoch": 0.29, "learning_rate": 7.085233277970709e-05, "loss": 1.02, "step": 116870 }, { "epoch": 0.29, "learning_rate": 7.085107453822538e-05, "loss": 1.024, "step": 116875 }, { "epoch": 0.29, "learning_rate": 7.084981629674367e-05, "loss": 1.0219, "step": 116880 }, { "epoch": 0.29, "learning_rate": 7.084855805526198e-05, "loss": 1.0219, "step": 116885 }, { "epoch": 0.29, "learning_rate": 7.084729981378027e-05, "loss": 1.0212, "step": 116890 }, { "epoch": 0.29, "learning_rate": 7.084604157229856e-05, "loss": 1.0224, "step": 116895 }, { "epoch": 0.29, "learning_rate": 7.084478333081685e-05, "loss": 1.0238, "step": 116900 }, { "epoch": 0.29, "learning_rate": 7.084352508933514e-05, "loss": 1.0206, "step": 116905 }, { "epoch": 0.29, "learning_rate": 7.084226684785345e-05, "loss": 1.0204, "step": 116910 }, { "epoch": 0.29, "learning_rate": 7.084100860637174e-05, "loss": 1.024, "step": 116915 }, { "epoch": 0.29, "learning_rate": 7.083975036489003e-05, "loss": 1.0234, "step": 116920 }, { "epoch": 0.29, "learning_rate": 7.083849212340832e-05, "loss": 1.023, "step": 116925 }, { "epoch": 0.29, "learning_rate": 7.083723388192663e-05, "loss": 1.0213, "step": 116930 }, { "epoch": 0.29, "learning_rate": 7.083597564044492e-05, "loss": 1.0213, "step": 116935 }, { "epoch": 0.29, "learning_rate": 7.083471739896321e-05, "loss": 1.0224, "step": 116940 }, { "epoch": 0.29, "learning_rate": 7.08334591574815e-05, "loss": 1.0212, "step": 116945 }, { "epoch": 0.29, "learning_rate": 7.083220091599981e-05, "loss": 1.0254, "step": 116950 }, { "epoch": 0.29, "learning_rate": 7.08309426745181e-05, "loss": 1.0222, "step": 116955 }, { "epoch": 0.29, "learning_rate": 7.082968443303639e-05, "loss": 1.0227, "step": 116960 }, { "epoch": 0.29, "learning_rate": 7.082842619155468e-05, "loss": 1.0441, "step": 116965 }, { "epoch": 0.29, "learning_rate": 7.082716795007298e-05, "loss": 1.021, "step": 116970 }, { "epoch": 0.29, "learning_rate": 7.082590970859128e-05, "loss": 1.0235, "step": 116975 }, { "epoch": 0.29, "learning_rate": 7.082465146710957e-05, "loss": 1.0223, "step": 116980 }, { "epoch": 0.29, "learning_rate": 7.082339322562786e-05, "loss": 1.023, "step": 116985 }, { "epoch": 0.29, "learning_rate": 7.082213498414616e-05, "loss": 1.0211, "step": 116990 }, { "epoch": 0.29, "learning_rate": 7.082087674266446e-05, "loss": 1.023, "step": 116995 }, { "epoch": 0.29, "learning_rate": 7.081961850118275e-05, "loss": 1.0227, "step": 117000 }, { "epoch": 0.29, "learning_rate": 7.081836025970104e-05, "loss": 1.0223, "step": 117005 }, { "epoch": 0.29, "learning_rate": 7.081710201821934e-05, "loss": 1.0215, "step": 117010 }, { "epoch": 0.29, "learning_rate": 7.081584377673764e-05, "loss": 1.0218, "step": 117015 }, { "epoch": 0.29, "learning_rate": 7.081458553525593e-05, "loss": 1.0225, "step": 117020 }, { "epoch": 0.29, "learning_rate": 7.081332729377422e-05, "loss": 1.0205, "step": 117025 }, { "epoch": 0.29, "learning_rate": 7.081206905229252e-05, "loss": 1.0216, "step": 117030 }, { "epoch": 0.29, "learning_rate": 7.081081081081081e-05, "loss": 1.0228, "step": 117035 }, { "epoch": 0.29, "learning_rate": 7.080955256932911e-05, "loss": 1.0236, "step": 117040 }, { "epoch": 0.29, "learning_rate": 7.08082943278474e-05, "loss": 1.0205, "step": 117045 }, { "epoch": 0.29, "learning_rate": 7.08070360863657e-05, "loss": 1.0223, "step": 117050 }, { "epoch": 0.29, "learning_rate": 7.080577784488399e-05, "loss": 1.0204, "step": 117055 }, { "epoch": 0.29, "learning_rate": 7.080451960340229e-05, "loss": 1.022, "step": 117060 }, { "epoch": 0.29, "learning_rate": 7.080326136192058e-05, "loss": 1.0217, "step": 117065 }, { "epoch": 0.29, "learning_rate": 7.080200312043888e-05, "loss": 1.0206, "step": 117070 }, { "epoch": 0.29, "learning_rate": 7.080074487895717e-05, "loss": 1.0228, "step": 117075 }, { "epoch": 0.29, "learning_rate": 7.079948663747546e-05, "loss": 1.024, "step": 117080 }, { "epoch": 0.29, "learning_rate": 7.079822839599376e-05, "loss": 1.0209, "step": 117085 }, { "epoch": 0.29, "learning_rate": 7.079697015451206e-05, "loss": 1.021, "step": 117090 }, { "epoch": 0.29, "learning_rate": 7.079571191303035e-05, "loss": 1.0241, "step": 117095 }, { "epoch": 0.29, "learning_rate": 7.079445367154864e-05, "loss": 1.0222, "step": 117100 }, { "epoch": 0.29, "learning_rate": 7.079319543006694e-05, "loss": 1.0213, "step": 117105 }, { "epoch": 0.29, "learning_rate": 7.079193718858523e-05, "loss": 1.0228, "step": 117110 }, { "epoch": 0.29, "learning_rate": 7.079067894710353e-05, "loss": 1.0218, "step": 117115 }, { "epoch": 0.29, "learning_rate": 7.078942070562182e-05, "loss": 1.0232, "step": 117120 }, { "epoch": 0.29, "learning_rate": 7.078816246414012e-05, "loss": 1.0213, "step": 117125 }, { "epoch": 0.29, "learning_rate": 7.078690422265841e-05, "loss": 1.0239, "step": 117130 }, { "epoch": 0.29, "learning_rate": 7.07856459811767e-05, "loss": 1.0214, "step": 117135 }, { "epoch": 0.29, "learning_rate": 7.0784387739695e-05, "loss": 1.0209, "step": 117140 }, { "epoch": 0.29, "learning_rate": 7.078312949821329e-05, "loss": 1.0353, "step": 117145 }, { "epoch": 0.29, "learning_rate": 7.07818712567316e-05, "loss": 1.0238, "step": 117150 }, { "epoch": 0.29, "learning_rate": 7.078061301524989e-05, "loss": 1.0222, "step": 117155 }, { "epoch": 0.29, "learning_rate": 7.077935477376818e-05, "loss": 1.0236, "step": 117160 }, { "epoch": 0.29, "learning_rate": 7.077809653228647e-05, "loss": 1.0223, "step": 117165 }, { "epoch": 0.29, "learning_rate": 7.077683829080477e-05, "loss": 1.0209, "step": 117170 }, { "epoch": 0.29, "learning_rate": 7.077558004932307e-05, "loss": 1.0219, "step": 117175 }, { "epoch": 0.29, "learning_rate": 7.077432180784136e-05, "loss": 1.0224, "step": 117180 }, { "epoch": 0.29, "learning_rate": 7.077306356635965e-05, "loss": 1.0229, "step": 117185 }, { "epoch": 0.29, "learning_rate": 7.077180532487795e-05, "loss": 1.0239, "step": 117190 }, { "epoch": 0.29, "learning_rate": 7.077079873169259e-05, "loss": 1.0226, "step": 117195 }, { "epoch": 0.29, "learning_rate": 7.076954049021088e-05, "loss": 1.0227, "step": 117200 }, { "epoch": 0.29, "learning_rate": 7.076828224872917e-05, "loss": 1.0223, "step": 117205 }, { "epoch": 0.29, "learning_rate": 7.076702400724748e-05, "loss": 1.0216, "step": 117210 }, { "epoch": 0.29, "learning_rate": 7.076576576576577e-05, "loss": 1.0218, "step": 117215 }, { "epoch": 0.29, "learning_rate": 7.076450752428406e-05, "loss": 1.0219, "step": 117220 }, { "epoch": 0.29, "learning_rate": 7.076324928280235e-05, "loss": 1.0235, "step": 117225 }, { "epoch": 0.29, "learning_rate": 7.076199104132065e-05, "loss": 1.0207, "step": 117230 }, { "epoch": 0.29, "learning_rate": 7.076073279983895e-05, "loss": 1.0253, "step": 117235 }, { "epoch": 0.29, "learning_rate": 7.075947455835724e-05, "loss": 1.0214, "step": 117240 }, { "epoch": 0.29, "learning_rate": 7.075821631687553e-05, "loss": 1.022, "step": 117245 }, { "epoch": 0.29, "learning_rate": 7.075695807539383e-05, "loss": 1.0224, "step": 117250 }, { "epoch": 0.29, "learning_rate": 7.075569983391213e-05, "loss": 1.0225, "step": 117255 }, { "epoch": 0.29, "learning_rate": 7.075444159243042e-05, "loss": 1.0222, "step": 117260 }, { "epoch": 0.29, "learning_rate": 7.075318335094871e-05, "loss": 1.0215, "step": 117265 }, { "epoch": 0.29, "learning_rate": 7.0751925109467e-05, "loss": 1.0216, "step": 117270 }, { "epoch": 0.29, "learning_rate": 7.075066686798531e-05, "loss": 1.0218, "step": 117275 }, { "epoch": 0.29, "learning_rate": 7.07494086265036e-05, "loss": 1.0219, "step": 117280 }, { "epoch": 0.29, "learning_rate": 7.07481503850219e-05, "loss": 1.0243, "step": 117285 }, { "epoch": 0.29, "learning_rate": 7.074689214354019e-05, "loss": 1.0242, "step": 117290 }, { "epoch": 0.29, "learning_rate": 7.074563390205848e-05, "loss": 1.0225, "step": 117295 }, { "epoch": 0.29, "learning_rate": 7.074437566057678e-05, "loss": 1.0234, "step": 117300 }, { "epoch": 0.29, "learning_rate": 7.074311741909507e-05, "loss": 1.0215, "step": 117305 }, { "epoch": 0.29, "learning_rate": 7.074185917761337e-05, "loss": 1.0206, "step": 117310 }, { "epoch": 0.29, "learning_rate": 7.074060093613166e-05, "loss": 1.0222, "step": 117315 }, { "epoch": 0.29, "learning_rate": 7.073934269464996e-05, "loss": 1.0218, "step": 117320 }, { "epoch": 0.29, "learning_rate": 7.073808445316825e-05, "loss": 1.0235, "step": 117325 }, { "epoch": 0.29, "learning_rate": 7.073682621168655e-05, "loss": 1.0219, "step": 117330 }, { "epoch": 0.29, "learning_rate": 7.073556797020484e-05, "loss": 1.0203, "step": 117335 }, { "epoch": 0.29, "learning_rate": 7.073430972872314e-05, "loss": 1.0218, "step": 117340 }, { "epoch": 0.29, "learning_rate": 7.073305148724143e-05, "loss": 1.0227, "step": 117345 }, { "epoch": 0.29, "learning_rate": 7.073179324575973e-05, "loss": 1.0202, "step": 117350 }, { "epoch": 0.29, "learning_rate": 7.073053500427802e-05, "loss": 1.0216, "step": 117355 }, { "epoch": 0.29, "learning_rate": 7.072927676279631e-05, "loss": 1.0218, "step": 117360 }, { "epoch": 0.29, "learning_rate": 7.072801852131461e-05, "loss": 1.0229, "step": 117365 }, { "epoch": 0.29, "learning_rate": 7.07267602798329e-05, "loss": 1.0219, "step": 117370 }, { "epoch": 0.29, "learning_rate": 7.07255020383512e-05, "loss": 1.0236, "step": 117375 }, { "epoch": 0.29, "learning_rate": 7.072424379686949e-05, "loss": 1.0222, "step": 117380 }, { "epoch": 0.29, "learning_rate": 7.07229855553878e-05, "loss": 1.0218, "step": 117385 }, { "epoch": 0.29, "learning_rate": 7.072172731390609e-05, "loss": 1.0227, "step": 117390 }, { "epoch": 0.29, "learning_rate": 7.072046907242438e-05, "loss": 1.0222, "step": 117395 }, { "epoch": 0.29, "learning_rate": 7.071921083094267e-05, "loss": 1.0209, "step": 117400 }, { "epoch": 0.29, "learning_rate": 7.071795258946097e-05, "loss": 1.0211, "step": 117405 }, { "epoch": 0.29, "learning_rate": 7.071669434797926e-05, "loss": 1.0222, "step": 117410 }, { "epoch": 0.29, "learning_rate": 7.071543610649757e-05, "loss": 1.0205, "step": 117415 }, { "epoch": 0.29, "learning_rate": 7.071417786501586e-05, "loss": 1.0224, "step": 117420 }, { "epoch": 0.29, "learning_rate": 7.071291962353415e-05, "loss": 1.0242, "step": 117425 }, { "epoch": 0.29, "learning_rate": 7.071166138205246e-05, "loss": 1.0217, "step": 117430 }, { "epoch": 0.29, "learning_rate": 7.071040314057075e-05, "loss": 1.0474, "step": 117435 }, { "epoch": 0.29, "learning_rate": 7.070914489908904e-05, "loss": 1.0221, "step": 117440 }, { "epoch": 0.29, "learning_rate": 7.070788665760733e-05, "loss": 1.0239, "step": 117445 }, { "epoch": 0.29, "learning_rate": 7.070662841612562e-05, "loss": 1.0217, "step": 117450 }, { "epoch": 0.29, "learning_rate": 7.070537017464393e-05, "loss": 1.023, "step": 117455 }, { "epoch": 0.29, "learning_rate": 7.070411193316222e-05, "loss": 1.0228, "step": 117460 }, { "epoch": 0.29, "learning_rate": 7.070285369168051e-05, "loss": 1.0193, "step": 117465 }, { "epoch": 0.29, "learning_rate": 7.07015954501988e-05, "loss": 1.0216, "step": 117470 }, { "epoch": 0.29, "learning_rate": 7.070033720871711e-05, "loss": 1.024, "step": 117475 }, { "epoch": 0.29, "learning_rate": 7.06990789672354e-05, "loss": 1.0199, "step": 117480 }, { "epoch": 0.29, "learning_rate": 7.069782072575369e-05, "loss": 1.0221, "step": 117485 }, { "epoch": 0.29, "learning_rate": 7.069656248427198e-05, "loss": 1.0233, "step": 117490 }, { "epoch": 0.29, "learning_rate": 7.069530424279029e-05, "loss": 1.0442, "step": 117495 }, { "epoch": 0.29, "learning_rate": 7.069404600130858e-05, "loss": 1.023, "step": 117500 }, { "epoch": 0.29, "learning_rate": 7.069278775982687e-05, "loss": 1.0225, "step": 117505 }, { "epoch": 0.29, "learning_rate": 7.069152951834516e-05, "loss": 1.0201, "step": 117510 }, { "epoch": 0.29, "learning_rate": 7.069027127686346e-05, "loss": 1.0197, "step": 117515 }, { "epoch": 0.29, "learning_rate": 7.068901303538176e-05, "loss": 1.0214, "step": 117520 }, { "epoch": 0.3, "learning_rate": 7.068775479390005e-05, "loss": 1.0244, "step": 117525 }, { "epoch": 0.3, "learning_rate": 7.068649655241834e-05, "loss": 1.0233, "step": 117530 }, { "epoch": 0.3, "learning_rate": 7.068523831093664e-05, "loss": 1.0228, "step": 117535 }, { "epoch": 0.3, "learning_rate": 7.068398006945494e-05, "loss": 1.0224, "step": 117540 }, { "epoch": 0.3, "learning_rate": 7.068272182797323e-05, "loss": 1.0204, "step": 117545 }, { "epoch": 0.3, "learning_rate": 7.068146358649152e-05, "loss": 1.024, "step": 117550 }, { "epoch": 0.3, "learning_rate": 7.068020534500982e-05, "loss": 1.0223, "step": 117555 }, { "epoch": 0.3, "learning_rate": 7.067894710352812e-05, "loss": 1.0211, "step": 117560 }, { "epoch": 0.3, "learning_rate": 7.067768886204641e-05, "loss": 1.0208, "step": 117565 }, { "epoch": 0.3, "learning_rate": 7.06764306205647e-05, "loss": 1.0201, "step": 117570 }, { "epoch": 0.3, "learning_rate": 7.0675172379083e-05, "loss": 1.0216, "step": 117575 }, { "epoch": 0.3, "learning_rate": 7.067391413760129e-05, "loss": 1.0215, "step": 117580 }, { "epoch": 0.3, "learning_rate": 7.067265589611959e-05, "loss": 1.0212, "step": 117585 }, { "epoch": 0.3, "learning_rate": 7.067139765463788e-05, "loss": 1.0198, "step": 117590 }, { "epoch": 0.3, "learning_rate": 7.067013941315618e-05, "loss": 1.0192, "step": 117595 }, { "epoch": 0.3, "learning_rate": 7.066888117167447e-05, "loss": 1.0202, "step": 117600 }, { "epoch": 0.3, "learning_rate": 7.066762293019277e-05, "loss": 1.0227, "step": 117605 }, { "epoch": 0.3, "learning_rate": 7.066636468871106e-05, "loss": 1.0227, "step": 117610 }, { "epoch": 0.3, "learning_rate": 7.066510644722936e-05, "loss": 1.0229, "step": 117615 }, { "epoch": 0.3, "learning_rate": 7.066384820574765e-05, "loss": 1.023, "step": 117620 }, { "epoch": 0.3, "learning_rate": 7.066258996426595e-05, "loss": 1.0226, "step": 117625 }, { "epoch": 0.3, "learning_rate": 7.066133172278424e-05, "loss": 1.0205, "step": 117630 }, { "epoch": 0.3, "learning_rate": 7.066007348130254e-05, "loss": 1.0219, "step": 117635 }, { "epoch": 0.3, "learning_rate": 7.065881523982083e-05, "loss": 1.0224, "step": 117640 }, { "epoch": 0.3, "learning_rate": 7.065755699833912e-05, "loss": 1.0199, "step": 117645 }, { "epoch": 0.3, "learning_rate": 7.065629875685742e-05, "loss": 1.0242, "step": 117650 }, { "epoch": 0.3, "learning_rate": 7.065504051537572e-05, "loss": 1.0221, "step": 117655 }, { "epoch": 0.3, "learning_rate": 7.065378227389401e-05, "loss": 1.0216, "step": 117660 }, { "epoch": 0.3, "learning_rate": 7.06525240324123e-05, "loss": 1.0222, "step": 117665 }, { "epoch": 0.3, "learning_rate": 7.06512657909306e-05, "loss": 1.0193, "step": 117670 }, { "epoch": 0.3, "learning_rate": 7.06500075494489e-05, "loss": 1.022, "step": 117675 }, { "epoch": 0.3, "learning_rate": 7.064874930796719e-05, "loss": 1.0231, "step": 117680 }, { "epoch": 0.3, "learning_rate": 7.064749106648548e-05, "loss": 1.0226, "step": 117685 }, { "epoch": 0.3, "learning_rate": 7.064623282500378e-05, "loss": 1.023, "step": 117690 }, { "epoch": 0.3, "learning_rate": 7.064497458352208e-05, "loss": 1.0218, "step": 117695 }, { "epoch": 0.3, "learning_rate": 7.064371634204037e-05, "loss": 1.0231, "step": 117700 }, { "epoch": 0.3, "learning_rate": 7.064245810055866e-05, "loss": 1.0202, "step": 117705 }, { "epoch": 0.3, "learning_rate": 7.064119985907695e-05, "loss": 1.0221, "step": 117710 }, { "epoch": 0.3, "learning_rate": 7.063994161759526e-05, "loss": 1.0218, "step": 117715 }, { "epoch": 0.3, "learning_rate": 7.063868337611355e-05, "loss": 1.022, "step": 117720 }, { "epoch": 0.3, "learning_rate": 7.063742513463184e-05, "loss": 1.0229, "step": 117725 }, { "epoch": 0.3, "learning_rate": 7.063616689315013e-05, "loss": 1.0212, "step": 117730 }, { "epoch": 0.3, "learning_rate": 7.063490865166844e-05, "loss": 1.0228, "step": 117735 }, { "epoch": 0.3, "learning_rate": 7.063365041018673e-05, "loss": 1.0419, "step": 117740 }, { "epoch": 0.3, "learning_rate": 7.063239216870502e-05, "loss": 1.0221, "step": 117745 }, { "epoch": 0.3, "learning_rate": 7.063113392722331e-05, "loss": 1.0207, "step": 117750 }, { "epoch": 0.3, "learning_rate": 7.062987568574162e-05, "loss": 1.0242, "step": 117755 }, { "epoch": 0.3, "learning_rate": 7.062861744425991e-05, "loss": 1.0217, "step": 117760 }, { "epoch": 0.3, "learning_rate": 7.06273592027782e-05, "loss": 1.0213, "step": 117765 }, { "epoch": 0.3, "learning_rate": 7.062610096129649e-05, "loss": 1.0224, "step": 117770 }, { "epoch": 0.3, "learning_rate": 7.062484271981478e-05, "loss": 1.0211, "step": 117775 }, { "epoch": 0.3, "learning_rate": 7.062358447833309e-05, "loss": 1.0219, "step": 117780 }, { "epoch": 0.3, "learning_rate": 7.062232623685138e-05, "loss": 1.0229, "step": 117785 }, { "epoch": 0.3, "learning_rate": 7.062106799536967e-05, "loss": 1.0241, "step": 117790 }, { "epoch": 0.3, "learning_rate": 7.061980975388796e-05, "loss": 1.0229, "step": 117795 }, { "epoch": 0.3, "learning_rate": 7.061855151240627e-05, "loss": 1.0219, "step": 117800 }, { "epoch": 0.3, "learning_rate": 7.061729327092456e-05, "loss": 1.0219, "step": 117805 }, { "epoch": 0.3, "learning_rate": 7.061603502944285e-05, "loss": 1.023, "step": 117810 }, { "epoch": 0.3, "learning_rate": 7.061477678796114e-05, "loss": 1.0231, "step": 117815 }, { "epoch": 0.3, "learning_rate": 7.061351854647943e-05, "loss": 1.021, "step": 117820 }, { "epoch": 0.3, "learning_rate": 7.061226030499774e-05, "loss": 1.0222, "step": 117825 }, { "epoch": 0.3, "learning_rate": 7.061100206351603e-05, "loss": 1.0203, "step": 117830 }, { "epoch": 0.3, "learning_rate": 7.060974382203432e-05, "loss": 1.0231, "step": 117835 }, { "epoch": 0.3, "learning_rate": 7.060848558055261e-05, "loss": 1.0206, "step": 117840 }, { "epoch": 0.3, "learning_rate": 7.060722733907092e-05, "loss": 1.02, "step": 117845 }, { "epoch": 0.3, "learning_rate": 7.060596909758921e-05, "loss": 1.0217, "step": 117850 }, { "epoch": 0.3, "learning_rate": 7.06047108561075e-05, "loss": 1.0212, "step": 117855 }, { "epoch": 0.3, "learning_rate": 7.06034526146258e-05, "loss": 1.0239, "step": 117860 }, { "epoch": 0.3, "learning_rate": 7.06021943731441e-05, "loss": 1.0229, "step": 117865 }, { "epoch": 0.3, "learning_rate": 7.060093613166239e-05, "loss": 1.0202, "step": 117870 }, { "epoch": 0.3, "learning_rate": 7.059967789018068e-05, "loss": 1.0206, "step": 117875 }, { "epoch": 0.3, "learning_rate": 7.059841964869897e-05, "loss": 1.0208, "step": 117880 }, { "epoch": 0.3, "learning_rate": 7.059716140721726e-05, "loss": 1.0227, "step": 117885 }, { "epoch": 0.3, "learning_rate": 7.059590316573557e-05, "loss": 1.0206, "step": 117890 }, { "epoch": 0.3, "learning_rate": 7.059464492425386e-05, "loss": 1.0223, "step": 117895 }, { "epoch": 0.3, "learning_rate": 7.059338668277215e-05, "loss": 1.0235, "step": 117900 }, { "epoch": 0.3, "learning_rate": 7.059212844129044e-05, "loss": 1.0229, "step": 117905 }, { "epoch": 0.3, "learning_rate": 7.059087019980875e-05, "loss": 1.0228, "step": 117910 }, { "epoch": 0.3, "learning_rate": 7.058961195832706e-05, "loss": 1.024, "step": 117915 }, { "epoch": 0.3, "learning_rate": 7.058835371684535e-05, "loss": 1.0228, "step": 117920 }, { "epoch": 0.3, "learning_rate": 7.058709547536364e-05, "loss": 1.0216, "step": 117925 }, { "epoch": 0.3, "learning_rate": 7.058583723388193e-05, "loss": 1.0201, "step": 117930 }, { "epoch": 0.3, "learning_rate": 7.058457899240024e-05, "loss": 1.0218, "step": 117935 }, { "epoch": 0.3, "learning_rate": 7.058332075091853e-05, "loss": 1.0225, "step": 117940 }, { "epoch": 0.3, "learning_rate": 7.058206250943682e-05, "loss": 1.0227, "step": 117945 }, { "epoch": 0.3, "learning_rate": 7.058080426795511e-05, "loss": 1.0211, "step": 117950 }, { "epoch": 0.3, "learning_rate": 7.057954602647341e-05, "loss": 1.0221, "step": 117955 }, { "epoch": 0.3, "learning_rate": 7.05782877849917e-05, "loss": 1.0227, "step": 117960 }, { "epoch": 0.3, "learning_rate": 7.057702954351e-05, "loss": 1.0209, "step": 117965 }, { "epoch": 0.3, "learning_rate": 7.057577130202829e-05, "loss": 1.0222, "step": 117970 }, { "epoch": 0.3, "learning_rate": 7.057451306054658e-05, "loss": 1.0221, "step": 117975 }, { "epoch": 0.3, "learning_rate": 7.057325481906489e-05, "loss": 1.0229, "step": 117980 }, { "epoch": 0.3, "learning_rate": 7.057199657758318e-05, "loss": 1.0242, "step": 117985 }, { "epoch": 0.3, "learning_rate": 7.057073833610147e-05, "loss": 1.0218, "step": 117990 }, { "epoch": 0.3, "learning_rate": 7.056948009461976e-05, "loss": 1.0202, "step": 117995 }, { "epoch": 0.3, "learning_rate": 7.05684735014344e-05, "loss": 1.0232, "step": 118000 }, { "epoch": 0.3, "learning_rate": 7.056721525995269e-05, "loss": 1.0232, "step": 118005 }, { "epoch": 0.3, "learning_rate": 7.056595701847098e-05, "loss": 1.0387, "step": 118010 }, { "epoch": 0.3, "learning_rate": 7.056469877698929e-05, "loss": 1.0229, "step": 118015 }, { "epoch": 0.3, "learning_rate": 7.056344053550758e-05, "loss": 1.0248, "step": 118020 }, { "epoch": 0.3, "learning_rate": 7.056218229402587e-05, "loss": 1.0207, "step": 118025 }, { "epoch": 0.3, "learning_rate": 7.056092405254416e-05, "loss": 1.0203, "step": 118030 }, { "epoch": 0.3, "learning_rate": 7.055966581106245e-05, "loss": 1.0229, "step": 118035 }, { "epoch": 0.3, "learning_rate": 7.055840756958076e-05, "loss": 1.0228, "step": 118040 }, { "epoch": 0.3, "learning_rate": 7.055714932809905e-05, "loss": 1.0214, "step": 118045 }, { "epoch": 0.3, "learning_rate": 7.055589108661734e-05, "loss": 1.0221, "step": 118050 }, { "epoch": 0.3, "learning_rate": 7.055463284513563e-05, "loss": 1.0199, "step": 118055 }, { "epoch": 0.3, "learning_rate": 7.055337460365394e-05, "loss": 1.0219, "step": 118060 }, { "epoch": 0.3, "learning_rate": 7.055211636217223e-05, "loss": 1.0194, "step": 118065 }, { "epoch": 0.3, "learning_rate": 7.055085812069052e-05, "loss": 1.022, "step": 118070 }, { "epoch": 0.3, "learning_rate": 7.054959987920881e-05, "loss": 1.0215, "step": 118075 }, { "epoch": 0.3, "learning_rate": 7.054834163772712e-05, "loss": 1.0208, "step": 118080 }, { "epoch": 0.3, "learning_rate": 7.054708339624541e-05, "loss": 1.02, "step": 118085 }, { "epoch": 0.3, "learning_rate": 7.05458251547637e-05, "loss": 1.0216, "step": 118090 }, { "epoch": 0.3, "learning_rate": 7.054456691328199e-05, "loss": 1.0219, "step": 118095 }, { "epoch": 0.3, "learning_rate": 7.054330867180028e-05, "loss": 1.0195, "step": 118100 }, { "epoch": 0.3, "learning_rate": 7.054205043031859e-05, "loss": 1.0213, "step": 118105 }, { "epoch": 0.3, "learning_rate": 7.054079218883688e-05, "loss": 1.022, "step": 118110 }, { "epoch": 0.3, "learning_rate": 7.053953394735519e-05, "loss": 1.0199, "step": 118115 }, { "epoch": 0.3, "learning_rate": 7.053827570587348e-05, "loss": 1.0228, "step": 118120 }, { "epoch": 0.3, "learning_rate": 7.053701746439177e-05, "loss": 1.0224, "step": 118125 }, { "epoch": 0.3, "learning_rate": 7.053575922291007e-05, "loss": 1.022, "step": 118130 }, { "epoch": 0.3, "learning_rate": 7.053450098142837e-05, "loss": 1.0226, "step": 118135 }, { "epoch": 0.3, "learning_rate": 7.053324273994666e-05, "loss": 1.0231, "step": 118140 }, { "epoch": 0.3, "learning_rate": 7.053198449846495e-05, "loss": 1.0228, "step": 118145 }, { "epoch": 0.3, "learning_rate": 7.053072625698325e-05, "loss": 1.0232, "step": 118150 }, { "epoch": 0.3, "learning_rate": 7.052946801550155e-05, "loss": 1.0229, "step": 118155 }, { "epoch": 0.3, "learning_rate": 7.052820977401984e-05, "loss": 1.0214, "step": 118160 }, { "epoch": 0.3, "learning_rate": 7.052695153253813e-05, "loss": 1.0218, "step": 118165 }, { "epoch": 0.3, "learning_rate": 7.052569329105643e-05, "loss": 1.0223, "step": 118170 }, { "epoch": 0.3, "learning_rate": 7.052443504957473e-05, "loss": 1.0232, "step": 118175 }, { "epoch": 0.3, "learning_rate": 7.052317680809302e-05, "loss": 1.0209, "step": 118180 }, { "epoch": 0.3, "learning_rate": 7.052191856661131e-05, "loss": 1.0219, "step": 118185 }, { "epoch": 0.3, "learning_rate": 7.05206603251296e-05, "loss": 1.0193, "step": 118190 }, { "epoch": 0.3, "learning_rate": 7.05194020836479e-05, "loss": 1.0243, "step": 118195 }, { "epoch": 0.3, "learning_rate": 7.05181438421662e-05, "loss": 1.024, "step": 118200 }, { "epoch": 0.3, "learning_rate": 7.051688560068449e-05, "loss": 1.0227, "step": 118205 }, { "epoch": 0.3, "learning_rate": 7.051562735920278e-05, "loss": 1.0227, "step": 118210 }, { "epoch": 0.3, "learning_rate": 7.051436911772109e-05, "loss": 1.022, "step": 118215 }, { "epoch": 0.3, "learning_rate": 7.051311087623938e-05, "loss": 1.0211, "step": 118220 }, { "epoch": 0.3, "learning_rate": 7.051185263475767e-05, "loss": 1.0239, "step": 118225 }, { "epoch": 0.3, "learning_rate": 7.051059439327596e-05, "loss": 1.0214, "step": 118230 }, { "epoch": 0.3, "learning_rate": 7.050933615179427e-05, "loss": 1.0224, "step": 118235 }, { "epoch": 0.3, "learning_rate": 7.050807791031256e-05, "loss": 1.0211, "step": 118240 }, { "epoch": 0.3, "learning_rate": 7.050681966883085e-05, "loss": 1.0215, "step": 118245 }, { "epoch": 0.3, "learning_rate": 7.050556142734914e-05, "loss": 1.0227, "step": 118250 }, { "epoch": 0.3, "learning_rate": 7.050430318586743e-05, "loss": 1.022, "step": 118255 }, { "epoch": 0.3, "learning_rate": 7.050304494438574e-05, "loss": 1.0218, "step": 118260 }, { "epoch": 0.3, "learning_rate": 7.050178670290403e-05, "loss": 1.0203, "step": 118265 }, { "epoch": 0.3, "learning_rate": 7.050052846142232e-05, "loss": 1.0225, "step": 118270 }, { "epoch": 0.3, "learning_rate": 7.049927021994061e-05, "loss": 1.0257, "step": 118275 }, { "epoch": 0.3, "learning_rate": 7.049801197845892e-05, "loss": 1.0222, "step": 118280 }, { "epoch": 0.3, "learning_rate": 7.049675373697721e-05, "loss": 1.0216, "step": 118285 }, { "epoch": 0.3, "learning_rate": 7.04954954954955e-05, "loss": 1.021, "step": 118290 }, { "epoch": 0.3, "learning_rate": 7.049423725401379e-05, "loss": 1.023, "step": 118295 }, { "epoch": 0.3, "learning_rate": 7.04929790125321e-05, "loss": 1.0209, "step": 118300 }, { "epoch": 0.3, "learning_rate": 7.049172077105039e-05, "loss": 1.0216, "step": 118305 }, { "epoch": 0.3, "learning_rate": 7.049046252956868e-05, "loss": 1.021, "step": 118310 }, { "epoch": 0.3, "learning_rate": 7.048920428808697e-05, "loss": 1.0239, "step": 118315 }, { "epoch": 0.3, "learning_rate": 7.048794604660526e-05, "loss": 1.024, "step": 118320 }, { "epoch": 0.3, "learning_rate": 7.048668780512357e-05, "loss": 1.0224, "step": 118325 }, { "epoch": 0.3, "learning_rate": 7.048542956364186e-05, "loss": 1.0203, "step": 118330 }, { "epoch": 0.3, "learning_rate": 7.048417132216015e-05, "loss": 1.0206, "step": 118335 }, { "epoch": 0.3, "learning_rate": 7.048291308067844e-05, "loss": 1.0239, "step": 118340 }, { "epoch": 0.3, "learning_rate": 7.048165483919675e-05, "loss": 1.0228, "step": 118345 }, { "epoch": 0.3, "learning_rate": 7.048039659771504e-05, "loss": 1.0226, "step": 118350 }, { "epoch": 0.3, "learning_rate": 7.047913835623333e-05, "loss": 1.0246, "step": 118355 }, { "epoch": 0.3, "learning_rate": 7.047788011475162e-05, "loss": 1.0298, "step": 118360 }, { "epoch": 0.3, "learning_rate": 7.047662187326993e-05, "loss": 1.0247, "step": 118365 }, { "epoch": 0.3, "learning_rate": 7.047536363178822e-05, "loss": 1.0214, "step": 118370 }, { "epoch": 0.3, "learning_rate": 7.047410539030651e-05, "loss": 1.0232, "step": 118375 }, { "epoch": 0.3, "learning_rate": 7.04728471488248e-05, "loss": 1.0234, "step": 118380 }, { "epoch": 0.3, "learning_rate": 7.04715889073431e-05, "loss": 1.0235, "step": 118385 }, { "epoch": 0.3, "learning_rate": 7.04703306658614e-05, "loss": 1.0243, "step": 118390 }, { "epoch": 0.3, "learning_rate": 7.046907242437969e-05, "loss": 1.0202, "step": 118395 }, { "epoch": 0.3, "learning_rate": 7.046781418289798e-05, "loss": 1.0206, "step": 118400 }, { "epoch": 0.3, "learning_rate": 7.046655594141627e-05, "loss": 1.0207, "step": 118405 }, { "epoch": 0.3, "learning_rate": 7.046529769993458e-05, "loss": 1.022, "step": 118410 }, { "epoch": 0.3, "learning_rate": 7.046403945845287e-05, "loss": 1.0234, "step": 118415 }, { "epoch": 0.3, "learning_rate": 7.046278121697116e-05, "loss": 1.0214, "step": 118420 }, { "epoch": 0.3, "learning_rate": 7.046152297548945e-05, "loss": 1.0224, "step": 118425 }, { "epoch": 0.3, "learning_rate": 7.046026473400776e-05, "loss": 1.024, "step": 118430 }, { "epoch": 0.3, "learning_rate": 7.045900649252605e-05, "loss": 1.0218, "step": 118435 }, { "epoch": 0.3, "learning_rate": 7.045774825104434e-05, "loss": 1.0232, "step": 118440 }, { "epoch": 0.3, "learning_rate": 7.045649000956263e-05, "loss": 1.024, "step": 118445 }, { "epoch": 0.3, "learning_rate": 7.045523176808093e-05, "loss": 1.0295, "step": 118450 }, { "epoch": 0.3, "learning_rate": 7.045397352659923e-05, "loss": 1.0223, "step": 118455 }, { "epoch": 0.3, "learning_rate": 7.045271528511752e-05, "loss": 1.0216, "step": 118460 }, { "epoch": 0.3, "learning_rate": 7.045145704363581e-05, "loss": 1.0211, "step": 118465 }, { "epoch": 0.3, "learning_rate": 7.04501988021541e-05, "loss": 1.022, "step": 118470 }, { "epoch": 0.3, "learning_rate": 7.044894056067241e-05, "loss": 1.0231, "step": 118475 }, { "epoch": 0.3, "learning_rate": 7.04476823191907e-05, "loss": 1.0211, "step": 118480 }, { "epoch": 0.3, "learning_rate": 7.0446424077709e-05, "loss": 1.0211, "step": 118485 }, { "epoch": 0.3, "learning_rate": 7.044516583622729e-05, "loss": 1.0207, "step": 118490 }, { "epoch": 0.3, "learning_rate": 7.044390759474559e-05, "loss": 1.0423, "step": 118495 }, { "epoch": 0.3, "learning_rate": 7.044264935326388e-05, "loss": 1.0238, "step": 118500 }, { "epoch": 0.3, "learning_rate": 7.044139111178217e-05, "loss": 1.0217, "step": 118505 }, { "epoch": 0.3, "learning_rate": 7.044013287030047e-05, "loss": 1.0225, "step": 118510 }, { "epoch": 0.3, "learning_rate": 7.043887462881876e-05, "loss": 1.0236, "step": 118515 }, { "epoch": 0.3, "learning_rate": 7.043761638733706e-05, "loss": 1.0213, "step": 118520 }, { "epoch": 0.3, "learning_rate": 7.043635814585535e-05, "loss": 1.0212, "step": 118525 }, { "epoch": 0.3, "learning_rate": 7.043509990437365e-05, "loss": 1.0251, "step": 118530 }, { "epoch": 0.3, "learning_rate": 7.043384166289194e-05, "loss": 1.0229, "step": 118535 }, { "epoch": 0.3, "learning_rate": 7.043258342141024e-05, "loss": 1.0239, "step": 118540 }, { "epoch": 0.3, "learning_rate": 7.043132517992853e-05, "loss": 1.0228, "step": 118545 }, { "epoch": 0.3, "learning_rate": 7.043006693844683e-05, "loss": 1.0234, "step": 118550 }, { "epoch": 0.3, "learning_rate": 7.042880869696512e-05, "loss": 1.023, "step": 118555 }, { "epoch": 0.3, "learning_rate": 7.042755045548341e-05, "loss": 1.022, "step": 118560 }, { "epoch": 0.3, "learning_rate": 7.042629221400171e-05, "loss": 1.0228, "step": 118565 }, { "epoch": 0.3, "learning_rate": 7.042503397252e-05, "loss": 1.0233, "step": 118570 }, { "epoch": 0.3, "learning_rate": 7.04237757310383e-05, "loss": 1.0216, "step": 118575 }, { "epoch": 0.3, "learning_rate": 7.042251748955659e-05, "loss": 1.0222, "step": 118580 }, { "epoch": 0.3, "learning_rate": 7.04212592480749e-05, "loss": 1.0214, "step": 118585 }, { "epoch": 0.3, "learning_rate": 7.042000100659319e-05, "loss": 1.0215, "step": 118590 }, { "epoch": 0.3, "learning_rate": 7.041874276511148e-05, "loss": 1.0217, "step": 118595 }, { "epoch": 0.3, "learning_rate": 7.041748452362977e-05, "loss": 1.0205, "step": 118600 }, { "epoch": 0.3, "learning_rate": 7.041622628214807e-05, "loss": 1.0189, "step": 118605 }, { "epoch": 0.3, "learning_rate": 7.041496804066637e-05, "loss": 1.023, "step": 118610 }, { "epoch": 0.3, "learning_rate": 7.041370979918467e-05, "loss": 1.0223, "step": 118615 }, { "epoch": 0.3, "learning_rate": 7.041245155770296e-05, "loss": 1.0225, "step": 118620 }, { "epoch": 0.3, "learning_rate": 7.041119331622125e-05, "loss": 1.0211, "step": 118625 }, { "epoch": 0.3, "learning_rate": 7.040993507473956e-05, "loss": 1.021, "step": 118630 }, { "epoch": 0.3, "learning_rate": 7.040867683325785e-05, "loss": 1.0208, "step": 118635 }, { "epoch": 0.3, "learning_rate": 7.040741859177614e-05, "loss": 1.0222, "step": 118640 }, { "epoch": 0.3, "learning_rate": 7.040616035029443e-05, "loss": 1.0218, "step": 118645 }, { "epoch": 0.3, "learning_rate": 7.040490210881273e-05, "loss": 1.0227, "step": 118650 }, { "epoch": 0.3, "learning_rate": 7.040364386733103e-05, "loss": 1.0231, "step": 118655 }, { "epoch": 0.3, "learning_rate": 7.040238562584932e-05, "loss": 1.0232, "step": 118660 }, { "epoch": 0.3, "learning_rate": 7.040112738436761e-05, "loss": 1.0185, "step": 118665 }, { "epoch": 0.3, "learning_rate": 7.03998691428859e-05, "loss": 1.0234, "step": 118670 }, { "epoch": 0.3, "learning_rate": 7.039861090140421e-05, "loss": 1.0235, "step": 118675 }, { "epoch": 0.3, "learning_rate": 7.03973526599225e-05, "loss": 1.0226, "step": 118680 }, { "epoch": 0.3, "learning_rate": 7.03960944184408e-05, "loss": 1.0204, "step": 118685 }, { "epoch": 0.3, "learning_rate": 7.039483617695908e-05, "loss": 1.0213, "step": 118690 }, { "epoch": 0.3, "learning_rate": 7.039357793547739e-05, "loss": 1.0226, "step": 118695 }, { "epoch": 0.3, "learning_rate": 7.039231969399568e-05, "loss": 1.0205, "step": 118700 }, { "epoch": 0.3, "learning_rate": 7.039106145251397e-05, "loss": 1.0195, "step": 118705 }, { "epoch": 0.3, "learning_rate": 7.038980321103226e-05, "loss": 1.0206, "step": 118710 }, { "epoch": 0.3, "learning_rate": 7.038854496955056e-05, "loss": 1.0219, "step": 118715 }, { "epoch": 0.3, "learning_rate": 7.038728672806886e-05, "loss": 1.0198, "step": 118720 }, { "epoch": 0.3, "learning_rate": 7.038602848658715e-05, "loss": 1.0216, "step": 118725 }, { "epoch": 0.3, "learning_rate": 7.038477024510544e-05, "loss": 1.0198, "step": 118730 }, { "epoch": 0.3, "learning_rate": 7.038351200362374e-05, "loss": 1.0227, "step": 118735 }, { "epoch": 0.3, "learning_rate": 7.038225376214204e-05, "loss": 1.0227, "step": 118740 }, { "epoch": 0.3, "learning_rate": 7.038099552066033e-05, "loss": 1.0212, "step": 118745 }, { "epoch": 0.3, "learning_rate": 7.037973727917862e-05, "loss": 1.0206, "step": 118750 }, { "epoch": 0.3, "learning_rate": 7.037847903769692e-05, "loss": 1.0235, "step": 118755 }, { "epoch": 0.3, "learning_rate": 7.037722079621522e-05, "loss": 1.0215, "step": 118760 }, { "epoch": 0.3, "learning_rate": 7.037596255473351e-05, "loss": 1.0208, "step": 118765 }, { "epoch": 0.3, "learning_rate": 7.03747043132518e-05, "loss": 1.0237, "step": 118770 }, { "epoch": 0.3, "learning_rate": 7.03734460717701e-05, "loss": 1.0216, "step": 118775 }, { "epoch": 0.3, "learning_rate": 7.037218783028839e-05, "loss": 1.0239, "step": 118780 }, { "epoch": 0.3, "learning_rate": 7.037092958880669e-05, "loss": 1.0207, "step": 118785 }, { "epoch": 0.3, "learning_rate": 7.036967134732498e-05, "loss": 1.0217, "step": 118790 }, { "epoch": 0.3, "learning_rate": 7.036841310584328e-05, "loss": 1.0201, "step": 118795 }, { "epoch": 0.3, "learning_rate": 7.036715486436157e-05, "loss": 1.0384, "step": 118800 }, { "epoch": 0.3, "learning_rate": 7.036589662287987e-05, "loss": 1.0234, "step": 118805 }, { "epoch": 0.3, "learning_rate": 7.036463838139816e-05, "loss": 1.0248, "step": 118810 }, { "epoch": 0.3, "learning_rate": 7.036338013991646e-05, "loss": 1.0227, "step": 118815 }, { "epoch": 0.3, "learning_rate": 7.036212189843475e-05, "loss": 1.0245, "step": 118820 }, { "epoch": 0.3, "learning_rate": 7.036086365695305e-05, "loss": 1.021, "step": 118825 }, { "epoch": 0.3, "learning_rate": 7.035960541547134e-05, "loss": 1.0234, "step": 118830 }, { "epoch": 0.3, "learning_rate": 7.035834717398964e-05, "loss": 1.0213, "step": 118835 }, { "epoch": 0.3, "learning_rate": 7.035708893250793e-05, "loss": 1.0222, "step": 118840 }, { "epoch": 0.3, "learning_rate": 7.035583069102622e-05, "loss": 1.022, "step": 118845 }, { "epoch": 0.3, "learning_rate": 7.035457244954452e-05, "loss": 1.0224, "step": 118850 }, { "epoch": 0.3, "learning_rate": 7.035331420806282e-05, "loss": 1.0203, "step": 118855 }, { "epoch": 0.3, "learning_rate": 7.035205596658111e-05, "loss": 1.0219, "step": 118860 }, { "epoch": 0.3, "learning_rate": 7.03507977250994e-05, "loss": 1.0181, "step": 118865 }, { "epoch": 0.3, "learning_rate": 7.03495394836177e-05, "loss": 1.0243, "step": 118870 }, { "epoch": 0.3, "learning_rate": 7.0348281242136e-05, "loss": 1.0207, "step": 118875 }, { "epoch": 0.3, "learning_rate": 7.034702300065429e-05, "loss": 1.0224, "step": 118880 }, { "epoch": 0.3, "learning_rate": 7.034576475917258e-05, "loss": 1.0214, "step": 118885 }, { "epoch": 0.3, "learning_rate": 7.034450651769088e-05, "loss": 1.0213, "step": 118890 }, { "epoch": 0.3, "learning_rate": 7.034324827620918e-05, "loss": 1.0228, "step": 118895 }, { "epoch": 0.3, "learning_rate": 7.034199003472747e-05, "loss": 1.0225, "step": 118900 }, { "epoch": 0.3, "learning_rate": 7.034073179324576e-05, "loss": 1.0221, "step": 118905 }, { "epoch": 0.3, "learning_rate": 7.033947355176405e-05, "loss": 1.0203, "step": 118910 }, { "epoch": 0.3, "learning_rate": 7.033821531028236e-05, "loss": 1.0247, "step": 118915 }, { "epoch": 0.3, "learning_rate": 7.033695706880065e-05, "loss": 1.022, "step": 118920 }, { "epoch": 0.3, "learning_rate": 7.033569882731894e-05, "loss": 1.0209, "step": 118925 }, { "epoch": 0.3, "learning_rate": 7.033444058583723e-05, "loss": 1.0202, "step": 118930 }, { "epoch": 0.3, "learning_rate": 7.033318234435554e-05, "loss": 1.0249, "step": 118935 }, { "epoch": 0.3, "learning_rate": 7.033192410287383e-05, "loss": 1.0208, "step": 118940 }, { "epoch": 0.3, "learning_rate": 7.033066586139212e-05, "loss": 1.0221, "step": 118945 }, { "epoch": 0.3, "learning_rate": 7.032940761991041e-05, "loss": 1.0216, "step": 118950 }, { "epoch": 0.3, "learning_rate": 7.032814937842872e-05, "loss": 1.024, "step": 118955 }, { "epoch": 0.3, "learning_rate": 7.032689113694701e-05, "loss": 1.0245, "step": 118960 }, { "epoch": 0.3, "learning_rate": 7.03256328954653e-05, "loss": 1.0211, "step": 118965 }, { "epoch": 0.3, "learning_rate": 7.032437465398359e-05, "loss": 1.0228, "step": 118970 }, { "epoch": 0.3, "learning_rate": 7.032311641250188e-05, "loss": 1.0202, "step": 118975 }, { "epoch": 0.3, "learning_rate": 7.032185817102019e-05, "loss": 1.0238, "step": 118980 }, { "epoch": 0.3, "learning_rate": 7.032059992953848e-05, "loss": 1.0222, "step": 118985 }, { "epoch": 0.3, "learning_rate": 7.031934168805677e-05, "loss": 1.0243, "step": 118990 }, { "epoch": 0.3, "learning_rate": 7.031808344657506e-05, "loss": 1.0256, "step": 118995 }, { "epoch": 0.3, "learning_rate": 7.031682520509337e-05, "loss": 1.0216, "step": 119000 }, { "epoch": 0.3, "learning_rate": 7.031556696361166e-05, "loss": 1.0217, "step": 119005 }, { "epoch": 0.3, "learning_rate": 7.031430872212995e-05, "loss": 1.0207, "step": 119010 }, { "epoch": 0.3, "learning_rate": 7.031305048064824e-05, "loss": 1.0239, "step": 119015 }, { "epoch": 0.3, "learning_rate": 7.031179223916655e-05, "loss": 1.0217, "step": 119020 }, { "epoch": 0.3, "learning_rate": 7.031053399768484e-05, "loss": 1.0201, "step": 119025 }, { "epoch": 0.3, "learning_rate": 7.030927575620313e-05, "loss": 1.0235, "step": 119030 }, { "epoch": 0.3, "learning_rate": 7.030801751472142e-05, "loss": 1.0228, "step": 119035 }, { "epoch": 0.3, "learning_rate": 7.030675927323971e-05, "loss": 1.0219, "step": 119040 }, { "epoch": 0.3, "learning_rate": 7.030550103175802e-05, "loss": 1.0245, "step": 119045 }, { "epoch": 0.3, "learning_rate": 7.030424279027631e-05, "loss": 1.0205, "step": 119050 }, { "epoch": 0.3, "learning_rate": 7.03029845487946e-05, "loss": 1.0433, "step": 119055 }, { "epoch": 0.3, "learning_rate": 7.03017263073129e-05, "loss": 1.0203, "step": 119060 }, { "epoch": 0.3, "learning_rate": 7.03004680658312e-05, "loss": 1.0246, "step": 119065 }, { "epoch": 0.3, "learning_rate": 7.029920982434949e-05, "loss": 1.0219, "step": 119070 }, { "epoch": 0.3, "learning_rate": 7.029795158286778e-05, "loss": 1.0209, "step": 119075 }, { "epoch": 0.3, "learning_rate": 7.029669334138607e-05, "loss": 1.0238, "step": 119080 }, { "epoch": 0.3, "learning_rate": 7.029543509990438e-05, "loss": 1.0237, "step": 119085 }, { "epoch": 0.3, "learning_rate": 7.029417685842267e-05, "loss": 1.0233, "step": 119090 }, { "epoch": 0.3, "learning_rate": 7.029291861694096e-05, "loss": 1.0218, "step": 119095 }, { "epoch": 0.3, "learning_rate": 7.029166037545925e-05, "loss": 1.0222, "step": 119100 }, { "epoch": 0.3, "learning_rate": 7.029040213397755e-05, "loss": 1.022, "step": 119105 }, { "epoch": 0.3, "learning_rate": 7.028914389249585e-05, "loss": 1.0221, "step": 119110 }, { "epoch": 0.3, "learning_rate": 7.028788565101416e-05, "loss": 1.0207, "step": 119115 }, { "epoch": 0.3, "learning_rate": 7.028662740953245e-05, "loss": 1.0245, "step": 119120 }, { "epoch": 0.3, "learning_rate": 7.028536916805074e-05, "loss": 1.0241, "step": 119125 }, { "epoch": 0.3, "learning_rate": 7.028411092656903e-05, "loss": 1.0225, "step": 119130 }, { "epoch": 0.3, "learning_rate": 7.028285268508734e-05, "loss": 1.0186, "step": 119135 }, { "epoch": 0.3, "learning_rate": 7.028159444360563e-05, "loss": 1.0208, "step": 119140 }, { "epoch": 0.3, "learning_rate": 7.028033620212392e-05, "loss": 1.0201, "step": 119145 }, { "epoch": 0.3, "learning_rate": 7.027907796064221e-05, "loss": 1.0215, "step": 119150 }, { "epoch": 0.3, "learning_rate": 7.027781971916052e-05, "loss": 1.0219, "step": 119155 }, { "epoch": 0.3, "learning_rate": 7.027656147767881e-05, "loss": 1.0221, "step": 119160 }, { "epoch": 0.3, "learning_rate": 7.02753032361971e-05, "loss": 1.0235, "step": 119165 }, { "epoch": 0.3, "learning_rate": 7.027404499471539e-05, "loss": 1.024, "step": 119170 }, { "epoch": 0.3, "learning_rate": 7.027278675323368e-05, "loss": 1.0218, "step": 119175 }, { "epoch": 0.3, "learning_rate": 7.027152851175199e-05, "loss": 1.0217, "step": 119180 }, { "epoch": 0.3, "learning_rate": 7.027027027027028e-05, "loss": 1.0446, "step": 119185 }, { "epoch": 0.3, "learning_rate": 7.026901202878857e-05, "loss": 1.02, "step": 119190 }, { "epoch": 0.3, "learning_rate": 7.026775378730686e-05, "loss": 1.0242, "step": 119195 }, { "epoch": 0.3, "learning_rate": 7.026649554582517e-05, "loss": 1.022, "step": 119200 }, { "epoch": 0.3, "learning_rate": 7.026523730434346e-05, "loss": 1.0216, "step": 119205 }, { "epoch": 0.3, "learning_rate": 7.026397906286175e-05, "loss": 1.0192, "step": 119210 }, { "epoch": 0.3, "learning_rate": 7.026272082138004e-05, "loss": 1.0241, "step": 119215 }, { "epoch": 0.3, "learning_rate": 7.026146257989835e-05, "loss": 1.0211, "step": 119220 }, { "epoch": 0.3, "learning_rate": 7.026020433841664e-05, "loss": 1.0205, "step": 119225 }, { "epoch": 0.3, "learning_rate": 7.025894609693493e-05, "loss": 1.0221, "step": 119230 }, { "epoch": 0.3, "learning_rate": 7.025768785545322e-05, "loss": 1.0204, "step": 119235 }, { "epoch": 0.3, "learning_rate": 7.025642961397151e-05, "loss": 1.0199, "step": 119240 }, { "epoch": 0.3, "learning_rate": 7.025517137248982e-05, "loss": 1.0214, "step": 119245 }, { "epoch": 0.3, "learning_rate": 7.025391313100811e-05, "loss": 1.0234, "step": 119250 }, { "epoch": 0.3, "learning_rate": 7.02526548895264e-05, "loss": 1.0207, "step": 119255 }, { "epoch": 0.3, "learning_rate": 7.025139664804469e-05, "loss": 1.0219, "step": 119260 }, { "epoch": 0.3, "learning_rate": 7.0250138406563e-05, "loss": 1.0226, "step": 119265 }, { "epoch": 0.3, "learning_rate": 7.024888016508129e-05, "loss": 1.0219, "step": 119270 }, { "epoch": 0.3, "learning_rate": 7.024762192359958e-05, "loss": 1.0214, "step": 119275 }, { "epoch": 0.3, "learning_rate": 7.024636368211787e-05, "loss": 1.0222, "step": 119280 }, { "epoch": 0.3, "learning_rate": 7.024510544063618e-05, "loss": 1.0238, "step": 119285 }, { "epoch": 0.3, "learning_rate": 7.024384719915447e-05, "loss": 1.0221, "step": 119290 }, { "epoch": 0.3, "learning_rate": 7.024258895767276e-05, "loss": 1.023, "step": 119295 }, { "epoch": 0.3, "learning_rate": 7.024133071619105e-05, "loss": 1.0223, "step": 119300 }, { "epoch": 0.3, "learning_rate": 7.024007247470934e-05, "loss": 1.0221, "step": 119305 }, { "epoch": 0.3, "learning_rate": 7.023881423322765e-05, "loss": 1.0194, "step": 119310 }, { "epoch": 0.3, "learning_rate": 7.023755599174594e-05, "loss": 1.0246, "step": 119315 }, { "epoch": 0.3, "learning_rate": 7.023629775026423e-05, "loss": 1.0225, "step": 119320 }, { "epoch": 0.3, "learning_rate": 7.023503950878252e-05, "loss": 1.0227, "step": 119325 }, { "epoch": 0.3, "learning_rate": 7.023378126730083e-05, "loss": 1.022, "step": 119330 }, { "epoch": 0.3, "learning_rate": 7.023252302581912e-05, "loss": 1.0217, "step": 119335 }, { "epoch": 0.3, "learning_rate": 7.023126478433741e-05, "loss": 1.0191, "step": 119340 }, { "epoch": 0.3, "learning_rate": 7.02300065428557e-05, "loss": 1.0216, "step": 119345 }, { "epoch": 0.3, "learning_rate": 7.022874830137401e-05, "loss": 1.0215, "step": 119350 }, { "epoch": 0.3, "learning_rate": 7.02274900598923e-05, "loss": 1.0225, "step": 119355 }, { "epoch": 0.3, "learning_rate": 7.022623181841059e-05, "loss": 1.0194, "step": 119360 }, { "epoch": 0.3, "learning_rate": 7.022497357692888e-05, "loss": 1.0215, "step": 119365 }, { "epoch": 0.3, "learning_rate": 7.022371533544718e-05, "loss": 1.0207, "step": 119370 }, { "epoch": 0.3, "learning_rate": 7.022245709396548e-05, "loss": 1.0216, "step": 119375 }, { "epoch": 0.3, "learning_rate": 7.022119885248377e-05, "loss": 1.0212, "step": 119380 }, { "epoch": 0.3, "learning_rate": 7.021994061100206e-05, "loss": 1.0195, "step": 119385 }, { "epoch": 0.3, "learning_rate": 7.021868236952036e-05, "loss": 1.0244, "step": 119390 }, { "epoch": 0.3, "learning_rate": 7.021742412803866e-05, "loss": 1.0214, "step": 119395 }, { "epoch": 0.3, "learning_rate": 7.021616588655695e-05, "loss": 1.0208, "step": 119400 }, { "epoch": 0.3, "learning_rate": 7.021490764507524e-05, "loss": 1.0189, "step": 119405 }, { "epoch": 0.3, "learning_rate": 7.021364940359354e-05, "loss": 1.0235, "step": 119410 }, { "epoch": 0.3, "learning_rate": 7.021239116211184e-05, "loss": 1.0216, "step": 119415 }, { "epoch": 0.3, "learning_rate": 7.021113292063013e-05, "loss": 1.0223, "step": 119420 }, { "epoch": 0.3, "learning_rate": 7.020987467914842e-05, "loss": 1.0243, "step": 119425 }, { "epoch": 0.3, "learning_rate": 7.020861643766672e-05, "loss": 1.0224, "step": 119430 }, { "epoch": 0.3, "learning_rate": 7.020735819618501e-05, "loss": 1.023, "step": 119435 }, { "epoch": 0.3, "learning_rate": 7.020609995470331e-05, "loss": 1.02, "step": 119440 }, { "epoch": 0.3, "learning_rate": 7.02048417132216e-05, "loss": 1.022, "step": 119445 }, { "epoch": 0.3, "learning_rate": 7.02035834717399e-05, "loss": 1.0223, "step": 119450 }, { "epoch": 0.3, "learning_rate": 7.020232523025819e-05, "loss": 1.0236, "step": 119455 }, { "epoch": 0.3, "learning_rate": 7.020106698877649e-05, "loss": 1.0223, "step": 119460 }, { "epoch": 0.3, "learning_rate": 7.019980874729478e-05, "loss": 1.0215, "step": 119465 }, { "epoch": 0.3, "learning_rate": 7.019855050581308e-05, "loss": 1.0204, "step": 119470 }, { "epoch": 0.3, "learning_rate": 7.019729226433137e-05, "loss": 1.0214, "step": 119475 }, { "epoch": 0.3, "learning_rate": 7.019603402284967e-05, "loss": 1.0203, "step": 119480 }, { "epoch": 0.3, "learning_rate": 7.019477578136796e-05, "loss": 1.0194, "step": 119485 }, { "epoch": 0.3, "learning_rate": 7.019351753988626e-05, "loss": 1.0214, "step": 119490 }, { "epoch": 0.3, "learning_rate": 7.019225929840455e-05, "loss": 1.0247, "step": 119495 }, { "epoch": 0.3, "learning_rate": 7.019100105692284e-05, "loss": 1.0224, "step": 119500 }, { "epoch": 0.3, "learning_rate": 7.018974281544114e-05, "loss": 1.0228, "step": 119505 }, { "epoch": 0.3, "learning_rate": 7.018848457395944e-05, "loss": 1.0205, "step": 119510 }, { "epoch": 0.3, "learning_rate": 7.018722633247773e-05, "loss": 1.0215, "step": 119515 }, { "epoch": 0.3, "learning_rate": 7.018596809099602e-05, "loss": 1.0231, "step": 119520 }, { "epoch": 0.3, "learning_rate": 7.018470984951432e-05, "loss": 1.0227, "step": 119525 }, { "epoch": 0.3, "learning_rate": 7.018345160803262e-05, "loss": 1.0229, "step": 119530 }, { "epoch": 0.3, "learning_rate": 7.018219336655091e-05, "loss": 1.024, "step": 119535 }, { "epoch": 0.3, "learning_rate": 7.01809351250692e-05, "loss": 1.0233, "step": 119540 }, { "epoch": 0.3, "learning_rate": 7.01796768835875e-05, "loss": 1.0257, "step": 119545 }, { "epoch": 0.3, "learning_rate": 7.01784186421058e-05, "loss": 1.0208, "step": 119550 }, { "epoch": 0.3, "learning_rate": 7.017716040062409e-05, "loss": 1.0416, "step": 119555 }, { "epoch": 0.3, "learning_rate": 7.017590215914238e-05, "loss": 1.0223, "step": 119560 }, { "epoch": 0.3, "learning_rate": 7.017464391766067e-05, "loss": 1.022, "step": 119565 }, { "epoch": 0.3, "learning_rate": 7.017338567617898e-05, "loss": 1.0208, "step": 119570 }, { "epoch": 0.3, "learning_rate": 7.017212743469727e-05, "loss": 1.0206, "step": 119575 }, { "epoch": 0.3, "learning_rate": 7.017086919321556e-05, "loss": 1.0228, "step": 119580 }, { "epoch": 0.3, "learning_rate": 7.016961095173385e-05, "loss": 1.0263, "step": 119585 }, { "epoch": 0.3, "learning_rate": 7.016835271025216e-05, "loss": 1.0201, "step": 119590 }, { "epoch": 0.3, "learning_rate": 7.016709446877045e-05, "loss": 1.0196, "step": 119595 }, { "epoch": 0.3, "learning_rate": 7.016583622728874e-05, "loss": 1.0205, "step": 119600 }, { "epoch": 0.3, "learning_rate": 7.016457798580703e-05, "loss": 1.0243, "step": 119605 }, { "epoch": 0.3, "learning_rate": 7.016331974432534e-05, "loss": 1.0203, "step": 119610 }, { "epoch": 0.3, "learning_rate": 7.016206150284364e-05, "loss": 1.0227, "step": 119615 }, { "epoch": 0.3, "learning_rate": 7.016080326136193e-05, "loss": 1.0221, "step": 119620 }, { "epoch": 0.3, "learning_rate": 7.015954501988022e-05, "loss": 1.0225, "step": 119625 }, { "epoch": 0.3, "learning_rate": 7.015828677839852e-05, "loss": 1.0219, "step": 119630 }, { "epoch": 0.3, "learning_rate": 7.01570285369168e-05, "loss": 1.0244, "step": 119635 }, { "epoch": 0.3, "learning_rate": 7.015577029543511e-05, "loss": 1.0232, "step": 119640 }, { "epoch": 0.3, "learning_rate": 7.01545120539534e-05, "loss": 1.0196, "step": 119645 }, { "epoch": 0.3, "learning_rate": 7.01532538124717e-05, "loss": 1.0204, "step": 119650 }, { "epoch": 0.3, "learning_rate": 7.015199557098999e-05, "loss": 1.0217, "step": 119655 }, { "epoch": 0.3, "learning_rate": 7.015073732950829e-05, "loss": 1.0206, "step": 119660 }, { "epoch": 0.3, "learning_rate": 7.014947908802658e-05, "loss": 1.0171, "step": 119665 }, { "epoch": 0.3, "learning_rate": 7.014822084654488e-05, "loss": 1.0225, "step": 119670 }, { "epoch": 0.3, "learning_rate": 7.014696260506317e-05, "loss": 1.0219, "step": 119675 }, { "epoch": 0.3, "learning_rate": 7.014570436358147e-05, "loss": 1.0229, "step": 119680 }, { "epoch": 0.3, "learning_rate": 7.014444612209976e-05, "loss": 1.0223, "step": 119685 }, { "epoch": 0.3, "learning_rate": 7.014318788061806e-05, "loss": 1.0229, "step": 119690 }, { "epoch": 0.3, "learning_rate": 7.014192963913635e-05, "loss": 1.0214, "step": 119695 }, { "epoch": 0.3, "learning_rate": 7.014067139765464e-05, "loss": 1.0237, "step": 119700 }, { "epoch": 0.3, "learning_rate": 7.013941315617294e-05, "loss": 1.0229, "step": 119705 }, { "epoch": 0.3, "learning_rate": 7.013815491469123e-05, "loss": 1.0226, "step": 119710 }, { "epoch": 0.3, "learning_rate": 7.013689667320953e-05, "loss": 1.0212, "step": 119715 }, { "epoch": 0.3, "learning_rate": 7.013563843172782e-05, "loss": 1.0222, "step": 119720 }, { "epoch": 0.3, "learning_rate": 7.013438019024612e-05, "loss": 1.0223, "step": 119725 }, { "epoch": 0.3, "learning_rate": 7.013312194876441e-05, "loss": 1.023, "step": 119730 }, { "epoch": 0.3, "learning_rate": 7.01318637072827e-05, "loss": 1.0221, "step": 119735 }, { "epoch": 0.3, "learning_rate": 7.0130605465801e-05, "loss": 1.0219, "step": 119740 }, { "epoch": 0.3, "learning_rate": 7.01293472243193e-05, "loss": 1.0211, "step": 119745 }, { "epoch": 0.3, "learning_rate": 7.01280889828376e-05, "loss": 1.0231, "step": 119750 }, { "epoch": 0.3, "learning_rate": 7.012683074135589e-05, "loss": 1.0232, "step": 119755 }, { "epoch": 0.3, "learning_rate": 7.012557249987418e-05, "loss": 1.0222, "step": 119760 }, { "epoch": 0.3, "learning_rate": 7.012431425839247e-05, "loss": 1.0226, "step": 119765 }, { "epoch": 0.3, "learning_rate": 7.012305601691077e-05, "loss": 1.0223, "step": 119770 }, { "epoch": 0.3, "learning_rate": 7.012179777542907e-05, "loss": 1.021, "step": 119775 }, { "epoch": 0.3, "learning_rate": 7.012053953394736e-05, "loss": 1.0195, "step": 119780 }, { "epoch": 0.3, "learning_rate": 7.011928129246565e-05, "loss": 1.021, "step": 119785 }, { "epoch": 0.3, "learning_rate": 7.011802305098395e-05, "loss": 1.023, "step": 119790 }, { "epoch": 0.3, "learning_rate": 7.011676480950225e-05, "loss": 1.0194, "step": 119795 }, { "epoch": 0.3, "learning_rate": 7.011550656802054e-05, "loss": 1.0199, "step": 119800 }, { "epoch": 0.3, "learning_rate": 7.011424832653883e-05, "loss": 1.0195, "step": 119805 }, { "epoch": 0.3, "learning_rate": 7.011299008505713e-05, "loss": 1.021, "step": 119810 }, { "epoch": 0.3, "learning_rate": 7.011173184357543e-05, "loss": 1.0225, "step": 119815 }, { "epoch": 0.3, "learning_rate": 7.011047360209372e-05, "loss": 1.0227, "step": 119820 }, { "epoch": 0.3, "learning_rate": 7.010921536061201e-05, "loss": 1.0226, "step": 119825 }, { "epoch": 0.3, "learning_rate": 7.01079571191303e-05, "loss": 1.0217, "step": 119830 }, { "epoch": 0.3, "learning_rate": 7.01066988776486e-05, "loss": 1.0203, "step": 119835 }, { "epoch": 0.3, "learning_rate": 7.01054406361669e-05, "loss": 1.0223, "step": 119840 }, { "epoch": 0.3, "learning_rate": 7.010418239468519e-05, "loss": 1.0234, "step": 119845 }, { "epoch": 0.3, "learning_rate": 7.010292415320348e-05, "loss": 1.0217, "step": 119850 }, { "epoch": 0.3, "learning_rate": 7.010166591172179e-05, "loss": 1.022, "step": 119855 }, { "epoch": 0.3, "learning_rate": 7.010040767024008e-05, "loss": 1.022, "step": 119860 }, { "epoch": 0.3, "learning_rate": 7.009914942875837e-05, "loss": 1.0191, "step": 119865 }, { "epoch": 0.3, "learning_rate": 7.009789118727666e-05, "loss": 1.0237, "step": 119870 }, { "epoch": 0.3, "learning_rate": 7.009663294579497e-05, "loss": 1.0209, "step": 119875 }, { "epoch": 0.3, "learning_rate": 7.009537470431326e-05, "loss": 1.0235, "step": 119880 }, { "epoch": 0.3, "learning_rate": 7.009411646283155e-05, "loss": 1.0219, "step": 119885 }, { "epoch": 0.3, "learning_rate": 7.009285822134984e-05, "loss": 1.0421, "step": 119890 }, { "epoch": 0.3, "learning_rate": 7.009159997986813e-05, "loss": 1.0218, "step": 119895 }, { "epoch": 0.3, "learning_rate": 7.009034173838644e-05, "loss": 1.0215, "step": 119900 }, { "epoch": 0.3, "learning_rate": 7.008908349690473e-05, "loss": 1.0211, "step": 119905 }, { "epoch": 0.3, "learning_rate": 7.008782525542302e-05, "loss": 1.021, "step": 119910 }, { "epoch": 0.3, "learning_rate": 7.008656701394131e-05, "loss": 1.0203, "step": 119915 }, { "epoch": 0.3, "learning_rate": 7.008530877245962e-05, "loss": 1.0231, "step": 119920 }, { "epoch": 0.3, "learning_rate": 7.008405053097791e-05, "loss": 1.0222, "step": 119925 }, { "epoch": 0.3, "learning_rate": 7.00827922894962e-05, "loss": 1.0217, "step": 119930 }, { "epoch": 0.3, "learning_rate": 7.008153404801449e-05, "loss": 1.0206, "step": 119935 }, { "epoch": 0.3, "learning_rate": 7.00802758065328e-05, "loss": 1.0413, "step": 119940 }, { "epoch": 0.3, "learning_rate": 7.007901756505109e-05, "loss": 1.0215, "step": 119945 }, { "epoch": 0.3, "learning_rate": 7.007775932356938e-05, "loss": 1.022, "step": 119950 }, { "epoch": 0.3, "learning_rate": 7.007650108208767e-05, "loss": 1.0205, "step": 119955 }, { "epoch": 0.3, "learning_rate": 7.007524284060596e-05, "loss": 1.0233, "step": 119960 }, { "epoch": 0.3, "learning_rate": 7.007398459912427e-05, "loss": 1.0237, "step": 119965 }, { "epoch": 0.3, "learning_rate": 7.007272635764256e-05, "loss": 1.0213, "step": 119970 }, { "epoch": 0.3, "learning_rate": 7.007146811616085e-05, "loss": 1.0217, "step": 119975 }, { "epoch": 0.3, "learning_rate": 7.007020987467914e-05, "loss": 1.0223, "step": 119980 }, { "epoch": 0.3, "learning_rate": 7.006895163319745e-05, "loss": 1.0234, "step": 119985 }, { "epoch": 0.3, "learning_rate": 7.006769339171574e-05, "loss": 1.0204, "step": 119990 }, { "epoch": 0.3, "learning_rate": 7.006643515023403e-05, "loss": 1.0204, "step": 119995 }, { "epoch": 0.3, "learning_rate": 7.006517690875232e-05, "loss": 1.0189, "step": 120000 }, { "epoch": 0.3, "learning_rate": 7.006391866727063e-05, "loss": 1.0212, "step": 120005 }, { "epoch": 0.3, "learning_rate": 7.006266042578892e-05, "loss": 1.0235, "step": 120010 }, { "epoch": 0.3, "learning_rate": 7.006140218430721e-05, "loss": 1.0208, "step": 120015 }, { "epoch": 0.3, "learning_rate": 7.00601439428255e-05, "loss": 1.0234, "step": 120020 }, { "epoch": 0.3, "learning_rate": 7.00588857013438e-05, "loss": 1.0231, "step": 120025 }, { "epoch": 0.3, "learning_rate": 7.00576274598621e-05, "loss": 1.0394, "step": 120030 }, { "epoch": 0.3, "learning_rate": 7.005636921838039e-05, "loss": 1.0211, "step": 120035 }, { "epoch": 0.3, "learning_rate": 7.005511097689868e-05, "loss": 1.0216, "step": 120040 }, { "epoch": 0.3, "learning_rate": 7.005385273541698e-05, "loss": 1.023, "step": 120045 }, { "epoch": 0.3, "learning_rate": 7.005259449393528e-05, "loss": 1.0229, "step": 120050 }, { "epoch": 0.3, "learning_rate": 7.005133625245357e-05, "loss": 1.0227, "step": 120055 }, { "epoch": 0.3, "learning_rate": 7.005007801097186e-05, "loss": 1.0226, "step": 120060 }, { "epoch": 0.3, "learning_rate": 7.004881976949016e-05, "loss": 1.0219, "step": 120065 }, { "epoch": 0.3, "learning_rate": 7.004756152800846e-05, "loss": 1.0238, "step": 120070 }, { "epoch": 0.3, "learning_rate": 7.004630328652675e-05, "loss": 1.0242, "step": 120075 }, { "epoch": 0.3, "learning_rate": 7.004504504504504e-05, "loss": 1.0212, "step": 120080 }, { "epoch": 0.3, "learning_rate": 7.004378680356334e-05, "loss": 1.0233, "step": 120085 }, { "epoch": 0.3, "learning_rate": 7.004252856208163e-05, "loss": 1.021, "step": 120090 }, { "epoch": 0.3, "learning_rate": 7.004127032059993e-05, "loss": 1.0237, "step": 120095 }, { "epoch": 0.3, "learning_rate": 7.004001207911822e-05, "loss": 1.0226, "step": 120100 }, { "epoch": 0.3, "learning_rate": 7.003875383763652e-05, "loss": 1.0247, "step": 120105 }, { "epoch": 0.3, "learning_rate": 7.003749559615482e-05, "loss": 1.0205, "step": 120110 }, { "epoch": 0.3, "learning_rate": 7.003623735467311e-05, "loss": 1.0256, "step": 120115 }, { "epoch": 0.3, "learning_rate": 7.003497911319142e-05, "loss": 1.0227, "step": 120120 }, { "epoch": 0.3, "learning_rate": 7.003372087170971e-05, "loss": 1.0223, "step": 120125 }, { "epoch": 0.3, "learning_rate": 7.0032462630228e-05, "loss": 1.022, "step": 120130 }, { "epoch": 0.3, "learning_rate": 7.003120438874629e-05, "loss": 1.0222, "step": 120135 }, { "epoch": 0.3, "learning_rate": 7.00299461472646e-05, "loss": 1.0215, "step": 120140 }, { "epoch": 0.3, "learning_rate": 7.002868790578289e-05, "loss": 1.0185, "step": 120145 }, { "epoch": 0.3, "learning_rate": 7.002742966430118e-05, "loss": 1.0221, "step": 120150 }, { "epoch": 0.3, "learning_rate": 7.002617142281947e-05, "loss": 1.0199, "step": 120155 }, { "epoch": 0.3, "learning_rate": 7.002491318133778e-05, "loss": 1.0205, "step": 120160 }, { "epoch": 0.3, "learning_rate": 7.002365493985607e-05, "loss": 1.0235, "step": 120165 }, { "epoch": 0.3, "learning_rate": 7.002239669837436e-05, "loss": 1.0211, "step": 120170 }, { "epoch": 0.3, "learning_rate": 7.002113845689265e-05, "loss": 1.0221, "step": 120175 }, { "epoch": 0.3, "learning_rate": 7.001988021541094e-05, "loss": 1.0219, "step": 120180 }, { "epoch": 0.3, "learning_rate": 7.001862197392925e-05, "loss": 1.0329, "step": 120185 }, { "epoch": 0.3, "learning_rate": 7.001736373244754e-05, "loss": 1.0218, "step": 120190 }, { "epoch": 0.3, "learning_rate": 7.001610549096583e-05, "loss": 1.0435, "step": 120195 }, { "epoch": 0.3, "learning_rate": 7.001484724948412e-05, "loss": 1.0205, "step": 120200 }, { "epoch": 0.3, "learning_rate": 7.001358900800243e-05, "loss": 1.0235, "step": 120205 }, { "epoch": 0.3, "learning_rate": 7.001233076652072e-05, "loss": 1.0232, "step": 120210 }, { "epoch": 0.3, "learning_rate": 7.001107252503901e-05, "loss": 1.0207, "step": 120215 }, { "epoch": 0.3, "learning_rate": 7.00098142835573e-05, "loss": 1.022, "step": 120220 }, { "epoch": 0.3, "learning_rate": 7.00085560420756e-05, "loss": 1.0218, "step": 120225 }, { "epoch": 0.3, "learning_rate": 7.00072978005939e-05, "loss": 1.0225, "step": 120230 }, { "epoch": 0.3, "learning_rate": 7.000603955911219e-05, "loss": 1.02, "step": 120235 }, { "epoch": 0.3, "learning_rate": 7.000478131763048e-05, "loss": 1.0222, "step": 120240 }, { "epoch": 0.3, "learning_rate": 7.000352307614877e-05, "loss": 1.0222, "step": 120245 }, { "epoch": 0.3, "learning_rate": 7.000226483466708e-05, "loss": 1.0214, "step": 120250 }, { "epoch": 0.3, "learning_rate": 7.000100659318537e-05, "loss": 1.0252, "step": 120255 }, { "epoch": 0.3, "learning_rate": 6.999974835170366e-05, "loss": 1.0199, "step": 120260 }, { "epoch": 0.3, "learning_rate": 6.999849011022195e-05, "loss": 1.0215, "step": 120265 }, { "epoch": 0.3, "learning_rate": 6.999723186874026e-05, "loss": 1.0205, "step": 120270 }, { "epoch": 0.3, "learning_rate": 6.999597362725855e-05, "loss": 1.0216, "step": 120275 }, { "epoch": 0.3, "learning_rate": 6.999471538577684e-05, "loss": 1.0207, "step": 120280 }, { "epoch": 0.3, "learning_rate": 6.999345714429513e-05, "loss": 1.0192, "step": 120285 }, { "epoch": 0.3, "learning_rate": 6.999219890281343e-05, "loss": 1.0217, "step": 120290 }, { "epoch": 0.3, "learning_rate": 6.999094066133173e-05, "loss": 1.0224, "step": 120295 }, { "epoch": 0.3, "learning_rate": 6.998968241985002e-05, "loss": 1.0231, "step": 120300 }, { "epoch": 0.3, "learning_rate": 6.998842417836831e-05, "loss": 1.0206, "step": 120305 }, { "epoch": 0.3, "learning_rate": 6.99871659368866e-05, "loss": 1.0215, "step": 120310 }, { "epoch": 0.3, "learning_rate": 6.998590769540491e-05, "loss": 1.0209, "step": 120315 }, { "epoch": 0.3, "learning_rate": 6.99846494539232e-05, "loss": 1.0225, "step": 120320 }, { "epoch": 0.3, "learning_rate": 6.99833912124415e-05, "loss": 1.0188, "step": 120325 }, { "epoch": 0.3, "learning_rate": 6.998213297095979e-05, "loss": 1.0229, "step": 120330 }, { "epoch": 0.3, "learning_rate": 6.998087472947809e-05, "loss": 1.022, "step": 120335 }, { "epoch": 0.3, "learning_rate": 6.997961648799638e-05, "loss": 1.0217, "step": 120340 }, { "epoch": 0.3, "learning_rate": 6.997835824651467e-05, "loss": 1.0251, "step": 120345 }, { "epoch": 0.3, "learning_rate": 6.997710000503297e-05, "loss": 1.0209, "step": 120350 }, { "epoch": 0.3, "learning_rate": 6.997584176355126e-05, "loss": 1.0216, "step": 120355 }, { "epoch": 0.3, "learning_rate": 6.997458352206956e-05, "loss": 1.0231, "step": 120360 }, { "epoch": 0.3, "learning_rate": 6.997332528058785e-05, "loss": 1.0327, "step": 120365 }, { "epoch": 0.3, "learning_rate": 6.997206703910615e-05, "loss": 1.0195, "step": 120370 }, { "epoch": 0.3, "learning_rate": 6.997080879762444e-05, "loss": 1.0209, "step": 120375 }, { "epoch": 0.3, "learning_rate": 6.996955055614274e-05, "loss": 1.0206, "step": 120380 }, { "epoch": 0.3, "learning_rate": 6.996829231466103e-05, "loss": 1.0205, "step": 120385 }, { "epoch": 0.3, "learning_rate": 6.996703407317933e-05, "loss": 1.0208, "step": 120390 }, { "epoch": 0.3, "learning_rate": 6.996577583169762e-05, "loss": 1.0207, "step": 120395 }, { "epoch": 0.3, "learning_rate": 6.996451759021592e-05, "loss": 1.0215, "step": 120400 }, { "epoch": 0.3, "learning_rate": 6.996325934873421e-05, "loss": 1.0218, "step": 120405 }, { "epoch": 0.3, "learning_rate": 6.99620011072525e-05, "loss": 1.0205, "step": 120410 }, { "epoch": 0.3, "learning_rate": 6.99607428657708e-05, "loss": 1.022, "step": 120415 }, { "epoch": 0.3, "learning_rate": 6.995948462428909e-05, "loss": 1.0241, "step": 120420 }, { "epoch": 0.3, "learning_rate": 6.99582263828074e-05, "loss": 1.021, "step": 120425 }, { "epoch": 0.3, "learning_rate": 6.995696814132569e-05, "loss": 1.0201, "step": 120430 }, { "epoch": 0.3, "learning_rate": 6.995570989984398e-05, "loss": 1.023, "step": 120435 }, { "epoch": 0.3, "learning_rate": 6.995445165836227e-05, "loss": 1.0188, "step": 120440 }, { "epoch": 0.3, "learning_rate": 6.995319341688057e-05, "loss": 1.0227, "step": 120445 }, { "epoch": 0.3, "learning_rate": 6.995193517539887e-05, "loss": 1.0233, "step": 120450 }, { "epoch": 0.3, "learning_rate": 6.995067693391716e-05, "loss": 1.0206, "step": 120455 }, { "epoch": 0.3, "learning_rate": 6.994941869243545e-05, "loss": 1.0224, "step": 120460 }, { "epoch": 0.3, "learning_rate": 6.994816045095375e-05, "loss": 1.0226, "step": 120465 }, { "epoch": 0.3, "learning_rate": 6.994690220947205e-05, "loss": 1.0229, "step": 120470 }, { "epoch": 0.3, "learning_rate": 6.994564396799034e-05, "loss": 1.0211, "step": 120475 }, { "epoch": 0.3, "learning_rate": 6.994438572650863e-05, "loss": 1.0211, "step": 120480 }, { "epoch": 0.3, "learning_rate": 6.994312748502692e-05, "loss": 1.0438, "step": 120485 }, { "epoch": 0.3, "learning_rate": 6.994186924354523e-05, "loss": 1.0238, "step": 120490 }, { "epoch": 0.3, "learning_rate": 6.994061100206352e-05, "loss": 1.0215, "step": 120495 }, { "epoch": 0.3, "learning_rate": 6.993935276058181e-05, "loss": 1.0227, "step": 120500 }, { "epoch": 0.3, "learning_rate": 6.99380945191001e-05, "loss": 1.0205, "step": 120505 }, { "epoch": 0.3, "learning_rate": 6.99368362776184e-05, "loss": 1.0227, "step": 120510 }, { "epoch": 0.3, "learning_rate": 6.99355780361367e-05, "loss": 1.0216, "step": 120515 }, { "epoch": 0.3, "learning_rate": 6.993431979465499e-05, "loss": 1.0207, "step": 120520 }, { "epoch": 0.3, "learning_rate": 6.993306155317328e-05, "loss": 1.0235, "step": 120525 }, { "epoch": 0.3, "learning_rate": 6.993180331169159e-05, "loss": 1.0214, "step": 120530 }, { "epoch": 0.3, "learning_rate": 6.993054507020988e-05, "loss": 1.0229, "step": 120535 }, { "epoch": 0.3, "learning_rate": 6.992928682872817e-05, "loss": 1.0242, "step": 120540 }, { "epoch": 0.3, "learning_rate": 6.992802858724646e-05, "loss": 1.0222, "step": 120545 }, { "epoch": 0.3, "learning_rate": 6.992677034576475e-05, "loss": 1.0236, "step": 120550 }, { "epoch": 0.3, "learning_rate": 6.992551210428306e-05, "loss": 1.022, "step": 120555 }, { "epoch": 0.3, "learning_rate": 6.992475715939404e-05, "loss": 1.0217, "step": 120560 }, { "epoch": 0.3, "learning_rate": 6.992349891791233e-05, "loss": 1.0207, "step": 120565 }, { "epoch": 0.3, "learning_rate": 6.992224067643063e-05, "loss": 1.0355, "step": 120570 }, { "epoch": 0.3, "learning_rate": 6.992098243494893e-05, "loss": 1.0449, "step": 120575 }, { "epoch": 0.3, "learning_rate": 6.991972419346722e-05, "loss": 1.0354, "step": 120580 }, { "epoch": 0.3, "learning_rate": 6.991846595198551e-05, "loss": 1.0216, "step": 120585 }, { "epoch": 0.3, "learning_rate": 6.99172077105038e-05, "loss": 1.0224, "step": 120590 }, { "epoch": 0.3, "learning_rate": 6.99159494690221e-05, "loss": 1.0215, "step": 120595 }, { "epoch": 0.3, "learning_rate": 6.99146912275404e-05, "loss": 1.0215, "step": 120600 }, { "epoch": 0.3, "learning_rate": 6.991343298605869e-05, "loss": 1.0216, "step": 120605 }, { "epoch": 0.3, "learning_rate": 6.991217474457698e-05, "loss": 1.0218, "step": 120610 }, { "epoch": 0.3, "learning_rate": 6.991091650309529e-05, "loss": 1.0205, "step": 120615 }, { "epoch": 0.3, "learning_rate": 6.990965826161358e-05, "loss": 1.0179, "step": 120620 }, { "epoch": 0.3, "learning_rate": 6.990840002013187e-05, "loss": 1.024, "step": 120625 }, { "epoch": 0.3, "learning_rate": 6.990714177865016e-05, "loss": 1.0208, "step": 120630 }, { "epoch": 0.3, "learning_rate": 6.990588353716847e-05, "loss": 1.0203, "step": 120635 }, { "epoch": 0.3, "learning_rate": 6.990462529568676e-05, "loss": 1.024, "step": 120640 }, { "epoch": 0.3, "learning_rate": 6.990336705420505e-05, "loss": 1.0231, "step": 120645 }, { "epoch": 0.3, "learning_rate": 6.990210881272334e-05, "loss": 1.0206, "step": 120650 }, { "epoch": 0.3, "learning_rate": 6.990085057124163e-05, "loss": 1.0232, "step": 120655 }, { "epoch": 0.3, "learning_rate": 6.989959232975994e-05, "loss": 1.0205, "step": 120660 }, { "epoch": 0.3, "learning_rate": 6.989833408827823e-05, "loss": 1.0221, "step": 120665 }, { "epoch": 0.3, "learning_rate": 6.989707584679652e-05, "loss": 1.022, "step": 120670 }, { "epoch": 0.3, "learning_rate": 6.989581760531481e-05, "loss": 1.0225, "step": 120675 }, { "epoch": 0.3, "learning_rate": 6.989455936383312e-05, "loss": 1.0235, "step": 120680 }, { "epoch": 0.3, "learning_rate": 6.989330112235141e-05, "loss": 1.0232, "step": 120685 }, { "epoch": 0.3, "learning_rate": 6.98920428808697e-05, "loss": 1.0225, "step": 120690 }, { "epoch": 0.3, "learning_rate": 6.989078463938799e-05, "loss": 1.022, "step": 120695 }, { "epoch": 0.3, "learning_rate": 6.98895263979063e-05, "loss": 1.0221, "step": 120700 }, { "epoch": 0.3, "learning_rate": 6.988826815642459e-05, "loss": 1.0216, "step": 120705 }, { "epoch": 0.3, "learning_rate": 6.988700991494288e-05, "loss": 1.0224, "step": 120710 }, { "epoch": 0.3, "learning_rate": 6.988575167346117e-05, "loss": 1.0209, "step": 120715 }, { "epoch": 0.3, "learning_rate": 6.988449343197946e-05, "loss": 1.0224, "step": 120720 }, { "epoch": 0.3, "learning_rate": 6.988323519049777e-05, "loss": 1.0212, "step": 120725 }, { "epoch": 0.3, "learning_rate": 6.988197694901606e-05, "loss": 1.0204, "step": 120730 }, { "epoch": 0.3, "learning_rate": 6.988071870753435e-05, "loss": 1.0182, "step": 120735 }, { "epoch": 0.3, "learning_rate": 6.987946046605264e-05, "loss": 1.0243, "step": 120740 }, { "epoch": 0.3, "learning_rate": 6.987820222457095e-05, "loss": 1.0208, "step": 120745 }, { "epoch": 0.3, "learning_rate": 6.987694398308924e-05, "loss": 1.0209, "step": 120750 }, { "epoch": 0.3, "learning_rate": 6.987568574160753e-05, "loss": 1.0221, "step": 120755 }, { "epoch": 0.3, "learning_rate": 6.987442750012582e-05, "loss": 1.0225, "step": 120760 }, { "epoch": 0.3, "learning_rate": 6.987316925864413e-05, "loss": 1.0235, "step": 120765 }, { "epoch": 0.3, "learning_rate": 6.987191101716242e-05, "loss": 1.0233, "step": 120770 }, { "epoch": 0.3, "learning_rate": 6.987065277568071e-05, "loss": 1.0219, "step": 120775 }, { "epoch": 0.3, "learning_rate": 6.9869394534199e-05, "loss": 1.0242, "step": 120780 }, { "epoch": 0.3, "learning_rate": 6.98681362927173e-05, "loss": 1.0215, "step": 120785 }, { "epoch": 0.3, "learning_rate": 6.98668780512356e-05, "loss": 1.0235, "step": 120790 }, { "epoch": 0.3, "learning_rate": 6.986561980975389e-05, "loss": 1.0226, "step": 120795 }, { "epoch": 0.3, "learning_rate": 6.986436156827218e-05, "loss": 1.0227, "step": 120800 }, { "epoch": 0.3, "learning_rate": 6.986310332679047e-05, "loss": 1.022, "step": 120805 }, { "epoch": 0.3, "learning_rate": 6.986184508530878e-05, "loss": 1.0195, "step": 120810 }, { "epoch": 0.3, "learning_rate": 6.986058684382707e-05, "loss": 1.0201, "step": 120815 }, { "epoch": 0.3, "learning_rate": 6.985932860234536e-05, "loss": 1.0218, "step": 120820 }, { "epoch": 0.3, "learning_rate": 6.985807036086365e-05, "loss": 1.0211, "step": 120825 }, { "epoch": 0.3, "learning_rate": 6.985681211938196e-05, "loss": 1.0221, "step": 120830 }, { "epoch": 0.3, "learning_rate": 6.985555387790025e-05, "loss": 1.0216, "step": 120835 }, { "epoch": 0.3, "learning_rate": 6.985429563641854e-05, "loss": 1.0231, "step": 120840 }, { "epoch": 0.3, "learning_rate": 6.985303739493683e-05, "loss": 1.0198, "step": 120845 }, { "epoch": 0.3, "learning_rate": 6.985177915345513e-05, "loss": 1.022, "step": 120850 }, { "epoch": 0.3, "learning_rate": 6.985052091197343e-05, "loss": 1.0217, "step": 120855 }, { "epoch": 0.3, "learning_rate": 6.984926267049172e-05, "loss": 1.0215, "step": 120860 }, { "epoch": 0.3, "learning_rate": 6.984800442901001e-05, "loss": 1.0251, "step": 120865 }, { "epoch": 0.3, "learning_rate": 6.98467461875283e-05, "loss": 1.0234, "step": 120870 }, { "epoch": 0.3, "learning_rate": 6.984548794604661e-05, "loss": 1.0243, "step": 120875 }, { "epoch": 0.3, "learning_rate": 6.98442297045649e-05, "loss": 1.0232, "step": 120880 }, { "epoch": 0.3, "learning_rate": 6.98429714630832e-05, "loss": 1.0245, "step": 120885 }, { "epoch": 0.3, "learning_rate": 6.984171322160149e-05, "loss": 1.023, "step": 120890 }, { "epoch": 0.3, "learning_rate": 6.984045498011979e-05, "loss": 1.0205, "step": 120895 }, { "epoch": 0.3, "learning_rate": 6.983919673863808e-05, "loss": 1.0221, "step": 120900 }, { "epoch": 0.3, "learning_rate": 6.983793849715637e-05, "loss": 1.0219, "step": 120905 }, { "epoch": 0.3, "learning_rate": 6.983668025567467e-05, "loss": 1.0197, "step": 120910 }, { "epoch": 0.3, "learning_rate": 6.983542201419296e-05, "loss": 1.0193, "step": 120915 }, { "epoch": 0.3, "learning_rate": 6.983416377271126e-05, "loss": 1.02, "step": 120920 }, { "epoch": 0.3, "learning_rate": 6.983290553122955e-05, "loss": 1.0207, "step": 120925 }, { "epoch": 0.3, "learning_rate": 6.983164728974785e-05, "loss": 1.0214, "step": 120930 }, { "epoch": 0.3, "learning_rate": 6.983038904826614e-05, "loss": 1.0203, "step": 120935 }, { "epoch": 0.3, "learning_rate": 6.982913080678444e-05, "loss": 1.0228, "step": 120940 }, { "epoch": 0.3, "learning_rate": 6.982787256530273e-05, "loss": 1.0195, "step": 120945 }, { "epoch": 0.3, "learning_rate": 6.982661432382103e-05, "loss": 1.0212, "step": 120950 }, { "epoch": 0.3, "learning_rate": 6.982535608233932e-05, "loss": 1.0224, "step": 120955 }, { "epoch": 0.3, "learning_rate": 6.982409784085762e-05, "loss": 1.0208, "step": 120960 }, { "epoch": 0.3, "learning_rate": 6.982283959937591e-05, "loss": 1.0225, "step": 120965 }, { "epoch": 0.3, "learning_rate": 6.98215813578942e-05, "loss": 1.0218, "step": 120970 }, { "epoch": 0.3, "learning_rate": 6.98203231164125e-05, "loss": 1.0208, "step": 120975 }, { "epoch": 0.3, "learning_rate": 6.981906487493079e-05, "loss": 1.0228, "step": 120980 }, { "epoch": 0.3, "learning_rate": 6.98178066334491e-05, "loss": 1.0195, "step": 120985 }, { "epoch": 0.3, "learning_rate": 6.981654839196739e-05, "loss": 1.0246, "step": 120990 }, { "epoch": 0.3, "learning_rate": 6.981529015048568e-05, "loss": 1.0212, "step": 120995 }, { "epoch": 0.3, "learning_rate": 6.981403190900397e-05, "loss": 1.025, "step": 121000 }, { "epoch": 0.3, "learning_rate": 6.981277366752227e-05, "loss": 1.0204, "step": 121005 }, { "epoch": 0.3, "learning_rate": 6.981151542604058e-05, "loss": 1.0217, "step": 121010 }, { "epoch": 0.3, "learning_rate": 6.981025718455887e-05, "loss": 1.0223, "step": 121015 }, { "epoch": 0.3, "learning_rate": 6.980899894307716e-05, "loss": 1.0231, "step": 121020 }, { "epoch": 0.3, "learning_rate": 6.980774070159545e-05, "loss": 1.0193, "step": 121025 }, { "epoch": 0.3, "learning_rate": 6.980648246011376e-05, "loss": 1.0205, "step": 121030 }, { "epoch": 0.3, "learning_rate": 6.980522421863205e-05, "loss": 1.0208, "step": 121035 }, { "epoch": 0.3, "learning_rate": 6.980396597715034e-05, "loss": 1.0238, "step": 121040 }, { "epoch": 0.3, "learning_rate": 6.980270773566863e-05, "loss": 1.0237, "step": 121045 }, { "epoch": 0.3, "learning_rate": 6.980144949418693e-05, "loss": 1.0213, "step": 121050 }, { "epoch": 0.3, "learning_rate": 6.980019125270523e-05, "loss": 1.0229, "step": 121055 }, { "epoch": 0.3, "learning_rate": 6.979893301122352e-05, "loss": 1.0215, "step": 121060 }, { "epoch": 0.3, "learning_rate": 6.979767476974181e-05, "loss": 1.0222, "step": 121065 }, { "epoch": 0.3, "learning_rate": 6.97964165282601e-05, "loss": 1.0225, "step": 121070 }, { "epoch": 0.3, "learning_rate": 6.979515828677841e-05, "loss": 1.0218, "step": 121075 }, { "epoch": 0.3, "learning_rate": 6.97939000452967e-05, "loss": 1.0222, "step": 121080 }, { "epoch": 0.3, "learning_rate": 6.9792641803815e-05, "loss": 1.0201, "step": 121085 }, { "epoch": 0.3, "learning_rate": 6.979138356233329e-05, "loss": 1.0217, "step": 121090 }, { "epoch": 0.3, "learning_rate": 6.979012532085159e-05, "loss": 1.0223, "step": 121095 }, { "epoch": 0.3, "learning_rate": 6.978886707936988e-05, "loss": 1.0223, "step": 121100 }, { "epoch": 0.3, "learning_rate": 6.978760883788817e-05, "loss": 1.0205, "step": 121105 }, { "epoch": 0.3, "learning_rate": 6.978635059640647e-05, "loss": 1.0203, "step": 121110 }, { "epoch": 0.3, "learning_rate": 6.978509235492476e-05, "loss": 1.021, "step": 121115 }, { "epoch": 0.3, "learning_rate": 6.978383411344306e-05, "loss": 1.0234, "step": 121120 }, { "epoch": 0.3, "learning_rate": 6.978257587196135e-05, "loss": 1.0219, "step": 121125 }, { "epoch": 0.3, "learning_rate": 6.978131763047965e-05, "loss": 1.0221, "step": 121130 }, { "epoch": 0.3, "learning_rate": 6.978005938899794e-05, "loss": 1.0205, "step": 121135 }, { "epoch": 0.3, "learning_rate": 6.977880114751624e-05, "loss": 1.0226, "step": 121140 }, { "epoch": 0.3, "learning_rate": 6.977754290603453e-05, "loss": 1.0228, "step": 121145 }, { "epoch": 0.3, "learning_rate": 6.977628466455283e-05, "loss": 1.021, "step": 121150 }, { "epoch": 0.3, "learning_rate": 6.977502642307112e-05, "loss": 1.0214, "step": 121155 }, { "epoch": 0.3, "learning_rate": 6.977376818158942e-05, "loss": 1.0235, "step": 121160 }, { "epoch": 0.3, "learning_rate": 6.977250994010771e-05, "loss": 1.0209, "step": 121165 }, { "epoch": 0.3, "learning_rate": 6.9771251698626e-05, "loss": 1.0211, "step": 121170 }, { "epoch": 0.3, "learning_rate": 6.97699934571443e-05, "loss": 1.0212, "step": 121175 }, { "epoch": 0.3, "learning_rate": 6.976873521566259e-05, "loss": 1.0194, "step": 121180 }, { "epoch": 0.3, "learning_rate": 6.97674769741809e-05, "loss": 1.0214, "step": 121185 }, { "epoch": 0.3, "learning_rate": 6.976621873269919e-05, "loss": 1.0204, "step": 121190 }, { "epoch": 0.3, "learning_rate": 6.976496049121748e-05, "loss": 1.0234, "step": 121195 }, { "epoch": 0.3, "learning_rate": 6.976370224973577e-05, "loss": 1.0201, "step": 121200 }, { "epoch": 0.3, "learning_rate": 6.976244400825407e-05, "loss": 1.0472, "step": 121205 }, { "epoch": 0.3, "learning_rate": 6.976118576677237e-05, "loss": 1.0205, "step": 121210 }, { "epoch": 0.3, "learning_rate": 6.975992752529066e-05, "loss": 1.0207, "step": 121215 }, { "epoch": 0.3, "learning_rate": 6.975866928380895e-05, "loss": 1.0217, "step": 121220 }, { "epoch": 0.3, "learning_rate": 6.975741104232725e-05, "loss": 1.0221, "step": 121225 }, { "epoch": 0.3, "learning_rate": 6.975615280084555e-05, "loss": 1.02, "step": 121230 }, { "epoch": 0.3, "learning_rate": 6.975489455936384e-05, "loss": 1.0244, "step": 121235 }, { "epoch": 0.3, "learning_rate": 6.975363631788213e-05, "loss": 1.0226, "step": 121240 }, { "epoch": 0.3, "learning_rate": 6.975237807640042e-05, "loss": 1.0222, "step": 121245 }, { "epoch": 0.3, "learning_rate": 6.975111983491873e-05, "loss": 1.0224, "step": 121250 }, { "epoch": 0.3, "learning_rate": 6.974986159343702e-05, "loss": 1.0225, "step": 121255 }, { "epoch": 0.3, "learning_rate": 6.974860335195531e-05, "loss": 1.0221, "step": 121260 }, { "epoch": 0.3, "learning_rate": 6.97473451104736e-05, "loss": 1.0224, "step": 121265 }, { "epoch": 0.3, "learning_rate": 6.97460868689919e-05, "loss": 1.0237, "step": 121270 }, { "epoch": 0.3, "learning_rate": 6.97448286275102e-05, "loss": 1.0243, "step": 121275 }, { "epoch": 0.3, "learning_rate": 6.974357038602849e-05, "loss": 1.0232, "step": 121280 }, { "epoch": 0.3, "learning_rate": 6.974231214454678e-05, "loss": 1.0226, "step": 121285 }, { "epoch": 0.3, "learning_rate": 6.974105390306508e-05, "loss": 1.0203, "step": 121290 }, { "epoch": 0.3, "learning_rate": 6.973979566158338e-05, "loss": 1.0233, "step": 121295 }, { "epoch": 0.3, "learning_rate": 6.973853742010167e-05, "loss": 1.0222, "step": 121300 }, { "epoch": 0.3, "learning_rate": 6.973727917861996e-05, "loss": 1.0243, "step": 121305 }, { "epoch": 0.3, "learning_rate": 6.973602093713825e-05, "loss": 1.0225, "step": 121310 }, { "epoch": 0.3, "learning_rate": 6.973476269565656e-05, "loss": 1.0228, "step": 121315 }, { "epoch": 0.3, "learning_rate": 6.973350445417485e-05, "loss": 1.0222, "step": 121320 }, { "epoch": 0.3, "learning_rate": 6.973224621269314e-05, "loss": 1.0448, "step": 121325 }, { "epoch": 0.3, "learning_rate": 6.973098797121143e-05, "loss": 1.0225, "step": 121330 }, { "epoch": 0.3, "learning_rate": 6.972972972972974e-05, "loss": 1.0209, "step": 121335 }, { "epoch": 0.3, "learning_rate": 6.972847148824803e-05, "loss": 1.0223, "step": 121340 }, { "epoch": 0.3, "learning_rate": 6.972721324676632e-05, "loss": 1.0193, "step": 121345 }, { "epoch": 0.3, "learning_rate": 6.972595500528461e-05, "loss": 1.0202, "step": 121350 }, { "epoch": 0.3, "learning_rate": 6.972469676380292e-05, "loss": 1.0209, "step": 121355 }, { "epoch": 0.3, "learning_rate": 6.972343852232121e-05, "loss": 1.0208, "step": 121360 }, { "epoch": 0.3, "learning_rate": 6.97221802808395e-05, "loss": 1.0224, "step": 121365 }, { "epoch": 0.3, "learning_rate": 6.972092203935779e-05, "loss": 1.0233, "step": 121370 }, { "epoch": 0.3, "learning_rate": 6.971966379787608e-05, "loss": 1.0229, "step": 121375 }, { "epoch": 0.3, "learning_rate": 6.971840555639439e-05, "loss": 1.0198, "step": 121380 }, { "epoch": 0.3, "learning_rate": 6.971714731491268e-05, "loss": 1.0189, "step": 121385 }, { "epoch": 0.3, "learning_rate": 6.971588907343097e-05, "loss": 1.0203, "step": 121390 }, { "epoch": 0.3, "learning_rate": 6.971463083194926e-05, "loss": 1.0228, "step": 121395 }, { "epoch": 0.3, "learning_rate": 6.971337259046757e-05, "loss": 1.0199, "step": 121400 }, { "epoch": 0.3, "learning_rate": 6.971211434898586e-05, "loss": 1.0225, "step": 121405 }, { "epoch": 0.3, "learning_rate": 6.971085610750415e-05, "loss": 1.0208, "step": 121410 }, { "epoch": 0.3, "learning_rate": 6.970959786602244e-05, "loss": 1.0221, "step": 121415 }, { "epoch": 0.3, "learning_rate": 6.970833962454075e-05, "loss": 1.0231, "step": 121420 }, { "epoch": 0.3, "learning_rate": 6.970708138305904e-05, "loss": 1.022, "step": 121425 }, { "epoch": 0.3, "learning_rate": 6.970582314157733e-05, "loss": 1.0219, "step": 121430 }, { "epoch": 0.3, "learning_rate": 6.970456490009562e-05, "loss": 1.0206, "step": 121435 }, { "epoch": 0.3, "learning_rate": 6.970330665861391e-05, "loss": 1.023, "step": 121440 }, { "epoch": 0.3, "learning_rate": 6.970204841713222e-05, "loss": 1.0224, "step": 121445 }, { "epoch": 0.3, "learning_rate": 6.970079017565051e-05, "loss": 1.0201, "step": 121450 }, { "epoch": 0.3, "learning_rate": 6.96995319341688e-05, "loss": 1.0209, "step": 121455 }, { "epoch": 0.3, "learning_rate": 6.96982736926871e-05, "loss": 1.019, "step": 121460 }, { "epoch": 0.3, "learning_rate": 6.96970154512054e-05, "loss": 1.023, "step": 121465 }, { "epoch": 0.3, "learning_rate": 6.969575720972369e-05, "loss": 1.0201, "step": 121470 }, { "epoch": 0.3, "learning_rate": 6.969449896824198e-05, "loss": 1.0208, "step": 121475 }, { "epoch": 0.3, "learning_rate": 6.969324072676027e-05, "loss": 1.0241, "step": 121480 }, { "epoch": 0.3, "learning_rate": 6.969198248527858e-05, "loss": 1.0192, "step": 121485 }, { "epoch": 0.3, "learning_rate": 6.969072424379687e-05, "loss": 1.0178, "step": 121490 }, { "epoch": 0.3, "learning_rate": 6.968946600231516e-05, "loss": 1.021, "step": 121495 }, { "epoch": 0.3, "learning_rate": 6.968820776083345e-05, "loss": 1.0203, "step": 121500 }, { "epoch": 0.3, "learning_rate": 6.968694951935175e-05, "loss": 1.0187, "step": 121505 }, { "epoch": 0.31, "learning_rate": 6.968569127787006e-05, "loss": 1.0229, "step": 121510 }, { "epoch": 0.31, "learning_rate": 6.968443303638836e-05, "loss": 1.0229, "step": 121515 }, { "epoch": 0.31, "learning_rate": 6.968317479490665e-05, "loss": 1.0198, "step": 121520 }, { "epoch": 0.31, "learning_rate": 6.968191655342494e-05, "loss": 1.025, "step": 121525 }, { "epoch": 0.31, "learning_rate": 6.968065831194323e-05, "loss": 1.0189, "step": 121530 }, { "epoch": 0.31, "learning_rate": 6.967940007046154e-05, "loss": 1.0212, "step": 121535 }, { "epoch": 0.31, "learning_rate": 6.967814182897983e-05, "loss": 1.0181, "step": 121540 }, { "epoch": 0.31, "learning_rate": 6.967688358749812e-05, "loss": 1.0237, "step": 121545 }, { "epoch": 0.31, "learning_rate": 6.967562534601641e-05, "loss": 1.0218, "step": 121550 }, { "epoch": 0.31, "learning_rate": 6.967436710453472e-05, "loss": 1.0207, "step": 121555 }, { "epoch": 0.31, "learning_rate": 6.967310886305301e-05, "loss": 1.0204, "step": 121560 }, { "epoch": 0.31, "learning_rate": 6.96718506215713e-05, "loss": 1.0222, "step": 121565 }, { "epoch": 0.31, "learning_rate": 6.967059238008959e-05, "loss": 1.0226, "step": 121570 }, { "epoch": 0.31, "learning_rate": 6.96693341386079e-05, "loss": 1.0219, "step": 121575 }, { "epoch": 0.31, "learning_rate": 6.966807589712619e-05, "loss": 1.0218, "step": 121580 }, { "epoch": 0.31, "learning_rate": 6.966681765564448e-05, "loss": 1.0215, "step": 121585 }, { "epoch": 0.31, "learning_rate": 6.966555941416277e-05, "loss": 1.0226, "step": 121590 }, { "epoch": 0.31, "learning_rate": 6.966430117268106e-05, "loss": 1.0227, "step": 121595 }, { "epoch": 0.31, "learning_rate": 6.966304293119937e-05, "loss": 1.0222, "step": 121600 }, { "epoch": 0.31, "learning_rate": 6.966178468971766e-05, "loss": 1.042, "step": 121605 }, { "epoch": 0.31, "learning_rate": 6.966052644823595e-05, "loss": 1.0211, "step": 121610 }, { "epoch": 0.31, "learning_rate": 6.965926820675424e-05, "loss": 1.0224, "step": 121615 }, { "epoch": 0.31, "learning_rate": 6.965800996527255e-05, "loss": 1.022, "step": 121620 }, { "epoch": 0.31, "learning_rate": 6.965675172379084e-05, "loss": 1.019, "step": 121625 }, { "epoch": 0.31, "learning_rate": 6.965549348230913e-05, "loss": 1.021, "step": 121630 }, { "epoch": 0.31, "learning_rate": 6.965423524082742e-05, "loss": 1.0198, "step": 121635 }, { "epoch": 0.31, "learning_rate": 6.965297699934573e-05, "loss": 1.0229, "step": 121640 }, { "epoch": 0.31, "learning_rate": 6.965171875786402e-05, "loss": 1.0219, "step": 121645 }, { "epoch": 0.31, "learning_rate": 6.965046051638231e-05, "loss": 1.022, "step": 121650 }, { "epoch": 0.31, "learning_rate": 6.96492022749006e-05, "loss": 1.0246, "step": 121655 }, { "epoch": 0.31, "learning_rate": 6.96479440334189e-05, "loss": 1.0216, "step": 121660 }, { "epoch": 0.31, "learning_rate": 6.96466857919372e-05, "loss": 1.0209, "step": 121665 }, { "epoch": 0.31, "learning_rate": 6.964542755045549e-05, "loss": 1.022, "step": 121670 }, { "epoch": 0.31, "learning_rate": 6.964416930897378e-05, "loss": 1.0211, "step": 121675 }, { "epoch": 0.31, "learning_rate": 6.964291106749207e-05, "loss": 1.02, "step": 121680 }, { "epoch": 0.31, "learning_rate": 6.964165282601038e-05, "loss": 1.0205, "step": 121685 }, { "epoch": 0.31, "learning_rate": 6.964039458452867e-05, "loss": 1.0247, "step": 121690 }, { "epoch": 0.31, "learning_rate": 6.963913634304696e-05, "loss": 1.0186, "step": 121695 }, { "epoch": 0.31, "learning_rate": 6.963787810156525e-05, "loss": 1.021, "step": 121700 }, { "epoch": 0.31, "learning_rate": 6.963661986008355e-05, "loss": 1.0219, "step": 121705 }, { "epoch": 0.31, "learning_rate": 6.963536161860185e-05, "loss": 1.0213, "step": 121710 }, { "epoch": 0.31, "learning_rate": 6.963410337712014e-05, "loss": 1.0229, "step": 121715 }, { "epoch": 0.31, "learning_rate": 6.963284513563843e-05, "loss": 1.022, "step": 121720 }, { "epoch": 0.31, "learning_rate": 6.963158689415672e-05, "loss": 1.0226, "step": 121725 }, { "epoch": 0.31, "learning_rate": 6.963032865267503e-05, "loss": 1.0205, "step": 121730 }, { "epoch": 0.31, "learning_rate": 6.962907041119332e-05, "loss": 1.0218, "step": 121735 }, { "epoch": 0.31, "learning_rate": 6.962781216971161e-05, "loss": 1.0185, "step": 121740 }, { "epoch": 0.31, "learning_rate": 6.96265539282299e-05, "loss": 1.021, "step": 121745 }, { "epoch": 0.31, "learning_rate": 6.962529568674821e-05, "loss": 1.0197, "step": 121750 }, { "epoch": 0.31, "learning_rate": 6.96240374452665e-05, "loss": 1.0222, "step": 121755 }, { "epoch": 0.31, "learning_rate": 6.96227792037848e-05, "loss": 1.0199, "step": 121760 }, { "epoch": 0.31, "learning_rate": 6.962152096230308e-05, "loss": 1.0214, "step": 121765 }, { "epoch": 0.31, "learning_rate": 6.962026272082138e-05, "loss": 1.0216, "step": 121770 }, { "epoch": 0.31, "learning_rate": 6.961900447933968e-05, "loss": 1.0215, "step": 121775 }, { "epoch": 0.31, "learning_rate": 6.961774623785797e-05, "loss": 1.0234, "step": 121780 }, { "epoch": 0.31, "learning_rate": 6.961648799637626e-05, "loss": 1.0246, "step": 121785 }, { "epoch": 0.31, "learning_rate": 6.961522975489456e-05, "loss": 1.0234, "step": 121790 }, { "epoch": 0.31, "learning_rate": 6.961397151341286e-05, "loss": 1.022, "step": 121795 }, { "epoch": 0.31, "learning_rate": 6.961271327193115e-05, "loss": 1.0214, "step": 121800 }, { "epoch": 0.31, "learning_rate": 6.961145503044944e-05, "loss": 1.0219, "step": 121805 }, { "epoch": 0.31, "learning_rate": 6.961019678896774e-05, "loss": 1.0217, "step": 121810 }, { "epoch": 0.31, "learning_rate": 6.960893854748604e-05, "loss": 1.0218, "step": 121815 }, { "epoch": 0.31, "learning_rate": 6.960768030600433e-05, "loss": 1.0207, "step": 121820 }, { "epoch": 0.31, "learning_rate": 6.960642206452262e-05, "loss": 1.0222, "step": 121825 }, { "epoch": 0.31, "learning_rate": 6.960516382304092e-05, "loss": 1.045, "step": 121830 }, { "epoch": 0.31, "learning_rate": 6.960390558155921e-05, "loss": 1.0243, "step": 121835 }, { "epoch": 0.31, "learning_rate": 6.960264734007751e-05, "loss": 1.0207, "step": 121840 }, { "epoch": 0.31, "learning_rate": 6.96013890985958e-05, "loss": 1.0203, "step": 121845 }, { "epoch": 0.31, "learning_rate": 6.96001308571141e-05, "loss": 1.0218, "step": 121850 }, { "epoch": 0.31, "learning_rate": 6.959887261563239e-05, "loss": 1.0216, "step": 121855 }, { "epoch": 0.31, "learning_rate": 6.959761437415069e-05, "loss": 1.0208, "step": 121860 }, { "epoch": 0.31, "learning_rate": 6.959635613266898e-05, "loss": 1.0243, "step": 121865 }, { "epoch": 0.31, "learning_rate": 6.959509789118728e-05, "loss": 1.0227, "step": 121870 }, { "epoch": 0.31, "learning_rate": 6.959383964970557e-05, "loss": 1.0252, "step": 121875 }, { "epoch": 0.31, "learning_rate": 6.959258140822387e-05, "loss": 1.0221, "step": 121880 }, { "epoch": 0.31, "learning_rate": 6.959132316674216e-05, "loss": 1.0204, "step": 121885 }, { "epoch": 0.31, "learning_rate": 6.959006492526046e-05, "loss": 1.0207, "step": 121890 }, { "epoch": 0.31, "learning_rate": 6.958880668377875e-05, "loss": 1.0204, "step": 121895 }, { "epoch": 0.31, "learning_rate": 6.958754844229704e-05, "loss": 1.0215, "step": 121900 }, { "epoch": 0.31, "learning_rate": 6.958629020081534e-05, "loss": 1.0228, "step": 121905 }, { "epoch": 0.31, "learning_rate": 6.958503195933364e-05, "loss": 1.0216, "step": 121910 }, { "epoch": 0.31, "learning_rate": 6.958377371785193e-05, "loss": 1.0208, "step": 121915 }, { "epoch": 0.31, "learning_rate": 6.958251547637022e-05, "loss": 1.0221, "step": 121920 }, { "epoch": 0.31, "learning_rate": 6.958125723488852e-05, "loss": 1.0202, "step": 121925 }, { "epoch": 0.31, "learning_rate": 6.957999899340682e-05, "loss": 1.0224, "step": 121930 }, { "epoch": 0.31, "learning_rate": 6.957874075192511e-05, "loss": 1.0186, "step": 121935 }, { "epoch": 0.31, "learning_rate": 6.95774825104434e-05, "loss": 1.0218, "step": 121940 }, { "epoch": 0.31, "learning_rate": 6.95762242689617e-05, "loss": 1.0194, "step": 121945 }, { "epoch": 0.31, "learning_rate": 6.957496602748e-05, "loss": 1.0242, "step": 121950 }, { "epoch": 0.31, "learning_rate": 6.957370778599829e-05, "loss": 1.0464, "step": 121955 }, { "epoch": 0.31, "learning_rate": 6.957244954451658e-05, "loss": 1.0212, "step": 121960 }, { "epoch": 0.31, "learning_rate": 6.957119130303487e-05, "loss": 1.0234, "step": 121965 }, { "epoch": 0.31, "learning_rate": 6.956993306155318e-05, "loss": 1.023, "step": 121970 }, { "epoch": 0.31, "learning_rate": 6.956867482007147e-05, "loss": 1.0205, "step": 121975 }, { "epoch": 0.31, "learning_rate": 6.956741657858976e-05, "loss": 1.0222, "step": 121980 }, { "epoch": 0.31, "learning_rate": 6.956615833710805e-05, "loss": 1.0209, "step": 121985 }, { "epoch": 0.31, "learning_rate": 6.956490009562636e-05, "loss": 1.0212, "step": 121990 }, { "epoch": 0.31, "learning_rate": 6.956364185414465e-05, "loss": 1.0227, "step": 121995 }, { "epoch": 0.31, "learning_rate": 6.956238361266294e-05, "loss": 1.0198, "step": 122000 }, { "epoch": 0.31, "learning_rate": 6.956112537118123e-05, "loss": 1.023, "step": 122005 }, { "epoch": 0.31, "learning_rate": 6.955986712969954e-05, "loss": 1.0224, "step": 122010 }, { "epoch": 0.31, "learning_rate": 6.955860888821784e-05, "loss": 1.0228, "step": 122015 }, { "epoch": 0.31, "learning_rate": 6.955735064673613e-05, "loss": 1.0261, "step": 122020 }, { "epoch": 0.31, "learning_rate": 6.955609240525442e-05, "loss": 1.0224, "step": 122025 }, { "epoch": 0.31, "learning_rate": 6.955483416377272e-05, "loss": 1.0208, "step": 122030 }, { "epoch": 0.31, "learning_rate": 6.955357592229102e-05, "loss": 1.0204, "step": 122035 }, { "epoch": 0.31, "learning_rate": 6.955231768080931e-05, "loss": 1.0212, "step": 122040 }, { "epoch": 0.31, "learning_rate": 6.95510594393276e-05, "loss": 1.022, "step": 122045 }, { "epoch": 0.31, "learning_rate": 6.95498011978459e-05, "loss": 1.0238, "step": 122050 }, { "epoch": 0.31, "learning_rate": 6.954854295636419e-05, "loss": 1.0439, "step": 122055 }, { "epoch": 0.31, "learning_rate": 6.954728471488249e-05, "loss": 1.0216, "step": 122060 }, { "epoch": 0.31, "learning_rate": 6.954602647340078e-05, "loss": 1.0213, "step": 122065 }, { "epoch": 0.31, "learning_rate": 6.954476823191908e-05, "loss": 1.0219, "step": 122070 }, { "epoch": 0.31, "learning_rate": 6.954350999043737e-05, "loss": 1.0225, "step": 122075 }, { "epoch": 0.31, "learning_rate": 6.954225174895567e-05, "loss": 1.023, "step": 122080 }, { "epoch": 0.31, "learning_rate": 6.954099350747396e-05, "loss": 1.0232, "step": 122085 }, { "epoch": 0.31, "learning_rate": 6.953973526599226e-05, "loss": 1.0228, "step": 122090 }, { "epoch": 0.31, "learning_rate": 6.953847702451055e-05, "loss": 1.0209, "step": 122095 }, { "epoch": 0.31, "learning_rate": 6.953721878302885e-05, "loss": 1.0217, "step": 122100 }, { "epoch": 0.31, "learning_rate": 6.953596054154714e-05, "loss": 1.0223, "step": 122105 }, { "epoch": 0.31, "learning_rate": 6.953470230006544e-05, "loss": 1.0199, "step": 122110 }, { "epoch": 0.31, "learning_rate": 6.953344405858373e-05, "loss": 1.0211, "step": 122115 }, { "epoch": 0.31, "learning_rate": 6.953218581710202e-05, "loss": 1.0238, "step": 122120 }, { "epoch": 0.31, "learning_rate": 6.953092757562032e-05, "loss": 1.0212, "step": 122125 }, { "epoch": 0.31, "learning_rate": 6.952966933413862e-05, "loss": 1.0237, "step": 122130 }, { "epoch": 0.31, "learning_rate": 6.952841109265691e-05, "loss": 1.0223, "step": 122135 }, { "epoch": 0.31, "learning_rate": 6.95271528511752e-05, "loss": 1.0224, "step": 122140 }, { "epoch": 0.31, "learning_rate": 6.95258946096935e-05, "loss": 1.0201, "step": 122145 }, { "epoch": 0.31, "learning_rate": 6.95246363682118e-05, "loss": 1.0219, "step": 122150 }, { "epoch": 0.31, "learning_rate": 6.952337812673009e-05, "loss": 1.0199, "step": 122155 }, { "epoch": 0.31, "learning_rate": 6.952211988524838e-05, "loss": 1.0207, "step": 122160 }, { "epoch": 0.31, "learning_rate": 6.952086164376668e-05, "loss": 1.0207, "step": 122165 }, { "epoch": 0.31, "learning_rate": 6.951960340228498e-05, "loss": 1.0235, "step": 122170 }, { "epoch": 0.31, "learning_rate": 6.951834516080327e-05, "loss": 1.0238, "step": 122175 }, { "epoch": 0.31, "learning_rate": 6.951708691932156e-05, "loss": 1.0215, "step": 122180 }, { "epoch": 0.31, "learning_rate": 6.951582867783985e-05, "loss": 1.0223, "step": 122185 }, { "epoch": 0.31, "learning_rate": 6.951457043635816e-05, "loss": 1.0226, "step": 122190 }, { "epoch": 0.31, "learning_rate": 6.951331219487645e-05, "loss": 1.022, "step": 122195 }, { "epoch": 0.31, "learning_rate": 6.951205395339474e-05, "loss": 1.0192, "step": 122200 }, { "epoch": 0.31, "learning_rate": 6.951079571191303e-05, "loss": 1.02, "step": 122205 }, { "epoch": 0.31, "learning_rate": 6.950953747043134e-05, "loss": 1.0213, "step": 122210 }, { "epoch": 0.31, "learning_rate": 6.950827922894963e-05, "loss": 1.023, "step": 122215 }, { "epoch": 0.31, "learning_rate": 6.950702098746792e-05, "loss": 1.0237, "step": 122220 }, { "epoch": 0.31, "learning_rate": 6.950576274598621e-05, "loss": 1.0226, "step": 122225 }, { "epoch": 0.31, "learning_rate": 6.950450450450452e-05, "loss": 1.0237, "step": 122230 }, { "epoch": 0.31, "learning_rate": 6.95032462630228e-05, "loss": 1.0214, "step": 122235 }, { "epoch": 0.31, "learning_rate": 6.95019880215411e-05, "loss": 1.0215, "step": 122240 }, { "epoch": 0.31, "learning_rate": 6.950072978005939e-05, "loss": 1.0209, "step": 122245 }, { "epoch": 0.31, "learning_rate": 6.949947153857768e-05, "loss": 1.0224, "step": 122250 }, { "epoch": 0.31, "learning_rate": 6.949821329709599e-05, "loss": 1.0227, "step": 122255 }, { "epoch": 0.31, "learning_rate": 6.949695505561428e-05, "loss": 1.0219, "step": 122260 }, { "epoch": 0.31, "learning_rate": 6.949569681413257e-05, "loss": 1.0216, "step": 122265 }, { "epoch": 0.31, "learning_rate": 6.949443857265086e-05, "loss": 1.0233, "step": 122270 }, { "epoch": 0.31, "learning_rate": 6.949318033116917e-05, "loss": 1.0218, "step": 122275 }, { "epoch": 0.31, "learning_rate": 6.949192208968746e-05, "loss": 1.0205, "step": 122280 }, { "epoch": 0.31, "learning_rate": 6.949066384820575e-05, "loss": 1.0184, "step": 122285 }, { "epoch": 0.31, "learning_rate": 6.948940560672404e-05, "loss": 1.0217, "step": 122290 }, { "epoch": 0.31, "learning_rate": 6.948814736524235e-05, "loss": 1.0219, "step": 122295 }, { "epoch": 0.31, "learning_rate": 6.948688912376064e-05, "loss": 1.0232, "step": 122300 }, { "epoch": 0.31, "learning_rate": 6.948563088227893e-05, "loss": 1.0235, "step": 122305 }, { "epoch": 0.31, "learning_rate": 6.948437264079722e-05, "loss": 1.0212, "step": 122310 }, { "epoch": 0.31, "learning_rate": 6.948311439931551e-05, "loss": 1.0189, "step": 122315 }, { "epoch": 0.31, "learning_rate": 6.948185615783382e-05, "loss": 1.0221, "step": 122320 }, { "epoch": 0.31, "learning_rate": 6.948059791635211e-05, "loss": 1.02, "step": 122325 }, { "epoch": 0.31, "learning_rate": 6.94793396748704e-05, "loss": 1.0214, "step": 122330 }, { "epoch": 0.31, "learning_rate": 6.947808143338869e-05, "loss": 1.0242, "step": 122335 }, { "epoch": 0.31, "learning_rate": 6.9476823191907e-05, "loss": 1.0201, "step": 122340 }, { "epoch": 0.31, "learning_rate": 6.947556495042529e-05, "loss": 1.0213, "step": 122345 }, { "epoch": 0.31, "learning_rate": 6.947430670894358e-05, "loss": 1.0211, "step": 122350 }, { "epoch": 0.31, "learning_rate": 6.947304846746187e-05, "loss": 1.0211, "step": 122355 }, { "epoch": 0.31, "learning_rate": 6.947179022598016e-05, "loss": 1.0223, "step": 122360 }, { "epoch": 0.31, "learning_rate": 6.947053198449847e-05, "loss": 1.0222, "step": 122365 }, { "epoch": 0.31, "learning_rate": 6.946927374301676e-05, "loss": 1.0214, "step": 122370 }, { "epoch": 0.31, "learning_rate": 6.946801550153505e-05, "loss": 1.0225, "step": 122375 }, { "epoch": 0.31, "learning_rate": 6.946675726005334e-05, "loss": 1.0222, "step": 122380 }, { "epoch": 0.31, "learning_rate": 6.946549901857165e-05, "loss": 1.0453, "step": 122385 }, { "epoch": 0.31, "learning_rate": 6.946424077708994e-05, "loss": 1.0232, "step": 122390 }, { "epoch": 0.31, "learning_rate": 6.946298253560823e-05, "loss": 1.0198, "step": 122395 }, { "epoch": 0.31, "learning_rate": 6.946172429412652e-05, "loss": 1.0189, "step": 122400 }, { "epoch": 0.31, "learning_rate": 6.946046605264483e-05, "loss": 1.0246, "step": 122405 }, { "epoch": 0.31, "learning_rate": 6.945920781116312e-05, "loss": 1.0224, "step": 122410 }, { "epoch": 0.31, "learning_rate": 6.945794956968141e-05, "loss": 1.0214, "step": 122415 }, { "epoch": 0.31, "learning_rate": 6.94566913281997e-05, "loss": 1.0216, "step": 122420 }, { "epoch": 0.31, "learning_rate": 6.9455433086718e-05, "loss": 1.0224, "step": 122425 }, { "epoch": 0.31, "learning_rate": 6.94541748452363e-05, "loss": 1.0216, "step": 122430 }, { "epoch": 0.31, "learning_rate": 6.945291660375459e-05, "loss": 1.0238, "step": 122435 }, { "epoch": 0.31, "learning_rate": 6.945165836227288e-05, "loss": 1.0226, "step": 122440 }, { "epoch": 0.31, "learning_rate": 6.945040012079118e-05, "loss": 1.0204, "step": 122445 }, { "epoch": 0.31, "learning_rate": 6.944914187930948e-05, "loss": 1.02, "step": 122450 }, { "epoch": 0.31, "learning_rate": 6.944788363782777e-05, "loss": 1.02, "step": 122455 }, { "epoch": 0.31, "learning_rate": 6.944662539634606e-05, "loss": 1.022, "step": 122460 }, { "epoch": 0.31, "learning_rate": 6.944536715486436e-05, "loss": 1.0228, "step": 122465 }, { "epoch": 0.31, "learning_rate": 6.944410891338266e-05, "loss": 1.0202, "step": 122470 }, { "epoch": 0.31, "learning_rate": 6.944285067190095e-05, "loss": 1.0203, "step": 122475 }, { "epoch": 0.31, "learning_rate": 6.944159243041924e-05, "loss": 1.0204, "step": 122480 }, { "epoch": 0.31, "learning_rate": 6.944033418893754e-05, "loss": 1.0218, "step": 122485 }, { "epoch": 0.31, "learning_rate": 6.943907594745583e-05, "loss": 1.0227, "step": 122490 }, { "epoch": 0.31, "learning_rate": 6.943781770597413e-05, "loss": 1.0228, "step": 122495 }, { "epoch": 0.31, "learning_rate": 6.943655946449242e-05, "loss": 1.0215, "step": 122500 }, { "epoch": 0.31, "learning_rate": 6.943530122301072e-05, "loss": 1.0215, "step": 122505 }, { "epoch": 0.31, "learning_rate": 6.943404298152902e-05, "loss": 1.0216, "step": 122510 }, { "epoch": 0.31, "learning_rate": 6.943278474004731e-05, "loss": 1.0209, "step": 122515 }, { "epoch": 0.31, "learning_rate": 6.943152649856562e-05, "loss": 1.0221, "step": 122520 }, { "epoch": 0.31, "learning_rate": 6.943026825708391e-05, "loss": 1.0215, "step": 122525 }, { "epoch": 0.31, "learning_rate": 6.94290100156022e-05, "loss": 1.0237, "step": 122530 }, { "epoch": 0.31, "learning_rate": 6.942775177412049e-05, "loss": 1.0219, "step": 122535 }, { "epoch": 0.31, "learning_rate": 6.94264935326388e-05, "loss": 1.0219, "step": 122540 }, { "epoch": 0.31, "learning_rate": 6.942523529115709e-05, "loss": 1.0237, "step": 122545 }, { "epoch": 0.31, "learning_rate": 6.942397704967538e-05, "loss": 1.0239, "step": 122550 }, { "epoch": 0.31, "learning_rate": 6.942271880819367e-05, "loss": 1.0228, "step": 122555 }, { "epoch": 0.31, "learning_rate": 6.942146056671198e-05, "loss": 1.0228, "step": 122560 }, { "epoch": 0.31, "learning_rate": 6.942020232523027e-05, "loss": 1.0214, "step": 122565 }, { "epoch": 0.31, "learning_rate": 6.941894408374856e-05, "loss": 1.0206, "step": 122570 }, { "epoch": 0.31, "learning_rate": 6.941768584226685e-05, "loss": 1.0212, "step": 122575 }, { "epoch": 0.31, "learning_rate": 6.941642760078514e-05, "loss": 1.0223, "step": 122580 }, { "epoch": 0.31, "learning_rate": 6.941516935930345e-05, "loss": 1.0206, "step": 122585 }, { "epoch": 0.31, "learning_rate": 6.941391111782174e-05, "loss": 1.022, "step": 122590 }, { "epoch": 0.31, "learning_rate": 6.941265287634003e-05, "loss": 1.0222, "step": 122595 }, { "epoch": 0.31, "learning_rate": 6.941139463485832e-05, "loss": 1.0206, "step": 122600 }, { "epoch": 0.31, "learning_rate": 6.941013639337663e-05, "loss": 1.0219, "step": 122605 }, { "epoch": 0.31, "learning_rate": 6.940887815189492e-05, "loss": 1.0225, "step": 122610 }, { "epoch": 0.31, "learning_rate": 6.940761991041321e-05, "loss": 1.0202, "step": 122615 }, { "epoch": 0.31, "learning_rate": 6.94063616689315e-05, "loss": 1.0213, "step": 122620 }, { "epoch": 0.31, "learning_rate": 6.940510342744981e-05, "loss": 1.0229, "step": 122625 }, { "epoch": 0.31, "learning_rate": 6.94038451859681e-05, "loss": 1.0225, "step": 122630 }, { "epoch": 0.31, "learning_rate": 6.940258694448639e-05, "loss": 1.0211, "step": 122635 }, { "epoch": 0.31, "learning_rate": 6.940132870300468e-05, "loss": 1.0442, "step": 122640 }, { "epoch": 0.31, "learning_rate": 6.940007046152298e-05, "loss": 1.0227, "step": 122645 }, { "epoch": 0.31, "learning_rate": 6.939881222004128e-05, "loss": 1.0211, "step": 122650 }, { "epoch": 0.31, "learning_rate": 6.939755397855957e-05, "loss": 1.0251, "step": 122655 }, { "epoch": 0.31, "learning_rate": 6.939629573707786e-05, "loss": 1.0197, "step": 122660 }, { "epoch": 0.31, "learning_rate": 6.939503749559616e-05, "loss": 1.0221, "step": 122665 }, { "epoch": 0.31, "learning_rate": 6.939377925411446e-05, "loss": 1.0222, "step": 122670 }, { "epoch": 0.31, "learning_rate": 6.939252101263275e-05, "loss": 1.0189, "step": 122675 }, { "epoch": 0.31, "learning_rate": 6.939126277115104e-05, "loss": 1.0192, "step": 122680 }, { "epoch": 0.31, "learning_rate": 6.939000452966934e-05, "loss": 1.0213, "step": 122685 }, { "epoch": 0.31, "learning_rate": 6.938874628818764e-05, "loss": 1.021, "step": 122690 }, { "epoch": 0.31, "learning_rate": 6.938748804670593e-05, "loss": 1.0227, "step": 122695 }, { "epoch": 0.31, "learning_rate": 6.938622980522422e-05, "loss": 1.0231, "step": 122700 }, { "epoch": 0.31, "learning_rate": 6.938497156374252e-05, "loss": 1.0196, "step": 122705 }, { "epoch": 0.31, "learning_rate": 6.93837133222608e-05, "loss": 1.0211, "step": 122710 }, { "epoch": 0.31, "learning_rate": 6.938245508077911e-05, "loss": 1.0222, "step": 122715 }, { "epoch": 0.31, "learning_rate": 6.93811968392974e-05, "loss": 1.0217, "step": 122720 }, { "epoch": 0.31, "learning_rate": 6.93799385978157e-05, "loss": 1.0194, "step": 122725 }, { "epoch": 0.31, "learning_rate": 6.937868035633399e-05, "loss": 1.0206, "step": 122730 }, { "epoch": 0.31, "learning_rate": 6.937742211485229e-05, "loss": 1.0215, "step": 122735 }, { "epoch": 0.31, "learning_rate": 6.937616387337058e-05, "loss": 1.02, "step": 122740 }, { "epoch": 0.31, "learning_rate": 6.937490563188887e-05, "loss": 1.0192, "step": 122745 }, { "epoch": 0.31, "learning_rate": 6.937364739040717e-05, "loss": 1.0198, "step": 122750 }, { "epoch": 0.31, "learning_rate": 6.937238914892547e-05, "loss": 1.0201, "step": 122755 }, { "epoch": 0.31, "learning_rate": 6.937113090744376e-05, "loss": 1.0203, "step": 122760 }, { "epoch": 0.31, "learning_rate": 6.936987266596205e-05, "loss": 1.0212, "step": 122765 }, { "epoch": 0.31, "learning_rate": 6.936861442448035e-05, "loss": 1.0215, "step": 122770 }, { "epoch": 0.31, "learning_rate": 6.936735618299864e-05, "loss": 1.0213, "step": 122775 }, { "epoch": 0.31, "learning_rate": 6.936609794151694e-05, "loss": 1.0212, "step": 122780 }, { "epoch": 0.31, "learning_rate": 6.936483970003523e-05, "loss": 1.0178, "step": 122785 }, { "epoch": 0.31, "learning_rate": 6.936358145855353e-05, "loss": 1.0216, "step": 122790 }, { "epoch": 0.31, "learning_rate": 6.936232321707182e-05, "loss": 1.022, "step": 122795 }, { "epoch": 0.31, "learning_rate": 6.936106497559012e-05, "loss": 1.023, "step": 122800 }, { "epoch": 0.31, "learning_rate": 6.935980673410841e-05, "loss": 1.0221, "step": 122805 }, { "epoch": 0.31, "learning_rate": 6.93585484926267e-05, "loss": 1.0213, "step": 122810 }, { "epoch": 0.31, "learning_rate": 6.9357290251145e-05, "loss": 1.0233, "step": 122815 }, { "epoch": 0.31, "learning_rate": 6.93560320096633e-05, "loss": 1.021, "step": 122820 }, { "epoch": 0.31, "learning_rate": 6.93547737681816e-05, "loss": 1.0233, "step": 122825 }, { "epoch": 0.31, "learning_rate": 6.935351552669989e-05, "loss": 1.0212, "step": 122830 }, { "epoch": 0.31, "learning_rate": 6.935225728521818e-05, "loss": 1.0212, "step": 122835 }, { "epoch": 0.31, "learning_rate": 6.935099904373647e-05, "loss": 1.0229, "step": 122840 }, { "epoch": 0.31, "learning_rate": 6.934974080225477e-05, "loss": 1.0213, "step": 122845 }, { "epoch": 0.31, "learning_rate": 6.934848256077307e-05, "loss": 1.0238, "step": 122850 }, { "epoch": 0.31, "learning_rate": 6.934722431929136e-05, "loss": 1.0223, "step": 122855 }, { "epoch": 0.31, "learning_rate": 6.934596607780965e-05, "loss": 1.0214, "step": 122860 }, { "epoch": 0.31, "learning_rate": 6.934470783632795e-05, "loss": 1.0215, "step": 122865 }, { "epoch": 0.31, "learning_rate": 6.934344959484625e-05, "loss": 1.0213, "step": 122870 }, { "epoch": 0.31, "learning_rate": 6.934219135336454e-05, "loss": 1.0223, "step": 122875 }, { "epoch": 0.31, "learning_rate": 6.934093311188283e-05, "loss": 1.0225, "step": 122880 }, { "epoch": 0.31, "learning_rate": 6.933967487040113e-05, "loss": 1.0228, "step": 122885 }, { "epoch": 0.31, "learning_rate": 6.933841662891943e-05, "loss": 1.0449, "step": 122890 }, { "epoch": 0.31, "learning_rate": 6.933715838743772e-05, "loss": 1.0232, "step": 122895 }, { "epoch": 0.31, "learning_rate": 6.933590014595601e-05, "loss": 1.0208, "step": 122900 }, { "epoch": 0.31, "learning_rate": 6.93346419044743e-05, "loss": 1.0206, "step": 122905 }, { "epoch": 0.31, "learning_rate": 6.93333836629926e-05, "loss": 1.0223, "step": 122910 }, { "epoch": 0.31, "learning_rate": 6.93321254215109e-05, "loss": 1.0209, "step": 122915 }, { "epoch": 0.31, "learning_rate": 6.933086718002919e-05, "loss": 1.022, "step": 122920 }, { "epoch": 0.31, "learning_rate": 6.932960893854748e-05, "loss": 1.0177, "step": 122925 }, { "epoch": 0.31, "learning_rate": 6.932835069706579e-05, "loss": 1.0197, "step": 122930 }, { "epoch": 0.31, "learning_rate": 6.932709245558408e-05, "loss": 1.0211, "step": 122935 }, { "epoch": 0.31, "learning_rate": 6.932583421410237e-05, "loss": 1.0194, "step": 122940 }, { "epoch": 0.31, "learning_rate": 6.932457597262066e-05, "loss": 1.0205, "step": 122945 }, { "epoch": 0.31, "learning_rate": 6.932331773113897e-05, "loss": 1.022, "step": 122950 }, { "epoch": 0.31, "learning_rate": 6.932205948965726e-05, "loss": 1.0233, "step": 122955 }, { "epoch": 0.31, "learning_rate": 6.932080124817555e-05, "loss": 1.0172, "step": 122960 }, { "epoch": 0.31, "learning_rate": 6.931954300669384e-05, "loss": 1.0224, "step": 122965 }, { "epoch": 0.31, "learning_rate": 6.931828476521213e-05, "loss": 1.0204, "step": 122970 }, { "epoch": 0.31, "learning_rate": 6.931702652373044e-05, "loss": 1.0237, "step": 122975 }, { "epoch": 0.31, "learning_rate": 6.931576828224873e-05, "loss": 1.0199, "step": 122980 }, { "epoch": 0.31, "learning_rate": 6.931451004076702e-05, "loss": 1.02, "step": 122985 }, { "epoch": 0.31, "learning_rate": 6.931325179928531e-05, "loss": 1.0226, "step": 122990 }, { "epoch": 0.31, "learning_rate": 6.931199355780362e-05, "loss": 1.0209, "step": 122995 }, { "epoch": 0.31, "learning_rate": 6.931073531632191e-05, "loss": 1.0197, "step": 123000 }, { "epoch": 0.31, "learning_rate": 6.93094770748402e-05, "loss": 1.0224, "step": 123005 }, { "epoch": 0.31, "learning_rate": 6.93082188333585e-05, "loss": 1.0215, "step": 123010 }, { "epoch": 0.31, "learning_rate": 6.93069605918768e-05, "loss": 1.0201, "step": 123015 }, { "epoch": 0.31, "learning_rate": 6.93057023503951e-05, "loss": 1.0233, "step": 123020 }, { "epoch": 0.31, "learning_rate": 6.93044441089134e-05, "loss": 1.0225, "step": 123025 }, { "epoch": 0.31, "learning_rate": 6.930318586743169e-05, "loss": 1.0201, "step": 123030 }, { "epoch": 0.31, "learning_rate": 6.930192762594998e-05, "loss": 1.0225, "step": 123035 }, { "epoch": 0.31, "learning_rate": 6.930066938446827e-05, "loss": 1.0211, "step": 123040 }, { "epoch": 0.31, "learning_rate": 6.929941114298657e-05, "loss": 1.0202, "step": 123045 }, { "epoch": 0.31, "learning_rate": 6.929815290150487e-05, "loss": 1.0203, "step": 123050 }, { "epoch": 0.31, "learning_rate": 6.929689466002316e-05, "loss": 1.0236, "step": 123055 }, { "epoch": 0.31, "learning_rate": 6.929563641854145e-05, "loss": 1.0229, "step": 123060 }, { "epoch": 0.31, "learning_rate": 6.929437817705975e-05, "loss": 1.0212, "step": 123065 }, { "epoch": 0.31, "learning_rate": 6.929311993557805e-05, "loss": 1.021, "step": 123070 }, { "epoch": 0.31, "learning_rate": 6.929186169409634e-05, "loss": 1.022, "step": 123075 }, { "epoch": 0.31, "learning_rate": 6.929060345261463e-05, "loss": 1.0215, "step": 123080 }, { "epoch": 0.31, "learning_rate": 6.928934521113293e-05, "loss": 1.0217, "step": 123085 }, { "epoch": 0.31, "learning_rate": 6.928808696965123e-05, "loss": 1.0229, "step": 123090 }, { "epoch": 0.31, "learning_rate": 6.928682872816952e-05, "loss": 1.023, "step": 123095 }, { "epoch": 0.31, "learning_rate": 6.928557048668781e-05, "loss": 1.0228, "step": 123100 }, { "epoch": 0.31, "learning_rate": 6.92843122452061e-05, "loss": 1.0214, "step": 123105 }, { "epoch": 0.31, "learning_rate": 6.92830540037244e-05, "loss": 1.0205, "step": 123110 }, { "epoch": 0.31, "learning_rate": 6.92817957622427e-05, "loss": 1.0231, "step": 123115 }, { "epoch": 0.31, "learning_rate": 6.928053752076099e-05, "loss": 1.0214, "step": 123120 }, { "epoch": 0.31, "learning_rate": 6.927927927927928e-05, "loss": 1.043, "step": 123125 }, { "epoch": 0.31, "learning_rate": 6.927802103779759e-05, "loss": 1.0209, "step": 123130 }, { "epoch": 0.31, "learning_rate": 6.927676279631588e-05, "loss": 1.0218, "step": 123135 }, { "epoch": 0.31, "learning_rate": 6.927550455483417e-05, "loss": 1.0222, "step": 123140 }, { "epoch": 0.31, "learning_rate": 6.927424631335246e-05, "loss": 1.0212, "step": 123145 }, { "epoch": 0.31, "learning_rate": 6.927298807187077e-05, "loss": 1.0204, "step": 123150 }, { "epoch": 0.31, "learning_rate": 6.927172983038906e-05, "loss": 1.0239, "step": 123155 }, { "epoch": 0.31, "learning_rate": 6.927047158890735e-05, "loss": 1.022, "step": 123160 }, { "epoch": 0.31, "learning_rate": 6.926921334742564e-05, "loss": 1.0222, "step": 123165 }, { "epoch": 0.31, "learning_rate": 6.926795510594393e-05, "loss": 1.0221, "step": 123170 }, { "epoch": 0.31, "learning_rate": 6.926669686446224e-05, "loss": 1.0213, "step": 123175 }, { "epoch": 0.31, "learning_rate": 6.926543862298053e-05, "loss": 1.0232, "step": 123180 }, { "epoch": 0.31, "learning_rate": 6.926418038149882e-05, "loss": 1.0224, "step": 123185 }, { "epoch": 0.31, "learning_rate": 6.926292214001711e-05, "loss": 1.0254, "step": 123190 }, { "epoch": 0.31, "learning_rate": 6.926166389853542e-05, "loss": 1.0246, "step": 123195 }, { "epoch": 0.31, "learning_rate": 6.926040565705371e-05, "loss": 1.0196, "step": 123200 }, { "epoch": 0.31, "learning_rate": 6.9259147415572e-05, "loss": 1.0205, "step": 123205 }, { "epoch": 0.31, "learning_rate": 6.925788917409029e-05, "loss": 1.0224, "step": 123210 }, { "epoch": 0.31, "learning_rate": 6.92566309326086e-05, "loss": 1.02, "step": 123215 }, { "epoch": 0.31, "learning_rate": 6.925537269112689e-05, "loss": 1.0427, "step": 123220 }, { "epoch": 0.31, "learning_rate": 6.925411444964518e-05, "loss": 1.0221, "step": 123225 }, { "epoch": 0.31, "learning_rate": 6.925285620816347e-05, "loss": 1.0219, "step": 123230 }, { "epoch": 0.31, "learning_rate": 6.925159796668176e-05, "loss": 1.0211, "step": 123235 }, { "epoch": 0.31, "learning_rate": 6.925033972520007e-05, "loss": 1.0219, "step": 123240 }, { "epoch": 0.31, "learning_rate": 6.924908148371836e-05, "loss": 1.0236, "step": 123245 }, { "epoch": 0.31, "learning_rate": 6.924782324223665e-05, "loss": 1.0234, "step": 123250 }, { "epoch": 0.31, "learning_rate": 6.924656500075494e-05, "loss": 1.0211, "step": 123255 }, { "epoch": 0.31, "learning_rate": 6.924530675927325e-05, "loss": 1.0202, "step": 123260 }, { "epoch": 0.31, "learning_rate": 6.924404851779154e-05, "loss": 1.0224, "step": 123265 }, { "epoch": 0.31, "learning_rate": 6.924279027630983e-05, "loss": 1.0205, "step": 123270 }, { "epoch": 0.31, "learning_rate": 6.924153203482812e-05, "loss": 1.021, "step": 123275 }, { "epoch": 0.31, "learning_rate": 6.924027379334643e-05, "loss": 1.0231, "step": 123280 }, { "epoch": 0.31, "learning_rate": 6.923901555186472e-05, "loss": 1.0232, "step": 123285 }, { "epoch": 0.31, "learning_rate": 6.923775731038301e-05, "loss": 1.0232, "step": 123290 }, { "epoch": 0.31, "learning_rate": 6.92364990689013e-05, "loss": 1.0208, "step": 123295 }, { "epoch": 0.31, "learning_rate": 6.92352408274196e-05, "loss": 1.0222, "step": 123300 }, { "epoch": 0.31, "learning_rate": 6.92339825859379e-05, "loss": 1.0209, "step": 123305 }, { "epoch": 0.31, "learning_rate": 6.923272434445619e-05, "loss": 1.0207, "step": 123310 }, { "epoch": 0.31, "learning_rate": 6.923146610297448e-05, "loss": 1.0233, "step": 123315 }, { "epoch": 0.31, "learning_rate": 6.923020786149277e-05, "loss": 1.0231, "step": 123320 }, { "epoch": 0.31, "learning_rate": 6.922894962001108e-05, "loss": 1.0214, "step": 123325 }, { "epoch": 0.31, "learning_rate": 6.922769137852937e-05, "loss": 1.0252, "step": 123330 }, { "epoch": 0.31, "learning_rate": 6.922643313704766e-05, "loss": 1.02, "step": 123335 }, { "epoch": 0.31, "learning_rate": 6.922517489556595e-05, "loss": 1.0246, "step": 123340 }, { "epoch": 0.31, "learning_rate": 6.922391665408426e-05, "loss": 1.0242, "step": 123345 }, { "epoch": 0.31, "learning_rate": 6.922265841260255e-05, "loss": 1.0221, "step": 123350 }, { "epoch": 0.31, "learning_rate": 6.922140017112084e-05, "loss": 1.021, "step": 123355 }, { "epoch": 0.31, "learning_rate": 6.922014192963913e-05, "loss": 1.0225, "step": 123360 }, { "epoch": 0.31, "learning_rate": 6.921888368815743e-05, "loss": 1.0184, "step": 123365 }, { "epoch": 0.31, "learning_rate": 6.921762544667573e-05, "loss": 1.0231, "step": 123370 }, { "epoch": 0.31, "learning_rate": 6.921636720519402e-05, "loss": 1.0217, "step": 123375 }, { "epoch": 0.31, "learning_rate": 6.921510896371231e-05, "loss": 1.0361, "step": 123380 }, { "epoch": 0.31, "learning_rate": 6.92138507222306e-05, "loss": 1.0203, "step": 123385 }, { "epoch": 0.31, "learning_rate": 6.921259248074891e-05, "loss": 1.0207, "step": 123390 }, { "epoch": 0.31, "learning_rate": 6.92113342392672e-05, "loss": 1.0203, "step": 123395 }, { "epoch": 0.31, "learning_rate": 6.92100759977855e-05, "loss": 1.0215, "step": 123400 }, { "epoch": 0.31, "learning_rate": 6.920881775630379e-05, "loss": 1.0228, "step": 123405 }, { "epoch": 0.31, "learning_rate": 6.920755951482209e-05, "loss": 1.0204, "step": 123410 }, { "epoch": 0.31, "learning_rate": 6.920630127334038e-05, "loss": 1.041, "step": 123415 }, { "epoch": 0.31, "learning_rate": 6.920504303185867e-05, "loss": 1.022, "step": 123420 }, { "epoch": 0.31, "learning_rate": 6.920378479037697e-05, "loss": 1.0207, "step": 123425 }, { "epoch": 0.31, "learning_rate": 6.920252654889526e-05, "loss": 1.0214, "step": 123430 }, { "epoch": 0.31, "learning_rate": 6.920126830741356e-05, "loss": 1.0236, "step": 123435 }, { "epoch": 0.31, "learning_rate": 6.920001006593185e-05, "loss": 1.0239, "step": 123440 }, { "epoch": 0.31, "learning_rate": 6.919875182445015e-05, "loss": 1.0215, "step": 123445 }, { "epoch": 0.31, "learning_rate": 6.919749358296844e-05, "loss": 1.0203, "step": 123450 }, { "epoch": 0.31, "learning_rate": 6.919623534148674e-05, "loss": 1.0191, "step": 123455 }, { "epoch": 0.31, "learning_rate": 6.919497710000503e-05, "loss": 1.0218, "step": 123460 }, { "epoch": 0.31, "learning_rate": 6.919371885852333e-05, "loss": 1.0218, "step": 123465 }, { "epoch": 0.31, "learning_rate": 6.919246061704162e-05, "loss": 1.0211, "step": 123470 }, { "epoch": 0.31, "learning_rate": 6.919120237555992e-05, "loss": 1.022, "step": 123475 }, { "epoch": 0.31, "learning_rate": 6.918994413407821e-05, "loss": 1.0233, "step": 123480 }, { "epoch": 0.31, "learning_rate": 6.91886858925965e-05, "loss": 1.0197, "step": 123485 }, { "epoch": 0.31, "learning_rate": 6.91874276511148e-05, "loss": 1.0235, "step": 123490 }, { "epoch": 0.31, "learning_rate": 6.918616940963309e-05, "loss": 1.0218, "step": 123495 }, { "epoch": 0.31, "learning_rate": 6.91849111681514e-05, "loss": 1.0215, "step": 123500 }, { "epoch": 0.31, "learning_rate": 6.918365292666969e-05, "loss": 1.0229, "step": 123505 }, { "epoch": 0.31, "learning_rate": 6.918239468518799e-05, "loss": 1.0452, "step": 123510 }, { "epoch": 0.31, "learning_rate": 6.918113644370628e-05, "loss": 1.0238, "step": 123515 }, { "epoch": 0.31, "learning_rate": 6.917987820222457e-05, "loss": 1.0227, "step": 123520 }, { "epoch": 0.31, "learning_rate": 6.917861996074288e-05, "loss": 1.0228, "step": 123525 }, { "epoch": 0.31, "learning_rate": 6.917736171926117e-05, "loss": 1.0219, "step": 123530 }, { "epoch": 0.31, "learning_rate": 6.917610347777946e-05, "loss": 1.0212, "step": 123535 }, { "epoch": 0.31, "learning_rate": 6.917484523629775e-05, "loss": 1.02, "step": 123540 }, { "epoch": 0.31, "learning_rate": 6.917358699481606e-05, "loss": 1.0206, "step": 123545 }, { "epoch": 0.31, "learning_rate": 6.917232875333435e-05, "loss": 1.0391, "step": 123550 }, { "epoch": 0.31, "learning_rate": 6.917107051185264e-05, "loss": 1.0239, "step": 123555 }, { "epoch": 0.31, "learning_rate": 6.916981227037093e-05, "loss": 1.0215, "step": 123560 }, { "epoch": 0.31, "learning_rate": 6.916855402888923e-05, "loss": 1.021, "step": 123565 }, { "epoch": 0.31, "learning_rate": 6.916729578740753e-05, "loss": 1.0206, "step": 123570 }, { "epoch": 0.31, "learning_rate": 6.916603754592582e-05, "loss": 1.0212, "step": 123575 }, { "epoch": 0.31, "learning_rate": 6.916477930444411e-05, "loss": 1.0222, "step": 123580 }, { "epoch": 0.31, "learning_rate": 6.91635210629624e-05, "loss": 1.023, "step": 123585 }, { "epoch": 0.31, "learning_rate": 6.916226282148071e-05, "loss": 1.021, "step": 123590 }, { "epoch": 0.31, "learning_rate": 6.9161004579999e-05, "loss": 1.0243, "step": 123595 }, { "epoch": 0.31, "learning_rate": 6.91597463385173e-05, "loss": 1.0203, "step": 123600 }, { "epoch": 0.31, "learning_rate": 6.915848809703559e-05, "loss": 1.02, "step": 123605 }, { "epoch": 0.31, "learning_rate": 6.915722985555389e-05, "loss": 1.0216, "step": 123610 }, { "epoch": 0.31, "learning_rate": 6.915597161407218e-05, "loss": 1.0235, "step": 123615 }, { "epoch": 0.31, "learning_rate": 6.915471337259047e-05, "loss": 1.0233, "step": 123620 }, { "epoch": 0.31, "learning_rate": 6.915345513110877e-05, "loss": 1.0203, "step": 123625 }, { "epoch": 0.31, "learning_rate": 6.915219688962706e-05, "loss": 1.0204, "step": 123630 }, { "epoch": 0.31, "learning_rate": 6.915093864814536e-05, "loss": 1.0207, "step": 123635 }, { "epoch": 0.31, "learning_rate": 6.914968040666365e-05, "loss": 1.0223, "step": 123640 }, { "epoch": 0.31, "learning_rate": 6.914842216518195e-05, "loss": 1.0229, "step": 123645 }, { "epoch": 0.31, "learning_rate": 6.914716392370024e-05, "loss": 1.0219, "step": 123650 }, { "epoch": 0.31, "learning_rate": 6.914590568221854e-05, "loss": 1.0214, "step": 123655 }, { "epoch": 0.31, "learning_rate": 6.914464744073683e-05, "loss": 1.0211, "step": 123660 }, { "epoch": 0.31, "learning_rate": 6.914338919925513e-05, "loss": 1.0233, "step": 123665 }, { "epoch": 0.31, "learning_rate": 6.914213095777342e-05, "loss": 1.0233, "step": 123670 }, { "epoch": 0.31, "learning_rate": 6.914087271629172e-05, "loss": 1.0195, "step": 123675 }, { "epoch": 0.31, "learning_rate": 6.913961447481001e-05, "loss": 1.0212, "step": 123680 }, { "epoch": 0.31, "learning_rate": 6.91383562333283e-05, "loss": 1.0216, "step": 123685 }, { "epoch": 0.31, "learning_rate": 6.91370979918466e-05, "loss": 1.0207, "step": 123690 }, { "epoch": 0.31, "learning_rate": 6.913583975036489e-05, "loss": 1.0221, "step": 123695 }, { "epoch": 0.31, "learning_rate": 6.91345815088832e-05, "loss": 1.0209, "step": 123700 }, { "epoch": 0.31, "learning_rate": 6.913332326740149e-05, "loss": 1.021, "step": 123705 }, { "epoch": 0.31, "learning_rate": 6.913206502591978e-05, "loss": 1.0233, "step": 123710 }, { "epoch": 0.31, "learning_rate": 6.913080678443807e-05, "loss": 1.0218, "step": 123715 }, { "epoch": 0.31, "learning_rate": 6.912954854295637e-05, "loss": 1.0212, "step": 123720 }, { "epoch": 0.31, "learning_rate": 6.912829030147467e-05, "loss": 1.0221, "step": 123725 }, { "epoch": 0.31, "learning_rate": 6.912703205999296e-05, "loss": 1.024, "step": 123730 }, { "epoch": 0.31, "learning_rate": 6.912577381851125e-05, "loss": 1.0227, "step": 123735 }, { "epoch": 0.31, "learning_rate": 6.912451557702955e-05, "loss": 1.0215, "step": 123740 }, { "epoch": 0.31, "learning_rate": 6.912325733554784e-05, "loss": 1.0198, "step": 123745 }, { "epoch": 0.31, "learning_rate": 6.912199909406614e-05, "loss": 1.0202, "step": 123750 }, { "epoch": 0.31, "learning_rate": 6.912074085258443e-05, "loss": 1.0215, "step": 123755 }, { "epoch": 0.31, "learning_rate": 6.911948261110272e-05, "loss": 1.0232, "step": 123760 }, { "epoch": 0.31, "learning_rate": 6.911822436962102e-05, "loss": 1.023, "step": 123765 }, { "epoch": 0.31, "learning_rate": 6.911696612813932e-05, "loss": 1.0217, "step": 123770 }, { "epoch": 0.31, "learning_rate": 6.911570788665761e-05, "loss": 1.0205, "step": 123775 }, { "epoch": 0.31, "learning_rate": 6.91144496451759e-05, "loss": 1.0211, "step": 123780 }, { "epoch": 0.31, "learning_rate": 6.91131914036942e-05, "loss": 1.0217, "step": 123785 }, { "epoch": 0.31, "learning_rate": 6.91119331622125e-05, "loss": 1.0218, "step": 123790 }, { "epoch": 0.31, "learning_rate": 6.911067492073079e-05, "loss": 1.0216, "step": 123795 }, { "epoch": 0.31, "learning_rate": 6.910941667924908e-05, "loss": 1.0221, "step": 123800 }, { "epoch": 0.31, "learning_rate": 6.910815843776738e-05, "loss": 1.0223, "step": 123805 }, { "epoch": 0.31, "learning_rate": 6.910690019628568e-05, "loss": 1.0221, "step": 123810 }, { "epoch": 0.31, "learning_rate": 6.910564195480397e-05, "loss": 1.0221, "step": 123815 }, { "epoch": 0.31, "learning_rate": 6.910438371332226e-05, "loss": 1.0224, "step": 123820 }, { "epoch": 0.31, "learning_rate": 6.910312547184055e-05, "loss": 1.023, "step": 123825 }, { "epoch": 0.31, "learning_rate": 6.910186723035886e-05, "loss": 1.022, "step": 123830 }, { "epoch": 0.31, "learning_rate": 6.910060898887715e-05, "loss": 1.0206, "step": 123835 }, { "epoch": 0.31, "learning_rate": 6.909935074739544e-05, "loss": 1.0227, "step": 123840 }, { "epoch": 0.31, "learning_rate": 6.909809250591373e-05, "loss": 1.0226, "step": 123845 }, { "epoch": 0.31, "learning_rate": 6.909683426443204e-05, "loss": 1.0227, "step": 123850 }, { "epoch": 0.31, "learning_rate": 6.909557602295033e-05, "loss": 1.0191, "step": 123855 }, { "epoch": 0.31, "learning_rate": 6.909431778146862e-05, "loss": 1.0198, "step": 123860 }, { "epoch": 0.31, "learning_rate": 6.909305953998691e-05, "loss": 1.0195, "step": 123865 }, { "epoch": 0.31, "learning_rate": 6.909180129850522e-05, "loss": 1.0221, "step": 123870 }, { "epoch": 0.31, "learning_rate": 6.909054305702351e-05, "loss": 1.0221, "step": 123875 }, { "epoch": 0.31, "learning_rate": 6.90892848155418e-05, "loss": 1.0227, "step": 123880 }, { "epoch": 0.31, "learning_rate": 6.908802657406009e-05, "loss": 1.024, "step": 123885 }, { "epoch": 0.31, "learning_rate": 6.908676833257838e-05, "loss": 1.0212, "step": 123890 }, { "epoch": 0.31, "learning_rate": 6.908551009109669e-05, "loss": 1.0223, "step": 123895 }, { "epoch": 0.31, "learning_rate": 6.908425184961498e-05, "loss": 1.0209, "step": 123900 }, { "epoch": 0.31, "learning_rate": 6.908299360813327e-05, "loss": 1.0239, "step": 123905 }, { "epoch": 0.31, "learning_rate": 6.908173536665156e-05, "loss": 1.0206, "step": 123910 }, { "epoch": 0.31, "learning_rate": 6.908047712516987e-05, "loss": 1.022, "step": 123915 }, { "epoch": 0.31, "learning_rate": 6.907921888368816e-05, "loss": 1.0223, "step": 123920 }, { "epoch": 0.31, "learning_rate": 6.907796064220645e-05, "loss": 1.0214, "step": 123925 }, { "epoch": 0.31, "learning_rate": 6.907670240072474e-05, "loss": 1.0217, "step": 123930 }, { "epoch": 0.31, "learning_rate": 6.907544415924305e-05, "loss": 1.0202, "step": 123935 }, { "epoch": 0.31, "learning_rate": 6.907418591776134e-05, "loss": 1.0218, "step": 123940 }, { "epoch": 0.31, "learning_rate": 6.907292767627963e-05, "loss": 1.0236, "step": 123945 }, { "epoch": 0.31, "learning_rate": 6.907166943479792e-05, "loss": 1.0246, "step": 123950 }, { "epoch": 0.31, "learning_rate": 6.907041119331621e-05, "loss": 1.0222, "step": 123955 }, { "epoch": 0.31, "learning_rate": 6.906915295183452e-05, "loss": 1.0412, "step": 123960 }, { "epoch": 0.31, "learning_rate": 6.906789471035281e-05, "loss": 1.0229, "step": 123965 }, { "epoch": 0.31, "learning_rate": 6.90666364688711e-05, "loss": 1.0213, "step": 123970 }, { "epoch": 0.31, "learning_rate": 6.90653782273894e-05, "loss": 1.021, "step": 123975 }, { "epoch": 0.31, "learning_rate": 6.90641199859077e-05, "loss": 1.0217, "step": 123980 }, { "epoch": 0.31, "learning_rate": 6.906286174442599e-05, "loss": 1.0232, "step": 123985 }, { "epoch": 0.31, "learning_rate": 6.906160350294428e-05, "loss": 1.0209, "step": 123990 }, { "epoch": 0.31, "learning_rate": 6.906034526146257e-05, "loss": 1.021, "step": 123995 }, { "epoch": 0.31, "learning_rate": 6.905908701998088e-05, "loss": 1.0194, "step": 124000 }, { "epoch": 0.31, "learning_rate": 6.905782877849917e-05, "loss": 1.0212, "step": 124005 }, { "epoch": 0.31, "learning_rate": 6.905657053701748e-05, "loss": 1.0206, "step": 124010 }, { "epoch": 0.31, "learning_rate": 6.905531229553577e-05, "loss": 1.0208, "step": 124015 }, { "epoch": 0.31, "learning_rate": 6.905405405405406e-05, "loss": 1.0225, "step": 124020 }, { "epoch": 0.31, "learning_rate": 6.905279581257235e-05, "loss": 1.0249, "step": 124025 }, { "epoch": 0.31, "learning_rate": 6.905153757109066e-05, "loss": 1.0206, "step": 124030 }, { "epoch": 0.31, "learning_rate": 6.905027932960895e-05, "loss": 1.0225, "step": 124035 }, { "epoch": 0.31, "learning_rate": 6.904902108812724e-05, "loss": 1.0221, "step": 124040 }, { "epoch": 0.31, "learning_rate": 6.904776284664553e-05, "loss": 1.0231, "step": 124045 }, { "epoch": 0.31, "learning_rate": 6.904650460516384e-05, "loss": 1.02, "step": 124050 }, { "epoch": 0.31, "learning_rate": 6.904524636368213e-05, "loss": 1.0204, "step": 124055 }, { "epoch": 0.31, "learning_rate": 6.904398812220042e-05, "loss": 1.0227, "step": 124060 }, { "epoch": 0.31, "learning_rate": 6.904272988071871e-05, "loss": 1.0229, "step": 124065 }, { "epoch": 0.31, "learning_rate": 6.904147163923702e-05, "loss": 1.0223, "step": 124070 }, { "epoch": 0.31, "learning_rate": 6.904021339775531e-05, "loss": 1.0213, "step": 124075 }, { "epoch": 0.31, "learning_rate": 6.90389551562736e-05, "loss": 1.0201, "step": 124080 }, { "epoch": 0.31, "learning_rate": 6.903769691479189e-05, "loss": 1.0345, "step": 124085 }, { "epoch": 0.31, "learning_rate": 6.903643867331018e-05, "loss": 1.0201, "step": 124090 }, { "epoch": 0.31, "learning_rate": 6.903518043182849e-05, "loss": 1.0204, "step": 124095 }, { "epoch": 0.31, "learning_rate": 6.903392219034678e-05, "loss": 1.0222, "step": 124100 }, { "epoch": 0.31, "learning_rate": 6.903266394886507e-05, "loss": 1.0219, "step": 124105 }, { "epoch": 0.31, "learning_rate": 6.903140570738336e-05, "loss": 1.0279, "step": 124110 }, { "epoch": 0.31, "learning_rate": 6.903014746590167e-05, "loss": 1.0317, "step": 124115 }, { "epoch": 0.31, "learning_rate": 6.902888922441996e-05, "loss": 1.0217, "step": 124120 }, { "epoch": 0.31, "learning_rate": 6.902763098293825e-05, "loss": 1.019, "step": 124125 }, { "epoch": 0.31, "learning_rate": 6.902637274145654e-05, "loss": 1.0206, "step": 124130 }, { "epoch": 0.31, "learning_rate": 6.902511449997485e-05, "loss": 1.0208, "step": 124135 }, { "epoch": 0.31, "learning_rate": 6.902385625849314e-05, "loss": 1.0209, "step": 124140 }, { "epoch": 0.31, "learning_rate": 6.902259801701143e-05, "loss": 1.0225, "step": 124145 }, { "epoch": 0.31, "learning_rate": 6.902133977552972e-05, "loss": 1.0235, "step": 124150 }, { "epoch": 0.31, "learning_rate": 6.902008153404801e-05, "loss": 1.0213, "step": 124155 }, { "epoch": 0.31, "learning_rate": 6.901882329256632e-05, "loss": 1.0224, "step": 124160 }, { "epoch": 0.31, "learning_rate": 6.901756505108461e-05, "loss": 1.0223, "step": 124165 }, { "epoch": 0.31, "learning_rate": 6.90163068096029e-05, "loss": 1.0183, "step": 124170 }, { "epoch": 0.31, "learning_rate": 6.90150485681212e-05, "loss": 1.0209, "step": 124175 }, { "epoch": 0.31, "learning_rate": 6.90137903266395e-05, "loss": 1.02, "step": 124180 }, { "epoch": 0.31, "learning_rate": 6.901253208515779e-05, "loss": 1.0226, "step": 124185 }, { "epoch": 0.31, "learning_rate": 6.901127384367608e-05, "loss": 1.0231, "step": 124190 }, { "epoch": 0.31, "learning_rate": 6.901001560219437e-05, "loss": 1.0221, "step": 124195 }, { "epoch": 0.31, "learning_rate": 6.900875736071268e-05, "loss": 1.0224, "step": 124200 }, { "epoch": 0.31, "learning_rate": 6.900749911923097e-05, "loss": 1.0223, "step": 124205 }, { "epoch": 0.31, "learning_rate": 6.900624087774926e-05, "loss": 1.0238, "step": 124210 }, { "epoch": 0.31, "learning_rate": 6.900498263626755e-05, "loss": 1.0204, "step": 124215 }, { "epoch": 0.31, "learning_rate": 6.900372439478584e-05, "loss": 1.021, "step": 124220 }, { "epoch": 0.31, "learning_rate": 6.900246615330415e-05, "loss": 1.0222, "step": 124225 }, { "epoch": 0.31, "learning_rate": 6.900120791182244e-05, "loss": 1.0174, "step": 124230 }, { "epoch": 0.31, "learning_rate": 6.899994967034073e-05, "loss": 1.0231, "step": 124235 }, { "epoch": 0.31, "learning_rate": 6.899869142885902e-05, "loss": 1.0202, "step": 124240 }, { "epoch": 0.31, "learning_rate": 6.899743318737733e-05, "loss": 1.0192, "step": 124245 }, { "epoch": 0.31, "learning_rate": 6.899617494589562e-05, "loss": 1.0248, "step": 124250 }, { "epoch": 0.31, "learning_rate": 6.899491670441391e-05, "loss": 1.0211, "step": 124255 }, { "epoch": 0.31, "learning_rate": 6.89936584629322e-05, "loss": 1.0209, "step": 124260 }, { "epoch": 0.31, "learning_rate": 6.899240022145051e-05, "loss": 1.0212, "step": 124265 }, { "epoch": 0.31, "learning_rate": 6.89911419799688e-05, "loss": 1.0217, "step": 124270 }, { "epoch": 0.31, "learning_rate": 6.898988373848709e-05, "loss": 1.021, "step": 124275 }, { "epoch": 0.31, "learning_rate": 6.898862549700538e-05, "loss": 1.0226, "step": 124280 }, { "epoch": 0.31, "learning_rate": 6.898736725552368e-05, "loss": 1.019, "step": 124285 }, { "epoch": 0.31, "learning_rate": 6.898610901404198e-05, "loss": 1.0214, "step": 124290 }, { "epoch": 0.31, "learning_rate": 6.898485077256027e-05, "loss": 1.0215, "step": 124295 }, { "epoch": 0.31, "learning_rate": 6.898359253107856e-05, "loss": 1.0221, "step": 124300 }, { "epoch": 0.31, "learning_rate": 6.898233428959686e-05, "loss": 1.0235, "step": 124305 }, { "epoch": 0.31, "learning_rate": 6.898107604811516e-05, "loss": 1.0223, "step": 124310 }, { "epoch": 0.31, "learning_rate": 6.897981780663345e-05, "loss": 1.0206, "step": 124315 }, { "epoch": 0.31, "learning_rate": 6.897855956515174e-05, "loss": 1.0196, "step": 124320 }, { "epoch": 0.31, "learning_rate": 6.897730132367004e-05, "loss": 1.0195, "step": 124325 }, { "epoch": 0.31, "learning_rate": 6.897604308218834e-05, "loss": 1.0221, "step": 124330 }, { "epoch": 0.31, "learning_rate": 6.897478484070663e-05, "loss": 1.0208, "step": 124335 }, { "epoch": 0.31, "learning_rate": 6.897352659922492e-05, "loss": 1.0211, "step": 124340 }, { "epoch": 0.31, "learning_rate": 6.897226835774322e-05, "loss": 1.0203, "step": 124345 }, { "epoch": 0.31, "learning_rate": 6.897101011626151e-05, "loss": 1.02, "step": 124350 }, { "epoch": 0.31, "learning_rate": 6.896975187477981e-05, "loss": 1.0232, "step": 124355 }, { "epoch": 0.31, "learning_rate": 6.89684936332981e-05, "loss": 1.0211, "step": 124360 }, { "epoch": 0.31, "learning_rate": 6.89672353918164e-05, "loss": 1.0238, "step": 124365 }, { "epoch": 0.31, "learning_rate": 6.896597715033469e-05, "loss": 1.0231, "step": 124370 }, { "epoch": 0.31, "learning_rate": 6.896471890885299e-05, "loss": 1.0214, "step": 124375 }, { "epoch": 0.31, "learning_rate": 6.896346066737128e-05, "loss": 1.0218, "step": 124380 }, { "epoch": 0.31, "learning_rate": 6.896220242588958e-05, "loss": 1.021, "step": 124385 }, { "epoch": 0.31, "learning_rate": 6.896094418440787e-05, "loss": 1.0228, "step": 124390 }, { "epoch": 0.31, "learning_rate": 6.895968594292617e-05, "loss": 1.0221, "step": 124395 }, { "epoch": 0.31, "learning_rate": 6.895842770144446e-05, "loss": 1.021, "step": 124400 }, { "epoch": 0.31, "learning_rate": 6.895716945996276e-05, "loss": 1.0208, "step": 124405 }, { "epoch": 0.31, "learning_rate": 6.895591121848105e-05, "loss": 1.0239, "step": 124410 }, { "epoch": 0.31, "learning_rate": 6.895465297699934e-05, "loss": 1.0206, "step": 124415 }, { "epoch": 0.31, "learning_rate": 6.895339473551764e-05, "loss": 1.0215, "step": 124420 }, { "epoch": 0.31, "learning_rate": 6.895213649403594e-05, "loss": 1.0214, "step": 124425 }, { "epoch": 0.31, "learning_rate": 6.895087825255423e-05, "loss": 1.0224, "step": 124430 }, { "epoch": 0.31, "learning_rate": 6.894962001107252e-05, "loss": 1.0207, "step": 124435 }, { "epoch": 0.31, "learning_rate": 6.894836176959082e-05, "loss": 1.0246, "step": 124440 }, { "epoch": 0.31, "learning_rate": 6.894710352810912e-05, "loss": 1.0223, "step": 124445 }, { "epoch": 0.31, "learning_rate": 6.894584528662741e-05, "loss": 1.0216, "step": 124450 }, { "epoch": 0.31, "learning_rate": 6.89445870451457e-05, "loss": 1.02, "step": 124455 }, { "epoch": 0.31, "learning_rate": 6.8943328803664e-05, "loss": 1.0224, "step": 124460 }, { "epoch": 0.31, "learning_rate": 6.89420705621823e-05, "loss": 1.0209, "step": 124465 }, { "epoch": 0.31, "learning_rate": 6.894081232070059e-05, "loss": 1.0212, "step": 124470 }, { "epoch": 0.31, "learning_rate": 6.893955407921888e-05, "loss": 1.0218, "step": 124475 }, { "epoch": 0.31, "learning_rate": 6.893829583773717e-05, "loss": 1.0222, "step": 124480 }, { "epoch": 0.31, "learning_rate": 6.893703759625548e-05, "loss": 1.0249, "step": 124485 }, { "epoch": 0.31, "learning_rate": 6.893577935477377e-05, "loss": 1.02, "step": 124490 }, { "epoch": 0.31, "learning_rate": 6.893452111329206e-05, "loss": 1.0214, "step": 124495 }, { "epoch": 0.31, "learning_rate": 6.893326287181035e-05, "loss": 1.023, "step": 124500 }, { "epoch": 0.31, "learning_rate": 6.893200463032866e-05, "loss": 1.0204, "step": 124505 }, { "epoch": 0.31, "learning_rate": 6.893074638884696e-05, "loss": 1.0228, "step": 124510 }, { "epoch": 0.31, "learning_rate": 6.892948814736525e-05, "loss": 1.0208, "step": 124515 }, { "epoch": 0.31, "learning_rate": 6.892822990588354e-05, "loss": 1.0193, "step": 124520 }, { "epoch": 0.31, "learning_rate": 6.892697166440184e-05, "loss": 1.0232, "step": 124525 }, { "epoch": 0.31, "learning_rate": 6.892571342292014e-05, "loss": 1.02, "step": 124530 }, { "epoch": 0.31, "learning_rate": 6.892445518143843e-05, "loss": 1.0226, "step": 124535 }, { "epoch": 0.31, "learning_rate": 6.892319693995672e-05, "loss": 1.0218, "step": 124540 }, { "epoch": 0.31, "learning_rate": 6.892193869847502e-05, "loss": 1.0203, "step": 124545 }, { "epoch": 0.31, "learning_rate": 6.892068045699332e-05, "loss": 1.024, "step": 124550 }, { "epoch": 0.31, "learning_rate": 6.891942221551161e-05, "loss": 1.0224, "step": 124555 }, { "epoch": 0.31, "learning_rate": 6.89181639740299e-05, "loss": 1.0187, "step": 124560 }, { "epoch": 0.31, "learning_rate": 6.89169057325482e-05, "loss": 1.0231, "step": 124565 }, { "epoch": 0.31, "learning_rate": 6.891564749106649e-05, "loss": 1.0224, "step": 124570 }, { "epoch": 0.31, "learning_rate": 6.891438924958479e-05, "loss": 1.0216, "step": 124575 }, { "epoch": 0.31, "learning_rate": 6.891313100810308e-05, "loss": 1.0243, "step": 124580 }, { "epoch": 0.31, "learning_rate": 6.891187276662138e-05, "loss": 1.0212, "step": 124585 }, { "epoch": 0.31, "learning_rate": 6.891061452513967e-05, "loss": 1.0215, "step": 124590 }, { "epoch": 0.31, "learning_rate": 6.890935628365797e-05, "loss": 1.0221, "step": 124595 }, { "epoch": 0.31, "learning_rate": 6.890809804217626e-05, "loss": 1.022, "step": 124600 }, { "epoch": 0.31, "learning_rate": 6.890683980069456e-05, "loss": 1.0231, "step": 124605 }, { "epoch": 0.31, "learning_rate": 6.890558155921285e-05, "loss": 1.022, "step": 124610 }, { "epoch": 0.31, "learning_rate": 6.890432331773115e-05, "loss": 1.022, "step": 124615 }, { "epoch": 0.31, "learning_rate": 6.890306507624944e-05, "loss": 1.0217, "step": 124620 }, { "epoch": 0.31, "learning_rate": 6.890180683476774e-05, "loss": 1.0218, "step": 124625 }, { "epoch": 0.31, "learning_rate": 6.890054859328603e-05, "loss": 1.0222, "step": 124630 }, { "epoch": 0.31, "learning_rate": 6.889929035180432e-05, "loss": 1.021, "step": 124635 }, { "epoch": 0.31, "learning_rate": 6.889803211032262e-05, "loss": 1.022, "step": 124640 }, { "epoch": 0.31, "learning_rate": 6.889677386884092e-05, "loss": 1.0236, "step": 124645 }, { "epoch": 0.31, "learning_rate": 6.889551562735921e-05, "loss": 1.0249, "step": 124650 }, { "epoch": 0.31, "learning_rate": 6.88942573858775e-05, "loss": 1.0207, "step": 124655 }, { "epoch": 0.31, "learning_rate": 6.88929991443958e-05, "loss": 1.0214, "step": 124660 }, { "epoch": 0.31, "learning_rate": 6.88917409029141e-05, "loss": 1.0224, "step": 124665 }, { "epoch": 0.31, "learning_rate": 6.889048266143239e-05, "loss": 1.0215, "step": 124670 }, { "epoch": 0.31, "learning_rate": 6.888922441995068e-05, "loss": 1.0201, "step": 124675 }, { "epoch": 0.31, "learning_rate": 6.888796617846897e-05, "loss": 1.0212, "step": 124680 }, { "epoch": 0.31, "learning_rate": 6.888670793698728e-05, "loss": 1.0205, "step": 124685 }, { "epoch": 0.31, "learning_rate": 6.888544969550557e-05, "loss": 1.0217, "step": 124690 }, { "epoch": 0.31, "learning_rate": 6.888419145402386e-05, "loss": 1.0216, "step": 124695 }, { "epoch": 0.31, "learning_rate": 6.888293321254215e-05, "loss": 1.0252, "step": 124700 }, { "epoch": 0.31, "learning_rate": 6.888167497106046e-05, "loss": 1.0215, "step": 124705 }, { "epoch": 0.31, "learning_rate": 6.888041672957875e-05, "loss": 1.0189, "step": 124710 }, { "epoch": 0.31, "learning_rate": 6.887915848809704e-05, "loss": 1.0229, "step": 124715 }, { "epoch": 0.31, "learning_rate": 6.887790024661533e-05, "loss": 1.0252, "step": 124720 }, { "epoch": 0.31, "learning_rate": 6.887664200513364e-05, "loss": 1.0221, "step": 124725 }, { "epoch": 0.31, "learning_rate": 6.887538376365193e-05, "loss": 1.0215, "step": 124730 }, { "epoch": 0.31, "learning_rate": 6.887412552217022e-05, "loss": 1.0217, "step": 124735 }, { "epoch": 0.31, "learning_rate": 6.887286728068851e-05, "loss": 1.0231, "step": 124740 }, { "epoch": 0.31, "learning_rate": 6.88716090392068e-05, "loss": 1.0219, "step": 124745 }, { "epoch": 0.31, "learning_rate": 6.88703507977251e-05, "loss": 1.0216, "step": 124750 }, { "epoch": 0.31, "learning_rate": 6.88690925562434e-05, "loss": 1.0443, "step": 124755 }, { "epoch": 0.31, "learning_rate": 6.886783431476169e-05, "loss": 1.0208, "step": 124760 }, { "epoch": 0.31, "learning_rate": 6.886657607327998e-05, "loss": 1.0212, "step": 124765 }, { "epoch": 0.31, "learning_rate": 6.886531783179829e-05, "loss": 1.0216, "step": 124770 }, { "epoch": 0.31, "learning_rate": 6.886405959031658e-05, "loss": 1.0235, "step": 124775 }, { "epoch": 0.31, "learning_rate": 6.886280134883487e-05, "loss": 1.0203, "step": 124780 }, { "epoch": 0.31, "learning_rate": 6.886154310735316e-05, "loss": 1.0209, "step": 124785 }, { "epoch": 0.31, "learning_rate": 6.886028486587147e-05, "loss": 1.0209, "step": 124790 }, { "epoch": 0.31, "learning_rate": 6.885902662438976e-05, "loss": 1.0205, "step": 124795 }, { "epoch": 0.31, "learning_rate": 6.885776838290805e-05, "loss": 1.0411, "step": 124800 }, { "epoch": 0.31, "learning_rate": 6.885651014142634e-05, "loss": 1.0225, "step": 124805 }, { "epoch": 0.31, "learning_rate": 6.885525189994463e-05, "loss": 1.0236, "step": 124810 }, { "epoch": 0.31, "learning_rate": 6.885399365846294e-05, "loss": 1.0208, "step": 124815 }, { "epoch": 0.31, "learning_rate": 6.885273541698123e-05, "loss": 1.0204, "step": 124820 }, { "epoch": 0.31, "learning_rate": 6.885147717549952e-05, "loss": 1.0234, "step": 124825 }, { "epoch": 0.31, "learning_rate": 6.885021893401781e-05, "loss": 1.0219, "step": 124830 }, { "epoch": 0.31, "learning_rate": 6.884896069253612e-05, "loss": 1.0203, "step": 124835 }, { "epoch": 0.31, "learning_rate": 6.884770245105441e-05, "loss": 1.0197, "step": 124840 }, { "epoch": 0.31, "learning_rate": 6.88464442095727e-05, "loss": 1.0203, "step": 124845 }, { "epoch": 0.31, "learning_rate": 6.884518596809099e-05, "loss": 1.0226, "step": 124850 }, { "epoch": 0.31, "learning_rate": 6.88439277266093e-05, "loss": 1.0203, "step": 124855 }, { "epoch": 0.31, "learning_rate": 6.884266948512759e-05, "loss": 1.0205, "step": 124860 }, { "epoch": 0.31, "learning_rate": 6.884141124364588e-05, "loss": 1.0466, "step": 124865 }, { "epoch": 0.31, "learning_rate": 6.884015300216417e-05, "loss": 1.0205, "step": 124870 }, { "epoch": 0.31, "learning_rate": 6.883889476068246e-05, "loss": 1.022, "step": 124875 }, { "epoch": 0.31, "learning_rate": 6.883763651920077e-05, "loss": 1.0239, "step": 124880 }, { "epoch": 0.31, "learning_rate": 6.883637827771906e-05, "loss": 1.0225, "step": 124885 }, { "epoch": 0.31, "learning_rate": 6.883512003623735e-05, "loss": 1.0248, "step": 124890 }, { "epoch": 0.31, "learning_rate": 6.883386179475564e-05, "loss": 1.0207, "step": 124895 }, { "epoch": 0.31, "learning_rate": 6.883260355327395e-05, "loss": 1.0208, "step": 124900 }, { "epoch": 0.31, "learning_rate": 6.883134531179224e-05, "loss": 1.0228, "step": 124905 }, { "epoch": 0.31, "learning_rate": 6.883008707031053e-05, "loss": 1.0238, "step": 124910 }, { "epoch": 0.31, "learning_rate": 6.882882882882882e-05, "loss": 1.0205, "step": 124915 }, { "epoch": 0.31, "learning_rate": 6.882757058734713e-05, "loss": 1.0204, "step": 124920 }, { "epoch": 0.31, "learning_rate": 6.882631234586542e-05, "loss": 1.0201, "step": 124925 }, { "epoch": 0.31, "learning_rate": 6.882505410438371e-05, "loss": 1.0203, "step": 124930 }, { "epoch": 0.31, "learning_rate": 6.8823795862902e-05, "loss": 1.0232, "step": 124935 }, { "epoch": 0.31, "learning_rate": 6.88225376214203e-05, "loss": 1.0225, "step": 124940 }, { "epoch": 0.31, "learning_rate": 6.88212793799386e-05, "loss": 1.0226, "step": 124945 }, { "epoch": 0.31, "learning_rate": 6.882002113845689e-05, "loss": 1.0207, "step": 124950 }, { "epoch": 0.31, "learning_rate": 6.881876289697518e-05, "loss": 1.0203, "step": 124955 }, { "epoch": 0.31, "learning_rate": 6.881750465549348e-05, "loss": 1.022, "step": 124960 }, { "epoch": 0.31, "learning_rate": 6.881624641401178e-05, "loss": 1.021, "step": 124965 }, { "epoch": 0.31, "learning_rate": 6.881498817253007e-05, "loss": 1.0208, "step": 124970 }, { "epoch": 0.31, "learning_rate": 6.881372993104836e-05, "loss": 1.0207, "step": 124975 }, { "epoch": 0.31, "learning_rate": 6.881247168956666e-05, "loss": 1.0224, "step": 124980 }, { "epoch": 0.31, "learning_rate": 6.881121344808496e-05, "loss": 1.0207, "step": 124985 }, { "epoch": 0.31, "learning_rate": 6.880995520660325e-05, "loss": 1.0217, "step": 124990 }, { "epoch": 0.31, "learning_rate": 6.880869696512154e-05, "loss": 1.0199, "step": 124995 }, { "epoch": 0.31, "learning_rate": 6.880743872363984e-05, "loss": 1.0238, "step": 125000 }, { "epoch": 0.31, "learning_rate": 6.880618048215814e-05, "loss": 1.0219, "step": 125005 }, { "epoch": 0.31, "learning_rate": 6.880492224067645e-05, "loss": 1.0209, "step": 125010 }, { "epoch": 0.31, "learning_rate": 6.880366399919474e-05, "loss": 1.0191, "step": 125015 }, { "epoch": 0.31, "learning_rate": 6.880240575771303e-05, "loss": 1.035, "step": 125020 }, { "epoch": 0.31, "learning_rate": 6.880114751623132e-05, "loss": 1.02, "step": 125025 }, { "epoch": 0.31, "learning_rate": 6.879988927474961e-05, "loss": 1.0209, "step": 125030 }, { "epoch": 0.31, "learning_rate": 6.879863103326792e-05, "loss": 1.0219, "step": 125035 }, { "epoch": 0.31, "learning_rate": 6.879737279178621e-05, "loss": 1.0233, "step": 125040 }, { "epoch": 0.31, "learning_rate": 6.87961145503045e-05, "loss": 1.0245, "step": 125045 }, { "epoch": 0.31, "learning_rate": 6.879485630882279e-05, "loss": 1.0206, "step": 125050 }, { "epoch": 0.31, "learning_rate": 6.87935980673411e-05, "loss": 1.0206, "step": 125055 }, { "epoch": 0.31, "learning_rate": 6.879233982585939e-05, "loss": 1.0217, "step": 125060 }, { "epoch": 0.31, "learning_rate": 6.879108158437768e-05, "loss": 1.021, "step": 125065 }, { "epoch": 0.31, "learning_rate": 6.878982334289597e-05, "loss": 1.0234, "step": 125070 }, { "epoch": 0.31, "learning_rate": 6.878856510141428e-05, "loss": 1.0225, "step": 125075 }, { "epoch": 0.31, "learning_rate": 6.878730685993257e-05, "loss": 1.0212, "step": 125080 }, { "epoch": 0.31, "learning_rate": 6.878604861845086e-05, "loss": 1.0217, "step": 125085 }, { "epoch": 0.31, "learning_rate": 6.878479037696915e-05, "loss": 1.0218, "step": 125090 }, { "epoch": 0.31, "learning_rate": 6.878353213548744e-05, "loss": 1.0215, "step": 125095 }, { "epoch": 0.31, "learning_rate": 6.878227389400575e-05, "loss": 1.0219, "step": 125100 }, { "epoch": 0.31, "learning_rate": 6.878101565252404e-05, "loss": 1.023, "step": 125105 }, { "epoch": 0.31, "learning_rate": 6.877975741104233e-05, "loss": 1.0215, "step": 125110 }, { "epoch": 0.31, "learning_rate": 6.877849916956062e-05, "loss": 1.0413, "step": 125115 }, { "epoch": 0.31, "learning_rate": 6.877724092807893e-05, "loss": 1.0221, "step": 125120 }, { "epoch": 0.31, "learning_rate": 6.877598268659722e-05, "loss": 1.0215, "step": 125125 }, { "epoch": 0.31, "learning_rate": 6.877472444511551e-05, "loss": 1.024, "step": 125130 }, { "epoch": 0.31, "learning_rate": 6.87734662036338e-05, "loss": 1.0229, "step": 125135 }, { "epoch": 0.31, "learning_rate": 6.877220796215211e-05, "loss": 1.0205, "step": 125140 }, { "epoch": 0.31, "learning_rate": 6.87709497206704e-05, "loss": 1.0187, "step": 125145 }, { "epoch": 0.31, "learning_rate": 6.876969147918869e-05, "loss": 1.0211, "step": 125150 }, { "epoch": 0.31, "learning_rate": 6.876843323770698e-05, "loss": 1.021, "step": 125155 }, { "epoch": 0.31, "learning_rate": 6.876717499622528e-05, "loss": 1.0198, "step": 125160 }, { "epoch": 0.31, "learning_rate": 6.876591675474358e-05, "loss": 1.0222, "step": 125165 }, { "epoch": 0.31, "learning_rate": 6.876465851326187e-05, "loss": 1.0221, "step": 125170 }, { "epoch": 0.31, "learning_rate": 6.876340027178016e-05, "loss": 1.021, "step": 125175 }, { "epoch": 0.31, "learning_rate": 6.876214203029845e-05, "loss": 1.0235, "step": 125180 }, { "epoch": 0.31, "learning_rate": 6.876088378881676e-05, "loss": 1.0199, "step": 125185 }, { "epoch": 0.31, "learning_rate": 6.875962554733505e-05, "loss": 1.022, "step": 125190 }, { "epoch": 0.31, "learning_rate": 6.875836730585334e-05, "loss": 1.0202, "step": 125195 }, { "epoch": 0.31, "learning_rate": 6.875710906437163e-05, "loss": 1.0202, "step": 125200 }, { "epoch": 0.31, "learning_rate": 6.875585082288994e-05, "loss": 1.0216, "step": 125205 }, { "epoch": 0.31, "learning_rate": 6.875459258140823e-05, "loss": 1.0181, "step": 125210 }, { "epoch": 0.31, "learning_rate": 6.875333433992652e-05, "loss": 1.0229, "step": 125215 }, { "epoch": 0.31, "learning_rate": 6.875207609844481e-05, "loss": 1.0198, "step": 125220 }, { "epoch": 0.31, "learning_rate": 6.87508178569631e-05, "loss": 1.0205, "step": 125225 }, { "epoch": 0.31, "learning_rate": 6.874955961548141e-05, "loss": 1.0206, "step": 125230 }, { "epoch": 0.31, "learning_rate": 6.87483013739997e-05, "loss": 1.0209, "step": 125235 }, { "epoch": 0.31, "learning_rate": 6.8747043132518e-05, "loss": 1.0205, "step": 125240 }, { "epoch": 0.31, "learning_rate": 6.874578489103629e-05, "loss": 1.0223, "step": 125245 }, { "epoch": 0.31, "learning_rate": 6.874452664955459e-05, "loss": 1.0234, "step": 125250 }, { "epoch": 0.31, "learning_rate": 6.874326840807288e-05, "loss": 1.021, "step": 125255 }, { "epoch": 0.31, "learning_rate": 6.874201016659117e-05, "loss": 1.0212, "step": 125260 }, { "epoch": 0.31, "learning_rate": 6.874075192510947e-05, "loss": 1.0208, "step": 125265 }, { "epoch": 0.31, "learning_rate": 6.873949368362777e-05, "loss": 1.0214, "step": 125270 }, { "epoch": 0.31, "learning_rate": 6.873823544214606e-05, "loss": 1.0207, "step": 125275 }, { "epoch": 0.31, "learning_rate": 6.873697720066435e-05, "loss": 1.0212, "step": 125280 }, { "epoch": 0.31, "learning_rate": 6.873571895918265e-05, "loss": 1.0223, "step": 125285 }, { "epoch": 0.31, "learning_rate": 6.873446071770094e-05, "loss": 1.0247, "step": 125290 }, { "epoch": 0.31, "learning_rate": 6.873320247621924e-05, "loss": 1.0213, "step": 125295 }, { "epoch": 0.31, "learning_rate": 6.873194423473753e-05, "loss": 1.0183, "step": 125300 }, { "epoch": 0.31, "learning_rate": 6.873068599325583e-05, "loss": 1.0221, "step": 125305 }, { "epoch": 0.31, "learning_rate": 6.872942775177412e-05, "loss": 1.0224, "step": 125310 }, { "epoch": 0.31, "learning_rate": 6.872816951029242e-05, "loss": 1.0223, "step": 125315 }, { "epoch": 0.31, "learning_rate": 6.872691126881071e-05, "loss": 1.0212, "step": 125320 }, { "epoch": 0.31, "learning_rate": 6.8725653027329e-05, "loss": 1.022, "step": 125325 }, { "epoch": 0.31, "learning_rate": 6.87243947858473e-05, "loss": 1.0232, "step": 125330 }, { "epoch": 0.31, "learning_rate": 6.872313654436559e-05, "loss": 1.0181, "step": 125335 }, { "epoch": 0.31, "learning_rate": 6.87218783028839e-05, "loss": 1.021, "step": 125340 }, { "epoch": 0.31, "learning_rate": 6.872062006140219e-05, "loss": 1.0224, "step": 125345 }, { "epoch": 0.31, "learning_rate": 6.871936181992048e-05, "loss": 1.0212, "step": 125350 }, { "epoch": 0.31, "learning_rate": 6.871810357843877e-05, "loss": 1.0182, "step": 125355 }, { "epoch": 0.31, "learning_rate": 6.871684533695707e-05, "loss": 1.021, "step": 125360 }, { "epoch": 0.31, "learning_rate": 6.871558709547537e-05, "loss": 1.0229, "step": 125365 }, { "epoch": 0.31, "learning_rate": 6.871432885399366e-05, "loss": 1.0213, "step": 125370 }, { "epoch": 0.31, "learning_rate": 6.871307061251195e-05, "loss": 1.0208, "step": 125375 }, { "epoch": 0.31, "learning_rate": 6.871181237103025e-05, "loss": 1.022, "step": 125380 }, { "epoch": 0.31, "learning_rate": 6.871055412954855e-05, "loss": 1.0204, "step": 125385 }, { "epoch": 0.31, "learning_rate": 6.870929588806684e-05, "loss": 1.0214, "step": 125390 }, { "epoch": 0.31, "learning_rate": 6.870803764658513e-05, "loss": 1.0204, "step": 125395 }, { "epoch": 0.31, "learning_rate": 6.870677940510342e-05, "loss": 1.0205, "step": 125400 }, { "epoch": 0.31, "learning_rate": 6.870552116362173e-05, "loss": 1.0237, "step": 125405 }, { "epoch": 0.31, "learning_rate": 6.870426292214002e-05, "loss": 1.0206, "step": 125410 }, { "epoch": 0.31, "learning_rate": 6.870300468065831e-05, "loss": 1.02, "step": 125415 }, { "epoch": 0.31, "learning_rate": 6.87017464391766e-05, "loss": 1.0237, "step": 125420 }, { "epoch": 0.31, "learning_rate": 6.87004881976949e-05, "loss": 1.0211, "step": 125425 }, { "epoch": 0.31, "learning_rate": 6.86992299562132e-05, "loss": 1.0214, "step": 125430 }, { "epoch": 0.31, "learning_rate": 6.869797171473149e-05, "loss": 1.0197, "step": 125435 }, { "epoch": 0.31, "learning_rate": 6.869671347324978e-05, "loss": 1.0231, "step": 125440 }, { "epoch": 0.31, "learning_rate": 6.869545523176809e-05, "loss": 1.0216, "step": 125445 }, { "epoch": 0.31, "learning_rate": 6.869419699028638e-05, "loss": 1.0204, "step": 125450 }, { "epoch": 0.31, "learning_rate": 6.869293874880467e-05, "loss": 1.0227, "step": 125455 }, { "epoch": 0.31, "learning_rate": 6.869168050732296e-05, "loss": 1.0472, "step": 125460 }, { "epoch": 0.31, "learning_rate": 6.869042226584125e-05, "loss": 1.0186, "step": 125465 }, { "epoch": 0.31, "learning_rate": 6.868916402435956e-05, "loss": 1.0223, "step": 125470 }, { "epoch": 0.31, "learning_rate": 6.868790578287785e-05, "loss": 1.0219, "step": 125475 }, { "epoch": 0.31, "learning_rate": 6.868664754139614e-05, "loss": 1.0218, "step": 125480 }, { "epoch": 0.31, "learning_rate": 6.868538929991443e-05, "loss": 1.0207, "step": 125485 }, { "epoch": 0.32, "learning_rate": 6.868413105843274e-05, "loss": 1.0223, "step": 125490 }, { "epoch": 0.32, "learning_rate": 6.868287281695103e-05, "loss": 1.0218, "step": 125495 }, { "epoch": 0.32, "learning_rate": 6.868161457546932e-05, "loss": 1.0191, "step": 125500 }, { "epoch": 0.32, "learning_rate": 6.868035633398763e-05, "loss": 1.0233, "step": 125505 }, { "epoch": 0.32, "learning_rate": 6.867909809250592e-05, "loss": 1.0227, "step": 125510 }, { "epoch": 0.32, "learning_rate": 6.867783985102422e-05, "loss": 1.0236, "step": 125515 }, { "epoch": 0.32, "learning_rate": 6.867658160954251e-05, "loss": 1.0229, "step": 125520 }, { "epoch": 0.32, "learning_rate": 6.86753233680608e-05, "loss": 1.0212, "step": 125525 }, { "epoch": 0.32, "learning_rate": 6.86740651265791e-05, "loss": 1.021, "step": 125530 }, { "epoch": 0.32, "learning_rate": 6.86728068850974e-05, "loss": 1.0259, "step": 125535 }, { "epoch": 0.32, "learning_rate": 6.86715486436157e-05, "loss": 1.023, "step": 125540 }, { "epoch": 0.32, "learning_rate": 6.867029040213399e-05, "loss": 1.0226, "step": 125545 }, { "epoch": 0.32, "learning_rate": 6.866903216065228e-05, "loss": 1.0213, "step": 125550 }, { "epoch": 0.32, "learning_rate": 6.866777391917057e-05, "loss": 1.0222, "step": 125555 }, { "epoch": 0.32, "learning_rate": 6.866651567768887e-05, "loss": 1.0216, "step": 125560 }, { "epoch": 0.32, "learning_rate": 6.866525743620717e-05, "loss": 1.0174, "step": 125565 }, { "epoch": 0.32, "learning_rate": 6.866399919472546e-05, "loss": 1.0211, "step": 125570 }, { "epoch": 0.32, "learning_rate": 6.866274095324375e-05, "loss": 1.0219, "step": 125575 }, { "epoch": 0.32, "learning_rate": 6.866148271176205e-05, "loss": 1.0227, "step": 125580 }, { "epoch": 0.32, "learning_rate": 6.866022447028035e-05, "loss": 1.0206, "step": 125585 }, { "epoch": 0.32, "learning_rate": 6.865896622879864e-05, "loss": 1.0211, "step": 125590 }, { "epoch": 0.32, "learning_rate": 6.865770798731693e-05, "loss": 1.0213, "step": 125595 }, { "epoch": 0.32, "learning_rate": 6.865644974583523e-05, "loss": 1.0214, "step": 125600 }, { "epoch": 0.32, "learning_rate": 6.865519150435353e-05, "loss": 1.0233, "step": 125605 }, { "epoch": 0.32, "learning_rate": 6.865393326287182e-05, "loss": 1.0219, "step": 125610 }, { "epoch": 0.32, "learning_rate": 6.865267502139011e-05, "loss": 1.022, "step": 125615 }, { "epoch": 0.32, "learning_rate": 6.86514167799084e-05, "loss": 1.0214, "step": 125620 }, { "epoch": 0.32, "learning_rate": 6.86501585384267e-05, "loss": 1.0214, "step": 125625 }, { "epoch": 0.32, "learning_rate": 6.8648900296945e-05, "loss": 1.0209, "step": 125630 }, { "epoch": 0.32, "learning_rate": 6.864764205546329e-05, "loss": 1.0213, "step": 125635 }, { "epoch": 0.32, "learning_rate": 6.864638381398158e-05, "loss": 1.0196, "step": 125640 }, { "epoch": 0.32, "learning_rate": 6.864512557249989e-05, "loss": 1.0229, "step": 125645 }, { "epoch": 0.32, "learning_rate": 6.864386733101818e-05, "loss": 1.0225, "step": 125650 }, { "epoch": 0.32, "learning_rate": 6.864260908953647e-05, "loss": 1.0232, "step": 125655 }, { "epoch": 0.32, "learning_rate": 6.864135084805476e-05, "loss": 1.0216, "step": 125660 }, { "epoch": 0.32, "learning_rate": 6.864009260657307e-05, "loss": 1.0225, "step": 125665 }, { "epoch": 0.32, "learning_rate": 6.863883436509136e-05, "loss": 1.0234, "step": 125670 }, { "epoch": 0.32, "learning_rate": 6.863757612360965e-05, "loss": 1.0235, "step": 125675 }, { "epoch": 0.32, "learning_rate": 6.863631788212794e-05, "loss": 1.0228, "step": 125680 }, { "epoch": 0.32, "learning_rate": 6.863505964064623e-05, "loss": 1.0225, "step": 125685 }, { "epoch": 0.32, "learning_rate": 6.863380139916454e-05, "loss": 1.0224, "step": 125690 }, { "epoch": 0.32, "learning_rate": 6.863254315768283e-05, "loss": 1.0225, "step": 125695 }, { "epoch": 0.32, "learning_rate": 6.863128491620112e-05, "loss": 1.0218, "step": 125700 }, { "epoch": 0.32, "learning_rate": 6.863002667471941e-05, "loss": 1.0206, "step": 125705 }, { "epoch": 0.32, "learning_rate": 6.862876843323772e-05, "loss": 1.0224, "step": 125710 }, { "epoch": 0.32, "learning_rate": 6.862751019175601e-05, "loss": 1.0192, "step": 125715 }, { "epoch": 0.32, "learning_rate": 6.86262519502743e-05, "loss": 1.0231, "step": 125720 }, { "epoch": 0.32, "learning_rate": 6.862499370879259e-05, "loss": 1.0208, "step": 125725 }, { "epoch": 0.32, "learning_rate": 6.86237354673109e-05, "loss": 1.0223, "step": 125730 }, { "epoch": 0.32, "learning_rate": 6.862247722582919e-05, "loss": 1.0227, "step": 125735 }, { "epoch": 0.32, "learning_rate": 6.862121898434748e-05, "loss": 1.0204, "step": 125740 }, { "epoch": 0.32, "learning_rate": 6.861996074286577e-05, "loss": 1.0227, "step": 125745 }, { "epoch": 0.32, "learning_rate": 6.861870250138406e-05, "loss": 1.0221, "step": 125750 }, { "epoch": 0.32, "learning_rate": 6.861744425990237e-05, "loss": 1.0212, "step": 125755 }, { "epoch": 0.32, "learning_rate": 6.861618601842066e-05, "loss": 1.022, "step": 125760 }, { "epoch": 0.32, "learning_rate": 6.861492777693895e-05, "loss": 1.0231, "step": 125765 }, { "epoch": 0.32, "learning_rate": 6.861366953545724e-05, "loss": 1.0203, "step": 125770 }, { "epoch": 0.32, "learning_rate": 6.861241129397555e-05, "loss": 1.0234, "step": 125775 }, { "epoch": 0.32, "learning_rate": 6.861115305249384e-05, "loss": 1.0226, "step": 125780 }, { "epoch": 0.32, "learning_rate": 6.860989481101213e-05, "loss": 1.0233, "step": 125785 }, { "epoch": 0.32, "learning_rate": 6.860863656953042e-05, "loss": 1.023, "step": 125790 }, { "epoch": 0.32, "learning_rate": 6.860737832804873e-05, "loss": 1.0231, "step": 125795 }, { "epoch": 0.32, "learning_rate": 6.860612008656702e-05, "loss": 1.0203, "step": 125800 }, { "epoch": 0.32, "learning_rate": 6.860486184508531e-05, "loss": 1.0197, "step": 125805 }, { "epoch": 0.32, "learning_rate": 6.86036036036036e-05, "loss": 1.0216, "step": 125810 }, { "epoch": 0.32, "learning_rate": 6.86023453621219e-05, "loss": 1.0176, "step": 125815 }, { "epoch": 0.32, "learning_rate": 6.86010871206402e-05, "loss": 1.0203, "step": 125820 }, { "epoch": 0.32, "learning_rate": 6.859982887915849e-05, "loss": 1.0228, "step": 125825 }, { "epoch": 0.32, "learning_rate": 6.859857063767678e-05, "loss": 1.0213, "step": 125830 }, { "epoch": 0.32, "learning_rate": 6.859731239619507e-05, "loss": 1.0236, "step": 125835 }, { "epoch": 0.32, "learning_rate": 6.859605415471338e-05, "loss": 1.0209, "step": 125840 }, { "epoch": 0.32, "learning_rate": 6.859479591323167e-05, "loss": 1.0234, "step": 125845 }, { "epoch": 0.32, "learning_rate": 6.859353767174996e-05, "loss": 1.0206, "step": 125850 }, { "epoch": 0.32, "learning_rate": 6.859227943026825e-05, "loss": 1.0245, "step": 125855 }, { "epoch": 0.32, "learning_rate": 6.859102118878656e-05, "loss": 1.0203, "step": 125860 }, { "epoch": 0.32, "learning_rate": 6.858976294730485e-05, "loss": 1.0226, "step": 125865 }, { "epoch": 0.32, "learning_rate": 6.858850470582314e-05, "loss": 1.0206, "step": 125870 }, { "epoch": 0.32, "learning_rate": 6.858724646434143e-05, "loss": 1.0218, "step": 125875 }, { "epoch": 0.32, "learning_rate": 6.858598822285973e-05, "loss": 1.0217, "step": 125880 }, { "epoch": 0.32, "learning_rate": 6.858472998137803e-05, "loss": 1.0217, "step": 125885 }, { "epoch": 0.32, "learning_rate": 6.858347173989632e-05, "loss": 1.0204, "step": 125890 }, { "epoch": 0.32, "learning_rate": 6.858221349841461e-05, "loss": 1.0211, "step": 125895 }, { "epoch": 0.32, "learning_rate": 6.85809552569329e-05, "loss": 1.0183, "step": 125900 }, { "epoch": 0.32, "learning_rate": 6.857969701545121e-05, "loss": 1.02, "step": 125905 }, { "epoch": 0.32, "learning_rate": 6.85784387739695e-05, "loss": 1.0226, "step": 125910 }, { "epoch": 0.32, "learning_rate": 6.85771805324878e-05, "loss": 1.0206, "step": 125915 }, { "epoch": 0.32, "learning_rate": 6.857592229100609e-05, "loss": 1.0223, "step": 125920 }, { "epoch": 0.32, "learning_rate": 6.857466404952439e-05, "loss": 1.0442, "step": 125925 }, { "epoch": 0.32, "learning_rate": 6.857340580804268e-05, "loss": 1.0208, "step": 125930 }, { "epoch": 0.32, "learning_rate": 6.857214756656097e-05, "loss": 1.023, "step": 125935 }, { "epoch": 0.32, "learning_rate": 6.857088932507927e-05, "loss": 1.0225, "step": 125940 }, { "epoch": 0.32, "learning_rate": 6.856963108359756e-05, "loss": 1.0207, "step": 125945 }, { "epoch": 0.32, "learning_rate": 6.856837284211586e-05, "loss": 1.0221, "step": 125950 }, { "epoch": 0.32, "learning_rate": 6.856711460063415e-05, "loss": 1.0223, "step": 125955 }, { "epoch": 0.32, "learning_rate": 6.856585635915245e-05, "loss": 1.0463, "step": 125960 }, { "epoch": 0.32, "learning_rate": 6.856459811767074e-05, "loss": 1.0217, "step": 125965 }, { "epoch": 0.32, "learning_rate": 6.856333987618904e-05, "loss": 1.0231, "step": 125970 }, { "epoch": 0.32, "learning_rate": 6.856208163470733e-05, "loss": 1.0232, "step": 125975 }, { "epoch": 0.32, "learning_rate": 6.856082339322563e-05, "loss": 1.0208, "step": 125980 }, { "epoch": 0.32, "learning_rate": 6.855956515174392e-05, "loss": 1.0197, "step": 125985 }, { "epoch": 0.32, "learning_rate": 6.855830691026221e-05, "loss": 1.0228, "step": 125990 }, { "epoch": 0.32, "learning_rate": 6.855704866878051e-05, "loss": 1.0224, "step": 125995 }, { "epoch": 0.32, "learning_rate": 6.85557904272988e-05, "loss": 1.0206, "step": 126000 }, { "epoch": 0.32, "learning_rate": 6.855453218581711e-05, "loss": 1.0214, "step": 126005 }, { "epoch": 0.32, "learning_rate": 6.85532739443354e-05, "loss": 1.0195, "step": 126010 }, { "epoch": 0.32, "learning_rate": 6.85520157028537e-05, "loss": 1.0211, "step": 126015 }, { "epoch": 0.32, "learning_rate": 6.8550757461372e-05, "loss": 1.0216, "step": 126020 }, { "epoch": 0.32, "learning_rate": 6.854949921989029e-05, "loss": 1.022, "step": 126025 }, { "epoch": 0.32, "learning_rate": 6.854824097840858e-05, "loss": 1.0211, "step": 126030 }, { "epoch": 0.32, "learning_rate": 6.854698273692687e-05, "loss": 1.0228, "step": 126035 }, { "epoch": 0.32, "learning_rate": 6.854572449544518e-05, "loss": 1.0212, "step": 126040 }, { "epoch": 0.32, "learning_rate": 6.854446625396347e-05, "loss": 1.0238, "step": 126045 }, { "epoch": 0.32, "learning_rate": 6.854320801248176e-05, "loss": 1.0221, "step": 126050 }, { "epoch": 0.32, "learning_rate": 6.854194977100005e-05, "loss": 1.0226, "step": 126055 }, { "epoch": 0.32, "learning_rate": 6.854069152951836e-05, "loss": 1.0202, "step": 126060 }, { "epoch": 0.32, "learning_rate": 6.853943328803665e-05, "loss": 1.0423, "step": 126065 }, { "epoch": 0.32, "learning_rate": 6.853817504655494e-05, "loss": 1.021, "step": 126070 }, { "epoch": 0.32, "learning_rate": 6.853691680507323e-05, "loss": 1.0212, "step": 126075 }, { "epoch": 0.32, "learning_rate": 6.853565856359153e-05, "loss": 1.0191, "step": 126080 }, { "epoch": 0.32, "learning_rate": 6.853440032210983e-05, "loss": 1.0203, "step": 126085 }, { "epoch": 0.32, "learning_rate": 6.853314208062812e-05, "loss": 1.0217, "step": 126090 }, { "epoch": 0.32, "learning_rate": 6.853188383914641e-05, "loss": 1.0203, "step": 126095 }, { "epoch": 0.32, "learning_rate": 6.85306255976647e-05, "loss": 1.0191, "step": 126100 }, { "epoch": 0.32, "learning_rate": 6.852936735618301e-05, "loss": 1.0218, "step": 126105 }, { "epoch": 0.32, "learning_rate": 6.85281091147013e-05, "loss": 1.0223, "step": 126110 }, { "epoch": 0.32, "learning_rate": 6.85268508732196e-05, "loss": 1.0208, "step": 126115 }, { "epoch": 0.32, "learning_rate": 6.852559263173789e-05, "loss": 1.0203, "step": 126120 }, { "epoch": 0.32, "learning_rate": 6.852433439025619e-05, "loss": 1.0204, "step": 126125 }, { "epoch": 0.32, "learning_rate": 6.852307614877448e-05, "loss": 1.0212, "step": 126130 }, { "epoch": 0.32, "learning_rate": 6.852181790729277e-05, "loss": 1.023, "step": 126135 }, { "epoch": 0.32, "learning_rate": 6.852055966581107e-05, "loss": 1.0219, "step": 126140 }, { "epoch": 0.32, "learning_rate": 6.851930142432936e-05, "loss": 1.0229, "step": 126145 }, { "epoch": 0.32, "learning_rate": 6.851804318284766e-05, "loss": 1.0233, "step": 126150 }, { "epoch": 0.32, "learning_rate": 6.851678494136595e-05, "loss": 1.0199, "step": 126155 }, { "epoch": 0.32, "learning_rate": 6.851552669988425e-05, "loss": 1.0207, "step": 126160 }, { "epoch": 0.32, "learning_rate": 6.851426845840254e-05, "loss": 1.0231, "step": 126165 }, { "epoch": 0.32, "learning_rate": 6.851301021692084e-05, "loss": 1.0207, "step": 126170 }, { "epoch": 0.32, "learning_rate": 6.851175197543913e-05, "loss": 1.0194, "step": 126175 }, { "epoch": 0.32, "learning_rate": 6.851049373395743e-05, "loss": 1.0231, "step": 126180 }, { "epoch": 0.32, "learning_rate": 6.850923549247572e-05, "loss": 1.0232, "step": 126185 }, { "epoch": 0.32, "learning_rate": 6.850797725099402e-05, "loss": 1.0242, "step": 126190 }, { "epoch": 0.32, "learning_rate": 6.850671900951231e-05, "loss": 1.0208, "step": 126195 }, { "epoch": 0.32, "learning_rate": 6.85054607680306e-05, "loss": 1.0221, "step": 126200 }, { "epoch": 0.32, "learning_rate": 6.85042025265489e-05, "loss": 1.0226, "step": 126205 }, { "epoch": 0.32, "learning_rate": 6.850294428506719e-05, "loss": 1.0236, "step": 126210 }, { "epoch": 0.32, "learning_rate": 6.85016860435855e-05, "loss": 1.0182, "step": 126215 }, { "epoch": 0.32, "learning_rate": 6.850042780210378e-05, "loss": 1.0416, "step": 126220 }, { "epoch": 0.32, "learning_rate": 6.849916956062208e-05, "loss": 1.0205, "step": 126225 }, { "epoch": 0.32, "learning_rate": 6.849791131914037e-05, "loss": 1.0226, "step": 126230 }, { "epoch": 0.32, "learning_rate": 6.849665307765867e-05, "loss": 1.0194, "step": 126235 }, { "epoch": 0.32, "learning_rate": 6.849539483617696e-05, "loss": 1.0225, "step": 126240 }, { "epoch": 0.32, "learning_rate": 6.849413659469526e-05, "loss": 1.0219, "step": 126245 }, { "epoch": 0.32, "learning_rate": 6.849287835321355e-05, "loss": 1.0212, "step": 126250 }, { "epoch": 0.32, "learning_rate": 6.849162011173185e-05, "loss": 1.0215, "step": 126255 }, { "epoch": 0.32, "learning_rate": 6.849036187025014e-05, "loss": 1.0221, "step": 126260 }, { "epoch": 0.32, "learning_rate": 6.848910362876844e-05, "loss": 1.0198, "step": 126265 }, { "epoch": 0.32, "learning_rate": 6.848784538728673e-05, "loss": 1.0205, "step": 126270 }, { "epoch": 0.32, "learning_rate": 6.848658714580502e-05, "loss": 1.0193, "step": 126275 }, { "epoch": 0.32, "learning_rate": 6.848532890432332e-05, "loss": 1.0218, "step": 126280 }, { "epoch": 0.32, "learning_rate": 6.848407066284162e-05, "loss": 1.0214, "step": 126285 }, { "epoch": 0.32, "learning_rate": 6.848281242135991e-05, "loss": 1.0207, "step": 126290 }, { "epoch": 0.32, "learning_rate": 6.84815541798782e-05, "loss": 1.021, "step": 126295 }, { "epoch": 0.32, "learning_rate": 6.84802959383965e-05, "loss": 1.0217, "step": 126300 }, { "epoch": 0.32, "learning_rate": 6.84790376969148e-05, "loss": 1.02, "step": 126305 }, { "epoch": 0.32, "learning_rate": 6.847777945543309e-05, "loss": 1.0203, "step": 126310 }, { "epoch": 0.32, "learning_rate": 6.847652121395138e-05, "loss": 1.0211, "step": 126315 }, { "epoch": 0.32, "learning_rate": 6.847526297246968e-05, "loss": 1.0228, "step": 126320 }, { "epoch": 0.32, "learning_rate": 6.847400473098798e-05, "loss": 1.0217, "step": 126325 }, { "epoch": 0.32, "learning_rate": 6.847274648950627e-05, "loss": 1.021, "step": 126330 }, { "epoch": 0.32, "learning_rate": 6.847148824802456e-05, "loss": 1.0243, "step": 126335 }, { "epoch": 0.32, "learning_rate": 6.847023000654285e-05, "loss": 1.024, "step": 126340 }, { "epoch": 0.32, "learning_rate": 6.846897176506116e-05, "loss": 1.0213, "step": 126345 }, { "epoch": 0.32, "learning_rate": 6.846771352357945e-05, "loss": 1.0215, "step": 126350 }, { "epoch": 0.32, "learning_rate": 6.846645528209774e-05, "loss": 1.0205, "step": 126355 }, { "epoch": 0.32, "learning_rate": 6.846519704061603e-05, "loss": 1.0199, "step": 126360 }, { "epoch": 0.32, "learning_rate": 6.846393879913434e-05, "loss": 1.0196, "step": 126365 }, { "epoch": 0.32, "learning_rate": 6.846268055765263e-05, "loss": 1.0217, "step": 126370 }, { "epoch": 0.32, "learning_rate": 6.846142231617092e-05, "loss": 1.0229, "step": 126375 }, { "epoch": 0.32, "learning_rate": 6.846016407468921e-05, "loss": 1.0206, "step": 126380 }, { "epoch": 0.32, "learning_rate": 6.845890583320752e-05, "loss": 1.0446, "step": 126385 }, { "epoch": 0.32, "learning_rate": 6.845764759172581e-05, "loss": 1.023, "step": 126390 }, { "epoch": 0.32, "learning_rate": 6.84563893502441e-05, "loss": 1.0203, "step": 126395 }, { "epoch": 0.32, "learning_rate": 6.845513110876239e-05, "loss": 1.0225, "step": 126400 }, { "epoch": 0.32, "learning_rate": 6.845387286728068e-05, "loss": 1.0211, "step": 126405 }, { "epoch": 0.32, "learning_rate": 6.845261462579899e-05, "loss": 1.0206, "step": 126410 }, { "epoch": 0.32, "learning_rate": 6.845135638431728e-05, "loss": 1.0218, "step": 126415 }, { "epoch": 0.32, "learning_rate": 6.845009814283557e-05, "loss": 1.023, "step": 126420 }, { "epoch": 0.32, "learning_rate": 6.844883990135386e-05, "loss": 1.0211, "step": 126425 }, { "epoch": 0.32, "learning_rate": 6.844758165987217e-05, "loss": 1.0435, "step": 126430 }, { "epoch": 0.32, "learning_rate": 6.844632341839046e-05, "loss": 1.0224, "step": 126435 }, { "epoch": 0.32, "learning_rate": 6.844506517690875e-05, "loss": 1.0211, "step": 126440 }, { "epoch": 0.32, "learning_rate": 6.844380693542704e-05, "loss": 1.0214, "step": 126445 }, { "epoch": 0.32, "learning_rate": 6.844254869394535e-05, "loss": 1.022, "step": 126450 }, { "epoch": 0.32, "learning_rate": 6.844129045246364e-05, "loss": 1.0226, "step": 126455 }, { "epoch": 0.32, "learning_rate": 6.844003221098193e-05, "loss": 1.0218, "step": 126460 }, { "epoch": 0.32, "learning_rate": 6.843877396950022e-05, "loss": 1.0247, "step": 126465 }, { "epoch": 0.32, "learning_rate": 6.843751572801851e-05, "loss": 1.02, "step": 126470 }, { "epoch": 0.32, "learning_rate": 6.843625748653682e-05, "loss": 1.0229, "step": 126475 }, { "epoch": 0.32, "learning_rate": 6.843499924505511e-05, "loss": 1.0218, "step": 126480 }, { "epoch": 0.32, "learning_rate": 6.84337410035734e-05, "loss": 1.0211, "step": 126485 }, { "epoch": 0.32, "learning_rate": 6.84324827620917e-05, "loss": 1.022, "step": 126490 }, { "epoch": 0.32, "learning_rate": 6.843122452061e-05, "loss": 1.0196, "step": 126495 }, { "epoch": 0.32, "learning_rate": 6.842996627912829e-05, "loss": 1.0198, "step": 126500 }, { "epoch": 0.32, "learning_rate": 6.84287080376466e-05, "loss": 1.0239, "step": 126505 }, { "epoch": 0.32, "learning_rate": 6.842744979616489e-05, "loss": 1.02, "step": 126510 }, { "epoch": 0.32, "learning_rate": 6.842619155468318e-05, "loss": 1.021, "step": 126515 }, { "epoch": 0.32, "learning_rate": 6.842493331320148e-05, "loss": 1.0207, "step": 126520 }, { "epoch": 0.32, "learning_rate": 6.842367507171978e-05, "loss": 1.0208, "step": 126525 }, { "epoch": 0.32, "learning_rate": 6.842241683023807e-05, "loss": 1.0215, "step": 126530 }, { "epoch": 0.32, "learning_rate": 6.842115858875636e-05, "loss": 1.0206, "step": 126535 }, { "epoch": 0.32, "learning_rate": 6.841990034727465e-05, "loss": 1.0216, "step": 126540 }, { "epoch": 0.32, "learning_rate": 6.841864210579296e-05, "loss": 1.0205, "step": 126545 }, { "epoch": 0.32, "learning_rate": 6.841738386431125e-05, "loss": 1.0216, "step": 126550 }, { "epoch": 0.32, "learning_rate": 6.841612562282954e-05, "loss": 1.021, "step": 126555 }, { "epoch": 0.32, "learning_rate": 6.841486738134783e-05, "loss": 1.0204, "step": 126560 }, { "epoch": 0.32, "learning_rate": 6.841360913986614e-05, "loss": 1.0204, "step": 126565 }, { "epoch": 0.32, "learning_rate": 6.841235089838443e-05, "loss": 1.0228, "step": 126570 }, { "epoch": 0.32, "learning_rate": 6.841109265690272e-05, "loss": 1.0207, "step": 126575 }, { "epoch": 0.32, "learning_rate": 6.840983441542101e-05, "loss": 1.0205, "step": 126580 }, { "epoch": 0.32, "learning_rate": 6.840857617393932e-05, "loss": 1.0206, "step": 126585 }, { "epoch": 0.32, "learning_rate": 6.840731793245761e-05, "loss": 1.0196, "step": 126590 }, { "epoch": 0.32, "learning_rate": 6.84060596909759e-05, "loss": 1.0203, "step": 126595 }, { "epoch": 0.32, "learning_rate": 6.840480144949419e-05, "loss": 1.0209, "step": 126600 }, { "epoch": 0.32, "learning_rate": 6.840354320801248e-05, "loss": 1.0238, "step": 126605 }, { "epoch": 0.32, "learning_rate": 6.840228496653079e-05, "loss": 1.0236, "step": 126610 }, { "epoch": 0.32, "learning_rate": 6.840102672504908e-05, "loss": 1.0196, "step": 126615 }, { "epoch": 0.32, "learning_rate": 6.839976848356737e-05, "loss": 1.0214, "step": 126620 }, { "epoch": 0.32, "learning_rate": 6.839851024208566e-05, "loss": 1.0218, "step": 126625 }, { "epoch": 0.32, "learning_rate": 6.839725200060397e-05, "loss": 1.0246, "step": 126630 }, { "epoch": 0.32, "learning_rate": 6.839599375912226e-05, "loss": 1.0225, "step": 126635 }, { "epoch": 0.32, "learning_rate": 6.839473551764055e-05, "loss": 1.023, "step": 126640 }, { "epoch": 0.32, "learning_rate": 6.839347727615884e-05, "loss": 1.0212, "step": 126645 }, { "epoch": 0.32, "learning_rate": 6.839221903467715e-05, "loss": 1.0211, "step": 126650 }, { "epoch": 0.32, "learning_rate": 6.839096079319544e-05, "loss": 1.0221, "step": 126655 }, { "epoch": 0.32, "learning_rate": 6.838970255171373e-05, "loss": 1.0208, "step": 126660 }, { "epoch": 0.32, "learning_rate": 6.838844431023202e-05, "loss": 1.0219, "step": 126665 }, { "epoch": 0.32, "learning_rate": 6.838718606875031e-05, "loss": 1.0198, "step": 126670 }, { "epoch": 0.32, "learning_rate": 6.838592782726862e-05, "loss": 1.0204, "step": 126675 }, { "epoch": 0.32, "learning_rate": 6.838466958578691e-05, "loss": 1.0204, "step": 126680 }, { "epoch": 0.32, "learning_rate": 6.83834113443052e-05, "loss": 1.0249, "step": 126685 }, { "epoch": 0.32, "learning_rate": 6.838215310282349e-05, "loss": 1.0241, "step": 126690 }, { "epoch": 0.32, "learning_rate": 6.83808948613418e-05, "loss": 1.0216, "step": 126695 }, { "epoch": 0.32, "learning_rate": 6.837963661986009e-05, "loss": 1.0228, "step": 126700 }, { "epoch": 0.32, "learning_rate": 6.837837837837838e-05, "loss": 1.0198, "step": 126705 }, { "epoch": 0.32, "learning_rate": 6.837712013689667e-05, "loss": 1.0212, "step": 126710 }, { "epoch": 0.32, "learning_rate": 6.837586189541498e-05, "loss": 1.0245, "step": 126715 }, { "epoch": 0.32, "learning_rate": 6.837460365393327e-05, "loss": 1.0217, "step": 126720 }, { "epoch": 0.32, "learning_rate": 6.837334541245156e-05, "loss": 1.0231, "step": 126725 }, { "epoch": 0.32, "learning_rate": 6.837208717096985e-05, "loss": 1.0244, "step": 126730 }, { "epoch": 0.32, "learning_rate": 6.837082892948814e-05, "loss": 1.0208, "step": 126735 }, { "epoch": 0.32, "learning_rate": 6.836957068800645e-05, "loss": 1.0222, "step": 126740 }, { "epoch": 0.32, "learning_rate": 6.836831244652474e-05, "loss": 1.0218, "step": 126745 }, { "epoch": 0.32, "learning_rate": 6.836705420504303e-05, "loss": 1.0225, "step": 126750 }, { "epoch": 0.32, "learning_rate": 6.836579596356132e-05, "loss": 1.0214, "step": 126755 }, { "epoch": 0.32, "learning_rate": 6.836453772207963e-05, "loss": 1.0236, "step": 126760 }, { "epoch": 0.32, "learning_rate": 6.836327948059792e-05, "loss": 1.022, "step": 126765 }, { "epoch": 0.32, "learning_rate": 6.836202123911621e-05, "loss": 1.0227, "step": 126770 }, { "epoch": 0.32, "learning_rate": 6.83607629976345e-05, "loss": 1.022, "step": 126775 }, { "epoch": 0.32, "learning_rate": 6.835950475615281e-05, "loss": 1.0223, "step": 126780 }, { "epoch": 0.32, "learning_rate": 6.83582465146711e-05, "loss": 1.02, "step": 126785 }, { "epoch": 0.32, "learning_rate": 6.835698827318939e-05, "loss": 1.0222, "step": 126790 }, { "epoch": 0.32, "learning_rate": 6.835573003170768e-05, "loss": 1.0213, "step": 126795 }, { "epoch": 0.32, "learning_rate": 6.835447179022598e-05, "loss": 1.0224, "step": 126800 }, { "epoch": 0.32, "learning_rate": 6.835321354874428e-05, "loss": 1.0411, "step": 126805 }, { "epoch": 0.32, "learning_rate": 6.835195530726257e-05, "loss": 1.0193, "step": 126810 }, { "epoch": 0.32, "learning_rate": 6.835069706578086e-05, "loss": 1.0203, "step": 126815 }, { "epoch": 0.32, "learning_rate": 6.834943882429916e-05, "loss": 1.0226, "step": 126820 }, { "epoch": 0.32, "learning_rate": 6.834818058281746e-05, "loss": 1.0221, "step": 126825 }, { "epoch": 0.32, "learning_rate": 6.834692234133575e-05, "loss": 1.0238, "step": 126830 }, { "epoch": 0.32, "learning_rate": 6.834566409985404e-05, "loss": 1.0196, "step": 126835 }, { "epoch": 0.32, "learning_rate": 6.834440585837234e-05, "loss": 1.0234, "step": 126840 }, { "epoch": 0.32, "learning_rate": 6.834314761689064e-05, "loss": 1.0215, "step": 126845 }, { "epoch": 0.32, "learning_rate": 6.834188937540893e-05, "loss": 1.0215, "step": 126850 }, { "epoch": 0.32, "learning_rate": 6.834063113392722e-05, "loss": 1.0205, "step": 126855 }, { "epoch": 0.32, "learning_rate": 6.833937289244552e-05, "loss": 1.0229, "step": 126860 }, { "epoch": 0.32, "learning_rate": 6.833811465096381e-05, "loss": 1.0212, "step": 126865 }, { "epoch": 0.32, "learning_rate": 6.833685640948211e-05, "loss": 1.0224, "step": 126870 }, { "epoch": 0.32, "learning_rate": 6.83355981680004e-05, "loss": 1.0225, "step": 126875 }, { "epoch": 0.32, "learning_rate": 6.83343399265187e-05, "loss": 1.0208, "step": 126880 }, { "epoch": 0.32, "learning_rate": 6.833308168503699e-05, "loss": 1.0217, "step": 126885 }, { "epoch": 0.32, "learning_rate": 6.833182344355529e-05, "loss": 1.0204, "step": 126890 }, { "epoch": 0.32, "learning_rate": 6.833056520207358e-05, "loss": 1.023, "step": 126895 }, { "epoch": 0.32, "learning_rate": 6.832930696059188e-05, "loss": 1.0189, "step": 126900 }, { "epoch": 0.32, "learning_rate": 6.832804871911017e-05, "loss": 1.0229, "step": 126905 }, { "epoch": 0.32, "learning_rate": 6.832679047762847e-05, "loss": 1.0231, "step": 126910 }, { "epoch": 0.32, "learning_rate": 6.832553223614676e-05, "loss": 1.0215, "step": 126915 }, { "epoch": 0.32, "learning_rate": 6.832427399466506e-05, "loss": 1.0217, "step": 126920 }, { "epoch": 0.32, "learning_rate": 6.832301575318335e-05, "loss": 1.0212, "step": 126925 }, { "epoch": 0.32, "learning_rate": 6.832175751170164e-05, "loss": 1.0218, "step": 126930 }, { "epoch": 0.32, "learning_rate": 6.832049927021994e-05, "loss": 1.0222, "step": 126935 }, { "epoch": 0.32, "learning_rate": 6.831924102873824e-05, "loss": 1.0202, "step": 126940 }, { "epoch": 0.32, "learning_rate": 6.831798278725653e-05, "loss": 1.0219, "step": 126945 }, { "epoch": 0.32, "learning_rate": 6.831672454577482e-05, "loss": 1.0199, "step": 126950 }, { "epoch": 0.32, "learning_rate": 6.831546630429312e-05, "loss": 1.0232, "step": 126955 }, { "epoch": 0.32, "learning_rate": 6.831420806281142e-05, "loss": 1.0209, "step": 126960 }, { "epoch": 0.32, "learning_rate": 6.831294982132971e-05, "loss": 1.02, "step": 126965 }, { "epoch": 0.32, "learning_rate": 6.8311691579848e-05, "loss": 1.0209, "step": 126970 }, { "epoch": 0.32, "learning_rate": 6.83104333383663e-05, "loss": 1.0217, "step": 126975 }, { "epoch": 0.32, "learning_rate": 6.83091750968846e-05, "loss": 1.0207, "step": 126980 }, { "epoch": 0.32, "learning_rate": 6.830791685540289e-05, "loss": 1.022, "step": 126985 }, { "epoch": 0.32, "learning_rate": 6.830665861392118e-05, "loss": 1.022, "step": 126990 }, { "epoch": 0.32, "learning_rate": 6.830540037243947e-05, "loss": 1.0211, "step": 126995 }, { "epoch": 0.32, "learning_rate": 6.830414213095778e-05, "loss": 1.0212, "step": 127000 }, { "epoch": 0.32, "learning_rate": 6.830288388947608e-05, "loss": 1.0227, "step": 127005 }, { "epoch": 0.32, "learning_rate": 6.830162564799437e-05, "loss": 1.0207, "step": 127010 }, { "epoch": 0.32, "learning_rate": 6.830036740651266e-05, "loss": 1.0222, "step": 127015 }, { "epoch": 0.32, "learning_rate": 6.829910916503096e-05, "loss": 1.0233, "step": 127020 }, { "epoch": 0.32, "learning_rate": 6.829785092354926e-05, "loss": 1.0218, "step": 127025 }, { "epoch": 0.32, "learning_rate": 6.829659268206755e-05, "loss": 1.021, "step": 127030 }, { "epoch": 0.32, "learning_rate": 6.829533444058584e-05, "loss": 1.0233, "step": 127035 }, { "epoch": 0.32, "learning_rate": 6.829407619910414e-05, "loss": 1.0239, "step": 127040 }, { "epoch": 0.32, "learning_rate": 6.829281795762244e-05, "loss": 1.0218, "step": 127045 }, { "epoch": 0.32, "learning_rate": 6.829155971614073e-05, "loss": 1.021, "step": 127050 }, { "epoch": 0.32, "learning_rate": 6.829030147465902e-05, "loss": 1.0198, "step": 127055 }, { "epoch": 0.32, "learning_rate": 6.828904323317732e-05, "loss": 1.0205, "step": 127060 }, { "epoch": 0.32, "learning_rate": 6.828778499169561e-05, "loss": 1.0209, "step": 127065 }, { "epoch": 0.32, "learning_rate": 6.828652675021391e-05, "loss": 1.0186, "step": 127070 }, { "epoch": 0.32, "learning_rate": 6.82852685087322e-05, "loss": 1.0216, "step": 127075 }, { "epoch": 0.32, "learning_rate": 6.82840102672505e-05, "loss": 1.0453, "step": 127080 }, { "epoch": 0.32, "learning_rate": 6.828275202576879e-05, "loss": 1.0201, "step": 127085 }, { "epoch": 0.32, "learning_rate": 6.828149378428709e-05, "loss": 1.0202, "step": 127090 }, { "epoch": 0.32, "learning_rate": 6.828023554280538e-05, "loss": 1.022, "step": 127095 }, { "epoch": 0.32, "learning_rate": 6.827897730132368e-05, "loss": 1.0214, "step": 127100 }, { "epoch": 0.32, "learning_rate": 6.827771905984197e-05, "loss": 1.024, "step": 127105 }, { "epoch": 0.32, "learning_rate": 6.827646081836027e-05, "loss": 1.0218, "step": 127110 }, { "epoch": 0.32, "learning_rate": 6.827520257687856e-05, "loss": 1.0205, "step": 127115 }, { "epoch": 0.32, "learning_rate": 6.827394433539686e-05, "loss": 1.0198, "step": 127120 }, { "epoch": 0.32, "learning_rate": 6.827268609391515e-05, "loss": 1.0203, "step": 127125 }, { "epoch": 0.32, "learning_rate": 6.827142785243344e-05, "loss": 1.0221, "step": 127130 }, { "epoch": 0.32, "learning_rate": 6.827016961095174e-05, "loss": 1.021, "step": 127135 }, { "epoch": 0.32, "learning_rate": 6.826891136947004e-05, "loss": 1.0219, "step": 127140 }, { "epoch": 0.32, "learning_rate": 6.826765312798833e-05, "loss": 1.0213, "step": 127145 }, { "epoch": 0.32, "learning_rate": 6.826639488650662e-05, "loss": 1.0198, "step": 127150 }, { "epoch": 0.32, "learning_rate": 6.826513664502492e-05, "loss": 1.0225, "step": 127155 }, { "epoch": 0.32, "learning_rate": 6.826387840354322e-05, "loss": 1.0227, "step": 127160 }, { "epoch": 0.32, "learning_rate": 6.82626201620615e-05, "loss": 1.0232, "step": 127165 }, { "epoch": 0.32, "learning_rate": 6.82613619205798e-05, "loss": 1.0214, "step": 127170 }, { "epoch": 0.32, "learning_rate": 6.82601036790981e-05, "loss": 1.0207, "step": 127175 }, { "epoch": 0.32, "learning_rate": 6.82588454376164e-05, "loss": 1.021, "step": 127180 }, { "epoch": 0.32, "learning_rate": 6.825758719613469e-05, "loss": 1.0217, "step": 127185 }, { "epoch": 0.32, "learning_rate": 6.825632895465298e-05, "loss": 1.0237, "step": 127190 }, { "epoch": 0.32, "learning_rate": 6.825507071317127e-05, "loss": 1.0231, "step": 127195 }, { "epoch": 0.32, "learning_rate": 6.825381247168957e-05, "loss": 1.0194, "step": 127200 }, { "epoch": 0.32, "learning_rate": 6.825255423020787e-05, "loss": 1.0226, "step": 127205 }, { "epoch": 0.32, "learning_rate": 6.825129598872616e-05, "loss": 1.0225, "step": 127210 }, { "epoch": 0.32, "learning_rate": 6.825003774724445e-05, "loss": 1.021, "step": 127215 }, { "epoch": 0.32, "learning_rate": 6.824877950576275e-05, "loss": 1.0232, "step": 127220 }, { "epoch": 0.32, "learning_rate": 6.824752126428105e-05, "loss": 1.0202, "step": 127225 }, { "epoch": 0.32, "learning_rate": 6.824626302279934e-05, "loss": 1.0204, "step": 127230 }, { "epoch": 0.32, "learning_rate": 6.824500478131763e-05, "loss": 1.0222, "step": 127235 }, { "epoch": 0.32, "learning_rate": 6.824374653983593e-05, "loss": 1.0198, "step": 127240 }, { "epoch": 0.32, "learning_rate": 6.824248829835423e-05, "loss": 1.0223, "step": 127245 }, { "epoch": 0.32, "learning_rate": 6.824123005687252e-05, "loss": 1.0217, "step": 127250 }, { "epoch": 0.32, "learning_rate": 6.823997181539081e-05, "loss": 1.0217, "step": 127255 }, { "epoch": 0.32, "learning_rate": 6.82387135739091e-05, "loss": 1.0234, "step": 127260 }, { "epoch": 0.32, "learning_rate": 6.82374553324274e-05, "loss": 1.0206, "step": 127265 }, { "epoch": 0.32, "learning_rate": 6.82361970909457e-05, "loss": 1.0198, "step": 127270 }, { "epoch": 0.32, "learning_rate": 6.823493884946399e-05, "loss": 1.0217, "step": 127275 }, { "epoch": 0.32, "learning_rate": 6.823368060798228e-05, "loss": 1.0206, "step": 127280 }, { "epoch": 0.32, "learning_rate": 6.823242236650059e-05, "loss": 1.0236, "step": 127285 }, { "epoch": 0.32, "learning_rate": 6.823116412501888e-05, "loss": 1.0206, "step": 127290 }, { "epoch": 0.32, "learning_rate": 6.822990588353717e-05, "loss": 1.0225, "step": 127295 }, { "epoch": 0.32, "learning_rate": 6.822864764205546e-05, "loss": 1.0228, "step": 127300 }, { "epoch": 0.32, "learning_rate": 6.822738940057377e-05, "loss": 1.0213, "step": 127305 }, { "epoch": 0.32, "learning_rate": 6.822613115909206e-05, "loss": 1.0217, "step": 127310 }, { "epoch": 0.32, "learning_rate": 6.822487291761035e-05, "loss": 1.0232, "step": 127315 }, { "epoch": 0.32, "learning_rate": 6.822361467612864e-05, "loss": 1.0219, "step": 127320 }, { "epoch": 0.32, "learning_rate": 6.822235643464693e-05, "loss": 1.0215, "step": 127325 }, { "epoch": 0.32, "learning_rate": 6.822109819316524e-05, "loss": 1.021, "step": 127330 }, { "epoch": 0.32, "learning_rate": 6.821983995168353e-05, "loss": 1.02, "step": 127335 }, { "epoch": 0.32, "learning_rate": 6.821858171020182e-05, "loss": 1.0215, "step": 127340 }, { "epoch": 0.32, "learning_rate": 6.821732346872011e-05, "loss": 1.0214, "step": 127345 }, { "epoch": 0.32, "learning_rate": 6.821606522723842e-05, "loss": 1.0224, "step": 127350 }, { "epoch": 0.32, "learning_rate": 6.821480698575671e-05, "loss": 1.0235, "step": 127355 }, { "epoch": 0.32, "learning_rate": 6.8213548744275e-05, "loss": 1.0186, "step": 127360 }, { "epoch": 0.32, "learning_rate": 6.821229050279329e-05, "loss": 1.0219, "step": 127365 }, { "epoch": 0.32, "learning_rate": 6.82110322613116e-05, "loss": 1.023, "step": 127370 }, { "epoch": 0.32, "learning_rate": 6.820977401982989e-05, "loss": 1.0215, "step": 127375 }, { "epoch": 0.32, "learning_rate": 6.820851577834818e-05, "loss": 1.023, "step": 127380 }, { "epoch": 0.32, "learning_rate": 6.820725753686647e-05, "loss": 1.0233, "step": 127385 }, { "epoch": 0.32, "learning_rate": 6.820599929538476e-05, "loss": 1.0195, "step": 127390 }, { "epoch": 0.32, "learning_rate": 6.820474105390307e-05, "loss": 1.0206, "step": 127395 }, { "epoch": 0.32, "learning_rate": 6.820348281242136e-05, "loss": 1.0248, "step": 127400 }, { "epoch": 0.32, "learning_rate": 6.820222457093965e-05, "loss": 1.0209, "step": 127405 }, { "epoch": 0.32, "learning_rate": 6.820096632945794e-05, "loss": 1.0225, "step": 127410 }, { "epoch": 0.32, "learning_rate": 6.819970808797625e-05, "loss": 1.0219, "step": 127415 }, { "epoch": 0.32, "learning_rate": 6.819844984649454e-05, "loss": 1.0209, "step": 127420 }, { "epoch": 0.32, "learning_rate": 6.819719160501283e-05, "loss": 1.025, "step": 127425 }, { "epoch": 0.32, "learning_rate": 6.819593336353112e-05, "loss": 1.0213, "step": 127430 }, { "epoch": 0.32, "learning_rate": 6.819467512204943e-05, "loss": 1.0205, "step": 127435 }, { "epoch": 0.32, "learning_rate": 6.819341688056772e-05, "loss": 1.0205, "step": 127440 }, { "epoch": 0.32, "learning_rate": 6.819215863908601e-05, "loss": 1.0219, "step": 127445 }, { "epoch": 0.32, "learning_rate": 6.81909003976043e-05, "loss": 1.0235, "step": 127450 }, { "epoch": 0.32, "learning_rate": 6.81896421561226e-05, "loss": 1.0219, "step": 127455 }, { "epoch": 0.32, "learning_rate": 6.81883839146409e-05, "loss": 1.0218, "step": 127460 }, { "epoch": 0.32, "learning_rate": 6.818712567315919e-05, "loss": 1.023, "step": 127465 }, { "epoch": 0.32, "learning_rate": 6.818586743167748e-05, "loss": 1.0224, "step": 127470 }, { "epoch": 0.32, "learning_rate": 6.818460919019578e-05, "loss": 1.0439, "step": 127475 }, { "epoch": 0.32, "learning_rate": 6.818335094871408e-05, "loss": 1.0225, "step": 127480 }, { "epoch": 0.32, "learning_rate": 6.818209270723237e-05, "loss": 1.0448, "step": 127485 }, { "epoch": 0.32, "learning_rate": 6.818083446575066e-05, "loss": 1.0204, "step": 127490 }, { "epoch": 0.32, "learning_rate": 6.817957622426896e-05, "loss": 1.0203, "step": 127495 }, { "epoch": 0.32, "learning_rate": 6.817831798278726e-05, "loss": 1.019, "step": 127500 }, { "epoch": 0.32, "learning_rate": 6.817705974130557e-05, "loss": 1.0191, "step": 127505 }, { "epoch": 0.32, "learning_rate": 6.817580149982386e-05, "loss": 1.0219, "step": 127510 }, { "epoch": 0.32, "learning_rate": 6.817454325834215e-05, "loss": 1.0205, "step": 127515 }, { "epoch": 0.32, "learning_rate": 6.817328501686044e-05, "loss": 1.0245, "step": 127520 }, { "epoch": 0.32, "learning_rate": 6.817202677537875e-05, "loss": 1.0197, "step": 127525 }, { "epoch": 0.32, "learning_rate": 6.817076853389704e-05, "loss": 1.0244, "step": 127530 }, { "epoch": 0.32, "learning_rate": 6.816951029241533e-05, "loss": 1.023, "step": 127535 }, { "epoch": 0.32, "learning_rate": 6.816825205093362e-05, "loss": 1.0235, "step": 127540 }, { "epoch": 0.32, "learning_rate": 6.816699380945191e-05, "loss": 1.0224, "step": 127545 }, { "epoch": 0.32, "learning_rate": 6.816573556797022e-05, "loss": 1.0182, "step": 127550 }, { "epoch": 0.32, "learning_rate": 6.816447732648851e-05, "loss": 1.0231, "step": 127555 }, { "epoch": 0.32, "learning_rate": 6.81632190850068e-05, "loss": 1.0203, "step": 127560 }, { "epoch": 0.32, "learning_rate": 6.816196084352509e-05, "loss": 1.0234, "step": 127565 }, { "epoch": 0.32, "learning_rate": 6.81607026020434e-05, "loss": 1.0217, "step": 127570 }, { "epoch": 0.32, "learning_rate": 6.815944436056169e-05, "loss": 1.0207, "step": 127575 }, { "epoch": 0.32, "learning_rate": 6.815818611907998e-05, "loss": 1.0212, "step": 127580 }, { "epoch": 0.32, "learning_rate": 6.815692787759827e-05, "loss": 1.024, "step": 127585 }, { "epoch": 0.32, "learning_rate": 6.815566963611658e-05, "loss": 1.0225, "step": 127590 }, { "epoch": 0.32, "learning_rate": 6.815441139463487e-05, "loss": 1.0204, "step": 127595 }, { "epoch": 0.32, "learning_rate": 6.815315315315316e-05, "loss": 1.0222, "step": 127600 }, { "epoch": 0.32, "learning_rate": 6.815189491167145e-05, "loss": 1.0209, "step": 127605 }, { "epoch": 0.32, "learning_rate": 6.815063667018974e-05, "loss": 1.0208, "step": 127610 }, { "epoch": 0.32, "learning_rate": 6.814937842870805e-05, "loss": 1.022, "step": 127615 }, { "epoch": 0.32, "learning_rate": 6.814812018722634e-05, "loss": 1.0202, "step": 127620 }, { "epoch": 0.32, "learning_rate": 6.814686194574463e-05, "loss": 1.0202, "step": 127625 }, { "epoch": 0.32, "learning_rate": 6.814560370426292e-05, "loss": 1.0207, "step": 127630 }, { "epoch": 0.32, "learning_rate": 6.814434546278123e-05, "loss": 1.0202, "step": 127635 }, { "epoch": 0.32, "learning_rate": 6.814308722129952e-05, "loss": 1.0222, "step": 127640 }, { "epoch": 0.32, "learning_rate": 6.814182897981781e-05, "loss": 1.0212, "step": 127645 }, { "epoch": 0.32, "learning_rate": 6.81405707383361e-05, "loss": 1.0233, "step": 127650 }, { "epoch": 0.32, "learning_rate": 6.81393124968544e-05, "loss": 1.0204, "step": 127655 }, { "epoch": 0.32, "learning_rate": 6.81380542553727e-05, "loss": 1.0218, "step": 127660 }, { "epoch": 0.32, "learning_rate": 6.813679601389099e-05, "loss": 1.0218, "step": 127665 }, { "epoch": 0.32, "learning_rate": 6.813553777240928e-05, "loss": 1.0218, "step": 127670 }, { "epoch": 0.32, "learning_rate": 6.813427953092757e-05, "loss": 1.021, "step": 127675 }, { "epoch": 0.32, "learning_rate": 6.813302128944588e-05, "loss": 1.0206, "step": 127680 }, { "epoch": 0.32, "learning_rate": 6.813176304796417e-05, "loss": 1.0218, "step": 127685 }, { "epoch": 0.32, "learning_rate": 6.813050480648246e-05, "loss": 1.0234, "step": 127690 }, { "epoch": 0.32, "learning_rate": 6.812924656500075e-05, "loss": 1.0211, "step": 127695 }, { "epoch": 0.32, "learning_rate": 6.812798832351906e-05, "loss": 1.0209, "step": 127700 }, { "epoch": 0.32, "learning_rate": 6.812673008203735e-05, "loss": 1.0238, "step": 127705 }, { "epoch": 0.32, "learning_rate": 6.812547184055564e-05, "loss": 1.0214, "step": 127710 }, { "epoch": 0.32, "learning_rate": 6.812421359907393e-05, "loss": 1.0203, "step": 127715 }, { "epoch": 0.32, "learning_rate": 6.812295535759223e-05, "loss": 1.0201, "step": 127720 }, { "epoch": 0.32, "learning_rate": 6.812169711611053e-05, "loss": 1.0198, "step": 127725 }, { "epoch": 0.32, "learning_rate": 6.812043887462882e-05, "loss": 1.0219, "step": 127730 }, { "epoch": 0.32, "learning_rate": 6.811918063314711e-05, "loss": 1.0229, "step": 127735 }, { "epoch": 0.32, "learning_rate": 6.81179223916654e-05, "loss": 1.0188, "step": 127740 }, { "epoch": 0.32, "learning_rate": 6.811666415018371e-05, "loss": 1.0192, "step": 127745 }, { "epoch": 0.32, "learning_rate": 6.8115405908702e-05, "loss": 1.0204, "step": 127750 }, { "epoch": 0.32, "learning_rate": 6.81141476672203e-05, "loss": 1.0408, "step": 127755 }, { "epoch": 0.32, "learning_rate": 6.811288942573859e-05, "loss": 1.022, "step": 127760 }, { "epoch": 0.32, "learning_rate": 6.811163118425689e-05, "loss": 1.0214, "step": 127765 }, { "epoch": 0.32, "learning_rate": 6.811037294277518e-05, "loss": 1.0221, "step": 127770 }, { "epoch": 0.32, "learning_rate": 6.810911470129347e-05, "loss": 1.0213, "step": 127775 }, { "epoch": 0.32, "learning_rate": 6.810785645981177e-05, "loss": 1.0194, "step": 127780 }, { "epoch": 0.32, "learning_rate": 6.810659821833006e-05, "loss": 1.0195, "step": 127785 }, { "epoch": 0.32, "learning_rate": 6.810533997684836e-05, "loss": 1.0216, "step": 127790 }, { "epoch": 0.32, "learning_rate": 6.810408173536665e-05, "loss": 1.0229, "step": 127795 }, { "epoch": 0.32, "learning_rate": 6.810282349388495e-05, "loss": 1.0212, "step": 127800 }, { "epoch": 0.32, "learning_rate": 6.810156525240324e-05, "loss": 1.0229, "step": 127805 }, { "epoch": 0.32, "learning_rate": 6.810030701092154e-05, "loss": 1.0233, "step": 127810 }, { "epoch": 0.32, "learning_rate": 6.809904876943983e-05, "loss": 1.0238, "step": 127815 }, { "epoch": 0.32, "learning_rate": 6.809779052795813e-05, "loss": 1.0203, "step": 127820 }, { "epoch": 0.32, "learning_rate": 6.809653228647642e-05, "loss": 1.0225, "step": 127825 }, { "epoch": 0.32, "learning_rate": 6.809527404499472e-05, "loss": 1.0459, "step": 127830 }, { "epoch": 0.32, "learning_rate": 6.809401580351301e-05, "loss": 1.0222, "step": 127835 }, { "epoch": 0.32, "learning_rate": 6.80927575620313e-05, "loss": 1.0214, "step": 127840 }, { "epoch": 0.32, "learning_rate": 6.80914993205496e-05, "loss": 1.0196, "step": 127845 }, { "epoch": 0.32, "learning_rate": 6.809024107906789e-05, "loss": 1.0224, "step": 127850 }, { "epoch": 0.32, "learning_rate": 6.80889828375862e-05, "loss": 1.0233, "step": 127855 }, { "epoch": 0.32, "learning_rate": 6.808772459610449e-05, "loss": 1.0225, "step": 127860 }, { "epoch": 0.32, "learning_rate": 6.808646635462278e-05, "loss": 1.022, "step": 127865 }, { "epoch": 0.32, "learning_rate": 6.808520811314107e-05, "loss": 1.023, "step": 127870 }, { "epoch": 0.32, "learning_rate": 6.808394987165937e-05, "loss": 1.0206, "step": 127875 }, { "epoch": 0.32, "learning_rate": 6.808269163017767e-05, "loss": 1.0219, "step": 127880 }, { "epoch": 0.32, "learning_rate": 6.808143338869596e-05, "loss": 1.0219, "step": 127885 }, { "epoch": 0.32, "learning_rate": 6.808017514721425e-05, "loss": 1.0202, "step": 127890 }, { "epoch": 0.32, "learning_rate": 6.807891690573255e-05, "loss": 1.0201, "step": 127895 }, { "epoch": 0.32, "learning_rate": 6.807765866425085e-05, "loss": 1.0222, "step": 127900 }, { "epoch": 0.32, "learning_rate": 6.807640042276914e-05, "loss": 1.0209, "step": 127905 }, { "epoch": 0.32, "learning_rate": 6.807514218128743e-05, "loss": 1.0204, "step": 127910 }, { "epoch": 0.32, "learning_rate": 6.807388393980572e-05, "loss": 1.0202, "step": 127915 }, { "epoch": 0.32, "learning_rate": 6.807262569832403e-05, "loss": 1.0248, "step": 127920 }, { "epoch": 0.32, "learning_rate": 6.807136745684232e-05, "loss": 1.0212, "step": 127925 }, { "epoch": 0.32, "learning_rate": 6.807010921536061e-05, "loss": 1.0217, "step": 127930 }, { "epoch": 0.32, "learning_rate": 6.80688509738789e-05, "loss": 1.0224, "step": 127935 }, { "epoch": 0.32, "learning_rate": 6.80675927323972e-05, "loss": 1.0213, "step": 127940 }, { "epoch": 0.32, "learning_rate": 6.80663344909155e-05, "loss": 1.0212, "step": 127945 }, { "epoch": 0.32, "learning_rate": 6.806507624943379e-05, "loss": 1.0229, "step": 127950 }, { "epoch": 0.32, "learning_rate": 6.806381800795208e-05, "loss": 1.0231, "step": 127955 }, { "epoch": 0.32, "learning_rate": 6.806255976647039e-05, "loss": 1.0205, "step": 127960 }, { "epoch": 0.32, "learning_rate": 6.806130152498868e-05, "loss": 1.0224, "step": 127965 }, { "epoch": 0.32, "learning_rate": 6.806004328350697e-05, "loss": 1.0197, "step": 127970 }, { "epoch": 0.32, "learning_rate": 6.805878504202526e-05, "loss": 1.0221, "step": 127975 }, { "epoch": 0.32, "learning_rate": 6.805752680054355e-05, "loss": 1.0203, "step": 127980 }, { "epoch": 0.32, "learning_rate": 6.805626855906186e-05, "loss": 1.0207, "step": 127985 }, { "epoch": 0.32, "learning_rate": 6.805501031758015e-05, "loss": 1.02, "step": 127990 }, { "epoch": 0.32, "learning_rate": 6.805375207609844e-05, "loss": 1.0219, "step": 127995 }, { "epoch": 0.32, "learning_rate": 6.805249383461675e-05, "loss": 1.0227, "step": 128000 }, { "epoch": 0.32, "learning_rate": 6.805123559313504e-05, "loss": 1.0211, "step": 128005 }, { "epoch": 0.32, "learning_rate": 6.804997735165334e-05, "loss": 1.0358, "step": 128010 }, { "epoch": 0.32, "learning_rate": 6.804871911017163e-05, "loss": 1.0202, "step": 128015 }, { "epoch": 0.32, "learning_rate": 6.804746086868993e-05, "loss": 1.0185, "step": 128020 }, { "epoch": 0.32, "learning_rate": 6.804620262720822e-05, "loss": 1.0224, "step": 128025 }, { "epoch": 0.32, "learning_rate": 6.804494438572652e-05, "loss": 1.0184, "step": 128030 }, { "epoch": 0.32, "learning_rate": 6.804368614424481e-05, "loss": 1.0205, "step": 128035 }, { "epoch": 0.32, "learning_rate": 6.80424279027631e-05, "loss": 1.0213, "step": 128040 }, { "epoch": 0.32, "learning_rate": 6.80411696612814e-05, "loss": 1.0204, "step": 128045 }, { "epoch": 0.32, "learning_rate": 6.80399114197997e-05, "loss": 1.0185, "step": 128050 }, { "epoch": 0.32, "learning_rate": 6.8038653178318e-05, "loss": 1.0246, "step": 128055 }, { "epoch": 0.32, "learning_rate": 6.803739493683629e-05, "loss": 1.0226, "step": 128060 }, { "epoch": 0.32, "learning_rate": 6.803613669535458e-05, "loss": 1.0231, "step": 128065 }, { "epoch": 0.32, "learning_rate": 6.803487845387287e-05, "loss": 1.0219, "step": 128070 }, { "epoch": 0.32, "learning_rate": 6.803362021239117e-05, "loss": 1.0213, "step": 128075 }, { "epoch": 0.32, "learning_rate": 6.803236197090947e-05, "loss": 1.0209, "step": 128080 }, { "epoch": 0.32, "learning_rate": 6.803110372942776e-05, "loss": 1.0238, "step": 128085 }, { "epoch": 0.32, "learning_rate": 6.802984548794605e-05, "loss": 1.0221, "step": 128090 }, { "epoch": 0.32, "learning_rate": 6.802858724646435e-05, "loss": 1.0213, "step": 128095 }, { "epoch": 0.32, "learning_rate": 6.802732900498265e-05, "loss": 1.0216, "step": 128100 }, { "epoch": 0.32, "learning_rate": 6.802607076350094e-05, "loss": 1.0247, "step": 128105 }, { "epoch": 0.32, "learning_rate": 6.802481252201923e-05, "loss": 1.022, "step": 128110 }, { "epoch": 0.32, "learning_rate": 6.802355428053753e-05, "loss": 1.0215, "step": 128115 }, { "epoch": 0.32, "learning_rate": 6.802229603905583e-05, "loss": 1.0215, "step": 128120 }, { "epoch": 0.32, "learning_rate": 6.802103779757412e-05, "loss": 1.0226, "step": 128125 }, { "epoch": 0.32, "learning_rate": 6.801977955609241e-05, "loss": 1.0228, "step": 128130 }, { "epoch": 0.32, "learning_rate": 6.80185213146107e-05, "loss": 1.0223, "step": 128135 }, { "epoch": 0.32, "learning_rate": 6.8017263073129e-05, "loss": 1.0236, "step": 128140 }, { "epoch": 0.32, "learning_rate": 6.80160048316473e-05, "loss": 1.0213, "step": 128145 }, { "epoch": 0.32, "learning_rate": 6.801474659016559e-05, "loss": 1.0215, "step": 128150 }, { "epoch": 0.32, "learning_rate": 6.801348834868388e-05, "loss": 1.0406, "step": 128155 }, { "epoch": 0.32, "learning_rate": 6.801223010720219e-05, "loss": 1.0195, "step": 128160 }, { "epoch": 0.32, "learning_rate": 6.801097186572048e-05, "loss": 1.0226, "step": 128165 }, { "epoch": 0.32, "learning_rate": 6.800971362423877e-05, "loss": 1.0215, "step": 128170 }, { "epoch": 0.32, "learning_rate": 6.800845538275706e-05, "loss": 1.024, "step": 128175 }, { "epoch": 0.32, "learning_rate": 6.800719714127537e-05, "loss": 1.0214, "step": 128180 }, { "epoch": 0.32, "learning_rate": 6.800593889979366e-05, "loss": 1.0204, "step": 128185 }, { "epoch": 0.32, "learning_rate": 6.800468065831195e-05, "loss": 1.0211, "step": 128190 }, { "epoch": 0.32, "learning_rate": 6.800342241683024e-05, "loss": 1.0221, "step": 128195 }, { "epoch": 0.32, "learning_rate": 6.800216417534853e-05, "loss": 1.0225, "step": 128200 }, { "epoch": 0.32, "learning_rate": 6.800090593386684e-05, "loss": 1.0219, "step": 128205 }, { "epoch": 0.32, "learning_rate": 6.799964769238513e-05, "loss": 1.0194, "step": 128210 }, { "epoch": 0.32, "learning_rate": 6.799838945090342e-05, "loss": 1.0201, "step": 128215 }, { "epoch": 0.32, "learning_rate": 6.799713120942171e-05, "loss": 1.0203, "step": 128220 }, { "epoch": 0.32, "learning_rate": 6.799587296794002e-05, "loss": 1.0219, "step": 128225 }, { "epoch": 0.32, "learning_rate": 6.799461472645831e-05, "loss": 1.0225, "step": 128230 }, { "epoch": 0.32, "learning_rate": 6.79933564849766e-05, "loss": 1.0214, "step": 128235 }, { "epoch": 0.32, "learning_rate": 6.799209824349489e-05, "loss": 1.0219, "step": 128240 }, { "epoch": 0.32, "learning_rate": 6.79908400020132e-05, "loss": 1.0243, "step": 128245 }, { "epoch": 0.32, "learning_rate": 6.798958176053149e-05, "loss": 1.0198, "step": 128250 }, { "epoch": 0.32, "learning_rate": 6.798832351904978e-05, "loss": 1.023, "step": 128255 }, { "epoch": 0.32, "learning_rate": 6.798706527756807e-05, "loss": 1.0217, "step": 128260 }, { "epoch": 0.32, "learning_rate": 6.798580703608636e-05, "loss": 1.0229, "step": 128265 }, { "epoch": 0.32, "learning_rate": 6.798454879460467e-05, "loss": 1.021, "step": 128270 }, { "epoch": 0.32, "learning_rate": 6.798329055312296e-05, "loss": 1.0213, "step": 128275 }, { "epoch": 0.32, "learning_rate": 6.798203231164125e-05, "loss": 1.0222, "step": 128280 }, { "epoch": 0.32, "learning_rate": 6.798077407015954e-05, "loss": 1.0197, "step": 128285 }, { "epoch": 0.32, "learning_rate": 6.797951582867785e-05, "loss": 1.0225, "step": 128290 }, { "epoch": 0.32, "learning_rate": 6.797825758719614e-05, "loss": 1.0204, "step": 128295 }, { "epoch": 0.32, "learning_rate": 6.797699934571443e-05, "loss": 1.0213, "step": 128300 }, { "epoch": 0.32, "learning_rate": 6.797574110423272e-05, "loss": 1.0219, "step": 128305 }, { "epoch": 0.32, "learning_rate": 6.797448286275101e-05, "loss": 1.0208, "step": 128310 }, { "epoch": 0.32, "learning_rate": 6.797322462126932e-05, "loss": 1.0214, "step": 128315 }, { "epoch": 0.32, "learning_rate": 6.797221802808396e-05, "loss": 1.0194, "step": 128320 }, { "epoch": 0.32, "learning_rate": 6.797095978660225e-05, "loss": 1.0206, "step": 128325 }, { "epoch": 0.32, "learning_rate": 6.796970154512055e-05, "loss": 1.0228, "step": 128330 }, { "epoch": 0.32, "learning_rate": 6.796844330363884e-05, "loss": 1.021, "step": 128335 }, { "epoch": 0.32, "learning_rate": 6.796718506215714e-05, "loss": 1.0217, "step": 128340 }, { "epoch": 0.32, "learning_rate": 6.796592682067543e-05, "loss": 1.0203, "step": 128345 }, { "epoch": 0.32, "learning_rate": 6.796466857919372e-05, "loss": 1.0194, "step": 128350 }, { "epoch": 0.32, "learning_rate": 6.796341033771202e-05, "loss": 1.0226, "step": 128355 }, { "epoch": 0.32, "learning_rate": 6.796215209623032e-05, "loss": 1.0211, "step": 128360 }, { "epoch": 0.32, "learning_rate": 6.796089385474861e-05, "loss": 1.0223, "step": 128365 }, { "epoch": 0.32, "learning_rate": 6.79596356132669e-05, "loss": 1.0209, "step": 128370 }, { "epoch": 0.32, "learning_rate": 6.79583773717852e-05, "loss": 1.0229, "step": 128375 }, { "epoch": 0.32, "learning_rate": 6.79571191303035e-05, "loss": 1.0214, "step": 128380 }, { "epoch": 0.32, "learning_rate": 6.795586088882179e-05, "loss": 1.0222, "step": 128385 }, { "epoch": 0.32, "learning_rate": 6.795460264734008e-05, "loss": 1.0188, "step": 128390 }, { "epoch": 0.32, "learning_rate": 6.795334440585837e-05, "loss": 1.0202, "step": 128395 }, { "epoch": 0.32, "learning_rate": 6.795208616437668e-05, "loss": 1.0216, "step": 128400 }, { "epoch": 0.32, "learning_rate": 6.795082792289497e-05, "loss": 1.0233, "step": 128405 }, { "epoch": 0.32, "learning_rate": 6.794956968141326e-05, "loss": 1.0191, "step": 128410 }, { "epoch": 0.32, "learning_rate": 6.794831143993155e-05, "loss": 1.0201, "step": 128415 }, { "epoch": 0.32, "learning_rate": 6.794705319844986e-05, "loss": 1.0217, "step": 128420 }, { "epoch": 0.32, "learning_rate": 6.794579495696815e-05, "loss": 1.023, "step": 128425 }, { "epoch": 0.32, "learning_rate": 6.794453671548644e-05, "loss": 1.0204, "step": 128430 }, { "epoch": 0.32, "learning_rate": 6.794327847400473e-05, "loss": 1.0217, "step": 128435 }, { "epoch": 0.32, "learning_rate": 6.794202023252304e-05, "loss": 1.0443, "step": 128440 }, { "epoch": 0.32, "learning_rate": 6.794076199104133e-05, "loss": 1.0235, "step": 128445 }, { "epoch": 0.32, "learning_rate": 6.793950374955962e-05, "loss": 1.0219, "step": 128450 }, { "epoch": 0.32, "learning_rate": 6.793824550807791e-05, "loss": 1.0242, "step": 128455 }, { "epoch": 0.32, "learning_rate": 6.79369872665962e-05, "loss": 1.0207, "step": 128460 }, { "epoch": 0.32, "learning_rate": 6.79357290251145e-05, "loss": 1.021, "step": 128465 }, { "epoch": 0.32, "learning_rate": 6.79344707836328e-05, "loss": 1.0215, "step": 128470 }, { "epoch": 0.32, "learning_rate": 6.793321254215109e-05, "loss": 1.0223, "step": 128475 }, { "epoch": 0.32, "learning_rate": 6.793195430066938e-05, "loss": 1.0224, "step": 128480 }, { "epoch": 0.32, "learning_rate": 6.793069605918769e-05, "loss": 1.0205, "step": 128485 }, { "epoch": 0.32, "learning_rate": 6.792943781770598e-05, "loss": 1.0228, "step": 128490 }, { "epoch": 0.32, "learning_rate": 6.792817957622427e-05, "loss": 1.0214, "step": 128495 }, { "epoch": 0.32, "learning_rate": 6.792692133474256e-05, "loss": 1.0216, "step": 128500 }, { "epoch": 0.32, "learning_rate": 6.792566309326087e-05, "loss": 1.0238, "step": 128505 }, { "epoch": 0.32, "learning_rate": 6.792440485177916e-05, "loss": 1.0202, "step": 128510 }, { "epoch": 0.32, "learning_rate": 6.792314661029745e-05, "loss": 1.0231, "step": 128515 }, { "epoch": 0.32, "learning_rate": 6.792188836881574e-05, "loss": 1.0221, "step": 128520 }, { "epoch": 0.32, "learning_rate": 6.792063012733403e-05, "loss": 1.0231, "step": 128525 }, { "epoch": 0.32, "learning_rate": 6.791937188585234e-05, "loss": 1.0367, "step": 128530 }, { "epoch": 0.32, "learning_rate": 6.791811364437063e-05, "loss": 1.022, "step": 128535 }, { "epoch": 0.32, "learning_rate": 6.791685540288892e-05, "loss": 1.021, "step": 128540 }, { "epoch": 0.32, "learning_rate": 6.791559716140721e-05, "loss": 1.0207, "step": 128545 }, { "epoch": 0.32, "learning_rate": 6.791433891992552e-05, "loss": 1.0227, "step": 128550 }, { "epoch": 0.32, "learning_rate": 6.791308067844381e-05, "loss": 1.0219, "step": 128555 }, { "epoch": 0.32, "learning_rate": 6.79118224369621e-05, "loss": 1.0198, "step": 128560 }, { "epoch": 0.32, "learning_rate": 6.791056419548039e-05, "loss": 1.0408, "step": 128565 }, { "epoch": 0.32, "learning_rate": 6.79093059539987e-05, "loss": 1.0206, "step": 128570 }, { "epoch": 0.32, "learning_rate": 6.790804771251699e-05, "loss": 1.0217, "step": 128575 }, { "epoch": 0.32, "learning_rate": 6.790678947103528e-05, "loss": 1.0198, "step": 128580 }, { "epoch": 0.32, "learning_rate": 6.790553122955357e-05, "loss": 1.022, "step": 128585 }, { "epoch": 0.32, "learning_rate": 6.790427298807186e-05, "loss": 1.021, "step": 128590 }, { "epoch": 0.32, "learning_rate": 6.790301474659017e-05, "loss": 1.0479, "step": 128595 }, { "epoch": 0.32, "learning_rate": 6.790175650510846e-05, "loss": 1.0217, "step": 128600 }, { "epoch": 0.32, "learning_rate": 6.790049826362675e-05, "loss": 1.0229, "step": 128605 }, { "epoch": 0.32, "learning_rate": 6.789924002214504e-05, "loss": 1.0194, "step": 128610 }, { "epoch": 0.32, "learning_rate": 6.789798178066335e-05, "loss": 1.0214, "step": 128615 }, { "epoch": 0.32, "learning_rate": 6.789672353918164e-05, "loss": 1.0208, "step": 128620 }, { "epoch": 0.32, "learning_rate": 6.789546529769993e-05, "loss": 1.0229, "step": 128625 }, { "epoch": 0.32, "learning_rate": 6.789420705621822e-05, "loss": 1.0236, "step": 128630 }, { "epoch": 0.32, "learning_rate": 6.789294881473653e-05, "loss": 1.022, "step": 128635 }, { "epoch": 0.32, "learning_rate": 6.789169057325482e-05, "loss": 1.0207, "step": 128640 }, { "epoch": 0.32, "learning_rate": 6.789043233177311e-05, "loss": 1.0223, "step": 128645 }, { "epoch": 0.32, "learning_rate": 6.78891740902914e-05, "loss": 1.0429, "step": 128650 }, { "epoch": 0.32, "learning_rate": 6.78879158488097e-05, "loss": 1.0194, "step": 128655 }, { "epoch": 0.32, "learning_rate": 6.7886657607328e-05, "loss": 1.0217, "step": 128660 }, { "epoch": 0.32, "learning_rate": 6.788539936584629e-05, "loss": 1.0208, "step": 128665 }, { "epoch": 0.32, "learning_rate": 6.788414112436458e-05, "loss": 1.0218, "step": 128670 }, { "epoch": 0.32, "learning_rate": 6.788288288288288e-05, "loss": 1.0216, "step": 128675 }, { "epoch": 0.32, "learning_rate": 6.788162464140118e-05, "loss": 1.0208, "step": 128680 }, { "epoch": 0.32, "learning_rate": 6.788036639991947e-05, "loss": 1.0198, "step": 128685 }, { "epoch": 0.32, "learning_rate": 6.787910815843776e-05, "loss": 1.0213, "step": 128690 }, { "epoch": 0.32, "learning_rate": 6.787784991695606e-05, "loss": 1.0207, "step": 128695 }, { "epoch": 0.32, "learning_rate": 6.787659167547436e-05, "loss": 1.021, "step": 128700 }, { "epoch": 0.32, "learning_rate": 6.787533343399267e-05, "loss": 1.0192, "step": 128705 }, { "epoch": 0.32, "learning_rate": 6.787407519251096e-05, "loss": 1.0211, "step": 128710 }, { "epoch": 0.32, "learning_rate": 6.787281695102925e-05, "loss": 1.0213, "step": 128715 }, { "epoch": 0.32, "learning_rate": 6.787155870954754e-05, "loss": 1.0242, "step": 128720 }, { "epoch": 0.32, "learning_rate": 6.787030046806585e-05, "loss": 1.0216, "step": 128725 }, { "epoch": 0.32, "learning_rate": 6.786904222658414e-05, "loss": 1.0213, "step": 128730 }, { "epoch": 0.32, "learning_rate": 6.786778398510243e-05, "loss": 1.0199, "step": 128735 }, { "epoch": 0.32, "learning_rate": 6.786652574362072e-05, "loss": 1.0189, "step": 128740 }, { "epoch": 0.32, "learning_rate": 6.786526750213901e-05, "loss": 1.0228, "step": 128745 }, { "epoch": 0.32, "learning_rate": 6.786400926065732e-05, "loss": 1.0234, "step": 128750 }, { "epoch": 0.32, "learning_rate": 6.786275101917561e-05, "loss": 1.0218, "step": 128755 }, { "epoch": 0.32, "learning_rate": 6.78614927776939e-05, "loss": 1.0214, "step": 128760 }, { "epoch": 0.32, "learning_rate": 6.786023453621219e-05, "loss": 1.0206, "step": 128765 }, { "epoch": 0.32, "learning_rate": 6.78589762947305e-05, "loss": 1.02, "step": 128770 }, { "epoch": 0.32, "learning_rate": 6.785771805324879e-05, "loss": 1.0168, "step": 128775 }, { "epoch": 0.32, "learning_rate": 6.785645981176708e-05, "loss": 1.0222, "step": 128780 }, { "epoch": 0.32, "learning_rate": 6.785520157028537e-05, "loss": 1.0202, "step": 128785 }, { "epoch": 0.32, "learning_rate": 6.785394332880368e-05, "loss": 1.0199, "step": 128790 }, { "epoch": 0.32, "learning_rate": 6.785268508732197e-05, "loss": 1.02, "step": 128795 }, { "epoch": 0.32, "learning_rate": 6.785142684584026e-05, "loss": 1.0193, "step": 128800 }, { "epoch": 0.32, "learning_rate": 6.785016860435855e-05, "loss": 1.0237, "step": 128805 }, { "epoch": 0.32, "learning_rate": 6.784891036287684e-05, "loss": 1.0223, "step": 128810 }, { "epoch": 0.32, "learning_rate": 6.784765212139515e-05, "loss": 1.0208, "step": 128815 }, { "epoch": 0.32, "learning_rate": 6.784639387991344e-05, "loss": 1.0196, "step": 128820 }, { "epoch": 0.32, "learning_rate": 6.784513563843173e-05, "loss": 1.0223, "step": 128825 }, { "epoch": 0.32, "learning_rate": 6.784387739695002e-05, "loss": 1.0218, "step": 128830 }, { "epoch": 0.32, "learning_rate": 6.784261915546833e-05, "loss": 1.0222, "step": 128835 }, { "epoch": 0.32, "learning_rate": 6.784136091398662e-05, "loss": 1.0198, "step": 128840 }, { "epoch": 0.32, "learning_rate": 6.784010267250491e-05, "loss": 1.0198, "step": 128845 }, { "epoch": 0.32, "learning_rate": 6.78388444310232e-05, "loss": 1.0219, "step": 128850 }, { "epoch": 0.32, "learning_rate": 6.783758618954151e-05, "loss": 1.0219, "step": 128855 }, { "epoch": 0.32, "learning_rate": 6.78363279480598e-05, "loss": 1.0202, "step": 128860 }, { "epoch": 0.32, "learning_rate": 6.783506970657809e-05, "loss": 1.0229, "step": 128865 }, { "epoch": 0.32, "learning_rate": 6.783381146509638e-05, "loss": 1.021, "step": 128870 }, { "epoch": 0.32, "learning_rate": 6.783255322361468e-05, "loss": 1.0211, "step": 128875 }, { "epoch": 0.32, "learning_rate": 6.783129498213298e-05, "loss": 1.0223, "step": 128880 }, { "epoch": 0.32, "learning_rate": 6.783003674065127e-05, "loss": 1.0206, "step": 128885 }, { "epoch": 0.32, "learning_rate": 6.782877849916956e-05, "loss": 1.0225, "step": 128890 }, { "epoch": 0.32, "learning_rate": 6.782752025768786e-05, "loss": 1.0208, "step": 128895 }, { "epoch": 0.32, "learning_rate": 6.782626201620616e-05, "loss": 1.0211, "step": 128900 }, { "epoch": 0.32, "learning_rate": 6.782500377472445e-05, "loss": 1.022, "step": 128905 }, { "epoch": 0.32, "learning_rate": 6.782374553324274e-05, "loss": 1.0204, "step": 128910 }, { "epoch": 0.32, "learning_rate": 6.782248729176104e-05, "loss": 1.019, "step": 128915 }, { "epoch": 0.32, "learning_rate": 6.782122905027934e-05, "loss": 1.0204, "step": 128920 }, { "epoch": 0.32, "learning_rate": 6.781997080879763e-05, "loss": 1.0229, "step": 128925 }, { "epoch": 0.32, "learning_rate": 6.781871256731592e-05, "loss": 1.0236, "step": 128930 }, { "epoch": 0.32, "learning_rate": 6.781745432583421e-05, "loss": 1.0199, "step": 128935 }, { "epoch": 0.32, "learning_rate": 6.78161960843525e-05, "loss": 1.0222, "step": 128940 }, { "epoch": 0.32, "learning_rate": 6.781493784287081e-05, "loss": 1.0393, "step": 128945 }, { "epoch": 0.32, "learning_rate": 6.78136796013891e-05, "loss": 1.0214, "step": 128950 }, { "epoch": 0.32, "learning_rate": 6.78124213599074e-05, "loss": 1.0219, "step": 128955 }, { "epoch": 0.32, "learning_rate": 6.781116311842569e-05, "loss": 1.0184, "step": 128960 }, { "epoch": 0.32, "learning_rate": 6.780990487694399e-05, "loss": 1.0238, "step": 128965 }, { "epoch": 0.32, "learning_rate": 6.780864663546228e-05, "loss": 1.0243, "step": 128970 }, { "epoch": 0.32, "learning_rate": 6.780738839398057e-05, "loss": 1.0233, "step": 128975 }, { "epoch": 0.32, "learning_rate": 6.780613015249887e-05, "loss": 1.0202, "step": 128980 }, { "epoch": 0.32, "learning_rate": 6.780487191101717e-05, "loss": 1.0211, "step": 128985 }, { "epoch": 0.32, "learning_rate": 6.780361366953546e-05, "loss": 1.0192, "step": 128990 }, { "epoch": 0.32, "learning_rate": 6.780235542805375e-05, "loss": 1.0214, "step": 128995 }, { "epoch": 0.32, "learning_rate": 6.780109718657205e-05, "loss": 1.023, "step": 129000 }, { "epoch": 0.32, "learning_rate": 6.779983894509034e-05, "loss": 1.0204, "step": 129005 }, { "epoch": 0.32, "learning_rate": 6.779858070360864e-05, "loss": 1.0209, "step": 129010 }, { "epoch": 0.32, "learning_rate": 6.779732246212693e-05, "loss": 1.0229, "step": 129015 }, { "epoch": 0.32, "learning_rate": 6.779606422064523e-05, "loss": 1.022, "step": 129020 }, { "epoch": 0.32, "learning_rate": 6.779480597916352e-05, "loss": 1.0214, "step": 129025 }, { "epoch": 0.32, "learning_rate": 6.779354773768182e-05, "loss": 1.0211, "step": 129030 }, { "epoch": 0.32, "learning_rate": 6.779228949620011e-05, "loss": 1.0221, "step": 129035 }, { "epoch": 0.32, "learning_rate": 6.77910312547184e-05, "loss": 1.0199, "step": 129040 }, { "epoch": 0.32, "learning_rate": 6.77897730132367e-05, "loss": 1.0412, "step": 129045 }, { "epoch": 0.32, "learning_rate": 6.778851477175499e-05, "loss": 1.0231, "step": 129050 }, { "epoch": 0.32, "learning_rate": 6.77872565302733e-05, "loss": 1.0201, "step": 129055 }, { "epoch": 0.32, "learning_rate": 6.778599828879159e-05, "loss": 1.0205, "step": 129060 }, { "epoch": 0.32, "learning_rate": 6.778474004730988e-05, "loss": 1.0208, "step": 129065 }, { "epoch": 0.32, "learning_rate": 6.778348180582817e-05, "loss": 1.0211, "step": 129070 }, { "epoch": 0.32, "learning_rate": 6.778222356434647e-05, "loss": 1.0194, "step": 129075 }, { "epoch": 0.32, "learning_rate": 6.778096532286477e-05, "loss": 1.0215, "step": 129080 }, { "epoch": 0.32, "learning_rate": 6.777970708138306e-05, "loss": 1.0215, "step": 129085 }, { "epoch": 0.32, "learning_rate": 6.777844883990135e-05, "loss": 1.0215, "step": 129090 }, { "epoch": 0.32, "learning_rate": 6.777719059841965e-05, "loss": 1.0238, "step": 129095 }, { "epoch": 0.32, "learning_rate": 6.777593235693795e-05, "loss": 1.0218, "step": 129100 }, { "epoch": 0.32, "learning_rate": 6.777467411545624e-05, "loss": 1.0199, "step": 129105 }, { "epoch": 0.32, "learning_rate": 6.777341587397453e-05, "loss": 1.022, "step": 129110 }, { "epoch": 0.32, "learning_rate": 6.777215763249282e-05, "loss": 1.0232, "step": 129115 }, { "epoch": 0.32, "learning_rate": 6.777089939101113e-05, "loss": 1.0216, "step": 129120 }, { "epoch": 0.32, "learning_rate": 6.776964114952942e-05, "loss": 1.0213, "step": 129125 }, { "epoch": 0.32, "learning_rate": 6.776838290804771e-05, "loss": 1.0218, "step": 129130 }, { "epoch": 0.32, "learning_rate": 6.7767124666566e-05, "loss": 1.0217, "step": 129135 }, { "epoch": 0.32, "learning_rate": 6.77658664250843e-05, "loss": 1.0235, "step": 129140 }, { "epoch": 0.32, "learning_rate": 6.77646081836026e-05, "loss": 1.0211, "step": 129145 }, { "epoch": 0.32, "learning_rate": 6.776334994212089e-05, "loss": 1.0212, "step": 129150 }, { "epoch": 0.32, "learning_rate": 6.776209170063918e-05, "loss": 1.0203, "step": 129155 }, { "epoch": 0.32, "learning_rate": 6.776083345915749e-05, "loss": 1.0206, "step": 129160 }, { "epoch": 0.32, "learning_rate": 6.775957521767578e-05, "loss": 1.0211, "step": 129165 }, { "epoch": 0.32, "learning_rate": 6.775831697619407e-05, "loss": 1.0222, "step": 129170 }, { "epoch": 0.32, "learning_rate": 6.775705873471236e-05, "loss": 1.0212, "step": 129175 }, { "epoch": 0.32, "learning_rate": 6.775580049323065e-05, "loss": 1.0202, "step": 129180 }, { "epoch": 0.32, "learning_rate": 6.775454225174896e-05, "loss": 1.0237, "step": 129185 }, { "epoch": 0.32, "learning_rate": 6.775328401026725e-05, "loss": 1.0221, "step": 129190 }, { "epoch": 0.32, "learning_rate": 6.775202576878554e-05, "loss": 1.0198, "step": 129195 }, { "epoch": 0.32, "learning_rate": 6.775076752730385e-05, "loss": 1.0213, "step": 129200 }, { "epoch": 0.32, "learning_rate": 6.774950928582214e-05, "loss": 1.0229, "step": 129205 }, { "epoch": 0.32, "learning_rate": 6.774825104434044e-05, "loss": 1.0204, "step": 129210 }, { "epoch": 0.32, "learning_rate": 6.774699280285873e-05, "loss": 1.0227, "step": 129215 }, { "epoch": 0.32, "learning_rate": 6.774573456137703e-05, "loss": 1.0206, "step": 129220 }, { "epoch": 0.32, "learning_rate": 6.774447631989532e-05, "loss": 1.0196, "step": 129225 }, { "epoch": 0.32, "learning_rate": 6.774321807841362e-05, "loss": 1.022, "step": 129230 }, { "epoch": 0.32, "learning_rate": 6.774195983693191e-05, "loss": 1.0236, "step": 129235 }, { "epoch": 0.32, "learning_rate": 6.77407015954502e-05, "loss": 1.0223, "step": 129240 }, { "epoch": 0.32, "learning_rate": 6.77394433539685e-05, "loss": 1.0201, "step": 129245 }, { "epoch": 0.32, "learning_rate": 6.77381851124868e-05, "loss": 1.0229, "step": 129250 }, { "epoch": 0.32, "learning_rate": 6.77369268710051e-05, "loss": 1.0197, "step": 129255 }, { "epoch": 0.32, "learning_rate": 6.773566862952339e-05, "loss": 1.0211, "step": 129260 }, { "epoch": 0.32, "learning_rate": 6.773441038804168e-05, "loss": 1.0217, "step": 129265 }, { "epoch": 0.32, "learning_rate": 6.773315214655997e-05, "loss": 1.0208, "step": 129270 }, { "epoch": 0.32, "learning_rate": 6.773189390507827e-05, "loss": 1.0225, "step": 129275 }, { "epoch": 0.32, "learning_rate": 6.773063566359657e-05, "loss": 1.0232, "step": 129280 }, { "epoch": 0.32, "learning_rate": 6.772937742211486e-05, "loss": 1.0216, "step": 129285 }, { "epoch": 0.32, "learning_rate": 6.772811918063315e-05, "loss": 1.0209, "step": 129290 }, { "epoch": 0.32, "learning_rate": 6.772686093915145e-05, "loss": 1.0222, "step": 129295 }, { "epoch": 0.32, "learning_rate": 6.772560269766975e-05, "loss": 1.0215, "step": 129300 }, { "epoch": 0.32, "learning_rate": 6.772434445618804e-05, "loss": 1.0209, "step": 129305 }, { "epoch": 0.32, "learning_rate": 6.772308621470633e-05, "loss": 1.0212, "step": 129310 }, { "epoch": 0.32, "learning_rate": 6.772182797322463e-05, "loss": 1.0214, "step": 129315 }, { "epoch": 0.32, "learning_rate": 6.772056973174293e-05, "loss": 1.0203, "step": 129320 }, { "epoch": 0.32, "learning_rate": 6.771931149026122e-05, "loss": 1.0216, "step": 129325 }, { "epoch": 0.32, "learning_rate": 6.771805324877951e-05, "loss": 1.0217, "step": 129330 }, { "epoch": 0.32, "learning_rate": 6.77167950072978e-05, "loss": 1.021, "step": 129335 }, { "epoch": 0.32, "learning_rate": 6.77155367658161e-05, "loss": 1.019, "step": 129340 }, { "epoch": 0.32, "learning_rate": 6.77142785243344e-05, "loss": 1.02, "step": 129345 }, { "epoch": 0.32, "learning_rate": 6.771302028285269e-05, "loss": 1.0236, "step": 129350 }, { "epoch": 0.32, "learning_rate": 6.771176204137098e-05, "loss": 1.0191, "step": 129355 }, { "epoch": 0.32, "learning_rate": 6.771050379988929e-05, "loss": 1.0192, "step": 129360 }, { "epoch": 0.32, "learning_rate": 6.770924555840758e-05, "loss": 1.0197, "step": 129365 }, { "epoch": 0.32, "learning_rate": 6.770798731692587e-05, "loss": 1.0231, "step": 129370 }, { "epoch": 0.32, "learning_rate": 6.770672907544416e-05, "loss": 1.0225, "step": 129375 }, { "epoch": 0.32, "learning_rate": 6.770547083396247e-05, "loss": 1.0213, "step": 129380 }, { "epoch": 0.32, "learning_rate": 6.770421259248076e-05, "loss": 1.0228, "step": 129385 }, { "epoch": 0.32, "learning_rate": 6.770295435099905e-05, "loss": 1.0218, "step": 129390 }, { "epoch": 0.32, "learning_rate": 6.770169610951734e-05, "loss": 1.0217, "step": 129395 }, { "epoch": 0.32, "learning_rate": 6.770043786803563e-05, "loss": 1.02, "step": 129400 }, { "epoch": 0.32, "learning_rate": 6.769917962655394e-05, "loss": 1.0217, "step": 129405 }, { "epoch": 0.32, "learning_rate": 6.769792138507223e-05, "loss": 1.0222, "step": 129410 }, { "epoch": 0.32, "learning_rate": 6.769666314359052e-05, "loss": 1.0212, "step": 129415 }, { "epoch": 0.32, "learning_rate": 6.769540490210881e-05, "loss": 1.0212, "step": 129420 }, { "epoch": 0.32, "learning_rate": 6.769414666062712e-05, "loss": 1.0196, "step": 129425 }, { "epoch": 0.32, "learning_rate": 6.769288841914541e-05, "loss": 1.0381, "step": 129430 }, { "epoch": 0.32, "learning_rate": 6.76916301776637e-05, "loss": 1.0229, "step": 129435 }, { "epoch": 0.32, "learning_rate": 6.769037193618199e-05, "loss": 1.0229, "step": 129440 }, { "epoch": 0.32, "learning_rate": 6.76891136947003e-05, "loss": 1.0218, "step": 129445 }, { "epoch": 0.32, "learning_rate": 6.768785545321859e-05, "loss": 1.0229, "step": 129450 }, { "epoch": 0.32, "learning_rate": 6.768659721173688e-05, "loss": 1.0201, "step": 129455 }, { "epoch": 0.32, "learning_rate": 6.768533897025517e-05, "loss": 1.0219, "step": 129460 }, { "epoch": 0.32, "learning_rate": 6.768408072877346e-05, "loss": 1.0229, "step": 129465 }, { "epoch": 0.32, "learning_rate": 6.768282248729177e-05, "loss": 1.0212, "step": 129470 }, { "epoch": 0.33, "learning_rate": 6.768156424581006e-05, "loss": 1.0206, "step": 129475 }, { "epoch": 0.33, "learning_rate": 6.768030600432835e-05, "loss": 1.0216, "step": 129480 }, { "epoch": 0.33, "learning_rate": 6.767904776284664e-05, "loss": 1.0445, "step": 129485 }, { "epoch": 0.33, "learning_rate": 6.767778952136495e-05, "loss": 1.0222, "step": 129490 }, { "epoch": 0.33, "learning_rate": 6.767653127988324e-05, "loss": 1.0199, "step": 129495 }, { "epoch": 0.33, "learning_rate": 6.767527303840153e-05, "loss": 1.021, "step": 129500 }, { "epoch": 0.33, "learning_rate": 6.767401479691982e-05, "loss": 1.0187, "step": 129505 }, { "epoch": 0.33, "learning_rate": 6.767275655543813e-05, "loss": 1.02, "step": 129510 }, { "epoch": 0.33, "learning_rate": 6.767149831395642e-05, "loss": 1.021, "step": 129515 }, { "epoch": 0.33, "learning_rate": 6.767024007247471e-05, "loss": 1.0227, "step": 129520 }, { "epoch": 0.33, "learning_rate": 6.7668981830993e-05, "loss": 1.0489, "step": 129525 }, { "epoch": 0.33, "learning_rate": 6.76677235895113e-05, "loss": 1.0205, "step": 129530 }, { "epoch": 0.33, "learning_rate": 6.76664653480296e-05, "loss": 1.0222, "step": 129535 }, { "epoch": 0.33, "learning_rate": 6.766520710654789e-05, "loss": 1.0207, "step": 129540 }, { "epoch": 0.33, "learning_rate": 6.766394886506618e-05, "loss": 1.0202, "step": 129545 }, { "epoch": 0.33, "learning_rate": 6.766269062358447e-05, "loss": 1.0236, "step": 129550 }, { "epoch": 0.33, "learning_rate": 6.766143238210278e-05, "loss": 1.0217, "step": 129555 }, { "epoch": 0.33, "learning_rate": 6.766017414062107e-05, "loss": 1.0201, "step": 129560 }, { "epoch": 0.33, "learning_rate": 6.765891589913936e-05, "loss": 1.0212, "step": 129565 }, { "epoch": 0.33, "learning_rate": 6.765765765765765e-05, "loss": 1.0212, "step": 129570 }, { "epoch": 0.33, "learning_rate": 6.765639941617596e-05, "loss": 1.0231, "step": 129575 }, { "epoch": 0.33, "learning_rate": 6.765514117469425e-05, "loss": 1.0198, "step": 129580 }, { "epoch": 0.33, "learning_rate": 6.765388293321254e-05, "loss": 1.0233, "step": 129585 }, { "epoch": 0.33, "learning_rate": 6.765262469173083e-05, "loss": 1.0224, "step": 129590 }, { "epoch": 0.33, "learning_rate": 6.765136645024913e-05, "loss": 1.0212, "step": 129595 }, { "epoch": 0.33, "learning_rate": 6.765010820876743e-05, "loss": 1.022, "step": 129600 }, { "epoch": 0.33, "learning_rate": 6.764884996728572e-05, "loss": 1.0201, "step": 129605 }, { "epoch": 0.33, "learning_rate": 6.764759172580401e-05, "loss": 1.0212, "step": 129610 }, { "epoch": 0.33, "learning_rate": 6.76463334843223e-05, "loss": 1.0215, "step": 129615 }, { "epoch": 0.33, "learning_rate": 6.764507524284061e-05, "loss": 1.0213, "step": 129620 }, { "epoch": 0.33, "learning_rate": 6.76438170013589e-05, "loss": 1.0207, "step": 129625 }, { "epoch": 0.33, "learning_rate": 6.76425587598772e-05, "loss": 1.0225, "step": 129630 }, { "epoch": 0.33, "learning_rate": 6.764130051839549e-05, "loss": 1.0214, "step": 129635 }, { "epoch": 0.33, "learning_rate": 6.764004227691379e-05, "loss": 1.0198, "step": 129640 }, { "epoch": 0.33, "learning_rate": 6.763878403543208e-05, "loss": 1.0219, "step": 129645 }, { "epoch": 0.33, "learning_rate": 6.763752579395037e-05, "loss": 1.0445, "step": 129650 }, { "epoch": 0.33, "learning_rate": 6.763626755246867e-05, "loss": 1.0202, "step": 129655 }, { "epoch": 0.33, "learning_rate": 6.763500931098696e-05, "loss": 1.0198, "step": 129660 }, { "epoch": 0.33, "learning_rate": 6.763375106950526e-05, "loss": 1.0197, "step": 129665 }, { "epoch": 0.33, "learning_rate": 6.763249282802355e-05, "loss": 1.0209, "step": 129670 }, { "epoch": 0.33, "learning_rate": 6.763123458654185e-05, "loss": 1.0195, "step": 129675 }, { "epoch": 0.33, "learning_rate": 6.762997634506014e-05, "loss": 1.0202, "step": 129680 }, { "epoch": 0.33, "learning_rate": 6.762871810357844e-05, "loss": 1.0219, "step": 129685 }, { "epoch": 0.33, "learning_rate": 6.762745986209673e-05, "loss": 1.0209, "step": 129690 }, { "epoch": 0.33, "learning_rate": 6.762620162061503e-05, "loss": 1.0216, "step": 129695 }, { "epoch": 0.33, "learning_rate": 6.762494337913333e-05, "loss": 1.0218, "step": 129700 }, { "epoch": 0.33, "learning_rate": 6.762368513765162e-05, "loss": 1.0233, "step": 129705 }, { "epoch": 0.33, "learning_rate": 6.762242689616993e-05, "loss": 1.0232, "step": 129710 }, { "epoch": 0.33, "learning_rate": 6.762116865468822e-05, "loss": 1.0211, "step": 129715 }, { "epoch": 0.33, "learning_rate": 6.761991041320651e-05, "loss": 1.0197, "step": 129720 }, { "epoch": 0.33, "learning_rate": 6.76186521717248e-05, "loss": 1.0216, "step": 129725 }, { "epoch": 0.33, "learning_rate": 6.76173939302431e-05, "loss": 1.0442, "step": 129730 }, { "epoch": 0.33, "learning_rate": 6.76161356887614e-05, "loss": 1.0188, "step": 129735 }, { "epoch": 0.33, "learning_rate": 6.761487744727969e-05, "loss": 1.0228, "step": 129740 }, { "epoch": 0.33, "learning_rate": 6.761361920579798e-05, "loss": 1.0468, "step": 129745 }, { "epoch": 0.33, "learning_rate": 6.761236096431627e-05, "loss": 1.0219, "step": 129750 }, { "epoch": 0.33, "learning_rate": 6.761110272283458e-05, "loss": 1.0214, "step": 129755 }, { "epoch": 0.33, "learning_rate": 6.760984448135287e-05, "loss": 1.0222, "step": 129760 }, { "epoch": 0.33, "learning_rate": 6.760858623987116e-05, "loss": 1.0197, "step": 129765 }, { "epoch": 0.33, "learning_rate": 6.760732799838945e-05, "loss": 1.0205, "step": 129770 }, { "epoch": 0.33, "learning_rate": 6.760606975690776e-05, "loss": 1.0224, "step": 129775 }, { "epoch": 0.33, "learning_rate": 6.760481151542605e-05, "loss": 1.0229, "step": 129780 }, { "epoch": 0.33, "learning_rate": 6.760355327394434e-05, "loss": 1.0247, "step": 129785 }, { "epoch": 0.33, "learning_rate": 6.760229503246263e-05, "loss": 1.0232, "step": 129790 }, { "epoch": 0.33, "learning_rate": 6.760103679098093e-05, "loss": 1.0209, "step": 129795 }, { "epoch": 0.33, "learning_rate": 6.759977854949923e-05, "loss": 1.0201, "step": 129800 }, { "epoch": 0.33, "learning_rate": 6.759852030801752e-05, "loss": 1.0196, "step": 129805 }, { "epoch": 0.33, "learning_rate": 6.759726206653581e-05, "loss": 1.0211, "step": 129810 }, { "epoch": 0.33, "learning_rate": 6.75960038250541e-05, "loss": 1.0209, "step": 129815 }, { "epoch": 0.33, "learning_rate": 6.759474558357241e-05, "loss": 1.0213, "step": 129820 }, { "epoch": 0.33, "learning_rate": 6.75934873420907e-05, "loss": 1.0243, "step": 129825 }, { "epoch": 0.33, "learning_rate": 6.7592229100609e-05, "loss": 1.0223, "step": 129830 }, { "epoch": 0.33, "learning_rate": 6.759097085912729e-05, "loss": 1.021, "step": 129835 }, { "epoch": 0.33, "learning_rate": 6.758971261764559e-05, "loss": 1.0222, "step": 129840 }, { "epoch": 0.33, "learning_rate": 6.758845437616388e-05, "loss": 1.0215, "step": 129845 }, { "epoch": 0.33, "learning_rate": 6.758719613468217e-05, "loss": 1.0229, "step": 129850 }, { "epoch": 0.33, "learning_rate": 6.758593789320047e-05, "loss": 1.0208, "step": 129855 }, { "epoch": 0.33, "learning_rate": 6.758467965171876e-05, "loss": 1.0204, "step": 129860 }, { "epoch": 0.33, "learning_rate": 6.758342141023706e-05, "loss": 1.0207, "step": 129865 }, { "epoch": 0.33, "learning_rate": 6.758216316875535e-05, "loss": 1.0193, "step": 129870 }, { "epoch": 0.33, "learning_rate": 6.758090492727365e-05, "loss": 1.0199, "step": 129875 }, { "epoch": 0.33, "learning_rate": 6.757964668579194e-05, "loss": 1.0215, "step": 129880 }, { "epoch": 0.33, "learning_rate": 6.757838844431024e-05, "loss": 1.0222, "step": 129885 }, { "epoch": 0.33, "learning_rate": 6.757713020282853e-05, "loss": 1.0217, "step": 129890 }, { "epoch": 0.33, "learning_rate": 6.757587196134683e-05, "loss": 1.021, "step": 129895 }, { "epoch": 0.33, "learning_rate": 6.757461371986512e-05, "loss": 1.0218, "step": 129900 }, { "epoch": 0.33, "learning_rate": 6.757335547838342e-05, "loss": 1.0235, "step": 129905 }, { "epoch": 0.33, "learning_rate": 6.757209723690171e-05, "loss": 1.024, "step": 129910 }, { "epoch": 0.33, "learning_rate": 6.757083899542e-05, "loss": 1.0207, "step": 129915 }, { "epoch": 0.33, "learning_rate": 6.75695807539383e-05, "loss": 1.0228, "step": 129920 }, { "epoch": 0.33, "learning_rate": 6.756832251245659e-05, "loss": 1.0221, "step": 129925 }, { "epoch": 0.33, "learning_rate": 6.75670642709749e-05, "loss": 1.0212, "step": 129930 }, { "epoch": 0.33, "learning_rate": 6.756580602949318e-05, "loss": 1.0251, "step": 129935 }, { "epoch": 0.33, "learning_rate": 6.756454778801148e-05, "loss": 1.0229, "step": 129940 }, { "epoch": 0.33, "learning_rate": 6.756328954652977e-05, "loss": 1.036, "step": 129945 }, { "epoch": 0.33, "learning_rate": 6.756203130504807e-05, "loss": 1.0199, "step": 129950 }, { "epoch": 0.33, "learning_rate": 6.756077306356636e-05, "loss": 1.0205, "step": 129955 }, { "epoch": 0.33, "learning_rate": 6.755951482208466e-05, "loss": 1.0248, "step": 129960 }, { "epoch": 0.33, "learning_rate": 6.755825658060295e-05, "loss": 1.0212, "step": 129965 }, { "epoch": 0.33, "learning_rate": 6.755699833912125e-05, "loss": 1.0222, "step": 129970 }, { "epoch": 0.33, "learning_rate": 6.755574009763954e-05, "loss": 1.0229, "step": 129975 }, { "epoch": 0.33, "learning_rate": 6.755448185615784e-05, "loss": 1.0176, "step": 129980 }, { "epoch": 0.33, "learning_rate": 6.755322361467613e-05, "loss": 1.0189, "step": 129985 }, { "epoch": 0.33, "learning_rate": 6.755196537319442e-05, "loss": 1.0412, "step": 129990 }, { "epoch": 0.33, "learning_rate": 6.755070713171272e-05, "loss": 1.0201, "step": 129995 }, { "epoch": 0.33, "learning_rate": 6.754944889023102e-05, "loss": 1.0197, "step": 130000 }, { "epoch": 0.33, "learning_rate": 6.754819064874931e-05, "loss": 1.0206, "step": 130005 }, { "epoch": 0.33, "learning_rate": 6.75469324072676e-05, "loss": 1.0231, "step": 130010 }, { "epoch": 0.33, "learning_rate": 6.75456741657859e-05, "loss": 1.0195, "step": 130015 }, { "epoch": 0.33, "learning_rate": 6.75444159243042e-05, "loss": 1.0238, "step": 130020 }, { "epoch": 0.33, "learning_rate": 6.754315768282249e-05, "loss": 1.0214, "step": 130025 }, { "epoch": 0.33, "learning_rate": 6.754189944134078e-05, "loss": 1.0215, "step": 130030 }, { "epoch": 0.33, "learning_rate": 6.754064119985908e-05, "loss": 1.0228, "step": 130035 }, { "epoch": 0.33, "learning_rate": 6.753938295837738e-05, "loss": 1.0224, "step": 130040 }, { "epoch": 0.33, "learning_rate": 6.753812471689567e-05, "loss": 1.0216, "step": 130045 }, { "epoch": 0.33, "learning_rate": 6.753686647541396e-05, "loss": 1.0217, "step": 130050 }, { "epoch": 0.33, "learning_rate": 6.753560823393225e-05, "loss": 1.021, "step": 130055 }, { "epoch": 0.33, "learning_rate": 6.753434999245056e-05, "loss": 1.022, "step": 130060 }, { "epoch": 0.33, "learning_rate": 6.753309175096885e-05, "loss": 1.0234, "step": 130065 }, { "epoch": 0.33, "learning_rate": 6.753183350948714e-05, "loss": 1.0199, "step": 130070 }, { "epoch": 0.33, "learning_rate": 6.753057526800543e-05, "loss": 1.023, "step": 130075 }, { "epoch": 0.33, "learning_rate": 6.752931702652374e-05, "loss": 1.0214, "step": 130080 }, { "epoch": 0.33, "learning_rate": 6.752805878504203e-05, "loss": 1.0217, "step": 130085 }, { "epoch": 0.33, "learning_rate": 6.752680054356032e-05, "loss": 1.02, "step": 130090 }, { "epoch": 0.33, "learning_rate": 6.752554230207861e-05, "loss": 1.0218, "step": 130095 }, { "epoch": 0.33, "learning_rate": 6.752428406059692e-05, "loss": 1.0216, "step": 130100 }, { "epoch": 0.33, "learning_rate": 6.752302581911521e-05, "loss": 1.0181, "step": 130105 }, { "epoch": 0.33, "learning_rate": 6.75217675776335e-05, "loss": 1.021, "step": 130110 }, { "epoch": 0.33, "learning_rate": 6.752050933615179e-05, "loss": 1.0203, "step": 130115 }, { "epoch": 0.33, "learning_rate": 6.751925109467008e-05, "loss": 1.0197, "step": 130120 }, { "epoch": 0.33, "learning_rate": 6.751799285318839e-05, "loss": 1.0209, "step": 130125 }, { "epoch": 0.33, "learning_rate": 6.751673461170668e-05, "loss": 1.0211, "step": 130130 }, { "epoch": 0.33, "learning_rate": 6.751547637022497e-05, "loss": 1.0212, "step": 130135 }, { "epoch": 0.33, "learning_rate": 6.751421812874326e-05, "loss": 1.0185, "step": 130140 }, { "epoch": 0.33, "learning_rate": 6.751295988726157e-05, "loss": 1.022, "step": 130145 }, { "epoch": 0.33, "learning_rate": 6.751170164577986e-05, "loss": 1.0229, "step": 130150 }, { "epoch": 0.33, "learning_rate": 6.751044340429815e-05, "loss": 1.0467, "step": 130155 }, { "epoch": 0.33, "learning_rate": 6.750918516281644e-05, "loss": 1.0197, "step": 130160 }, { "epoch": 0.33, "learning_rate": 6.750792692133475e-05, "loss": 1.0216, "step": 130165 }, { "epoch": 0.33, "learning_rate": 6.750666867985304e-05, "loss": 1.0219, "step": 130170 }, { "epoch": 0.33, "learning_rate": 6.750541043837133e-05, "loss": 1.0223, "step": 130175 }, { "epoch": 0.33, "learning_rate": 6.750415219688962e-05, "loss": 1.018, "step": 130180 }, { "epoch": 0.33, "learning_rate": 6.750289395540791e-05, "loss": 1.0228, "step": 130185 }, { "epoch": 0.33, "learning_rate": 6.750163571392622e-05, "loss": 1.0216, "step": 130190 }, { "epoch": 0.33, "learning_rate": 6.750037747244451e-05, "loss": 1.0191, "step": 130195 }, { "epoch": 0.33, "learning_rate": 6.749911923096282e-05, "loss": 1.022, "step": 130200 }, { "epoch": 0.33, "learning_rate": 6.749786098948111e-05, "loss": 1.0216, "step": 130205 }, { "epoch": 0.33, "learning_rate": 6.74966027479994e-05, "loss": 1.0224, "step": 130210 }, { "epoch": 0.33, "learning_rate": 6.74953445065177e-05, "loss": 1.0198, "step": 130215 }, { "epoch": 0.33, "learning_rate": 6.7494086265036e-05, "loss": 1.0439, "step": 130220 }, { "epoch": 0.33, "learning_rate": 6.749282802355429e-05, "loss": 1.0194, "step": 130225 }, { "epoch": 0.33, "learning_rate": 6.749156978207258e-05, "loss": 1.0191, "step": 130230 }, { "epoch": 0.33, "learning_rate": 6.749031154059088e-05, "loss": 1.0204, "step": 130235 }, { "epoch": 0.33, "learning_rate": 6.748905329910918e-05, "loss": 1.0229, "step": 130240 }, { "epoch": 0.33, "learning_rate": 6.748779505762747e-05, "loss": 1.0206, "step": 130245 }, { "epoch": 0.33, "learning_rate": 6.748653681614576e-05, "loss": 1.0228, "step": 130250 }, { "epoch": 0.33, "learning_rate": 6.748527857466405e-05, "loss": 1.0223, "step": 130255 }, { "epoch": 0.33, "learning_rate": 6.748402033318236e-05, "loss": 1.0442, "step": 130260 }, { "epoch": 0.33, "learning_rate": 6.748276209170065e-05, "loss": 1.0251, "step": 130265 }, { "epoch": 0.33, "learning_rate": 6.748150385021894e-05, "loss": 1.0228, "step": 130270 }, { "epoch": 0.33, "learning_rate": 6.748024560873723e-05, "loss": 1.0231, "step": 130275 }, { "epoch": 0.33, "learning_rate": 6.747898736725554e-05, "loss": 1.0203, "step": 130280 }, { "epoch": 0.33, "learning_rate": 6.747772912577383e-05, "loss": 1.021, "step": 130285 }, { "epoch": 0.33, "learning_rate": 6.747647088429212e-05, "loss": 1.0226, "step": 130290 }, { "epoch": 0.33, "learning_rate": 6.747521264281041e-05, "loss": 1.0214, "step": 130295 }, { "epoch": 0.33, "learning_rate": 6.747395440132872e-05, "loss": 1.0207, "step": 130300 }, { "epoch": 0.33, "learning_rate": 6.747269615984701e-05, "loss": 1.0235, "step": 130305 }, { "epoch": 0.33, "learning_rate": 6.74714379183653e-05, "loss": 1.0212, "step": 130310 }, { "epoch": 0.33, "learning_rate": 6.747017967688359e-05, "loss": 1.0207, "step": 130315 }, { "epoch": 0.33, "learning_rate": 6.746892143540188e-05, "loss": 1.0192, "step": 130320 }, { "epoch": 0.33, "learning_rate": 6.746791484221652e-05, "loss": 1.0221, "step": 130325 }, { "epoch": 0.33, "learning_rate": 6.746665660073481e-05, "loss": 1.0388, "step": 130330 }, { "epoch": 0.33, "learning_rate": 6.74653983592531e-05, "loss": 1.0199, "step": 130335 }, { "epoch": 0.33, "learning_rate": 6.74641401177714e-05, "loss": 1.0202, "step": 130340 }, { "epoch": 0.33, "learning_rate": 6.74628818762897e-05, "loss": 1.0185, "step": 130345 }, { "epoch": 0.33, "learning_rate": 6.746162363480799e-05, "loss": 1.0225, "step": 130350 }, { "epoch": 0.33, "learning_rate": 6.746036539332628e-05, "loss": 1.022, "step": 130355 }, { "epoch": 0.33, "learning_rate": 6.745910715184459e-05, "loss": 1.0218, "step": 130360 }, { "epoch": 0.33, "learning_rate": 6.745784891036288e-05, "loss": 1.0205, "step": 130365 }, { "epoch": 0.33, "learning_rate": 6.745659066888117e-05, "loss": 1.0191, "step": 130370 }, { "epoch": 0.33, "learning_rate": 6.745533242739946e-05, "loss": 1.0229, "step": 130375 }, { "epoch": 0.33, "learning_rate": 6.745407418591777e-05, "loss": 1.0201, "step": 130380 }, { "epoch": 0.33, "learning_rate": 6.745281594443606e-05, "loss": 1.0239, "step": 130385 }, { "epoch": 0.33, "learning_rate": 6.745155770295435e-05, "loss": 1.0207, "step": 130390 }, { "epoch": 0.33, "learning_rate": 6.745029946147264e-05, "loss": 1.0195, "step": 130395 }, { "epoch": 0.33, "learning_rate": 6.744904121999095e-05, "loss": 1.0217, "step": 130400 }, { "epoch": 0.33, "learning_rate": 6.744778297850924e-05, "loss": 1.0415, "step": 130405 }, { "epoch": 0.33, "learning_rate": 6.744652473702754e-05, "loss": 1.0206, "step": 130410 }, { "epoch": 0.33, "learning_rate": 6.744526649554583e-05, "loss": 1.0225, "step": 130415 }, { "epoch": 0.33, "learning_rate": 6.744400825406413e-05, "loss": 1.0201, "step": 130420 }, { "epoch": 0.33, "learning_rate": 6.744275001258242e-05, "loss": 1.0245, "step": 130425 }, { "epoch": 0.33, "learning_rate": 6.744149177110072e-05, "loss": 1.0204, "step": 130430 }, { "epoch": 0.33, "learning_rate": 6.744023352961901e-05, "loss": 1.0198, "step": 130435 }, { "epoch": 0.33, "learning_rate": 6.74389752881373e-05, "loss": 1.0247, "step": 130440 }, { "epoch": 0.33, "learning_rate": 6.74377170466556e-05, "loss": 1.021, "step": 130445 }, { "epoch": 0.33, "learning_rate": 6.74364588051739e-05, "loss": 1.0211, "step": 130450 }, { "epoch": 0.33, "learning_rate": 6.74352005636922e-05, "loss": 1.0225, "step": 130455 }, { "epoch": 0.33, "learning_rate": 6.743394232221049e-05, "loss": 1.0215, "step": 130460 }, { "epoch": 0.33, "learning_rate": 6.743268408072878e-05, "loss": 1.021, "step": 130465 }, { "epoch": 0.33, "learning_rate": 6.743142583924707e-05, "loss": 1.022, "step": 130470 }, { "epoch": 0.33, "learning_rate": 6.743016759776537e-05, "loss": 1.0214, "step": 130475 }, { "epoch": 0.33, "learning_rate": 6.742890935628367e-05, "loss": 1.0215, "step": 130480 }, { "epoch": 0.33, "learning_rate": 6.742765111480196e-05, "loss": 1.0191, "step": 130485 }, { "epoch": 0.33, "learning_rate": 6.742639287332025e-05, "loss": 1.0227, "step": 130490 }, { "epoch": 0.33, "learning_rate": 6.742513463183855e-05, "loss": 1.0211, "step": 130495 }, { "epoch": 0.33, "learning_rate": 6.742387639035685e-05, "loss": 1.0211, "step": 130500 }, { "epoch": 0.33, "learning_rate": 6.742261814887514e-05, "loss": 1.0229, "step": 130505 }, { "epoch": 0.33, "learning_rate": 6.742135990739343e-05, "loss": 1.0198, "step": 130510 }, { "epoch": 0.33, "learning_rate": 6.742010166591173e-05, "loss": 1.02, "step": 130515 }, { "epoch": 0.33, "learning_rate": 6.741884342443003e-05, "loss": 1.0231, "step": 130520 }, { "epoch": 0.33, "learning_rate": 6.741758518294832e-05, "loss": 1.0201, "step": 130525 }, { "epoch": 0.33, "learning_rate": 6.741632694146661e-05, "loss": 1.0205, "step": 130530 }, { "epoch": 0.33, "learning_rate": 6.74150686999849e-05, "loss": 1.0218, "step": 130535 }, { "epoch": 0.33, "learning_rate": 6.74138104585032e-05, "loss": 1.0237, "step": 130540 }, { "epoch": 0.33, "learning_rate": 6.74125522170215e-05, "loss": 1.02, "step": 130545 }, { "epoch": 0.33, "learning_rate": 6.741129397553979e-05, "loss": 1.0199, "step": 130550 }, { "epoch": 0.33, "learning_rate": 6.741003573405808e-05, "loss": 1.0228, "step": 130555 }, { "epoch": 0.33, "learning_rate": 6.740877749257639e-05, "loss": 1.0211, "step": 130560 }, { "epoch": 0.33, "learning_rate": 6.740751925109468e-05, "loss": 1.0228, "step": 130565 }, { "epoch": 0.33, "learning_rate": 6.740626100961297e-05, "loss": 1.0208, "step": 130570 }, { "epoch": 0.33, "learning_rate": 6.74052544164276e-05, "loss": 1.0235, "step": 130575 }, { "epoch": 0.33, "learning_rate": 6.74039961749459e-05, "loss": 1.0224, "step": 130580 }, { "epoch": 0.33, "learning_rate": 6.740273793346419e-05, "loss": 1.0227, "step": 130585 }, { "epoch": 0.33, "learning_rate": 6.740147969198248e-05, "loss": 1.0237, "step": 130590 }, { "epoch": 0.33, "learning_rate": 6.740022145050077e-05, "loss": 1.0219, "step": 130595 }, { "epoch": 0.33, "learning_rate": 6.739896320901909e-05, "loss": 1.0207, "step": 130600 }, { "epoch": 0.33, "learning_rate": 6.739770496753738e-05, "loss": 1.0205, "step": 130605 }, { "epoch": 0.33, "learning_rate": 6.739644672605567e-05, "loss": 1.0227, "step": 130610 }, { "epoch": 0.33, "learning_rate": 6.739518848457396e-05, "loss": 1.0227, "step": 130615 }, { "epoch": 0.33, "learning_rate": 6.739393024309226e-05, "loss": 1.0197, "step": 130620 }, { "epoch": 0.33, "learning_rate": 6.739267200161056e-05, "loss": 1.0232, "step": 130625 }, { "epoch": 0.33, "learning_rate": 6.739141376012885e-05, "loss": 1.021, "step": 130630 }, { "epoch": 0.33, "learning_rate": 6.739015551864714e-05, "loss": 1.0182, "step": 130635 }, { "epoch": 0.33, "learning_rate": 6.738889727716544e-05, "loss": 1.0182, "step": 130640 }, { "epoch": 0.33, "learning_rate": 6.738763903568374e-05, "loss": 1.0223, "step": 130645 }, { "epoch": 0.33, "learning_rate": 6.738638079420203e-05, "loss": 1.0213, "step": 130650 }, { "epoch": 0.33, "learning_rate": 6.738512255272032e-05, "loss": 1.0221, "step": 130655 }, { "epoch": 0.33, "learning_rate": 6.738386431123862e-05, "loss": 1.0202, "step": 130660 }, { "epoch": 0.33, "learning_rate": 6.738260606975692e-05, "loss": 1.0213, "step": 130665 }, { "epoch": 0.33, "learning_rate": 6.738134782827521e-05, "loss": 1.0214, "step": 130670 }, { "epoch": 0.33, "learning_rate": 6.73800895867935e-05, "loss": 1.022, "step": 130675 }, { "epoch": 0.33, "learning_rate": 6.73788313453118e-05, "loss": 1.022, "step": 130680 }, { "epoch": 0.33, "learning_rate": 6.737757310383009e-05, "loss": 1.022, "step": 130685 }, { "epoch": 0.33, "learning_rate": 6.737631486234839e-05, "loss": 1.0244, "step": 130690 }, { "epoch": 0.33, "learning_rate": 6.737505662086668e-05, "loss": 1.0231, "step": 130695 }, { "epoch": 0.33, "learning_rate": 6.737379837938498e-05, "loss": 1.0212, "step": 130700 }, { "epoch": 0.33, "learning_rate": 6.737254013790327e-05, "loss": 1.0216, "step": 130705 }, { "epoch": 0.33, "learning_rate": 6.737128189642157e-05, "loss": 1.0201, "step": 130710 }, { "epoch": 0.33, "learning_rate": 6.737002365493986e-05, "loss": 1.0215, "step": 130715 }, { "epoch": 0.33, "learning_rate": 6.736876541345816e-05, "loss": 1.0429, "step": 130720 }, { "epoch": 0.33, "learning_rate": 6.736750717197645e-05, "loss": 1.0226, "step": 130725 }, { "epoch": 0.33, "learning_rate": 6.736624893049475e-05, "loss": 1.0195, "step": 130730 }, { "epoch": 0.33, "learning_rate": 6.736499068901304e-05, "loss": 1.0199, "step": 130735 }, { "epoch": 0.33, "learning_rate": 6.736373244753134e-05, "loss": 1.0226, "step": 130740 }, { "epoch": 0.33, "learning_rate": 6.736247420604963e-05, "loss": 1.0228, "step": 130745 }, { "epoch": 0.33, "learning_rate": 6.736121596456792e-05, "loss": 1.02, "step": 130750 }, { "epoch": 0.33, "learning_rate": 6.735995772308622e-05, "loss": 1.0219, "step": 130755 }, { "epoch": 0.33, "learning_rate": 6.735869948160452e-05, "loss": 1.0216, "step": 130760 }, { "epoch": 0.33, "learning_rate": 6.735744124012281e-05, "loss": 1.0242, "step": 130765 }, { "epoch": 0.33, "learning_rate": 6.73561829986411e-05, "loss": 1.0231, "step": 130770 }, { "epoch": 0.33, "learning_rate": 6.73549247571594e-05, "loss": 1.0396, "step": 130775 }, { "epoch": 0.33, "learning_rate": 6.73536665156777e-05, "loss": 1.0222, "step": 130780 }, { "epoch": 0.33, "learning_rate": 6.735240827419599e-05, "loss": 1.0211, "step": 130785 }, { "epoch": 0.33, "learning_rate": 6.735115003271428e-05, "loss": 1.0217, "step": 130790 }, { "epoch": 0.33, "learning_rate": 6.734989179123258e-05, "loss": 1.0226, "step": 130795 }, { "epoch": 0.33, "learning_rate": 6.734863354975088e-05, "loss": 1.0196, "step": 130800 }, { "epoch": 0.33, "learning_rate": 6.734737530826917e-05, "loss": 1.0197, "step": 130805 }, { "epoch": 0.33, "learning_rate": 6.734611706678746e-05, "loss": 1.0208, "step": 130810 }, { "epoch": 0.33, "learning_rate": 6.734485882530575e-05, "loss": 1.0174, "step": 130815 }, { "epoch": 0.33, "learning_rate": 6.734360058382406e-05, "loss": 1.0212, "step": 130820 }, { "epoch": 0.33, "learning_rate": 6.734234234234235e-05, "loss": 1.0196, "step": 130825 }, { "epoch": 0.33, "learning_rate": 6.734108410086064e-05, "loss": 1.0218, "step": 130830 }, { "epoch": 0.33, "learning_rate": 6.733982585937893e-05, "loss": 1.0205, "step": 130835 }, { "epoch": 0.33, "learning_rate": 6.733856761789724e-05, "loss": 1.0216, "step": 130840 }, { "epoch": 0.33, "learning_rate": 6.733730937641553e-05, "loss": 1.0196, "step": 130845 }, { "epoch": 0.33, "learning_rate": 6.733605113493382e-05, "loss": 1.0216, "step": 130850 }, { "epoch": 0.33, "learning_rate": 6.733479289345211e-05, "loss": 1.02, "step": 130855 }, { "epoch": 0.33, "learning_rate": 6.733353465197042e-05, "loss": 1.0218, "step": 130860 }, { "epoch": 0.33, "learning_rate": 6.733227641048871e-05, "loss": 1.0204, "step": 130865 }, { "epoch": 0.33, "learning_rate": 6.7331018169007e-05, "loss": 1.02, "step": 130870 }, { "epoch": 0.33, "learning_rate": 6.732975992752529e-05, "loss": 1.0206, "step": 130875 }, { "epoch": 0.33, "learning_rate": 6.732850168604358e-05, "loss": 1.0214, "step": 130880 }, { "epoch": 0.33, "learning_rate": 6.732724344456189e-05, "loss": 1.0206, "step": 130885 }, { "epoch": 0.33, "learning_rate": 6.732598520308018e-05, "loss": 1.0216, "step": 130890 }, { "epoch": 0.33, "learning_rate": 6.732472696159847e-05, "loss": 1.0203, "step": 130895 }, { "epoch": 0.33, "learning_rate": 6.732346872011676e-05, "loss": 1.0207, "step": 130900 }, { "epoch": 0.33, "learning_rate": 6.732221047863507e-05, "loss": 1.019, "step": 130905 }, { "epoch": 0.33, "learning_rate": 6.732095223715336e-05, "loss": 1.0206, "step": 130910 }, { "epoch": 0.33, "learning_rate": 6.731969399567165e-05, "loss": 1.0242, "step": 130915 }, { "epoch": 0.33, "learning_rate": 6.731843575418994e-05, "loss": 1.023, "step": 130920 }, { "epoch": 0.33, "learning_rate": 6.731717751270825e-05, "loss": 1.0216, "step": 130925 }, { "epoch": 0.33, "learning_rate": 6.731591927122654e-05, "loss": 1.0204, "step": 130930 }, { "epoch": 0.33, "learning_rate": 6.731466102974483e-05, "loss": 1.0216, "step": 130935 }, { "epoch": 0.33, "learning_rate": 6.731340278826312e-05, "loss": 1.0225, "step": 130940 }, { "epoch": 0.33, "learning_rate": 6.731214454678141e-05, "loss": 1.0207, "step": 130945 }, { "epoch": 0.33, "learning_rate": 6.731088630529972e-05, "loss": 1.0223, "step": 130950 }, { "epoch": 0.33, "learning_rate": 6.730962806381801e-05, "loss": 1.0195, "step": 130955 }, { "epoch": 0.33, "learning_rate": 6.73083698223363e-05, "loss": 1.0186, "step": 130960 }, { "epoch": 0.33, "learning_rate": 6.73071115808546e-05, "loss": 1.0234, "step": 130965 }, { "epoch": 0.33, "learning_rate": 6.73058533393729e-05, "loss": 1.0208, "step": 130970 }, { "epoch": 0.33, "learning_rate": 6.730459509789119e-05, "loss": 1.0233, "step": 130975 }, { "epoch": 0.33, "learning_rate": 6.730333685640948e-05, "loss": 1.0203, "step": 130980 }, { "epoch": 0.33, "learning_rate": 6.730207861492777e-05, "loss": 1.0209, "step": 130985 }, { "epoch": 0.33, "learning_rate": 6.730082037344608e-05, "loss": 1.0214, "step": 130990 }, { "epoch": 0.33, "learning_rate": 6.729956213196437e-05, "loss": 1.0213, "step": 130995 }, { "epoch": 0.33, "learning_rate": 6.729830389048266e-05, "loss": 1.0233, "step": 131000 }, { "epoch": 0.33, "learning_rate": 6.729704564900095e-05, "loss": 1.0214, "step": 131005 }, { "epoch": 0.33, "learning_rate": 6.729578740751924e-05, "loss": 1.0186, "step": 131010 }, { "epoch": 0.33, "learning_rate": 6.729452916603755e-05, "loss": 1.0244, "step": 131015 }, { "epoch": 0.33, "learning_rate": 6.729327092455584e-05, "loss": 1.0202, "step": 131020 }, { "epoch": 0.33, "learning_rate": 6.729201268307413e-05, "loss": 1.0208, "step": 131025 }, { "epoch": 0.33, "learning_rate": 6.729075444159242e-05, "loss": 1.0223, "step": 131030 }, { "epoch": 0.33, "learning_rate": 6.728949620011073e-05, "loss": 1.039, "step": 131035 }, { "epoch": 0.33, "learning_rate": 6.728823795862902e-05, "loss": 1.0227, "step": 131040 }, { "epoch": 0.33, "learning_rate": 6.728697971714731e-05, "loss": 1.0212, "step": 131045 }, { "epoch": 0.33, "learning_rate": 6.72857214756656e-05, "loss": 1.0211, "step": 131050 }, { "epoch": 0.33, "learning_rate": 6.728446323418391e-05, "loss": 1.0199, "step": 131055 }, { "epoch": 0.33, "learning_rate": 6.72832049927022e-05, "loss": 1.0202, "step": 131060 }, { "epoch": 0.33, "learning_rate": 6.728194675122049e-05, "loss": 1.0207, "step": 131065 }, { "epoch": 0.33, "learning_rate": 6.728068850973878e-05, "loss": 1.0221, "step": 131070 }, { "epoch": 0.33, "learning_rate": 6.727943026825708e-05, "loss": 1.0209, "step": 131075 }, { "epoch": 0.33, "learning_rate": 6.727817202677538e-05, "loss": 1.0214, "step": 131080 }, { "epoch": 0.33, "learning_rate": 6.727691378529367e-05, "loss": 1.0187, "step": 131085 }, { "epoch": 0.33, "learning_rate": 6.727565554381196e-05, "loss": 1.0201, "step": 131090 }, { "epoch": 0.33, "learning_rate": 6.727439730233026e-05, "loss": 1.0231, "step": 131095 }, { "epoch": 0.33, "learning_rate": 6.727313906084856e-05, "loss": 1.0206, "step": 131100 }, { "epoch": 0.33, "learning_rate": 6.727188081936687e-05, "loss": 1.0208, "step": 131105 }, { "epoch": 0.33, "learning_rate": 6.727062257788516e-05, "loss": 1.0201, "step": 131110 }, { "epoch": 0.33, "learning_rate": 6.726936433640345e-05, "loss": 1.0204, "step": 131115 }, { "epoch": 0.33, "learning_rate": 6.726810609492174e-05, "loss": 1.0184, "step": 131120 }, { "epoch": 0.33, "learning_rate": 6.726684785344005e-05, "loss": 1.0213, "step": 131125 }, { "epoch": 0.33, "learning_rate": 6.726558961195834e-05, "loss": 1.0235, "step": 131130 }, { "epoch": 0.33, "learning_rate": 6.726433137047663e-05, "loss": 1.0223, "step": 131135 }, { "epoch": 0.33, "learning_rate": 6.726307312899492e-05, "loss": 1.0203, "step": 131140 }, { "epoch": 0.33, "learning_rate": 6.726181488751321e-05, "loss": 1.0231, "step": 131145 }, { "epoch": 0.33, "learning_rate": 6.726055664603152e-05, "loss": 1.022, "step": 131150 }, { "epoch": 0.33, "learning_rate": 6.725929840454981e-05, "loss": 1.0217, "step": 131155 }, { "epoch": 0.33, "learning_rate": 6.72580401630681e-05, "loss": 1.022, "step": 131160 }, { "epoch": 0.33, "learning_rate": 6.725678192158639e-05, "loss": 1.0223, "step": 131165 }, { "epoch": 0.33, "learning_rate": 6.72555236801047e-05, "loss": 1.0236, "step": 131170 }, { "epoch": 0.33, "learning_rate": 6.725426543862299e-05, "loss": 1.0236, "step": 131175 }, { "epoch": 0.33, "learning_rate": 6.725300719714128e-05, "loss": 1.0198, "step": 131180 }, { "epoch": 0.33, "learning_rate": 6.725174895565957e-05, "loss": 1.0205, "step": 131185 }, { "epoch": 0.33, "learning_rate": 6.725049071417788e-05, "loss": 1.0185, "step": 131190 }, { "epoch": 0.33, "learning_rate": 6.724923247269617e-05, "loss": 1.021, "step": 131195 }, { "epoch": 0.33, "learning_rate": 6.724797423121446e-05, "loss": 1.0216, "step": 131200 }, { "epoch": 0.33, "learning_rate": 6.724671598973275e-05, "loss": 1.0444, "step": 131205 }, { "epoch": 0.33, "learning_rate": 6.724545774825104e-05, "loss": 1.023, "step": 131210 }, { "epoch": 0.33, "learning_rate": 6.724419950676935e-05, "loss": 1.0215, "step": 131215 }, { "epoch": 0.33, "learning_rate": 6.724294126528764e-05, "loss": 1.0196, "step": 131220 }, { "epoch": 0.33, "learning_rate": 6.724168302380593e-05, "loss": 1.0205, "step": 131225 }, { "epoch": 0.33, "learning_rate": 6.724042478232422e-05, "loss": 1.0228, "step": 131230 }, { "epoch": 0.33, "learning_rate": 6.723916654084253e-05, "loss": 1.0225, "step": 131235 }, { "epoch": 0.33, "learning_rate": 6.723790829936082e-05, "loss": 1.0217, "step": 131240 }, { "epoch": 0.33, "learning_rate": 6.723665005787911e-05, "loss": 1.0205, "step": 131245 }, { "epoch": 0.33, "learning_rate": 6.72353918163974e-05, "loss": 1.0215, "step": 131250 }, { "epoch": 0.33, "learning_rate": 6.723413357491571e-05, "loss": 1.0225, "step": 131255 }, { "epoch": 0.33, "learning_rate": 6.7232875333434e-05, "loss": 1.0204, "step": 131260 }, { "epoch": 0.33, "learning_rate": 6.723161709195229e-05, "loss": 1.0215, "step": 131265 }, { "epoch": 0.33, "learning_rate": 6.723035885047058e-05, "loss": 1.0202, "step": 131270 }, { "epoch": 0.33, "learning_rate": 6.722910060898888e-05, "loss": 1.0221, "step": 131275 }, { "epoch": 0.33, "learning_rate": 6.722784236750718e-05, "loss": 1.0195, "step": 131280 }, { "epoch": 0.33, "learning_rate": 6.722658412602547e-05, "loss": 1.0209, "step": 131285 }, { "epoch": 0.33, "learning_rate": 6.722532588454376e-05, "loss": 1.0227, "step": 131290 }, { "epoch": 0.33, "learning_rate": 6.722406764306206e-05, "loss": 1.022, "step": 131295 }, { "epoch": 0.33, "learning_rate": 6.722280940158036e-05, "loss": 1.0227, "step": 131300 }, { "epoch": 0.33, "learning_rate": 6.722155116009865e-05, "loss": 1.0228, "step": 131305 }, { "epoch": 0.33, "learning_rate": 6.722029291861694e-05, "loss": 1.0188, "step": 131310 }, { "epoch": 0.33, "learning_rate": 6.721903467713524e-05, "loss": 1.023, "step": 131315 }, { "epoch": 0.33, "learning_rate": 6.721777643565354e-05, "loss": 1.0196, "step": 131320 }, { "epoch": 0.33, "learning_rate": 6.721651819417183e-05, "loss": 1.0208, "step": 131325 }, { "epoch": 0.33, "learning_rate": 6.721525995269012e-05, "loss": 1.0376, "step": 131330 }, { "epoch": 0.33, "learning_rate": 6.721400171120842e-05, "loss": 1.0208, "step": 131335 }, { "epoch": 0.33, "learning_rate": 6.721274346972671e-05, "loss": 1.02, "step": 131340 }, { "epoch": 0.33, "learning_rate": 6.721148522824501e-05, "loss": 1.0232, "step": 131345 }, { "epoch": 0.33, "learning_rate": 6.72102269867633e-05, "loss": 1.0208, "step": 131350 }, { "epoch": 0.33, "learning_rate": 6.72089687452816e-05, "loss": 1.0212, "step": 131355 }, { "epoch": 0.33, "learning_rate": 6.720771050379989e-05, "loss": 1.0232, "step": 131360 }, { "epoch": 0.33, "learning_rate": 6.720645226231819e-05, "loss": 1.0222, "step": 131365 }, { "epoch": 0.33, "learning_rate": 6.720519402083648e-05, "loss": 1.0206, "step": 131370 }, { "epoch": 0.33, "learning_rate": 6.720393577935478e-05, "loss": 1.0228, "step": 131375 }, { "epoch": 0.33, "learning_rate": 6.720267753787307e-05, "loss": 1.0219, "step": 131380 }, { "epoch": 0.33, "learning_rate": 6.720141929639137e-05, "loss": 1.0212, "step": 131385 }, { "epoch": 0.33, "learning_rate": 6.720016105490966e-05, "loss": 1.0241, "step": 131390 }, { "epoch": 0.33, "learning_rate": 6.719890281342796e-05, "loss": 1.0218, "step": 131395 }, { "epoch": 0.33, "learning_rate": 6.719764457194625e-05, "loss": 1.0222, "step": 131400 }, { "epoch": 0.33, "learning_rate": 6.719638633046454e-05, "loss": 1.0227, "step": 131405 }, { "epoch": 0.33, "learning_rate": 6.719512808898284e-05, "loss": 1.0203, "step": 131410 }, { "epoch": 0.33, "learning_rate": 6.719386984750114e-05, "loss": 1.0192, "step": 131415 }, { "epoch": 0.33, "learning_rate": 6.719261160601943e-05, "loss": 1.021, "step": 131420 }, { "epoch": 0.33, "learning_rate": 6.719135336453772e-05, "loss": 1.021, "step": 131425 }, { "epoch": 0.33, "learning_rate": 6.719009512305602e-05, "loss": 1.0216, "step": 131430 }, { "epoch": 0.33, "learning_rate": 6.718883688157432e-05, "loss": 1.0206, "step": 131435 }, { "epoch": 0.33, "learning_rate": 6.718757864009261e-05, "loss": 1.0207, "step": 131440 }, { "epoch": 0.33, "learning_rate": 6.71863203986109e-05, "loss": 1.0223, "step": 131445 }, { "epoch": 0.33, "learning_rate": 6.71850621571292e-05, "loss": 1.0226, "step": 131450 }, { "epoch": 0.33, "learning_rate": 6.71838039156475e-05, "loss": 1.0229, "step": 131455 }, { "epoch": 0.33, "learning_rate": 6.718254567416579e-05, "loss": 1.02, "step": 131460 }, { "epoch": 0.33, "learning_rate": 6.718128743268408e-05, "loss": 1.0436, "step": 131465 }, { "epoch": 0.33, "learning_rate": 6.718002919120237e-05, "loss": 1.022, "step": 131470 }, { "epoch": 0.33, "learning_rate": 6.717877094972068e-05, "loss": 1.0195, "step": 131475 }, { "epoch": 0.33, "learning_rate": 6.717751270823897e-05, "loss": 1.0233, "step": 131480 }, { "epoch": 0.33, "learning_rate": 6.717625446675726e-05, "loss": 1.0209, "step": 131485 }, { "epoch": 0.33, "learning_rate": 6.717499622527555e-05, "loss": 1.021, "step": 131490 }, { "epoch": 0.33, "learning_rate": 6.717373798379386e-05, "loss": 1.0206, "step": 131495 }, { "epoch": 0.33, "learning_rate": 6.717247974231215e-05, "loss": 1.0218, "step": 131500 }, { "epoch": 0.33, "learning_rate": 6.717122150083044e-05, "loss": 1.0211, "step": 131505 }, { "epoch": 0.33, "learning_rate": 6.716996325934873e-05, "loss": 1.0222, "step": 131510 }, { "epoch": 0.33, "learning_rate": 6.716870501786703e-05, "loss": 1.02, "step": 131515 }, { "epoch": 0.33, "learning_rate": 6.716744677638533e-05, "loss": 1.0174, "step": 131520 }, { "epoch": 0.33, "learning_rate": 6.716618853490362e-05, "loss": 1.0208, "step": 131525 }, { "epoch": 0.33, "learning_rate": 6.716493029342191e-05, "loss": 1.0202, "step": 131530 }, { "epoch": 0.33, "learning_rate": 6.71636720519402e-05, "loss": 1.0203, "step": 131535 }, { "epoch": 0.33, "learning_rate": 6.71624138104585e-05, "loss": 1.0215, "step": 131540 }, { "epoch": 0.33, "learning_rate": 6.71611555689768e-05, "loss": 1.0419, "step": 131545 }, { "epoch": 0.33, "learning_rate": 6.715989732749509e-05, "loss": 1.0218, "step": 131550 }, { "epoch": 0.33, "learning_rate": 6.715863908601338e-05, "loss": 1.0224, "step": 131555 }, { "epoch": 0.33, "learning_rate": 6.715738084453169e-05, "loss": 1.021, "step": 131560 }, { "epoch": 0.33, "learning_rate": 6.715612260304998e-05, "loss": 1.0195, "step": 131565 }, { "epoch": 0.33, "learning_rate": 6.715486436156827e-05, "loss": 1.0202, "step": 131570 }, { "epoch": 0.33, "learning_rate": 6.715360612008656e-05, "loss": 1.0216, "step": 131575 }, { "epoch": 0.33, "learning_rate": 6.715234787860487e-05, "loss": 1.0176, "step": 131580 }, { "epoch": 0.33, "learning_rate": 6.715108963712316e-05, "loss": 1.0216, "step": 131585 }, { "epoch": 0.33, "learning_rate": 6.714983139564145e-05, "loss": 1.024, "step": 131590 }, { "epoch": 0.33, "learning_rate": 6.714857315415974e-05, "loss": 1.0178, "step": 131595 }, { "epoch": 0.33, "learning_rate": 6.714731491267805e-05, "loss": 1.0177, "step": 131600 }, { "epoch": 0.33, "learning_rate": 6.714605667119634e-05, "loss": 1.0231, "step": 131605 }, { "epoch": 0.33, "learning_rate": 6.714479842971464e-05, "loss": 1.023, "step": 131610 }, { "epoch": 0.33, "learning_rate": 6.714354018823293e-05, "loss": 1.02, "step": 131615 }, { "epoch": 0.33, "learning_rate": 6.714228194675123e-05, "loss": 1.0213, "step": 131620 }, { "epoch": 0.33, "learning_rate": 6.714102370526952e-05, "loss": 1.0221, "step": 131625 }, { "epoch": 0.33, "learning_rate": 6.713976546378782e-05, "loss": 1.02, "step": 131630 }, { "epoch": 0.33, "learning_rate": 6.713850722230611e-05, "loss": 1.0205, "step": 131635 }, { "epoch": 0.33, "learning_rate": 6.71372489808244e-05, "loss": 1.0209, "step": 131640 }, { "epoch": 0.33, "learning_rate": 6.71359907393427e-05, "loss": 1.022, "step": 131645 }, { "epoch": 0.33, "learning_rate": 6.7134732497861e-05, "loss": 1.0211, "step": 131650 }, { "epoch": 0.33, "learning_rate": 6.71334742563793e-05, "loss": 1.0219, "step": 131655 }, { "epoch": 0.33, "learning_rate": 6.713221601489759e-05, "loss": 1.021, "step": 131660 }, { "epoch": 0.33, "learning_rate": 6.713095777341588e-05, "loss": 1.0202, "step": 131665 }, { "epoch": 0.33, "learning_rate": 6.712969953193417e-05, "loss": 1.0232, "step": 131670 }, { "epoch": 0.33, "learning_rate": 6.712844129045247e-05, "loss": 1.0207, "step": 131675 }, { "epoch": 0.33, "learning_rate": 6.712718304897077e-05, "loss": 1.0199, "step": 131680 }, { "epoch": 0.33, "learning_rate": 6.712592480748906e-05, "loss": 1.0205, "step": 131685 }, { "epoch": 0.33, "learning_rate": 6.712466656600735e-05, "loss": 1.0213, "step": 131690 }, { "epoch": 0.33, "learning_rate": 6.712340832452565e-05, "loss": 1.0207, "step": 131695 }, { "epoch": 0.33, "learning_rate": 6.712215008304395e-05, "loss": 1.0193, "step": 131700 }, { "epoch": 0.33, "learning_rate": 6.712089184156224e-05, "loss": 1.0214, "step": 131705 }, { "epoch": 0.33, "learning_rate": 6.711963360008053e-05, "loss": 1.0213, "step": 131710 }, { "epoch": 0.33, "learning_rate": 6.711837535859883e-05, "loss": 1.0203, "step": 131715 }, { "epoch": 0.33, "learning_rate": 6.711711711711713e-05, "loss": 1.0245, "step": 131720 }, { "epoch": 0.33, "learning_rate": 6.711585887563542e-05, "loss": 1.023, "step": 131725 }, { "epoch": 0.33, "learning_rate": 6.711460063415371e-05, "loss": 1.0208, "step": 131730 }, { "epoch": 0.33, "learning_rate": 6.7113342392672e-05, "loss": 1.0183, "step": 131735 }, { "epoch": 0.33, "learning_rate": 6.71120841511903e-05, "loss": 1.0209, "step": 131740 }, { "epoch": 0.33, "learning_rate": 6.71108259097086e-05, "loss": 1.0199, "step": 131745 }, { "epoch": 0.33, "learning_rate": 6.710956766822689e-05, "loss": 1.0226, "step": 131750 }, { "epoch": 0.33, "learning_rate": 6.710830942674518e-05, "loss": 1.0201, "step": 131755 }, { "epoch": 0.33, "learning_rate": 6.710705118526349e-05, "loss": 1.0182, "step": 131760 }, { "epoch": 0.33, "learning_rate": 6.710579294378178e-05, "loss": 1.0174, "step": 131765 }, { "epoch": 0.33, "learning_rate": 6.710453470230007e-05, "loss": 1.0205, "step": 131770 }, { "epoch": 0.33, "learning_rate": 6.710327646081836e-05, "loss": 1.0232, "step": 131775 }, { "epoch": 0.33, "learning_rate": 6.710201821933667e-05, "loss": 1.0188, "step": 131780 }, { "epoch": 0.33, "learning_rate": 6.710075997785496e-05, "loss": 1.0227, "step": 131785 }, { "epoch": 0.33, "learning_rate": 6.709950173637325e-05, "loss": 1.0222, "step": 131790 }, { "epoch": 0.33, "learning_rate": 6.709824349489154e-05, "loss": 1.0223, "step": 131795 }, { "epoch": 0.33, "learning_rate": 6.709698525340983e-05, "loss": 1.019, "step": 131800 }, { "epoch": 0.33, "learning_rate": 6.709572701192814e-05, "loss": 1.0212, "step": 131805 }, { "epoch": 0.33, "learning_rate": 6.709446877044643e-05, "loss": 1.0229, "step": 131810 }, { "epoch": 0.33, "learning_rate": 6.709321052896472e-05, "loss": 1.0224, "step": 131815 }, { "epoch": 0.33, "learning_rate": 6.709195228748301e-05, "loss": 1.0204, "step": 131820 }, { "epoch": 0.33, "learning_rate": 6.709069404600132e-05, "loss": 1.0248, "step": 131825 }, { "epoch": 0.33, "learning_rate": 6.708943580451961e-05, "loss": 1.0192, "step": 131830 }, { "epoch": 0.33, "learning_rate": 6.70881775630379e-05, "loss": 1.022, "step": 131835 }, { "epoch": 0.33, "learning_rate": 6.708691932155619e-05, "loss": 1.0206, "step": 131840 }, { "epoch": 0.33, "learning_rate": 6.70856610800745e-05, "loss": 1.0448, "step": 131845 }, { "epoch": 0.33, "learning_rate": 6.708440283859279e-05, "loss": 1.0377, "step": 131850 }, { "epoch": 0.33, "learning_rate": 6.708314459711108e-05, "loss": 1.0222, "step": 131855 }, { "epoch": 0.33, "learning_rate": 6.708188635562937e-05, "loss": 1.0201, "step": 131860 }, { "epoch": 0.33, "learning_rate": 6.708062811414766e-05, "loss": 1.0216, "step": 131865 }, { "epoch": 0.33, "learning_rate": 6.707936987266597e-05, "loss": 1.0222, "step": 131870 }, { "epoch": 0.33, "learning_rate": 6.707811163118426e-05, "loss": 1.0199, "step": 131875 }, { "epoch": 0.33, "learning_rate": 6.707685338970255e-05, "loss": 1.0225, "step": 131880 }, { "epoch": 0.33, "learning_rate": 6.707559514822084e-05, "loss": 1.0193, "step": 131885 }, { "epoch": 0.33, "learning_rate": 6.707433690673915e-05, "loss": 1.0208, "step": 131890 }, { "epoch": 0.33, "learning_rate": 6.707307866525744e-05, "loss": 1.0229, "step": 131895 }, { "epoch": 0.33, "learning_rate": 6.707182042377573e-05, "loss": 1.0217, "step": 131900 }, { "epoch": 0.33, "learning_rate": 6.707056218229402e-05, "loss": 1.0215, "step": 131905 }, { "epoch": 0.33, "learning_rate": 6.706930394081233e-05, "loss": 1.0192, "step": 131910 }, { "epoch": 0.33, "learning_rate": 6.706804569933062e-05, "loss": 1.0233, "step": 131915 }, { "epoch": 0.33, "learning_rate": 6.706678745784891e-05, "loss": 1.0208, "step": 131920 }, { "epoch": 0.33, "learning_rate": 6.70655292163672e-05, "loss": 1.0218, "step": 131925 }, { "epoch": 0.33, "learning_rate": 6.70642709748855e-05, "loss": 1.0448, "step": 131930 }, { "epoch": 0.33, "learning_rate": 6.70630127334038e-05, "loss": 1.02, "step": 131935 }, { "epoch": 0.33, "learning_rate": 6.706175449192209e-05, "loss": 1.0195, "step": 131940 }, { "epoch": 0.33, "learning_rate": 6.706049625044038e-05, "loss": 1.0205, "step": 131945 }, { "epoch": 0.33, "learning_rate": 6.705923800895868e-05, "loss": 1.019, "step": 131950 }, { "epoch": 0.33, "learning_rate": 6.705797976747698e-05, "loss": 1.0192, "step": 131955 }, { "epoch": 0.33, "learning_rate": 6.705672152599527e-05, "loss": 1.0223, "step": 131960 }, { "epoch": 0.33, "learning_rate": 6.705546328451356e-05, "loss": 1.0214, "step": 131965 }, { "epoch": 0.33, "learning_rate": 6.705420504303185e-05, "loss": 1.0214, "step": 131970 }, { "epoch": 0.33, "learning_rate": 6.705294680155016e-05, "loss": 1.021, "step": 131975 }, { "epoch": 0.33, "learning_rate": 6.705168856006845e-05, "loss": 1.0204, "step": 131980 }, { "epoch": 0.33, "learning_rate": 6.705043031858674e-05, "loss": 1.0219, "step": 131985 }, { "epoch": 0.33, "learning_rate": 6.704917207710503e-05, "loss": 1.0235, "step": 131990 }, { "epoch": 0.33, "learning_rate": 6.704791383562333e-05, "loss": 1.0239, "step": 131995 }, { "epoch": 0.33, "learning_rate": 6.704665559414163e-05, "loss": 1.0208, "step": 132000 }, { "epoch": 0.33, "learning_rate": 6.704539735265992e-05, "loss": 1.0208, "step": 132005 }, { "epoch": 0.33, "learning_rate": 6.704413911117821e-05, "loss": 1.0205, "step": 132010 }, { "epoch": 0.33, "learning_rate": 6.70428808696965e-05, "loss": 1.0236, "step": 132015 }, { "epoch": 0.33, "learning_rate": 6.704162262821481e-05, "loss": 1.0223, "step": 132020 }, { "epoch": 0.33, "learning_rate": 6.70403643867331e-05, "loss": 1.0205, "step": 132025 }, { "epoch": 0.33, "learning_rate": 6.70391061452514e-05, "loss": 1.0211, "step": 132030 }, { "epoch": 0.33, "learning_rate": 6.703784790376969e-05, "loss": 1.0203, "step": 132035 }, { "epoch": 0.33, "learning_rate": 6.703658966228799e-05, "loss": 1.0392, "step": 132040 }, { "epoch": 0.33, "learning_rate": 6.703533142080628e-05, "loss": 1.0201, "step": 132045 }, { "epoch": 0.33, "learning_rate": 6.703407317932457e-05, "loss": 1.0219, "step": 132050 }, { "epoch": 0.33, "learning_rate": 6.703281493784287e-05, "loss": 1.0205, "step": 132055 }, { "epoch": 0.33, "learning_rate": 6.703155669636116e-05, "loss": 1.0227, "step": 132060 }, { "epoch": 0.33, "learning_rate": 6.703029845487946e-05, "loss": 1.0219, "step": 132065 }, { "epoch": 0.33, "learning_rate": 6.702904021339775e-05, "loss": 1.0205, "step": 132070 }, { "epoch": 0.33, "learning_rate": 6.702778197191605e-05, "loss": 1.0204, "step": 132075 }, { "epoch": 0.33, "learning_rate": 6.702652373043434e-05, "loss": 1.023, "step": 132080 }, { "epoch": 0.33, "learning_rate": 6.702526548895264e-05, "loss": 1.021, "step": 132085 }, { "epoch": 0.33, "learning_rate": 6.702400724747093e-05, "loss": 1.0211, "step": 132090 }, { "epoch": 0.33, "learning_rate": 6.702274900598923e-05, "loss": 1.0212, "step": 132095 }, { "epoch": 0.33, "learning_rate": 6.702149076450753e-05, "loss": 1.023, "step": 132100 }, { "epoch": 0.33, "learning_rate": 6.702023252302582e-05, "loss": 1.0225, "step": 132105 }, { "epoch": 0.33, "learning_rate": 6.701897428154413e-05, "loss": 1.0224, "step": 132110 }, { "epoch": 0.33, "learning_rate": 6.701771604006242e-05, "loss": 1.0206, "step": 132115 }, { "epoch": 0.33, "learning_rate": 6.701645779858071e-05, "loss": 1.022, "step": 132120 }, { "epoch": 0.33, "learning_rate": 6.7015199557099e-05, "loss": 1.0211, "step": 132125 }, { "epoch": 0.33, "learning_rate": 6.70139413156173e-05, "loss": 1.0233, "step": 132130 }, { "epoch": 0.33, "learning_rate": 6.70126830741356e-05, "loss": 1.0219, "step": 132135 }, { "epoch": 0.33, "learning_rate": 6.701142483265389e-05, "loss": 1.0209, "step": 132140 }, { "epoch": 0.33, "learning_rate": 6.701016659117218e-05, "loss": 1.0212, "step": 132145 }, { "epoch": 0.33, "learning_rate": 6.700890834969047e-05, "loss": 1.0197, "step": 132150 }, { "epoch": 0.33, "learning_rate": 6.700765010820878e-05, "loss": 1.0186, "step": 132155 }, { "epoch": 0.33, "learning_rate": 6.700639186672707e-05, "loss": 1.0194, "step": 132160 }, { "epoch": 0.33, "learning_rate": 6.700513362524536e-05, "loss": 1.0198, "step": 132165 }, { "epoch": 0.33, "learning_rate": 6.700387538376365e-05, "loss": 1.0229, "step": 132170 }, { "epoch": 0.33, "learning_rate": 6.700261714228196e-05, "loss": 1.0232, "step": 132175 }, { "epoch": 0.33, "learning_rate": 6.700135890080025e-05, "loss": 1.0214, "step": 132180 }, { "epoch": 0.33, "learning_rate": 6.700010065931854e-05, "loss": 1.0206, "step": 132185 }, { "epoch": 0.33, "learning_rate": 6.699884241783683e-05, "loss": 1.023, "step": 132190 }, { "epoch": 0.33, "learning_rate": 6.699758417635513e-05, "loss": 1.0198, "step": 132195 }, { "epoch": 0.33, "learning_rate": 6.699632593487343e-05, "loss": 1.0243, "step": 132200 }, { "epoch": 0.33, "learning_rate": 6.699506769339172e-05, "loss": 1.0245, "step": 132205 }, { "epoch": 0.33, "learning_rate": 6.699380945191001e-05, "loss": 1.0202, "step": 132210 }, { "epoch": 0.33, "learning_rate": 6.69925512104283e-05, "loss": 1.0219, "step": 132215 }, { "epoch": 0.33, "learning_rate": 6.699129296894661e-05, "loss": 1.0406, "step": 132220 }, { "epoch": 0.33, "learning_rate": 6.69900347274649e-05, "loss": 1.0232, "step": 132225 }, { "epoch": 0.33, "learning_rate": 6.69887764859832e-05, "loss": 1.0234, "step": 132230 }, { "epoch": 0.33, "learning_rate": 6.698751824450149e-05, "loss": 1.0227, "step": 132235 }, { "epoch": 0.33, "learning_rate": 6.698626000301979e-05, "loss": 1.0192, "step": 132240 }, { "epoch": 0.33, "learning_rate": 6.698500176153808e-05, "loss": 1.0429, "step": 132245 }, { "epoch": 0.33, "learning_rate": 6.698374352005637e-05, "loss": 1.0222, "step": 132250 }, { "epoch": 0.33, "learning_rate": 6.698248527857467e-05, "loss": 1.023, "step": 132255 }, { "epoch": 0.33, "learning_rate": 6.698122703709296e-05, "loss": 1.0472, "step": 132260 }, { "epoch": 0.33, "learning_rate": 6.697996879561126e-05, "loss": 1.0206, "step": 132265 }, { "epoch": 0.33, "learning_rate": 6.697871055412955e-05, "loss": 1.0234, "step": 132270 }, { "epoch": 0.33, "learning_rate": 6.697745231264785e-05, "loss": 1.0217, "step": 132275 }, { "epoch": 0.33, "learning_rate": 6.697619407116614e-05, "loss": 1.0233, "step": 132280 }, { "epoch": 0.33, "learning_rate": 6.697493582968444e-05, "loss": 1.0222, "step": 132285 }, { "epoch": 0.33, "learning_rate": 6.697367758820273e-05, "loss": 1.0195, "step": 132290 }, { "epoch": 0.33, "learning_rate": 6.697241934672103e-05, "loss": 1.0219, "step": 132295 }, { "epoch": 0.33, "learning_rate": 6.697116110523932e-05, "loss": 1.0429, "step": 132300 }, { "epoch": 0.33, "learning_rate": 6.696990286375762e-05, "loss": 1.0219, "step": 132305 }, { "epoch": 0.33, "learning_rate": 6.696864462227591e-05, "loss": 1.0221, "step": 132310 }, { "epoch": 0.33, "learning_rate": 6.69673863807942e-05, "loss": 1.0209, "step": 132315 }, { "epoch": 0.33, "learning_rate": 6.69661281393125e-05, "loss": 1.02, "step": 132320 }, { "epoch": 0.33, "learning_rate": 6.696486989783079e-05, "loss": 1.0211, "step": 132325 }, { "epoch": 0.33, "learning_rate": 6.69636116563491e-05, "loss": 1.0217, "step": 132330 }, { "epoch": 0.33, "learning_rate": 6.696235341486739e-05, "loss": 1.0226, "step": 132335 }, { "epoch": 0.33, "learning_rate": 6.696109517338568e-05, "loss": 1.0217, "step": 132340 }, { "epoch": 0.33, "learning_rate": 6.695983693190397e-05, "loss": 1.0211, "step": 132345 }, { "epoch": 0.33, "learning_rate": 6.695857869042227e-05, "loss": 1.0231, "step": 132350 }, { "epoch": 0.33, "learning_rate": 6.695732044894057e-05, "loss": 1.0199, "step": 132355 }, { "epoch": 0.33, "learning_rate": 6.695606220745886e-05, "loss": 1.0218, "step": 132360 }, { "epoch": 0.33, "learning_rate": 6.695480396597715e-05, "loss": 1.02, "step": 132365 }, { "epoch": 0.33, "learning_rate": 6.695354572449545e-05, "loss": 1.0241, "step": 132370 }, { "epoch": 0.33, "learning_rate": 6.695228748301375e-05, "loss": 1.022, "step": 132375 }, { "epoch": 0.33, "learning_rate": 6.695102924153204e-05, "loss": 1.0213, "step": 132380 }, { "epoch": 0.33, "learning_rate": 6.694977100005033e-05, "loss": 1.0217, "step": 132385 }, { "epoch": 0.33, "learning_rate": 6.694851275856862e-05, "loss": 1.0212, "step": 132390 }, { "epoch": 0.33, "learning_rate": 6.694725451708693e-05, "loss": 1.0211, "step": 132395 }, { "epoch": 0.33, "learning_rate": 6.694599627560522e-05, "loss": 1.0206, "step": 132400 }, { "epoch": 0.33, "learning_rate": 6.694473803412351e-05, "loss": 1.0215, "step": 132405 }, { "epoch": 0.33, "learning_rate": 6.69434797926418e-05, "loss": 1.02, "step": 132410 }, { "epoch": 0.33, "learning_rate": 6.69422215511601e-05, "loss": 1.0225, "step": 132415 }, { "epoch": 0.33, "learning_rate": 6.69409633096784e-05, "loss": 1.0198, "step": 132420 }, { "epoch": 0.33, "learning_rate": 6.693970506819669e-05, "loss": 1.0198, "step": 132425 }, { "epoch": 0.33, "learning_rate": 6.693844682671498e-05, "loss": 1.023, "step": 132430 }, { "epoch": 0.33, "learning_rate": 6.693718858523329e-05, "loss": 1.0222, "step": 132435 }, { "epoch": 0.33, "learning_rate": 6.693593034375158e-05, "loss": 1.0219, "step": 132440 }, { "epoch": 0.33, "learning_rate": 6.693467210226987e-05, "loss": 1.021, "step": 132445 }, { "epoch": 0.33, "learning_rate": 6.693341386078816e-05, "loss": 1.0196, "step": 132450 }, { "epoch": 0.33, "learning_rate": 6.693215561930645e-05, "loss": 1.021, "step": 132455 }, { "epoch": 0.33, "learning_rate": 6.693089737782476e-05, "loss": 1.0434, "step": 132460 }, { "epoch": 0.33, "learning_rate": 6.692963913634305e-05, "loss": 1.0239, "step": 132465 }, { "epoch": 0.33, "learning_rate": 6.692838089486134e-05, "loss": 1.0457, "step": 132470 }, { "epoch": 0.33, "learning_rate": 6.692712265337963e-05, "loss": 1.0211, "step": 132475 }, { "epoch": 0.33, "learning_rate": 6.692586441189794e-05, "loss": 1.0206, "step": 132480 }, { "epoch": 0.33, "learning_rate": 6.692460617041623e-05, "loss": 1.0224, "step": 132485 }, { "epoch": 0.33, "learning_rate": 6.692334792893452e-05, "loss": 1.022, "step": 132490 }, { "epoch": 0.33, "learning_rate": 6.692208968745281e-05, "loss": 1.0217, "step": 132495 }, { "epoch": 0.33, "learning_rate": 6.692083144597112e-05, "loss": 1.0205, "step": 132500 }, { "epoch": 0.33, "learning_rate": 6.691957320448941e-05, "loss": 1.0194, "step": 132505 }, { "epoch": 0.33, "learning_rate": 6.69183149630077e-05, "loss": 1.0212, "step": 132510 }, { "epoch": 0.33, "learning_rate": 6.691705672152599e-05, "loss": 1.0207, "step": 132515 }, { "epoch": 0.33, "learning_rate": 6.691579848004428e-05, "loss": 1.0217, "step": 132520 }, { "epoch": 0.33, "learning_rate": 6.691454023856259e-05, "loss": 1.0202, "step": 132525 }, { "epoch": 0.33, "learning_rate": 6.691328199708088e-05, "loss": 1.0218, "step": 132530 }, { "epoch": 0.33, "learning_rate": 6.691202375559917e-05, "loss": 1.0213, "step": 132535 }, { "epoch": 0.33, "learning_rate": 6.691076551411746e-05, "loss": 1.0228, "step": 132540 }, { "epoch": 0.33, "learning_rate": 6.690950727263577e-05, "loss": 1.0213, "step": 132545 }, { "epoch": 0.33, "learning_rate": 6.690824903115406e-05, "loss": 1.0227, "step": 132550 }, { "epoch": 0.33, "learning_rate": 6.690699078967235e-05, "loss": 1.0219, "step": 132555 }, { "epoch": 0.33, "learning_rate": 6.690573254819064e-05, "loss": 1.0192, "step": 132560 }, { "epoch": 0.33, "learning_rate": 6.690447430670895e-05, "loss": 1.0225, "step": 132565 }, { "epoch": 0.33, "learning_rate": 6.690321606522724e-05, "loss": 1.0204, "step": 132570 }, { "epoch": 0.33, "learning_rate": 6.690195782374553e-05, "loss": 1.023, "step": 132575 }, { "epoch": 0.33, "learning_rate": 6.690069958226382e-05, "loss": 1.0213, "step": 132580 }, { "epoch": 0.33, "learning_rate": 6.689944134078211e-05, "loss": 1.02, "step": 132585 }, { "epoch": 0.33, "learning_rate": 6.689818309930042e-05, "loss": 1.0212, "step": 132590 }, { "epoch": 0.33, "learning_rate": 6.689692485781872e-05, "loss": 1.0393, "step": 132595 }, { "epoch": 0.33, "learning_rate": 6.689566661633702e-05, "loss": 1.0222, "step": 132600 }, { "epoch": 0.33, "learning_rate": 6.689440837485531e-05, "loss": 1.0224, "step": 132605 }, { "epoch": 0.33, "learning_rate": 6.68931501333736e-05, "loss": 1.0209, "step": 132610 }, { "epoch": 0.33, "learning_rate": 6.68918918918919e-05, "loss": 1.0209, "step": 132615 }, { "epoch": 0.33, "learning_rate": 6.68906336504102e-05, "loss": 1.0203, "step": 132620 }, { "epoch": 0.33, "learning_rate": 6.688937540892849e-05, "loss": 1.021, "step": 132625 }, { "epoch": 0.33, "learning_rate": 6.688811716744678e-05, "loss": 1.0191, "step": 132630 }, { "epoch": 0.33, "learning_rate": 6.688685892596508e-05, "loss": 1.0222, "step": 132635 }, { "epoch": 0.33, "learning_rate": 6.688560068448338e-05, "loss": 1.0215, "step": 132640 }, { "epoch": 0.33, "learning_rate": 6.688434244300167e-05, "loss": 1.0211, "step": 132645 }, { "epoch": 0.33, "learning_rate": 6.688308420151996e-05, "loss": 1.0205, "step": 132650 }, { "epoch": 0.33, "learning_rate": 6.688182596003826e-05, "loss": 1.0204, "step": 132655 }, { "epoch": 0.33, "learning_rate": 6.688056771855656e-05, "loss": 1.0208, "step": 132660 }, { "epoch": 0.33, "learning_rate": 6.687930947707485e-05, "loss": 1.0211, "step": 132665 }, { "epoch": 0.33, "learning_rate": 6.687805123559314e-05, "loss": 1.0195, "step": 132670 }, { "epoch": 0.33, "learning_rate": 6.687679299411143e-05, "loss": 1.0398, "step": 132675 }, { "epoch": 0.33, "learning_rate": 6.687553475262974e-05, "loss": 1.0192, "step": 132680 }, { "epoch": 0.33, "learning_rate": 6.687427651114803e-05, "loss": 1.0192, "step": 132685 }, { "epoch": 0.33, "learning_rate": 6.687301826966632e-05, "loss": 1.0221, "step": 132690 }, { "epoch": 0.33, "learning_rate": 6.687176002818461e-05, "loss": 1.0213, "step": 132695 }, { "epoch": 0.33, "learning_rate": 6.687050178670292e-05, "loss": 1.0208, "step": 132700 }, { "epoch": 0.33, "learning_rate": 6.686924354522121e-05, "loss": 1.0207, "step": 132705 }, { "epoch": 0.33, "learning_rate": 6.68679853037395e-05, "loss": 1.0206, "step": 132710 }, { "epoch": 0.33, "learning_rate": 6.686672706225779e-05, "loss": 1.0219, "step": 132715 }, { "epoch": 0.33, "learning_rate": 6.68654688207761e-05, "loss": 1.0228, "step": 132720 }, { "epoch": 0.33, "learning_rate": 6.686421057929439e-05, "loss": 1.0193, "step": 132725 }, { "epoch": 0.33, "learning_rate": 6.686295233781268e-05, "loss": 1.0198, "step": 132730 }, { "epoch": 0.33, "learning_rate": 6.686169409633097e-05, "loss": 1.0214, "step": 132735 }, { "epoch": 0.33, "learning_rate": 6.686043585484926e-05, "loss": 1.0225, "step": 132740 }, { "epoch": 0.33, "learning_rate": 6.685917761336757e-05, "loss": 1.0237, "step": 132745 }, { "epoch": 0.33, "learning_rate": 6.685791937188586e-05, "loss": 1.0409, "step": 132750 }, { "epoch": 0.33, "learning_rate": 6.685666113040415e-05, "loss": 1.0426, "step": 132755 }, { "epoch": 0.33, "learning_rate": 6.685540288892244e-05, "loss": 1.0217, "step": 132760 }, { "epoch": 0.33, "learning_rate": 6.685414464744075e-05, "loss": 1.0232, "step": 132765 }, { "epoch": 0.33, "learning_rate": 6.685288640595904e-05, "loss": 1.0225, "step": 132770 }, { "epoch": 0.33, "learning_rate": 6.685162816447733e-05, "loss": 1.0196, "step": 132775 }, { "epoch": 0.33, "learning_rate": 6.685036992299562e-05, "loss": 1.0217, "step": 132780 }, { "epoch": 0.33, "learning_rate": 6.684911168151393e-05, "loss": 1.0212, "step": 132785 }, { "epoch": 0.33, "learning_rate": 6.684785344003222e-05, "loss": 1.02, "step": 132790 }, { "epoch": 0.33, "learning_rate": 6.684659519855051e-05, "loss": 1.022, "step": 132795 }, { "epoch": 0.33, "learning_rate": 6.68453369570688e-05, "loss": 1.0456, "step": 132800 }, { "epoch": 0.33, "learning_rate": 6.68440787155871e-05, "loss": 1.0206, "step": 132805 }, { "epoch": 0.33, "learning_rate": 6.68428204741054e-05, "loss": 1.0185, "step": 132810 }, { "epoch": 0.33, "learning_rate": 6.684156223262369e-05, "loss": 1.0209, "step": 132815 }, { "epoch": 0.33, "learning_rate": 6.684030399114198e-05, "loss": 1.0199, "step": 132820 }, { "epoch": 0.33, "learning_rate": 6.683904574966027e-05, "loss": 1.0219, "step": 132825 }, { "epoch": 0.33, "learning_rate": 6.683778750817858e-05, "loss": 1.0207, "step": 132830 }, { "epoch": 0.33, "learning_rate": 6.683652926669687e-05, "loss": 1.0235, "step": 132835 }, { "epoch": 0.33, "learning_rate": 6.683527102521516e-05, "loss": 1.0224, "step": 132840 }, { "epoch": 0.33, "learning_rate": 6.683401278373345e-05, "loss": 1.0207, "step": 132845 }, { "epoch": 0.33, "learning_rate": 6.683275454225175e-05, "loss": 1.0463, "step": 132850 }, { "epoch": 0.33, "learning_rate": 6.683149630077005e-05, "loss": 1.022, "step": 132855 }, { "epoch": 0.33, "learning_rate": 6.683023805928834e-05, "loss": 1.0213, "step": 132860 }, { "epoch": 0.33, "learning_rate": 6.682897981780663e-05, "loss": 1.0206, "step": 132865 }, { "epoch": 0.33, "learning_rate": 6.682772157632493e-05, "loss": 1.0203, "step": 132870 }, { "epoch": 0.33, "learning_rate": 6.682646333484323e-05, "loss": 1.0219, "step": 132875 }, { "epoch": 0.33, "learning_rate": 6.682520509336152e-05, "loss": 1.0218, "step": 132880 }, { "epoch": 0.33, "learning_rate": 6.682394685187981e-05, "loss": 1.0221, "step": 132885 }, { "epoch": 0.33, "learning_rate": 6.68226886103981e-05, "loss": 1.0215, "step": 132890 }, { "epoch": 0.33, "learning_rate": 6.682143036891641e-05, "loss": 1.022, "step": 132895 }, { "epoch": 0.33, "learning_rate": 6.68201721274347e-05, "loss": 1.0213, "step": 132900 }, { "epoch": 0.33, "learning_rate": 6.6818913885953e-05, "loss": 1.0199, "step": 132905 }, { "epoch": 0.33, "learning_rate": 6.681765564447129e-05, "loss": 1.0191, "step": 132910 }, { "epoch": 0.33, "learning_rate": 6.681639740298958e-05, "loss": 1.0218, "step": 132915 }, { "epoch": 0.33, "learning_rate": 6.681513916150788e-05, "loss": 1.0224, "step": 132920 }, { "epoch": 0.33, "learning_rate": 6.681388092002617e-05, "loss": 1.0211, "step": 132925 }, { "epoch": 0.33, "learning_rate": 6.681262267854447e-05, "loss": 1.021, "step": 132930 }, { "epoch": 0.33, "learning_rate": 6.681136443706276e-05, "loss": 1.019, "step": 132935 }, { "epoch": 0.33, "learning_rate": 6.681010619558106e-05, "loss": 1.0222, "step": 132940 }, { "epoch": 0.33, "learning_rate": 6.680884795409935e-05, "loss": 1.0232, "step": 132945 }, { "epoch": 0.33, "learning_rate": 6.680758971261765e-05, "loss": 1.0196, "step": 132950 }, { "epoch": 0.33, "learning_rate": 6.680633147113594e-05, "loss": 1.0212, "step": 132955 }, { "epoch": 0.33, "learning_rate": 6.680507322965424e-05, "loss": 1.0231, "step": 132960 }, { "epoch": 0.33, "learning_rate": 6.680381498817253e-05, "loss": 1.0213, "step": 132965 }, { "epoch": 0.33, "learning_rate": 6.680255674669082e-05, "loss": 1.0209, "step": 132970 }, { "epoch": 0.33, "learning_rate": 6.680129850520912e-05, "loss": 1.0208, "step": 132975 }, { "epoch": 0.33, "learning_rate": 6.680004026372741e-05, "loss": 1.0449, "step": 132980 }, { "epoch": 0.33, "learning_rate": 6.679878202224571e-05, "loss": 1.0209, "step": 132985 }, { "epoch": 0.33, "learning_rate": 6.6797523780764e-05, "loss": 1.0223, "step": 132990 }, { "epoch": 0.33, "learning_rate": 6.67962655392823e-05, "loss": 1.0211, "step": 132995 }, { "epoch": 0.33, "learning_rate": 6.679500729780059e-05, "loss": 1.0247, "step": 133000 }, { "epoch": 0.33, "learning_rate": 6.67937490563189e-05, "loss": 1.0189, "step": 133005 }, { "epoch": 0.33, "learning_rate": 6.679249081483718e-05, "loss": 1.023, "step": 133010 }, { "epoch": 0.33, "learning_rate": 6.679123257335548e-05, "loss": 1.0235, "step": 133015 }, { "epoch": 0.33, "learning_rate": 6.678997433187377e-05, "loss": 1.0204, "step": 133020 }, { "epoch": 0.33, "learning_rate": 6.678871609039207e-05, "loss": 1.0201, "step": 133025 }, { "epoch": 0.33, "learning_rate": 6.678745784891036e-05, "loss": 1.0457, "step": 133030 }, { "epoch": 0.33, "learning_rate": 6.678619960742866e-05, "loss": 1.0216, "step": 133035 }, { "epoch": 0.33, "learning_rate": 6.678494136594695e-05, "loss": 1.0214, "step": 133040 }, { "epoch": 0.33, "learning_rate": 6.678368312446524e-05, "loss": 1.0222, "step": 133045 }, { "epoch": 0.33, "learning_rate": 6.678242488298354e-05, "loss": 1.0247, "step": 133050 }, { "epoch": 0.33, "learning_rate": 6.678116664150184e-05, "loss": 1.0198, "step": 133055 }, { "epoch": 0.33, "learning_rate": 6.677990840002013e-05, "loss": 1.0208, "step": 133060 }, { "epoch": 0.33, "learning_rate": 6.677865015853842e-05, "loss": 1.0232, "step": 133065 }, { "epoch": 0.33, "learning_rate": 6.677739191705672e-05, "loss": 1.0225, "step": 133070 }, { "epoch": 0.33, "learning_rate": 6.677613367557502e-05, "loss": 1.0225, "step": 133075 }, { "epoch": 0.33, "learning_rate": 6.677487543409331e-05, "loss": 1.0203, "step": 133080 }, { "epoch": 0.33, "learning_rate": 6.67736171926116e-05, "loss": 1.0232, "step": 133085 }, { "epoch": 0.33, "learning_rate": 6.67723589511299e-05, "loss": 1.0219, "step": 133090 }, { "epoch": 0.33, "learning_rate": 6.677110070964821e-05, "loss": 1.0201, "step": 133095 }, { "epoch": 0.33, "learning_rate": 6.67698424681665e-05, "loss": 1.0227, "step": 133100 }, { "epoch": 0.33, "learning_rate": 6.676858422668479e-05, "loss": 1.0204, "step": 133105 }, { "epoch": 0.33, "learning_rate": 6.676732598520308e-05, "loss": 1.0223, "step": 133110 }, { "epoch": 0.33, "learning_rate": 6.676606774372139e-05, "loss": 1.0213, "step": 133115 }, { "epoch": 0.33, "learning_rate": 6.676480950223968e-05, "loss": 1.0218, "step": 133120 }, { "epoch": 0.33, "learning_rate": 6.676355126075797e-05, "loss": 1.0224, "step": 133125 }, { "epoch": 0.33, "learning_rate": 6.676229301927626e-05, "loss": 1.0192, "step": 133130 }, { "epoch": 0.33, "learning_rate": 6.676103477779456e-05, "loss": 1.0217, "step": 133135 }, { "epoch": 0.33, "learning_rate": 6.675977653631286e-05, "loss": 1.0221, "step": 133140 }, { "epoch": 0.33, "learning_rate": 6.675851829483115e-05, "loss": 1.0188, "step": 133145 }, { "epoch": 0.33, "learning_rate": 6.675726005334944e-05, "loss": 1.0439, "step": 133150 }, { "epoch": 0.33, "learning_rate": 6.675600181186774e-05, "loss": 1.0215, "step": 133155 }, { "epoch": 0.33, "learning_rate": 6.675474357038604e-05, "loss": 1.0201, "step": 133160 }, { "epoch": 0.33, "learning_rate": 6.675348532890433e-05, "loss": 1.0202, "step": 133165 }, { "epoch": 0.33, "learning_rate": 6.675222708742262e-05, "loss": 1.0195, "step": 133170 }, { "epoch": 0.33, "learning_rate": 6.675096884594092e-05, "loss": 1.0211, "step": 133175 }, { "epoch": 0.33, "learning_rate": 6.674971060445922e-05, "loss": 1.0215, "step": 133180 }, { "epoch": 0.33, "learning_rate": 6.674845236297751e-05, "loss": 1.0199, "step": 133185 }, { "epoch": 0.33, "learning_rate": 6.67471941214958e-05, "loss": 1.0229, "step": 133190 }, { "epoch": 0.33, "learning_rate": 6.67459358800141e-05, "loss": 1.022, "step": 133195 }, { "epoch": 0.33, "learning_rate": 6.674467763853239e-05, "loss": 1.0207, "step": 133200 }, { "epoch": 0.33, "learning_rate": 6.674341939705069e-05, "loss": 1.0223, "step": 133205 }, { "epoch": 0.33, "learning_rate": 6.674216115556898e-05, "loss": 1.0208, "step": 133210 }, { "epoch": 0.33, "learning_rate": 6.674090291408728e-05, "loss": 1.0204, "step": 133215 }, { "epoch": 0.33, "learning_rate": 6.673964467260557e-05, "loss": 1.0216, "step": 133220 }, { "epoch": 0.33, "learning_rate": 6.673838643112387e-05, "loss": 1.0206, "step": 133225 }, { "epoch": 0.33, "learning_rate": 6.673712818964216e-05, "loss": 1.0205, "step": 133230 }, { "epoch": 0.33, "learning_rate": 6.673586994816046e-05, "loss": 1.021, "step": 133235 }, { "epoch": 0.33, "learning_rate": 6.673461170667875e-05, "loss": 1.0183, "step": 133240 }, { "epoch": 0.33, "learning_rate": 6.673335346519705e-05, "loss": 1.0197, "step": 133245 }, { "epoch": 0.33, "learning_rate": 6.673209522371534e-05, "loss": 1.0202, "step": 133250 }, { "epoch": 0.33, "learning_rate": 6.673083698223364e-05, "loss": 1.0202, "step": 133255 }, { "epoch": 0.33, "learning_rate": 6.672957874075193e-05, "loss": 1.0234, "step": 133260 }, { "epoch": 0.33, "learning_rate": 6.672832049927022e-05, "loss": 1.0212, "step": 133265 }, { "epoch": 0.33, "learning_rate": 6.672706225778852e-05, "loss": 1.0209, "step": 133270 }, { "epoch": 0.33, "learning_rate": 6.672580401630682e-05, "loss": 1.0233, "step": 133275 }, { "epoch": 0.33, "learning_rate": 6.672454577482511e-05, "loss": 1.0218, "step": 133280 }, { "epoch": 0.33, "learning_rate": 6.67232875333434e-05, "loss": 1.0206, "step": 133285 }, { "epoch": 0.33, "learning_rate": 6.672228094015803e-05, "loss": 1.0206, "step": 133290 }, { "epoch": 0.33, "learning_rate": 6.672102269867634e-05, "loss": 1.0237, "step": 133295 }, { "epoch": 0.33, "learning_rate": 6.671976445719463e-05, "loss": 1.0225, "step": 133300 }, { "epoch": 0.33, "learning_rate": 6.671850621571292e-05, "loss": 1.0245, "step": 133305 }, { "epoch": 0.33, "learning_rate": 6.671724797423123e-05, "loss": 1.0208, "step": 133310 }, { "epoch": 0.33, "learning_rate": 6.671598973274952e-05, "loss": 1.0201, "step": 133315 }, { "epoch": 0.33, "learning_rate": 6.671473149126781e-05, "loss": 1.0172, "step": 133320 }, { "epoch": 0.33, "learning_rate": 6.67134732497861e-05, "loss": 1.021, "step": 133325 }, { "epoch": 0.33, "learning_rate": 6.671221500830441e-05, "loss": 1.0211, "step": 133330 }, { "epoch": 0.33, "learning_rate": 6.67109567668227e-05, "loss": 1.0222, "step": 133335 }, { "epoch": 0.33, "learning_rate": 6.670969852534099e-05, "loss": 1.0216, "step": 133340 }, { "epoch": 0.33, "learning_rate": 6.670844028385928e-05, "loss": 1.0216, "step": 133345 }, { "epoch": 0.33, "learning_rate": 6.670718204237757e-05, "loss": 1.0226, "step": 133350 }, { "epoch": 0.33, "learning_rate": 6.670592380089588e-05, "loss": 1.0232, "step": 133355 }, { "epoch": 0.33, "learning_rate": 6.670466555941417e-05, "loss": 1.0212, "step": 133360 }, { "epoch": 0.33, "learning_rate": 6.670340731793246e-05, "loss": 1.0221, "step": 133365 }, { "epoch": 0.33, "learning_rate": 6.670214907645075e-05, "loss": 1.0349, "step": 133370 }, { "epoch": 0.33, "learning_rate": 6.670089083496906e-05, "loss": 1.0202, "step": 133375 }, { "epoch": 0.33, "learning_rate": 6.669963259348735e-05, "loss": 1.0219, "step": 133380 }, { "epoch": 0.33, "learning_rate": 6.669837435200564e-05, "loss": 1.0201, "step": 133385 }, { "epoch": 0.33, "learning_rate": 6.669711611052393e-05, "loss": 1.0208, "step": 133390 }, { "epoch": 0.33, "learning_rate": 6.669585786904224e-05, "loss": 1.022, "step": 133395 }, { "epoch": 0.33, "learning_rate": 6.669459962756053e-05, "loss": 1.0228, "step": 133400 }, { "epoch": 0.33, "learning_rate": 6.669334138607882e-05, "loss": 1.0364, "step": 133405 }, { "epoch": 0.33, "learning_rate": 6.669208314459711e-05, "loss": 1.0199, "step": 133410 }, { "epoch": 0.33, "learning_rate": 6.66908249031154e-05, "loss": 1.0213, "step": 133415 }, { "epoch": 0.33, "learning_rate": 6.668956666163371e-05, "loss": 1.0216, "step": 133420 }, { "epoch": 0.33, "learning_rate": 6.6688308420152e-05, "loss": 1.022, "step": 133425 }, { "epoch": 0.33, "learning_rate": 6.66870501786703e-05, "loss": 1.0223, "step": 133430 }, { "epoch": 0.33, "learning_rate": 6.668579193718859e-05, "loss": 1.0247, "step": 133435 }, { "epoch": 0.33, "learning_rate": 6.668453369570689e-05, "loss": 1.021, "step": 133440 }, { "epoch": 0.33, "learning_rate": 6.668327545422518e-05, "loss": 1.0212, "step": 133445 }, { "epoch": 0.33, "learning_rate": 6.668201721274347e-05, "loss": 1.0233, "step": 133450 }, { "epoch": 0.33, "learning_rate": 6.668075897126177e-05, "loss": 1.0222, "step": 133455 }, { "epoch": 0.34, "learning_rate": 6.667950072978007e-05, "loss": 1.0218, "step": 133460 }, { "epoch": 0.34, "learning_rate": 6.667824248829836e-05, "loss": 1.0199, "step": 133465 }, { "epoch": 0.34, "learning_rate": 6.667698424681665e-05, "loss": 1.0219, "step": 133470 }, { "epoch": 0.34, "learning_rate": 6.667572600533495e-05, "loss": 1.0204, "step": 133475 }, { "epoch": 0.34, "learning_rate": 6.667446776385324e-05, "loss": 1.0223, "step": 133480 }, { "epoch": 0.34, "learning_rate": 6.667320952237154e-05, "loss": 1.0209, "step": 133485 }, { "epoch": 0.34, "learning_rate": 6.667195128088983e-05, "loss": 1.0207, "step": 133490 }, { "epoch": 0.34, "learning_rate": 6.667069303940813e-05, "loss": 1.0243, "step": 133495 }, { "epoch": 0.34, "learning_rate": 6.666943479792642e-05, "loss": 1.021, "step": 133500 }, { "epoch": 0.34, "learning_rate": 6.666817655644472e-05, "loss": 1.0222, "step": 133505 }, { "epoch": 0.34, "learning_rate": 6.666691831496301e-05, "loss": 1.0227, "step": 133510 }, { "epoch": 0.34, "learning_rate": 6.66656600734813e-05, "loss": 1.0198, "step": 133515 }, { "epoch": 0.34, "learning_rate": 6.66644018319996e-05, "loss": 1.022, "step": 133520 }, { "epoch": 0.34, "learning_rate": 6.666314359051789e-05, "loss": 1.0211, "step": 133525 }, { "epoch": 0.34, "learning_rate": 6.66618853490362e-05, "loss": 1.0218, "step": 133530 }, { "epoch": 0.34, "learning_rate": 6.666062710755449e-05, "loss": 1.0216, "step": 133535 }, { "epoch": 0.34, "learning_rate": 6.665936886607278e-05, "loss": 1.0197, "step": 133540 }, { "epoch": 0.34, "learning_rate": 6.665811062459107e-05, "loss": 1.021, "step": 133545 }, { "epoch": 0.34, "learning_rate": 6.665685238310937e-05, "loss": 1.0214, "step": 133550 }, { "epoch": 0.34, "learning_rate": 6.665559414162767e-05, "loss": 1.0194, "step": 133555 }, { "epoch": 0.34, "learning_rate": 6.665433590014596e-05, "loss": 1.0228, "step": 133560 }, { "epoch": 0.34, "learning_rate": 6.665307765866425e-05, "loss": 1.0194, "step": 133565 }, { "epoch": 0.34, "learning_rate": 6.665181941718255e-05, "loss": 1.0193, "step": 133570 }, { "epoch": 0.34, "learning_rate": 6.665056117570085e-05, "loss": 1.0214, "step": 133575 }, { "epoch": 0.34, "learning_rate": 6.664930293421914e-05, "loss": 1.0214, "step": 133580 }, { "epoch": 0.34, "learning_rate": 6.664804469273743e-05, "loss": 1.0225, "step": 133585 }, { "epoch": 0.34, "learning_rate": 6.664678645125572e-05, "loss": 1.0196, "step": 133590 }, { "epoch": 0.34, "learning_rate": 6.664552820977403e-05, "loss": 1.021, "step": 133595 }, { "epoch": 0.34, "learning_rate": 6.664426996829232e-05, "loss": 1.0183, "step": 133600 }, { "epoch": 0.34, "learning_rate": 6.664301172681061e-05, "loss": 1.0214, "step": 133605 }, { "epoch": 0.34, "learning_rate": 6.66417534853289e-05, "loss": 1.0206, "step": 133610 }, { "epoch": 0.34, "learning_rate": 6.66404952438472e-05, "loss": 1.041, "step": 133615 }, { "epoch": 0.34, "learning_rate": 6.66392370023655e-05, "loss": 1.0213, "step": 133620 }, { "epoch": 0.34, "learning_rate": 6.663797876088379e-05, "loss": 1.0232, "step": 133625 }, { "epoch": 0.34, "learning_rate": 6.663672051940208e-05, "loss": 1.0198, "step": 133630 }, { "epoch": 0.34, "learning_rate": 6.663546227792039e-05, "loss": 1.0214, "step": 133635 }, { "epoch": 0.34, "learning_rate": 6.663420403643868e-05, "loss": 1.0194, "step": 133640 }, { "epoch": 0.34, "learning_rate": 6.663294579495697e-05, "loss": 1.0242, "step": 133645 }, { "epoch": 0.34, "learning_rate": 6.663168755347526e-05, "loss": 1.0197, "step": 133650 }, { "epoch": 0.34, "learning_rate": 6.663042931199355e-05, "loss": 1.0209, "step": 133655 }, { "epoch": 0.34, "learning_rate": 6.662917107051186e-05, "loss": 1.0246, "step": 133660 }, { "epoch": 0.34, "learning_rate": 6.662791282903015e-05, "loss": 1.042, "step": 133665 }, { "epoch": 0.34, "learning_rate": 6.662665458754844e-05, "loss": 1.0201, "step": 133670 }, { "epoch": 0.34, "learning_rate": 6.662539634606673e-05, "loss": 1.0211, "step": 133675 }, { "epoch": 0.34, "learning_rate": 6.662413810458504e-05, "loss": 1.0212, "step": 133680 }, { "epoch": 0.34, "learning_rate": 6.662287986310333e-05, "loss": 1.0234, "step": 133685 }, { "epoch": 0.34, "learning_rate": 6.662162162162162e-05, "loss": 1.0209, "step": 133690 }, { "epoch": 0.34, "learning_rate": 6.662036338013991e-05, "loss": 1.0213, "step": 133695 }, { "epoch": 0.34, "learning_rate": 6.661910513865822e-05, "loss": 1.0226, "step": 133700 }, { "epoch": 0.34, "learning_rate": 6.661784689717651e-05, "loss": 1.0182, "step": 133705 }, { "epoch": 0.34, "learning_rate": 6.66165886556948e-05, "loss": 1.0213, "step": 133710 }, { "epoch": 0.34, "learning_rate": 6.661533041421309e-05, "loss": 1.0205, "step": 133715 }, { "epoch": 0.34, "learning_rate": 6.661407217273138e-05, "loss": 1.0374, "step": 133720 }, { "epoch": 0.34, "learning_rate": 6.661281393124969e-05, "loss": 1.0221, "step": 133725 }, { "epoch": 0.34, "learning_rate": 6.661155568976798e-05, "loss": 1.0229, "step": 133730 }, { "epoch": 0.34, "learning_rate": 6.661029744828627e-05, "loss": 1.0216, "step": 133735 }, { "epoch": 0.34, "learning_rate": 6.660903920680456e-05, "loss": 1.0225, "step": 133740 }, { "epoch": 0.34, "learning_rate": 6.660778096532287e-05, "loss": 1.0215, "step": 133745 }, { "epoch": 0.34, "learning_rate": 6.660652272384116e-05, "loss": 1.0211, "step": 133750 }, { "epoch": 0.34, "learning_rate": 6.660526448235945e-05, "loss": 1.02, "step": 133755 }, { "epoch": 0.34, "learning_rate": 6.660400624087774e-05, "loss": 1.0212, "step": 133760 }, { "epoch": 0.34, "learning_rate": 6.660274799939605e-05, "loss": 1.0204, "step": 133765 }, { "epoch": 0.34, "learning_rate": 6.660148975791434e-05, "loss": 1.0192, "step": 133770 }, { "epoch": 0.34, "learning_rate": 6.660023151643263e-05, "loss": 1.0212, "step": 133775 }, { "epoch": 0.34, "learning_rate": 6.659897327495092e-05, "loss": 1.0199, "step": 133780 }, { "epoch": 0.34, "learning_rate": 6.659771503346921e-05, "loss": 1.0189, "step": 133785 }, { "epoch": 0.34, "learning_rate": 6.659645679198752e-05, "loss": 1.021, "step": 133790 }, { "epoch": 0.34, "learning_rate": 6.659519855050583e-05, "loss": 1.0211, "step": 133795 }, { "epoch": 0.34, "learning_rate": 6.659394030902412e-05, "loss": 1.0241, "step": 133800 }, { "epoch": 0.34, "learning_rate": 6.659268206754241e-05, "loss": 1.0195, "step": 133805 }, { "epoch": 0.34, "learning_rate": 6.65914238260607e-05, "loss": 1.021, "step": 133810 }, { "epoch": 0.34, "learning_rate": 6.6590165584579e-05, "loss": 1.0212, "step": 133815 }, { "epoch": 0.34, "learning_rate": 6.65889073430973e-05, "loss": 1.0208, "step": 133820 }, { "epoch": 0.34, "learning_rate": 6.658764910161559e-05, "loss": 1.0206, "step": 133825 }, { "epoch": 0.34, "learning_rate": 6.658639086013388e-05, "loss": 1.0212, "step": 133830 }, { "epoch": 0.34, "learning_rate": 6.658513261865218e-05, "loss": 1.0203, "step": 133835 }, { "epoch": 0.34, "learning_rate": 6.658387437717048e-05, "loss": 1.0225, "step": 133840 }, { "epoch": 0.34, "learning_rate": 6.658261613568877e-05, "loss": 1.0235, "step": 133845 }, { "epoch": 0.34, "learning_rate": 6.658135789420706e-05, "loss": 1.0194, "step": 133850 }, { "epoch": 0.34, "learning_rate": 6.658009965272536e-05, "loss": 1.0223, "step": 133855 }, { "epoch": 0.34, "learning_rate": 6.657884141124366e-05, "loss": 1.0196, "step": 133860 }, { "epoch": 0.34, "learning_rate": 6.657758316976195e-05, "loss": 1.0229, "step": 133865 }, { "epoch": 0.34, "learning_rate": 6.657632492828024e-05, "loss": 1.0189, "step": 133870 }, { "epoch": 0.34, "learning_rate": 6.657506668679853e-05, "loss": 1.0222, "step": 133875 }, { "epoch": 0.34, "learning_rate": 6.657380844531684e-05, "loss": 1.0485, "step": 133880 }, { "epoch": 0.34, "learning_rate": 6.657255020383513e-05, "loss": 1.018, "step": 133885 }, { "epoch": 0.34, "learning_rate": 6.657129196235342e-05, "loss": 1.0225, "step": 133890 }, { "epoch": 0.34, "learning_rate": 6.657003372087171e-05, "loss": 1.0221, "step": 133895 }, { "epoch": 0.34, "learning_rate": 6.656877547939002e-05, "loss": 1.0234, "step": 133900 }, { "epoch": 0.34, "learning_rate": 6.656751723790831e-05, "loss": 1.0219, "step": 133905 }, { "epoch": 0.34, "learning_rate": 6.65662589964266e-05, "loss": 1.0207, "step": 133910 }, { "epoch": 0.34, "learning_rate": 6.656500075494489e-05, "loss": 1.0214, "step": 133915 }, { "epoch": 0.34, "learning_rate": 6.65637425134632e-05, "loss": 1.0203, "step": 133920 }, { "epoch": 0.34, "learning_rate": 6.656248427198149e-05, "loss": 1.0221, "step": 133925 }, { "epoch": 0.34, "learning_rate": 6.656122603049978e-05, "loss": 1.0201, "step": 133930 }, { "epoch": 0.34, "learning_rate": 6.655996778901807e-05, "loss": 1.0442, "step": 133935 }, { "epoch": 0.34, "learning_rate": 6.655870954753636e-05, "loss": 1.0226, "step": 133940 }, { "epoch": 0.34, "learning_rate": 6.655745130605467e-05, "loss": 1.0209, "step": 133945 }, { "epoch": 0.34, "learning_rate": 6.655619306457296e-05, "loss": 1.0219, "step": 133950 }, { "epoch": 0.34, "learning_rate": 6.655493482309125e-05, "loss": 1.0175, "step": 133955 }, { "epoch": 0.34, "learning_rate": 6.655367658160954e-05, "loss": 1.0199, "step": 133960 }, { "epoch": 0.34, "learning_rate": 6.655241834012785e-05, "loss": 1.0201, "step": 133965 }, { "epoch": 0.34, "learning_rate": 6.655116009864614e-05, "loss": 1.0218, "step": 133970 }, { "epoch": 0.34, "learning_rate": 6.654990185716443e-05, "loss": 1.02, "step": 133975 }, { "epoch": 0.34, "learning_rate": 6.654864361568272e-05, "loss": 1.0216, "step": 133980 }, { "epoch": 0.34, "learning_rate": 6.654738537420103e-05, "loss": 1.021, "step": 133985 }, { "epoch": 0.34, "learning_rate": 6.654612713271932e-05, "loss": 1.024, "step": 133990 }, { "epoch": 0.34, "learning_rate": 6.654486889123761e-05, "loss": 1.0204, "step": 133995 }, { "epoch": 0.34, "learning_rate": 6.65436106497559e-05, "loss": 1.022, "step": 134000 }, { "epoch": 0.34, "learning_rate": 6.65423524082742e-05, "loss": 1.0219, "step": 134005 }, { "epoch": 0.34, "learning_rate": 6.65410941667925e-05, "loss": 1.0218, "step": 134010 }, { "epoch": 0.34, "learning_rate": 6.653983592531079e-05, "loss": 1.0234, "step": 134015 }, { "epoch": 0.34, "learning_rate": 6.653857768382908e-05, "loss": 1.0203, "step": 134020 }, { "epoch": 0.34, "learning_rate": 6.653731944234737e-05, "loss": 1.0227, "step": 134025 }, { "epoch": 0.34, "learning_rate": 6.653606120086568e-05, "loss": 1.0219, "step": 134030 }, { "epoch": 0.34, "learning_rate": 6.653480295938397e-05, "loss": 1.0233, "step": 134035 }, { "epoch": 0.34, "learning_rate": 6.653354471790226e-05, "loss": 1.0212, "step": 134040 }, { "epoch": 0.34, "learning_rate": 6.653228647642055e-05, "loss": 1.0199, "step": 134045 }, { "epoch": 0.34, "learning_rate": 6.653102823493886e-05, "loss": 1.0204, "step": 134050 }, { "epoch": 0.34, "learning_rate": 6.652976999345715e-05, "loss": 1.0225, "step": 134055 }, { "epoch": 0.34, "learning_rate": 6.652851175197544e-05, "loss": 1.0231, "step": 134060 }, { "epoch": 0.34, "learning_rate": 6.652725351049373e-05, "loss": 1.0206, "step": 134065 }, { "epoch": 0.34, "learning_rate": 6.652599526901203e-05, "loss": 1.02, "step": 134070 }, { "epoch": 0.34, "learning_rate": 6.652473702753033e-05, "loss": 1.0199, "step": 134075 }, { "epoch": 0.34, "learning_rate": 6.652347878604862e-05, "loss": 1.0222, "step": 134080 }, { "epoch": 0.34, "learning_rate": 6.652222054456691e-05, "loss": 1.0202, "step": 134085 }, { "epoch": 0.34, "learning_rate": 6.65209623030852e-05, "loss": 1.021, "step": 134090 }, { "epoch": 0.34, "learning_rate": 6.651970406160351e-05, "loss": 1.0206, "step": 134095 }, { "epoch": 0.34, "learning_rate": 6.65184458201218e-05, "loss": 1.0209, "step": 134100 }, { "epoch": 0.34, "learning_rate": 6.65171875786401e-05, "loss": 1.0223, "step": 134105 }, { "epoch": 0.34, "learning_rate": 6.651592933715839e-05, "loss": 1.0547, "step": 134110 }, { "epoch": 0.34, "learning_rate": 6.651467109567669e-05, "loss": 1.0198, "step": 134115 }, { "epoch": 0.34, "learning_rate": 6.651341285419498e-05, "loss": 1.0216, "step": 134120 }, { "epoch": 0.34, "learning_rate": 6.651215461271327e-05, "loss": 1.0221, "step": 134125 }, { "epoch": 0.34, "learning_rate": 6.651089637123157e-05, "loss": 1.0214, "step": 134130 }, { "epoch": 0.34, "learning_rate": 6.650963812974986e-05, "loss": 1.0208, "step": 134135 }, { "epoch": 0.34, "learning_rate": 6.650837988826816e-05, "loss": 1.0202, "step": 134140 }, { "epoch": 0.34, "learning_rate": 6.650712164678645e-05, "loss": 1.0217, "step": 134145 }, { "epoch": 0.34, "learning_rate": 6.650586340530475e-05, "loss": 1.02, "step": 134150 }, { "epoch": 0.34, "learning_rate": 6.650460516382304e-05, "loss": 1.0219, "step": 134155 }, { "epoch": 0.34, "learning_rate": 6.650334692234134e-05, "loss": 1.0213, "step": 134160 }, { "epoch": 0.34, "learning_rate": 6.650208868085963e-05, "loss": 1.021, "step": 134165 }, { "epoch": 0.34, "learning_rate": 6.650083043937793e-05, "loss": 1.0215, "step": 134170 }, { "epoch": 0.34, "learning_rate": 6.649957219789622e-05, "loss": 1.0214, "step": 134175 }, { "epoch": 0.34, "learning_rate": 6.649831395641452e-05, "loss": 1.0206, "step": 134180 }, { "epoch": 0.34, "learning_rate": 6.649705571493281e-05, "loss": 1.0214, "step": 134185 }, { "epoch": 0.34, "learning_rate": 6.64957974734511e-05, "loss": 1.0196, "step": 134190 }, { "epoch": 0.34, "learning_rate": 6.64945392319694e-05, "loss": 1.021, "step": 134195 }, { "epoch": 0.34, "learning_rate": 6.649328099048769e-05, "loss": 1.0203, "step": 134200 }, { "epoch": 0.34, "learning_rate": 6.6492022749006e-05, "loss": 1.02, "step": 134205 }, { "epoch": 0.34, "learning_rate": 6.649076450752429e-05, "loss": 1.0214, "step": 134210 }, { "epoch": 0.34, "learning_rate": 6.648950626604258e-05, "loss": 1.0253, "step": 134215 }, { "epoch": 0.34, "learning_rate": 6.648824802456087e-05, "loss": 1.0204, "step": 134220 }, { "epoch": 0.34, "learning_rate": 6.648698978307917e-05, "loss": 1.0255, "step": 134225 }, { "epoch": 0.34, "learning_rate": 6.648573154159747e-05, "loss": 1.0201, "step": 134230 }, { "epoch": 0.34, "learning_rate": 6.648447330011576e-05, "loss": 1.021, "step": 134235 }, { "epoch": 0.34, "learning_rate": 6.648321505863405e-05, "loss": 1.0213, "step": 134240 }, { "epoch": 0.34, "learning_rate": 6.648195681715234e-05, "loss": 1.0227, "step": 134245 }, { "epoch": 0.34, "learning_rate": 6.648069857567065e-05, "loss": 1.0215, "step": 134250 }, { "epoch": 0.34, "learning_rate": 6.647944033418894e-05, "loss": 1.0222, "step": 134255 }, { "epoch": 0.34, "learning_rate": 6.647818209270723e-05, "loss": 1.0213, "step": 134260 }, { "epoch": 0.34, "learning_rate": 6.647692385122552e-05, "loss": 1.0214, "step": 134265 }, { "epoch": 0.34, "learning_rate": 6.647566560974382e-05, "loss": 1.0207, "step": 134270 }, { "epoch": 0.34, "learning_rate": 6.647440736826212e-05, "loss": 1.0183, "step": 134275 }, { "epoch": 0.34, "learning_rate": 6.647314912678041e-05, "loss": 1.0237, "step": 134280 }, { "epoch": 0.34, "learning_rate": 6.64718908852987e-05, "loss": 1.0208, "step": 134285 }, { "epoch": 0.34, "learning_rate": 6.6470632643817e-05, "loss": 1.0215, "step": 134290 }, { "epoch": 0.34, "learning_rate": 6.646937440233531e-05, "loss": 1.021, "step": 134295 }, { "epoch": 0.34, "learning_rate": 6.64681161608536e-05, "loss": 1.0204, "step": 134300 }, { "epoch": 0.34, "learning_rate": 6.64668579193719e-05, "loss": 1.0227, "step": 134305 }, { "epoch": 0.34, "learning_rate": 6.646559967789018e-05, "loss": 1.0206, "step": 134310 }, { "epoch": 0.34, "learning_rate": 6.646434143640849e-05, "loss": 1.021, "step": 134315 }, { "epoch": 0.34, "learning_rate": 6.646308319492678e-05, "loss": 1.0216, "step": 134320 }, { "epoch": 0.34, "learning_rate": 6.646182495344507e-05, "loss": 1.019, "step": 134325 }, { "epoch": 0.34, "learning_rate": 6.646056671196336e-05, "loss": 1.0424, "step": 134330 }, { "epoch": 0.34, "learning_rate": 6.645930847048166e-05, "loss": 1.0218, "step": 134335 }, { "epoch": 0.34, "learning_rate": 6.645805022899996e-05, "loss": 1.0214, "step": 134340 }, { "epoch": 0.34, "learning_rate": 6.645679198751825e-05, "loss": 1.0204, "step": 134345 }, { "epoch": 0.34, "learning_rate": 6.645553374603654e-05, "loss": 1.0223, "step": 134350 }, { "epoch": 0.34, "learning_rate": 6.645427550455484e-05, "loss": 1.0203, "step": 134355 }, { "epoch": 0.34, "learning_rate": 6.645301726307314e-05, "loss": 1.0207, "step": 134360 }, { "epoch": 0.34, "learning_rate": 6.645175902159143e-05, "loss": 1.0205, "step": 134365 }, { "epoch": 0.34, "learning_rate": 6.645050078010972e-05, "loss": 1.0211, "step": 134370 }, { "epoch": 0.34, "learning_rate": 6.644924253862802e-05, "loss": 1.024, "step": 134375 }, { "epoch": 0.34, "learning_rate": 6.644798429714632e-05, "loss": 1.0219, "step": 134380 }, { "epoch": 0.34, "learning_rate": 6.644672605566461e-05, "loss": 1.0209, "step": 134385 }, { "epoch": 0.34, "learning_rate": 6.64454678141829e-05, "loss": 1.0216, "step": 134390 }, { "epoch": 0.34, "learning_rate": 6.64442095727012e-05, "loss": 1.0198, "step": 134395 }, { "epoch": 0.34, "learning_rate": 6.644295133121949e-05, "loss": 1.0207, "step": 134400 }, { "epoch": 0.34, "learning_rate": 6.644169308973779e-05, "loss": 1.0216, "step": 134405 }, { "epoch": 0.34, "learning_rate": 6.644043484825608e-05, "loss": 1.0212, "step": 134410 }, { "epoch": 0.34, "learning_rate": 6.643917660677438e-05, "loss": 1.0189, "step": 134415 }, { "epoch": 0.34, "learning_rate": 6.643791836529267e-05, "loss": 1.0234, "step": 134420 }, { "epoch": 0.34, "learning_rate": 6.643666012381097e-05, "loss": 1.0223, "step": 134425 }, { "epoch": 0.34, "learning_rate": 6.643540188232926e-05, "loss": 1.0204, "step": 134430 }, { "epoch": 0.34, "learning_rate": 6.643439528914389e-05, "loss": 1.047, "step": 134435 }, { "epoch": 0.34, "learning_rate": 6.643313704766219e-05, "loss": 1.0215, "step": 134440 }, { "epoch": 0.34, "learning_rate": 6.643187880618048e-05, "loss": 1.02, "step": 134445 }, { "epoch": 0.34, "learning_rate": 6.643062056469878e-05, "loss": 1.0205, "step": 134450 }, { "epoch": 0.34, "learning_rate": 6.642936232321707e-05, "loss": 1.0203, "step": 134455 }, { "epoch": 0.34, "learning_rate": 6.642810408173536e-05, "loss": 1.0204, "step": 134460 }, { "epoch": 0.34, "learning_rate": 6.642684584025366e-05, "loss": 1.0219, "step": 134465 }, { "epoch": 0.34, "learning_rate": 6.642558759877196e-05, "loss": 1.0213, "step": 134470 }, { "epoch": 0.34, "learning_rate": 6.642432935729025e-05, "loss": 1.0217, "step": 134475 }, { "epoch": 0.34, "learning_rate": 6.642307111580854e-05, "loss": 1.0211, "step": 134480 }, { "epoch": 0.34, "learning_rate": 6.642181287432684e-05, "loss": 1.0197, "step": 134485 }, { "epoch": 0.34, "learning_rate": 6.642055463284514e-05, "loss": 1.0223, "step": 134490 }, { "epoch": 0.34, "learning_rate": 6.641929639136344e-05, "loss": 1.0215, "step": 134495 }, { "epoch": 0.34, "learning_rate": 6.641803814988173e-05, "loss": 1.0213, "step": 134500 }, { "epoch": 0.34, "learning_rate": 6.641677990840002e-05, "loss": 1.0218, "step": 134505 }, { "epoch": 0.34, "learning_rate": 6.641552166691833e-05, "loss": 1.0216, "step": 134510 }, { "epoch": 0.34, "learning_rate": 6.641426342543662e-05, "loss": 1.0209, "step": 134515 }, { "epoch": 0.34, "learning_rate": 6.641300518395491e-05, "loss": 1.0235, "step": 134520 }, { "epoch": 0.34, "learning_rate": 6.64117469424732e-05, "loss": 1.02, "step": 134525 }, { "epoch": 0.34, "learning_rate": 6.641048870099151e-05, "loss": 1.0201, "step": 134530 }, { "epoch": 0.34, "learning_rate": 6.64092304595098e-05, "loss": 1.0211, "step": 134535 }, { "epoch": 0.34, "learning_rate": 6.640797221802809e-05, "loss": 1.0193, "step": 134540 }, { "epoch": 0.34, "learning_rate": 6.640671397654638e-05, "loss": 1.0196, "step": 134545 }, { "epoch": 0.34, "learning_rate": 6.640545573506467e-05, "loss": 1.0212, "step": 134550 }, { "epoch": 0.34, "learning_rate": 6.640419749358298e-05, "loss": 1.0206, "step": 134555 }, { "epoch": 0.34, "learning_rate": 6.640293925210127e-05, "loss": 1.0198, "step": 134560 }, { "epoch": 0.34, "learning_rate": 6.640168101061956e-05, "loss": 1.022, "step": 134565 }, { "epoch": 0.34, "learning_rate": 6.640042276913785e-05, "loss": 1.0211, "step": 134570 }, { "epoch": 0.34, "learning_rate": 6.639916452765616e-05, "loss": 1.0201, "step": 134575 }, { "epoch": 0.34, "learning_rate": 6.639790628617445e-05, "loss": 1.023, "step": 134580 }, { "epoch": 0.34, "learning_rate": 6.639664804469274e-05, "loss": 1.021, "step": 134585 }, { "epoch": 0.34, "learning_rate": 6.639538980321103e-05, "loss": 1.0232, "step": 134590 }, { "epoch": 0.34, "learning_rate": 6.639413156172934e-05, "loss": 1.0192, "step": 134595 }, { "epoch": 0.34, "learning_rate": 6.639287332024763e-05, "loss": 1.0231, "step": 134600 }, { "epoch": 0.34, "learning_rate": 6.639161507876592e-05, "loss": 1.0193, "step": 134605 }, { "epoch": 0.34, "learning_rate": 6.639035683728421e-05, "loss": 1.0196, "step": 134610 }, { "epoch": 0.34, "learning_rate": 6.63890985958025e-05, "loss": 1.0205, "step": 134615 }, { "epoch": 0.34, "learning_rate": 6.638784035432081e-05, "loss": 1.0227, "step": 134620 }, { "epoch": 0.34, "learning_rate": 6.63865821128391e-05, "loss": 1.0221, "step": 134625 }, { "epoch": 0.34, "learning_rate": 6.63853238713574e-05, "loss": 1.0191, "step": 134630 }, { "epoch": 0.34, "learning_rate": 6.638406562987569e-05, "loss": 1.0189, "step": 134635 }, { "epoch": 0.34, "learning_rate": 6.638280738839399e-05, "loss": 1.0247, "step": 134640 }, { "epoch": 0.34, "learning_rate": 6.638154914691228e-05, "loss": 1.0206, "step": 134645 }, { "epoch": 0.34, "learning_rate": 6.638029090543057e-05, "loss": 1.021, "step": 134650 }, { "epoch": 0.34, "learning_rate": 6.637903266394887e-05, "loss": 1.0234, "step": 134655 }, { "epoch": 0.34, "learning_rate": 6.637777442246717e-05, "loss": 1.0226, "step": 134660 }, { "epoch": 0.34, "learning_rate": 6.637651618098546e-05, "loss": 1.022, "step": 134665 }, { "epoch": 0.34, "learning_rate": 6.637525793950375e-05, "loss": 1.0246, "step": 134670 }, { "epoch": 0.34, "learning_rate": 6.637399969802205e-05, "loss": 1.0245, "step": 134675 }, { "epoch": 0.34, "learning_rate": 6.637274145654034e-05, "loss": 1.0224, "step": 134680 }, { "epoch": 0.34, "learning_rate": 6.637148321505864e-05, "loss": 1.0205, "step": 134685 }, { "epoch": 0.34, "learning_rate": 6.637022497357693e-05, "loss": 1.0215, "step": 134690 }, { "epoch": 0.34, "learning_rate": 6.636896673209523e-05, "loss": 1.0212, "step": 134695 }, { "epoch": 0.34, "learning_rate": 6.636770849061352e-05, "loss": 1.0232, "step": 134700 }, { "epoch": 0.34, "learning_rate": 6.636645024913182e-05, "loss": 1.0216, "step": 134705 }, { "epoch": 0.34, "learning_rate": 6.636519200765011e-05, "loss": 1.0233, "step": 134710 }, { "epoch": 0.34, "learning_rate": 6.63639337661684e-05, "loss": 1.0229, "step": 134715 }, { "epoch": 0.34, "learning_rate": 6.63626755246867e-05, "loss": 1.0206, "step": 134720 }, { "epoch": 0.34, "learning_rate": 6.6361417283205e-05, "loss": 1.0193, "step": 134725 }, { "epoch": 0.34, "learning_rate": 6.63601590417233e-05, "loss": 1.022, "step": 134730 }, { "epoch": 0.34, "learning_rate": 6.635890080024159e-05, "loss": 1.0196, "step": 134735 }, { "epoch": 0.34, "learning_rate": 6.635764255875988e-05, "loss": 1.0218, "step": 134740 }, { "epoch": 0.34, "learning_rate": 6.635638431727817e-05, "loss": 1.0203, "step": 134745 }, { "epoch": 0.34, "learning_rate": 6.635512607579647e-05, "loss": 1.0211, "step": 134750 }, { "epoch": 0.34, "learning_rate": 6.635386783431477e-05, "loss": 1.0224, "step": 134755 }, { "epoch": 0.34, "learning_rate": 6.635260959283306e-05, "loss": 1.0204, "step": 134760 }, { "epoch": 0.34, "learning_rate": 6.635135135135135e-05, "loss": 1.0207, "step": 134765 }, { "epoch": 0.34, "learning_rate": 6.635009310986965e-05, "loss": 1.0211, "step": 134770 }, { "epoch": 0.34, "learning_rate": 6.634883486838795e-05, "loss": 1.0213, "step": 134775 }, { "epoch": 0.34, "learning_rate": 6.634757662690624e-05, "loss": 1.019, "step": 134780 }, { "epoch": 0.34, "learning_rate": 6.634631838542453e-05, "loss": 1.0228, "step": 134785 }, { "epoch": 0.34, "learning_rate": 6.634506014394283e-05, "loss": 1.0211, "step": 134790 }, { "epoch": 0.34, "learning_rate": 6.634380190246113e-05, "loss": 1.0211, "step": 134795 }, { "epoch": 0.34, "learning_rate": 6.634254366097942e-05, "loss": 1.0197, "step": 134800 }, { "epoch": 0.34, "learning_rate": 6.634128541949771e-05, "loss": 1.0186, "step": 134805 }, { "epoch": 0.34, "learning_rate": 6.6340027178016e-05, "loss": 1.0204, "step": 134810 }, { "epoch": 0.34, "learning_rate": 6.63387689365343e-05, "loss": 1.02, "step": 134815 }, { "epoch": 0.34, "learning_rate": 6.63375106950526e-05, "loss": 1.021, "step": 134820 }, { "epoch": 0.34, "learning_rate": 6.633625245357089e-05, "loss": 1.0226, "step": 134825 }, { "epoch": 0.34, "learning_rate": 6.633499421208918e-05, "loss": 1.0243, "step": 134830 }, { "epoch": 0.34, "learning_rate": 6.633373597060749e-05, "loss": 1.0226, "step": 134835 }, { "epoch": 0.34, "learning_rate": 6.633247772912578e-05, "loss": 1.0212, "step": 134840 }, { "epoch": 0.34, "learning_rate": 6.633121948764407e-05, "loss": 1.0204, "step": 134845 }, { "epoch": 0.34, "learning_rate": 6.632996124616236e-05, "loss": 1.0199, "step": 134850 }, { "epoch": 0.34, "learning_rate": 6.632870300468067e-05, "loss": 1.021, "step": 134855 }, { "epoch": 0.34, "learning_rate": 6.632744476319896e-05, "loss": 1.0195, "step": 134860 }, { "epoch": 0.34, "learning_rate": 6.632618652171725e-05, "loss": 1.0222, "step": 134865 }, { "epoch": 0.34, "learning_rate": 6.632492828023554e-05, "loss": 1.0227, "step": 134870 }, { "epoch": 0.34, "learning_rate": 6.632367003875383e-05, "loss": 1.0218, "step": 134875 }, { "epoch": 0.34, "learning_rate": 6.632241179727214e-05, "loss": 1.0193, "step": 134880 }, { "epoch": 0.34, "learning_rate": 6.632115355579043e-05, "loss": 1.0216, "step": 134885 }, { "epoch": 0.34, "learning_rate": 6.631989531430872e-05, "loss": 1.019, "step": 134890 }, { "epoch": 0.34, "learning_rate": 6.631863707282701e-05, "loss": 1.0236, "step": 134895 }, { "epoch": 0.34, "learning_rate": 6.631737883134532e-05, "loss": 1.0219, "step": 134900 }, { "epoch": 0.34, "learning_rate": 6.631612058986361e-05, "loss": 1.0214, "step": 134905 }, { "epoch": 0.34, "learning_rate": 6.63148623483819e-05, "loss": 1.0207, "step": 134910 }, { "epoch": 0.34, "learning_rate": 6.631360410690019e-05, "loss": 1.0211, "step": 134915 }, { "epoch": 0.34, "learning_rate": 6.63123458654185e-05, "loss": 1.0217, "step": 134920 }, { "epoch": 0.34, "learning_rate": 6.631108762393679e-05, "loss": 1.0203, "step": 134925 }, { "epoch": 0.34, "learning_rate": 6.630982938245508e-05, "loss": 1.0217, "step": 134930 }, { "epoch": 0.34, "learning_rate": 6.630857114097337e-05, "loss": 1.0206, "step": 134935 }, { "epoch": 0.34, "learning_rate": 6.630731289949166e-05, "loss": 1.0206, "step": 134940 }, { "epoch": 0.34, "learning_rate": 6.630605465800997e-05, "loss": 1.0205, "step": 134945 }, { "epoch": 0.34, "learning_rate": 6.630479641652826e-05, "loss": 1.0201, "step": 134950 }, { "epoch": 0.34, "learning_rate": 6.630353817504655e-05, "loss": 1.0236, "step": 134955 }, { "epoch": 0.34, "learning_rate": 6.630227993356484e-05, "loss": 1.0218, "step": 134960 }, { "epoch": 0.34, "learning_rate": 6.630102169208315e-05, "loss": 1.0214, "step": 134965 }, { "epoch": 0.34, "learning_rate": 6.629976345060144e-05, "loss": 1.0222, "step": 134970 }, { "epoch": 0.34, "learning_rate": 6.629850520911973e-05, "loss": 1.0392, "step": 134975 }, { "epoch": 0.34, "learning_rate": 6.629724696763802e-05, "loss": 1.0207, "step": 134980 }, { "epoch": 0.34, "learning_rate": 6.629598872615632e-05, "loss": 1.0374, "step": 134985 }, { "epoch": 0.34, "learning_rate": 6.629473048467462e-05, "loss": 1.0205, "step": 134990 }, { "epoch": 0.34, "learning_rate": 6.629347224319293e-05, "loss": 1.0197, "step": 134995 }, { "epoch": 0.34, "learning_rate": 6.629221400171122e-05, "loss": 1.022, "step": 135000 }, { "epoch": 0.34, "learning_rate": 6.629095576022951e-05, "loss": 1.0208, "step": 135005 }, { "epoch": 0.34, "learning_rate": 6.62896975187478e-05, "loss": 1.0225, "step": 135010 }, { "epoch": 0.34, "learning_rate": 6.62884392772661e-05, "loss": 1.0188, "step": 135015 }, { "epoch": 0.34, "learning_rate": 6.62871810357844e-05, "loss": 1.021, "step": 135020 }, { "epoch": 0.34, "learning_rate": 6.628592279430269e-05, "loss": 1.0203, "step": 135025 }, { "epoch": 0.34, "learning_rate": 6.628466455282098e-05, "loss": 1.0206, "step": 135030 }, { "epoch": 0.34, "learning_rate": 6.628340631133929e-05, "loss": 1.021, "step": 135035 }, { "epoch": 0.34, "learning_rate": 6.628214806985758e-05, "loss": 1.022, "step": 135040 }, { "epoch": 0.34, "learning_rate": 6.628088982837587e-05, "loss": 1.0196, "step": 135045 }, { "epoch": 0.34, "learning_rate": 6.627963158689416e-05, "loss": 1.0202, "step": 135050 }, { "epoch": 0.34, "learning_rate": 6.627837334541247e-05, "loss": 1.0194, "step": 135055 }, { "epoch": 0.34, "learning_rate": 6.627711510393076e-05, "loss": 1.0203, "step": 135060 }, { "epoch": 0.34, "learning_rate": 6.627585686244905e-05, "loss": 1.0206, "step": 135065 }, { "epoch": 0.34, "learning_rate": 6.627459862096734e-05, "loss": 1.0227, "step": 135070 }, { "epoch": 0.34, "learning_rate": 6.627334037948563e-05, "loss": 1.0195, "step": 135075 }, { "epoch": 0.34, "learning_rate": 6.627208213800394e-05, "loss": 1.0213, "step": 135080 }, { "epoch": 0.34, "learning_rate": 6.627082389652223e-05, "loss": 1.0198, "step": 135085 }, { "epoch": 0.34, "learning_rate": 6.626956565504052e-05, "loss": 1.0224, "step": 135090 }, { "epoch": 0.34, "learning_rate": 6.626830741355881e-05, "loss": 1.0216, "step": 135095 }, { "epoch": 0.34, "learning_rate": 6.626704917207712e-05, "loss": 1.0205, "step": 135100 }, { "epoch": 0.34, "learning_rate": 6.626579093059541e-05, "loss": 1.0209, "step": 135105 }, { "epoch": 0.34, "learning_rate": 6.62645326891137e-05, "loss": 1.018, "step": 135110 }, { "epoch": 0.34, "learning_rate": 6.626327444763199e-05, "loss": 1.0213, "step": 135115 }, { "epoch": 0.34, "learning_rate": 6.62620162061503e-05, "loss": 1.0228, "step": 135120 }, { "epoch": 0.34, "learning_rate": 6.626075796466859e-05, "loss": 1.0185, "step": 135125 }, { "epoch": 0.34, "learning_rate": 6.625949972318688e-05, "loss": 1.0214, "step": 135130 }, { "epoch": 0.34, "learning_rate": 6.625824148170517e-05, "loss": 1.0225, "step": 135135 }, { "epoch": 0.34, "learning_rate": 6.625698324022346e-05, "loss": 1.0236, "step": 135140 }, { "epoch": 0.34, "learning_rate": 6.625572499874177e-05, "loss": 1.0214, "step": 135145 }, { "epoch": 0.34, "learning_rate": 6.625446675726006e-05, "loss": 1.0235, "step": 135150 }, { "epoch": 0.34, "learning_rate": 6.625320851577835e-05, "loss": 1.0226, "step": 135155 }, { "epoch": 0.34, "learning_rate": 6.625195027429664e-05, "loss": 1.0232, "step": 135160 }, { "epoch": 0.34, "learning_rate": 6.625069203281495e-05, "loss": 1.0211, "step": 135165 }, { "epoch": 0.34, "learning_rate": 6.624943379133324e-05, "loss": 1.0227, "step": 135170 }, { "epoch": 0.34, "learning_rate": 6.624817554985153e-05, "loss": 1.0181, "step": 135175 }, { "epoch": 0.34, "learning_rate": 6.624691730836982e-05, "loss": 1.022, "step": 135180 }, { "epoch": 0.34, "learning_rate": 6.624565906688813e-05, "loss": 1.0213, "step": 135185 }, { "epoch": 0.34, "learning_rate": 6.624440082540642e-05, "loss": 1.0217, "step": 135190 }, { "epoch": 0.34, "learning_rate": 6.624314258392471e-05, "loss": 1.0222, "step": 135195 }, { "epoch": 0.34, "learning_rate": 6.6241884342443e-05, "loss": 1.0182, "step": 135200 }, { "epoch": 0.34, "learning_rate": 6.62406261009613e-05, "loss": 1.0204, "step": 135205 }, { "epoch": 0.34, "learning_rate": 6.62393678594796e-05, "loss": 1.0219, "step": 135210 }, { "epoch": 0.34, "learning_rate": 6.623810961799789e-05, "loss": 1.0232, "step": 135215 }, { "epoch": 0.34, "learning_rate": 6.623685137651618e-05, "loss": 1.022, "step": 135220 }, { "epoch": 0.34, "learning_rate": 6.623559313503447e-05, "loss": 1.0247, "step": 135225 }, { "epoch": 0.34, "learning_rate": 6.623433489355278e-05, "loss": 1.0234, "step": 135230 }, { "epoch": 0.34, "learning_rate": 6.623307665207107e-05, "loss": 1.0181, "step": 135235 }, { "epoch": 0.34, "learning_rate": 6.623181841058936e-05, "loss": 1.0213, "step": 135240 }, { "epoch": 0.34, "learning_rate": 6.623056016910765e-05, "loss": 1.023, "step": 135245 }, { "epoch": 0.34, "learning_rate": 6.622930192762596e-05, "loss": 1.0214, "step": 135250 }, { "epoch": 0.34, "learning_rate": 6.622804368614425e-05, "loss": 1.0219, "step": 135255 }, { "epoch": 0.34, "learning_rate": 6.622678544466254e-05, "loss": 1.0208, "step": 135260 }, { "epoch": 0.34, "learning_rate": 6.622552720318083e-05, "loss": 1.021, "step": 135265 }, { "epoch": 0.34, "learning_rate": 6.622426896169913e-05, "loss": 1.021, "step": 135270 }, { "epoch": 0.34, "learning_rate": 6.622301072021743e-05, "loss": 1.0199, "step": 135275 }, { "epoch": 0.34, "learning_rate": 6.622175247873572e-05, "loss": 1.0216, "step": 135280 }, { "epoch": 0.34, "learning_rate": 6.622049423725401e-05, "loss": 1.0219, "step": 135285 }, { "epoch": 0.34, "learning_rate": 6.62192359957723e-05, "loss": 1.0227, "step": 135290 }, { "epoch": 0.34, "learning_rate": 6.621797775429061e-05, "loss": 1.0227, "step": 135295 }, { "epoch": 0.34, "learning_rate": 6.62167195128089e-05, "loss": 1.0209, "step": 135300 }, { "epoch": 0.34, "learning_rate": 6.62154612713272e-05, "loss": 1.0234, "step": 135305 }, { "epoch": 0.34, "learning_rate": 6.621420302984549e-05, "loss": 1.0212, "step": 135310 }, { "epoch": 0.34, "learning_rate": 6.621294478836379e-05, "loss": 1.0215, "step": 135315 }, { "epoch": 0.34, "learning_rate": 6.621168654688208e-05, "loss": 1.0198, "step": 135320 }, { "epoch": 0.34, "learning_rate": 6.621042830540037e-05, "loss": 1.0222, "step": 135325 }, { "epoch": 0.34, "learning_rate": 6.620917006391867e-05, "loss": 1.0237, "step": 135330 }, { "epoch": 0.34, "learning_rate": 6.620791182243696e-05, "loss": 1.0196, "step": 135335 }, { "epoch": 0.34, "learning_rate": 6.620665358095526e-05, "loss": 1.0199, "step": 135340 }, { "epoch": 0.34, "learning_rate": 6.620539533947355e-05, "loss": 1.022, "step": 135345 }, { "epoch": 0.34, "learning_rate": 6.620413709799185e-05, "loss": 1.021, "step": 135350 }, { "epoch": 0.34, "learning_rate": 6.620287885651014e-05, "loss": 1.0209, "step": 135355 }, { "epoch": 0.34, "learning_rate": 6.620162061502844e-05, "loss": 1.0205, "step": 135360 }, { "epoch": 0.34, "learning_rate": 6.620036237354673e-05, "loss": 1.0192, "step": 135365 }, { "epoch": 0.34, "learning_rate": 6.619910413206503e-05, "loss": 1.0196, "step": 135370 }, { "epoch": 0.34, "learning_rate": 6.619784589058332e-05, "loss": 1.0195, "step": 135375 }, { "epoch": 0.34, "learning_rate": 6.619658764910162e-05, "loss": 1.022, "step": 135380 }, { "epoch": 0.34, "learning_rate": 6.619532940761991e-05, "loss": 1.0208, "step": 135385 }, { "epoch": 0.34, "learning_rate": 6.61940711661382e-05, "loss": 1.0212, "step": 135390 }, { "epoch": 0.34, "learning_rate": 6.61928129246565e-05, "loss": 1.0415, "step": 135395 }, { "epoch": 0.34, "learning_rate": 6.619155468317479e-05, "loss": 1.0213, "step": 135400 }, { "epoch": 0.34, "learning_rate": 6.61902964416931e-05, "loss": 1.0216, "step": 135405 }, { "epoch": 0.34, "learning_rate": 6.618903820021139e-05, "loss": 1.0198, "step": 135410 }, { "epoch": 0.34, "learning_rate": 6.618777995872968e-05, "loss": 1.0214, "step": 135415 }, { "epoch": 0.34, "learning_rate": 6.618652171724797e-05, "loss": 1.0212, "step": 135420 }, { "epoch": 0.34, "learning_rate": 6.618526347576627e-05, "loss": 1.0226, "step": 135425 }, { "epoch": 0.34, "learning_rate": 6.618400523428457e-05, "loss": 1.019, "step": 135430 }, { "epoch": 0.34, "learning_rate": 6.618274699280286e-05, "loss": 1.0181, "step": 135435 }, { "epoch": 0.34, "learning_rate": 6.618148875132115e-05, "loss": 1.0193, "step": 135440 }, { "epoch": 0.34, "learning_rate": 6.618023050983945e-05, "loss": 1.0227, "step": 135445 }, { "epoch": 0.34, "learning_rate": 6.617897226835775e-05, "loss": 1.0195, "step": 135450 }, { "epoch": 0.34, "learning_rate": 6.617771402687604e-05, "loss": 1.0206, "step": 135455 }, { "epoch": 0.34, "learning_rate": 6.617645578539433e-05, "loss": 1.0232, "step": 135460 }, { "epoch": 0.34, "learning_rate": 6.617519754391262e-05, "loss": 1.019, "step": 135465 }, { "epoch": 0.34, "learning_rate": 6.617393930243093e-05, "loss": 1.019, "step": 135470 }, { "epoch": 0.34, "learning_rate": 6.617268106094922e-05, "loss": 1.0223, "step": 135475 }, { "epoch": 0.34, "learning_rate": 6.617142281946751e-05, "loss": 1.0216, "step": 135480 }, { "epoch": 0.34, "learning_rate": 6.61701645779858e-05, "loss": 1.0212, "step": 135485 }, { "epoch": 0.34, "learning_rate": 6.61689063365041e-05, "loss": 1.0215, "step": 135490 }, { "epoch": 0.34, "learning_rate": 6.616764809502241e-05, "loss": 1.0205, "step": 135495 }, { "epoch": 0.34, "learning_rate": 6.61663898535407e-05, "loss": 1.0204, "step": 135500 }, { "epoch": 0.34, "learning_rate": 6.6165131612059e-05, "loss": 1.0207, "step": 135505 }, { "epoch": 0.34, "learning_rate": 6.616387337057729e-05, "loss": 1.0223, "step": 135510 }, { "epoch": 0.34, "learning_rate": 6.616261512909559e-05, "loss": 1.0192, "step": 135515 }, { "epoch": 0.34, "learning_rate": 6.616135688761388e-05, "loss": 1.0197, "step": 135520 }, { "epoch": 0.34, "learning_rate": 6.616009864613217e-05, "loss": 1.0212, "step": 135525 }, { "epoch": 0.34, "learning_rate": 6.615884040465047e-05, "loss": 1.0198, "step": 135530 }, { "epoch": 0.34, "learning_rate": 6.615758216316876e-05, "loss": 1.0237, "step": 135535 }, { "epoch": 0.34, "learning_rate": 6.615632392168706e-05, "loss": 1.0212, "step": 135540 }, { "epoch": 0.34, "learning_rate": 6.615506568020535e-05, "loss": 1.0226, "step": 135545 }, { "epoch": 0.34, "learning_rate": 6.615380743872364e-05, "loss": 1.0198, "step": 135550 }, { "epoch": 0.34, "learning_rate": 6.615254919724194e-05, "loss": 1.0196, "step": 135555 }, { "epoch": 0.34, "learning_rate": 6.615129095576024e-05, "loss": 1.0215, "step": 135560 }, { "epoch": 0.34, "learning_rate": 6.615003271427853e-05, "loss": 1.0211, "step": 135565 }, { "epoch": 0.34, "learning_rate": 6.614877447279682e-05, "loss": 1.0227, "step": 135570 }, { "epoch": 0.34, "learning_rate": 6.614751623131512e-05, "loss": 1.023, "step": 135575 }, { "epoch": 0.34, "learning_rate": 6.614625798983342e-05, "loss": 1.0208, "step": 135580 }, { "epoch": 0.34, "learning_rate": 6.614499974835171e-05, "loss": 1.0196, "step": 135585 }, { "epoch": 0.34, "learning_rate": 6.614374150687e-05, "loss": 1.021, "step": 135590 }, { "epoch": 0.34, "learning_rate": 6.61424832653883e-05, "loss": 1.0217, "step": 135595 }, { "epoch": 0.34, "learning_rate": 6.614122502390659e-05, "loss": 1.0228, "step": 135600 }, { "epoch": 0.34, "learning_rate": 6.61399667824249e-05, "loss": 1.0195, "step": 135605 }, { "epoch": 0.34, "learning_rate": 6.613870854094318e-05, "loss": 1.0205, "step": 135610 }, { "epoch": 0.34, "learning_rate": 6.613745029946148e-05, "loss": 1.0229, "step": 135615 }, { "epoch": 0.34, "learning_rate": 6.613619205797977e-05, "loss": 1.0207, "step": 135620 }, { "epoch": 0.34, "learning_rate": 6.613493381649807e-05, "loss": 1.0223, "step": 135625 }, { "epoch": 0.34, "learning_rate": 6.613367557501636e-05, "loss": 1.0186, "step": 135630 }, { "epoch": 0.34, "learning_rate": 6.613241733353466e-05, "loss": 1.0221, "step": 135635 }, { "epoch": 0.34, "learning_rate": 6.613115909205295e-05, "loss": 1.0206, "step": 135640 }, { "epoch": 0.34, "learning_rate": 6.612990085057125e-05, "loss": 1.02, "step": 135645 }, { "epoch": 0.34, "learning_rate": 6.612864260908954e-05, "loss": 1.0221, "step": 135650 }, { "epoch": 0.34, "learning_rate": 6.612738436760784e-05, "loss": 1.0411, "step": 135655 }, { "epoch": 0.34, "learning_rate": 6.612612612612613e-05, "loss": 1.0343, "step": 135660 }, { "epoch": 0.34, "learning_rate": 6.612486788464442e-05, "loss": 1.0214, "step": 135665 }, { "epoch": 0.34, "learning_rate": 6.612360964316272e-05, "loss": 1.018, "step": 135670 }, { "epoch": 0.34, "learning_rate": 6.612235140168102e-05, "loss": 1.0226, "step": 135675 }, { "epoch": 0.34, "learning_rate": 6.612109316019931e-05, "loss": 1.0221, "step": 135680 }, { "epoch": 0.34, "learning_rate": 6.61198349187176e-05, "loss": 1.0232, "step": 135685 }, { "epoch": 0.34, "learning_rate": 6.61185766772359e-05, "loss": 1.0211, "step": 135690 }, { "epoch": 0.34, "learning_rate": 6.61173184357542e-05, "loss": 1.0222, "step": 135695 }, { "epoch": 0.34, "learning_rate": 6.611606019427249e-05, "loss": 1.0211, "step": 135700 }, { "epoch": 0.34, "learning_rate": 6.611480195279078e-05, "loss": 1.021, "step": 135705 }, { "epoch": 0.34, "learning_rate": 6.611354371130908e-05, "loss": 1.0191, "step": 135710 }, { "epoch": 0.34, "learning_rate": 6.611228546982738e-05, "loss": 1.0226, "step": 135715 }, { "epoch": 0.34, "learning_rate": 6.611102722834567e-05, "loss": 1.0202, "step": 135720 }, { "epoch": 0.34, "learning_rate": 6.610976898686396e-05, "loss": 1.0218, "step": 135725 }, { "epoch": 0.34, "learning_rate": 6.610851074538225e-05, "loss": 1.0184, "step": 135730 }, { "epoch": 0.34, "learning_rate": 6.610725250390056e-05, "loss": 1.0213, "step": 135735 }, { "epoch": 0.34, "learning_rate": 6.610599426241885e-05, "loss": 1.0198, "step": 135740 }, { "epoch": 0.34, "learning_rate": 6.610473602093714e-05, "loss": 1.0218, "step": 135745 }, { "epoch": 0.34, "learning_rate": 6.610347777945543e-05, "loss": 1.0208, "step": 135750 }, { "epoch": 0.34, "learning_rate": 6.610221953797374e-05, "loss": 1.0219, "step": 135755 }, { "epoch": 0.34, "learning_rate": 6.610096129649203e-05, "loss": 1.0237, "step": 135760 }, { "epoch": 0.34, "learning_rate": 6.609970305501032e-05, "loss": 1.0218, "step": 135765 }, { "epoch": 0.34, "learning_rate": 6.609844481352861e-05, "loss": 1.0201, "step": 135770 }, { "epoch": 0.34, "learning_rate": 6.609718657204692e-05, "loss": 1.0211, "step": 135775 }, { "epoch": 0.34, "learning_rate": 6.609592833056521e-05, "loss": 1.0213, "step": 135780 }, { "epoch": 0.34, "learning_rate": 6.60946700890835e-05, "loss": 1.0227, "step": 135785 }, { "epoch": 0.34, "learning_rate": 6.609341184760179e-05, "loss": 1.0231, "step": 135790 }, { "epoch": 0.34, "learning_rate": 6.609215360612008e-05, "loss": 1.0226, "step": 135795 }, { "epoch": 0.34, "learning_rate": 6.609089536463839e-05, "loss": 1.0206, "step": 135800 }, { "epoch": 0.34, "learning_rate": 6.608963712315668e-05, "loss": 1.0233, "step": 135805 }, { "epoch": 0.34, "learning_rate": 6.608837888167497e-05, "loss": 1.0216, "step": 135810 }, { "epoch": 0.34, "learning_rate": 6.608712064019326e-05, "loss": 1.0217, "step": 135815 }, { "epoch": 0.34, "learning_rate": 6.608586239871157e-05, "loss": 1.02, "step": 135820 }, { "epoch": 0.34, "learning_rate": 6.608460415722986e-05, "loss": 1.021, "step": 135825 }, { "epoch": 0.34, "learning_rate": 6.608334591574815e-05, "loss": 1.0229, "step": 135830 }, { "epoch": 0.34, "learning_rate": 6.608208767426644e-05, "loss": 1.0204, "step": 135835 }, { "epoch": 0.34, "learning_rate": 6.608082943278475e-05, "loss": 1.0219, "step": 135840 }, { "epoch": 0.34, "learning_rate": 6.607957119130304e-05, "loss": 1.0244, "step": 135845 }, { "epoch": 0.34, "learning_rate": 6.607831294982133e-05, "loss": 1.0201, "step": 135850 }, { "epoch": 0.34, "learning_rate": 6.607705470833962e-05, "loss": 1.0231, "step": 135855 }, { "epoch": 0.34, "learning_rate": 6.607579646685791e-05, "loss": 1.0218, "step": 135860 }, { "epoch": 0.34, "learning_rate": 6.607453822537622e-05, "loss": 1.0215, "step": 135865 }, { "epoch": 0.34, "learning_rate": 6.607327998389451e-05, "loss": 1.0211, "step": 135870 }, { "epoch": 0.34, "learning_rate": 6.60720217424128e-05, "loss": 1.0218, "step": 135875 }, { "epoch": 0.34, "learning_rate": 6.60707635009311e-05, "loss": 1.0223, "step": 135880 }, { "epoch": 0.34, "learning_rate": 6.60695052594494e-05, "loss": 1.0211, "step": 135885 }, { "epoch": 0.34, "learning_rate": 6.606824701796769e-05, "loss": 1.0219, "step": 135890 }, { "epoch": 0.34, "learning_rate": 6.606698877648598e-05, "loss": 1.0199, "step": 135895 }, { "epoch": 0.34, "learning_rate": 6.606573053500427e-05, "loss": 1.0205, "step": 135900 }, { "epoch": 0.34, "learning_rate": 6.606447229352258e-05, "loss": 1.0218, "step": 135905 }, { "epoch": 0.34, "learning_rate": 6.606321405204087e-05, "loss": 1.0228, "step": 135910 }, { "epoch": 0.34, "learning_rate": 6.606195581055916e-05, "loss": 1.0342, "step": 135915 }, { "epoch": 0.34, "learning_rate": 6.606069756907745e-05, "loss": 1.0216, "step": 135920 }, { "epoch": 0.34, "learning_rate": 6.605943932759575e-05, "loss": 1.0209, "step": 135925 }, { "epoch": 0.34, "learning_rate": 6.605818108611405e-05, "loss": 1.0201, "step": 135930 }, { "epoch": 0.34, "learning_rate": 6.605692284463234e-05, "loss": 1.0241, "step": 135935 }, { "epoch": 0.34, "learning_rate": 6.605566460315063e-05, "loss": 1.021, "step": 135940 }, { "epoch": 0.34, "learning_rate": 6.605440636166893e-05, "loss": 1.0227, "step": 135945 }, { "epoch": 0.34, "learning_rate": 6.605314812018723e-05, "loss": 1.022, "step": 135950 }, { "epoch": 0.34, "learning_rate": 6.605188987870552e-05, "loss": 1.0222, "step": 135955 }, { "epoch": 0.34, "learning_rate": 6.605063163722381e-05, "loss": 1.0214, "step": 135960 }, { "epoch": 0.34, "learning_rate": 6.60493733957421e-05, "loss": 1.0238, "step": 135965 }, { "epoch": 0.34, "learning_rate": 6.604811515426041e-05, "loss": 1.0234, "step": 135970 }, { "epoch": 0.34, "learning_rate": 6.60468569127787e-05, "loss": 1.0207, "step": 135975 }, { "epoch": 0.34, "learning_rate": 6.6045598671297e-05, "loss": 1.0223, "step": 135980 }, { "epoch": 0.34, "learning_rate": 6.604434042981529e-05, "loss": 1.0238, "step": 135985 }, { "epoch": 0.34, "learning_rate": 6.604308218833358e-05, "loss": 1.0218, "step": 135990 }, { "epoch": 0.34, "learning_rate": 6.604182394685188e-05, "loss": 1.0207, "step": 135995 }, { "epoch": 0.34, "learning_rate": 6.604056570537019e-05, "loss": 1.0227, "step": 136000 }, { "epoch": 0.34, "learning_rate": 6.603930746388848e-05, "loss": 1.041, "step": 136005 }, { "epoch": 0.34, "learning_rate": 6.603804922240677e-05, "loss": 1.0457, "step": 136010 }, { "epoch": 0.34, "learning_rate": 6.603679098092506e-05, "loss": 1.0223, "step": 136015 }, { "epoch": 0.34, "learning_rate": 6.603553273944337e-05, "loss": 1.0216, "step": 136020 }, { "epoch": 0.34, "learning_rate": 6.603427449796166e-05, "loss": 1.0207, "step": 136025 }, { "epoch": 0.34, "learning_rate": 6.603301625647995e-05, "loss": 1.0218, "step": 136030 }, { "epoch": 0.34, "learning_rate": 6.603175801499824e-05, "loss": 1.024, "step": 136035 }, { "epoch": 0.34, "learning_rate": 6.603049977351655e-05, "loss": 1.0204, "step": 136040 }, { "epoch": 0.34, "learning_rate": 6.602924153203484e-05, "loss": 1.0214, "step": 136045 }, { "epoch": 0.34, "learning_rate": 6.602798329055313e-05, "loss": 1.0213, "step": 136050 }, { "epoch": 0.34, "learning_rate": 6.602672504907142e-05, "loss": 1.0199, "step": 136055 }, { "epoch": 0.34, "learning_rate": 6.602546680758971e-05, "loss": 1.0219, "step": 136060 }, { "epoch": 0.34, "learning_rate": 6.602420856610802e-05, "loss": 1.0201, "step": 136065 }, { "epoch": 0.34, "learning_rate": 6.602295032462631e-05, "loss": 1.022, "step": 136070 }, { "epoch": 0.34, "learning_rate": 6.60216920831446e-05, "loss": 1.02, "step": 136075 }, { "epoch": 0.34, "learning_rate": 6.602043384166289e-05, "loss": 1.0208, "step": 136080 }, { "epoch": 0.34, "learning_rate": 6.60191756001812e-05, "loss": 1.0212, "step": 136085 }, { "epoch": 0.34, "learning_rate": 6.601791735869949e-05, "loss": 1.0218, "step": 136090 }, { "epoch": 0.34, "learning_rate": 6.601665911721778e-05, "loss": 1.0214, "step": 136095 }, { "epoch": 0.34, "learning_rate": 6.601540087573607e-05, "loss": 1.0216, "step": 136100 }, { "epoch": 0.34, "learning_rate": 6.601414263425438e-05, "loss": 1.0186, "step": 136105 }, { "epoch": 0.34, "learning_rate": 6.601288439277267e-05, "loss": 1.0212, "step": 136110 }, { "epoch": 0.34, "learning_rate": 6.601162615129096e-05, "loss": 1.0212, "step": 136115 }, { "epoch": 0.34, "learning_rate": 6.601036790980925e-05, "loss": 1.0218, "step": 136120 }, { "epoch": 0.34, "learning_rate": 6.600910966832754e-05, "loss": 1.0234, "step": 136125 }, { "epoch": 0.34, "learning_rate": 6.600785142684585e-05, "loss": 1.0206, "step": 136130 }, { "epoch": 0.34, "learning_rate": 6.600659318536414e-05, "loss": 1.0423, "step": 136135 }, { "epoch": 0.34, "learning_rate": 6.600533494388243e-05, "loss": 1.0213, "step": 136140 }, { "epoch": 0.34, "learning_rate": 6.600407670240072e-05, "loss": 1.0207, "step": 136145 }, { "epoch": 0.34, "learning_rate": 6.600281846091903e-05, "loss": 1.0207, "step": 136150 }, { "epoch": 0.34, "learning_rate": 6.600156021943732e-05, "loss": 1.0226, "step": 136155 }, { "epoch": 0.34, "learning_rate": 6.600030197795561e-05, "loss": 1.0184, "step": 136160 }, { "epoch": 0.34, "learning_rate": 6.59990437364739e-05, "loss": 1.0224, "step": 136165 }, { "epoch": 0.34, "learning_rate": 6.599778549499221e-05, "loss": 1.016, "step": 136170 }, { "epoch": 0.34, "learning_rate": 6.59965272535105e-05, "loss": 1.0201, "step": 136175 }, { "epoch": 0.34, "learning_rate": 6.599526901202879e-05, "loss": 1.0222, "step": 136180 }, { "epoch": 0.34, "learning_rate": 6.599401077054708e-05, "loss": 1.0226, "step": 136185 }, { "epoch": 0.34, "learning_rate": 6.599275252906538e-05, "loss": 1.0244, "step": 136190 }, { "epoch": 0.34, "learning_rate": 6.599149428758368e-05, "loss": 1.0234, "step": 136195 }, { "epoch": 0.34, "learning_rate": 6.599023604610197e-05, "loss": 1.0212, "step": 136200 }, { "epoch": 0.34, "learning_rate": 6.598897780462026e-05, "loss": 1.0209, "step": 136205 }, { "epoch": 0.34, "learning_rate": 6.598771956313856e-05, "loss": 1.0212, "step": 136210 }, { "epoch": 0.34, "learning_rate": 6.598646132165686e-05, "loss": 1.0189, "step": 136215 }, { "epoch": 0.34, "learning_rate": 6.598520308017515e-05, "loss": 1.0223, "step": 136220 }, { "epoch": 0.34, "learning_rate": 6.598394483869344e-05, "loss": 1.0189, "step": 136225 }, { "epoch": 0.34, "learning_rate": 6.598268659721174e-05, "loss": 1.0199, "step": 136230 }, { "epoch": 0.34, "learning_rate": 6.598142835573004e-05, "loss": 1.0237, "step": 136235 }, { "epoch": 0.34, "learning_rate": 6.598017011424833e-05, "loss": 1.0206, "step": 136240 }, { "epoch": 0.34, "learning_rate": 6.597891187276662e-05, "loss": 1.0208, "step": 136245 }, { "epoch": 0.34, "learning_rate": 6.597765363128492e-05, "loss": 1.0217, "step": 136250 }, { "epoch": 0.34, "learning_rate": 6.597639538980321e-05, "loss": 1.0392, "step": 136255 }, { "epoch": 0.34, "learning_rate": 6.597513714832151e-05, "loss": 1.0182, "step": 136260 }, { "epoch": 0.34, "learning_rate": 6.59738789068398e-05, "loss": 1.0436, "step": 136265 }, { "epoch": 0.34, "learning_rate": 6.59726206653581e-05, "loss": 1.0204, "step": 136270 }, { "epoch": 0.34, "learning_rate": 6.597136242387639e-05, "loss": 1.0208, "step": 136275 }, { "epoch": 0.34, "learning_rate": 6.597010418239469e-05, "loss": 1.0238, "step": 136280 }, { "epoch": 0.34, "learning_rate": 6.596884594091298e-05, "loss": 1.0203, "step": 136285 }, { "epoch": 0.34, "learning_rate": 6.596758769943128e-05, "loss": 1.0195, "step": 136290 }, { "epoch": 0.34, "learning_rate": 6.596632945794957e-05, "loss": 1.022, "step": 136295 }, { "epoch": 0.34, "learning_rate": 6.596507121646787e-05, "loss": 1.0232, "step": 136300 }, { "epoch": 0.34, "learning_rate": 6.596381297498616e-05, "loss": 1.0231, "step": 136305 }, { "epoch": 0.34, "learning_rate": 6.596255473350446e-05, "loss": 1.0208, "step": 136310 }, { "epoch": 0.34, "learning_rate": 6.596129649202275e-05, "loss": 1.021, "step": 136315 }, { "epoch": 0.34, "learning_rate": 6.596003825054104e-05, "loss": 1.0211, "step": 136320 }, { "epoch": 0.34, "learning_rate": 6.595878000905934e-05, "loss": 1.022, "step": 136325 }, { "epoch": 0.34, "learning_rate": 6.595752176757764e-05, "loss": 1.0217, "step": 136330 }, { "epoch": 0.34, "learning_rate": 6.595626352609593e-05, "loss": 1.023, "step": 136335 }, { "epoch": 0.34, "learning_rate": 6.595500528461422e-05, "loss": 1.0205, "step": 136340 }, { "epoch": 0.34, "learning_rate": 6.595374704313252e-05, "loss": 1.0226, "step": 136345 }, { "epoch": 0.34, "learning_rate": 6.595248880165082e-05, "loss": 1.0227, "step": 136350 }, { "epoch": 0.34, "learning_rate": 6.595123056016911e-05, "loss": 1.0217, "step": 136355 }, { "epoch": 0.34, "learning_rate": 6.59499723186874e-05, "loss": 1.0206, "step": 136360 }, { "epoch": 0.34, "learning_rate": 6.59487140772057e-05, "loss": 1.0206, "step": 136365 }, { "epoch": 0.34, "learning_rate": 6.5947455835724e-05, "loss": 1.0234, "step": 136370 }, { "epoch": 0.34, "learning_rate": 6.594619759424229e-05, "loss": 1.0205, "step": 136375 }, { "epoch": 0.34, "learning_rate": 6.594493935276058e-05, "loss": 1.0214, "step": 136380 }, { "epoch": 0.34, "learning_rate": 6.594368111127887e-05, "loss": 1.0212, "step": 136385 }, { "epoch": 0.34, "learning_rate": 6.594242286979718e-05, "loss": 1.023, "step": 136390 }, { "epoch": 0.34, "learning_rate": 6.594116462831547e-05, "loss": 1.0212, "step": 136395 }, { "epoch": 0.34, "learning_rate": 6.593990638683376e-05, "loss": 1.0214, "step": 136400 }, { "epoch": 0.34, "learning_rate": 6.593864814535205e-05, "loss": 1.0193, "step": 136405 }, { "epoch": 0.34, "learning_rate": 6.593738990387036e-05, "loss": 1.0203, "step": 136410 }, { "epoch": 0.34, "learning_rate": 6.593613166238865e-05, "loss": 1.0206, "step": 136415 }, { "epoch": 0.34, "learning_rate": 6.593487342090694e-05, "loss": 1.0228, "step": 136420 }, { "epoch": 0.34, "learning_rate": 6.593361517942523e-05, "loss": 1.0197, "step": 136425 }, { "epoch": 0.34, "learning_rate": 6.593235693794354e-05, "loss": 1.0197, "step": 136430 }, { "epoch": 0.34, "learning_rate": 6.593109869646183e-05, "loss": 1.0441, "step": 136435 }, { "epoch": 0.34, "learning_rate": 6.592984045498012e-05, "loss": 1.0223, "step": 136440 }, { "epoch": 0.34, "learning_rate": 6.592858221349841e-05, "loss": 1.0211, "step": 136445 }, { "epoch": 0.34, "learning_rate": 6.59273239720167e-05, "loss": 1.0211, "step": 136450 }, { "epoch": 0.34, "learning_rate": 6.592606573053501e-05, "loss": 1.0219, "step": 136455 }, { "epoch": 0.34, "learning_rate": 6.59248074890533e-05, "loss": 1.023, "step": 136460 }, { "epoch": 0.34, "learning_rate": 6.592354924757159e-05, "loss": 1.0189, "step": 136465 }, { "epoch": 0.34, "learning_rate": 6.592229100608988e-05, "loss": 1.0422, "step": 136470 }, { "epoch": 0.34, "learning_rate": 6.592103276460819e-05, "loss": 1.0211, "step": 136475 }, { "epoch": 0.34, "learning_rate": 6.591977452312648e-05, "loss": 1.0217, "step": 136480 }, { "epoch": 0.34, "learning_rate": 6.591851628164477e-05, "loss": 1.0206, "step": 136485 }, { "epoch": 0.34, "learning_rate": 6.591725804016306e-05, "loss": 1.0214, "step": 136490 }, { "epoch": 0.34, "learning_rate": 6.591599979868137e-05, "loss": 1.0209, "step": 136495 }, { "epoch": 0.34, "learning_rate": 6.591474155719967e-05, "loss": 1.0202, "step": 136500 }, { "epoch": 0.34, "learning_rate": 6.591348331571796e-05, "loss": 1.0218, "step": 136505 }, { "epoch": 0.34, "learning_rate": 6.591222507423626e-05, "loss": 1.0191, "step": 136510 }, { "epoch": 0.34, "learning_rate": 6.591096683275455e-05, "loss": 1.0221, "step": 136515 }, { "epoch": 0.34, "learning_rate": 6.590970859127285e-05, "loss": 1.0197, "step": 136520 }, { "epoch": 0.34, "learning_rate": 6.590845034979114e-05, "loss": 1.0199, "step": 136525 }, { "epoch": 0.34, "learning_rate": 6.590719210830944e-05, "loss": 1.0223, "step": 136530 }, { "epoch": 0.34, "learning_rate": 6.590593386682773e-05, "loss": 1.0209, "step": 136535 }, { "epoch": 0.34, "learning_rate": 6.590467562534602e-05, "loss": 1.0216, "step": 136540 }, { "epoch": 0.34, "learning_rate": 6.590341738386432e-05, "loss": 1.0217, "step": 136545 }, { "epoch": 0.34, "learning_rate": 6.590215914238262e-05, "loss": 1.0217, "step": 136550 }, { "epoch": 0.34, "learning_rate": 6.59009009009009e-05, "loss": 1.0228, "step": 136555 }, { "epoch": 0.34, "learning_rate": 6.58996426594192e-05, "loss": 1.0194, "step": 136560 }, { "epoch": 0.34, "learning_rate": 6.58983844179375e-05, "loss": 1.0218, "step": 136565 }, { "epoch": 0.34, "learning_rate": 6.58971261764558e-05, "loss": 1.0185, "step": 136570 }, { "epoch": 0.34, "learning_rate": 6.589586793497409e-05, "loss": 1.0211, "step": 136575 }, { "epoch": 0.34, "learning_rate": 6.589460969349238e-05, "loss": 1.0216, "step": 136580 }, { "epoch": 0.34, "learning_rate": 6.589335145201067e-05, "loss": 1.0211, "step": 136585 }, { "epoch": 0.34, "learning_rate": 6.589209321052897e-05, "loss": 1.0224, "step": 136590 }, { "epoch": 0.34, "learning_rate": 6.589083496904727e-05, "loss": 1.0212, "step": 136595 }, { "epoch": 0.34, "learning_rate": 6.588957672756556e-05, "loss": 1.0201, "step": 136600 }, { "epoch": 0.34, "learning_rate": 6.588831848608385e-05, "loss": 1.0212, "step": 136605 }, { "epoch": 0.34, "learning_rate": 6.588706024460215e-05, "loss": 1.0221, "step": 136610 }, { "epoch": 0.34, "learning_rate": 6.588580200312045e-05, "loss": 1.022, "step": 136615 }, { "epoch": 0.34, "learning_rate": 6.588454376163874e-05, "loss": 1.0221, "step": 136620 }, { "epoch": 0.34, "learning_rate": 6.588328552015703e-05, "loss": 1.0195, "step": 136625 }, { "epoch": 0.34, "learning_rate": 6.588202727867533e-05, "loss": 1.0188, "step": 136630 }, { "epoch": 0.34, "learning_rate": 6.588076903719363e-05, "loss": 1.0233, "step": 136635 }, { "epoch": 0.34, "learning_rate": 6.587951079571192e-05, "loss": 1.0189, "step": 136640 }, { "epoch": 0.34, "learning_rate": 6.587825255423021e-05, "loss": 1.0207, "step": 136645 }, { "epoch": 0.34, "learning_rate": 6.58769943127485e-05, "loss": 1.0231, "step": 136650 }, { "epoch": 0.34, "learning_rate": 6.58757360712668e-05, "loss": 1.0219, "step": 136655 }, { "epoch": 0.34, "learning_rate": 6.58744778297851e-05, "loss": 1.0213, "step": 136660 }, { "epoch": 0.34, "learning_rate": 6.587321958830339e-05, "loss": 1.0179, "step": 136665 }, { "epoch": 0.34, "learning_rate": 6.587196134682168e-05, "loss": 1.0224, "step": 136670 }, { "epoch": 0.34, "learning_rate": 6.587070310533999e-05, "loss": 1.0217, "step": 136675 }, { "epoch": 0.34, "learning_rate": 6.586944486385828e-05, "loss": 1.0237, "step": 136680 }, { "epoch": 0.34, "learning_rate": 6.586818662237657e-05, "loss": 1.0205, "step": 136685 }, { "epoch": 0.34, "learning_rate": 6.586692838089486e-05, "loss": 1.0201, "step": 136690 }, { "epoch": 0.34, "learning_rate": 6.586567013941317e-05, "loss": 1.0234, "step": 136695 }, { "epoch": 0.34, "learning_rate": 6.586441189793146e-05, "loss": 1.0225, "step": 136700 }, { "epoch": 0.34, "learning_rate": 6.586315365644975e-05, "loss": 1.0218, "step": 136705 }, { "epoch": 0.34, "learning_rate": 6.586189541496804e-05, "loss": 1.0205, "step": 136710 }, { "epoch": 0.34, "learning_rate": 6.586063717348633e-05, "loss": 1.0227, "step": 136715 }, { "epoch": 0.34, "learning_rate": 6.585937893200464e-05, "loss": 1.021, "step": 136720 }, { "epoch": 0.34, "learning_rate": 6.585812069052293e-05, "loss": 1.0188, "step": 136725 }, { "epoch": 0.34, "learning_rate": 6.585686244904122e-05, "loss": 1.024, "step": 136730 }, { "epoch": 0.34, "learning_rate": 6.585560420755951e-05, "loss": 1.0221, "step": 136735 }, { "epoch": 0.34, "learning_rate": 6.585434596607782e-05, "loss": 1.0183, "step": 136740 }, { "epoch": 0.34, "learning_rate": 6.585308772459611e-05, "loss": 1.0321, "step": 136745 }, { "epoch": 0.34, "learning_rate": 6.58518294831144e-05, "loss": 1.0203, "step": 136750 }, { "epoch": 0.34, "learning_rate": 6.585057124163269e-05, "loss": 1.0195, "step": 136755 }, { "epoch": 0.34, "learning_rate": 6.5849313000151e-05, "loss": 1.0209, "step": 136760 }, { "epoch": 0.34, "learning_rate": 6.584805475866929e-05, "loss": 1.0213, "step": 136765 }, { "epoch": 0.34, "learning_rate": 6.584679651718758e-05, "loss": 1.0216, "step": 136770 }, { "epoch": 0.34, "learning_rate": 6.584553827570587e-05, "loss": 1.0223, "step": 136775 }, { "epoch": 0.34, "learning_rate": 6.584428003422416e-05, "loss": 1.0242, "step": 136780 }, { "epoch": 0.34, "learning_rate": 6.584302179274247e-05, "loss": 1.0215, "step": 136785 }, { "epoch": 0.34, "learning_rate": 6.584176355126076e-05, "loss": 1.0215, "step": 136790 }, { "epoch": 0.34, "learning_rate": 6.584050530977905e-05, "loss": 1.0205, "step": 136795 }, { "epoch": 0.34, "learning_rate": 6.583924706829734e-05, "loss": 1.0207, "step": 136800 }, { "epoch": 0.34, "learning_rate": 6.583798882681565e-05, "loss": 1.0196, "step": 136805 }, { "epoch": 0.34, "learning_rate": 6.583673058533394e-05, "loss": 1.0218, "step": 136810 }, { "epoch": 0.34, "learning_rate": 6.583547234385223e-05, "loss": 1.0202, "step": 136815 }, { "epoch": 0.34, "learning_rate": 6.583421410237052e-05, "loss": 1.0214, "step": 136820 }, { "epoch": 0.34, "learning_rate": 6.583295586088883e-05, "loss": 1.0176, "step": 136825 }, { "epoch": 0.34, "learning_rate": 6.583169761940712e-05, "loss": 1.0221, "step": 136830 }, { "epoch": 0.34, "learning_rate": 6.583043937792541e-05, "loss": 1.0214, "step": 136835 }, { "epoch": 0.34, "learning_rate": 6.58291811364437e-05, "loss": 1.0225, "step": 136840 }, { "epoch": 0.34, "learning_rate": 6.5827922894962e-05, "loss": 1.0192, "step": 136845 }, { "epoch": 0.34, "learning_rate": 6.58266646534803e-05, "loss": 1.0187, "step": 136850 }, { "epoch": 0.34, "learning_rate": 6.582540641199859e-05, "loss": 1.0218, "step": 136855 }, { "epoch": 0.34, "learning_rate": 6.582414817051688e-05, "loss": 1.0201, "step": 136860 }, { "epoch": 0.34, "learning_rate": 6.582288992903518e-05, "loss": 1.0188, "step": 136865 }, { "epoch": 0.34, "learning_rate": 6.582163168755348e-05, "loss": 1.0225, "step": 136870 }, { "epoch": 0.34, "learning_rate": 6.582037344607177e-05, "loss": 1.0227, "step": 136875 }, { "epoch": 0.34, "learning_rate": 6.581911520459006e-05, "loss": 1.0207, "step": 136880 }, { "epoch": 0.34, "learning_rate": 6.581785696310836e-05, "loss": 1.0203, "step": 136885 }, { "epoch": 0.34, "learning_rate": 6.581659872162666e-05, "loss": 1.0216, "step": 136890 }, { "epoch": 0.34, "learning_rate": 6.581534048014495e-05, "loss": 1.0206, "step": 136895 }, { "epoch": 0.34, "learning_rate": 6.581408223866324e-05, "loss": 1.0215, "step": 136900 }, { "epoch": 0.34, "learning_rate": 6.581282399718154e-05, "loss": 1.0198, "step": 136905 }, { "epoch": 0.34, "learning_rate": 6.581156575569983e-05, "loss": 1.0227, "step": 136910 }, { "epoch": 0.34, "learning_rate": 6.581030751421813e-05, "loss": 1.0212, "step": 136915 }, { "epoch": 0.34, "learning_rate": 6.580904927273642e-05, "loss": 1.0188, "step": 136920 }, { "epoch": 0.34, "learning_rate": 6.580779103125472e-05, "loss": 1.0186, "step": 136925 }, { "epoch": 0.34, "learning_rate": 6.580653278977301e-05, "loss": 1.0213, "step": 136930 }, { "epoch": 0.34, "learning_rate": 6.580527454829131e-05, "loss": 1.0232, "step": 136935 }, { "epoch": 0.34, "learning_rate": 6.58040163068096e-05, "loss": 1.0213, "step": 136940 }, { "epoch": 0.34, "learning_rate": 6.58027580653279e-05, "loss": 1.0229, "step": 136945 }, { "epoch": 0.34, "learning_rate": 6.580149982384619e-05, "loss": 1.0207, "step": 136950 }, { "epoch": 0.34, "learning_rate": 6.580024158236449e-05, "loss": 1.0234, "step": 136955 }, { "epoch": 0.34, "learning_rate": 6.579898334088278e-05, "loss": 1.0218, "step": 136960 }, { "epoch": 0.34, "learning_rate": 6.579772509940108e-05, "loss": 1.0217, "step": 136965 }, { "epoch": 0.34, "learning_rate": 6.579646685791937e-05, "loss": 1.0225, "step": 136970 }, { "epoch": 0.34, "learning_rate": 6.579520861643766e-05, "loss": 1.0211, "step": 136975 }, { "epoch": 0.34, "learning_rate": 6.579395037495596e-05, "loss": 1.0223, "step": 136980 }, { "epoch": 0.34, "learning_rate": 6.579269213347426e-05, "loss": 1.0203, "step": 136985 }, { "epoch": 0.34, "learning_rate": 6.579143389199256e-05, "loss": 1.0209, "step": 136990 }, { "epoch": 0.34, "learning_rate": 6.579017565051085e-05, "loss": 1.0232, "step": 136995 }, { "epoch": 0.34, "learning_rate": 6.578891740902914e-05, "loss": 1.0227, "step": 137000 }, { "epoch": 0.34, "learning_rate": 6.578765916754745e-05, "loss": 1.0228, "step": 137005 }, { "epoch": 0.34, "learning_rate": 6.578640092606574e-05, "loss": 1.0197, "step": 137010 }, { "epoch": 0.34, "learning_rate": 6.578514268458403e-05, "loss": 1.0204, "step": 137015 }, { "epoch": 0.34, "learning_rate": 6.578388444310232e-05, "loss": 1.022, "step": 137020 }, { "epoch": 0.34, "learning_rate": 6.578262620162063e-05, "loss": 1.0209, "step": 137025 }, { "epoch": 0.34, "learning_rate": 6.578136796013892e-05, "loss": 1.0209, "step": 137030 }, { "epoch": 0.34, "learning_rate": 6.578010971865721e-05, "loss": 1.0225, "step": 137035 }, { "epoch": 0.34, "learning_rate": 6.57788514771755e-05, "loss": 1.0227, "step": 137040 }, { "epoch": 0.34, "learning_rate": 6.577759323569381e-05, "loss": 1.0223, "step": 137045 }, { "epoch": 0.34, "learning_rate": 6.57763349942121e-05, "loss": 1.0218, "step": 137050 }, { "epoch": 0.34, "learning_rate": 6.577507675273039e-05, "loss": 1.0228, "step": 137055 }, { "epoch": 0.34, "learning_rate": 6.577381851124868e-05, "loss": 1.0209, "step": 137060 }, { "epoch": 0.34, "learning_rate": 6.577256026976697e-05, "loss": 1.0206, "step": 137065 }, { "epoch": 0.34, "learning_rate": 6.577130202828528e-05, "loss": 1.0234, "step": 137070 }, { "epoch": 0.34, "learning_rate": 6.577004378680357e-05, "loss": 1.0198, "step": 137075 }, { "epoch": 0.34, "learning_rate": 6.576878554532186e-05, "loss": 1.0189, "step": 137080 }, { "epoch": 0.34, "learning_rate": 6.576752730384015e-05, "loss": 1.0213, "step": 137085 }, { "epoch": 0.34, "learning_rate": 6.576626906235846e-05, "loss": 1.0217, "step": 137090 }, { "epoch": 0.34, "learning_rate": 6.576501082087675e-05, "loss": 1.0405, "step": 137095 }, { "epoch": 0.34, "learning_rate": 6.576375257939504e-05, "loss": 1.0226, "step": 137100 }, { "epoch": 0.34, "learning_rate": 6.576249433791333e-05, "loss": 1.0224, "step": 137105 }, { "epoch": 0.34, "learning_rate": 6.576123609643164e-05, "loss": 1.0207, "step": 137110 }, { "epoch": 0.34, "learning_rate": 6.575997785494993e-05, "loss": 1.0224, "step": 137115 }, { "epoch": 0.34, "learning_rate": 6.575871961346822e-05, "loss": 1.0191, "step": 137120 }, { "epoch": 0.34, "learning_rate": 6.575746137198651e-05, "loss": 1.037, "step": 137125 }, { "epoch": 0.34, "learning_rate": 6.57562031305048e-05, "loss": 1.0196, "step": 137130 }, { "epoch": 0.34, "learning_rate": 6.575494488902311e-05, "loss": 1.02, "step": 137135 }, { "epoch": 0.34, "learning_rate": 6.57536866475414e-05, "loss": 1.0209, "step": 137140 }, { "epoch": 0.34, "learning_rate": 6.57524284060597e-05, "loss": 1.0227, "step": 137145 }, { "epoch": 0.34, "learning_rate": 6.575117016457799e-05, "loss": 1.0222, "step": 137150 }, { "epoch": 0.34, "learning_rate": 6.574991192309629e-05, "loss": 1.0202, "step": 137155 }, { "epoch": 0.34, "learning_rate": 6.574865368161458e-05, "loss": 1.0186, "step": 137160 }, { "epoch": 0.34, "learning_rate": 6.574739544013287e-05, "loss": 1.0204, "step": 137165 }, { "epoch": 0.34, "learning_rate": 6.574613719865117e-05, "loss": 1.0208, "step": 137170 }, { "epoch": 0.34, "learning_rate": 6.574487895716947e-05, "loss": 1.021, "step": 137175 }, { "epoch": 0.34, "learning_rate": 6.574362071568776e-05, "loss": 1.0224, "step": 137180 }, { "epoch": 0.34, "learning_rate": 6.574236247420605e-05, "loss": 1.0201, "step": 137185 }, { "epoch": 0.34, "learning_rate": 6.574110423272435e-05, "loss": 1.0204, "step": 137190 }, { "epoch": 0.34, "learning_rate": 6.573984599124264e-05, "loss": 1.0234, "step": 137195 }, { "epoch": 0.34, "learning_rate": 6.573858774976094e-05, "loss": 1.022, "step": 137200 }, { "epoch": 0.34, "learning_rate": 6.573732950827923e-05, "loss": 1.0196, "step": 137205 }, { "epoch": 0.34, "learning_rate": 6.573607126679753e-05, "loss": 1.0223, "step": 137210 }, { "epoch": 0.34, "learning_rate": 6.573481302531582e-05, "loss": 1.0215, "step": 137215 }, { "epoch": 0.34, "learning_rate": 6.573355478383412e-05, "loss": 1.0213, "step": 137220 }, { "epoch": 0.34, "learning_rate": 6.573229654235241e-05, "loss": 1.0217, "step": 137225 }, { "epoch": 0.34, "learning_rate": 6.57310383008707e-05, "loss": 1.0201, "step": 137230 }, { "epoch": 0.34, "learning_rate": 6.5729780059389e-05, "loss": 1.0187, "step": 137235 }, { "epoch": 0.34, "learning_rate": 6.572852181790729e-05, "loss": 1.0222, "step": 137240 }, { "epoch": 0.34, "learning_rate": 6.57272635764256e-05, "loss": 1.0197, "step": 137245 }, { "epoch": 0.34, "learning_rate": 6.572600533494389e-05, "loss": 1.0205, "step": 137250 }, { "epoch": 0.34, "learning_rate": 6.572474709346218e-05, "loss": 1.0231, "step": 137255 }, { "epoch": 0.34, "learning_rate": 6.572348885198047e-05, "loss": 1.0186, "step": 137260 }, { "epoch": 0.34, "learning_rate": 6.572223061049877e-05, "loss": 1.0183, "step": 137265 }, { "epoch": 0.34, "learning_rate": 6.572097236901707e-05, "loss": 1.0222, "step": 137270 }, { "epoch": 0.34, "learning_rate": 6.571971412753536e-05, "loss": 1.0191, "step": 137275 }, { "epoch": 0.34, "learning_rate": 6.571845588605365e-05, "loss": 1.0212, "step": 137280 }, { "epoch": 0.34, "learning_rate": 6.571719764457195e-05, "loss": 1.0206, "step": 137285 }, { "epoch": 0.34, "learning_rate": 6.571593940309025e-05, "loss": 1.0219, "step": 137290 }, { "epoch": 0.34, "learning_rate": 6.571468116160854e-05, "loss": 1.0399, "step": 137295 }, { "epoch": 0.34, "learning_rate": 6.571342292012683e-05, "loss": 1.0224, "step": 137300 }, { "epoch": 0.34, "learning_rate": 6.571216467864512e-05, "loss": 1.0195, "step": 137305 }, { "epoch": 0.34, "learning_rate": 6.571090643716343e-05, "loss": 1.023, "step": 137310 }, { "epoch": 0.34, "learning_rate": 6.570964819568172e-05, "loss": 1.0211, "step": 137315 }, { "epoch": 0.34, "learning_rate": 6.570838995420001e-05, "loss": 1.0212, "step": 137320 }, { "epoch": 0.34, "learning_rate": 6.57071317127183e-05, "loss": 1.0194, "step": 137325 }, { "epoch": 0.34, "learning_rate": 6.57058734712366e-05, "loss": 1.0197, "step": 137330 }, { "epoch": 0.34, "learning_rate": 6.57046152297549e-05, "loss": 1.0187, "step": 137335 }, { "epoch": 0.34, "learning_rate": 6.570335698827319e-05, "loss": 1.0222, "step": 137340 }, { "epoch": 0.34, "learning_rate": 6.570209874679148e-05, "loss": 1.0215, "step": 137345 }, { "epoch": 0.34, "learning_rate": 6.570084050530979e-05, "loss": 1.0225, "step": 137350 }, { "epoch": 0.34, "learning_rate": 6.569958226382808e-05, "loss": 1.0213, "step": 137355 }, { "epoch": 0.34, "learning_rate": 6.569832402234637e-05, "loss": 1.0216, "step": 137360 }, { "epoch": 0.34, "learning_rate": 6.569706578086466e-05, "loss": 1.0195, "step": 137365 }, { "epoch": 0.34, "learning_rate": 6.569580753938295e-05, "loss": 1.0227, "step": 137370 }, { "epoch": 0.34, "learning_rate": 6.569454929790126e-05, "loss": 1.0411, "step": 137375 }, { "epoch": 0.34, "learning_rate": 6.569329105641955e-05, "loss": 1.0201, "step": 137380 }, { "epoch": 0.34, "learning_rate": 6.569203281493784e-05, "loss": 1.043, "step": 137385 }, { "epoch": 0.34, "learning_rate": 6.569077457345613e-05, "loss": 1.0235, "step": 137390 }, { "epoch": 0.34, "learning_rate": 6.568951633197444e-05, "loss": 1.0206, "step": 137395 }, { "epoch": 0.34, "learning_rate": 6.568825809049273e-05, "loss": 1.021, "step": 137400 }, { "epoch": 0.34, "learning_rate": 6.568699984901102e-05, "loss": 1.0205, "step": 137405 }, { "epoch": 0.34, "learning_rate": 6.568574160752931e-05, "loss": 1.0225, "step": 137410 }, { "epoch": 0.34, "learning_rate": 6.568448336604762e-05, "loss": 1.0219, "step": 137415 }, { "epoch": 0.34, "learning_rate": 6.568322512456591e-05, "loss": 1.0224, "step": 137420 }, { "epoch": 0.34, "learning_rate": 6.56819668830842e-05, "loss": 1.0199, "step": 137425 }, { "epoch": 0.34, "learning_rate": 6.568070864160249e-05, "loss": 1.0221, "step": 137430 }, { "epoch": 0.34, "learning_rate": 6.567945040012078e-05, "loss": 1.0217, "step": 137435 }, { "epoch": 0.34, "learning_rate": 6.567819215863909e-05, "loss": 1.0218, "step": 137440 }, { "epoch": 0.35, "learning_rate": 6.567693391715738e-05, "loss": 1.0205, "step": 137445 }, { "epoch": 0.35, "learning_rate": 6.567567567567567e-05, "loss": 1.021, "step": 137450 }, { "epoch": 0.35, "learning_rate": 6.567441743419396e-05, "loss": 1.0208, "step": 137455 }, { "epoch": 0.35, "learning_rate": 6.567315919271227e-05, "loss": 1.0222, "step": 137460 }, { "epoch": 0.35, "learning_rate": 6.567190095123056e-05, "loss": 1.0411, "step": 137465 }, { "epoch": 0.35, "learning_rate": 6.567064270974885e-05, "loss": 1.0208, "step": 137470 }, { "epoch": 0.35, "learning_rate": 6.566938446826714e-05, "loss": 1.0224, "step": 137475 }, { "epoch": 0.35, "learning_rate": 6.566812622678545e-05, "loss": 1.0205, "step": 137480 }, { "epoch": 0.35, "learning_rate": 6.566686798530374e-05, "loss": 1.0197, "step": 137485 }, { "epoch": 0.35, "learning_rate": 6.566560974382205e-05, "loss": 1.0211, "step": 137490 }, { "epoch": 0.35, "learning_rate": 6.566435150234034e-05, "loss": 1.0236, "step": 137495 }, { "epoch": 0.35, "learning_rate": 6.566309326085863e-05, "loss": 1.0205, "step": 137500 }, { "epoch": 0.35, "learning_rate": 6.566183501937693e-05, "loss": 1.023, "step": 137505 }, { "epoch": 0.35, "learning_rate": 6.566057677789523e-05, "loss": 1.0198, "step": 137510 }, { "epoch": 0.35, "learning_rate": 6.565931853641352e-05, "loss": 1.0221, "step": 137515 }, { "epoch": 0.35, "learning_rate": 6.565806029493181e-05, "loss": 1.0206, "step": 137520 }, { "epoch": 0.35, "learning_rate": 6.56568020534501e-05, "loss": 1.0205, "step": 137525 }, { "epoch": 0.35, "learning_rate": 6.56555438119684e-05, "loss": 1.0184, "step": 137530 }, { "epoch": 0.35, "learning_rate": 6.56542855704867e-05, "loss": 1.0213, "step": 137535 }, { "epoch": 0.35, "learning_rate": 6.565302732900499e-05, "loss": 1.0208, "step": 137540 }, { "epoch": 0.35, "learning_rate": 6.565176908752328e-05, "loss": 1.0241, "step": 137545 }, { "epoch": 0.35, "learning_rate": 6.565051084604159e-05, "loss": 1.0212, "step": 137550 }, { "epoch": 0.35, "learning_rate": 6.564925260455988e-05, "loss": 1.0205, "step": 137555 }, { "epoch": 0.35, "learning_rate": 6.564799436307817e-05, "loss": 1.0209, "step": 137560 }, { "epoch": 0.35, "learning_rate": 6.564673612159646e-05, "loss": 1.0219, "step": 137565 }, { "epoch": 0.35, "learning_rate": 6.564547788011476e-05, "loss": 1.0205, "step": 137570 }, { "epoch": 0.35, "learning_rate": 6.564421963863306e-05, "loss": 1.0194, "step": 137575 }, { "epoch": 0.35, "learning_rate": 6.564296139715135e-05, "loss": 1.0211, "step": 137580 }, { "epoch": 0.35, "learning_rate": 6.564170315566964e-05, "loss": 1.0199, "step": 137585 }, { "epoch": 0.35, "learning_rate": 6.564044491418793e-05, "loss": 1.021, "step": 137590 }, { "epoch": 0.35, "learning_rate": 6.563918667270624e-05, "loss": 1.0225, "step": 137595 }, { "epoch": 0.35, "learning_rate": 6.563792843122453e-05, "loss": 1.0221, "step": 137600 }, { "epoch": 0.35, "learning_rate": 6.563667018974282e-05, "loss": 1.0208, "step": 137605 }, { "epoch": 0.35, "learning_rate": 6.563541194826111e-05, "loss": 1.0218, "step": 137610 }, { "epoch": 0.35, "learning_rate": 6.563415370677942e-05, "loss": 1.0215, "step": 137615 }, { "epoch": 0.35, "learning_rate": 6.563289546529771e-05, "loss": 1.0196, "step": 137620 }, { "epoch": 0.35, "learning_rate": 6.5631637223816e-05, "loss": 1.023, "step": 137625 }, { "epoch": 0.35, "learning_rate": 6.563037898233429e-05, "loss": 1.0222, "step": 137630 }, { "epoch": 0.35, "learning_rate": 6.56291207408526e-05, "loss": 1.0233, "step": 137635 }, { "epoch": 0.35, "learning_rate": 6.562786249937089e-05, "loss": 1.0191, "step": 137640 }, { "epoch": 0.35, "learning_rate": 6.562660425788918e-05, "loss": 1.0232, "step": 137645 }, { "epoch": 0.35, "learning_rate": 6.562534601640747e-05, "loss": 1.0212, "step": 137650 }, { "epoch": 0.35, "learning_rate": 6.562408777492576e-05, "loss": 1.0197, "step": 137655 }, { "epoch": 0.35, "learning_rate": 6.562282953344407e-05, "loss": 1.0189, "step": 137660 }, { "epoch": 0.35, "learning_rate": 6.562157129196236e-05, "loss": 1.0198, "step": 137665 }, { "epoch": 0.35, "learning_rate": 6.562031305048065e-05, "loss": 1.0222, "step": 137670 }, { "epoch": 0.35, "learning_rate": 6.561905480899894e-05, "loss": 1.0218, "step": 137675 }, { "epoch": 0.35, "learning_rate": 6.561779656751725e-05, "loss": 1.018, "step": 137680 }, { "epoch": 0.35, "learning_rate": 6.561653832603554e-05, "loss": 1.0207, "step": 137685 }, { "epoch": 0.35, "learning_rate": 6.561528008455383e-05, "loss": 1.0221, "step": 137690 }, { "epoch": 0.35, "learning_rate": 6.561402184307212e-05, "loss": 1.0192, "step": 137695 }, { "epoch": 0.35, "learning_rate": 6.561276360159043e-05, "loss": 1.0216, "step": 137700 }, { "epoch": 0.35, "learning_rate": 6.561150536010872e-05, "loss": 1.0219, "step": 137705 }, { "epoch": 0.35, "learning_rate": 6.561024711862701e-05, "loss": 1.0226, "step": 137710 }, { "epoch": 0.35, "learning_rate": 6.56089888771453e-05, "loss": 1.0218, "step": 137715 }, { "epoch": 0.35, "learning_rate": 6.56077306356636e-05, "loss": 1.0227, "step": 137720 }, { "epoch": 0.35, "learning_rate": 6.56064723941819e-05, "loss": 1.0194, "step": 137725 }, { "epoch": 0.35, "learning_rate": 6.560521415270019e-05, "loss": 1.0207, "step": 137730 }, { "epoch": 0.35, "learning_rate": 6.560395591121848e-05, "loss": 1.0219, "step": 137735 }, { "epoch": 0.35, "learning_rate": 6.560269766973677e-05, "loss": 1.0442, "step": 137740 }, { "epoch": 0.35, "learning_rate": 6.560143942825508e-05, "loss": 1.0215, "step": 137745 }, { "epoch": 0.35, "learning_rate": 6.560018118677337e-05, "loss": 1.0205, "step": 137750 }, { "epoch": 0.35, "learning_rate": 6.559892294529166e-05, "loss": 1.0204, "step": 137755 }, { "epoch": 0.35, "learning_rate": 6.559766470380995e-05, "loss": 1.0213, "step": 137760 }, { "epoch": 0.35, "learning_rate": 6.559640646232826e-05, "loss": 1.0214, "step": 137765 }, { "epoch": 0.35, "learning_rate": 6.559514822084655e-05, "loss": 1.0205, "step": 137770 }, { "epoch": 0.35, "learning_rate": 6.559388997936484e-05, "loss": 1.0179, "step": 137775 }, { "epoch": 0.35, "learning_rate": 6.559263173788313e-05, "loss": 1.0213, "step": 137780 }, { "epoch": 0.35, "learning_rate": 6.559137349640143e-05, "loss": 1.0194, "step": 137785 }, { "epoch": 0.35, "learning_rate": 6.559011525491973e-05, "loss": 1.0206, "step": 137790 }, { "epoch": 0.35, "learning_rate": 6.558885701343802e-05, "loss": 1.0205, "step": 137795 }, { "epoch": 0.35, "learning_rate": 6.558759877195631e-05, "loss": 1.0214, "step": 137800 }, { "epoch": 0.35, "learning_rate": 6.55863405304746e-05, "loss": 1.0197, "step": 137805 }, { "epoch": 0.35, "learning_rate": 6.558508228899291e-05, "loss": 1.0212, "step": 137810 }, { "epoch": 0.35, "learning_rate": 6.55838240475112e-05, "loss": 1.0214, "step": 137815 }, { "epoch": 0.35, "learning_rate": 6.55825658060295e-05, "loss": 1.0215, "step": 137820 }, { "epoch": 0.35, "learning_rate": 6.558130756454779e-05, "loss": 1.0218, "step": 137825 }, { "epoch": 0.35, "learning_rate": 6.558004932306609e-05, "loss": 1.0221, "step": 137830 }, { "epoch": 0.35, "learning_rate": 6.557879108158438e-05, "loss": 1.0182, "step": 137835 }, { "epoch": 0.35, "learning_rate": 6.557753284010267e-05, "loss": 1.0212, "step": 137840 }, { "epoch": 0.35, "learning_rate": 6.557627459862097e-05, "loss": 1.0214, "step": 137845 }, { "epoch": 0.35, "learning_rate": 6.557501635713926e-05, "loss": 1.0246, "step": 137850 }, { "epoch": 0.35, "learning_rate": 6.557375811565756e-05, "loss": 1.0193, "step": 137855 }, { "epoch": 0.35, "learning_rate": 6.557249987417585e-05, "loss": 1.0212, "step": 137860 }, { "epoch": 0.35, "learning_rate": 6.557124163269415e-05, "loss": 1.0405, "step": 137865 }, { "epoch": 0.35, "learning_rate": 6.556998339121244e-05, "loss": 1.022, "step": 137870 }, { "epoch": 0.35, "learning_rate": 6.556872514973074e-05, "loss": 1.0196, "step": 137875 }, { "epoch": 0.35, "learning_rate": 6.556746690824903e-05, "loss": 1.02, "step": 137880 }, { "epoch": 0.35, "learning_rate": 6.556620866676733e-05, "loss": 1.0216, "step": 137885 }, { "epoch": 0.35, "learning_rate": 6.556495042528562e-05, "loss": 1.019, "step": 137890 }, { "epoch": 0.35, "learning_rate": 6.556369218380392e-05, "loss": 1.02, "step": 137895 }, { "epoch": 0.35, "learning_rate": 6.556243394232221e-05, "loss": 1.0458, "step": 137900 }, { "epoch": 0.35, "learning_rate": 6.55611757008405e-05, "loss": 1.0231, "step": 137905 }, { "epoch": 0.35, "learning_rate": 6.55599174593588e-05, "loss": 1.0184, "step": 137910 }, { "epoch": 0.35, "learning_rate": 6.555865921787709e-05, "loss": 1.0284, "step": 137915 }, { "epoch": 0.35, "learning_rate": 6.55574009763954e-05, "loss": 1.0204, "step": 137920 }, { "epoch": 0.35, "learning_rate": 6.555614273491369e-05, "loss": 1.0193, "step": 137925 }, { "epoch": 0.35, "learning_rate": 6.555488449343198e-05, "loss": 1.0208, "step": 137930 }, { "epoch": 0.35, "learning_rate": 6.555362625195027e-05, "loss": 1.0223, "step": 137935 }, { "epoch": 0.35, "learning_rate": 6.555236801046857e-05, "loss": 1.0213, "step": 137940 }, { "epoch": 0.35, "learning_rate": 6.555110976898687e-05, "loss": 1.0224, "step": 137945 }, { "epoch": 0.35, "learning_rate": 6.554985152750516e-05, "loss": 1.0235, "step": 137950 }, { "epoch": 0.35, "learning_rate": 6.554859328602345e-05, "loss": 1.0187, "step": 137955 }, { "epoch": 0.35, "learning_rate": 6.554733504454174e-05, "loss": 1.0232, "step": 137960 }, { "epoch": 0.35, "learning_rate": 6.554607680306005e-05, "loss": 1.0211, "step": 137965 }, { "epoch": 0.35, "learning_rate": 6.554481856157834e-05, "loss": 1.0419, "step": 137970 }, { "epoch": 0.35, "learning_rate": 6.554356032009663e-05, "loss": 1.0201, "step": 137975 }, { "epoch": 0.35, "learning_rate": 6.554230207861492e-05, "loss": 1.023, "step": 137980 }, { "epoch": 0.35, "learning_rate": 6.554104383713323e-05, "loss": 1.0201, "step": 137985 }, { "epoch": 0.35, "learning_rate": 6.553978559565153e-05, "loss": 1.0202, "step": 137990 }, { "epoch": 0.35, "learning_rate": 6.553852735416982e-05, "loss": 1.0214, "step": 137995 }, { "epoch": 0.35, "learning_rate": 6.553726911268811e-05, "loss": 1.0225, "step": 138000 }, { "epoch": 0.35, "learning_rate": 6.55360108712064e-05, "loss": 1.0239, "step": 138005 }, { "epoch": 0.35, "learning_rate": 6.553475262972471e-05, "loss": 1.0216, "step": 138010 }, { "epoch": 0.35, "learning_rate": 6.5533494388243e-05, "loss": 1.0224, "step": 138015 }, { "epoch": 0.35, "learning_rate": 6.55322361467613e-05, "loss": 1.0223, "step": 138020 }, { "epoch": 0.35, "learning_rate": 6.553097790527958e-05, "loss": 1.0223, "step": 138025 }, { "epoch": 0.35, "learning_rate": 6.552971966379789e-05, "loss": 1.0208, "step": 138030 }, { "epoch": 0.35, "learning_rate": 6.552846142231618e-05, "loss": 1.0228, "step": 138035 }, { "epoch": 0.35, "learning_rate": 6.552720318083447e-05, "loss": 1.0217, "step": 138040 }, { "epoch": 0.35, "learning_rate": 6.552594493935276e-05, "loss": 1.0228, "step": 138045 }, { "epoch": 0.35, "learning_rate": 6.552468669787106e-05, "loss": 1.0215, "step": 138050 }, { "epoch": 0.35, "learning_rate": 6.552342845638936e-05, "loss": 1.0208, "step": 138055 }, { "epoch": 0.35, "learning_rate": 6.552217021490765e-05, "loss": 1.0209, "step": 138060 }, { "epoch": 0.35, "learning_rate": 6.552091197342594e-05, "loss": 1.0204, "step": 138065 }, { "epoch": 0.35, "learning_rate": 6.551965373194424e-05, "loss": 1.0224, "step": 138070 }, { "epoch": 0.35, "learning_rate": 6.551839549046254e-05, "loss": 1.041, "step": 138075 }, { "epoch": 0.35, "learning_rate": 6.551713724898083e-05, "loss": 1.0212, "step": 138080 }, { "epoch": 0.35, "learning_rate": 6.551587900749912e-05, "loss": 1.0213, "step": 138085 }, { "epoch": 0.35, "learning_rate": 6.551462076601742e-05, "loss": 1.0201, "step": 138090 }, { "epoch": 0.35, "learning_rate": 6.551336252453572e-05, "loss": 1.023, "step": 138095 }, { "epoch": 0.35, "learning_rate": 6.551210428305401e-05, "loss": 1.0224, "step": 138100 }, { "epoch": 0.35, "learning_rate": 6.55108460415723e-05, "loss": 1.0314, "step": 138105 }, { "epoch": 0.35, "learning_rate": 6.55095878000906e-05, "loss": 1.021, "step": 138110 }, { "epoch": 0.35, "learning_rate": 6.550832955860889e-05, "loss": 1.0202, "step": 138115 }, { "epoch": 0.35, "learning_rate": 6.550707131712719e-05, "loss": 1.0173, "step": 138120 }, { "epoch": 0.35, "learning_rate": 6.550581307564548e-05, "loss": 1.0206, "step": 138125 }, { "epoch": 0.35, "learning_rate": 6.550455483416378e-05, "loss": 1.0218, "step": 138130 }, { "epoch": 0.35, "learning_rate": 6.550329659268207e-05, "loss": 1.0235, "step": 138135 }, { "epoch": 0.35, "learning_rate": 6.550203835120037e-05, "loss": 1.0175, "step": 138140 }, { "epoch": 0.35, "learning_rate": 6.550078010971866e-05, "loss": 1.0217, "step": 138145 }, { "epoch": 0.35, "learning_rate": 6.549952186823696e-05, "loss": 1.0197, "step": 138150 }, { "epoch": 0.35, "learning_rate": 6.549826362675525e-05, "loss": 1.0214, "step": 138155 }, { "epoch": 0.35, "learning_rate": 6.549700538527355e-05, "loss": 1.0204, "step": 138160 }, { "epoch": 0.35, "learning_rate": 6.549574714379184e-05, "loss": 1.0207, "step": 138165 }, { "epoch": 0.35, "learning_rate": 6.549448890231014e-05, "loss": 1.0205, "step": 138170 }, { "epoch": 0.35, "learning_rate": 6.549323066082843e-05, "loss": 1.0225, "step": 138175 }, { "epoch": 0.35, "learning_rate": 6.549197241934672e-05, "loss": 1.0191, "step": 138180 }, { "epoch": 0.35, "learning_rate": 6.549071417786502e-05, "loss": 1.0204, "step": 138185 }, { "epoch": 0.35, "learning_rate": 6.548945593638332e-05, "loss": 1.0222, "step": 138190 }, { "epoch": 0.35, "learning_rate": 6.548819769490161e-05, "loss": 1.0207, "step": 138195 }, { "epoch": 0.35, "learning_rate": 6.54869394534199e-05, "loss": 1.023, "step": 138200 }, { "epoch": 0.35, "learning_rate": 6.54856812119382e-05, "loss": 1.0207, "step": 138205 }, { "epoch": 0.35, "learning_rate": 6.54844229704565e-05, "loss": 1.0217, "step": 138210 }, { "epoch": 0.35, "learning_rate": 6.548316472897479e-05, "loss": 1.0207, "step": 138215 }, { "epoch": 0.35, "learning_rate": 6.548190648749308e-05, "loss": 1.0225, "step": 138220 }, { "epoch": 0.35, "learning_rate": 6.548064824601138e-05, "loss": 1.021, "step": 138225 }, { "epoch": 0.35, "learning_rate": 6.547939000452968e-05, "loss": 1.043, "step": 138230 }, { "epoch": 0.35, "learning_rate": 6.547813176304797e-05, "loss": 1.0224, "step": 138235 }, { "epoch": 0.35, "learning_rate": 6.547687352156626e-05, "loss": 1.0221, "step": 138240 }, { "epoch": 0.35, "learning_rate": 6.547561528008455e-05, "loss": 1.0203, "step": 138245 }, { "epoch": 0.35, "learning_rate": 6.547435703860286e-05, "loss": 1.0226, "step": 138250 }, { "epoch": 0.35, "learning_rate": 6.547309879712115e-05, "loss": 1.0209, "step": 138255 }, { "epoch": 0.35, "learning_rate": 6.547184055563944e-05, "loss": 1.0193, "step": 138260 }, { "epoch": 0.35, "learning_rate": 6.547058231415773e-05, "loss": 1.0201, "step": 138265 }, { "epoch": 0.35, "learning_rate": 6.546932407267604e-05, "loss": 1.0212, "step": 138270 }, { "epoch": 0.35, "learning_rate": 6.546806583119433e-05, "loss": 1.0192, "step": 138275 }, { "epoch": 0.35, "learning_rate": 6.546680758971262e-05, "loss": 1.0209, "step": 138280 }, { "epoch": 0.35, "learning_rate": 6.546554934823091e-05, "loss": 1.0211, "step": 138285 }, { "epoch": 0.35, "learning_rate": 6.546429110674922e-05, "loss": 1.019, "step": 138290 }, { "epoch": 0.35, "learning_rate": 6.546303286526751e-05, "loss": 1.0199, "step": 138295 }, { "epoch": 0.35, "learning_rate": 6.54617746237858e-05, "loss": 1.0227, "step": 138300 }, { "epoch": 0.35, "learning_rate": 6.546051638230409e-05, "loss": 1.0206, "step": 138305 }, { "epoch": 0.35, "learning_rate": 6.545925814082238e-05, "loss": 1.0204, "step": 138310 }, { "epoch": 0.35, "learning_rate": 6.545799989934069e-05, "loss": 1.0191, "step": 138315 }, { "epoch": 0.35, "learning_rate": 6.545674165785898e-05, "loss": 1.0214, "step": 138320 }, { "epoch": 0.35, "learning_rate": 6.545548341637727e-05, "loss": 1.0226, "step": 138325 }, { "epoch": 0.35, "learning_rate": 6.545422517489556e-05, "loss": 1.0199, "step": 138330 }, { "epoch": 0.35, "learning_rate": 6.545296693341387e-05, "loss": 1.0203, "step": 138335 }, { "epoch": 0.35, "learning_rate": 6.545170869193216e-05, "loss": 1.045, "step": 138340 }, { "epoch": 0.35, "learning_rate": 6.545045045045045e-05, "loss": 1.0206, "step": 138345 }, { "epoch": 0.35, "learning_rate": 6.544919220896874e-05, "loss": 1.0183, "step": 138350 }, { "epoch": 0.35, "learning_rate": 6.544793396748705e-05, "loss": 1.0207, "step": 138355 }, { "epoch": 0.35, "learning_rate": 6.544667572600534e-05, "loss": 1.0218, "step": 138360 }, { "epoch": 0.35, "learning_rate": 6.544541748452363e-05, "loss": 1.0196, "step": 138365 }, { "epoch": 0.35, "learning_rate": 6.544415924304192e-05, "loss": 1.0224, "step": 138370 }, { "epoch": 0.35, "learning_rate": 6.544290100156021e-05, "loss": 1.0237, "step": 138375 }, { "epoch": 0.35, "learning_rate": 6.544164276007852e-05, "loss": 1.0187, "step": 138380 }, { "epoch": 0.35, "learning_rate": 6.544038451859681e-05, "loss": 1.0206, "step": 138385 }, { "epoch": 0.35, "learning_rate": 6.54391262771151e-05, "loss": 1.02, "step": 138390 }, { "epoch": 0.35, "learning_rate": 6.54378680356334e-05, "loss": 1.0219, "step": 138395 }, { "epoch": 0.35, "learning_rate": 6.54366097941517e-05, "loss": 1.0204, "step": 138400 }, { "epoch": 0.35, "learning_rate": 6.543535155266999e-05, "loss": 1.0204, "step": 138405 }, { "epoch": 0.35, "learning_rate": 6.543409331118828e-05, "loss": 1.021, "step": 138410 }, { "epoch": 0.35, "learning_rate": 6.543283506970657e-05, "loss": 1.0212, "step": 138415 }, { "epoch": 0.35, "learning_rate": 6.543157682822488e-05, "loss": 1.0202, "step": 138420 }, { "epoch": 0.35, "learning_rate": 6.543031858674317e-05, "loss": 1.0217, "step": 138425 }, { "epoch": 0.35, "learning_rate": 6.542906034526146e-05, "loss": 1.0213, "step": 138430 }, { "epoch": 0.35, "learning_rate": 6.542780210377975e-05, "loss": 1.0213, "step": 138435 }, { "epoch": 0.35, "learning_rate": 6.542654386229805e-05, "loss": 1.0235, "step": 138440 }, { "epoch": 0.35, "learning_rate": 6.542528562081635e-05, "loss": 1.0201, "step": 138445 }, { "epoch": 0.35, "learning_rate": 6.542402737933464e-05, "loss": 1.0346, "step": 138450 }, { "epoch": 0.35, "learning_rate": 6.542276913785293e-05, "loss": 1.0206, "step": 138455 }, { "epoch": 0.35, "learning_rate": 6.542151089637122e-05, "loss": 1.0409, "step": 138460 }, { "epoch": 0.35, "learning_rate": 6.542025265488953e-05, "loss": 1.0216, "step": 138465 }, { "epoch": 0.35, "learning_rate": 6.541899441340782e-05, "loss": 1.0229, "step": 138470 }, { "epoch": 0.35, "learning_rate": 6.541773617192611e-05, "loss": 1.0224, "step": 138475 }, { "epoch": 0.35, "learning_rate": 6.54164779304444e-05, "loss": 1.0195, "step": 138480 }, { "epoch": 0.35, "learning_rate": 6.541521968896271e-05, "loss": 1.0183, "step": 138485 }, { "epoch": 0.35, "learning_rate": 6.541396144748102e-05, "loss": 1.0206, "step": 138490 }, { "epoch": 0.35, "learning_rate": 6.541270320599931e-05, "loss": 1.0218, "step": 138495 }, { "epoch": 0.35, "learning_rate": 6.54114449645176e-05, "loss": 1.0206, "step": 138500 }, { "epoch": 0.35, "learning_rate": 6.541018672303589e-05, "loss": 1.0212, "step": 138505 }, { "epoch": 0.35, "learning_rate": 6.540892848155418e-05, "loss": 1.0217, "step": 138510 }, { "epoch": 0.35, "learning_rate": 6.540767024007249e-05, "loss": 1.0208, "step": 138515 }, { "epoch": 0.35, "learning_rate": 6.540641199859078e-05, "loss": 1.0233, "step": 138520 }, { "epoch": 0.35, "learning_rate": 6.540515375710907e-05, "loss": 1.0215, "step": 138525 }, { "epoch": 0.35, "learning_rate": 6.540389551562736e-05, "loss": 1.0237, "step": 138530 }, { "epoch": 0.35, "learning_rate": 6.540263727414567e-05, "loss": 1.022, "step": 138535 }, { "epoch": 0.35, "learning_rate": 6.540137903266396e-05, "loss": 1.0208, "step": 138540 }, { "epoch": 0.35, "learning_rate": 6.540012079118225e-05, "loss": 1.0208, "step": 138545 }, { "epoch": 0.35, "learning_rate": 6.539886254970054e-05, "loss": 1.0203, "step": 138550 }, { "epoch": 0.35, "learning_rate": 6.539760430821885e-05, "loss": 1.0203, "step": 138555 }, { "epoch": 0.35, "learning_rate": 6.539634606673714e-05, "loss": 1.0196, "step": 138560 }, { "epoch": 0.35, "learning_rate": 6.539508782525543e-05, "loss": 1.0242, "step": 138565 }, { "epoch": 0.35, "learning_rate": 6.539382958377372e-05, "loss": 1.0211, "step": 138570 }, { "epoch": 0.35, "learning_rate": 6.539257134229201e-05, "loss": 1.0206, "step": 138575 }, { "epoch": 0.35, "learning_rate": 6.539131310081032e-05, "loss": 1.0206, "step": 138580 }, { "epoch": 0.35, "learning_rate": 6.539005485932861e-05, "loss": 1.0223, "step": 138585 }, { "epoch": 0.35, "learning_rate": 6.53887966178469e-05, "loss": 1.0207, "step": 138590 }, { "epoch": 0.35, "learning_rate": 6.538753837636519e-05, "loss": 1.0192, "step": 138595 }, { "epoch": 0.35, "learning_rate": 6.53862801348835e-05, "loss": 1.0201, "step": 138600 }, { "epoch": 0.35, "learning_rate": 6.538502189340179e-05, "loss": 1.0219, "step": 138605 }, { "epoch": 0.35, "learning_rate": 6.538376365192008e-05, "loss": 1.0233, "step": 138610 }, { "epoch": 0.35, "learning_rate": 6.538250541043837e-05, "loss": 1.0215, "step": 138615 }, { "epoch": 0.35, "learning_rate": 6.538124716895668e-05, "loss": 1.0219, "step": 138620 }, { "epoch": 0.35, "learning_rate": 6.537998892747497e-05, "loss": 1.0193, "step": 138625 }, { "epoch": 0.35, "learning_rate": 6.537873068599326e-05, "loss": 1.0207, "step": 138630 }, { "epoch": 0.35, "learning_rate": 6.537747244451155e-05, "loss": 1.0201, "step": 138635 }, { "epoch": 0.35, "learning_rate": 6.537621420302984e-05, "loss": 1.0198, "step": 138640 }, { "epoch": 0.35, "learning_rate": 6.537495596154815e-05, "loss": 1.0236, "step": 138645 }, { "epoch": 0.35, "learning_rate": 6.537369772006644e-05, "loss": 1.0205, "step": 138650 }, { "epoch": 0.35, "learning_rate": 6.537243947858473e-05, "loss": 1.0214, "step": 138655 }, { "epoch": 0.35, "learning_rate": 6.537118123710302e-05, "loss": 1.0213, "step": 138660 }, { "epoch": 0.35, "learning_rate": 6.536992299562133e-05, "loss": 1.0221, "step": 138665 }, { "epoch": 0.35, "learning_rate": 6.536866475413962e-05, "loss": 1.023, "step": 138670 }, { "epoch": 0.35, "learning_rate": 6.536740651265791e-05, "loss": 1.0195, "step": 138675 }, { "epoch": 0.35, "learning_rate": 6.53661482711762e-05, "loss": 1.0195, "step": 138680 }, { "epoch": 0.35, "learning_rate": 6.536489002969451e-05, "loss": 1.0218, "step": 138685 }, { "epoch": 0.35, "learning_rate": 6.53636317882128e-05, "loss": 1.0213, "step": 138690 }, { "epoch": 0.35, "learning_rate": 6.536237354673109e-05, "loss": 1.0213, "step": 138695 }, { "epoch": 0.35, "learning_rate": 6.536111530524938e-05, "loss": 1.0207, "step": 138700 }, { "epoch": 0.35, "learning_rate": 6.535985706376768e-05, "loss": 1.0202, "step": 138705 }, { "epoch": 0.35, "learning_rate": 6.535859882228598e-05, "loss": 1.0198, "step": 138710 }, { "epoch": 0.35, "learning_rate": 6.535734058080427e-05, "loss": 1.019, "step": 138715 }, { "epoch": 0.35, "learning_rate": 6.535608233932256e-05, "loss": 1.0195, "step": 138720 }, { "epoch": 0.35, "learning_rate": 6.535482409784086e-05, "loss": 1.0198, "step": 138725 }, { "epoch": 0.35, "learning_rate": 6.535356585635916e-05, "loss": 1.0216, "step": 138730 }, { "epoch": 0.35, "learning_rate": 6.535230761487745e-05, "loss": 1.0186, "step": 138735 }, { "epoch": 0.35, "learning_rate": 6.535104937339574e-05, "loss": 1.0199, "step": 138740 }, { "epoch": 0.35, "learning_rate": 6.534979113191404e-05, "loss": 1.044, "step": 138745 }, { "epoch": 0.35, "learning_rate": 6.534853289043234e-05, "loss": 1.0189, "step": 138750 }, { "epoch": 0.35, "learning_rate": 6.534727464895063e-05, "loss": 1.0221, "step": 138755 }, { "epoch": 0.35, "learning_rate": 6.534601640746892e-05, "loss": 1.0199, "step": 138760 }, { "epoch": 0.35, "learning_rate": 6.534475816598722e-05, "loss": 1.0227, "step": 138765 }, { "epoch": 0.35, "learning_rate": 6.534349992450551e-05, "loss": 1.0201, "step": 138770 }, { "epoch": 0.35, "learning_rate": 6.534224168302381e-05, "loss": 1.0202, "step": 138775 }, { "epoch": 0.35, "learning_rate": 6.53409834415421e-05, "loss": 1.0223, "step": 138780 }, { "epoch": 0.35, "learning_rate": 6.53397252000604e-05, "loss": 1.0227, "step": 138785 }, { "epoch": 0.35, "learning_rate": 6.533846695857869e-05, "loss": 1.0216, "step": 138790 }, { "epoch": 0.35, "learning_rate": 6.533720871709699e-05, "loss": 1.0223, "step": 138795 }, { "epoch": 0.35, "learning_rate": 6.533595047561528e-05, "loss": 1.0212, "step": 138800 }, { "epoch": 0.35, "learning_rate": 6.533469223413358e-05, "loss": 1.0216, "step": 138805 }, { "epoch": 0.35, "learning_rate": 6.533343399265187e-05, "loss": 1.0226, "step": 138810 }, { "epoch": 0.35, "learning_rate": 6.533217575117017e-05, "loss": 1.0179, "step": 138815 }, { "epoch": 0.35, "learning_rate": 6.533091750968846e-05, "loss": 1.0229, "step": 138820 }, { "epoch": 0.35, "learning_rate": 6.532965926820676e-05, "loss": 1.0218, "step": 138825 }, { "epoch": 0.35, "learning_rate": 6.532840102672505e-05, "loss": 1.0211, "step": 138830 }, { "epoch": 0.35, "learning_rate": 6.532714278524334e-05, "loss": 1.0225, "step": 138835 }, { "epoch": 0.35, "learning_rate": 6.532588454376164e-05, "loss": 1.0244, "step": 138840 }, { "epoch": 0.35, "learning_rate": 6.532462630227994e-05, "loss": 1.0209, "step": 138845 }, { "epoch": 0.35, "learning_rate": 6.532336806079823e-05, "loss": 1.0247, "step": 138850 }, { "epoch": 0.35, "learning_rate": 6.532210981931652e-05, "loss": 1.0197, "step": 138855 }, { "epoch": 0.35, "learning_rate": 6.532085157783482e-05, "loss": 1.0219, "step": 138860 }, { "epoch": 0.35, "learning_rate": 6.531959333635312e-05, "loss": 1.021, "step": 138865 }, { "epoch": 0.35, "learning_rate": 6.531833509487141e-05, "loss": 1.0204, "step": 138870 }, { "epoch": 0.35, "learning_rate": 6.53170768533897e-05, "loss": 1.0194, "step": 138875 }, { "epoch": 0.35, "learning_rate": 6.5315818611908e-05, "loss": 1.024, "step": 138880 }, { "epoch": 0.35, "learning_rate": 6.53145603704263e-05, "loss": 1.022, "step": 138885 }, { "epoch": 0.35, "learning_rate": 6.531330212894459e-05, "loss": 1.0208, "step": 138890 }, { "epoch": 0.35, "learning_rate": 6.531204388746288e-05, "loss": 1.021, "step": 138895 }, { "epoch": 0.35, "learning_rate": 6.531078564598117e-05, "loss": 1.022, "step": 138900 }, { "epoch": 0.35, "learning_rate": 6.530952740449948e-05, "loss": 1.022, "step": 138905 }, { "epoch": 0.35, "learning_rate": 6.530826916301777e-05, "loss": 1.0198, "step": 138910 }, { "epoch": 0.35, "learning_rate": 6.530701092153606e-05, "loss": 1.022, "step": 138915 }, { "epoch": 0.35, "learning_rate": 6.530575268005435e-05, "loss": 1.0413, "step": 138920 }, { "epoch": 0.35, "learning_rate": 6.530449443857266e-05, "loss": 1.0207, "step": 138925 }, { "epoch": 0.35, "learning_rate": 6.530323619709095e-05, "loss": 1.0222, "step": 138930 }, { "epoch": 0.35, "learning_rate": 6.530197795560924e-05, "loss": 1.0232, "step": 138935 }, { "epoch": 0.35, "learning_rate": 6.530071971412753e-05, "loss": 1.0209, "step": 138940 }, { "epoch": 0.35, "learning_rate": 6.529946147264584e-05, "loss": 1.0223, "step": 138945 }, { "epoch": 0.35, "learning_rate": 6.529820323116413e-05, "loss": 1.0182, "step": 138950 }, { "epoch": 0.35, "learning_rate": 6.529694498968242e-05, "loss": 1.0435, "step": 138955 }, { "epoch": 0.35, "learning_rate": 6.529568674820071e-05, "loss": 1.0213, "step": 138960 }, { "epoch": 0.35, "learning_rate": 6.5294428506719e-05, "loss": 1.0201, "step": 138965 }, { "epoch": 0.35, "learning_rate": 6.52931702652373e-05, "loss": 1.0194, "step": 138970 }, { "epoch": 0.35, "learning_rate": 6.52919120237556e-05, "loss": 1.0218, "step": 138975 }, { "epoch": 0.35, "learning_rate": 6.529065378227389e-05, "loss": 1.0223, "step": 138980 }, { "epoch": 0.35, "learning_rate": 6.528939554079218e-05, "loss": 1.023, "step": 138985 }, { "epoch": 0.35, "learning_rate": 6.528813729931049e-05, "loss": 1.0208, "step": 138990 }, { "epoch": 0.35, "learning_rate": 6.528687905782879e-05, "loss": 1.0222, "step": 138995 }, { "epoch": 0.35, "learning_rate": 6.528562081634708e-05, "loss": 1.0234, "step": 139000 }, { "epoch": 0.35, "learning_rate": 6.528436257486537e-05, "loss": 1.0204, "step": 139005 }, { "epoch": 0.35, "learning_rate": 6.528310433338367e-05, "loss": 1.0218, "step": 139010 }, { "epoch": 0.35, "learning_rate": 6.528184609190197e-05, "loss": 1.0433, "step": 139015 }, { "epoch": 0.35, "learning_rate": 6.528058785042026e-05, "loss": 1.0202, "step": 139020 }, { "epoch": 0.35, "learning_rate": 6.527932960893855e-05, "loss": 1.0195, "step": 139025 }, { "epoch": 0.35, "learning_rate": 6.527807136745685e-05, "loss": 1.0192, "step": 139030 }, { "epoch": 0.35, "learning_rate": 6.527681312597514e-05, "loss": 1.0214, "step": 139035 }, { "epoch": 0.35, "learning_rate": 6.527555488449344e-05, "loss": 1.0201, "step": 139040 }, { "epoch": 0.35, "learning_rate": 6.527429664301173e-05, "loss": 1.0188, "step": 139045 }, { "epoch": 0.35, "learning_rate": 6.527303840153003e-05, "loss": 1.019, "step": 139050 }, { "epoch": 0.35, "learning_rate": 6.527178016004832e-05, "loss": 1.021, "step": 139055 }, { "epoch": 0.35, "learning_rate": 6.527052191856662e-05, "loss": 1.025, "step": 139060 }, { "epoch": 0.35, "learning_rate": 6.526926367708491e-05, "loss": 1.0219, "step": 139065 }, { "epoch": 0.35, "learning_rate": 6.52680054356032e-05, "loss": 1.0203, "step": 139070 }, { "epoch": 0.35, "learning_rate": 6.52667471941215e-05, "loss": 1.02, "step": 139075 }, { "epoch": 0.35, "learning_rate": 6.52654889526398e-05, "loss": 1.0209, "step": 139080 }, { "epoch": 0.35, "learning_rate": 6.52642307111581e-05, "loss": 1.0209, "step": 139085 }, { "epoch": 0.35, "learning_rate": 6.526297246967639e-05, "loss": 1.0208, "step": 139090 }, { "epoch": 0.35, "learning_rate": 6.526171422819468e-05, "loss": 1.0194, "step": 139095 }, { "epoch": 0.35, "learning_rate": 6.526045598671297e-05, "loss": 1.0199, "step": 139100 }, { "epoch": 0.35, "learning_rate": 6.525919774523127e-05, "loss": 1.0194, "step": 139105 }, { "epoch": 0.35, "learning_rate": 6.525793950374957e-05, "loss": 1.0224, "step": 139110 }, { "epoch": 0.35, "learning_rate": 6.525668126226786e-05, "loss": 1.0227, "step": 139115 }, { "epoch": 0.35, "learning_rate": 6.525542302078615e-05, "loss": 1.02, "step": 139120 }, { "epoch": 0.35, "learning_rate": 6.525416477930445e-05, "loss": 1.0193, "step": 139125 }, { "epoch": 0.35, "learning_rate": 6.525290653782275e-05, "loss": 1.0188, "step": 139130 }, { "epoch": 0.35, "learning_rate": 6.525164829634104e-05, "loss": 1.0211, "step": 139135 }, { "epoch": 0.35, "learning_rate": 6.525039005485933e-05, "loss": 1.0201, "step": 139140 }, { "epoch": 0.35, "learning_rate": 6.524913181337763e-05, "loss": 1.0178, "step": 139145 }, { "epoch": 0.35, "learning_rate": 6.524787357189593e-05, "loss": 1.0212, "step": 139150 }, { "epoch": 0.35, "learning_rate": 6.524661533041422e-05, "loss": 1.0229, "step": 139155 }, { "epoch": 0.35, "learning_rate": 6.524535708893251e-05, "loss": 1.0425, "step": 139160 }, { "epoch": 0.35, "learning_rate": 6.52440988474508e-05, "loss": 1.0215, "step": 139165 }, { "epoch": 0.35, "learning_rate": 6.52428406059691e-05, "loss": 1.02, "step": 139170 }, { "epoch": 0.35, "learning_rate": 6.52415823644874e-05, "loss": 1.0216, "step": 139175 }, { "epoch": 0.35, "learning_rate": 6.524032412300569e-05, "loss": 1.0197, "step": 139180 }, { "epoch": 0.35, "learning_rate": 6.523906588152398e-05, "loss": 1.0191, "step": 139185 }, { "epoch": 0.35, "learning_rate": 6.523780764004229e-05, "loss": 1.0203, "step": 139190 }, { "epoch": 0.35, "learning_rate": 6.523654939856058e-05, "loss": 1.023, "step": 139195 }, { "epoch": 0.35, "learning_rate": 6.523529115707887e-05, "loss": 1.0185, "step": 139200 }, { "epoch": 0.35, "learning_rate": 6.523403291559716e-05, "loss": 1.0219, "step": 139205 }, { "epoch": 0.35, "learning_rate": 6.523277467411547e-05, "loss": 1.0191, "step": 139210 }, { "epoch": 0.35, "learning_rate": 6.523151643263376e-05, "loss": 1.0205, "step": 139215 }, { "epoch": 0.35, "learning_rate": 6.523025819115205e-05, "loss": 1.02, "step": 139220 }, { "epoch": 0.35, "learning_rate": 6.522899994967034e-05, "loss": 1.022, "step": 139225 }, { "epoch": 0.35, "learning_rate": 6.522774170818863e-05, "loss": 1.022, "step": 139230 }, { "epoch": 0.35, "learning_rate": 6.522648346670694e-05, "loss": 1.0213, "step": 139235 }, { "epoch": 0.35, "learning_rate": 6.522522522522523e-05, "loss": 1.0214, "step": 139240 }, { "epoch": 0.35, "learning_rate": 6.522396698374352e-05, "loss": 1.0224, "step": 139245 }, { "epoch": 0.35, "learning_rate": 6.522270874226181e-05, "loss": 1.0197, "step": 139250 }, { "epoch": 0.35, "learning_rate": 6.522145050078012e-05, "loss": 1.0207, "step": 139255 }, { "epoch": 0.35, "learning_rate": 6.522019225929841e-05, "loss": 1.0218, "step": 139260 }, { "epoch": 0.35, "learning_rate": 6.52189340178167e-05, "loss": 1.0217, "step": 139265 }, { "epoch": 0.35, "learning_rate": 6.521767577633499e-05, "loss": 1.0187, "step": 139270 }, { "epoch": 0.35, "learning_rate": 6.52164175348533e-05, "loss": 1.0187, "step": 139275 }, { "epoch": 0.35, "learning_rate": 6.521515929337159e-05, "loss": 1.0205, "step": 139280 }, { "epoch": 0.35, "learning_rate": 6.521390105188988e-05, "loss": 1.0209, "step": 139285 }, { "epoch": 0.35, "learning_rate": 6.521264281040817e-05, "loss": 1.021, "step": 139290 }, { "epoch": 0.35, "learning_rate": 6.521138456892646e-05, "loss": 1.0197, "step": 139295 }, { "epoch": 0.35, "learning_rate": 6.521012632744477e-05, "loss": 1.0209, "step": 139300 }, { "epoch": 0.35, "learning_rate": 6.520886808596306e-05, "loss": 1.0219, "step": 139305 }, { "epoch": 0.35, "learning_rate": 6.520760984448135e-05, "loss": 1.022, "step": 139310 }, { "epoch": 0.35, "learning_rate": 6.520635160299964e-05, "loss": 1.0188, "step": 139315 }, { "epoch": 0.35, "learning_rate": 6.520509336151795e-05, "loss": 1.0228, "step": 139320 }, { "epoch": 0.35, "learning_rate": 6.520383512003624e-05, "loss": 1.0236, "step": 139325 }, { "epoch": 0.35, "learning_rate": 6.520257687855453e-05, "loss": 1.0222, "step": 139330 }, { "epoch": 0.35, "learning_rate": 6.520131863707282e-05, "loss": 1.022, "step": 139335 }, { "epoch": 0.35, "learning_rate": 6.520006039559113e-05, "loss": 1.0219, "step": 139340 }, { "epoch": 0.35, "learning_rate": 6.519880215410942e-05, "loss": 1.0231, "step": 139345 }, { "epoch": 0.35, "learning_rate": 6.519754391262771e-05, "loss": 1.023, "step": 139350 }, { "epoch": 0.35, "learning_rate": 6.5196285671146e-05, "loss": 1.0197, "step": 139355 }, { "epoch": 0.35, "learning_rate": 6.51950274296643e-05, "loss": 1.0211, "step": 139360 }, { "epoch": 0.35, "learning_rate": 6.51937691881826e-05, "loss": 1.0224, "step": 139365 }, { "epoch": 0.35, "learning_rate": 6.519251094670089e-05, "loss": 1.0202, "step": 139370 }, { "epoch": 0.35, "learning_rate": 6.519125270521918e-05, "loss": 1.0208, "step": 139375 }, { "epoch": 0.35, "learning_rate": 6.518999446373748e-05, "loss": 1.0212, "step": 139380 }, { "epoch": 0.35, "learning_rate": 6.518873622225578e-05, "loss": 1.0267, "step": 139385 }, { "epoch": 0.35, "learning_rate": 6.518747798077407e-05, "loss": 1.0205, "step": 139390 }, { "epoch": 0.35, "learning_rate": 6.518621973929236e-05, "loss": 1.0229, "step": 139395 }, { "epoch": 0.35, "learning_rate": 6.518496149781066e-05, "loss": 1.0223, "step": 139400 }, { "epoch": 0.35, "learning_rate": 6.518370325632896e-05, "loss": 1.0217, "step": 139405 }, { "epoch": 0.35, "learning_rate": 6.518244501484725e-05, "loss": 1.0216, "step": 139410 }, { "epoch": 0.35, "learning_rate": 6.518118677336554e-05, "loss": 1.0179, "step": 139415 }, { "epoch": 0.35, "learning_rate": 6.517992853188384e-05, "loss": 1.0214, "step": 139420 }, { "epoch": 0.35, "learning_rate": 6.517867029040213e-05, "loss": 1.0211, "step": 139425 }, { "epoch": 0.35, "learning_rate": 6.517741204892043e-05, "loss": 1.0197, "step": 139430 }, { "epoch": 0.35, "learning_rate": 6.517615380743872e-05, "loss": 1.02, "step": 139435 }, { "epoch": 0.35, "learning_rate": 6.517489556595702e-05, "loss": 1.0235, "step": 139440 }, { "epoch": 0.35, "learning_rate": 6.51736373244753e-05, "loss": 1.0186, "step": 139445 }, { "epoch": 0.35, "learning_rate": 6.517237908299361e-05, "loss": 1.0208, "step": 139450 }, { "epoch": 0.35, "learning_rate": 6.51711208415119e-05, "loss": 1.021, "step": 139455 }, { "epoch": 0.35, "learning_rate": 6.51698626000302e-05, "loss": 1.0203, "step": 139460 }, { "epoch": 0.35, "learning_rate": 6.516860435854849e-05, "loss": 1.0202, "step": 139465 }, { "epoch": 0.35, "learning_rate": 6.516734611706679e-05, "loss": 1.018, "step": 139470 }, { "epoch": 0.35, "learning_rate": 6.516608787558508e-05, "loss": 1.0209, "step": 139475 }, { "epoch": 0.35, "learning_rate": 6.516482963410337e-05, "loss": 1.0201, "step": 139480 }, { "epoch": 0.35, "learning_rate": 6.516357139262167e-05, "loss": 1.0215, "step": 139485 }, { "epoch": 0.35, "learning_rate": 6.516231315113997e-05, "loss": 1.0225, "step": 139490 }, { "epoch": 0.35, "learning_rate": 6.516105490965828e-05, "loss": 1.0219, "step": 139495 }, { "epoch": 0.35, "learning_rate": 6.515979666817657e-05, "loss": 1.0217, "step": 139500 }, { "epoch": 0.35, "learning_rate": 6.515853842669486e-05, "loss": 1.0212, "step": 139505 }, { "epoch": 0.35, "learning_rate": 6.515728018521315e-05, "loss": 1.0226, "step": 139510 }, { "epoch": 0.35, "learning_rate": 6.515602194373144e-05, "loss": 1.0208, "step": 139515 }, { "epoch": 0.35, "learning_rate": 6.515476370224975e-05, "loss": 1.0211, "step": 139520 }, { "epoch": 0.35, "learning_rate": 6.515350546076804e-05, "loss": 1.0229, "step": 139525 }, { "epoch": 0.35, "learning_rate": 6.515224721928633e-05, "loss": 1.0215, "step": 139530 }, { "epoch": 0.35, "learning_rate": 6.515098897780462e-05, "loss": 1.0213, "step": 139535 }, { "epoch": 0.35, "learning_rate": 6.514973073632293e-05, "loss": 1.0208, "step": 139540 }, { "epoch": 0.35, "learning_rate": 6.514847249484122e-05, "loss": 1.0198, "step": 139545 }, { "epoch": 0.35, "learning_rate": 6.514721425335951e-05, "loss": 1.0216, "step": 139550 }, { "epoch": 0.35, "learning_rate": 6.51459560118778e-05, "loss": 1.0212, "step": 139555 }, { "epoch": 0.35, "learning_rate": 6.51446977703961e-05, "loss": 1.0203, "step": 139560 }, { "epoch": 0.35, "learning_rate": 6.51434395289144e-05, "loss": 1.0193, "step": 139565 }, { "epoch": 0.35, "learning_rate": 6.514218128743269e-05, "loss": 1.0204, "step": 139570 }, { "epoch": 0.35, "learning_rate": 6.514092304595098e-05, "loss": 1.0197, "step": 139575 }, { "epoch": 0.35, "learning_rate": 6.513966480446927e-05, "loss": 1.0222, "step": 139580 }, { "epoch": 0.35, "learning_rate": 6.513840656298758e-05, "loss": 1.0173, "step": 139585 }, { "epoch": 0.35, "learning_rate": 6.513714832150587e-05, "loss": 1.0214, "step": 139590 }, { "epoch": 0.35, "learning_rate": 6.513589008002416e-05, "loss": 1.0195, "step": 139595 }, { "epoch": 0.35, "learning_rate": 6.513463183854245e-05, "loss": 1.0201, "step": 139600 }, { "epoch": 0.35, "learning_rate": 6.513337359706076e-05, "loss": 1.0211, "step": 139605 }, { "epoch": 0.35, "learning_rate": 6.513211535557905e-05, "loss": 1.0198, "step": 139610 }, { "epoch": 0.35, "learning_rate": 6.513085711409734e-05, "loss": 1.0204, "step": 139615 }, { "epoch": 0.35, "learning_rate": 6.512959887261563e-05, "loss": 1.022, "step": 139620 }, { "epoch": 0.35, "learning_rate": 6.512834063113393e-05, "loss": 1.0221, "step": 139625 }, { "epoch": 0.35, "learning_rate": 6.512708238965223e-05, "loss": 1.0187, "step": 139630 }, { "epoch": 0.35, "learning_rate": 6.512582414817052e-05, "loss": 1.0201, "step": 139635 }, { "epoch": 0.35, "learning_rate": 6.512456590668881e-05, "loss": 1.0201, "step": 139640 }, { "epoch": 0.35, "learning_rate": 6.51233076652071e-05, "loss": 1.0204, "step": 139645 }, { "epoch": 0.35, "learning_rate": 6.512204942372541e-05, "loss": 1.0227, "step": 139650 }, { "epoch": 0.35, "learning_rate": 6.51207911822437e-05, "loss": 1.02, "step": 139655 }, { "epoch": 0.35, "learning_rate": 6.5119532940762e-05, "loss": 1.0199, "step": 139660 }, { "epoch": 0.35, "learning_rate": 6.511827469928029e-05, "loss": 1.0226, "step": 139665 }, { "epoch": 0.35, "learning_rate": 6.511701645779859e-05, "loss": 1.0226, "step": 139670 }, { "epoch": 0.35, "learning_rate": 6.511575821631688e-05, "loss": 1.0213, "step": 139675 }, { "epoch": 0.35, "learning_rate": 6.511449997483517e-05, "loss": 1.022, "step": 139680 }, { "epoch": 0.35, "learning_rate": 6.511324173335347e-05, "loss": 1.0202, "step": 139685 }, { "epoch": 0.35, "learning_rate": 6.511198349187176e-05, "loss": 1.0212, "step": 139690 }, { "epoch": 0.35, "learning_rate": 6.511072525039006e-05, "loss": 1.0162, "step": 139695 }, { "epoch": 0.35, "learning_rate": 6.510946700890835e-05, "loss": 1.0222, "step": 139700 }, { "epoch": 0.35, "learning_rate": 6.510820876742665e-05, "loss": 1.021, "step": 139705 }, { "epoch": 0.35, "learning_rate": 6.510695052594494e-05, "loss": 1.0195, "step": 139710 }, { "epoch": 0.35, "learning_rate": 6.510569228446324e-05, "loss": 1.0222, "step": 139715 }, { "epoch": 0.35, "learning_rate": 6.510443404298153e-05, "loss": 1.0216, "step": 139720 }, { "epoch": 0.35, "learning_rate": 6.510317580149983e-05, "loss": 1.0206, "step": 139725 }, { "epoch": 0.35, "learning_rate": 6.510191756001812e-05, "loss": 1.0203, "step": 139730 }, { "epoch": 0.35, "learning_rate": 6.510065931853642e-05, "loss": 1.0209, "step": 139735 }, { "epoch": 0.35, "learning_rate": 6.509940107705471e-05, "loss": 1.0223, "step": 139740 }, { "epoch": 0.35, "learning_rate": 6.5098142835573e-05, "loss": 1.022, "step": 139745 }, { "epoch": 0.35, "learning_rate": 6.50968845940913e-05, "loss": 1.0217, "step": 139750 }, { "epoch": 0.35, "learning_rate": 6.509562635260959e-05, "loss": 1.0228, "step": 139755 }, { "epoch": 0.35, "learning_rate": 6.50943681111279e-05, "loss": 1.019, "step": 139760 }, { "epoch": 0.35, "learning_rate": 6.509310986964619e-05, "loss": 1.0211, "step": 139765 }, { "epoch": 0.35, "learning_rate": 6.509185162816448e-05, "loss": 1.0209, "step": 139770 }, { "epoch": 0.35, "learning_rate": 6.509059338668277e-05, "loss": 1.0208, "step": 139775 }, { "epoch": 0.35, "learning_rate": 6.508933514520107e-05, "loss": 1.0207, "step": 139780 }, { "epoch": 0.35, "learning_rate": 6.508807690371937e-05, "loss": 1.0226, "step": 139785 }, { "epoch": 0.35, "learning_rate": 6.508681866223766e-05, "loss": 1.0183, "step": 139790 }, { "epoch": 0.35, "learning_rate": 6.508556042075595e-05, "loss": 1.0196, "step": 139795 }, { "epoch": 0.35, "learning_rate": 6.508430217927425e-05, "loss": 1.0209, "step": 139800 }, { "epoch": 0.35, "learning_rate": 6.508304393779255e-05, "loss": 1.0224, "step": 139805 }, { "epoch": 0.35, "learning_rate": 6.508178569631084e-05, "loss": 1.021, "step": 139810 }, { "epoch": 0.35, "learning_rate": 6.508052745482913e-05, "loss": 1.0167, "step": 139815 }, { "epoch": 0.35, "learning_rate": 6.507926921334742e-05, "loss": 1.0205, "step": 139820 }, { "epoch": 0.35, "learning_rate": 6.507801097186573e-05, "loss": 1.0191, "step": 139825 }, { "epoch": 0.35, "learning_rate": 6.507675273038402e-05, "loss": 1.0216, "step": 139830 }, { "epoch": 0.35, "learning_rate": 6.507549448890231e-05, "loss": 1.0186, "step": 139835 }, { "epoch": 0.35, "learning_rate": 6.50742362474206e-05, "loss": 1.0403, "step": 139840 }, { "epoch": 0.35, "learning_rate": 6.50729780059389e-05, "loss": 1.0209, "step": 139845 }, { "epoch": 0.35, "learning_rate": 6.50717197644572e-05, "loss": 1.0227, "step": 139850 }, { "epoch": 0.35, "learning_rate": 6.507046152297549e-05, "loss": 1.0205, "step": 139855 }, { "epoch": 0.35, "learning_rate": 6.506920328149378e-05, "loss": 1.0182, "step": 139860 }, { "epoch": 0.35, "learning_rate": 6.506794504001209e-05, "loss": 1.0183, "step": 139865 }, { "epoch": 0.35, "learning_rate": 6.506668679853038e-05, "loss": 1.0206, "step": 139870 }, { "epoch": 0.35, "learning_rate": 6.506542855704867e-05, "loss": 1.0218, "step": 139875 }, { "epoch": 0.35, "learning_rate": 6.506417031556696e-05, "loss": 1.0184, "step": 139880 }, { "epoch": 0.35, "learning_rate": 6.506291207408525e-05, "loss": 1.0192, "step": 139885 }, { "epoch": 0.35, "learning_rate": 6.506165383260356e-05, "loss": 1.0204, "step": 139890 }, { "epoch": 0.35, "learning_rate": 6.506039559112185e-05, "loss": 1.0195, "step": 139895 }, { "epoch": 0.35, "learning_rate": 6.505913734964014e-05, "loss": 1.0196, "step": 139900 }, { "epoch": 0.35, "learning_rate": 6.505787910815843e-05, "loss": 1.0213, "step": 139905 }, { "epoch": 0.35, "learning_rate": 6.505662086667674e-05, "loss": 1.0202, "step": 139910 }, { "epoch": 0.35, "learning_rate": 6.505536262519503e-05, "loss": 1.0193, "step": 139915 }, { "epoch": 0.35, "learning_rate": 6.505410438371332e-05, "loss": 1.021, "step": 139920 }, { "epoch": 0.35, "learning_rate": 6.505284614223161e-05, "loss": 1.0193, "step": 139925 }, { "epoch": 0.35, "learning_rate": 6.505158790074992e-05, "loss": 1.0192, "step": 139930 }, { "epoch": 0.35, "learning_rate": 6.505032965926821e-05, "loss": 1.0205, "step": 139935 }, { "epoch": 0.35, "learning_rate": 6.50490714177865e-05, "loss": 1.0197, "step": 139940 }, { "epoch": 0.35, "learning_rate": 6.504781317630479e-05, "loss": 1.0191, "step": 139945 }, { "epoch": 0.35, "learning_rate": 6.504655493482308e-05, "loss": 1.0205, "step": 139950 }, { "epoch": 0.35, "learning_rate": 6.504529669334139e-05, "loss": 1.0208, "step": 139955 }, { "epoch": 0.35, "learning_rate": 6.504403845185968e-05, "loss": 1.0209, "step": 139960 }, { "epoch": 0.35, "learning_rate": 6.504278021037797e-05, "loss": 1.0204, "step": 139965 }, { "epoch": 0.35, "learning_rate": 6.504152196889626e-05, "loss": 1.0196, "step": 139970 }, { "epoch": 0.35, "learning_rate": 6.504026372741457e-05, "loss": 1.0222, "step": 139975 }, { "epoch": 0.35, "learning_rate": 6.503900548593286e-05, "loss": 1.0208, "step": 139980 }, { "epoch": 0.35, "learning_rate": 6.503774724445115e-05, "loss": 1.0197, "step": 139985 }, { "epoch": 0.35, "learning_rate": 6.503648900296946e-05, "loss": 1.0223, "step": 139990 }, { "epoch": 0.35, "learning_rate": 6.503523076148775e-05, "loss": 1.0195, "step": 139995 }, { "epoch": 0.35, "learning_rate": 6.503397252000605e-05, "loss": 1.0205, "step": 140000 }, { "epoch": 0.35, "learning_rate": 6.503271427852435e-05, "loss": 1.0219, "step": 140005 }, { "epoch": 0.35, "learning_rate": 6.503145603704264e-05, "loss": 1.0198, "step": 140010 }, { "epoch": 0.35, "learning_rate": 6.503019779556093e-05, "loss": 1.0199, "step": 140015 }, { "epoch": 0.35, "learning_rate": 6.502893955407923e-05, "loss": 1.0214, "step": 140020 }, { "epoch": 0.35, "learning_rate": 6.502768131259752e-05, "loss": 1.0209, "step": 140025 }, { "epoch": 0.35, "learning_rate": 6.502642307111582e-05, "loss": 1.0208, "step": 140030 }, { "epoch": 0.35, "learning_rate": 6.502516482963411e-05, "loss": 1.0224, "step": 140035 }, { "epoch": 0.35, "learning_rate": 6.50239065881524e-05, "loss": 1.0206, "step": 140040 }, { "epoch": 0.35, "learning_rate": 6.50226483466707e-05, "loss": 1.0204, "step": 140045 }, { "epoch": 0.35, "learning_rate": 6.5021390105189e-05, "loss": 1.0189, "step": 140050 }, { "epoch": 0.35, "learning_rate": 6.502013186370729e-05, "loss": 1.0204, "step": 140055 }, { "epoch": 0.35, "learning_rate": 6.501887362222558e-05, "loss": 1.0189, "step": 140060 }, { "epoch": 0.35, "learning_rate": 6.501761538074388e-05, "loss": 1.0219, "step": 140065 }, { "epoch": 0.35, "learning_rate": 6.501635713926218e-05, "loss": 1.0212, "step": 140070 }, { "epoch": 0.35, "learning_rate": 6.501509889778047e-05, "loss": 1.0208, "step": 140075 }, { "epoch": 0.35, "learning_rate": 6.501384065629876e-05, "loss": 1.0223, "step": 140080 }, { "epoch": 0.35, "learning_rate": 6.501258241481706e-05, "loss": 1.0212, "step": 140085 }, { "epoch": 0.35, "learning_rate": 6.501132417333536e-05, "loss": 1.0219, "step": 140090 }, { "epoch": 0.35, "learning_rate": 6.501006593185365e-05, "loss": 1.0224, "step": 140095 }, { "epoch": 0.35, "learning_rate": 6.500880769037194e-05, "loss": 1.0221, "step": 140100 }, { "epoch": 0.35, "learning_rate": 6.500754944889023e-05, "loss": 1.022, "step": 140105 }, { "epoch": 0.35, "learning_rate": 6.500629120740854e-05, "loss": 1.0216, "step": 140110 }, { "epoch": 0.35, "learning_rate": 6.500503296592683e-05, "loss": 1.0213, "step": 140115 }, { "epoch": 0.35, "learning_rate": 6.500377472444512e-05, "loss": 1.0212, "step": 140120 }, { "epoch": 0.35, "learning_rate": 6.500251648296341e-05, "loss": 1.0218, "step": 140125 }, { "epoch": 0.35, "learning_rate": 6.500125824148172e-05, "loss": 1.0193, "step": 140130 }, { "epoch": 0.35, "learning_rate": 6.500000000000001e-05, "loss": 1.0238, "step": 140135 }, { "epoch": 0.35, "learning_rate": 6.49987417585183e-05, "loss": 1.0235, "step": 140140 }, { "epoch": 0.35, "learning_rate": 6.499748351703659e-05, "loss": 1.0217, "step": 140145 }, { "epoch": 0.35, "learning_rate": 6.49962252755549e-05, "loss": 1.0209, "step": 140150 }, { "epoch": 0.35, "learning_rate": 6.499496703407319e-05, "loss": 1.0223, "step": 140155 }, { "epoch": 0.35, "learning_rate": 6.499370879259148e-05, "loss": 1.0192, "step": 140160 }, { "epoch": 0.35, "learning_rate": 6.499245055110977e-05, "loss": 1.0189, "step": 140165 }, { "epoch": 0.35, "learning_rate": 6.499119230962806e-05, "loss": 1.019, "step": 140170 }, { "epoch": 0.35, "learning_rate": 6.498993406814637e-05, "loss": 1.0217, "step": 140175 }, { "epoch": 0.35, "learning_rate": 6.498867582666466e-05, "loss": 1.0199, "step": 140180 }, { "epoch": 0.35, "learning_rate": 6.498741758518295e-05, "loss": 1.0216, "step": 140185 }, { "epoch": 0.35, "learning_rate": 6.498615934370124e-05, "loss": 1.0193, "step": 140190 }, { "epoch": 0.35, "learning_rate": 6.498490110221955e-05, "loss": 1.0228, "step": 140195 }, { "epoch": 0.35, "learning_rate": 6.498364286073784e-05, "loss": 1.0216, "step": 140200 }, { "epoch": 0.35, "learning_rate": 6.498238461925613e-05, "loss": 1.02, "step": 140205 }, { "epoch": 0.35, "learning_rate": 6.498112637777442e-05, "loss": 1.0206, "step": 140210 }, { "epoch": 0.35, "learning_rate": 6.497986813629273e-05, "loss": 1.0205, "step": 140215 }, { "epoch": 0.35, "learning_rate": 6.497860989481102e-05, "loss": 1.0326, "step": 140220 }, { "epoch": 0.35, "learning_rate": 6.497735165332931e-05, "loss": 1.0215, "step": 140225 }, { "epoch": 0.35, "learning_rate": 6.49760934118476e-05, "loss": 1.0209, "step": 140230 }, { "epoch": 0.35, "learning_rate": 6.49748351703659e-05, "loss": 1.0197, "step": 140235 }, { "epoch": 0.35, "learning_rate": 6.49735769288842e-05, "loss": 1.0211, "step": 140240 }, { "epoch": 0.35, "learning_rate": 6.497231868740249e-05, "loss": 1.0247, "step": 140245 }, { "epoch": 0.35, "learning_rate": 6.497106044592078e-05, "loss": 1.043, "step": 140250 }, { "epoch": 0.35, "learning_rate": 6.496980220443907e-05, "loss": 1.02, "step": 140255 }, { "epoch": 0.35, "learning_rate": 6.496854396295738e-05, "loss": 1.0209, "step": 140260 }, { "epoch": 0.35, "learning_rate": 6.496728572147567e-05, "loss": 1.0222, "step": 140265 }, { "epoch": 0.35, "learning_rate": 6.496602747999396e-05, "loss": 1.0443, "step": 140270 }, { "epoch": 0.35, "learning_rate": 6.496476923851225e-05, "loss": 1.0217, "step": 140275 }, { "epoch": 0.35, "learning_rate": 6.496351099703055e-05, "loss": 1.0222, "step": 140280 }, { "epoch": 0.35, "learning_rate": 6.496225275554885e-05, "loss": 1.0216, "step": 140285 }, { "epoch": 0.35, "learning_rate": 6.496099451406714e-05, "loss": 1.0216, "step": 140290 }, { "epoch": 0.35, "learning_rate": 6.495973627258543e-05, "loss": 1.0218, "step": 140295 }, { "epoch": 0.35, "learning_rate": 6.495847803110373e-05, "loss": 1.0199, "step": 140300 }, { "epoch": 0.35, "learning_rate": 6.495721978962203e-05, "loss": 1.0216, "step": 140305 }, { "epoch": 0.35, "learning_rate": 6.495596154814032e-05, "loss": 1.0204, "step": 140310 }, { "epoch": 0.35, "learning_rate": 6.495470330665861e-05, "loss": 1.02, "step": 140315 }, { "epoch": 0.35, "learning_rate": 6.49534450651769e-05, "loss": 1.0192, "step": 140320 }, { "epoch": 0.35, "learning_rate": 6.495218682369521e-05, "loss": 1.0223, "step": 140325 }, { "epoch": 0.35, "learning_rate": 6.49509285822135e-05, "loss": 1.0207, "step": 140330 }, { "epoch": 0.35, "learning_rate": 6.49496703407318e-05, "loss": 1.0195, "step": 140335 }, { "epoch": 0.35, "learning_rate": 6.494841209925009e-05, "loss": 1.0223, "step": 140340 }, { "epoch": 0.35, "learning_rate": 6.494715385776838e-05, "loss": 1.0208, "step": 140345 }, { "epoch": 0.35, "learning_rate": 6.494589561628668e-05, "loss": 1.0216, "step": 140350 }, { "epoch": 0.35, "learning_rate": 6.494463737480497e-05, "loss": 1.0212, "step": 140355 }, { "epoch": 0.35, "learning_rate": 6.494337913332327e-05, "loss": 1.023, "step": 140360 }, { "epoch": 0.35, "learning_rate": 6.494212089184156e-05, "loss": 1.0237, "step": 140365 }, { "epoch": 0.35, "learning_rate": 6.494086265035986e-05, "loss": 1.02, "step": 140370 }, { "epoch": 0.35, "learning_rate": 6.493960440887815e-05, "loss": 1.0203, "step": 140375 }, { "epoch": 0.35, "learning_rate": 6.493834616739645e-05, "loss": 1.021, "step": 140380 }, { "epoch": 0.35, "learning_rate": 6.493708792591474e-05, "loss": 1.0197, "step": 140385 }, { "epoch": 0.35, "learning_rate": 6.493582968443304e-05, "loss": 1.0217, "step": 140390 }, { "epoch": 0.35, "learning_rate": 6.493457144295133e-05, "loss": 1.0205, "step": 140395 }, { "epoch": 0.35, "learning_rate": 6.493331320146963e-05, "loss": 1.0208, "step": 140400 }, { "epoch": 0.35, "learning_rate": 6.493205495998792e-05, "loss": 1.0216, "step": 140405 }, { "epoch": 0.35, "learning_rate": 6.493079671850621e-05, "loss": 1.0203, "step": 140410 }, { "epoch": 0.35, "learning_rate": 6.492953847702451e-05, "loss": 1.0196, "step": 140415 }, { "epoch": 0.35, "learning_rate": 6.49282802355428e-05, "loss": 1.0231, "step": 140420 }, { "epoch": 0.35, "learning_rate": 6.49270219940611e-05, "loss": 1.0189, "step": 140425 }, { "epoch": 0.35, "learning_rate": 6.492576375257939e-05, "loss": 1.0192, "step": 140430 }, { "epoch": 0.35, "learning_rate": 6.49245055110977e-05, "loss": 1.0199, "step": 140435 }, { "epoch": 0.35, "learning_rate": 6.492324726961599e-05, "loss": 1.0206, "step": 140440 }, { "epoch": 0.35, "learning_rate": 6.492198902813428e-05, "loss": 1.0222, "step": 140445 }, { "epoch": 0.35, "learning_rate": 6.492073078665257e-05, "loss": 1.0214, "step": 140450 }, { "epoch": 0.35, "learning_rate": 6.491972419346722e-05, "loss": 1.0219, "step": 140455 }, { "epoch": 0.35, "learning_rate": 6.491846595198551e-05, "loss": 1.0217, "step": 140460 }, { "epoch": 0.35, "learning_rate": 6.49172077105038e-05, "loss": 1.0199, "step": 140465 }, { "epoch": 0.35, "learning_rate": 6.491594946902209e-05, "loss": 1.0215, "step": 140470 }, { "epoch": 0.35, "learning_rate": 6.49146912275404e-05, "loss": 1.0207, "step": 140475 }, { "epoch": 0.35, "learning_rate": 6.491343298605869e-05, "loss": 1.0208, "step": 140480 }, { "epoch": 0.35, "learning_rate": 6.491217474457698e-05, "loss": 1.0204, "step": 140485 }, { "epoch": 0.35, "learning_rate": 6.491091650309527e-05, "loss": 1.0206, "step": 140490 }, { "epoch": 0.35, "learning_rate": 6.490965826161356e-05, "loss": 1.0213, "step": 140495 }, { "epoch": 0.35, "learning_rate": 6.490840002013187e-05, "loss": 1.0214, "step": 140500 }, { "epoch": 0.35, "learning_rate": 6.490714177865016e-05, "loss": 1.02, "step": 140505 }, { "epoch": 0.35, "learning_rate": 6.490588353716845e-05, "loss": 1.0195, "step": 140510 }, { "epoch": 0.35, "learning_rate": 6.490462529568674e-05, "loss": 1.0205, "step": 140515 }, { "epoch": 0.35, "learning_rate": 6.490336705420505e-05, "loss": 1.0211, "step": 140520 }, { "epoch": 0.35, "learning_rate": 6.490210881272334e-05, "loss": 1.0195, "step": 140525 }, { "epoch": 0.35, "learning_rate": 6.490085057124163e-05, "loss": 1.0208, "step": 140530 }, { "epoch": 0.35, "learning_rate": 6.489959232975992e-05, "loss": 1.0207, "step": 140535 }, { "epoch": 0.35, "learning_rate": 6.489833408827823e-05, "loss": 1.0209, "step": 140540 }, { "epoch": 0.35, "learning_rate": 6.489707584679652e-05, "loss": 1.0206, "step": 140545 }, { "epoch": 0.35, "learning_rate": 6.489581760531481e-05, "loss": 1.0211, "step": 140550 }, { "epoch": 0.35, "learning_rate": 6.48945593638331e-05, "loss": 1.02, "step": 140555 }, { "epoch": 0.35, "learning_rate": 6.48933011223514e-05, "loss": 1.0214, "step": 140560 }, { "epoch": 0.35, "learning_rate": 6.48920428808697e-05, "loss": 1.0215, "step": 140565 }, { "epoch": 0.35, "learning_rate": 6.489078463938799e-05, "loss": 1.023, "step": 140570 }, { "epoch": 0.35, "learning_rate": 6.488952639790628e-05, "loss": 1.0199, "step": 140575 }, { "epoch": 0.35, "learning_rate": 6.488826815642458e-05, "loss": 1.02, "step": 140580 }, { "epoch": 0.35, "learning_rate": 6.488700991494288e-05, "loss": 1.0191, "step": 140585 }, { "epoch": 0.35, "learning_rate": 6.488575167346117e-05, "loss": 1.0207, "step": 140590 }, { "epoch": 0.35, "learning_rate": 6.488449343197946e-05, "loss": 1.0211, "step": 140595 }, { "epoch": 0.35, "learning_rate": 6.488323519049776e-05, "loss": 1.0195, "step": 140600 }, { "epoch": 0.35, "learning_rate": 6.488197694901606e-05, "loss": 1.0176, "step": 140605 }, { "epoch": 0.35, "learning_rate": 6.488071870753435e-05, "loss": 1.0205, "step": 140610 }, { "epoch": 0.35, "learning_rate": 6.487946046605264e-05, "loss": 1.0193, "step": 140615 }, { "epoch": 0.35, "learning_rate": 6.487820222457094e-05, "loss": 1.0211, "step": 140620 }, { "epoch": 0.35, "learning_rate": 6.487694398308923e-05, "loss": 1.0217, "step": 140625 }, { "epoch": 0.35, "learning_rate": 6.487568574160753e-05, "loss": 1.0221, "step": 140630 }, { "epoch": 0.35, "learning_rate": 6.487442750012582e-05, "loss": 1.0205, "step": 140635 }, { "epoch": 0.35, "learning_rate": 6.487316925864412e-05, "loss": 1.0208, "step": 140640 }, { "epoch": 0.35, "learning_rate": 6.487191101716241e-05, "loss": 1.0191, "step": 140645 }, { "epoch": 0.35, "learning_rate": 6.487065277568071e-05, "loss": 1.0207, "step": 140650 }, { "epoch": 0.35, "learning_rate": 6.4869394534199e-05, "loss": 1.0217, "step": 140655 }, { "epoch": 0.35, "learning_rate": 6.48681362927173e-05, "loss": 1.0198, "step": 140660 }, { "epoch": 0.35, "learning_rate": 6.486687805123559e-05, "loss": 1.0239, "step": 140665 }, { "epoch": 0.35, "learning_rate": 6.486561980975389e-05, "loss": 1.0205, "step": 140670 }, { "epoch": 0.35, "learning_rate": 6.486436156827218e-05, "loss": 1.0208, "step": 140675 }, { "epoch": 0.35, "learning_rate": 6.486310332679048e-05, "loss": 1.0205, "step": 140680 }, { "epoch": 0.35, "learning_rate": 6.486184508530877e-05, "loss": 1.0235, "step": 140685 }, { "epoch": 0.35, "learning_rate": 6.486058684382707e-05, "loss": 1.0239, "step": 140690 }, { "epoch": 0.35, "learning_rate": 6.485932860234538e-05, "loss": 1.0214, "step": 140695 }, { "epoch": 0.35, "learning_rate": 6.485807036086367e-05, "loss": 1.02, "step": 140700 }, { "epoch": 0.35, "learning_rate": 6.485681211938196e-05, "loss": 1.0226, "step": 140705 }, { "epoch": 0.35, "learning_rate": 6.485555387790025e-05, "loss": 1.02, "step": 140710 }, { "epoch": 0.35, "learning_rate": 6.485429563641854e-05, "loss": 1.0191, "step": 140715 }, { "epoch": 0.35, "learning_rate": 6.485303739493685e-05, "loss": 1.0227, "step": 140720 }, { "epoch": 0.35, "learning_rate": 6.485177915345514e-05, "loss": 1.0207, "step": 140725 }, { "epoch": 0.35, "learning_rate": 6.485052091197343e-05, "loss": 1.0217, "step": 140730 }, { "epoch": 0.35, "learning_rate": 6.484926267049172e-05, "loss": 1.0198, "step": 140735 }, { "epoch": 0.35, "learning_rate": 6.484800442901003e-05, "loss": 1.022, "step": 140740 }, { "epoch": 0.35, "learning_rate": 6.484674618752832e-05, "loss": 1.0207, "step": 140745 }, { "epoch": 0.35, "learning_rate": 6.484548794604661e-05, "loss": 1.0204, "step": 140750 }, { "epoch": 0.35, "learning_rate": 6.48442297045649e-05, "loss": 1.0231, "step": 140755 }, { "epoch": 0.35, "learning_rate": 6.484297146308321e-05, "loss": 1.0205, "step": 140760 }, { "epoch": 0.35, "learning_rate": 6.48417132216015e-05, "loss": 1.0216, "step": 140765 }, { "epoch": 0.35, "learning_rate": 6.484045498011979e-05, "loss": 1.0182, "step": 140770 }, { "epoch": 0.35, "learning_rate": 6.483919673863808e-05, "loss": 1.0193, "step": 140775 }, { "epoch": 0.35, "learning_rate": 6.483793849715637e-05, "loss": 1.0199, "step": 140780 }, { "epoch": 0.35, "learning_rate": 6.483668025567468e-05, "loss": 1.0213, "step": 140785 }, { "epoch": 0.35, "learning_rate": 6.483542201419297e-05, "loss": 1.0246, "step": 140790 }, { "epoch": 0.35, "learning_rate": 6.483416377271126e-05, "loss": 1.0228, "step": 140795 }, { "epoch": 0.35, "learning_rate": 6.483290553122955e-05, "loss": 1.0191, "step": 140800 }, { "epoch": 0.35, "learning_rate": 6.483164728974786e-05, "loss": 1.0227, "step": 140805 }, { "epoch": 0.35, "learning_rate": 6.483038904826615e-05, "loss": 1.0223, "step": 140810 }, { "epoch": 0.35, "learning_rate": 6.482913080678444e-05, "loss": 1.0189, "step": 140815 }, { "epoch": 0.35, "learning_rate": 6.482787256530273e-05, "loss": 1.0213, "step": 140820 }, { "epoch": 0.35, "learning_rate": 6.482661432382104e-05, "loss": 1.0213, "step": 140825 }, { "epoch": 0.35, "learning_rate": 6.482535608233933e-05, "loss": 1.0211, "step": 140830 }, { "epoch": 0.35, "learning_rate": 6.482409784085762e-05, "loss": 1.0213, "step": 140835 }, { "epoch": 0.35, "learning_rate": 6.482283959937591e-05, "loss": 1.0195, "step": 140840 }, { "epoch": 0.35, "learning_rate": 6.48215813578942e-05, "loss": 1.0199, "step": 140845 }, { "epoch": 0.35, "learning_rate": 6.482032311641251e-05, "loss": 1.0217, "step": 140850 }, { "epoch": 0.35, "learning_rate": 6.48190648749308e-05, "loss": 1.0203, "step": 140855 }, { "epoch": 0.35, "learning_rate": 6.48178066334491e-05, "loss": 1.0208, "step": 140860 }, { "epoch": 0.35, "learning_rate": 6.481654839196739e-05, "loss": 1.0206, "step": 140865 }, { "epoch": 0.35, "learning_rate": 6.481529015048569e-05, "loss": 1.0207, "step": 140870 }, { "epoch": 0.35, "learning_rate": 6.481403190900398e-05, "loss": 1.0214, "step": 140875 }, { "epoch": 0.35, "learning_rate": 6.481277366752227e-05, "loss": 1.0204, "step": 140880 }, { "epoch": 0.35, "learning_rate": 6.481151542604057e-05, "loss": 1.0222, "step": 140885 }, { "epoch": 0.35, "learning_rate": 6.481025718455887e-05, "loss": 1.0215, "step": 140890 }, { "epoch": 0.35, "learning_rate": 6.480899894307716e-05, "loss": 1.0203, "step": 140895 }, { "epoch": 0.35, "learning_rate": 6.480774070159545e-05, "loss": 1.0222, "step": 140900 }, { "epoch": 0.35, "learning_rate": 6.480648246011375e-05, "loss": 1.0197, "step": 140905 }, { "epoch": 0.35, "learning_rate": 6.480522421863204e-05, "loss": 1.0196, "step": 140910 }, { "epoch": 0.35, "learning_rate": 6.480396597715034e-05, "loss": 1.0436, "step": 140915 }, { "epoch": 0.35, "learning_rate": 6.480270773566863e-05, "loss": 1.0219, "step": 140920 }, { "epoch": 0.35, "learning_rate": 6.480144949418693e-05, "loss": 1.0211, "step": 140925 }, { "epoch": 0.35, "learning_rate": 6.480019125270522e-05, "loss": 1.0228, "step": 140930 }, { "epoch": 0.35, "learning_rate": 6.479893301122352e-05, "loss": 1.0184, "step": 140935 }, { "epoch": 0.35, "learning_rate": 6.479767476974181e-05, "loss": 1.02, "step": 140940 }, { "epoch": 0.35, "learning_rate": 6.47964165282601e-05, "loss": 1.0197, "step": 140945 }, { "epoch": 0.35, "learning_rate": 6.47951582867784e-05, "loss": 1.0203, "step": 140950 }, { "epoch": 0.35, "learning_rate": 6.479390004529669e-05, "loss": 1.0215, "step": 140955 }, { "epoch": 0.35, "learning_rate": 6.4792641803815e-05, "loss": 1.0208, "step": 140960 }, { "epoch": 0.35, "learning_rate": 6.479138356233329e-05, "loss": 1.0242, "step": 140965 }, { "epoch": 0.35, "learning_rate": 6.479012532085158e-05, "loss": 1.02, "step": 140970 }, { "epoch": 0.35, "learning_rate": 6.478886707936987e-05, "loss": 1.0217, "step": 140975 }, { "epoch": 0.35, "learning_rate": 6.478760883788817e-05, "loss": 1.0205, "step": 140980 }, { "epoch": 0.35, "learning_rate": 6.478635059640647e-05, "loss": 1.0202, "step": 140985 }, { "epoch": 0.35, "learning_rate": 6.478509235492476e-05, "loss": 1.0195, "step": 140990 }, { "epoch": 0.35, "learning_rate": 6.478383411344305e-05, "loss": 1.0209, "step": 140995 }, { "epoch": 0.35, "learning_rate": 6.478257587196135e-05, "loss": 1.0387, "step": 141000 }, { "epoch": 0.35, "learning_rate": 6.478131763047965e-05, "loss": 1.0191, "step": 141005 }, { "epoch": 0.35, "learning_rate": 6.478005938899794e-05, "loss": 1.0206, "step": 141010 }, { "epoch": 0.35, "learning_rate": 6.477880114751623e-05, "loss": 1.0476, "step": 141015 }, { "epoch": 0.35, "learning_rate": 6.477754290603452e-05, "loss": 1.0216, "step": 141020 }, { "epoch": 0.35, "learning_rate": 6.477628466455283e-05, "loss": 1.0212, "step": 141025 }, { "epoch": 0.35, "learning_rate": 6.477502642307112e-05, "loss": 1.0209, "step": 141030 }, { "epoch": 0.35, "learning_rate": 6.477376818158941e-05, "loss": 1.0202, "step": 141035 }, { "epoch": 0.35, "learning_rate": 6.47725099401077e-05, "loss": 1.0249, "step": 141040 }, { "epoch": 0.35, "learning_rate": 6.4771251698626e-05, "loss": 1.0434, "step": 141045 }, { "epoch": 0.35, "learning_rate": 6.47699934571443e-05, "loss": 1.0201, "step": 141050 }, { "epoch": 0.35, "learning_rate": 6.476873521566259e-05, "loss": 1.0197, "step": 141055 }, { "epoch": 0.35, "learning_rate": 6.476747697418088e-05, "loss": 1.0197, "step": 141060 }, { "epoch": 0.35, "learning_rate": 6.476621873269919e-05, "loss": 1.0213, "step": 141065 }, { "epoch": 0.35, "learning_rate": 6.476496049121748e-05, "loss": 1.0226, "step": 141070 }, { "epoch": 0.35, "learning_rate": 6.476370224973577e-05, "loss": 1.0191, "step": 141075 }, { "epoch": 0.35, "learning_rate": 6.476244400825406e-05, "loss": 1.0219, "step": 141080 }, { "epoch": 0.35, "learning_rate": 6.476118576677235e-05, "loss": 1.0213, "step": 141085 }, { "epoch": 0.35, "learning_rate": 6.475992752529066e-05, "loss": 1.0209, "step": 141090 }, { "epoch": 0.35, "learning_rate": 6.475866928380895e-05, "loss": 1.0192, "step": 141095 }, { "epoch": 0.35, "learning_rate": 6.475741104232724e-05, "loss": 1.0203, "step": 141100 }, { "epoch": 0.35, "learning_rate": 6.475615280084553e-05, "loss": 1.044, "step": 141105 }, { "epoch": 0.35, "learning_rate": 6.475489455936384e-05, "loss": 1.0195, "step": 141110 }, { "epoch": 0.35, "learning_rate": 6.475363631788213e-05, "loss": 1.0226, "step": 141115 }, { "epoch": 0.35, "learning_rate": 6.475237807640042e-05, "loss": 1.0205, "step": 141120 }, { "epoch": 0.35, "learning_rate": 6.475111983491871e-05, "loss": 1.0193, "step": 141125 }, { "epoch": 0.35, "learning_rate": 6.474986159343702e-05, "loss": 1.0205, "step": 141130 }, { "epoch": 0.35, "learning_rate": 6.474860335195531e-05, "loss": 1.0206, "step": 141135 }, { "epoch": 0.35, "learning_rate": 6.47473451104736e-05, "loss": 1.0224, "step": 141140 }, { "epoch": 0.35, "learning_rate": 6.474608686899189e-05, "loss": 1.0212, "step": 141145 }, { "epoch": 0.35, "learning_rate": 6.474482862751018e-05, "loss": 1.0201, "step": 141150 }, { "epoch": 0.35, "learning_rate": 6.474357038602849e-05, "loss": 1.0191, "step": 141155 }, { "epoch": 0.35, "learning_rate": 6.474231214454678e-05, "loss": 1.0204, "step": 141160 }, { "epoch": 0.35, "learning_rate": 6.474105390306507e-05, "loss": 1.0215, "step": 141165 }, { "epoch": 0.35, "learning_rate": 6.473979566158336e-05, "loss": 1.0194, "step": 141170 }, { "epoch": 0.35, "learning_rate": 6.473853742010167e-05, "loss": 1.0173, "step": 141175 }, { "epoch": 0.35, "learning_rate": 6.473727917861996e-05, "loss": 1.021, "step": 141180 }, { "epoch": 0.35, "learning_rate": 6.473602093713827e-05, "loss": 1.0209, "step": 141185 }, { "epoch": 0.35, "learning_rate": 6.473476269565656e-05, "loss": 1.0217, "step": 141190 }, { "epoch": 0.35, "learning_rate": 6.473350445417485e-05, "loss": 1.0204, "step": 141195 }, { "epoch": 0.35, "learning_rate": 6.473224621269315e-05, "loss": 1.0187, "step": 141200 }, { "epoch": 0.35, "learning_rate": 6.473098797121145e-05, "loss": 1.0201, "step": 141205 }, { "epoch": 0.35, "learning_rate": 6.472972972972974e-05, "loss": 1.0223, "step": 141210 }, { "epoch": 0.35, "learning_rate": 6.472847148824803e-05, "loss": 1.0196, "step": 141215 }, { "epoch": 0.35, "learning_rate": 6.472721324676633e-05, "loss": 1.0209, "step": 141220 }, { "epoch": 0.35, "learning_rate": 6.472595500528463e-05, "loss": 1.0229, "step": 141225 }, { "epoch": 0.35, "learning_rate": 6.472469676380292e-05, "loss": 1.0204, "step": 141230 }, { "epoch": 0.35, "learning_rate": 6.472343852232121e-05, "loss": 1.018, "step": 141235 }, { "epoch": 0.35, "learning_rate": 6.47221802808395e-05, "loss": 1.0232, "step": 141240 }, { "epoch": 0.35, "learning_rate": 6.47209220393578e-05, "loss": 1.0221, "step": 141245 }, { "epoch": 0.35, "learning_rate": 6.47196637978761e-05, "loss": 1.0226, "step": 141250 }, { "epoch": 0.35, "learning_rate": 6.471840555639439e-05, "loss": 1.0186, "step": 141255 }, { "epoch": 0.35, "learning_rate": 6.471714731491268e-05, "loss": 1.0206, "step": 141260 }, { "epoch": 0.35, "learning_rate": 6.471588907343099e-05, "loss": 1.0205, "step": 141265 }, { "epoch": 0.35, "learning_rate": 6.471463083194928e-05, "loss": 1.0201, "step": 141270 }, { "epoch": 0.35, "learning_rate": 6.471337259046757e-05, "loss": 1.0231, "step": 141275 }, { "epoch": 0.35, "learning_rate": 6.471211434898586e-05, "loss": 1.0199, "step": 141280 }, { "epoch": 0.35, "learning_rate": 6.471085610750417e-05, "loss": 1.0215, "step": 141285 }, { "epoch": 0.35, "learning_rate": 6.470959786602246e-05, "loss": 1.0207, "step": 141290 }, { "epoch": 0.35, "learning_rate": 6.470833962454075e-05, "loss": 1.0192, "step": 141295 }, { "epoch": 0.35, "learning_rate": 6.470708138305904e-05, "loss": 1.0212, "step": 141300 }, { "epoch": 0.35, "learning_rate": 6.470582314157733e-05, "loss": 1.0215, "step": 141305 }, { "epoch": 0.35, "learning_rate": 6.470456490009564e-05, "loss": 1.0199, "step": 141310 }, { "epoch": 0.35, "learning_rate": 6.470330665861393e-05, "loss": 1.0417, "step": 141315 }, { "epoch": 0.35, "learning_rate": 6.470204841713222e-05, "loss": 1.0219, "step": 141320 }, { "epoch": 0.35, "learning_rate": 6.470079017565051e-05, "loss": 1.0219, "step": 141325 }, { "epoch": 0.35, "learning_rate": 6.469953193416882e-05, "loss": 1.0201, "step": 141330 }, { "epoch": 0.35, "learning_rate": 6.469827369268711e-05, "loss": 1.0192, "step": 141335 }, { "epoch": 0.35, "learning_rate": 6.46970154512054e-05, "loss": 1.0189, "step": 141340 }, { "epoch": 0.35, "learning_rate": 6.469575720972369e-05, "loss": 1.0211, "step": 141345 }, { "epoch": 0.35, "learning_rate": 6.4694498968242e-05, "loss": 1.0223, "step": 141350 }, { "epoch": 0.35, "learning_rate": 6.469324072676029e-05, "loss": 1.0225, "step": 141355 }, { "epoch": 0.35, "learning_rate": 6.469198248527858e-05, "loss": 1.0198, "step": 141360 }, { "epoch": 0.35, "learning_rate": 6.469072424379687e-05, "loss": 1.0202, "step": 141365 }, { "epoch": 0.35, "learning_rate": 6.468946600231516e-05, "loss": 1.018, "step": 141370 }, { "epoch": 0.35, "learning_rate": 6.468820776083347e-05, "loss": 1.0228, "step": 141375 }, { "epoch": 0.35, "learning_rate": 6.468694951935176e-05, "loss": 1.02, "step": 141380 }, { "epoch": 0.35, "learning_rate": 6.468569127787005e-05, "loss": 1.0193, "step": 141385 }, { "epoch": 0.35, "learning_rate": 6.468443303638834e-05, "loss": 1.0201, "step": 141390 }, { "epoch": 0.35, "learning_rate": 6.468317479490665e-05, "loss": 1.0224, "step": 141395 }, { "epoch": 0.35, "learning_rate": 6.468191655342494e-05, "loss": 1.0201, "step": 141400 }, { "epoch": 0.35, "learning_rate": 6.468065831194323e-05, "loss": 1.0229, "step": 141405 }, { "epoch": 0.35, "learning_rate": 6.467940007046152e-05, "loss": 1.0203, "step": 141410 }, { "epoch": 0.35, "learning_rate": 6.467814182897983e-05, "loss": 1.02, "step": 141415 }, { "epoch": 0.35, "learning_rate": 6.467688358749812e-05, "loss": 1.023, "step": 141420 }, { "epoch": 0.35, "learning_rate": 6.467562534601641e-05, "loss": 1.0201, "step": 141425 }, { "epoch": 0.36, "learning_rate": 6.46743671045347e-05, "loss": 1.0204, "step": 141430 }, { "epoch": 0.36, "learning_rate": 6.4673108863053e-05, "loss": 1.021, "step": 141435 }, { "epoch": 0.36, "learning_rate": 6.46718506215713e-05, "loss": 1.0203, "step": 141440 }, { "epoch": 0.36, "learning_rate": 6.467059238008959e-05, "loss": 1.0222, "step": 141445 }, { "epoch": 0.36, "learning_rate": 6.466933413860788e-05, "loss": 1.0224, "step": 141450 }, { "epoch": 0.36, "learning_rate": 6.466807589712617e-05, "loss": 1.0193, "step": 141455 }, { "epoch": 0.36, "learning_rate": 6.466681765564448e-05, "loss": 1.0204, "step": 141460 }, { "epoch": 0.36, "learning_rate": 6.466555941416277e-05, "loss": 1.0202, "step": 141465 }, { "epoch": 0.36, "learning_rate": 6.466430117268106e-05, "loss": 1.0214, "step": 141470 }, { "epoch": 0.36, "learning_rate": 6.466304293119935e-05, "loss": 1.0213, "step": 141475 }, { "epoch": 0.36, "learning_rate": 6.466178468971766e-05, "loss": 1.0373, "step": 141480 }, { "epoch": 0.36, "learning_rate": 6.466052644823595e-05, "loss": 1.0186, "step": 141485 }, { "epoch": 0.36, "learning_rate": 6.465926820675424e-05, "loss": 1.0416, "step": 141490 }, { "epoch": 0.36, "learning_rate": 6.465800996527253e-05, "loss": 1.0195, "step": 141495 }, { "epoch": 0.36, "learning_rate": 6.465675172379083e-05, "loss": 1.0215, "step": 141500 }, { "epoch": 0.36, "learning_rate": 6.465549348230913e-05, "loss": 1.0206, "step": 141505 }, { "epoch": 0.36, "learning_rate": 6.465423524082742e-05, "loss": 1.0225, "step": 141510 }, { "epoch": 0.36, "learning_rate": 6.465297699934571e-05, "loss": 1.0233, "step": 141515 }, { "epoch": 0.36, "learning_rate": 6.4651718757864e-05, "loss": 1.0204, "step": 141520 }, { "epoch": 0.36, "learning_rate": 6.465046051638231e-05, "loss": 1.0215, "step": 141525 }, { "epoch": 0.36, "learning_rate": 6.46492022749006e-05, "loss": 1.0217, "step": 141530 }, { "epoch": 0.36, "learning_rate": 6.46479440334189e-05, "loss": 1.0191, "step": 141535 }, { "epoch": 0.36, "learning_rate": 6.464668579193719e-05, "loss": 1.0198, "step": 141540 }, { "epoch": 0.36, "learning_rate": 6.464542755045549e-05, "loss": 1.0201, "step": 141545 }, { "epoch": 0.36, "learning_rate": 6.464416930897378e-05, "loss": 1.0211, "step": 141550 }, { "epoch": 0.36, "learning_rate": 6.464291106749207e-05, "loss": 1.0198, "step": 141555 }, { "epoch": 0.36, "learning_rate": 6.464165282601037e-05, "loss": 1.0221, "step": 141560 }, { "epoch": 0.36, "learning_rate": 6.464039458452866e-05, "loss": 1.0236, "step": 141565 }, { "epoch": 0.36, "learning_rate": 6.463913634304696e-05, "loss": 1.0198, "step": 141570 }, { "epoch": 0.36, "learning_rate": 6.463787810156525e-05, "loss": 1.0212, "step": 141575 }, { "epoch": 0.36, "learning_rate": 6.463661986008355e-05, "loss": 1.0189, "step": 141580 }, { "epoch": 0.36, "learning_rate": 6.463536161860184e-05, "loss": 1.0208, "step": 141585 }, { "epoch": 0.36, "learning_rate": 6.463410337712014e-05, "loss": 1.0201, "step": 141590 }, { "epoch": 0.36, "learning_rate": 6.463284513563843e-05, "loss": 1.023, "step": 141595 }, { "epoch": 0.36, "learning_rate": 6.463158689415673e-05, "loss": 1.021, "step": 141600 }, { "epoch": 0.36, "learning_rate": 6.463032865267502e-05, "loss": 1.02, "step": 141605 }, { "epoch": 0.36, "learning_rate": 6.462907041119332e-05, "loss": 1.0189, "step": 141610 }, { "epoch": 0.36, "learning_rate": 6.462781216971161e-05, "loss": 1.021, "step": 141615 }, { "epoch": 0.36, "learning_rate": 6.46265539282299e-05, "loss": 1.0207, "step": 141620 }, { "epoch": 0.36, "learning_rate": 6.46252956867482e-05, "loss": 1.0191, "step": 141625 }, { "epoch": 0.36, "learning_rate": 6.462403744526649e-05, "loss": 1.018, "step": 141630 }, { "epoch": 0.36, "learning_rate": 6.46227792037848e-05, "loss": 1.0215, "step": 141635 }, { "epoch": 0.36, "learning_rate": 6.462152096230309e-05, "loss": 1.0213, "step": 141640 }, { "epoch": 0.36, "learning_rate": 6.462026272082138e-05, "loss": 1.0233, "step": 141645 }, { "epoch": 0.36, "learning_rate": 6.461900447933967e-05, "loss": 1.0204, "step": 141650 }, { "epoch": 0.36, "learning_rate": 6.461774623785797e-05, "loss": 1.0242, "step": 141655 }, { "epoch": 0.36, "learning_rate": 6.461648799637627e-05, "loss": 1.0201, "step": 141660 }, { "epoch": 0.36, "learning_rate": 6.461522975489456e-05, "loss": 1.0216, "step": 141665 }, { "epoch": 0.36, "learning_rate": 6.461397151341285e-05, "loss": 1.019, "step": 141670 }, { "epoch": 0.36, "learning_rate": 6.461271327193114e-05, "loss": 1.0214, "step": 141675 }, { "epoch": 0.36, "learning_rate": 6.461145503044945e-05, "loss": 1.0232, "step": 141680 }, { "epoch": 0.36, "learning_rate": 6.461019678896775e-05, "loss": 1.0202, "step": 141685 }, { "epoch": 0.36, "learning_rate": 6.460893854748604e-05, "loss": 1.0204, "step": 141690 }, { "epoch": 0.36, "learning_rate": 6.460768030600433e-05, "loss": 1.02, "step": 141695 }, { "epoch": 0.36, "learning_rate": 6.460642206452263e-05, "loss": 1.0202, "step": 141700 }, { "epoch": 0.36, "learning_rate": 6.460516382304093e-05, "loss": 1.0183, "step": 141705 }, { "epoch": 0.36, "learning_rate": 6.460390558155922e-05, "loss": 1.0222, "step": 141710 }, { "epoch": 0.36, "learning_rate": 6.460264734007751e-05, "loss": 1.0196, "step": 141715 }, { "epoch": 0.36, "learning_rate": 6.46013890985958e-05, "loss": 1.0208, "step": 141720 }, { "epoch": 0.36, "learning_rate": 6.460013085711411e-05, "loss": 1.0183, "step": 141725 }, { "epoch": 0.36, "learning_rate": 6.45988726156324e-05, "loss": 1.0469, "step": 141730 }, { "epoch": 0.36, "learning_rate": 6.45976143741507e-05, "loss": 1.0196, "step": 141735 }, { "epoch": 0.36, "learning_rate": 6.459635613266899e-05, "loss": 1.0211, "step": 141740 }, { "epoch": 0.36, "learning_rate": 6.459509789118729e-05, "loss": 1.0461, "step": 141745 }, { "epoch": 0.36, "learning_rate": 6.459383964970558e-05, "loss": 1.0211, "step": 141750 }, { "epoch": 0.36, "learning_rate": 6.459258140822387e-05, "loss": 1.0228, "step": 141755 }, { "epoch": 0.36, "learning_rate": 6.459132316674216e-05, "loss": 1.023, "step": 141760 }, { "epoch": 0.36, "learning_rate": 6.459006492526046e-05, "loss": 1.0232, "step": 141765 }, { "epoch": 0.36, "learning_rate": 6.458880668377876e-05, "loss": 1.0217, "step": 141770 }, { "epoch": 0.36, "learning_rate": 6.458754844229705e-05, "loss": 1.0216, "step": 141775 }, { "epoch": 0.36, "learning_rate": 6.458629020081534e-05, "loss": 1.0222, "step": 141780 }, { "epoch": 0.36, "learning_rate": 6.458503195933364e-05, "loss": 1.0215, "step": 141785 }, { "epoch": 0.36, "learning_rate": 6.458377371785194e-05, "loss": 1.0197, "step": 141790 }, { "epoch": 0.36, "learning_rate": 6.458251547637023e-05, "loss": 1.0225, "step": 141795 }, { "epoch": 0.36, "learning_rate": 6.458125723488852e-05, "loss": 1.0198, "step": 141800 }, { "epoch": 0.36, "learning_rate": 6.457999899340682e-05, "loss": 1.0221, "step": 141805 }, { "epoch": 0.36, "learning_rate": 6.457874075192512e-05, "loss": 1.024, "step": 141810 }, { "epoch": 0.36, "learning_rate": 6.457748251044341e-05, "loss": 1.018, "step": 141815 }, { "epoch": 0.36, "learning_rate": 6.45762242689617e-05, "loss": 1.0207, "step": 141820 }, { "epoch": 0.36, "learning_rate": 6.457496602748e-05, "loss": 1.0222, "step": 141825 }, { "epoch": 0.36, "learning_rate": 6.457370778599829e-05, "loss": 1.0208, "step": 141830 }, { "epoch": 0.36, "learning_rate": 6.457244954451659e-05, "loss": 1.021, "step": 141835 }, { "epoch": 0.36, "learning_rate": 6.457119130303488e-05, "loss": 1.0201, "step": 141840 }, { "epoch": 0.36, "learning_rate": 6.456993306155318e-05, "loss": 1.0203, "step": 141845 }, { "epoch": 0.36, "learning_rate": 6.456867482007147e-05, "loss": 1.0202, "step": 141850 }, { "epoch": 0.36, "learning_rate": 6.456741657858977e-05, "loss": 1.0214, "step": 141855 }, { "epoch": 0.36, "learning_rate": 6.456615833710806e-05, "loss": 1.041, "step": 141860 }, { "epoch": 0.36, "learning_rate": 6.456490009562636e-05, "loss": 1.0215, "step": 141865 }, { "epoch": 0.36, "learning_rate": 6.456364185414465e-05, "loss": 1.0217, "step": 141870 }, { "epoch": 0.36, "learning_rate": 6.456238361266295e-05, "loss": 1.0202, "step": 141875 }, { "epoch": 0.36, "learning_rate": 6.456112537118124e-05, "loss": 1.0394, "step": 141880 }, { "epoch": 0.36, "learning_rate": 6.455986712969954e-05, "loss": 1.0192, "step": 141885 }, { "epoch": 0.36, "learning_rate": 6.455860888821783e-05, "loss": 1.0449, "step": 141890 }, { "epoch": 0.36, "learning_rate": 6.455735064673612e-05, "loss": 1.0238, "step": 141895 }, { "epoch": 0.36, "learning_rate": 6.455609240525442e-05, "loss": 1.0219, "step": 141900 }, { "epoch": 0.36, "learning_rate": 6.455483416377272e-05, "loss": 1.0227, "step": 141905 }, { "epoch": 0.36, "learning_rate": 6.455357592229101e-05, "loss": 1.0222, "step": 141910 }, { "epoch": 0.36, "learning_rate": 6.45523176808093e-05, "loss": 1.0206, "step": 141915 }, { "epoch": 0.36, "learning_rate": 6.45510594393276e-05, "loss": 1.0198, "step": 141920 }, { "epoch": 0.36, "learning_rate": 6.45498011978459e-05, "loss": 1.0214, "step": 141925 }, { "epoch": 0.36, "learning_rate": 6.454854295636419e-05, "loss": 1.021, "step": 141930 }, { "epoch": 0.36, "learning_rate": 6.454728471488248e-05, "loss": 1.021, "step": 141935 }, { "epoch": 0.36, "learning_rate": 6.454602647340078e-05, "loss": 1.0211, "step": 141940 }, { "epoch": 0.36, "learning_rate": 6.454476823191908e-05, "loss": 1.0206, "step": 141945 }, { "epoch": 0.36, "learning_rate": 6.454350999043737e-05, "loss": 1.0202, "step": 141950 }, { "epoch": 0.36, "learning_rate": 6.454225174895566e-05, "loss": 1.0212, "step": 141955 }, { "epoch": 0.36, "learning_rate": 6.454099350747395e-05, "loss": 1.022, "step": 141960 }, { "epoch": 0.36, "learning_rate": 6.453973526599226e-05, "loss": 1.0191, "step": 141965 }, { "epoch": 0.36, "learning_rate": 6.453847702451055e-05, "loss": 1.0201, "step": 141970 }, { "epoch": 0.36, "learning_rate": 6.453721878302884e-05, "loss": 1.0187, "step": 141975 }, { "epoch": 0.36, "learning_rate": 6.453596054154713e-05, "loss": 1.0206, "step": 141980 }, { "epoch": 0.36, "learning_rate": 6.453470230006544e-05, "loss": 1.0207, "step": 141985 }, { "epoch": 0.36, "learning_rate": 6.453344405858373e-05, "loss": 1.0222, "step": 141990 }, { "epoch": 0.36, "learning_rate": 6.453218581710202e-05, "loss": 1.0213, "step": 141995 }, { "epoch": 0.36, "learning_rate": 6.453092757562031e-05, "loss": 1.0216, "step": 142000 }, { "epoch": 0.36, "learning_rate": 6.452966933413862e-05, "loss": 1.0198, "step": 142005 }, { "epoch": 0.36, "learning_rate": 6.452841109265691e-05, "loss": 1.0187, "step": 142010 }, { "epoch": 0.36, "learning_rate": 6.45271528511752e-05, "loss": 1.0228, "step": 142015 }, { "epoch": 0.36, "learning_rate": 6.452589460969349e-05, "loss": 1.0201, "step": 142020 }, { "epoch": 0.36, "learning_rate": 6.452463636821178e-05, "loss": 1.0214, "step": 142025 }, { "epoch": 0.36, "learning_rate": 6.452337812673009e-05, "loss": 1.0202, "step": 142030 }, { "epoch": 0.36, "learning_rate": 6.452211988524838e-05, "loss": 1.0225, "step": 142035 }, { "epoch": 0.36, "learning_rate": 6.452086164376667e-05, "loss": 1.0203, "step": 142040 }, { "epoch": 0.36, "learning_rate": 6.451960340228496e-05, "loss": 1.0206, "step": 142045 }, { "epoch": 0.36, "learning_rate": 6.451834516080327e-05, "loss": 1.0185, "step": 142050 }, { "epoch": 0.36, "learning_rate": 6.451708691932156e-05, "loss": 1.0197, "step": 142055 }, { "epoch": 0.36, "learning_rate": 6.451582867783985e-05, "loss": 1.0203, "step": 142060 }, { "epoch": 0.36, "learning_rate": 6.451457043635814e-05, "loss": 1.0203, "step": 142065 }, { "epoch": 0.36, "learning_rate": 6.451331219487645e-05, "loss": 1.0209, "step": 142070 }, { "epoch": 0.36, "learning_rate": 6.451205395339474e-05, "loss": 1.0209, "step": 142075 }, { "epoch": 0.36, "learning_rate": 6.451079571191303e-05, "loss": 1.021, "step": 142080 }, { "epoch": 0.36, "learning_rate": 6.450953747043132e-05, "loss": 1.0192, "step": 142085 }, { "epoch": 0.36, "learning_rate": 6.450827922894961e-05, "loss": 1.0199, "step": 142090 }, { "epoch": 0.36, "learning_rate": 6.450702098746792e-05, "loss": 1.0192, "step": 142095 }, { "epoch": 0.36, "learning_rate": 6.450576274598621e-05, "loss": 1.022, "step": 142100 }, { "epoch": 0.36, "learning_rate": 6.45045045045045e-05, "loss": 1.0199, "step": 142105 }, { "epoch": 0.36, "learning_rate": 6.45032462630228e-05, "loss": 1.0209, "step": 142110 }, { "epoch": 0.36, "learning_rate": 6.45019880215411e-05, "loss": 1.0224, "step": 142115 }, { "epoch": 0.36, "learning_rate": 6.450072978005939e-05, "loss": 1.0225, "step": 142120 }, { "epoch": 0.36, "learning_rate": 6.449947153857768e-05, "loss": 1.0424, "step": 142125 }, { "epoch": 0.36, "learning_rate": 6.449821329709597e-05, "loss": 1.0191, "step": 142130 }, { "epoch": 0.36, "learning_rate": 6.449695505561428e-05, "loss": 1.0202, "step": 142135 }, { "epoch": 0.36, "learning_rate": 6.449569681413257e-05, "loss": 1.0184, "step": 142140 }, { "epoch": 0.36, "learning_rate": 6.449443857265086e-05, "loss": 1.0199, "step": 142145 }, { "epoch": 0.36, "learning_rate": 6.449318033116915e-05, "loss": 1.0203, "step": 142150 }, { "epoch": 0.36, "learning_rate": 6.449192208968745e-05, "loss": 1.0193, "step": 142155 }, { "epoch": 0.36, "learning_rate": 6.449066384820575e-05, "loss": 1.0228, "step": 142160 }, { "epoch": 0.36, "learning_rate": 6.448940560672404e-05, "loss": 1.022, "step": 142165 }, { "epoch": 0.36, "learning_rate": 6.448814736524233e-05, "loss": 1.0213, "step": 142170 }, { "epoch": 0.36, "learning_rate": 6.448688912376063e-05, "loss": 1.0187, "step": 142175 }, { "epoch": 0.36, "learning_rate": 6.448563088227893e-05, "loss": 1.0194, "step": 142180 }, { "epoch": 0.36, "learning_rate": 6.448437264079724e-05, "loss": 1.0201, "step": 142185 }, { "epoch": 0.36, "learning_rate": 6.448311439931553e-05, "loss": 1.0224, "step": 142190 }, { "epoch": 0.36, "learning_rate": 6.448185615783382e-05, "loss": 1.0216, "step": 142195 }, { "epoch": 0.36, "learning_rate": 6.448059791635211e-05, "loss": 1.0219, "step": 142200 }, { "epoch": 0.36, "learning_rate": 6.447933967487042e-05, "loss": 1.021, "step": 142205 }, { "epoch": 0.36, "learning_rate": 6.447808143338871e-05, "loss": 1.0205, "step": 142210 }, { "epoch": 0.36, "learning_rate": 6.4476823191907e-05, "loss": 1.0218, "step": 142215 }, { "epoch": 0.36, "learning_rate": 6.447556495042529e-05, "loss": 1.0206, "step": 142220 }, { "epoch": 0.36, "learning_rate": 6.447430670894358e-05, "loss": 1.0195, "step": 142225 }, { "epoch": 0.36, "learning_rate": 6.447304846746189e-05, "loss": 1.0193, "step": 142230 }, { "epoch": 0.36, "learning_rate": 6.447179022598018e-05, "loss": 1.0216, "step": 142235 }, { "epoch": 0.36, "learning_rate": 6.447053198449847e-05, "loss": 1.0203, "step": 142240 }, { "epoch": 0.36, "learning_rate": 6.446927374301676e-05, "loss": 1.0207, "step": 142245 }, { "epoch": 0.36, "learning_rate": 6.446801550153507e-05, "loss": 1.0175, "step": 142250 }, { "epoch": 0.36, "learning_rate": 6.446675726005336e-05, "loss": 1.0193, "step": 142255 }, { "epoch": 0.36, "learning_rate": 6.446549901857165e-05, "loss": 1.0227, "step": 142260 }, { "epoch": 0.36, "learning_rate": 6.446424077708994e-05, "loss": 1.0217, "step": 142265 }, { "epoch": 0.36, "learning_rate": 6.446298253560825e-05, "loss": 1.0211, "step": 142270 }, { "epoch": 0.36, "learning_rate": 6.446172429412654e-05, "loss": 1.0225, "step": 142275 }, { "epoch": 0.36, "learning_rate": 6.446046605264483e-05, "loss": 1.0247, "step": 142280 }, { "epoch": 0.36, "learning_rate": 6.445920781116312e-05, "loss": 1.023, "step": 142285 }, { "epoch": 0.36, "learning_rate": 6.445794956968141e-05, "loss": 1.0183, "step": 142290 }, { "epoch": 0.36, "learning_rate": 6.445669132819972e-05, "loss": 1.0241, "step": 142295 }, { "epoch": 0.36, "learning_rate": 6.445543308671801e-05, "loss": 1.0225, "step": 142300 }, { "epoch": 0.36, "learning_rate": 6.44541748452363e-05, "loss": 1.0188, "step": 142305 }, { "epoch": 0.36, "learning_rate": 6.445291660375459e-05, "loss": 1.021, "step": 142310 }, { "epoch": 0.36, "learning_rate": 6.44516583622729e-05, "loss": 1.0213, "step": 142315 }, { "epoch": 0.36, "learning_rate": 6.445040012079119e-05, "loss": 1.0196, "step": 142320 }, { "epoch": 0.36, "learning_rate": 6.444914187930948e-05, "loss": 1.0248, "step": 142325 }, { "epoch": 0.36, "learning_rate": 6.444788363782777e-05, "loss": 1.0198, "step": 142330 }, { "epoch": 0.36, "learning_rate": 6.444662539634608e-05, "loss": 1.0215, "step": 142335 }, { "epoch": 0.36, "learning_rate": 6.444536715486437e-05, "loss": 1.0221, "step": 142340 }, { "epoch": 0.36, "learning_rate": 6.444410891338266e-05, "loss": 1.0203, "step": 142345 }, { "epoch": 0.36, "learning_rate": 6.444285067190095e-05, "loss": 1.0216, "step": 142350 }, { "epoch": 0.36, "learning_rate": 6.444159243041924e-05, "loss": 1.0377, "step": 142355 }, { "epoch": 0.36, "learning_rate": 6.444033418893755e-05, "loss": 1.0242, "step": 142360 }, { "epoch": 0.36, "learning_rate": 6.443907594745584e-05, "loss": 1.0207, "step": 142365 }, { "epoch": 0.36, "learning_rate": 6.443781770597413e-05, "loss": 1.035, "step": 142370 }, { "epoch": 0.36, "learning_rate": 6.443655946449242e-05, "loss": 1.0382, "step": 142375 }, { "epoch": 0.36, "learning_rate": 6.443530122301073e-05, "loss": 1.0198, "step": 142380 }, { "epoch": 0.36, "learning_rate": 6.443404298152902e-05, "loss": 1.0208, "step": 142385 }, { "epoch": 0.36, "learning_rate": 6.443278474004731e-05, "loss": 1.0203, "step": 142390 }, { "epoch": 0.36, "learning_rate": 6.44315264985656e-05, "loss": 1.0217, "step": 142395 }, { "epoch": 0.36, "learning_rate": 6.443026825708391e-05, "loss": 1.0227, "step": 142400 }, { "epoch": 0.36, "learning_rate": 6.44290100156022e-05, "loss": 1.0194, "step": 142405 }, { "epoch": 0.36, "learning_rate": 6.442775177412049e-05, "loss": 1.0428, "step": 142410 }, { "epoch": 0.36, "learning_rate": 6.442649353263878e-05, "loss": 1.019, "step": 142415 }, { "epoch": 0.36, "learning_rate": 6.442523529115708e-05, "loss": 1.019, "step": 142420 }, { "epoch": 0.36, "learning_rate": 6.442397704967538e-05, "loss": 1.0217, "step": 142425 }, { "epoch": 0.36, "learning_rate": 6.442271880819367e-05, "loss": 1.0226, "step": 142430 }, { "epoch": 0.36, "learning_rate": 6.442146056671196e-05, "loss": 1.0219, "step": 142435 }, { "epoch": 0.36, "learning_rate": 6.442020232523026e-05, "loss": 1.0197, "step": 142440 }, { "epoch": 0.36, "learning_rate": 6.441894408374856e-05, "loss": 1.0193, "step": 142445 }, { "epoch": 0.36, "learning_rate": 6.441768584226685e-05, "loss": 1.0212, "step": 142450 }, { "epoch": 0.36, "learning_rate": 6.441642760078514e-05, "loss": 1.0201, "step": 142455 }, { "epoch": 0.36, "learning_rate": 6.441516935930344e-05, "loss": 1.0197, "step": 142460 }, { "epoch": 0.36, "learning_rate": 6.441391111782174e-05, "loss": 1.0212, "step": 142465 }, { "epoch": 0.36, "learning_rate": 6.441265287634003e-05, "loss": 1.0176, "step": 142470 }, { "epoch": 0.36, "learning_rate": 6.441139463485832e-05, "loss": 1.0227, "step": 142475 }, { "epoch": 0.36, "learning_rate": 6.441013639337662e-05, "loss": 1.0209, "step": 142480 }, { "epoch": 0.36, "learning_rate": 6.440887815189491e-05, "loss": 1.0219, "step": 142485 }, { "epoch": 0.36, "learning_rate": 6.440761991041321e-05, "loss": 1.0202, "step": 142490 }, { "epoch": 0.36, "learning_rate": 6.44063616689315e-05, "loss": 1.0188, "step": 142495 }, { "epoch": 0.36, "learning_rate": 6.44051034274498e-05, "loss": 1.0223, "step": 142500 }, { "epoch": 0.36, "learning_rate": 6.440384518596809e-05, "loss": 1.0419, "step": 142505 }, { "epoch": 0.36, "learning_rate": 6.440258694448639e-05, "loss": 1.0223, "step": 142510 }, { "epoch": 0.36, "learning_rate": 6.440132870300468e-05, "loss": 1.0208, "step": 142515 }, { "epoch": 0.36, "learning_rate": 6.440007046152298e-05, "loss": 1.0208, "step": 142520 }, { "epoch": 0.36, "learning_rate": 6.439881222004127e-05, "loss": 1.0202, "step": 142525 }, { "epoch": 0.36, "learning_rate": 6.439755397855957e-05, "loss": 1.0206, "step": 142530 }, { "epoch": 0.36, "learning_rate": 6.439629573707786e-05, "loss": 1.0215, "step": 142535 }, { "epoch": 0.36, "learning_rate": 6.439503749559616e-05, "loss": 1.0192, "step": 142540 }, { "epoch": 0.36, "learning_rate": 6.439377925411445e-05, "loss": 1.0218, "step": 142545 }, { "epoch": 0.36, "learning_rate": 6.439252101263274e-05, "loss": 1.022, "step": 142550 }, { "epoch": 0.36, "learning_rate": 6.439126277115104e-05, "loss": 1.0212, "step": 142555 }, { "epoch": 0.36, "learning_rate": 6.439000452966934e-05, "loss": 1.0222, "step": 142560 }, { "epoch": 0.36, "learning_rate": 6.438874628818763e-05, "loss": 1.0191, "step": 142565 }, { "epoch": 0.36, "learning_rate": 6.438748804670592e-05, "loss": 1.0203, "step": 142570 }, { "epoch": 0.36, "learning_rate": 6.438622980522422e-05, "loss": 1.019, "step": 142575 }, { "epoch": 0.36, "learning_rate": 6.438497156374252e-05, "loss": 1.019, "step": 142580 }, { "epoch": 0.36, "learning_rate": 6.438371332226081e-05, "loss": 1.0197, "step": 142585 }, { "epoch": 0.36, "learning_rate": 6.43824550807791e-05, "loss": 1.0213, "step": 142590 }, { "epoch": 0.36, "learning_rate": 6.43811968392974e-05, "loss": 1.0208, "step": 142595 }, { "epoch": 0.36, "learning_rate": 6.43799385978157e-05, "loss": 1.0208, "step": 142600 }, { "epoch": 0.36, "learning_rate": 6.437868035633399e-05, "loss": 1.0215, "step": 142605 }, { "epoch": 0.36, "learning_rate": 6.437742211485228e-05, "loss": 1.0205, "step": 142610 }, { "epoch": 0.36, "learning_rate": 6.437641552166693e-05, "loss": 1.0215, "step": 142615 }, { "epoch": 0.36, "learning_rate": 6.437515728018522e-05, "loss": 1.0207, "step": 142620 }, { "epoch": 0.36, "learning_rate": 6.437389903870351e-05, "loss": 1.0203, "step": 142625 }, { "epoch": 0.36, "learning_rate": 6.43726407972218e-05, "loss": 1.0197, "step": 142630 }, { "epoch": 0.36, "learning_rate": 6.43713825557401e-05, "loss": 1.0209, "step": 142635 }, { "epoch": 0.36, "learning_rate": 6.43701243142584e-05, "loss": 1.021, "step": 142640 }, { "epoch": 0.36, "learning_rate": 6.436886607277669e-05, "loss": 1.0209, "step": 142645 }, { "epoch": 0.36, "learning_rate": 6.436760783129498e-05, "loss": 1.0221, "step": 142650 }, { "epoch": 0.36, "learning_rate": 6.436634958981327e-05, "loss": 1.0216, "step": 142655 }, { "epoch": 0.36, "learning_rate": 6.436509134833158e-05, "loss": 1.0227, "step": 142660 }, { "epoch": 0.36, "learning_rate": 6.436383310684987e-05, "loss": 1.0204, "step": 142665 }, { "epoch": 0.36, "learning_rate": 6.436257486536816e-05, "loss": 1.0224, "step": 142670 }, { "epoch": 0.36, "learning_rate": 6.436131662388645e-05, "loss": 1.0203, "step": 142675 }, { "epoch": 0.36, "learning_rate": 6.436005838240476e-05, "loss": 1.0182, "step": 142680 }, { "epoch": 0.36, "learning_rate": 6.435880014092305e-05, "loss": 1.0193, "step": 142685 }, { "epoch": 0.36, "learning_rate": 6.435754189944134e-05, "loss": 1.0202, "step": 142690 }, { "epoch": 0.36, "learning_rate": 6.435628365795963e-05, "loss": 1.0198, "step": 142695 }, { "epoch": 0.36, "learning_rate": 6.435502541647793e-05, "loss": 1.0218, "step": 142700 }, { "epoch": 0.36, "learning_rate": 6.435376717499623e-05, "loss": 1.0216, "step": 142705 }, { "epoch": 0.36, "learning_rate": 6.435250893351452e-05, "loss": 1.0188, "step": 142710 }, { "epoch": 0.36, "learning_rate": 6.435125069203281e-05, "loss": 1.0224, "step": 142715 }, { "epoch": 0.36, "learning_rate": 6.43499924505511e-05, "loss": 1.0202, "step": 142720 }, { "epoch": 0.36, "learning_rate": 6.434873420906941e-05, "loss": 1.0213, "step": 142725 }, { "epoch": 0.36, "learning_rate": 6.43474759675877e-05, "loss": 1.0218, "step": 142730 }, { "epoch": 0.36, "learning_rate": 6.4346217726106e-05, "loss": 1.0216, "step": 142735 }, { "epoch": 0.36, "learning_rate": 6.434495948462429e-05, "loss": 1.0388, "step": 142740 }, { "epoch": 0.36, "learning_rate": 6.434370124314259e-05, "loss": 1.0194, "step": 142745 }, { "epoch": 0.36, "learning_rate": 6.434244300166088e-05, "loss": 1.0205, "step": 142750 }, { "epoch": 0.36, "learning_rate": 6.434118476017917e-05, "loss": 1.0206, "step": 142755 }, { "epoch": 0.36, "learning_rate": 6.433992651869747e-05, "loss": 1.0216, "step": 142760 }, { "epoch": 0.36, "learning_rate": 6.433866827721576e-05, "loss": 1.0234, "step": 142765 }, { "epoch": 0.36, "learning_rate": 6.433741003573406e-05, "loss": 1.02, "step": 142770 }, { "epoch": 0.36, "learning_rate": 6.433615179425235e-05, "loss": 1.021, "step": 142775 }, { "epoch": 0.36, "learning_rate": 6.433489355277065e-05, "loss": 1.0193, "step": 142780 }, { "epoch": 0.36, "learning_rate": 6.433363531128894e-05, "loss": 1.0197, "step": 142785 }, { "epoch": 0.36, "learning_rate": 6.433237706980724e-05, "loss": 1.0217, "step": 142790 }, { "epoch": 0.36, "learning_rate": 6.433111882832553e-05, "loss": 1.0212, "step": 142795 }, { "epoch": 0.36, "learning_rate": 6.432986058684383e-05, "loss": 1.0216, "step": 142800 }, { "epoch": 0.36, "learning_rate": 6.432860234536212e-05, "loss": 1.0202, "step": 142805 }, { "epoch": 0.36, "learning_rate": 6.432734410388042e-05, "loss": 1.0202, "step": 142810 }, { "epoch": 0.36, "learning_rate": 6.432608586239871e-05, "loss": 1.02, "step": 142815 }, { "epoch": 0.36, "learning_rate": 6.4324827620917e-05, "loss": 1.0212, "step": 142820 }, { "epoch": 0.36, "learning_rate": 6.43235693794353e-05, "loss": 1.0196, "step": 142825 }, { "epoch": 0.36, "learning_rate": 6.432231113795359e-05, "loss": 1.0211, "step": 142830 }, { "epoch": 0.36, "learning_rate": 6.43210528964719e-05, "loss": 1.0238, "step": 142835 }, { "epoch": 0.36, "learning_rate": 6.431979465499019e-05, "loss": 1.0203, "step": 142840 }, { "epoch": 0.36, "learning_rate": 6.431853641350848e-05, "loss": 1.0202, "step": 142845 }, { "epoch": 0.36, "learning_rate": 6.431727817202677e-05, "loss": 1.0214, "step": 142850 }, { "epoch": 0.36, "learning_rate": 6.431601993054507e-05, "loss": 1.039, "step": 142855 }, { "epoch": 0.36, "learning_rate": 6.431476168906337e-05, "loss": 1.0193, "step": 142860 }, { "epoch": 0.36, "learning_rate": 6.431350344758166e-05, "loss": 1.0214, "step": 142865 }, { "epoch": 0.36, "learning_rate": 6.431224520609995e-05, "loss": 1.0225, "step": 142870 }, { "epoch": 0.36, "learning_rate": 6.431098696461825e-05, "loss": 1.0215, "step": 142875 }, { "epoch": 0.36, "learning_rate": 6.430972872313655e-05, "loss": 1.0189, "step": 142880 }, { "epoch": 0.36, "learning_rate": 6.430847048165485e-05, "loss": 1.0191, "step": 142885 }, { "epoch": 0.36, "learning_rate": 6.430721224017314e-05, "loss": 1.0218, "step": 142890 }, { "epoch": 0.36, "learning_rate": 6.430595399869143e-05, "loss": 1.0207, "step": 142895 }, { "epoch": 0.36, "learning_rate": 6.430469575720973e-05, "loss": 1.018, "step": 142900 }, { "epoch": 0.36, "learning_rate": 6.430343751572803e-05, "loss": 1.0224, "step": 142905 }, { "epoch": 0.36, "learning_rate": 6.430217927424632e-05, "loss": 1.021, "step": 142910 }, { "epoch": 0.36, "learning_rate": 6.430092103276461e-05, "loss": 1.0189, "step": 142915 }, { "epoch": 0.36, "learning_rate": 6.42996627912829e-05, "loss": 1.0209, "step": 142920 }, { "epoch": 0.36, "learning_rate": 6.429840454980121e-05, "loss": 1.0205, "step": 142925 }, { "epoch": 0.36, "learning_rate": 6.42971463083195e-05, "loss": 1.0213, "step": 142930 }, { "epoch": 0.36, "learning_rate": 6.42958880668378e-05, "loss": 1.0216, "step": 142935 }, { "epoch": 0.36, "learning_rate": 6.429462982535609e-05, "loss": 1.0205, "step": 142940 }, { "epoch": 0.36, "learning_rate": 6.429337158387439e-05, "loss": 1.0206, "step": 142945 }, { "epoch": 0.36, "learning_rate": 6.429211334239268e-05, "loss": 1.0209, "step": 142950 }, { "epoch": 0.36, "learning_rate": 6.429085510091097e-05, "loss": 1.0206, "step": 142955 }, { "epoch": 0.36, "learning_rate": 6.428959685942927e-05, "loss": 1.0209, "step": 142960 }, { "epoch": 0.36, "learning_rate": 6.428833861794756e-05, "loss": 1.0217, "step": 142965 }, { "epoch": 0.36, "learning_rate": 6.428708037646586e-05, "loss": 1.0339, "step": 142970 }, { "epoch": 0.36, "learning_rate": 6.428582213498415e-05, "loss": 1.0182, "step": 142975 }, { "epoch": 0.36, "learning_rate": 6.428456389350245e-05, "loss": 1.0205, "step": 142980 }, { "epoch": 0.36, "learning_rate": 6.428330565202074e-05, "loss": 1.0204, "step": 142985 }, { "epoch": 0.36, "learning_rate": 6.428204741053904e-05, "loss": 1.0231, "step": 142990 }, { "epoch": 0.36, "learning_rate": 6.428078916905733e-05, "loss": 1.0194, "step": 142995 }, { "epoch": 0.36, "learning_rate": 6.427953092757563e-05, "loss": 1.0219, "step": 143000 }, { "epoch": 0.36, "learning_rate": 6.427827268609392e-05, "loss": 1.0211, "step": 143005 }, { "epoch": 0.36, "learning_rate": 6.427701444461222e-05, "loss": 1.022, "step": 143010 }, { "epoch": 0.36, "learning_rate": 6.427575620313051e-05, "loss": 1.0213, "step": 143015 }, { "epoch": 0.36, "learning_rate": 6.42744979616488e-05, "loss": 1.0198, "step": 143020 }, { "epoch": 0.36, "learning_rate": 6.42732397201671e-05, "loss": 1.0226, "step": 143025 }, { "epoch": 0.36, "learning_rate": 6.427198147868539e-05, "loss": 1.0233, "step": 143030 }, { "epoch": 0.36, "learning_rate": 6.42707232372037e-05, "loss": 1.0193, "step": 143035 }, { "epoch": 0.36, "learning_rate": 6.426946499572199e-05, "loss": 1.0209, "step": 143040 }, { "epoch": 0.36, "learning_rate": 6.426820675424028e-05, "loss": 1.0176, "step": 143045 }, { "epoch": 0.36, "learning_rate": 6.426694851275857e-05, "loss": 1.0193, "step": 143050 }, { "epoch": 0.36, "learning_rate": 6.426569027127687e-05, "loss": 1.0209, "step": 143055 }, { "epoch": 0.36, "learning_rate": 6.426443202979516e-05, "loss": 1.0209, "step": 143060 }, { "epoch": 0.36, "learning_rate": 6.426317378831346e-05, "loss": 1.0211, "step": 143065 }, { "epoch": 0.36, "learning_rate": 6.426191554683175e-05, "loss": 1.0206, "step": 143070 }, { "epoch": 0.36, "learning_rate": 6.426090895364638e-05, "loss": 1.0208, "step": 143075 }, { "epoch": 0.36, "learning_rate": 6.425965071216468e-05, "loss": 1.0211, "step": 143080 }, { "epoch": 0.36, "learning_rate": 6.425839247068298e-05, "loss": 1.0201, "step": 143085 }, { "epoch": 0.36, "learning_rate": 6.425713422920127e-05, "loss": 1.0194, "step": 143090 }, { "epoch": 0.36, "learning_rate": 6.425587598771958e-05, "loss": 1.0208, "step": 143095 }, { "epoch": 0.36, "learning_rate": 6.425461774623787e-05, "loss": 1.0198, "step": 143100 }, { "epoch": 0.36, "learning_rate": 6.425335950475616e-05, "loss": 1.0199, "step": 143105 }, { "epoch": 0.36, "learning_rate": 6.425210126327445e-05, "loss": 1.0233, "step": 143110 }, { "epoch": 0.36, "learning_rate": 6.425084302179274e-05, "loss": 1.0205, "step": 143115 }, { "epoch": 0.36, "learning_rate": 6.424958478031105e-05, "loss": 1.0203, "step": 143120 }, { "epoch": 0.36, "learning_rate": 6.424832653882934e-05, "loss": 1.0191, "step": 143125 }, { "epoch": 0.36, "learning_rate": 6.424706829734763e-05, "loss": 1.0235, "step": 143130 }, { "epoch": 0.36, "learning_rate": 6.424581005586592e-05, "loss": 1.0208, "step": 143135 }, { "epoch": 0.36, "learning_rate": 6.424455181438423e-05, "loss": 1.0212, "step": 143140 }, { "epoch": 0.36, "learning_rate": 6.424329357290252e-05, "loss": 1.019, "step": 143145 }, { "epoch": 0.36, "learning_rate": 6.424203533142081e-05, "loss": 1.0176, "step": 143150 }, { "epoch": 0.36, "learning_rate": 6.42407770899391e-05, "loss": 1.0183, "step": 143155 }, { "epoch": 0.36, "learning_rate": 6.423951884845741e-05, "loss": 1.0196, "step": 143160 }, { "epoch": 0.36, "learning_rate": 6.42382606069757e-05, "loss": 1.0207, "step": 143165 }, { "epoch": 0.36, "learning_rate": 6.423700236549399e-05, "loss": 1.0197, "step": 143170 }, { "epoch": 0.36, "learning_rate": 6.423574412401228e-05, "loss": 1.0456, "step": 143175 }, { "epoch": 0.36, "learning_rate": 6.423448588253058e-05, "loss": 1.0181, "step": 143180 }, { "epoch": 0.36, "learning_rate": 6.423322764104888e-05, "loss": 1.0191, "step": 143185 }, { "epoch": 0.36, "learning_rate": 6.423196939956717e-05, "loss": 1.0216, "step": 143190 }, { "epoch": 0.36, "learning_rate": 6.423071115808546e-05, "loss": 1.0202, "step": 143195 }, { "epoch": 0.36, "learning_rate": 6.422945291660376e-05, "loss": 1.0214, "step": 143200 }, { "epoch": 0.36, "learning_rate": 6.422819467512206e-05, "loss": 1.0193, "step": 143205 }, { "epoch": 0.36, "learning_rate": 6.422693643364035e-05, "loss": 1.0209, "step": 143210 }, { "epoch": 0.36, "learning_rate": 6.422567819215864e-05, "loss": 1.0221, "step": 143215 }, { "epoch": 0.36, "learning_rate": 6.422441995067694e-05, "loss": 1.023, "step": 143220 }, { "epoch": 0.36, "learning_rate": 6.422316170919524e-05, "loss": 1.0201, "step": 143225 }, { "epoch": 0.36, "learning_rate": 6.422190346771353e-05, "loss": 1.0218, "step": 143230 }, { "epoch": 0.36, "learning_rate": 6.422064522623182e-05, "loss": 1.018, "step": 143235 }, { "epoch": 0.36, "learning_rate": 6.421938698475012e-05, "loss": 1.0202, "step": 143240 }, { "epoch": 0.36, "learning_rate": 6.421812874326841e-05, "loss": 1.0199, "step": 143245 }, { "epoch": 0.36, "learning_rate": 6.421687050178671e-05, "loss": 1.0214, "step": 143250 }, { "epoch": 0.36, "learning_rate": 6.4215612260305e-05, "loss": 1.0209, "step": 143255 }, { "epoch": 0.36, "learning_rate": 6.42143540188233e-05, "loss": 1.0198, "step": 143260 }, { "epoch": 0.36, "learning_rate": 6.421309577734159e-05, "loss": 1.0208, "step": 143265 }, { "epoch": 0.36, "learning_rate": 6.421183753585989e-05, "loss": 1.0211, "step": 143270 }, { "epoch": 0.36, "learning_rate": 6.421057929437818e-05, "loss": 1.0227, "step": 143275 }, { "epoch": 0.36, "learning_rate": 6.420932105289648e-05, "loss": 1.0215, "step": 143280 }, { "epoch": 0.36, "learning_rate": 6.420806281141477e-05, "loss": 1.0208, "step": 143285 }, { "epoch": 0.36, "learning_rate": 6.420680456993307e-05, "loss": 1.0202, "step": 143290 }, { "epoch": 0.36, "learning_rate": 6.420554632845136e-05, "loss": 1.0205, "step": 143295 }, { "epoch": 0.36, "learning_rate": 6.420428808696966e-05, "loss": 1.0206, "step": 143300 }, { "epoch": 0.36, "learning_rate": 6.420302984548795e-05, "loss": 1.0251, "step": 143305 }, { "epoch": 0.36, "learning_rate": 6.420177160400624e-05, "loss": 1.0184, "step": 143310 }, { "epoch": 0.36, "learning_rate": 6.420051336252454e-05, "loss": 1.0209, "step": 143315 }, { "epoch": 0.36, "learning_rate": 6.419925512104284e-05, "loss": 1.0205, "step": 143320 }, { "epoch": 0.36, "learning_rate": 6.419799687956113e-05, "loss": 1.0216, "step": 143325 }, { "epoch": 0.36, "learning_rate": 6.419673863807942e-05, "loss": 1.0199, "step": 143330 }, { "epoch": 0.36, "learning_rate": 6.419548039659772e-05, "loss": 1.0208, "step": 143335 }, { "epoch": 0.36, "learning_rate": 6.419422215511601e-05, "loss": 1.0196, "step": 143340 }, { "epoch": 0.36, "learning_rate": 6.41929639136343e-05, "loss": 1.0195, "step": 143345 }, { "epoch": 0.36, "learning_rate": 6.41917056721526e-05, "loss": 1.0208, "step": 143350 }, { "epoch": 0.36, "learning_rate": 6.41904474306709e-05, "loss": 1.0214, "step": 143355 }, { "epoch": 0.36, "learning_rate": 6.41891891891892e-05, "loss": 1.0219, "step": 143360 }, { "epoch": 0.36, "learning_rate": 6.418793094770749e-05, "loss": 1.0207, "step": 143365 }, { "epoch": 0.36, "learning_rate": 6.418667270622578e-05, "loss": 1.0225, "step": 143370 }, { "epoch": 0.36, "learning_rate": 6.418541446474407e-05, "loss": 1.0227, "step": 143375 }, { "epoch": 0.36, "learning_rate": 6.418415622326237e-05, "loss": 1.0194, "step": 143380 }, { "epoch": 0.36, "learning_rate": 6.418289798178067e-05, "loss": 1.0217, "step": 143385 }, { "epoch": 0.36, "learning_rate": 6.418163974029896e-05, "loss": 1.0191, "step": 143390 }, { "epoch": 0.36, "learning_rate": 6.418038149881725e-05, "loss": 1.0213, "step": 143395 }, { "epoch": 0.36, "learning_rate": 6.417912325733555e-05, "loss": 1.0222, "step": 143400 }, { "epoch": 0.36, "learning_rate": 6.417786501585385e-05, "loss": 1.0193, "step": 143405 }, { "epoch": 0.36, "learning_rate": 6.417660677437214e-05, "loss": 1.0212, "step": 143410 }, { "epoch": 0.36, "learning_rate": 6.417534853289043e-05, "loss": 1.0206, "step": 143415 }, { "epoch": 0.36, "learning_rate": 6.417409029140873e-05, "loss": 1.022, "step": 143420 }, { "epoch": 0.36, "learning_rate": 6.417283204992703e-05, "loss": 1.0208, "step": 143425 }, { "epoch": 0.36, "learning_rate": 6.417157380844532e-05, "loss": 1.0199, "step": 143430 }, { "epoch": 0.36, "learning_rate": 6.417031556696361e-05, "loss": 1.0217, "step": 143435 }, { "epoch": 0.36, "learning_rate": 6.41690573254819e-05, "loss": 1.0201, "step": 143440 }, { "epoch": 0.36, "learning_rate": 6.41677990840002e-05, "loss": 1.0222, "step": 143445 }, { "epoch": 0.36, "learning_rate": 6.41665408425185e-05, "loss": 1.0199, "step": 143450 }, { "epoch": 0.36, "learning_rate": 6.416528260103679e-05, "loss": 1.0224, "step": 143455 }, { "epoch": 0.36, "learning_rate": 6.416402435955508e-05, "loss": 1.0227, "step": 143460 }, { "epoch": 0.36, "learning_rate": 6.416276611807339e-05, "loss": 1.0205, "step": 143465 }, { "epoch": 0.36, "learning_rate": 6.416150787659168e-05, "loss": 1.0221, "step": 143470 }, { "epoch": 0.36, "learning_rate": 6.416024963510997e-05, "loss": 1.0364, "step": 143475 }, { "epoch": 0.36, "learning_rate": 6.415899139362826e-05, "loss": 1.0406, "step": 143480 }, { "epoch": 0.36, "learning_rate": 6.415773315214657e-05, "loss": 1.0213, "step": 143485 }, { "epoch": 0.36, "learning_rate": 6.415647491066486e-05, "loss": 1.0184, "step": 143490 }, { "epoch": 0.36, "learning_rate": 6.415521666918315e-05, "loss": 1.0197, "step": 143495 }, { "epoch": 0.36, "learning_rate": 6.415395842770144e-05, "loss": 1.02, "step": 143500 }, { "epoch": 0.36, "learning_rate": 6.415270018621973e-05, "loss": 1.0202, "step": 143505 }, { "epoch": 0.36, "learning_rate": 6.415144194473804e-05, "loss": 1.0225, "step": 143510 }, { "epoch": 0.36, "learning_rate": 6.415018370325633e-05, "loss": 1.0222, "step": 143515 }, { "epoch": 0.36, "learning_rate": 6.414892546177462e-05, "loss": 1.0209, "step": 143520 }, { "epoch": 0.36, "learning_rate": 6.414766722029291e-05, "loss": 1.0207, "step": 143525 }, { "epoch": 0.36, "learning_rate": 6.414640897881122e-05, "loss": 1.0208, "step": 143530 }, { "epoch": 0.36, "learning_rate": 6.414515073732951e-05, "loss": 1.0215, "step": 143535 }, { "epoch": 0.36, "learning_rate": 6.41438924958478e-05, "loss": 1.0214, "step": 143540 }, { "epoch": 0.36, "learning_rate": 6.414263425436609e-05, "loss": 1.0216, "step": 143545 }, { "epoch": 0.36, "learning_rate": 6.41413760128844e-05, "loss": 1.0215, "step": 143550 }, { "epoch": 0.36, "learning_rate": 6.414011777140269e-05, "loss": 1.0203, "step": 143555 }, { "epoch": 0.36, "learning_rate": 6.413885952992098e-05, "loss": 1.0455, "step": 143560 }, { "epoch": 0.36, "learning_rate": 6.413760128843927e-05, "loss": 1.021, "step": 143565 }, { "epoch": 0.36, "learning_rate": 6.413634304695756e-05, "loss": 1.022, "step": 143570 }, { "epoch": 0.36, "learning_rate": 6.413508480547587e-05, "loss": 1.0232, "step": 143575 }, { "epoch": 0.36, "learning_rate": 6.413382656399416e-05, "loss": 1.0201, "step": 143580 }, { "epoch": 0.36, "learning_rate": 6.413256832251247e-05, "loss": 1.0181, "step": 143585 }, { "epoch": 0.36, "learning_rate": 6.413131008103076e-05, "loss": 1.0214, "step": 143590 }, { "epoch": 0.36, "learning_rate": 6.413005183954905e-05, "loss": 1.0194, "step": 143595 }, { "epoch": 0.36, "learning_rate": 6.412879359806735e-05, "loss": 1.0232, "step": 143600 }, { "epoch": 0.36, "learning_rate": 6.412753535658565e-05, "loss": 1.0233, "step": 143605 }, { "epoch": 0.36, "learning_rate": 6.412627711510394e-05, "loss": 1.0221, "step": 143610 }, { "epoch": 0.36, "learning_rate": 6.412501887362223e-05, "loss": 1.0244, "step": 143615 }, { "epoch": 0.36, "learning_rate": 6.412376063214053e-05, "loss": 1.0195, "step": 143620 }, { "epoch": 0.36, "learning_rate": 6.412250239065883e-05, "loss": 1.0218, "step": 143625 }, { "epoch": 0.36, "learning_rate": 6.412124414917712e-05, "loss": 1.0213, "step": 143630 }, { "epoch": 0.36, "learning_rate": 6.411998590769541e-05, "loss": 1.0211, "step": 143635 }, { "epoch": 0.36, "learning_rate": 6.41187276662137e-05, "loss": 1.023, "step": 143640 }, { "epoch": 0.36, "learning_rate": 6.4117469424732e-05, "loss": 1.0232, "step": 143645 }, { "epoch": 0.36, "learning_rate": 6.41162111832503e-05, "loss": 1.0218, "step": 143650 }, { "epoch": 0.36, "learning_rate": 6.411495294176859e-05, "loss": 1.0207, "step": 143655 }, { "epoch": 0.36, "learning_rate": 6.411369470028688e-05, "loss": 1.0193, "step": 143660 }, { "epoch": 0.36, "learning_rate": 6.411243645880519e-05, "loss": 1.0214, "step": 143665 }, { "epoch": 0.36, "learning_rate": 6.411117821732348e-05, "loss": 1.0219, "step": 143670 }, { "epoch": 0.36, "learning_rate": 6.410991997584177e-05, "loss": 1.0202, "step": 143675 }, { "epoch": 0.36, "learning_rate": 6.410866173436006e-05, "loss": 1.022, "step": 143680 }, { "epoch": 0.36, "learning_rate": 6.410740349287837e-05, "loss": 1.0215, "step": 143685 }, { "epoch": 0.36, "learning_rate": 6.410614525139666e-05, "loss": 1.0231, "step": 143690 }, { "epoch": 0.36, "learning_rate": 6.410488700991495e-05, "loss": 1.0208, "step": 143695 }, { "epoch": 0.36, "learning_rate": 6.410362876843324e-05, "loss": 1.0195, "step": 143700 }, { "epoch": 0.36, "learning_rate": 6.410237052695153e-05, "loss": 1.0221, "step": 143705 }, { "epoch": 0.36, "learning_rate": 6.410111228546984e-05, "loss": 1.0217, "step": 143710 }, { "epoch": 0.36, "learning_rate": 6.409985404398813e-05, "loss": 1.0204, "step": 143715 }, { "epoch": 0.36, "learning_rate": 6.409859580250642e-05, "loss": 1.0197, "step": 143720 }, { "epoch": 0.36, "learning_rate": 6.409733756102471e-05, "loss": 1.0212, "step": 143725 }, { "epoch": 0.36, "learning_rate": 6.409607931954302e-05, "loss": 1.0212, "step": 143730 }, { "epoch": 0.36, "learning_rate": 6.409482107806131e-05, "loss": 1.0208, "step": 143735 }, { "epoch": 0.36, "learning_rate": 6.40935628365796e-05, "loss": 1.0189, "step": 143740 }, { "epoch": 0.36, "learning_rate": 6.409230459509789e-05, "loss": 1.0225, "step": 143745 }, { "epoch": 0.36, "learning_rate": 6.40910463536162e-05, "loss": 1.0218, "step": 143750 }, { "epoch": 0.36, "learning_rate": 6.408978811213449e-05, "loss": 1.0217, "step": 143755 }, { "epoch": 0.36, "learning_rate": 6.408852987065278e-05, "loss": 1.0196, "step": 143760 }, { "epoch": 0.36, "learning_rate": 6.408727162917107e-05, "loss": 1.0199, "step": 143765 }, { "epoch": 0.36, "learning_rate": 6.408601338768936e-05, "loss": 1.02, "step": 143770 }, { "epoch": 0.36, "learning_rate": 6.408475514620767e-05, "loss": 1.0204, "step": 143775 }, { "epoch": 0.36, "learning_rate": 6.408349690472596e-05, "loss": 1.0204, "step": 143780 }, { "epoch": 0.36, "learning_rate": 6.408223866324425e-05, "loss": 1.0202, "step": 143785 }, { "epoch": 0.36, "learning_rate": 6.408098042176254e-05, "loss": 1.0199, "step": 143790 }, { "epoch": 0.36, "learning_rate": 6.407972218028085e-05, "loss": 1.0212, "step": 143795 }, { "epoch": 0.36, "learning_rate": 6.407846393879914e-05, "loss": 1.0211, "step": 143800 }, { "epoch": 0.36, "learning_rate": 6.407720569731743e-05, "loss": 1.0214, "step": 143805 }, { "epoch": 0.36, "learning_rate": 6.407594745583572e-05, "loss": 1.0214, "step": 143810 }, { "epoch": 0.36, "learning_rate": 6.407468921435403e-05, "loss": 1.0214, "step": 143815 }, { "epoch": 0.36, "learning_rate": 6.407343097287232e-05, "loss": 1.0217, "step": 143820 }, { "epoch": 0.36, "learning_rate": 6.407217273139061e-05, "loss": 1.0202, "step": 143825 }, { "epoch": 0.36, "learning_rate": 6.40709144899089e-05, "loss": 1.0198, "step": 143830 }, { "epoch": 0.36, "learning_rate": 6.40696562484272e-05, "loss": 1.022, "step": 143835 }, { "epoch": 0.36, "learning_rate": 6.40683980069455e-05, "loss": 1.0219, "step": 143840 }, { "epoch": 0.36, "learning_rate": 6.406713976546379e-05, "loss": 1.0201, "step": 143845 }, { "epoch": 0.36, "learning_rate": 6.406588152398208e-05, "loss": 1.0208, "step": 143850 }, { "epoch": 0.36, "learning_rate": 6.406462328250037e-05, "loss": 1.0202, "step": 143855 }, { "epoch": 0.36, "learning_rate": 6.406336504101868e-05, "loss": 1.0194, "step": 143860 }, { "epoch": 0.36, "learning_rate": 6.406210679953697e-05, "loss": 1.0431, "step": 143865 }, { "epoch": 0.36, "learning_rate": 6.406084855805526e-05, "loss": 1.0217, "step": 143870 }, { "epoch": 0.36, "learning_rate": 6.405959031657355e-05, "loss": 1.0205, "step": 143875 }, { "epoch": 0.36, "learning_rate": 6.405833207509186e-05, "loss": 1.0234, "step": 143880 }, { "epoch": 0.36, "learning_rate": 6.405707383361015e-05, "loss": 1.0214, "step": 143885 }, { "epoch": 0.36, "learning_rate": 6.405581559212844e-05, "loss": 1.0204, "step": 143890 }, { "epoch": 0.36, "learning_rate": 6.405455735064673e-05, "loss": 1.0226, "step": 143895 }, { "epoch": 0.36, "learning_rate": 6.405329910916503e-05, "loss": 1.0217, "step": 143900 }, { "epoch": 0.36, "learning_rate": 6.405204086768333e-05, "loss": 1.019, "step": 143905 }, { "epoch": 0.36, "learning_rate": 6.405078262620162e-05, "loss": 1.0216, "step": 143910 }, { "epoch": 0.36, "learning_rate": 6.404952438471991e-05, "loss": 1.0198, "step": 143915 }, { "epoch": 0.36, "learning_rate": 6.40482661432382e-05, "loss": 1.0206, "step": 143920 }, { "epoch": 0.36, "learning_rate": 6.404700790175651e-05, "loss": 1.02, "step": 143925 }, { "epoch": 0.36, "learning_rate": 6.40457496602748e-05, "loss": 1.0205, "step": 143930 }, { "epoch": 0.36, "learning_rate": 6.40444914187931e-05, "loss": 1.0224, "step": 143935 }, { "epoch": 0.36, "learning_rate": 6.404323317731139e-05, "loss": 1.0223, "step": 143940 }, { "epoch": 0.36, "learning_rate": 6.404197493582969e-05, "loss": 1.0218, "step": 143945 }, { "epoch": 0.36, "learning_rate": 6.404071669434798e-05, "loss": 1.021, "step": 143950 }, { "epoch": 0.36, "learning_rate": 6.403945845286627e-05, "loss": 1.0197, "step": 143955 }, { "epoch": 0.36, "learning_rate": 6.403820021138457e-05, "loss": 1.02, "step": 143960 }, { "epoch": 0.36, "learning_rate": 6.403694196990286e-05, "loss": 1.0206, "step": 143965 }, { "epoch": 0.36, "learning_rate": 6.403568372842116e-05, "loss": 1.0201, "step": 143970 }, { "epoch": 0.36, "learning_rate": 6.403442548693945e-05, "loss": 1.0215, "step": 143975 }, { "epoch": 0.36, "learning_rate": 6.403316724545775e-05, "loss": 1.0222, "step": 143980 }, { "epoch": 0.36, "learning_rate": 6.403190900397604e-05, "loss": 1.0213, "step": 143985 }, { "epoch": 0.36, "learning_rate": 6.403065076249434e-05, "loss": 1.0245, "step": 143990 }, { "epoch": 0.36, "learning_rate": 6.402939252101263e-05, "loss": 1.0199, "step": 143995 }, { "epoch": 0.36, "learning_rate": 6.402813427953093e-05, "loss": 1.0192, "step": 144000 }, { "epoch": 0.36, "learning_rate": 6.402687603804922e-05, "loss": 1.0204, "step": 144005 }, { "epoch": 0.36, "learning_rate": 6.402561779656752e-05, "loss": 1.022, "step": 144010 }, { "epoch": 0.36, "learning_rate": 6.402435955508581e-05, "loss": 1.0219, "step": 144015 }, { "epoch": 0.36, "learning_rate": 6.40231013136041e-05, "loss": 1.021, "step": 144020 }, { "epoch": 0.36, "learning_rate": 6.40218430721224e-05, "loss": 1.0188, "step": 144025 }, { "epoch": 0.36, "learning_rate": 6.402058483064069e-05, "loss": 1.0186, "step": 144030 }, { "epoch": 0.36, "learning_rate": 6.4019326589159e-05, "loss": 1.0199, "step": 144035 }, { "epoch": 0.36, "learning_rate": 6.401806834767729e-05, "loss": 1.0199, "step": 144040 }, { "epoch": 0.36, "learning_rate": 6.401681010619558e-05, "loss": 1.0218, "step": 144045 }, { "epoch": 0.36, "learning_rate": 6.401555186471387e-05, "loss": 1.021, "step": 144050 }, { "epoch": 0.36, "learning_rate": 6.401429362323217e-05, "loss": 1.0216, "step": 144055 }, { "epoch": 0.36, "learning_rate": 6.401303538175047e-05, "loss": 1.0207, "step": 144060 }, { "epoch": 0.36, "learning_rate": 6.401177714026876e-05, "loss": 1.0193, "step": 144065 }, { "epoch": 0.36, "learning_rate": 6.401051889878705e-05, "loss": 1.021, "step": 144070 }, { "epoch": 0.36, "learning_rate": 6.400926065730535e-05, "loss": 1.0216, "step": 144075 }, { "epoch": 0.36, "learning_rate": 6.400800241582365e-05, "loss": 1.0212, "step": 144080 }, { "epoch": 0.36, "learning_rate": 6.400674417434195e-05, "loss": 1.0184, "step": 144085 }, { "epoch": 0.36, "learning_rate": 6.400548593286024e-05, "loss": 1.0208, "step": 144090 }, { "epoch": 0.36, "learning_rate": 6.400422769137853e-05, "loss": 1.0179, "step": 144095 }, { "epoch": 0.36, "learning_rate": 6.400296944989683e-05, "loss": 1.0202, "step": 144100 }, { "epoch": 0.36, "learning_rate": 6.400171120841513e-05, "loss": 1.0223, "step": 144105 }, { "epoch": 0.36, "learning_rate": 6.400045296693342e-05, "loss": 1.0178, "step": 144110 }, { "epoch": 0.36, "learning_rate": 6.399919472545171e-05, "loss": 1.0225, "step": 144115 }, { "epoch": 0.36, "learning_rate": 6.399793648397e-05, "loss": 1.0212, "step": 144120 }, { "epoch": 0.36, "learning_rate": 6.399667824248831e-05, "loss": 1.0194, "step": 144125 }, { "epoch": 0.36, "learning_rate": 6.39954200010066e-05, "loss": 1.0219, "step": 144130 }, { "epoch": 0.36, "learning_rate": 6.39941617595249e-05, "loss": 1.0223, "step": 144135 }, { "epoch": 0.36, "learning_rate": 6.399290351804319e-05, "loss": 1.0201, "step": 144140 }, { "epoch": 0.36, "learning_rate": 6.399164527656149e-05, "loss": 1.021, "step": 144145 }, { "epoch": 0.36, "learning_rate": 6.399038703507978e-05, "loss": 1.025, "step": 144150 }, { "epoch": 0.36, "learning_rate": 6.398912879359807e-05, "loss": 1.0205, "step": 144155 }, { "epoch": 0.36, "learning_rate": 6.398787055211637e-05, "loss": 1.0198, "step": 144160 }, { "epoch": 0.36, "learning_rate": 6.398661231063466e-05, "loss": 1.0201, "step": 144165 }, { "epoch": 0.36, "learning_rate": 6.398535406915296e-05, "loss": 1.0203, "step": 144170 }, { "epoch": 0.36, "learning_rate": 6.398409582767125e-05, "loss": 1.0216, "step": 144175 }, { "epoch": 0.36, "learning_rate": 6.398283758618955e-05, "loss": 1.0215, "step": 144180 }, { "epoch": 0.36, "learning_rate": 6.398157934470784e-05, "loss": 1.0205, "step": 144185 }, { "epoch": 0.36, "learning_rate": 6.398032110322614e-05, "loss": 1.0438, "step": 144190 }, { "epoch": 0.36, "learning_rate": 6.397906286174443e-05, "loss": 1.0195, "step": 144195 }, { "epoch": 0.36, "learning_rate": 6.397780462026273e-05, "loss": 1.0202, "step": 144200 }, { "epoch": 0.36, "learning_rate": 6.397654637878102e-05, "loss": 1.0227, "step": 144205 }, { "epoch": 0.36, "learning_rate": 6.397528813729932e-05, "loss": 1.0212, "step": 144210 }, { "epoch": 0.36, "learning_rate": 6.397402989581761e-05, "loss": 1.0202, "step": 144215 }, { "epoch": 0.36, "learning_rate": 6.39727716543359e-05, "loss": 1.0199, "step": 144220 }, { "epoch": 0.36, "learning_rate": 6.39715134128542e-05, "loss": 1.0212, "step": 144225 }, { "epoch": 0.36, "learning_rate": 6.397025517137249e-05, "loss": 1.0186, "step": 144230 }, { "epoch": 0.36, "learning_rate": 6.39689969298908e-05, "loss": 1.0216, "step": 144235 }, { "epoch": 0.36, "learning_rate": 6.396773868840909e-05, "loss": 1.0186, "step": 144240 }, { "epoch": 0.36, "learning_rate": 6.396648044692738e-05, "loss": 1.0437, "step": 144245 }, { "epoch": 0.36, "learning_rate": 6.396522220544567e-05, "loss": 1.0214, "step": 144250 }, { "epoch": 0.36, "learning_rate": 6.396396396396397e-05, "loss": 1.0196, "step": 144255 }, { "epoch": 0.36, "learning_rate": 6.396270572248227e-05, "loss": 1.0192, "step": 144260 }, { "epoch": 0.36, "learning_rate": 6.396144748100056e-05, "loss": 1.0214, "step": 144265 }, { "epoch": 0.36, "learning_rate": 6.396018923951885e-05, "loss": 1.0177, "step": 144270 }, { "epoch": 0.36, "learning_rate": 6.395893099803715e-05, "loss": 1.0197, "step": 144275 }, { "epoch": 0.36, "learning_rate": 6.395767275655545e-05, "loss": 1.0262, "step": 144280 }, { "epoch": 0.36, "learning_rate": 6.395641451507374e-05, "loss": 1.0212, "step": 144285 }, { "epoch": 0.36, "learning_rate": 6.395515627359203e-05, "loss": 1.0211, "step": 144290 }, { "epoch": 0.36, "learning_rate": 6.395389803211032e-05, "loss": 1.0209, "step": 144295 }, { "epoch": 0.36, "learning_rate": 6.395263979062863e-05, "loss": 1.0208, "step": 144300 }, { "epoch": 0.36, "learning_rate": 6.395138154914692e-05, "loss": 1.0218, "step": 144305 }, { "epoch": 0.36, "learning_rate": 6.395012330766521e-05, "loss": 1.0207, "step": 144310 }, { "epoch": 0.36, "learning_rate": 6.39488650661835e-05, "loss": 1.0195, "step": 144315 }, { "epoch": 0.36, "learning_rate": 6.39476068247018e-05, "loss": 1.0214, "step": 144320 }, { "epoch": 0.36, "learning_rate": 6.39463485832201e-05, "loss": 1.0206, "step": 144325 }, { "epoch": 0.36, "learning_rate": 6.394509034173839e-05, "loss": 1.0209, "step": 144330 }, { "epoch": 0.36, "learning_rate": 6.394383210025668e-05, "loss": 1.0223, "step": 144335 }, { "epoch": 0.36, "learning_rate": 6.394257385877498e-05, "loss": 1.0212, "step": 144340 }, { "epoch": 0.36, "learning_rate": 6.394131561729328e-05, "loss": 1.0187, "step": 144345 }, { "epoch": 0.36, "learning_rate": 6.394005737581157e-05, "loss": 1.0205, "step": 144350 }, { "epoch": 0.36, "learning_rate": 6.393879913432986e-05, "loss": 1.0172, "step": 144355 }, { "epoch": 0.36, "learning_rate": 6.393754089284815e-05, "loss": 1.0223, "step": 144360 }, { "epoch": 0.36, "learning_rate": 6.393628265136646e-05, "loss": 1.0214, "step": 144365 }, { "epoch": 0.36, "learning_rate": 6.393502440988475e-05, "loss": 1.0225, "step": 144370 }, { "epoch": 0.36, "learning_rate": 6.393376616840304e-05, "loss": 1.0191, "step": 144375 }, { "epoch": 0.36, "learning_rate": 6.393250792692133e-05, "loss": 1.0209, "step": 144380 }, { "epoch": 0.36, "learning_rate": 6.393124968543964e-05, "loss": 1.0223, "step": 144385 }, { "epoch": 0.36, "learning_rate": 6.392999144395793e-05, "loss": 1.0233, "step": 144390 }, { "epoch": 0.36, "learning_rate": 6.392873320247622e-05, "loss": 1.0221, "step": 144395 }, { "epoch": 0.36, "learning_rate": 6.392747496099451e-05, "loss": 1.0189, "step": 144400 }, { "epoch": 0.36, "learning_rate": 6.392621671951282e-05, "loss": 1.0207, "step": 144405 }, { "epoch": 0.36, "learning_rate": 6.392495847803111e-05, "loss": 1.0222, "step": 144410 }, { "epoch": 0.36, "learning_rate": 6.39237002365494e-05, "loss": 1.022, "step": 144415 }, { "epoch": 0.36, "learning_rate": 6.392244199506769e-05, "loss": 1.0206, "step": 144420 }, { "epoch": 0.36, "learning_rate": 6.392118375358598e-05, "loss": 1.0196, "step": 144425 }, { "epoch": 0.36, "learning_rate": 6.391992551210429e-05, "loss": 1.0188, "step": 144430 }, { "epoch": 0.36, "learning_rate": 6.391866727062258e-05, "loss": 1.0201, "step": 144435 }, { "epoch": 0.36, "learning_rate": 6.391740902914087e-05, "loss": 1.0201, "step": 144440 }, { "epoch": 0.36, "learning_rate": 6.391615078765916e-05, "loss": 1.0198, "step": 144445 }, { "epoch": 0.36, "learning_rate": 6.391489254617747e-05, "loss": 1.0189, "step": 144450 }, { "epoch": 0.36, "learning_rate": 6.391363430469576e-05, "loss": 1.0199, "step": 144455 }, { "epoch": 0.36, "learning_rate": 6.391237606321405e-05, "loss": 1.0182, "step": 144460 }, { "epoch": 0.36, "learning_rate": 6.391111782173234e-05, "loss": 1.022, "step": 144465 }, { "epoch": 0.36, "learning_rate": 6.390985958025065e-05, "loss": 1.0191, "step": 144470 }, { "epoch": 0.36, "learning_rate": 6.390860133876894e-05, "loss": 1.0196, "step": 144475 }, { "epoch": 0.36, "learning_rate": 6.390734309728723e-05, "loss": 1.0212, "step": 144480 }, { "epoch": 0.36, "learning_rate": 6.390608485580552e-05, "loss": 1.0191, "step": 144485 }, { "epoch": 0.36, "learning_rate": 6.390482661432381e-05, "loss": 1.0212, "step": 144490 }, { "epoch": 0.36, "learning_rate": 6.390356837284212e-05, "loss": 1.0219, "step": 144495 }, { "epoch": 0.36, "learning_rate": 6.390231013136041e-05, "loss": 1.0214, "step": 144500 }, { "epoch": 0.36, "learning_rate": 6.39010518898787e-05, "loss": 1.0206, "step": 144505 }, { "epoch": 0.36, "learning_rate": 6.3899793648397e-05, "loss": 1.0185, "step": 144510 }, { "epoch": 0.36, "learning_rate": 6.38985354069153e-05, "loss": 1.0206, "step": 144515 }, { "epoch": 0.36, "learning_rate": 6.389727716543359e-05, "loss": 1.0198, "step": 144520 }, { "epoch": 0.36, "learning_rate": 6.389601892395188e-05, "loss": 1.0219, "step": 144525 }, { "epoch": 0.36, "learning_rate": 6.389476068247017e-05, "loss": 1.0196, "step": 144530 }, { "epoch": 0.36, "learning_rate": 6.389350244098848e-05, "loss": 1.0213, "step": 144535 }, { "epoch": 0.36, "learning_rate": 6.389224419950677e-05, "loss": 1.0197, "step": 144540 }, { "epoch": 0.36, "learning_rate": 6.389098595802506e-05, "loss": 1.0199, "step": 144545 }, { "epoch": 0.36, "learning_rate": 6.388972771654335e-05, "loss": 1.0186, "step": 144550 }, { "epoch": 0.36, "learning_rate": 6.388846947506165e-05, "loss": 1.0202, "step": 144555 }, { "epoch": 0.36, "learning_rate": 6.388721123357995e-05, "loss": 1.0203, "step": 144560 }, { "epoch": 0.36, "learning_rate": 6.388595299209824e-05, "loss": 1.0232, "step": 144565 }, { "epoch": 0.36, "learning_rate": 6.388469475061653e-05, "loss": 1.0213, "step": 144570 }, { "epoch": 0.36, "learning_rate": 6.388343650913483e-05, "loss": 1.0218, "step": 144575 }, { "epoch": 0.36, "learning_rate": 6.388217826765313e-05, "loss": 1.0212, "step": 144580 }, { "epoch": 0.36, "learning_rate": 6.388092002617144e-05, "loss": 1.0208, "step": 144585 }, { "epoch": 0.36, "learning_rate": 6.387966178468973e-05, "loss": 1.0205, "step": 144590 }, { "epoch": 0.36, "learning_rate": 6.387840354320802e-05, "loss": 1.0209, "step": 144595 }, { "epoch": 0.36, "learning_rate": 6.387714530172631e-05, "loss": 1.022, "step": 144600 }, { "epoch": 0.36, "learning_rate": 6.387588706024462e-05, "loss": 1.0213, "step": 144605 }, { "epoch": 0.36, "learning_rate": 6.387462881876291e-05, "loss": 1.0193, "step": 144610 }, { "epoch": 0.36, "learning_rate": 6.38733705772812e-05, "loss": 1.0217, "step": 144615 }, { "epoch": 0.36, "learning_rate": 6.387211233579949e-05, "loss": 1.0203, "step": 144620 }, { "epoch": 0.36, "learning_rate": 6.38708540943178e-05, "loss": 1.021, "step": 144625 }, { "epoch": 0.36, "learning_rate": 6.386959585283609e-05, "loss": 1.022, "step": 144630 }, { "epoch": 0.36, "learning_rate": 6.386833761135438e-05, "loss": 1.0229, "step": 144635 }, { "epoch": 0.36, "learning_rate": 6.386707936987267e-05, "loss": 1.0187, "step": 144640 }, { "epoch": 0.36, "learning_rate": 6.386582112839096e-05, "loss": 1.0205, "step": 144645 }, { "epoch": 0.36, "learning_rate": 6.386456288690927e-05, "loss": 1.0203, "step": 144650 }, { "epoch": 0.36, "learning_rate": 6.386330464542756e-05, "loss": 1.0189, "step": 144655 }, { "epoch": 0.36, "learning_rate": 6.386204640394585e-05, "loss": 1.0194, "step": 144660 }, { "epoch": 0.36, "learning_rate": 6.386078816246414e-05, "loss": 1.0201, "step": 144665 }, { "epoch": 0.36, "learning_rate": 6.385952992098245e-05, "loss": 1.0199, "step": 144670 }, { "epoch": 0.36, "learning_rate": 6.385827167950074e-05, "loss": 1.0207, "step": 144675 }, { "epoch": 0.36, "learning_rate": 6.385701343801903e-05, "loss": 1.0195, "step": 144680 }, { "epoch": 0.36, "learning_rate": 6.385575519653732e-05, "loss": 1.022, "step": 144685 }, { "epoch": 0.36, "learning_rate": 6.385449695505563e-05, "loss": 1.0201, "step": 144690 }, { "epoch": 0.36, "learning_rate": 6.385323871357392e-05, "loss": 1.023, "step": 144695 }, { "epoch": 0.36, "learning_rate": 6.385198047209221e-05, "loss": 1.0204, "step": 144700 }, { "epoch": 0.36, "learning_rate": 6.38507222306105e-05, "loss": 1.0219, "step": 144705 }, { "epoch": 0.36, "learning_rate": 6.38494639891288e-05, "loss": 1.0205, "step": 144710 }, { "epoch": 0.36, "learning_rate": 6.38482057476471e-05, "loss": 1.0198, "step": 144715 }, { "epoch": 0.36, "learning_rate": 6.384694750616539e-05, "loss": 1.0205, "step": 144720 }, { "epoch": 0.36, "learning_rate": 6.384568926468368e-05, "loss": 1.0215, "step": 144725 }, { "epoch": 0.36, "learning_rate": 6.384443102320197e-05, "loss": 1.0195, "step": 144730 }, { "epoch": 0.36, "learning_rate": 6.384317278172028e-05, "loss": 1.0232, "step": 144735 }, { "epoch": 0.36, "learning_rate": 6.384191454023857e-05, "loss": 1.0205, "step": 144740 }, { "epoch": 0.36, "learning_rate": 6.384065629875686e-05, "loss": 1.0199, "step": 144745 }, { "epoch": 0.36, "learning_rate": 6.383939805727515e-05, "loss": 1.0183, "step": 144750 }, { "epoch": 0.36, "learning_rate": 6.383813981579345e-05, "loss": 1.019, "step": 144755 }, { "epoch": 0.36, "learning_rate": 6.383688157431175e-05, "loss": 1.0206, "step": 144760 }, { "epoch": 0.36, "learning_rate": 6.383562333283004e-05, "loss": 1.0194, "step": 144765 }, { "epoch": 0.36, "learning_rate": 6.383436509134833e-05, "loss": 1.0193, "step": 144770 }, { "epoch": 0.36, "learning_rate": 6.383310684986663e-05, "loss": 1.0199, "step": 144775 }, { "epoch": 0.36, "learning_rate": 6.383184860838493e-05, "loss": 1.0219, "step": 144780 }, { "epoch": 0.36, "learning_rate": 6.383059036690322e-05, "loss": 1.019, "step": 144785 }, { "epoch": 0.36, "learning_rate": 6.382933212542151e-05, "loss": 1.0227, "step": 144790 }, { "epoch": 0.36, "learning_rate": 6.38280738839398e-05, "loss": 1.021, "step": 144795 }, { "epoch": 0.36, "learning_rate": 6.382681564245811e-05, "loss": 1.019, "step": 144800 }, { "epoch": 0.36, "learning_rate": 6.38255574009764e-05, "loss": 1.0231, "step": 144805 }, { "epoch": 0.36, "learning_rate": 6.38242991594947e-05, "loss": 1.0182, "step": 144810 }, { "epoch": 0.36, "learning_rate": 6.382304091801298e-05, "loss": 1.0207, "step": 144815 }, { "epoch": 0.36, "learning_rate": 6.382178267653128e-05, "loss": 1.0175, "step": 144820 }, { "epoch": 0.36, "learning_rate": 6.382052443504958e-05, "loss": 1.0214, "step": 144825 }, { "epoch": 0.36, "learning_rate": 6.381926619356787e-05, "loss": 1.0211, "step": 144830 }, { "epoch": 0.36, "learning_rate": 6.381800795208616e-05, "loss": 1.0241, "step": 144835 }, { "epoch": 0.36, "learning_rate": 6.381674971060446e-05, "loss": 1.0198, "step": 144840 }, { "epoch": 0.36, "learning_rate": 6.381549146912276e-05, "loss": 1.0201, "step": 144845 }, { "epoch": 0.36, "learning_rate": 6.381423322764105e-05, "loss": 1.0211, "step": 144850 }, { "epoch": 0.36, "learning_rate": 6.381297498615934e-05, "loss": 1.0209, "step": 144855 }, { "epoch": 0.36, "learning_rate": 6.381171674467764e-05, "loss": 1.0206, "step": 144860 }, { "epoch": 0.36, "learning_rate": 6.381045850319594e-05, "loss": 1.0213, "step": 144865 }, { "epoch": 0.36, "learning_rate": 6.380920026171423e-05, "loss": 1.0205, "step": 144870 }, { "epoch": 0.36, "learning_rate": 6.380794202023252e-05, "loss": 1.0205, "step": 144875 }, { "epoch": 0.36, "learning_rate": 6.380668377875082e-05, "loss": 1.02, "step": 144880 }, { "epoch": 0.36, "learning_rate": 6.380542553726911e-05, "loss": 1.0194, "step": 144885 }, { "epoch": 0.36, "learning_rate": 6.380416729578741e-05, "loss": 1.0205, "step": 144890 }, { "epoch": 0.36, "learning_rate": 6.38029090543057e-05, "loss": 1.021, "step": 144895 }, { "epoch": 0.36, "learning_rate": 6.3801650812824e-05, "loss": 1.0204, "step": 144900 }, { "epoch": 0.36, "learning_rate": 6.380039257134229e-05, "loss": 1.0208, "step": 144905 }, { "epoch": 0.36, "learning_rate": 6.379913432986059e-05, "loss": 1.0185, "step": 144910 }, { "epoch": 0.36, "learning_rate": 6.379787608837888e-05, "loss": 1.0197, "step": 144915 }, { "epoch": 0.36, "learning_rate": 6.379661784689718e-05, "loss": 1.0196, "step": 144920 }, { "epoch": 0.36, "learning_rate": 6.379535960541547e-05, "loss": 1.0219, "step": 144925 }, { "epoch": 0.36, "learning_rate": 6.379410136393377e-05, "loss": 1.0197, "step": 144930 }, { "epoch": 0.36, "learning_rate": 6.379284312245206e-05, "loss": 1.0212, "step": 144935 }, { "epoch": 0.36, "learning_rate": 6.379158488097036e-05, "loss": 1.0216, "step": 144940 }, { "epoch": 0.36, "learning_rate": 6.379032663948865e-05, "loss": 1.0198, "step": 144945 }, { "epoch": 0.36, "learning_rate": 6.378906839800694e-05, "loss": 1.0213, "step": 144950 }, { "epoch": 0.36, "learning_rate": 6.378781015652524e-05, "loss": 1.023, "step": 144955 }, { "epoch": 0.36, "learning_rate": 6.378655191504354e-05, "loss": 1.02, "step": 144960 }, { "epoch": 0.36, "learning_rate": 6.378529367356183e-05, "loss": 1.0212, "step": 144965 }, { "epoch": 0.36, "learning_rate": 6.378403543208012e-05, "loss": 1.0241, "step": 144970 }, { "epoch": 0.36, "learning_rate": 6.378277719059842e-05, "loss": 1.0194, "step": 144975 }, { "epoch": 0.36, "learning_rate": 6.378151894911672e-05, "loss": 1.0466, "step": 144980 }, { "epoch": 0.36, "learning_rate": 6.378026070763501e-05, "loss": 1.0211, "step": 144985 }, { "epoch": 0.36, "learning_rate": 6.37790024661533e-05, "loss": 1.0206, "step": 144990 }, { "epoch": 0.36, "learning_rate": 6.37777442246716e-05, "loss": 1.02, "step": 144995 }, { "epoch": 0.36, "learning_rate": 6.37764859831899e-05, "loss": 1.0199, "step": 145000 }, { "epoch": 0.36, "learning_rate": 6.377522774170819e-05, "loss": 1.021, "step": 145005 }, { "epoch": 0.36, "learning_rate": 6.377396950022648e-05, "loss": 1.0224, "step": 145010 }, { "epoch": 0.36, "learning_rate": 6.377271125874477e-05, "loss": 1.0198, "step": 145015 }, { "epoch": 0.36, "learning_rate": 6.377145301726308e-05, "loss": 1.0195, "step": 145020 }, { "epoch": 0.36, "learning_rate": 6.377019477578137e-05, "loss": 1.0209, "step": 145025 }, { "epoch": 0.36, "learning_rate": 6.376893653429966e-05, "loss": 1.0197, "step": 145030 }, { "epoch": 0.36, "learning_rate": 6.376767829281795e-05, "loss": 1.0197, "step": 145035 }, { "epoch": 0.36, "learning_rate": 6.376642005133626e-05, "loss": 1.0217, "step": 145040 }, { "epoch": 0.36, "learning_rate": 6.376516180985455e-05, "loss": 1.0203, "step": 145045 }, { "epoch": 0.36, "learning_rate": 6.376390356837284e-05, "loss": 1.0195, "step": 145050 }, { "epoch": 0.36, "learning_rate": 6.376264532689113e-05, "loss": 1.0231, "step": 145055 }, { "epoch": 0.36, "learning_rate": 6.376138708540944e-05, "loss": 1.0219, "step": 145060 }, { "epoch": 0.36, "learning_rate": 6.376012884392773e-05, "loss": 1.0194, "step": 145065 }, { "epoch": 0.36, "learning_rate": 6.375887060244602e-05, "loss": 1.0183, "step": 145070 }, { "epoch": 0.36, "learning_rate": 6.375761236096431e-05, "loss": 1.0227, "step": 145075 }, { "epoch": 0.36, "learning_rate": 6.37563541194826e-05, "loss": 1.0213, "step": 145080 }, { "epoch": 0.36, "learning_rate": 6.375509587800092e-05, "loss": 1.0193, "step": 145085 }, { "epoch": 0.36, "learning_rate": 6.375383763651921e-05, "loss": 1.0204, "step": 145090 }, { "epoch": 0.36, "learning_rate": 6.37525793950375e-05, "loss": 1.0205, "step": 145095 }, { "epoch": 0.36, "learning_rate": 6.37513211535558e-05, "loss": 1.0211, "step": 145100 }, { "epoch": 0.36, "learning_rate": 6.375006291207409e-05, "loss": 1.0218, "step": 145105 }, { "epoch": 0.36, "learning_rate": 6.374880467059239e-05, "loss": 1.0214, "step": 145110 }, { "epoch": 0.36, "learning_rate": 6.374754642911068e-05, "loss": 1.021, "step": 145115 }, { "epoch": 0.36, "learning_rate": 6.374628818762898e-05, "loss": 1.0208, "step": 145120 }, { "epoch": 0.36, "learning_rate": 6.374502994614727e-05, "loss": 1.0202, "step": 145125 }, { "epoch": 0.36, "learning_rate": 6.374377170466557e-05, "loss": 1.0216, "step": 145130 }, { "epoch": 0.36, "learning_rate": 6.374251346318386e-05, "loss": 1.0195, "step": 145135 }, { "epoch": 0.36, "learning_rate": 6.374125522170216e-05, "loss": 1.0188, "step": 145140 }, { "epoch": 0.36, "learning_rate": 6.373999698022045e-05, "loss": 1.0201, "step": 145145 }, { "epoch": 0.36, "learning_rate": 6.373873873873875e-05, "loss": 1.021, "step": 145150 }, { "epoch": 0.36, "learning_rate": 6.373748049725704e-05, "loss": 1.0221, "step": 145155 }, { "epoch": 0.36, "learning_rate": 6.373622225577534e-05, "loss": 1.0203, "step": 145160 }, { "epoch": 0.36, "learning_rate": 6.373496401429363e-05, "loss": 1.0207, "step": 145165 }, { "epoch": 0.36, "learning_rate": 6.373370577281192e-05, "loss": 1.0204, "step": 145170 }, { "epoch": 0.36, "learning_rate": 6.373244753133022e-05, "loss": 1.0208, "step": 145175 }, { "epoch": 0.36, "learning_rate": 6.373118928984852e-05, "loss": 1.0215, "step": 145180 }, { "epoch": 0.36, "learning_rate": 6.372993104836681e-05, "loss": 1.0206, "step": 145185 }, { "epoch": 0.36, "learning_rate": 6.37286728068851e-05, "loss": 1.0227, "step": 145190 }, { "epoch": 0.36, "learning_rate": 6.37274145654034e-05, "loss": 1.0216, "step": 145195 }, { "epoch": 0.36, "learning_rate": 6.37261563239217e-05, "loss": 1.0217, "step": 145200 }, { "epoch": 0.36, "learning_rate": 6.372489808243999e-05, "loss": 1.02, "step": 145205 }, { "epoch": 0.36, "learning_rate": 6.372363984095828e-05, "loss": 1.0202, "step": 145210 }, { "epoch": 0.36, "learning_rate": 6.372238159947658e-05, "loss": 1.0195, "step": 145215 }, { "epoch": 0.36, "learning_rate": 6.372112335799488e-05, "loss": 1.0219, "step": 145220 }, { "epoch": 0.36, "learning_rate": 6.371986511651317e-05, "loss": 1.0192, "step": 145225 }, { "epoch": 0.36, "learning_rate": 6.371860687503146e-05, "loss": 1.0211, "step": 145230 }, { "epoch": 0.36, "learning_rate": 6.371734863354975e-05, "loss": 1.0366, "step": 145235 }, { "epoch": 0.36, "learning_rate": 6.371609039206806e-05, "loss": 1.0207, "step": 145240 }, { "epoch": 0.36, "learning_rate": 6.371483215058635e-05, "loss": 1.0193, "step": 145245 }, { "epoch": 0.36, "learning_rate": 6.371357390910464e-05, "loss": 1.0215, "step": 145250 }, { "epoch": 0.36, "learning_rate": 6.371231566762293e-05, "loss": 1.0209, "step": 145255 }, { "epoch": 0.36, "learning_rate": 6.371105742614124e-05, "loss": 1.0213, "step": 145260 }, { "epoch": 0.36, "learning_rate": 6.370979918465953e-05, "loss": 1.0191, "step": 145265 }, { "epoch": 0.36, "learning_rate": 6.370854094317782e-05, "loss": 1.0213, "step": 145270 }, { "epoch": 0.36, "learning_rate": 6.370728270169611e-05, "loss": 1.0208, "step": 145275 }, { "epoch": 0.36, "learning_rate": 6.370602446021442e-05, "loss": 1.022, "step": 145280 }, { "epoch": 0.36, "learning_rate": 6.37047662187327e-05, "loss": 1.0205, "step": 145285 }, { "epoch": 0.36, "learning_rate": 6.3703507977251e-05, "loss": 1.0204, "step": 145290 }, { "epoch": 0.36, "learning_rate": 6.370224973576929e-05, "loss": 1.0231, "step": 145295 }, { "epoch": 0.36, "learning_rate": 6.370099149428758e-05, "loss": 1.0215, "step": 145300 }, { "epoch": 0.36, "learning_rate": 6.369973325280589e-05, "loss": 1.0217, "step": 145305 }, { "epoch": 0.36, "learning_rate": 6.369847501132418e-05, "loss": 1.0214, "step": 145310 }, { "epoch": 0.36, "learning_rate": 6.369721676984247e-05, "loss": 1.0193, "step": 145315 }, { "epoch": 0.36, "learning_rate": 6.369595852836076e-05, "loss": 1.0216, "step": 145320 }, { "epoch": 0.36, "learning_rate": 6.369470028687907e-05, "loss": 1.0215, "step": 145325 }, { "epoch": 0.36, "learning_rate": 6.369344204539736e-05, "loss": 1.0194, "step": 145330 }, { "epoch": 0.36, "learning_rate": 6.369218380391565e-05, "loss": 1.0184, "step": 145335 }, { "epoch": 0.36, "learning_rate": 6.369092556243394e-05, "loss": 1.0209, "step": 145340 }, { "epoch": 0.36, "learning_rate": 6.368966732095225e-05, "loss": 1.0198, "step": 145345 }, { "epoch": 0.36, "learning_rate": 6.368840907947054e-05, "loss": 1.0206, "step": 145350 }, { "epoch": 0.36, "learning_rate": 6.368715083798883e-05, "loss": 1.0216, "step": 145355 }, { "epoch": 0.36, "learning_rate": 6.368589259650712e-05, "loss": 1.0213, "step": 145360 }, { "epoch": 0.36, "learning_rate": 6.368463435502541e-05, "loss": 1.0206, "step": 145365 }, { "epoch": 0.36, "learning_rate": 6.368337611354372e-05, "loss": 1.0227, "step": 145370 }, { "epoch": 0.36, "learning_rate": 6.368211787206201e-05, "loss": 1.0213, "step": 145375 }, { "epoch": 0.36, "learning_rate": 6.36808596305803e-05, "loss": 1.0199, "step": 145380 }, { "epoch": 0.36, "learning_rate": 6.367960138909859e-05, "loss": 1.0209, "step": 145385 }, { "epoch": 0.36, "learning_rate": 6.36783431476169e-05, "loss": 1.0211, "step": 145390 }, { "epoch": 0.36, "learning_rate": 6.367708490613519e-05, "loss": 1.0227, "step": 145395 }, { "epoch": 0.36, "learning_rate": 6.367582666465348e-05, "loss": 1.0231, "step": 145400 }, { "epoch": 0.36, "learning_rate": 6.367456842317177e-05, "loss": 1.0235, "step": 145405 }, { "epoch": 0.37, "learning_rate": 6.367331018169006e-05, "loss": 1.0185, "step": 145410 }, { "epoch": 0.37, "learning_rate": 6.367205194020837e-05, "loss": 1.022, "step": 145415 }, { "epoch": 0.37, "learning_rate": 6.367079369872666e-05, "loss": 1.0224, "step": 145420 }, { "epoch": 0.37, "learning_rate": 6.366953545724495e-05, "loss": 1.0195, "step": 145425 }, { "epoch": 0.37, "learning_rate": 6.366827721576324e-05, "loss": 1.0202, "step": 145430 }, { "epoch": 0.37, "learning_rate": 6.366701897428155e-05, "loss": 1.0194, "step": 145435 }, { "epoch": 0.37, "learning_rate": 6.366576073279984e-05, "loss": 1.0218, "step": 145440 }, { "epoch": 0.37, "learning_rate": 6.366450249131813e-05, "loss": 1.0205, "step": 145445 }, { "epoch": 0.37, "learning_rate": 6.366324424983642e-05, "loss": 1.0205, "step": 145450 }, { "epoch": 0.37, "learning_rate": 6.366198600835473e-05, "loss": 1.0223, "step": 145455 }, { "epoch": 0.37, "learning_rate": 6.366072776687302e-05, "loss": 1.0204, "step": 145460 }, { "epoch": 0.37, "learning_rate": 6.365946952539131e-05, "loss": 1.0221, "step": 145465 }, { "epoch": 0.37, "learning_rate": 6.36582112839096e-05, "loss": 1.0224, "step": 145470 }, { "epoch": 0.37, "learning_rate": 6.36569530424279e-05, "loss": 1.021, "step": 145475 }, { "epoch": 0.37, "learning_rate": 6.36556948009462e-05, "loss": 1.0226, "step": 145480 }, { "epoch": 0.37, "learning_rate": 6.365443655946449e-05, "loss": 1.0193, "step": 145485 }, { "epoch": 0.37, "learning_rate": 6.365317831798278e-05, "loss": 1.0218, "step": 145490 }, { "epoch": 0.37, "learning_rate": 6.365192007650108e-05, "loss": 1.0202, "step": 145495 }, { "epoch": 0.37, "learning_rate": 6.365066183501938e-05, "loss": 1.0196, "step": 145500 }, { "epoch": 0.37, "learning_rate": 6.364940359353767e-05, "loss": 1.0218, "step": 145505 }, { "epoch": 0.37, "learning_rate": 6.364814535205596e-05, "loss": 1.0207, "step": 145510 }, { "epoch": 0.37, "learning_rate": 6.364688711057426e-05, "loss": 1.0204, "step": 145515 }, { "epoch": 0.37, "learning_rate": 6.364562886909256e-05, "loss": 1.0441, "step": 145520 }, { "epoch": 0.37, "learning_rate": 6.364437062761085e-05, "loss": 1.0224, "step": 145525 }, { "epoch": 0.37, "learning_rate": 6.364311238612914e-05, "loss": 1.0218, "step": 145530 }, { "epoch": 0.37, "learning_rate": 6.364185414464744e-05, "loss": 1.0237, "step": 145535 }, { "epoch": 0.37, "learning_rate": 6.364059590316573e-05, "loss": 1.0228, "step": 145540 }, { "epoch": 0.37, "learning_rate": 6.363933766168403e-05, "loss": 1.0201, "step": 145545 }, { "epoch": 0.37, "learning_rate": 6.363807942020232e-05, "loss": 1.0202, "step": 145550 }, { "epoch": 0.37, "learning_rate": 6.363682117872062e-05, "loss": 1.0216, "step": 145555 }, { "epoch": 0.37, "learning_rate": 6.363556293723891e-05, "loss": 1.0188, "step": 145560 }, { "epoch": 0.37, "learning_rate": 6.363430469575721e-05, "loss": 1.0231, "step": 145565 }, { "epoch": 0.37, "learning_rate": 6.36330464542755e-05, "loss": 1.0208, "step": 145570 }, { "epoch": 0.37, "learning_rate": 6.36317882127938e-05, "loss": 1.0192, "step": 145575 }, { "epoch": 0.37, "learning_rate": 6.36305299713121e-05, "loss": 1.0207, "step": 145580 }, { "epoch": 0.37, "learning_rate": 6.362927172983039e-05, "loss": 1.0208, "step": 145585 }, { "epoch": 0.37, "learning_rate": 6.36280134883487e-05, "loss": 1.0212, "step": 145590 }, { "epoch": 0.37, "learning_rate": 6.362675524686699e-05, "loss": 1.0225, "step": 145595 }, { "epoch": 0.37, "learning_rate": 6.362549700538528e-05, "loss": 1.0214, "step": 145600 }, { "epoch": 0.37, "learning_rate": 6.362423876390357e-05, "loss": 1.0189, "step": 145605 }, { "epoch": 0.37, "learning_rate": 6.362298052242188e-05, "loss": 1.0203, "step": 145610 }, { "epoch": 0.37, "learning_rate": 6.362172228094017e-05, "loss": 1.0192, "step": 145615 }, { "epoch": 0.37, "learning_rate": 6.362046403945846e-05, "loss": 1.0219, "step": 145620 }, { "epoch": 0.37, "learning_rate": 6.361920579797675e-05, "loss": 1.0204, "step": 145625 }, { "epoch": 0.37, "learning_rate": 6.361794755649504e-05, "loss": 1.0197, "step": 145630 }, { "epoch": 0.37, "learning_rate": 6.361668931501335e-05, "loss": 1.0189, "step": 145635 }, { "epoch": 0.37, "learning_rate": 6.361543107353164e-05, "loss": 1.0195, "step": 145640 }, { "epoch": 0.37, "learning_rate": 6.361417283204993e-05, "loss": 1.0211, "step": 145645 }, { "epoch": 0.37, "learning_rate": 6.361291459056822e-05, "loss": 1.0195, "step": 145650 }, { "epoch": 0.37, "learning_rate": 6.361165634908653e-05, "loss": 1.0214, "step": 145655 }, { "epoch": 0.37, "learning_rate": 6.361039810760482e-05, "loss": 1.0196, "step": 145660 }, { "epoch": 0.37, "learning_rate": 6.360913986612311e-05, "loss": 1.0219, "step": 145665 }, { "epoch": 0.37, "learning_rate": 6.36078816246414e-05, "loss": 1.0203, "step": 145670 }, { "epoch": 0.37, "learning_rate": 6.360662338315971e-05, "loss": 1.0226, "step": 145675 }, { "epoch": 0.37, "learning_rate": 6.3605365141678e-05, "loss": 1.0207, "step": 145680 }, { "epoch": 0.37, "learning_rate": 6.360410690019629e-05, "loss": 1.0215, "step": 145685 }, { "epoch": 0.37, "learning_rate": 6.360284865871458e-05, "loss": 1.0202, "step": 145690 }, { "epoch": 0.37, "learning_rate": 6.360159041723288e-05, "loss": 1.0218, "step": 145695 }, { "epoch": 0.37, "learning_rate": 6.360033217575118e-05, "loss": 1.0214, "step": 145700 }, { "epoch": 0.37, "learning_rate": 6.359907393426947e-05, "loss": 1.0225, "step": 145705 }, { "epoch": 0.37, "learning_rate": 6.359781569278776e-05, "loss": 1.0219, "step": 145710 }, { "epoch": 0.37, "learning_rate": 6.359655745130606e-05, "loss": 1.0224, "step": 145715 }, { "epoch": 0.37, "learning_rate": 6.359529920982436e-05, "loss": 1.0205, "step": 145720 }, { "epoch": 0.37, "learning_rate": 6.359404096834265e-05, "loss": 1.0208, "step": 145725 }, { "epoch": 0.37, "learning_rate": 6.359278272686094e-05, "loss": 1.0205, "step": 145730 }, { "epoch": 0.37, "learning_rate": 6.359152448537924e-05, "loss": 1.022, "step": 145735 }, { "epoch": 0.37, "learning_rate": 6.359026624389754e-05, "loss": 1.0194, "step": 145740 }, { "epoch": 0.37, "learning_rate": 6.358900800241583e-05, "loss": 1.0194, "step": 145745 }, { "epoch": 0.37, "learning_rate": 6.358774976093412e-05, "loss": 1.0191, "step": 145750 }, { "epoch": 0.37, "learning_rate": 6.358649151945242e-05, "loss": 1.0216, "step": 145755 }, { "epoch": 0.37, "learning_rate": 6.35852332779707e-05, "loss": 1.0196, "step": 145760 }, { "epoch": 0.37, "learning_rate": 6.358397503648901e-05, "loss": 1.0202, "step": 145765 }, { "epoch": 0.37, "learning_rate": 6.35827167950073e-05, "loss": 1.0206, "step": 145770 }, { "epoch": 0.37, "learning_rate": 6.35814585535256e-05, "loss": 1.0181, "step": 145775 }, { "epoch": 0.37, "learning_rate": 6.358020031204389e-05, "loss": 1.0183, "step": 145780 }, { "epoch": 0.37, "learning_rate": 6.357894207056219e-05, "loss": 1.0218, "step": 145785 }, { "epoch": 0.37, "learning_rate": 6.357768382908048e-05, "loss": 1.0204, "step": 145790 }, { "epoch": 0.37, "learning_rate": 6.357642558759877e-05, "loss": 1.0206, "step": 145795 }, { "epoch": 0.37, "learning_rate": 6.357516734611707e-05, "loss": 1.0201, "step": 145800 }, { "epoch": 0.37, "learning_rate": 6.357390910463537e-05, "loss": 1.0204, "step": 145805 }, { "epoch": 0.37, "learning_rate": 6.357265086315366e-05, "loss": 1.0344, "step": 145810 }, { "epoch": 0.37, "learning_rate": 6.357139262167195e-05, "loss": 1.0211, "step": 145815 }, { "epoch": 0.37, "learning_rate": 6.357013438019025e-05, "loss": 1.0201, "step": 145820 }, { "epoch": 0.37, "learning_rate": 6.356887613870854e-05, "loss": 1.0207, "step": 145825 }, { "epoch": 0.37, "learning_rate": 6.356761789722684e-05, "loss": 1.0183, "step": 145830 }, { "epoch": 0.37, "learning_rate": 6.356635965574513e-05, "loss": 1.0209, "step": 145835 }, { "epoch": 0.37, "learning_rate": 6.356510141426343e-05, "loss": 1.021, "step": 145840 }, { "epoch": 0.37, "learning_rate": 6.356384317278172e-05, "loss": 1.0207, "step": 145845 }, { "epoch": 0.37, "learning_rate": 6.356258493130002e-05, "loss": 1.0198, "step": 145850 }, { "epoch": 0.37, "learning_rate": 6.356132668981831e-05, "loss": 1.0216, "step": 145855 }, { "epoch": 0.37, "learning_rate": 6.35600684483366e-05, "loss": 1.0213, "step": 145860 }, { "epoch": 0.37, "learning_rate": 6.35588102068549e-05, "loss": 1.0197, "step": 145865 }, { "epoch": 0.37, "learning_rate": 6.35575519653732e-05, "loss": 1.0192, "step": 145870 }, { "epoch": 0.37, "learning_rate": 6.35562937238915e-05, "loss": 1.0197, "step": 145875 }, { "epoch": 0.37, "learning_rate": 6.355503548240979e-05, "loss": 1.0206, "step": 145880 }, { "epoch": 0.37, "learning_rate": 6.355377724092808e-05, "loss": 1.0213, "step": 145885 }, { "epoch": 0.37, "learning_rate": 6.355251899944637e-05, "loss": 1.0238, "step": 145890 }, { "epoch": 0.37, "learning_rate": 6.355126075796467e-05, "loss": 1.0208, "step": 145895 }, { "epoch": 0.37, "learning_rate": 6.355000251648297e-05, "loss": 1.0224, "step": 145900 }, { "epoch": 0.37, "learning_rate": 6.354874427500126e-05, "loss": 1.0193, "step": 145905 }, { "epoch": 0.37, "learning_rate": 6.354748603351955e-05, "loss": 1.0193, "step": 145910 }, { "epoch": 0.37, "learning_rate": 6.354622779203785e-05, "loss": 1.0179, "step": 145915 }, { "epoch": 0.37, "learning_rate": 6.354496955055615e-05, "loss": 1.021, "step": 145920 }, { "epoch": 0.37, "learning_rate": 6.354371130907444e-05, "loss": 1.0192, "step": 145925 }, { "epoch": 0.37, "learning_rate": 6.354245306759273e-05, "loss": 1.0221, "step": 145930 }, { "epoch": 0.37, "learning_rate": 6.354119482611103e-05, "loss": 1.0214, "step": 145935 }, { "epoch": 0.37, "learning_rate": 6.353993658462933e-05, "loss": 1.0231, "step": 145940 }, { "epoch": 0.37, "learning_rate": 6.353867834314762e-05, "loss": 1.0197, "step": 145945 }, { "epoch": 0.37, "learning_rate": 6.353742010166591e-05, "loss": 1.0201, "step": 145950 }, { "epoch": 0.37, "learning_rate": 6.35361618601842e-05, "loss": 1.0226, "step": 145955 }, { "epoch": 0.37, "learning_rate": 6.35349036187025e-05, "loss": 1.0196, "step": 145960 }, { "epoch": 0.37, "learning_rate": 6.35336453772208e-05, "loss": 1.0201, "step": 145965 }, { "epoch": 0.37, "learning_rate": 6.353238713573909e-05, "loss": 1.0196, "step": 145970 }, { "epoch": 0.37, "learning_rate": 6.353112889425738e-05, "loss": 1.0452, "step": 145975 }, { "epoch": 0.37, "learning_rate": 6.352987065277569e-05, "loss": 1.0211, "step": 145980 }, { "epoch": 0.37, "learning_rate": 6.352861241129398e-05, "loss": 1.0197, "step": 145985 }, { "epoch": 0.37, "learning_rate": 6.352735416981227e-05, "loss": 1.0209, "step": 145990 }, { "epoch": 0.37, "learning_rate": 6.352609592833056e-05, "loss": 1.0238, "step": 145995 }, { "epoch": 0.37, "learning_rate": 6.352483768684887e-05, "loss": 1.0189, "step": 146000 }, { "epoch": 0.37, "learning_rate": 6.352357944536716e-05, "loss": 1.0216, "step": 146005 }, { "epoch": 0.37, "learning_rate": 6.352232120388545e-05, "loss": 1.0233, "step": 146010 }, { "epoch": 0.37, "learning_rate": 6.352106296240374e-05, "loss": 1.0226, "step": 146015 }, { "epoch": 0.37, "learning_rate": 6.351980472092203e-05, "loss": 1.0225, "step": 146020 }, { "epoch": 0.37, "learning_rate": 6.351854647944034e-05, "loss": 1.0212, "step": 146025 }, { "epoch": 0.37, "learning_rate": 6.351728823795863e-05, "loss": 1.0211, "step": 146030 }, { "epoch": 0.37, "learning_rate": 6.351602999647692e-05, "loss": 1.0372, "step": 146035 }, { "epoch": 0.37, "learning_rate": 6.351477175499521e-05, "loss": 1.0201, "step": 146040 }, { "epoch": 0.37, "learning_rate": 6.351351351351352e-05, "loss": 1.02, "step": 146045 }, { "epoch": 0.37, "learning_rate": 6.351225527203181e-05, "loss": 1.0218, "step": 146050 }, { "epoch": 0.37, "learning_rate": 6.35109970305501e-05, "loss": 1.0217, "step": 146055 }, { "epoch": 0.37, "learning_rate": 6.350973878906839e-05, "loss": 1.0214, "step": 146060 }, { "epoch": 0.37, "learning_rate": 6.350848054758668e-05, "loss": 1.0218, "step": 146065 }, { "epoch": 0.37, "learning_rate": 6.350722230610499e-05, "loss": 1.0191, "step": 146070 }, { "epoch": 0.37, "learning_rate": 6.350596406462328e-05, "loss": 1.0218, "step": 146075 }, { "epoch": 0.37, "learning_rate": 6.350470582314159e-05, "loss": 1.0201, "step": 146080 }, { "epoch": 0.37, "learning_rate": 6.350344758165988e-05, "loss": 1.0213, "step": 146085 }, { "epoch": 0.37, "learning_rate": 6.350218934017817e-05, "loss": 1.0238, "step": 146090 }, { "epoch": 0.37, "learning_rate": 6.350093109869647e-05, "loss": 1.0209, "step": 146095 }, { "epoch": 0.37, "learning_rate": 6.349967285721477e-05, "loss": 1.0211, "step": 146100 }, { "epoch": 0.37, "learning_rate": 6.349841461573306e-05, "loss": 1.0215, "step": 146105 }, { "epoch": 0.37, "learning_rate": 6.349715637425135e-05, "loss": 1.0439, "step": 146110 }, { "epoch": 0.37, "learning_rate": 6.349589813276965e-05, "loss": 1.02, "step": 146115 }, { "epoch": 0.37, "learning_rate": 6.349463989128795e-05, "loss": 1.0204, "step": 146120 }, { "epoch": 0.37, "learning_rate": 6.349338164980624e-05, "loss": 1.02, "step": 146125 }, { "epoch": 0.37, "learning_rate": 6.349212340832453e-05, "loss": 1.021, "step": 146130 }, { "epoch": 0.37, "learning_rate": 6.349086516684283e-05, "loss": 1.0204, "step": 146135 }, { "epoch": 0.37, "learning_rate": 6.348960692536113e-05, "loss": 1.0219, "step": 146140 }, { "epoch": 0.37, "learning_rate": 6.348834868387942e-05, "loss": 1.0191, "step": 146145 }, { "epoch": 0.37, "learning_rate": 6.348709044239771e-05, "loss": 1.0212, "step": 146150 }, { "epoch": 0.37, "learning_rate": 6.3485832200916e-05, "loss": 1.0207, "step": 146155 }, { "epoch": 0.37, "learning_rate": 6.34845739594343e-05, "loss": 1.0205, "step": 146160 }, { "epoch": 0.37, "learning_rate": 6.34833157179526e-05, "loss": 1.0213, "step": 146165 }, { "epoch": 0.37, "learning_rate": 6.348205747647089e-05, "loss": 1.0209, "step": 146170 }, { "epoch": 0.37, "learning_rate": 6.348079923498918e-05, "loss": 1.0173, "step": 146175 }, { "epoch": 0.37, "learning_rate": 6.347954099350749e-05, "loss": 1.0213, "step": 146180 }, { "epoch": 0.37, "learning_rate": 6.347828275202578e-05, "loss": 1.0193, "step": 146185 }, { "epoch": 0.37, "learning_rate": 6.347702451054407e-05, "loss": 1.0201, "step": 146190 }, { "epoch": 0.37, "learning_rate": 6.347576626906236e-05, "loss": 1.0208, "step": 146195 }, { "epoch": 0.37, "learning_rate": 6.347450802758067e-05, "loss": 1.0187, "step": 146200 }, { "epoch": 0.37, "learning_rate": 6.347324978609896e-05, "loss": 1.0184, "step": 146205 }, { "epoch": 0.37, "learning_rate": 6.347199154461725e-05, "loss": 1.0208, "step": 146210 }, { "epoch": 0.37, "learning_rate": 6.347073330313554e-05, "loss": 1.0229, "step": 146215 }, { "epoch": 0.37, "learning_rate": 6.346947506165383e-05, "loss": 1.0196, "step": 146220 }, { "epoch": 0.37, "learning_rate": 6.346821682017214e-05, "loss": 1.0201, "step": 146225 }, { "epoch": 0.37, "learning_rate": 6.346695857869043e-05, "loss": 1.0212, "step": 146230 }, { "epoch": 0.37, "learning_rate": 6.346570033720872e-05, "loss": 1.0211, "step": 146235 }, { "epoch": 0.37, "learning_rate": 6.346444209572701e-05, "loss": 1.0207, "step": 146240 }, { "epoch": 0.37, "learning_rate": 6.346318385424532e-05, "loss": 1.0224, "step": 146245 }, { "epoch": 0.37, "learning_rate": 6.346192561276361e-05, "loss": 1.0199, "step": 146250 }, { "epoch": 0.37, "learning_rate": 6.34606673712819e-05, "loss": 1.022, "step": 146255 }, { "epoch": 0.37, "learning_rate": 6.345940912980019e-05, "loss": 1.0195, "step": 146260 }, { "epoch": 0.37, "learning_rate": 6.34581508883185e-05, "loss": 1.0207, "step": 146265 }, { "epoch": 0.37, "learning_rate": 6.345689264683679e-05, "loss": 1.0215, "step": 146270 }, { "epoch": 0.37, "learning_rate": 6.345563440535508e-05, "loss": 1.0397, "step": 146275 }, { "epoch": 0.37, "learning_rate": 6.345437616387337e-05, "loss": 1.019, "step": 146280 }, { "epoch": 0.37, "learning_rate": 6.345311792239166e-05, "loss": 1.0209, "step": 146285 }, { "epoch": 0.37, "learning_rate": 6.345185968090997e-05, "loss": 1.0225, "step": 146290 }, { "epoch": 0.37, "learning_rate": 6.345060143942826e-05, "loss": 1.0222, "step": 146295 }, { "epoch": 0.37, "learning_rate": 6.344934319794655e-05, "loss": 1.0193, "step": 146300 }, { "epoch": 0.37, "learning_rate": 6.344808495646484e-05, "loss": 1.0186, "step": 146305 }, { "epoch": 0.37, "learning_rate": 6.344682671498315e-05, "loss": 1.0363, "step": 146310 }, { "epoch": 0.37, "learning_rate": 6.344556847350144e-05, "loss": 1.0444, "step": 146315 }, { "epoch": 0.37, "learning_rate": 6.344431023201973e-05, "loss": 1.0217, "step": 146320 }, { "epoch": 0.37, "learning_rate": 6.344305199053802e-05, "loss": 1.0225, "step": 146325 }, { "epoch": 0.37, "learning_rate": 6.344179374905633e-05, "loss": 1.0211, "step": 146330 }, { "epoch": 0.37, "learning_rate": 6.344053550757462e-05, "loss": 1.0178, "step": 146335 }, { "epoch": 0.37, "learning_rate": 6.343927726609291e-05, "loss": 1.0182, "step": 146340 }, { "epoch": 0.37, "learning_rate": 6.34380190246112e-05, "loss": 1.02, "step": 146345 }, { "epoch": 0.37, "learning_rate": 6.34367607831295e-05, "loss": 1.0203, "step": 146350 }, { "epoch": 0.37, "learning_rate": 6.34355025416478e-05, "loss": 1.021, "step": 146355 }, { "epoch": 0.37, "learning_rate": 6.343424430016609e-05, "loss": 1.0188, "step": 146360 }, { "epoch": 0.37, "learning_rate": 6.343298605868438e-05, "loss": 1.0192, "step": 146365 }, { "epoch": 0.37, "learning_rate": 6.343172781720267e-05, "loss": 1.022, "step": 146370 }, { "epoch": 0.37, "learning_rate": 6.343046957572098e-05, "loss": 1.0382, "step": 146375 }, { "epoch": 0.37, "learning_rate": 6.342921133423927e-05, "loss": 1.0202, "step": 146380 }, { "epoch": 0.37, "learning_rate": 6.342795309275756e-05, "loss": 1.0203, "step": 146385 }, { "epoch": 0.37, "learning_rate": 6.342669485127585e-05, "loss": 1.0204, "step": 146390 }, { "epoch": 0.37, "learning_rate": 6.342543660979416e-05, "loss": 1.0199, "step": 146395 }, { "epoch": 0.37, "learning_rate": 6.342417836831245e-05, "loss": 1.02, "step": 146400 }, { "epoch": 0.37, "learning_rate": 6.342292012683074e-05, "loss": 1.0207, "step": 146405 }, { "epoch": 0.37, "learning_rate": 6.342166188534903e-05, "loss": 1.0218, "step": 146410 }, { "epoch": 0.37, "learning_rate": 6.342040364386733e-05, "loss": 1.0208, "step": 146415 }, { "epoch": 0.37, "learning_rate": 6.341914540238563e-05, "loss": 1.0193, "step": 146420 }, { "epoch": 0.37, "learning_rate": 6.341788716090392e-05, "loss": 1.0212, "step": 146425 }, { "epoch": 0.37, "learning_rate": 6.341662891942221e-05, "loss": 1.0225, "step": 146430 }, { "epoch": 0.37, "learning_rate": 6.34153706779405e-05, "loss": 1.0221, "step": 146435 }, { "epoch": 0.37, "learning_rate": 6.341411243645881e-05, "loss": 1.0193, "step": 146440 }, { "epoch": 0.37, "learning_rate": 6.34128541949771e-05, "loss": 1.0189, "step": 146445 }, { "epoch": 0.37, "learning_rate": 6.34115959534954e-05, "loss": 1.0211, "step": 146450 }, { "epoch": 0.37, "learning_rate": 6.341033771201369e-05, "loss": 1.0204, "step": 146455 }, { "epoch": 0.37, "learning_rate": 6.340907947053199e-05, "loss": 1.0205, "step": 146460 }, { "epoch": 0.37, "learning_rate": 6.340782122905028e-05, "loss": 1.0213, "step": 146465 }, { "epoch": 0.37, "learning_rate": 6.340656298756857e-05, "loss": 1.0174, "step": 146470 }, { "epoch": 0.37, "learning_rate": 6.340530474608687e-05, "loss": 1.0204, "step": 146475 }, { "epoch": 0.37, "learning_rate": 6.340404650460516e-05, "loss": 1.0189, "step": 146480 }, { "epoch": 0.37, "learning_rate": 6.340278826312346e-05, "loss": 1.0213, "step": 146485 }, { "epoch": 0.37, "learning_rate": 6.340153002164175e-05, "loss": 1.0193, "step": 146490 }, { "epoch": 0.37, "learning_rate": 6.340027178016005e-05, "loss": 1.0215, "step": 146495 }, { "epoch": 0.37, "learning_rate": 6.339901353867834e-05, "loss": 1.0198, "step": 146500 }, { "epoch": 0.37, "learning_rate": 6.339775529719664e-05, "loss": 1.0215, "step": 146505 }, { "epoch": 0.37, "learning_rate": 6.339649705571493e-05, "loss": 1.0202, "step": 146510 }, { "epoch": 0.37, "learning_rate": 6.339523881423323e-05, "loss": 1.0204, "step": 146515 }, { "epoch": 0.37, "learning_rate": 6.339398057275152e-05, "loss": 1.0191, "step": 146520 }, { "epoch": 0.37, "learning_rate": 6.339272233126982e-05, "loss": 1.0213, "step": 146525 }, { "epoch": 0.37, "learning_rate": 6.339146408978811e-05, "loss": 1.0188, "step": 146530 }, { "epoch": 0.37, "learning_rate": 6.33902058483064e-05, "loss": 1.0194, "step": 146535 }, { "epoch": 0.37, "learning_rate": 6.33889476068247e-05, "loss": 1.019, "step": 146540 }, { "epoch": 0.37, "learning_rate": 6.338768936534299e-05, "loss": 1.0229, "step": 146545 }, { "epoch": 0.37, "learning_rate": 6.33864311238613e-05, "loss": 1.0197, "step": 146550 }, { "epoch": 0.37, "learning_rate": 6.338517288237959e-05, "loss": 1.0225, "step": 146555 }, { "epoch": 0.37, "learning_rate": 6.338391464089788e-05, "loss": 1.0204, "step": 146560 }, { "epoch": 0.37, "learning_rate": 6.338265639941617e-05, "loss": 1.0231, "step": 146565 }, { "epoch": 0.37, "learning_rate": 6.338139815793447e-05, "loss": 1.0234, "step": 146570 }, { "epoch": 0.37, "learning_rate": 6.338013991645277e-05, "loss": 1.022, "step": 146575 }, { "epoch": 0.37, "learning_rate": 6.337888167497107e-05, "loss": 1.021, "step": 146580 }, { "epoch": 0.37, "learning_rate": 6.337762343348936e-05, "loss": 1.0216, "step": 146585 }, { "epoch": 0.37, "learning_rate": 6.337636519200765e-05, "loss": 1.0217, "step": 146590 }, { "epoch": 0.37, "learning_rate": 6.337510695052596e-05, "loss": 1.0207, "step": 146595 }, { "epoch": 0.37, "learning_rate": 6.337384870904425e-05, "loss": 1.0201, "step": 146600 }, { "epoch": 0.37, "learning_rate": 6.337259046756254e-05, "loss": 1.0196, "step": 146605 }, { "epoch": 0.37, "learning_rate": 6.337133222608083e-05, "loss": 1.0191, "step": 146610 }, { "epoch": 0.37, "learning_rate": 6.337007398459913e-05, "loss": 1.0202, "step": 146615 }, { "epoch": 0.37, "learning_rate": 6.336881574311743e-05, "loss": 1.0194, "step": 146620 }, { "epoch": 0.37, "learning_rate": 6.336755750163572e-05, "loss": 1.0185, "step": 146625 }, { "epoch": 0.37, "learning_rate": 6.336629926015401e-05, "loss": 1.0197, "step": 146630 }, { "epoch": 0.37, "learning_rate": 6.33650410186723e-05, "loss": 1.0203, "step": 146635 }, { "epoch": 0.37, "learning_rate": 6.336378277719061e-05, "loss": 1.0207, "step": 146640 }, { "epoch": 0.37, "learning_rate": 6.33625245357089e-05, "loss": 1.0194, "step": 146645 }, { "epoch": 0.37, "learning_rate": 6.33612662942272e-05, "loss": 1.0204, "step": 146650 }, { "epoch": 0.37, "learning_rate": 6.336000805274549e-05, "loss": 1.0211, "step": 146655 }, { "epoch": 0.37, "learning_rate": 6.335874981126379e-05, "loss": 1.0203, "step": 146660 }, { "epoch": 0.37, "learning_rate": 6.335749156978208e-05, "loss": 1.0221, "step": 146665 }, { "epoch": 0.37, "learning_rate": 6.335623332830037e-05, "loss": 1.0231, "step": 146670 }, { "epoch": 0.37, "learning_rate": 6.335497508681867e-05, "loss": 1.0213, "step": 146675 }, { "epoch": 0.37, "learning_rate": 6.335371684533696e-05, "loss": 1.0191, "step": 146680 }, { "epoch": 0.37, "learning_rate": 6.335245860385526e-05, "loss": 1.0225, "step": 146685 }, { "epoch": 0.37, "learning_rate": 6.335120036237355e-05, "loss": 1.0204, "step": 146690 }, { "epoch": 0.37, "learning_rate": 6.334994212089185e-05, "loss": 1.0213, "step": 146695 }, { "epoch": 0.37, "learning_rate": 6.334868387941014e-05, "loss": 1.0168, "step": 146700 }, { "epoch": 0.37, "learning_rate": 6.334742563792844e-05, "loss": 1.0216, "step": 146705 }, { "epoch": 0.37, "learning_rate": 6.334616739644673e-05, "loss": 1.0205, "step": 146710 }, { "epoch": 0.37, "learning_rate": 6.334490915496503e-05, "loss": 1.0192, "step": 146715 }, { "epoch": 0.37, "learning_rate": 6.334365091348332e-05, "loss": 1.0192, "step": 146720 }, { "epoch": 0.37, "learning_rate": 6.334239267200162e-05, "loss": 1.0202, "step": 146725 }, { "epoch": 0.37, "learning_rate": 6.334113443051991e-05, "loss": 1.0193, "step": 146730 }, { "epoch": 0.37, "learning_rate": 6.33398761890382e-05, "loss": 1.0233, "step": 146735 }, { "epoch": 0.37, "learning_rate": 6.33386179475565e-05, "loss": 1.0212, "step": 146740 }, { "epoch": 0.37, "learning_rate": 6.333735970607479e-05, "loss": 1.0221, "step": 146745 }, { "epoch": 0.37, "learning_rate": 6.33361014645931e-05, "loss": 1.0201, "step": 146750 }, { "epoch": 0.37, "learning_rate": 6.333484322311139e-05, "loss": 1.023, "step": 146755 }, { "epoch": 0.37, "learning_rate": 6.333358498162968e-05, "loss": 1.0187, "step": 146760 }, { "epoch": 0.37, "learning_rate": 6.333232674014797e-05, "loss": 1.0179, "step": 146765 }, { "epoch": 0.37, "learning_rate": 6.333106849866627e-05, "loss": 1.0214, "step": 146770 }, { "epoch": 0.37, "learning_rate": 6.332981025718457e-05, "loss": 1.0206, "step": 146775 }, { "epoch": 0.37, "learning_rate": 6.332855201570286e-05, "loss": 1.0217, "step": 146780 }, { "epoch": 0.37, "learning_rate": 6.332729377422115e-05, "loss": 1.0196, "step": 146785 }, { "epoch": 0.37, "learning_rate": 6.332603553273945e-05, "loss": 1.0211, "step": 146790 }, { "epoch": 0.37, "learning_rate": 6.332477729125774e-05, "loss": 1.0199, "step": 146795 }, { "epoch": 0.37, "learning_rate": 6.332351904977604e-05, "loss": 1.0213, "step": 146800 }, { "epoch": 0.37, "learning_rate": 6.332226080829433e-05, "loss": 1.0183, "step": 146805 }, { "epoch": 0.37, "learning_rate": 6.332100256681262e-05, "loss": 1.022, "step": 146810 }, { "epoch": 0.37, "learning_rate": 6.331974432533092e-05, "loss": 1.0228, "step": 146815 }, { "epoch": 0.37, "learning_rate": 6.331848608384922e-05, "loss": 1.0204, "step": 146820 }, { "epoch": 0.37, "learning_rate": 6.331722784236751e-05, "loss": 1.0238, "step": 146825 }, { "epoch": 0.37, "learning_rate": 6.33159696008858e-05, "loss": 1.0199, "step": 146830 }, { "epoch": 0.37, "learning_rate": 6.33147113594041e-05, "loss": 1.0231, "step": 146835 }, { "epoch": 0.37, "learning_rate": 6.33134531179224e-05, "loss": 1.0196, "step": 146840 }, { "epoch": 0.37, "learning_rate": 6.331219487644069e-05, "loss": 1.0217, "step": 146845 }, { "epoch": 0.37, "learning_rate": 6.331093663495898e-05, "loss": 1.0217, "step": 146850 }, { "epoch": 0.37, "learning_rate": 6.330967839347728e-05, "loss": 1.0208, "step": 146855 }, { "epoch": 0.37, "learning_rate": 6.330842015199558e-05, "loss": 1.0203, "step": 146860 }, { "epoch": 0.37, "learning_rate": 6.330716191051387e-05, "loss": 1.0202, "step": 146865 }, { "epoch": 0.37, "learning_rate": 6.330590366903216e-05, "loss": 1.0215, "step": 146870 }, { "epoch": 0.37, "learning_rate": 6.330464542755045e-05, "loss": 1.0184, "step": 146875 }, { "epoch": 0.37, "learning_rate": 6.330338718606876e-05, "loss": 1.0239, "step": 146880 }, { "epoch": 0.37, "learning_rate": 6.330212894458705e-05, "loss": 1.0194, "step": 146885 }, { "epoch": 0.37, "learning_rate": 6.330087070310534e-05, "loss": 1.0211, "step": 146890 }, { "epoch": 0.37, "learning_rate": 6.329961246162363e-05, "loss": 1.0231, "step": 146895 }, { "epoch": 0.37, "learning_rate": 6.329835422014194e-05, "loss": 1.0223, "step": 146900 }, { "epoch": 0.37, "learning_rate": 6.329709597866023e-05, "loss": 1.0242, "step": 146905 }, { "epoch": 0.37, "learning_rate": 6.329583773717852e-05, "loss": 1.021, "step": 146910 }, { "epoch": 0.37, "learning_rate": 6.329457949569681e-05, "loss": 1.0202, "step": 146915 }, { "epoch": 0.37, "learning_rate": 6.329332125421512e-05, "loss": 1.022, "step": 146920 }, { "epoch": 0.37, "learning_rate": 6.329206301273341e-05, "loss": 1.02, "step": 146925 }, { "epoch": 0.37, "learning_rate": 6.32908047712517e-05, "loss": 1.0199, "step": 146930 }, { "epoch": 0.37, "learning_rate": 6.328954652976999e-05, "loss": 1.0225, "step": 146935 }, { "epoch": 0.37, "learning_rate": 6.328828828828828e-05, "loss": 1.0196, "step": 146940 }, { "epoch": 0.37, "learning_rate": 6.328703004680659e-05, "loss": 1.0204, "step": 146945 }, { "epoch": 0.37, "learning_rate": 6.328577180532488e-05, "loss": 1.0218, "step": 146950 }, { "epoch": 0.37, "learning_rate": 6.328451356384317e-05, "loss": 1.0224, "step": 146955 }, { "epoch": 0.37, "learning_rate": 6.328325532236146e-05, "loss": 1.0198, "step": 146960 }, { "epoch": 0.37, "learning_rate": 6.328199708087977e-05, "loss": 1.0202, "step": 146965 }, { "epoch": 0.37, "learning_rate": 6.328073883939806e-05, "loss": 1.0212, "step": 146970 }, { "epoch": 0.37, "learning_rate": 6.327948059791635e-05, "loss": 1.0198, "step": 146975 }, { "epoch": 0.37, "learning_rate": 6.327822235643464e-05, "loss": 1.0406, "step": 146980 }, { "epoch": 0.37, "learning_rate": 6.327696411495295e-05, "loss": 1.0188, "step": 146985 }, { "epoch": 0.37, "learning_rate": 6.327570587347124e-05, "loss": 1.0177, "step": 146990 }, { "epoch": 0.37, "learning_rate": 6.327444763198953e-05, "loss": 1.0192, "step": 146995 }, { "epoch": 0.37, "learning_rate": 6.327318939050782e-05, "loss": 1.0222, "step": 147000 }, { "epoch": 0.37, "learning_rate": 6.327193114902611e-05, "loss": 1.02, "step": 147005 }, { "epoch": 0.37, "learning_rate": 6.327067290754442e-05, "loss": 1.0198, "step": 147010 }, { "epoch": 0.37, "learning_rate": 6.326941466606271e-05, "loss": 1.0213, "step": 147015 }, { "epoch": 0.37, "learning_rate": 6.3268156424581e-05, "loss": 1.0465, "step": 147020 }, { "epoch": 0.37, "learning_rate": 6.32668981830993e-05, "loss": 1.0195, "step": 147025 }, { "epoch": 0.37, "learning_rate": 6.32656399416176e-05, "loss": 1.0189, "step": 147030 }, { "epoch": 0.37, "learning_rate": 6.326438170013589e-05, "loss": 1.0222, "step": 147035 }, { "epoch": 0.37, "learning_rate": 6.326312345865418e-05, "loss": 1.0185, "step": 147040 }, { "epoch": 0.37, "learning_rate": 6.326186521717247e-05, "loss": 1.0211, "step": 147045 }, { "epoch": 0.37, "learning_rate": 6.326060697569078e-05, "loss": 1.0182, "step": 147050 }, { "epoch": 0.37, "learning_rate": 6.325934873420907e-05, "loss": 1.0202, "step": 147055 }, { "epoch": 0.37, "learning_rate": 6.325809049272736e-05, "loss": 1.0208, "step": 147060 }, { "epoch": 0.37, "learning_rate": 6.325683225124565e-05, "loss": 1.0212, "step": 147065 }, { "epoch": 0.37, "learning_rate": 6.325557400976395e-05, "loss": 1.0246, "step": 147070 }, { "epoch": 0.37, "learning_rate": 6.325431576828225e-05, "loss": 1.0381, "step": 147075 }, { "epoch": 0.37, "learning_rate": 6.325305752680056e-05, "loss": 1.0193, "step": 147080 }, { "epoch": 0.37, "learning_rate": 6.325179928531885e-05, "loss": 1.0202, "step": 147085 }, { "epoch": 0.37, "learning_rate": 6.325054104383714e-05, "loss": 1.0212, "step": 147090 }, { "epoch": 0.37, "learning_rate": 6.324928280235543e-05, "loss": 1.0423, "step": 147095 }, { "epoch": 0.37, "learning_rate": 6.324802456087374e-05, "loss": 1.0214, "step": 147100 }, { "epoch": 0.37, "learning_rate": 6.324676631939203e-05, "loss": 1.0188, "step": 147105 }, { "epoch": 0.37, "learning_rate": 6.324550807791032e-05, "loss": 1.0203, "step": 147110 }, { "epoch": 0.37, "learning_rate": 6.324424983642861e-05, "loss": 1.0206, "step": 147115 }, { "epoch": 0.37, "learning_rate": 6.324299159494692e-05, "loss": 1.021, "step": 147120 }, { "epoch": 0.37, "learning_rate": 6.324173335346521e-05, "loss": 1.0219, "step": 147125 }, { "epoch": 0.37, "learning_rate": 6.32404751119835e-05, "loss": 1.0218, "step": 147130 }, { "epoch": 0.37, "learning_rate": 6.323921687050179e-05, "loss": 1.0198, "step": 147135 }, { "epoch": 0.37, "learning_rate": 6.323795862902008e-05, "loss": 1.0216, "step": 147140 }, { "epoch": 0.37, "learning_rate": 6.323670038753839e-05, "loss": 1.0221, "step": 147145 }, { "epoch": 0.37, "learning_rate": 6.323544214605668e-05, "loss": 1.0202, "step": 147150 }, { "epoch": 0.37, "learning_rate": 6.323418390457497e-05, "loss": 1.019, "step": 147155 }, { "epoch": 0.37, "learning_rate": 6.323292566309326e-05, "loss": 1.0197, "step": 147160 }, { "epoch": 0.37, "learning_rate": 6.323166742161157e-05, "loss": 1.0181, "step": 147165 }, { "epoch": 0.37, "learning_rate": 6.323040918012986e-05, "loss": 1.0205, "step": 147170 }, { "epoch": 0.37, "learning_rate": 6.322915093864815e-05, "loss": 1.0196, "step": 147175 }, { "epoch": 0.37, "learning_rate": 6.322789269716644e-05, "loss": 1.0193, "step": 147180 }, { "epoch": 0.37, "learning_rate": 6.322663445568475e-05, "loss": 1.0203, "step": 147185 }, { "epoch": 0.37, "learning_rate": 6.322537621420304e-05, "loss": 1.0224, "step": 147190 }, { "epoch": 0.37, "learning_rate": 6.322411797272133e-05, "loss": 1.0219, "step": 147195 }, { "epoch": 0.37, "learning_rate": 6.322285973123962e-05, "loss": 1.0204, "step": 147200 }, { "epoch": 0.37, "learning_rate": 6.322160148975791e-05, "loss": 1.0187, "step": 147205 }, { "epoch": 0.37, "learning_rate": 6.322034324827622e-05, "loss": 1.0209, "step": 147210 }, { "epoch": 0.37, "learning_rate": 6.321908500679451e-05, "loss": 1.0204, "step": 147215 }, { "epoch": 0.37, "learning_rate": 6.32178267653128e-05, "loss": 1.0209, "step": 147220 }, { "epoch": 0.37, "learning_rate": 6.32165685238311e-05, "loss": 1.0204, "step": 147225 }, { "epoch": 0.37, "learning_rate": 6.32153102823494e-05, "loss": 1.0199, "step": 147230 }, { "epoch": 0.37, "learning_rate": 6.321405204086769e-05, "loss": 1.0441, "step": 147235 }, { "epoch": 0.37, "learning_rate": 6.321279379938598e-05, "loss": 1.0207, "step": 147240 }, { "epoch": 0.37, "learning_rate": 6.321153555790427e-05, "loss": 1.0202, "step": 147245 }, { "epoch": 0.37, "learning_rate": 6.321027731642258e-05, "loss": 1.02, "step": 147250 }, { "epoch": 0.37, "learning_rate": 6.320901907494087e-05, "loss": 1.0213, "step": 147255 }, { "epoch": 0.37, "learning_rate": 6.320776083345916e-05, "loss": 1.0195, "step": 147260 }, { "epoch": 0.37, "learning_rate": 6.320650259197745e-05, "loss": 1.0194, "step": 147265 }, { "epoch": 0.37, "learning_rate": 6.320524435049574e-05, "loss": 1.0224, "step": 147270 }, { "epoch": 0.37, "learning_rate": 6.320398610901405e-05, "loss": 1.0166, "step": 147275 }, { "epoch": 0.37, "learning_rate": 6.320272786753234e-05, "loss": 1.0189, "step": 147280 }, { "epoch": 0.37, "learning_rate": 6.320146962605063e-05, "loss": 1.0213, "step": 147285 }, { "epoch": 0.37, "learning_rate": 6.320021138456892e-05, "loss": 1.0223, "step": 147290 }, { "epoch": 0.37, "learning_rate": 6.319895314308723e-05, "loss": 1.0199, "step": 147295 }, { "epoch": 0.37, "learning_rate": 6.319769490160552e-05, "loss": 1.0215, "step": 147300 }, { "epoch": 0.37, "learning_rate": 6.319643666012381e-05, "loss": 1.0415, "step": 147305 }, { "epoch": 0.37, "learning_rate": 6.31951784186421e-05, "loss": 1.0197, "step": 147310 }, { "epoch": 0.37, "learning_rate": 6.319392017716041e-05, "loss": 1.0198, "step": 147315 }, { "epoch": 0.37, "learning_rate": 6.31926619356787e-05, "loss": 1.0226, "step": 147320 }, { "epoch": 0.37, "learning_rate": 6.319140369419699e-05, "loss": 1.019, "step": 147325 }, { "epoch": 0.37, "learning_rate": 6.319014545271528e-05, "loss": 1.0215, "step": 147330 }, { "epoch": 0.37, "learning_rate": 6.318888721123358e-05, "loss": 1.0205, "step": 147335 }, { "epoch": 0.37, "learning_rate": 6.318762896975188e-05, "loss": 1.0187, "step": 147340 }, { "epoch": 0.37, "learning_rate": 6.318637072827017e-05, "loss": 1.0193, "step": 147345 }, { "epoch": 0.37, "learning_rate": 6.318511248678846e-05, "loss": 1.0201, "step": 147350 }, { "epoch": 0.37, "learning_rate": 6.318385424530676e-05, "loss": 1.0201, "step": 147355 }, { "epoch": 0.37, "learning_rate": 6.318259600382506e-05, "loss": 1.0183, "step": 147360 }, { "epoch": 0.37, "learning_rate": 6.318133776234335e-05, "loss": 1.0203, "step": 147365 }, { "epoch": 0.37, "learning_rate": 6.318007952086164e-05, "loss": 1.0195, "step": 147370 }, { "epoch": 0.37, "learning_rate": 6.317882127937994e-05, "loss": 1.021, "step": 147375 }, { "epoch": 0.37, "learning_rate": 6.317756303789824e-05, "loss": 1.0214, "step": 147380 }, { "epoch": 0.37, "learning_rate": 6.317630479641653e-05, "loss": 1.0199, "step": 147385 }, { "epoch": 0.37, "learning_rate": 6.317504655493482e-05, "loss": 1.0213, "step": 147390 }, { "epoch": 0.37, "learning_rate": 6.317378831345312e-05, "loss": 1.0189, "step": 147395 }, { "epoch": 0.37, "learning_rate": 6.317253007197141e-05, "loss": 1.0195, "step": 147400 }, { "epoch": 0.37, "learning_rate": 6.317127183048971e-05, "loss": 1.0284, "step": 147405 }, { "epoch": 0.37, "learning_rate": 6.3170013589008e-05, "loss": 1.022, "step": 147410 }, { "epoch": 0.37, "learning_rate": 6.31687553475263e-05, "loss": 1.0192, "step": 147415 }, { "epoch": 0.37, "learning_rate": 6.316749710604459e-05, "loss": 1.0207, "step": 147420 }, { "epoch": 0.37, "learning_rate": 6.316623886456289e-05, "loss": 1.0232, "step": 147425 }, { "epoch": 0.37, "learning_rate": 6.316498062308118e-05, "loss": 1.0183, "step": 147430 }, { "epoch": 0.37, "learning_rate": 6.316372238159948e-05, "loss": 1.0207, "step": 147435 }, { "epoch": 0.37, "learning_rate": 6.316246414011777e-05, "loss": 1.0179, "step": 147440 }, { "epoch": 0.37, "learning_rate": 6.316120589863607e-05, "loss": 1.0202, "step": 147445 }, { "epoch": 0.37, "learning_rate": 6.315994765715436e-05, "loss": 1.0227, "step": 147450 }, { "epoch": 0.37, "learning_rate": 6.315868941567266e-05, "loss": 1.0177, "step": 147455 }, { "epoch": 0.37, "learning_rate": 6.315743117419095e-05, "loss": 1.0403, "step": 147460 }, { "epoch": 0.37, "learning_rate": 6.315617293270924e-05, "loss": 1.0202, "step": 147465 }, { "epoch": 0.37, "learning_rate": 6.315491469122754e-05, "loss": 1.0183, "step": 147470 }, { "epoch": 0.37, "learning_rate": 6.315365644974584e-05, "loss": 1.0209, "step": 147475 }, { "epoch": 0.37, "learning_rate": 6.315239820826413e-05, "loss": 1.0196, "step": 147480 }, { "epoch": 0.37, "learning_rate": 6.315113996678242e-05, "loss": 1.0192, "step": 147485 }, { "epoch": 0.37, "learning_rate": 6.314988172530072e-05, "loss": 1.0213, "step": 147490 }, { "epoch": 0.37, "learning_rate": 6.314862348381902e-05, "loss": 1.0224, "step": 147495 }, { "epoch": 0.37, "learning_rate": 6.314736524233731e-05, "loss": 1.0213, "step": 147500 }, { "epoch": 0.37, "learning_rate": 6.31461070008556e-05, "loss": 1.0217, "step": 147505 }, { "epoch": 0.37, "learning_rate": 6.31448487593739e-05, "loss": 1.0207, "step": 147510 }, { "epoch": 0.37, "learning_rate": 6.31435905178922e-05, "loss": 1.0224, "step": 147515 }, { "epoch": 0.37, "learning_rate": 6.314233227641049e-05, "loss": 1.018, "step": 147520 }, { "epoch": 0.37, "learning_rate": 6.314107403492878e-05, "loss": 1.0213, "step": 147525 }, { "epoch": 0.37, "learning_rate": 6.313981579344707e-05, "loss": 1.021, "step": 147530 }, { "epoch": 0.37, "learning_rate": 6.313855755196538e-05, "loss": 1.0211, "step": 147535 }, { "epoch": 0.37, "learning_rate": 6.313729931048367e-05, "loss": 1.0217, "step": 147540 }, { "epoch": 0.37, "learning_rate": 6.313604106900196e-05, "loss": 1.0204, "step": 147545 }, { "epoch": 0.37, "learning_rate": 6.313478282752025e-05, "loss": 1.0205, "step": 147550 }, { "epoch": 0.37, "learning_rate": 6.313352458603856e-05, "loss": 1.0191, "step": 147555 }, { "epoch": 0.37, "learning_rate": 6.313226634455685e-05, "loss": 1.0215, "step": 147560 }, { "epoch": 0.37, "learning_rate": 6.313100810307514e-05, "loss": 1.0198, "step": 147565 }, { "epoch": 0.37, "learning_rate": 6.312974986159343e-05, "loss": 1.0212, "step": 147570 }, { "epoch": 0.37, "learning_rate": 6.312849162011174e-05, "loss": 1.022, "step": 147575 }, { "epoch": 0.37, "learning_rate": 6.312723337863004e-05, "loss": 1.019, "step": 147580 }, { "epoch": 0.37, "learning_rate": 6.312597513714833e-05, "loss": 1.0198, "step": 147585 }, { "epoch": 0.37, "learning_rate": 6.312471689566662e-05, "loss": 1.0219, "step": 147590 }, { "epoch": 0.37, "learning_rate": 6.312345865418492e-05, "loss": 1.0214, "step": 147595 }, { "epoch": 0.37, "learning_rate": 6.312220041270322e-05, "loss": 1.02, "step": 147600 }, { "epoch": 0.37, "learning_rate": 6.312094217122151e-05, "loss": 1.0198, "step": 147605 }, { "epoch": 0.37, "learning_rate": 6.31196839297398e-05, "loss": 1.0221, "step": 147610 }, { "epoch": 0.37, "learning_rate": 6.31184256882581e-05, "loss": 1.0203, "step": 147615 }, { "epoch": 0.37, "learning_rate": 6.311716744677639e-05, "loss": 1.0217, "step": 147620 }, { "epoch": 0.37, "learning_rate": 6.311590920529469e-05, "loss": 1.0197, "step": 147625 }, { "epoch": 0.37, "learning_rate": 6.311465096381298e-05, "loss": 1.0212, "step": 147630 }, { "epoch": 0.37, "learning_rate": 6.311339272233128e-05, "loss": 1.0211, "step": 147635 }, { "epoch": 0.37, "learning_rate": 6.311213448084957e-05, "loss": 1.0211, "step": 147640 }, { "epoch": 0.37, "learning_rate": 6.311087623936787e-05, "loss": 1.0204, "step": 147645 }, { "epoch": 0.37, "learning_rate": 6.310961799788616e-05, "loss": 1.0201, "step": 147650 }, { "epoch": 0.37, "learning_rate": 6.310835975640446e-05, "loss": 1.0217, "step": 147655 }, { "epoch": 0.37, "learning_rate": 6.310710151492275e-05, "loss": 1.0189, "step": 147660 }, { "epoch": 0.37, "learning_rate": 6.310584327344105e-05, "loss": 1.0199, "step": 147665 }, { "epoch": 0.37, "learning_rate": 6.310458503195934e-05, "loss": 1.018, "step": 147670 }, { "epoch": 0.37, "learning_rate": 6.310332679047764e-05, "loss": 1.02, "step": 147675 }, { "epoch": 0.37, "learning_rate": 6.310206854899593e-05, "loss": 1.0435, "step": 147680 }, { "epoch": 0.37, "learning_rate": 6.310081030751422e-05, "loss": 1.0196, "step": 147685 }, { "epoch": 0.37, "learning_rate": 6.309955206603252e-05, "loss": 1.0204, "step": 147690 }, { "epoch": 0.37, "learning_rate": 6.309829382455082e-05, "loss": 1.0213, "step": 147695 }, { "epoch": 0.37, "learning_rate": 6.309703558306911e-05, "loss": 1.022, "step": 147700 }, { "epoch": 0.37, "learning_rate": 6.30957773415874e-05, "loss": 1.0182, "step": 147705 }, { "epoch": 0.37, "learning_rate": 6.30945191001057e-05, "loss": 1.0204, "step": 147710 }, { "epoch": 0.37, "learning_rate": 6.3093260858624e-05, "loss": 1.0204, "step": 147715 }, { "epoch": 0.37, "learning_rate": 6.309200261714229e-05, "loss": 1.0201, "step": 147720 }, { "epoch": 0.37, "learning_rate": 6.309074437566058e-05, "loss": 1.0223, "step": 147725 }, { "epoch": 0.37, "learning_rate": 6.308948613417887e-05, "loss": 1.0201, "step": 147730 }, { "epoch": 0.37, "learning_rate": 6.308822789269718e-05, "loss": 1.0194, "step": 147735 }, { "epoch": 0.37, "learning_rate": 6.308696965121547e-05, "loss": 1.0191, "step": 147740 }, { "epoch": 0.37, "learning_rate": 6.308571140973376e-05, "loss": 1.0195, "step": 147745 }, { "epoch": 0.37, "learning_rate": 6.308445316825205e-05, "loss": 1.0208, "step": 147750 }, { "epoch": 0.37, "learning_rate": 6.308319492677036e-05, "loss": 1.0202, "step": 147755 }, { "epoch": 0.37, "learning_rate": 6.308193668528865e-05, "loss": 1.0179, "step": 147760 }, { "epoch": 0.37, "learning_rate": 6.308067844380694e-05, "loss": 1.0203, "step": 147765 }, { "epoch": 0.37, "learning_rate": 6.307942020232523e-05, "loss": 1.0207, "step": 147770 }, { "epoch": 0.37, "learning_rate": 6.307816196084354e-05, "loss": 1.0218, "step": 147775 }, { "epoch": 0.37, "learning_rate": 6.307690371936183e-05, "loss": 1.021, "step": 147780 }, { "epoch": 0.37, "learning_rate": 6.307564547788012e-05, "loss": 1.0206, "step": 147785 }, { "epoch": 0.37, "learning_rate": 6.307438723639841e-05, "loss": 1.0222, "step": 147790 }, { "epoch": 0.37, "learning_rate": 6.30731289949167e-05, "loss": 1.0198, "step": 147795 }, { "epoch": 0.37, "learning_rate": 6.3071870753435e-05, "loss": 1.0217, "step": 147800 }, { "epoch": 0.37, "learning_rate": 6.30706125119533e-05, "loss": 1.0188, "step": 147805 }, { "epoch": 0.37, "learning_rate": 6.306935427047159e-05, "loss": 1.0218, "step": 147810 }, { "epoch": 0.37, "learning_rate": 6.306809602898988e-05, "loss": 1.0203, "step": 147815 }, { "epoch": 0.37, "learning_rate": 6.306683778750819e-05, "loss": 1.0193, "step": 147820 }, { "epoch": 0.37, "learning_rate": 6.306557954602648e-05, "loss": 1.0182, "step": 147825 }, { "epoch": 0.37, "learning_rate": 6.306432130454477e-05, "loss": 1.0185, "step": 147830 }, { "epoch": 0.37, "learning_rate": 6.306306306306306e-05, "loss": 1.0205, "step": 147835 }, { "epoch": 0.37, "learning_rate": 6.306180482158137e-05, "loss": 1.0195, "step": 147840 }, { "epoch": 0.37, "learning_rate": 6.306054658009966e-05, "loss": 1.0203, "step": 147845 }, { "epoch": 0.37, "learning_rate": 6.305928833861795e-05, "loss": 1.0208, "step": 147850 }, { "epoch": 0.37, "learning_rate": 6.305803009713624e-05, "loss": 1.0207, "step": 147855 }, { "epoch": 0.37, "learning_rate": 6.305677185565453e-05, "loss": 1.0199, "step": 147860 }, { "epoch": 0.37, "learning_rate": 6.305551361417284e-05, "loss": 1.0219, "step": 147865 }, { "epoch": 0.37, "learning_rate": 6.305425537269113e-05, "loss": 1.0196, "step": 147870 }, { "epoch": 0.37, "learning_rate": 6.305299713120942e-05, "loss": 1.0219, "step": 147875 }, { "epoch": 0.37, "learning_rate": 6.305173888972771e-05, "loss": 1.0232, "step": 147880 }, { "epoch": 0.37, "learning_rate": 6.305048064824602e-05, "loss": 1.0185, "step": 147885 }, { "epoch": 0.37, "learning_rate": 6.304922240676431e-05, "loss": 1.0436, "step": 147890 }, { "epoch": 0.37, "learning_rate": 6.30479641652826e-05, "loss": 1.0183, "step": 147895 }, { "epoch": 0.37, "learning_rate": 6.304670592380089e-05, "loss": 1.024, "step": 147900 }, { "epoch": 0.37, "learning_rate": 6.30454476823192e-05, "loss": 1.0202, "step": 147905 }, { "epoch": 0.37, "learning_rate": 6.304418944083749e-05, "loss": 1.0407, "step": 147910 }, { "epoch": 0.37, "learning_rate": 6.304293119935578e-05, "loss": 1.0217, "step": 147915 }, { "epoch": 0.37, "learning_rate": 6.304167295787407e-05, "loss": 1.0206, "step": 147920 }, { "epoch": 0.37, "learning_rate": 6.304041471639236e-05, "loss": 1.0225, "step": 147925 }, { "epoch": 0.37, "learning_rate": 6.303915647491067e-05, "loss": 1.0405, "step": 147930 }, { "epoch": 0.37, "learning_rate": 6.303789823342896e-05, "loss": 1.0206, "step": 147935 }, { "epoch": 0.37, "learning_rate": 6.303663999194725e-05, "loss": 1.019, "step": 147940 }, { "epoch": 0.37, "learning_rate": 6.303538175046554e-05, "loss": 1.0213, "step": 147945 }, { "epoch": 0.37, "learning_rate": 6.303412350898385e-05, "loss": 1.0213, "step": 147950 }, { "epoch": 0.37, "learning_rate": 6.303286526750214e-05, "loss": 1.0185, "step": 147955 }, { "epoch": 0.37, "learning_rate": 6.303160702602043e-05, "loss": 1.0432, "step": 147960 }, { "epoch": 0.37, "learning_rate": 6.303034878453872e-05, "loss": 1.0173, "step": 147965 }, { "epoch": 0.37, "learning_rate": 6.302909054305703e-05, "loss": 1.0233, "step": 147970 }, { "epoch": 0.37, "learning_rate": 6.302783230157532e-05, "loss": 1.0207, "step": 147975 }, { "epoch": 0.37, "learning_rate": 6.302657406009361e-05, "loss": 1.0201, "step": 147980 }, { "epoch": 0.37, "learning_rate": 6.30253158186119e-05, "loss": 1.0187, "step": 147985 }, { "epoch": 0.37, "learning_rate": 6.30240575771302e-05, "loss": 1.019, "step": 147990 }, { "epoch": 0.37, "learning_rate": 6.30227993356485e-05, "loss": 1.0195, "step": 147995 }, { "epoch": 0.37, "learning_rate": 6.302154109416679e-05, "loss": 1.0198, "step": 148000 }, { "epoch": 0.37, "learning_rate": 6.302028285268508e-05, "loss": 1.0214, "step": 148005 }, { "epoch": 0.37, "learning_rate": 6.301902461120338e-05, "loss": 1.0188, "step": 148010 }, { "epoch": 0.37, "learning_rate": 6.301776636972168e-05, "loss": 1.021, "step": 148015 }, { "epoch": 0.37, "learning_rate": 6.301650812823997e-05, "loss": 1.0202, "step": 148020 }, { "epoch": 0.37, "learning_rate": 6.301524988675826e-05, "loss": 1.0192, "step": 148025 }, { "epoch": 0.37, "learning_rate": 6.301399164527656e-05, "loss": 1.0211, "step": 148030 }, { "epoch": 0.37, "learning_rate": 6.301273340379486e-05, "loss": 1.0209, "step": 148035 }, { "epoch": 0.37, "learning_rate": 6.301147516231315e-05, "loss": 1.0216, "step": 148040 }, { "epoch": 0.37, "learning_rate": 6.301021692083144e-05, "loss": 1.0232, "step": 148045 }, { "epoch": 0.37, "learning_rate": 6.300895867934974e-05, "loss": 1.021, "step": 148050 }, { "epoch": 0.37, "learning_rate": 6.300770043786803e-05, "loss": 1.022, "step": 148055 }, { "epoch": 0.37, "learning_rate": 6.300644219638633e-05, "loss": 1.0213, "step": 148060 }, { "epoch": 0.37, "learning_rate": 6.300518395490462e-05, "loss": 1.0219, "step": 148065 }, { "epoch": 0.37, "learning_rate": 6.300392571342292e-05, "loss": 1.0203, "step": 148070 }, { "epoch": 0.37, "learning_rate": 6.300266747194121e-05, "loss": 1.021, "step": 148075 }, { "epoch": 0.37, "learning_rate": 6.300140923045951e-05, "loss": 1.0188, "step": 148080 }, { "epoch": 0.37, "learning_rate": 6.300015098897782e-05, "loss": 1.0213, "step": 148085 }, { "epoch": 0.37, "learning_rate": 6.299889274749611e-05, "loss": 1.0216, "step": 148090 }, { "epoch": 0.37, "learning_rate": 6.29976345060144e-05, "loss": 1.0388, "step": 148095 }, { "epoch": 0.37, "learning_rate": 6.299637626453269e-05, "loss": 1.0393, "step": 148100 }, { "epoch": 0.37, "learning_rate": 6.2995118023051e-05, "loss": 1.0203, "step": 148105 }, { "epoch": 0.37, "learning_rate": 6.299385978156929e-05, "loss": 1.0207, "step": 148110 }, { "epoch": 0.37, "learning_rate": 6.299260154008758e-05, "loss": 1.0195, "step": 148115 }, { "epoch": 0.37, "learning_rate": 6.299134329860587e-05, "loss": 1.0217, "step": 148120 }, { "epoch": 0.37, "learning_rate": 6.299008505712418e-05, "loss": 1.0192, "step": 148125 }, { "epoch": 0.37, "learning_rate": 6.298882681564247e-05, "loss": 1.0202, "step": 148130 }, { "epoch": 0.37, "learning_rate": 6.298756857416076e-05, "loss": 1.0207, "step": 148135 }, { "epoch": 0.37, "learning_rate": 6.298631033267905e-05, "loss": 1.0195, "step": 148140 }, { "epoch": 0.37, "learning_rate": 6.298505209119734e-05, "loss": 1.0174, "step": 148145 }, { "epoch": 0.37, "learning_rate": 6.298379384971565e-05, "loss": 1.0198, "step": 148150 }, { "epoch": 0.37, "learning_rate": 6.298253560823394e-05, "loss": 1.0212, "step": 148155 }, { "epoch": 0.37, "learning_rate": 6.298127736675223e-05, "loss": 1.0213, "step": 148160 }, { "epoch": 0.37, "learning_rate": 6.298001912527052e-05, "loss": 1.022, "step": 148165 }, { "epoch": 0.37, "learning_rate": 6.297876088378883e-05, "loss": 1.0191, "step": 148170 }, { "epoch": 0.37, "learning_rate": 6.297750264230712e-05, "loss": 1.0224, "step": 148175 }, { "epoch": 0.37, "learning_rate": 6.297624440082541e-05, "loss": 1.0185, "step": 148180 }, { "epoch": 0.37, "learning_rate": 6.29749861593437e-05, "loss": 1.0212, "step": 148185 }, { "epoch": 0.37, "learning_rate": 6.297372791786201e-05, "loss": 1.0211, "step": 148190 }, { "epoch": 0.37, "learning_rate": 6.29724696763803e-05, "loss": 1.0403, "step": 148195 }, { "epoch": 0.37, "learning_rate": 6.297121143489859e-05, "loss": 1.0215, "step": 148200 }, { "epoch": 0.37, "learning_rate": 6.296995319341688e-05, "loss": 1.0198, "step": 148205 }, { "epoch": 0.37, "learning_rate": 6.296869495193518e-05, "loss": 1.0217, "step": 148210 }, { "epoch": 0.37, "learning_rate": 6.296743671045348e-05, "loss": 1.021, "step": 148215 }, { "epoch": 0.37, "learning_rate": 6.296617846897177e-05, "loss": 1.0219, "step": 148220 }, { "epoch": 0.37, "learning_rate": 6.296492022749006e-05, "loss": 1.0192, "step": 148225 }, { "epoch": 0.37, "learning_rate": 6.296366198600836e-05, "loss": 1.0221, "step": 148230 }, { "epoch": 0.37, "learning_rate": 6.296240374452666e-05, "loss": 1.0207, "step": 148235 }, { "epoch": 0.37, "learning_rate": 6.296114550304495e-05, "loss": 1.0201, "step": 148240 }, { "epoch": 0.37, "learning_rate": 6.295988726156324e-05, "loss": 1.0181, "step": 148245 }, { "epoch": 0.37, "learning_rate": 6.295862902008153e-05, "loss": 1.0387, "step": 148250 }, { "epoch": 0.37, "learning_rate": 6.295737077859984e-05, "loss": 1.0223, "step": 148255 }, { "epoch": 0.37, "learning_rate": 6.295611253711813e-05, "loss": 1.0217, "step": 148260 }, { "epoch": 0.37, "learning_rate": 6.295485429563642e-05, "loss": 1.0224, "step": 148265 }, { "epoch": 0.37, "learning_rate": 6.295359605415471e-05, "loss": 1.0202, "step": 148270 }, { "epoch": 0.37, "learning_rate": 6.2952337812673e-05, "loss": 1.0227, "step": 148275 }, { "epoch": 0.37, "learning_rate": 6.295107957119131e-05, "loss": 1.0205, "step": 148280 }, { "epoch": 0.37, "learning_rate": 6.29498213297096e-05, "loss": 1.0179, "step": 148285 }, { "epoch": 0.37, "learning_rate": 6.29485630882279e-05, "loss": 1.0205, "step": 148290 }, { "epoch": 0.37, "learning_rate": 6.294730484674619e-05, "loss": 1.0208, "step": 148295 }, { "epoch": 0.37, "learning_rate": 6.294604660526449e-05, "loss": 1.0212, "step": 148300 }, { "epoch": 0.37, "learning_rate": 6.294478836378278e-05, "loss": 1.0194, "step": 148305 }, { "epoch": 0.37, "learning_rate": 6.294353012230107e-05, "loss": 1.019, "step": 148310 }, { "epoch": 0.37, "learning_rate": 6.294227188081937e-05, "loss": 1.0194, "step": 148315 }, { "epoch": 0.37, "learning_rate": 6.294101363933767e-05, "loss": 1.0192, "step": 148320 }, { "epoch": 0.37, "learning_rate": 6.293975539785596e-05, "loss": 1.0175, "step": 148325 }, { "epoch": 0.37, "learning_rate": 6.293849715637425e-05, "loss": 1.0192, "step": 148330 }, { "epoch": 0.37, "learning_rate": 6.293723891489255e-05, "loss": 1.0198, "step": 148335 }, { "epoch": 0.37, "learning_rate": 6.293598067341084e-05, "loss": 1.0188, "step": 148340 }, { "epoch": 0.37, "learning_rate": 6.293472243192914e-05, "loss": 1.0196, "step": 148345 }, { "epoch": 0.37, "learning_rate": 6.293346419044743e-05, "loss": 1.0199, "step": 148350 }, { "epoch": 0.37, "learning_rate": 6.293220594896573e-05, "loss": 1.0204, "step": 148355 }, { "epoch": 0.37, "learning_rate": 6.293094770748402e-05, "loss": 1.0232, "step": 148360 }, { "epoch": 0.37, "learning_rate": 6.292968946600232e-05, "loss": 1.0431, "step": 148365 }, { "epoch": 0.37, "learning_rate": 6.292843122452061e-05, "loss": 1.0224, "step": 148370 }, { "epoch": 0.37, "learning_rate": 6.29271729830389e-05, "loss": 1.0224, "step": 148375 }, { "epoch": 0.37, "learning_rate": 6.292616638985354e-05, "loss": 1.0203, "step": 148380 }, { "epoch": 0.37, "learning_rate": 6.292490814837185e-05, "loss": 1.0216, "step": 148385 }, { "epoch": 0.37, "learning_rate": 6.292364990689014e-05, "loss": 1.0199, "step": 148390 }, { "epoch": 0.37, "learning_rate": 6.292239166540843e-05, "loss": 1.0215, "step": 148395 }, { "epoch": 0.37, "learning_rate": 6.292113342392672e-05, "loss": 1.0231, "step": 148400 }, { "epoch": 0.37, "learning_rate": 6.291987518244503e-05, "loss": 1.0206, "step": 148405 }, { "epoch": 0.37, "learning_rate": 6.291861694096332e-05, "loss": 1.0207, "step": 148410 }, { "epoch": 0.37, "learning_rate": 6.291735869948161e-05, "loss": 1.0215, "step": 148415 }, { "epoch": 0.37, "learning_rate": 6.29161004579999e-05, "loss": 1.0215, "step": 148420 }, { "epoch": 0.37, "learning_rate": 6.29148422165182e-05, "loss": 1.0457, "step": 148425 }, { "epoch": 0.37, "learning_rate": 6.29135839750365e-05, "loss": 1.0213, "step": 148430 }, { "epoch": 0.37, "learning_rate": 6.291232573355479e-05, "loss": 1.0208, "step": 148435 }, { "epoch": 0.37, "learning_rate": 6.291106749207308e-05, "loss": 1.0205, "step": 148440 }, { "epoch": 0.37, "learning_rate": 6.290980925059137e-05, "loss": 1.0188, "step": 148445 }, { "epoch": 0.37, "learning_rate": 6.290855100910968e-05, "loss": 1.0222, "step": 148450 }, { "epoch": 0.37, "learning_rate": 6.290729276762797e-05, "loss": 1.021, "step": 148455 }, { "epoch": 0.37, "learning_rate": 6.290603452614626e-05, "loss": 1.0213, "step": 148460 }, { "epoch": 0.37, "learning_rate": 6.290477628466455e-05, "loss": 1.0213, "step": 148465 }, { "epoch": 0.37, "learning_rate": 6.290351804318285e-05, "loss": 1.0208, "step": 148470 }, { "epoch": 0.37, "learning_rate": 6.290225980170115e-05, "loss": 1.0211, "step": 148475 }, { "epoch": 0.37, "learning_rate": 6.290100156021944e-05, "loss": 1.0175, "step": 148480 }, { "epoch": 0.37, "learning_rate": 6.289974331873773e-05, "loss": 1.0196, "step": 148485 }, { "epoch": 0.37, "learning_rate": 6.289848507725603e-05, "loss": 1.0409, "step": 148490 }, { "epoch": 0.37, "learning_rate": 6.289722683577433e-05, "loss": 1.0202, "step": 148495 }, { "epoch": 0.37, "learning_rate": 6.289596859429262e-05, "loss": 1.0201, "step": 148500 }, { "epoch": 0.37, "learning_rate": 6.289471035281091e-05, "loss": 1.0189, "step": 148505 }, { "epoch": 0.37, "learning_rate": 6.28934521113292e-05, "loss": 1.0212, "step": 148510 }, { "epoch": 0.37, "learning_rate": 6.289219386984751e-05, "loss": 1.0219, "step": 148515 }, { "epoch": 0.37, "learning_rate": 6.28909356283658e-05, "loss": 1.0212, "step": 148520 }, { "epoch": 0.37, "learning_rate": 6.28896773868841e-05, "loss": 1.019, "step": 148525 }, { "epoch": 0.37, "learning_rate": 6.288841914540238e-05, "loss": 1.0205, "step": 148530 }, { "epoch": 0.37, "learning_rate": 6.288716090392068e-05, "loss": 1.0201, "step": 148535 }, { "epoch": 0.37, "learning_rate": 6.288590266243898e-05, "loss": 1.0211, "step": 148540 }, { "epoch": 0.37, "learning_rate": 6.288464442095727e-05, "loss": 1.021, "step": 148545 }, { "epoch": 0.37, "learning_rate": 6.288338617947556e-05, "loss": 1.0198, "step": 148550 }, { "epoch": 0.37, "learning_rate": 6.288212793799386e-05, "loss": 1.0192, "step": 148555 }, { "epoch": 0.37, "learning_rate": 6.288086969651216e-05, "loss": 1.0184, "step": 148560 }, { "epoch": 0.37, "learning_rate": 6.287961145503045e-05, "loss": 1.0191, "step": 148565 }, { "epoch": 0.37, "learning_rate": 6.287835321354874e-05, "loss": 1.0191, "step": 148570 }, { "epoch": 0.37, "learning_rate": 6.287709497206704e-05, "loss": 1.0206, "step": 148575 }, { "epoch": 0.37, "learning_rate": 6.287583673058534e-05, "loss": 1.0375, "step": 148580 }, { "epoch": 0.37, "learning_rate": 6.287457848910363e-05, "loss": 1.0215, "step": 148585 }, { "epoch": 0.37, "learning_rate": 6.287332024762192e-05, "loss": 1.0225, "step": 148590 }, { "epoch": 0.37, "learning_rate": 6.287206200614022e-05, "loss": 1.0201, "step": 148595 }, { "epoch": 0.37, "learning_rate": 6.287080376465851e-05, "loss": 1.021, "step": 148600 }, { "epoch": 0.37, "learning_rate": 6.286954552317681e-05, "loss": 1.0444, "step": 148605 }, { "epoch": 0.37, "learning_rate": 6.28682872816951e-05, "loss": 1.0203, "step": 148610 }, { "epoch": 0.37, "learning_rate": 6.28670290402134e-05, "loss": 1.0214, "step": 148615 }, { "epoch": 0.37, "learning_rate": 6.286577079873169e-05, "loss": 1.0195, "step": 148620 }, { "epoch": 0.37, "learning_rate": 6.286451255724999e-05, "loss": 1.0189, "step": 148625 }, { "epoch": 0.37, "learning_rate": 6.286325431576828e-05, "loss": 1.0211, "step": 148630 }, { "epoch": 0.37, "learning_rate": 6.286199607428658e-05, "loss": 1.0213, "step": 148635 }, { "epoch": 0.37, "learning_rate": 6.286073783280487e-05, "loss": 1.0208, "step": 148640 }, { "epoch": 0.37, "learning_rate": 6.285947959132317e-05, "loss": 1.0212, "step": 148645 }, { "epoch": 0.37, "learning_rate": 6.285822134984146e-05, "loss": 1.0191, "step": 148650 }, { "epoch": 0.37, "learning_rate": 6.285696310835976e-05, "loss": 1.0192, "step": 148655 }, { "epoch": 0.37, "learning_rate": 6.285570486687805e-05, "loss": 1.0189, "step": 148660 }, { "epoch": 0.37, "learning_rate": 6.285444662539634e-05, "loss": 1.0195, "step": 148665 }, { "epoch": 0.37, "learning_rate": 6.285318838391464e-05, "loss": 1.0216, "step": 148670 }, { "epoch": 0.37, "learning_rate": 6.285193014243294e-05, "loss": 1.0194, "step": 148675 }, { "epoch": 0.37, "learning_rate": 6.285067190095123e-05, "loss": 1.0176, "step": 148680 }, { "epoch": 0.37, "learning_rate": 6.284941365946952e-05, "loss": 1.0195, "step": 148685 }, { "epoch": 0.37, "learning_rate": 6.284815541798782e-05, "loss": 1.0197, "step": 148690 }, { "epoch": 0.37, "learning_rate": 6.284689717650612e-05, "loss": 1.0217, "step": 148695 }, { "epoch": 0.37, "learning_rate": 6.284563893502441e-05, "loss": 1.0222, "step": 148700 }, { "epoch": 0.37, "learning_rate": 6.28443806935427e-05, "loss": 1.0183, "step": 148705 }, { "epoch": 0.37, "learning_rate": 6.2843122452061e-05, "loss": 1.019, "step": 148710 }, { "epoch": 0.37, "learning_rate": 6.28418642105793e-05, "loss": 1.0203, "step": 148715 }, { "epoch": 0.37, "learning_rate": 6.284060596909759e-05, "loss": 1.0223, "step": 148720 }, { "epoch": 0.37, "learning_rate": 6.283934772761588e-05, "loss": 1.0216, "step": 148725 }, { "epoch": 0.37, "learning_rate": 6.283808948613417e-05, "loss": 1.0197, "step": 148730 }, { "epoch": 0.37, "learning_rate": 6.283683124465248e-05, "loss": 1.0225, "step": 148735 }, { "epoch": 0.37, "learning_rate": 6.283557300317077e-05, "loss": 1.0191, "step": 148740 }, { "epoch": 0.37, "learning_rate": 6.283431476168906e-05, "loss": 1.0221, "step": 148745 }, { "epoch": 0.37, "learning_rate": 6.283305652020735e-05, "loss": 1.0202, "step": 148750 }, { "epoch": 0.37, "learning_rate": 6.283179827872566e-05, "loss": 1.0199, "step": 148755 }, { "epoch": 0.37, "learning_rate": 6.283054003724395e-05, "loss": 1.0211, "step": 148760 }, { "epoch": 0.37, "learning_rate": 6.282928179576224e-05, "loss": 1.0205, "step": 148765 }, { "epoch": 0.37, "learning_rate": 6.282802355428053e-05, "loss": 1.0223, "step": 148770 }, { "epoch": 0.37, "learning_rate": 6.282676531279884e-05, "loss": 1.0231, "step": 148775 }, { "epoch": 0.37, "learning_rate": 6.282550707131714e-05, "loss": 1.0213, "step": 148780 }, { "epoch": 0.37, "learning_rate": 6.282424882983543e-05, "loss": 1.0206, "step": 148785 }, { "epoch": 0.37, "learning_rate": 6.282299058835372e-05, "loss": 1.0203, "step": 148790 }, { "epoch": 0.37, "learning_rate": 6.282173234687202e-05, "loss": 1.0188, "step": 148795 }, { "epoch": 0.37, "learning_rate": 6.282047410539032e-05, "loss": 1.02, "step": 148800 }, { "epoch": 0.37, "learning_rate": 6.281921586390861e-05, "loss": 1.0188, "step": 148805 }, { "epoch": 0.37, "learning_rate": 6.28179576224269e-05, "loss": 1.0211, "step": 148810 }, { "epoch": 0.37, "learning_rate": 6.28166993809452e-05, "loss": 1.0194, "step": 148815 }, { "epoch": 0.37, "learning_rate": 6.281544113946349e-05, "loss": 1.0192, "step": 148820 }, { "epoch": 0.37, "learning_rate": 6.281418289798179e-05, "loss": 1.0207, "step": 148825 }, { "epoch": 0.37, "learning_rate": 6.281292465650008e-05, "loss": 1.0193, "step": 148830 }, { "epoch": 0.37, "learning_rate": 6.281166641501838e-05, "loss": 1.0197, "step": 148835 }, { "epoch": 0.37, "learning_rate": 6.281040817353667e-05, "loss": 1.0214, "step": 148840 }, { "epoch": 0.37, "learning_rate": 6.280914993205497e-05, "loss": 1.0208, "step": 148845 }, { "epoch": 0.37, "learning_rate": 6.280789169057326e-05, "loss": 1.0199, "step": 148850 }, { "epoch": 0.37, "learning_rate": 6.280663344909156e-05, "loss": 1.0181, "step": 148855 }, { "epoch": 0.37, "learning_rate": 6.280537520760985e-05, "loss": 1.0221, "step": 148860 }, { "epoch": 0.37, "learning_rate": 6.280411696612815e-05, "loss": 1.0204, "step": 148865 }, { "epoch": 0.37, "learning_rate": 6.280285872464644e-05, "loss": 1.0226, "step": 148870 }, { "epoch": 0.37, "learning_rate": 6.280160048316474e-05, "loss": 1.0214, "step": 148875 }, { "epoch": 0.37, "learning_rate": 6.280034224168303e-05, "loss": 1.0201, "step": 148880 }, { "epoch": 0.37, "learning_rate": 6.279908400020132e-05, "loss": 1.0216, "step": 148885 }, { "epoch": 0.37, "learning_rate": 6.279782575871962e-05, "loss": 1.0198, "step": 148890 }, { "epoch": 0.37, "learning_rate": 6.279656751723792e-05, "loss": 1.0217, "step": 148895 }, { "epoch": 0.37, "learning_rate": 6.279530927575621e-05, "loss": 1.0208, "step": 148900 }, { "epoch": 0.37, "learning_rate": 6.27940510342745e-05, "loss": 1.0173, "step": 148905 }, { "epoch": 0.37, "learning_rate": 6.27927927927928e-05, "loss": 1.0463, "step": 148910 }, { "epoch": 0.37, "learning_rate": 6.27915345513111e-05, "loss": 1.0213, "step": 148915 }, { "epoch": 0.37, "learning_rate": 6.279027630982939e-05, "loss": 1.0206, "step": 148920 }, { "epoch": 0.37, "learning_rate": 6.278901806834768e-05, "loss": 1.0186, "step": 148925 }, { "epoch": 0.37, "learning_rate": 6.278775982686598e-05, "loss": 1.0221, "step": 148930 }, { "epoch": 0.37, "learning_rate": 6.278650158538428e-05, "loss": 1.0201, "step": 148935 }, { "epoch": 0.37, "learning_rate": 6.278524334390257e-05, "loss": 1.0221, "step": 148940 }, { "epoch": 0.37, "learning_rate": 6.278398510242086e-05, "loss": 1.0187, "step": 148945 }, { "epoch": 0.37, "learning_rate": 6.278272686093915e-05, "loss": 1.0207, "step": 148950 }, { "epoch": 0.37, "learning_rate": 6.278146861945746e-05, "loss": 1.022, "step": 148955 }, { "epoch": 0.37, "learning_rate": 6.278021037797575e-05, "loss": 1.0203, "step": 148960 }, { "epoch": 0.37, "learning_rate": 6.277895213649404e-05, "loss": 1.022, "step": 148965 }, { "epoch": 0.37, "learning_rate": 6.277769389501233e-05, "loss": 1.0218, "step": 148970 }, { "epoch": 0.37, "learning_rate": 6.277643565353064e-05, "loss": 1.0189, "step": 148975 }, { "epoch": 0.37, "learning_rate": 6.277517741204893e-05, "loss": 1.0218, "step": 148980 }, { "epoch": 0.37, "learning_rate": 6.277391917056722e-05, "loss": 1.0222, "step": 148985 }, { "epoch": 0.37, "learning_rate": 6.277266092908551e-05, "loss": 1.0205, "step": 148990 }, { "epoch": 0.37, "learning_rate": 6.277140268760382e-05, "loss": 1.0214, "step": 148995 }, { "epoch": 0.37, "learning_rate": 6.277014444612211e-05, "loss": 1.0239, "step": 149000 }, { "epoch": 0.37, "learning_rate": 6.27688862046404e-05, "loss": 1.0207, "step": 149005 }, { "epoch": 0.37, "learning_rate": 6.276762796315869e-05, "loss": 1.0219, "step": 149010 }, { "epoch": 0.37, "learning_rate": 6.276636972167698e-05, "loss": 1.0209, "step": 149015 }, { "epoch": 0.37, "learning_rate": 6.276511148019529e-05, "loss": 1.0192, "step": 149020 }, { "epoch": 0.37, "learning_rate": 6.276385323871358e-05, "loss": 1.019, "step": 149025 }, { "epoch": 0.37, "learning_rate": 6.276259499723187e-05, "loss": 1.0219, "step": 149030 }, { "epoch": 0.37, "learning_rate": 6.276133675575016e-05, "loss": 1.02, "step": 149035 }, { "epoch": 0.37, "learning_rate": 6.276007851426847e-05, "loss": 1.0215, "step": 149040 }, { "epoch": 0.37, "learning_rate": 6.275882027278676e-05, "loss": 1.0223, "step": 149045 }, { "epoch": 0.37, "learning_rate": 6.275756203130505e-05, "loss": 1.0198, "step": 149050 }, { "epoch": 0.37, "learning_rate": 6.275630378982334e-05, "loss": 1.0192, "step": 149055 }, { "epoch": 0.37, "learning_rate": 6.275504554834165e-05, "loss": 1.0194, "step": 149060 }, { "epoch": 0.37, "learning_rate": 6.275378730685994e-05, "loss": 1.0197, "step": 149065 }, { "epoch": 0.37, "learning_rate": 6.275252906537823e-05, "loss": 1.0206, "step": 149070 }, { "epoch": 0.37, "learning_rate": 6.275127082389652e-05, "loss": 1.0199, "step": 149075 }, { "epoch": 0.37, "learning_rate": 6.275001258241481e-05, "loss": 1.0222, "step": 149080 }, { "epoch": 0.37, "learning_rate": 6.274875434093312e-05, "loss": 1.02, "step": 149085 }, { "epoch": 0.37, "learning_rate": 6.274749609945141e-05, "loss": 1.0212, "step": 149090 }, { "epoch": 0.37, "learning_rate": 6.27462378579697e-05, "loss": 1.0217, "step": 149095 }, { "epoch": 0.37, "learning_rate": 6.274497961648799e-05, "loss": 1.0204, "step": 149100 }, { "epoch": 0.37, "learning_rate": 6.27437213750063e-05, "loss": 1.0205, "step": 149105 }, { "epoch": 0.37, "learning_rate": 6.274246313352459e-05, "loss": 1.0212, "step": 149110 }, { "epoch": 0.37, "learning_rate": 6.274120489204288e-05, "loss": 1.0216, "step": 149115 }, { "epoch": 0.37, "learning_rate": 6.273994665056117e-05, "loss": 1.0213, "step": 149120 }, { "epoch": 0.37, "learning_rate": 6.273868840907946e-05, "loss": 1.0216, "step": 149125 }, { "epoch": 0.37, "learning_rate": 6.273743016759777e-05, "loss": 1.0185, "step": 149130 }, { "epoch": 0.37, "learning_rate": 6.273617192611606e-05, "loss": 1.0194, "step": 149135 }, { "epoch": 0.37, "learning_rate": 6.273491368463435e-05, "loss": 1.0202, "step": 149140 }, { "epoch": 0.37, "learning_rate": 6.273365544315264e-05, "loss": 1.021, "step": 149145 }, { "epoch": 0.37, "learning_rate": 6.273239720167095e-05, "loss": 1.0224, "step": 149150 }, { "epoch": 0.37, "learning_rate": 6.273113896018924e-05, "loss": 1.0185, "step": 149155 }, { "epoch": 0.37, "learning_rate": 6.272988071870753e-05, "loss": 1.023, "step": 149160 }, { "epoch": 0.37, "learning_rate": 6.272862247722582e-05, "loss": 1.0215, "step": 149165 }, { "epoch": 0.37, "learning_rate": 6.272736423574413e-05, "loss": 1.0204, "step": 149170 }, { "epoch": 0.37, "learning_rate": 6.272610599426242e-05, "loss": 1.0183, "step": 149175 }, { "epoch": 0.37, "learning_rate": 6.272484775278071e-05, "loss": 1.0197, "step": 149180 }, { "epoch": 0.37, "learning_rate": 6.2723589511299e-05, "loss": 1.0191, "step": 149185 }, { "epoch": 0.37, "learning_rate": 6.27223312698173e-05, "loss": 1.0221, "step": 149190 }, { "epoch": 0.37, "learning_rate": 6.27210730283356e-05, "loss": 1.0197, "step": 149195 }, { "epoch": 0.37, "learning_rate": 6.271981478685389e-05, "loss": 1.0202, "step": 149200 }, { "epoch": 0.37, "learning_rate": 6.271855654537218e-05, "loss": 1.0222, "step": 149205 }, { "epoch": 0.37, "learning_rate": 6.271729830389048e-05, "loss": 1.0217, "step": 149210 }, { "epoch": 0.37, "learning_rate": 6.271604006240878e-05, "loss": 1.0205, "step": 149215 }, { "epoch": 0.37, "learning_rate": 6.271478182092707e-05, "loss": 1.0178, "step": 149220 }, { "epoch": 0.37, "learning_rate": 6.271352357944536e-05, "loss": 1.021, "step": 149225 }, { "epoch": 0.37, "learning_rate": 6.271226533796366e-05, "loss": 1.0207, "step": 149230 }, { "epoch": 0.37, "learning_rate": 6.271100709648196e-05, "loss": 1.0187, "step": 149235 }, { "epoch": 0.37, "learning_rate": 6.270974885500025e-05, "loss": 1.0206, "step": 149240 }, { "epoch": 0.37, "learning_rate": 6.270849061351854e-05, "loss": 1.0189, "step": 149245 }, { "epoch": 0.37, "learning_rate": 6.270723237203684e-05, "loss": 1.0431, "step": 149250 }, { "epoch": 0.37, "learning_rate": 6.270597413055513e-05, "loss": 1.0206, "step": 149255 }, { "epoch": 0.37, "learning_rate": 6.270471588907343e-05, "loss": 1.0206, "step": 149260 }, { "epoch": 0.37, "learning_rate": 6.270345764759172e-05, "loss": 1.0218, "step": 149265 }, { "epoch": 0.37, "learning_rate": 6.270219940611002e-05, "loss": 1.0197, "step": 149270 }, { "epoch": 0.37, "learning_rate": 6.270094116462832e-05, "loss": 1.0249, "step": 149275 }, { "epoch": 0.37, "learning_rate": 6.269968292314661e-05, "loss": 1.021, "step": 149280 }, { "epoch": 0.37, "learning_rate": 6.269842468166492e-05, "loss": 1.0186, "step": 149285 }, { "epoch": 0.37, "learning_rate": 6.269716644018321e-05, "loss": 1.0211, "step": 149290 }, { "epoch": 0.37, "learning_rate": 6.26959081987015e-05, "loss": 1.0155, "step": 149295 }, { "epoch": 0.37, "learning_rate": 6.269464995721979e-05, "loss": 1.0192, "step": 149300 }, { "epoch": 0.37, "learning_rate": 6.26933917157381e-05, "loss": 1.0199, "step": 149305 }, { "epoch": 0.37, "learning_rate": 6.269213347425639e-05, "loss": 1.0191, "step": 149310 }, { "epoch": 0.37, "learning_rate": 6.269087523277468e-05, "loss": 1.0198, "step": 149315 }, { "epoch": 0.37, "learning_rate": 6.268961699129297e-05, "loss": 1.0179, "step": 149320 }, { "epoch": 0.37, "learning_rate": 6.268835874981128e-05, "loss": 1.0194, "step": 149325 }, { "epoch": 0.37, "learning_rate": 6.268710050832957e-05, "loss": 1.0202, "step": 149330 }, { "epoch": 0.37, "learning_rate": 6.268584226684786e-05, "loss": 1.021, "step": 149335 }, { "epoch": 0.37, "learning_rate": 6.268458402536615e-05, "loss": 1.0186, "step": 149340 }, { "epoch": 0.37, "learning_rate": 6.268332578388444e-05, "loss": 1.0219, "step": 149345 }, { "epoch": 0.37, "learning_rate": 6.268206754240275e-05, "loss": 1.0206, "step": 149350 }, { "epoch": 0.37, "learning_rate": 6.268080930092104e-05, "loss": 1.0559, "step": 149355 }, { "epoch": 0.37, "learning_rate": 6.267955105943933e-05, "loss": 1.0219, "step": 149360 }, { "epoch": 0.37, "learning_rate": 6.267829281795762e-05, "loss": 1.0207, "step": 149365 }, { "epoch": 0.37, "learning_rate": 6.267703457647593e-05, "loss": 1.0218, "step": 149370 }, { "epoch": 0.37, "learning_rate": 6.267577633499422e-05, "loss": 1.0211, "step": 149375 }, { "epoch": 0.37, "learning_rate": 6.267451809351251e-05, "loss": 1.0216, "step": 149380 }, { "epoch": 0.37, "learning_rate": 6.26732598520308e-05, "loss": 1.0208, "step": 149385 }, { "epoch": 0.37, "learning_rate": 6.267200161054911e-05, "loss": 1.0195, "step": 149390 }, { "epoch": 0.38, "learning_rate": 6.26707433690674e-05, "loss": 1.0227, "step": 149395 }, { "epoch": 0.38, "learning_rate": 6.266948512758569e-05, "loss": 1.0193, "step": 149400 }, { "epoch": 0.38, "learning_rate": 6.266822688610398e-05, "loss": 1.02, "step": 149405 }, { "epoch": 0.38, "learning_rate": 6.266696864462228e-05, "loss": 1.0214, "step": 149410 }, { "epoch": 0.38, "learning_rate": 6.266571040314058e-05, "loss": 1.019, "step": 149415 }, { "epoch": 0.38, "learning_rate": 6.266445216165887e-05, "loss": 1.0201, "step": 149420 }, { "epoch": 0.38, "learning_rate": 6.266319392017716e-05, "loss": 1.0188, "step": 149425 }, { "epoch": 0.38, "learning_rate": 6.266193567869546e-05, "loss": 1.0186, "step": 149430 }, { "epoch": 0.38, "learning_rate": 6.266067743721376e-05, "loss": 1.0211, "step": 149435 }, { "epoch": 0.38, "learning_rate": 6.265941919573205e-05, "loss": 1.0219, "step": 149440 }, { "epoch": 0.38, "learning_rate": 6.265816095425034e-05, "loss": 1.0211, "step": 149445 }, { "epoch": 0.38, "learning_rate": 6.265690271276864e-05, "loss": 1.0204, "step": 149450 }, { "epoch": 0.38, "learning_rate": 6.265564447128694e-05, "loss": 1.0196, "step": 149455 }, { "epoch": 0.38, "learning_rate": 6.265438622980523e-05, "loss": 1.0219, "step": 149460 }, { "epoch": 0.38, "learning_rate": 6.265312798832352e-05, "loss": 1.0189, "step": 149465 }, { "epoch": 0.38, "learning_rate": 6.265186974684182e-05, "loss": 1.0193, "step": 149470 }, { "epoch": 0.38, "learning_rate": 6.26506115053601e-05, "loss": 1.0195, "step": 149475 }, { "epoch": 0.38, "learning_rate": 6.264935326387841e-05, "loss": 1.0407, "step": 149480 }, { "epoch": 0.38, "learning_rate": 6.26480950223967e-05, "loss": 1.0236, "step": 149485 }, { "epoch": 0.38, "learning_rate": 6.2646836780915e-05, "loss": 1.0234, "step": 149490 }, { "epoch": 0.38, "learning_rate": 6.264557853943329e-05, "loss": 1.0212, "step": 149495 }, { "epoch": 0.38, "learning_rate": 6.264432029795159e-05, "loss": 1.0202, "step": 149500 }, { "epoch": 0.38, "learning_rate": 6.264306205646988e-05, "loss": 1.0206, "step": 149505 }, { "epoch": 0.38, "learning_rate": 6.264180381498818e-05, "loss": 1.0197, "step": 149510 }, { "epoch": 0.38, "learning_rate": 6.264054557350647e-05, "loss": 1.0208, "step": 149515 }, { "epoch": 0.38, "learning_rate": 6.263928733202477e-05, "loss": 1.0216, "step": 149520 }, { "epoch": 0.38, "learning_rate": 6.263802909054306e-05, "loss": 1.019, "step": 149525 }, { "epoch": 0.38, "learning_rate": 6.263677084906136e-05, "loss": 1.0187, "step": 149530 }, { "epoch": 0.38, "learning_rate": 6.263551260757965e-05, "loss": 1.0204, "step": 149535 }, { "epoch": 0.38, "learning_rate": 6.263425436609794e-05, "loss": 1.0206, "step": 149540 }, { "epoch": 0.38, "learning_rate": 6.263299612461624e-05, "loss": 1.0458, "step": 149545 }, { "epoch": 0.38, "learning_rate": 6.263173788313453e-05, "loss": 1.0211, "step": 149550 }, { "epoch": 0.38, "learning_rate": 6.263047964165283e-05, "loss": 1.0198, "step": 149555 }, { "epoch": 0.38, "learning_rate": 6.262922140017112e-05, "loss": 1.0216, "step": 149560 }, { "epoch": 0.38, "learning_rate": 6.262796315868942e-05, "loss": 1.0217, "step": 149565 }, { "epoch": 0.38, "learning_rate": 6.262670491720771e-05, "loss": 1.0212, "step": 149570 }, { "epoch": 0.38, "learning_rate": 6.2625446675726e-05, "loss": 1.0197, "step": 149575 }, { "epoch": 0.38, "learning_rate": 6.26241884342443e-05, "loss": 1.0201, "step": 149580 }, { "epoch": 0.38, "learning_rate": 6.26229301927626e-05, "loss": 1.021, "step": 149585 }, { "epoch": 0.38, "learning_rate": 6.26216719512809e-05, "loss": 1.021, "step": 149590 }, { "epoch": 0.38, "learning_rate": 6.262041370979919e-05, "loss": 1.0191, "step": 149595 }, { "epoch": 0.38, "learning_rate": 6.261915546831748e-05, "loss": 1.0209, "step": 149600 }, { "epoch": 0.38, "learning_rate": 6.261789722683577e-05, "loss": 1.0219, "step": 149605 }, { "epoch": 0.38, "learning_rate": 6.261663898535407e-05, "loss": 1.0226, "step": 149610 }, { "epoch": 0.38, "learning_rate": 6.261538074387237e-05, "loss": 1.0192, "step": 149615 }, { "epoch": 0.38, "learning_rate": 6.261412250239066e-05, "loss": 1.02, "step": 149620 }, { "epoch": 0.38, "learning_rate": 6.261286426090895e-05, "loss": 1.0208, "step": 149625 }, { "epoch": 0.38, "learning_rate": 6.261160601942725e-05, "loss": 1.0224, "step": 149630 }, { "epoch": 0.38, "learning_rate": 6.261034777794555e-05, "loss": 1.0218, "step": 149635 }, { "epoch": 0.38, "learning_rate": 6.260908953646384e-05, "loss": 1.021, "step": 149640 }, { "epoch": 0.38, "learning_rate": 6.260783129498213e-05, "loss": 1.0204, "step": 149645 }, { "epoch": 0.38, "learning_rate": 6.260657305350043e-05, "loss": 1.0211, "step": 149650 }, { "epoch": 0.38, "learning_rate": 6.260531481201873e-05, "loss": 1.0232, "step": 149655 }, { "epoch": 0.38, "learning_rate": 6.260405657053702e-05, "loss": 1.0194, "step": 149660 }, { "epoch": 0.38, "learning_rate": 6.260279832905531e-05, "loss": 1.0211, "step": 149665 }, { "epoch": 0.38, "learning_rate": 6.26015400875736e-05, "loss": 1.0187, "step": 149670 }, { "epoch": 0.38, "learning_rate": 6.26002818460919e-05, "loss": 1.0208, "step": 149675 }, { "epoch": 0.38, "learning_rate": 6.25990236046102e-05, "loss": 1.0192, "step": 149680 }, { "epoch": 0.38, "learning_rate": 6.259776536312849e-05, "loss": 1.0214, "step": 149685 }, { "epoch": 0.38, "learning_rate": 6.259650712164678e-05, "loss": 1.0204, "step": 149690 }, { "epoch": 0.38, "learning_rate": 6.259524888016509e-05, "loss": 1.0189, "step": 149695 }, { "epoch": 0.38, "learning_rate": 6.259399063868338e-05, "loss": 1.0203, "step": 149700 }, { "epoch": 0.38, "learning_rate": 6.259273239720167e-05, "loss": 1.0211, "step": 149705 }, { "epoch": 0.38, "learning_rate": 6.259147415571996e-05, "loss": 1.0172, "step": 149710 }, { "epoch": 0.38, "learning_rate": 6.259021591423827e-05, "loss": 1.0172, "step": 149715 }, { "epoch": 0.38, "learning_rate": 6.258895767275656e-05, "loss": 1.0211, "step": 149720 }, { "epoch": 0.38, "learning_rate": 6.258769943127485e-05, "loss": 1.0188, "step": 149725 }, { "epoch": 0.38, "learning_rate": 6.258644118979314e-05, "loss": 1.0218, "step": 149730 }, { "epoch": 0.38, "learning_rate": 6.258518294831143e-05, "loss": 1.0207, "step": 149735 }, { "epoch": 0.38, "learning_rate": 6.258392470682974e-05, "loss": 1.0211, "step": 149740 }, { "epoch": 0.38, "learning_rate": 6.258266646534803e-05, "loss": 1.0197, "step": 149745 }, { "epoch": 0.38, "learning_rate": 6.258140822386632e-05, "loss": 1.02, "step": 149750 }, { "epoch": 0.38, "learning_rate": 6.258014998238461e-05, "loss": 1.0228, "step": 149755 }, { "epoch": 0.38, "learning_rate": 6.257889174090292e-05, "loss": 1.0225, "step": 149760 }, { "epoch": 0.38, "learning_rate": 6.257763349942121e-05, "loss": 1.017, "step": 149765 }, { "epoch": 0.38, "learning_rate": 6.25763752579395e-05, "loss": 1.0205, "step": 149770 }, { "epoch": 0.38, "learning_rate": 6.25751170164578e-05, "loss": 1.0206, "step": 149775 }, { "epoch": 0.38, "learning_rate": 6.25738587749761e-05, "loss": 1.0202, "step": 149780 }, { "epoch": 0.38, "learning_rate": 6.25726005334944e-05, "loss": 1.0246, "step": 149785 }, { "epoch": 0.38, "learning_rate": 6.25713422920127e-05, "loss": 1.0188, "step": 149790 }, { "epoch": 0.38, "learning_rate": 6.257008405053099e-05, "loss": 1.0227, "step": 149795 }, { "epoch": 0.38, "learning_rate": 6.256882580904928e-05, "loss": 1.0208, "step": 149800 }, { "epoch": 0.38, "learning_rate": 6.256756756756757e-05, "loss": 1.0183, "step": 149805 }, { "epoch": 0.38, "learning_rate": 6.256630932608587e-05, "loss": 1.0197, "step": 149810 }, { "epoch": 0.38, "learning_rate": 6.256505108460417e-05, "loss": 1.0173, "step": 149815 }, { "epoch": 0.38, "learning_rate": 6.256379284312246e-05, "loss": 1.0199, "step": 149820 }, { "epoch": 0.38, "learning_rate": 6.256253460164075e-05, "loss": 1.0175, "step": 149825 }, { "epoch": 0.38, "learning_rate": 6.256127636015905e-05, "loss": 1.0207, "step": 149830 }, { "epoch": 0.38, "learning_rate": 6.256001811867735e-05, "loss": 1.0192, "step": 149835 }, { "epoch": 0.38, "learning_rate": 6.255875987719564e-05, "loss": 1.0182, "step": 149840 }, { "epoch": 0.38, "learning_rate": 6.255750163571393e-05, "loss": 1.0216, "step": 149845 }, { "epoch": 0.38, "learning_rate": 6.255624339423223e-05, "loss": 1.0228, "step": 149850 }, { "epoch": 0.38, "learning_rate": 6.255498515275053e-05, "loss": 1.0207, "step": 149855 }, { "epoch": 0.38, "learning_rate": 6.255372691126882e-05, "loss": 1.0195, "step": 149860 }, { "epoch": 0.38, "learning_rate": 6.255246866978711e-05, "loss": 1.0192, "step": 149865 }, { "epoch": 0.38, "learning_rate": 6.25512104283054e-05, "loss": 1.0198, "step": 149870 }, { "epoch": 0.38, "learning_rate": 6.25499521868237e-05, "loss": 1.02, "step": 149875 }, { "epoch": 0.38, "learning_rate": 6.2548693945342e-05, "loss": 1.0191, "step": 149880 }, { "epoch": 0.38, "learning_rate": 6.254743570386029e-05, "loss": 1.0197, "step": 149885 }, { "epoch": 0.38, "learning_rate": 6.254617746237858e-05, "loss": 1.0192, "step": 149890 }, { "epoch": 0.38, "learning_rate": 6.254491922089689e-05, "loss": 1.0218, "step": 149895 }, { "epoch": 0.38, "learning_rate": 6.254366097941518e-05, "loss": 1.0204, "step": 149900 }, { "epoch": 0.38, "learning_rate": 6.254240273793347e-05, "loss": 1.0181, "step": 149905 }, { "epoch": 0.38, "learning_rate": 6.254114449645176e-05, "loss": 1.0197, "step": 149910 }, { "epoch": 0.38, "learning_rate": 6.253988625497007e-05, "loss": 1.0196, "step": 149915 }, { "epoch": 0.38, "learning_rate": 6.253862801348836e-05, "loss": 1.0194, "step": 149920 }, { "epoch": 0.38, "learning_rate": 6.253736977200665e-05, "loss": 1.0191, "step": 149925 }, { "epoch": 0.38, "learning_rate": 6.253611153052494e-05, "loss": 1.0214, "step": 149930 }, { "epoch": 0.38, "learning_rate": 6.253485328904323e-05, "loss": 1.0207, "step": 149935 }, { "epoch": 0.38, "learning_rate": 6.253359504756154e-05, "loss": 1.0199, "step": 149940 }, { "epoch": 0.38, "learning_rate": 6.253233680607983e-05, "loss": 1.0201, "step": 149945 }, { "epoch": 0.38, "learning_rate": 6.253107856459812e-05, "loss": 1.0206, "step": 149950 }, { "epoch": 0.38, "learning_rate": 6.252982032311641e-05, "loss": 1.022, "step": 149955 }, { "epoch": 0.38, "learning_rate": 6.252856208163472e-05, "loss": 1.0209, "step": 149960 }, { "epoch": 0.38, "learning_rate": 6.252730384015301e-05, "loss": 1.0205, "step": 149965 }, { "epoch": 0.38, "learning_rate": 6.25260455986713e-05, "loss": 1.0202, "step": 149970 }, { "epoch": 0.38, "learning_rate": 6.252478735718959e-05, "loss": 1.0196, "step": 149975 }, { "epoch": 0.38, "learning_rate": 6.25235291157079e-05, "loss": 1.0193, "step": 149980 }, { "epoch": 0.38, "learning_rate": 6.252227087422619e-05, "loss": 1.0199, "step": 149985 }, { "epoch": 0.38, "learning_rate": 6.252101263274448e-05, "loss": 1.0426, "step": 149990 }, { "epoch": 0.38, "learning_rate": 6.251975439126277e-05, "loss": 1.021, "step": 149995 }, { "epoch": 0.38, "learning_rate": 6.251849614978106e-05, "loss": 1.0204, "step": 150000 }, { "epoch": 0.38, "learning_rate": 6.251723790829937e-05, "loss": 1.0182, "step": 150005 }, { "epoch": 0.38, "learning_rate": 6.251597966681766e-05, "loss": 1.0202, "step": 150010 }, { "epoch": 0.38, "learning_rate": 6.251472142533595e-05, "loss": 1.0227, "step": 150015 }, { "epoch": 0.38, "learning_rate": 6.251346318385424e-05, "loss": 1.0214, "step": 150020 }, { "epoch": 0.38, "learning_rate": 6.251220494237255e-05, "loss": 1.0216, "step": 150025 }, { "epoch": 0.38, "learning_rate": 6.251094670089084e-05, "loss": 1.0209, "step": 150030 }, { "epoch": 0.38, "learning_rate": 6.250968845940913e-05, "loss": 1.0209, "step": 150035 }, { "epoch": 0.38, "learning_rate": 6.250843021792742e-05, "loss": 1.0408, "step": 150040 }, { "epoch": 0.38, "learning_rate": 6.250717197644573e-05, "loss": 1.0219, "step": 150045 }, { "epoch": 0.38, "learning_rate": 6.250591373496402e-05, "loss": 1.0188, "step": 150050 }, { "epoch": 0.38, "learning_rate": 6.250465549348231e-05, "loss": 1.0212, "step": 150055 }, { "epoch": 0.38, "learning_rate": 6.25033972520006e-05, "loss": 1.0211, "step": 150060 }, { "epoch": 0.38, "learning_rate": 6.25021390105189e-05, "loss": 1.0196, "step": 150065 }, { "epoch": 0.38, "learning_rate": 6.25008807690372e-05, "loss": 1.0203, "step": 150070 }, { "epoch": 0.38, "learning_rate": 6.249962252755549e-05, "loss": 1.0204, "step": 150075 }, { "epoch": 0.38, "learning_rate": 6.249836428607378e-05, "loss": 1.0216, "step": 150080 }, { "epoch": 0.38, "learning_rate": 6.249710604459207e-05, "loss": 1.0186, "step": 150085 }, { "epoch": 0.38, "learning_rate": 6.249584780311038e-05, "loss": 1.022, "step": 150090 }, { "epoch": 0.38, "learning_rate": 6.249458956162867e-05, "loss": 1.0204, "step": 150095 }, { "epoch": 0.38, "learning_rate": 6.249333132014696e-05, "loss": 1.02, "step": 150100 }, { "epoch": 0.38, "learning_rate": 6.249207307866525e-05, "loss": 1.0199, "step": 150105 }, { "epoch": 0.38, "learning_rate": 6.249081483718356e-05, "loss": 1.021, "step": 150110 }, { "epoch": 0.38, "learning_rate": 6.248955659570185e-05, "loss": 1.0188, "step": 150115 }, { "epoch": 0.38, "learning_rate": 6.248829835422014e-05, "loss": 1.0186, "step": 150120 }, { "epoch": 0.38, "learning_rate": 6.248704011273843e-05, "loss": 1.0192, "step": 150125 }, { "epoch": 0.38, "learning_rate": 6.248578187125673e-05, "loss": 1.019, "step": 150130 }, { "epoch": 0.38, "learning_rate": 6.248452362977503e-05, "loss": 1.0202, "step": 150135 }, { "epoch": 0.38, "learning_rate": 6.248326538829332e-05, "loss": 1.0383, "step": 150140 }, { "epoch": 0.38, "learning_rate": 6.248200714681161e-05, "loss": 1.0218, "step": 150145 }, { "epoch": 0.38, "learning_rate": 6.24807489053299e-05, "loss": 1.0329, "step": 150150 }, { "epoch": 0.38, "learning_rate": 6.247949066384821e-05, "loss": 1.0193, "step": 150155 }, { "epoch": 0.38, "learning_rate": 6.24782324223665e-05, "loss": 1.0228, "step": 150160 }, { "epoch": 0.38, "learning_rate": 6.24769741808848e-05, "loss": 1.0211, "step": 150165 }, { "epoch": 0.38, "learning_rate": 6.247571593940309e-05, "loss": 1.0207, "step": 150170 }, { "epoch": 0.38, "learning_rate": 6.247445769792139e-05, "loss": 1.0209, "step": 150175 }, { "epoch": 0.38, "learning_rate": 6.247319945643968e-05, "loss": 1.0198, "step": 150180 }, { "epoch": 0.38, "learning_rate": 6.247194121495797e-05, "loss": 1.0209, "step": 150185 }, { "epoch": 0.38, "learning_rate": 6.247068297347627e-05, "loss": 1.022, "step": 150190 }, { "epoch": 0.38, "learning_rate": 6.246942473199456e-05, "loss": 1.0208, "step": 150195 }, { "epoch": 0.38, "learning_rate": 6.246816649051286e-05, "loss": 1.017, "step": 150200 }, { "epoch": 0.38, "learning_rate": 6.246690824903115e-05, "loss": 1.0211, "step": 150205 }, { "epoch": 0.38, "learning_rate": 6.246565000754945e-05, "loss": 1.0212, "step": 150210 }, { "epoch": 0.38, "learning_rate": 6.246439176606774e-05, "loss": 1.0193, "step": 150215 }, { "epoch": 0.38, "learning_rate": 6.246313352458604e-05, "loss": 1.0226, "step": 150220 }, { "epoch": 0.38, "learning_rate": 6.246187528310433e-05, "loss": 1.02, "step": 150225 }, { "epoch": 0.38, "learning_rate": 6.246061704162263e-05, "loss": 1.0196, "step": 150230 }, { "epoch": 0.38, "learning_rate": 6.245935880014092e-05, "loss": 1.0206, "step": 150235 }, { "epoch": 0.38, "learning_rate": 6.245810055865922e-05, "loss": 1.0181, "step": 150240 }, { "epoch": 0.38, "learning_rate": 6.245684231717751e-05, "loss": 1.0224, "step": 150245 }, { "epoch": 0.38, "learning_rate": 6.24555840756958e-05, "loss": 1.0201, "step": 150250 }, { "epoch": 0.38, "learning_rate": 6.24543258342141e-05, "loss": 1.0186, "step": 150255 }, { "epoch": 0.38, "learning_rate": 6.245306759273239e-05, "loss": 1.0198, "step": 150260 }, { "epoch": 0.38, "learning_rate": 6.24518093512507e-05, "loss": 1.0206, "step": 150265 }, { "epoch": 0.38, "learning_rate": 6.245055110976899e-05, "loss": 1.0204, "step": 150270 }, { "epoch": 0.38, "learning_rate": 6.244929286828729e-05, "loss": 1.0209, "step": 150275 }, { "epoch": 0.38, "learning_rate": 6.244803462680558e-05, "loss": 1.0199, "step": 150280 }, { "epoch": 0.38, "learning_rate": 6.244677638532387e-05, "loss": 1.0195, "step": 150285 }, { "epoch": 0.38, "learning_rate": 6.244551814384218e-05, "loss": 1.0177, "step": 150290 }, { "epoch": 0.38, "learning_rate": 6.244425990236047e-05, "loss": 1.0413, "step": 150295 }, { "epoch": 0.38, "learning_rate": 6.244300166087876e-05, "loss": 1.0403, "step": 150300 }, { "epoch": 0.38, "learning_rate": 6.244174341939705e-05, "loss": 1.0201, "step": 150305 }, { "epoch": 0.38, "learning_rate": 6.244048517791536e-05, "loss": 1.021, "step": 150310 }, { "epoch": 0.38, "learning_rate": 6.243922693643365e-05, "loss": 1.0206, "step": 150315 }, { "epoch": 0.38, "learning_rate": 6.243796869495194e-05, "loss": 1.0185, "step": 150320 }, { "epoch": 0.38, "learning_rate": 6.243671045347023e-05, "loss": 1.0213, "step": 150325 }, { "epoch": 0.38, "learning_rate": 6.243545221198853e-05, "loss": 1.0212, "step": 150330 }, { "epoch": 0.38, "learning_rate": 6.243419397050683e-05, "loss": 1.0198, "step": 150335 }, { "epoch": 0.38, "learning_rate": 6.243293572902512e-05, "loss": 1.021, "step": 150340 }, { "epoch": 0.38, "learning_rate": 6.243167748754341e-05, "loss": 1.0208, "step": 150345 }, { "epoch": 0.38, "learning_rate": 6.24304192460617e-05, "loss": 1.0213, "step": 150350 }, { "epoch": 0.38, "learning_rate": 6.242916100458001e-05, "loss": 1.0195, "step": 150355 }, { "epoch": 0.38, "learning_rate": 6.24279027630983e-05, "loss": 1.0172, "step": 150360 }, { "epoch": 0.38, "learning_rate": 6.24266445216166e-05, "loss": 1.0208, "step": 150365 }, { "epoch": 0.38, "learning_rate": 6.242538628013489e-05, "loss": 1.02, "step": 150370 }, { "epoch": 0.38, "learning_rate": 6.242412803865319e-05, "loss": 1.0185, "step": 150375 }, { "epoch": 0.38, "learning_rate": 6.242286979717148e-05, "loss": 1.0217, "step": 150380 }, { "epoch": 0.38, "learning_rate": 6.242161155568977e-05, "loss": 1.0223, "step": 150385 }, { "epoch": 0.38, "learning_rate": 6.242035331420807e-05, "loss": 1.0187, "step": 150390 }, { "epoch": 0.38, "learning_rate": 6.241909507272636e-05, "loss": 1.0208, "step": 150395 }, { "epoch": 0.38, "learning_rate": 6.241783683124466e-05, "loss": 1.0224, "step": 150400 }, { "epoch": 0.38, "learning_rate": 6.241657858976295e-05, "loss": 1.0191, "step": 150405 }, { "epoch": 0.38, "learning_rate": 6.241532034828125e-05, "loss": 1.0235, "step": 150410 }, { "epoch": 0.38, "learning_rate": 6.241406210679954e-05, "loss": 1.0187, "step": 150415 }, { "epoch": 0.38, "learning_rate": 6.241280386531784e-05, "loss": 1.0197, "step": 150420 }, { "epoch": 0.38, "learning_rate": 6.241154562383613e-05, "loss": 1.0232, "step": 150425 }, { "epoch": 0.38, "learning_rate": 6.241028738235443e-05, "loss": 1.0243, "step": 150430 }, { "epoch": 0.38, "learning_rate": 6.240902914087272e-05, "loss": 1.0213, "step": 150435 }, { "epoch": 0.38, "learning_rate": 6.240777089939102e-05, "loss": 1.022, "step": 150440 }, { "epoch": 0.38, "learning_rate": 6.240651265790931e-05, "loss": 1.0199, "step": 150445 }, { "epoch": 0.38, "learning_rate": 6.24052544164276e-05, "loss": 1.0187, "step": 150450 }, { "epoch": 0.38, "learning_rate": 6.24039961749459e-05, "loss": 1.0211, "step": 150455 }, { "epoch": 0.38, "learning_rate": 6.240273793346419e-05, "loss": 1.0215, "step": 150460 }, { "epoch": 0.38, "learning_rate": 6.24014796919825e-05, "loss": 1.0407, "step": 150465 }, { "epoch": 0.38, "learning_rate": 6.240022145050079e-05, "loss": 1.0202, "step": 150470 }, { "epoch": 0.38, "learning_rate": 6.239896320901908e-05, "loss": 1.0199, "step": 150475 }, { "epoch": 0.38, "learning_rate": 6.239770496753737e-05, "loss": 1.0207, "step": 150480 }, { "epoch": 0.38, "learning_rate": 6.239644672605567e-05, "loss": 1.0184, "step": 150485 }, { "epoch": 0.38, "learning_rate": 6.239518848457397e-05, "loss": 1.0189, "step": 150490 }, { "epoch": 0.38, "learning_rate": 6.239393024309226e-05, "loss": 1.0213, "step": 150495 }, { "epoch": 0.38, "learning_rate": 6.239267200161055e-05, "loss": 1.0207, "step": 150500 }, { "epoch": 0.38, "learning_rate": 6.239141376012885e-05, "loss": 1.0201, "step": 150505 }, { "epoch": 0.38, "learning_rate": 6.239015551864715e-05, "loss": 1.0212, "step": 150510 }, { "epoch": 0.38, "learning_rate": 6.238889727716544e-05, "loss": 1.0185, "step": 150515 }, { "epoch": 0.38, "learning_rate": 6.238763903568373e-05, "loss": 1.0347, "step": 150520 }, { "epoch": 0.38, "learning_rate": 6.238638079420202e-05, "loss": 1.0194, "step": 150525 }, { "epoch": 0.38, "learning_rate": 6.238512255272033e-05, "loss": 1.0203, "step": 150530 }, { "epoch": 0.38, "learning_rate": 6.238386431123862e-05, "loss": 1.0201, "step": 150535 }, { "epoch": 0.38, "learning_rate": 6.238260606975691e-05, "loss": 1.0202, "step": 150540 }, { "epoch": 0.38, "learning_rate": 6.23813478282752e-05, "loss": 1.0216, "step": 150545 }, { "epoch": 0.38, "learning_rate": 6.23800895867935e-05, "loss": 1.0186, "step": 150550 }, { "epoch": 0.38, "learning_rate": 6.23788313453118e-05, "loss": 1.0201, "step": 150555 }, { "epoch": 0.38, "learning_rate": 6.237757310383009e-05, "loss": 1.0199, "step": 150560 }, { "epoch": 0.38, "learning_rate": 6.237631486234838e-05, "loss": 1.0188, "step": 150565 }, { "epoch": 0.38, "learning_rate": 6.237505662086668e-05, "loss": 1.0207, "step": 150570 }, { "epoch": 0.38, "learning_rate": 6.237379837938498e-05, "loss": 1.0219, "step": 150575 }, { "epoch": 0.38, "learning_rate": 6.237254013790327e-05, "loss": 1.0203, "step": 150580 }, { "epoch": 0.38, "learning_rate": 6.237128189642156e-05, "loss": 1.0211, "step": 150585 }, { "epoch": 0.38, "learning_rate": 6.237002365493985e-05, "loss": 1.0232, "step": 150590 }, { "epoch": 0.38, "learning_rate": 6.236876541345816e-05, "loss": 1.0214, "step": 150595 }, { "epoch": 0.38, "learning_rate": 6.236750717197645e-05, "loss": 1.0203, "step": 150600 }, { "epoch": 0.38, "learning_rate": 6.236624893049474e-05, "loss": 1.0209, "step": 150605 }, { "epoch": 0.38, "learning_rate": 6.236499068901303e-05, "loss": 1.0217, "step": 150610 }, { "epoch": 0.38, "learning_rate": 6.236373244753134e-05, "loss": 1.0222, "step": 150615 }, { "epoch": 0.38, "learning_rate": 6.236247420604963e-05, "loss": 1.0193, "step": 150620 }, { "epoch": 0.38, "learning_rate": 6.236121596456792e-05, "loss": 1.0216, "step": 150625 }, { "epoch": 0.38, "learning_rate": 6.235995772308621e-05, "loss": 1.019, "step": 150630 }, { "epoch": 0.38, "learning_rate": 6.235869948160452e-05, "loss": 1.018, "step": 150635 }, { "epoch": 0.38, "learning_rate": 6.235744124012281e-05, "loss": 1.0204, "step": 150640 }, { "epoch": 0.38, "learning_rate": 6.23561829986411e-05, "loss": 1.0206, "step": 150645 }, { "epoch": 0.38, "learning_rate": 6.235492475715939e-05, "loss": 1.0198, "step": 150650 }, { "epoch": 0.38, "learning_rate": 6.235366651567768e-05, "loss": 1.0199, "step": 150655 }, { "epoch": 0.38, "learning_rate": 6.235240827419599e-05, "loss": 1.0194, "step": 150660 }, { "epoch": 0.38, "learning_rate": 6.235115003271428e-05, "loss": 1.0188, "step": 150665 }, { "epoch": 0.38, "learning_rate": 6.234989179123257e-05, "loss": 1.0224, "step": 150670 }, { "epoch": 0.38, "learning_rate": 6.234863354975086e-05, "loss": 1.0217, "step": 150675 }, { "epoch": 0.38, "learning_rate": 6.234737530826917e-05, "loss": 1.0225, "step": 150680 }, { "epoch": 0.38, "learning_rate": 6.234611706678746e-05, "loss": 1.0208, "step": 150685 }, { "epoch": 0.38, "learning_rate": 6.234485882530575e-05, "loss": 1.0224, "step": 150690 }, { "epoch": 0.38, "learning_rate": 6.234360058382404e-05, "loss": 1.0205, "step": 150695 }, { "epoch": 0.38, "learning_rate": 6.234234234234235e-05, "loss": 1.0201, "step": 150700 }, { "epoch": 0.38, "learning_rate": 6.234108410086064e-05, "loss": 1.0334, "step": 150705 }, { "epoch": 0.38, "learning_rate": 6.233982585937893e-05, "loss": 1.0204, "step": 150710 }, { "epoch": 0.38, "learning_rate": 6.233856761789722e-05, "loss": 1.0177, "step": 150715 }, { "epoch": 0.38, "learning_rate": 6.233730937641551e-05, "loss": 1.0201, "step": 150720 }, { "epoch": 0.38, "learning_rate": 6.233605113493382e-05, "loss": 1.022, "step": 150725 }, { "epoch": 0.38, "learning_rate": 6.233479289345211e-05, "loss": 1.045, "step": 150730 }, { "epoch": 0.38, "learning_rate": 6.23335346519704e-05, "loss": 1.0204, "step": 150735 }, { "epoch": 0.38, "learning_rate": 6.23322764104887e-05, "loss": 1.0217, "step": 150740 }, { "epoch": 0.38, "learning_rate": 6.2331018169007e-05, "loss": 1.0203, "step": 150745 }, { "epoch": 0.38, "learning_rate": 6.232975992752529e-05, "loss": 1.0172, "step": 150750 }, { "epoch": 0.38, "learning_rate": 6.232875333433993e-05, "loss": 1.0177, "step": 150755 }, { "epoch": 0.38, "learning_rate": 6.232749509285822e-05, "loss": 1.021, "step": 150760 }, { "epoch": 0.38, "learning_rate": 6.232623685137652e-05, "loss": 1.0209, "step": 150765 }, { "epoch": 0.38, "learning_rate": 6.232497860989482e-05, "loss": 1.0208, "step": 150770 }, { "epoch": 0.38, "learning_rate": 6.23237203684131e-05, "loss": 1.0195, "step": 150775 }, { "epoch": 0.38, "learning_rate": 6.23224621269314e-05, "loss": 1.0191, "step": 150780 }, { "epoch": 0.38, "learning_rate": 6.23212038854497e-05, "loss": 1.0197, "step": 150785 }, { "epoch": 0.38, "learning_rate": 6.2319945643968e-05, "loss": 1.0228, "step": 150790 }, { "epoch": 0.38, "learning_rate": 6.231868740248629e-05, "loss": 1.0205, "step": 150795 }, { "epoch": 0.38, "learning_rate": 6.231742916100458e-05, "loss": 1.0197, "step": 150800 }, { "epoch": 0.38, "learning_rate": 6.231617091952287e-05, "loss": 1.0207, "step": 150805 }, { "epoch": 0.38, "learning_rate": 6.231491267804118e-05, "loss": 1.0201, "step": 150810 }, { "epoch": 0.38, "learning_rate": 6.231365443655947e-05, "loss": 1.0185, "step": 150815 }, { "epoch": 0.38, "learning_rate": 6.231239619507776e-05, "loss": 1.0187, "step": 150820 }, { "epoch": 0.38, "learning_rate": 6.231113795359605e-05, "loss": 1.0395, "step": 150825 }, { "epoch": 0.38, "learning_rate": 6.230987971211435e-05, "loss": 1.0186, "step": 150830 }, { "epoch": 0.38, "learning_rate": 6.230862147063265e-05, "loss": 1.0375, "step": 150835 }, { "epoch": 0.38, "learning_rate": 6.230736322915094e-05, "loss": 1.0218, "step": 150840 }, { "epoch": 0.38, "learning_rate": 6.230610498766923e-05, "loss": 1.0194, "step": 150845 }, { "epoch": 0.38, "learning_rate": 6.230484674618753e-05, "loss": 1.0205, "step": 150850 }, { "epoch": 0.38, "learning_rate": 6.230358850470583e-05, "loss": 1.0187, "step": 150855 }, { "epoch": 0.38, "learning_rate": 6.230233026322412e-05, "loss": 1.0179, "step": 150860 }, { "epoch": 0.38, "learning_rate": 6.230107202174241e-05, "loss": 1.0206, "step": 150865 }, { "epoch": 0.38, "learning_rate": 6.22998137802607e-05, "loss": 1.0219, "step": 150870 }, { "epoch": 0.38, "learning_rate": 6.2298555538779e-05, "loss": 1.0215, "step": 150875 }, { "epoch": 0.38, "learning_rate": 6.22972972972973e-05, "loss": 1.0191, "step": 150880 }, { "epoch": 0.38, "learning_rate": 6.229603905581559e-05, "loss": 1.0206, "step": 150885 }, { "epoch": 0.38, "learning_rate": 6.229478081433388e-05, "loss": 1.0212, "step": 150890 }, { "epoch": 0.38, "learning_rate": 6.229352257285219e-05, "loss": 1.0234, "step": 150895 }, { "epoch": 0.38, "learning_rate": 6.229226433137048e-05, "loss": 1.0208, "step": 150900 }, { "epoch": 0.38, "learning_rate": 6.229100608988877e-05, "loss": 1.0215, "step": 150905 }, { "epoch": 0.38, "learning_rate": 6.228974784840706e-05, "loss": 1.0202, "step": 150910 }, { "epoch": 0.38, "learning_rate": 6.228848960692537e-05, "loss": 1.0214, "step": 150915 }, { "epoch": 0.38, "learning_rate": 6.228723136544366e-05, "loss": 1.0188, "step": 150920 }, { "epoch": 0.38, "learning_rate": 6.228597312396195e-05, "loss": 1.0218, "step": 150925 }, { "epoch": 0.38, "learning_rate": 6.228471488248024e-05, "loss": 1.0365, "step": 150930 }, { "epoch": 0.38, "learning_rate": 6.228345664099853e-05, "loss": 1.0211, "step": 150935 }, { "epoch": 0.38, "learning_rate": 6.228219839951684e-05, "loss": 1.0186, "step": 150940 }, { "epoch": 0.38, "learning_rate": 6.228094015803513e-05, "loss": 1.0199, "step": 150945 }, { "epoch": 0.38, "learning_rate": 6.227968191655342e-05, "loss": 1.0229, "step": 150950 }, { "epoch": 0.38, "learning_rate": 6.227842367507171e-05, "loss": 1.0225, "step": 150955 }, { "epoch": 0.38, "learning_rate": 6.227716543359002e-05, "loss": 1.0193, "step": 150960 }, { "epoch": 0.38, "learning_rate": 6.227590719210831e-05, "loss": 1.0211, "step": 150965 }, { "epoch": 0.38, "learning_rate": 6.22746489506266e-05, "loss": 1.0201, "step": 150970 }, { "epoch": 0.38, "learning_rate": 6.22733907091449e-05, "loss": 1.0213, "step": 150975 }, { "epoch": 0.38, "learning_rate": 6.22721324676632e-05, "loss": 1.0204, "step": 150980 }, { "epoch": 0.38, "learning_rate": 6.22708742261815e-05, "loss": 1.0208, "step": 150985 }, { "epoch": 0.38, "learning_rate": 6.22696159846998e-05, "loss": 1.0197, "step": 150990 }, { "epoch": 0.38, "learning_rate": 6.226835774321809e-05, "loss": 1.0189, "step": 150995 }, { "epoch": 0.38, "learning_rate": 6.226709950173638e-05, "loss": 1.0226, "step": 151000 }, { "epoch": 0.38, "learning_rate": 6.226584126025467e-05, "loss": 1.0217, "step": 151005 }, { "epoch": 0.38, "learning_rate": 6.226458301877297e-05, "loss": 1.021, "step": 151010 }, { "epoch": 0.38, "learning_rate": 6.226332477729127e-05, "loss": 1.0194, "step": 151015 }, { "epoch": 0.38, "learning_rate": 6.226206653580956e-05, "loss": 1.0218, "step": 151020 }, { "epoch": 0.38, "learning_rate": 6.226080829432785e-05, "loss": 1.0203, "step": 151025 }, { "epoch": 0.38, "learning_rate": 6.225955005284615e-05, "loss": 1.0193, "step": 151030 }, { "epoch": 0.38, "learning_rate": 6.225829181136445e-05, "loss": 1.0194, "step": 151035 }, { "epoch": 0.38, "learning_rate": 6.225703356988274e-05, "loss": 1.0196, "step": 151040 }, { "epoch": 0.38, "learning_rate": 6.225577532840103e-05, "loss": 1.0199, "step": 151045 }, { "epoch": 0.38, "learning_rate": 6.225451708691933e-05, "loss": 1.0205, "step": 151050 }, { "epoch": 0.38, "learning_rate": 6.225325884543763e-05, "loss": 1.021, "step": 151055 }, { "epoch": 0.38, "learning_rate": 6.225200060395592e-05, "loss": 1.0186, "step": 151060 }, { "epoch": 0.38, "learning_rate": 6.225074236247421e-05, "loss": 1.0206, "step": 151065 }, { "epoch": 0.38, "learning_rate": 6.22494841209925e-05, "loss": 1.0217, "step": 151070 }, { "epoch": 0.38, "learning_rate": 6.22482258795108e-05, "loss": 1.0194, "step": 151075 }, { "epoch": 0.38, "learning_rate": 6.22469676380291e-05, "loss": 1.0416, "step": 151080 }, { "epoch": 0.38, "learning_rate": 6.224570939654739e-05, "loss": 1.021, "step": 151085 }, { "epoch": 0.38, "learning_rate": 6.224445115506568e-05, "loss": 1.0199, "step": 151090 }, { "epoch": 0.38, "learning_rate": 6.224319291358399e-05, "loss": 1.0223, "step": 151095 }, { "epoch": 0.38, "learning_rate": 6.224193467210228e-05, "loss": 1.0213, "step": 151100 }, { "epoch": 0.38, "learning_rate": 6.224067643062057e-05, "loss": 1.0224, "step": 151105 }, { "epoch": 0.38, "learning_rate": 6.223941818913886e-05, "loss": 1.0219, "step": 151110 }, { "epoch": 0.38, "learning_rate": 6.223815994765717e-05, "loss": 1.0224, "step": 151115 }, { "epoch": 0.38, "learning_rate": 6.223690170617546e-05, "loss": 1.0194, "step": 151120 }, { "epoch": 0.38, "learning_rate": 6.223564346469375e-05, "loss": 1.0187, "step": 151125 }, { "epoch": 0.38, "learning_rate": 6.223438522321204e-05, "loss": 1.0189, "step": 151130 }, { "epoch": 0.38, "learning_rate": 6.223312698173033e-05, "loss": 1.0194, "step": 151135 }, { "epoch": 0.38, "learning_rate": 6.223186874024864e-05, "loss": 1.0221, "step": 151140 }, { "epoch": 0.38, "learning_rate": 6.223061049876693e-05, "loss": 1.0199, "step": 151145 }, { "epoch": 0.38, "learning_rate": 6.222935225728522e-05, "loss": 1.0187, "step": 151150 }, { "epoch": 0.38, "learning_rate": 6.222809401580351e-05, "loss": 1.0193, "step": 151155 }, { "epoch": 0.38, "learning_rate": 6.222683577432182e-05, "loss": 1.0205, "step": 151160 }, { "epoch": 0.38, "learning_rate": 6.222557753284011e-05, "loss": 1.0191, "step": 151165 }, { "epoch": 0.38, "learning_rate": 6.22243192913584e-05, "loss": 1.0188, "step": 151170 }, { "epoch": 0.38, "learning_rate": 6.222306104987669e-05, "loss": 1.0207, "step": 151175 }, { "epoch": 0.38, "learning_rate": 6.2221802808395e-05, "loss": 1.0213, "step": 151180 }, { "epoch": 0.38, "learning_rate": 6.222054456691329e-05, "loss": 1.0191, "step": 151185 }, { "epoch": 0.38, "learning_rate": 6.221928632543158e-05, "loss": 1.0386, "step": 151190 }, { "epoch": 0.38, "learning_rate": 6.221802808394987e-05, "loss": 1.0198, "step": 151195 }, { "epoch": 0.38, "learning_rate": 6.221676984246816e-05, "loss": 1.0201, "step": 151200 }, { "epoch": 0.38, "learning_rate": 6.221551160098647e-05, "loss": 1.0196, "step": 151205 }, { "epoch": 0.38, "learning_rate": 6.221425335950476e-05, "loss": 1.0212, "step": 151210 }, { "epoch": 0.38, "learning_rate": 6.221299511802305e-05, "loss": 1.0202, "step": 151215 }, { "epoch": 0.38, "learning_rate": 6.221173687654134e-05, "loss": 1.0166, "step": 151220 }, { "epoch": 0.38, "learning_rate": 6.221047863505965e-05, "loss": 1.0202, "step": 151225 }, { "epoch": 0.38, "learning_rate": 6.220922039357794e-05, "loss": 1.0203, "step": 151230 }, { "epoch": 0.38, "learning_rate": 6.220796215209623e-05, "loss": 1.0205, "step": 151235 }, { "epoch": 0.38, "learning_rate": 6.220670391061452e-05, "loss": 1.0226, "step": 151240 }, { "epoch": 0.38, "learning_rate": 6.220544566913283e-05, "loss": 1.0203, "step": 151245 }, { "epoch": 0.38, "learning_rate": 6.220418742765112e-05, "loss": 1.023, "step": 151250 }, { "epoch": 0.38, "learning_rate": 6.220292918616941e-05, "loss": 1.0179, "step": 151255 }, { "epoch": 0.38, "learning_rate": 6.22016709446877e-05, "loss": 1.02, "step": 151260 }, { "epoch": 0.38, "learning_rate": 6.2200412703206e-05, "loss": 1.0187, "step": 151265 }, { "epoch": 0.38, "learning_rate": 6.21991544617243e-05, "loss": 1.02, "step": 151270 }, { "epoch": 0.38, "learning_rate": 6.219789622024259e-05, "loss": 1.0198, "step": 151275 }, { "epoch": 0.38, "learning_rate": 6.219663797876088e-05, "loss": 1.0217, "step": 151280 }, { "epoch": 0.38, "learning_rate": 6.219537973727917e-05, "loss": 1.02, "step": 151285 }, { "epoch": 0.38, "learning_rate": 6.219412149579748e-05, "loss": 1.0206, "step": 151290 }, { "epoch": 0.38, "learning_rate": 6.219286325431577e-05, "loss": 1.02, "step": 151295 }, { "epoch": 0.38, "learning_rate": 6.219160501283406e-05, "loss": 1.0192, "step": 151300 }, { "epoch": 0.38, "learning_rate": 6.219034677135235e-05, "loss": 1.0193, "step": 151305 }, { "epoch": 0.38, "learning_rate": 6.218908852987066e-05, "loss": 1.023, "step": 151310 }, { "epoch": 0.38, "learning_rate": 6.218783028838895e-05, "loss": 1.0198, "step": 151315 }, { "epoch": 0.38, "learning_rate": 6.218657204690724e-05, "loss": 1.0203, "step": 151320 }, { "epoch": 0.38, "learning_rate": 6.218531380542553e-05, "loss": 1.0195, "step": 151325 }, { "epoch": 0.38, "learning_rate": 6.218405556394383e-05, "loss": 1.0198, "step": 151330 }, { "epoch": 0.38, "learning_rate": 6.218279732246213e-05, "loss": 1.0215, "step": 151335 }, { "epoch": 0.38, "learning_rate": 6.218153908098042e-05, "loss": 1.0205, "step": 151340 }, { "epoch": 0.38, "learning_rate": 6.218028083949871e-05, "loss": 1.0222, "step": 151345 }, { "epoch": 0.38, "learning_rate": 6.2179022598017e-05, "loss": 1.0268, "step": 151350 }, { "epoch": 0.38, "learning_rate": 6.217776435653531e-05, "loss": 1.0207, "step": 151355 }, { "epoch": 0.38, "learning_rate": 6.21765061150536e-05, "loss": 1.0198, "step": 151360 }, { "epoch": 0.38, "learning_rate": 6.21752478735719e-05, "loss": 1.0199, "step": 151365 }, { "epoch": 0.38, "learning_rate": 6.217398963209019e-05, "loss": 1.0216, "step": 151370 }, { "epoch": 0.38, "learning_rate": 6.217273139060849e-05, "loss": 1.023, "step": 151375 }, { "epoch": 0.38, "learning_rate": 6.217147314912678e-05, "loss": 1.0202, "step": 151380 }, { "epoch": 0.38, "learning_rate": 6.217021490764507e-05, "loss": 1.0194, "step": 151385 }, { "epoch": 0.38, "learning_rate": 6.216895666616337e-05, "loss": 1.0198, "step": 151390 }, { "epoch": 0.38, "learning_rate": 6.216769842468166e-05, "loss": 1.0215, "step": 151395 }, { "epoch": 0.38, "learning_rate": 6.216644018319996e-05, "loss": 1.0199, "step": 151400 }, { "epoch": 0.38, "learning_rate": 6.216518194171825e-05, "loss": 1.0198, "step": 151405 }, { "epoch": 0.38, "learning_rate": 6.216392370023655e-05, "loss": 1.0204, "step": 151410 }, { "epoch": 0.38, "learning_rate": 6.216266545875484e-05, "loss": 1.0216, "step": 151415 }, { "epoch": 0.38, "learning_rate": 6.216140721727314e-05, "loss": 1.0198, "step": 151420 }, { "epoch": 0.38, "learning_rate": 6.216014897579143e-05, "loss": 1.018, "step": 151425 }, { "epoch": 0.38, "learning_rate": 6.215889073430973e-05, "loss": 1.0215, "step": 151430 }, { "epoch": 0.38, "learning_rate": 6.215763249282802e-05, "loss": 1.0196, "step": 151435 }, { "epoch": 0.38, "learning_rate": 6.215637425134632e-05, "loss": 1.019, "step": 151440 }, { "epoch": 0.38, "learning_rate": 6.215511600986461e-05, "loss": 1.0188, "step": 151445 }, { "epoch": 0.38, "learning_rate": 6.21538577683829e-05, "loss": 1.0221, "step": 151450 }, { "epoch": 0.38, "learning_rate": 6.21525995269012e-05, "loss": 1.0215, "step": 151455 }, { "epoch": 0.38, "learning_rate": 6.215134128541949e-05, "loss": 1.0211, "step": 151460 }, { "epoch": 0.38, "learning_rate": 6.21500830439378e-05, "loss": 1.0219, "step": 151465 }, { "epoch": 0.38, "learning_rate": 6.214882480245609e-05, "loss": 1.0209, "step": 151470 }, { "epoch": 0.38, "learning_rate": 6.214756656097439e-05, "loss": 1.0217, "step": 151475 }, { "epoch": 0.38, "learning_rate": 6.214630831949268e-05, "loss": 1.0194, "step": 151480 }, { "epoch": 0.38, "learning_rate": 6.214505007801097e-05, "loss": 1.0207, "step": 151485 }, { "epoch": 0.38, "learning_rate": 6.214379183652928e-05, "loss": 1.0208, "step": 151490 }, { "epoch": 0.38, "learning_rate": 6.214253359504757e-05, "loss": 1.0191, "step": 151495 }, { "epoch": 0.38, "learning_rate": 6.214127535356586e-05, "loss": 1.023, "step": 151500 }, { "epoch": 0.38, "learning_rate": 6.214001711208415e-05, "loss": 1.0195, "step": 151505 }, { "epoch": 0.38, "learning_rate": 6.213875887060246e-05, "loss": 1.0209, "step": 151510 }, { "epoch": 0.38, "learning_rate": 6.213750062912075e-05, "loss": 1.0202, "step": 151515 }, { "epoch": 0.38, "learning_rate": 6.213624238763904e-05, "loss": 1.0211, "step": 151520 }, { "epoch": 0.38, "learning_rate": 6.213498414615733e-05, "loss": 1.0175, "step": 151525 }, { "epoch": 0.38, "learning_rate": 6.213372590467563e-05, "loss": 1.0198, "step": 151530 }, { "epoch": 0.38, "learning_rate": 6.213246766319393e-05, "loss": 1.0194, "step": 151535 }, { "epoch": 0.38, "learning_rate": 6.213120942171222e-05, "loss": 1.0222, "step": 151540 }, { "epoch": 0.38, "learning_rate": 6.212995118023051e-05, "loss": 1.0195, "step": 151545 }, { "epoch": 0.38, "learning_rate": 6.21286929387488e-05, "loss": 1.0195, "step": 151550 }, { "epoch": 0.38, "learning_rate": 6.212743469726711e-05, "loss": 1.0437, "step": 151555 }, { "epoch": 0.38, "learning_rate": 6.21261764557854e-05, "loss": 1.0193, "step": 151560 }, { "epoch": 0.38, "learning_rate": 6.21249182143037e-05, "loss": 1.0222, "step": 151565 }, { "epoch": 0.38, "learning_rate": 6.212365997282199e-05, "loss": 1.0225, "step": 151570 }, { "epoch": 0.38, "learning_rate": 6.212240173134029e-05, "loss": 1.0219, "step": 151575 }, { "epoch": 0.38, "learning_rate": 6.212114348985858e-05, "loss": 1.0234, "step": 151580 }, { "epoch": 0.38, "learning_rate": 6.211988524837687e-05, "loss": 1.0202, "step": 151585 }, { "epoch": 0.38, "learning_rate": 6.211862700689517e-05, "loss": 1.0208, "step": 151590 }, { "epoch": 0.38, "learning_rate": 6.211736876541346e-05, "loss": 1.0231, "step": 151595 }, { "epoch": 0.38, "learning_rate": 6.211611052393176e-05, "loss": 1.0205, "step": 151600 }, { "epoch": 0.38, "learning_rate": 6.211485228245005e-05, "loss": 1.019, "step": 151605 }, { "epoch": 0.38, "learning_rate": 6.211359404096835e-05, "loss": 1.0224, "step": 151610 }, { "epoch": 0.38, "learning_rate": 6.211233579948664e-05, "loss": 1.0197, "step": 151615 }, { "epoch": 0.38, "learning_rate": 6.211107755800494e-05, "loss": 1.022, "step": 151620 }, { "epoch": 0.38, "learning_rate": 6.210981931652323e-05, "loss": 1.0217, "step": 151625 }, { "epoch": 0.38, "learning_rate": 6.210856107504153e-05, "loss": 1.0213, "step": 151630 }, { "epoch": 0.38, "learning_rate": 6.210730283355982e-05, "loss": 1.0193, "step": 151635 }, { "epoch": 0.38, "learning_rate": 6.210604459207812e-05, "loss": 1.0208, "step": 151640 }, { "epoch": 0.38, "learning_rate": 6.210478635059641e-05, "loss": 1.02, "step": 151645 }, { "epoch": 0.38, "learning_rate": 6.21035281091147e-05, "loss": 1.0202, "step": 151650 }, { "epoch": 0.38, "learning_rate": 6.2102269867633e-05, "loss": 1.0208, "step": 151655 }, { "epoch": 0.38, "learning_rate": 6.210101162615129e-05, "loss": 1.022, "step": 151660 }, { "epoch": 0.38, "learning_rate": 6.20997533846696e-05, "loss": 1.018, "step": 151665 }, { "epoch": 0.38, "learning_rate": 6.209849514318789e-05, "loss": 1.0198, "step": 151670 }, { "epoch": 0.38, "learning_rate": 6.209723690170618e-05, "loss": 1.0216, "step": 151675 }, { "epoch": 0.38, "learning_rate": 6.209597866022447e-05, "loss": 1.0193, "step": 151680 }, { "epoch": 0.38, "learning_rate": 6.209472041874277e-05, "loss": 1.0191, "step": 151685 }, { "epoch": 0.38, "learning_rate": 6.209346217726107e-05, "loss": 1.0208, "step": 151690 }, { "epoch": 0.38, "learning_rate": 6.209220393577936e-05, "loss": 1.0198, "step": 151695 }, { "epoch": 0.38, "learning_rate": 6.209094569429765e-05, "loss": 1.0194, "step": 151700 }, { "epoch": 0.38, "learning_rate": 6.208968745281595e-05, "loss": 1.0194, "step": 151705 }, { "epoch": 0.38, "learning_rate": 6.208842921133425e-05, "loss": 1.0223, "step": 151710 }, { "epoch": 0.38, "learning_rate": 6.208717096985254e-05, "loss": 1.0211, "step": 151715 }, { "epoch": 0.38, "learning_rate": 6.208591272837083e-05, "loss": 1.0189, "step": 151720 }, { "epoch": 0.38, "learning_rate": 6.208465448688912e-05, "loss": 1.0199, "step": 151725 }, { "epoch": 0.38, "learning_rate": 6.208339624540743e-05, "loss": 1.0209, "step": 151730 }, { "epoch": 0.38, "learning_rate": 6.208213800392572e-05, "loss": 1.0205, "step": 151735 }, { "epoch": 0.38, "learning_rate": 6.208087976244401e-05, "loss": 1.0185, "step": 151740 }, { "epoch": 0.38, "learning_rate": 6.20796215209623e-05, "loss": 1.0255, "step": 151745 }, { "epoch": 0.38, "learning_rate": 6.20783632794806e-05, "loss": 1.0442, "step": 151750 }, { "epoch": 0.38, "learning_rate": 6.20771050379989e-05, "loss": 1.0203, "step": 151755 }, { "epoch": 0.38, "learning_rate": 6.207584679651719e-05, "loss": 1.0199, "step": 151760 }, { "epoch": 0.38, "learning_rate": 6.207458855503548e-05, "loss": 1.0191, "step": 151765 }, { "epoch": 0.38, "learning_rate": 6.207333031355379e-05, "loss": 1.022, "step": 151770 }, { "epoch": 0.38, "learning_rate": 6.207207207207208e-05, "loss": 1.02, "step": 151775 }, { "epoch": 0.38, "learning_rate": 6.207081383059037e-05, "loss": 1.0223, "step": 151780 }, { "epoch": 0.38, "learning_rate": 6.206955558910866e-05, "loss": 1.0222, "step": 151785 }, { "epoch": 0.38, "learning_rate": 6.206829734762695e-05, "loss": 1.0219, "step": 151790 }, { "epoch": 0.38, "learning_rate": 6.206703910614526e-05, "loss": 1.0228, "step": 151795 }, { "epoch": 0.38, "learning_rate": 6.206578086466355e-05, "loss": 1.0209, "step": 151800 }, { "epoch": 0.38, "learning_rate": 6.206452262318184e-05, "loss": 1.0227, "step": 151805 }, { "epoch": 0.38, "learning_rate": 6.206326438170013e-05, "loss": 1.0227, "step": 151810 }, { "epoch": 0.38, "learning_rate": 6.206200614021844e-05, "loss": 1.0184, "step": 151815 }, { "epoch": 0.38, "learning_rate": 6.206074789873673e-05, "loss": 1.0218, "step": 151820 }, { "epoch": 0.38, "learning_rate": 6.205948965725502e-05, "loss": 1.02, "step": 151825 }, { "epoch": 0.38, "learning_rate": 6.205823141577331e-05, "loss": 1.0198, "step": 151830 }, { "epoch": 0.38, "learning_rate": 6.205697317429162e-05, "loss": 1.0188, "step": 151835 }, { "epoch": 0.38, "learning_rate": 6.205571493280991e-05, "loss": 1.0198, "step": 151840 }, { "epoch": 0.38, "learning_rate": 6.20544566913282e-05, "loss": 1.0224, "step": 151845 }, { "epoch": 0.38, "learning_rate": 6.205319844984649e-05, "loss": 1.0204, "step": 151850 }, { "epoch": 0.38, "learning_rate": 6.205194020836478e-05, "loss": 1.0176, "step": 151855 }, { "epoch": 0.38, "learning_rate": 6.205068196688309e-05, "loss": 1.0199, "step": 151860 }, { "epoch": 0.38, "learning_rate": 6.204942372540138e-05, "loss": 1.0189, "step": 151865 }, { "epoch": 0.38, "learning_rate": 6.204816548391967e-05, "loss": 1.0194, "step": 151870 }, { "epoch": 0.38, "learning_rate": 6.204690724243796e-05, "loss": 1.0198, "step": 151875 }, { "epoch": 0.38, "learning_rate": 6.204564900095627e-05, "loss": 1.0208, "step": 151880 }, { "epoch": 0.38, "learning_rate": 6.204439075947456e-05, "loss": 1.0198, "step": 151885 }, { "epoch": 0.38, "learning_rate": 6.204313251799285e-05, "loss": 1.0189, "step": 151890 }, { "epoch": 0.38, "learning_rate": 6.204187427651114e-05, "loss": 1.0194, "step": 151895 }, { "epoch": 0.38, "learning_rate": 6.204061603502945e-05, "loss": 1.0199, "step": 151900 }, { "epoch": 0.38, "learning_rate": 6.203935779354774e-05, "loss": 1.0205, "step": 151905 }, { "epoch": 0.38, "learning_rate": 6.203809955206603e-05, "loss": 1.0191, "step": 151910 }, { "epoch": 0.38, "learning_rate": 6.203684131058432e-05, "loss": 1.0183, "step": 151915 }, { "epoch": 0.38, "learning_rate": 6.203558306910261e-05, "loss": 1.0209, "step": 151920 }, { "epoch": 0.38, "learning_rate": 6.203432482762092e-05, "loss": 1.0215, "step": 151925 }, { "epoch": 0.38, "learning_rate": 6.203306658613921e-05, "loss": 1.0382, "step": 151930 }, { "epoch": 0.38, "learning_rate": 6.20318083446575e-05, "loss": 1.0216, "step": 151935 }, { "epoch": 0.38, "learning_rate": 6.20305501031758e-05, "loss": 1.0207, "step": 151940 }, { "epoch": 0.38, "learning_rate": 6.20292918616941e-05, "loss": 1.0212, "step": 151945 }, { "epoch": 0.38, "learning_rate": 6.202803362021239e-05, "loss": 1.0195, "step": 151950 }, { "epoch": 0.38, "learning_rate": 6.202677537873068e-05, "loss": 1.0211, "step": 151955 }, { "epoch": 0.38, "learning_rate": 6.202551713724897e-05, "loss": 1.0187, "step": 151960 }, { "epoch": 0.38, "learning_rate": 6.202425889576728e-05, "loss": 1.0415, "step": 151965 }, { "epoch": 0.38, "learning_rate": 6.202300065428557e-05, "loss": 1.0196, "step": 151970 }, { "epoch": 0.38, "learning_rate": 6.202174241280388e-05, "loss": 1.0202, "step": 151975 }, { "epoch": 0.38, "learning_rate": 6.202048417132217e-05, "loss": 1.0409, "step": 151980 }, { "epoch": 0.38, "learning_rate": 6.201922592984046e-05, "loss": 1.0197, "step": 151985 }, { "epoch": 0.38, "learning_rate": 6.201796768835876e-05, "loss": 1.0237, "step": 151990 }, { "epoch": 0.38, "learning_rate": 6.201670944687706e-05, "loss": 1.0195, "step": 151995 }, { "epoch": 0.38, "learning_rate": 6.201545120539535e-05, "loss": 1.0206, "step": 152000 }, { "epoch": 0.38, "learning_rate": 6.201419296391364e-05, "loss": 1.0184, "step": 152005 }, { "epoch": 0.38, "learning_rate": 6.201293472243193e-05, "loss": 1.0197, "step": 152010 }, { "epoch": 0.38, "learning_rate": 6.201167648095024e-05, "loss": 1.0208, "step": 152015 }, { "epoch": 0.38, "learning_rate": 6.201041823946853e-05, "loss": 1.0189, "step": 152020 }, { "epoch": 0.38, "learning_rate": 6.200915999798682e-05, "loss": 1.0191, "step": 152025 }, { "epoch": 0.38, "learning_rate": 6.200790175650511e-05, "loss": 1.0208, "step": 152030 }, { "epoch": 0.38, "learning_rate": 6.200664351502342e-05, "loss": 1.0189, "step": 152035 }, { "epoch": 0.38, "learning_rate": 6.200538527354171e-05, "loss": 1.0215, "step": 152040 }, { "epoch": 0.38, "learning_rate": 6.200412703206e-05, "loss": 1.0179, "step": 152045 }, { "epoch": 0.38, "learning_rate": 6.200286879057829e-05, "loss": 1.0198, "step": 152050 }, { "epoch": 0.38, "learning_rate": 6.20016105490966e-05, "loss": 1.022, "step": 152055 }, { "epoch": 0.38, "learning_rate": 6.200035230761489e-05, "loss": 1.0234, "step": 152060 }, { "epoch": 0.38, "learning_rate": 6.199909406613318e-05, "loss": 1.0219, "step": 152065 }, { "epoch": 0.38, "learning_rate": 6.199783582465147e-05, "loss": 1.0198, "step": 152070 }, { "epoch": 0.38, "learning_rate": 6.199657758316976e-05, "loss": 1.0203, "step": 152075 }, { "epoch": 0.38, "learning_rate": 6.199531934168807e-05, "loss": 1.0294, "step": 152080 }, { "epoch": 0.38, "learning_rate": 6.199406110020636e-05, "loss": 1.0205, "step": 152085 }, { "epoch": 0.38, "learning_rate": 6.199280285872465e-05, "loss": 1.0214, "step": 152090 }, { "epoch": 0.38, "learning_rate": 6.199154461724294e-05, "loss": 1.0222, "step": 152095 }, { "epoch": 0.38, "learning_rate": 6.199028637576125e-05, "loss": 1.02, "step": 152100 }, { "epoch": 0.38, "learning_rate": 6.198902813427954e-05, "loss": 1.0187, "step": 152105 }, { "epoch": 0.38, "learning_rate": 6.198776989279783e-05, "loss": 1.0192, "step": 152110 }, { "epoch": 0.38, "learning_rate": 6.198651165131612e-05, "loss": 1.0201, "step": 152115 }, { "epoch": 0.38, "learning_rate": 6.198525340983443e-05, "loss": 1.0192, "step": 152120 }, { "epoch": 0.38, "learning_rate": 6.198399516835272e-05, "loss": 1.0402, "step": 152125 }, { "epoch": 0.38, "learning_rate": 6.198273692687101e-05, "loss": 1.0199, "step": 152130 }, { "epoch": 0.38, "learning_rate": 6.19814786853893e-05, "loss": 1.0197, "step": 152135 }, { "epoch": 0.38, "learning_rate": 6.19802204439076e-05, "loss": 1.0197, "step": 152140 }, { "epoch": 0.38, "learning_rate": 6.19789622024259e-05, "loss": 1.0218, "step": 152145 }, { "epoch": 0.38, "learning_rate": 6.197770396094419e-05, "loss": 1.0212, "step": 152150 }, { "epoch": 0.38, "learning_rate": 6.197644571946248e-05, "loss": 1.0222, "step": 152155 }, { "epoch": 0.38, "learning_rate": 6.197518747798077e-05, "loss": 1.0203, "step": 152160 }, { "epoch": 0.38, "learning_rate": 6.197392923649908e-05, "loss": 1.021, "step": 152165 }, { "epoch": 0.38, "learning_rate": 6.197267099501737e-05, "loss": 1.0211, "step": 152170 }, { "epoch": 0.38, "learning_rate": 6.197141275353566e-05, "loss": 1.0205, "step": 152175 }, { "epoch": 0.38, "learning_rate": 6.197015451205395e-05, "loss": 1.0205, "step": 152180 }, { "epoch": 0.38, "learning_rate": 6.196889627057225e-05, "loss": 1.0198, "step": 152185 }, { "epoch": 0.38, "learning_rate": 6.196763802909055e-05, "loss": 1.0201, "step": 152190 }, { "epoch": 0.38, "learning_rate": 6.196637978760884e-05, "loss": 1.0197, "step": 152195 }, { "epoch": 0.38, "learning_rate": 6.196512154612713e-05, "loss": 1.0201, "step": 152200 }, { "epoch": 0.38, "learning_rate": 6.196386330464543e-05, "loss": 1.0246, "step": 152205 }, { "epoch": 0.38, "learning_rate": 6.196260506316373e-05, "loss": 1.0191, "step": 152210 }, { "epoch": 0.38, "learning_rate": 6.196134682168202e-05, "loss": 1.0206, "step": 152215 }, { "epoch": 0.38, "learning_rate": 6.196008858020031e-05, "loss": 1.0205, "step": 152220 }, { "epoch": 0.38, "learning_rate": 6.19588303387186e-05, "loss": 1.0203, "step": 152225 }, { "epoch": 0.38, "learning_rate": 6.195757209723691e-05, "loss": 1.0202, "step": 152230 }, { "epoch": 0.38, "learning_rate": 6.19563138557552e-05, "loss": 1.0216, "step": 152235 }, { "epoch": 0.38, "learning_rate": 6.19550556142735e-05, "loss": 1.0214, "step": 152240 }, { "epoch": 0.38, "learning_rate": 6.195379737279179e-05, "loss": 1.019, "step": 152245 }, { "epoch": 0.38, "learning_rate": 6.195253913131008e-05, "loss": 1.0198, "step": 152250 }, { "epoch": 0.38, "learning_rate": 6.195128088982838e-05, "loss": 1.019, "step": 152255 }, { "epoch": 0.38, "learning_rate": 6.195002264834667e-05, "loss": 1.0193, "step": 152260 }, { "epoch": 0.38, "learning_rate": 6.194876440686497e-05, "loss": 1.0232, "step": 152265 }, { "epoch": 0.38, "learning_rate": 6.194750616538326e-05, "loss": 1.0211, "step": 152270 }, { "epoch": 0.38, "learning_rate": 6.194624792390156e-05, "loss": 1.0214, "step": 152275 }, { "epoch": 0.38, "learning_rate": 6.194498968241985e-05, "loss": 1.02, "step": 152280 }, { "epoch": 0.38, "learning_rate": 6.194373144093814e-05, "loss": 1.0206, "step": 152285 }, { "epoch": 0.38, "learning_rate": 6.194247319945644e-05, "loss": 1.0202, "step": 152290 }, { "epoch": 0.38, "learning_rate": 6.194121495797474e-05, "loss": 1.0218, "step": 152295 }, { "epoch": 0.38, "learning_rate": 6.193995671649303e-05, "loss": 1.0183, "step": 152300 }, { "epoch": 0.38, "learning_rate": 6.193869847501132e-05, "loss": 1.0211, "step": 152305 }, { "epoch": 0.38, "learning_rate": 6.193744023352962e-05, "loss": 1.02, "step": 152310 }, { "epoch": 0.38, "learning_rate": 6.193618199204791e-05, "loss": 1.0206, "step": 152315 }, { "epoch": 0.38, "learning_rate": 6.193492375056621e-05, "loss": 1.0222, "step": 152320 }, { "epoch": 0.38, "learning_rate": 6.19336655090845e-05, "loss": 1.0201, "step": 152325 }, { "epoch": 0.38, "learning_rate": 6.19324072676028e-05, "loss": 1.0208, "step": 152330 }, { "epoch": 0.38, "learning_rate": 6.193114902612109e-05, "loss": 1.023, "step": 152335 }, { "epoch": 0.38, "learning_rate": 6.192989078463939e-05, "loss": 1.0177, "step": 152340 }, { "epoch": 0.38, "learning_rate": 6.192863254315768e-05, "loss": 1.0422, "step": 152345 }, { "epoch": 0.38, "learning_rate": 6.192737430167598e-05, "loss": 1.0187, "step": 152350 }, { "epoch": 0.38, "learning_rate": 6.192611606019427e-05, "loss": 1.0216, "step": 152355 }, { "epoch": 0.38, "learning_rate": 6.192485781871257e-05, "loss": 1.0213, "step": 152360 }, { "epoch": 0.38, "learning_rate": 6.192359957723086e-05, "loss": 1.0217, "step": 152365 }, { "epoch": 0.38, "learning_rate": 6.192234133574916e-05, "loss": 1.02, "step": 152370 }, { "epoch": 0.38, "learning_rate": 6.192108309426745e-05, "loss": 1.0218, "step": 152375 }, { "epoch": 0.38, "learning_rate": 6.191982485278574e-05, "loss": 1.0197, "step": 152380 }, { "epoch": 0.38, "learning_rate": 6.191856661130404e-05, "loss": 1.0192, "step": 152385 }, { "epoch": 0.38, "learning_rate": 6.191730836982234e-05, "loss": 1.0201, "step": 152390 }, { "epoch": 0.38, "learning_rate": 6.191605012834063e-05, "loss": 1.021, "step": 152395 }, { "epoch": 0.38, "learning_rate": 6.191479188685892e-05, "loss": 1.0196, "step": 152400 }, { "epoch": 0.38, "learning_rate": 6.191353364537722e-05, "loss": 1.0208, "step": 152405 }, { "epoch": 0.38, "learning_rate": 6.191227540389552e-05, "loss": 1.0208, "step": 152410 }, { "epoch": 0.38, "learning_rate": 6.191101716241381e-05, "loss": 1.0212, "step": 152415 }, { "epoch": 0.38, "learning_rate": 6.19097589209321e-05, "loss": 1.0188, "step": 152420 }, { "epoch": 0.38, "learning_rate": 6.19085006794504e-05, "loss": 1.0213, "step": 152425 }, { "epoch": 0.38, "learning_rate": 6.19072424379687e-05, "loss": 1.0205, "step": 152430 }, { "epoch": 0.38, "learning_rate": 6.190598419648699e-05, "loss": 1.0183, "step": 152435 }, { "epoch": 0.38, "learning_rate": 6.190472595500528e-05, "loss": 1.0206, "step": 152440 }, { "epoch": 0.38, "learning_rate": 6.190346771352357e-05, "loss": 1.0187, "step": 152445 }, { "epoch": 0.38, "learning_rate": 6.190220947204188e-05, "loss": 1.0189, "step": 152450 }, { "epoch": 0.38, "learning_rate": 6.190095123056017e-05, "loss": 1.0184, "step": 152455 }, { "epoch": 0.38, "learning_rate": 6.189969298907846e-05, "loss": 1.0208, "step": 152460 }, { "epoch": 0.38, "learning_rate": 6.189843474759675e-05, "loss": 1.0177, "step": 152465 }, { "epoch": 0.38, "learning_rate": 6.189717650611506e-05, "loss": 1.0195, "step": 152470 }, { "epoch": 0.38, "learning_rate": 6.189591826463336e-05, "loss": 1.0211, "step": 152475 }, { "epoch": 0.38, "learning_rate": 6.189466002315165e-05, "loss": 1.0181, "step": 152480 }, { "epoch": 0.38, "learning_rate": 6.189340178166994e-05, "loss": 1.0174, "step": 152485 }, { "epoch": 0.38, "learning_rate": 6.189214354018824e-05, "loss": 1.0185, "step": 152490 }, { "epoch": 0.38, "learning_rate": 6.189088529870654e-05, "loss": 1.021, "step": 152495 }, { "epoch": 0.38, "learning_rate": 6.188962705722483e-05, "loss": 1.0189, "step": 152500 }, { "epoch": 0.38, "learning_rate": 6.188836881574312e-05, "loss": 1.0225, "step": 152505 }, { "epoch": 0.38, "learning_rate": 6.188711057426142e-05, "loss": 1.0196, "step": 152510 }, { "epoch": 0.38, "learning_rate": 6.188585233277972e-05, "loss": 1.0178, "step": 152515 }, { "epoch": 0.38, "learning_rate": 6.188459409129801e-05, "loss": 1.0203, "step": 152520 }, { "epoch": 0.38, "learning_rate": 6.18833358498163e-05, "loss": 1.0209, "step": 152525 }, { "epoch": 0.38, "learning_rate": 6.18820776083346e-05, "loss": 1.0212, "step": 152530 }, { "epoch": 0.38, "learning_rate": 6.188081936685289e-05, "loss": 1.0195, "step": 152535 }, { "epoch": 0.38, "learning_rate": 6.187956112537119e-05, "loss": 1.0216, "step": 152540 }, { "epoch": 0.38, "learning_rate": 6.187830288388948e-05, "loss": 1.0209, "step": 152545 }, { "epoch": 0.38, "learning_rate": 6.187704464240778e-05, "loss": 1.0199, "step": 152550 }, { "epoch": 0.38, "learning_rate": 6.187578640092607e-05, "loss": 1.0211, "step": 152555 }, { "epoch": 0.38, "learning_rate": 6.187452815944437e-05, "loss": 1.02, "step": 152560 }, { "epoch": 0.38, "learning_rate": 6.187326991796266e-05, "loss": 1.0203, "step": 152565 }, { "epoch": 0.38, "learning_rate": 6.187201167648096e-05, "loss": 1.0203, "step": 152570 }, { "epoch": 0.38, "learning_rate": 6.187075343499925e-05, "loss": 1.0172, "step": 152575 }, { "epoch": 0.38, "learning_rate": 6.186949519351755e-05, "loss": 1.0204, "step": 152580 }, { "epoch": 0.38, "learning_rate": 6.186823695203584e-05, "loss": 1.0206, "step": 152585 }, { "epoch": 0.38, "learning_rate": 6.186697871055414e-05, "loss": 1.0198, "step": 152590 }, { "epoch": 0.38, "learning_rate": 6.186572046907243e-05, "loss": 1.0218, "step": 152595 }, { "epoch": 0.38, "learning_rate": 6.186446222759072e-05, "loss": 1.021, "step": 152600 }, { "epoch": 0.38, "learning_rate": 6.186320398610902e-05, "loss": 1.0199, "step": 152605 }, { "epoch": 0.38, "learning_rate": 6.186194574462732e-05, "loss": 1.0222, "step": 152610 }, { "epoch": 0.38, "learning_rate": 6.186068750314561e-05, "loss": 1.0198, "step": 152615 }, { "epoch": 0.38, "learning_rate": 6.18594292616639e-05, "loss": 1.0219, "step": 152620 }, { "epoch": 0.38, "learning_rate": 6.18581710201822e-05, "loss": 1.019, "step": 152625 }, { "epoch": 0.38, "learning_rate": 6.18569127787005e-05, "loss": 1.0176, "step": 152630 }, { "epoch": 0.38, "learning_rate": 6.185565453721879e-05, "loss": 1.0202, "step": 152635 }, { "epoch": 0.38, "learning_rate": 6.185439629573708e-05, "loss": 1.0177, "step": 152640 }, { "epoch": 0.38, "learning_rate": 6.185313805425538e-05, "loss": 1.0223, "step": 152645 }, { "epoch": 0.38, "learning_rate": 6.185187981277368e-05, "loss": 1.0198, "step": 152650 }, { "epoch": 0.38, "learning_rate": 6.185062157129197e-05, "loss": 1.0235, "step": 152655 }, { "epoch": 0.38, "learning_rate": 6.184936332981026e-05, "loss": 1.0214, "step": 152660 }, { "epoch": 0.38, "learning_rate": 6.184810508832855e-05, "loss": 1.0216, "step": 152665 }, { "epoch": 0.38, "learning_rate": 6.184684684684686e-05, "loss": 1.0217, "step": 152670 }, { "epoch": 0.38, "learning_rate": 6.184558860536515e-05, "loss": 1.0181, "step": 152675 }, { "epoch": 0.38, "learning_rate": 6.184433036388344e-05, "loss": 1.0197, "step": 152680 }, { "epoch": 0.38, "learning_rate": 6.184307212240173e-05, "loss": 1.042, "step": 152685 }, { "epoch": 0.38, "learning_rate": 6.184181388092004e-05, "loss": 1.0207, "step": 152690 }, { "epoch": 0.38, "learning_rate": 6.184055563943833e-05, "loss": 1.0193, "step": 152695 }, { "epoch": 0.38, "learning_rate": 6.183929739795662e-05, "loss": 1.0181, "step": 152700 }, { "epoch": 0.38, "learning_rate": 6.183803915647491e-05, "loss": 1.0196, "step": 152705 }, { "epoch": 0.38, "learning_rate": 6.183678091499322e-05, "loss": 1.0227, "step": 152710 }, { "epoch": 0.38, "learning_rate": 6.183552267351151e-05, "loss": 1.0184, "step": 152715 }, { "epoch": 0.38, "learning_rate": 6.18342644320298e-05, "loss": 1.0202, "step": 152720 }, { "epoch": 0.38, "learning_rate": 6.183300619054809e-05, "loss": 1.0188, "step": 152725 }, { "epoch": 0.38, "learning_rate": 6.183174794906638e-05, "loss": 1.0228, "step": 152730 }, { "epoch": 0.38, "learning_rate": 6.183048970758469e-05, "loss": 1.0195, "step": 152735 }, { "epoch": 0.38, "learning_rate": 6.182923146610298e-05, "loss": 1.0436, "step": 152740 }, { "epoch": 0.38, "learning_rate": 6.182797322462127e-05, "loss": 1.0215, "step": 152745 }, { "epoch": 0.38, "learning_rate": 6.182671498313956e-05, "loss": 1.0226, "step": 152750 }, { "epoch": 0.38, "learning_rate": 6.182545674165787e-05, "loss": 1.0213, "step": 152755 }, { "epoch": 0.38, "learning_rate": 6.182419850017616e-05, "loss": 1.0208, "step": 152760 }, { "epoch": 0.38, "learning_rate": 6.182294025869445e-05, "loss": 1.0217, "step": 152765 }, { "epoch": 0.38, "learning_rate": 6.182168201721274e-05, "loss": 1.0191, "step": 152770 }, { "epoch": 0.38, "learning_rate": 6.182042377573105e-05, "loss": 1.0202, "step": 152775 }, { "epoch": 0.38, "learning_rate": 6.181916553424934e-05, "loss": 1.0207, "step": 152780 }, { "epoch": 0.38, "learning_rate": 6.181790729276763e-05, "loss": 1.0189, "step": 152785 }, { "epoch": 0.38, "learning_rate": 6.181664905128592e-05, "loss": 1.0211, "step": 152790 }, { "epoch": 0.38, "learning_rate": 6.181539080980421e-05, "loss": 1.0187, "step": 152795 }, { "epoch": 0.38, "learning_rate": 6.181413256832252e-05, "loss": 1.0194, "step": 152800 }, { "epoch": 0.38, "learning_rate": 6.181287432684081e-05, "loss": 1.0206, "step": 152805 }, { "epoch": 0.38, "learning_rate": 6.18116160853591e-05, "loss": 1.0195, "step": 152810 }, { "epoch": 0.38, "learning_rate": 6.181035784387739e-05, "loss": 1.0206, "step": 152815 }, { "epoch": 0.38, "learning_rate": 6.18090996023957e-05, "loss": 1.0222, "step": 152820 }, { "epoch": 0.38, "learning_rate": 6.180784136091399e-05, "loss": 1.0185, "step": 152825 }, { "epoch": 0.38, "learning_rate": 6.180658311943228e-05, "loss": 1.02, "step": 152830 }, { "epoch": 0.38, "learning_rate": 6.180532487795057e-05, "loss": 1.0213, "step": 152835 }, { "epoch": 0.38, "learning_rate": 6.180406663646886e-05, "loss": 1.0206, "step": 152840 }, { "epoch": 0.38, "learning_rate": 6.180280839498717e-05, "loss": 1.0192, "step": 152845 }, { "epoch": 0.38, "learning_rate": 6.180155015350546e-05, "loss": 1.0188, "step": 152850 }, { "epoch": 0.38, "learning_rate": 6.180029191202375e-05, "loss": 1.021, "step": 152855 }, { "epoch": 0.38, "learning_rate": 6.179903367054204e-05, "loss": 1.0195, "step": 152860 }, { "epoch": 0.38, "learning_rate": 6.179777542906035e-05, "loss": 1.021, "step": 152865 }, { "epoch": 0.38, "learning_rate": 6.179651718757864e-05, "loss": 1.023, "step": 152870 }, { "epoch": 0.38, "learning_rate": 6.179525894609693e-05, "loss": 1.0203, "step": 152875 }, { "epoch": 0.38, "learning_rate": 6.179400070461522e-05, "loss": 1.0195, "step": 152880 }, { "epoch": 0.38, "learning_rate": 6.179274246313353e-05, "loss": 1.0189, "step": 152885 }, { "epoch": 0.38, "learning_rate": 6.179148422165182e-05, "loss": 1.0212, "step": 152890 }, { "epoch": 0.38, "learning_rate": 6.179022598017011e-05, "loss": 1.0187, "step": 152895 }, { "epoch": 0.38, "learning_rate": 6.17889677386884e-05, "loss": 1.0201, "step": 152900 }, { "epoch": 0.38, "learning_rate": 6.17877094972067e-05, "loss": 1.0235, "step": 152905 }, { "epoch": 0.38, "learning_rate": 6.1786451255725e-05, "loss": 1.0184, "step": 152910 }, { "epoch": 0.38, "learning_rate": 6.178519301424329e-05, "loss": 1.0213, "step": 152915 }, { "epoch": 0.38, "learning_rate": 6.178393477276158e-05, "loss": 1.021, "step": 152920 }, { "epoch": 0.38, "learning_rate": 6.178267653127988e-05, "loss": 1.0218, "step": 152925 }, { "epoch": 0.38, "learning_rate": 6.178141828979818e-05, "loss": 1.0197, "step": 152930 }, { "epoch": 0.38, "learning_rate": 6.178016004831647e-05, "loss": 1.0218, "step": 152935 }, { "epoch": 0.38, "learning_rate": 6.177890180683476e-05, "loss": 1.0194, "step": 152940 }, { "epoch": 0.38, "learning_rate": 6.177764356535306e-05, "loss": 1.0201, "step": 152945 }, { "epoch": 0.38, "learning_rate": 6.177638532387136e-05, "loss": 1.0204, "step": 152950 }, { "epoch": 0.38, "learning_rate": 6.177512708238965e-05, "loss": 1.0221, "step": 152955 }, { "epoch": 0.38, "learning_rate": 6.177386884090794e-05, "loss": 1.0208, "step": 152960 }, { "epoch": 0.38, "learning_rate": 6.177261059942624e-05, "loss": 1.0173, "step": 152965 }, { "epoch": 0.38, "learning_rate": 6.177135235794454e-05, "loss": 1.0166, "step": 152970 }, { "epoch": 0.38, "learning_rate": 6.177009411646285e-05, "loss": 1.0177, "step": 152975 }, { "epoch": 0.38, "learning_rate": 6.176883587498114e-05, "loss": 1.0201, "step": 152980 }, { "epoch": 0.38, "learning_rate": 6.176757763349943e-05, "loss": 1.0184, "step": 152985 }, { "epoch": 0.38, "learning_rate": 6.176631939201772e-05, "loss": 1.0428, "step": 152990 }, { "epoch": 0.38, "learning_rate": 6.176506115053601e-05, "loss": 1.0197, "step": 152995 }, { "epoch": 0.38, "learning_rate": 6.176380290905432e-05, "loss": 1.019, "step": 153000 }, { "epoch": 0.38, "learning_rate": 6.176254466757261e-05, "loss": 1.0185, "step": 153005 }, { "epoch": 0.38, "learning_rate": 6.17612864260909e-05, "loss": 1.0164, "step": 153010 }, { "epoch": 0.38, "learning_rate": 6.176002818460919e-05, "loss": 1.0225, "step": 153015 }, { "epoch": 0.38, "learning_rate": 6.17587699431275e-05, "loss": 1.0211, "step": 153020 }, { "epoch": 0.38, "learning_rate": 6.175751170164579e-05, "loss": 1.0247, "step": 153025 }, { "epoch": 0.38, "learning_rate": 6.175625346016408e-05, "loss": 1.019, "step": 153030 }, { "epoch": 0.38, "learning_rate": 6.175499521868237e-05, "loss": 1.0214, "step": 153035 }, { "epoch": 0.38, "learning_rate": 6.175373697720068e-05, "loss": 1.0192, "step": 153040 }, { "epoch": 0.38, "learning_rate": 6.175247873571897e-05, "loss": 1.0201, "step": 153045 }, { "epoch": 0.38, "learning_rate": 6.175122049423726e-05, "loss": 1.0184, "step": 153050 }, { "epoch": 0.38, "learning_rate": 6.174996225275555e-05, "loss": 1.0207, "step": 153055 }, { "epoch": 0.38, "learning_rate": 6.174870401127384e-05, "loss": 1.0197, "step": 153060 }, { "epoch": 0.38, "learning_rate": 6.174744576979215e-05, "loss": 1.0228, "step": 153065 }, { "epoch": 0.38, "learning_rate": 6.174618752831044e-05, "loss": 1.0198, "step": 153070 }, { "epoch": 0.38, "learning_rate": 6.174492928682873e-05, "loss": 1.0427, "step": 153075 }, { "epoch": 0.38, "learning_rate": 6.174367104534702e-05, "loss": 1.021, "step": 153080 }, { "epoch": 0.38, "learning_rate": 6.174241280386533e-05, "loss": 1.0221, "step": 153085 }, { "epoch": 0.38, "learning_rate": 6.174115456238362e-05, "loss": 1.0203, "step": 153090 }, { "epoch": 0.38, "learning_rate": 6.173989632090191e-05, "loss": 1.0205, "step": 153095 }, { "epoch": 0.38, "learning_rate": 6.17386380794202e-05, "loss": 1.0208, "step": 153100 }, { "epoch": 0.38, "learning_rate": 6.173737983793851e-05, "loss": 1.0221, "step": 153105 }, { "epoch": 0.38, "learning_rate": 6.17361215964568e-05, "loss": 1.018, "step": 153110 }, { "epoch": 0.38, "learning_rate": 6.173486335497509e-05, "loss": 1.0179, "step": 153115 }, { "epoch": 0.38, "learning_rate": 6.173360511349338e-05, "loss": 1.02, "step": 153120 }, { "epoch": 0.38, "learning_rate": 6.173234687201168e-05, "loss": 1.0207, "step": 153125 }, { "epoch": 0.38, "learning_rate": 6.173108863052998e-05, "loss": 1.0188, "step": 153130 }, { "epoch": 0.38, "learning_rate": 6.172983038904827e-05, "loss": 1.0229, "step": 153135 }, { "epoch": 0.38, "learning_rate": 6.172857214756656e-05, "loss": 1.0229, "step": 153140 }, { "epoch": 0.38, "learning_rate": 6.172731390608486e-05, "loss": 1.0221, "step": 153145 }, { "epoch": 0.38, "learning_rate": 6.172605566460316e-05, "loss": 1.0215, "step": 153150 }, { "epoch": 0.38, "learning_rate": 6.172479742312145e-05, "loss": 1.0214, "step": 153155 }, { "epoch": 0.38, "learning_rate": 6.172353918163974e-05, "loss": 1.0204, "step": 153160 }, { "epoch": 0.38, "learning_rate": 6.172228094015804e-05, "loss": 1.0216, "step": 153165 }, { "epoch": 0.38, "learning_rate": 6.172102269867634e-05, "loss": 1.0185, "step": 153170 }, { "epoch": 0.38, "learning_rate": 6.171976445719463e-05, "loss": 1.017, "step": 153175 }, { "epoch": 0.38, "learning_rate": 6.171850621571292e-05, "loss": 1.0204, "step": 153180 }, { "epoch": 0.38, "learning_rate": 6.171724797423122e-05, "loss": 1.0204, "step": 153185 }, { "epoch": 0.38, "learning_rate": 6.171598973274951e-05, "loss": 1.02, "step": 153190 }, { "epoch": 0.38, "learning_rate": 6.171473149126781e-05, "loss": 1.0184, "step": 153195 }, { "epoch": 0.38, "learning_rate": 6.17134732497861e-05, "loss": 1.0187, "step": 153200 }, { "epoch": 0.38, "learning_rate": 6.17122150083044e-05, "loss": 1.0205, "step": 153205 }, { "epoch": 0.38, "learning_rate": 6.171095676682269e-05, "loss": 1.0184, "step": 153210 }, { "epoch": 0.38, "learning_rate": 6.170969852534099e-05, "loss": 1.0201, "step": 153215 }, { "epoch": 0.38, "learning_rate": 6.170844028385928e-05, "loss": 1.0221, "step": 153220 }, { "epoch": 0.38, "learning_rate": 6.170718204237758e-05, "loss": 1.0201, "step": 153225 }, { "epoch": 0.38, "learning_rate": 6.170592380089587e-05, "loss": 1.0201, "step": 153230 }, { "epoch": 0.38, "learning_rate": 6.170466555941417e-05, "loss": 1.0205, "step": 153235 }, { "epoch": 0.38, "learning_rate": 6.170340731793246e-05, "loss": 1.0205, "step": 153240 }, { "epoch": 0.38, "learning_rate": 6.170214907645076e-05, "loss": 1.0218, "step": 153245 }, { "epoch": 0.38, "learning_rate": 6.170089083496905e-05, "loss": 1.0203, "step": 153250 }, { "epoch": 0.38, "learning_rate": 6.169963259348734e-05, "loss": 1.0171, "step": 153255 }, { "epoch": 0.38, "learning_rate": 6.169837435200564e-05, "loss": 1.0192, "step": 153260 }, { "epoch": 0.38, "learning_rate": 6.169711611052394e-05, "loss": 1.0203, "step": 153265 }, { "epoch": 0.38, "learning_rate": 6.169585786904223e-05, "loss": 1.022, "step": 153270 }, { "epoch": 0.38, "learning_rate": 6.169459962756052e-05, "loss": 1.021, "step": 153275 }, { "epoch": 0.38, "learning_rate": 6.169334138607882e-05, "loss": 1.0205, "step": 153280 }, { "epoch": 0.38, "learning_rate": 6.169208314459711e-05, "loss": 1.0196, "step": 153285 }, { "epoch": 0.38, "learning_rate": 6.16908249031154e-05, "loss": 1.0179, "step": 153290 }, { "epoch": 0.38, "learning_rate": 6.16895666616337e-05, "loss": 1.018, "step": 153295 }, { "epoch": 0.38, "learning_rate": 6.1688308420152e-05, "loss": 1.0181, "step": 153300 }, { "epoch": 0.38, "learning_rate": 6.16870501786703e-05, "loss": 1.018, "step": 153305 }, { "epoch": 0.38, "learning_rate": 6.168579193718859e-05, "loss": 1.019, "step": 153310 }, { "epoch": 0.38, "learning_rate": 6.168453369570688e-05, "loss": 1.0204, "step": 153315 }, { "epoch": 0.38, "learning_rate": 6.168327545422517e-05, "loss": 1.0201, "step": 153320 }, { "epoch": 0.38, "learning_rate": 6.168201721274347e-05, "loss": 1.0199, "step": 153325 }, { "epoch": 0.38, "learning_rate": 6.168075897126177e-05, "loss": 1.0221, "step": 153330 }, { "epoch": 0.38, "learning_rate": 6.167950072978006e-05, "loss": 1.0186, "step": 153335 }, { "epoch": 0.38, "learning_rate": 6.167824248829835e-05, "loss": 1.0217, "step": 153340 }, { "epoch": 0.38, "learning_rate": 6.167698424681665e-05, "loss": 1.0201, "step": 153345 }, { "epoch": 0.38, "learning_rate": 6.167572600533495e-05, "loss": 1.0201, "step": 153350 }, { "epoch": 0.38, "learning_rate": 6.167446776385324e-05, "loss": 1.0204, "step": 153355 }, { "epoch": 0.38, "learning_rate": 6.167320952237153e-05, "loss": 1.0196, "step": 153360 }, { "epoch": 0.38, "learning_rate": 6.167195128088983e-05, "loss": 1.0233, "step": 153365 }, { "epoch": 0.38, "learning_rate": 6.167069303940813e-05, "loss": 1.0207, "step": 153370 }, { "epoch": 0.38, "learning_rate": 6.166943479792642e-05, "loss": 1.021, "step": 153375 }, { "epoch": 0.39, "learning_rate": 6.166817655644471e-05, "loss": 1.0197, "step": 153380 }, { "epoch": 0.39, "learning_rate": 6.1666918314963e-05, "loss": 1.0644, "step": 153385 }, { "epoch": 0.39, "learning_rate": 6.16656600734813e-05, "loss": 1.0195, "step": 153390 }, { "epoch": 0.39, "learning_rate": 6.16644018319996e-05, "loss": 1.0193, "step": 153395 }, { "epoch": 0.39, "learning_rate": 6.166314359051789e-05, "loss": 1.019, "step": 153400 }, { "epoch": 0.39, "learning_rate": 6.166188534903618e-05, "loss": 1.0195, "step": 153405 }, { "epoch": 0.39, "learning_rate": 6.166062710755449e-05, "loss": 1.0202, "step": 153410 }, { "epoch": 0.39, "learning_rate": 6.165936886607278e-05, "loss": 1.0175, "step": 153415 }, { "epoch": 0.39, "learning_rate": 6.165811062459107e-05, "loss": 1.0192, "step": 153420 }, { "epoch": 0.39, "learning_rate": 6.165685238310936e-05, "loss": 1.0207, "step": 153425 }, { "epoch": 0.39, "learning_rate": 6.165559414162767e-05, "loss": 1.0185, "step": 153430 }, { "epoch": 0.39, "learning_rate": 6.165433590014596e-05, "loss": 1.0208, "step": 153435 }, { "epoch": 0.39, "learning_rate": 6.165307765866425e-05, "loss": 1.0197, "step": 153440 }, { "epoch": 0.39, "learning_rate": 6.165181941718254e-05, "loss": 1.0209, "step": 153445 }, { "epoch": 0.39, "learning_rate": 6.165056117570083e-05, "loss": 1.0208, "step": 153450 }, { "epoch": 0.39, "learning_rate": 6.164930293421914e-05, "loss": 1.0168, "step": 153455 }, { "epoch": 0.39, "learning_rate": 6.164804469273743e-05, "loss": 1.0221, "step": 153460 }, { "epoch": 0.39, "learning_rate": 6.164678645125572e-05, "loss": 1.022, "step": 153465 }, { "epoch": 0.39, "learning_rate": 6.164552820977403e-05, "loss": 1.025, "step": 153470 }, { "epoch": 0.39, "learning_rate": 6.164426996829232e-05, "loss": 1.0209, "step": 153475 }, { "epoch": 0.39, "learning_rate": 6.164326337510695e-05, "loss": 1.0214, "step": 153480 }, { "epoch": 0.39, "learning_rate": 6.164200513362525e-05, "loss": 1.0193, "step": 153485 }, { "epoch": 0.39, "learning_rate": 6.164074689214354e-05, "loss": 1.0198, "step": 153490 }, { "epoch": 0.39, "learning_rate": 6.163948865066184e-05, "loss": 1.0196, "step": 153495 }, { "epoch": 0.39, "learning_rate": 6.163823040918013e-05, "loss": 1.0218, "step": 153500 }, { "epoch": 0.39, "learning_rate": 6.163697216769843e-05, "loss": 1.0408, "step": 153505 }, { "epoch": 0.39, "learning_rate": 6.163571392621672e-05, "loss": 1.0203, "step": 153510 }, { "epoch": 0.39, "learning_rate": 6.163445568473502e-05, "loss": 1.0207, "step": 153515 }, { "epoch": 0.39, "learning_rate": 6.163319744325331e-05, "loss": 1.0232, "step": 153520 }, { "epoch": 0.39, "learning_rate": 6.16319392017716e-05, "loss": 1.0199, "step": 153525 }, { "epoch": 0.39, "learning_rate": 6.16306809602899e-05, "loss": 1.019, "step": 153530 }, { "epoch": 0.39, "learning_rate": 6.162942271880819e-05, "loss": 1.0191, "step": 153535 }, { "epoch": 0.39, "learning_rate": 6.16281644773265e-05, "loss": 1.0226, "step": 153540 }, { "epoch": 0.39, "learning_rate": 6.162690623584479e-05, "loss": 1.0212, "step": 153545 }, { "epoch": 0.39, "learning_rate": 6.162564799436308e-05, "loss": 1.0219, "step": 153550 }, { "epoch": 0.39, "learning_rate": 6.162438975288137e-05, "loss": 1.0217, "step": 153555 }, { "epoch": 0.39, "learning_rate": 6.162313151139967e-05, "loss": 1.0183, "step": 153560 }, { "epoch": 0.39, "learning_rate": 6.162187326991797e-05, "loss": 1.0223, "step": 153565 }, { "epoch": 0.39, "learning_rate": 6.162061502843626e-05, "loss": 1.0217, "step": 153570 }, { "epoch": 0.39, "learning_rate": 6.161935678695455e-05, "loss": 1.0209, "step": 153575 }, { "epoch": 0.39, "learning_rate": 6.161809854547284e-05, "loss": 1.0176, "step": 153580 }, { "epoch": 0.39, "learning_rate": 6.161684030399114e-05, "loss": 1.0207, "step": 153585 }, { "epoch": 0.39, "learning_rate": 6.161558206250944e-05, "loss": 1.018, "step": 153590 }, { "epoch": 0.39, "learning_rate": 6.161432382102773e-05, "loss": 1.0178, "step": 153595 }, { "epoch": 0.39, "learning_rate": 6.161306557954602e-05, "loss": 1.0226, "step": 153600 }, { "epoch": 0.39, "learning_rate": 6.161180733806432e-05, "loss": 1.0191, "step": 153605 }, { "epoch": 0.39, "learning_rate": 6.161054909658262e-05, "loss": 1.0194, "step": 153610 }, { "epoch": 0.39, "learning_rate": 6.160929085510091e-05, "loss": 1.0419, "step": 153615 }, { "epoch": 0.39, "learning_rate": 6.16080326136192e-05, "loss": 1.0193, "step": 153620 }, { "epoch": 0.39, "learning_rate": 6.16067743721375e-05, "loss": 1.0172, "step": 153625 }, { "epoch": 0.39, "learning_rate": 6.16055161306558e-05, "loss": 1.0225, "step": 153630 }, { "epoch": 0.39, "learning_rate": 6.160425788917409e-05, "loss": 1.0194, "step": 153635 }, { "epoch": 0.39, "learning_rate": 6.160299964769238e-05, "loss": 1.02, "step": 153640 }, { "epoch": 0.39, "learning_rate": 6.160174140621067e-05, "loss": 1.0182, "step": 153645 }, { "epoch": 0.39, "learning_rate": 6.160048316472898e-05, "loss": 1.0209, "step": 153650 }, { "epoch": 0.39, "learning_rate": 6.159922492324727e-05, "loss": 1.0214, "step": 153655 }, { "epoch": 0.39, "learning_rate": 6.159796668176556e-05, "loss": 1.0198, "step": 153660 }, { "epoch": 0.39, "learning_rate": 6.159670844028385e-05, "loss": 1.0174, "step": 153665 }, { "epoch": 0.39, "learning_rate": 6.159545019880216e-05, "loss": 1.0203, "step": 153670 }, { "epoch": 0.39, "learning_rate": 6.159419195732046e-05, "loss": 1.0215, "step": 153675 }, { "epoch": 0.39, "learning_rate": 6.159293371583875e-05, "loss": 1.0219, "step": 153680 }, { "epoch": 0.39, "learning_rate": 6.159167547435704e-05, "loss": 1.0183, "step": 153685 }, { "epoch": 0.39, "learning_rate": 6.159041723287534e-05, "loss": 1.0187, "step": 153690 }, { "epoch": 0.39, "learning_rate": 6.158915899139364e-05, "loss": 1.0196, "step": 153695 }, { "epoch": 0.39, "learning_rate": 6.158790074991193e-05, "loss": 1.0215, "step": 153700 }, { "epoch": 0.39, "learning_rate": 6.158664250843022e-05, "loss": 1.0192, "step": 153705 }, { "epoch": 0.39, "learning_rate": 6.158538426694852e-05, "loss": 1.0202, "step": 153710 }, { "epoch": 0.39, "learning_rate": 6.158412602546682e-05, "loss": 1.0222, "step": 153715 }, { "epoch": 0.39, "learning_rate": 6.158286778398511e-05, "loss": 1.0227, "step": 153720 }, { "epoch": 0.39, "learning_rate": 6.15816095425034e-05, "loss": 1.0228, "step": 153725 }, { "epoch": 0.39, "learning_rate": 6.15803513010217e-05, "loss": 1.0217, "step": 153730 }, { "epoch": 0.39, "learning_rate": 6.157909305953999e-05, "loss": 1.0219, "step": 153735 }, { "epoch": 0.39, "learning_rate": 6.157783481805829e-05, "loss": 1.0197, "step": 153740 }, { "epoch": 0.39, "learning_rate": 6.157657657657658e-05, "loss": 1.0206, "step": 153745 }, { "epoch": 0.39, "learning_rate": 6.157531833509488e-05, "loss": 1.0227, "step": 153750 }, { "epoch": 0.39, "learning_rate": 6.157406009361317e-05, "loss": 1.0214, "step": 153755 }, { "epoch": 0.39, "learning_rate": 6.157280185213147e-05, "loss": 1.0179, "step": 153760 }, { "epoch": 0.39, "learning_rate": 6.157154361064976e-05, "loss": 1.0221, "step": 153765 }, { "epoch": 0.39, "learning_rate": 6.157028536916806e-05, "loss": 1.0201, "step": 153770 }, { "epoch": 0.39, "learning_rate": 6.156902712768635e-05, "loss": 1.0206, "step": 153775 }, { "epoch": 0.39, "learning_rate": 6.156776888620465e-05, "loss": 1.0204, "step": 153780 }, { "epoch": 0.39, "learning_rate": 6.156651064472294e-05, "loss": 1.0194, "step": 153785 }, { "epoch": 0.39, "learning_rate": 6.156525240324124e-05, "loss": 1.0208, "step": 153790 }, { "epoch": 0.39, "learning_rate": 6.156399416175953e-05, "loss": 1.0209, "step": 153795 }, { "epoch": 0.39, "learning_rate": 6.156273592027782e-05, "loss": 1.0189, "step": 153800 }, { "epoch": 0.39, "learning_rate": 6.156147767879612e-05, "loss": 1.0193, "step": 153805 }, { "epoch": 0.39, "learning_rate": 6.156021943731442e-05, "loss": 1.0368, "step": 153810 }, { "epoch": 0.39, "learning_rate": 6.155896119583271e-05, "loss": 1.0185, "step": 153815 }, { "epoch": 0.39, "learning_rate": 6.1557702954351e-05, "loss": 1.0191, "step": 153820 }, { "epoch": 0.39, "learning_rate": 6.15564447128693e-05, "loss": 1.02, "step": 153825 }, { "epoch": 0.39, "learning_rate": 6.15551864713876e-05, "loss": 1.0197, "step": 153830 }, { "epoch": 0.39, "learning_rate": 6.155392822990589e-05, "loss": 1.0198, "step": 153835 }, { "epoch": 0.39, "learning_rate": 6.155266998842418e-05, "loss": 1.0218, "step": 153840 }, { "epoch": 0.39, "learning_rate": 6.155141174694248e-05, "loss": 1.0217, "step": 153845 }, { "epoch": 0.39, "learning_rate": 6.155015350546078e-05, "loss": 1.0203, "step": 153850 }, { "epoch": 0.39, "learning_rate": 6.154889526397907e-05, "loss": 1.0198, "step": 153855 }, { "epoch": 0.39, "learning_rate": 6.154763702249736e-05, "loss": 1.0195, "step": 153860 }, { "epoch": 0.39, "learning_rate": 6.154637878101565e-05, "loss": 1.0197, "step": 153865 }, { "epoch": 0.39, "learning_rate": 6.154512053953396e-05, "loss": 1.0184, "step": 153870 }, { "epoch": 0.39, "learning_rate": 6.154386229805225e-05, "loss": 1.0198, "step": 153875 }, { "epoch": 0.39, "learning_rate": 6.154260405657054e-05, "loss": 1.02, "step": 153880 }, { "epoch": 0.39, "learning_rate": 6.154134581508883e-05, "loss": 1.0216, "step": 153885 }, { "epoch": 0.39, "learning_rate": 6.154008757360714e-05, "loss": 1.0212, "step": 153890 }, { "epoch": 0.39, "learning_rate": 6.153882933212543e-05, "loss": 1.0209, "step": 153895 }, { "epoch": 0.39, "learning_rate": 6.153757109064372e-05, "loss": 1.0202, "step": 153900 }, { "epoch": 0.39, "learning_rate": 6.153631284916201e-05, "loss": 1.035, "step": 153905 }, { "epoch": 0.39, "learning_rate": 6.153505460768032e-05, "loss": 1.0205, "step": 153910 }, { "epoch": 0.39, "learning_rate": 6.153379636619861e-05, "loss": 1.0202, "step": 153915 }, { "epoch": 0.39, "learning_rate": 6.15325381247169e-05, "loss": 1.0201, "step": 153920 }, { "epoch": 0.39, "learning_rate": 6.153127988323519e-05, "loss": 1.0198, "step": 153925 }, { "epoch": 0.39, "learning_rate": 6.153002164175348e-05, "loss": 1.0198, "step": 153930 }, { "epoch": 0.39, "learning_rate": 6.152876340027179e-05, "loss": 1.0192, "step": 153935 }, { "epoch": 0.39, "learning_rate": 6.152750515879008e-05, "loss": 1.0188, "step": 153940 }, { "epoch": 0.39, "learning_rate": 6.152624691730837e-05, "loss": 1.0208, "step": 153945 }, { "epoch": 0.39, "learning_rate": 6.152498867582666e-05, "loss": 1.0194, "step": 153950 }, { "epoch": 0.39, "learning_rate": 6.152373043434497e-05, "loss": 1.0205, "step": 153955 }, { "epoch": 0.39, "learning_rate": 6.152247219286326e-05, "loss": 1.0211, "step": 153960 }, { "epoch": 0.39, "learning_rate": 6.152121395138155e-05, "loss": 1.0195, "step": 153965 }, { "epoch": 0.39, "learning_rate": 6.151995570989984e-05, "loss": 1.0226, "step": 153970 }, { "epoch": 0.39, "learning_rate": 6.151869746841815e-05, "loss": 1.0201, "step": 153975 }, { "epoch": 0.39, "learning_rate": 6.151743922693644e-05, "loss": 1.0191, "step": 153980 }, { "epoch": 0.39, "learning_rate": 6.151618098545473e-05, "loss": 1.0209, "step": 153985 }, { "epoch": 0.39, "learning_rate": 6.151492274397302e-05, "loss": 1.0193, "step": 153990 }, { "epoch": 0.39, "learning_rate": 6.151366450249131e-05, "loss": 1.0204, "step": 153995 }, { "epoch": 0.39, "learning_rate": 6.151240626100962e-05, "loss": 1.0183, "step": 154000 }, { "epoch": 0.39, "learning_rate": 6.151114801952791e-05, "loss": 1.02, "step": 154005 }, { "epoch": 0.39, "learning_rate": 6.15098897780462e-05, "loss": 1.0188, "step": 154010 }, { "epoch": 0.39, "learning_rate": 6.15086315365645e-05, "loss": 1.0231, "step": 154015 }, { "epoch": 0.39, "learning_rate": 6.15073732950828e-05, "loss": 1.0201, "step": 154020 }, { "epoch": 0.39, "learning_rate": 6.150611505360109e-05, "loss": 1.0187, "step": 154025 }, { "epoch": 0.39, "learning_rate": 6.150485681211938e-05, "loss": 1.0169, "step": 154030 }, { "epoch": 0.39, "learning_rate": 6.150359857063767e-05, "loss": 1.0211, "step": 154035 }, { "epoch": 0.39, "learning_rate": 6.150234032915598e-05, "loss": 1.0209, "step": 154040 }, { "epoch": 0.39, "learning_rate": 6.150108208767427e-05, "loss": 1.0216, "step": 154045 }, { "epoch": 0.39, "learning_rate": 6.149982384619256e-05, "loss": 1.0199, "step": 154050 }, { "epoch": 0.39, "learning_rate": 6.149856560471085e-05, "loss": 1.0229, "step": 154055 }, { "epoch": 0.39, "learning_rate": 6.149730736322914e-05, "loss": 1.0207, "step": 154060 }, { "epoch": 0.39, "learning_rate": 6.149604912174745e-05, "loss": 1.0208, "step": 154065 }, { "epoch": 0.39, "learning_rate": 6.149479088026574e-05, "loss": 1.0205, "step": 154070 }, { "epoch": 0.39, "learning_rate": 6.149353263878403e-05, "loss": 1.019, "step": 154075 }, { "epoch": 0.39, "learning_rate": 6.149227439730232e-05, "loss": 1.0183, "step": 154080 }, { "epoch": 0.39, "learning_rate": 6.149101615582063e-05, "loss": 1.0205, "step": 154085 }, { "epoch": 0.39, "learning_rate": 6.148975791433892e-05, "loss": 1.0217, "step": 154090 }, { "epoch": 0.39, "learning_rate": 6.148849967285721e-05, "loss": 1.0203, "step": 154095 }, { "epoch": 0.39, "learning_rate": 6.14872414313755e-05, "loss": 1.0455, "step": 154100 }, { "epoch": 0.39, "learning_rate": 6.148598318989381e-05, "loss": 1.0186, "step": 154105 }, { "epoch": 0.39, "learning_rate": 6.14847249484121e-05, "loss": 1.0213, "step": 154110 }, { "epoch": 0.39, "learning_rate": 6.148346670693039e-05, "loss": 1.0212, "step": 154115 }, { "epoch": 0.39, "learning_rate": 6.148220846544868e-05, "loss": 1.0209, "step": 154120 }, { "epoch": 0.39, "learning_rate": 6.148095022396698e-05, "loss": 1.0217, "step": 154125 }, { "epoch": 0.39, "learning_rate": 6.147969198248528e-05, "loss": 1.0186, "step": 154130 }, { "epoch": 0.39, "learning_rate": 6.147843374100357e-05, "loss": 1.0184, "step": 154135 }, { "epoch": 0.39, "learning_rate": 6.147717549952186e-05, "loss": 1.0201, "step": 154140 }, { "epoch": 0.39, "learning_rate": 6.147591725804016e-05, "loss": 1.0214, "step": 154145 }, { "epoch": 0.39, "learning_rate": 6.147465901655846e-05, "loss": 1.0222, "step": 154150 }, { "epoch": 0.39, "learning_rate": 6.147340077507675e-05, "loss": 1.0193, "step": 154155 }, { "epoch": 0.39, "learning_rate": 6.147214253359504e-05, "loss": 1.0193, "step": 154160 }, { "epoch": 0.39, "learning_rate": 6.147088429211334e-05, "loss": 1.0198, "step": 154165 }, { "epoch": 0.39, "learning_rate": 6.146962605063164e-05, "loss": 1.0224, "step": 154170 }, { "epoch": 0.39, "learning_rate": 6.146836780914995e-05, "loss": 1.0188, "step": 154175 }, { "epoch": 0.39, "learning_rate": 6.146710956766824e-05, "loss": 1.0216, "step": 154180 }, { "epoch": 0.39, "learning_rate": 6.146585132618653e-05, "loss": 1.0194, "step": 154185 }, { "epoch": 0.39, "learning_rate": 6.146459308470482e-05, "loss": 1.0217, "step": 154190 }, { "epoch": 0.39, "learning_rate": 6.146333484322311e-05, "loss": 1.0214, "step": 154195 }, { "epoch": 0.39, "learning_rate": 6.146207660174142e-05, "loss": 1.0202, "step": 154200 }, { "epoch": 0.39, "learning_rate": 6.146081836025971e-05, "loss": 1.0223, "step": 154205 }, { "epoch": 0.39, "learning_rate": 6.1459560118778e-05, "loss": 1.0238, "step": 154210 }, { "epoch": 0.39, "learning_rate": 6.145830187729629e-05, "loss": 1.0217, "step": 154215 }, { "epoch": 0.39, "learning_rate": 6.14570436358146e-05, "loss": 1.0202, "step": 154220 }, { "epoch": 0.39, "learning_rate": 6.145578539433289e-05, "loss": 1.0235, "step": 154225 }, { "epoch": 0.39, "learning_rate": 6.145452715285118e-05, "loss": 1.0194, "step": 154230 }, { "epoch": 0.39, "learning_rate": 6.145326891136947e-05, "loss": 1.0202, "step": 154235 }, { "epoch": 0.39, "learning_rate": 6.145201066988778e-05, "loss": 1.0191, "step": 154240 }, { "epoch": 0.39, "learning_rate": 6.145075242840607e-05, "loss": 1.0185, "step": 154245 }, { "epoch": 0.39, "learning_rate": 6.144949418692436e-05, "loss": 1.0178, "step": 154250 }, { "epoch": 0.39, "learning_rate": 6.144823594544265e-05, "loss": 1.0206, "step": 154255 }, { "epoch": 0.39, "learning_rate": 6.144697770396094e-05, "loss": 1.0199, "step": 154260 }, { "epoch": 0.39, "learning_rate": 6.144571946247925e-05, "loss": 1.0203, "step": 154265 }, { "epoch": 0.39, "learning_rate": 6.144446122099754e-05, "loss": 1.0183, "step": 154270 }, { "epoch": 0.39, "learning_rate": 6.144320297951583e-05, "loss": 1.0449, "step": 154275 }, { "epoch": 0.39, "learning_rate": 6.144194473803412e-05, "loss": 1.0197, "step": 154280 }, { "epoch": 0.39, "learning_rate": 6.144068649655243e-05, "loss": 1.033, "step": 154285 }, { "epoch": 0.39, "learning_rate": 6.143942825507072e-05, "loss": 1.0206, "step": 154290 }, { "epoch": 0.39, "learning_rate": 6.143817001358901e-05, "loss": 1.02, "step": 154295 }, { "epoch": 0.39, "learning_rate": 6.14369117721073e-05, "loss": 1.0201, "step": 154300 }, { "epoch": 0.39, "learning_rate": 6.143565353062561e-05, "loss": 1.0199, "step": 154305 }, { "epoch": 0.39, "learning_rate": 6.14343952891439e-05, "loss": 1.0202, "step": 154310 }, { "epoch": 0.39, "learning_rate": 6.143313704766219e-05, "loss": 1.0196, "step": 154315 }, { "epoch": 0.39, "learning_rate": 6.143187880618048e-05, "loss": 1.0191, "step": 154320 }, { "epoch": 0.39, "learning_rate": 6.143062056469878e-05, "loss": 1.0208, "step": 154325 }, { "epoch": 0.39, "learning_rate": 6.142936232321708e-05, "loss": 1.0213, "step": 154330 }, { "epoch": 0.39, "learning_rate": 6.142810408173537e-05, "loss": 1.0183, "step": 154335 }, { "epoch": 0.39, "learning_rate": 6.142684584025366e-05, "loss": 1.0201, "step": 154340 }, { "epoch": 0.39, "learning_rate": 6.142558759877196e-05, "loss": 1.02, "step": 154345 }, { "epoch": 0.39, "learning_rate": 6.142432935729026e-05, "loss": 1.0197, "step": 154350 }, { "epoch": 0.39, "learning_rate": 6.142307111580855e-05, "loss": 1.0201, "step": 154355 }, { "epoch": 0.39, "learning_rate": 6.142181287432684e-05, "loss": 1.0211, "step": 154360 }, { "epoch": 0.39, "learning_rate": 6.142055463284514e-05, "loss": 1.0187, "step": 154365 }, { "epoch": 0.39, "learning_rate": 6.141929639136344e-05, "loss": 1.0217, "step": 154370 }, { "epoch": 0.39, "learning_rate": 6.141803814988173e-05, "loss": 1.0188, "step": 154375 }, { "epoch": 0.39, "learning_rate": 6.141677990840002e-05, "loss": 1.02, "step": 154380 }, { "epoch": 0.39, "learning_rate": 6.141552166691832e-05, "loss": 1.019, "step": 154385 }, { "epoch": 0.39, "learning_rate": 6.141426342543661e-05, "loss": 1.0188, "step": 154390 }, { "epoch": 0.39, "learning_rate": 6.141300518395491e-05, "loss": 1.0212, "step": 154395 }, { "epoch": 0.39, "learning_rate": 6.14117469424732e-05, "loss": 1.0177, "step": 154400 }, { "epoch": 0.39, "learning_rate": 6.14104887009915e-05, "loss": 1.0197, "step": 154405 }, { "epoch": 0.39, "learning_rate": 6.140923045950979e-05, "loss": 1.0197, "step": 154410 }, { "epoch": 0.39, "learning_rate": 6.140797221802809e-05, "loss": 1.0199, "step": 154415 }, { "epoch": 0.39, "learning_rate": 6.140671397654638e-05, "loss": 1.02, "step": 154420 }, { "epoch": 0.39, "learning_rate": 6.140545573506468e-05, "loss": 1.0215, "step": 154425 }, { "epoch": 0.39, "learning_rate": 6.140419749358297e-05, "loss": 1.0189, "step": 154430 }, { "epoch": 0.39, "learning_rate": 6.140293925210127e-05, "loss": 1.0203, "step": 154435 }, { "epoch": 0.39, "learning_rate": 6.140168101061956e-05, "loss": 1.0224, "step": 154440 }, { "epoch": 0.39, "learning_rate": 6.140042276913786e-05, "loss": 1.0201, "step": 154445 }, { "epoch": 0.39, "learning_rate": 6.139916452765615e-05, "loss": 1.0198, "step": 154450 }, { "epoch": 0.39, "learning_rate": 6.139790628617444e-05, "loss": 1.0201, "step": 154455 }, { "epoch": 0.39, "learning_rate": 6.139664804469274e-05, "loss": 1.0185, "step": 154460 }, { "epoch": 0.39, "learning_rate": 6.139538980321104e-05, "loss": 1.0219, "step": 154465 }, { "epoch": 0.39, "learning_rate": 6.139413156172933e-05, "loss": 1.0183, "step": 154470 }, { "epoch": 0.39, "learning_rate": 6.139287332024762e-05, "loss": 1.0188, "step": 154475 }, { "epoch": 0.39, "learning_rate": 6.139161507876592e-05, "loss": 1.0209, "step": 154480 }, { "epoch": 0.39, "learning_rate": 6.139035683728422e-05, "loss": 1.0219, "step": 154485 }, { "epoch": 0.39, "learning_rate": 6.138909859580251e-05, "loss": 1.021, "step": 154490 }, { "epoch": 0.39, "learning_rate": 6.13878403543208e-05, "loss": 1.0193, "step": 154495 }, { "epoch": 0.39, "learning_rate": 6.13865821128391e-05, "loss": 1.0194, "step": 154500 }, { "epoch": 0.39, "learning_rate": 6.13853238713574e-05, "loss": 1.0214, "step": 154505 }, { "epoch": 0.39, "learning_rate": 6.138406562987569e-05, "loss": 1.024, "step": 154510 }, { "epoch": 0.39, "learning_rate": 6.138280738839398e-05, "loss": 1.0216, "step": 154515 }, { "epoch": 0.39, "learning_rate": 6.138154914691227e-05, "loss": 1.0222, "step": 154520 }, { "epoch": 0.39, "learning_rate": 6.138029090543058e-05, "loss": 1.0212, "step": 154525 }, { "epoch": 0.39, "learning_rate": 6.137903266394887e-05, "loss": 1.0232, "step": 154530 }, { "epoch": 0.39, "learning_rate": 6.137777442246716e-05, "loss": 1.0223, "step": 154535 }, { "epoch": 0.39, "learning_rate": 6.137651618098545e-05, "loss": 1.02, "step": 154540 }, { "epoch": 0.39, "learning_rate": 6.137525793950376e-05, "loss": 1.0206, "step": 154545 }, { "epoch": 0.39, "learning_rate": 6.137399969802205e-05, "loss": 1.0217, "step": 154550 }, { "epoch": 0.39, "learning_rate": 6.137274145654034e-05, "loss": 1.0189, "step": 154555 }, { "epoch": 0.39, "learning_rate": 6.137148321505863e-05, "loss": 1.0192, "step": 154560 }, { "epoch": 0.39, "learning_rate": 6.137022497357694e-05, "loss": 1.0203, "step": 154565 }, { "epoch": 0.39, "learning_rate": 6.136896673209523e-05, "loss": 1.0207, "step": 154570 }, { "epoch": 0.39, "learning_rate": 6.136770849061352e-05, "loss": 1.0193, "step": 154575 }, { "epoch": 0.39, "learning_rate": 6.136645024913181e-05, "loss": 1.0174, "step": 154580 }, { "epoch": 0.39, "learning_rate": 6.13651920076501e-05, "loss": 1.0194, "step": 154585 }, { "epoch": 0.39, "learning_rate": 6.13639337661684e-05, "loss": 1.023, "step": 154590 }, { "epoch": 0.39, "learning_rate": 6.13626755246867e-05, "loss": 1.0179, "step": 154595 }, { "epoch": 0.39, "learning_rate": 6.136141728320499e-05, "loss": 1.0206, "step": 154600 }, { "epoch": 0.39, "learning_rate": 6.136015904172328e-05, "loss": 1.0196, "step": 154605 }, { "epoch": 0.39, "learning_rate": 6.135890080024159e-05, "loss": 1.0195, "step": 154610 }, { "epoch": 0.39, "learning_rate": 6.135764255875988e-05, "loss": 1.0211, "step": 154615 }, { "epoch": 0.39, "learning_rate": 6.135638431727817e-05, "loss": 1.0207, "step": 154620 }, { "epoch": 0.39, "learning_rate": 6.135512607579646e-05, "loss": 1.0182, "step": 154625 }, { "epoch": 0.39, "learning_rate": 6.135386783431477e-05, "loss": 1.0202, "step": 154630 }, { "epoch": 0.39, "learning_rate": 6.135260959283306e-05, "loss": 1.0193, "step": 154635 }, { "epoch": 0.39, "learning_rate": 6.135135135135135e-05, "loss": 1.0201, "step": 154640 }, { "epoch": 0.39, "learning_rate": 6.135009310986964e-05, "loss": 1.0202, "step": 154645 }, { "epoch": 0.39, "learning_rate": 6.134883486838793e-05, "loss": 1.0194, "step": 154650 }, { "epoch": 0.39, "learning_rate": 6.134757662690624e-05, "loss": 1.0214, "step": 154655 }, { "epoch": 0.39, "learning_rate": 6.134631838542453e-05, "loss": 1.0207, "step": 154660 }, { "epoch": 0.39, "learning_rate": 6.134506014394282e-05, "loss": 1.017, "step": 154665 }, { "epoch": 0.39, "learning_rate": 6.134380190246113e-05, "loss": 1.0207, "step": 154670 }, { "epoch": 0.39, "learning_rate": 6.134254366097942e-05, "loss": 1.0203, "step": 154675 }, { "epoch": 0.39, "learning_rate": 6.134128541949772e-05, "loss": 1.02, "step": 154680 }, { "epoch": 0.39, "learning_rate": 6.134002717801601e-05, "loss": 1.0179, "step": 154685 }, { "epoch": 0.39, "learning_rate": 6.13387689365343e-05, "loss": 1.0202, "step": 154690 }, { "epoch": 0.39, "learning_rate": 6.13375106950526e-05, "loss": 1.0204, "step": 154695 }, { "epoch": 0.39, "learning_rate": 6.13362524535709e-05, "loss": 1.0211, "step": 154700 }, { "epoch": 0.39, "learning_rate": 6.13349942120892e-05, "loss": 1.0187, "step": 154705 }, { "epoch": 0.39, "learning_rate": 6.133373597060749e-05, "loss": 1.0206, "step": 154710 }, { "epoch": 0.39, "learning_rate": 6.133247772912578e-05, "loss": 1.0222, "step": 154715 }, { "epoch": 0.39, "learning_rate": 6.133121948764407e-05, "loss": 1.0211, "step": 154720 }, { "epoch": 0.39, "learning_rate": 6.132996124616237e-05, "loss": 1.0195, "step": 154725 }, { "epoch": 0.39, "learning_rate": 6.132870300468067e-05, "loss": 1.0205, "step": 154730 }, { "epoch": 0.39, "learning_rate": 6.132744476319896e-05, "loss": 1.022, "step": 154735 }, { "epoch": 0.39, "learning_rate": 6.132618652171725e-05, "loss": 1.019, "step": 154740 }, { "epoch": 0.39, "learning_rate": 6.132492828023555e-05, "loss": 1.0203, "step": 154745 }, { "epoch": 0.39, "learning_rate": 6.132367003875385e-05, "loss": 1.0364, "step": 154750 }, { "epoch": 0.39, "learning_rate": 6.132241179727214e-05, "loss": 1.0207, "step": 154755 }, { "epoch": 0.39, "learning_rate": 6.132115355579043e-05, "loss": 1.0199, "step": 154760 }, { "epoch": 0.39, "learning_rate": 6.131989531430873e-05, "loss": 1.0197, "step": 154765 }, { "epoch": 0.39, "learning_rate": 6.131863707282703e-05, "loss": 1.0205, "step": 154770 }, { "epoch": 0.39, "learning_rate": 6.131737883134532e-05, "loss": 1.0203, "step": 154775 }, { "epoch": 0.39, "learning_rate": 6.131612058986361e-05, "loss": 1.0307, "step": 154780 }, { "epoch": 0.39, "learning_rate": 6.13148623483819e-05, "loss": 1.0187, "step": 154785 }, { "epoch": 0.39, "learning_rate": 6.13136041069002e-05, "loss": 1.0196, "step": 154790 }, { "epoch": 0.39, "learning_rate": 6.13123458654185e-05, "loss": 1.0181, "step": 154795 }, { "epoch": 0.39, "learning_rate": 6.131108762393679e-05, "loss": 1.0201, "step": 154800 }, { "epoch": 0.39, "learning_rate": 6.130982938245508e-05, "loss": 1.0212, "step": 154805 }, { "epoch": 0.39, "learning_rate": 6.130857114097339e-05, "loss": 1.0203, "step": 154810 }, { "epoch": 0.39, "learning_rate": 6.130731289949168e-05, "loss": 1.0404, "step": 154815 }, { "epoch": 0.39, "learning_rate": 6.130605465800997e-05, "loss": 1.0213, "step": 154820 }, { "epoch": 0.39, "learning_rate": 6.130479641652826e-05, "loss": 1.0205, "step": 154825 }, { "epoch": 0.39, "learning_rate": 6.130353817504657e-05, "loss": 1.0216, "step": 154830 }, { "epoch": 0.39, "learning_rate": 6.130227993356486e-05, "loss": 1.0188, "step": 154835 }, { "epoch": 0.39, "learning_rate": 6.130102169208315e-05, "loss": 1.0196, "step": 154840 }, { "epoch": 0.39, "learning_rate": 6.129976345060144e-05, "loss": 1.0213, "step": 154845 }, { "epoch": 0.39, "learning_rate": 6.129850520911973e-05, "loss": 1.0195, "step": 154850 }, { "epoch": 0.39, "learning_rate": 6.129724696763804e-05, "loss": 1.02, "step": 154855 }, { "epoch": 0.39, "learning_rate": 6.129598872615633e-05, "loss": 1.0208, "step": 154860 }, { "epoch": 0.39, "learning_rate": 6.129473048467462e-05, "loss": 1.023, "step": 154865 }, { "epoch": 0.39, "learning_rate": 6.129347224319291e-05, "loss": 1.0184, "step": 154870 }, { "epoch": 0.39, "learning_rate": 6.129221400171122e-05, "loss": 1.02, "step": 154875 }, { "epoch": 0.39, "learning_rate": 6.129095576022951e-05, "loss": 1.0192, "step": 154880 }, { "epoch": 0.39, "learning_rate": 6.12896975187478e-05, "loss": 1.0198, "step": 154885 }, { "epoch": 0.39, "learning_rate": 6.128843927726609e-05, "loss": 1.0214, "step": 154890 }, { "epoch": 0.39, "learning_rate": 6.12871810357844e-05, "loss": 1.0186, "step": 154895 }, { "epoch": 0.39, "learning_rate": 6.128592279430269e-05, "loss": 1.0213, "step": 154900 }, { "epoch": 0.39, "learning_rate": 6.128466455282098e-05, "loss": 1.0211, "step": 154905 }, { "epoch": 0.39, "learning_rate": 6.128340631133927e-05, "loss": 1.0209, "step": 154910 }, { "epoch": 0.39, "learning_rate": 6.128214806985756e-05, "loss": 1.0193, "step": 154915 }, { "epoch": 0.39, "learning_rate": 6.128088982837587e-05, "loss": 1.0193, "step": 154920 }, { "epoch": 0.39, "learning_rate": 6.127963158689416e-05, "loss": 1.0182, "step": 154925 }, { "epoch": 0.39, "learning_rate": 6.127837334541245e-05, "loss": 1.0183, "step": 154930 }, { "epoch": 0.39, "learning_rate": 6.127711510393074e-05, "loss": 1.0202, "step": 154935 }, { "epoch": 0.39, "learning_rate": 6.127585686244905e-05, "loss": 1.0212, "step": 154940 }, { "epoch": 0.39, "learning_rate": 6.127459862096734e-05, "loss": 1.0201, "step": 154945 }, { "epoch": 0.39, "learning_rate": 6.127334037948563e-05, "loss": 1.0216, "step": 154950 }, { "epoch": 0.39, "learning_rate": 6.127208213800392e-05, "loss": 1.0191, "step": 154955 }, { "epoch": 0.39, "learning_rate": 6.127082389652223e-05, "loss": 1.0182, "step": 154960 }, { "epoch": 0.39, "learning_rate": 6.126956565504052e-05, "loss": 1.0203, "step": 154965 }, { "epoch": 0.39, "learning_rate": 6.126830741355881e-05, "loss": 1.0223, "step": 154970 }, { "epoch": 0.39, "learning_rate": 6.12670491720771e-05, "loss": 1.0179, "step": 154975 }, { "epoch": 0.39, "learning_rate": 6.12657909305954e-05, "loss": 1.0221, "step": 154980 }, { "epoch": 0.39, "learning_rate": 6.12645326891137e-05, "loss": 1.0216, "step": 154985 }, { "epoch": 0.39, "learning_rate": 6.126327444763199e-05, "loss": 1.0199, "step": 154990 }, { "epoch": 0.39, "learning_rate": 6.126201620615028e-05, "loss": 1.018, "step": 154995 }, { "epoch": 0.39, "learning_rate": 6.126075796466858e-05, "loss": 1.0184, "step": 155000 }, { "epoch": 0.39, "learning_rate": 6.125949972318688e-05, "loss": 1.0201, "step": 155005 }, { "epoch": 0.39, "learning_rate": 6.125824148170517e-05, "loss": 1.0192, "step": 155010 }, { "epoch": 0.39, "learning_rate": 6.125698324022346e-05, "loss": 1.0206, "step": 155015 }, { "epoch": 0.39, "learning_rate": 6.125572499874175e-05, "loss": 1.0217, "step": 155020 }, { "epoch": 0.39, "learning_rate": 6.125446675726006e-05, "loss": 1.0172, "step": 155025 }, { "epoch": 0.39, "learning_rate": 6.125320851577835e-05, "loss": 1.021, "step": 155030 }, { "epoch": 0.39, "learning_rate": 6.125195027429664e-05, "loss": 1.0204, "step": 155035 }, { "epoch": 0.39, "learning_rate": 6.125069203281493e-05, "loss": 1.0189, "step": 155040 }, { "epoch": 0.39, "learning_rate": 6.124943379133323e-05, "loss": 1.0174, "step": 155045 }, { "epoch": 0.39, "learning_rate": 6.124817554985153e-05, "loss": 1.0199, "step": 155050 }, { "epoch": 0.39, "learning_rate": 6.124691730836982e-05, "loss": 1.0189, "step": 155055 }, { "epoch": 0.39, "learning_rate": 6.124565906688811e-05, "loss": 1.0214, "step": 155060 }, { "epoch": 0.39, "learning_rate": 6.12444008254064e-05, "loss": 1.0205, "step": 155065 }, { "epoch": 0.39, "learning_rate": 6.124314258392471e-05, "loss": 1.0216, "step": 155070 }, { "epoch": 0.39, "learning_rate": 6.1241884342443e-05, "loss": 1.0177, "step": 155075 }, { "epoch": 0.39, "learning_rate": 6.12406261009613e-05, "loss": 1.019, "step": 155080 }, { "epoch": 0.39, "learning_rate": 6.123936785947959e-05, "loss": 1.0198, "step": 155085 }, { "epoch": 0.39, "learning_rate": 6.123810961799789e-05, "loss": 1.0198, "step": 155090 }, { "epoch": 0.39, "learning_rate": 6.123685137651618e-05, "loss": 1.0203, "step": 155095 }, { "epoch": 0.39, "learning_rate": 6.123559313503447e-05, "loss": 1.0205, "step": 155100 }, { "epoch": 0.39, "learning_rate": 6.123433489355277e-05, "loss": 1.0414, "step": 155105 }, { "epoch": 0.39, "learning_rate": 6.123307665207106e-05, "loss": 1.0204, "step": 155110 }, { "epoch": 0.39, "learning_rate": 6.123181841058936e-05, "loss": 1.0218, "step": 155115 }, { "epoch": 0.39, "learning_rate": 6.123056016910765e-05, "loss": 1.0223, "step": 155120 }, { "epoch": 0.39, "learning_rate": 6.122955357592229e-05, "loss": 1.043, "step": 155125 }, { "epoch": 0.39, "learning_rate": 6.122829533444058e-05, "loss": 1.0336, "step": 155130 }, { "epoch": 0.39, "learning_rate": 6.122703709295889e-05, "loss": 1.0202, "step": 155135 }, { "epoch": 0.39, "learning_rate": 6.122577885147718e-05, "loss": 1.0235, "step": 155140 }, { "epoch": 0.39, "learning_rate": 6.122477225829182e-05, "loss": 1.0397, "step": 155145 }, { "epoch": 0.39, "learning_rate": 6.12235140168101e-05, "loss": 1.0201, "step": 155150 }, { "epoch": 0.39, "learning_rate": 6.122225577532841e-05, "loss": 1.0446, "step": 155155 }, { "epoch": 0.39, "learning_rate": 6.12209975338467e-05, "loss": 1.0222, "step": 155160 }, { "epoch": 0.39, "learning_rate": 6.1219739292365e-05, "loss": 1.0209, "step": 155165 }, { "epoch": 0.39, "learning_rate": 6.121848105088329e-05, "loss": 1.0212, "step": 155170 }, { "epoch": 0.39, "learning_rate": 6.121722280940159e-05, "loss": 1.0419, "step": 155175 }, { "epoch": 0.39, "learning_rate": 6.121596456791988e-05, "loss": 1.0212, "step": 155180 }, { "epoch": 0.39, "learning_rate": 6.121470632643817e-05, "loss": 1.0188, "step": 155185 }, { "epoch": 0.39, "learning_rate": 6.121369973325281e-05, "loss": 1.0366, "step": 155190 }, { "epoch": 0.39, "learning_rate": 6.12124414917711e-05, "loss": 1.0205, "step": 155195 }, { "epoch": 0.39, "learning_rate": 6.12111832502894e-05, "loss": 1.0191, "step": 155200 }, { "epoch": 0.39, "learning_rate": 6.120992500880769e-05, "loss": 1.0198, "step": 155205 }, { "epoch": 0.39, "learning_rate": 6.120866676732598e-05, "loss": 1.0195, "step": 155210 }, { "epoch": 0.39, "learning_rate": 6.120740852584428e-05, "loss": 1.0199, "step": 155215 }, { "epoch": 0.39, "learning_rate": 6.120615028436257e-05, "loss": 1.0215, "step": 155220 }, { "epoch": 0.39, "learning_rate": 6.120489204288087e-05, "loss": 1.0196, "step": 155225 }, { "epoch": 0.39, "learning_rate": 6.120363380139916e-05, "loss": 1.0226, "step": 155230 }, { "epoch": 0.39, "learning_rate": 6.120237555991746e-05, "loss": 1.0199, "step": 155235 }, { "epoch": 0.39, "learning_rate": 6.120111731843575e-05, "loss": 1.0357, "step": 155240 }, { "epoch": 0.39, "learning_rate": 6.119985907695405e-05, "loss": 1.0228, "step": 155245 }, { "epoch": 0.39, "learning_rate": 6.119860083547234e-05, "loss": 1.0187, "step": 155250 }, { "epoch": 0.39, "learning_rate": 6.119734259399064e-05, "loss": 1.0214, "step": 155255 }, { "epoch": 0.39, "learning_rate": 6.119608435250893e-05, "loss": 1.0215, "step": 155260 }, { "epoch": 0.39, "learning_rate": 6.119482611102723e-05, "loss": 1.0188, "step": 155265 }, { "epoch": 0.39, "learning_rate": 6.119356786954552e-05, "loss": 1.018, "step": 155270 }, { "epoch": 0.39, "learning_rate": 6.119230962806382e-05, "loss": 1.0228, "step": 155275 }, { "epoch": 0.39, "learning_rate": 6.119105138658213e-05, "loss": 1.0211, "step": 155280 }, { "epoch": 0.39, "learning_rate": 6.118979314510042e-05, "loss": 1.0204, "step": 155285 }, { "epoch": 0.39, "learning_rate": 6.118853490361871e-05, "loss": 1.0208, "step": 155290 }, { "epoch": 0.39, "learning_rate": 6.1187276662137e-05, "loss": 1.0201, "step": 155295 }, { "epoch": 0.39, "learning_rate": 6.11860184206553e-05, "loss": 1.017, "step": 155300 }, { "epoch": 0.39, "learning_rate": 6.11847601791736e-05, "loss": 1.0202, "step": 155305 }, { "epoch": 0.39, "learning_rate": 6.118350193769189e-05, "loss": 1.0182, "step": 155310 }, { "epoch": 0.39, "learning_rate": 6.118224369621018e-05, "loss": 1.0185, "step": 155315 }, { "epoch": 0.39, "learning_rate": 6.118098545472847e-05, "loss": 1.0198, "step": 155320 }, { "epoch": 0.39, "learning_rate": 6.117972721324678e-05, "loss": 1.0193, "step": 155325 }, { "epoch": 0.39, "learning_rate": 6.117846897176507e-05, "loss": 1.0211, "step": 155330 }, { "epoch": 0.39, "learning_rate": 6.117721073028336e-05, "loss": 1.0209, "step": 155335 }, { "epoch": 0.39, "learning_rate": 6.117595248880165e-05, "loss": 1.0166, "step": 155340 }, { "epoch": 0.39, "learning_rate": 6.117469424731996e-05, "loss": 1.0196, "step": 155345 }, { "epoch": 0.39, "learning_rate": 6.117343600583825e-05, "loss": 1.0203, "step": 155350 }, { "epoch": 0.39, "learning_rate": 6.117217776435654e-05, "loss": 1.0195, "step": 155355 }, { "epoch": 0.39, "learning_rate": 6.117091952287483e-05, "loss": 1.023, "step": 155360 }, { "epoch": 0.39, "learning_rate": 6.116966128139313e-05, "loss": 1.0202, "step": 155365 }, { "epoch": 0.39, "learning_rate": 6.116840303991143e-05, "loss": 1.022, "step": 155370 }, { "epoch": 0.39, "learning_rate": 6.116714479842972e-05, "loss": 1.02, "step": 155375 }, { "epoch": 0.39, "learning_rate": 6.116588655694801e-05, "loss": 1.0339, "step": 155380 }, { "epoch": 0.39, "learning_rate": 6.11646283154663e-05, "loss": 1.0208, "step": 155385 }, { "epoch": 0.39, "learning_rate": 6.116337007398461e-05, "loss": 1.0201, "step": 155390 }, { "epoch": 0.39, "learning_rate": 6.11621118325029e-05, "loss": 1.02, "step": 155395 }, { "epoch": 0.39, "learning_rate": 6.11608535910212e-05, "loss": 1.0191, "step": 155400 }, { "epoch": 0.39, "learning_rate": 6.115959534953949e-05, "loss": 1.0214, "step": 155405 }, { "epoch": 0.39, "learning_rate": 6.115833710805779e-05, "loss": 1.0205, "step": 155410 }, { "epoch": 0.39, "learning_rate": 6.115707886657608e-05, "loss": 1.0196, "step": 155415 }, { "epoch": 0.39, "learning_rate": 6.115582062509437e-05, "loss": 1.0202, "step": 155420 }, { "epoch": 0.39, "learning_rate": 6.115456238361267e-05, "loss": 1.0182, "step": 155425 }, { "epoch": 0.39, "learning_rate": 6.115330414213096e-05, "loss": 1.0213, "step": 155430 }, { "epoch": 0.39, "learning_rate": 6.115204590064926e-05, "loss": 1.0214, "step": 155435 }, { "epoch": 0.39, "learning_rate": 6.115078765916755e-05, "loss": 1.0219, "step": 155440 }, { "epoch": 0.39, "learning_rate": 6.114952941768584e-05, "loss": 1.0196, "step": 155445 }, { "epoch": 0.39, "learning_rate": 6.114827117620414e-05, "loss": 1.0381, "step": 155450 }, { "epoch": 0.39, "learning_rate": 6.114701293472244e-05, "loss": 1.0202, "step": 155455 }, { "epoch": 0.39, "learning_rate": 6.114575469324073e-05, "loss": 1.017, "step": 155460 }, { "epoch": 0.39, "learning_rate": 6.114449645175902e-05, "loss": 1.0225, "step": 155465 }, { "epoch": 0.39, "learning_rate": 6.114323821027732e-05, "loss": 1.0186, "step": 155470 }, { "epoch": 0.39, "learning_rate": 6.114197996879562e-05, "loss": 1.019, "step": 155475 }, { "epoch": 0.39, "learning_rate": 6.114072172731391e-05, "loss": 1.0208, "step": 155480 }, { "epoch": 0.39, "learning_rate": 6.11394634858322e-05, "loss": 1.02, "step": 155485 }, { "epoch": 0.39, "learning_rate": 6.11382052443505e-05, "loss": 1.0197, "step": 155490 }, { "epoch": 0.39, "learning_rate": 6.113694700286879e-05, "loss": 1.0205, "step": 155495 }, { "epoch": 0.39, "learning_rate": 6.11356887613871e-05, "loss": 1.021, "step": 155500 }, { "epoch": 0.39, "learning_rate": 6.113443051990538e-05, "loss": 1.0191, "step": 155505 }, { "epoch": 0.39, "learning_rate": 6.113317227842368e-05, "loss": 1.0212, "step": 155510 }, { "epoch": 0.39, "learning_rate": 6.113191403694197e-05, "loss": 1.0179, "step": 155515 }, { "epoch": 0.39, "learning_rate": 6.113065579546027e-05, "loss": 1.0191, "step": 155520 }, { "epoch": 0.39, "learning_rate": 6.112939755397856e-05, "loss": 1.0371, "step": 155525 }, { "epoch": 0.39, "learning_rate": 6.112813931249686e-05, "loss": 1.0212, "step": 155530 }, { "epoch": 0.39, "learning_rate": 6.112688107101515e-05, "loss": 1.0196, "step": 155535 }, { "epoch": 0.39, "learning_rate": 6.112562282953345e-05, "loss": 1.0201, "step": 155540 }, { "epoch": 0.39, "learning_rate": 6.112436458805174e-05, "loss": 1.0211, "step": 155545 }, { "epoch": 0.39, "learning_rate": 6.112310634657004e-05, "loss": 1.043, "step": 155550 }, { "epoch": 0.39, "learning_rate": 6.112184810508833e-05, "loss": 1.0219, "step": 155555 }, { "epoch": 0.39, "learning_rate": 6.112058986360662e-05, "loss": 1.0201, "step": 155560 }, { "epoch": 0.39, "learning_rate": 6.111933162212492e-05, "loss": 1.0228, "step": 155565 }, { "epoch": 0.39, "learning_rate": 6.111807338064322e-05, "loss": 1.0183, "step": 155570 }, { "epoch": 0.39, "learning_rate": 6.111681513916151e-05, "loss": 1.0206, "step": 155575 }, { "epoch": 0.39, "learning_rate": 6.11155568976798e-05, "loss": 1.0186, "step": 155580 }, { "epoch": 0.39, "learning_rate": 6.11142986561981e-05, "loss": 1.02, "step": 155585 }, { "epoch": 0.39, "learning_rate": 6.11130404147164e-05, "loss": 1.0206, "step": 155590 }, { "epoch": 0.39, "learning_rate": 6.111178217323469e-05, "loss": 1.0202, "step": 155595 }, { "epoch": 0.39, "learning_rate": 6.111052393175298e-05, "loss": 1.02, "step": 155600 }, { "epoch": 0.39, "learning_rate": 6.110926569027128e-05, "loss": 1.0192, "step": 155605 }, { "epoch": 0.39, "learning_rate": 6.110800744878958e-05, "loss": 1.0204, "step": 155610 }, { "epoch": 0.39, "learning_rate": 6.110674920730787e-05, "loss": 1.0214, "step": 155615 }, { "epoch": 0.39, "learning_rate": 6.110549096582616e-05, "loss": 1.0205, "step": 155620 }, { "epoch": 0.39, "learning_rate": 6.110423272434445e-05, "loss": 1.0195, "step": 155625 }, { "epoch": 0.39, "learning_rate": 6.110297448286276e-05, "loss": 1.0191, "step": 155630 }, { "epoch": 0.39, "learning_rate": 6.110171624138105e-05, "loss": 1.0221, "step": 155635 }, { "epoch": 0.39, "learning_rate": 6.110045799989934e-05, "loss": 1.0205, "step": 155640 }, { "epoch": 0.39, "learning_rate": 6.109919975841763e-05, "loss": 1.0217, "step": 155645 }, { "epoch": 0.39, "learning_rate": 6.109794151693594e-05, "loss": 1.0218, "step": 155650 }, { "epoch": 0.39, "learning_rate": 6.109668327545423e-05, "loss": 1.0218, "step": 155655 }, { "epoch": 0.39, "learning_rate": 6.109542503397252e-05, "loss": 1.0203, "step": 155660 }, { "epoch": 0.39, "learning_rate": 6.109416679249081e-05, "loss": 1.0219, "step": 155665 }, { "epoch": 0.39, "learning_rate": 6.109290855100912e-05, "loss": 1.0207, "step": 155670 }, { "epoch": 0.39, "learning_rate": 6.109165030952741e-05, "loss": 1.0187, "step": 155675 }, { "epoch": 0.39, "learning_rate": 6.10903920680457e-05, "loss": 1.0202, "step": 155680 }, { "epoch": 0.39, "learning_rate": 6.108913382656399e-05, "loss": 1.0199, "step": 155685 }, { "epoch": 0.39, "learning_rate": 6.108787558508228e-05, "loss": 1.0197, "step": 155690 }, { "epoch": 0.39, "learning_rate": 6.108661734360059e-05, "loss": 1.0197, "step": 155695 }, { "epoch": 0.39, "learning_rate": 6.108535910211888e-05, "loss": 1.0195, "step": 155700 }, { "epoch": 0.39, "learning_rate": 6.108410086063717e-05, "loss": 1.0182, "step": 155705 }, { "epoch": 0.39, "learning_rate": 6.108284261915546e-05, "loss": 1.0186, "step": 155710 }, { "epoch": 0.39, "learning_rate": 6.108158437767377e-05, "loss": 1.0205, "step": 155715 }, { "epoch": 0.39, "learning_rate": 6.108032613619206e-05, "loss": 1.0221, "step": 155720 }, { "epoch": 0.39, "learning_rate": 6.107906789471035e-05, "loss": 1.0212, "step": 155725 }, { "epoch": 0.39, "learning_rate": 6.107780965322864e-05, "loss": 1.0204, "step": 155730 }, { "epoch": 0.39, "learning_rate": 6.107655141174695e-05, "loss": 1.0204, "step": 155735 }, { "epoch": 0.39, "learning_rate": 6.107529317026524e-05, "loss": 1.0192, "step": 155740 }, { "epoch": 0.39, "learning_rate": 6.107403492878353e-05, "loss": 1.0198, "step": 155745 }, { "epoch": 0.39, "learning_rate": 6.107277668730182e-05, "loss": 1.0224, "step": 155750 }, { "epoch": 0.39, "learning_rate": 6.107151844582011e-05, "loss": 1.0202, "step": 155755 }, { "epoch": 0.39, "learning_rate": 6.107026020433842e-05, "loss": 1.0201, "step": 155760 }, { "epoch": 0.39, "learning_rate": 6.106900196285671e-05, "loss": 1.021, "step": 155765 }, { "epoch": 0.39, "learning_rate": 6.106774372137502e-05, "loss": 1.0226, "step": 155770 }, { "epoch": 0.39, "learning_rate": 6.106648547989331e-05, "loss": 1.0204, "step": 155775 }, { "epoch": 0.39, "learning_rate": 6.10652272384116e-05, "loss": 1.0219, "step": 155780 }, { "epoch": 0.39, "learning_rate": 6.10639689969299e-05, "loss": 1.0228, "step": 155785 }, { "epoch": 0.39, "learning_rate": 6.10627107554482e-05, "loss": 1.022, "step": 155790 }, { "epoch": 0.39, "learning_rate": 6.106145251396649e-05, "loss": 1.0207, "step": 155795 }, { "epoch": 0.39, "learning_rate": 6.106019427248478e-05, "loss": 1.0198, "step": 155800 }, { "epoch": 0.39, "learning_rate": 6.105893603100308e-05, "loss": 1.0203, "step": 155805 }, { "epoch": 0.39, "learning_rate": 6.105767778952138e-05, "loss": 1.0408, "step": 155810 }, { "epoch": 0.39, "learning_rate": 6.105641954803967e-05, "loss": 1.0206, "step": 155815 }, { "epoch": 0.39, "learning_rate": 6.105516130655796e-05, "loss": 1.0173, "step": 155820 }, { "epoch": 0.39, "learning_rate": 6.105390306507625e-05, "loss": 1.021, "step": 155825 }, { "epoch": 0.39, "learning_rate": 6.105264482359456e-05, "loss": 1.0224, "step": 155830 }, { "epoch": 0.39, "learning_rate": 6.105138658211285e-05, "loss": 1.0193, "step": 155835 }, { "epoch": 0.39, "learning_rate": 6.105012834063114e-05, "loss": 1.019, "step": 155840 }, { "epoch": 0.39, "learning_rate": 6.104887009914943e-05, "loss": 1.0201, "step": 155845 }, { "epoch": 0.39, "learning_rate": 6.104761185766774e-05, "loss": 1.0195, "step": 155850 }, { "epoch": 0.39, "learning_rate": 6.104635361618603e-05, "loss": 1.0208, "step": 155855 }, { "epoch": 0.39, "learning_rate": 6.104509537470432e-05, "loss": 1.0191, "step": 155860 }, { "epoch": 0.39, "learning_rate": 6.104383713322261e-05, "loss": 1.0203, "step": 155865 }, { "epoch": 0.39, "learning_rate": 6.104257889174092e-05, "loss": 1.0219, "step": 155870 }, { "epoch": 0.39, "learning_rate": 6.104132065025921e-05, "loss": 1.02, "step": 155875 }, { "epoch": 0.39, "learning_rate": 6.10400624087775e-05, "loss": 1.0206, "step": 155880 }, { "epoch": 0.39, "learning_rate": 6.103880416729579e-05, "loss": 1.0188, "step": 155885 }, { "epoch": 0.39, "learning_rate": 6.103754592581408e-05, "loss": 1.0187, "step": 155890 }, { "epoch": 0.39, "learning_rate": 6.103628768433239e-05, "loss": 1.0194, "step": 155895 }, { "epoch": 0.39, "learning_rate": 6.103502944285068e-05, "loss": 1.0215, "step": 155900 }, { "epoch": 0.39, "learning_rate": 6.103377120136897e-05, "loss": 1.0235, "step": 155905 }, { "epoch": 0.39, "learning_rate": 6.103251295988727e-05, "loss": 1.0189, "step": 155910 }, { "epoch": 0.39, "learning_rate": 6.103125471840556e-05, "loss": 1.0228, "step": 155915 }, { "epoch": 0.39, "learning_rate": 6.102999647692386e-05, "loss": 1.0209, "step": 155920 }, { "epoch": 0.39, "learning_rate": 6.102873823544215e-05, "loss": 1.0214, "step": 155925 }, { "epoch": 0.39, "learning_rate": 6.102747999396045e-05, "loss": 1.022, "step": 155930 }, { "epoch": 0.39, "learning_rate": 6.102622175247874e-05, "loss": 1.0199, "step": 155935 }, { "epoch": 0.39, "learning_rate": 6.102496351099704e-05, "loss": 1.0193, "step": 155940 }, { "epoch": 0.39, "learning_rate": 6.102370526951533e-05, "loss": 1.0209, "step": 155945 }, { "epoch": 0.39, "learning_rate": 6.102244702803362e-05, "loss": 1.0187, "step": 155950 }, { "epoch": 0.39, "learning_rate": 6.102118878655192e-05, "loss": 1.022, "step": 155955 }, { "epoch": 0.39, "learning_rate": 6.101993054507021e-05, "loss": 1.0207, "step": 155960 }, { "epoch": 0.39, "learning_rate": 6.101867230358851e-05, "loss": 1.0424, "step": 155965 }, { "epoch": 0.39, "learning_rate": 6.10174140621068e-05, "loss": 1.0194, "step": 155970 }, { "epoch": 0.39, "learning_rate": 6.10161558206251e-05, "loss": 1.0205, "step": 155975 }, { "epoch": 0.39, "learning_rate": 6.101489757914339e-05, "loss": 1.019, "step": 155980 }, { "epoch": 0.39, "learning_rate": 6.101363933766169e-05, "loss": 1.0185, "step": 155985 }, { "epoch": 0.39, "learning_rate": 6.101238109617998e-05, "loss": 1.0178, "step": 155990 }, { "epoch": 0.39, "learning_rate": 6.101112285469828e-05, "loss": 1.0206, "step": 155995 }, { "epoch": 0.39, "learning_rate": 6.100986461321657e-05, "loss": 1.0186, "step": 156000 }, { "epoch": 0.39, "learning_rate": 6.100860637173487e-05, "loss": 1.0195, "step": 156005 }, { "epoch": 0.39, "learning_rate": 6.100734813025316e-05, "loss": 1.0203, "step": 156010 }, { "epoch": 0.39, "learning_rate": 6.100608988877145e-05, "loss": 1.0205, "step": 156015 }, { "epoch": 0.39, "learning_rate": 6.100483164728975e-05, "loss": 1.0201, "step": 156020 }, { "epoch": 0.39, "learning_rate": 6.100357340580804e-05, "loss": 1.0206, "step": 156025 }, { "epoch": 0.39, "learning_rate": 6.100231516432634e-05, "loss": 1.0197, "step": 156030 }, { "epoch": 0.39, "learning_rate": 6.100105692284463e-05, "loss": 1.0205, "step": 156035 }, { "epoch": 0.39, "learning_rate": 6.099979868136293e-05, "loss": 1.0199, "step": 156040 }, { "epoch": 0.39, "learning_rate": 6.099854043988122e-05, "loss": 1.0195, "step": 156045 }, { "epoch": 0.39, "learning_rate": 6.099728219839952e-05, "loss": 1.0205, "step": 156050 }, { "epoch": 0.39, "learning_rate": 6.099602395691781e-05, "loss": 1.022, "step": 156055 }, { "epoch": 0.39, "learning_rate": 6.099476571543611e-05, "loss": 1.0189, "step": 156060 }, { "epoch": 0.39, "learning_rate": 6.09935074739544e-05, "loss": 1.0184, "step": 156065 }, { "epoch": 0.39, "learning_rate": 6.09922492324727e-05, "loss": 1.0342, "step": 156070 }, { "epoch": 0.39, "learning_rate": 6.099099099099099e-05, "loss": 1.0185, "step": 156075 }, { "epoch": 0.39, "learning_rate": 6.0989732749509284e-05, "loss": 1.0205, "step": 156080 }, { "epoch": 0.39, "learning_rate": 6.098847450802758e-05, "loss": 1.0211, "step": 156085 }, { "epoch": 0.39, "learning_rate": 6.0987216266545874e-05, "loss": 1.0204, "step": 156090 }, { "epoch": 0.39, "learning_rate": 6.098595802506417e-05, "loss": 1.0218, "step": 156095 }, { "epoch": 0.39, "learning_rate": 6.0984699783582464e-05, "loss": 1.0188, "step": 156100 }, { "epoch": 0.39, "learning_rate": 6.098344154210076e-05, "loss": 1.0158, "step": 156105 }, { "epoch": 0.39, "learning_rate": 6.0982183300619054e-05, "loss": 1.0187, "step": 156110 }, { "epoch": 0.39, "learning_rate": 6.098092505913735e-05, "loss": 1.02, "step": 156115 }, { "epoch": 0.39, "learning_rate": 6.0979666817655644e-05, "loss": 1.0189, "step": 156120 }, { "epoch": 0.39, "learning_rate": 6.097840857617394e-05, "loss": 1.0206, "step": 156125 }, { "epoch": 0.39, "learning_rate": 6.0977150334692234e-05, "loss": 1.0218, "step": 156130 }, { "epoch": 0.39, "learning_rate": 6.097589209321053e-05, "loss": 1.0191, "step": 156135 }, { "epoch": 0.39, "learning_rate": 6.0974633851728824e-05, "loss": 1.0182, "step": 156140 }, { "epoch": 0.39, "learning_rate": 6.0973375610247116e-05, "loss": 1.0187, "step": 156145 }, { "epoch": 0.39, "learning_rate": 6.0972117368765414e-05, "loss": 1.0196, "step": 156150 }, { "epoch": 0.39, "learning_rate": 6.0970859127283706e-05, "loss": 1.0201, "step": 156155 }, { "epoch": 0.39, "learning_rate": 6.0969600885802004e-05, "loss": 1.0215, "step": 156160 }, { "epoch": 0.39, "learning_rate": 6.0968342644320296e-05, "loss": 1.0209, "step": 156165 }, { "epoch": 0.39, "learning_rate": 6.0967084402838594e-05, "loss": 1.0195, "step": 156170 }, { "epoch": 0.39, "learning_rate": 6.0965826161356886e-05, "loss": 1.0215, "step": 156175 }, { "epoch": 0.39, "learning_rate": 6.0964567919875184e-05, "loss": 1.0204, "step": 156180 }, { "epoch": 0.39, "learning_rate": 6.0963309678393476e-05, "loss": 1.0221, "step": 156185 }, { "epoch": 0.39, "learning_rate": 6.0962051436911774e-05, "loss": 1.0203, "step": 156190 }, { "epoch": 0.39, "learning_rate": 6.0960793195430066e-05, "loss": 1.0202, "step": 156195 }, { "epoch": 0.39, "learning_rate": 6.095953495394836e-05, "loss": 1.0211, "step": 156200 }, { "epoch": 0.39, "learning_rate": 6.0958276712466656e-05, "loss": 1.0405, "step": 156205 }, { "epoch": 0.39, "learning_rate": 6.095701847098495e-05, "loss": 1.0191, "step": 156210 }, { "epoch": 0.39, "learning_rate": 6.0955760229503246e-05, "loss": 1.0226, "step": 156215 }, { "epoch": 0.39, "learning_rate": 6.095450198802154e-05, "loss": 1.02, "step": 156220 }, { "epoch": 0.39, "learning_rate": 6.0953243746539836e-05, "loss": 1.0213, "step": 156225 }, { "epoch": 0.39, "learning_rate": 6.095198550505813e-05, "loss": 1.019, "step": 156230 }, { "epoch": 0.39, "learning_rate": 6.0950727263576426e-05, "loss": 1.0213, "step": 156235 }, { "epoch": 0.39, "learning_rate": 6.094946902209472e-05, "loss": 1.0216, "step": 156240 }, { "epoch": 0.39, "learning_rate": 6.0948210780613016e-05, "loss": 1.0199, "step": 156245 }, { "epoch": 0.39, "learning_rate": 6.094695253913131e-05, "loss": 1.0193, "step": 156250 }, { "epoch": 0.39, "learning_rate": 6.0945694297649606e-05, "loss": 1.0194, "step": 156255 }, { "epoch": 0.39, "learning_rate": 6.09444360561679e-05, "loss": 1.0208, "step": 156260 }, { "epoch": 0.39, "learning_rate": 6.094317781468619e-05, "loss": 1.0208, "step": 156265 }, { "epoch": 0.39, "learning_rate": 6.09419195732045e-05, "loss": 1.0196, "step": 156270 }, { "epoch": 0.39, "learning_rate": 6.094066133172279e-05, "loss": 1.0219, "step": 156275 }, { "epoch": 0.39, "learning_rate": 6.093940309024109e-05, "loss": 1.0207, "step": 156280 }, { "epoch": 0.39, "learning_rate": 6.093814484875938e-05, "loss": 1.0215, "step": 156285 }, { "epoch": 0.39, "learning_rate": 6.0936886607277674e-05, "loss": 1.0196, "step": 156290 }, { "epoch": 0.39, "learning_rate": 6.093562836579597e-05, "loss": 1.0213, "step": 156295 }, { "epoch": 0.39, "learning_rate": 6.0934370124314264e-05, "loss": 1.0207, "step": 156300 }, { "epoch": 0.39, "learning_rate": 6.093311188283256e-05, "loss": 1.02, "step": 156305 }, { "epoch": 0.39, "learning_rate": 6.0931853641350854e-05, "loss": 1.0189, "step": 156310 }, { "epoch": 0.39, "learning_rate": 6.093059539986915e-05, "loss": 1.0201, "step": 156315 }, { "epoch": 0.39, "learning_rate": 6.0929337158387444e-05, "loss": 1.018, "step": 156320 }, { "epoch": 0.39, "learning_rate": 6.092807891690574e-05, "loss": 1.0215, "step": 156325 }, { "epoch": 0.39, "learning_rate": 6.0926820675424034e-05, "loss": 1.0203, "step": 156330 }, { "epoch": 0.39, "learning_rate": 6.092556243394233e-05, "loss": 1.0188, "step": 156335 }, { "epoch": 0.39, "learning_rate": 6.0924304192460624e-05, "loss": 1.0202, "step": 156340 }, { "epoch": 0.39, "learning_rate": 6.0923045950978915e-05, "loss": 1.0207, "step": 156345 }, { "epoch": 0.39, "learning_rate": 6.0921787709497214e-05, "loss": 1.0214, "step": 156350 }, { "epoch": 0.39, "learning_rate": 6.0920529468015505e-05, "loss": 1.0232, "step": 156355 }, { "epoch": 0.39, "learning_rate": 6.0919271226533804e-05, "loss": 1.0186, "step": 156360 }, { "epoch": 0.39, "learning_rate": 6.0918012985052095e-05, "loss": 1.0222, "step": 156365 }, { "epoch": 0.39, "learning_rate": 6.0916754743570394e-05, "loss": 1.0217, "step": 156370 }, { "epoch": 0.39, "learning_rate": 6.0915496502088685e-05, "loss": 1.0187, "step": 156375 }, { "epoch": 0.39, "learning_rate": 6.0914238260606984e-05, "loss": 1.0235, "step": 156380 }, { "epoch": 0.39, "learning_rate": 6.0912980019125275e-05, "loss": 1.0205, "step": 156385 }, { "epoch": 0.39, "learning_rate": 6.0911721777643574e-05, "loss": 1.0211, "step": 156390 }, { "epoch": 0.39, "learning_rate": 6.0910463536161865e-05, "loss": 1.0201, "step": 156395 }, { "epoch": 0.39, "learning_rate": 6.0909205294680164e-05, "loss": 1.0201, "step": 156400 }, { "epoch": 0.39, "learning_rate": 6.0907947053198455e-05, "loss": 1.0212, "step": 156405 }, { "epoch": 0.39, "learning_rate": 6.090668881171675e-05, "loss": 1.0188, "step": 156410 }, { "epoch": 0.39, "learning_rate": 6.0905430570235045e-05, "loss": 1.0204, "step": 156415 }, { "epoch": 0.39, "learning_rate": 6.090417232875334e-05, "loss": 1.0185, "step": 156420 }, { "epoch": 0.39, "learning_rate": 6.0902914087271635e-05, "loss": 1.0213, "step": 156425 }, { "epoch": 0.39, "learning_rate": 6.090165584578993e-05, "loss": 1.0194, "step": 156430 }, { "epoch": 0.39, "learning_rate": 6.0900397604308225e-05, "loss": 1.0208, "step": 156435 }, { "epoch": 0.39, "learning_rate": 6.089913936282652e-05, "loss": 1.0176, "step": 156440 }, { "epoch": 0.39, "learning_rate": 6.0897881121344815e-05, "loss": 1.0221, "step": 156445 }, { "epoch": 0.39, "learning_rate": 6.0896622879863107e-05, "loss": 1.0198, "step": 156450 }, { "epoch": 0.39, "learning_rate": 6.0895364638381405e-05, "loss": 1.0201, "step": 156455 }, { "epoch": 0.39, "learning_rate": 6.0894106396899697e-05, "loss": 1.0212, "step": 156460 }, { "epoch": 0.39, "learning_rate": 6.0892848155417995e-05, "loss": 1.0188, "step": 156465 }, { "epoch": 0.39, "learning_rate": 6.0891589913936287e-05, "loss": 1.0198, "step": 156470 }, { "epoch": 0.39, "learning_rate": 6.089033167245458e-05, "loss": 1.0219, "step": 156475 }, { "epoch": 0.39, "learning_rate": 6.0889073430972877e-05, "loss": 1.0196, "step": 156480 }, { "epoch": 0.39, "learning_rate": 6.088781518949117e-05, "loss": 1.0193, "step": 156485 }, { "epoch": 0.39, "learning_rate": 6.0886556948009466e-05, "loss": 1.0444, "step": 156490 }, { "epoch": 0.39, "learning_rate": 6.088529870652776e-05, "loss": 1.0199, "step": 156495 }, { "epoch": 0.39, "learning_rate": 6.0884040465046056e-05, "loss": 1.0203, "step": 156500 }, { "epoch": 0.39, "learning_rate": 6.088278222356435e-05, "loss": 1.0207, "step": 156505 }, { "epoch": 0.39, "learning_rate": 6.0881523982082646e-05, "loss": 1.0197, "step": 156510 }, { "epoch": 0.39, "learning_rate": 6.088026574060094e-05, "loss": 1.0415, "step": 156515 }, { "epoch": 0.39, "learning_rate": 6.0879007499119236e-05, "loss": 1.0216, "step": 156520 }, { "epoch": 0.39, "learning_rate": 6.087774925763753e-05, "loss": 1.0194, "step": 156525 }, { "epoch": 0.39, "learning_rate": 6.0876491016155826e-05, "loss": 1.0219, "step": 156530 }, { "epoch": 0.39, "learning_rate": 6.087523277467412e-05, "loss": 1.0207, "step": 156535 }, { "epoch": 0.39, "learning_rate": 6.087397453319241e-05, "loss": 1.0207, "step": 156540 }, { "epoch": 0.39, "learning_rate": 6.087271629171071e-05, "loss": 1.017, "step": 156545 }, { "epoch": 0.39, "learning_rate": 6.0871458050229e-05, "loss": 1.018, "step": 156550 }, { "epoch": 0.39, "learning_rate": 6.08701998087473e-05, "loss": 1.0214, "step": 156555 }, { "epoch": 0.39, "learning_rate": 6.086894156726559e-05, "loss": 1.0211, "step": 156560 }, { "epoch": 0.39, "learning_rate": 6.086768332578389e-05, "loss": 1.021, "step": 156565 }, { "epoch": 0.39, "learning_rate": 6.086642508430218e-05, "loss": 1.0189, "step": 156570 }, { "epoch": 0.39, "learning_rate": 6.086516684282048e-05, "loss": 1.0181, "step": 156575 }, { "epoch": 0.39, "learning_rate": 6.086390860133877e-05, "loss": 1.0206, "step": 156580 }, { "epoch": 0.39, "learning_rate": 6.086265035985707e-05, "loss": 1.0196, "step": 156585 }, { "epoch": 0.39, "learning_rate": 6.086139211837536e-05, "loss": 1.0191, "step": 156590 }, { "epoch": 0.39, "learning_rate": 6.086013387689366e-05, "loss": 1.0199, "step": 156595 }, { "epoch": 0.39, "learning_rate": 6.085887563541195e-05, "loss": 1.0186, "step": 156600 }, { "epoch": 0.39, "learning_rate": 6.085761739393024e-05, "loss": 1.0204, "step": 156605 }, { "epoch": 0.39, "learning_rate": 6.085635915244854e-05, "loss": 1.0188, "step": 156610 }, { "epoch": 0.39, "learning_rate": 6.085510091096683e-05, "loss": 1.0202, "step": 156615 }, { "epoch": 0.39, "learning_rate": 6.085384266948513e-05, "loss": 1.0208, "step": 156620 }, { "epoch": 0.39, "learning_rate": 6.085258442800342e-05, "loss": 1.0219, "step": 156625 }, { "epoch": 0.39, "learning_rate": 6.085132618652172e-05, "loss": 1.0225, "step": 156630 }, { "epoch": 0.39, "learning_rate": 6.085006794504001e-05, "loss": 1.0224, "step": 156635 }, { "epoch": 0.39, "learning_rate": 6.084880970355831e-05, "loss": 1.021, "step": 156640 }, { "epoch": 0.39, "learning_rate": 6.08475514620766e-05, "loss": 1.0152, "step": 156645 }, { "epoch": 0.39, "learning_rate": 6.08462932205949e-05, "loss": 1.0194, "step": 156650 }, { "epoch": 0.39, "learning_rate": 6.084503497911319e-05, "loss": 1.0215, "step": 156655 }, { "epoch": 0.39, "learning_rate": 6.084377673763149e-05, "loss": 1.0202, "step": 156660 }, { "epoch": 0.39, "learning_rate": 6.084251849614978e-05, "loss": 1.0194, "step": 156665 }, { "epoch": 0.39, "learning_rate": 6.084126025466807e-05, "loss": 1.0197, "step": 156670 }, { "epoch": 0.39, "learning_rate": 6.084000201318637e-05, "loss": 1.0223, "step": 156675 }, { "epoch": 0.39, "learning_rate": 6.083874377170466e-05, "loss": 1.0204, "step": 156680 }, { "epoch": 0.39, "learning_rate": 6.083748553022296e-05, "loss": 1.0199, "step": 156685 }, { "epoch": 0.39, "learning_rate": 6.083622728874125e-05, "loss": 1.0188, "step": 156690 }, { "epoch": 0.39, "learning_rate": 6.083496904725955e-05, "loss": 1.0226, "step": 156695 }, { "epoch": 0.39, "learning_rate": 6.083371080577784e-05, "loss": 1.0175, "step": 156700 }, { "epoch": 0.39, "learning_rate": 6.083245256429614e-05, "loss": 1.0196, "step": 156705 }, { "epoch": 0.39, "learning_rate": 6.083119432281443e-05, "loss": 1.0206, "step": 156710 }, { "epoch": 0.39, "learning_rate": 6.082993608133273e-05, "loss": 1.0179, "step": 156715 }, { "epoch": 0.39, "learning_rate": 6.082867783985102e-05, "loss": 1.0216, "step": 156720 }, { "epoch": 0.39, "learning_rate": 6.082741959836932e-05, "loss": 1.0178, "step": 156725 }, { "epoch": 0.39, "learning_rate": 6.082616135688761e-05, "loss": 1.0175, "step": 156730 }, { "epoch": 0.39, "learning_rate": 6.0824903115405904e-05, "loss": 1.0177, "step": 156735 }, { "epoch": 0.39, "learning_rate": 6.08236448739242e-05, "loss": 1.0188, "step": 156740 }, { "epoch": 0.39, "learning_rate": 6.0822386632442494e-05, "loss": 1.0199, "step": 156745 }, { "epoch": 0.39, "learning_rate": 6.082112839096079e-05, "loss": 1.0201, "step": 156750 }, { "epoch": 0.39, "learning_rate": 6.0819870149479084e-05, "loss": 1.0209, "step": 156755 }, { "epoch": 0.39, "learning_rate": 6.081861190799738e-05, "loss": 1.0185, "step": 156760 }, { "epoch": 0.39, "learning_rate": 6.0817353666515674e-05, "loss": 1.019, "step": 156765 }, { "epoch": 0.39, "learning_rate": 6.081609542503398e-05, "loss": 1.0206, "step": 156770 }, { "epoch": 0.39, "learning_rate": 6.081483718355228e-05, "loss": 1.0195, "step": 156775 }, { "epoch": 0.39, "learning_rate": 6.081357894207057e-05, "loss": 1.0231, "step": 156780 }, { "epoch": 0.39, "learning_rate": 6.081232070058887e-05, "loss": 1.0182, "step": 156785 }, { "epoch": 0.39, "learning_rate": 6.081106245910716e-05, "loss": 1.0194, "step": 156790 }, { "epoch": 0.39, "learning_rate": 6.080980421762546e-05, "loss": 1.0213, "step": 156795 }, { "epoch": 0.39, "learning_rate": 6.080854597614375e-05, "loss": 1.0192, "step": 156800 }, { "epoch": 0.39, "learning_rate": 6.080728773466205e-05, "loss": 1.0194, "step": 156805 }, { "epoch": 0.39, "learning_rate": 6.080602949318034e-05, "loss": 1.021, "step": 156810 }, { "epoch": 0.39, "learning_rate": 6.080477125169863e-05, "loss": 1.0211, "step": 156815 }, { "epoch": 0.39, "learning_rate": 6.080351301021693e-05, "loss": 1.0205, "step": 156820 }, { "epoch": 0.39, "learning_rate": 6.080225476873522e-05, "loss": 1.0191, "step": 156825 }, { "epoch": 0.39, "learning_rate": 6.080099652725352e-05, "loss": 1.0177, "step": 156830 }, { "epoch": 0.39, "learning_rate": 6.079973828577181e-05, "loss": 1.0199, "step": 156835 }, { "epoch": 0.39, "learning_rate": 6.079848004429011e-05, "loss": 1.0206, "step": 156840 }, { "epoch": 0.39, "learning_rate": 6.07972218028084e-05, "loss": 1.019, "step": 156845 }, { "epoch": 0.39, "learning_rate": 6.07959635613267e-05, "loss": 1.0169, "step": 156850 }, { "epoch": 0.39, "learning_rate": 6.079470531984499e-05, "loss": 1.0211, "step": 156855 }, { "epoch": 0.39, "learning_rate": 6.079344707836329e-05, "loss": 1.0212, "step": 156860 }, { "epoch": 0.39, "learning_rate": 6.079218883688158e-05, "loss": 1.0216, "step": 156865 }, { "epoch": 0.39, "learning_rate": 6.079093059539988e-05, "loss": 1.0214, "step": 156870 }, { "epoch": 0.39, "learning_rate": 6.078967235391817e-05, "loss": 1.0353, "step": 156875 }, { "epoch": 0.39, "learning_rate": 6.078841411243646e-05, "loss": 1.0203, "step": 156880 }, { "epoch": 0.39, "learning_rate": 6.078715587095476e-05, "loss": 1.0175, "step": 156885 }, { "epoch": 0.39, "learning_rate": 6.078589762947305e-05, "loss": 1.0195, "step": 156890 }, { "epoch": 0.39, "learning_rate": 6.078463938799135e-05, "loss": 1.0217, "step": 156895 }, { "epoch": 0.39, "learning_rate": 6.078338114650964e-05, "loss": 1.0212, "step": 156900 }, { "epoch": 0.39, "learning_rate": 6.078212290502794e-05, "loss": 1.0208, "step": 156905 }, { "epoch": 0.39, "learning_rate": 6.078086466354623e-05, "loss": 1.0212, "step": 156910 }, { "epoch": 0.39, "learning_rate": 6.077960642206453e-05, "loss": 1.019, "step": 156915 }, { "epoch": 0.39, "learning_rate": 6.077834818058282e-05, "loss": 1.0185, "step": 156920 }, { "epoch": 0.39, "learning_rate": 6.077708993910112e-05, "loss": 1.0203, "step": 156925 }, { "epoch": 0.39, "learning_rate": 6.077583169761941e-05, "loss": 1.0199, "step": 156930 }, { "epoch": 0.39, "learning_rate": 6.077457345613771e-05, "loss": 1.0181, "step": 156935 }, { "epoch": 0.39, "learning_rate": 6.0773315214656e-05, "loss": 1.0189, "step": 156940 }, { "epoch": 0.39, "learning_rate": 6.077205697317429e-05, "loss": 1.0181, "step": 156945 }, { "epoch": 0.39, "learning_rate": 6.077079873169259e-05, "loss": 1.0225, "step": 156950 }, { "epoch": 0.39, "learning_rate": 6.076954049021088e-05, "loss": 1.0188, "step": 156955 }, { "epoch": 0.39, "learning_rate": 6.076828224872918e-05, "loss": 1.0211, "step": 156960 }, { "epoch": 0.39, "learning_rate": 6.076702400724747e-05, "loss": 1.0202, "step": 156965 }, { "epoch": 0.39, "learning_rate": 6.076576576576577e-05, "loss": 1.0228, "step": 156970 }, { "epoch": 0.39, "learning_rate": 6.076450752428406e-05, "loss": 1.0192, "step": 156975 }, { "epoch": 0.39, "learning_rate": 6.076324928280236e-05, "loss": 1.0181, "step": 156980 }, { "epoch": 0.39, "learning_rate": 6.076199104132065e-05, "loss": 1.0188, "step": 156985 }, { "epoch": 0.39, "learning_rate": 6.076073279983895e-05, "loss": 1.0201, "step": 156990 }, { "epoch": 0.39, "learning_rate": 6.075947455835724e-05, "loss": 1.0209, "step": 156995 }, { "epoch": 0.39, "learning_rate": 6.0758216316875535e-05, "loss": 1.0182, "step": 157000 }, { "epoch": 0.39, "learning_rate": 6.075695807539383e-05, "loss": 1.0196, "step": 157005 }, { "epoch": 0.39, "learning_rate": 6.0755699833912125e-05, "loss": 1.0197, "step": 157010 }, { "epoch": 0.39, "learning_rate": 6.075444159243042e-05, "loss": 1.0205, "step": 157015 }, { "epoch": 0.39, "learning_rate": 6.0753183350948715e-05, "loss": 1.0201, "step": 157020 }, { "epoch": 0.39, "learning_rate": 6.075192510946701e-05, "loss": 1.0191, "step": 157025 }, { "epoch": 0.39, "learning_rate": 6.0750666867985305e-05, "loss": 1.0203, "step": 157030 }, { "epoch": 0.39, "learning_rate": 6.07494086265036e-05, "loss": 1.0239, "step": 157035 }, { "epoch": 0.39, "learning_rate": 6.0748150385021895e-05, "loss": 1.0211, "step": 157040 }, { "epoch": 0.39, "learning_rate": 6.074689214354019e-05, "loss": 1.0191, "step": 157045 }, { "epoch": 0.39, "learning_rate": 6.0745633902058485e-05, "loss": 1.021, "step": 157050 }, { "epoch": 0.39, "learning_rate": 6.074437566057678e-05, "loss": 1.0219, "step": 157055 }, { "epoch": 0.39, "learning_rate": 6.0743117419095075e-05, "loss": 1.0207, "step": 157060 }, { "epoch": 0.39, "learning_rate": 6.0741859177613366e-05, "loss": 1.0203, "step": 157065 }, { "epoch": 0.39, "learning_rate": 6.0740600936131664e-05, "loss": 1.0195, "step": 157070 }, { "epoch": 0.39, "learning_rate": 6.0739342694649956e-05, "loss": 1.0214, "step": 157075 }, { "epoch": 0.39, "learning_rate": 6.0738084453168254e-05, "loss": 1.02, "step": 157080 }, { "epoch": 0.39, "learning_rate": 6.0736826211686546e-05, "loss": 1.0203, "step": 157085 }, { "epoch": 0.39, "learning_rate": 6.0735567970204844e-05, "loss": 1.0211, "step": 157090 }, { "epoch": 0.39, "learning_rate": 6.0734309728723136e-05, "loss": 1.0212, "step": 157095 }, { "epoch": 0.39, "learning_rate": 6.0733051487241434e-05, "loss": 1.017, "step": 157100 }, { "epoch": 0.39, "learning_rate": 6.0731793245759726e-05, "loss": 1.0202, "step": 157105 }, { "epoch": 0.39, "learning_rate": 6.0730535004278024e-05, "loss": 1.0222, "step": 157110 }, { "epoch": 0.39, "learning_rate": 6.0729276762796316e-05, "loss": 1.0212, "step": 157115 }, { "epoch": 0.39, "learning_rate": 6.0728018521314614e-05, "loss": 1.0213, "step": 157120 }, { "epoch": 0.39, "learning_rate": 6.0726760279832906e-05, "loss": 1.019, "step": 157125 }, { "epoch": 0.39, "learning_rate": 6.07255020383512e-05, "loss": 1.0192, "step": 157130 }, { "epoch": 0.39, "learning_rate": 6.0724243796869496e-05, "loss": 1.0216, "step": 157135 }, { "epoch": 0.39, "learning_rate": 6.072298555538779e-05, "loss": 1.0205, "step": 157140 }, { "epoch": 0.39, "learning_rate": 6.0721727313906086e-05, "loss": 1.0214, "step": 157145 }, { "epoch": 0.39, "learning_rate": 6.072046907242438e-05, "loss": 1.0198, "step": 157150 }, { "epoch": 0.39, "learning_rate": 6.0719210830942676e-05, "loss": 1.0181, "step": 157155 }, { "epoch": 0.39, "learning_rate": 6.071795258946097e-05, "loss": 1.0195, "step": 157160 }, { "epoch": 0.39, "learning_rate": 6.0716694347979266e-05, "loss": 1.0206, "step": 157165 }, { "epoch": 0.39, "learning_rate": 6.071543610649756e-05, "loss": 1.0172, "step": 157170 }, { "epoch": 0.39, "learning_rate": 6.0714177865015856e-05, "loss": 1.0192, "step": 157175 }, { "epoch": 0.39, "learning_rate": 6.071291962353415e-05, "loss": 1.0182, "step": 157180 }, { "epoch": 0.39, "learning_rate": 6.0711661382052446e-05, "loss": 1.034, "step": 157185 }, { "epoch": 0.39, "learning_rate": 6.071040314057074e-05, "loss": 1.0207, "step": 157190 }, { "epoch": 0.39, "learning_rate": 6.070914489908903e-05, "loss": 1.0192, "step": 157195 }, { "epoch": 0.39, "learning_rate": 6.070788665760733e-05, "loss": 1.0196, "step": 157200 }, { "epoch": 0.39, "learning_rate": 6.070662841612562e-05, "loss": 1.0191, "step": 157205 }, { "epoch": 0.39, "learning_rate": 6.070537017464392e-05, "loss": 1.0402, "step": 157210 }, { "epoch": 0.39, "learning_rate": 6.070411193316221e-05, "loss": 1.0204, "step": 157215 }, { "epoch": 0.39, "learning_rate": 6.070285369168051e-05, "loss": 1.0408, "step": 157220 }, { "epoch": 0.39, "learning_rate": 6.07015954501988e-05, "loss": 1.0178, "step": 157225 }, { "epoch": 0.39, "learning_rate": 6.07003372087171e-05, "loss": 1.0206, "step": 157230 }, { "epoch": 0.39, "learning_rate": 6.069907896723539e-05, "loss": 1.0226, "step": 157235 }, { "epoch": 0.39, "learning_rate": 6.069782072575369e-05, "loss": 1.0224, "step": 157240 }, { "epoch": 0.39, "learning_rate": 6.069656248427198e-05, "loss": 1.0189, "step": 157245 }, { "epoch": 0.39, "learning_rate": 6.069530424279028e-05, "loss": 1.023, "step": 157250 }, { "epoch": 0.39, "learning_rate": 6.069404600130857e-05, "loss": 1.02, "step": 157255 }, { "epoch": 0.39, "learning_rate": 6.069278775982686e-05, "loss": 1.0194, "step": 157260 }, { "epoch": 0.39, "learning_rate": 6.069152951834516e-05, "loss": 1.0197, "step": 157265 }, { "epoch": 0.39, "learning_rate": 6.0690271276863464e-05, "loss": 1.0213, "step": 157270 }, { "epoch": 0.39, "learning_rate": 6.0689013035381756e-05, "loss": 1.0221, "step": 157275 }, { "epoch": 0.39, "learning_rate": 6.0687754793900054e-05, "loss": 1.0211, "step": 157280 }, { "epoch": 0.39, "learning_rate": 6.0686496552418345e-05, "loss": 1.0224, "step": 157285 }, { "epoch": 0.39, "learning_rate": 6.0685238310936644e-05, "loss": 1.0203, "step": 157290 }, { "epoch": 0.39, "learning_rate": 6.0683980069454935e-05, "loss": 1.0193, "step": 157295 }, { "epoch": 0.39, "learning_rate": 6.0682721827973234e-05, "loss": 1.0189, "step": 157300 }, { "epoch": 0.39, "learning_rate": 6.0681463586491525e-05, "loss": 1.0222, "step": 157305 }, { "epoch": 0.39, "learning_rate": 6.0680205345009824e-05, "loss": 1.0201, "step": 157310 }, { "epoch": 0.39, "learning_rate": 6.0678947103528115e-05, "loss": 1.0182, "step": 157315 }, { "epoch": 0.39, "learning_rate": 6.0677688862046414e-05, "loss": 1.0191, "step": 157320 }, { "epoch": 0.39, "learning_rate": 6.0676430620564705e-05, "loss": 1.0196, "step": 157325 }, { "epoch": 0.39, "learning_rate": 6.0675172379083004e-05, "loss": 1.0202, "step": 157330 }, { "epoch": 0.39, "learning_rate": 6.0673914137601295e-05, "loss": 1.0212, "step": 157335 }, { "epoch": 0.39, "learning_rate": 6.067265589611959e-05, "loss": 1.0191, "step": 157340 }, { "epoch": 0.39, "learning_rate": 6.0671397654637885e-05, "loss": 1.0189, "step": 157345 }, { "epoch": 0.39, "learning_rate": 6.067013941315618e-05, "loss": 1.0192, "step": 157350 }, { "epoch": 0.39, "learning_rate": 6.0668881171674475e-05, "loss": 1.0196, "step": 157355 }, { "epoch": 0.39, "learning_rate": 6.066762293019277e-05, "loss": 1.0198, "step": 157360 }, { "epoch": 0.4, "learning_rate": 6.0666364688711065e-05, "loss": 1.022, "step": 157365 }, { "epoch": 0.4, "learning_rate": 6.066510644722936e-05, "loss": 1.0181, "step": 157370 }, { "epoch": 0.4, "learning_rate": 6.0663848205747655e-05, "loss": 1.0208, "step": 157375 }, { "epoch": 0.4, "learning_rate": 6.066258996426595e-05, "loss": 1.0193, "step": 157380 }, { "epoch": 0.4, "learning_rate": 6.0661331722784245e-05, "loss": 1.0223, "step": 157385 }, { "epoch": 0.4, "learning_rate": 6.066007348130254e-05, "loss": 1.0207, "step": 157390 }, { "epoch": 0.4, "learning_rate": 6.0658815239820835e-05, "loss": 1.0194, "step": 157395 }, { "epoch": 0.4, "learning_rate": 6.065755699833913e-05, "loss": 1.0194, "step": 157400 }, { "epoch": 0.4, "learning_rate": 6.065629875685742e-05, "loss": 1.0203, "step": 157405 }, { "epoch": 0.4, "learning_rate": 6.065504051537572e-05, "loss": 1.0184, "step": 157410 }, { "epoch": 0.4, "learning_rate": 6.065378227389401e-05, "loss": 1.0222, "step": 157415 }, { "epoch": 0.4, "learning_rate": 6.065252403241231e-05, "loss": 1.0202, "step": 157420 }, { "epoch": 0.4, "learning_rate": 6.06512657909306e-05, "loss": 1.019, "step": 157425 }, { "epoch": 0.4, "learning_rate": 6.06500075494489e-05, "loss": 1.0219, "step": 157430 }, { "epoch": 0.4, "learning_rate": 6.064874930796719e-05, "loss": 1.0222, "step": 157435 }, { "epoch": 0.4, "learning_rate": 6.064749106648549e-05, "loss": 1.0195, "step": 157440 }, { "epoch": 0.4, "learning_rate": 6.064623282500378e-05, "loss": 1.0221, "step": 157445 }, { "epoch": 0.4, "learning_rate": 6.064497458352208e-05, "loss": 1.021, "step": 157450 }, { "epoch": 0.4, "learning_rate": 6.064371634204037e-05, "loss": 1.0194, "step": 157455 }, { "epoch": 0.4, "learning_rate": 6.0642458100558667e-05, "loss": 1.0211, "step": 157460 }, { "epoch": 0.4, "learning_rate": 6.064119985907696e-05, "loss": 1.02, "step": 157465 }, { "epoch": 0.4, "learning_rate": 6.063994161759525e-05, "loss": 1.0398, "step": 157470 }, { "epoch": 0.4, "learning_rate": 6.063868337611355e-05, "loss": 1.0197, "step": 157475 }, { "epoch": 0.4, "learning_rate": 6.063742513463184e-05, "loss": 1.0211, "step": 157480 }, { "epoch": 0.4, "learning_rate": 6.063616689315014e-05, "loss": 1.0208, "step": 157485 }, { "epoch": 0.4, "learning_rate": 6.063490865166843e-05, "loss": 1.022, "step": 157490 }, { "epoch": 0.4, "learning_rate": 6.063365041018673e-05, "loss": 1.0201, "step": 157495 }, { "epoch": 0.4, "learning_rate": 6.063239216870502e-05, "loss": 1.021, "step": 157500 }, { "epoch": 0.4, "learning_rate": 6.063113392722332e-05, "loss": 1.0216, "step": 157505 }, { "epoch": 0.4, "learning_rate": 6.062987568574161e-05, "loss": 1.0405, "step": 157510 }, { "epoch": 0.4, "learning_rate": 6.062861744425991e-05, "loss": 1.0173, "step": 157515 }, { "epoch": 0.4, "learning_rate": 6.06273592027782e-05, "loss": 1.0194, "step": 157520 }, { "epoch": 0.4, "learning_rate": 6.06261009612965e-05, "loss": 1.0202, "step": 157525 }, { "epoch": 0.4, "learning_rate": 6.062484271981479e-05, "loss": 1.0428, "step": 157530 }, { "epoch": 0.4, "learning_rate": 6.062358447833308e-05, "loss": 1.022, "step": 157535 }, { "epoch": 0.4, "learning_rate": 6.062232623685138e-05, "loss": 1.0213, "step": 157540 }, { "epoch": 0.4, "learning_rate": 6.062106799536967e-05, "loss": 1.0204, "step": 157545 }, { "epoch": 0.4, "learning_rate": 6.061980975388797e-05, "loss": 1.0201, "step": 157550 }, { "epoch": 0.4, "learning_rate": 6.061855151240626e-05, "loss": 1.0182, "step": 157555 }, { "epoch": 0.4, "learning_rate": 6.061729327092456e-05, "loss": 1.0172, "step": 157560 }, { "epoch": 0.4, "learning_rate": 6.061603502944285e-05, "loss": 1.0214, "step": 157565 }, { "epoch": 0.4, "learning_rate": 6.061477678796115e-05, "loss": 1.0208, "step": 157570 }, { "epoch": 0.4, "learning_rate": 6.061351854647944e-05, "loss": 1.0209, "step": 157575 }, { "epoch": 0.4, "learning_rate": 6.061226030499774e-05, "loss": 1.022, "step": 157580 }, { "epoch": 0.4, "learning_rate": 6.061100206351603e-05, "loss": 1.0185, "step": 157585 }, { "epoch": 0.4, "learning_rate": 6.060974382203433e-05, "loss": 1.0189, "step": 157590 }, { "epoch": 0.4, "learning_rate": 6.060848558055262e-05, "loss": 1.0194, "step": 157595 }, { "epoch": 0.4, "learning_rate": 6.060722733907091e-05, "loss": 1.0205, "step": 157600 }, { "epoch": 0.4, "learning_rate": 6.060596909758921e-05, "loss": 1.019, "step": 157605 }, { "epoch": 0.4, "learning_rate": 6.06047108561075e-05, "loss": 1.0212, "step": 157610 }, { "epoch": 0.4, "learning_rate": 6.06034526146258e-05, "loss": 1.0204, "step": 157615 }, { "epoch": 0.4, "learning_rate": 6.060219437314409e-05, "loss": 1.0199, "step": 157620 }, { "epoch": 0.4, "learning_rate": 6.060093613166239e-05, "loss": 1.0195, "step": 157625 }, { "epoch": 0.4, "learning_rate": 6.059967789018068e-05, "loss": 1.0203, "step": 157630 }, { "epoch": 0.4, "learning_rate": 6.059841964869898e-05, "loss": 1.02, "step": 157635 }, { "epoch": 0.4, "learning_rate": 6.059716140721727e-05, "loss": 1.0204, "step": 157640 }, { "epoch": 0.4, "learning_rate": 6.059590316573557e-05, "loss": 1.0196, "step": 157645 }, { "epoch": 0.4, "learning_rate": 6.059464492425386e-05, "loss": 1.0187, "step": 157650 }, { "epoch": 0.4, "learning_rate": 6.0593386682772154e-05, "loss": 1.0197, "step": 157655 }, { "epoch": 0.4, "learning_rate": 6.059212844129045e-05, "loss": 1.0217, "step": 157660 }, { "epoch": 0.4, "learning_rate": 6.0590870199808744e-05, "loss": 1.019, "step": 157665 }, { "epoch": 0.4, "learning_rate": 6.058961195832704e-05, "loss": 1.0218, "step": 157670 }, { "epoch": 0.4, "learning_rate": 6.0588353716845334e-05, "loss": 1.0189, "step": 157675 }, { "epoch": 0.4, "learning_rate": 6.058709547536363e-05, "loss": 1.0193, "step": 157680 }, { "epoch": 0.4, "learning_rate": 6.0585837233881924e-05, "loss": 1.0188, "step": 157685 }, { "epoch": 0.4, "learning_rate": 6.058457899240022e-05, "loss": 1.0165, "step": 157690 }, { "epoch": 0.4, "learning_rate": 6.0583320750918514e-05, "loss": 1.0201, "step": 157695 }, { "epoch": 0.4, "learning_rate": 6.058206250943681e-05, "loss": 1.0216, "step": 157700 }, { "epoch": 0.4, "learning_rate": 6.0580804267955104e-05, "loss": 1.0198, "step": 157705 }, { "epoch": 0.4, "learning_rate": 6.05795460264734e-05, "loss": 1.0206, "step": 157710 }, { "epoch": 0.4, "learning_rate": 6.0578287784991694e-05, "loss": 1.0188, "step": 157715 }, { "epoch": 0.4, "learning_rate": 6.0577029543509986e-05, "loss": 1.0205, "step": 157720 }, { "epoch": 0.4, "learning_rate": 6.0575771302028284e-05, "loss": 1.0195, "step": 157725 }, { "epoch": 0.4, "learning_rate": 6.0574513060546575e-05, "loss": 1.0187, "step": 157730 }, { "epoch": 0.4, "learning_rate": 6.0573254819064874e-05, "loss": 1.0201, "step": 157735 }, { "epoch": 0.4, "learning_rate": 6.0571996577583165e-05, "loss": 1.0184, "step": 157740 }, { "epoch": 0.4, "learning_rate": 6.0570738336101464e-05, "loss": 1.0185, "step": 157745 }, { "epoch": 0.4, "learning_rate": 6.0569480094619755e-05, "loss": 1.0404, "step": 157750 }, { "epoch": 0.4, "learning_rate": 6.0568221853138054e-05, "loss": 1.0205, "step": 157755 }, { "epoch": 0.4, "learning_rate": 6.0566963611656345e-05, "loss": 1.0208, "step": 157760 }, { "epoch": 0.4, "learning_rate": 6.0565705370174644e-05, "loss": 1.0203, "step": 157765 }, { "epoch": 0.4, "learning_rate": 6.056444712869295e-05, "loss": 1.021, "step": 157770 }, { "epoch": 0.4, "learning_rate": 6.056318888721124e-05, "loss": 1.0204, "step": 157775 }, { "epoch": 0.4, "learning_rate": 6.056193064572954e-05, "loss": 1.0198, "step": 157780 }, { "epoch": 0.4, "learning_rate": 6.056067240424783e-05, "loss": 1.0207, "step": 157785 }, { "epoch": 0.4, "learning_rate": 6.055941416276613e-05, "loss": 1.0202, "step": 157790 }, { "epoch": 0.4, "learning_rate": 6.055815592128442e-05, "loss": 1.0204, "step": 157795 }, { "epoch": 0.4, "learning_rate": 6.055689767980272e-05, "loss": 1.0202, "step": 157800 }, { "epoch": 0.4, "learning_rate": 6.055563943832101e-05, "loss": 1.0191, "step": 157805 }, { "epoch": 0.4, "learning_rate": 6.05543811968393e-05, "loss": 1.0207, "step": 157810 }, { "epoch": 0.4, "learning_rate": 6.05531229553576e-05, "loss": 1.0221, "step": 157815 }, { "epoch": 0.4, "learning_rate": 6.055186471387589e-05, "loss": 1.0192, "step": 157820 }, { "epoch": 0.4, "learning_rate": 6.055060647239419e-05, "loss": 1.021, "step": 157825 }, { "epoch": 0.4, "learning_rate": 6.054934823091248e-05, "loss": 1.0375, "step": 157830 }, { "epoch": 0.4, "learning_rate": 6.054808998943078e-05, "loss": 1.0179, "step": 157835 }, { "epoch": 0.4, "learning_rate": 6.054683174794907e-05, "loss": 1.0214, "step": 157840 }, { "epoch": 0.4, "learning_rate": 6.054557350646737e-05, "loss": 1.0215, "step": 157845 }, { "epoch": 0.4, "learning_rate": 6.054431526498566e-05, "loss": 1.0206, "step": 157850 }, { "epoch": 0.4, "learning_rate": 6.054305702350396e-05, "loss": 1.0213, "step": 157855 }, { "epoch": 0.4, "learning_rate": 6.054179878202225e-05, "loss": 1.02, "step": 157860 }, { "epoch": 0.4, "learning_rate": 6.0540540540540543e-05, "loss": 1.0179, "step": 157865 }, { "epoch": 0.4, "learning_rate": 6.053928229905884e-05, "loss": 1.0176, "step": 157870 }, { "epoch": 0.4, "learning_rate": 6.0538024057577133e-05, "loss": 1.0212, "step": 157875 }, { "epoch": 0.4, "learning_rate": 6.053676581609543e-05, "loss": 1.0201, "step": 157880 }, { "epoch": 0.4, "learning_rate": 6.0535507574613723e-05, "loss": 1.0225, "step": 157885 }, { "epoch": 0.4, "learning_rate": 6.053424933313202e-05, "loss": 1.0212, "step": 157890 }, { "epoch": 0.4, "learning_rate": 6.0532991091650313e-05, "loss": 1.0192, "step": 157895 }, { "epoch": 0.4, "learning_rate": 6.053173285016861e-05, "loss": 1.0217, "step": 157900 }, { "epoch": 0.4, "learning_rate": 6.05304746086869e-05, "loss": 1.0206, "step": 157905 }, { "epoch": 0.4, "learning_rate": 6.05292163672052e-05, "loss": 1.0169, "step": 157910 }, { "epoch": 0.4, "learning_rate": 6.052795812572349e-05, "loss": 1.0209, "step": 157915 }, { "epoch": 0.4, "learning_rate": 6.052669988424179e-05, "loss": 1.0182, "step": 157920 }, { "epoch": 0.4, "learning_rate": 6.052544164276008e-05, "loss": 1.0201, "step": 157925 }, { "epoch": 0.4, "learning_rate": 6.0524183401278375e-05, "loss": 1.0228, "step": 157930 }, { "epoch": 0.4, "learning_rate": 6.052292515979667e-05, "loss": 1.0194, "step": 157935 }, { "epoch": 0.4, "learning_rate": 6.0521666918314965e-05, "loss": 1.0197, "step": 157940 }, { "epoch": 0.4, "learning_rate": 6.052040867683326e-05, "loss": 1.0199, "step": 157945 }, { "epoch": 0.4, "learning_rate": 6.0519150435351555e-05, "loss": 1.0189, "step": 157950 }, { "epoch": 0.4, "learning_rate": 6.051789219386985e-05, "loss": 1.0214, "step": 157955 }, { "epoch": 0.4, "learning_rate": 6.0516633952388145e-05, "loss": 1.0214, "step": 157960 }, { "epoch": 0.4, "learning_rate": 6.051537571090644e-05, "loss": 1.0202, "step": 157965 }, { "epoch": 0.4, "learning_rate": 6.0514117469424735e-05, "loss": 1.0181, "step": 157970 }, { "epoch": 0.4, "learning_rate": 6.051285922794303e-05, "loss": 1.0193, "step": 157975 }, { "epoch": 0.4, "learning_rate": 6.0511600986461325e-05, "loss": 1.021, "step": 157980 }, { "epoch": 0.4, "learning_rate": 6.051034274497962e-05, "loss": 1.0205, "step": 157985 }, { "epoch": 0.4, "learning_rate": 6.0509084503497915e-05, "loss": 1.0205, "step": 157990 }, { "epoch": 0.4, "learning_rate": 6.0507826262016206e-05, "loss": 1.023, "step": 157995 }, { "epoch": 0.4, "learning_rate": 6.0506568020534505e-05, "loss": 1.0206, "step": 158000 }, { "epoch": 0.4, "learning_rate": 6.0505309779052796e-05, "loss": 1.0213, "step": 158005 }, { "epoch": 0.4, "learning_rate": 6.0504051537571095e-05, "loss": 1.0325, "step": 158010 }, { "epoch": 0.4, "learning_rate": 6.0502793296089386e-05, "loss": 1.0207, "step": 158015 }, { "epoch": 0.4, "learning_rate": 6.0501535054607685e-05, "loss": 1.0196, "step": 158020 }, { "epoch": 0.4, "learning_rate": 6.0500276813125976e-05, "loss": 1.0223, "step": 158025 }, { "epoch": 0.4, "learning_rate": 6.0499018571644275e-05, "loss": 1.0195, "step": 158030 }, { "epoch": 0.4, "learning_rate": 6.0497760330162566e-05, "loss": 1.0209, "step": 158035 }, { "epoch": 0.4, "learning_rate": 6.0496502088680865e-05, "loss": 1.0192, "step": 158040 }, { "epoch": 0.4, "learning_rate": 6.0495243847199156e-05, "loss": 1.0195, "step": 158045 }, { "epoch": 0.4, "learning_rate": 6.0493985605717455e-05, "loss": 1.0197, "step": 158050 }, { "epoch": 0.4, "learning_rate": 6.0492727364235746e-05, "loss": 1.0216, "step": 158055 }, { "epoch": 0.4, "learning_rate": 6.049146912275404e-05, "loss": 1.0197, "step": 158060 }, { "epoch": 0.4, "learning_rate": 6.0490210881272336e-05, "loss": 1.0209, "step": 158065 }, { "epoch": 0.4, "learning_rate": 6.048895263979063e-05, "loss": 1.0195, "step": 158070 }, { "epoch": 0.4, "learning_rate": 6.0487694398308926e-05, "loss": 1.0226, "step": 158075 }, { "epoch": 0.4, "learning_rate": 6.048643615682722e-05, "loss": 1.0197, "step": 158080 }, { "epoch": 0.4, "learning_rate": 6.0485177915345516e-05, "loss": 1.0216, "step": 158085 }, { "epoch": 0.4, "learning_rate": 6.048391967386381e-05, "loss": 1.0207, "step": 158090 }, { "epoch": 0.4, "learning_rate": 6.0482661432382106e-05, "loss": 1.0189, "step": 158095 }, { "epoch": 0.4, "learning_rate": 6.04814031909004e-05, "loss": 1.021, "step": 158100 }, { "epoch": 0.4, "learning_rate": 6.0480144949418696e-05, "loss": 1.0216, "step": 158105 }, { "epoch": 0.4, "learning_rate": 6.047888670793699e-05, "loss": 1.0201, "step": 158110 }, { "epoch": 0.4, "learning_rate": 6.0477628466455286e-05, "loss": 1.0208, "step": 158115 }, { "epoch": 0.4, "learning_rate": 6.047637022497358e-05, "loss": 1.0194, "step": 158120 }, { "epoch": 0.4, "learning_rate": 6.047511198349187e-05, "loss": 1.0188, "step": 158125 }, { "epoch": 0.4, "learning_rate": 6.047385374201017e-05, "loss": 1.0207, "step": 158130 }, { "epoch": 0.4, "learning_rate": 6.047259550052846e-05, "loss": 1.0208, "step": 158135 }, { "epoch": 0.4, "learning_rate": 6.047133725904676e-05, "loss": 1.0198, "step": 158140 }, { "epoch": 0.4, "learning_rate": 6.047007901756505e-05, "loss": 1.0191, "step": 158145 }, { "epoch": 0.4, "learning_rate": 6.046882077608335e-05, "loss": 1.0199, "step": 158150 }, { "epoch": 0.4, "learning_rate": 6.046756253460164e-05, "loss": 1.0179, "step": 158155 }, { "epoch": 0.4, "learning_rate": 6.046630429311994e-05, "loss": 1.0196, "step": 158160 }, { "epoch": 0.4, "learning_rate": 6.046504605163823e-05, "loss": 1.0191, "step": 158165 }, { "epoch": 0.4, "learning_rate": 6.046378781015653e-05, "loss": 1.0225, "step": 158170 }, { "epoch": 0.4, "learning_rate": 6.046252956867482e-05, "loss": 1.0194, "step": 158175 }, { "epoch": 0.4, "learning_rate": 6.046127132719312e-05, "loss": 1.0202, "step": 158180 }, { "epoch": 0.4, "learning_rate": 6.046001308571141e-05, "loss": 1.0181, "step": 158185 }, { "epoch": 0.4, "learning_rate": 6.04587548442297e-05, "loss": 1.0185, "step": 158190 }, { "epoch": 0.4, "learning_rate": 6.0457496602748e-05, "loss": 1.0186, "step": 158195 }, { "epoch": 0.4, "learning_rate": 6.045623836126629e-05, "loss": 1.018, "step": 158200 }, { "epoch": 0.4, "learning_rate": 6.045498011978459e-05, "loss": 1.0183, "step": 158205 }, { "epoch": 0.4, "learning_rate": 6.045372187830288e-05, "loss": 1.0178, "step": 158210 }, { "epoch": 0.4, "learning_rate": 6.045246363682118e-05, "loss": 1.0193, "step": 158215 }, { "epoch": 0.4, "learning_rate": 6.045120539533947e-05, "loss": 1.0214, "step": 158220 }, { "epoch": 0.4, "learning_rate": 6.044994715385777e-05, "loss": 1.0198, "step": 158225 }, { "epoch": 0.4, "learning_rate": 6.044868891237606e-05, "loss": 1.0193, "step": 158230 }, { "epoch": 0.4, "learning_rate": 6.044743067089436e-05, "loss": 1.0188, "step": 158235 }, { "epoch": 0.4, "learning_rate": 6.044617242941265e-05, "loss": 1.0189, "step": 158240 }, { "epoch": 0.4, "learning_rate": 6.044491418793095e-05, "loss": 1.0208, "step": 158245 }, { "epoch": 0.4, "learning_rate": 6.044365594644924e-05, "loss": 1.0223, "step": 158250 }, { "epoch": 0.4, "learning_rate": 6.044239770496753e-05, "loss": 1.019, "step": 158255 }, { "epoch": 0.4, "learning_rate": 6.044113946348583e-05, "loss": 1.0197, "step": 158260 }, { "epoch": 0.4, "learning_rate": 6.043988122200412e-05, "loss": 1.0201, "step": 158265 }, { "epoch": 0.4, "learning_rate": 6.043862298052243e-05, "loss": 1.0208, "step": 158270 }, { "epoch": 0.4, "learning_rate": 6.0437364739040726e-05, "loss": 1.0209, "step": 158275 }, { "epoch": 0.4, "learning_rate": 6.043610649755902e-05, "loss": 1.0204, "step": 158280 }, { "epoch": 0.4, "learning_rate": 6.0434848256077315e-05, "loss": 1.0196, "step": 158285 }, { "epoch": 0.4, "learning_rate": 6.043359001459561e-05, "loss": 1.0188, "step": 158290 }, { "epoch": 0.4, "learning_rate": 6.0432331773113905e-05, "loss": 1.0175, "step": 158295 }, { "epoch": 0.4, "learning_rate": 6.04310735316322e-05, "loss": 1.0216, "step": 158300 }, { "epoch": 0.4, "learning_rate": 6.0429815290150495e-05, "loss": 1.0217, "step": 158305 }, { "epoch": 0.4, "learning_rate": 6.042855704866879e-05, "loss": 1.0416, "step": 158310 }, { "epoch": 0.4, "learning_rate": 6.0427298807187085e-05, "loss": 1.0205, "step": 158315 }, { "epoch": 0.4, "learning_rate": 6.042604056570538e-05, "loss": 1.0209, "step": 158320 }, { "epoch": 0.4, "learning_rate": 6.0424782324223675e-05, "loss": 1.0193, "step": 158325 }, { "epoch": 0.4, "learning_rate": 6.042352408274197e-05, "loss": 1.0218, "step": 158330 }, { "epoch": 0.4, "learning_rate": 6.042226584126026e-05, "loss": 1.0199, "step": 158335 }, { "epoch": 0.4, "learning_rate": 6.042100759977856e-05, "loss": 1.0194, "step": 158340 }, { "epoch": 0.4, "learning_rate": 6.041974935829685e-05, "loss": 1.0202, "step": 158345 }, { "epoch": 0.4, "learning_rate": 6.041849111681515e-05, "loss": 1.0189, "step": 158350 }, { "epoch": 0.4, "learning_rate": 6.041723287533344e-05, "loss": 1.0187, "step": 158355 }, { "epoch": 0.4, "learning_rate": 6.041597463385174e-05, "loss": 1.0193, "step": 158360 }, { "epoch": 0.4, "learning_rate": 6.041471639237003e-05, "loss": 1.0204, "step": 158365 }, { "epoch": 0.4, "learning_rate": 6.041345815088833e-05, "loss": 1.0206, "step": 158370 }, { "epoch": 0.4, "learning_rate": 6.041219990940662e-05, "loss": 1.02, "step": 158375 }, { "epoch": 0.4, "learning_rate": 6.041094166792492e-05, "loss": 1.0205, "step": 158380 }, { "epoch": 0.4, "learning_rate": 6.040968342644321e-05, "loss": 1.0203, "step": 158385 }, { "epoch": 0.4, "learning_rate": 6.040842518496151e-05, "loss": 1.0204, "step": 158390 }, { "epoch": 0.4, "learning_rate": 6.04071669434798e-05, "loss": 1.0198, "step": 158395 }, { "epoch": 0.4, "learning_rate": 6.040590870199809e-05, "loss": 1.0184, "step": 158400 }, { "epoch": 0.4, "learning_rate": 6.040465046051639e-05, "loss": 1.0213, "step": 158405 }, { "epoch": 0.4, "learning_rate": 6.040339221903468e-05, "loss": 1.016, "step": 158410 }, { "epoch": 0.4, "learning_rate": 6.040213397755298e-05, "loss": 1.0177, "step": 158415 }, { "epoch": 0.4, "learning_rate": 6.040087573607127e-05, "loss": 1.0212, "step": 158420 }, { "epoch": 0.4, "learning_rate": 6.039961749458957e-05, "loss": 1.0191, "step": 158425 }, { "epoch": 0.4, "learning_rate": 6.039835925310786e-05, "loss": 1.0183, "step": 158430 }, { "epoch": 0.4, "learning_rate": 6.039710101162616e-05, "loss": 1.0184, "step": 158435 }, { "epoch": 0.4, "learning_rate": 6.039584277014445e-05, "loss": 1.0202, "step": 158440 }, { "epoch": 0.4, "learning_rate": 6.039458452866275e-05, "loss": 1.0429, "step": 158445 }, { "epoch": 0.4, "learning_rate": 6.039332628718104e-05, "loss": 1.0211, "step": 158450 }, { "epoch": 0.4, "learning_rate": 6.039206804569934e-05, "loss": 1.0455, "step": 158455 }, { "epoch": 0.4, "learning_rate": 6.039080980421763e-05, "loss": 1.0222, "step": 158460 }, { "epoch": 0.4, "learning_rate": 6.038955156273592e-05, "loss": 1.0181, "step": 158465 }, { "epoch": 0.4, "learning_rate": 6.038829332125422e-05, "loss": 1.0192, "step": 158470 }, { "epoch": 0.4, "learning_rate": 6.038703507977251e-05, "loss": 1.0223, "step": 158475 }, { "epoch": 0.4, "learning_rate": 6.038577683829081e-05, "loss": 1.0196, "step": 158480 }, { "epoch": 0.4, "learning_rate": 6.03845185968091e-05, "loss": 1.0179, "step": 158485 }, { "epoch": 0.4, "learning_rate": 6.03832603553274e-05, "loss": 1.0208, "step": 158490 }, { "epoch": 0.4, "learning_rate": 6.038200211384569e-05, "loss": 1.0169, "step": 158495 }, { "epoch": 0.4, "learning_rate": 6.038074387236399e-05, "loss": 1.0308, "step": 158500 }, { "epoch": 0.4, "learning_rate": 6.037948563088228e-05, "loss": 1.0188, "step": 158505 }, { "epoch": 0.4, "learning_rate": 6.037822738940058e-05, "loss": 1.0215, "step": 158510 }, { "epoch": 0.4, "learning_rate": 6.037696914791887e-05, "loss": 1.0214, "step": 158515 }, { "epoch": 0.4, "learning_rate": 6.037571090643716e-05, "loss": 1.0224, "step": 158520 }, { "epoch": 0.4, "learning_rate": 6.037445266495546e-05, "loss": 1.0422, "step": 158525 }, { "epoch": 0.4, "learning_rate": 6.037319442347375e-05, "loss": 1.0202, "step": 158530 }, { "epoch": 0.4, "learning_rate": 6.037193618199205e-05, "loss": 1.0207, "step": 158535 }, { "epoch": 0.4, "learning_rate": 6.037067794051034e-05, "loss": 1.0202, "step": 158540 }, { "epoch": 0.4, "learning_rate": 6.036941969902864e-05, "loss": 1.0185, "step": 158545 }, { "epoch": 0.4, "learning_rate": 6.036816145754693e-05, "loss": 1.0198, "step": 158550 }, { "epoch": 0.4, "learning_rate": 6.036690321606523e-05, "loss": 1.0204, "step": 158555 }, { "epoch": 0.4, "learning_rate": 6.036564497458352e-05, "loss": 1.0196, "step": 158560 }, { "epoch": 0.4, "learning_rate": 6.036438673310182e-05, "loss": 1.019, "step": 158565 }, { "epoch": 0.4, "learning_rate": 6.036312849162011e-05, "loss": 1.0405, "step": 158570 }, { "epoch": 0.4, "learning_rate": 6.036187025013841e-05, "loss": 1.0192, "step": 158575 }, { "epoch": 0.4, "learning_rate": 6.03606120086567e-05, "loss": 1.0193, "step": 158580 }, { "epoch": 0.4, "learning_rate": 6.0359353767174994e-05, "loss": 1.0202, "step": 158585 }, { "epoch": 0.4, "learning_rate": 6.035809552569329e-05, "loss": 1.0195, "step": 158590 }, { "epoch": 0.4, "learning_rate": 6.0356837284211584e-05, "loss": 1.0191, "step": 158595 }, { "epoch": 0.4, "learning_rate": 6.035557904272988e-05, "loss": 1.0191, "step": 158600 }, { "epoch": 0.4, "learning_rate": 6.0354320801248174e-05, "loss": 1.0212, "step": 158605 }, { "epoch": 0.4, "learning_rate": 6.035306255976647e-05, "loss": 1.0205, "step": 158610 }, { "epoch": 0.4, "learning_rate": 6.0351804318284764e-05, "loss": 1.019, "step": 158615 }, { "epoch": 0.4, "learning_rate": 6.035054607680306e-05, "loss": 1.0185, "step": 158620 }, { "epoch": 0.4, "learning_rate": 6.0349287835321354e-05, "loss": 1.0217, "step": 158625 }, { "epoch": 0.4, "learning_rate": 6.034802959383965e-05, "loss": 1.0213, "step": 158630 }, { "epoch": 0.4, "learning_rate": 6.0346771352357944e-05, "loss": 1.02, "step": 158635 }, { "epoch": 0.4, "learning_rate": 6.034551311087624e-05, "loss": 1.0181, "step": 158640 }, { "epoch": 0.4, "learning_rate": 6.0344254869394534e-05, "loss": 1.0198, "step": 158645 }, { "epoch": 0.4, "learning_rate": 6.0342996627912826e-05, "loss": 1.0203, "step": 158650 }, { "epoch": 0.4, "learning_rate": 6.0341738386431124e-05, "loss": 1.0202, "step": 158655 }, { "epoch": 0.4, "learning_rate": 6.0340480144949416e-05, "loss": 1.0176, "step": 158660 }, { "epoch": 0.4, "learning_rate": 6.0339221903467714e-05, "loss": 1.0222, "step": 158665 }, { "epoch": 0.4, "learning_rate": 6.0337963661986006e-05, "loss": 1.019, "step": 158670 }, { "epoch": 0.4, "learning_rate": 6.0336705420504304e-05, "loss": 1.0203, "step": 158675 }, { "epoch": 0.4, "learning_rate": 6.0335447179022596e-05, "loss": 1.0187, "step": 158680 }, { "epoch": 0.4, "learning_rate": 6.0334188937540894e-05, "loss": 1.0212, "step": 158685 }, { "epoch": 0.4, "learning_rate": 6.0332930696059186e-05, "loss": 1.0182, "step": 158690 }, { "epoch": 0.4, "learning_rate": 6.0331672454577484e-05, "loss": 1.0203, "step": 158695 }, { "epoch": 0.4, "learning_rate": 6.0330414213095776e-05, "loss": 1.021, "step": 158700 }, { "epoch": 0.4, "learning_rate": 6.0329155971614074e-05, "loss": 1.0204, "step": 158705 }, { "epoch": 0.4, "learning_rate": 6.0327897730132366e-05, "loss": 1.0225, "step": 158710 }, { "epoch": 0.4, "learning_rate": 6.032663948865066e-05, "loss": 1.0182, "step": 158715 }, { "epoch": 0.4, "learning_rate": 6.0325381247168956e-05, "loss": 1.017, "step": 158720 }, { "epoch": 0.4, "learning_rate": 6.032412300568725e-05, "loss": 1.0201, "step": 158725 }, { "epoch": 0.4, "learning_rate": 6.0322864764205545e-05, "loss": 1.02, "step": 158730 }, { "epoch": 0.4, "learning_rate": 6.032160652272384e-05, "loss": 1.0201, "step": 158735 }, { "epoch": 0.4, "learning_rate": 6.0320348281242135e-05, "loss": 1.0183, "step": 158740 }, { "epoch": 0.4, "learning_rate": 6.031909003976043e-05, "loss": 1.0177, "step": 158745 }, { "epoch": 0.4, "learning_rate": 6.0317831798278725e-05, "loss": 1.0427, "step": 158750 }, { "epoch": 0.4, "learning_rate": 6.031657355679702e-05, "loss": 1.0194, "step": 158755 }, { "epoch": 0.4, "learning_rate": 6.0315315315315315e-05, "loss": 1.0217, "step": 158760 }, { "epoch": 0.4, "learning_rate": 6.031405707383362e-05, "loss": 1.0196, "step": 158765 }, { "epoch": 0.4, "learning_rate": 6.031279883235191e-05, "loss": 1.0194, "step": 158770 }, { "epoch": 0.4, "learning_rate": 6.031154059087021e-05, "loss": 1.0197, "step": 158775 }, { "epoch": 0.4, "learning_rate": 6.03102823493885e-05, "loss": 1.0198, "step": 158780 }, { "epoch": 0.4, "learning_rate": 6.03090241079068e-05, "loss": 1.0221, "step": 158785 }, { "epoch": 0.4, "learning_rate": 6.030776586642509e-05, "loss": 1.0185, "step": 158790 }, { "epoch": 0.4, "learning_rate": 6.0306507624943384e-05, "loss": 1.0187, "step": 158795 }, { "epoch": 0.4, "learning_rate": 6.030524938346168e-05, "loss": 1.0208, "step": 158800 }, { "epoch": 0.4, "learning_rate": 6.0303991141979974e-05, "loss": 1.0208, "step": 158805 }, { "epoch": 0.4, "learning_rate": 6.030273290049827e-05, "loss": 1.0194, "step": 158810 }, { "epoch": 0.4, "learning_rate": 6.0301474659016564e-05, "loss": 1.0216, "step": 158815 }, { "epoch": 0.4, "learning_rate": 6.030021641753486e-05, "loss": 1.0218, "step": 158820 }, { "epoch": 0.4, "learning_rate": 6.0298958176053154e-05, "loss": 1.0181, "step": 158825 }, { "epoch": 0.4, "learning_rate": 6.029769993457145e-05, "loss": 1.0214, "step": 158830 }, { "epoch": 0.4, "learning_rate": 6.0296441693089744e-05, "loss": 1.0202, "step": 158835 }, { "epoch": 0.4, "learning_rate": 6.029518345160804e-05, "loss": 1.0205, "step": 158840 }, { "epoch": 0.4, "learning_rate": 6.0293925210126334e-05, "loss": 1.022, "step": 158845 }, { "epoch": 0.4, "learning_rate": 6.029266696864463e-05, "loss": 1.0206, "step": 158850 }, { "epoch": 0.4, "learning_rate": 6.0291408727162924e-05, "loss": 1.0183, "step": 158855 }, { "epoch": 0.4, "learning_rate": 6.0290150485681215e-05, "loss": 1.039, "step": 158860 }, { "epoch": 0.4, "learning_rate": 6.0288892244199513e-05, "loss": 1.0185, "step": 158865 }, { "epoch": 0.4, "learning_rate": 6.0287634002717805e-05, "loss": 1.0191, "step": 158870 }, { "epoch": 0.4, "learning_rate": 6.0286375761236103e-05, "loss": 1.0452, "step": 158875 }, { "epoch": 0.4, "learning_rate": 6.0285117519754395e-05, "loss": 1.0233, "step": 158880 }, { "epoch": 0.4, "learning_rate": 6.0283859278272693e-05, "loss": 1.0217, "step": 158885 }, { "epoch": 0.4, "learning_rate": 6.0282601036790985e-05, "loss": 1.02, "step": 158890 }, { "epoch": 0.4, "learning_rate": 6.0281342795309283e-05, "loss": 1.0192, "step": 158895 }, { "epoch": 0.4, "learning_rate": 6.0280084553827575e-05, "loss": 1.0193, "step": 158900 }, { "epoch": 0.4, "learning_rate": 6.027882631234587e-05, "loss": 1.0207, "step": 158905 }, { "epoch": 0.4, "learning_rate": 6.0277568070864165e-05, "loss": 1.0182, "step": 158910 }, { "epoch": 0.4, "learning_rate": 6.027630982938246e-05, "loss": 1.0189, "step": 158915 }, { "epoch": 0.4, "learning_rate": 6.0275051587900755e-05, "loss": 1.0228, "step": 158920 }, { "epoch": 0.4, "learning_rate": 6.0273793346419047e-05, "loss": 1.0221, "step": 158925 }, { "epoch": 0.4, "learning_rate": 6.0272535104937345e-05, "loss": 1.0211, "step": 158930 }, { "epoch": 0.4, "learning_rate": 6.0271276863455637e-05, "loss": 1.0178, "step": 158935 }, { "epoch": 0.4, "learning_rate": 6.0270018621973935e-05, "loss": 1.0204, "step": 158940 }, { "epoch": 0.4, "learning_rate": 6.0268760380492226e-05, "loss": 1.0204, "step": 158945 }, { "epoch": 0.4, "learning_rate": 6.0267502139010525e-05, "loss": 1.0197, "step": 158950 }, { "epoch": 0.4, "learning_rate": 6.0266243897528816e-05, "loss": 1.0198, "step": 158955 }, { "epoch": 0.4, "learning_rate": 6.0264985656047115e-05, "loss": 1.019, "step": 158960 }, { "epoch": 0.4, "learning_rate": 6.0263727414565406e-05, "loss": 1.0207, "step": 158965 }, { "epoch": 0.4, "learning_rate": 6.0262469173083705e-05, "loss": 1.0205, "step": 158970 }, { "epoch": 0.4, "learning_rate": 6.0261210931601996e-05, "loss": 1.0195, "step": 158975 }, { "epoch": 0.4, "learning_rate": 6.0259952690120295e-05, "loss": 1.0172, "step": 158980 }, { "epoch": 0.4, "learning_rate": 6.0258694448638586e-05, "loss": 1.0194, "step": 158985 }, { "epoch": 0.4, "learning_rate": 6.025743620715688e-05, "loss": 1.0187, "step": 158990 }, { "epoch": 0.4, "learning_rate": 6.0256177965675176e-05, "loss": 1.0194, "step": 158995 }, { "epoch": 0.4, "learning_rate": 6.025491972419347e-05, "loss": 1.0196, "step": 159000 }, { "epoch": 0.4, "learning_rate": 6.0253661482711766e-05, "loss": 1.0191, "step": 159005 }, { "epoch": 0.4, "learning_rate": 6.025240324123006e-05, "loss": 1.0175, "step": 159010 }, { "epoch": 0.4, "learning_rate": 6.0251144999748356e-05, "loss": 1.0201, "step": 159015 }, { "epoch": 0.4, "learning_rate": 6.024988675826665e-05, "loss": 1.0231, "step": 159020 }, { "epoch": 0.4, "learning_rate": 6.0248628516784946e-05, "loss": 1.021, "step": 159025 }, { "epoch": 0.4, "learning_rate": 6.024737027530324e-05, "loss": 1.0194, "step": 159030 }, { "epoch": 0.4, "learning_rate": 6.0246112033821536e-05, "loss": 1.0202, "step": 159035 }, { "epoch": 0.4, "learning_rate": 6.024485379233983e-05, "loss": 1.0201, "step": 159040 }, { "epoch": 0.4, "learning_rate": 6.0243595550858126e-05, "loss": 1.0386, "step": 159045 }, { "epoch": 0.4, "learning_rate": 6.024233730937642e-05, "loss": 1.0185, "step": 159050 }, { "epoch": 0.4, "learning_rate": 6.024107906789471e-05, "loss": 1.0195, "step": 159055 }, { "epoch": 0.4, "learning_rate": 6.023982082641301e-05, "loss": 1.0197, "step": 159060 }, { "epoch": 0.4, "learning_rate": 6.02385625849313e-05, "loss": 1.0197, "step": 159065 }, { "epoch": 0.4, "learning_rate": 6.02373043434496e-05, "loss": 1.0188, "step": 159070 }, { "epoch": 0.4, "learning_rate": 6.023604610196789e-05, "loss": 1.0205, "step": 159075 }, { "epoch": 0.4, "learning_rate": 6.023478786048619e-05, "loss": 1.0221, "step": 159080 }, { "epoch": 0.4, "learning_rate": 6.023352961900448e-05, "loss": 1.0193, "step": 159085 }, { "epoch": 0.4, "learning_rate": 6.023227137752278e-05, "loss": 1.0177, "step": 159090 }, { "epoch": 0.4, "learning_rate": 6.023101313604107e-05, "loss": 1.0204, "step": 159095 }, { "epoch": 0.4, "learning_rate": 6.022975489455937e-05, "loss": 1.0195, "step": 159100 }, { "epoch": 0.4, "learning_rate": 6.022849665307766e-05, "loss": 1.0205, "step": 159105 }, { "epoch": 0.4, "learning_rate": 6.022723841159596e-05, "loss": 1.0204, "step": 159110 }, { "epoch": 0.4, "learning_rate": 6.022598017011425e-05, "loss": 1.0219, "step": 159115 }, { "epoch": 0.4, "learning_rate": 6.022472192863254e-05, "loss": 1.0185, "step": 159120 }, { "epoch": 0.4, "learning_rate": 6.022346368715084e-05, "loss": 1.0201, "step": 159125 }, { "epoch": 0.4, "learning_rate": 6.022220544566913e-05, "loss": 1.02, "step": 159130 }, { "epoch": 0.4, "learning_rate": 6.022094720418743e-05, "loss": 1.0202, "step": 159135 }, { "epoch": 0.4, "learning_rate": 6.021968896270572e-05, "loss": 1.022, "step": 159140 }, { "epoch": 0.4, "learning_rate": 6.021843072122402e-05, "loss": 1.0198, "step": 159145 }, { "epoch": 0.4, "learning_rate": 6.021717247974231e-05, "loss": 1.0199, "step": 159150 }, { "epoch": 0.4, "learning_rate": 6.021591423826061e-05, "loss": 1.0208, "step": 159155 }, { "epoch": 0.4, "learning_rate": 6.02146559967789e-05, "loss": 1.0198, "step": 159160 }, { "epoch": 0.4, "learning_rate": 6.02133977552972e-05, "loss": 1.0207, "step": 159165 }, { "epoch": 0.4, "learning_rate": 6.021213951381549e-05, "loss": 1.0187, "step": 159170 }, { "epoch": 0.4, "learning_rate": 6.021088127233378e-05, "loss": 1.018, "step": 159175 }, { "epoch": 0.4, "learning_rate": 6.020962303085208e-05, "loss": 1.0215, "step": 159180 }, { "epoch": 0.4, "learning_rate": 6.020836478937037e-05, "loss": 1.0205, "step": 159185 }, { "epoch": 0.4, "learning_rate": 6.020710654788867e-05, "loss": 1.0215, "step": 159190 }, { "epoch": 0.4, "learning_rate": 6.020584830640696e-05, "loss": 1.0199, "step": 159195 }, { "epoch": 0.4, "learning_rate": 6.020459006492526e-05, "loss": 1.0198, "step": 159200 }, { "epoch": 0.4, "learning_rate": 6.020333182344355e-05, "loss": 1.0363, "step": 159205 }, { "epoch": 0.4, "learning_rate": 6.020207358196185e-05, "loss": 1.0199, "step": 159210 }, { "epoch": 0.4, "learning_rate": 6.020081534048014e-05, "loss": 1.0213, "step": 159215 }, { "epoch": 0.4, "learning_rate": 6.019955709899844e-05, "loss": 1.0199, "step": 159220 }, { "epoch": 0.4, "learning_rate": 6.019829885751673e-05, "loss": 1.0208, "step": 159225 }, { "epoch": 0.4, "learning_rate": 6.019704061603503e-05, "loss": 1.0214, "step": 159230 }, { "epoch": 0.4, "learning_rate": 6.019578237455332e-05, "loss": 1.0198, "step": 159235 }, { "epoch": 0.4, "learning_rate": 6.0194524133071614e-05, "loss": 1.0204, "step": 159240 }, { "epoch": 0.4, "learning_rate": 6.019326589158991e-05, "loss": 1.0198, "step": 159245 }, { "epoch": 0.4, "learning_rate": 6.0192007650108204e-05, "loss": 1.0207, "step": 159250 }, { "epoch": 0.4, "learning_rate": 6.01907494086265e-05, "loss": 1.0217, "step": 159255 }, { "epoch": 0.4, "learning_rate": 6.0189491167144794e-05, "loss": 1.0188, "step": 159260 }, { "epoch": 0.4, "learning_rate": 6.01882329256631e-05, "loss": 1.0182, "step": 159265 }, { "epoch": 0.4, "learning_rate": 6.01869746841814e-05, "loss": 1.0209, "step": 159270 }, { "epoch": 0.4, "learning_rate": 6.018571644269969e-05, "loss": 1.0208, "step": 159275 }, { "epoch": 0.4, "learning_rate": 6.018445820121799e-05, "loss": 1.0181, "step": 159280 }, { "epoch": 0.4, "learning_rate": 6.018319995973628e-05, "loss": 1.0204, "step": 159285 }, { "epoch": 0.4, "learning_rate": 6.018194171825458e-05, "loss": 1.0208, "step": 159290 }, { "epoch": 0.4, "learning_rate": 6.018068347677287e-05, "loss": 1.0189, "step": 159295 }, { "epoch": 0.4, "learning_rate": 6.017942523529117e-05, "loss": 1.0196, "step": 159300 }, { "epoch": 0.4, "learning_rate": 6.017816699380946e-05, "loss": 1.0188, "step": 159305 }, { "epoch": 0.4, "learning_rate": 6.017690875232776e-05, "loss": 1.0202, "step": 159310 }, { "epoch": 0.4, "learning_rate": 6.017565051084605e-05, "loss": 1.0216, "step": 159315 }, { "epoch": 0.4, "learning_rate": 6.017439226936434e-05, "loss": 1.0204, "step": 159320 }, { "epoch": 0.4, "learning_rate": 6.017313402788264e-05, "loss": 1.019, "step": 159325 }, { "epoch": 0.4, "learning_rate": 6.017187578640093e-05, "loss": 1.02, "step": 159330 }, { "epoch": 0.4, "learning_rate": 6.017061754491923e-05, "loss": 1.0206, "step": 159335 }, { "epoch": 0.4, "learning_rate": 6.016935930343752e-05, "loss": 1.0198, "step": 159340 }, { "epoch": 0.4, "learning_rate": 6.016810106195582e-05, "loss": 1.0186, "step": 159345 }, { "epoch": 0.4, "learning_rate": 6.016684282047411e-05, "loss": 1.0185, "step": 159350 }, { "epoch": 0.4, "learning_rate": 6.016558457899241e-05, "loss": 1.02, "step": 159355 }, { "epoch": 0.4, "learning_rate": 6.01643263375107e-05, "loss": 1.0182, "step": 159360 }, { "epoch": 0.4, "learning_rate": 6.0163068096029e-05, "loss": 1.0213, "step": 159365 }, { "epoch": 0.4, "learning_rate": 6.016180985454729e-05, "loss": 1.0205, "step": 159370 }, { "epoch": 0.4, "learning_rate": 6.016055161306559e-05, "loss": 1.0187, "step": 159375 }, { "epoch": 0.4, "learning_rate": 6.015929337158388e-05, "loss": 1.0192, "step": 159380 }, { "epoch": 0.4, "learning_rate": 6.015803513010217e-05, "loss": 1.0173, "step": 159385 }, { "epoch": 0.4, "learning_rate": 6.015677688862047e-05, "loss": 1.0187, "step": 159390 }, { "epoch": 0.4, "learning_rate": 6.015551864713876e-05, "loss": 1.0226, "step": 159395 }, { "epoch": 0.4, "learning_rate": 6.015426040565706e-05, "loss": 1.0194, "step": 159400 }, { "epoch": 0.4, "learning_rate": 6.015300216417535e-05, "loss": 1.0208, "step": 159405 }, { "epoch": 0.4, "learning_rate": 6.015174392269365e-05, "loss": 1.0539, "step": 159410 }, { "epoch": 0.4, "learning_rate": 6.015048568121194e-05, "loss": 1.0202, "step": 159415 }, { "epoch": 0.4, "learning_rate": 6.014922743973024e-05, "loss": 1.0189, "step": 159420 }, { "epoch": 0.4, "learning_rate": 6.014796919824853e-05, "loss": 1.0204, "step": 159425 }, { "epoch": 0.4, "learning_rate": 6.014671095676683e-05, "loss": 1.0201, "step": 159430 }, { "epoch": 0.4, "learning_rate": 6.014545271528512e-05, "loss": 1.0188, "step": 159435 }, { "epoch": 0.4, "learning_rate": 6.014419447380342e-05, "loss": 1.0206, "step": 159440 }, { "epoch": 0.4, "learning_rate": 6.014293623232171e-05, "loss": 1.0199, "step": 159445 }, { "epoch": 0.4, "learning_rate": 6.014167799084e-05, "loss": 1.0205, "step": 159450 }, { "epoch": 0.4, "learning_rate": 6.01404197493583e-05, "loss": 1.0207, "step": 159455 }, { "epoch": 0.4, "learning_rate": 6.013916150787659e-05, "loss": 1.0419, "step": 159460 }, { "epoch": 0.4, "learning_rate": 6.013790326639489e-05, "loss": 1.0201, "step": 159465 }, { "epoch": 0.4, "learning_rate": 6.013664502491318e-05, "loss": 1.0191, "step": 159470 }, { "epoch": 0.4, "learning_rate": 6.013538678343148e-05, "loss": 1.0221, "step": 159475 }, { "epoch": 0.4, "learning_rate": 6.013412854194977e-05, "loss": 1.0191, "step": 159480 }, { "epoch": 0.4, "learning_rate": 6.013287030046807e-05, "loss": 1.0197, "step": 159485 }, { "epoch": 0.4, "learning_rate": 6.013161205898636e-05, "loss": 1.0227, "step": 159490 }, { "epoch": 0.4, "learning_rate": 6.013035381750466e-05, "loss": 1.0213, "step": 159495 }, { "epoch": 0.4, "learning_rate": 6.012909557602295e-05, "loss": 1.0203, "step": 159500 }, { "epoch": 0.4, "learning_rate": 6.012783733454125e-05, "loss": 1.0207, "step": 159505 }, { "epoch": 0.4, "learning_rate": 6.012657909305954e-05, "loss": 1.0184, "step": 159510 }, { "epoch": 0.4, "learning_rate": 6.0125320851577835e-05, "loss": 1.0186, "step": 159515 }, { "epoch": 0.4, "learning_rate": 6.012406261009613e-05, "loss": 1.0184, "step": 159520 }, { "epoch": 0.4, "learning_rate": 6.0122804368614424e-05, "loss": 1.0196, "step": 159525 }, { "epoch": 0.4, "learning_rate": 6.012154612713272e-05, "loss": 1.022, "step": 159530 }, { "epoch": 0.4, "learning_rate": 6.0120287885651014e-05, "loss": 1.0211, "step": 159535 }, { "epoch": 0.4, "learning_rate": 6.011902964416931e-05, "loss": 1.0213, "step": 159540 }, { "epoch": 0.4, "learning_rate": 6.0117771402687604e-05, "loss": 1.0184, "step": 159545 }, { "epoch": 0.4, "learning_rate": 6.01165131612059e-05, "loss": 1.0211, "step": 159550 }, { "epoch": 0.4, "learning_rate": 6.0115254919724194e-05, "loss": 1.0196, "step": 159555 }, { "epoch": 0.4, "learning_rate": 6.011399667824249e-05, "loss": 1.0197, "step": 159560 }, { "epoch": 0.4, "learning_rate": 6.0112738436760784e-05, "loss": 1.0395, "step": 159565 }, { "epoch": 0.4, "learning_rate": 6.011148019527908e-05, "loss": 1.0189, "step": 159570 }, { "epoch": 0.4, "learning_rate": 6.0110221953797374e-05, "loss": 1.0203, "step": 159575 }, { "epoch": 0.4, "learning_rate": 6.0108963712315666e-05, "loss": 1.0209, "step": 159580 }, { "epoch": 0.4, "learning_rate": 6.0107705470833964e-05, "loss": 1.0207, "step": 159585 }, { "epoch": 0.4, "learning_rate": 6.0106447229352256e-05, "loss": 1.0206, "step": 159590 }, { "epoch": 0.4, "learning_rate": 6.0105188987870554e-05, "loss": 1.0177, "step": 159595 }, { "epoch": 0.4, "learning_rate": 6.0103930746388846e-05, "loss": 1.018, "step": 159600 }, { "epoch": 0.4, "learning_rate": 6.0102672504907144e-05, "loss": 1.0193, "step": 159605 }, { "epoch": 0.4, "learning_rate": 6.0101414263425436e-05, "loss": 1.02, "step": 159610 }, { "epoch": 0.4, "learning_rate": 6.0100156021943734e-05, "loss": 1.0192, "step": 159615 }, { "epoch": 0.4, "learning_rate": 6.0098897780462026e-05, "loss": 1.0216, "step": 159620 }, { "epoch": 0.4, "learning_rate": 6.0097639538980324e-05, "loss": 1.0212, "step": 159625 }, { "epoch": 0.4, "learning_rate": 6.0096381297498616e-05, "loss": 1.0197, "step": 159630 }, { "epoch": 0.4, "learning_rate": 6.0095123056016914e-05, "loss": 1.0191, "step": 159635 }, { "epoch": 0.4, "learning_rate": 6.0093864814535206e-05, "loss": 1.0202, "step": 159640 }, { "epoch": 0.4, "learning_rate": 6.00926065730535e-05, "loss": 1.0203, "step": 159645 }, { "epoch": 0.4, "learning_rate": 6.0091348331571796e-05, "loss": 1.0196, "step": 159650 }, { "epoch": 0.4, "learning_rate": 6.009009009009009e-05, "loss": 1.0183, "step": 159655 }, { "epoch": 0.4, "learning_rate": 6.0088831848608386e-05, "loss": 1.0195, "step": 159660 }, { "epoch": 0.4, "learning_rate": 6.008757360712668e-05, "loss": 1.0194, "step": 159665 }, { "epoch": 0.4, "learning_rate": 6.0086315365644976e-05, "loss": 1.0454, "step": 159670 }, { "epoch": 0.4, "learning_rate": 6.008505712416327e-05, "loss": 1.0192, "step": 159675 }, { "epoch": 0.4, "learning_rate": 6.0083798882681566e-05, "loss": 1.019, "step": 159680 }, { "epoch": 0.4, "learning_rate": 6.008254064119986e-05, "loss": 1.0193, "step": 159685 }, { "epoch": 0.4, "learning_rate": 6.0081282399718156e-05, "loss": 1.0203, "step": 159690 }, { "epoch": 0.4, "learning_rate": 6.008002415823645e-05, "loss": 1.0189, "step": 159695 }, { "epoch": 0.4, "learning_rate": 6.0078765916754746e-05, "loss": 1.0197, "step": 159700 }, { "epoch": 0.4, "learning_rate": 6.007750767527304e-05, "loss": 1.0224, "step": 159705 }, { "epoch": 0.4, "learning_rate": 6.007624943379133e-05, "loss": 1.0192, "step": 159710 }, { "epoch": 0.4, "learning_rate": 6.007499119230963e-05, "loss": 1.0174, "step": 159715 }, { "epoch": 0.4, "learning_rate": 6.007373295082792e-05, "loss": 1.0208, "step": 159720 }, { "epoch": 0.4, "learning_rate": 6.007247470934622e-05, "loss": 1.0178, "step": 159725 }, { "epoch": 0.4, "learning_rate": 6.007121646786451e-05, "loss": 1.0217, "step": 159730 }, { "epoch": 0.4, "learning_rate": 6.006995822638281e-05, "loss": 1.0209, "step": 159735 }, { "epoch": 0.4, "learning_rate": 6.00686999849011e-05, "loss": 1.0179, "step": 159740 }, { "epoch": 0.4, "learning_rate": 6.00674417434194e-05, "loss": 1.0209, "step": 159745 }, { "epoch": 0.4, "learning_rate": 6.006618350193769e-05, "loss": 1.0201, "step": 159750 }, { "epoch": 0.4, "learning_rate": 6.006492526045599e-05, "loss": 1.0186, "step": 159755 }, { "epoch": 0.4, "learning_rate": 6.006366701897428e-05, "loss": 1.0207, "step": 159760 }, { "epoch": 0.4, "learning_rate": 6.0062408777492584e-05, "loss": 1.0201, "step": 159765 }, { "epoch": 0.4, "learning_rate": 6.006115053601088e-05, "loss": 1.0186, "step": 159770 }, { "epoch": 0.4, "learning_rate": 6.0059892294529174e-05, "loss": 1.0198, "step": 159775 }, { "epoch": 0.4, "learning_rate": 6.005863405304747e-05, "loss": 1.0197, "step": 159780 }, { "epoch": 0.4, "learning_rate": 6.0057375811565764e-05, "loss": 1.0192, "step": 159785 }, { "epoch": 0.4, "learning_rate": 6.0056117570084055e-05, "loss": 1.0182, "step": 159790 }, { "epoch": 0.4, "learning_rate": 6.0054859328602354e-05, "loss": 1.0198, "step": 159795 }, { "epoch": 0.4, "learning_rate": 6.0053601087120645e-05, "loss": 1.0199, "step": 159800 }, { "epoch": 0.4, "learning_rate": 6.0052342845638944e-05, "loss": 1.019, "step": 159805 }, { "epoch": 0.4, "learning_rate": 6.0051084604157235e-05, "loss": 1.0202, "step": 159810 }, { "epoch": 0.4, "learning_rate": 6.0049826362675534e-05, "loss": 1.0206, "step": 159815 }, { "epoch": 0.4, "learning_rate": 6.0048568121193825e-05, "loss": 1.0181, "step": 159820 }, { "epoch": 0.4, "learning_rate": 6.0047309879712124e-05, "loss": 1.0206, "step": 159825 }, { "epoch": 0.4, "learning_rate": 6.0046051638230415e-05, "loss": 1.0221, "step": 159830 }, { "epoch": 0.4, "learning_rate": 6.0044793396748714e-05, "loss": 1.0218, "step": 159835 }, { "epoch": 0.4, "learning_rate": 6.0043535155267005e-05, "loss": 1.0204, "step": 159840 }, { "epoch": 0.4, "learning_rate": 6.0042276913785304e-05, "loss": 1.0195, "step": 159845 }, { "epoch": 0.4, "learning_rate": 6.0041018672303595e-05, "loss": 1.0202, "step": 159850 }, { "epoch": 0.4, "learning_rate": 6.003976043082189e-05, "loss": 1.0194, "step": 159855 }, { "epoch": 0.4, "learning_rate": 6.0038502189340185e-05, "loss": 1.0205, "step": 159860 }, { "epoch": 0.4, "learning_rate": 6.003724394785848e-05, "loss": 1.0183, "step": 159865 }, { "epoch": 0.4, "learning_rate": 6.0035985706376775e-05, "loss": 1.0191, "step": 159870 }, { "epoch": 0.4, "learning_rate": 6.003472746489507e-05, "loss": 1.021, "step": 159875 }, { "epoch": 0.4, "learning_rate": 6.0033469223413365e-05, "loss": 1.0215, "step": 159880 }, { "epoch": 0.4, "learning_rate": 6.003221098193166e-05, "loss": 1.0199, "step": 159885 }, { "epoch": 0.4, "learning_rate": 6.0030952740449955e-05, "loss": 1.0194, "step": 159890 }, { "epoch": 0.4, "learning_rate": 6.002969449896825e-05, "loss": 1.0199, "step": 159895 }, { "epoch": 0.4, "learning_rate": 6.0028436257486545e-05, "loss": 1.0216, "step": 159900 }, { "epoch": 0.4, "learning_rate": 6.0027178016004837e-05, "loss": 1.0205, "step": 159905 }, { "epoch": 0.4, "learning_rate": 6.0025919774523135e-05, "loss": 1.0194, "step": 159910 }, { "epoch": 0.4, "learning_rate": 6.0024661533041427e-05, "loss": 1.0183, "step": 159915 }, { "epoch": 0.4, "learning_rate": 6.002340329155972e-05, "loss": 1.0252, "step": 159920 }, { "epoch": 0.4, "learning_rate": 6.0022145050078017e-05, "loss": 1.0224, "step": 159925 }, { "epoch": 0.4, "learning_rate": 6.002088680859631e-05, "loss": 1.0212, "step": 159930 }, { "epoch": 0.4, "learning_rate": 6.0019628567114607e-05, "loss": 1.0189, "step": 159935 }, { "epoch": 0.4, "learning_rate": 6.00183703256329e-05, "loss": 1.0183, "step": 159940 }, { "epoch": 0.4, "learning_rate": 6.0017112084151196e-05, "loss": 1.0204, "step": 159945 }, { "epoch": 0.4, "learning_rate": 6.001585384266949e-05, "loss": 1.0385, "step": 159950 }, { "epoch": 0.4, "learning_rate": 6.0014595601187786e-05, "loss": 1.0204, "step": 159955 }, { "epoch": 0.4, "learning_rate": 6.001333735970608e-05, "loss": 1.0205, "step": 159960 }, { "epoch": 0.4, "learning_rate": 6.0012079118224376e-05, "loss": 1.0178, "step": 159965 }, { "epoch": 0.4, "learning_rate": 6.001082087674267e-05, "loss": 1.0199, "step": 159970 }, { "epoch": 0.4, "learning_rate": 6.000956263526096e-05, "loss": 1.0228, "step": 159975 }, { "epoch": 0.4, "learning_rate": 6.000830439377926e-05, "loss": 1.0196, "step": 159980 }, { "epoch": 0.4, "learning_rate": 6.000704615229755e-05, "loss": 1.0209, "step": 159985 }, { "epoch": 0.4, "learning_rate": 6.000578791081585e-05, "loss": 1.0207, "step": 159990 }, { "epoch": 0.4, "learning_rate": 6.000452966933414e-05, "loss": 1.02, "step": 159995 }, { "epoch": 0.4, "learning_rate": 6.000327142785244e-05, "loss": 1.022, "step": 160000 }, { "epoch": 0.4, "learning_rate": 6.000201318637073e-05, "loss": 1.0209, "step": 160005 }, { "epoch": 0.4, "learning_rate": 6.000075494488903e-05, "loss": 1.0428, "step": 160010 }, { "epoch": 0.4, "learning_rate": 5.999949670340732e-05, "loss": 1.0216, "step": 160015 }, { "epoch": 0.4, "learning_rate": 5.999823846192562e-05, "loss": 1.0196, "step": 160020 }, { "epoch": 0.4, "learning_rate": 5.999698022044391e-05, "loss": 1.02, "step": 160025 }, { "epoch": 0.4, "learning_rate": 5.999572197896221e-05, "loss": 1.0225, "step": 160030 }, { "epoch": 0.4, "learning_rate": 5.99944637374805e-05, "loss": 1.0197, "step": 160035 }, { "epoch": 0.4, "learning_rate": 5.999320549599879e-05, "loss": 1.0379, "step": 160040 }, { "epoch": 0.4, "learning_rate": 5.999194725451709e-05, "loss": 1.0201, "step": 160045 }, { "epoch": 0.4, "learning_rate": 5.999068901303538e-05, "loss": 1.022, "step": 160050 }, { "epoch": 0.4, "learning_rate": 5.998943077155368e-05, "loss": 1.0185, "step": 160055 }, { "epoch": 0.4, "learning_rate": 5.998817253007197e-05, "loss": 1.0218, "step": 160060 }, { "epoch": 0.4, "learning_rate": 5.998691428859027e-05, "loss": 1.0202, "step": 160065 }, { "epoch": 0.4, "learning_rate": 5.998565604710856e-05, "loss": 1.019, "step": 160070 }, { "epoch": 0.4, "learning_rate": 5.998439780562686e-05, "loss": 1.0172, "step": 160075 }, { "epoch": 0.4, "learning_rate": 5.998313956414515e-05, "loss": 1.0199, "step": 160080 }, { "epoch": 0.4, "learning_rate": 5.998188132266345e-05, "loss": 1.0182, "step": 160085 }, { "epoch": 0.4, "learning_rate": 5.998062308118174e-05, "loss": 1.0217, "step": 160090 }, { "epoch": 0.4, "learning_rate": 5.997936483970004e-05, "loss": 1.039, "step": 160095 }, { "epoch": 0.4, "learning_rate": 5.997810659821833e-05, "loss": 1.0211, "step": 160100 }, { "epoch": 0.4, "learning_rate": 5.997684835673662e-05, "loss": 1.0213, "step": 160105 }, { "epoch": 0.4, "learning_rate": 5.997559011525492e-05, "loss": 1.0193, "step": 160110 }, { "epoch": 0.4, "learning_rate": 5.997433187377321e-05, "loss": 1.019, "step": 160115 }, { "epoch": 0.4, "learning_rate": 5.997307363229151e-05, "loss": 1.0212, "step": 160120 }, { "epoch": 0.4, "learning_rate": 5.99718153908098e-05, "loss": 1.0413, "step": 160125 }, { "epoch": 0.4, "learning_rate": 5.99705571493281e-05, "loss": 1.0197, "step": 160130 }, { "epoch": 0.4, "learning_rate": 5.996929890784639e-05, "loss": 1.0207, "step": 160135 }, { "epoch": 0.4, "learning_rate": 5.996804066636469e-05, "loss": 1.0195, "step": 160140 }, { "epoch": 0.4, "learning_rate": 5.996678242488298e-05, "loss": 1.0191, "step": 160145 }, { "epoch": 0.4, "learning_rate": 5.996552418340128e-05, "loss": 1.0166, "step": 160150 }, { "epoch": 0.4, "learning_rate": 5.996426594191957e-05, "loss": 1.0196, "step": 160155 }, { "epoch": 0.4, "learning_rate": 5.996300770043787e-05, "loss": 1.0222, "step": 160160 }, { "epoch": 0.4, "learning_rate": 5.996174945895616e-05, "loss": 1.0204, "step": 160165 }, { "epoch": 0.4, "learning_rate": 5.9960491217474454e-05, "loss": 1.0208, "step": 160170 }, { "epoch": 0.4, "learning_rate": 5.995923297599275e-05, "loss": 1.0189, "step": 160175 }, { "epoch": 0.4, "learning_rate": 5.9957974734511044e-05, "loss": 1.0186, "step": 160180 }, { "epoch": 0.4, "learning_rate": 5.995671649302934e-05, "loss": 1.02, "step": 160185 }, { "epoch": 0.4, "learning_rate": 5.9955458251547634e-05, "loss": 1.0212, "step": 160190 }, { "epoch": 0.4, "learning_rate": 5.995420001006593e-05, "loss": 1.0208, "step": 160195 }, { "epoch": 0.4, "learning_rate": 5.9952941768584224e-05, "loss": 1.0199, "step": 160200 }, { "epoch": 0.4, "learning_rate": 5.995168352710252e-05, "loss": 1.0203, "step": 160205 }, { "epoch": 0.4, "learning_rate": 5.9950425285620814e-05, "loss": 1.0184, "step": 160210 }, { "epoch": 0.4, "learning_rate": 5.994916704413911e-05, "loss": 1.0225, "step": 160215 }, { "epoch": 0.4, "learning_rate": 5.9947908802657404e-05, "loss": 1.0179, "step": 160220 }, { "epoch": 0.4, "learning_rate": 5.99466505611757e-05, "loss": 1.0178, "step": 160225 }, { "epoch": 0.4, "learning_rate": 5.9945392319693994e-05, "loss": 1.0195, "step": 160230 }, { "epoch": 0.4, "learning_rate": 5.9944134078212285e-05, "loss": 1.0204, "step": 160235 }, { "epoch": 0.4, "learning_rate": 5.9942875836730584e-05, "loss": 1.0182, "step": 160240 }, { "epoch": 0.4, "learning_rate": 5.9941617595248875e-05, "loss": 1.019, "step": 160245 }, { "epoch": 0.4, "learning_rate": 5.9940359353767174e-05, "loss": 1.0191, "step": 160250 }, { "epoch": 0.4, "learning_rate": 5.9939101112285465e-05, "loss": 1.0189, "step": 160255 }, { "epoch": 0.4, "learning_rate": 5.9937842870803764e-05, "loss": 1.0169, "step": 160260 }, { "epoch": 0.4, "learning_rate": 5.993658462932207e-05, "loss": 1.0192, "step": 160265 }, { "epoch": 0.4, "learning_rate": 5.993532638784036e-05, "loss": 1.0199, "step": 160270 }, { "epoch": 0.4, "learning_rate": 5.993406814635866e-05, "loss": 1.0211, "step": 160275 }, { "epoch": 0.4, "learning_rate": 5.993280990487695e-05, "loss": 1.022, "step": 160280 }, { "epoch": 0.4, "learning_rate": 5.993155166339525e-05, "loss": 1.0188, "step": 160285 }, { "epoch": 0.4, "learning_rate": 5.993029342191354e-05, "loss": 1.0185, "step": 160290 }, { "epoch": 0.4, "learning_rate": 5.992903518043184e-05, "loss": 1.0186, "step": 160295 }, { "epoch": 0.4, "learning_rate": 5.992777693895013e-05, "loss": 1.0206, "step": 160300 }, { "epoch": 0.4, "learning_rate": 5.992651869746843e-05, "loss": 1.0198, "step": 160305 }, { "epoch": 0.4, "learning_rate": 5.992526045598672e-05, "loss": 1.0203, "step": 160310 }, { "epoch": 0.4, "learning_rate": 5.992400221450501e-05, "loss": 1.019, "step": 160315 }, { "epoch": 0.4, "learning_rate": 5.992274397302331e-05, "loss": 1.0183, "step": 160320 }, { "epoch": 0.4, "learning_rate": 5.99214857315416e-05, "loss": 1.0198, "step": 160325 }, { "epoch": 0.4, "learning_rate": 5.99202274900599e-05, "loss": 1.0216, "step": 160330 }, { "epoch": 0.4, "learning_rate": 5.991896924857819e-05, "loss": 1.0192, "step": 160335 }, { "epoch": 0.4, "learning_rate": 5.991771100709649e-05, "loss": 1.0188, "step": 160340 }, { "epoch": 0.4, "learning_rate": 5.991645276561478e-05, "loss": 1.0219, "step": 160345 }, { "epoch": 0.4, "learning_rate": 5.991519452413308e-05, "loss": 1.0222, "step": 160350 }, { "epoch": 0.4, "learning_rate": 5.991393628265137e-05, "loss": 1.0192, "step": 160355 }, { "epoch": 0.4, "learning_rate": 5.991267804116967e-05, "loss": 1.02, "step": 160360 }, { "epoch": 0.4, "learning_rate": 5.991141979968796e-05, "loss": 1.021, "step": 160365 }, { "epoch": 0.4, "learning_rate": 5.991016155820626e-05, "loss": 1.0221, "step": 160370 }, { "epoch": 0.4, "learning_rate": 5.990890331672455e-05, "loss": 1.0197, "step": 160375 }, { "epoch": 0.4, "learning_rate": 5.990764507524284e-05, "loss": 1.0191, "step": 160380 }, { "epoch": 0.4, "learning_rate": 5.990638683376114e-05, "loss": 1.02, "step": 160385 }, { "epoch": 0.4, "learning_rate": 5.990512859227943e-05, "loss": 1.0197, "step": 160390 }, { "epoch": 0.4, "learning_rate": 5.990387035079773e-05, "loss": 1.0199, "step": 160395 }, { "epoch": 0.4, "learning_rate": 5.990261210931602e-05, "loss": 1.0207, "step": 160400 }, { "epoch": 0.4, "learning_rate": 5.990135386783432e-05, "loss": 1.0188, "step": 160405 }, { "epoch": 0.4, "learning_rate": 5.990009562635261e-05, "loss": 1.0188, "step": 160410 }, { "epoch": 0.4, "learning_rate": 5.989883738487091e-05, "loss": 1.0209, "step": 160415 }, { "epoch": 0.4, "learning_rate": 5.98975791433892e-05, "loss": 1.0203, "step": 160420 }, { "epoch": 0.4, "learning_rate": 5.98963209019075e-05, "loss": 1.0213, "step": 160425 }, { "epoch": 0.4, "learning_rate": 5.989506266042579e-05, "loss": 1.0198, "step": 160430 }, { "epoch": 0.4, "learning_rate": 5.989380441894409e-05, "loss": 1.0188, "step": 160435 }, { "epoch": 0.4, "learning_rate": 5.989254617746238e-05, "loss": 1.0193, "step": 160440 }, { "epoch": 0.4, "learning_rate": 5.9891287935980675e-05, "loss": 1.0229, "step": 160445 }, { "epoch": 0.4, "learning_rate": 5.989002969449897e-05, "loss": 1.0178, "step": 160450 }, { "epoch": 0.4, "learning_rate": 5.9888771453017265e-05, "loss": 1.0208, "step": 160455 }, { "epoch": 0.4, "learning_rate": 5.988751321153556e-05, "loss": 1.021, "step": 160460 }, { "epoch": 0.4, "learning_rate": 5.9886254970053855e-05, "loss": 1.021, "step": 160465 }, { "epoch": 0.4, "learning_rate": 5.988499672857215e-05, "loss": 1.0208, "step": 160470 }, { "epoch": 0.4, "learning_rate": 5.9883738487090445e-05, "loss": 1.0217, "step": 160475 }, { "epoch": 0.4, "learning_rate": 5.988248024560874e-05, "loss": 1.0213, "step": 160480 }, { "epoch": 0.4, "learning_rate": 5.9881222004127035e-05, "loss": 1.021, "step": 160485 }, { "epoch": 0.4, "learning_rate": 5.987996376264533e-05, "loss": 1.023, "step": 160490 }, { "epoch": 0.4, "learning_rate": 5.9878705521163625e-05, "loss": 1.0212, "step": 160495 }, { "epoch": 0.4, "learning_rate": 5.987744727968192e-05, "loss": 1.0222, "step": 160500 }, { "epoch": 0.4, "learning_rate": 5.9876189038200215e-05, "loss": 1.0193, "step": 160505 }, { "epoch": 0.4, "learning_rate": 5.9874930796718506e-05, "loss": 1.0215, "step": 160510 }, { "epoch": 0.4, "learning_rate": 5.9873672555236805e-05, "loss": 1.0206, "step": 160515 }, { "epoch": 0.4, "learning_rate": 5.9872414313755096e-05, "loss": 1.0223, "step": 160520 }, { "epoch": 0.4, "learning_rate": 5.9871156072273394e-05, "loss": 1.0196, "step": 160525 }, { "epoch": 0.4, "learning_rate": 5.9869897830791686e-05, "loss": 1.0215, "step": 160530 }, { "epoch": 0.4, "learning_rate": 5.9868639589309984e-05, "loss": 1.0211, "step": 160535 }, { "epoch": 0.4, "learning_rate": 5.9867381347828276e-05, "loss": 1.0202, "step": 160540 }, { "epoch": 0.4, "learning_rate": 5.9866123106346574e-05, "loss": 1.021, "step": 160545 }, { "epoch": 0.4, "learning_rate": 5.9864864864864866e-05, "loss": 1.019, "step": 160550 }, { "epoch": 0.4, "learning_rate": 5.9863606623383164e-05, "loss": 1.0211, "step": 160555 }, { "epoch": 0.4, "learning_rate": 5.9862348381901456e-05, "loss": 1.0203, "step": 160560 }, { "epoch": 0.4, "learning_rate": 5.9861090140419754e-05, "loss": 1.0195, "step": 160565 }, { "epoch": 0.4, "learning_rate": 5.9859831898938046e-05, "loss": 1.021, "step": 160570 }, { "epoch": 0.4, "learning_rate": 5.985857365745634e-05, "loss": 1.0185, "step": 160575 }, { "epoch": 0.4, "learning_rate": 5.9857315415974636e-05, "loss": 1.0205, "step": 160580 }, { "epoch": 0.4, "learning_rate": 5.985605717449293e-05, "loss": 1.0265, "step": 160585 }, { "epoch": 0.4, "learning_rate": 5.9854798933011226e-05, "loss": 1.046, "step": 160590 }, { "epoch": 0.4, "learning_rate": 5.985354069152952e-05, "loss": 1.0195, "step": 160595 }, { "epoch": 0.4, "learning_rate": 5.9852282450047816e-05, "loss": 1.0192, "step": 160600 }, { "epoch": 0.4, "learning_rate": 5.985102420856611e-05, "loss": 1.0208, "step": 160605 }, { "epoch": 0.4, "learning_rate": 5.9849765967084406e-05, "loss": 1.0203, "step": 160610 }, { "epoch": 0.4, "learning_rate": 5.98485077256027e-05, "loss": 1.021, "step": 160615 }, { "epoch": 0.4, "learning_rate": 5.9847249484120996e-05, "loss": 1.0222, "step": 160620 }, { "epoch": 0.4, "learning_rate": 5.984599124263929e-05, "loss": 1.0205, "step": 160625 }, { "epoch": 0.4, "learning_rate": 5.9844733001157586e-05, "loss": 1.0211, "step": 160630 }, { "epoch": 0.4, "learning_rate": 5.984347475967588e-05, "loss": 1.02, "step": 160635 }, { "epoch": 0.4, "learning_rate": 5.984221651819417e-05, "loss": 1.017, "step": 160640 }, { "epoch": 0.4, "learning_rate": 5.984095827671247e-05, "loss": 1.0228, "step": 160645 }, { "epoch": 0.4, "learning_rate": 5.983970003523076e-05, "loss": 1.0207, "step": 160650 }, { "epoch": 0.4, "learning_rate": 5.983844179374906e-05, "loss": 1.035, "step": 160655 }, { "epoch": 0.4, "learning_rate": 5.983718355226735e-05, "loss": 1.0193, "step": 160660 }, { "epoch": 0.4, "learning_rate": 5.983592531078565e-05, "loss": 1.0192, "step": 160665 }, { "epoch": 0.4, "learning_rate": 5.983466706930394e-05, "loss": 1.0213, "step": 160670 }, { "epoch": 0.4, "learning_rate": 5.983340882782224e-05, "loss": 1.0201, "step": 160675 }, { "epoch": 0.4, "learning_rate": 5.983215058634053e-05, "loss": 1.0245, "step": 160680 }, { "epoch": 0.4, "learning_rate": 5.983089234485883e-05, "loss": 1.0209, "step": 160685 }, { "epoch": 0.4, "learning_rate": 5.982963410337712e-05, "loss": 1.0199, "step": 160690 }, { "epoch": 0.4, "learning_rate": 5.982837586189541e-05, "loss": 1.0218, "step": 160695 }, { "epoch": 0.4, "learning_rate": 5.982711762041371e-05, "loss": 1.0456, "step": 160700 }, { "epoch": 0.4, "learning_rate": 5.9825859378932e-05, "loss": 1.0375, "step": 160705 }, { "epoch": 0.4, "learning_rate": 5.98246011374503e-05, "loss": 1.0214, "step": 160710 }, { "epoch": 0.4, "learning_rate": 5.982334289596859e-05, "loss": 1.0219, "step": 160715 }, { "epoch": 0.4, "learning_rate": 5.982208465448689e-05, "loss": 1.021, "step": 160720 }, { "epoch": 0.4, "learning_rate": 5.982082641300518e-05, "loss": 1.0182, "step": 160725 }, { "epoch": 0.4, "learning_rate": 5.981956817152348e-05, "loss": 1.0181, "step": 160730 }, { "epoch": 0.4, "learning_rate": 5.981830993004177e-05, "loss": 1.0215, "step": 160735 }, { "epoch": 0.4, "learning_rate": 5.981705168856007e-05, "loss": 1.0172, "step": 160740 }, { "epoch": 0.4, "learning_rate": 5.981579344707836e-05, "loss": 1.0199, "step": 160745 }, { "epoch": 0.4, "learning_rate": 5.981453520559666e-05, "loss": 1.0192, "step": 160750 }, { "epoch": 0.4, "learning_rate": 5.981327696411495e-05, "loss": 1.0203, "step": 160755 }, { "epoch": 0.4, "learning_rate": 5.981201872263324e-05, "loss": 1.02, "step": 160760 }, { "epoch": 0.4, "learning_rate": 5.9810760481151554e-05, "loss": 1.0219, "step": 160765 }, { "epoch": 0.4, "learning_rate": 5.9809502239669845e-05, "loss": 1.0209, "step": 160770 }, { "epoch": 0.4, "learning_rate": 5.9808243998188144e-05, "loss": 1.0179, "step": 160775 }, { "epoch": 0.4, "learning_rate": 5.9806985756706435e-05, "loss": 1.0193, "step": 160780 }, { "epoch": 0.4, "learning_rate": 5.980572751522473e-05, "loss": 1.0197, "step": 160785 }, { "epoch": 0.4, "learning_rate": 5.9804469273743025e-05, "loss": 1.0203, "step": 160790 }, { "epoch": 0.4, "learning_rate": 5.980321103226132e-05, "loss": 1.0192, "step": 160795 }, { "epoch": 0.4, "learning_rate": 5.9801952790779615e-05, "loss": 1.0193, "step": 160800 }, { "epoch": 0.4, "learning_rate": 5.980069454929791e-05, "loss": 1.0207, "step": 160805 }, { "epoch": 0.4, "learning_rate": 5.9799436307816205e-05, "loss": 1.0214, "step": 160810 }, { "epoch": 0.4, "learning_rate": 5.97981780663345e-05, "loss": 1.0203, "step": 160815 }, { "epoch": 0.4, "learning_rate": 5.9796919824852795e-05, "loss": 1.0204, "step": 160820 }, { "epoch": 0.4, "learning_rate": 5.979566158337109e-05, "loss": 1.0416, "step": 160825 }, { "epoch": 0.4, "learning_rate": 5.9794403341889385e-05, "loss": 1.0195, "step": 160830 }, { "epoch": 0.4, "learning_rate": 5.979314510040768e-05, "loss": 1.0199, "step": 160835 }, { "epoch": 0.4, "learning_rate": 5.979188685892597e-05, "loss": 1.0193, "step": 160840 }, { "epoch": 0.4, "learning_rate": 5.979062861744427e-05, "loss": 1.0418, "step": 160845 }, { "epoch": 0.4, "learning_rate": 5.978937037596256e-05, "loss": 1.0195, "step": 160850 }, { "epoch": 0.4, "learning_rate": 5.978811213448086e-05, "loss": 1.0184, "step": 160855 }, { "epoch": 0.4, "learning_rate": 5.978685389299915e-05, "loss": 1.0182, "step": 160860 }, { "epoch": 0.4, "learning_rate": 5.978559565151745e-05, "loss": 1.0196, "step": 160865 }, { "epoch": 0.4, "learning_rate": 5.978433741003574e-05, "loss": 1.0205, "step": 160870 }, { "epoch": 0.4, "learning_rate": 5.978307916855404e-05, "loss": 1.0219, "step": 160875 }, { "epoch": 0.4, "learning_rate": 5.978182092707233e-05, "loss": 1.0207, "step": 160880 }, { "epoch": 0.4, "learning_rate": 5.978056268559063e-05, "loss": 1.0192, "step": 160885 }, { "epoch": 0.4, "learning_rate": 5.977930444410892e-05, "loss": 1.019, "step": 160890 }, { "epoch": 0.4, "learning_rate": 5.977804620262722e-05, "loss": 1.02, "step": 160895 }, { "epoch": 0.4, "learning_rate": 5.977678796114551e-05, "loss": 1.0216, "step": 160900 }, { "epoch": 0.4, "learning_rate": 5.97755297196638e-05, "loss": 1.0198, "step": 160905 }, { "epoch": 0.4, "learning_rate": 5.97742714781821e-05, "loss": 1.0199, "step": 160910 }, { "epoch": 0.4, "learning_rate": 5.977301323670039e-05, "loss": 1.0214, "step": 160915 }, { "epoch": 0.4, "learning_rate": 5.977175499521869e-05, "loss": 1.0202, "step": 160920 }, { "epoch": 0.4, "learning_rate": 5.977049675373698e-05, "loss": 1.0202, "step": 160925 }, { "epoch": 0.4, "learning_rate": 5.976923851225528e-05, "loss": 1.0201, "step": 160930 }, { "epoch": 0.4, "learning_rate": 5.976798027077357e-05, "loss": 1.0222, "step": 160935 }, { "epoch": 0.4, "learning_rate": 5.976672202929187e-05, "loss": 1.0193, "step": 160940 }, { "epoch": 0.4, "learning_rate": 5.976546378781016e-05, "loss": 1.0178, "step": 160945 }, { "epoch": 0.4, "learning_rate": 5.976420554632846e-05, "loss": 1.0206, "step": 160950 }, { "epoch": 0.4, "learning_rate": 5.976294730484675e-05, "loss": 1.0193, "step": 160955 }, { "epoch": 0.4, "learning_rate": 5.976168906336505e-05, "loss": 1.0187, "step": 160960 }, { "epoch": 0.4, "learning_rate": 5.976043082188334e-05, "loss": 1.0215, "step": 160965 }, { "epoch": 0.4, "learning_rate": 5.975917258040163e-05, "loss": 1.0204, "step": 160970 }, { "epoch": 0.4, "learning_rate": 5.975791433891993e-05, "loss": 1.0206, "step": 160975 }, { "epoch": 0.4, "learning_rate": 5.975665609743822e-05, "loss": 1.0198, "step": 160980 }, { "epoch": 0.4, "learning_rate": 5.975539785595652e-05, "loss": 1.02, "step": 160985 }, { "epoch": 0.4, "learning_rate": 5.975413961447481e-05, "loss": 1.0403, "step": 160990 }, { "epoch": 0.4, "learning_rate": 5.975288137299311e-05, "loss": 1.0193, "step": 160995 }, { "epoch": 0.4, "learning_rate": 5.97516231315114e-05, "loss": 1.019, "step": 161000 }, { "epoch": 0.4, "learning_rate": 5.97503648900297e-05, "loss": 1.0197, "step": 161005 }, { "epoch": 0.4, "learning_rate": 5.974910664854799e-05, "loss": 1.0185, "step": 161010 }, { "epoch": 0.4, "learning_rate": 5.974784840706629e-05, "loss": 1.0188, "step": 161015 }, { "epoch": 0.4, "learning_rate": 5.974659016558458e-05, "loss": 1.0199, "step": 161020 }, { "epoch": 0.4, "learning_rate": 5.974533192410288e-05, "loss": 1.0212, "step": 161025 }, { "epoch": 0.4, "learning_rate": 5.974407368262117e-05, "loss": 1.0175, "step": 161030 }, { "epoch": 0.4, "learning_rate": 5.974281544113946e-05, "loss": 1.0178, "step": 161035 }, { "epoch": 0.4, "learning_rate": 5.974155719965776e-05, "loss": 1.0215, "step": 161040 }, { "epoch": 0.4, "learning_rate": 5.974029895817605e-05, "loss": 1.022, "step": 161045 }, { "epoch": 0.4, "learning_rate": 5.973904071669435e-05, "loss": 1.0216, "step": 161050 }, { "epoch": 0.4, "learning_rate": 5.973778247521264e-05, "loss": 1.0185, "step": 161055 }, { "epoch": 0.4, "learning_rate": 5.973652423373094e-05, "loss": 1.0211, "step": 161060 }, { "epoch": 0.4, "learning_rate": 5.973526599224923e-05, "loss": 1.0204, "step": 161065 }, { "epoch": 0.4, "learning_rate": 5.973400775076753e-05, "loss": 1.0191, "step": 161070 }, { "epoch": 0.4, "learning_rate": 5.973274950928582e-05, "loss": 1.0194, "step": 161075 }, { "epoch": 0.4, "learning_rate": 5.973149126780412e-05, "loss": 1.0186, "step": 161080 }, { "epoch": 0.4, "learning_rate": 5.973023302632241e-05, "loss": 1.0209, "step": 161085 }, { "epoch": 0.4, "learning_rate": 5.972897478484071e-05, "loss": 1.0198, "step": 161090 }, { "epoch": 0.4, "learning_rate": 5.9727716543359e-05, "loss": 1.0196, "step": 161095 }, { "epoch": 0.4, "learning_rate": 5.9726458301877294e-05, "loss": 1.02, "step": 161100 }, { "epoch": 0.4, "learning_rate": 5.972520006039559e-05, "loss": 1.0234, "step": 161105 }, { "epoch": 0.4, "learning_rate": 5.9723941818913884e-05, "loss": 1.0189, "step": 161110 }, { "epoch": 0.4, "learning_rate": 5.972268357743218e-05, "loss": 1.021, "step": 161115 }, { "epoch": 0.4, "learning_rate": 5.9721425335950474e-05, "loss": 1.0207, "step": 161120 }, { "epoch": 0.4, "learning_rate": 5.972016709446877e-05, "loss": 1.02, "step": 161125 }, { "epoch": 0.4, "learning_rate": 5.9718908852987064e-05, "loss": 1.0188, "step": 161130 }, { "epoch": 0.4, "learning_rate": 5.971765061150536e-05, "loss": 1.0201, "step": 161135 }, { "epoch": 0.4, "learning_rate": 5.9716392370023654e-05, "loss": 1.0184, "step": 161140 }, { "epoch": 0.4, "learning_rate": 5.971513412854195e-05, "loss": 1.0207, "step": 161145 }, { "epoch": 0.4, "learning_rate": 5.9713875887060244e-05, "loss": 1.0201, "step": 161150 }, { "epoch": 0.4, "learning_rate": 5.971261764557854e-05, "loss": 1.0196, "step": 161155 }, { "epoch": 0.4, "learning_rate": 5.9711359404096834e-05, "loss": 1.0204, "step": 161160 }, { "epoch": 0.4, "learning_rate": 5.9710101162615126e-05, "loss": 1.017, "step": 161165 }, { "epoch": 0.4, "learning_rate": 5.9708842921133424e-05, "loss": 1.0207, "step": 161170 }, { "epoch": 0.4, "learning_rate": 5.9707584679651716e-05, "loss": 1.0212, "step": 161175 }, { "epoch": 0.4, "learning_rate": 5.9706326438170014e-05, "loss": 1.0205, "step": 161180 }, { "epoch": 0.4, "learning_rate": 5.9705068196688305e-05, "loss": 1.0206, "step": 161185 }, { "epoch": 0.4, "learning_rate": 5.9703809955206604e-05, "loss": 1.0208, "step": 161190 }, { "epoch": 0.4, "learning_rate": 5.9702551713724895e-05, "loss": 1.0173, "step": 161195 }, { "epoch": 0.4, "learning_rate": 5.9701293472243194e-05, "loss": 1.0202, "step": 161200 }, { "epoch": 0.4, "learning_rate": 5.9700035230761485e-05, "loss": 1.0393, "step": 161205 }, { "epoch": 0.4, "learning_rate": 5.9698776989279784e-05, "loss": 1.0202, "step": 161210 }, { "epoch": 0.4, "learning_rate": 5.9697518747798075e-05, "loss": 1.0206, "step": 161215 }, { "epoch": 0.4, "learning_rate": 5.9696260506316374e-05, "loss": 1.0363, "step": 161220 }, { "epoch": 0.4, "learning_rate": 5.9695002264834665e-05, "loss": 1.02, "step": 161225 }, { "epoch": 0.4, "learning_rate": 5.969374402335296e-05, "loss": 1.0437, "step": 161230 }, { "epoch": 0.4, "learning_rate": 5.9692485781871255e-05, "loss": 1.0184, "step": 161235 }, { "epoch": 0.4, "learning_rate": 5.969122754038955e-05, "loss": 1.02, "step": 161240 }, { "epoch": 0.4, "learning_rate": 5.9689969298907845e-05, "loss": 1.0221, "step": 161245 }, { "epoch": 0.4, "learning_rate": 5.968871105742614e-05, "loss": 1.0195, "step": 161250 }, { "epoch": 0.4, "learning_rate": 5.9687452815944435e-05, "loss": 1.0212, "step": 161255 }, { "epoch": 0.4, "learning_rate": 5.968619457446273e-05, "loss": 1.0221, "step": 161260 }, { "epoch": 0.4, "learning_rate": 5.968493633298103e-05, "loss": 1.0276, "step": 161265 }, { "epoch": 0.4, "learning_rate": 5.968367809149933e-05, "loss": 1.0215, "step": 161270 }, { "epoch": 0.4, "learning_rate": 5.968241985001762e-05, "loss": 1.0187, "step": 161275 }, { "epoch": 0.4, "learning_rate": 5.968116160853592e-05, "loss": 1.0186, "step": 161280 }, { "epoch": 0.4, "learning_rate": 5.967990336705421e-05, "loss": 1.0196, "step": 161285 }, { "epoch": 0.4, "learning_rate": 5.967864512557251e-05, "loss": 1.0214, "step": 161290 }, { "epoch": 0.4, "learning_rate": 5.96773868840908e-05, "loss": 1.0206, "step": 161295 }, { "epoch": 0.4, "learning_rate": 5.96761286426091e-05, "loss": 1.0181, "step": 161300 }, { "epoch": 0.4, "learning_rate": 5.967487040112739e-05, "loss": 1.0164, "step": 161305 }, { "epoch": 0.4, "learning_rate": 5.9673612159645684e-05, "loss": 1.0201, "step": 161310 }, { "epoch": 0.4, "learning_rate": 5.967235391816398e-05, "loss": 1.0207, "step": 161315 }, { "epoch": 0.4, "learning_rate": 5.9671095676682273e-05, "loss": 1.0193, "step": 161320 }, { "epoch": 0.4, "learning_rate": 5.966983743520057e-05, "loss": 1.0182, "step": 161325 }, { "epoch": 0.4, "learning_rate": 5.9668579193718863e-05, "loss": 1.0207, "step": 161330 }, { "epoch": 0.4, "learning_rate": 5.966732095223716e-05, "loss": 1.0184, "step": 161335 }, { "epoch": 0.4, "learning_rate": 5.9666062710755453e-05, "loss": 1.0199, "step": 161340 }, { "epoch": 0.41, "learning_rate": 5.966480446927375e-05, "loss": 1.019, "step": 161345 }, { "epoch": 0.41, "learning_rate": 5.9663546227792043e-05, "loss": 1.0197, "step": 161350 }, { "epoch": 0.41, "learning_rate": 5.966228798631034e-05, "loss": 1.0178, "step": 161355 }, { "epoch": 0.41, "learning_rate": 5.966102974482863e-05, "loss": 1.0195, "step": 161360 }, { "epoch": 0.41, "learning_rate": 5.965977150334693e-05, "loss": 1.0217, "step": 161365 }, { "epoch": 0.41, "learning_rate": 5.965851326186522e-05, "loss": 1.0182, "step": 161370 }, { "epoch": 0.41, "learning_rate": 5.9657255020383515e-05, "loss": 1.0238, "step": 161375 }, { "epoch": 0.41, "learning_rate": 5.965599677890181e-05, "loss": 1.0173, "step": 161380 }, { "epoch": 0.41, "learning_rate": 5.9654738537420105e-05, "loss": 1.019, "step": 161385 }, { "epoch": 0.41, "learning_rate": 5.96534802959384e-05, "loss": 1.0208, "step": 161390 }, { "epoch": 0.41, "learning_rate": 5.9652222054456695e-05, "loss": 1.0177, "step": 161395 }, { "epoch": 0.41, "learning_rate": 5.965096381297499e-05, "loss": 1.0176, "step": 161400 }, { "epoch": 0.41, "learning_rate": 5.9649705571493285e-05, "loss": 1.0191, "step": 161405 }, { "epoch": 0.41, "learning_rate": 5.964844733001158e-05, "loss": 1.0212, "step": 161410 }, { "epoch": 0.41, "learning_rate": 5.9647189088529875e-05, "loss": 1.0187, "step": 161415 }, { "epoch": 0.41, "learning_rate": 5.964593084704817e-05, "loss": 1.0214, "step": 161420 }, { "epoch": 0.41, "learning_rate": 5.9644672605566465e-05, "loss": 1.0202, "step": 161425 }, { "epoch": 0.41, "learning_rate": 5.964341436408476e-05, "loss": 1.0196, "step": 161430 }, { "epoch": 0.41, "learning_rate": 5.9642156122603055e-05, "loss": 1.0198, "step": 161435 }, { "epoch": 0.41, "learning_rate": 5.9640897881121346e-05, "loss": 1.0206, "step": 161440 }, { "epoch": 0.41, "learning_rate": 5.9639639639639645e-05, "loss": 1.0204, "step": 161445 }, { "epoch": 0.41, "learning_rate": 5.9638381398157936e-05, "loss": 1.0387, "step": 161450 }, { "epoch": 0.41, "learning_rate": 5.9637123156676235e-05, "loss": 1.0214, "step": 161455 }, { "epoch": 0.41, "learning_rate": 5.9635864915194526e-05, "loss": 1.0181, "step": 161460 }, { "epoch": 0.41, "learning_rate": 5.9634606673712825e-05, "loss": 1.0189, "step": 161465 }, { "epoch": 0.41, "learning_rate": 5.9633348432231116e-05, "loss": 1.0218, "step": 161470 }, { "epoch": 0.41, "learning_rate": 5.9632090190749415e-05, "loss": 1.0207, "step": 161475 }, { "epoch": 0.41, "learning_rate": 5.9630831949267706e-05, "loss": 1.0246, "step": 161480 }, { "epoch": 0.41, "learning_rate": 5.9629573707786005e-05, "loss": 1.021, "step": 161485 }, { "epoch": 0.41, "learning_rate": 5.9628315466304296e-05, "loss": 1.019, "step": 161490 }, { "epoch": 0.41, "learning_rate": 5.962705722482259e-05, "loss": 1.0212, "step": 161495 }, { "epoch": 0.41, "learning_rate": 5.9625798983340886e-05, "loss": 1.0188, "step": 161500 }, { "epoch": 0.41, "learning_rate": 5.962454074185918e-05, "loss": 1.0224, "step": 161505 }, { "epoch": 0.41, "learning_rate": 5.9623282500377476e-05, "loss": 1.0183, "step": 161510 }, { "epoch": 0.41, "learning_rate": 5.962202425889577e-05, "loss": 1.021, "step": 161515 }, { "epoch": 0.41, "learning_rate": 5.9620766017414066e-05, "loss": 1.0205, "step": 161520 }, { "epoch": 0.41, "learning_rate": 5.961950777593236e-05, "loss": 1.0216, "step": 161525 }, { "epoch": 0.41, "learning_rate": 5.9618249534450656e-05, "loss": 1.0205, "step": 161530 }, { "epoch": 0.41, "learning_rate": 5.961699129296895e-05, "loss": 1.0173, "step": 161535 }, { "epoch": 0.41, "learning_rate": 5.9615733051487246e-05, "loss": 1.0201, "step": 161540 }, { "epoch": 0.41, "learning_rate": 5.961447481000554e-05, "loss": 1.0182, "step": 161545 }, { "epoch": 0.41, "learning_rate": 5.9613216568523836e-05, "loss": 1.0197, "step": 161550 }, { "epoch": 0.41, "learning_rate": 5.961195832704213e-05, "loss": 1.0193, "step": 161555 }, { "epoch": 0.41, "learning_rate": 5.961070008556042e-05, "loss": 1.0185, "step": 161560 }, { "epoch": 0.41, "learning_rate": 5.960944184407872e-05, "loss": 1.0196, "step": 161565 }, { "epoch": 0.41, "learning_rate": 5.960818360259701e-05, "loss": 1.0179, "step": 161570 }, { "epoch": 0.41, "learning_rate": 5.960692536111531e-05, "loss": 1.0212, "step": 161575 }, { "epoch": 0.41, "learning_rate": 5.96056671196336e-05, "loss": 1.0221, "step": 161580 }, { "epoch": 0.41, "learning_rate": 5.96044088781519e-05, "loss": 1.0238, "step": 161585 }, { "epoch": 0.41, "learning_rate": 5.960315063667019e-05, "loss": 1.0182, "step": 161590 }, { "epoch": 0.41, "learning_rate": 5.960189239518849e-05, "loss": 1.0187, "step": 161595 }, { "epoch": 0.41, "learning_rate": 5.960063415370678e-05, "loss": 1.0199, "step": 161600 }, { "epoch": 0.41, "learning_rate": 5.959937591222508e-05, "loss": 1.0191, "step": 161605 }, { "epoch": 0.41, "learning_rate": 5.959811767074337e-05, "loss": 1.021, "step": 161610 }, { "epoch": 0.41, "learning_rate": 5.959685942926167e-05, "loss": 1.0188, "step": 161615 }, { "epoch": 0.41, "learning_rate": 5.959560118777996e-05, "loss": 1.0185, "step": 161620 }, { "epoch": 0.41, "learning_rate": 5.959434294629825e-05, "loss": 1.0218, "step": 161625 }, { "epoch": 0.41, "learning_rate": 5.959308470481655e-05, "loss": 1.019, "step": 161630 }, { "epoch": 0.41, "learning_rate": 5.959182646333484e-05, "loss": 1.019, "step": 161635 }, { "epoch": 0.41, "learning_rate": 5.959056822185314e-05, "loss": 1.0187, "step": 161640 }, { "epoch": 0.41, "learning_rate": 5.958930998037143e-05, "loss": 1.0194, "step": 161645 }, { "epoch": 0.41, "learning_rate": 5.958805173888973e-05, "loss": 1.0207, "step": 161650 }, { "epoch": 0.41, "learning_rate": 5.958679349740802e-05, "loss": 1.0215, "step": 161655 }, { "epoch": 0.41, "learning_rate": 5.958553525592632e-05, "loss": 1.021, "step": 161660 }, { "epoch": 0.41, "learning_rate": 5.958427701444461e-05, "loss": 1.0188, "step": 161665 }, { "epoch": 0.41, "learning_rate": 5.958301877296291e-05, "loss": 1.0182, "step": 161670 }, { "epoch": 0.41, "learning_rate": 5.95817605314812e-05, "loss": 1.0212, "step": 161675 }, { "epoch": 0.41, "learning_rate": 5.95805022899995e-05, "loss": 1.0199, "step": 161680 }, { "epoch": 0.41, "learning_rate": 5.957924404851779e-05, "loss": 1.0469, "step": 161685 }, { "epoch": 0.41, "learning_rate": 5.957798580703608e-05, "loss": 1.02, "step": 161690 }, { "epoch": 0.41, "learning_rate": 5.957672756555438e-05, "loss": 1.02, "step": 161695 }, { "epoch": 0.41, "learning_rate": 5.957546932407267e-05, "loss": 1.0184, "step": 161700 }, { "epoch": 0.41, "learning_rate": 5.957421108259097e-05, "loss": 1.0194, "step": 161705 }, { "epoch": 0.41, "learning_rate": 5.957295284110926e-05, "loss": 1.0197, "step": 161710 }, { "epoch": 0.41, "learning_rate": 5.957169459962756e-05, "loss": 1.0179, "step": 161715 }, { "epoch": 0.41, "learning_rate": 5.957043635814585e-05, "loss": 1.0177, "step": 161720 }, { "epoch": 0.41, "learning_rate": 5.956917811666415e-05, "loss": 1.0202, "step": 161725 }, { "epoch": 0.41, "learning_rate": 5.956791987518244e-05, "loss": 1.0208, "step": 161730 }, { "epoch": 0.41, "learning_rate": 5.956666163370074e-05, "loss": 1.0207, "step": 161735 }, { "epoch": 0.41, "learning_rate": 5.956540339221903e-05, "loss": 1.0178, "step": 161740 }, { "epoch": 0.41, "learning_rate": 5.956414515073733e-05, "loss": 1.0186, "step": 161745 }, { "epoch": 0.41, "learning_rate": 5.956288690925562e-05, "loss": 1.0221, "step": 161750 }, { "epoch": 0.41, "learning_rate": 5.9561628667773914e-05, "loss": 1.0193, "step": 161755 }, { "epoch": 0.41, "learning_rate": 5.956037042629221e-05, "loss": 1.0217, "step": 161760 }, { "epoch": 0.41, "learning_rate": 5.955911218481052e-05, "loss": 1.0209, "step": 161765 }, { "epoch": 0.41, "learning_rate": 5.955785394332881e-05, "loss": 1.021, "step": 161770 }, { "epoch": 0.41, "learning_rate": 5.955659570184711e-05, "loss": 1.0225, "step": 161775 }, { "epoch": 0.41, "learning_rate": 5.95553374603654e-05, "loss": 1.0237, "step": 161780 }, { "epoch": 0.41, "learning_rate": 5.95540792188837e-05, "loss": 1.0185, "step": 161785 }, { "epoch": 0.41, "learning_rate": 5.955282097740199e-05, "loss": 1.0189, "step": 161790 }, { "epoch": 0.41, "learning_rate": 5.955156273592029e-05, "loss": 1.0214, "step": 161795 }, { "epoch": 0.41, "learning_rate": 5.955030449443858e-05, "loss": 1.0185, "step": 161800 }, { "epoch": 0.41, "learning_rate": 5.954904625295688e-05, "loss": 1.0203, "step": 161805 }, { "epoch": 0.41, "learning_rate": 5.954778801147517e-05, "loss": 1.0196, "step": 161810 }, { "epoch": 0.41, "learning_rate": 5.954652976999347e-05, "loss": 1.0173, "step": 161815 }, { "epoch": 0.41, "learning_rate": 5.954527152851176e-05, "loss": 1.0216, "step": 161820 }, { "epoch": 0.41, "learning_rate": 5.954401328703006e-05, "loss": 1.0198, "step": 161825 }, { "epoch": 0.41, "learning_rate": 5.954275504554835e-05, "loss": 1.0196, "step": 161830 }, { "epoch": 0.41, "learning_rate": 5.954149680406664e-05, "loss": 1.0198, "step": 161835 }, { "epoch": 0.41, "learning_rate": 5.954023856258494e-05, "loss": 1.021, "step": 161840 }, { "epoch": 0.41, "learning_rate": 5.953898032110323e-05, "loss": 1.0197, "step": 161845 }, { "epoch": 0.41, "learning_rate": 5.953772207962153e-05, "loss": 1.0191, "step": 161850 }, { "epoch": 0.41, "learning_rate": 5.953646383813982e-05, "loss": 1.0202, "step": 161855 }, { "epoch": 0.41, "learning_rate": 5.953520559665812e-05, "loss": 1.0217, "step": 161860 }, { "epoch": 0.41, "learning_rate": 5.953394735517641e-05, "loss": 1.0176, "step": 161865 }, { "epoch": 0.41, "learning_rate": 5.953268911369471e-05, "loss": 1.0184, "step": 161870 }, { "epoch": 0.41, "learning_rate": 5.9531430872213e-05, "loss": 1.0199, "step": 161875 }, { "epoch": 0.41, "learning_rate": 5.95301726307313e-05, "loss": 1.0209, "step": 161880 }, { "epoch": 0.41, "learning_rate": 5.952891438924959e-05, "loss": 1.0215, "step": 161885 }, { "epoch": 0.41, "learning_rate": 5.952765614776789e-05, "loss": 1.0172, "step": 161890 }, { "epoch": 0.41, "learning_rate": 5.952639790628618e-05, "loss": 1.0212, "step": 161895 }, { "epoch": 0.41, "learning_rate": 5.952513966480447e-05, "loss": 1.0206, "step": 161900 }, { "epoch": 0.41, "learning_rate": 5.952388142332277e-05, "loss": 1.0211, "step": 161905 }, { "epoch": 0.41, "learning_rate": 5.952262318184106e-05, "loss": 1.0212, "step": 161910 }, { "epoch": 0.41, "learning_rate": 5.952136494035936e-05, "loss": 1.0178, "step": 161915 }, { "epoch": 0.41, "learning_rate": 5.952010669887765e-05, "loss": 1.0185, "step": 161920 }, { "epoch": 0.41, "learning_rate": 5.951884845739595e-05, "loss": 1.0219, "step": 161925 }, { "epoch": 0.41, "learning_rate": 5.951759021591424e-05, "loss": 1.0207, "step": 161930 }, { "epoch": 0.41, "learning_rate": 5.951633197443254e-05, "loss": 1.0195, "step": 161935 }, { "epoch": 0.41, "learning_rate": 5.951507373295083e-05, "loss": 1.0205, "step": 161940 }, { "epoch": 0.41, "learning_rate": 5.951381549146913e-05, "loss": 1.0186, "step": 161945 }, { "epoch": 0.41, "learning_rate": 5.951255724998742e-05, "loss": 1.0188, "step": 161950 }, { "epoch": 0.41, "learning_rate": 5.951129900850572e-05, "loss": 1.0219, "step": 161955 }, { "epoch": 0.41, "learning_rate": 5.951004076702401e-05, "loss": 1.0185, "step": 161960 }, { "epoch": 0.41, "learning_rate": 5.95087825255423e-05, "loss": 1.0209, "step": 161965 }, { "epoch": 0.41, "learning_rate": 5.95075242840606e-05, "loss": 1.0202, "step": 161970 }, { "epoch": 0.41, "learning_rate": 5.950626604257889e-05, "loss": 1.0211, "step": 161975 }, { "epoch": 0.41, "learning_rate": 5.950500780109719e-05, "loss": 1.0216, "step": 161980 }, { "epoch": 0.41, "learning_rate": 5.950374955961548e-05, "loss": 1.019, "step": 161985 }, { "epoch": 0.41, "learning_rate": 5.950249131813378e-05, "loss": 1.0204, "step": 161990 }, { "epoch": 0.41, "learning_rate": 5.950123307665207e-05, "loss": 1.019, "step": 161995 }, { "epoch": 0.41, "learning_rate": 5.949997483517037e-05, "loss": 1.0182, "step": 162000 }, { "epoch": 0.41, "learning_rate": 5.949871659368866e-05, "loss": 1.0186, "step": 162005 }, { "epoch": 0.41, "learning_rate": 5.949745835220696e-05, "loss": 1.0204, "step": 162010 }, { "epoch": 0.41, "learning_rate": 5.949620011072525e-05, "loss": 1.0198, "step": 162015 }, { "epoch": 0.41, "learning_rate": 5.949494186924355e-05, "loss": 1.0195, "step": 162020 }, { "epoch": 0.41, "learning_rate": 5.949368362776184e-05, "loss": 1.0205, "step": 162025 }, { "epoch": 0.41, "learning_rate": 5.9492425386280134e-05, "loss": 1.0217, "step": 162030 }, { "epoch": 0.41, "learning_rate": 5.949116714479843e-05, "loss": 1.0194, "step": 162035 }, { "epoch": 0.41, "learning_rate": 5.9489908903316724e-05, "loss": 1.0181, "step": 162040 }, { "epoch": 0.41, "learning_rate": 5.948865066183502e-05, "loss": 1.0219, "step": 162045 }, { "epoch": 0.41, "learning_rate": 5.9487392420353314e-05, "loss": 1.0199, "step": 162050 }, { "epoch": 0.41, "learning_rate": 5.948613417887161e-05, "loss": 1.02, "step": 162055 }, { "epoch": 0.41, "learning_rate": 5.9484875937389904e-05, "loss": 1.0181, "step": 162060 }, { "epoch": 0.41, "learning_rate": 5.94836176959082e-05, "loss": 1.0193, "step": 162065 }, { "epoch": 0.41, "learning_rate": 5.9482359454426494e-05, "loss": 1.0192, "step": 162070 }, { "epoch": 0.41, "learning_rate": 5.948110121294479e-05, "loss": 1.0176, "step": 162075 }, { "epoch": 0.41, "learning_rate": 5.9479842971463084e-05, "loss": 1.0211, "step": 162080 }, { "epoch": 0.41, "learning_rate": 5.947858472998138e-05, "loss": 1.0212, "step": 162085 }, { "epoch": 0.41, "learning_rate": 5.9477326488499674e-05, "loss": 1.0197, "step": 162090 }, { "epoch": 0.41, "learning_rate": 5.9476068247017966e-05, "loss": 1.0212, "step": 162095 }, { "epoch": 0.41, "learning_rate": 5.9474810005536264e-05, "loss": 1.0176, "step": 162100 }, { "epoch": 0.41, "learning_rate": 5.9473551764054556e-05, "loss": 1.0198, "step": 162105 }, { "epoch": 0.41, "learning_rate": 5.9472293522572854e-05, "loss": 1.0193, "step": 162110 }, { "epoch": 0.41, "learning_rate": 5.9471035281091146e-05, "loss": 1.0213, "step": 162115 }, { "epoch": 0.41, "learning_rate": 5.9469777039609444e-05, "loss": 1.0203, "step": 162120 }, { "epoch": 0.41, "learning_rate": 5.9468518798127736e-05, "loss": 1.0201, "step": 162125 }, { "epoch": 0.41, "learning_rate": 5.9467260556646034e-05, "loss": 1.0216, "step": 162130 }, { "epoch": 0.41, "learning_rate": 5.9466002315164326e-05, "loss": 1.0185, "step": 162135 }, { "epoch": 0.41, "learning_rate": 5.9464744073682624e-05, "loss": 1.0196, "step": 162140 }, { "epoch": 0.41, "learning_rate": 5.9463485832200916e-05, "loss": 1.0217, "step": 162145 }, { "epoch": 0.41, "learning_rate": 5.946222759071921e-05, "loss": 1.019, "step": 162150 }, { "epoch": 0.41, "learning_rate": 5.9460969349237506e-05, "loss": 1.0207, "step": 162155 }, { "epoch": 0.41, "learning_rate": 5.94597111077558e-05, "loss": 1.0198, "step": 162160 }, { "epoch": 0.41, "learning_rate": 5.9458452866274096e-05, "loss": 1.0194, "step": 162165 }, { "epoch": 0.41, "learning_rate": 5.945719462479239e-05, "loss": 1.0218, "step": 162170 }, { "epoch": 0.41, "learning_rate": 5.9455936383310686e-05, "loss": 1.0213, "step": 162175 }, { "epoch": 0.41, "learning_rate": 5.945467814182898e-05, "loss": 1.0215, "step": 162180 }, { "epoch": 0.41, "learning_rate": 5.9453419900347275e-05, "loss": 1.0203, "step": 162185 }, { "epoch": 0.41, "learning_rate": 5.945216165886557e-05, "loss": 1.0199, "step": 162190 }, { "epoch": 0.41, "learning_rate": 5.9450903417383865e-05, "loss": 1.0392, "step": 162195 }, { "epoch": 0.41, "learning_rate": 5.944964517590216e-05, "loss": 1.0203, "step": 162200 }, { "epoch": 0.41, "learning_rate": 5.9448386934420455e-05, "loss": 1.0199, "step": 162205 }, { "epoch": 0.41, "learning_rate": 5.944712869293875e-05, "loss": 1.0186, "step": 162210 }, { "epoch": 0.41, "learning_rate": 5.944587045145704e-05, "loss": 1.0225, "step": 162215 }, { "epoch": 0.41, "learning_rate": 5.944461220997534e-05, "loss": 1.0196, "step": 162220 }, { "epoch": 0.41, "learning_rate": 5.944335396849363e-05, "loss": 1.0196, "step": 162225 }, { "epoch": 0.41, "learning_rate": 5.944209572701193e-05, "loss": 1.0206, "step": 162230 }, { "epoch": 0.41, "learning_rate": 5.944083748553022e-05, "loss": 1.0438, "step": 162235 }, { "epoch": 0.41, "learning_rate": 5.943957924404852e-05, "loss": 1.0165, "step": 162240 }, { "epoch": 0.41, "learning_rate": 5.943832100256681e-05, "loss": 1.0186, "step": 162245 }, { "epoch": 0.41, "learning_rate": 5.943706276108511e-05, "loss": 1.0177, "step": 162250 }, { "epoch": 0.41, "learning_rate": 5.94358045196034e-05, "loss": 1.021, "step": 162255 }, { "epoch": 0.41, "learning_rate": 5.9434546278121704e-05, "loss": 1.0212, "step": 162260 }, { "epoch": 0.41, "learning_rate": 5.943328803664e-05, "loss": 1.0204, "step": 162265 }, { "epoch": 0.41, "learning_rate": 5.9432029795158294e-05, "loss": 1.0189, "step": 162270 }, { "epoch": 0.41, "learning_rate": 5.943077155367659e-05, "loss": 1.0202, "step": 162275 }, { "epoch": 0.41, "learning_rate": 5.9429513312194884e-05, "loss": 1.0195, "step": 162280 }, { "epoch": 0.41, "learning_rate": 5.942825507071318e-05, "loss": 1.02, "step": 162285 }, { "epoch": 0.41, "learning_rate": 5.9426996829231474e-05, "loss": 1.0197, "step": 162290 }, { "epoch": 0.41, "learning_rate": 5.9425738587749765e-05, "loss": 1.0193, "step": 162295 }, { "epoch": 0.41, "learning_rate": 5.9424480346268064e-05, "loss": 1.0179, "step": 162300 }, { "epoch": 0.41, "learning_rate": 5.9423222104786355e-05, "loss": 1.0187, "step": 162305 }, { "epoch": 0.41, "learning_rate": 5.9421963863304654e-05, "loss": 1.0211, "step": 162310 }, { "epoch": 0.41, "learning_rate": 5.9420705621822945e-05, "loss": 1.0191, "step": 162315 }, { "epoch": 0.41, "learning_rate": 5.9419447380341243e-05, "loss": 1.0192, "step": 162320 }, { "epoch": 0.41, "learning_rate": 5.9418189138859535e-05, "loss": 1.0191, "step": 162325 }, { "epoch": 0.41, "learning_rate": 5.9416930897377833e-05, "loss": 1.0176, "step": 162330 }, { "epoch": 0.41, "learning_rate": 5.9415672655896125e-05, "loss": 1.0212, "step": 162335 }, { "epoch": 0.41, "learning_rate": 5.9414414414414423e-05, "loss": 1.0215, "step": 162340 }, { "epoch": 0.41, "learning_rate": 5.9413156172932715e-05, "loss": 1.0189, "step": 162345 }, { "epoch": 0.41, "learning_rate": 5.9411897931451013e-05, "loss": 1.0192, "step": 162350 }, { "epoch": 0.41, "learning_rate": 5.9410639689969305e-05, "loss": 1.0206, "step": 162355 }, { "epoch": 0.41, "learning_rate": 5.9409381448487597e-05, "loss": 1.0221, "step": 162360 }, { "epoch": 0.41, "learning_rate": 5.9408123207005895e-05, "loss": 1.0214, "step": 162365 }, { "epoch": 0.41, "learning_rate": 5.9406864965524187e-05, "loss": 1.0202, "step": 162370 }, { "epoch": 0.41, "learning_rate": 5.9405606724042485e-05, "loss": 1.0197, "step": 162375 }, { "epoch": 0.41, "learning_rate": 5.9404348482560777e-05, "loss": 1.0192, "step": 162380 }, { "epoch": 0.41, "learning_rate": 5.9403090241079075e-05, "loss": 1.0215, "step": 162385 }, { "epoch": 0.41, "learning_rate": 5.9401831999597367e-05, "loss": 1.0199, "step": 162390 }, { "epoch": 0.41, "learning_rate": 5.9400573758115665e-05, "loss": 1.0211, "step": 162395 }, { "epoch": 0.41, "learning_rate": 5.9399315516633956e-05, "loss": 1.0203, "step": 162400 }, { "epoch": 0.41, "learning_rate": 5.9398057275152255e-05, "loss": 1.0206, "step": 162405 }, { "epoch": 0.41, "learning_rate": 5.9396799033670546e-05, "loss": 1.0193, "step": 162410 }, { "epoch": 0.41, "learning_rate": 5.9395540792188845e-05, "loss": 1.0224, "step": 162415 }, { "epoch": 0.41, "learning_rate": 5.9394282550707136e-05, "loss": 1.02, "step": 162420 }, { "epoch": 0.41, "learning_rate": 5.939302430922543e-05, "loss": 1.0223, "step": 162425 }, { "epoch": 0.41, "learning_rate": 5.9391766067743726e-05, "loss": 1.0188, "step": 162430 }, { "epoch": 0.41, "learning_rate": 5.939050782626202e-05, "loss": 1.0215, "step": 162435 }, { "epoch": 0.41, "learning_rate": 5.9389249584780316e-05, "loss": 1.0193, "step": 162440 }, { "epoch": 0.41, "learning_rate": 5.938799134329861e-05, "loss": 1.0201, "step": 162445 }, { "epoch": 0.41, "learning_rate": 5.9386733101816906e-05, "loss": 1.018, "step": 162450 }, { "epoch": 0.41, "learning_rate": 5.93854748603352e-05, "loss": 1.0214, "step": 162455 }, { "epoch": 0.41, "learning_rate": 5.9384216618853496e-05, "loss": 1.0202, "step": 162460 }, { "epoch": 0.41, "learning_rate": 5.938295837737179e-05, "loss": 1.0212, "step": 162465 }, { "epoch": 0.41, "learning_rate": 5.9381700135890086e-05, "loss": 1.0195, "step": 162470 }, { "epoch": 0.41, "learning_rate": 5.938044189440838e-05, "loss": 1.0174, "step": 162475 }, { "epoch": 0.41, "learning_rate": 5.9379183652926676e-05, "loss": 1.0214, "step": 162480 }, { "epoch": 0.41, "learning_rate": 5.937792541144497e-05, "loss": 1.0189, "step": 162485 }, { "epoch": 0.41, "learning_rate": 5.937666716996326e-05, "loss": 1.0209, "step": 162490 }, { "epoch": 0.41, "learning_rate": 5.937540892848156e-05, "loss": 1.0211, "step": 162495 }, { "epoch": 0.41, "learning_rate": 5.937415068699985e-05, "loss": 1.0192, "step": 162500 }, { "epoch": 0.41, "learning_rate": 5.937289244551815e-05, "loss": 1.0212, "step": 162505 }, { "epoch": 0.41, "learning_rate": 5.937163420403644e-05, "loss": 1.0179, "step": 162510 }, { "epoch": 0.41, "learning_rate": 5.937037596255474e-05, "loss": 1.0406, "step": 162515 }, { "epoch": 0.41, "learning_rate": 5.936911772107303e-05, "loss": 1.0195, "step": 162520 }, { "epoch": 0.41, "learning_rate": 5.936785947959133e-05, "loss": 1.019, "step": 162525 }, { "epoch": 0.41, "learning_rate": 5.936660123810962e-05, "loss": 1.0196, "step": 162530 }, { "epoch": 0.41, "learning_rate": 5.936534299662792e-05, "loss": 1.0196, "step": 162535 }, { "epoch": 0.41, "learning_rate": 5.936408475514621e-05, "loss": 1.0179, "step": 162540 }, { "epoch": 0.41, "learning_rate": 5.936282651366451e-05, "loss": 1.018, "step": 162545 }, { "epoch": 0.41, "learning_rate": 5.93615682721828e-05, "loss": 1.0194, "step": 162550 }, { "epoch": 0.41, "learning_rate": 5.936031003070109e-05, "loss": 1.0174, "step": 162555 }, { "epoch": 0.41, "learning_rate": 5.935905178921939e-05, "loss": 1.0205, "step": 162560 }, { "epoch": 0.41, "learning_rate": 5.935779354773768e-05, "loss": 1.0203, "step": 162565 }, { "epoch": 0.41, "learning_rate": 5.935653530625598e-05, "loss": 1.0189, "step": 162570 }, { "epoch": 0.41, "learning_rate": 5.935527706477427e-05, "loss": 1.0194, "step": 162575 }, { "epoch": 0.41, "learning_rate": 5.935401882329257e-05, "loss": 1.0213, "step": 162580 }, { "epoch": 0.41, "learning_rate": 5.935276058181086e-05, "loss": 1.0164, "step": 162585 }, { "epoch": 0.41, "learning_rate": 5.935150234032916e-05, "loss": 1.0197, "step": 162590 }, { "epoch": 0.41, "learning_rate": 5.935024409884745e-05, "loss": 1.0187, "step": 162595 }, { "epoch": 0.41, "learning_rate": 5.934898585736575e-05, "loss": 1.0195, "step": 162600 }, { "epoch": 0.41, "learning_rate": 5.934772761588404e-05, "loss": 1.0221, "step": 162605 }, { "epoch": 0.41, "learning_rate": 5.934646937440234e-05, "loss": 1.0218, "step": 162610 }, { "epoch": 0.41, "learning_rate": 5.934521113292063e-05, "loss": 1.019, "step": 162615 }, { "epoch": 0.41, "learning_rate": 5.934395289143892e-05, "loss": 1.0187, "step": 162620 }, { "epoch": 0.41, "learning_rate": 5.934269464995722e-05, "loss": 1.0205, "step": 162625 }, { "epoch": 0.41, "learning_rate": 5.934143640847551e-05, "loss": 1.0214, "step": 162630 }, { "epoch": 0.41, "learning_rate": 5.934017816699381e-05, "loss": 1.0204, "step": 162635 }, { "epoch": 0.41, "learning_rate": 5.93389199255121e-05, "loss": 1.0196, "step": 162640 }, { "epoch": 0.41, "learning_rate": 5.93376616840304e-05, "loss": 1.0191, "step": 162645 }, { "epoch": 0.41, "learning_rate": 5.933640344254869e-05, "loss": 1.0216, "step": 162650 }, { "epoch": 0.41, "learning_rate": 5.933514520106699e-05, "loss": 1.0198, "step": 162655 }, { "epoch": 0.41, "learning_rate": 5.933388695958528e-05, "loss": 1.02, "step": 162660 }, { "epoch": 0.41, "learning_rate": 5.933262871810358e-05, "loss": 1.0189, "step": 162665 }, { "epoch": 0.41, "learning_rate": 5.933137047662187e-05, "loss": 1.0221, "step": 162670 }, { "epoch": 0.41, "learning_rate": 5.933011223514017e-05, "loss": 1.0192, "step": 162675 }, { "epoch": 0.41, "learning_rate": 5.932885399365846e-05, "loss": 1.0184, "step": 162680 }, { "epoch": 0.41, "learning_rate": 5.9327595752176754e-05, "loss": 1.0202, "step": 162685 }, { "epoch": 0.41, "learning_rate": 5.932633751069505e-05, "loss": 1.0212, "step": 162690 }, { "epoch": 0.41, "learning_rate": 5.9325079269213344e-05, "loss": 1.0187, "step": 162695 }, { "epoch": 0.41, "learning_rate": 5.932382102773164e-05, "loss": 1.0208, "step": 162700 }, { "epoch": 0.41, "learning_rate": 5.9322562786249934e-05, "loss": 1.0205, "step": 162705 }, { "epoch": 0.41, "learning_rate": 5.932130454476823e-05, "loss": 1.0198, "step": 162710 }, { "epoch": 0.41, "learning_rate": 5.9320046303286524e-05, "loss": 1.0198, "step": 162715 }, { "epoch": 0.41, "learning_rate": 5.931878806180482e-05, "loss": 1.0186, "step": 162720 }, { "epoch": 0.41, "learning_rate": 5.9317529820323114e-05, "loss": 1.0195, "step": 162725 }, { "epoch": 0.41, "learning_rate": 5.931627157884141e-05, "loss": 1.0182, "step": 162730 }, { "epoch": 0.41, "learning_rate": 5.9315013337359704e-05, "loss": 1.0183, "step": 162735 }, { "epoch": 0.41, "learning_rate": 5.9313755095878e-05, "loss": 1.0405, "step": 162740 }, { "epoch": 0.41, "learning_rate": 5.9312496854396294e-05, "loss": 1.0203, "step": 162745 }, { "epoch": 0.41, "learning_rate": 5.9311238612914585e-05, "loss": 1.0211, "step": 162750 }, { "epoch": 0.41, "learning_rate": 5.9309980371432884e-05, "loss": 1.0199, "step": 162755 }, { "epoch": 0.41, "learning_rate": 5.930872212995119e-05, "loss": 1.0185, "step": 162760 }, { "epoch": 0.41, "learning_rate": 5.930746388846948e-05, "loss": 1.0194, "step": 162765 }, { "epoch": 0.41, "learning_rate": 5.930620564698778e-05, "loss": 1.0214, "step": 162770 }, { "epoch": 0.41, "learning_rate": 5.930494740550607e-05, "loss": 1.0195, "step": 162775 }, { "epoch": 0.41, "learning_rate": 5.930368916402437e-05, "loss": 1.0198, "step": 162780 }, { "epoch": 0.41, "learning_rate": 5.930243092254266e-05, "loss": 1.018, "step": 162785 }, { "epoch": 0.41, "learning_rate": 5.930117268106096e-05, "loss": 1.0206, "step": 162790 }, { "epoch": 0.41, "learning_rate": 5.929991443957925e-05, "loss": 1.0179, "step": 162795 }, { "epoch": 0.41, "learning_rate": 5.929865619809755e-05, "loss": 1.0194, "step": 162800 }, { "epoch": 0.41, "learning_rate": 5.929739795661584e-05, "loss": 1.0208, "step": 162805 }, { "epoch": 0.41, "learning_rate": 5.929613971513414e-05, "loss": 1.019, "step": 162810 }, { "epoch": 0.41, "learning_rate": 5.929488147365243e-05, "loss": 1.0212, "step": 162815 }, { "epoch": 0.41, "learning_rate": 5.929362323217073e-05, "loss": 1.0192, "step": 162820 }, { "epoch": 0.41, "learning_rate": 5.929236499068902e-05, "loss": 1.0201, "step": 162825 }, { "epoch": 0.41, "learning_rate": 5.929110674920731e-05, "loss": 1.0213, "step": 162830 }, { "epoch": 0.41, "learning_rate": 5.928984850772561e-05, "loss": 1.0199, "step": 162835 }, { "epoch": 0.41, "learning_rate": 5.92885902662439e-05, "loss": 1.0191, "step": 162840 }, { "epoch": 0.41, "learning_rate": 5.92873320247622e-05, "loss": 1.0192, "step": 162845 }, { "epoch": 0.41, "learning_rate": 5.928607378328049e-05, "loss": 1.0201, "step": 162850 }, { "epoch": 0.41, "learning_rate": 5.928481554179879e-05, "loss": 1.0199, "step": 162855 }, { "epoch": 0.41, "learning_rate": 5.928380894861342e-05, "loss": 1.0193, "step": 162860 }, { "epoch": 0.41, "learning_rate": 5.928255070713171e-05, "loss": 1.0189, "step": 162865 }, { "epoch": 0.41, "learning_rate": 5.928129246565001e-05, "loss": 1.0362, "step": 162870 }, { "epoch": 0.41, "learning_rate": 5.92800342241683e-05, "loss": 1.021, "step": 162875 }, { "epoch": 0.41, "learning_rate": 5.92787759826866e-05, "loss": 1.0215, "step": 162880 }, { "epoch": 0.41, "learning_rate": 5.927751774120489e-05, "loss": 1.0186, "step": 162885 }, { "epoch": 0.41, "learning_rate": 5.927625949972318e-05, "loss": 1.0199, "step": 162890 }, { "epoch": 0.41, "learning_rate": 5.927500125824148e-05, "loss": 1.0192, "step": 162895 }, { "epoch": 0.41, "learning_rate": 5.927374301675977e-05, "loss": 1.037, "step": 162900 }, { "epoch": 0.41, "learning_rate": 5.927248477527807e-05, "loss": 1.0205, "step": 162905 }, { "epoch": 0.41, "learning_rate": 5.927122653379636e-05, "loss": 1.0186, "step": 162910 }, { "epoch": 0.41, "learning_rate": 5.926996829231466e-05, "loss": 1.0188, "step": 162915 }, { "epoch": 0.41, "learning_rate": 5.926871005083295e-05, "loss": 1.02, "step": 162920 }, { "epoch": 0.41, "learning_rate": 5.926745180935125e-05, "loss": 1.0185, "step": 162925 }, { "epoch": 0.41, "learning_rate": 5.926619356786954e-05, "loss": 1.0189, "step": 162930 }, { "epoch": 0.41, "learning_rate": 5.926493532638784e-05, "loss": 1.0178, "step": 162935 }, { "epoch": 0.41, "learning_rate": 5.9263928733202477e-05, "loss": 1.019, "step": 162940 }, { "epoch": 0.41, "learning_rate": 5.9262670491720775e-05, "loss": 1.0212, "step": 162945 }, { "epoch": 0.41, "learning_rate": 5.9261412250239067e-05, "loss": 1.021, "step": 162950 }, { "epoch": 0.41, "learning_rate": 5.926040565705371e-05, "loss": 1.0213, "step": 162955 }, { "epoch": 0.41, "learning_rate": 5.9259147415572e-05, "loss": 1.0199, "step": 162960 }, { "epoch": 0.41, "learning_rate": 5.92578891740903e-05, "loss": 1.0199, "step": 162965 }, { "epoch": 0.41, "learning_rate": 5.925663093260859e-05, "loss": 1.036, "step": 162970 }, { "epoch": 0.41, "learning_rate": 5.925537269112689e-05, "loss": 1.0177, "step": 162975 }, { "epoch": 0.41, "learning_rate": 5.925436609794152e-05, "loss": 1.0183, "step": 162980 }, { "epoch": 0.41, "learning_rate": 5.925310785645981e-05, "loss": 1.0196, "step": 162985 }, { "epoch": 0.41, "learning_rate": 5.925184961497811e-05, "loss": 1.0205, "step": 162990 }, { "epoch": 0.41, "learning_rate": 5.92505913734964e-05, "loss": 1.0199, "step": 162995 }, { "epoch": 0.41, "learning_rate": 5.924933313201469e-05, "loss": 1.0207, "step": 163000 }, { "epoch": 0.41, "learning_rate": 5.924807489053299e-05, "loss": 1.0189, "step": 163005 }, { "epoch": 0.41, "learning_rate": 5.924681664905128e-05, "loss": 1.0205, "step": 163010 }, { "epoch": 0.41, "learning_rate": 5.9245810055865924e-05, "loss": 1.0409, "step": 163015 }, { "epoch": 0.41, "learning_rate": 5.9244551814384216e-05, "loss": 1.0178, "step": 163020 }, { "epoch": 0.41, "learning_rate": 5.9243293572902514e-05, "loss": 1.0214, "step": 163025 }, { "epoch": 0.41, "learning_rate": 5.9242035331420806e-05, "loss": 1.0187, "step": 163030 }, { "epoch": 0.41, "learning_rate": 5.9240777089939104e-05, "loss": 1.0185, "step": 163035 }, { "epoch": 0.41, "learning_rate": 5.9239518848457396e-05, "loss": 1.0218, "step": 163040 }, { "epoch": 0.41, "learning_rate": 5.9238260606975694e-05, "loss": 1.0208, "step": 163045 }, { "epoch": 0.41, "learning_rate": 5.9237002365493986e-05, "loss": 1.0209, "step": 163050 }, { "epoch": 0.41, "learning_rate": 5.9235744124012284e-05, "loss": 1.0199, "step": 163055 }, { "epoch": 0.41, "learning_rate": 5.9234485882530576e-05, "loss": 1.0202, "step": 163060 }, { "epoch": 0.41, "learning_rate": 5.9233227641048874e-05, "loss": 1.0186, "step": 163065 }, { "epoch": 0.41, "learning_rate": 5.9231969399567166e-05, "loss": 1.0182, "step": 163070 }, { "epoch": 0.41, "learning_rate": 5.9230711158085464e-05, "loss": 1.0215, "step": 163075 }, { "epoch": 0.41, "learning_rate": 5.9229452916603756e-05, "loss": 1.0175, "step": 163080 }, { "epoch": 0.41, "learning_rate": 5.922819467512205e-05, "loss": 1.0197, "step": 163085 }, { "epoch": 0.41, "learning_rate": 5.9226936433640346e-05, "loss": 1.0179, "step": 163090 }, { "epoch": 0.41, "learning_rate": 5.922567819215864e-05, "loss": 1.021, "step": 163095 }, { "epoch": 0.41, "learning_rate": 5.9224419950676936e-05, "loss": 1.02, "step": 163100 }, { "epoch": 0.41, "learning_rate": 5.922316170919523e-05, "loss": 1.0182, "step": 163105 }, { "epoch": 0.41, "learning_rate": 5.9221903467713526e-05, "loss": 1.0191, "step": 163110 }, { "epoch": 0.41, "learning_rate": 5.922064522623182e-05, "loss": 1.0187, "step": 163115 }, { "epoch": 0.41, "learning_rate": 5.9219386984750116e-05, "loss": 1.022, "step": 163120 }, { "epoch": 0.41, "learning_rate": 5.921812874326841e-05, "loss": 1.0185, "step": 163125 }, { "epoch": 0.41, "learning_rate": 5.9216870501786706e-05, "loss": 1.0194, "step": 163130 }, { "epoch": 0.41, "learning_rate": 5.9215612260305e-05, "loss": 1.0191, "step": 163135 }, { "epoch": 0.41, "learning_rate": 5.9214354018823296e-05, "loss": 1.0213, "step": 163140 }, { "epoch": 0.41, "learning_rate": 5.921309577734159e-05, "loss": 1.0212, "step": 163145 }, { "epoch": 0.41, "learning_rate": 5.921183753585988e-05, "loss": 1.0196, "step": 163150 }, { "epoch": 0.41, "learning_rate": 5.921057929437818e-05, "loss": 1.0189, "step": 163155 }, { "epoch": 0.41, "learning_rate": 5.920932105289647e-05, "loss": 1.0188, "step": 163160 }, { "epoch": 0.41, "learning_rate": 5.920806281141477e-05, "loss": 1.0194, "step": 163165 }, { "epoch": 0.41, "learning_rate": 5.920680456993306e-05, "loss": 1.0194, "step": 163170 }, { "epoch": 0.41, "learning_rate": 5.920554632845136e-05, "loss": 1.0196, "step": 163175 }, { "epoch": 0.41, "learning_rate": 5.920428808696965e-05, "loss": 1.0219, "step": 163180 }, { "epoch": 0.41, "learning_rate": 5.920302984548795e-05, "loss": 1.0194, "step": 163185 }, { "epoch": 0.41, "learning_rate": 5.920177160400624e-05, "loss": 1.0213, "step": 163190 }, { "epoch": 0.41, "learning_rate": 5.920051336252454e-05, "loss": 1.021, "step": 163195 }, { "epoch": 0.41, "learning_rate": 5.919925512104283e-05, "loss": 1.0201, "step": 163200 }, { "epoch": 0.41, "learning_rate": 5.919799687956113e-05, "loss": 1.0205, "step": 163205 }, { "epoch": 0.41, "learning_rate": 5.919673863807942e-05, "loss": 1.0176, "step": 163210 }, { "epoch": 0.41, "learning_rate": 5.919548039659771e-05, "loss": 1.0208, "step": 163215 }, { "epoch": 0.41, "learning_rate": 5.919422215511601e-05, "loss": 1.0186, "step": 163220 }, { "epoch": 0.41, "learning_rate": 5.91929639136343e-05, "loss": 1.0222, "step": 163225 }, { "epoch": 0.41, "learning_rate": 5.91917056721526e-05, "loss": 1.0228, "step": 163230 }, { "epoch": 0.41, "learning_rate": 5.919044743067089e-05, "loss": 1.0204, "step": 163235 }, { "epoch": 0.41, "learning_rate": 5.918918918918919e-05, "loss": 1.0198, "step": 163240 }, { "epoch": 0.41, "learning_rate": 5.918793094770748e-05, "loss": 1.02, "step": 163245 }, { "epoch": 0.41, "learning_rate": 5.918667270622578e-05, "loss": 1.022, "step": 163250 }, { "epoch": 0.41, "learning_rate": 5.918541446474407e-05, "loss": 1.0196, "step": 163255 }, { "epoch": 0.41, "learning_rate": 5.918415622326237e-05, "loss": 1.0217, "step": 163260 }, { "epoch": 0.41, "learning_rate": 5.9182897981780674e-05, "loss": 1.0207, "step": 163265 }, { "epoch": 0.41, "learning_rate": 5.9181639740298965e-05, "loss": 1.0216, "step": 163270 }, { "epoch": 0.41, "learning_rate": 5.9180381498817264e-05, "loss": 1.021, "step": 163275 }, { "epoch": 0.41, "learning_rate": 5.9179123257335555e-05, "loss": 1.019, "step": 163280 }, { "epoch": 0.41, "learning_rate": 5.9177865015853854e-05, "loss": 1.0196, "step": 163285 }, { "epoch": 0.41, "learning_rate": 5.9176606774372145e-05, "loss": 1.0206, "step": 163290 }, { "epoch": 0.41, "learning_rate": 5.917534853289044e-05, "loss": 1.02, "step": 163295 }, { "epoch": 0.41, "learning_rate": 5.9174090291408735e-05, "loss": 1.0184, "step": 163300 }, { "epoch": 0.41, "learning_rate": 5.917283204992703e-05, "loss": 1.0216, "step": 163305 }, { "epoch": 0.41, "learning_rate": 5.9171573808445325e-05, "loss": 1.0205, "step": 163310 }, { "epoch": 0.41, "learning_rate": 5.917031556696362e-05, "loss": 1.0433, "step": 163315 }, { "epoch": 0.41, "learning_rate": 5.9169057325481915e-05, "loss": 1.0183, "step": 163320 }, { "epoch": 0.41, "learning_rate": 5.916779908400021e-05, "loss": 1.02, "step": 163325 }, { "epoch": 0.41, "learning_rate": 5.9166540842518505e-05, "loss": 1.0199, "step": 163330 }, { "epoch": 0.41, "learning_rate": 5.91652826010368e-05, "loss": 1.0178, "step": 163335 }, { "epoch": 0.41, "learning_rate": 5.9164024359555095e-05, "loss": 1.0214, "step": 163340 }, { "epoch": 0.41, "learning_rate": 5.916276611807339e-05, "loss": 1.0195, "step": 163345 }, { "epoch": 0.41, "learning_rate": 5.9161507876591685e-05, "loss": 1.02, "step": 163350 }, { "epoch": 0.41, "learning_rate": 5.916024963510998e-05, "loss": 1.0212, "step": 163355 }, { "epoch": 0.41, "learning_rate": 5.915899139362827e-05, "loss": 1.021, "step": 163360 }, { "epoch": 0.41, "learning_rate": 5.9157733152146567e-05, "loss": 1.0183, "step": 163365 }, { "epoch": 0.41, "learning_rate": 5.915647491066486e-05, "loss": 1.0193, "step": 163370 }, { "epoch": 0.41, "learning_rate": 5.9155216669183157e-05, "loss": 1.0423, "step": 163375 }, { "epoch": 0.41, "learning_rate": 5.915395842770145e-05, "loss": 1.0194, "step": 163380 }, { "epoch": 0.41, "learning_rate": 5.9152700186219747e-05, "loss": 1.0206, "step": 163385 }, { "epoch": 0.41, "learning_rate": 5.915144194473804e-05, "loss": 1.0219, "step": 163390 }, { "epoch": 0.41, "learning_rate": 5.9150183703256337e-05, "loss": 1.0203, "step": 163395 }, { "epoch": 0.41, "learning_rate": 5.914892546177463e-05, "loss": 1.0204, "step": 163400 }, { "epoch": 0.41, "learning_rate": 5.9147667220292926e-05, "loss": 1.0219, "step": 163405 }, { "epoch": 0.41, "learning_rate": 5.914640897881122e-05, "loss": 1.0167, "step": 163410 }, { "epoch": 0.41, "learning_rate": 5.9145150737329516e-05, "loss": 1.0215, "step": 163415 }, { "epoch": 0.41, "learning_rate": 5.914389249584781e-05, "loss": 1.0213, "step": 163420 }, { "epoch": 0.41, "learning_rate": 5.91426342543661e-05, "loss": 1.0201, "step": 163425 }, { "epoch": 0.41, "learning_rate": 5.91413760128844e-05, "loss": 1.0204, "step": 163430 }, { "epoch": 0.41, "learning_rate": 5.914011777140269e-05, "loss": 1.0163, "step": 163435 }, { "epoch": 0.41, "learning_rate": 5.913885952992099e-05, "loss": 1.0197, "step": 163440 }, { "epoch": 0.41, "learning_rate": 5.913760128843928e-05, "loss": 1.0207, "step": 163445 }, { "epoch": 0.41, "learning_rate": 5.913634304695758e-05, "loss": 1.02, "step": 163450 }, { "epoch": 0.41, "learning_rate": 5.913508480547587e-05, "loss": 1.0189, "step": 163455 }, { "epoch": 0.41, "learning_rate": 5.913382656399417e-05, "loss": 1.02, "step": 163460 }, { "epoch": 0.41, "learning_rate": 5.913256832251246e-05, "loss": 1.021, "step": 163465 }, { "epoch": 0.41, "learning_rate": 5.913131008103076e-05, "loss": 1.0181, "step": 163470 }, { "epoch": 0.41, "learning_rate": 5.913005183954905e-05, "loss": 1.0216, "step": 163475 }, { "epoch": 0.41, "learning_rate": 5.912879359806735e-05, "loss": 1.02, "step": 163480 }, { "epoch": 0.41, "learning_rate": 5.912753535658564e-05, "loss": 1.0193, "step": 163485 }, { "epoch": 0.41, "learning_rate": 5.912627711510393e-05, "loss": 1.0211, "step": 163490 }, { "epoch": 0.41, "learning_rate": 5.912501887362223e-05, "loss": 1.0186, "step": 163495 }, { "epoch": 0.41, "learning_rate": 5.912376063214052e-05, "loss": 1.0213, "step": 163500 }, { "epoch": 0.41, "learning_rate": 5.912250239065882e-05, "loss": 1.0191, "step": 163505 }, { "epoch": 0.41, "learning_rate": 5.912124414917711e-05, "loss": 1.0177, "step": 163510 }, { "epoch": 0.41, "learning_rate": 5.911998590769541e-05, "loss": 1.0189, "step": 163515 }, { "epoch": 0.41, "learning_rate": 5.91187276662137e-05, "loss": 1.0185, "step": 163520 }, { "epoch": 0.41, "learning_rate": 5.9117469424732e-05, "loss": 1.0211, "step": 163525 }, { "epoch": 0.41, "learning_rate": 5.911621118325029e-05, "loss": 1.0206, "step": 163530 }, { "epoch": 0.41, "learning_rate": 5.911495294176859e-05, "loss": 1.0205, "step": 163535 }, { "epoch": 0.41, "learning_rate": 5.911369470028688e-05, "loss": 1.016, "step": 163540 }, { "epoch": 0.41, "learning_rate": 5.911243645880518e-05, "loss": 1.0183, "step": 163545 }, { "epoch": 0.41, "learning_rate": 5.911117821732347e-05, "loss": 1.0201, "step": 163550 }, { "epoch": 0.41, "learning_rate": 5.910991997584176e-05, "loss": 1.0176, "step": 163555 }, { "epoch": 0.41, "learning_rate": 5.910866173436006e-05, "loss": 1.0205, "step": 163560 }, { "epoch": 0.41, "learning_rate": 5.910740349287835e-05, "loss": 1.0203, "step": 163565 }, { "epoch": 0.41, "learning_rate": 5.910614525139665e-05, "loss": 1.018, "step": 163570 }, { "epoch": 0.41, "learning_rate": 5.910488700991494e-05, "loss": 1.0196, "step": 163575 }, { "epoch": 0.41, "learning_rate": 5.910362876843324e-05, "loss": 1.0228, "step": 163580 }, { "epoch": 0.41, "learning_rate": 5.910237052695153e-05, "loss": 1.021, "step": 163585 }, { "epoch": 0.41, "learning_rate": 5.910111228546983e-05, "loss": 1.0182, "step": 163590 }, { "epoch": 0.41, "learning_rate": 5.909985404398812e-05, "loss": 1.0177, "step": 163595 }, { "epoch": 0.41, "learning_rate": 5.909859580250642e-05, "loss": 1.0386, "step": 163600 }, { "epoch": 0.41, "learning_rate": 5.909733756102471e-05, "loss": 1.0202, "step": 163605 }, { "epoch": 0.41, "learning_rate": 5.909607931954301e-05, "loss": 1.0212, "step": 163610 }, { "epoch": 0.41, "learning_rate": 5.90948210780613e-05, "loss": 1.0195, "step": 163615 }, { "epoch": 0.41, "learning_rate": 5.9093562836579594e-05, "loss": 1.0204, "step": 163620 }, { "epoch": 0.41, "learning_rate": 5.909230459509789e-05, "loss": 1.0184, "step": 163625 }, { "epoch": 0.41, "learning_rate": 5.9091046353616184e-05, "loss": 1.0205, "step": 163630 }, { "epoch": 0.41, "learning_rate": 5.908978811213448e-05, "loss": 1.0184, "step": 163635 }, { "epoch": 0.41, "learning_rate": 5.9088529870652774e-05, "loss": 1.0173, "step": 163640 }, { "epoch": 0.41, "learning_rate": 5.908727162917107e-05, "loss": 1.0198, "step": 163645 }, { "epoch": 0.41, "learning_rate": 5.9086013387689364e-05, "loss": 1.0184, "step": 163650 }, { "epoch": 0.41, "learning_rate": 5.908475514620766e-05, "loss": 1.0194, "step": 163655 }, { "epoch": 0.41, "learning_rate": 5.9083496904725954e-05, "loss": 1.0198, "step": 163660 }, { "epoch": 0.41, "learning_rate": 5.908223866324425e-05, "loss": 1.0204, "step": 163665 }, { "epoch": 0.41, "learning_rate": 5.9080980421762544e-05, "loss": 1.0224, "step": 163670 }, { "epoch": 0.41, "learning_rate": 5.9079722180280835e-05, "loss": 1.021, "step": 163675 }, { "epoch": 0.41, "learning_rate": 5.9078463938799134e-05, "loss": 1.0187, "step": 163680 }, { "epoch": 0.41, "learning_rate": 5.9077205697317425e-05, "loss": 1.0181, "step": 163685 }, { "epoch": 0.41, "learning_rate": 5.9075947455835724e-05, "loss": 1.0196, "step": 163690 }, { "epoch": 0.41, "learning_rate": 5.9074689214354015e-05, "loss": 1.0192, "step": 163695 }, { "epoch": 0.41, "learning_rate": 5.9073430972872314e-05, "loss": 1.022, "step": 163700 }, { "epoch": 0.41, "learning_rate": 5.9072172731390605e-05, "loss": 1.0197, "step": 163705 }, { "epoch": 0.41, "learning_rate": 5.9070914489908904e-05, "loss": 1.0205, "step": 163710 }, { "epoch": 0.41, "learning_rate": 5.9069656248427195e-05, "loss": 1.0168, "step": 163715 }, { "epoch": 0.41, "learning_rate": 5.9068398006945494e-05, "loss": 1.0196, "step": 163720 }, { "epoch": 0.41, "learning_rate": 5.9067139765463785e-05, "loss": 1.0201, "step": 163725 }, { "epoch": 0.41, "learning_rate": 5.9065881523982084e-05, "loss": 1.0189, "step": 163730 }, { "epoch": 0.41, "learning_rate": 5.9064623282500375e-05, "loss": 1.0213, "step": 163735 }, { "epoch": 0.41, "learning_rate": 5.906336504101867e-05, "loss": 1.0185, "step": 163740 }, { "epoch": 0.41, "learning_rate": 5.9062106799536965e-05, "loss": 1.0352, "step": 163745 }, { "epoch": 0.41, "learning_rate": 5.906084855805526e-05, "loss": 1.0197, "step": 163750 }, { "epoch": 0.41, "learning_rate": 5.9059590316573555e-05, "loss": 1.0181, "step": 163755 }, { "epoch": 0.41, "learning_rate": 5.905833207509185e-05, "loss": 1.0179, "step": 163760 }, { "epoch": 0.41, "learning_rate": 5.905707383361015e-05, "loss": 1.0184, "step": 163765 }, { "epoch": 0.41, "learning_rate": 5.905581559212845e-05, "loss": 1.021, "step": 163770 }, { "epoch": 0.41, "learning_rate": 5.905455735064674e-05, "loss": 1.0185, "step": 163775 }, { "epoch": 0.41, "learning_rate": 5.905329910916504e-05, "loss": 1.02, "step": 163780 }, { "epoch": 0.41, "learning_rate": 5.905204086768333e-05, "loss": 1.0205, "step": 163785 }, { "epoch": 0.41, "learning_rate": 5.905078262620163e-05, "loss": 1.0217, "step": 163790 }, { "epoch": 0.41, "learning_rate": 5.904952438471992e-05, "loss": 1.0184, "step": 163795 }, { "epoch": 0.41, "learning_rate": 5.904826614323822e-05, "loss": 1.0195, "step": 163800 }, { "epoch": 0.41, "learning_rate": 5.904700790175651e-05, "loss": 1.0203, "step": 163805 }, { "epoch": 0.41, "learning_rate": 5.904574966027481e-05, "loss": 1.0217, "step": 163810 }, { "epoch": 0.41, "learning_rate": 5.90444914187931e-05, "loss": 1.0192, "step": 163815 }, { "epoch": 0.41, "learning_rate": 5.904323317731139e-05, "loss": 1.02, "step": 163820 }, { "epoch": 0.41, "learning_rate": 5.904197493582969e-05, "loss": 1.018, "step": 163825 }, { "epoch": 0.41, "learning_rate": 5.904071669434798e-05, "loss": 1.0182, "step": 163830 }, { "epoch": 0.41, "learning_rate": 5.903945845286628e-05, "loss": 1.0213, "step": 163835 }, { "epoch": 0.41, "learning_rate": 5.903820021138457e-05, "loss": 1.0249, "step": 163840 }, { "epoch": 0.41, "learning_rate": 5.903694196990287e-05, "loss": 1.0226, "step": 163845 }, { "epoch": 0.41, "learning_rate": 5.903568372842116e-05, "loss": 1.0208, "step": 163850 }, { "epoch": 0.41, "learning_rate": 5.903442548693946e-05, "loss": 1.0215, "step": 163855 }, { "epoch": 0.41, "learning_rate": 5.903316724545775e-05, "loss": 1.0202, "step": 163860 }, { "epoch": 0.41, "learning_rate": 5.903190900397605e-05, "loss": 1.0189, "step": 163865 }, { "epoch": 0.41, "learning_rate": 5.903065076249434e-05, "loss": 1.0204, "step": 163870 }, { "epoch": 0.41, "learning_rate": 5.902939252101264e-05, "loss": 1.0217, "step": 163875 }, { "epoch": 0.41, "learning_rate": 5.902813427953093e-05, "loss": 1.0208, "step": 163880 }, { "epoch": 0.41, "learning_rate": 5.9026876038049225e-05, "loss": 1.0218, "step": 163885 }, { "epoch": 0.41, "learning_rate": 5.902561779656752e-05, "loss": 1.0216, "step": 163890 }, { "epoch": 0.41, "learning_rate": 5.9024359555085815e-05, "loss": 1.0206, "step": 163895 }, { "epoch": 0.41, "learning_rate": 5.902310131360411e-05, "loss": 1.0193, "step": 163900 }, { "epoch": 0.41, "learning_rate": 5.9021843072122405e-05, "loss": 1.0195, "step": 163905 }, { "epoch": 0.41, "learning_rate": 5.90205848306407e-05, "loss": 1.0191, "step": 163910 }, { "epoch": 0.41, "learning_rate": 5.9019326589158995e-05, "loss": 1.0207, "step": 163915 }, { "epoch": 0.41, "learning_rate": 5.901806834767729e-05, "loss": 1.0201, "step": 163920 }, { "epoch": 0.41, "learning_rate": 5.9016810106195585e-05, "loss": 1.019, "step": 163925 }, { "epoch": 0.41, "learning_rate": 5.901555186471388e-05, "loss": 1.0205, "step": 163930 }, { "epoch": 0.41, "learning_rate": 5.9014293623232175e-05, "loss": 1.0207, "step": 163935 }, { "epoch": 0.41, "learning_rate": 5.901303538175047e-05, "loss": 1.0174, "step": 163940 }, { "epoch": 0.41, "learning_rate": 5.9011777140268765e-05, "loss": 1.0198, "step": 163945 }, { "epoch": 0.41, "learning_rate": 5.9010518898787056e-05, "loss": 1.0198, "step": 163950 }, { "epoch": 0.41, "learning_rate": 5.9009260657305355e-05, "loss": 1.0202, "step": 163955 }, { "epoch": 0.41, "learning_rate": 5.9008002415823646e-05, "loss": 1.0185, "step": 163960 }, { "epoch": 0.41, "learning_rate": 5.9006744174341945e-05, "loss": 1.0202, "step": 163965 }, { "epoch": 0.41, "learning_rate": 5.9005485932860236e-05, "loss": 1.0177, "step": 163970 }, { "epoch": 0.41, "learning_rate": 5.9004227691378535e-05, "loss": 1.0201, "step": 163975 }, { "epoch": 0.41, "learning_rate": 5.9002969449896826e-05, "loss": 1.0194, "step": 163980 }, { "epoch": 0.41, "learning_rate": 5.9001711208415124e-05, "loss": 1.0209, "step": 163985 }, { "epoch": 0.41, "learning_rate": 5.9000452966933416e-05, "loss": 1.0217, "step": 163990 }, { "epoch": 0.41, "learning_rate": 5.8999194725451714e-05, "loss": 1.0191, "step": 163995 }, { "epoch": 0.41, "learning_rate": 5.8997936483970006e-05, "loss": 1.0172, "step": 164000 }, { "epoch": 0.41, "learning_rate": 5.8996678242488304e-05, "loss": 1.0195, "step": 164005 }, { "epoch": 0.41, "learning_rate": 5.8995420001006596e-05, "loss": 1.0201, "step": 164010 }, { "epoch": 0.41, "learning_rate": 5.899416175952489e-05, "loss": 1.0172, "step": 164015 }, { "epoch": 0.41, "learning_rate": 5.8992903518043186e-05, "loss": 1.0167, "step": 164020 }, { "epoch": 0.41, "learning_rate": 5.899164527656148e-05, "loss": 1.0202, "step": 164025 }, { "epoch": 0.41, "learning_rate": 5.8990387035079776e-05, "loss": 1.0228, "step": 164030 }, { "epoch": 0.41, "learning_rate": 5.898912879359807e-05, "loss": 1.0203, "step": 164035 }, { "epoch": 0.41, "learning_rate": 5.8987870552116366e-05, "loss": 1.0216, "step": 164040 }, { "epoch": 0.41, "learning_rate": 5.898661231063466e-05, "loss": 1.0229, "step": 164045 }, { "epoch": 0.41, "learning_rate": 5.8985354069152956e-05, "loss": 1.0182, "step": 164050 }, { "epoch": 0.41, "learning_rate": 5.898409582767125e-05, "loss": 1.0192, "step": 164055 }, { "epoch": 0.41, "learning_rate": 5.8982837586189546e-05, "loss": 1.0182, "step": 164060 }, { "epoch": 0.41, "learning_rate": 5.898157934470784e-05, "loss": 1.02, "step": 164065 }, { "epoch": 0.41, "learning_rate": 5.8980321103226136e-05, "loss": 1.0195, "step": 164070 }, { "epoch": 0.41, "learning_rate": 5.897906286174443e-05, "loss": 1.019, "step": 164075 }, { "epoch": 0.41, "learning_rate": 5.897780462026272e-05, "loss": 1.0202, "step": 164080 }, { "epoch": 0.41, "learning_rate": 5.897654637878102e-05, "loss": 1.0199, "step": 164085 }, { "epoch": 0.41, "learning_rate": 5.897528813729931e-05, "loss": 1.0192, "step": 164090 }, { "epoch": 0.41, "learning_rate": 5.897402989581761e-05, "loss": 1.0206, "step": 164095 }, { "epoch": 0.41, "learning_rate": 5.89727716543359e-05, "loss": 1.019, "step": 164100 }, { "epoch": 0.41, "learning_rate": 5.89715134128542e-05, "loss": 1.0166, "step": 164105 }, { "epoch": 0.41, "learning_rate": 5.897025517137249e-05, "loss": 1.0196, "step": 164110 }, { "epoch": 0.41, "learning_rate": 5.896899692989079e-05, "loss": 1.0345, "step": 164115 }, { "epoch": 0.41, "learning_rate": 5.896773868840908e-05, "loss": 1.0193, "step": 164120 }, { "epoch": 0.41, "learning_rate": 5.896648044692738e-05, "loss": 1.0217, "step": 164125 }, { "epoch": 0.41, "learning_rate": 5.896522220544567e-05, "loss": 1.0186, "step": 164130 }, { "epoch": 0.41, "learning_rate": 5.896396396396397e-05, "loss": 1.02, "step": 164135 }, { "epoch": 0.41, "learning_rate": 5.896270572248226e-05, "loss": 1.0217, "step": 164140 }, { "epoch": 0.41, "learning_rate": 5.896144748100055e-05, "loss": 1.0178, "step": 164145 }, { "epoch": 0.41, "learning_rate": 5.896018923951885e-05, "loss": 1.0212, "step": 164150 }, { "epoch": 0.41, "learning_rate": 5.895893099803714e-05, "loss": 1.0196, "step": 164155 }, { "epoch": 0.41, "learning_rate": 5.895767275655544e-05, "loss": 1.0208, "step": 164160 }, { "epoch": 0.41, "learning_rate": 5.895641451507373e-05, "loss": 1.0204, "step": 164165 }, { "epoch": 0.41, "learning_rate": 5.895515627359203e-05, "loss": 1.0201, "step": 164170 }, { "epoch": 0.41, "learning_rate": 5.895389803211032e-05, "loss": 1.0205, "step": 164175 }, { "epoch": 0.41, "learning_rate": 5.895263979062862e-05, "loss": 1.0197, "step": 164180 }, { "epoch": 0.41, "learning_rate": 5.895138154914691e-05, "loss": 1.0203, "step": 164185 }, { "epoch": 0.41, "learning_rate": 5.895012330766521e-05, "loss": 1.0208, "step": 164190 }, { "epoch": 0.41, "learning_rate": 5.89488650661835e-05, "loss": 1.0334, "step": 164195 }, { "epoch": 0.41, "learning_rate": 5.89476068247018e-05, "loss": 1.0193, "step": 164200 }, { "epoch": 0.41, "learning_rate": 5.894634858322009e-05, "loss": 1.0341, "step": 164205 }, { "epoch": 0.41, "learning_rate": 5.894509034173838e-05, "loss": 1.0226, "step": 164210 }, { "epoch": 0.41, "learning_rate": 5.894383210025668e-05, "loss": 1.0199, "step": 164215 }, { "epoch": 0.41, "learning_rate": 5.894257385877497e-05, "loss": 1.0191, "step": 164220 }, { "epoch": 0.41, "learning_rate": 5.894131561729327e-05, "loss": 1.0171, "step": 164225 }, { "epoch": 0.41, "learning_rate": 5.894005737581156e-05, "loss": 1.0215, "step": 164230 }, { "epoch": 0.41, "learning_rate": 5.893879913432986e-05, "loss": 1.0201, "step": 164235 }, { "epoch": 0.41, "learning_rate": 5.893754089284815e-05, "loss": 1.0185, "step": 164240 }, { "epoch": 0.41, "learning_rate": 5.893628265136645e-05, "loss": 1.0191, "step": 164245 }, { "epoch": 0.41, "learning_rate": 5.893502440988474e-05, "loss": 1.0197, "step": 164250 }, { "epoch": 0.41, "learning_rate": 5.893376616840304e-05, "loss": 1.0176, "step": 164255 }, { "epoch": 0.41, "learning_rate": 5.893250792692133e-05, "loss": 1.0188, "step": 164260 }, { "epoch": 0.41, "learning_rate": 5.893124968543964e-05, "loss": 1.0207, "step": 164265 }, { "epoch": 0.41, "learning_rate": 5.8929991443957935e-05, "loss": 1.0214, "step": 164270 }, { "epoch": 0.41, "learning_rate": 5.892873320247623e-05, "loss": 1.0219, "step": 164275 }, { "epoch": 0.41, "learning_rate": 5.8927474960994525e-05, "loss": 1.0214, "step": 164280 }, { "epoch": 0.41, "learning_rate": 5.892621671951282e-05, "loss": 1.0205, "step": 164285 }, { "epoch": 0.41, "learning_rate": 5.892495847803111e-05, "loss": 1.0203, "step": 164290 }, { "epoch": 0.41, "learning_rate": 5.892370023654941e-05, "loss": 1.0421, "step": 164295 }, { "epoch": 0.41, "learning_rate": 5.89224419950677e-05, "loss": 1.0193, "step": 164300 }, { "epoch": 0.41, "learning_rate": 5.8921183753586e-05, "loss": 1.0216, "step": 164305 }, { "epoch": 0.41, "learning_rate": 5.891992551210429e-05, "loss": 1.0187, "step": 164310 }, { "epoch": 0.41, "learning_rate": 5.891866727062259e-05, "loss": 1.0203, "step": 164315 }, { "epoch": 0.41, "learning_rate": 5.891740902914088e-05, "loss": 1.0192, "step": 164320 }, { "epoch": 0.41, "learning_rate": 5.891615078765918e-05, "loss": 1.0191, "step": 164325 }, { "epoch": 0.41, "learning_rate": 5.891489254617747e-05, "loss": 1.0213, "step": 164330 }, { "epoch": 0.41, "learning_rate": 5.891363430469577e-05, "loss": 1.0437, "step": 164335 }, { "epoch": 0.41, "learning_rate": 5.891237606321406e-05, "loss": 1.0405, "step": 164340 }, { "epoch": 0.41, "learning_rate": 5.891111782173236e-05, "loss": 1.022, "step": 164345 }, { "epoch": 0.41, "learning_rate": 5.890985958025065e-05, "loss": 1.0189, "step": 164350 }, { "epoch": 0.41, "learning_rate": 5.890860133876894e-05, "loss": 1.0217, "step": 164355 }, { "epoch": 0.41, "learning_rate": 5.890734309728724e-05, "loss": 1.0212, "step": 164360 }, { "epoch": 0.41, "learning_rate": 5.890608485580553e-05, "loss": 1.0168, "step": 164365 }, { "epoch": 0.41, "learning_rate": 5.890482661432383e-05, "loss": 1.0183, "step": 164370 }, { "epoch": 0.41, "learning_rate": 5.890356837284212e-05, "loss": 1.0195, "step": 164375 }, { "epoch": 0.41, "learning_rate": 5.890231013136042e-05, "loss": 1.0206, "step": 164380 }, { "epoch": 0.41, "learning_rate": 5.890105188987871e-05, "loss": 1.0207, "step": 164385 }, { "epoch": 0.41, "learning_rate": 5.889979364839701e-05, "loss": 1.0182, "step": 164390 }, { "epoch": 0.41, "learning_rate": 5.88985354069153e-05, "loss": 1.0219, "step": 164395 }, { "epoch": 0.41, "learning_rate": 5.88972771654336e-05, "loss": 1.0205, "step": 164400 }, { "epoch": 0.41, "learning_rate": 5.889601892395189e-05, "loss": 1.0198, "step": 164405 }, { "epoch": 0.41, "learning_rate": 5.889476068247019e-05, "loss": 1.0195, "step": 164410 }, { "epoch": 0.41, "learning_rate": 5.889350244098848e-05, "loss": 1.0381, "step": 164415 }, { "epoch": 0.41, "learning_rate": 5.889224419950677e-05, "loss": 1.0206, "step": 164420 }, { "epoch": 0.41, "learning_rate": 5.889098595802507e-05, "loss": 1.0188, "step": 164425 }, { "epoch": 0.41, "learning_rate": 5.888972771654336e-05, "loss": 1.0197, "step": 164430 }, { "epoch": 0.41, "learning_rate": 5.888846947506166e-05, "loss": 1.021, "step": 164435 }, { "epoch": 0.41, "learning_rate": 5.888721123357995e-05, "loss": 1.0203, "step": 164440 }, { "epoch": 0.41, "learning_rate": 5.888595299209825e-05, "loss": 1.0222, "step": 164445 }, { "epoch": 0.41, "learning_rate": 5.888469475061654e-05, "loss": 1.0405, "step": 164450 }, { "epoch": 0.41, "learning_rate": 5.888343650913484e-05, "loss": 1.022, "step": 164455 }, { "epoch": 0.41, "learning_rate": 5.888217826765313e-05, "loss": 1.02, "step": 164460 }, { "epoch": 0.41, "learning_rate": 5.888092002617143e-05, "loss": 1.0235, "step": 164465 }, { "epoch": 0.41, "learning_rate": 5.887966178468972e-05, "loss": 1.0181, "step": 164470 }, { "epoch": 0.41, "learning_rate": 5.887840354320801e-05, "loss": 1.0229, "step": 164475 }, { "epoch": 0.41, "learning_rate": 5.887714530172631e-05, "loss": 1.021, "step": 164480 }, { "epoch": 0.41, "learning_rate": 5.88758870602446e-05, "loss": 1.0199, "step": 164485 }, { "epoch": 0.41, "learning_rate": 5.88746288187629e-05, "loss": 1.0186, "step": 164490 }, { "epoch": 0.41, "learning_rate": 5.887337057728119e-05, "loss": 1.0216, "step": 164495 }, { "epoch": 0.41, "learning_rate": 5.887211233579949e-05, "loss": 1.0189, "step": 164500 }, { "epoch": 0.41, "learning_rate": 5.887085409431778e-05, "loss": 1.0187, "step": 164505 }, { "epoch": 0.41, "learning_rate": 5.886959585283608e-05, "loss": 1.0197, "step": 164510 }, { "epoch": 0.41, "learning_rate": 5.886833761135437e-05, "loss": 1.0191, "step": 164515 }, { "epoch": 0.41, "learning_rate": 5.886707936987267e-05, "loss": 1.0202, "step": 164520 }, { "epoch": 0.41, "learning_rate": 5.886582112839096e-05, "loss": 1.0203, "step": 164525 }, { "epoch": 0.41, "learning_rate": 5.886456288690926e-05, "loss": 1.0197, "step": 164530 }, { "epoch": 0.41, "learning_rate": 5.886330464542755e-05, "loss": 1.0171, "step": 164535 }, { "epoch": 0.41, "learning_rate": 5.8862046403945844e-05, "loss": 1.0189, "step": 164540 }, { "epoch": 0.41, "learning_rate": 5.886078816246414e-05, "loss": 1.0218, "step": 164545 }, { "epoch": 0.41, "learning_rate": 5.8859529920982434e-05, "loss": 1.0202, "step": 164550 }, { "epoch": 0.41, "learning_rate": 5.885827167950073e-05, "loss": 1.0193, "step": 164555 }, { "epoch": 0.41, "learning_rate": 5.8857013438019024e-05, "loss": 1.0166, "step": 164560 }, { "epoch": 0.41, "learning_rate": 5.885575519653732e-05, "loss": 1.0442, "step": 164565 }, { "epoch": 0.41, "learning_rate": 5.8854496955055614e-05, "loss": 1.0191, "step": 164570 }, { "epoch": 0.41, "learning_rate": 5.885323871357391e-05, "loss": 1.0206, "step": 164575 }, { "epoch": 0.41, "learning_rate": 5.8851980472092204e-05, "loss": 1.02, "step": 164580 }, { "epoch": 0.41, "learning_rate": 5.88507222306105e-05, "loss": 1.0192, "step": 164585 }, { "epoch": 0.41, "learning_rate": 5.8849463989128794e-05, "loss": 1.0197, "step": 164590 }, { "epoch": 0.41, "learning_rate": 5.884820574764709e-05, "loss": 1.0179, "step": 164595 }, { "epoch": 0.41, "learning_rate": 5.8846947506165384e-05, "loss": 1.0198, "step": 164600 }, { "epoch": 0.41, "learning_rate": 5.8845689264683676e-05, "loss": 1.0189, "step": 164605 }, { "epoch": 0.41, "learning_rate": 5.8844431023201974e-05, "loss": 1.0219, "step": 164610 }, { "epoch": 0.41, "learning_rate": 5.8843172781720266e-05, "loss": 1.0187, "step": 164615 }, { "epoch": 0.41, "learning_rate": 5.8841914540238564e-05, "loss": 1.0186, "step": 164620 }, { "epoch": 0.41, "learning_rate": 5.8840656298756856e-05, "loss": 1.02, "step": 164625 }, { "epoch": 0.41, "learning_rate": 5.8839398057275154e-05, "loss": 1.0209, "step": 164630 }, { "epoch": 0.41, "learning_rate": 5.8838139815793446e-05, "loss": 1.0192, "step": 164635 }, { "epoch": 0.41, "learning_rate": 5.8836881574311744e-05, "loss": 1.0197, "step": 164640 }, { "epoch": 0.41, "learning_rate": 5.8835623332830035e-05, "loss": 1.0201, "step": 164645 }, { "epoch": 0.41, "learning_rate": 5.8834365091348334e-05, "loss": 1.0436, "step": 164650 }, { "epoch": 0.41, "learning_rate": 5.8833106849866625e-05, "loss": 1.0214, "step": 164655 }, { "epoch": 0.41, "learning_rate": 5.8831848608384924e-05, "loss": 1.0215, "step": 164660 }, { "epoch": 0.41, "learning_rate": 5.8830590366903215e-05, "loss": 1.0195, "step": 164665 }, { "epoch": 0.41, "learning_rate": 5.882933212542151e-05, "loss": 1.0217, "step": 164670 }, { "epoch": 0.41, "learning_rate": 5.8828073883939805e-05, "loss": 1.0366, "step": 164675 }, { "epoch": 0.41, "learning_rate": 5.88268156424581e-05, "loss": 1.0204, "step": 164680 }, { "epoch": 0.41, "learning_rate": 5.8825557400976395e-05, "loss": 1.0203, "step": 164685 }, { "epoch": 0.41, "learning_rate": 5.882429915949469e-05, "loss": 1.0188, "step": 164690 }, { "epoch": 0.41, "learning_rate": 5.8823040918012985e-05, "loss": 1.0194, "step": 164695 }, { "epoch": 0.41, "learning_rate": 5.882178267653128e-05, "loss": 1.0199, "step": 164700 }, { "epoch": 0.41, "learning_rate": 5.8820524435049575e-05, "loss": 1.0186, "step": 164705 }, { "epoch": 0.41, "learning_rate": 5.881926619356787e-05, "loss": 1.0205, "step": 164710 }, { "epoch": 0.41, "learning_rate": 5.8818007952086165e-05, "loss": 1.0394, "step": 164715 }, { "epoch": 0.41, "learning_rate": 5.881674971060446e-05, "loss": 1.0199, "step": 164720 }, { "epoch": 0.41, "learning_rate": 5.8815491469122755e-05, "loss": 1.0192, "step": 164725 }, { "epoch": 0.41, "learning_rate": 5.881423322764105e-05, "loss": 1.022, "step": 164730 }, { "epoch": 0.41, "learning_rate": 5.881297498615934e-05, "loss": 1.0217, "step": 164735 }, { "epoch": 0.41, "learning_rate": 5.881171674467764e-05, "loss": 1.0196, "step": 164740 }, { "epoch": 0.41, "learning_rate": 5.881045850319593e-05, "loss": 1.0184, "step": 164745 }, { "epoch": 0.41, "learning_rate": 5.880920026171423e-05, "loss": 1.0202, "step": 164750 }, { "epoch": 0.41, "learning_rate": 5.880794202023252e-05, "loss": 1.0184, "step": 164755 }, { "epoch": 0.41, "learning_rate": 5.880668377875082e-05, "loss": 1.0183, "step": 164760 }, { "epoch": 0.41, "learning_rate": 5.880542553726912e-05, "loss": 1.0182, "step": 164765 }, { "epoch": 0.41, "learning_rate": 5.8804167295787414e-05, "loss": 1.02, "step": 164770 }, { "epoch": 0.41, "learning_rate": 5.880290905430571e-05, "loss": 1.0202, "step": 164775 }, { "epoch": 0.41, "learning_rate": 5.8801650812824003e-05, "loss": 1.0353, "step": 164780 }, { "epoch": 0.41, "learning_rate": 5.88003925713423e-05, "loss": 1.0202, "step": 164785 }, { "epoch": 0.41, "learning_rate": 5.8799134329860593e-05, "loss": 1.0202, "step": 164790 }, { "epoch": 0.41, "learning_rate": 5.879787608837889e-05, "loss": 1.0203, "step": 164795 }, { "epoch": 0.41, "learning_rate": 5.8796617846897183e-05, "loss": 1.0202, "step": 164800 }, { "epoch": 0.41, "learning_rate": 5.879535960541548e-05, "loss": 1.0199, "step": 164805 }, { "epoch": 0.41, "learning_rate": 5.8794101363933773e-05, "loss": 1.0178, "step": 164810 }, { "epoch": 0.41, "learning_rate": 5.8792843122452065e-05, "loss": 1.0185, "step": 164815 }, { "epoch": 0.41, "learning_rate": 5.879158488097036e-05, "loss": 1.0199, "step": 164820 }, { "epoch": 0.41, "learning_rate": 5.8790326639488655e-05, "loss": 1.0206, "step": 164825 }, { "epoch": 0.41, "learning_rate": 5.878906839800695e-05, "loss": 1.0223, "step": 164830 }, { "epoch": 0.41, "learning_rate": 5.8787810156525245e-05, "loss": 1.0201, "step": 164835 }, { "epoch": 0.41, "learning_rate": 5.878655191504354e-05, "loss": 1.0215, "step": 164840 }, { "epoch": 0.41, "learning_rate": 5.8785293673561835e-05, "loss": 1.0182, "step": 164845 }, { "epoch": 0.41, "learning_rate": 5.878403543208013e-05, "loss": 1.0215, "step": 164850 }, { "epoch": 0.41, "learning_rate": 5.8782777190598425e-05, "loss": 1.0198, "step": 164855 }, { "epoch": 0.41, "learning_rate": 5.878151894911672e-05, "loss": 1.0191, "step": 164860 }, { "epoch": 0.41, "learning_rate": 5.8780260707635015e-05, "loss": 1.0201, "step": 164865 }, { "epoch": 0.41, "learning_rate": 5.877900246615331e-05, "loss": 1.0206, "step": 164870 }, { "epoch": 0.41, "learning_rate": 5.8777744224671605e-05, "loss": 1.017, "step": 164875 }, { "epoch": 0.41, "learning_rate": 5.8776485983189896e-05, "loss": 1.0183, "step": 164880 }, { "epoch": 0.41, "learning_rate": 5.8775227741708195e-05, "loss": 1.0213, "step": 164885 }, { "epoch": 0.41, "learning_rate": 5.8773969500226486e-05, "loss": 1.0202, "step": 164890 }, { "epoch": 0.41, "learning_rate": 5.8772711258744785e-05, "loss": 1.0409, "step": 164895 }, { "epoch": 0.41, "learning_rate": 5.8771453017263076e-05, "loss": 1.0195, "step": 164900 }, { "epoch": 0.41, "learning_rate": 5.8770194775781375e-05, "loss": 1.0187, "step": 164905 }, { "epoch": 0.41, "learning_rate": 5.8768936534299666e-05, "loss": 1.0201, "step": 164910 }, { "epoch": 0.41, "learning_rate": 5.8767678292817965e-05, "loss": 1.0206, "step": 164915 }, { "epoch": 0.41, "learning_rate": 5.8766420051336256e-05, "loss": 1.02, "step": 164920 }, { "epoch": 0.41, "learning_rate": 5.8765161809854555e-05, "loss": 1.0185, "step": 164925 }, { "epoch": 0.41, "learning_rate": 5.8763903568372846e-05, "loss": 1.0206, "step": 164930 }, { "epoch": 0.41, "learning_rate": 5.8762645326891145e-05, "loss": 1.0199, "step": 164935 }, { "epoch": 0.41, "learning_rate": 5.8761387085409436e-05, "loss": 1.0184, "step": 164940 }, { "epoch": 0.41, "learning_rate": 5.876012884392773e-05, "loss": 1.019, "step": 164945 }, { "epoch": 0.41, "learning_rate": 5.8758870602446026e-05, "loss": 1.019, "step": 164950 }, { "epoch": 0.41, "learning_rate": 5.875761236096432e-05, "loss": 1.0196, "step": 164955 }, { "epoch": 0.41, "learning_rate": 5.8756354119482616e-05, "loss": 1.0187, "step": 164960 }, { "epoch": 0.41, "learning_rate": 5.875509587800091e-05, "loss": 1.0198, "step": 164965 }, { "epoch": 0.41, "learning_rate": 5.8753837636519206e-05, "loss": 1.021, "step": 164970 }, { "epoch": 0.41, "learning_rate": 5.87525793950375e-05, "loss": 1.0198, "step": 164975 }, { "epoch": 0.41, "learning_rate": 5.8751321153555796e-05, "loss": 1.0211, "step": 164980 }, { "epoch": 0.41, "learning_rate": 5.875006291207409e-05, "loss": 1.0176, "step": 164985 }, { "epoch": 0.41, "learning_rate": 5.8748804670592386e-05, "loss": 1.0184, "step": 164990 }, { "epoch": 0.41, "learning_rate": 5.874754642911068e-05, "loss": 1.0194, "step": 164995 }, { "epoch": 0.41, "learning_rate": 5.8746288187628976e-05, "loss": 1.0203, "step": 165000 }, { "epoch": 0.41, "learning_rate": 5.874502994614727e-05, "loss": 1.0197, "step": 165005 }, { "epoch": 0.41, "learning_rate": 5.874377170466556e-05, "loss": 1.0191, "step": 165010 }, { "epoch": 0.41, "learning_rate": 5.874251346318386e-05, "loss": 1.0191, "step": 165015 }, { "epoch": 0.41, "learning_rate": 5.874125522170215e-05, "loss": 1.0196, "step": 165020 }, { "epoch": 0.41, "learning_rate": 5.873999698022045e-05, "loss": 1.0188, "step": 165025 }, { "epoch": 0.41, "learning_rate": 5.873873873873874e-05, "loss": 1.0193, "step": 165030 }, { "epoch": 0.41, "learning_rate": 5.873748049725704e-05, "loss": 1.0192, "step": 165035 }, { "epoch": 0.41, "learning_rate": 5.873622225577533e-05, "loss": 1.0196, "step": 165040 }, { "epoch": 0.41, "learning_rate": 5.873496401429363e-05, "loss": 1.0218, "step": 165045 }, { "epoch": 0.41, "learning_rate": 5.873370577281192e-05, "loss": 1.0202, "step": 165050 }, { "epoch": 0.41, "learning_rate": 5.873244753133022e-05, "loss": 1.019, "step": 165055 }, { "epoch": 0.41, "learning_rate": 5.873118928984851e-05, "loss": 1.0181, "step": 165060 }, { "epoch": 0.41, "learning_rate": 5.872993104836681e-05, "loss": 1.0195, "step": 165065 }, { "epoch": 0.41, "learning_rate": 5.87286728068851e-05, "loss": 1.02, "step": 165070 }, { "epoch": 0.41, "learning_rate": 5.872741456540339e-05, "loss": 1.0205, "step": 165075 }, { "epoch": 0.41, "learning_rate": 5.872615632392169e-05, "loss": 1.0206, "step": 165080 }, { "epoch": 0.41, "learning_rate": 5.872489808243998e-05, "loss": 1.0193, "step": 165085 }, { "epoch": 0.41, "learning_rate": 5.872363984095828e-05, "loss": 1.0193, "step": 165090 }, { "epoch": 0.41, "learning_rate": 5.872238159947657e-05, "loss": 1.0206, "step": 165095 }, { "epoch": 0.41, "learning_rate": 5.872112335799487e-05, "loss": 1.0212, "step": 165100 }, { "epoch": 0.41, "learning_rate": 5.871986511651316e-05, "loss": 1.0205, "step": 165105 }, { "epoch": 0.41, "learning_rate": 5.871860687503146e-05, "loss": 1.0197, "step": 165110 }, { "epoch": 0.41, "learning_rate": 5.871734863354975e-05, "loss": 1.0199, "step": 165115 }, { "epoch": 0.41, "learning_rate": 5.871609039206805e-05, "loss": 1.0184, "step": 165120 }, { "epoch": 0.41, "learning_rate": 5.871483215058634e-05, "loss": 1.0202, "step": 165125 }, { "epoch": 0.41, "learning_rate": 5.871357390910463e-05, "loss": 1.0178, "step": 165130 }, { "epoch": 0.41, "learning_rate": 5.871231566762293e-05, "loss": 1.0181, "step": 165135 }, { "epoch": 0.41, "learning_rate": 5.871105742614122e-05, "loss": 1.0198, "step": 165140 }, { "epoch": 0.41, "learning_rate": 5.870979918465952e-05, "loss": 1.0194, "step": 165145 }, { "epoch": 0.41, "learning_rate": 5.870854094317781e-05, "loss": 1.0183, "step": 165150 }, { "epoch": 0.41, "learning_rate": 5.870728270169611e-05, "loss": 1.0188, "step": 165155 }, { "epoch": 0.41, "learning_rate": 5.87060244602144e-05, "loss": 1.0201, "step": 165160 }, { "epoch": 0.41, "learning_rate": 5.87047662187327e-05, "loss": 1.0211, "step": 165165 }, { "epoch": 0.41, "learning_rate": 5.870350797725099e-05, "loss": 1.0214, "step": 165170 }, { "epoch": 0.41, "learning_rate": 5.870224973576929e-05, "loss": 1.0202, "step": 165175 }, { "epoch": 0.41, "learning_rate": 5.870099149428758e-05, "loss": 1.0192, "step": 165180 }, { "epoch": 0.41, "learning_rate": 5.869973325280588e-05, "loss": 1.0182, "step": 165185 }, { "epoch": 0.41, "learning_rate": 5.869847501132417e-05, "loss": 1.0189, "step": 165190 }, { "epoch": 0.41, "learning_rate": 5.8697216769842464e-05, "loss": 1.02, "step": 165195 }, { "epoch": 0.41, "learning_rate": 5.869595852836076e-05, "loss": 1.0213, "step": 165200 }, { "epoch": 0.41, "learning_rate": 5.8694700286879054e-05, "loss": 1.0207, "step": 165205 }, { "epoch": 0.41, "learning_rate": 5.869344204539735e-05, "loss": 1.0203, "step": 165210 }, { "epoch": 0.41, "learning_rate": 5.8692183803915644e-05, "loss": 1.0192, "step": 165215 }, { "epoch": 0.41, "learning_rate": 5.869092556243394e-05, "loss": 1.0209, "step": 165220 }, { "epoch": 0.41, "learning_rate": 5.8689667320952233e-05, "loss": 1.0189, "step": 165225 }, { "epoch": 0.41, "learning_rate": 5.868840907947053e-05, "loss": 1.0181, "step": 165230 }, { "epoch": 0.41, "learning_rate": 5.8687150837988823e-05, "loss": 1.0187, "step": 165235 }, { "epoch": 0.41, "learning_rate": 5.868589259650712e-05, "loss": 1.0213, "step": 165240 }, { "epoch": 0.41, "learning_rate": 5.8684634355025413e-05, "loss": 1.0218, "step": 165245 }, { "epoch": 0.41, "learning_rate": 5.868337611354371e-05, "loss": 1.0207, "step": 165250 }, { "epoch": 0.41, "learning_rate": 5.8682117872062003e-05, "loss": 1.0174, "step": 165255 }, { "epoch": 0.41, "learning_rate": 5.868085963058031e-05, "loss": 1.019, "step": 165260 }, { "epoch": 0.41, "learning_rate": 5.867960138909861e-05, "loss": 1.0185, "step": 165265 }, { "epoch": 0.41, "learning_rate": 5.86783431476169e-05, "loss": 1.0176, "step": 165270 }, { "epoch": 0.41, "learning_rate": 5.867708490613519e-05, "loss": 1.0195, "step": 165275 }, { "epoch": 0.41, "learning_rate": 5.867582666465349e-05, "loss": 1.0195, "step": 165280 }, { "epoch": 0.41, "learning_rate": 5.867456842317178e-05, "loss": 1.0187, "step": 165285 }, { "epoch": 0.41, "learning_rate": 5.867331018169008e-05, "loss": 1.019, "step": 165290 }, { "epoch": 0.41, "learning_rate": 5.867205194020837e-05, "loss": 1.0187, "step": 165295 }, { "epoch": 0.41, "learning_rate": 5.867079369872667e-05, "loss": 1.0211, "step": 165300 }, { "epoch": 0.41, "learning_rate": 5.866953545724496e-05, "loss": 1.0204, "step": 165305 }, { "epoch": 0.41, "learning_rate": 5.866827721576326e-05, "loss": 1.019, "step": 165310 }, { "epoch": 0.41, "learning_rate": 5.866701897428155e-05, "loss": 1.0212, "step": 165315 }, { "epoch": 0.41, "learning_rate": 5.866576073279985e-05, "loss": 1.0183, "step": 165320 }, { "epoch": 0.41, "learning_rate": 5.866450249131814e-05, "loss": 1.0205, "step": 165325 }, { "epoch": 0.42, "learning_rate": 5.866324424983644e-05, "loss": 1.0197, "step": 165330 }, { "epoch": 0.42, "learning_rate": 5.866198600835473e-05, "loss": 1.0216, "step": 165335 }, { "epoch": 0.42, "learning_rate": 5.866072776687302e-05, "loss": 1.0174, "step": 165340 }, { "epoch": 0.42, "learning_rate": 5.865946952539132e-05, "loss": 1.0181, "step": 165345 }, { "epoch": 0.42, "learning_rate": 5.865821128390961e-05, "loss": 1.0189, "step": 165350 }, { "epoch": 0.42, "learning_rate": 5.865695304242791e-05, "loss": 1.018, "step": 165355 }, { "epoch": 0.42, "learning_rate": 5.86556948009462e-05, "loss": 1.0193, "step": 165360 }, { "epoch": 0.42, "learning_rate": 5.86544365594645e-05, "loss": 1.0189, "step": 165365 }, { "epoch": 0.42, "learning_rate": 5.865317831798279e-05, "loss": 1.021, "step": 165370 }, { "epoch": 0.42, "learning_rate": 5.865192007650109e-05, "loss": 1.021, "step": 165375 }, { "epoch": 0.42, "learning_rate": 5.865066183501938e-05, "loss": 1.0173, "step": 165380 }, { "epoch": 0.42, "learning_rate": 5.864940359353768e-05, "loss": 1.0215, "step": 165385 }, { "epoch": 0.42, "learning_rate": 5.864814535205597e-05, "loss": 1.0195, "step": 165390 }, { "epoch": 0.42, "learning_rate": 5.864688711057427e-05, "loss": 1.0191, "step": 165395 }, { "epoch": 0.42, "learning_rate": 5.864562886909256e-05, "loss": 1.0192, "step": 165400 }, { "epoch": 0.42, "learning_rate": 5.864437062761085e-05, "loss": 1.0222, "step": 165405 }, { "epoch": 0.42, "learning_rate": 5.864311238612915e-05, "loss": 1.0206, "step": 165410 }, { "epoch": 0.42, "learning_rate": 5.864185414464744e-05, "loss": 1.0224, "step": 165415 }, { "epoch": 0.42, "learning_rate": 5.864059590316574e-05, "loss": 1.0189, "step": 165420 }, { "epoch": 0.42, "learning_rate": 5.863933766168403e-05, "loss": 1.0178, "step": 165425 }, { "epoch": 0.42, "learning_rate": 5.863807942020233e-05, "loss": 1.0176, "step": 165430 }, { "epoch": 0.42, "learning_rate": 5.863682117872062e-05, "loss": 1.0202, "step": 165435 }, { "epoch": 0.42, "learning_rate": 5.863556293723892e-05, "loss": 1.0262, "step": 165440 }, { "epoch": 0.42, "learning_rate": 5.863430469575721e-05, "loss": 1.0193, "step": 165445 }, { "epoch": 0.42, "learning_rate": 5.863304645427551e-05, "loss": 1.0199, "step": 165450 }, { "epoch": 0.42, "learning_rate": 5.86317882127938e-05, "loss": 1.0205, "step": 165455 }, { "epoch": 0.42, "learning_rate": 5.86305299713121e-05, "loss": 1.0205, "step": 165460 }, { "epoch": 0.42, "learning_rate": 5.862927172983039e-05, "loss": 1.0191, "step": 165465 }, { "epoch": 0.42, "learning_rate": 5.8628013488348684e-05, "loss": 1.0219, "step": 165470 }, { "epoch": 0.42, "learning_rate": 5.862675524686698e-05, "loss": 1.0181, "step": 165475 }, { "epoch": 0.42, "learning_rate": 5.8625497005385274e-05, "loss": 1.021, "step": 165480 }, { "epoch": 0.42, "learning_rate": 5.862423876390357e-05, "loss": 1.0206, "step": 165485 }, { "epoch": 0.42, "learning_rate": 5.8622980522421864e-05, "loss": 1.0188, "step": 165490 }, { "epoch": 0.42, "learning_rate": 5.862172228094016e-05, "loss": 1.0186, "step": 165495 }, { "epoch": 0.42, "learning_rate": 5.8620464039458454e-05, "loss": 1.0202, "step": 165500 }, { "epoch": 0.42, "learning_rate": 5.861920579797675e-05, "loss": 1.0227, "step": 165505 }, { "epoch": 0.42, "learning_rate": 5.8617947556495044e-05, "loss": 1.019, "step": 165510 }, { "epoch": 0.42, "learning_rate": 5.861668931501334e-05, "loss": 1.0219, "step": 165515 }, { "epoch": 0.42, "learning_rate": 5.8615431073531634e-05, "loss": 1.0186, "step": 165520 }, { "epoch": 0.42, "learning_rate": 5.861417283204993e-05, "loss": 1.0223, "step": 165525 }, { "epoch": 0.42, "learning_rate": 5.8612914590568224e-05, "loss": 1.0183, "step": 165530 }, { "epoch": 0.42, "learning_rate": 5.8611656349086516e-05, "loss": 1.0168, "step": 165535 }, { "epoch": 0.42, "learning_rate": 5.8610398107604814e-05, "loss": 1.0174, "step": 165540 }, { "epoch": 0.42, "learning_rate": 5.8609139866123106e-05, "loss": 1.0203, "step": 165545 }, { "epoch": 0.42, "learning_rate": 5.8607881624641404e-05, "loss": 1.0155, "step": 165550 }, { "epoch": 0.42, "learning_rate": 5.8606623383159696e-05, "loss": 1.0179, "step": 165555 }, { "epoch": 0.42, "learning_rate": 5.8605365141677994e-05, "loss": 1.0205, "step": 165560 }, { "epoch": 0.42, "learning_rate": 5.8604106900196286e-05, "loss": 1.0189, "step": 165565 }, { "epoch": 0.42, "learning_rate": 5.8602848658714584e-05, "loss": 1.0209, "step": 165570 }, { "epoch": 0.42, "learning_rate": 5.8601590417232876e-05, "loss": 1.0204, "step": 165575 }, { "epoch": 0.42, "learning_rate": 5.8600332175751174e-05, "loss": 1.0203, "step": 165580 }, { "epoch": 0.42, "learning_rate": 5.8599073934269466e-05, "loss": 1.0206, "step": 165585 }, { "epoch": 0.42, "learning_rate": 5.8597815692787764e-05, "loss": 1.0177, "step": 165590 }, { "epoch": 0.42, "learning_rate": 5.8596557451306056e-05, "loss": 1.0206, "step": 165595 }, { "epoch": 0.42, "learning_rate": 5.859529920982435e-05, "loss": 1.0183, "step": 165600 }, { "epoch": 0.42, "learning_rate": 5.8594040968342646e-05, "loss": 1.0165, "step": 165605 }, { "epoch": 0.42, "learning_rate": 5.859278272686094e-05, "loss": 1.0179, "step": 165610 }, { "epoch": 0.42, "learning_rate": 5.8591524485379236e-05, "loss": 1.0183, "step": 165615 }, { "epoch": 0.42, "learning_rate": 5.859026624389753e-05, "loss": 1.0188, "step": 165620 }, { "epoch": 0.42, "learning_rate": 5.8589008002415826e-05, "loss": 1.0198, "step": 165625 }, { "epoch": 0.42, "learning_rate": 5.858774976093412e-05, "loss": 1.0186, "step": 165630 }, { "epoch": 0.42, "learning_rate": 5.8586491519452416e-05, "loss": 1.0434, "step": 165635 }, { "epoch": 0.42, "learning_rate": 5.858523327797071e-05, "loss": 1.0204, "step": 165640 }, { "epoch": 0.42, "learning_rate": 5.8583975036489005e-05, "loss": 1.0221, "step": 165645 }, { "epoch": 0.42, "learning_rate": 5.85827167950073e-05, "loss": 1.021, "step": 165650 }, { "epoch": 0.42, "learning_rate": 5.8581458553525595e-05, "loss": 1.0207, "step": 165655 }, { "epoch": 0.42, "learning_rate": 5.858020031204389e-05, "loss": 1.021, "step": 165660 }, { "epoch": 0.42, "learning_rate": 5.857894207056218e-05, "loss": 1.0201, "step": 165665 }, { "epoch": 0.42, "learning_rate": 5.857768382908048e-05, "loss": 1.0226, "step": 165670 }, { "epoch": 0.42, "learning_rate": 5.857642558759877e-05, "loss": 1.0216, "step": 165675 }, { "epoch": 0.42, "learning_rate": 5.857516734611707e-05, "loss": 1.0196, "step": 165680 }, { "epoch": 0.42, "learning_rate": 5.857390910463536e-05, "loss": 1.0223, "step": 165685 }, { "epoch": 0.42, "learning_rate": 5.857265086315366e-05, "loss": 1.0391, "step": 165690 }, { "epoch": 0.42, "learning_rate": 5.857139262167195e-05, "loss": 1.0231, "step": 165695 }, { "epoch": 0.42, "learning_rate": 5.857013438019025e-05, "loss": 1.0201, "step": 165700 }, { "epoch": 0.42, "learning_rate": 5.856887613870854e-05, "loss": 1.0204, "step": 165705 }, { "epoch": 0.42, "learning_rate": 5.856761789722684e-05, "loss": 1.0177, "step": 165710 }, { "epoch": 0.42, "learning_rate": 5.856635965574513e-05, "loss": 1.0195, "step": 165715 }, { "epoch": 0.42, "learning_rate": 5.856510141426343e-05, "loss": 1.0186, "step": 165720 }, { "epoch": 0.42, "learning_rate": 5.856384317278172e-05, "loss": 1.0211, "step": 165725 }, { "epoch": 0.42, "learning_rate": 5.856258493130001e-05, "loss": 1.0178, "step": 165730 }, { "epoch": 0.42, "learning_rate": 5.856132668981831e-05, "loss": 1.0222, "step": 165735 }, { "epoch": 0.42, "learning_rate": 5.85600684483366e-05, "loss": 1.0189, "step": 165740 }, { "epoch": 0.42, "learning_rate": 5.85588102068549e-05, "loss": 1.0179, "step": 165745 }, { "epoch": 0.42, "learning_rate": 5.855755196537319e-05, "loss": 1.0206, "step": 165750 }, { "epoch": 0.42, "learning_rate": 5.855629372389149e-05, "loss": 1.0198, "step": 165755 }, { "epoch": 0.42, "learning_rate": 5.8555035482409794e-05, "loss": 1.0214, "step": 165760 }, { "epoch": 0.42, "learning_rate": 5.8553777240928085e-05, "loss": 1.021, "step": 165765 }, { "epoch": 0.42, "learning_rate": 5.8552518999446384e-05, "loss": 1.0186, "step": 165770 }, { "epoch": 0.42, "learning_rate": 5.8551260757964675e-05, "loss": 1.0189, "step": 165775 }, { "epoch": 0.42, "learning_rate": 5.8550002516482973e-05, "loss": 1.0188, "step": 165780 }, { "epoch": 0.42, "learning_rate": 5.8548744275001265e-05, "loss": 1.0194, "step": 165785 }, { "epoch": 0.42, "learning_rate": 5.8547486033519563e-05, "loss": 1.0191, "step": 165790 }, { "epoch": 0.42, "learning_rate": 5.8546227792037855e-05, "loss": 1.0204, "step": 165795 }, { "epoch": 0.42, "learning_rate": 5.8544969550556153e-05, "loss": 1.022, "step": 165800 }, { "epoch": 0.42, "learning_rate": 5.8543711309074445e-05, "loss": 1.019, "step": 165805 }, { "epoch": 0.42, "learning_rate": 5.854245306759274e-05, "loss": 1.0198, "step": 165810 }, { "epoch": 0.42, "learning_rate": 5.8541194826111035e-05, "loss": 1.0206, "step": 165815 }, { "epoch": 0.42, "learning_rate": 5.8539936584629327e-05, "loss": 1.0186, "step": 165820 }, { "epoch": 0.42, "learning_rate": 5.8538678343147625e-05, "loss": 1.0204, "step": 165825 }, { "epoch": 0.42, "learning_rate": 5.8537420101665917e-05, "loss": 1.0214, "step": 165830 }, { "epoch": 0.42, "learning_rate": 5.8536161860184215e-05, "loss": 1.0188, "step": 165835 }, { "epoch": 0.42, "learning_rate": 5.8534903618702507e-05, "loss": 1.0209, "step": 165840 }, { "epoch": 0.42, "learning_rate": 5.8533645377220805e-05, "loss": 1.018, "step": 165845 }, { "epoch": 0.42, "learning_rate": 5.8532387135739097e-05, "loss": 1.0185, "step": 165850 }, { "epoch": 0.42, "learning_rate": 5.8531128894257395e-05, "loss": 1.0189, "step": 165855 }, { "epoch": 0.42, "learning_rate": 5.8529870652775686e-05, "loss": 1.0171, "step": 165860 }, { "epoch": 0.42, "learning_rate": 5.8528612411293985e-05, "loss": 1.0218, "step": 165865 }, { "epoch": 0.42, "learning_rate": 5.8527354169812276e-05, "loss": 1.0199, "step": 165870 }, { "epoch": 0.42, "learning_rate": 5.852609592833057e-05, "loss": 1.0199, "step": 165875 }, { "epoch": 0.42, "learning_rate": 5.8524837686848866e-05, "loss": 1.0194, "step": 165880 }, { "epoch": 0.42, "learning_rate": 5.852357944536716e-05, "loss": 1.0183, "step": 165885 }, { "epoch": 0.42, "learning_rate": 5.8522321203885456e-05, "loss": 1.0192, "step": 165890 }, { "epoch": 0.42, "learning_rate": 5.852106296240375e-05, "loss": 1.0189, "step": 165895 }, { "epoch": 0.42, "learning_rate": 5.8519804720922046e-05, "loss": 1.0201, "step": 165900 }, { "epoch": 0.42, "learning_rate": 5.851854647944034e-05, "loss": 1.0183, "step": 165905 }, { "epoch": 0.42, "learning_rate": 5.8517288237958636e-05, "loss": 1.0198, "step": 165910 }, { "epoch": 0.42, "learning_rate": 5.851602999647693e-05, "loss": 1.0188, "step": 165915 }, { "epoch": 0.42, "learning_rate": 5.8514771754995226e-05, "loss": 1.0407, "step": 165920 }, { "epoch": 0.42, "learning_rate": 5.851351351351352e-05, "loss": 1.0206, "step": 165925 }, { "epoch": 0.42, "learning_rate": 5.8512255272031816e-05, "loss": 1.0215, "step": 165930 }, { "epoch": 0.42, "learning_rate": 5.851099703055011e-05, "loss": 1.0211, "step": 165935 }, { "epoch": 0.42, "learning_rate": 5.85097387890684e-05, "loss": 1.0184, "step": 165940 }, { "epoch": 0.42, "learning_rate": 5.85084805475867e-05, "loss": 1.0202, "step": 165945 }, { "epoch": 0.42, "learning_rate": 5.850722230610499e-05, "loss": 1.0248, "step": 165950 }, { "epoch": 0.42, "learning_rate": 5.850596406462329e-05, "loss": 1.0188, "step": 165955 }, { "epoch": 0.42, "learning_rate": 5.850470582314158e-05, "loss": 1.0205, "step": 165960 }, { "epoch": 0.42, "learning_rate": 5.850344758165988e-05, "loss": 1.0171, "step": 165965 }, { "epoch": 0.42, "learning_rate": 5.850218934017817e-05, "loss": 1.0191, "step": 165970 }, { "epoch": 0.42, "learning_rate": 5.850093109869647e-05, "loss": 1.0211, "step": 165975 }, { "epoch": 0.42, "learning_rate": 5.849967285721476e-05, "loss": 1.0203, "step": 165980 }, { "epoch": 0.42, "learning_rate": 5.849841461573306e-05, "loss": 1.0205, "step": 165985 }, { "epoch": 0.42, "learning_rate": 5.849715637425135e-05, "loss": 1.0216, "step": 165990 }, { "epoch": 0.42, "learning_rate": 5.849589813276964e-05, "loss": 1.0186, "step": 165995 }, { "epoch": 0.42, "learning_rate": 5.849463989128794e-05, "loss": 1.0177, "step": 166000 }, { "epoch": 0.42, "learning_rate": 5.849338164980623e-05, "loss": 1.021, "step": 166005 }, { "epoch": 0.42, "learning_rate": 5.849212340832453e-05, "loss": 1.0203, "step": 166010 }, { "epoch": 0.42, "learning_rate": 5.849086516684282e-05, "loss": 1.0204, "step": 166015 }, { "epoch": 0.42, "learning_rate": 5.848960692536112e-05, "loss": 1.0203, "step": 166020 }, { "epoch": 0.42, "learning_rate": 5.848834868387941e-05, "loss": 1.02, "step": 166025 }, { "epoch": 0.42, "learning_rate": 5.848709044239771e-05, "loss": 1.0178, "step": 166030 }, { "epoch": 0.42, "learning_rate": 5.8485832200916e-05, "loss": 1.022, "step": 166035 }, { "epoch": 0.42, "learning_rate": 5.84845739594343e-05, "loss": 1.0213, "step": 166040 }, { "epoch": 0.42, "learning_rate": 5.848331571795259e-05, "loss": 1.0205, "step": 166045 }, { "epoch": 0.42, "learning_rate": 5.848205747647089e-05, "loss": 1.0367, "step": 166050 }, { "epoch": 0.42, "learning_rate": 5.848079923498918e-05, "loss": 1.02, "step": 166055 }, { "epoch": 0.42, "learning_rate": 5.847954099350747e-05, "loss": 1.018, "step": 166060 }, { "epoch": 0.42, "learning_rate": 5.847828275202577e-05, "loss": 1.02, "step": 166065 }, { "epoch": 0.42, "learning_rate": 5.847702451054406e-05, "loss": 1.0204, "step": 166070 }, { "epoch": 0.42, "learning_rate": 5.847576626906236e-05, "loss": 1.0199, "step": 166075 }, { "epoch": 0.42, "learning_rate": 5.847450802758065e-05, "loss": 1.0216, "step": 166080 }, { "epoch": 0.42, "learning_rate": 5.847324978609895e-05, "loss": 1.0186, "step": 166085 }, { "epoch": 0.42, "learning_rate": 5.847199154461724e-05, "loss": 1.0176, "step": 166090 }, { "epoch": 0.42, "learning_rate": 5.847073330313554e-05, "loss": 1.021, "step": 166095 }, { "epoch": 0.42, "learning_rate": 5.846947506165383e-05, "loss": 1.0185, "step": 166100 }, { "epoch": 0.42, "learning_rate": 5.846821682017213e-05, "loss": 1.0184, "step": 166105 }, { "epoch": 0.42, "learning_rate": 5.846695857869042e-05, "loss": 1.0159, "step": 166110 }, { "epoch": 0.42, "learning_rate": 5.846570033720872e-05, "loss": 1.0425, "step": 166115 }, { "epoch": 0.42, "learning_rate": 5.846444209572701e-05, "loss": 1.0198, "step": 166120 }, { "epoch": 0.42, "learning_rate": 5.8463183854245304e-05, "loss": 1.02, "step": 166125 }, { "epoch": 0.42, "learning_rate": 5.84619256127636e-05, "loss": 1.0275, "step": 166130 }, { "epoch": 0.42, "learning_rate": 5.8460667371281894e-05, "loss": 1.0217, "step": 166135 }, { "epoch": 0.42, "learning_rate": 5.845940912980019e-05, "loss": 1.0224, "step": 166140 }, { "epoch": 0.42, "learning_rate": 5.8458150888318484e-05, "loss": 1.0192, "step": 166145 }, { "epoch": 0.42, "learning_rate": 5.845689264683678e-05, "loss": 1.0211, "step": 166150 }, { "epoch": 0.42, "learning_rate": 5.8455634405355074e-05, "loss": 1.0204, "step": 166155 }, { "epoch": 0.42, "learning_rate": 5.845437616387337e-05, "loss": 1.0191, "step": 166160 }, { "epoch": 0.42, "learning_rate": 5.8453117922391664e-05, "loss": 1.0197, "step": 166165 }, { "epoch": 0.42, "learning_rate": 5.845185968090996e-05, "loss": 1.0206, "step": 166170 }, { "epoch": 0.42, "learning_rate": 5.8450601439428254e-05, "loss": 1.019, "step": 166175 }, { "epoch": 0.42, "learning_rate": 5.844934319794655e-05, "loss": 1.0194, "step": 166180 }, { "epoch": 0.42, "learning_rate": 5.8448084956464844e-05, "loss": 1.0192, "step": 166185 }, { "epoch": 0.42, "learning_rate": 5.8446826714983135e-05, "loss": 1.0191, "step": 166190 }, { "epoch": 0.42, "learning_rate": 5.8445568473501434e-05, "loss": 1.0213, "step": 166195 }, { "epoch": 0.42, "learning_rate": 5.8444310232019725e-05, "loss": 1.019, "step": 166200 }, { "epoch": 0.42, "learning_rate": 5.8443051990538024e-05, "loss": 1.0215, "step": 166205 }, { "epoch": 0.42, "learning_rate": 5.8441793749056315e-05, "loss": 1.0303, "step": 166210 }, { "epoch": 0.42, "learning_rate": 5.8440535507574614e-05, "loss": 1.0195, "step": 166215 }, { "epoch": 0.42, "learning_rate": 5.8439277266092905e-05, "loss": 1.02, "step": 166220 }, { "epoch": 0.42, "learning_rate": 5.8438019024611203e-05, "loss": 1.0194, "step": 166225 }, { "epoch": 0.42, "learning_rate": 5.8436760783129495e-05, "loss": 1.0188, "step": 166230 }, { "epoch": 0.42, "learning_rate": 5.8435502541647793e-05, "loss": 1.0191, "step": 166235 }, { "epoch": 0.42, "learning_rate": 5.8434244300166085e-05, "loss": 1.0213, "step": 166240 }, { "epoch": 0.42, "learning_rate": 5.8432986058684383e-05, "loss": 1.017, "step": 166245 }, { "epoch": 0.42, "learning_rate": 5.8431727817202675e-05, "loss": 1.0204, "step": 166250 }, { "epoch": 0.42, "learning_rate": 5.843046957572097e-05, "loss": 1.0216, "step": 166255 }, { "epoch": 0.42, "learning_rate": 5.842921133423928e-05, "loss": 1.0152, "step": 166260 }, { "epoch": 0.42, "learning_rate": 5.842795309275757e-05, "loss": 1.0206, "step": 166265 }, { "epoch": 0.42, "learning_rate": 5.842669485127586e-05, "loss": 1.0197, "step": 166270 }, { "epoch": 0.42, "learning_rate": 5.842543660979416e-05, "loss": 1.0217, "step": 166275 }, { "epoch": 0.42, "learning_rate": 5.842417836831245e-05, "loss": 1.0205, "step": 166280 }, { "epoch": 0.42, "learning_rate": 5.842292012683075e-05, "loss": 1.0196, "step": 166285 }, { "epoch": 0.42, "learning_rate": 5.842166188534904e-05, "loss": 1.0202, "step": 166290 }, { "epoch": 0.42, "learning_rate": 5.842040364386734e-05, "loss": 1.0208, "step": 166295 }, { "epoch": 0.42, "learning_rate": 5.841914540238563e-05, "loss": 1.0195, "step": 166300 }, { "epoch": 0.42, "learning_rate": 5.841788716090393e-05, "loss": 1.0169, "step": 166305 }, { "epoch": 0.42, "learning_rate": 5.841662891942222e-05, "loss": 1.0186, "step": 166310 }, { "epoch": 0.42, "learning_rate": 5.841537067794052e-05, "loss": 1.0216, "step": 166315 }, { "epoch": 0.42, "learning_rate": 5.841411243645881e-05, "loss": 1.0189, "step": 166320 }, { "epoch": 0.42, "learning_rate": 5.841285419497711e-05, "loss": 1.0199, "step": 166325 }, { "epoch": 0.42, "learning_rate": 5.84115959534954e-05, "loss": 1.021, "step": 166330 }, { "epoch": 0.42, "learning_rate": 5.841033771201369e-05, "loss": 1.019, "step": 166335 }, { "epoch": 0.42, "learning_rate": 5.840907947053199e-05, "loss": 1.035, "step": 166340 }, { "epoch": 0.42, "learning_rate": 5.840782122905028e-05, "loss": 1.0184, "step": 166345 }, { "epoch": 0.42, "learning_rate": 5.840656298756858e-05, "loss": 1.0166, "step": 166350 }, { "epoch": 0.42, "learning_rate": 5.840530474608687e-05, "loss": 1.0201, "step": 166355 }, { "epoch": 0.42, "learning_rate": 5.840404650460517e-05, "loss": 1.0195, "step": 166360 }, { "epoch": 0.42, "learning_rate": 5.840278826312346e-05, "loss": 1.0209, "step": 166365 }, { "epoch": 0.42, "learning_rate": 5.840153002164176e-05, "loss": 1.0184, "step": 166370 }, { "epoch": 0.42, "learning_rate": 5.840027178016005e-05, "loss": 1.0194, "step": 166375 }, { "epoch": 0.42, "learning_rate": 5.839901353867835e-05, "loss": 1.0192, "step": 166380 }, { "epoch": 0.42, "learning_rate": 5.839775529719664e-05, "loss": 1.0207, "step": 166385 }, { "epoch": 0.42, "learning_rate": 5.839649705571494e-05, "loss": 1.0191, "step": 166390 }, { "epoch": 0.42, "learning_rate": 5.839523881423323e-05, "loss": 1.0207, "step": 166395 }, { "epoch": 0.42, "learning_rate": 5.8393980572751525e-05, "loss": 1.0201, "step": 166400 }, { "epoch": 0.42, "learning_rate": 5.839272233126982e-05, "loss": 1.0208, "step": 166405 }, { "epoch": 0.42, "learning_rate": 5.8391464089788115e-05, "loss": 1.0212, "step": 166410 }, { "epoch": 0.42, "learning_rate": 5.839020584830641e-05, "loss": 1.0185, "step": 166415 }, { "epoch": 0.42, "learning_rate": 5.8388947606824705e-05, "loss": 1.0189, "step": 166420 }, { "epoch": 0.42, "learning_rate": 5.8387689365343e-05, "loss": 1.0207, "step": 166425 }, { "epoch": 0.42, "learning_rate": 5.8386431123861295e-05, "loss": 1.0193, "step": 166430 }, { "epoch": 0.42, "learning_rate": 5.838517288237959e-05, "loss": 1.02, "step": 166435 }, { "epoch": 0.42, "learning_rate": 5.8383914640897884e-05, "loss": 1.0193, "step": 166440 }, { "epoch": 0.42, "learning_rate": 5.838265639941618e-05, "loss": 1.0213, "step": 166445 }, { "epoch": 0.42, "learning_rate": 5.8381398157934474e-05, "loss": 1.0188, "step": 166450 }, { "epoch": 0.42, "learning_rate": 5.838013991645277e-05, "loss": 1.0204, "step": 166455 }, { "epoch": 0.42, "learning_rate": 5.8378881674971064e-05, "loss": 1.0187, "step": 166460 }, { "epoch": 0.42, "learning_rate": 5.8377623433489356e-05, "loss": 1.0215, "step": 166465 }, { "epoch": 0.42, "learning_rate": 5.8376365192007654e-05, "loss": 1.0182, "step": 166470 }, { "epoch": 0.42, "learning_rate": 5.8375106950525946e-05, "loss": 1.0182, "step": 166475 }, { "epoch": 0.42, "learning_rate": 5.8373848709044244e-05, "loss": 1.0207, "step": 166480 }, { "epoch": 0.42, "learning_rate": 5.8372590467562536e-05, "loss": 1.0228, "step": 166485 }, { "epoch": 0.42, "learning_rate": 5.8371332226080834e-05, "loss": 1.0389, "step": 166490 }, { "epoch": 0.42, "learning_rate": 5.8370073984599126e-05, "loss": 1.0202, "step": 166495 }, { "epoch": 0.42, "learning_rate": 5.8368815743117424e-05, "loss": 1.0417, "step": 166500 }, { "epoch": 0.42, "learning_rate": 5.8367557501635716e-05, "loss": 1.0202, "step": 166505 }, { "epoch": 0.42, "learning_rate": 5.8366299260154014e-05, "loss": 1.0192, "step": 166510 }, { "epoch": 0.42, "learning_rate": 5.8365041018672306e-05, "loss": 1.0212, "step": 166515 }, { "epoch": 0.42, "learning_rate": 5.8363782777190604e-05, "loss": 1.0172, "step": 166520 }, { "epoch": 0.42, "learning_rate": 5.8362524535708896e-05, "loss": 1.0184, "step": 166525 }, { "epoch": 0.42, "learning_rate": 5.836126629422719e-05, "loss": 1.0197, "step": 166530 }, { "epoch": 0.42, "learning_rate": 5.8360008052745486e-05, "loss": 1.0203, "step": 166535 }, { "epoch": 0.42, "learning_rate": 5.835874981126378e-05, "loss": 1.0208, "step": 166540 }, { "epoch": 0.42, "learning_rate": 5.8357491569782076e-05, "loss": 1.021, "step": 166545 }, { "epoch": 0.42, "learning_rate": 5.835623332830037e-05, "loss": 1.0199, "step": 166550 }, { "epoch": 0.42, "learning_rate": 5.8354975086818666e-05, "loss": 1.0201, "step": 166555 }, { "epoch": 0.42, "learning_rate": 5.835371684533696e-05, "loss": 1.0196, "step": 166560 }, { "epoch": 0.42, "learning_rate": 5.8352458603855256e-05, "loss": 1.019, "step": 166565 }, { "epoch": 0.42, "learning_rate": 5.835120036237355e-05, "loss": 1.0212, "step": 166570 }, { "epoch": 0.42, "learning_rate": 5.8349942120891846e-05, "loss": 1.0197, "step": 166575 }, { "epoch": 0.42, "learning_rate": 5.834868387941014e-05, "loss": 1.0206, "step": 166580 }, { "epoch": 0.42, "learning_rate": 5.8347425637928436e-05, "loss": 1.0199, "step": 166585 }, { "epoch": 0.42, "learning_rate": 5.834616739644673e-05, "loss": 1.0219, "step": 166590 }, { "epoch": 0.42, "learning_rate": 5.834490915496502e-05, "loss": 1.0195, "step": 166595 }, { "epoch": 0.42, "learning_rate": 5.834365091348332e-05, "loss": 1.0198, "step": 166600 }, { "epoch": 0.42, "learning_rate": 5.834239267200161e-05, "loss": 1.0198, "step": 166605 }, { "epoch": 0.42, "learning_rate": 5.834113443051991e-05, "loss": 1.0167, "step": 166610 }, { "epoch": 0.42, "learning_rate": 5.83398761890382e-05, "loss": 1.0213, "step": 166615 }, { "epoch": 0.42, "learning_rate": 5.83386179475565e-05, "loss": 1.0195, "step": 166620 }, { "epoch": 0.42, "learning_rate": 5.833735970607479e-05, "loss": 1.0191, "step": 166625 }, { "epoch": 0.42, "learning_rate": 5.833610146459309e-05, "loss": 1.0203, "step": 166630 }, { "epoch": 0.42, "learning_rate": 5.833484322311138e-05, "loss": 1.0214, "step": 166635 }, { "epoch": 0.42, "learning_rate": 5.833358498162968e-05, "loss": 1.0199, "step": 166640 }, { "epoch": 0.42, "learning_rate": 5.833232674014797e-05, "loss": 1.02, "step": 166645 }, { "epoch": 0.42, "learning_rate": 5.833106849866626e-05, "loss": 1.0178, "step": 166650 }, { "epoch": 0.42, "learning_rate": 5.832981025718456e-05, "loss": 1.0212, "step": 166655 }, { "epoch": 0.42, "learning_rate": 5.832855201570285e-05, "loss": 1.0187, "step": 166660 }, { "epoch": 0.42, "learning_rate": 5.832729377422115e-05, "loss": 1.0193, "step": 166665 }, { "epoch": 0.42, "learning_rate": 5.832603553273944e-05, "loss": 1.0175, "step": 166670 }, { "epoch": 0.42, "learning_rate": 5.832477729125774e-05, "loss": 1.0212, "step": 166675 }, { "epoch": 0.42, "learning_rate": 5.832351904977603e-05, "loss": 1.0199, "step": 166680 }, { "epoch": 0.42, "learning_rate": 5.832226080829433e-05, "loss": 1.0182, "step": 166685 }, { "epoch": 0.42, "learning_rate": 5.832100256681262e-05, "loss": 1.0224, "step": 166690 }, { "epoch": 0.42, "learning_rate": 5.831974432533092e-05, "loss": 1.0183, "step": 166695 }, { "epoch": 0.42, "learning_rate": 5.831848608384921e-05, "loss": 1.02, "step": 166700 }, { "epoch": 0.42, "learning_rate": 5.831722784236751e-05, "loss": 1.0176, "step": 166705 }, { "epoch": 0.42, "learning_rate": 5.83159696008858e-05, "loss": 1.0197, "step": 166710 }, { "epoch": 0.42, "learning_rate": 5.831471135940409e-05, "loss": 1.0198, "step": 166715 }, { "epoch": 0.42, "learning_rate": 5.831345311792239e-05, "loss": 1.021, "step": 166720 }, { "epoch": 0.42, "learning_rate": 5.831219487644068e-05, "loss": 1.0186, "step": 166725 }, { "epoch": 0.42, "learning_rate": 5.831093663495898e-05, "loss": 1.0196, "step": 166730 }, { "epoch": 0.42, "learning_rate": 5.830967839347727e-05, "loss": 1.0194, "step": 166735 }, { "epoch": 0.42, "learning_rate": 5.830842015199557e-05, "loss": 1.0207, "step": 166740 }, { "epoch": 0.42, "learning_rate": 5.830716191051386e-05, "loss": 1.0199, "step": 166745 }, { "epoch": 0.42, "learning_rate": 5.830590366903216e-05, "loss": 1.0367, "step": 166750 }, { "epoch": 0.42, "learning_rate": 5.830464542755045e-05, "loss": 1.0204, "step": 166755 }, { "epoch": 0.42, "learning_rate": 5.830338718606876e-05, "loss": 1.0227, "step": 166760 }, { "epoch": 0.42, "learning_rate": 5.8302128944587055e-05, "loss": 1.0225, "step": 166765 }, { "epoch": 0.42, "learning_rate": 5.830087070310535e-05, "loss": 1.0187, "step": 166770 }, { "epoch": 0.42, "learning_rate": 5.8299612461623645e-05, "loss": 1.0195, "step": 166775 }, { "epoch": 0.42, "learning_rate": 5.829835422014194e-05, "loss": 1.0198, "step": 166780 }, { "epoch": 0.42, "learning_rate": 5.8297095978660235e-05, "loss": 1.0216, "step": 166785 }, { "epoch": 0.42, "learning_rate": 5.829583773717853e-05, "loss": 1.0202, "step": 166790 }, { "epoch": 0.42, "learning_rate": 5.829457949569682e-05, "loss": 1.0194, "step": 166795 }, { "epoch": 0.42, "learning_rate": 5.829332125421512e-05, "loss": 1.0214, "step": 166800 }, { "epoch": 0.42, "learning_rate": 5.829206301273341e-05, "loss": 1.0184, "step": 166805 }, { "epoch": 0.42, "learning_rate": 5.829080477125171e-05, "loss": 1.0194, "step": 166810 }, { "epoch": 0.42, "learning_rate": 5.828954652977e-05, "loss": 1.0198, "step": 166815 }, { "epoch": 0.42, "learning_rate": 5.82882882882883e-05, "loss": 1.0191, "step": 166820 }, { "epoch": 0.42, "learning_rate": 5.828703004680659e-05, "loss": 1.0191, "step": 166825 }, { "epoch": 0.42, "learning_rate": 5.8285771805324887e-05, "loss": 1.0196, "step": 166830 }, { "epoch": 0.42, "learning_rate": 5.828451356384318e-05, "loss": 1.0208, "step": 166835 }, { "epoch": 0.42, "learning_rate": 5.8283255322361477e-05, "loss": 1.0406, "step": 166840 }, { "epoch": 0.42, "learning_rate": 5.828199708087977e-05, "loss": 1.0221, "step": 166845 }, { "epoch": 0.42, "learning_rate": 5.8280738839398067e-05, "loss": 1.0187, "step": 166850 }, { "epoch": 0.42, "learning_rate": 5.827948059791636e-05, "loss": 1.0198, "step": 166855 }, { "epoch": 0.42, "learning_rate": 5.827822235643465e-05, "loss": 1.0175, "step": 166860 }, { "epoch": 0.42, "learning_rate": 5.827696411495295e-05, "loss": 1.0201, "step": 166865 }, { "epoch": 0.42, "learning_rate": 5.827570587347124e-05, "loss": 1.0175, "step": 166870 }, { "epoch": 0.42, "learning_rate": 5.827444763198954e-05, "loss": 1.0206, "step": 166875 }, { "epoch": 0.42, "learning_rate": 5.827318939050783e-05, "loss": 1.0199, "step": 166880 }, { "epoch": 0.42, "learning_rate": 5.827193114902613e-05, "loss": 1.0191, "step": 166885 }, { "epoch": 0.42, "learning_rate": 5.827067290754442e-05, "loss": 1.0212, "step": 166890 }, { "epoch": 0.42, "learning_rate": 5.826941466606272e-05, "loss": 1.0206, "step": 166895 }, { "epoch": 0.42, "learning_rate": 5.826815642458101e-05, "loss": 1.0195, "step": 166900 }, { "epoch": 0.42, "learning_rate": 5.826689818309931e-05, "loss": 1.0195, "step": 166905 }, { "epoch": 0.42, "learning_rate": 5.82656399416176e-05, "loss": 1.0188, "step": 166910 }, { "epoch": 0.42, "learning_rate": 5.82643817001359e-05, "loss": 1.0187, "step": 166915 }, { "epoch": 0.42, "learning_rate": 5.826312345865419e-05, "loss": 1.0184, "step": 166920 }, { "epoch": 0.42, "learning_rate": 5.826186521717248e-05, "loss": 1.0196, "step": 166925 }, { "epoch": 0.42, "learning_rate": 5.826060697569078e-05, "loss": 1.0196, "step": 166930 }, { "epoch": 0.42, "learning_rate": 5.825934873420907e-05, "loss": 1.0191, "step": 166935 }, { "epoch": 0.42, "learning_rate": 5.825809049272737e-05, "loss": 1.0187, "step": 166940 }, { "epoch": 0.42, "learning_rate": 5.825683225124566e-05, "loss": 1.0183, "step": 166945 }, { "epoch": 0.42, "learning_rate": 5.825557400976396e-05, "loss": 1.0174, "step": 166950 }, { "epoch": 0.42, "learning_rate": 5.825431576828225e-05, "loss": 1.02, "step": 166955 }, { "epoch": 0.42, "learning_rate": 5.825305752680055e-05, "loss": 1.0196, "step": 166960 }, { "epoch": 0.42, "learning_rate": 5.825179928531884e-05, "loss": 1.0202, "step": 166965 }, { "epoch": 0.42, "learning_rate": 5.825054104383714e-05, "loss": 1.023, "step": 166970 }, { "epoch": 0.42, "learning_rate": 5.824928280235543e-05, "loss": 1.0182, "step": 166975 }, { "epoch": 0.42, "learning_rate": 5.824802456087373e-05, "loss": 1.043, "step": 166980 }, { "epoch": 0.42, "learning_rate": 5.824676631939202e-05, "loss": 1.0195, "step": 166985 }, { "epoch": 0.42, "learning_rate": 5.824550807791031e-05, "loss": 1.0217, "step": 166990 }, { "epoch": 0.42, "learning_rate": 5.824424983642861e-05, "loss": 1.0199, "step": 166995 }, { "epoch": 0.42, "learning_rate": 5.82429915949469e-05, "loss": 1.0189, "step": 167000 }, { "epoch": 0.42, "learning_rate": 5.82417333534652e-05, "loss": 1.0189, "step": 167005 }, { "epoch": 0.42, "learning_rate": 5.824047511198349e-05, "loss": 1.0215, "step": 167010 }, { "epoch": 0.42, "learning_rate": 5.823921687050179e-05, "loss": 1.0216, "step": 167015 }, { "epoch": 0.42, "learning_rate": 5.823795862902008e-05, "loss": 1.0201, "step": 167020 }, { "epoch": 0.42, "learning_rate": 5.823670038753838e-05, "loss": 1.0204, "step": 167025 }, { "epoch": 0.42, "learning_rate": 5.823544214605667e-05, "loss": 1.023, "step": 167030 }, { "epoch": 0.42, "learning_rate": 5.823418390457497e-05, "loss": 1.0208, "step": 167035 }, { "epoch": 0.42, "learning_rate": 5.823292566309326e-05, "loss": 1.0179, "step": 167040 }, { "epoch": 0.42, "learning_rate": 5.823166742161156e-05, "loss": 1.0201, "step": 167045 }, { "epoch": 0.42, "learning_rate": 5.823040918012985e-05, "loss": 1.0177, "step": 167050 }, { "epoch": 0.42, "learning_rate": 5.8229150938648144e-05, "loss": 1.0204, "step": 167055 }, { "epoch": 0.42, "learning_rate": 5.822789269716644e-05, "loss": 1.0216, "step": 167060 }, { "epoch": 0.42, "learning_rate": 5.8226634455684734e-05, "loss": 1.0382, "step": 167065 }, { "epoch": 0.42, "learning_rate": 5.822537621420303e-05, "loss": 1.0189, "step": 167070 }, { "epoch": 0.42, "learning_rate": 5.8224117972721324e-05, "loss": 1.0196, "step": 167075 }, { "epoch": 0.42, "learning_rate": 5.822285973123962e-05, "loss": 1.0191, "step": 167080 }, { "epoch": 0.42, "learning_rate": 5.8221601489757914e-05, "loss": 1.0209, "step": 167085 }, { "epoch": 0.42, "learning_rate": 5.822034324827621e-05, "loss": 1.0218, "step": 167090 }, { "epoch": 0.42, "learning_rate": 5.8219085006794504e-05, "loss": 1.0195, "step": 167095 }, { "epoch": 0.42, "learning_rate": 5.82178267653128e-05, "loss": 1.0229, "step": 167100 }, { "epoch": 0.42, "learning_rate": 5.8216568523831094e-05, "loss": 1.0195, "step": 167105 }, { "epoch": 0.42, "learning_rate": 5.821531028234939e-05, "loss": 1.0185, "step": 167110 }, { "epoch": 0.42, "learning_rate": 5.8214052040867684e-05, "loss": 1.0197, "step": 167115 }, { "epoch": 0.42, "learning_rate": 5.8212793799385975e-05, "loss": 1.0442, "step": 167120 }, { "epoch": 0.42, "learning_rate": 5.8211535557904274e-05, "loss": 1.0216, "step": 167125 }, { "epoch": 0.42, "learning_rate": 5.8210277316422565e-05, "loss": 1.0201, "step": 167130 }, { "epoch": 0.42, "learning_rate": 5.8209019074940864e-05, "loss": 1.0221, "step": 167135 }, { "epoch": 0.42, "learning_rate": 5.8207760833459155e-05, "loss": 1.0194, "step": 167140 }, { "epoch": 0.42, "learning_rate": 5.8206502591977454e-05, "loss": 1.0182, "step": 167145 }, { "epoch": 0.42, "learning_rate": 5.8205244350495745e-05, "loss": 1.0196, "step": 167150 }, { "epoch": 0.42, "learning_rate": 5.8203986109014044e-05, "loss": 1.0203, "step": 167155 }, { "epoch": 0.42, "learning_rate": 5.8202727867532335e-05, "loss": 1.0202, "step": 167160 }, { "epoch": 0.42, "learning_rate": 5.8201469626050634e-05, "loss": 1.0222, "step": 167165 }, { "epoch": 0.42, "learning_rate": 5.8200211384568925e-05, "loss": 1.021, "step": 167170 }, { "epoch": 0.42, "learning_rate": 5.8198953143087224e-05, "loss": 1.0215, "step": 167175 }, { "epoch": 0.42, "learning_rate": 5.8197694901605515e-05, "loss": 1.0183, "step": 167180 }, { "epoch": 0.42, "learning_rate": 5.819643666012381e-05, "loss": 1.0209, "step": 167185 }, { "epoch": 0.42, "learning_rate": 5.8195178418642105e-05, "loss": 1.0189, "step": 167190 }, { "epoch": 0.42, "learning_rate": 5.81939201771604e-05, "loss": 1.0202, "step": 167195 }, { "epoch": 0.42, "learning_rate": 5.8192661935678695e-05, "loss": 1.0211, "step": 167200 }, { "epoch": 0.42, "learning_rate": 5.819140369419699e-05, "loss": 1.0215, "step": 167205 }, { "epoch": 0.42, "learning_rate": 5.8190145452715285e-05, "loss": 1.0196, "step": 167210 }, { "epoch": 0.42, "learning_rate": 5.818888721123358e-05, "loss": 1.0194, "step": 167215 }, { "epoch": 0.42, "learning_rate": 5.8187628969751875e-05, "loss": 1.0179, "step": 167220 }, { "epoch": 0.42, "learning_rate": 5.818637072827017e-05, "loss": 1.0206, "step": 167225 }, { "epoch": 0.42, "learning_rate": 5.8185112486788465e-05, "loss": 1.0183, "step": 167230 }, { "epoch": 0.42, "learning_rate": 5.818385424530676e-05, "loss": 1.0225, "step": 167235 }, { "epoch": 0.42, "learning_rate": 5.8182596003825055e-05, "loss": 1.0197, "step": 167240 }, { "epoch": 0.42, "learning_rate": 5.818133776234335e-05, "loss": 1.0196, "step": 167245 }, { "epoch": 0.42, "learning_rate": 5.818007952086164e-05, "loss": 1.0195, "step": 167250 }, { "epoch": 0.42, "learning_rate": 5.817882127937994e-05, "loss": 1.0215, "step": 167255 }, { "epoch": 0.42, "learning_rate": 5.817756303789824e-05, "loss": 1.0202, "step": 167260 }, { "epoch": 0.42, "learning_rate": 5.8176304796416533e-05, "loss": 1.0143, "step": 167265 }, { "epoch": 0.42, "learning_rate": 5.817504655493483e-05, "loss": 1.0203, "step": 167270 }, { "epoch": 0.42, "learning_rate": 5.817378831345312e-05, "loss": 1.0213, "step": 167275 }, { "epoch": 0.42, "learning_rate": 5.817253007197142e-05, "loss": 1.0204, "step": 167280 }, { "epoch": 0.42, "learning_rate": 5.817127183048971e-05, "loss": 1.0211, "step": 167285 }, { "epoch": 0.42, "learning_rate": 5.817001358900801e-05, "loss": 1.0209, "step": 167290 }, { "epoch": 0.42, "learning_rate": 5.81687553475263e-05, "loss": 1.0196, "step": 167295 }, { "epoch": 0.42, "learning_rate": 5.81674971060446e-05, "loss": 1.0201, "step": 167300 }, { "epoch": 0.42, "learning_rate": 5.816623886456289e-05, "loss": 1.021, "step": 167305 }, { "epoch": 0.42, "learning_rate": 5.816498062308119e-05, "loss": 1.0166, "step": 167310 }, { "epoch": 0.42, "learning_rate": 5.816372238159948e-05, "loss": 1.0193, "step": 167315 }, { "epoch": 0.42, "learning_rate": 5.816246414011778e-05, "loss": 1.0191, "step": 167320 }, { "epoch": 0.42, "learning_rate": 5.816120589863607e-05, "loss": 1.0199, "step": 167325 }, { "epoch": 0.42, "learning_rate": 5.8159947657154365e-05, "loss": 1.0201, "step": 167330 }, { "epoch": 0.42, "learning_rate": 5.815868941567266e-05, "loss": 1.0226, "step": 167335 }, { "epoch": 0.42, "learning_rate": 5.8157431174190955e-05, "loss": 1.0196, "step": 167340 }, { "epoch": 0.42, "learning_rate": 5.815617293270925e-05, "loss": 1.0179, "step": 167345 }, { "epoch": 0.42, "learning_rate": 5.8154914691227545e-05, "loss": 1.0199, "step": 167350 }, { "epoch": 0.42, "learning_rate": 5.815365644974584e-05, "loss": 1.0195, "step": 167355 }, { "epoch": 0.42, "learning_rate": 5.8152398208264135e-05, "loss": 1.0221, "step": 167360 }, { "epoch": 0.42, "learning_rate": 5.815113996678243e-05, "loss": 1.0198, "step": 167365 }, { "epoch": 0.42, "learning_rate": 5.8149881725300725e-05, "loss": 1.0187, "step": 167370 }, { "epoch": 0.42, "learning_rate": 5.814862348381902e-05, "loss": 1.0198, "step": 167375 }, { "epoch": 0.42, "learning_rate": 5.8147365242337315e-05, "loss": 1.0205, "step": 167380 }, { "epoch": 0.42, "learning_rate": 5.814610700085561e-05, "loss": 1.0224, "step": 167385 }, { "epoch": 0.42, "learning_rate": 5.8144848759373905e-05, "loss": 1.019, "step": 167390 }, { "epoch": 0.42, "learning_rate": 5.8143590517892196e-05, "loss": 1.0171, "step": 167395 }, { "epoch": 0.42, "learning_rate": 5.8142332276410495e-05, "loss": 1.0171, "step": 167400 }, { "epoch": 0.42, "learning_rate": 5.8141074034928786e-05, "loss": 1.0211, "step": 167405 }, { "epoch": 0.42, "learning_rate": 5.8139815793447085e-05, "loss": 1.0178, "step": 167410 }, { "epoch": 0.42, "learning_rate": 5.8138557551965376e-05, "loss": 1.0183, "step": 167415 }, { "epoch": 0.42, "learning_rate": 5.8137299310483675e-05, "loss": 1.0184, "step": 167420 }, { "epoch": 0.42, "learning_rate": 5.8136041069001966e-05, "loss": 1.0203, "step": 167425 }, { "epoch": 0.42, "learning_rate": 5.8134782827520265e-05, "loss": 1.0189, "step": 167430 }, { "epoch": 0.42, "learning_rate": 5.8133524586038556e-05, "loss": 1.0196, "step": 167435 }, { "epoch": 0.42, "learning_rate": 5.8132266344556854e-05, "loss": 1.0209, "step": 167440 }, { "epoch": 0.42, "learning_rate": 5.8131008103075146e-05, "loss": 1.0204, "step": 167445 }, { "epoch": 0.42, "learning_rate": 5.812974986159344e-05, "loss": 1.0238, "step": 167450 }, { "epoch": 0.42, "learning_rate": 5.8128491620111736e-05, "loss": 1.02, "step": 167455 }, { "epoch": 0.42, "learning_rate": 5.812723337863003e-05, "loss": 1.0215, "step": 167460 }, { "epoch": 0.42, "learning_rate": 5.8125975137148326e-05, "loss": 1.0175, "step": 167465 }, { "epoch": 0.42, "learning_rate": 5.812471689566662e-05, "loss": 1.0201, "step": 167470 }, { "epoch": 0.42, "learning_rate": 5.8123458654184916e-05, "loss": 1.0187, "step": 167475 }, { "epoch": 0.42, "learning_rate": 5.812220041270321e-05, "loss": 1.0182, "step": 167480 }, { "epoch": 0.42, "learning_rate": 5.8120942171221506e-05, "loss": 1.0188, "step": 167485 }, { "epoch": 0.42, "learning_rate": 5.81196839297398e-05, "loss": 1.0195, "step": 167490 }, { "epoch": 0.42, "learning_rate": 5.8118425688258096e-05, "loss": 1.0183, "step": 167495 }, { "epoch": 0.42, "learning_rate": 5.811716744677639e-05, "loss": 1.0207, "step": 167500 }, { "epoch": 0.42, "learning_rate": 5.8115909205294686e-05, "loss": 1.0188, "step": 167505 }, { "epoch": 0.42, "learning_rate": 5.811465096381298e-05, "loss": 1.0181, "step": 167510 }, { "epoch": 0.42, "learning_rate": 5.811339272233127e-05, "loss": 1.0201, "step": 167515 }, { "epoch": 0.42, "learning_rate": 5.811213448084957e-05, "loss": 1.022, "step": 167520 }, { "epoch": 0.42, "learning_rate": 5.811087623936786e-05, "loss": 1.0189, "step": 167525 }, { "epoch": 0.42, "learning_rate": 5.810961799788616e-05, "loss": 1.0212, "step": 167530 }, { "epoch": 0.42, "learning_rate": 5.810835975640445e-05, "loss": 1.0196, "step": 167535 }, { "epoch": 0.42, "learning_rate": 5.810710151492275e-05, "loss": 1.0214, "step": 167540 }, { "epoch": 0.42, "learning_rate": 5.810584327344104e-05, "loss": 1.0206, "step": 167545 }, { "epoch": 0.42, "learning_rate": 5.810458503195934e-05, "loss": 1.0188, "step": 167550 }, { "epoch": 0.42, "learning_rate": 5.810332679047763e-05, "loss": 1.0189, "step": 167555 }, { "epoch": 0.42, "learning_rate": 5.810206854899593e-05, "loss": 1.0191, "step": 167560 }, { "epoch": 0.42, "learning_rate": 5.810081030751422e-05, "loss": 1.0195, "step": 167565 }, { "epoch": 0.42, "learning_rate": 5.809955206603252e-05, "loss": 1.0181, "step": 167570 }, { "epoch": 0.42, "learning_rate": 5.809829382455081e-05, "loss": 1.0187, "step": 167575 }, { "epoch": 0.42, "learning_rate": 5.80970355830691e-05, "loss": 1.0198, "step": 167580 }, { "epoch": 0.42, "learning_rate": 5.80957773415874e-05, "loss": 1.0185, "step": 167585 }, { "epoch": 0.42, "learning_rate": 5.809451910010569e-05, "loss": 1.0203, "step": 167590 }, { "epoch": 0.42, "learning_rate": 5.809326085862399e-05, "loss": 1.0194, "step": 167595 }, { "epoch": 0.42, "learning_rate": 5.809200261714228e-05, "loss": 1.0198, "step": 167600 }, { "epoch": 0.42, "learning_rate": 5.809074437566058e-05, "loss": 1.0212, "step": 167605 }, { "epoch": 0.42, "learning_rate": 5.808948613417887e-05, "loss": 1.0185, "step": 167610 }, { "epoch": 0.42, "learning_rate": 5.808822789269717e-05, "loss": 1.0197, "step": 167615 }, { "epoch": 0.42, "learning_rate": 5.808696965121546e-05, "loss": 1.0191, "step": 167620 }, { "epoch": 0.42, "learning_rate": 5.808571140973376e-05, "loss": 1.0177, "step": 167625 }, { "epoch": 0.42, "learning_rate": 5.808445316825205e-05, "loss": 1.0192, "step": 167630 }, { "epoch": 0.42, "learning_rate": 5.808319492677035e-05, "loss": 1.019, "step": 167635 }, { "epoch": 0.42, "learning_rate": 5.808193668528864e-05, "loss": 1.0218, "step": 167640 }, { "epoch": 0.42, "learning_rate": 5.808067844380693e-05, "loss": 1.0226, "step": 167645 }, { "epoch": 0.42, "learning_rate": 5.807942020232523e-05, "loss": 1.0166, "step": 167650 }, { "epoch": 0.42, "learning_rate": 5.807816196084352e-05, "loss": 1.02, "step": 167655 }, { "epoch": 0.42, "learning_rate": 5.807690371936182e-05, "loss": 1.0205, "step": 167660 }, { "epoch": 0.42, "learning_rate": 5.807564547788011e-05, "loss": 1.0201, "step": 167665 }, { "epoch": 0.42, "learning_rate": 5.807438723639841e-05, "loss": 1.0174, "step": 167670 }, { "epoch": 0.42, "learning_rate": 5.80731289949167e-05, "loss": 1.0222, "step": 167675 }, { "epoch": 0.42, "learning_rate": 5.8071870753435e-05, "loss": 1.0191, "step": 167680 }, { "epoch": 0.42, "learning_rate": 5.807061251195329e-05, "loss": 1.0194, "step": 167685 }, { "epoch": 0.42, "learning_rate": 5.806935427047159e-05, "loss": 1.0211, "step": 167690 }, { "epoch": 0.42, "learning_rate": 5.806809602898988e-05, "loss": 1.0175, "step": 167695 }, { "epoch": 0.42, "learning_rate": 5.806683778750818e-05, "loss": 1.0206, "step": 167700 }, { "epoch": 0.42, "learning_rate": 5.806557954602647e-05, "loss": 1.0192, "step": 167705 }, { "epoch": 0.42, "learning_rate": 5.806432130454476e-05, "loss": 1.019, "step": 167710 }, { "epoch": 0.42, "learning_rate": 5.806306306306306e-05, "loss": 1.0237, "step": 167715 }, { "epoch": 0.42, "learning_rate": 5.806180482158135e-05, "loss": 1.0192, "step": 167720 }, { "epoch": 0.42, "learning_rate": 5.806054658009965e-05, "loss": 1.0224, "step": 167725 }, { "epoch": 0.42, "learning_rate": 5.805928833861794e-05, "loss": 1.0215, "step": 167730 }, { "epoch": 0.42, "learning_rate": 5.805803009713624e-05, "loss": 1.0181, "step": 167735 }, { "epoch": 0.42, "learning_rate": 5.805677185565453e-05, "loss": 1.0194, "step": 167740 }, { "epoch": 0.42, "learning_rate": 5.805551361417283e-05, "loss": 1.0197, "step": 167745 }, { "epoch": 0.42, "learning_rate": 5.805425537269112e-05, "loss": 1.0202, "step": 167750 }, { "epoch": 0.42, "learning_rate": 5.805299713120942e-05, "loss": 1.0181, "step": 167755 }, { "epoch": 0.42, "learning_rate": 5.805173888972773e-05, "loss": 1.0175, "step": 167760 }, { "epoch": 0.42, "learning_rate": 5.805048064824602e-05, "loss": 1.0189, "step": 167765 }, { "epoch": 0.42, "learning_rate": 5.804922240676432e-05, "loss": 1.0186, "step": 167770 }, { "epoch": 0.42, "learning_rate": 5.804796416528261e-05, "loss": 1.0193, "step": 167775 }, { "epoch": 0.42, "learning_rate": 5.804670592380091e-05, "loss": 1.0205, "step": 167780 }, { "epoch": 0.42, "learning_rate": 5.80454476823192e-05, "loss": 1.0185, "step": 167785 }, { "epoch": 0.42, "learning_rate": 5.804418944083749e-05, "loss": 1.0192, "step": 167790 }, { "epoch": 0.42, "learning_rate": 5.804293119935579e-05, "loss": 1.019, "step": 167795 }, { "epoch": 0.42, "learning_rate": 5.804167295787408e-05, "loss": 1.0199, "step": 167800 }, { "epoch": 0.42, "learning_rate": 5.804041471639238e-05, "loss": 1.0192, "step": 167805 }, { "epoch": 0.42, "learning_rate": 5.803915647491067e-05, "loss": 1.0195, "step": 167810 }, { "epoch": 0.42, "learning_rate": 5.803789823342897e-05, "loss": 1.0215, "step": 167815 }, { "epoch": 0.42, "learning_rate": 5.803663999194726e-05, "loss": 1.0187, "step": 167820 }, { "epoch": 0.42, "learning_rate": 5.803538175046556e-05, "loss": 1.0181, "step": 167825 }, { "epoch": 0.42, "learning_rate": 5.803412350898385e-05, "loss": 1.0194, "step": 167830 }, { "epoch": 0.42, "learning_rate": 5.803286526750215e-05, "loss": 1.0213, "step": 167835 }, { "epoch": 0.42, "learning_rate": 5.803160702602044e-05, "loss": 1.0178, "step": 167840 }, { "epoch": 0.42, "learning_rate": 5.803034878453874e-05, "loss": 1.0203, "step": 167845 }, { "epoch": 0.42, "learning_rate": 5.802909054305703e-05, "loss": 1.0179, "step": 167850 }, { "epoch": 0.42, "learning_rate": 5.802783230157532e-05, "loss": 1.0187, "step": 167855 }, { "epoch": 0.42, "learning_rate": 5.802657406009362e-05, "loss": 1.0188, "step": 167860 }, { "epoch": 0.42, "learning_rate": 5.802531581861191e-05, "loss": 1.0433, "step": 167865 }, { "epoch": 0.42, "learning_rate": 5.802405757713021e-05, "loss": 1.0193, "step": 167870 }, { "epoch": 0.42, "learning_rate": 5.80227993356485e-05, "loss": 1.0207, "step": 167875 }, { "epoch": 0.42, "learning_rate": 5.80215410941668e-05, "loss": 1.0193, "step": 167880 }, { "epoch": 0.42, "learning_rate": 5.802028285268509e-05, "loss": 1.0172, "step": 167885 }, { "epoch": 0.42, "learning_rate": 5.801902461120339e-05, "loss": 1.0207, "step": 167890 }, { "epoch": 0.42, "learning_rate": 5.801776636972168e-05, "loss": 1.0191, "step": 167895 }, { "epoch": 0.42, "learning_rate": 5.801650812823998e-05, "loss": 1.0202, "step": 167900 }, { "epoch": 0.42, "learning_rate": 5.801524988675827e-05, "loss": 1.0187, "step": 167905 }, { "epoch": 0.42, "learning_rate": 5.801399164527657e-05, "loss": 1.0203, "step": 167910 }, { "epoch": 0.42, "learning_rate": 5.801273340379486e-05, "loss": 1.0193, "step": 167915 }, { "epoch": 0.42, "learning_rate": 5.801147516231315e-05, "loss": 1.0166, "step": 167920 }, { "epoch": 0.42, "learning_rate": 5.801021692083145e-05, "loss": 1.0223, "step": 167925 }, { "epoch": 0.42, "learning_rate": 5.800895867934974e-05, "loss": 1.0201, "step": 167930 }, { "epoch": 0.42, "learning_rate": 5.800770043786804e-05, "loss": 1.0193, "step": 167935 }, { "epoch": 0.42, "learning_rate": 5.800644219638633e-05, "loss": 1.0202, "step": 167940 }, { "epoch": 0.42, "learning_rate": 5.800518395490463e-05, "loss": 1.0232, "step": 167945 }, { "epoch": 0.42, "learning_rate": 5.800392571342292e-05, "loss": 1.0198, "step": 167950 }, { "epoch": 0.42, "learning_rate": 5.800266747194122e-05, "loss": 1.0186, "step": 167955 }, { "epoch": 0.42, "learning_rate": 5.800140923045951e-05, "loss": 1.0212, "step": 167960 }, { "epoch": 0.42, "learning_rate": 5.800015098897781e-05, "loss": 1.022, "step": 167965 }, { "epoch": 0.42, "learning_rate": 5.79988927474961e-05, "loss": 1.0205, "step": 167970 }, { "epoch": 0.42, "learning_rate": 5.79976345060144e-05, "loss": 1.0187, "step": 167975 }, { "epoch": 0.42, "learning_rate": 5.799637626453269e-05, "loss": 1.0189, "step": 167980 }, { "epoch": 0.42, "learning_rate": 5.7995118023050984e-05, "loss": 1.0212, "step": 167985 }, { "epoch": 0.42, "learning_rate": 5.799385978156928e-05, "loss": 1.0177, "step": 167990 }, { "epoch": 0.42, "learning_rate": 5.7992601540087574e-05, "loss": 1.0205, "step": 167995 }, { "epoch": 0.42, "learning_rate": 5.799134329860587e-05, "loss": 1.0195, "step": 168000 }, { "epoch": 0.42, "learning_rate": 5.7990085057124164e-05, "loss": 1.0175, "step": 168005 }, { "epoch": 0.42, "learning_rate": 5.798882681564246e-05, "loss": 1.0228, "step": 168010 }, { "epoch": 0.42, "learning_rate": 5.7987568574160754e-05, "loss": 1.0188, "step": 168015 }, { "epoch": 0.42, "learning_rate": 5.798631033267905e-05, "loss": 1.0206, "step": 168020 }, { "epoch": 0.42, "learning_rate": 5.7985052091197344e-05, "loss": 1.0178, "step": 168025 }, { "epoch": 0.42, "learning_rate": 5.798379384971564e-05, "loss": 1.0189, "step": 168030 }, { "epoch": 0.42, "learning_rate": 5.7982535608233934e-05, "loss": 1.017, "step": 168035 }, { "epoch": 0.42, "learning_rate": 5.798127736675223e-05, "loss": 1.0195, "step": 168040 }, { "epoch": 0.42, "learning_rate": 5.7980019125270524e-05, "loss": 1.0185, "step": 168045 }, { "epoch": 0.42, "learning_rate": 5.7978760883788816e-05, "loss": 1.0206, "step": 168050 }, { "epoch": 0.42, "learning_rate": 5.7977502642307114e-05, "loss": 1.0213, "step": 168055 }, { "epoch": 0.42, "learning_rate": 5.7976244400825406e-05, "loss": 1.0188, "step": 168060 }, { "epoch": 0.42, "learning_rate": 5.7974986159343704e-05, "loss": 1.0225, "step": 168065 }, { "epoch": 0.42, "learning_rate": 5.7973727917861996e-05, "loss": 1.0189, "step": 168070 }, { "epoch": 0.42, "learning_rate": 5.7972469676380294e-05, "loss": 1.019, "step": 168075 }, { "epoch": 0.42, "learning_rate": 5.7971211434898586e-05, "loss": 1.021, "step": 168080 }, { "epoch": 0.42, "learning_rate": 5.7969953193416884e-05, "loss": 1.0202, "step": 168085 }, { "epoch": 0.42, "learning_rate": 5.7968694951935176e-05, "loss": 1.0201, "step": 168090 }, { "epoch": 0.42, "learning_rate": 5.7967436710453474e-05, "loss": 1.02, "step": 168095 }, { "epoch": 0.42, "learning_rate": 5.7966178468971765e-05, "loss": 1.0193, "step": 168100 }, { "epoch": 0.42, "learning_rate": 5.796492022749006e-05, "loss": 1.0191, "step": 168105 }, { "epoch": 0.42, "learning_rate": 5.7963661986008355e-05, "loss": 1.019, "step": 168110 }, { "epoch": 0.42, "learning_rate": 5.796240374452665e-05, "loss": 1.0187, "step": 168115 }, { "epoch": 0.42, "learning_rate": 5.7961145503044945e-05, "loss": 1.0215, "step": 168120 }, { "epoch": 0.42, "learning_rate": 5.795988726156324e-05, "loss": 1.0181, "step": 168125 }, { "epoch": 0.42, "learning_rate": 5.7958629020081535e-05, "loss": 1.0203, "step": 168130 }, { "epoch": 0.42, "learning_rate": 5.795737077859983e-05, "loss": 1.019, "step": 168135 }, { "epoch": 0.42, "learning_rate": 5.7956112537118125e-05, "loss": 1.0177, "step": 168140 }, { "epoch": 0.42, "learning_rate": 5.795485429563642e-05, "loss": 1.0194, "step": 168145 }, { "epoch": 0.42, "learning_rate": 5.7953596054154715e-05, "loss": 1.022, "step": 168150 }, { "epoch": 0.42, "learning_rate": 5.795233781267301e-05, "loss": 1.0197, "step": 168155 }, { "epoch": 0.42, "learning_rate": 5.7951079571191305e-05, "loss": 1.0201, "step": 168160 }, { "epoch": 0.42, "learning_rate": 5.79498213297096e-05, "loss": 1.0217, "step": 168165 }, { "epoch": 0.42, "learning_rate": 5.794856308822789e-05, "loss": 1.0215, "step": 168170 }, { "epoch": 0.42, "learning_rate": 5.794730484674619e-05, "loss": 1.0196, "step": 168175 }, { "epoch": 0.42, "learning_rate": 5.794604660526448e-05, "loss": 1.02, "step": 168180 }, { "epoch": 0.42, "learning_rate": 5.794478836378278e-05, "loss": 1.0181, "step": 168185 }, { "epoch": 0.42, "learning_rate": 5.794353012230107e-05, "loss": 1.0202, "step": 168190 }, { "epoch": 0.42, "learning_rate": 5.794227188081937e-05, "loss": 1.021, "step": 168195 }, { "epoch": 0.42, "learning_rate": 5.794101363933766e-05, "loss": 1.0173, "step": 168200 }, { "epoch": 0.42, "learning_rate": 5.793975539785596e-05, "loss": 1.0196, "step": 168205 }, { "epoch": 0.42, "learning_rate": 5.793849715637425e-05, "loss": 1.0196, "step": 168210 }, { "epoch": 0.42, "learning_rate": 5.793723891489255e-05, "loss": 1.0185, "step": 168215 }, { "epoch": 0.42, "learning_rate": 5.793598067341084e-05, "loss": 1.0208, "step": 168220 }, { "epoch": 0.42, "learning_rate": 5.793472243192914e-05, "loss": 1.0214, "step": 168225 }, { "epoch": 0.42, "learning_rate": 5.793346419044743e-05, "loss": 1.0203, "step": 168230 }, { "epoch": 0.42, "learning_rate": 5.793220594896572e-05, "loss": 1.0198, "step": 168235 }, { "epoch": 0.42, "learning_rate": 5.793094770748402e-05, "loss": 1.0196, "step": 168240 }, { "epoch": 0.42, "learning_rate": 5.792968946600231e-05, "loss": 1.0188, "step": 168245 }, { "epoch": 0.42, "learning_rate": 5.792843122452061e-05, "loss": 1.0189, "step": 168250 }, { "epoch": 0.42, "learning_rate": 5.79271729830389e-05, "loss": 1.018, "step": 168255 }, { "epoch": 0.42, "learning_rate": 5.7925914741557205e-05, "loss": 1.0218, "step": 168260 }, { "epoch": 0.42, "learning_rate": 5.7924656500075503e-05, "loss": 1.0197, "step": 168265 }, { "epoch": 0.42, "learning_rate": 5.7923398258593795e-05, "loss": 1.0189, "step": 168270 }, { "epoch": 0.42, "learning_rate": 5.792214001711209e-05, "loss": 1.0191, "step": 168275 }, { "epoch": 0.42, "learning_rate": 5.7920881775630385e-05, "loss": 1.0182, "step": 168280 }, { "epoch": 0.42, "learning_rate": 5.791962353414868e-05, "loss": 1.0195, "step": 168285 }, { "epoch": 0.42, "learning_rate": 5.7918365292666975e-05, "loss": 1.0198, "step": 168290 }, { "epoch": 0.42, "learning_rate": 5.791710705118527e-05, "loss": 1.0191, "step": 168295 }, { "epoch": 0.42, "learning_rate": 5.7915848809703565e-05, "loss": 1.0198, "step": 168300 }, { "epoch": 0.42, "learning_rate": 5.791459056822186e-05, "loss": 1.0206, "step": 168305 }, { "epoch": 0.42, "learning_rate": 5.7913332326740155e-05, "loss": 1.0232, "step": 168310 }, { "epoch": 0.42, "learning_rate": 5.7912074085258446e-05, "loss": 1.0212, "step": 168315 }, { "epoch": 0.42, "learning_rate": 5.7910815843776745e-05, "loss": 1.0173, "step": 168320 }, { "epoch": 0.42, "learning_rate": 5.7909557602295036e-05, "loss": 1.0212, "step": 168325 }, { "epoch": 0.42, "learning_rate": 5.7908299360813335e-05, "loss": 1.0198, "step": 168330 }, { "epoch": 0.42, "learning_rate": 5.7907041119331626e-05, "loss": 1.0203, "step": 168335 }, { "epoch": 0.42, "learning_rate": 5.7905782877849925e-05, "loss": 1.0198, "step": 168340 }, { "epoch": 0.42, "learning_rate": 5.7904524636368216e-05, "loss": 1.0176, "step": 168345 }, { "epoch": 0.42, "learning_rate": 5.7903266394886515e-05, "loss": 1.0192, "step": 168350 }, { "epoch": 0.42, "learning_rate": 5.7902008153404806e-05, "loss": 1.0188, "step": 168355 }, { "epoch": 0.42, "learning_rate": 5.7900749911923105e-05, "loss": 1.0201, "step": 168360 }, { "epoch": 0.42, "learning_rate": 5.7899491670441396e-05, "loss": 1.0193, "step": 168365 }, { "epoch": 0.42, "learning_rate": 5.7898233428959695e-05, "loss": 1.02, "step": 168370 }, { "epoch": 0.42, "learning_rate": 5.7896975187477986e-05, "loss": 1.0198, "step": 168375 }, { "epoch": 0.42, "learning_rate": 5.789571694599628e-05, "loss": 1.0187, "step": 168380 }, { "epoch": 0.42, "learning_rate": 5.7894458704514576e-05, "loss": 1.02, "step": 168385 }, { "epoch": 0.42, "learning_rate": 5.789320046303287e-05, "loss": 1.0168, "step": 168390 }, { "epoch": 0.42, "learning_rate": 5.7891942221551166e-05, "loss": 1.0211, "step": 168395 }, { "epoch": 0.42, "learning_rate": 5.789068398006946e-05, "loss": 1.0205, "step": 168400 }, { "epoch": 0.42, "learning_rate": 5.7889425738587756e-05, "loss": 1.0194, "step": 168405 }, { "epoch": 0.42, "learning_rate": 5.788816749710605e-05, "loss": 1.0196, "step": 168410 }, { "epoch": 0.42, "learning_rate": 5.7886909255624346e-05, "loss": 1.02, "step": 168415 }, { "epoch": 0.42, "learning_rate": 5.788565101414264e-05, "loss": 1.0192, "step": 168420 }, { "epoch": 0.42, "learning_rate": 5.7884392772660936e-05, "loss": 1.0201, "step": 168425 }, { "epoch": 0.42, "learning_rate": 5.788313453117923e-05, "loss": 1.0177, "step": 168430 }, { "epoch": 0.42, "learning_rate": 5.7881876289697526e-05, "loss": 1.0187, "step": 168435 }, { "epoch": 0.42, "learning_rate": 5.788061804821582e-05, "loss": 1.0202, "step": 168440 }, { "epoch": 0.42, "learning_rate": 5.787935980673411e-05, "loss": 1.0211, "step": 168445 }, { "epoch": 0.42, "learning_rate": 5.787810156525241e-05, "loss": 1.0206, "step": 168450 }, { "epoch": 0.42, "learning_rate": 5.78768433237707e-05, "loss": 1.0202, "step": 168455 }, { "epoch": 0.42, "learning_rate": 5.7875585082289e-05, "loss": 1.0205, "step": 168460 }, { "epoch": 0.42, "learning_rate": 5.787432684080729e-05, "loss": 1.0198, "step": 168465 }, { "epoch": 0.42, "learning_rate": 5.787306859932559e-05, "loss": 1.0172, "step": 168470 }, { "epoch": 0.42, "learning_rate": 5.787181035784388e-05, "loss": 1.0207, "step": 168475 }, { "epoch": 0.42, "learning_rate": 5.787055211636218e-05, "loss": 1.0187, "step": 168480 }, { "epoch": 0.42, "learning_rate": 5.786929387488047e-05, "loss": 1.0181, "step": 168485 }, { "epoch": 0.42, "learning_rate": 5.786803563339877e-05, "loss": 1.0389, "step": 168490 }, { "epoch": 0.42, "learning_rate": 5.786677739191706e-05, "loss": 1.019, "step": 168495 }, { "epoch": 0.42, "learning_rate": 5.786551915043536e-05, "loss": 1.0189, "step": 168500 }, { "epoch": 0.42, "learning_rate": 5.786426090895365e-05, "loss": 1.0188, "step": 168505 }, { "epoch": 0.42, "learning_rate": 5.786300266747194e-05, "loss": 1.023, "step": 168510 }, { "epoch": 0.42, "learning_rate": 5.786174442599024e-05, "loss": 1.0198, "step": 168515 }, { "epoch": 0.42, "learning_rate": 5.786048618450853e-05, "loss": 1.0195, "step": 168520 }, { "epoch": 0.42, "learning_rate": 5.785922794302683e-05, "loss": 1.018, "step": 168525 }, { "epoch": 0.42, "learning_rate": 5.785796970154512e-05, "loss": 1.0192, "step": 168530 }, { "epoch": 0.42, "learning_rate": 5.785671146006342e-05, "loss": 1.0198, "step": 168535 }, { "epoch": 0.42, "learning_rate": 5.785545321858171e-05, "loss": 1.0222, "step": 168540 }, { "epoch": 0.42, "learning_rate": 5.785419497710001e-05, "loss": 1.0345, "step": 168545 }, { "epoch": 0.42, "learning_rate": 5.78529367356183e-05, "loss": 1.0216, "step": 168550 }, { "epoch": 0.42, "learning_rate": 5.78516784941366e-05, "loss": 1.0206, "step": 168555 }, { "epoch": 0.42, "learning_rate": 5.785042025265489e-05, "loss": 1.0185, "step": 168560 }, { "epoch": 0.42, "learning_rate": 5.784916201117319e-05, "loss": 1.0206, "step": 168565 }, { "epoch": 0.42, "learning_rate": 5.784790376969148e-05, "loss": 1.0219, "step": 168570 }, { "epoch": 0.42, "learning_rate": 5.784664552820977e-05, "loss": 1.0173, "step": 168575 }, { "epoch": 0.42, "learning_rate": 5.784538728672807e-05, "loss": 1.018, "step": 168580 }, { "epoch": 0.42, "learning_rate": 5.784412904524636e-05, "loss": 1.0209, "step": 168585 }, { "epoch": 0.42, "learning_rate": 5.784287080376466e-05, "loss": 1.0185, "step": 168590 }, { "epoch": 0.42, "learning_rate": 5.784161256228295e-05, "loss": 1.017, "step": 168595 }, { "epoch": 0.42, "learning_rate": 5.784035432080125e-05, "loss": 1.0274, "step": 168600 }, { "epoch": 0.42, "learning_rate": 5.783909607931954e-05, "loss": 1.0191, "step": 168605 }, { "epoch": 0.42, "learning_rate": 5.783783783783784e-05, "loss": 1.0191, "step": 168610 }, { "epoch": 0.42, "learning_rate": 5.783657959635613e-05, "loss": 1.0205, "step": 168615 }, { "epoch": 0.42, "learning_rate": 5.783532135487443e-05, "loss": 1.0196, "step": 168620 }, { "epoch": 0.42, "learning_rate": 5.783406311339272e-05, "loss": 1.018, "step": 168625 }, { "epoch": 0.42, "learning_rate": 5.783280487191102e-05, "loss": 1.0185, "step": 168630 }, { "epoch": 0.42, "learning_rate": 5.783154663042931e-05, "loss": 1.0222, "step": 168635 }, { "epoch": 0.42, "learning_rate": 5.7830288388947604e-05, "loss": 1.0182, "step": 168640 }, { "epoch": 0.42, "learning_rate": 5.78290301474659e-05, "loss": 1.0173, "step": 168645 }, { "epoch": 0.42, "learning_rate": 5.7827771905984194e-05, "loss": 1.0186, "step": 168650 }, { "epoch": 0.42, "learning_rate": 5.782651366450249e-05, "loss": 1.0191, "step": 168655 }, { "epoch": 0.42, "learning_rate": 5.7825255423020784e-05, "loss": 1.0198, "step": 168660 }, { "epoch": 0.42, "learning_rate": 5.782399718153908e-05, "loss": 1.0198, "step": 168665 }, { "epoch": 0.42, "learning_rate": 5.7822738940057374e-05, "loss": 1.0195, "step": 168670 }, { "epoch": 0.42, "learning_rate": 5.782148069857567e-05, "loss": 1.019, "step": 168675 }, { "epoch": 0.42, "learning_rate": 5.7820222457093963e-05, "loss": 1.0199, "step": 168680 }, { "epoch": 0.42, "learning_rate": 5.781896421561226e-05, "loss": 1.0201, "step": 168685 }, { "epoch": 0.42, "learning_rate": 5.7817705974130553e-05, "loss": 1.0207, "step": 168690 }, { "epoch": 0.42, "learning_rate": 5.781644773264885e-05, "loss": 1.0205, "step": 168695 }, { "epoch": 0.42, "learning_rate": 5.7815189491167143e-05, "loss": 1.0207, "step": 168700 }, { "epoch": 0.42, "learning_rate": 5.7813931249685435e-05, "loss": 1.0177, "step": 168705 }, { "epoch": 0.42, "learning_rate": 5.7812673008203733e-05, "loss": 1.0207, "step": 168710 }, { "epoch": 0.42, "learning_rate": 5.7811414766722025e-05, "loss": 1.021, "step": 168715 }, { "epoch": 0.42, "learning_rate": 5.781015652524032e-05, "loss": 1.0203, "step": 168720 }, { "epoch": 0.42, "learning_rate": 5.7808898283758615e-05, "loss": 1.0188, "step": 168725 }, { "epoch": 0.42, "learning_rate": 5.780764004227691e-05, "loss": 1.0212, "step": 168730 }, { "epoch": 0.42, "learning_rate": 5.7806381800795205e-05, "loss": 1.0188, "step": 168735 }, { "epoch": 0.42, "learning_rate": 5.78051235593135e-05, "loss": 1.021, "step": 168740 }, { "epoch": 0.42, "learning_rate": 5.7803865317831795e-05, "loss": 1.0219, "step": 168745 }, { "epoch": 0.42, "learning_rate": 5.780260707635009e-05, "loss": 1.0206, "step": 168750 }, { "epoch": 0.42, "learning_rate": 5.78013488348684e-05, "loss": 1.0206, "step": 168755 }, { "epoch": 0.42, "learning_rate": 5.780009059338669e-05, "loss": 1.0195, "step": 168760 }, { "epoch": 0.42, "learning_rate": 5.779883235190499e-05, "loss": 1.0205, "step": 168765 }, { "epoch": 0.42, "learning_rate": 5.779757411042328e-05, "loss": 1.0202, "step": 168770 }, { "epoch": 0.42, "learning_rate": 5.779631586894158e-05, "loss": 1.0188, "step": 168775 }, { "epoch": 0.42, "learning_rate": 5.779505762745987e-05, "loss": 1.0233, "step": 168780 }, { "epoch": 0.42, "learning_rate": 5.779379938597816e-05, "loss": 1.0195, "step": 168785 }, { "epoch": 0.42, "learning_rate": 5.779254114449646e-05, "loss": 1.0181, "step": 168790 }, { "epoch": 0.42, "learning_rate": 5.779128290301475e-05, "loss": 1.0208, "step": 168795 }, { "epoch": 0.42, "learning_rate": 5.779002466153305e-05, "loss": 1.0205, "step": 168800 }, { "epoch": 0.42, "learning_rate": 5.778876642005134e-05, "loss": 1.0213, "step": 168805 }, { "epoch": 0.42, "learning_rate": 5.778750817856964e-05, "loss": 1.0218, "step": 168810 }, { "epoch": 0.42, "learning_rate": 5.778624993708793e-05, "loss": 1.0221, "step": 168815 }, { "epoch": 0.42, "learning_rate": 5.778499169560623e-05, "loss": 1.0208, "step": 168820 }, { "epoch": 0.42, "learning_rate": 5.778373345412452e-05, "loss": 1.0202, "step": 168825 }, { "epoch": 0.42, "learning_rate": 5.778247521264282e-05, "loss": 1.0182, "step": 168830 }, { "epoch": 0.42, "learning_rate": 5.778121697116111e-05, "loss": 1.0196, "step": 168835 }, { "epoch": 0.42, "learning_rate": 5.777995872967941e-05, "loss": 1.0196, "step": 168840 }, { "epoch": 0.42, "learning_rate": 5.77787004881977e-05, "loss": 1.0258, "step": 168845 }, { "epoch": 0.42, "learning_rate": 5.777744224671599e-05, "loss": 1.0173, "step": 168850 }, { "epoch": 0.42, "learning_rate": 5.777618400523429e-05, "loss": 1.017, "step": 168855 }, { "epoch": 0.42, "learning_rate": 5.777492576375258e-05, "loss": 1.0183, "step": 168860 }, { "epoch": 0.42, "learning_rate": 5.777366752227088e-05, "loss": 1.0205, "step": 168865 }, { "epoch": 0.42, "learning_rate": 5.777240928078917e-05, "loss": 1.0208, "step": 168870 }, { "epoch": 0.42, "learning_rate": 5.777115103930747e-05, "loss": 1.0187, "step": 168875 }, { "epoch": 0.42, "learning_rate": 5.776989279782576e-05, "loss": 1.0202, "step": 168880 }, { "epoch": 0.42, "learning_rate": 5.776863455634406e-05, "loss": 1.0215, "step": 168885 }, { "epoch": 0.42, "learning_rate": 5.776737631486235e-05, "loss": 1.0195, "step": 168890 }, { "epoch": 0.42, "learning_rate": 5.776611807338065e-05, "loss": 1.0213, "step": 168895 }, { "epoch": 0.42, "learning_rate": 5.776485983189894e-05, "loss": 1.0188, "step": 168900 }, { "epoch": 0.42, "learning_rate": 5.776360159041724e-05, "loss": 1.0198, "step": 168905 }, { "epoch": 0.42, "learning_rate": 5.776234334893553e-05, "loss": 1.0185, "step": 168910 }, { "epoch": 0.42, "learning_rate": 5.7761085107453824e-05, "loss": 1.0185, "step": 168915 }, { "epoch": 0.42, "learning_rate": 5.775982686597212e-05, "loss": 1.0204, "step": 168920 }, { "epoch": 0.42, "learning_rate": 5.7758568624490414e-05, "loss": 1.0195, "step": 168925 }, { "epoch": 0.42, "learning_rate": 5.775731038300871e-05, "loss": 1.0181, "step": 168930 }, { "epoch": 0.42, "learning_rate": 5.7756052141527004e-05, "loss": 1.0186, "step": 168935 }, { "epoch": 0.42, "learning_rate": 5.77547939000453e-05, "loss": 1.0171, "step": 168940 }, { "epoch": 0.42, "learning_rate": 5.7753535658563594e-05, "loss": 1.0202, "step": 168945 }, { "epoch": 0.42, "learning_rate": 5.775227741708189e-05, "loss": 1.0206, "step": 168950 }, { "epoch": 0.42, "learning_rate": 5.7751019175600184e-05, "loss": 1.0205, "step": 168955 }, { "epoch": 0.42, "learning_rate": 5.774976093411848e-05, "loss": 1.0203, "step": 168960 }, { "epoch": 0.42, "learning_rate": 5.7748502692636774e-05, "loss": 1.0202, "step": 168965 }, { "epoch": 0.42, "learning_rate": 5.7747244451155066e-05, "loss": 1.0189, "step": 168970 }, { "epoch": 0.42, "learning_rate": 5.7745986209673364e-05, "loss": 1.0197, "step": 168975 }, { "epoch": 0.42, "learning_rate": 5.7744727968191656e-05, "loss": 1.02, "step": 168980 }, { "epoch": 0.42, "learning_rate": 5.7743469726709954e-05, "loss": 1.0207, "step": 168985 }, { "epoch": 0.42, "learning_rate": 5.7742211485228246e-05, "loss": 1.0187, "step": 168990 }, { "epoch": 0.42, "learning_rate": 5.7740953243746544e-05, "loss": 1.0171, "step": 168995 }, { "epoch": 0.42, "learning_rate": 5.7739695002264836e-05, "loss": 1.0201, "step": 169000 }, { "epoch": 0.42, "learning_rate": 5.7738436760783134e-05, "loss": 1.0204, "step": 169005 }, { "epoch": 0.42, "learning_rate": 5.7737178519301426e-05, "loss": 1.0179, "step": 169010 }, { "epoch": 0.42, "learning_rate": 5.7735920277819724e-05, "loss": 1.0215, "step": 169015 }, { "epoch": 0.42, "learning_rate": 5.7734662036338016e-05, "loss": 1.0179, "step": 169020 }, { "epoch": 0.42, "learning_rate": 5.7733403794856314e-05, "loss": 1.0215, "step": 169025 }, { "epoch": 0.42, "learning_rate": 5.7732145553374606e-05, "loss": 1.0195, "step": 169030 }, { "epoch": 0.42, "learning_rate": 5.77308873118929e-05, "loss": 1.0223, "step": 169035 }, { "epoch": 0.42, "learning_rate": 5.7729629070411196e-05, "loss": 1.0199, "step": 169040 }, { "epoch": 0.42, "learning_rate": 5.772837082892949e-05, "loss": 1.0215, "step": 169045 }, { "epoch": 0.42, "learning_rate": 5.7727112587447786e-05, "loss": 1.0181, "step": 169050 }, { "epoch": 0.42, "learning_rate": 5.772585434596608e-05, "loss": 1.0199, "step": 169055 }, { "epoch": 0.42, "learning_rate": 5.7724596104484376e-05, "loss": 1.0189, "step": 169060 }, { "epoch": 0.42, "learning_rate": 5.772333786300267e-05, "loss": 1.0185, "step": 169065 }, { "epoch": 0.42, "learning_rate": 5.7722079621520966e-05, "loss": 1.0196, "step": 169070 }, { "epoch": 0.42, "learning_rate": 5.772082138003926e-05, "loss": 1.0169, "step": 169075 }, { "epoch": 0.42, "learning_rate": 5.7719563138557556e-05, "loss": 1.018, "step": 169080 }, { "epoch": 0.42, "learning_rate": 5.771830489707585e-05, "loss": 1.0186, "step": 169085 }, { "epoch": 0.42, "learning_rate": 5.7717046655594146e-05, "loss": 1.0206, "step": 169090 }, { "epoch": 0.42, "learning_rate": 5.771578841411244e-05, "loss": 1.0209, "step": 169095 }, { "epoch": 0.42, "learning_rate": 5.771453017263073e-05, "loss": 1.0409, "step": 169100 }, { "epoch": 0.42, "learning_rate": 5.771327193114903e-05, "loss": 1.0328, "step": 169105 }, { "epoch": 0.42, "learning_rate": 5.771201368966732e-05, "loss": 1.0192, "step": 169110 }, { "epoch": 0.42, "learning_rate": 5.771075544818562e-05, "loss": 1.0195, "step": 169115 }, { "epoch": 0.42, "learning_rate": 5.770949720670391e-05, "loss": 1.0302, "step": 169120 }, { "epoch": 0.42, "learning_rate": 5.770823896522221e-05, "loss": 1.0188, "step": 169125 }, { "epoch": 0.42, "learning_rate": 5.77069807237405e-05, "loss": 1.0415, "step": 169130 }, { "epoch": 0.42, "learning_rate": 5.77057224822588e-05, "loss": 1.0194, "step": 169135 }, { "epoch": 0.42, "learning_rate": 5.770446424077709e-05, "loss": 1.0229, "step": 169140 }, { "epoch": 0.42, "learning_rate": 5.770320599929539e-05, "loss": 1.0213, "step": 169145 }, { "epoch": 0.42, "learning_rate": 5.770194775781368e-05, "loss": 1.0189, "step": 169150 }, { "epoch": 0.42, "learning_rate": 5.770068951633198e-05, "loss": 1.0227, "step": 169155 }, { "epoch": 0.42, "learning_rate": 5.769943127485027e-05, "loss": 1.0188, "step": 169160 }, { "epoch": 0.42, "learning_rate": 5.769817303336856e-05, "loss": 1.0162, "step": 169165 }, { "epoch": 0.42, "learning_rate": 5.769691479188686e-05, "loss": 1.0201, "step": 169170 }, { "epoch": 0.42, "learning_rate": 5.769565655040515e-05, "loss": 1.0201, "step": 169175 }, { "epoch": 0.42, "learning_rate": 5.769439830892345e-05, "loss": 1.0175, "step": 169180 }, { "epoch": 0.42, "learning_rate": 5.769314006744174e-05, "loss": 1.0206, "step": 169185 }, { "epoch": 0.42, "learning_rate": 5.769188182596004e-05, "loss": 1.0215, "step": 169190 }, { "epoch": 0.42, "learning_rate": 5.769062358447833e-05, "loss": 1.0205, "step": 169195 }, { "epoch": 0.42, "learning_rate": 5.768936534299663e-05, "loss": 1.0224, "step": 169200 }, { "epoch": 0.42, "learning_rate": 5.768810710151492e-05, "loss": 1.0203, "step": 169205 }, { "epoch": 0.42, "learning_rate": 5.768684886003322e-05, "loss": 1.0192, "step": 169210 }, { "epoch": 0.42, "learning_rate": 5.768559061855151e-05, "loss": 1.0199, "step": 169215 }, { "epoch": 0.42, "learning_rate": 5.768433237706981e-05, "loss": 1.0199, "step": 169220 }, { "epoch": 0.42, "learning_rate": 5.76830741355881e-05, "loss": 1.0205, "step": 169225 }, { "epoch": 0.42, "learning_rate": 5.768181589410639e-05, "loss": 1.0178, "step": 169230 }, { "epoch": 0.42, "learning_rate": 5.768055765262469e-05, "loss": 1.0191, "step": 169235 }, { "epoch": 0.42, "learning_rate": 5.767929941114298e-05, "loss": 1.0196, "step": 169240 }, { "epoch": 0.42, "learning_rate": 5.767804116966128e-05, "loss": 1.0229, "step": 169245 }, { "epoch": 0.42, "learning_rate": 5.767678292817957e-05, "loss": 1.019, "step": 169250 }, { "epoch": 0.42, "learning_rate": 5.767552468669788e-05, "loss": 1.021, "step": 169255 }, { "epoch": 0.42, "learning_rate": 5.7674266445216175e-05, "loss": 1.0179, "step": 169260 }, { "epoch": 0.42, "learning_rate": 5.767300820373447e-05, "loss": 1.0208, "step": 169265 }, { "epoch": 0.42, "learning_rate": 5.7671749962252765e-05, "loss": 1.0221, "step": 169270 }, { "epoch": 0.42, "learning_rate": 5.7670491720771057e-05, "loss": 1.0187, "step": 169275 }, { "epoch": 0.42, "learning_rate": 5.7669233479289355e-05, "loss": 1.0194, "step": 169280 }, { "epoch": 0.42, "learning_rate": 5.7667975237807647e-05, "loss": 1.0194, "step": 169285 }, { "epoch": 0.42, "learning_rate": 5.7666716996325945e-05, "loss": 1.0208, "step": 169290 }, { "epoch": 0.42, "learning_rate": 5.7665458754844237e-05, "loss": 1.0211, "step": 169295 }, { "epoch": 0.42, "learning_rate": 5.7664200513362535e-05, "loss": 1.0191, "step": 169300 }, { "epoch": 0.42, "learning_rate": 5.7662942271880827e-05, "loss": 1.0192, "step": 169305 }, { "epoch": 0.42, "learning_rate": 5.766168403039912e-05, "loss": 1.0187, "step": 169310 }, { "epoch": 0.43, "learning_rate": 5.7660425788917416e-05, "loss": 1.0187, "step": 169315 }, { "epoch": 0.43, "learning_rate": 5.765916754743571e-05, "loss": 1.0215, "step": 169320 }, { "epoch": 0.43, "learning_rate": 5.7657909305954006e-05, "loss": 1.0194, "step": 169325 }, { "epoch": 0.43, "learning_rate": 5.76566510644723e-05, "loss": 1.0206, "step": 169330 }, { "epoch": 0.43, "learning_rate": 5.7655392822990596e-05, "loss": 1.0285, "step": 169335 }, { "epoch": 0.43, "learning_rate": 5.765413458150889e-05, "loss": 1.0191, "step": 169340 }, { "epoch": 0.43, "learning_rate": 5.7652876340027186e-05, "loss": 1.0203, "step": 169345 }, { "epoch": 0.43, "learning_rate": 5.765161809854548e-05, "loss": 1.0193, "step": 169350 }, { "epoch": 0.43, "learning_rate": 5.7650359857063776e-05, "loss": 1.0214, "step": 169355 }, { "epoch": 0.43, "learning_rate": 5.764910161558207e-05, "loss": 1.0187, "step": 169360 }, { "epoch": 0.43, "learning_rate": 5.7647843374100366e-05, "loss": 1.0192, "step": 169365 }, { "epoch": 0.43, "learning_rate": 5.764658513261866e-05, "loss": 1.0208, "step": 169370 }, { "epoch": 0.43, "learning_rate": 5.764532689113695e-05, "loss": 1.0199, "step": 169375 }, { "epoch": 0.43, "learning_rate": 5.764406864965525e-05, "loss": 1.0198, "step": 169380 }, { "epoch": 0.43, "learning_rate": 5.764281040817354e-05, "loss": 1.02, "step": 169385 }, { "epoch": 0.43, "learning_rate": 5.764155216669184e-05, "loss": 1.0176, "step": 169390 }, { "epoch": 0.43, "learning_rate": 5.764029392521013e-05, "loss": 1.0187, "step": 169395 }, { "epoch": 0.43, "learning_rate": 5.763903568372843e-05, "loss": 1.0194, "step": 169400 }, { "epoch": 0.43, "learning_rate": 5.763777744224672e-05, "loss": 1.0198, "step": 169405 }, { "epoch": 0.43, "learning_rate": 5.763651920076502e-05, "loss": 1.0198, "step": 169410 }, { "epoch": 0.43, "learning_rate": 5.763526095928331e-05, "loss": 1.0196, "step": 169415 }, { "epoch": 0.43, "learning_rate": 5.763400271780161e-05, "loss": 1.0186, "step": 169420 }, { "epoch": 0.43, "learning_rate": 5.76327444763199e-05, "loss": 1.02, "step": 169425 }, { "epoch": 0.43, "learning_rate": 5.76314862348382e-05, "loss": 1.0194, "step": 169430 }, { "epoch": 0.43, "learning_rate": 5.763022799335649e-05, "loss": 1.0213, "step": 169435 }, { "epoch": 0.43, "learning_rate": 5.762896975187478e-05, "loss": 1.0194, "step": 169440 }, { "epoch": 0.43, "learning_rate": 5.762771151039308e-05, "loss": 1.0187, "step": 169445 }, { "epoch": 0.43, "learning_rate": 5.762645326891137e-05, "loss": 1.0191, "step": 169450 }, { "epoch": 0.43, "learning_rate": 5.762519502742967e-05, "loss": 1.0191, "step": 169455 }, { "epoch": 0.43, "learning_rate": 5.762393678594796e-05, "loss": 1.0192, "step": 169460 }, { "epoch": 0.43, "learning_rate": 5.762267854446626e-05, "loss": 1.0185, "step": 169465 }, { "epoch": 0.43, "learning_rate": 5.762142030298455e-05, "loss": 1.0199, "step": 169470 }, { "epoch": 0.43, "learning_rate": 5.762016206150285e-05, "loss": 1.0207, "step": 169475 }, { "epoch": 0.43, "learning_rate": 5.761890382002114e-05, "loss": 1.0211, "step": 169480 }, { "epoch": 0.43, "learning_rate": 5.761764557853944e-05, "loss": 1.0193, "step": 169485 }, { "epoch": 0.43, "learning_rate": 5.761638733705773e-05, "loss": 1.0219, "step": 169490 }, { "epoch": 0.43, "learning_rate": 5.761512909557603e-05, "loss": 1.021, "step": 169495 }, { "epoch": 0.43, "learning_rate": 5.761387085409432e-05, "loss": 1.0194, "step": 169500 }, { "epoch": 0.43, "learning_rate": 5.761261261261261e-05, "loss": 1.0198, "step": 169505 }, { "epoch": 0.43, "learning_rate": 5.761135437113091e-05, "loss": 1.0205, "step": 169510 }, { "epoch": 0.43, "learning_rate": 5.76100961296492e-05, "loss": 1.0184, "step": 169515 }, { "epoch": 0.43, "learning_rate": 5.76088378881675e-05, "loss": 1.0209, "step": 169520 }, { "epoch": 0.43, "learning_rate": 5.760757964668579e-05, "loss": 1.0196, "step": 169525 }, { "epoch": 0.43, "learning_rate": 5.760632140520409e-05, "loss": 1.0182, "step": 169530 }, { "epoch": 0.43, "learning_rate": 5.760506316372238e-05, "loss": 1.0188, "step": 169535 }, { "epoch": 0.43, "learning_rate": 5.760380492224068e-05, "loss": 1.0206, "step": 169540 }, { "epoch": 0.43, "learning_rate": 5.760254668075897e-05, "loss": 1.0211, "step": 169545 }, { "epoch": 0.43, "learning_rate": 5.760128843927727e-05, "loss": 1.0174, "step": 169550 }, { "epoch": 0.43, "learning_rate": 5.760003019779556e-05, "loss": 1.0194, "step": 169555 }, { "epoch": 0.43, "learning_rate": 5.759877195631386e-05, "loss": 1.0187, "step": 169560 }, { "epoch": 0.43, "learning_rate": 5.759751371483215e-05, "loss": 1.018, "step": 169565 }, { "epoch": 0.43, "learning_rate": 5.7596255473350444e-05, "loss": 1.0189, "step": 169570 }, { "epoch": 0.43, "learning_rate": 5.759499723186874e-05, "loss": 1.0211, "step": 169575 }, { "epoch": 0.43, "learning_rate": 5.7593738990387034e-05, "loss": 1.0204, "step": 169580 }, { "epoch": 0.43, "learning_rate": 5.759248074890533e-05, "loss": 1.021, "step": 169585 }, { "epoch": 0.43, "learning_rate": 5.7591222507423624e-05, "loss": 1.0172, "step": 169590 }, { "epoch": 0.43, "learning_rate": 5.758996426594192e-05, "loss": 1.0207, "step": 169595 }, { "epoch": 0.43, "learning_rate": 5.7588706024460214e-05, "loss": 1.0197, "step": 169600 }, { "epoch": 0.43, "learning_rate": 5.758744778297851e-05, "loss": 1.0179, "step": 169605 }, { "epoch": 0.43, "learning_rate": 5.7586189541496804e-05, "loss": 1.0196, "step": 169610 }, { "epoch": 0.43, "learning_rate": 5.75849313000151e-05, "loss": 1.0203, "step": 169615 }, { "epoch": 0.43, "learning_rate": 5.7583673058533394e-05, "loss": 1.0199, "step": 169620 }, { "epoch": 0.43, "learning_rate": 5.7582414817051685e-05, "loss": 1.0179, "step": 169625 }, { "epoch": 0.43, "learning_rate": 5.7581156575569984e-05, "loss": 1.0183, "step": 169630 }, { "epoch": 0.43, "learning_rate": 5.7579898334088275e-05, "loss": 1.0197, "step": 169635 }, { "epoch": 0.43, "learning_rate": 5.7578640092606574e-05, "loss": 1.0179, "step": 169640 }, { "epoch": 0.43, "learning_rate": 5.7577381851124865e-05, "loss": 1.0182, "step": 169645 }, { "epoch": 0.43, "learning_rate": 5.7576123609643164e-05, "loss": 1.0217, "step": 169650 }, { "epoch": 0.43, "learning_rate": 5.7574865368161455e-05, "loss": 1.0208, "step": 169655 }, { "epoch": 0.43, "learning_rate": 5.7573607126679754e-05, "loss": 1.0357, "step": 169660 }, { "epoch": 0.43, "learning_rate": 5.7572348885198045e-05, "loss": 1.0198, "step": 169665 }, { "epoch": 0.43, "learning_rate": 5.7571090643716344e-05, "loss": 1.0213, "step": 169670 }, { "epoch": 0.43, "learning_rate": 5.7569832402234635e-05, "loss": 1.0217, "step": 169675 }, { "epoch": 0.43, "learning_rate": 5.7568574160752933e-05, "loss": 1.021, "step": 169680 }, { "epoch": 0.43, "learning_rate": 5.7567315919271225e-05, "loss": 1.0216, "step": 169685 }, { "epoch": 0.43, "learning_rate": 5.756605767778952e-05, "loss": 1.0171, "step": 169690 }, { "epoch": 0.43, "learning_rate": 5.7564799436307815e-05, "loss": 1.0203, "step": 169695 }, { "epoch": 0.43, "learning_rate": 5.756354119482611e-05, "loss": 1.0193, "step": 169700 }, { "epoch": 0.43, "learning_rate": 5.7562282953344405e-05, "loss": 1.0194, "step": 169705 }, { "epoch": 0.43, "learning_rate": 5.75610247118627e-05, "loss": 1.0196, "step": 169710 }, { "epoch": 0.43, "learning_rate": 5.7559766470380995e-05, "loss": 1.0427, "step": 169715 }, { "epoch": 0.43, "learning_rate": 5.7558508228899287e-05, "loss": 1.0188, "step": 169720 }, { "epoch": 0.43, "learning_rate": 5.7557249987417585e-05, "loss": 1.0185, "step": 169725 }, { "epoch": 0.43, "learning_rate": 5.7555991745935877e-05, "loss": 1.0207, "step": 169730 }, { "epoch": 0.43, "learning_rate": 5.7554733504454175e-05, "loss": 1.0393, "step": 169735 }, { "epoch": 0.43, "learning_rate": 5.7553475262972467e-05, "loss": 1.0193, "step": 169740 }, { "epoch": 0.43, "learning_rate": 5.7552217021490765e-05, "loss": 1.0177, "step": 169745 }, { "epoch": 0.43, "learning_rate": 5.7550958780009057e-05, "loss": 1.0194, "step": 169750 }, { "epoch": 0.43, "learning_rate": 5.754970053852736e-05, "loss": 1.0213, "step": 169755 }, { "epoch": 0.43, "learning_rate": 5.754844229704566e-05, "loss": 1.0186, "step": 169760 }, { "epoch": 0.43, "learning_rate": 5.754718405556395e-05, "loss": 1.0212, "step": 169765 }, { "epoch": 0.43, "learning_rate": 5.754592581408224e-05, "loss": 1.019, "step": 169770 }, { "epoch": 0.43, "learning_rate": 5.754466757260054e-05, "loss": 1.0194, "step": 169775 }, { "epoch": 0.43, "learning_rate": 5.754340933111883e-05, "loss": 1.0187, "step": 169780 }, { "epoch": 0.43, "learning_rate": 5.754215108963713e-05, "loss": 1.0189, "step": 169785 }, { "epoch": 0.43, "learning_rate": 5.754089284815542e-05, "loss": 1.0206, "step": 169790 }, { "epoch": 0.43, "learning_rate": 5.753963460667372e-05, "loss": 1.0204, "step": 169795 }, { "epoch": 0.43, "learning_rate": 5.753837636519201e-05, "loss": 1.019, "step": 169800 }, { "epoch": 0.43, "learning_rate": 5.753711812371031e-05, "loss": 1.02, "step": 169805 }, { "epoch": 0.43, "learning_rate": 5.75358598822286e-05, "loss": 1.0206, "step": 169810 }, { "epoch": 0.43, "learning_rate": 5.75346016407469e-05, "loss": 1.0185, "step": 169815 }, { "epoch": 0.43, "learning_rate": 5.753334339926519e-05, "loss": 1.0198, "step": 169820 }, { "epoch": 0.43, "learning_rate": 5.753208515778349e-05, "loss": 1.0204, "step": 169825 }, { "epoch": 0.43, "learning_rate": 5.753082691630178e-05, "loss": 1.0189, "step": 169830 }, { "epoch": 0.43, "learning_rate": 5.7529568674820075e-05, "loss": 1.0187, "step": 169835 }, { "epoch": 0.43, "learning_rate": 5.752831043333837e-05, "loss": 1.0202, "step": 169840 }, { "epoch": 0.43, "learning_rate": 5.7527052191856665e-05, "loss": 1.018, "step": 169845 }, { "epoch": 0.43, "learning_rate": 5.752579395037496e-05, "loss": 1.0197, "step": 169850 }, { "epoch": 0.43, "learning_rate": 5.7524535708893255e-05, "loss": 1.0198, "step": 169855 }, { "epoch": 0.43, "learning_rate": 5.752327746741155e-05, "loss": 1.0198, "step": 169860 }, { "epoch": 0.43, "learning_rate": 5.7522019225929845e-05, "loss": 1.0184, "step": 169865 }, { "epoch": 0.43, "learning_rate": 5.752076098444814e-05, "loss": 1.02, "step": 169870 }, { "epoch": 0.43, "learning_rate": 5.7519502742966435e-05, "loss": 1.0202, "step": 169875 }, { "epoch": 0.43, "learning_rate": 5.751824450148473e-05, "loss": 1.0207, "step": 169880 }, { "epoch": 0.43, "learning_rate": 5.7516986260003025e-05, "loss": 1.0395, "step": 169885 }, { "epoch": 0.43, "learning_rate": 5.751572801852132e-05, "loss": 1.0207, "step": 169890 }, { "epoch": 0.43, "learning_rate": 5.7514469777039614e-05, "loss": 1.0178, "step": 169895 }, { "epoch": 0.43, "learning_rate": 5.7513211535557906e-05, "loss": 1.0351, "step": 169900 }, { "epoch": 0.43, "learning_rate": 5.7511953294076204e-05, "loss": 1.0186, "step": 169905 }, { "epoch": 0.43, "learning_rate": 5.7510695052594496e-05, "loss": 1.0346, "step": 169910 }, { "epoch": 0.43, "learning_rate": 5.7509436811112794e-05, "loss": 1.02, "step": 169915 }, { "epoch": 0.43, "learning_rate": 5.7508178569631086e-05, "loss": 1.0176, "step": 169920 }, { "epoch": 0.43, "learning_rate": 5.7506920328149384e-05, "loss": 1.0195, "step": 169925 }, { "epoch": 0.43, "learning_rate": 5.7505662086667676e-05, "loss": 1.0196, "step": 169930 }, { "epoch": 0.43, "learning_rate": 5.7504403845185974e-05, "loss": 1.0195, "step": 169935 }, { "epoch": 0.43, "learning_rate": 5.7503145603704266e-05, "loss": 1.0192, "step": 169940 }, { "epoch": 0.43, "learning_rate": 5.7501887362222564e-05, "loss": 1.0407, "step": 169945 }, { "epoch": 0.43, "learning_rate": 5.7500629120740856e-05, "loss": 1.0174, "step": 169950 }, { "epoch": 0.43, "learning_rate": 5.7499370879259154e-05, "loss": 1.0223, "step": 169955 }, { "epoch": 0.43, "learning_rate": 5.7498112637777446e-05, "loss": 1.0184, "step": 169960 }, { "epoch": 0.43, "learning_rate": 5.749685439629574e-05, "loss": 1.0192, "step": 169965 }, { "epoch": 0.43, "learning_rate": 5.7495596154814036e-05, "loss": 1.0197, "step": 169970 }, { "epoch": 0.43, "learning_rate": 5.749433791333233e-05, "loss": 1.0195, "step": 169975 }, { "epoch": 0.43, "learning_rate": 5.7493079671850626e-05, "loss": 1.0203, "step": 169980 }, { "epoch": 0.43, "learning_rate": 5.749182143036892e-05, "loss": 1.0213, "step": 169985 }, { "epoch": 0.43, "learning_rate": 5.7490563188887216e-05, "loss": 1.0199, "step": 169990 }, { "epoch": 0.43, "learning_rate": 5.748930494740551e-05, "loss": 1.0398, "step": 169995 }, { "epoch": 0.43, "learning_rate": 5.7488046705923806e-05, "loss": 1.0193, "step": 170000 }, { "epoch": 0.43, "learning_rate": 5.74867884644421e-05, "loss": 1.0439, "step": 170005 }, { "epoch": 0.43, "learning_rate": 5.7485530222960396e-05, "loss": 1.0466, "step": 170010 }, { "epoch": 0.43, "learning_rate": 5.748427198147869e-05, "loss": 1.045, "step": 170015 }, { "epoch": 0.43, "learning_rate": 5.7483013739996986e-05, "loss": 1.0455, "step": 170020 }, { "epoch": 0.43, "learning_rate": 5.748175549851528e-05, "loss": 1.0476, "step": 170025 }, { "epoch": 0.43, "learning_rate": 5.748049725703357e-05, "loss": 1.044, "step": 170030 }, { "epoch": 0.43, "learning_rate": 5.747923901555187e-05, "loss": 1.0438, "step": 170035 }, { "epoch": 0.43, "learning_rate": 5.747798077407016e-05, "loss": 1.0472, "step": 170040 }, { "epoch": 0.43, "learning_rate": 5.747672253258846e-05, "loss": 1.0456, "step": 170045 }, { "epoch": 0.43, "learning_rate": 5.747546429110675e-05, "loss": 1.0501, "step": 170050 }, { "epoch": 0.43, "learning_rate": 5.747420604962505e-05, "loss": 1.0441, "step": 170055 }, { "epoch": 0.43, "learning_rate": 5.747294780814334e-05, "loss": 1.0473, "step": 170060 }, { "epoch": 0.43, "learning_rate": 5.747168956666164e-05, "loss": 1.0691, "step": 170065 }, { "epoch": 0.43, "learning_rate": 5.747043132517993e-05, "loss": 1.0451, "step": 170070 }, { "epoch": 0.43, "learning_rate": 5.746917308369823e-05, "loss": 1.0434, "step": 170075 }, { "epoch": 0.43, "learning_rate": 5.746791484221652e-05, "loss": 1.0494, "step": 170080 }, { "epoch": 0.43, "learning_rate": 5.746665660073482e-05, "loss": 1.0463, "step": 170085 }, { "epoch": 0.43, "learning_rate": 5.746539835925311e-05, "loss": 1.0666, "step": 170090 }, { "epoch": 0.43, "learning_rate": 5.74641401177714e-05, "loss": 1.0487, "step": 170095 }, { "epoch": 0.43, "learning_rate": 5.74628818762897e-05, "loss": 1.0459, "step": 170100 }, { "epoch": 0.43, "learning_rate": 5.746162363480799e-05, "loss": 1.0467, "step": 170105 }, { "epoch": 0.43, "learning_rate": 5.746036539332629e-05, "loss": 1.0462, "step": 170110 }, { "epoch": 0.43, "learning_rate": 5.745910715184458e-05, "loss": 1.0465, "step": 170115 }, { "epoch": 0.43, "learning_rate": 5.745784891036288e-05, "loss": 1.0475, "step": 170120 }, { "epoch": 0.43, "learning_rate": 5.745659066888117e-05, "loss": 1.0478, "step": 170125 }, { "epoch": 0.43, "learning_rate": 5.745533242739947e-05, "loss": 1.0469, "step": 170130 }, { "epoch": 0.43, "learning_rate": 5.745407418591776e-05, "loss": 1.0478, "step": 170135 }, { "epoch": 0.43, "learning_rate": 5.745281594443606e-05, "loss": 1.0488, "step": 170140 }, { "epoch": 0.43, "learning_rate": 5.745155770295435e-05, "loss": 1.0469, "step": 170145 }, { "epoch": 0.43, "learning_rate": 5.745029946147265e-05, "loss": 1.0445, "step": 170150 }, { "epoch": 0.43, "learning_rate": 5.744904121999094e-05, "loss": 1.0479, "step": 170155 }, { "epoch": 0.43, "learning_rate": 5.744778297850923e-05, "loss": 1.0476, "step": 170160 }, { "epoch": 0.43, "learning_rate": 5.744652473702753e-05, "loss": 1.0479, "step": 170165 }, { "epoch": 0.43, "learning_rate": 5.744526649554582e-05, "loss": 1.0511, "step": 170170 }, { "epoch": 0.43, "learning_rate": 5.744400825406412e-05, "loss": 1.0458, "step": 170175 }, { "epoch": 0.43, "learning_rate": 5.744275001258241e-05, "loss": 1.047, "step": 170180 }, { "epoch": 0.43, "learning_rate": 5.744149177110071e-05, "loss": 1.0462, "step": 170185 }, { "epoch": 0.43, "learning_rate": 5.7440233529619e-05, "loss": 1.0487, "step": 170190 }, { "epoch": 0.43, "learning_rate": 5.74389752881373e-05, "loss": 1.0613, "step": 170195 }, { "epoch": 0.43, "learning_rate": 5.743771704665559e-05, "loss": 1.0481, "step": 170200 }, { "epoch": 0.43, "learning_rate": 5.743645880517389e-05, "loss": 1.0458, "step": 170205 }, { "epoch": 0.43, "learning_rate": 5.743520056369218e-05, "loss": 1.0482, "step": 170210 }, { "epoch": 0.43, "learning_rate": 5.743394232221048e-05, "loss": 1.0482, "step": 170215 }, { "epoch": 0.43, "learning_rate": 5.743268408072877e-05, "loss": 1.0473, "step": 170220 }, { "epoch": 0.43, "learning_rate": 5.743142583924706e-05, "loss": 1.0473, "step": 170225 }, { "epoch": 0.43, "learning_rate": 5.743016759776536e-05, "loss": 1.0469, "step": 170230 }, { "epoch": 0.43, "learning_rate": 5.742890935628365e-05, "loss": 1.0463, "step": 170235 }, { "epoch": 0.43, "learning_rate": 5.742765111480195e-05, "loss": 1.0459, "step": 170240 }, { "epoch": 0.43, "learning_rate": 5.742639287332024e-05, "loss": 1.0459, "step": 170245 }, { "epoch": 0.43, "learning_rate": 5.742513463183854e-05, "loss": 1.0469, "step": 170250 }, { "epoch": 0.43, "learning_rate": 5.742387639035685e-05, "loss": 1.0481, "step": 170255 }, { "epoch": 0.43, "learning_rate": 5.742261814887514e-05, "loss": 1.0454, "step": 170260 }, { "epoch": 0.43, "learning_rate": 5.742135990739344e-05, "loss": 1.0448, "step": 170265 }, { "epoch": 0.43, "learning_rate": 5.742010166591173e-05, "loss": 1.0477, "step": 170270 }, { "epoch": 0.43, "learning_rate": 5.741884342443003e-05, "loss": 1.0476, "step": 170275 }, { "epoch": 0.43, "learning_rate": 5.741758518294832e-05, "loss": 1.0462, "step": 170280 }, { "epoch": 0.43, "learning_rate": 5.7416326941466617e-05, "loss": 1.0487, "step": 170285 }, { "epoch": 0.43, "learning_rate": 5.741506869998491e-05, "loss": 1.0515, "step": 170290 }, { "epoch": 0.43, "learning_rate": 5.7413810458503207e-05, "loss": 1.0448, "step": 170295 }, { "epoch": 0.43, "learning_rate": 5.74125522170215e-05, "loss": 1.0685, "step": 170300 }, { "epoch": 0.43, "learning_rate": 5.741129397553979e-05, "loss": 1.0458, "step": 170305 }, { "epoch": 0.43, "learning_rate": 5.741003573405809e-05, "loss": 1.0437, "step": 170310 }, { "epoch": 0.43, "learning_rate": 5.740877749257638e-05, "loss": 1.0475, "step": 170315 }, { "epoch": 0.43, "learning_rate": 5.740751925109468e-05, "loss": 1.047, "step": 170320 }, { "epoch": 0.43, "learning_rate": 5.740626100961297e-05, "loss": 1.046, "step": 170325 }, { "epoch": 0.43, "learning_rate": 5.740500276813127e-05, "loss": 1.0458, "step": 170330 }, { "epoch": 0.43, "learning_rate": 5.740374452664956e-05, "loss": 1.0467, "step": 170335 }, { "epoch": 0.43, "learning_rate": 5.740248628516786e-05, "loss": 1.0447, "step": 170340 }, { "epoch": 0.43, "learning_rate": 5.740122804368615e-05, "loss": 1.047, "step": 170345 }, { "epoch": 0.43, "learning_rate": 5.739996980220445e-05, "loss": 1.043, "step": 170350 }, { "epoch": 0.43, "learning_rate": 5.739871156072274e-05, "loss": 1.047, "step": 170355 }, { "epoch": 0.43, "learning_rate": 5.739745331924104e-05, "loss": 1.0444, "step": 170360 }, { "epoch": 0.43, "learning_rate": 5.739619507775933e-05, "loss": 1.0451, "step": 170365 }, { "epoch": 0.43, "learning_rate": 5.739493683627762e-05, "loss": 1.0457, "step": 170370 }, { "epoch": 0.43, "learning_rate": 5.739367859479592e-05, "loss": 1.0457, "step": 170375 }, { "epoch": 0.43, "learning_rate": 5.739242035331421e-05, "loss": 1.0479, "step": 170380 }, { "epoch": 0.43, "learning_rate": 5.739116211183251e-05, "loss": 1.0485, "step": 170385 }, { "epoch": 0.43, "learning_rate": 5.73899038703508e-05, "loss": 1.0454, "step": 170390 }, { "epoch": 0.43, "learning_rate": 5.73886456288691e-05, "loss": 1.0453, "step": 170395 }, { "epoch": 0.43, "learning_rate": 5.738738738738739e-05, "loss": 1.047, "step": 170400 }, { "epoch": 0.43, "learning_rate": 5.738612914590569e-05, "loss": 1.0434, "step": 170405 }, { "epoch": 0.43, "learning_rate": 5.738487090442398e-05, "loss": 1.0442, "step": 170410 }, { "epoch": 0.43, "learning_rate": 5.738361266294228e-05, "loss": 1.0462, "step": 170415 }, { "epoch": 0.43, "learning_rate": 5.738235442146057e-05, "loss": 1.0457, "step": 170420 }, { "epoch": 0.43, "learning_rate": 5.738109617997886e-05, "loss": 1.0457, "step": 170425 }, { "epoch": 0.43, "learning_rate": 5.737983793849716e-05, "loss": 1.0465, "step": 170430 }, { "epoch": 0.43, "learning_rate": 5.737857969701545e-05, "loss": 1.0465, "step": 170435 }, { "epoch": 0.43, "learning_rate": 5.737732145553375e-05, "loss": 1.047, "step": 170440 }, { "epoch": 0.43, "learning_rate": 5.737606321405204e-05, "loss": 1.0478, "step": 170445 }, { "epoch": 0.43, "learning_rate": 5.737480497257034e-05, "loss": 1.0441, "step": 170450 }, { "epoch": 0.43, "learning_rate": 5.737354673108863e-05, "loss": 1.0452, "step": 170455 }, { "epoch": 0.43, "learning_rate": 5.737228848960693e-05, "loss": 1.0447, "step": 170460 }, { "epoch": 0.43, "learning_rate": 5.737103024812522e-05, "loss": 1.0444, "step": 170465 }, { "epoch": 0.43, "learning_rate": 5.736977200664352e-05, "loss": 1.0467, "step": 170470 }, { "epoch": 0.43, "learning_rate": 5.736851376516181e-05, "loss": 1.0469, "step": 170475 }, { "epoch": 0.43, "learning_rate": 5.736725552368011e-05, "loss": 1.0457, "step": 170480 }, { "epoch": 0.43, "learning_rate": 5.73659972821984e-05, "loss": 1.0438, "step": 170485 }, { "epoch": 0.43, "learning_rate": 5.7364739040716694e-05, "loss": 1.0466, "step": 170490 }, { "epoch": 0.43, "learning_rate": 5.736348079923499e-05, "loss": 1.0419, "step": 170495 }, { "epoch": 0.43, "learning_rate": 5.7362222557753284e-05, "loss": 1.0459, "step": 170500 }, { "epoch": 0.43, "learning_rate": 5.736096431627158e-05, "loss": 1.0454, "step": 170505 }, { "epoch": 0.43, "learning_rate": 5.7359706074789874e-05, "loss": 1.0684, "step": 170510 }, { "epoch": 0.43, "learning_rate": 5.735844783330817e-05, "loss": 1.0488, "step": 170515 }, { "epoch": 0.43, "learning_rate": 5.7357189591826464e-05, "loss": 1.0454, "step": 170520 }, { "epoch": 0.43, "learning_rate": 5.735593135034476e-05, "loss": 1.0435, "step": 170525 }, { "epoch": 0.43, "learning_rate": 5.7354673108863054e-05, "loss": 1.0459, "step": 170530 }, { "epoch": 0.43, "learning_rate": 5.735341486738135e-05, "loss": 1.0454, "step": 170535 }, { "epoch": 0.43, "learning_rate": 5.7352156625899644e-05, "loss": 1.0432, "step": 170540 }, { "epoch": 0.43, "learning_rate": 5.735089838441794e-05, "loss": 1.046, "step": 170545 }, { "epoch": 0.43, "learning_rate": 5.7349640142936234e-05, "loss": 1.0685, "step": 170550 }, { "epoch": 0.43, "learning_rate": 5.7348381901454525e-05, "loss": 1.0449, "step": 170555 }, { "epoch": 0.43, "learning_rate": 5.7347123659972824e-05, "loss": 1.0432, "step": 170560 }, { "epoch": 0.43, "learning_rate": 5.7345865418491115e-05, "loss": 1.0429, "step": 170565 }, { "epoch": 0.43, "learning_rate": 5.7344607177009414e-05, "loss": 1.0451, "step": 170570 }, { "epoch": 0.43, "learning_rate": 5.7343348935527705e-05, "loss": 1.0471, "step": 170575 }, { "epoch": 0.43, "learning_rate": 5.7342090694046004e-05, "loss": 1.0461, "step": 170580 }, { "epoch": 0.43, "learning_rate": 5.7340832452564295e-05, "loss": 1.0446, "step": 170585 }, { "epoch": 0.43, "learning_rate": 5.7339574211082594e-05, "loss": 1.0439, "step": 170590 }, { "epoch": 0.43, "learning_rate": 5.7338315969600885e-05, "loss": 1.0456, "step": 170595 }, { "epoch": 0.43, "learning_rate": 5.7337057728119184e-05, "loss": 1.0432, "step": 170600 }, { "epoch": 0.43, "learning_rate": 5.7335799486637475e-05, "loss": 1.0457, "step": 170605 }, { "epoch": 0.43, "learning_rate": 5.7334541245155774e-05, "loss": 1.0675, "step": 170610 }, { "epoch": 0.43, "learning_rate": 5.7333283003674065e-05, "loss": 1.0443, "step": 170615 }, { "epoch": 0.43, "learning_rate": 5.733202476219236e-05, "loss": 1.0445, "step": 170620 }, { "epoch": 0.43, "learning_rate": 5.7330766520710655e-05, "loss": 1.0463, "step": 170625 }, { "epoch": 0.43, "learning_rate": 5.732950827922895e-05, "loss": 1.0487, "step": 170630 }, { "epoch": 0.43, "learning_rate": 5.7328250037747245e-05, "loss": 1.0439, "step": 170635 }, { "epoch": 0.43, "learning_rate": 5.732699179626554e-05, "loss": 1.0445, "step": 170640 }, { "epoch": 0.43, "learning_rate": 5.7325733554783835e-05, "loss": 1.0433, "step": 170645 }, { "epoch": 0.43, "learning_rate": 5.732447531330213e-05, "loss": 1.0436, "step": 170650 }, { "epoch": 0.43, "learning_rate": 5.7323217071820425e-05, "loss": 1.0486, "step": 170655 }, { "epoch": 0.43, "learning_rate": 5.732195883033872e-05, "loss": 1.046, "step": 170660 }, { "epoch": 0.43, "learning_rate": 5.7320700588857015e-05, "loss": 1.0456, "step": 170665 }, { "epoch": 0.43, "learning_rate": 5.731944234737531e-05, "loss": 1.0447, "step": 170670 }, { "epoch": 0.43, "learning_rate": 5.7318184105893605e-05, "loss": 1.046, "step": 170675 }, { "epoch": 0.43, "learning_rate": 5.73169258644119e-05, "loss": 1.0463, "step": 170680 }, { "epoch": 0.43, "learning_rate": 5.731566762293019e-05, "loss": 1.047, "step": 170685 }, { "epoch": 0.43, "learning_rate": 5.731440938144849e-05, "loss": 1.0434, "step": 170690 }, { "epoch": 0.43, "learning_rate": 5.731315113996678e-05, "loss": 1.0465, "step": 170695 }, { "epoch": 0.43, "learning_rate": 5.731189289848508e-05, "loss": 1.0443, "step": 170700 }, { "epoch": 0.43, "learning_rate": 5.731063465700337e-05, "loss": 1.0598, "step": 170705 }, { "epoch": 0.43, "learning_rate": 5.730937641552167e-05, "loss": 1.066, "step": 170710 }, { "epoch": 0.43, "learning_rate": 5.730811817403996e-05, "loss": 1.0454, "step": 170715 }, { "epoch": 0.43, "learning_rate": 5.730685993255826e-05, "loss": 1.0454, "step": 170720 }, { "epoch": 0.43, "learning_rate": 5.730560169107655e-05, "loss": 1.0681, "step": 170725 }, { "epoch": 0.43, "learning_rate": 5.7304343449594847e-05, "loss": 1.0456, "step": 170730 }, { "epoch": 0.43, "learning_rate": 5.730308520811314e-05, "loss": 1.0457, "step": 170735 }, { "epoch": 0.43, "learning_rate": 5.7301826966631437e-05, "loss": 1.0462, "step": 170740 }, { "epoch": 0.43, "learning_rate": 5.730056872514973e-05, "loss": 1.0432, "step": 170745 }, { "epoch": 0.43, "learning_rate": 5.729931048366802e-05, "loss": 1.046, "step": 170750 }, { "epoch": 0.43, "learning_rate": 5.729805224218633e-05, "loss": 1.0447, "step": 170755 }, { "epoch": 0.43, "learning_rate": 5.729679400070462e-05, "loss": 1.0484, "step": 170760 }, { "epoch": 0.43, "learning_rate": 5.7295535759222915e-05, "loss": 1.048, "step": 170765 }, { "epoch": 0.43, "learning_rate": 5.729427751774121e-05, "loss": 1.0459, "step": 170770 }, { "epoch": 0.43, "learning_rate": 5.7293019276259505e-05, "loss": 1.0447, "step": 170775 }, { "epoch": 0.43, "learning_rate": 5.72917610347778e-05, "loss": 1.0447, "step": 170780 }, { "epoch": 0.43, "learning_rate": 5.7290502793296095e-05, "loss": 1.0463, "step": 170785 }, { "epoch": 0.43, "learning_rate": 5.728924455181439e-05, "loss": 1.0492, "step": 170790 }, { "epoch": 0.43, "learning_rate": 5.7287986310332685e-05, "loss": 1.0452, "step": 170795 }, { "epoch": 0.43, "learning_rate": 5.728672806885098e-05, "loss": 1.047, "step": 170800 }, { "epoch": 0.43, "learning_rate": 5.7285469827369275e-05, "loss": 1.0467, "step": 170805 }, { "epoch": 0.43, "learning_rate": 5.728421158588757e-05, "loss": 1.0442, "step": 170810 }, { "epoch": 0.43, "learning_rate": 5.7282953344405865e-05, "loss": 1.0473, "step": 170815 }, { "epoch": 0.43, "learning_rate": 5.728169510292416e-05, "loss": 1.0467, "step": 170820 }, { "epoch": 0.43, "learning_rate": 5.7280436861442455e-05, "loss": 1.042, "step": 170825 }, { "epoch": 0.43, "learning_rate": 5.7279178619960746e-05, "loss": 1.0431, "step": 170830 }, { "epoch": 0.43, "learning_rate": 5.7277920378479045e-05, "loss": 1.0444, "step": 170835 }, { "epoch": 0.43, "learning_rate": 5.7276662136997336e-05, "loss": 1.0701, "step": 170840 }, { "epoch": 0.43, "learning_rate": 5.7275403895515635e-05, "loss": 1.0441, "step": 170845 }, { "epoch": 0.43, "learning_rate": 5.7274145654033926e-05, "loss": 1.0471, "step": 170850 }, { "epoch": 0.43, "learning_rate": 5.7272887412552225e-05, "loss": 1.0458, "step": 170855 }, { "epoch": 0.43, "learning_rate": 5.7271629171070516e-05, "loss": 1.0453, "step": 170860 }, { "epoch": 0.43, "learning_rate": 5.7270370929588815e-05, "loss": 1.0484, "step": 170865 }, { "epoch": 0.43, "learning_rate": 5.7269112688107106e-05, "loss": 1.0443, "step": 170870 }, { "epoch": 0.43, "learning_rate": 5.7267854446625405e-05, "loss": 1.0469, "step": 170875 }, { "epoch": 0.43, "learning_rate": 5.7266596205143696e-05, "loss": 1.046, "step": 170880 }, { "epoch": 0.43, "learning_rate": 5.7265337963661995e-05, "loss": 1.044, "step": 170885 }, { "epoch": 0.43, "learning_rate": 5.7264079722180286e-05, "loss": 1.0481, "step": 170890 }, { "epoch": 0.43, "learning_rate": 5.726282148069858e-05, "loss": 1.0454, "step": 170895 }, { "epoch": 0.43, "learning_rate": 5.7261563239216876e-05, "loss": 1.0451, "step": 170900 }, { "epoch": 0.43, "learning_rate": 5.726030499773517e-05, "loss": 1.0487, "step": 170905 }, { "epoch": 0.43, "learning_rate": 5.7259046756253466e-05, "loss": 1.0476, "step": 170910 }, { "epoch": 0.43, "learning_rate": 5.725778851477176e-05, "loss": 1.0463, "step": 170915 }, { "epoch": 0.43, "learning_rate": 5.7256530273290056e-05, "loss": 1.0443, "step": 170920 }, { "epoch": 0.43, "learning_rate": 5.725527203180835e-05, "loss": 1.0417, "step": 170925 }, { "epoch": 0.43, "learning_rate": 5.7254013790326646e-05, "loss": 1.0436, "step": 170930 }, { "epoch": 0.43, "learning_rate": 5.725275554884494e-05, "loss": 1.0452, "step": 170935 }, { "epoch": 0.43, "learning_rate": 5.7251497307363236e-05, "loss": 1.0474, "step": 170940 }, { "epoch": 0.43, "learning_rate": 5.725023906588153e-05, "loss": 1.0447, "step": 170945 }, { "epoch": 0.43, "learning_rate": 5.7248980824399826e-05, "loss": 1.0471, "step": 170950 }, { "epoch": 0.43, "learning_rate": 5.724772258291812e-05, "loss": 1.0701, "step": 170955 }, { "epoch": 0.43, "learning_rate": 5.724646434143641e-05, "loss": 1.0458, "step": 170960 }, { "epoch": 0.43, "learning_rate": 5.724520609995471e-05, "loss": 1.0438, "step": 170965 }, { "epoch": 0.43, "learning_rate": 5.7243947858473e-05, "loss": 1.0462, "step": 170970 }, { "epoch": 0.43, "learning_rate": 5.72426896169913e-05, "loss": 1.0451, "step": 170975 }, { "epoch": 0.43, "learning_rate": 5.724143137550959e-05, "loss": 1.045, "step": 170980 }, { "epoch": 0.43, "learning_rate": 5.724017313402789e-05, "loss": 1.0418, "step": 170985 }, { "epoch": 0.43, "learning_rate": 5.723891489254618e-05, "loss": 1.0447, "step": 170990 }, { "epoch": 0.43, "learning_rate": 5.723765665106448e-05, "loss": 1.0457, "step": 170995 }, { "epoch": 0.43, "learning_rate": 5.723639840958277e-05, "loss": 1.0444, "step": 171000 }, { "epoch": 0.43, "learning_rate": 5.723514016810107e-05, "loss": 1.0471, "step": 171005 }, { "epoch": 0.43, "learning_rate": 5.723388192661936e-05, "loss": 1.0453, "step": 171010 }, { "epoch": 0.43, "learning_rate": 5.723262368513766e-05, "loss": 1.0429, "step": 171015 }, { "epoch": 0.43, "learning_rate": 5.723136544365595e-05, "loss": 1.0456, "step": 171020 }, { "epoch": 0.43, "learning_rate": 5.723010720217424e-05, "loss": 1.0461, "step": 171025 }, { "epoch": 0.43, "learning_rate": 5.722884896069254e-05, "loss": 1.0461, "step": 171030 }, { "epoch": 0.43, "learning_rate": 5.722759071921083e-05, "loss": 1.0479, "step": 171035 }, { "epoch": 0.43, "learning_rate": 5.722633247772913e-05, "loss": 1.0401, "step": 171040 }, { "epoch": 0.43, "learning_rate": 5.722507423624742e-05, "loss": 1.0475, "step": 171045 }, { "epoch": 0.43, "learning_rate": 5.722381599476572e-05, "loss": 1.0448, "step": 171050 }, { "epoch": 0.43, "learning_rate": 5.722255775328401e-05, "loss": 1.0432, "step": 171055 }, { "epoch": 0.43, "learning_rate": 5.722129951180231e-05, "loss": 1.0466, "step": 171060 }, { "epoch": 0.43, "learning_rate": 5.72200412703206e-05, "loss": 1.0477, "step": 171065 }, { "epoch": 0.43, "learning_rate": 5.72187830288389e-05, "loss": 1.0425, "step": 171070 }, { "epoch": 0.43, "learning_rate": 5.721752478735719e-05, "loss": 1.045, "step": 171075 }, { "epoch": 0.43, "learning_rate": 5.721626654587548e-05, "loss": 1.0458, "step": 171080 }, { "epoch": 0.43, "learning_rate": 5.721500830439378e-05, "loss": 1.0467, "step": 171085 }, { "epoch": 0.43, "learning_rate": 5.721375006291207e-05, "loss": 1.046, "step": 171090 }, { "epoch": 0.43, "learning_rate": 5.721249182143037e-05, "loss": 1.0719, "step": 171095 }, { "epoch": 0.43, "learning_rate": 5.721123357994866e-05, "loss": 1.0457, "step": 171100 }, { "epoch": 0.43, "learning_rate": 5.720997533846696e-05, "loss": 1.0474, "step": 171105 }, { "epoch": 0.43, "learning_rate": 5.720871709698525e-05, "loss": 1.0433, "step": 171110 }, { "epoch": 0.43, "learning_rate": 5.720745885550355e-05, "loss": 1.047, "step": 171115 }, { "epoch": 0.43, "learning_rate": 5.720620061402184e-05, "loss": 1.0459, "step": 171120 }, { "epoch": 0.43, "learning_rate": 5.720494237254014e-05, "loss": 1.0453, "step": 171125 }, { "epoch": 0.43, "learning_rate": 5.720368413105843e-05, "loss": 1.0477, "step": 171130 }, { "epoch": 0.43, "learning_rate": 5.720242588957673e-05, "loss": 1.0443, "step": 171135 }, { "epoch": 0.43, "learning_rate": 5.720116764809502e-05, "loss": 1.0445, "step": 171140 }, { "epoch": 0.43, "learning_rate": 5.7199909406613313e-05, "loss": 1.0443, "step": 171145 }, { "epoch": 0.43, "learning_rate": 5.719865116513161e-05, "loss": 1.0511, "step": 171150 }, { "epoch": 0.43, "learning_rate": 5.7197392923649903e-05, "loss": 1.0472, "step": 171155 }, { "epoch": 0.43, "learning_rate": 5.71961346821682e-05, "loss": 1.0489, "step": 171160 }, { "epoch": 0.43, "learning_rate": 5.7194876440686493e-05, "loss": 1.0461, "step": 171165 }, { "epoch": 0.43, "learning_rate": 5.719361819920479e-05, "loss": 1.0431, "step": 171170 }, { "epoch": 0.43, "learning_rate": 5.719235995772308e-05, "loss": 1.0642, "step": 171175 }, { "epoch": 0.43, "learning_rate": 5.719110171624138e-05, "loss": 1.0638, "step": 171180 }, { "epoch": 0.43, "learning_rate": 5.718984347475967e-05, "loss": 1.0444, "step": 171185 }, { "epoch": 0.43, "learning_rate": 5.718858523327797e-05, "loss": 1.0462, "step": 171190 }, { "epoch": 0.43, "learning_rate": 5.718732699179626e-05, "loss": 1.0442, "step": 171195 }, { "epoch": 0.43, "learning_rate": 5.718606875031456e-05, "loss": 1.0451, "step": 171200 }, { "epoch": 0.43, "learning_rate": 5.718481050883285e-05, "loss": 1.0457, "step": 171205 }, { "epoch": 0.43, "learning_rate": 5.7183552267351145e-05, "loss": 1.0466, "step": 171210 }, { "epoch": 0.43, "learning_rate": 5.718229402586944e-05, "loss": 1.0693, "step": 171215 }, { "epoch": 0.43, "learning_rate": 5.7181035784387735e-05, "loss": 1.0455, "step": 171220 }, { "epoch": 0.43, "learning_rate": 5.717977754290603e-05, "loss": 1.0449, "step": 171225 }, { "epoch": 0.43, "learning_rate": 5.7178519301424325e-05, "loss": 1.0471, "step": 171230 }, { "epoch": 0.43, "learning_rate": 5.717726105994262e-05, "loss": 1.0476, "step": 171235 }, { "epoch": 0.43, "learning_rate": 5.7176002818460915e-05, "loss": 1.0448, "step": 171240 }, { "epoch": 0.43, "learning_rate": 5.717474457697921e-05, "loss": 1.0439, "step": 171245 }, { "epoch": 0.43, "learning_rate": 5.7173486335497505e-05, "loss": 1.0468, "step": 171250 }, { "epoch": 0.43, "learning_rate": 5.717222809401581e-05, "loss": 1.0459, "step": 171255 }, { "epoch": 0.43, "learning_rate": 5.717096985253411e-05, "loss": 1.0469, "step": 171260 }, { "epoch": 0.43, "learning_rate": 5.71697116110524e-05, "loss": 1.0471, "step": 171265 }, { "epoch": 0.43, "learning_rate": 5.71684533695707e-05, "loss": 1.0449, "step": 171270 }, { "epoch": 0.43, "learning_rate": 5.716719512808899e-05, "loss": 1.0489, "step": 171275 }, { "epoch": 0.43, "learning_rate": 5.716593688660729e-05, "loss": 1.045, "step": 171280 }, { "epoch": 0.43, "learning_rate": 5.716467864512558e-05, "loss": 1.0479, "step": 171285 }, { "epoch": 0.43, "learning_rate": 5.716342040364387e-05, "loss": 1.043, "step": 171290 }, { "epoch": 0.43, "learning_rate": 5.716216216216217e-05, "loss": 1.0434, "step": 171295 }, { "epoch": 0.43, "learning_rate": 5.716090392068046e-05, "loss": 1.0443, "step": 171300 }, { "epoch": 0.43, "learning_rate": 5.715964567919876e-05, "loss": 1.0714, "step": 171305 }, { "epoch": 0.43, "learning_rate": 5.715838743771705e-05, "loss": 1.0464, "step": 171310 }, { "epoch": 0.43, "learning_rate": 5.715712919623535e-05, "loss": 1.0452, "step": 171315 }, { "epoch": 0.43, "learning_rate": 5.715587095475364e-05, "loss": 1.0456, "step": 171320 }, { "epoch": 0.43, "learning_rate": 5.715461271327194e-05, "loss": 1.0446, "step": 171325 }, { "epoch": 0.43, "learning_rate": 5.715335447179023e-05, "loss": 1.044, "step": 171330 }, { "epoch": 0.43, "learning_rate": 5.715209623030853e-05, "loss": 1.0444, "step": 171335 }, { "epoch": 0.43, "learning_rate": 5.715083798882682e-05, "loss": 1.0454, "step": 171340 }, { "epoch": 0.43, "learning_rate": 5.714957974734512e-05, "loss": 1.047, "step": 171345 }, { "epoch": 0.43, "learning_rate": 5.714832150586341e-05, "loss": 1.0458, "step": 171350 }, { "epoch": 0.43, "learning_rate": 5.71470632643817e-05, "loss": 1.0492, "step": 171355 }, { "epoch": 0.43, "learning_rate": 5.71458050229e-05, "loss": 1.0445, "step": 171360 }, { "epoch": 0.43, "learning_rate": 5.714454678141829e-05, "loss": 1.0453, "step": 171365 }, { "epoch": 0.43, "learning_rate": 5.714328853993659e-05, "loss": 1.0458, "step": 171370 }, { "epoch": 0.43, "learning_rate": 5.714203029845488e-05, "loss": 1.045, "step": 171375 }, { "epoch": 0.43, "learning_rate": 5.714077205697318e-05, "loss": 1.0444, "step": 171380 }, { "epoch": 0.43, "learning_rate": 5.713951381549147e-05, "loss": 1.0439, "step": 171385 }, { "epoch": 0.43, "learning_rate": 5.713825557400977e-05, "loss": 1.0467, "step": 171390 }, { "epoch": 0.43, "learning_rate": 5.713699733252806e-05, "loss": 1.0497, "step": 171395 }, { "epoch": 0.43, "learning_rate": 5.713573909104636e-05, "loss": 1.0484, "step": 171400 }, { "epoch": 0.43, "learning_rate": 5.713448084956465e-05, "loss": 1.0479, "step": 171405 }, { "epoch": 0.43, "learning_rate": 5.713322260808295e-05, "loss": 1.0437, "step": 171410 }, { "epoch": 0.43, "learning_rate": 5.713196436660124e-05, "loss": 1.043, "step": 171415 }, { "epoch": 0.43, "learning_rate": 5.7130706125119534e-05, "loss": 1.0469, "step": 171420 }, { "epoch": 0.43, "learning_rate": 5.712944788363783e-05, "loss": 1.0483, "step": 171425 }, { "epoch": 0.43, "learning_rate": 5.7128189642156124e-05, "loss": 1.0447, "step": 171430 }, { "epoch": 0.43, "learning_rate": 5.712693140067442e-05, "loss": 1.0431, "step": 171435 }, { "epoch": 0.43, "learning_rate": 5.7125673159192714e-05, "loss": 1.044, "step": 171440 }, { "epoch": 0.43, "learning_rate": 5.712441491771101e-05, "loss": 1.0451, "step": 171445 }, { "epoch": 0.43, "learning_rate": 5.7123156676229304e-05, "loss": 1.0479, "step": 171450 }, { "epoch": 0.43, "learning_rate": 5.71218984347476e-05, "loss": 1.0458, "step": 171455 }, { "epoch": 0.43, "learning_rate": 5.7120640193265894e-05, "loss": 1.0443, "step": 171460 }, { "epoch": 0.43, "learning_rate": 5.711938195178419e-05, "loss": 1.0458, "step": 171465 }, { "epoch": 0.43, "learning_rate": 5.7118123710302484e-05, "loss": 1.0472, "step": 171470 }, { "epoch": 0.43, "learning_rate": 5.711686546882078e-05, "loss": 1.0459, "step": 171475 }, { "epoch": 0.43, "learning_rate": 5.7115607227339074e-05, "loss": 1.0467, "step": 171480 }, { "epoch": 0.43, "learning_rate": 5.7114348985857366e-05, "loss": 1.0473, "step": 171485 }, { "epoch": 0.43, "learning_rate": 5.7113090744375664e-05, "loss": 1.0492, "step": 171490 }, { "epoch": 0.43, "learning_rate": 5.7111832502893956e-05, "loss": 1.0457, "step": 171495 }, { "epoch": 0.43, "learning_rate": 5.7110574261412254e-05, "loss": 1.0468, "step": 171500 }, { "epoch": 0.43, "learning_rate": 5.7109316019930546e-05, "loss": 1.0456, "step": 171505 }, { "epoch": 0.43, "learning_rate": 5.7108057778448844e-05, "loss": 1.0434, "step": 171510 }, { "epoch": 0.43, "learning_rate": 5.7106799536967136e-05, "loss": 1.042, "step": 171515 }, { "epoch": 0.43, "learning_rate": 5.7105541295485434e-05, "loss": 1.049, "step": 171520 }, { "epoch": 0.43, "learning_rate": 5.7104283054003726e-05, "loss": 1.0487, "step": 171525 }, { "epoch": 0.43, "learning_rate": 5.7103024812522024e-05, "loss": 1.0439, "step": 171530 }, { "epoch": 0.43, "learning_rate": 5.7101766571040316e-05, "loss": 1.0459, "step": 171535 }, { "epoch": 0.43, "learning_rate": 5.7100508329558614e-05, "loss": 1.0465, "step": 171540 }, { "epoch": 0.43, "learning_rate": 5.7099250088076906e-05, "loss": 1.0456, "step": 171545 }, { "epoch": 0.43, "learning_rate": 5.70979918465952e-05, "loss": 1.0472, "step": 171550 }, { "epoch": 0.43, "learning_rate": 5.7096733605113495e-05, "loss": 1.061, "step": 171555 }, { "epoch": 0.43, "learning_rate": 5.709547536363179e-05, "loss": 1.0688, "step": 171560 }, { "epoch": 0.43, "learning_rate": 5.7094217122150085e-05, "loss": 1.0474, "step": 171565 }, { "epoch": 0.43, "learning_rate": 5.709295888066838e-05, "loss": 1.0462, "step": 171570 }, { "epoch": 0.43, "learning_rate": 5.7091700639186675e-05, "loss": 1.0503, "step": 171575 }, { "epoch": 0.43, "learning_rate": 5.709044239770497e-05, "loss": 1.0445, "step": 171580 }, { "epoch": 0.43, "learning_rate": 5.7089184156223265e-05, "loss": 1.0462, "step": 171585 }, { "epoch": 0.43, "learning_rate": 5.708792591474156e-05, "loss": 1.0484, "step": 171590 }, { "epoch": 0.43, "learning_rate": 5.7086667673259855e-05, "loss": 1.0424, "step": 171595 }, { "epoch": 0.43, "learning_rate": 5.708540943177815e-05, "loss": 1.0441, "step": 171600 }, { "epoch": 0.43, "learning_rate": 5.7084151190296445e-05, "loss": 1.046, "step": 171605 }, { "epoch": 0.43, "learning_rate": 5.708289294881474e-05, "loss": 1.0469, "step": 171610 }, { "epoch": 0.43, "learning_rate": 5.708163470733303e-05, "loss": 1.0474, "step": 171615 }, { "epoch": 0.43, "learning_rate": 5.708037646585133e-05, "loss": 1.045, "step": 171620 }, { "epoch": 0.43, "learning_rate": 5.707911822436962e-05, "loss": 1.0442, "step": 171625 }, { "epoch": 0.43, "learning_rate": 5.707785998288792e-05, "loss": 1.0453, "step": 171630 }, { "epoch": 0.43, "learning_rate": 5.707660174140621e-05, "loss": 1.0413, "step": 171635 }, { "epoch": 0.43, "learning_rate": 5.707534349992451e-05, "loss": 1.0462, "step": 171640 }, { "epoch": 0.43, "learning_rate": 5.70740852584428e-05, "loss": 1.0488, "step": 171645 }, { "epoch": 0.43, "learning_rate": 5.70728270169611e-05, "loss": 1.0493, "step": 171650 }, { "epoch": 0.43, "learning_rate": 5.707156877547939e-05, "loss": 1.0435, "step": 171655 }, { "epoch": 0.43, "learning_rate": 5.707031053399769e-05, "loss": 1.0447, "step": 171660 }, { "epoch": 0.43, "learning_rate": 5.706905229251598e-05, "loss": 1.0405, "step": 171665 }, { "epoch": 0.43, "learning_rate": 5.706779405103428e-05, "loss": 1.0446, "step": 171670 }, { "epoch": 0.43, "learning_rate": 5.706653580955257e-05, "loss": 1.0451, "step": 171675 }, { "epoch": 0.43, "learning_rate": 5.706527756807086e-05, "loss": 1.0471, "step": 171680 }, { "epoch": 0.43, "learning_rate": 5.706401932658916e-05, "loss": 1.0482, "step": 171685 }, { "epoch": 0.43, "learning_rate": 5.706276108510745e-05, "loss": 1.0451, "step": 171690 }, { "epoch": 0.43, "learning_rate": 5.706150284362575e-05, "loss": 1.045, "step": 171695 }, { "epoch": 0.43, "learning_rate": 5.706024460214404e-05, "loss": 1.0462, "step": 171700 }, { "epoch": 0.43, "learning_rate": 5.705898636066234e-05, "loss": 1.0438, "step": 171705 }, { "epoch": 0.43, "learning_rate": 5.705772811918063e-05, "loss": 1.0469, "step": 171710 }, { "epoch": 0.43, "learning_rate": 5.705646987769893e-05, "loss": 1.0479, "step": 171715 }, { "epoch": 0.43, "learning_rate": 5.705521163621722e-05, "loss": 1.0646, "step": 171720 }, { "epoch": 0.43, "learning_rate": 5.705395339473552e-05, "loss": 1.0455, "step": 171725 }, { "epoch": 0.43, "learning_rate": 5.705269515325381e-05, "loss": 1.044, "step": 171730 }, { "epoch": 0.43, "learning_rate": 5.705143691177211e-05, "loss": 1.0478, "step": 171735 }, { "epoch": 0.43, "learning_rate": 5.70501786702904e-05, "loss": 1.0484, "step": 171740 }, { "epoch": 0.43, "learning_rate": 5.704892042880869e-05, "loss": 1.0443, "step": 171745 }, { "epoch": 0.43, "learning_rate": 5.7047662187327e-05, "loss": 1.0466, "step": 171750 }, { "epoch": 0.43, "learning_rate": 5.7046403945845295e-05, "loss": 1.0467, "step": 171755 }, { "epoch": 0.43, "learning_rate": 5.7045145704363587e-05, "loss": 1.0459, "step": 171760 }, { "epoch": 0.43, "learning_rate": 5.7043887462881885e-05, "loss": 1.0426, "step": 171765 }, { "epoch": 0.43, "learning_rate": 5.7042629221400176e-05, "loss": 1.0442, "step": 171770 }, { "epoch": 0.43, "learning_rate": 5.7041370979918475e-05, "loss": 1.0441, "step": 171775 }, { "epoch": 0.43, "learning_rate": 5.7040112738436766e-05, "loss": 1.0443, "step": 171780 }, { "epoch": 0.43, "learning_rate": 5.7038854496955065e-05, "loss": 1.0486, "step": 171785 }, { "epoch": 0.43, "learning_rate": 5.7037596255473356e-05, "loss": 1.0446, "step": 171790 }, { "epoch": 0.43, "learning_rate": 5.7036338013991655e-05, "loss": 1.0423, "step": 171795 }, { "epoch": 0.43, "learning_rate": 5.7035079772509946e-05, "loss": 1.0442, "step": 171800 }, { "epoch": 0.43, "learning_rate": 5.7033821531028245e-05, "loss": 1.047, "step": 171805 }, { "epoch": 0.43, "learning_rate": 5.7032563289546536e-05, "loss": 1.0495, "step": 171810 }, { "epoch": 0.43, "learning_rate": 5.7031305048064835e-05, "loss": 1.0684, "step": 171815 }, { "epoch": 0.43, "learning_rate": 5.7030046806583126e-05, "loss": 1.0481, "step": 171820 }, { "epoch": 0.43, "learning_rate": 5.702878856510142e-05, "loss": 1.0465, "step": 171825 }, { "epoch": 0.43, "learning_rate": 5.7027530323619716e-05, "loss": 1.0473, "step": 171830 }, { "epoch": 0.43, "learning_rate": 5.702627208213801e-05, "loss": 1.0454, "step": 171835 }, { "epoch": 0.43, "learning_rate": 5.7025013840656306e-05, "loss": 1.0455, "step": 171840 }, { "epoch": 0.43, "learning_rate": 5.70237555991746e-05, "loss": 1.046, "step": 171845 }, { "epoch": 0.43, "learning_rate": 5.7022497357692896e-05, "loss": 1.0459, "step": 171850 }, { "epoch": 0.43, "learning_rate": 5.702123911621119e-05, "loss": 1.0439, "step": 171855 }, { "epoch": 0.43, "learning_rate": 5.7019980874729486e-05, "loss": 1.0505, "step": 171860 }, { "epoch": 0.43, "learning_rate": 5.701872263324778e-05, "loss": 1.045, "step": 171865 }, { "epoch": 0.43, "learning_rate": 5.7017464391766076e-05, "loss": 1.0482, "step": 171870 }, { "epoch": 0.43, "learning_rate": 5.701620615028437e-05, "loss": 1.0432, "step": 171875 }, { "epoch": 0.43, "learning_rate": 5.7014947908802666e-05, "loss": 1.0461, "step": 171880 }, { "epoch": 0.43, "learning_rate": 5.701368966732096e-05, "loss": 1.0478, "step": 171885 }, { "epoch": 0.43, "learning_rate": 5.701243142583925e-05, "loss": 1.0457, "step": 171890 }, { "epoch": 0.43, "learning_rate": 5.701117318435755e-05, "loss": 1.043, "step": 171895 }, { "epoch": 0.43, "learning_rate": 5.700991494287584e-05, "loss": 1.0451, "step": 171900 }, { "epoch": 0.43, "learning_rate": 5.700865670139414e-05, "loss": 1.0448, "step": 171905 }, { "epoch": 0.43, "learning_rate": 5.700739845991243e-05, "loss": 1.0439, "step": 171910 }, { "epoch": 0.43, "learning_rate": 5.700614021843073e-05, "loss": 1.0446, "step": 171915 }, { "epoch": 0.43, "learning_rate": 5.700488197694902e-05, "loss": 1.0481, "step": 171920 }, { "epoch": 0.43, "learning_rate": 5.700362373546732e-05, "loss": 1.0457, "step": 171925 }, { "epoch": 0.43, "learning_rate": 5.700236549398561e-05, "loss": 1.0459, "step": 171930 }, { "epoch": 0.43, "learning_rate": 5.700110725250391e-05, "loss": 1.0481, "step": 171935 }, { "epoch": 0.43, "learning_rate": 5.69998490110222e-05, "loss": 1.0471, "step": 171940 }, { "epoch": 0.43, "learning_rate": 5.699859076954049e-05, "loss": 1.0443, "step": 171945 }, { "epoch": 0.43, "learning_rate": 5.699733252805879e-05, "loss": 1.0446, "step": 171950 }, { "epoch": 0.43, "learning_rate": 5.699607428657708e-05, "loss": 1.0442, "step": 171955 }, { "epoch": 0.43, "learning_rate": 5.699481604509538e-05, "loss": 1.0467, "step": 171960 }, { "epoch": 0.43, "learning_rate": 5.699355780361367e-05, "loss": 1.0461, "step": 171965 }, { "epoch": 0.43, "learning_rate": 5.699229956213197e-05, "loss": 1.0434, "step": 171970 }, { "epoch": 0.43, "learning_rate": 5.699104132065026e-05, "loss": 1.0442, "step": 171975 }, { "epoch": 0.43, "learning_rate": 5.698978307916856e-05, "loss": 1.0438, "step": 171980 }, { "epoch": 0.43, "learning_rate": 5.698852483768685e-05, "loss": 1.0461, "step": 171985 }, { "epoch": 0.43, "learning_rate": 5.698726659620515e-05, "loss": 1.0442, "step": 171990 }, { "epoch": 0.43, "learning_rate": 5.698600835472344e-05, "loss": 1.0486, "step": 171995 }, { "epoch": 0.43, "learning_rate": 5.698475011324174e-05, "loss": 1.045, "step": 172000 }, { "epoch": 0.43, "learning_rate": 5.698349187176003e-05, "loss": 1.0498, "step": 172005 }, { "epoch": 0.43, "learning_rate": 5.698223363027832e-05, "loss": 1.0442, "step": 172010 }, { "epoch": 0.43, "learning_rate": 5.698097538879662e-05, "loss": 1.0439, "step": 172015 }, { "epoch": 0.43, "learning_rate": 5.697971714731491e-05, "loss": 1.0657, "step": 172020 }, { "epoch": 0.43, "learning_rate": 5.697845890583321e-05, "loss": 1.0444, "step": 172025 }, { "epoch": 0.43, "learning_rate": 5.69772006643515e-05, "loss": 1.0445, "step": 172030 }, { "epoch": 0.43, "learning_rate": 5.69759424228698e-05, "loss": 1.0462, "step": 172035 }, { "epoch": 0.43, "learning_rate": 5.697468418138809e-05, "loss": 1.0469, "step": 172040 }, { "epoch": 0.43, "learning_rate": 5.697342593990639e-05, "loss": 1.0466, "step": 172045 }, { "epoch": 0.43, "learning_rate": 5.697216769842468e-05, "loss": 1.0472, "step": 172050 }, { "epoch": 0.43, "learning_rate": 5.697090945694298e-05, "loss": 1.0477, "step": 172055 }, { "epoch": 0.43, "learning_rate": 5.696965121546127e-05, "loss": 1.0486, "step": 172060 }, { "epoch": 0.43, "learning_rate": 5.696839297397957e-05, "loss": 1.0455, "step": 172065 }, { "epoch": 0.43, "learning_rate": 5.696713473249786e-05, "loss": 1.0466, "step": 172070 }, { "epoch": 0.43, "learning_rate": 5.6965876491016154e-05, "loss": 1.0429, "step": 172075 }, { "epoch": 0.43, "learning_rate": 5.696461824953445e-05, "loss": 1.0448, "step": 172080 }, { "epoch": 0.43, "learning_rate": 5.6963360008052744e-05, "loss": 1.0434, "step": 172085 }, { "epoch": 0.43, "learning_rate": 5.696210176657104e-05, "loss": 1.0482, "step": 172090 }, { "epoch": 0.43, "learning_rate": 5.6960843525089334e-05, "loss": 1.0439, "step": 172095 }, { "epoch": 0.43, "learning_rate": 5.695958528360763e-05, "loss": 1.0643, "step": 172100 }, { "epoch": 0.43, "learning_rate": 5.6958327042125924e-05, "loss": 1.0474, "step": 172105 }, { "epoch": 0.43, "learning_rate": 5.695706880064422e-05, "loss": 1.0443, "step": 172110 }, { "epoch": 0.43, "learning_rate": 5.6955810559162514e-05, "loss": 1.0493, "step": 172115 }, { "epoch": 0.43, "learning_rate": 5.695455231768081e-05, "loss": 1.0461, "step": 172120 }, { "epoch": 0.43, "learning_rate": 5.6953294076199104e-05, "loss": 1.0481, "step": 172125 }, { "epoch": 0.43, "learning_rate": 5.69520358347174e-05, "loss": 1.0482, "step": 172130 }, { "epoch": 0.43, "learning_rate": 5.6950777593235693e-05, "loss": 1.0418, "step": 172135 }, { "epoch": 0.43, "learning_rate": 5.6949519351753985e-05, "loss": 1.0433, "step": 172140 }, { "epoch": 0.43, "learning_rate": 5.6948261110272283e-05, "loss": 1.0493, "step": 172145 }, { "epoch": 0.43, "learning_rate": 5.6947002868790575e-05, "loss": 1.0457, "step": 172150 }, { "epoch": 0.43, "learning_rate": 5.6945744627308873e-05, "loss": 1.0426, "step": 172155 }, { "epoch": 0.43, "learning_rate": 5.6944486385827165e-05, "loss": 1.0448, "step": 172160 }, { "epoch": 0.43, "learning_rate": 5.6943228144345463e-05, "loss": 1.0446, "step": 172165 }, { "epoch": 0.43, "learning_rate": 5.6941969902863755e-05, "loss": 1.0478, "step": 172170 }, { "epoch": 0.43, "learning_rate": 5.694071166138205e-05, "loss": 1.0454, "step": 172175 }, { "epoch": 0.43, "learning_rate": 5.6939453419900345e-05, "loss": 1.0447, "step": 172180 }, { "epoch": 0.43, "learning_rate": 5.693819517841864e-05, "loss": 1.0448, "step": 172185 }, { "epoch": 0.43, "learning_rate": 5.6936936936936935e-05, "loss": 1.0464, "step": 172190 }, { "epoch": 0.43, "learning_rate": 5.693567869545523e-05, "loss": 1.0669, "step": 172195 }, { "epoch": 0.43, "learning_rate": 5.6934420453973525e-05, "loss": 1.0436, "step": 172200 }, { "epoch": 0.43, "learning_rate": 5.6933162212491817e-05, "loss": 1.0458, "step": 172205 }, { "epoch": 0.43, "learning_rate": 5.6931903971010115e-05, "loss": 1.0449, "step": 172210 }, { "epoch": 0.43, "learning_rate": 5.6930645729528406e-05, "loss": 1.0446, "step": 172215 }, { "epoch": 0.43, "learning_rate": 5.6929387488046705e-05, "loss": 1.0449, "step": 172220 }, { "epoch": 0.43, "learning_rate": 5.6928129246564996e-05, "loss": 1.0458, "step": 172225 }, { "epoch": 0.43, "learning_rate": 5.6926871005083295e-05, "loss": 1.0443, "step": 172230 }, { "epoch": 0.43, "learning_rate": 5.6925612763601586e-05, "loss": 1.047, "step": 172235 }, { "epoch": 0.43, "learning_rate": 5.6924354522119885e-05, "loss": 1.049, "step": 172240 }, { "epoch": 0.43, "learning_rate": 5.6923096280638176e-05, "loss": 1.043, "step": 172245 }, { "epoch": 0.43, "learning_rate": 5.692183803915648e-05, "loss": 1.0458, "step": 172250 }, { "epoch": 0.43, "learning_rate": 5.692057979767478e-05, "loss": 1.0451, "step": 172255 }, { "epoch": 0.43, "learning_rate": 5.691932155619307e-05, "loss": 1.046, "step": 172260 }, { "epoch": 0.43, "learning_rate": 5.691806331471137e-05, "loss": 1.048, "step": 172265 }, { "epoch": 0.43, "learning_rate": 5.691680507322966e-05, "loss": 1.0471, "step": 172270 }, { "epoch": 0.43, "learning_rate": 5.691554683174796e-05, "loss": 1.0477, "step": 172275 }, { "epoch": 0.43, "learning_rate": 5.691428859026625e-05, "loss": 1.0444, "step": 172280 }, { "epoch": 0.43, "learning_rate": 5.691303034878454e-05, "loss": 1.0429, "step": 172285 }, { "epoch": 0.43, "learning_rate": 5.691177210730284e-05, "loss": 1.0497, "step": 172290 }, { "epoch": 0.43, "learning_rate": 5.691051386582113e-05, "loss": 1.0452, "step": 172295 }, { "epoch": 0.43, "learning_rate": 5.690925562433943e-05, "loss": 1.0448, "step": 172300 }, { "epoch": 0.43, "learning_rate": 5.690799738285772e-05, "loss": 1.0468, "step": 172305 }, { "epoch": 0.43, "learning_rate": 5.690673914137602e-05, "loss": 1.0434, "step": 172310 }, { "epoch": 0.43, "learning_rate": 5.690548089989431e-05, "loss": 1.0459, "step": 172315 }, { "epoch": 0.43, "learning_rate": 5.690422265841261e-05, "loss": 1.0451, "step": 172320 }, { "epoch": 0.43, "learning_rate": 5.69029644169309e-05, "loss": 1.0473, "step": 172325 }, { "epoch": 0.43, "learning_rate": 5.69017061754492e-05, "loss": 1.0466, "step": 172330 }, { "epoch": 0.43, "learning_rate": 5.690044793396749e-05, "loss": 1.0486, "step": 172335 }, { "epoch": 0.43, "learning_rate": 5.689918969248579e-05, "loss": 1.0483, "step": 172340 }, { "epoch": 0.43, "learning_rate": 5.689793145100408e-05, "loss": 1.07, "step": 172345 }, { "epoch": 0.43, "learning_rate": 5.6896673209522374e-05, "loss": 1.0465, "step": 172350 }, { "epoch": 0.43, "learning_rate": 5.689541496804067e-05, "loss": 1.0457, "step": 172355 }, { "epoch": 0.43, "learning_rate": 5.6894156726558964e-05, "loss": 1.0426, "step": 172360 }, { "epoch": 0.43, "learning_rate": 5.689289848507726e-05, "loss": 1.0437, "step": 172365 }, { "epoch": 0.43, "learning_rate": 5.6891640243595554e-05, "loss": 1.0448, "step": 172370 }, { "epoch": 0.43, "learning_rate": 5.689038200211385e-05, "loss": 1.0425, "step": 172375 }, { "epoch": 0.43, "learning_rate": 5.6889123760632144e-05, "loss": 1.0439, "step": 172380 }, { "epoch": 0.43, "learning_rate": 5.688786551915044e-05, "loss": 1.0472, "step": 172385 }, { "epoch": 0.43, "learning_rate": 5.6886607277668734e-05, "loss": 1.0433, "step": 172390 }, { "epoch": 0.43, "learning_rate": 5.688534903618703e-05, "loss": 1.0496, "step": 172395 }, { "epoch": 0.43, "learning_rate": 5.6884090794705324e-05, "loss": 1.0467, "step": 172400 }, { "epoch": 0.43, "learning_rate": 5.688283255322362e-05, "loss": 1.0463, "step": 172405 }, { "epoch": 0.43, "learning_rate": 5.6881574311741914e-05, "loss": 1.0453, "step": 172410 }, { "epoch": 0.43, "learning_rate": 5.6880316070260206e-05, "loss": 1.0478, "step": 172415 }, { "epoch": 0.43, "learning_rate": 5.6879057828778504e-05, "loss": 1.0484, "step": 172420 }, { "epoch": 0.43, "learning_rate": 5.6877799587296796e-05, "loss": 1.0467, "step": 172425 }, { "epoch": 0.43, "learning_rate": 5.6876541345815094e-05, "loss": 1.0482, "step": 172430 }, { "epoch": 0.43, "learning_rate": 5.6875283104333386e-05, "loss": 1.0458, "step": 172435 }, { "epoch": 0.43, "learning_rate": 5.6874024862851684e-05, "loss": 1.0428, "step": 172440 }, { "epoch": 0.43, "learning_rate": 5.6872766621369976e-05, "loss": 1.0438, "step": 172445 }, { "epoch": 0.43, "learning_rate": 5.6871508379888274e-05, "loss": 1.0451, "step": 172450 }, { "epoch": 0.43, "learning_rate": 5.6870250138406566e-05, "loss": 1.0421, "step": 172455 }, { "epoch": 0.43, "learning_rate": 5.6868991896924864e-05, "loss": 1.0458, "step": 172460 }, { "epoch": 0.43, "learning_rate": 5.6867733655443156e-05, "loss": 1.044, "step": 172465 }, { "epoch": 0.43, "learning_rate": 5.6866475413961454e-05, "loss": 1.0444, "step": 172470 }, { "epoch": 0.43, "learning_rate": 5.6865217172479746e-05, "loss": 1.0482, "step": 172475 }, { "epoch": 0.43, "learning_rate": 5.686395893099804e-05, "loss": 1.0463, "step": 172480 }, { "epoch": 0.43, "learning_rate": 5.6862700689516336e-05, "loss": 1.0452, "step": 172485 }, { "epoch": 0.43, "learning_rate": 5.686169409633098e-05, "loss": 1.0493, "step": 172490 }, { "epoch": 0.43, "learning_rate": 5.686043585484927e-05, "loss": 1.0473, "step": 172495 }, { "epoch": 0.43, "learning_rate": 5.685917761336756e-05, "loss": 1.0447, "step": 172500 }, { "epoch": 0.43, "learning_rate": 5.685791937188586e-05, "loss": 1.0459, "step": 172505 }, { "epoch": 0.43, "learning_rate": 5.685666113040415e-05, "loss": 1.0689, "step": 172510 }, { "epoch": 0.43, "learning_rate": 5.685540288892245e-05, "loss": 1.0434, "step": 172515 }, { "epoch": 0.43, "learning_rate": 5.685414464744074e-05, "loss": 1.0453, "step": 172520 }, { "epoch": 0.43, "learning_rate": 5.685288640595904e-05, "loss": 1.0446, "step": 172525 }, { "epoch": 0.43, "learning_rate": 5.685162816447733e-05, "loss": 1.0444, "step": 172530 }, { "epoch": 0.43, "learning_rate": 5.685036992299563e-05, "loss": 1.0471, "step": 172535 }, { "epoch": 0.43, "learning_rate": 5.684911168151392e-05, "loss": 1.0715, "step": 172540 }, { "epoch": 0.43, "learning_rate": 5.684785344003222e-05, "loss": 1.057, "step": 172545 }, { "epoch": 0.43, "learning_rate": 5.684659519855051e-05, "loss": 1.0479, "step": 172550 }, { "epoch": 0.43, "learning_rate": 5.684533695706881e-05, "loss": 1.0454, "step": 172555 }, { "epoch": 0.43, "learning_rate": 5.68440787155871e-05, "loss": 1.0419, "step": 172560 }, { "epoch": 0.43, "learning_rate": 5.684282047410539e-05, "loss": 1.0455, "step": 172565 }, { "epoch": 0.43, "learning_rate": 5.684156223262369e-05, "loss": 1.046, "step": 172570 }, { "epoch": 0.43, "learning_rate": 5.684030399114198e-05, "loss": 1.0464, "step": 172575 }, { "epoch": 0.43, "learning_rate": 5.683904574966028e-05, "loss": 1.0483, "step": 172580 }, { "epoch": 0.43, "learning_rate": 5.683778750817857e-05, "loss": 1.0464, "step": 172585 }, { "epoch": 0.43, "learning_rate": 5.683652926669687e-05, "loss": 1.0442, "step": 172590 }, { "epoch": 0.43, "learning_rate": 5.683527102521516e-05, "loss": 1.0476, "step": 172595 }, { "epoch": 0.43, "learning_rate": 5.683401278373346e-05, "loss": 1.0438, "step": 172600 }, { "epoch": 0.43, "learning_rate": 5.683275454225175e-05, "loss": 1.0457, "step": 172605 }, { "epoch": 0.43, "learning_rate": 5.683149630077005e-05, "loss": 1.0458, "step": 172610 }, { "epoch": 0.43, "learning_rate": 5.683023805928834e-05, "loss": 1.0427, "step": 172615 }, { "epoch": 0.43, "learning_rate": 5.682897981780664e-05, "loss": 1.0451, "step": 172620 }, { "epoch": 0.43, "learning_rate": 5.682772157632493e-05, "loss": 1.044, "step": 172625 }, { "epoch": 0.43, "learning_rate": 5.6826463334843225e-05, "loss": 1.0431, "step": 172630 }, { "epoch": 0.43, "learning_rate": 5.682520509336152e-05, "loss": 1.0453, "step": 172635 }, { "epoch": 0.43, "learning_rate": 5.6823946851879814e-05, "loss": 1.0475, "step": 172640 }, { "epoch": 0.43, "learning_rate": 5.682268861039811e-05, "loss": 1.0511, "step": 172645 }, { "epoch": 0.43, "learning_rate": 5.6821430368916404e-05, "loss": 1.0453, "step": 172650 }, { "epoch": 0.43, "learning_rate": 5.68201721274347e-05, "loss": 1.043, "step": 172655 }, { "epoch": 0.43, "learning_rate": 5.6818913885952994e-05, "loss": 1.0695, "step": 172660 }, { "epoch": 0.43, "learning_rate": 5.681765564447129e-05, "loss": 1.0661, "step": 172665 }, { "epoch": 0.43, "learning_rate": 5.6816397402989584e-05, "loss": 1.0459, "step": 172670 }, { "epoch": 0.43, "learning_rate": 5.681513916150788e-05, "loss": 1.0466, "step": 172675 }, { "epoch": 0.43, "learning_rate": 5.6813880920026174e-05, "loss": 1.0469, "step": 172680 }, { "epoch": 0.43, "learning_rate": 5.6812622678544466e-05, "loss": 1.0476, "step": 172685 }, { "epoch": 0.43, "learning_rate": 5.6811364437062764e-05, "loss": 1.044, "step": 172690 }, { "epoch": 0.43, "learning_rate": 5.6810106195581056e-05, "loss": 1.0458, "step": 172695 }, { "epoch": 0.43, "learning_rate": 5.6808847954099354e-05, "loss": 1.0443, "step": 172700 }, { "epoch": 0.43, "learning_rate": 5.6807589712617646e-05, "loss": 1.0452, "step": 172705 }, { "epoch": 0.43, "learning_rate": 5.6806331471135944e-05, "loss": 1.0454, "step": 172710 }, { "epoch": 0.43, "learning_rate": 5.6805073229654236e-05, "loss": 1.0458, "step": 172715 }, { "epoch": 0.43, "learning_rate": 5.6803814988172534e-05, "loss": 1.0471, "step": 172720 }, { "epoch": 0.43, "learning_rate": 5.6802556746690826e-05, "loss": 1.0477, "step": 172725 }, { "epoch": 0.43, "learning_rate": 5.6801298505209124e-05, "loss": 1.0447, "step": 172730 }, { "epoch": 0.43, "learning_rate": 5.6800040263727416e-05, "loss": 1.0464, "step": 172735 }, { "epoch": 0.43, "learning_rate": 5.6798782022245714e-05, "loss": 1.0482, "step": 172740 }, { "epoch": 0.43, "learning_rate": 5.6797523780764006e-05, "loss": 1.0479, "step": 172745 }, { "epoch": 0.43, "learning_rate": 5.67962655392823e-05, "loss": 1.0448, "step": 172750 }, { "epoch": 0.43, "learning_rate": 5.6795007297800596e-05, "loss": 1.0451, "step": 172755 }, { "epoch": 0.43, "learning_rate": 5.679374905631889e-05, "loss": 1.045, "step": 172760 }, { "epoch": 0.43, "learning_rate": 5.6792490814837186e-05, "loss": 1.0491, "step": 172765 }, { "epoch": 0.43, "learning_rate": 5.679123257335548e-05, "loss": 1.0445, "step": 172770 }, { "epoch": 0.43, "learning_rate": 5.6789974331873776e-05, "loss": 1.0464, "step": 172775 }, { "epoch": 0.43, "learning_rate": 5.678871609039207e-05, "loss": 1.0476, "step": 172780 }, { "epoch": 0.43, "learning_rate": 5.6787457848910366e-05, "loss": 1.0436, "step": 172785 }, { "epoch": 0.43, "learning_rate": 5.678619960742866e-05, "loss": 1.048, "step": 172790 }, { "epoch": 0.43, "learning_rate": 5.6784941365946956e-05, "loss": 1.0488, "step": 172795 }, { "epoch": 0.43, "learning_rate": 5.678368312446525e-05, "loss": 1.0712, "step": 172800 }, { "epoch": 0.43, "learning_rate": 5.6782424882983546e-05, "loss": 1.0451, "step": 172805 }, { "epoch": 0.43, "learning_rate": 5.678116664150184e-05, "loss": 1.0443, "step": 172810 }, { "epoch": 0.43, "learning_rate": 5.677990840002013e-05, "loss": 1.0629, "step": 172815 }, { "epoch": 0.43, "learning_rate": 5.677865015853843e-05, "loss": 1.0441, "step": 172820 }, { "epoch": 0.43, "learning_rate": 5.677739191705672e-05, "loss": 1.0479, "step": 172825 }, { "epoch": 0.43, "learning_rate": 5.677613367557502e-05, "loss": 1.0448, "step": 172830 }, { "epoch": 0.43, "learning_rate": 5.677487543409331e-05, "loss": 1.0438, "step": 172835 }, { "epoch": 0.43, "learning_rate": 5.677361719261161e-05, "loss": 1.0488, "step": 172840 }, { "epoch": 0.43, "learning_rate": 5.67723589511299e-05, "loss": 1.0424, "step": 172845 }, { "epoch": 0.43, "learning_rate": 5.67711007096482e-05, "loss": 1.0726, "step": 172850 }, { "epoch": 0.43, "learning_rate": 5.676984246816649e-05, "loss": 1.0444, "step": 172855 }, { "epoch": 0.43, "learning_rate": 5.676858422668479e-05, "loss": 1.0445, "step": 172860 }, { "epoch": 0.43, "learning_rate": 5.676732598520308e-05, "loss": 1.0452, "step": 172865 }, { "epoch": 0.43, "learning_rate": 5.676606774372138e-05, "loss": 1.0497, "step": 172870 }, { "epoch": 0.43, "learning_rate": 5.676480950223967e-05, "loss": 1.0454, "step": 172875 }, { "epoch": 0.43, "learning_rate": 5.676355126075796e-05, "loss": 1.0483, "step": 172880 }, { "epoch": 0.43, "learning_rate": 5.676229301927626e-05, "loss": 1.0482, "step": 172885 }, { "epoch": 0.43, "learning_rate": 5.676103477779455e-05, "loss": 1.0481, "step": 172890 }, { "epoch": 0.43, "learning_rate": 5.675977653631285e-05, "loss": 1.0468, "step": 172895 }, { "epoch": 0.43, "learning_rate": 5.675851829483114e-05, "loss": 1.0451, "step": 172900 }, { "epoch": 0.43, "learning_rate": 5.675726005334944e-05, "loss": 1.0461, "step": 172905 }, { "epoch": 0.43, "learning_rate": 5.675600181186773e-05, "loss": 1.0436, "step": 172910 }, { "epoch": 0.43, "learning_rate": 5.675474357038603e-05, "loss": 1.0463, "step": 172915 }, { "epoch": 0.43, "learning_rate": 5.675348532890432e-05, "loss": 1.0477, "step": 172920 }, { "epoch": 0.43, "learning_rate": 5.675222708742262e-05, "loss": 1.0474, "step": 172925 }, { "epoch": 0.43, "learning_rate": 5.675096884594091e-05, "loss": 1.0475, "step": 172930 }, { "epoch": 0.43, "learning_rate": 5.674971060445921e-05, "loss": 1.0451, "step": 172935 }, { "epoch": 0.43, "learning_rate": 5.67484523629775e-05, "loss": 1.0448, "step": 172940 }, { "epoch": 0.43, "learning_rate": 5.674719412149579e-05, "loss": 1.0438, "step": 172945 }, { "epoch": 0.43, "learning_rate": 5.6745935880014104e-05, "loss": 1.0416, "step": 172950 }, { "epoch": 0.43, "learning_rate": 5.6744677638532395e-05, "loss": 1.0458, "step": 172955 }, { "epoch": 0.43, "learning_rate": 5.674341939705069e-05, "loss": 1.0477, "step": 172960 }, { "epoch": 0.43, "learning_rate": 5.6742161155568985e-05, "loss": 1.0473, "step": 172965 }, { "epoch": 0.43, "learning_rate": 5.674090291408728e-05, "loss": 1.0432, "step": 172970 }, { "epoch": 0.43, "learning_rate": 5.6739644672605575e-05, "loss": 1.0432, "step": 172975 }, { "epoch": 0.43, "learning_rate": 5.673838643112387e-05, "loss": 1.044, "step": 172980 }, { "epoch": 0.43, "learning_rate": 5.6737128189642165e-05, "loss": 1.0458, "step": 172985 }, { "epoch": 0.43, "learning_rate": 5.673586994816046e-05, "loss": 1.047, "step": 172990 }, { "epoch": 0.43, "learning_rate": 5.6734611706678755e-05, "loss": 1.047, "step": 172995 }, { "epoch": 0.43, "learning_rate": 5.673335346519705e-05, "loss": 1.0452, "step": 173000 }, { "epoch": 0.43, "learning_rate": 5.6732095223715345e-05, "loss": 1.0479, "step": 173005 }, { "epoch": 0.43, "learning_rate": 5.673083698223364e-05, "loss": 1.0466, "step": 173010 }, { "epoch": 0.43, "learning_rate": 5.6729578740751935e-05, "loss": 1.0425, "step": 173015 }, { "epoch": 0.43, "learning_rate": 5.6728320499270227e-05, "loss": 1.0429, "step": 173020 }, { "epoch": 0.43, "learning_rate": 5.672706225778852e-05, "loss": 1.069, "step": 173025 }, { "epoch": 0.43, "learning_rate": 5.6725804016306817e-05, "loss": 1.0463, "step": 173030 }, { "epoch": 0.43, "learning_rate": 5.672454577482511e-05, "loss": 1.0464, "step": 173035 }, { "epoch": 0.43, "learning_rate": 5.6723287533343407e-05, "loss": 1.0477, "step": 173040 }, { "epoch": 0.43, "learning_rate": 5.67220292918617e-05, "loss": 1.0443, "step": 173045 }, { "epoch": 0.43, "learning_rate": 5.6720771050379997e-05, "loss": 1.046, "step": 173050 }, { "epoch": 0.43, "learning_rate": 5.671951280889829e-05, "loss": 1.0454, "step": 173055 }, { "epoch": 0.43, "learning_rate": 5.6718254567416586e-05, "loss": 1.0458, "step": 173060 }, { "epoch": 0.43, "learning_rate": 5.671699632593488e-05, "loss": 1.0479, "step": 173065 }, { "epoch": 0.43, "learning_rate": 5.6715738084453176e-05, "loss": 1.0433, "step": 173070 }, { "epoch": 0.43, "learning_rate": 5.671447984297147e-05, "loss": 1.0444, "step": 173075 }, { "epoch": 0.43, "learning_rate": 5.6713221601489766e-05, "loss": 1.044, "step": 173080 }, { "epoch": 0.43, "learning_rate": 5.671196336000806e-05, "loss": 1.0466, "step": 173085 }, { "epoch": 0.43, "learning_rate": 5.671070511852635e-05, "loss": 1.0471, "step": 173090 }, { "epoch": 0.43, "learning_rate": 5.670944687704465e-05, "loss": 1.0422, "step": 173095 }, { "epoch": 0.43, "learning_rate": 5.670818863556294e-05, "loss": 1.0437, "step": 173100 }, { "epoch": 0.43, "learning_rate": 5.670693039408124e-05, "loss": 1.0463, "step": 173105 }, { "epoch": 0.43, "learning_rate": 5.670567215259953e-05, "loss": 1.0424, "step": 173110 }, { "epoch": 0.43, "learning_rate": 5.670441391111783e-05, "loss": 1.0463, "step": 173115 }, { "epoch": 0.43, "learning_rate": 5.670315566963612e-05, "loss": 1.0491, "step": 173120 }, { "epoch": 0.43, "learning_rate": 5.670189742815442e-05, "loss": 1.0468, "step": 173125 }, { "epoch": 0.43, "learning_rate": 5.670063918667271e-05, "loss": 1.0471, "step": 173130 }, { "epoch": 0.43, "learning_rate": 5.669938094519101e-05, "loss": 1.0475, "step": 173135 }, { "epoch": 0.43, "learning_rate": 5.66981227037093e-05, "loss": 1.0443, "step": 173140 }, { "epoch": 0.43, "learning_rate": 5.66968644622276e-05, "loss": 1.0454, "step": 173145 }, { "epoch": 0.43, "learning_rate": 5.669560622074589e-05, "loss": 1.0438, "step": 173150 }, { "epoch": 0.43, "learning_rate": 5.669434797926418e-05, "loss": 1.0465, "step": 173155 }, { "epoch": 0.43, "learning_rate": 5.669308973778248e-05, "loss": 1.0474, "step": 173160 }, { "epoch": 0.43, "learning_rate": 5.669183149630077e-05, "loss": 1.0451, "step": 173165 }, { "epoch": 0.43, "learning_rate": 5.669057325481907e-05, "loss": 1.0457, "step": 173170 }, { "epoch": 0.43, "learning_rate": 5.668931501333736e-05, "loss": 1.0443, "step": 173175 }, { "epoch": 0.43, "learning_rate": 5.668805677185566e-05, "loss": 1.0454, "step": 173180 }, { "epoch": 0.43, "learning_rate": 5.668679853037395e-05, "loss": 1.0434, "step": 173185 }, { "epoch": 0.43, "learning_rate": 5.668554028889225e-05, "loss": 1.0444, "step": 173190 }, { "epoch": 0.43, "learning_rate": 5.668428204741054e-05, "loss": 1.0439, "step": 173195 }, { "epoch": 0.43, "learning_rate": 5.668302380592884e-05, "loss": 1.0461, "step": 173200 }, { "epoch": 0.43, "learning_rate": 5.668176556444713e-05, "loss": 1.0475, "step": 173205 }, { "epoch": 0.43, "learning_rate": 5.668050732296543e-05, "loss": 1.0427, "step": 173210 }, { "epoch": 0.43, "learning_rate": 5.667924908148372e-05, "loss": 1.0497, "step": 173215 }, { "epoch": 0.43, "learning_rate": 5.667799084000201e-05, "loss": 1.0669, "step": 173220 }, { "epoch": 0.43, "learning_rate": 5.667673259852031e-05, "loss": 1.0455, "step": 173225 }, { "epoch": 0.43, "learning_rate": 5.66754743570386e-05, "loss": 1.0453, "step": 173230 }, { "epoch": 0.43, "learning_rate": 5.66742161155569e-05, "loss": 1.0438, "step": 173235 }, { "epoch": 0.43, "learning_rate": 5.667295787407519e-05, "loss": 1.0479, "step": 173240 }, { "epoch": 0.43, "learning_rate": 5.667169963259349e-05, "loss": 1.047, "step": 173245 }, { "epoch": 0.43, "learning_rate": 5.667044139111178e-05, "loss": 1.047, "step": 173250 }, { "epoch": 0.43, "learning_rate": 5.666918314963008e-05, "loss": 1.0481, "step": 173255 }, { "epoch": 0.43, "learning_rate": 5.666792490814837e-05, "loss": 1.0455, "step": 173260 }, { "epoch": 0.43, "learning_rate": 5.666666666666667e-05, "loss": 1.0468, "step": 173265 }, { "epoch": 0.43, "learning_rate": 5.666540842518496e-05, "loss": 1.0456, "step": 173270 }, { "epoch": 0.43, "learning_rate": 5.666415018370326e-05, "loss": 1.0474, "step": 173275 }, { "epoch": 0.43, "learning_rate": 5.666289194222155e-05, "loss": 1.0468, "step": 173280 }, { "epoch": 0.43, "learning_rate": 5.6661633700739844e-05, "loss": 1.047, "step": 173285 }, { "epoch": 0.43, "learning_rate": 5.666037545925814e-05, "loss": 1.0474, "step": 173290 }, { "epoch": 0.43, "learning_rate": 5.6659117217776434e-05, "loss": 1.0457, "step": 173295 }, { "epoch": 0.44, "learning_rate": 5.665785897629473e-05, "loss": 1.0436, "step": 173300 }, { "epoch": 0.44, "learning_rate": 5.6656600734813024e-05, "loss": 1.0431, "step": 173305 }, { "epoch": 0.44, "learning_rate": 5.665534249333132e-05, "loss": 1.0443, "step": 173310 }, { "epoch": 0.44, "learning_rate": 5.6654084251849614e-05, "loss": 1.0437, "step": 173315 }, { "epoch": 0.44, "learning_rate": 5.665282601036791e-05, "loss": 1.0419, "step": 173320 }, { "epoch": 0.44, "learning_rate": 5.6651567768886204e-05, "loss": 1.0453, "step": 173325 }, { "epoch": 0.44, "learning_rate": 5.66503095274045e-05, "loss": 1.043, "step": 173330 }, { "epoch": 0.44, "learning_rate": 5.6649051285922794e-05, "loss": 1.0474, "step": 173335 }, { "epoch": 0.44, "learning_rate": 5.6647793044441085e-05, "loss": 1.0448, "step": 173340 }, { "epoch": 0.44, "learning_rate": 5.6646534802959384e-05, "loss": 1.0469, "step": 173345 }, { "epoch": 0.44, "learning_rate": 5.6645276561477675e-05, "loss": 1.047, "step": 173350 }, { "epoch": 0.44, "learning_rate": 5.6644018319995974e-05, "loss": 1.046, "step": 173355 }, { "epoch": 0.44, "learning_rate": 5.6642760078514265e-05, "loss": 1.0442, "step": 173360 }, { "epoch": 0.44, "learning_rate": 5.6641501837032564e-05, "loss": 1.0465, "step": 173365 }, { "epoch": 0.44, "learning_rate": 5.6640243595550855e-05, "loss": 1.0465, "step": 173370 }, { "epoch": 0.44, "learning_rate": 5.6638985354069154e-05, "loss": 1.045, "step": 173375 }, { "epoch": 0.44, "learning_rate": 5.6637727112587445e-05, "loss": 1.0445, "step": 173380 }, { "epoch": 0.44, "learning_rate": 5.6636468871105744e-05, "loss": 1.0426, "step": 173385 }, { "epoch": 0.44, "learning_rate": 5.6635210629624035e-05, "loss": 1.0457, "step": 173390 }, { "epoch": 0.44, "learning_rate": 5.6633952388142334e-05, "loss": 1.0449, "step": 173395 }, { "epoch": 0.44, "learning_rate": 5.6632694146660625e-05, "loss": 1.0485, "step": 173400 }, { "epoch": 0.44, "learning_rate": 5.663143590517892e-05, "loss": 1.0441, "step": 173405 }, { "epoch": 0.44, "learning_rate": 5.6630177663697215e-05, "loss": 1.0459, "step": 173410 }, { "epoch": 0.44, "learning_rate": 5.662891942221551e-05, "loss": 1.0438, "step": 173415 }, { "epoch": 0.44, "learning_rate": 5.6627661180733805e-05, "loss": 1.0457, "step": 173420 }, { "epoch": 0.44, "learning_rate": 5.66264029392521e-05, "loss": 1.0472, "step": 173425 }, { "epoch": 0.44, "learning_rate": 5.6625144697770395e-05, "loss": 1.0504, "step": 173430 }, { "epoch": 0.44, "learning_rate": 5.662388645628869e-05, "loss": 1.0466, "step": 173435 }, { "epoch": 0.44, "learning_rate": 5.6622628214806985e-05, "loss": 1.0465, "step": 173440 }, { "epoch": 0.44, "learning_rate": 5.662136997332528e-05, "loss": 1.0461, "step": 173445 }, { "epoch": 0.44, "learning_rate": 5.662011173184358e-05, "loss": 1.0473, "step": 173450 }, { "epoch": 0.44, "learning_rate": 5.661885349036188e-05, "loss": 1.0431, "step": 173455 }, { "epoch": 0.44, "learning_rate": 5.661759524888017e-05, "loss": 1.046, "step": 173460 }, { "epoch": 0.44, "learning_rate": 5.661633700739847e-05, "loss": 1.0423, "step": 173465 }, { "epoch": 0.44, "learning_rate": 5.661507876591676e-05, "loss": 1.0453, "step": 173470 }, { "epoch": 0.44, "learning_rate": 5.661382052443506e-05, "loss": 1.0439, "step": 173475 }, { "epoch": 0.44, "learning_rate": 5.661256228295335e-05, "loss": 1.0459, "step": 173480 }, { "epoch": 0.44, "learning_rate": 5.661130404147164e-05, "loss": 1.0454, "step": 173485 }, { "epoch": 0.44, "learning_rate": 5.661004579998994e-05, "loss": 1.0477, "step": 173490 }, { "epoch": 0.44, "learning_rate": 5.660878755850823e-05, "loss": 1.0441, "step": 173495 }, { "epoch": 0.44, "learning_rate": 5.660752931702653e-05, "loss": 1.0467, "step": 173500 }, { "epoch": 0.44, "learning_rate": 5.660627107554482e-05, "loss": 1.0429, "step": 173505 }, { "epoch": 0.44, "learning_rate": 5.660501283406312e-05, "loss": 1.0454, "step": 173510 }, { "epoch": 0.44, "learning_rate": 5.660375459258141e-05, "loss": 1.0435, "step": 173515 }, { "epoch": 0.44, "learning_rate": 5.660249635109971e-05, "loss": 1.046, "step": 173520 }, { "epoch": 0.44, "learning_rate": 5.6601238109618e-05, "loss": 1.0431, "step": 173525 }, { "epoch": 0.44, "learning_rate": 5.65999798681363e-05, "loss": 1.046, "step": 173530 }, { "epoch": 0.44, "learning_rate": 5.659872162665459e-05, "loss": 1.0488, "step": 173535 }, { "epoch": 0.44, "learning_rate": 5.659746338517289e-05, "loss": 1.0455, "step": 173540 }, { "epoch": 0.44, "learning_rate": 5.659620514369118e-05, "loss": 1.0441, "step": 173545 }, { "epoch": 0.44, "learning_rate": 5.6594946902209475e-05, "loss": 1.0459, "step": 173550 }, { "epoch": 0.44, "learning_rate": 5.659368866072777e-05, "loss": 1.0473, "step": 173555 }, { "epoch": 0.44, "learning_rate": 5.6592430419246065e-05, "loss": 1.0449, "step": 173560 }, { "epoch": 0.44, "learning_rate": 5.659117217776436e-05, "loss": 1.0455, "step": 173565 }, { "epoch": 0.44, "learning_rate": 5.6589913936282655e-05, "loss": 1.0484, "step": 173570 }, { "epoch": 0.44, "learning_rate": 5.658865569480095e-05, "loss": 1.0475, "step": 173575 }, { "epoch": 0.44, "learning_rate": 5.6587397453319245e-05, "loss": 1.0485, "step": 173580 }, { "epoch": 0.44, "learning_rate": 5.658613921183754e-05, "loss": 1.0477, "step": 173585 }, { "epoch": 0.44, "learning_rate": 5.6584880970355835e-05, "loss": 1.0692, "step": 173590 }, { "epoch": 0.44, "learning_rate": 5.658362272887413e-05, "loss": 1.0483, "step": 173595 }, { "epoch": 0.44, "learning_rate": 5.6582364487392425e-05, "loss": 1.0566, "step": 173600 }, { "epoch": 0.44, "learning_rate": 5.658110624591072e-05, "loss": 1.0446, "step": 173605 }, { "epoch": 0.44, "learning_rate": 5.6579848004429015e-05, "loss": 1.0461, "step": 173610 }, { "epoch": 0.44, "learning_rate": 5.6578589762947306e-05, "loss": 1.0472, "step": 173615 }, { "epoch": 0.44, "learning_rate": 5.6577331521465605e-05, "loss": 1.0469, "step": 173620 }, { "epoch": 0.44, "learning_rate": 5.6576073279983896e-05, "loss": 1.0432, "step": 173625 }, { "epoch": 0.44, "learning_rate": 5.6574815038502195e-05, "loss": 1.0435, "step": 173630 }, { "epoch": 0.44, "learning_rate": 5.6573556797020486e-05, "loss": 1.0705, "step": 173635 }, { "epoch": 0.44, "learning_rate": 5.6572298555538784e-05, "loss": 1.0461, "step": 173640 }, { "epoch": 0.44, "learning_rate": 5.6571040314057076e-05, "loss": 1.0477, "step": 173645 }, { "epoch": 0.44, "learning_rate": 5.6569782072575374e-05, "loss": 1.0476, "step": 173650 }, { "epoch": 0.44, "learning_rate": 5.6568523831093666e-05, "loss": 1.0449, "step": 173655 }, { "epoch": 0.44, "learning_rate": 5.6567265589611964e-05, "loss": 1.0452, "step": 173660 }, { "epoch": 0.44, "learning_rate": 5.6566007348130256e-05, "loss": 1.0465, "step": 173665 }, { "epoch": 0.44, "learning_rate": 5.6564749106648554e-05, "loss": 1.0452, "step": 173670 }, { "epoch": 0.44, "learning_rate": 5.6563490865166846e-05, "loss": 1.0458, "step": 173675 }, { "epoch": 0.44, "learning_rate": 5.656223262368514e-05, "loss": 1.0457, "step": 173680 }, { "epoch": 0.44, "learning_rate": 5.6560974382203436e-05, "loss": 1.0457, "step": 173685 }, { "epoch": 0.44, "learning_rate": 5.655971614072173e-05, "loss": 1.0476, "step": 173690 }, { "epoch": 0.44, "learning_rate": 5.6558457899240026e-05, "loss": 1.0653, "step": 173695 }, { "epoch": 0.44, "learning_rate": 5.655719965775832e-05, "loss": 1.0463, "step": 173700 }, { "epoch": 0.44, "learning_rate": 5.6555941416276616e-05, "loss": 1.0437, "step": 173705 }, { "epoch": 0.44, "learning_rate": 5.655468317479491e-05, "loss": 1.0485, "step": 173710 }, { "epoch": 0.44, "learning_rate": 5.6553424933313206e-05, "loss": 1.0477, "step": 173715 }, { "epoch": 0.44, "learning_rate": 5.65521666918315e-05, "loss": 1.0457, "step": 173720 }, { "epoch": 0.44, "learning_rate": 5.6550908450349796e-05, "loss": 1.0461, "step": 173725 }, { "epoch": 0.44, "learning_rate": 5.654965020886809e-05, "loss": 1.0465, "step": 173730 }, { "epoch": 0.44, "learning_rate": 5.6548391967386386e-05, "loss": 1.0442, "step": 173735 }, { "epoch": 0.44, "learning_rate": 5.654713372590468e-05, "loss": 1.045, "step": 173740 }, { "epoch": 0.44, "learning_rate": 5.654587548442297e-05, "loss": 1.0459, "step": 173745 }, { "epoch": 0.44, "learning_rate": 5.654461724294127e-05, "loss": 1.0467, "step": 173750 }, { "epoch": 0.44, "learning_rate": 5.654335900145956e-05, "loss": 1.0431, "step": 173755 }, { "epoch": 0.44, "learning_rate": 5.654210075997786e-05, "loss": 1.0425, "step": 173760 }, { "epoch": 0.44, "learning_rate": 5.654084251849615e-05, "loss": 1.0457, "step": 173765 }, { "epoch": 0.44, "learning_rate": 5.653958427701445e-05, "loss": 1.0477, "step": 173770 }, { "epoch": 0.44, "learning_rate": 5.653832603553274e-05, "loss": 1.0441, "step": 173775 }, { "epoch": 0.44, "learning_rate": 5.653706779405104e-05, "loss": 1.0451, "step": 173780 }, { "epoch": 0.44, "learning_rate": 5.653580955256933e-05, "loss": 1.0466, "step": 173785 }, { "epoch": 0.44, "learning_rate": 5.653455131108763e-05, "loss": 1.045, "step": 173790 }, { "epoch": 0.44, "learning_rate": 5.653329306960592e-05, "loss": 1.0432, "step": 173795 }, { "epoch": 0.44, "learning_rate": 5.653203482812422e-05, "loss": 1.0456, "step": 173800 }, { "epoch": 0.44, "learning_rate": 5.653077658664251e-05, "loss": 1.0442, "step": 173805 }, { "epoch": 0.44, "learning_rate": 5.65295183451608e-05, "loss": 1.0457, "step": 173810 }, { "epoch": 0.44, "learning_rate": 5.65282601036791e-05, "loss": 1.0451, "step": 173815 }, { "epoch": 0.44, "learning_rate": 5.652700186219739e-05, "loss": 1.0443, "step": 173820 }, { "epoch": 0.44, "learning_rate": 5.652574362071569e-05, "loss": 1.0464, "step": 173825 }, { "epoch": 0.44, "learning_rate": 5.652448537923398e-05, "loss": 1.0476, "step": 173830 }, { "epoch": 0.44, "learning_rate": 5.652322713775228e-05, "loss": 1.0455, "step": 173835 }, { "epoch": 0.44, "learning_rate": 5.652196889627057e-05, "loss": 1.0469, "step": 173840 }, { "epoch": 0.44, "learning_rate": 5.652071065478887e-05, "loss": 1.0457, "step": 173845 }, { "epoch": 0.44, "learning_rate": 5.651945241330716e-05, "loss": 1.0495, "step": 173850 }, { "epoch": 0.44, "learning_rate": 5.651819417182546e-05, "loss": 1.0451, "step": 173855 }, { "epoch": 0.44, "learning_rate": 5.651693593034375e-05, "loss": 1.0449, "step": 173860 }, { "epoch": 0.44, "learning_rate": 5.651567768886205e-05, "loss": 1.0463, "step": 173865 }, { "epoch": 0.44, "learning_rate": 5.651441944738034e-05, "loss": 1.0464, "step": 173870 }, { "epoch": 0.44, "learning_rate": 5.651316120589863e-05, "loss": 1.045, "step": 173875 }, { "epoch": 0.44, "learning_rate": 5.651190296441693e-05, "loss": 1.0478, "step": 173880 }, { "epoch": 0.44, "learning_rate": 5.651064472293522e-05, "loss": 1.0465, "step": 173885 }, { "epoch": 0.44, "learning_rate": 5.650938648145352e-05, "loss": 1.0432, "step": 173890 }, { "epoch": 0.44, "learning_rate": 5.650812823997181e-05, "loss": 1.0681, "step": 173895 }, { "epoch": 0.44, "learning_rate": 5.650686999849011e-05, "loss": 1.0456, "step": 173900 }, { "epoch": 0.44, "learning_rate": 5.65056117570084e-05, "loss": 1.0462, "step": 173905 }, { "epoch": 0.44, "learning_rate": 5.65043535155267e-05, "loss": 1.0427, "step": 173910 }, { "epoch": 0.44, "learning_rate": 5.650309527404499e-05, "loss": 1.0488, "step": 173915 }, { "epoch": 0.44, "learning_rate": 5.650183703256329e-05, "loss": 1.0498, "step": 173920 }, { "epoch": 0.44, "learning_rate": 5.650057879108158e-05, "loss": 1.0483, "step": 173925 }, { "epoch": 0.44, "learning_rate": 5.649932054959988e-05, "loss": 1.0435, "step": 173930 }, { "epoch": 0.44, "learning_rate": 5.649806230811817e-05, "loss": 1.0442, "step": 173935 }, { "epoch": 0.44, "learning_rate": 5.649680406663646e-05, "loss": 1.0482, "step": 173940 }, { "epoch": 0.44, "learning_rate": 5.649554582515476e-05, "loss": 1.0477, "step": 173945 }, { "epoch": 0.44, "learning_rate": 5.649428758367307e-05, "loss": 1.063, "step": 173950 }, { "epoch": 0.44, "learning_rate": 5.649302934219136e-05, "loss": 1.0481, "step": 173955 }, { "epoch": 0.44, "learning_rate": 5.649177110070966e-05, "loss": 1.0474, "step": 173960 }, { "epoch": 0.44, "learning_rate": 5.649051285922795e-05, "loss": 1.0462, "step": 173965 }, { "epoch": 0.44, "learning_rate": 5.648925461774625e-05, "loss": 1.045, "step": 173970 }, { "epoch": 0.44, "learning_rate": 5.648799637626454e-05, "loss": 1.0437, "step": 173975 }, { "epoch": 0.44, "learning_rate": 5.648673813478284e-05, "loss": 1.0452, "step": 173980 }, { "epoch": 0.44, "learning_rate": 5.648547989330113e-05, "loss": 1.0459, "step": 173985 }, { "epoch": 0.44, "learning_rate": 5.648422165181943e-05, "loss": 1.0469, "step": 173990 }, { "epoch": 0.44, "learning_rate": 5.648296341033772e-05, "loss": 1.0446, "step": 173995 }, { "epoch": 0.44, "learning_rate": 5.648170516885602e-05, "loss": 1.0463, "step": 174000 }, { "epoch": 0.44, "learning_rate": 5.648044692737431e-05, "loss": 1.0434, "step": 174005 }, { "epoch": 0.44, "learning_rate": 5.647918868589261e-05, "loss": 1.046, "step": 174010 }, { "epoch": 0.44, "learning_rate": 5.64779304444109e-05, "loss": 1.044, "step": 174015 }, { "epoch": 0.44, "learning_rate": 5.647667220292919e-05, "loss": 1.0446, "step": 174020 }, { "epoch": 0.44, "learning_rate": 5.647541396144749e-05, "loss": 1.0456, "step": 174025 }, { "epoch": 0.44, "learning_rate": 5.647415571996578e-05, "loss": 1.0903, "step": 174030 }, { "epoch": 0.44, "learning_rate": 5.647289747848408e-05, "loss": 1.047, "step": 174035 }, { "epoch": 0.44, "learning_rate": 5.647163923700237e-05, "loss": 1.0484, "step": 174040 }, { "epoch": 0.44, "learning_rate": 5.647038099552067e-05, "loss": 1.0491, "step": 174045 }, { "epoch": 0.44, "learning_rate": 5.646912275403896e-05, "loss": 1.0443, "step": 174050 }, { "epoch": 0.44, "learning_rate": 5.646786451255726e-05, "loss": 1.0478, "step": 174055 }, { "epoch": 0.44, "learning_rate": 5.646660627107555e-05, "loss": 1.0421, "step": 174060 }, { "epoch": 0.44, "learning_rate": 5.646534802959385e-05, "loss": 1.0477, "step": 174065 }, { "epoch": 0.44, "learning_rate": 5.646408978811214e-05, "loss": 1.0445, "step": 174070 }, { "epoch": 0.44, "learning_rate": 5.646283154663044e-05, "loss": 1.0454, "step": 174075 }, { "epoch": 0.44, "learning_rate": 5.646157330514873e-05, "loss": 1.0474, "step": 174080 }, { "epoch": 0.44, "learning_rate": 5.646031506366702e-05, "loss": 1.0454, "step": 174085 }, { "epoch": 0.44, "learning_rate": 5.645905682218532e-05, "loss": 1.0445, "step": 174090 }, { "epoch": 0.44, "learning_rate": 5.645779858070361e-05, "loss": 1.0461, "step": 174095 }, { "epoch": 0.44, "learning_rate": 5.645654033922191e-05, "loss": 1.045, "step": 174100 }, { "epoch": 0.44, "learning_rate": 5.64552820977402e-05, "loss": 1.0484, "step": 174105 }, { "epoch": 0.44, "learning_rate": 5.64540238562585e-05, "loss": 1.0477, "step": 174110 }, { "epoch": 0.44, "learning_rate": 5.645276561477679e-05, "loss": 1.0439, "step": 174115 }, { "epoch": 0.44, "learning_rate": 5.645150737329509e-05, "loss": 1.0456, "step": 174120 }, { "epoch": 0.44, "learning_rate": 5.645024913181338e-05, "loss": 1.0445, "step": 174125 }, { "epoch": 0.44, "learning_rate": 5.644899089033168e-05, "loss": 1.047, "step": 174130 }, { "epoch": 0.44, "learning_rate": 5.644773264884997e-05, "loss": 1.0463, "step": 174135 }, { "epoch": 0.44, "learning_rate": 5.644647440736826e-05, "loss": 1.0469, "step": 174140 }, { "epoch": 0.44, "learning_rate": 5.644521616588656e-05, "loss": 1.0483, "step": 174145 }, { "epoch": 0.44, "learning_rate": 5.644395792440485e-05, "loss": 1.0467, "step": 174150 }, { "epoch": 0.44, "learning_rate": 5.644269968292315e-05, "loss": 1.0478, "step": 174155 }, { "epoch": 0.44, "learning_rate": 5.644144144144144e-05, "loss": 1.0434, "step": 174160 }, { "epoch": 0.44, "learning_rate": 5.644018319995974e-05, "loss": 1.0475, "step": 174165 }, { "epoch": 0.44, "learning_rate": 5.643892495847803e-05, "loss": 1.0486, "step": 174170 }, { "epoch": 0.44, "learning_rate": 5.643766671699633e-05, "loss": 1.0476, "step": 174175 }, { "epoch": 0.44, "learning_rate": 5.643640847551462e-05, "loss": 1.0472, "step": 174180 }, { "epoch": 0.44, "learning_rate": 5.643515023403292e-05, "loss": 1.0453, "step": 174185 }, { "epoch": 0.44, "learning_rate": 5.643389199255121e-05, "loss": 1.0487, "step": 174190 }, { "epoch": 0.44, "learning_rate": 5.643263375106951e-05, "loss": 1.046, "step": 174195 }, { "epoch": 0.44, "learning_rate": 5.64313755095878e-05, "loss": 1.0476, "step": 174200 }, { "epoch": 0.44, "learning_rate": 5.6430117268106094e-05, "loss": 1.0471, "step": 174205 }, { "epoch": 0.44, "learning_rate": 5.642885902662439e-05, "loss": 1.0466, "step": 174210 }, { "epoch": 0.44, "learning_rate": 5.6427600785142684e-05, "loss": 1.0442, "step": 174215 }, { "epoch": 0.44, "learning_rate": 5.642634254366098e-05, "loss": 1.0449, "step": 174220 }, { "epoch": 0.44, "learning_rate": 5.6425084302179274e-05, "loss": 1.0684, "step": 174225 }, { "epoch": 0.44, "learning_rate": 5.642382606069757e-05, "loss": 1.0484, "step": 174230 }, { "epoch": 0.44, "learning_rate": 5.6422567819215864e-05, "loss": 1.0457, "step": 174235 }, { "epoch": 0.44, "learning_rate": 5.642130957773416e-05, "loss": 1.0429, "step": 174240 }, { "epoch": 0.44, "learning_rate": 5.6420051336252454e-05, "loss": 1.0487, "step": 174245 }, { "epoch": 0.44, "learning_rate": 5.641879309477075e-05, "loss": 1.0452, "step": 174250 }, { "epoch": 0.44, "learning_rate": 5.6417534853289044e-05, "loss": 1.0456, "step": 174255 }, { "epoch": 0.44, "learning_rate": 5.641627661180734e-05, "loss": 1.0433, "step": 174260 }, { "epoch": 0.44, "learning_rate": 5.6415018370325634e-05, "loss": 1.0447, "step": 174265 }, { "epoch": 0.44, "learning_rate": 5.6413760128843926e-05, "loss": 1.0456, "step": 174270 }, { "epoch": 0.44, "learning_rate": 5.6412501887362224e-05, "loss": 1.0445, "step": 174275 }, { "epoch": 0.44, "learning_rate": 5.6411243645880516e-05, "loss": 1.0442, "step": 174280 }, { "epoch": 0.44, "learning_rate": 5.6409985404398814e-05, "loss": 1.0444, "step": 174285 }, { "epoch": 0.44, "learning_rate": 5.6408727162917106e-05, "loss": 1.0474, "step": 174290 }, { "epoch": 0.44, "learning_rate": 5.6407468921435404e-05, "loss": 1.0458, "step": 174295 }, { "epoch": 0.44, "learning_rate": 5.6406210679953695e-05, "loss": 1.0456, "step": 174300 }, { "epoch": 0.44, "learning_rate": 5.6404952438471994e-05, "loss": 1.0651, "step": 174305 }, { "epoch": 0.44, "learning_rate": 5.6403694196990285e-05, "loss": 1.0479, "step": 174310 }, { "epoch": 0.44, "learning_rate": 5.6402435955508584e-05, "loss": 1.0448, "step": 174315 }, { "epoch": 0.44, "learning_rate": 5.6401177714026875e-05, "loss": 1.0477, "step": 174320 }, { "epoch": 0.44, "learning_rate": 5.6399919472545174e-05, "loss": 1.0469, "step": 174325 }, { "epoch": 0.44, "learning_rate": 5.6398661231063465e-05, "loss": 1.0442, "step": 174330 }, { "epoch": 0.44, "learning_rate": 5.639740298958176e-05, "loss": 1.0424, "step": 174335 }, { "epoch": 0.44, "learning_rate": 5.6396144748100055e-05, "loss": 1.0451, "step": 174340 }, { "epoch": 0.44, "learning_rate": 5.639488650661835e-05, "loss": 1.0473, "step": 174345 }, { "epoch": 0.44, "learning_rate": 5.6393628265136645e-05, "loss": 1.0452, "step": 174350 }, { "epoch": 0.44, "learning_rate": 5.639237002365494e-05, "loss": 1.0476, "step": 174355 }, { "epoch": 0.44, "learning_rate": 5.6391111782173235e-05, "loss": 1.0462, "step": 174360 }, { "epoch": 0.44, "learning_rate": 5.638985354069153e-05, "loss": 1.0468, "step": 174365 }, { "epoch": 0.44, "learning_rate": 5.6388595299209825e-05, "loss": 1.0447, "step": 174370 }, { "epoch": 0.44, "learning_rate": 5.638733705772812e-05, "loss": 1.045, "step": 174375 }, { "epoch": 0.44, "learning_rate": 5.6386078816246415e-05, "loss": 1.0421, "step": 174380 }, { "epoch": 0.44, "learning_rate": 5.638482057476471e-05, "loss": 1.0468, "step": 174385 }, { "epoch": 0.44, "learning_rate": 5.6383562333283005e-05, "loss": 1.0511, "step": 174390 }, { "epoch": 0.44, "learning_rate": 5.63823040918013e-05, "loss": 1.0444, "step": 174395 }, { "epoch": 0.44, "learning_rate": 5.638104585031959e-05, "loss": 1.0441, "step": 174400 }, { "epoch": 0.44, "learning_rate": 5.637978760883789e-05, "loss": 1.0562, "step": 174405 }, { "epoch": 0.44, "learning_rate": 5.637852936735618e-05, "loss": 1.047, "step": 174410 }, { "epoch": 0.44, "learning_rate": 5.637727112587448e-05, "loss": 1.0445, "step": 174415 }, { "epoch": 0.44, "learning_rate": 5.637601288439277e-05, "loss": 1.0501, "step": 174420 }, { "epoch": 0.44, "learning_rate": 5.637475464291107e-05, "loss": 1.0482, "step": 174425 }, { "epoch": 0.44, "learning_rate": 5.637349640142936e-05, "loss": 1.0459, "step": 174430 }, { "epoch": 0.44, "learning_rate": 5.637223815994766e-05, "loss": 1.0486, "step": 174435 }, { "epoch": 0.44, "learning_rate": 5.637097991846595e-05, "loss": 1.0471, "step": 174440 }, { "epoch": 0.44, "learning_rate": 5.636972167698425e-05, "loss": 1.0475, "step": 174445 }, { "epoch": 0.44, "learning_rate": 5.636846343550255e-05, "loss": 1.0469, "step": 174450 }, { "epoch": 0.44, "learning_rate": 5.6367205194020843e-05, "loss": 1.0449, "step": 174455 }, { "epoch": 0.44, "learning_rate": 5.636594695253914e-05, "loss": 1.0437, "step": 174460 }, { "epoch": 0.44, "learning_rate": 5.6364688711057433e-05, "loss": 1.0684, "step": 174465 }, { "epoch": 0.44, "learning_rate": 5.636343046957573e-05, "loss": 1.0465, "step": 174470 }, { "epoch": 0.44, "learning_rate": 5.636217222809402e-05, "loss": 1.0451, "step": 174475 }, { "epoch": 0.44, "learning_rate": 5.6360913986612315e-05, "loss": 1.0483, "step": 174480 }, { "epoch": 0.44, "learning_rate": 5.635965574513061e-05, "loss": 1.0484, "step": 174485 }, { "epoch": 0.44, "learning_rate": 5.6358397503648905e-05, "loss": 1.0472, "step": 174490 }, { "epoch": 0.44, "learning_rate": 5.63571392621672e-05, "loss": 1.0431, "step": 174495 }, { "epoch": 0.44, "learning_rate": 5.6355881020685495e-05, "loss": 1.0463, "step": 174500 }, { "epoch": 0.44, "learning_rate": 5.635462277920379e-05, "loss": 1.045, "step": 174505 }, { "epoch": 0.44, "learning_rate": 5.6353364537722085e-05, "loss": 1.048, "step": 174510 }, { "epoch": 0.44, "learning_rate": 5.635210629624038e-05, "loss": 1.0437, "step": 174515 }, { "epoch": 0.44, "learning_rate": 5.6350848054758675e-05, "loss": 1.0469, "step": 174520 }, { "epoch": 0.44, "learning_rate": 5.634958981327697e-05, "loss": 1.0421, "step": 174525 }, { "epoch": 0.44, "learning_rate": 5.6348331571795265e-05, "loss": 1.046, "step": 174530 }, { "epoch": 0.44, "learning_rate": 5.634707333031356e-05, "loss": 1.0458, "step": 174535 }, { "epoch": 0.44, "learning_rate": 5.6345815088831855e-05, "loss": 1.047, "step": 174540 }, { "epoch": 0.44, "learning_rate": 5.6344556847350146e-05, "loss": 1.0454, "step": 174545 }, { "epoch": 0.44, "learning_rate": 5.6343298605868445e-05, "loss": 1.0451, "step": 174550 }, { "epoch": 0.44, "learning_rate": 5.6342040364386736e-05, "loss": 1.0455, "step": 174555 }, { "epoch": 0.44, "learning_rate": 5.6340782122905035e-05, "loss": 1.0434, "step": 174560 }, { "epoch": 0.44, "learning_rate": 5.6339523881423326e-05, "loss": 1.0461, "step": 174565 }, { "epoch": 0.44, "learning_rate": 5.6338265639941625e-05, "loss": 1.0442, "step": 174570 }, { "epoch": 0.44, "learning_rate": 5.6337007398459916e-05, "loss": 1.0439, "step": 174575 }, { "epoch": 0.44, "learning_rate": 5.6335749156978215e-05, "loss": 1.0443, "step": 174580 }, { "epoch": 0.44, "learning_rate": 5.6334490915496506e-05, "loss": 1.0452, "step": 174585 }, { "epoch": 0.44, "learning_rate": 5.6333232674014805e-05, "loss": 1.0464, "step": 174590 }, { "epoch": 0.44, "learning_rate": 5.6331974432533096e-05, "loss": 1.0469, "step": 174595 }, { "epoch": 0.44, "learning_rate": 5.6330716191051395e-05, "loss": 1.0491, "step": 174600 }, { "epoch": 0.44, "learning_rate": 5.6329457949569686e-05, "loss": 1.0441, "step": 174605 }, { "epoch": 0.44, "learning_rate": 5.632819970808798e-05, "loss": 1.0685, "step": 174610 }, { "epoch": 0.44, "learning_rate": 5.6326941466606276e-05, "loss": 1.0484, "step": 174615 }, { "epoch": 0.44, "learning_rate": 5.632568322512457e-05, "loss": 1.0466, "step": 174620 }, { "epoch": 0.44, "learning_rate": 5.6324424983642866e-05, "loss": 1.0472, "step": 174625 }, { "epoch": 0.44, "learning_rate": 5.632316674216116e-05, "loss": 1.043, "step": 174630 }, { "epoch": 0.44, "learning_rate": 5.6321908500679456e-05, "loss": 1.0456, "step": 174635 }, { "epoch": 0.44, "learning_rate": 5.632065025919775e-05, "loss": 1.0655, "step": 174640 }, { "epoch": 0.44, "learning_rate": 5.6319392017716046e-05, "loss": 1.0475, "step": 174645 }, { "epoch": 0.44, "learning_rate": 5.631813377623434e-05, "loss": 1.0424, "step": 174650 }, { "epoch": 0.44, "learning_rate": 5.6316875534752636e-05, "loss": 1.044, "step": 174655 }, { "epoch": 0.44, "learning_rate": 5.631561729327093e-05, "loss": 1.0456, "step": 174660 }, { "epoch": 0.44, "learning_rate": 5.6314359051789226e-05, "loss": 1.0483, "step": 174665 }, { "epoch": 0.44, "learning_rate": 5.631310081030752e-05, "loss": 1.0467, "step": 174670 }, { "epoch": 0.44, "learning_rate": 5.631184256882581e-05, "loss": 1.0477, "step": 174675 }, { "epoch": 0.44, "learning_rate": 5.631058432734411e-05, "loss": 1.0432, "step": 174680 }, { "epoch": 0.44, "learning_rate": 5.63093260858624e-05, "loss": 1.0661, "step": 174685 }, { "epoch": 0.44, "learning_rate": 5.63080678443807e-05, "loss": 1.0468, "step": 174690 }, { "epoch": 0.44, "learning_rate": 5.630680960289899e-05, "loss": 1.0462, "step": 174695 }, { "epoch": 0.44, "learning_rate": 5.630555136141729e-05, "loss": 1.0471, "step": 174700 }, { "epoch": 0.44, "learning_rate": 5.630429311993558e-05, "loss": 1.0445, "step": 174705 }, { "epoch": 0.44, "learning_rate": 5.630303487845388e-05, "loss": 1.0463, "step": 174710 }, { "epoch": 0.44, "learning_rate": 5.630177663697217e-05, "loss": 1.0478, "step": 174715 }, { "epoch": 0.44, "learning_rate": 5.630051839549047e-05, "loss": 1.0493, "step": 174720 }, { "epoch": 0.44, "learning_rate": 5.629926015400876e-05, "loss": 1.0452, "step": 174725 }, { "epoch": 0.44, "learning_rate": 5.629800191252706e-05, "loss": 1.0439, "step": 174730 }, { "epoch": 0.44, "learning_rate": 5.629674367104535e-05, "loss": 1.0467, "step": 174735 }, { "epoch": 0.44, "learning_rate": 5.629548542956364e-05, "loss": 1.0701, "step": 174740 }, { "epoch": 0.44, "learning_rate": 5.629422718808194e-05, "loss": 1.0458, "step": 174745 }, { "epoch": 0.44, "learning_rate": 5.629296894660023e-05, "loss": 1.0449, "step": 174750 }, { "epoch": 0.44, "learning_rate": 5.629171070511853e-05, "loss": 1.0504, "step": 174755 }, { "epoch": 0.44, "learning_rate": 5.629045246363682e-05, "loss": 1.0472, "step": 174760 }, { "epoch": 0.44, "learning_rate": 5.628919422215512e-05, "loss": 1.0494, "step": 174765 }, { "epoch": 0.44, "learning_rate": 5.628793598067341e-05, "loss": 1.0684, "step": 174770 }, { "epoch": 0.44, "learning_rate": 5.628667773919171e-05, "loss": 1.0469, "step": 174775 }, { "epoch": 0.44, "learning_rate": 5.628541949771e-05, "loss": 1.0502, "step": 174780 }, { "epoch": 0.44, "learning_rate": 5.62841612562283e-05, "loss": 1.0458, "step": 174785 }, { "epoch": 0.44, "learning_rate": 5.628290301474659e-05, "loss": 1.0486, "step": 174790 }, { "epoch": 0.44, "learning_rate": 5.628164477326488e-05, "loss": 1.0437, "step": 174795 }, { "epoch": 0.44, "learning_rate": 5.628038653178318e-05, "loss": 1.0451, "step": 174800 }, { "epoch": 0.44, "learning_rate": 5.627912829030147e-05, "loss": 1.0462, "step": 174805 }, { "epoch": 0.44, "learning_rate": 5.627787004881977e-05, "loss": 1.0459, "step": 174810 }, { "epoch": 0.44, "learning_rate": 5.627661180733806e-05, "loss": 1.0488, "step": 174815 }, { "epoch": 0.44, "learning_rate": 5.627535356585636e-05, "loss": 1.0479, "step": 174820 }, { "epoch": 0.44, "learning_rate": 5.627409532437465e-05, "loss": 1.0475, "step": 174825 }, { "epoch": 0.44, "learning_rate": 5.627283708289295e-05, "loss": 1.0464, "step": 174830 }, { "epoch": 0.44, "learning_rate": 5.627157884141124e-05, "loss": 1.0469, "step": 174835 }, { "epoch": 0.44, "learning_rate": 5.627032059992954e-05, "loss": 1.0454, "step": 174840 }, { "epoch": 0.44, "learning_rate": 5.626906235844783e-05, "loss": 1.059, "step": 174845 }, { "epoch": 0.44, "learning_rate": 5.626780411696613e-05, "loss": 1.0482, "step": 174850 }, { "epoch": 0.44, "learning_rate": 5.626654587548442e-05, "loss": 1.0494, "step": 174855 }, { "epoch": 0.44, "learning_rate": 5.6265287634002714e-05, "loss": 1.0447, "step": 174860 }, { "epoch": 0.44, "learning_rate": 5.626402939252101e-05, "loss": 1.0597, "step": 174865 }, { "epoch": 0.44, "learning_rate": 5.6262771151039304e-05, "loss": 1.0448, "step": 174870 }, { "epoch": 0.44, "learning_rate": 5.62615129095576e-05, "loss": 1.0462, "step": 174875 }, { "epoch": 0.44, "learning_rate": 5.6260254668075893e-05, "loss": 1.0473, "step": 174880 }, { "epoch": 0.44, "learning_rate": 5.625899642659419e-05, "loss": 1.045, "step": 174885 }, { "epoch": 0.44, "learning_rate": 5.6257738185112483e-05, "loss": 1.0485, "step": 174890 }, { "epoch": 0.44, "learning_rate": 5.625647994363078e-05, "loss": 1.0489, "step": 174895 }, { "epoch": 0.44, "learning_rate": 5.6255221702149073e-05, "loss": 1.0442, "step": 174900 }, { "epoch": 0.44, "learning_rate": 5.625396346066737e-05, "loss": 1.0444, "step": 174905 }, { "epoch": 0.44, "learning_rate": 5.6252705219185663e-05, "loss": 1.0459, "step": 174910 }, { "epoch": 0.44, "learning_rate": 5.625144697770396e-05, "loss": 1.0468, "step": 174915 }, { "epoch": 0.44, "learning_rate": 5.625018873622225e-05, "loss": 1.0434, "step": 174920 }, { "epoch": 0.44, "learning_rate": 5.6248930494740545e-05, "loss": 1.0452, "step": 174925 }, { "epoch": 0.44, "learning_rate": 5.624767225325884e-05, "loss": 1.0449, "step": 174930 }, { "epoch": 0.44, "learning_rate": 5.6246414011777135e-05, "loss": 1.0459, "step": 174935 }, { "epoch": 0.44, "learning_rate": 5.624515577029543e-05, "loss": 1.0472, "step": 174940 }, { "epoch": 0.44, "learning_rate": 5.6243897528813725e-05, "loss": 1.0459, "step": 174945 }, { "epoch": 0.44, "learning_rate": 5.624263928733203e-05, "loss": 1.0436, "step": 174950 }, { "epoch": 0.44, "learning_rate": 5.624138104585033e-05, "loss": 1.0456, "step": 174955 }, { "epoch": 0.44, "learning_rate": 5.624012280436862e-05, "loss": 1.0449, "step": 174960 }, { "epoch": 0.44, "learning_rate": 5.623886456288692e-05, "loss": 1.0479, "step": 174965 }, { "epoch": 0.44, "learning_rate": 5.623760632140521e-05, "loss": 1.0483, "step": 174970 }, { "epoch": 0.44, "learning_rate": 5.623634807992351e-05, "loss": 1.0468, "step": 174975 }, { "epoch": 0.44, "learning_rate": 5.62350898384418e-05, "loss": 1.0438, "step": 174980 }, { "epoch": 0.44, "learning_rate": 5.62338315969601e-05, "loss": 1.0462, "step": 174985 }, { "epoch": 0.44, "learning_rate": 5.623257335547839e-05, "loss": 1.0453, "step": 174990 }, { "epoch": 0.44, "learning_rate": 5.623131511399669e-05, "loss": 1.0462, "step": 174995 }, { "epoch": 0.44, "learning_rate": 5.623005687251498e-05, "loss": 1.0444, "step": 175000 }, { "epoch": 0.44, "learning_rate": 5.622879863103327e-05, "loss": 1.0454, "step": 175005 }, { "epoch": 0.44, "learning_rate": 5.622754038955157e-05, "loss": 1.0446, "step": 175010 }, { "epoch": 0.44, "learning_rate": 5.622628214806986e-05, "loss": 1.044, "step": 175015 }, { "epoch": 0.44, "learning_rate": 5.622502390658816e-05, "loss": 1.0456, "step": 175020 }, { "epoch": 0.44, "learning_rate": 5.622376566510645e-05, "loss": 1.0482, "step": 175025 }, { "epoch": 0.44, "learning_rate": 5.622250742362475e-05, "loss": 1.0485, "step": 175030 }, { "epoch": 0.44, "learning_rate": 5.622124918214304e-05, "loss": 1.0449, "step": 175035 }, { "epoch": 0.44, "learning_rate": 5.621999094066134e-05, "loss": 1.0458, "step": 175040 }, { "epoch": 0.44, "learning_rate": 5.621873269917963e-05, "loss": 1.0457, "step": 175045 }, { "epoch": 0.44, "learning_rate": 5.621747445769793e-05, "loss": 1.0455, "step": 175050 }, { "epoch": 0.44, "learning_rate": 5.621621621621622e-05, "loss": 1.045, "step": 175055 }, { "epoch": 0.44, "learning_rate": 5.621495797473452e-05, "loss": 1.0492, "step": 175060 }, { "epoch": 0.44, "learning_rate": 5.621369973325281e-05, "loss": 1.0463, "step": 175065 }, { "epoch": 0.44, "learning_rate": 5.62124414917711e-05, "loss": 1.0431, "step": 175070 }, { "epoch": 0.44, "learning_rate": 5.62111832502894e-05, "loss": 1.0457, "step": 175075 }, { "epoch": 0.44, "learning_rate": 5.620992500880769e-05, "loss": 1.0658, "step": 175080 }, { "epoch": 0.44, "learning_rate": 5.620866676732599e-05, "loss": 1.0452, "step": 175085 }, { "epoch": 0.44, "learning_rate": 5.620740852584428e-05, "loss": 1.0468, "step": 175090 }, { "epoch": 0.44, "learning_rate": 5.620615028436258e-05, "loss": 1.0625, "step": 175095 }, { "epoch": 0.44, "learning_rate": 5.620489204288087e-05, "loss": 1.0467, "step": 175100 }, { "epoch": 0.44, "learning_rate": 5.620363380139917e-05, "loss": 1.0667, "step": 175105 }, { "epoch": 0.44, "learning_rate": 5.620237555991746e-05, "loss": 1.0496, "step": 175110 }, { "epoch": 0.44, "learning_rate": 5.620111731843576e-05, "loss": 1.0477, "step": 175115 }, { "epoch": 0.44, "learning_rate": 5.619985907695405e-05, "loss": 1.0432, "step": 175120 }, { "epoch": 0.44, "learning_rate": 5.619860083547235e-05, "loss": 1.0489, "step": 175125 }, { "epoch": 0.44, "learning_rate": 5.619734259399064e-05, "loss": 1.0431, "step": 175130 }, { "epoch": 0.44, "learning_rate": 5.6196084352508934e-05, "loss": 1.0436, "step": 175135 }, { "epoch": 0.44, "learning_rate": 5.619482611102723e-05, "loss": 1.0471, "step": 175140 }, { "epoch": 0.44, "learning_rate": 5.6193567869545524e-05, "loss": 1.0472, "step": 175145 }, { "epoch": 0.44, "learning_rate": 5.619230962806382e-05, "loss": 1.0462, "step": 175150 }, { "epoch": 0.44, "learning_rate": 5.6191051386582114e-05, "loss": 1.0466, "step": 175155 }, { "epoch": 0.44, "learning_rate": 5.618979314510041e-05, "loss": 1.0476, "step": 175160 }, { "epoch": 0.44, "learning_rate": 5.6188534903618704e-05, "loss": 1.0424, "step": 175165 }, { "epoch": 0.44, "learning_rate": 5.6187276662137e-05, "loss": 1.0472, "step": 175170 }, { "epoch": 0.44, "learning_rate": 5.6186018420655294e-05, "loss": 1.0434, "step": 175175 }, { "epoch": 0.44, "learning_rate": 5.618476017917359e-05, "loss": 1.0464, "step": 175180 }, { "epoch": 0.44, "learning_rate": 5.6183501937691884e-05, "loss": 1.0449, "step": 175185 }, { "epoch": 0.44, "learning_rate": 5.618224369621018e-05, "loss": 1.0441, "step": 175190 }, { "epoch": 0.44, "learning_rate": 5.6180985454728474e-05, "loss": 1.0487, "step": 175195 }, { "epoch": 0.44, "learning_rate": 5.6179727213246766e-05, "loss": 1.0435, "step": 175200 }, { "epoch": 0.44, "learning_rate": 5.6178468971765064e-05, "loss": 1.0475, "step": 175205 }, { "epoch": 0.44, "learning_rate": 5.6177210730283356e-05, "loss": 1.0444, "step": 175210 }, { "epoch": 0.44, "learning_rate": 5.6175952488801654e-05, "loss": 1.043, "step": 175215 }, { "epoch": 0.44, "learning_rate": 5.6174694247319946e-05, "loss": 1.046, "step": 175220 }, { "epoch": 0.44, "learning_rate": 5.6173436005838244e-05, "loss": 1.0492, "step": 175225 }, { "epoch": 0.44, "learning_rate": 5.6172177764356536e-05, "loss": 1.0467, "step": 175230 }, { "epoch": 0.44, "learning_rate": 5.6170919522874834e-05, "loss": 1.0625, "step": 175235 }, { "epoch": 0.44, "learning_rate": 5.6169661281393126e-05, "loss": 1.044, "step": 175240 }, { "epoch": 0.44, "learning_rate": 5.6168403039911424e-05, "loss": 1.0502, "step": 175245 }, { "epoch": 0.44, "learning_rate": 5.6167144798429716e-05, "loss": 1.0678, "step": 175250 }, { "epoch": 0.44, "learning_rate": 5.616613820524436e-05, "loss": 1.0427, "step": 175255 }, { "epoch": 0.44, "learning_rate": 5.616487996376265e-05, "loss": 1.0449, "step": 175260 }, { "epoch": 0.44, "learning_rate": 5.616362172228095e-05, "loss": 1.0461, "step": 175265 }, { "epoch": 0.44, "learning_rate": 5.616236348079924e-05, "loss": 1.0462, "step": 175270 }, { "epoch": 0.44, "learning_rate": 5.616110523931754e-05, "loss": 1.0464, "step": 175275 }, { "epoch": 0.44, "learning_rate": 5.615984699783583e-05, "loss": 1.0478, "step": 175280 }, { "epoch": 0.44, "learning_rate": 5.615858875635412e-05, "loss": 1.0424, "step": 175285 }, { "epoch": 0.44, "learning_rate": 5.615733051487242e-05, "loss": 1.0463, "step": 175290 }, { "epoch": 0.44, "learning_rate": 5.615607227339071e-05, "loss": 1.0464, "step": 175295 }, { "epoch": 0.44, "learning_rate": 5.615481403190901e-05, "loss": 1.046, "step": 175300 }, { "epoch": 0.44, "learning_rate": 5.61535557904273e-05, "loss": 1.0417, "step": 175305 }, { "epoch": 0.44, "learning_rate": 5.61522975489456e-05, "loss": 1.0447, "step": 175310 }, { "epoch": 0.44, "learning_rate": 5.615103930746389e-05, "loss": 1.0465, "step": 175315 }, { "epoch": 0.44, "learning_rate": 5.614978106598219e-05, "loss": 1.0437, "step": 175320 }, { "epoch": 0.44, "learning_rate": 5.614877447279682e-05, "loss": 1.0682, "step": 175325 }, { "epoch": 0.44, "learning_rate": 5.614751623131511e-05, "loss": 1.0441, "step": 175330 }, { "epoch": 0.44, "learning_rate": 5.614625798983341e-05, "loss": 1.0472, "step": 175335 }, { "epoch": 0.44, "learning_rate": 5.61449997483517e-05, "loss": 1.0455, "step": 175340 }, { "epoch": 0.44, "learning_rate": 5.614374150687e-05, "loss": 1.0442, "step": 175345 }, { "epoch": 0.44, "learning_rate": 5.6142483265388304e-05, "loss": 1.0443, "step": 175350 }, { "epoch": 0.44, "learning_rate": 5.6141225023906596e-05, "loss": 1.047, "step": 175355 }, { "epoch": 0.44, "learning_rate": 5.6139966782424894e-05, "loss": 1.0432, "step": 175360 }, { "epoch": 0.44, "learning_rate": 5.6138708540943186e-05, "loss": 1.0436, "step": 175365 }, { "epoch": 0.44, "learning_rate": 5.613745029946148e-05, "loss": 1.0454, "step": 175370 }, { "epoch": 0.44, "learning_rate": 5.6136192057979776e-05, "loss": 1.0456, "step": 175375 }, { "epoch": 0.44, "learning_rate": 5.613493381649807e-05, "loss": 1.0459, "step": 175380 }, { "epoch": 0.44, "learning_rate": 5.6133675575016366e-05, "loss": 1.0524, "step": 175385 }, { "epoch": 0.44, "learning_rate": 5.613241733353466e-05, "loss": 1.0455, "step": 175390 }, { "epoch": 0.44, "learning_rate": 5.6131159092052956e-05, "loss": 1.0464, "step": 175395 }, { "epoch": 0.44, "learning_rate": 5.612990085057125e-05, "loss": 1.0455, "step": 175400 }, { "epoch": 0.44, "learning_rate": 5.6128642609089546e-05, "loss": 1.0428, "step": 175405 }, { "epoch": 0.44, "learning_rate": 5.612738436760784e-05, "loss": 1.0428, "step": 175410 }, { "epoch": 0.44, "learning_rate": 5.6126126126126136e-05, "loss": 1.0449, "step": 175415 }, { "epoch": 0.44, "learning_rate": 5.612486788464443e-05, "loss": 1.0435, "step": 175420 }, { "epoch": 0.44, "learning_rate": 5.6123609643162726e-05, "loss": 1.0455, "step": 175425 }, { "epoch": 0.44, "learning_rate": 5.612235140168102e-05, "loss": 1.0445, "step": 175430 }, { "epoch": 0.44, "learning_rate": 5.612109316019931e-05, "loss": 1.0459, "step": 175435 }, { "epoch": 0.44, "learning_rate": 5.611983491871761e-05, "loss": 1.0454, "step": 175440 }, { "epoch": 0.44, "learning_rate": 5.61185766772359e-05, "loss": 1.0422, "step": 175445 }, { "epoch": 0.44, "learning_rate": 5.61173184357542e-05, "loss": 1.0459, "step": 175450 }, { "epoch": 0.44, "learning_rate": 5.611606019427249e-05, "loss": 1.0478, "step": 175455 }, { "epoch": 0.44, "learning_rate": 5.611480195279079e-05, "loss": 1.0478, "step": 175460 }, { "epoch": 0.44, "learning_rate": 5.611354371130908e-05, "loss": 1.0669, "step": 175465 }, { "epoch": 0.44, "learning_rate": 5.611253711812371e-05, "loss": 1.0446, "step": 175470 }, { "epoch": 0.44, "learning_rate": 5.6111278876642006e-05, "loss": 1.0458, "step": 175475 }, { "epoch": 0.44, "learning_rate": 5.61100206351603e-05, "loss": 1.0473, "step": 175480 }, { "epoch": 0.44, "learning_rate": 5.6108762393678596e-05, "loss": 1.0455, "step": 175485 }, { "epoch": 0.44, "learning_rate": 5.610750415219689e-05, "loss": 1.045, "step": 175490 }, { "epoch": 0.44, "learning_rate": 5.610624591071518e-05, "loss": 1.0594, "step": 175495 }, { "epoch": 0.44, "learning_rate": 5.610498766923348e-05, "loss": 1.0647, "step": 175500 }, { "epoch": 0.44, "learning_rate": 5.610398107604812e-05, "loss": 1.0702, "step": 175505 }, { "epoch": 0.44, "learning_rate": 5.6102974482862757e-05, "loss": 1.0444, "step": 175510 }, { "epoch": 0.44, "learning_rate": 5.6101716241381055e-05, "loss": 1.0451, "step": 175515 }, { "epoch": 0.44, "learning_rate": 5.6100457999899346e-05, "loss": 1.0473, "step": 175520 }, { "epoch": 0.44, "learning_rate": 5.6099199758417645e-05, "loss": 1.0441, "step": 175525 }, { "epoch": 0.44, "learning_rate": 5.6097941516935936e-05, "loss": 1.0452, "step": 175530 }, { "epoch": 0.44, "learning_rate": 5.6096683275454235e-05, "loss": 1.0482, "step": 175535 }, { "epoch": 0.44, "learning_rate": 5.6095425033972526e-05, "loss": 1.0468, "step": 175540 }, { "epoch": 0.44, "learning_rate": 5.609416679249082e-05, "loss": 1.0478, "step": 175545 }, { "epoch": 0.44, "learning_rate": 5.6092908551009116e-05, "loss": 1.0426, "step": 175550 }, { "epoch": 0.44, "learning_rate": 5.609165030952741e-05, "loss": 1.0459, "step": 175555 }, { "epoch": 0.44, "learning_rate": 5.6090392068045706e-05, "loss": 1.0455, "step": 175560 }, { "epoch": 0.44, "learning_rate": 5.6089133826564e-05, "loss": 1.0475, "step": 175565 }, { "epoch": 0.44, "learning_rate": 5.6087875585082296e-05, "loss": 1.0452, "step": 175570 }, { "epoch": 0.44, "learning_rate": 5.608661734360059e-05, "loss": 1.0458, "step": 175575 }, { "epoch": 0.44, "learning_rate": 5.6085359102118886e-05, "loss": 1.0441, "step": 175580 }, { "epoch": 0.44, "learning_rate": 5.608410086063718e-05, "loss": 1.0452, "step": 175585 }, { "epoch": 0.44, "learning_rate": 5.6082842619155476e-05, "loss": 1.0481, "step": 175590 }, { "epoch": 0.44, "learning_rate": 5.608158437767377e-05, "loss": 1.0485, "step": 175595 }, { "epoch": 0.44, "learning_rate": 5.6080326136192066e-05, "loss": 1.0468, "step": 175600 }, { "epoch": 0.44, "learning_rate": 5.607906789471036e-05, "loss": 1.0432, "step": 175605 }, { "epoch": 0.44, "learning_rate": 5.607780965322865e-05, "loss": 1.0446, "step": 175610 }, { "epoch": 0.44, "learning_rate": 5.607655141174695e-05, "loss": 1.0442, "step": 175615 }, { "epoch": 0.44, "learning_rate": 5.607529317026524e-05, "loss": 1.0432, "step": 175620 }, { "epoch": 0.44, "learning_rate": 5.607403492878354e-05, "loss": 1.0446, "step": 175625 }, { "epoch": 0.44, "learning_rate": 5.607277668730183e-05, "loss": 1.0474, "step": 175630 }, { "epoch": 0.44, "learning_rate": 5.607151844582013e-05, "loss": 1.0443, "step": 175635 }, { "epoch": 0.44, "learning_rate": 5.607026020433842e-05, "loss": 1.0478, "step": 175640 }, { "epoch": 0.44, "learning_rate": 5.606900196285672e-05, "loss": 1.0469, "step": 175645 }, { "epoch": 0.44, "learning_rate": 5.606774372137501e-05, "loss": 1.0463, "step": 175650 }, { "epoch": 0.44, "learning_rate": 5.606648547989331e-05, "loss": 1.0467, "step": 175655 }, { "epoch": 0.44, "learning_rate": 5.60652272384116e-05, "loss": 1.0444, "step": 175660 }, { "epoch": 0.44, "learning_rate": 5.606396899692989e-05, "loss": 1.0479, "step": 175665 }, { "epoch": 0.44, "learning_rate": 5.606271075544819e-05, "loss": 1.0676, "step": 175670 }, { "epoch": 0.44, "learning_rate": 5.606145251396648e-05, "loss": 1.0572, "step": 175675 }, { "epoch": 0.44, "learning_rate": 5.606019427248478e-05, "loss": 1.047, "step": 175680 }, { "epoch": 0.44, "learning_rate": 5.605893603100307e-05, "loss": 1.0495, "step": 175685 }, { "epoch": 0.44, "learning_rate": 5.605767778952137e-05, "loss": 1.0408, "step": 175690 }, { "epoch": 0.44, "learning_rate": 5.605641954803966e-05, "loss": 1.0456, "step": 175695 }, { "epoch": 0.44, "learning_rate": 5.605516130655796e-05, "loss": 1.0439, "step": 175700 }, { "epoch": 0.44, "learning_rate": 5.605390306507625e-05, "loss": 1.0453, "step": 175705 }, { "epoch": 0.44, "learning_rate": 5.605264482359455e-05, "loss": 1.0458, "step": 175710 }, { "epoch": 0.44, "learning_rate": 5.605138658211284e-05, "loss": 1.0449, "step": 175715 }, { "epoch": 0.44, "learning_rate": 5.605012834063114e-05, "loss": 1.0445, "step": 175720 }, { "epoch": 0.44, "learning_rate": 5.604887009914943e-05, "loss": 1.0466, "step": 175725 }, { "epoch": 0.44, "learning_rate": 5.604761185766772e-05, "loss": 1.049, "step": 175730 }, { "epoch": 0.44, "learning_rate": 5.604635361618602e-05, "loss": 1.0472, "step": 175735 }, { "epoch": 0.44, "learning_rate": 5.604509537470431e-05, "loss": 1.0486, "step": 175740 }, { "epoch": 0.44, "learning_rate": 5.604383713322261e-05, "loss": 1.0434, "step": 175745 }, { "epoch": 0.44, "learning_rate": 5.60425788917409e-05, "loss": 1.0449, "step": 175750 }, { "epoch": 0.44, "learning_rate": 5.60413206502592e-05, "loss": 1.0459, "step": 175755 }, { "epoch": 0.44, "learning_rate": 5.604006240877749e-05, "loss": 1.046, "step": 175760 }, { "epoch": 0.44, "learning_rate": 5.603880416729579e-05, "loss": 1.0494, "step": 175765 }, { "epoch": 0.44, "learning_rate": 5.603754592581408e-05, "loss": 1.0461, "step": 175770 }, { "epoch": 0.44, "learning_rate": 5.603628768433238e-05, "loss": 1.0446, "step": 175775 }, { "epoch": 0.44, "learning_rate": 5.603502944285067e-05, "loss": 1.0487, "step": 175780 }, { "epoch": 0.44, "learning_rate": 5.603377120136897e-05, "loss": 1.0471, "step": 175785 }, { "epoch": 0.44, "learning_rate": 5.603251295988726e-05, "loss": 1.0476, "step": 175790 }, { "epoch": 0.44, "learning_rate": 5.6031254718405554e-05, "loss": 1.0465, "step": 175795 }, { "epoch": 0.44, "learning_rate": 5.602999647692385e-05, "loss": 1.0469, "step": 175800 }, { "epoch": 0.44, "learning_rate": 5.6028738235442144e-05, "loss": 1.0441, "step": 175805 }, { "epoch": 0.44, "learning_rate": 5.602747999396044e-05, "loss": 1.047, "step": 175810 }, { "epoch": 0.44, "learning_rate": 5.6026221752478734e-05, "loss": 1.0457, "step": 175815 }, { "epoch": 0.44, "learning_rate": 5.602496351099703e-05, "loss": 1.0438, "step": 175820 }, { "epoch": 0.44, "learning_rate": 5.6023705269515324e-05, "loss": 1.0485, "step": 175825 }, { "epoch": 0.44, "learning_rate": 5.602244702803362e-05, "loss": 1.044, "step": 175830 }, { "epoch": 0.44, "learning_rate": 5.6021188786551914e-05, "loss": 1.0475, "step": 175835 }, { "epoch": 0.44, "learning_rate": 5.601993054507021e-05, "loss": 1.0648, "step": 175840 }, { "epoch": 0.44, "learning_rate": 5.6018672303588504e-05, "loss": 1.0445, "step": 175845 }, { "epoch": 0.44, "learning_rate": 5.60174140621068e-05, "loss": 1.0451, "step": 175850 }, { "epoch": 0.44, "learning_rate": 5.6016155820625094e-05, "loss": 1.0444, "step": 175855 }, { "epoch": 0.44, "learning_rate": 5.6014897579143385e-05, "loss": 1.0468, "step": 175860 }, { "epoch": 0.44, "learning_rate": 5.6013639337661684e-05, "loss": 1.0473, "step": 175865 }, { "epoch": 0.44, "learning_rate": 5.6012381096179975e-05, "loss": 1.0439, "step": 175870 }, { "epoch": 0.44, "learning_rate": 5.6011122854698274e-05, "loss": 1.0422, "step": 175875 }, { "epoch": 0.44, "learning_rate": 5.6009864613216565e-05, "loss": 1.0458, "step": 175880 }, { "epoch": 0.44, "learning_rate": 5.6008606371734863e-05, "loss": 1.0427, "step": 175885 }, { "epoch": 0.44, "learning_rate": 5.6007348130253155e-05, "loss": 1.0452, "step": 175890 }, { "epoch": 0.44, "learning_rate": 5.6006089888771453e-05, "loss": 1.0472, "step": 175895 }, { "epoch": 0.44, "learning_rate": 5.6004831647289745e-05, "loss": 1.0464, "step": 175900 }, { "epoch": 0.44, "learning_rate": 5.6003573405808043e-05, "loss": 1.0431, "step": 175905 }, { "epoch": 0.44, "learning_rate": 5.6002315164326335e-05, "loss": 1.0479, "step": 175910 }, { "epoch": 0.44, "learning_rate": 5.6001056922844633e-05, "loss": 1.0464, "step": 175915 }, { "epoch": 0.44, "learning_rate": 5.5999798681362925e-05, "loss": 1.0455, "step": 175920 }, { "epoch": 0.44, "learning_rate": 5.5998540439881217e-05, "loss": 1.0461, "step": 175925 }, { "epoch": 0.44, "learning_rate": 5.5997282198399515e-05, "loss": 1.0458, "step": 175930 }, { "epoch": 0.44, "learning_rate": 5.5996023956917807e-05, "loss": 1.0451, "step": 175935 }, { "epoch": 0.44, "learning_rate": 5.5994765715436105e-05, "loss": 1.0437, "step": 175940 }, { "epoch": 0.44, "learning_rate": 5.5993507473954397e-05, "loss": 1.0423, "step": 175945 }, { "epoch": 0.44, "learning_rate": 5.59922492324727e-05, "loss": 1.0632, "step": 175950 }, { "epoch": 0.44, "learning_rate": 5.5990990990991e-05, "loss": 1.0454, "step": 175955 }, { "epoch": 0.44, "learning_rate": 5.598973274950929e-05, "loss": 1.042, "step": 175960 }, { "epoch": 0.44, "learning_rate": 5.598847450802759e-05, "loss": 1.0437, "step": 175965 }, { "epoch": 0.44, "learning_rate": 5.598721626654588e-05, "loss": 1.0471, "step": 175970 }, { "epoch": 0.44, "learning_rate": 5.598595802506418e-05, "loss": 1.0448, "step": 175975 }, { "epoch": 0.44, "learning_rate": 5.598469978358247e-05, "loss": 1.0596, "step": 175980 }, { "epoch": 0.44, "learning_rate": 5.598344154210077e-05, "loss": 1.0447, "step": 175985 }, { "epoch": 0.44, "learning_rate": 5.598218330061906e-05, "loss": 1.0459, "step": 175990 }, { "epoch": 0.44, "learning_rate": 5.598092505913736e-05, "loss": 1.0446, "step": 175995 }, { "epoch": 0.44, "learning_rate": 5.597966681765565e-05, "loss": 1.047, "step": 176000 }, { "epoch": 0.44, "learning_rate": 5.597840857617394e-05, "loss": 1.0475, "step": 176005 }, { "epoch": 0.44, "learning_rate": 5.597715033469224e-05, "loss": 1.0484, "step": 176010 }, { "epoch": 0.44, "learning_rate": 5.597589209321053e-05, "loss": 1.0475, "step": 176015 }, { "epoch": 0.44, "learning_rate": 5.597463385172883e-05, "loss": 1.0461, "step": 176020 }, { "epoch": 0.44, "learning_rate": 5.597337561024712e-05, "loss": 1.0449, "step": 176025 }, { "epoch": 0.44, "learning_rate": 5.597211736876542e-05, "loss": 1.0459, "step": 176030 }, { "epoch": 0.44, "learning_rate": 5.597085912728371e-05, "loss": 1.0478, "step": 176035 }, { "epoch": 0.44, "learning_rate": 5.596960088580201e-05, "loss": 1.0456, "step": 176040 }, { "epoch": 0.44, "learning_rate": 5.59683426443203e-05, "loss": 1.045, "step": 176045 }, { "epoch": 0.44, "learning_rate": 5.59670844028386e-05, "loss": 1.0464, "step": 176050 }, { "epoch": 0.44, "learning_rate": 5.596582616135689e-05, "loss": 1.049, "step": 176055 }, { "epoch": 0.44, "learning_rate": 5.596456791987519e-05, "loss": 1.0453, "step": 176060 }, { "epoch": 0.44, "learning_rate": 5.596330967839348e-05, "loss": 1.0476, "step": 176065 }, { "epoch": 0.44, "learning_rate": 5.5962051436911775e-05, "loss": 1.0435, "step": 176070 }, { "epoch": 0.44, "learning_rate": 5.596079319543007e-05, "loss": 1.0463, "step": 176075 }, { "epoch": 0.44, "learning_rate": 5.5959534953948365e-05, "loss": 1.044, "step": 176080 }, { "epoch": 0.44, "learning_rate": 5.595827671246666e-05, "loss": 1.0436, "step": 176085 }, { "epoch": 0.44, "learning_rate": 5.5957018470984955e-05, "loss": 1.0467, "step": 176090 }, { "epoch": 0.44, "learning_rate": 5.595576022950325e-05, "loss": 1.0476, "step": 176095 }, { "epoch": 0.44, "learning_rate": 5.5954501988021544e-05, "loss": 1.0452, "step": 176100 }, { "epoch": 0.44, "learning_rate": 5.595324374653984e-05, "loss": 1.0491, "step": 176105 }, { "epoch": 0.44, "learning_rate": 5.5951985505058134e-05, "loss": 1.0447, "step": 176110 }, { "epoch": 0.44, "learning_rate": 5.595072726357643e-05, "loss": 1.0438, "step": 176115 }, { "epoch": 0.44, "learning_rate": 5.5949469022094724e-05, "loss": 1.044, "step": 176120 }, { "epoch": 0.44, "learning_rate": 5.594821078061302e-05, "loss": 1.0446, "step": 176125 }, { "epoch": 0.44, "learning_rate": 5.5946952539131314e-05, "loss": 1.0453, "step": 176130 }, { "epoch": 0.44, "learning_rate": 5.5945694297649606e-05, "loss": 1.0475, "step": 176135 }, { "epoch": 0.44, "learning_rate": 5.5944436056167904e-05, "loss": 1.0453, "step": 176140 }, { "epoch": 0.44, "learning_rate": 5.5943177814686196e-05, "loss": 1.0444, "step": 176145 }, { "epoch": 0.44, "learning_rate": 5.5941919573204494e-05, "loss": 1.0461, "step": 176150 }, { "epoch": 0.44, "learning_rate": 5.5940661331722786e-05, "loss": 1.0443, "step": 176155 }, { "epoch": 0.44, "learning_rate": 5.5939403090241084e-05, "loss": 1.0478, "step": 176160 }, { "epoch": 0.44, "learning_rate": 5.5938144848759376e-05, "loss": 1.0477, "step": 176165 }, { "epoch": 0.44, "learning_rate": 5.5936886607277674e-05, "loss": 1.048, "step": 176170 }, { "epoch": 0.44, "learning_rate": 5.5935628365795966e-05, "loss": 1.0452, "step": 176175 }, { "epoch": 0.44, "learning_rate": 5.5934370124314264e-05, "loss": 1.0461, "step": 176180 }, { "epoch": 0.44, "learning_rate": 5.5933111882832556e-05, "loss": 1.0452, "step": 176185 }, { "epoch": 0.44, "learning_rate": 5.5931853641350854e-05, "loss": 1.0475, "step": 176190 }, { "epoch": 0.44, "learning_rate": 5.5930595399869146e-05, "loss": 1.0451, "step": 176195 }, { "epoch": 0.44, "learning_rate": 5.592933715838744e-05, "loss": 1.0483, "step": 176200 }, { "epoch": 0.44, "learning_rate": 5.5928078916905736e-05, "loss": 1.0474, "step": 176205 }, { "epoch": 0.44, "learning_rate": 5.592682067542403e-05, "loss": 1.0477, "step": 176210 }, { "epoch": 0.44, "learning_rate": 5.5925562433942326e-05, "loss": 1.0461, "step": 176215 }, { "epoch": 0.44, "learning_rate": 5.592430419246062e-05, "loss": 1.0457, "step": 176220 }, { "epoch": 0.44, "learning_rate": 5.5923045950978916e-05, "loss": 1.0436, "step": 176225 }, { "epoch": 0.44, "learning_rate": 5.592178770949721e-05, "loss": 1.043, "step": 176230 }, { "epoch": 0.44, "learning_rate": 5.5920529468015506e-05, "loss": 1.0426, "step": 176235 }, { "epoch": 0.44, "learning_rate": 5.59192712265338e-05, "loss": 1.0454, "step": 176240 }, { "epoch": 0.44, "learning_rate": 5.5918012985052096e-05, "loss": 1.0408, "step": 176245 }, { "epoch": 0.44, "learning_rate": 5.591675474357039e-05, "loss": 1.0438, "step": 176250 }, { "epoch": 0.44, "learning_rate": 5.5915496502088686e-05, "loss": 1.046, "step": 176255 }, { "epoch": 0.44, "learning_rate": 5.591423826060698e-05, "loss": 1.0446, "step": 176260 }, { "epoch": 0.44, "learning_rate": 5.591298001912527e-05, "loss": 1.0471, "step": 176265 }, { "epoch": 0.44, "learning_rate": 5.591172177764357e-05, "loss": 1.0492, "step": 176270 }, { "epoch": 0.44, "learning_rate": 5.591046353616186e-05, "loss": 1.0452, "step": 176275 }, { "epoch": 0.44, "learning_rate": 5.590920529468016e-05, "loss": 1.0455, "step": 176280 }, { "epoch": 0.44, "learning_rate": 5.590794705319845e-05, "loss": 1.0458, "step": 176285 }, { "epoch": 0.44, "learning_rate": 5.590668881171675e-05, "loss": 1.0454, "step": 176290 }, { "epoch": 0.44, "learning_rate": 5.590543057023504e-05, "loss": 1.0446, "step": 176295 }, { "epoch": 0.44, "learning_rate": 5.590417232875334e-05, "loss": 1.0444, "step": 176300 }, { "epoch": 0.44, "learning_rate": 5.590291408727163e-05, "loss": 1.0704, "step": 176305 }, { "epoch": 0.44, "learning_rate": 5.590165584578993e-05, "loss": 1.0646, "step": 176310 }, { "epoch": 0.44, "learning_rate": 5.590039760430822e-05, "loss": 1.0446, "step": 176315 }, { "epoch": 0.44, "learning_rate": 5.589913936282651e-05, "loss": 1.0443, "step": 176320 }, { "epoch": 0.44, "learning_rate": 5.589788112134481e-05, "loss": 1.0474, "step": 176325 }, { "epoch": 0.44, "learning_rate": 5.58966228798631e-05, "loss": 1.042, "step": 176330 }, { "epoch": 0.44, "learning_rate": 5.58953646383814e-05, "loss": 1.0481, "step": 176335 }, { "epoch": 0.44, "learning_rate": 5.589410639689969e-05, "loss": 1.0466, "step": 176340 }, { "epoch": 0.44, "learning_rate": 5.589284815541799e-05, "loss": 1.0433, "step": 176345 }, { "epoch": 0.44, "learning_rate": 5.589158991393628e-05, "loss": 1.0468, "step": 176350 }, { "epoch": 0.44, "learning_rate": 5.589033167245458e-05, "loss": 1.0433, "step": 176355 }, { "epoch": 0.44, "learning_rate": 5.588907343097287e-05, "loss": 1.043, "step": 176360 }, { "epoch": 0.44, "learning_rate": 5.588781518949117e-05, "loss": 1.0422, "step": 176365 }, { "epoch": 0.44, "learning_rate": 5.588655694800946e-05, "loss": 1.0457, "step": 176370 }, { "epoch": 0.44, "learning_rate": 5.588529870652776e-05, "loss": 1.0473, "step": 176375 }, { "epoch": 0.44, "learning_rate": 5.588404046504605e-05, "loss": 1.0458, "step": 176380 }, { "epoch": 0.44, "learning_rate": 5.588278222356434e-05, "loss": 1.0688, "step": 176385 }, { "epoch": 0.44, "learning_rate": 5.588152398208264e-05, "loss": 1.046, "step": 176390 }, { "epoch": 0.44, "learning_rate": 5.588026574060093e-05, "loss": 1.0656, "step": 176395 }, { "epoch": 0.44, "learning_rate": 5.587900749911923e-05, "loss": 1.042, "step": 176400 }, { "epoch": 0.44, "learning_rate": 5.587774925763752e-05, "loss": 1.0468, "step": 176405 }, { "epoch": 0.44, "learning_rate": 5.587649101615582e-05, "loss": 1.0659, "step": 176410 }, { "epoch": 0.44, "learning_rate": 5.587523277467411e-05, "loss": 1.0471, "step": 176415 }, { "epoch": 0.44, "learning_rate": 5.587397453319241e-05, "loss": 1.0434, "step": 176420 }, { "epoch": 0.44, "learning_rate": 5.58727162917107e-05, "loss": 1.0468, "step": 176425 }, { "epoch": 0.44, "learning_rate": 5.5871458050229e-05, "loss": 1.046, "step": 176430 }, { "epoch": 0.44, "learning_rate": 5.587019980874729e-05, "loss": 1.0463, "step": 176435 }, { "epoch": 0.44, "learning_rate": 5.586894156726559e-05, "loss": 1.0478, "step": 176440 }, { "epoch": 0.44, "learning_rate": 5.586768332578388e-05, "loss": 1.0432, "step": 176445 }, { "epoch": 0.44, "learning_rate": 5.586642508430219e-05, "loss": 1.0473, "step": 176450 }, { "epoch": 0.44, "learning_rate": 5.5865166842820485e-05, "loss": 1.0693, "step": 176455 }, { "epoch": 0.44, "learning_rate": 5.586390860133878e-05, "loss": 1.0448, "step": 176460 }, { "epoch": 0.44, "learning_rate": 5.586265035985707e-05, "loss": 1.045, "step": 176465 }, { "epoch": 0.44, "learning_rate": 5.586139211837537e-05, "loss": 1.0491, "step": 176470 }, { "epoch": 0.44, "learning_rate": 5.586013387689366e-05, "loss": 1.0446, "step": 176475 }, { "epoch": 0.44, "learning_rate": 5.5858875635411957e-05, "loss": 1.0494, "step": 176480 }, { "epoch": 0.44, "learning_rate": 5.585761739393025e-05, "loss": 1.0481, "step": 176485 }, { "epoch": 0.44, "learning_rate": 5.5856359152448547e-05, "loss": 1.0423, "step": 176490 }, { "epoch": 0.44, "learning_rate": 5.585510091096684e-05, "loss": 1.0447, "step": 176495 }, { "epoch": 0.44, "learning_rate": 5.5853842669485137e-05, "loss": 1.0436, "step": 176500 }, { "epoch": 0.44, "learning_rate": 5.585258442800343e-05, "loss": 1.0483, "step": 176505 }, { "epoch": 0.44, "learning_rate": 5.5851326186521727e-05, "loss": 1.0426, "step": 176510 }, { "epoch": 0.44, "learning_rate": 5.585006794504002e-05, "loss": 1.044, "step": 176515 }, { "epoch": 0.44, "learning_rate": 5.5848809703558316e-05, "loss": 1.0465, "step": 176520 }, { "epoch": 0.44, "learning_rate": 5.584755146207661e-05, "loss": 1.0451, "step": 176525 }, { "epoch": 0.44, "learning_rate": 5.58462932205949e-05, "loss": 1.0438, "step": 176530 }, { "epoch": 0.44, "learning_rate": 5.58450349791132e-05, "loss": 1.0406, "step": 176535 }, { "epoch": 0.44, "learning_rate": 5.584377673763149e-05, "loss": 1.0589, "step": 176540 }, { "epoch": 0.44, "learning_rate": 5.584251849614979e-05, "loss": 1.0461, "step": 176545 }, { "epoch": 0.44, "learning_rate": 5.584126025466808e-05, "loss": 1.0462, "step": 176550 }, { "epoch": 0.44, "learning_rate": 5.584000201318638e-05, "loss": 1.0454, "step": 176555 }, { "epoch": 0.44, "learning_rate": 5.583874377170467e-05, "loss": 1.0438, "step": 176560 }, { "epoch": 0.44, "learning_rate": 5.583748553022297e-05, "loss": 1.0435, "step": 176565 }, { "epoch": 0.44, "learning_rate": 5.583622728874126e-05, "loss": 1.046, "step": 176570 }, { "epoch": 0.44, "learning_rate": 5.583496904725956e-05, "loss": 1.0453, "step": 176575 }, { "epoch": 0.44, "learning_rate": 5.583371080577785e-05, "loss": 1.0442, "step": 176580 }, { "epoch": 0.44, "learning_rate": 5.583245256429615e-05, "loss": 1.0406, "step": 176585 }, { "epoch": 0.44, "learning_rate": 5.583119432281444e-05, "loss": 1.0479, "step": 176590 }, { "epoch": 0.44, "learning_rate": 5.582993608133273e-05, "loss": 1.0435, "step": 176595 }, { "epoch": 0.44, "learning_rate": 5.582867783985103e-05, "loss": 1.044, "step": 176600 }, { "epoch": 0.44, "learning_rate": 5.582741959836932e-05, "loss": 1.0421, "step": 176605 }, { "epoch": 0.44, "learning_rate": 5.582616135688762e-05, "loss": 1.0446, "step": 176610 }, { "epoch": 0.44, "learning_rate": 5.582490311540591e-05, "loss": 1.0436, "step": 176615 }, { "epoch": 0.44, "learning_rate": 5.582364487392421e-05, "loss": 1.0452, "step": 176620 }, { "epoch": 0.44, "learning_rate": 5.58223866324425e-05, "loss": 1.0471, "step": 176625 }, { "epoch": 0.44, "learning_rate": 5.58211283909608e-05, "loss": 1.0448, "step": 176630 }, { "epoch": 0.44, "learning_rate": 5.581987014947909e-05, "loss": 1.0461, "step": 176635 }, { "epoch": 0.44, "learning_rate": 5.581861190799739e-05, "loss": 1.0473, "step": 176640 }, { "epoch": 0.44, "learning_rate": 5.581735366651568e-05, "loss": 1.0457, "step": 176645 }, { "epoch": 0.44, "learning_rate": 5.581609542503398e-05, "loss": 1.0445, "step": 176650 }, { "epoch": 0.44, "learning_rate": 5.581483718355227e-05, "loss": 1.0444, "step": 176655 }, { "epoch": 0.44, "learning_rate": 5.581357894207056e-05, "loss": 1.0452, "step": 176660 }, { "epoch": 0.44, "learning_rate": 5.581232070058886e-05, "loss": 1.0448, "step": 176665 }, { "epoch": 0.44, "learning_rate": 5.581106245910715e-05, "loss": 1.043, "step": 176670 }, { "epoch": 0.44, "learning_rate": 5.580980421762545e-05, "loss": 1.0408, "step": 176675 }, { "epoch": 0.44, "learning_rate": 5.580854597614374e-05, "loss": 1.046, "step": 176680 }, { "epoch": 0.44, "learning_rate": 5.580728773466204e-05, "loss": 1.0725, "step": 176685 }, { "epoch": 0.44, "learning_rate": 5.580602949318033e-05, "loss": 1.0475, "step": 176690 }, { "epoch": 0.44, "learning_rate": 5.580477125169863e-05, "loss": 1.0439, "step": 176695 }, { "epoch": 0.44, "learning_rate": 5.580351301021692e-05, "loss": 1.047, "step": 176700 }, { "epoch": 0.44, "learning_rate": 5.580225476873522e-05, "loss": 1.0446, "step": 176705 }, { "epoch": 0.44, "learning_rate": 5.580099652725351e-05, "loss": 1.047, "step": 176710 }, { "epoch": 0.44, "learning_rate": 5.579973828577181e-05, "loss": 1.0454, "step": 176715 }, { "epoch": 0.44, "learning_rate": 5.57984800442901e-05, "loss": 1.0464, "step": 176720 }, { "epoch": 0.44, "learning_rate": 5.5797221802808394e-05, "loss": 1.0484, "step": 176725 }, { "epoch": 0.44, "learning_rate": 5.579596356132669e-05, "loss": 1.0462, "step": 176730 }, { "epoch": 0.44, "learning_rate": 5.5794705319844984e-05, "loss": 1.0479, "step": 176735 }, { "epoch": 0.44, "learning_rate": 5.579344707836328e-05, "loss": 1.0449, "step": 176740 }, { "epoch": 0.44, "learning_rate": 5.5792188836881574e-05, "loss": 1.0471, "step": 176745 }, { "epoch": 0.44, "learning_rate": 5.579093059539987e-05, "loss": 1.0449, "step": 176750 }, { "epoch": 0.44, "learning_rate": 5.5789672353918164e-05, "loss": 1.0475, "step": 176755 }, { "epoch": 0.44, "learning_rate": 5.578841411243646e-05, "loss": 1.0463, "step": 176760 }, { "epoch": 0.44, "learning_rate": 5.5787155870954754e-05, "loss": 1.0455, "step": 176765 }, { "epoch": 0.44, "learning_rate": 5.578589762947305e-05, "loss": 1.0465, "step": 176770 }, { "epoch": 0.44, "learning_rate": 5.5784639387991344e-05, "loss": 1.0444, "step": 176775 }, { "epoch": 0.44, "learning_rate": 5.578338114650964e-05, "loss": 1.0473, "step": 176780 }, { "epoch": 0.44, "learning_rate": 5.5782122905027934e-05, "loss": 1.0435, "step": 176785 }, { "epoch": 0.44, "learning_rate": 5.5780864663546225e-05, "loss": 1.0432, "step": 176790 }, { "epoch": 0.44, "learning_rate": 5.5779606422064524e-05, "loss": 1.0453, "step": 176795 }, { "epoch": 0.44, "learning_rate": 5.5778348180582815e-05, "loss": 1.0443, "step": 176800 }, { "epoch": 0.44, "learning_rate": 5.5777089939101114e-05, "loss": 1.0449, "step": 176805 }, { "epoch": 0.44, "learning_rate": 5.5775831697619405e-05, "loss": 1.0449, "step": 176810 }, { "epoch": 0.44, "learning_rate": 5.5774573456137704e-05, "loss": 1.0436, "step": 176815 }, { "epoch": 0.44, "learning_rate": 5.5773315214655995e-05, "loss": 1.0469, "step": 176820 }, { "epoch": 0.44, "learning_rate": 5.5772056973174294e-05, "loss": 1.0454, "step": 176825 }, { "epoch": 0.44, "learning_rate": 5.5770798731692585e-05, "loss": 1.0483, "step": 176830 }, { "epoch": 0.44, "learning_rate": 5.5769540490210884e-05, "loss": 1.045, "step": 176835 }, { "epoch": 0.44, "learning_rate": 5.5768282248729175e-05, "loss": 1.0429, "step": 176840 }, { "epoch": 0.44, "learning_rate": 5.5767024007247474e-05, "loss": 1.0434, "step": 176845 }, { "epoch": 0.44, "learning_rate": 5.5765765765765765e-05, "loss": 1.047, "step": 176850 }, { "epoch": 0.44, "learning_rate": 5.576450752428406e-05, "loss": 1.0692, "step": 176855 }, { "epoch": 0.44, "learning_rate": 5.5763249282802355e-05, "loss": 1.045, "step": 176860 }, { "epoch": 0.44, "learning_rate": 5.576199104132065e-05, "loss": 1.0461, "step": 176865 }, { "epoch": 0.44, "learning_rate": 5.5760732799838945e-05, "loss": 1.0485, "step": 176870 }, { "epoch": 0.44, "learning_rate": 5.575947455835724e-05, "loss": 1.0483, "step": 176875 }, { "epoch": 0.44, "learning_rate": 5.5758216316875535e-05, "loss": 1.0461, "step": 176880 }, { "epoch": 0.44, "learning_rate": 5.575695807539383e-05, "loss": 1.0465, "step": 176885 }, { "epoch": 0.44, "learning_rate": 5.5755699833912125e-05, "loss": 1.0487, "step": 176890 }, { "epoch": 0.44, "learning_rate": 5.575444159243042e-05, "loss": 1.0451, "step": 176895 }, { "epoch": 0.44, "learning_rate": 5.5753183350948715e-05, "loss": 1.0433, "step": 176900 }, { "epoch": 0.44, "learning_rate": 5.575192510946701e-05, "loss": 1.0455, "step": 176905 }, { "epoch": 0.44, "learning_rate": 5.5750666867985305e-05, "loss": 1.0505, "step": 176910 }, { "epoch": 0.44, "learning_rate": 5.57494086265036e-05, "loss": 1.0443, "step": 176915 }, { "epoch": 0.44, "learning_rate": 5.574815038502189e-05, "loss": 1.0447, "step": 176920 }, { "epoch": 0.44, "learning_rate": 5.5746892143540187e-05, "loss": 1.0434, "step": 176925 }, { "epoch": 0.44, "learning_rate": 5.574563390205848e-05, "loss": 1.0479, "step": 176930 }, { "epoch": 0.44, "learning_rate": 5.5744375660576777e-05, "loss": 1.0448, "step": 176935 }, { "epoch": 0.44, "learning_rate": 5.574311741909507e-05, "loss": 1.0435, "step": 176940 }, { "epoch": 0.44, "learning_rate": 5.5741859177613367e-05, "loss": 1.0452, "step": 176945 }, { "epoch": 0.44, "learning_rate": 5.574060093613167e-05, "loss": 1.0455, "step": 176950 }, { "epoch": 0.44, "learning_rate": 5.573934269464996e-05, "loss": 1.0457, "step": 176955 }, { "epoch": 0.44, "learning_rate": 5.573808445316826e-05, "loss": 1.0469, "step": 176960 }, { "epoch": 0.44, "learning_rate": 5.573682621168655e-05, "loss": 1.0422, "step": 176965 }, { "epoch": 0.44, "learning_rate": 5.573556797020485e-05, "loss": 1.0456, "step": 176970 }, { "epoch": 0.44, "learning_rate": 5.573430972872314e-05, "loss": 1.0463, "step": 176975 }, { "epoch": 0.44, "learning_rate": 5.573305148724144e-05, "loss": 1.044, "step": 176980 }, { "epoch": 0.44, "learning_rate": 5.573179324575973e-05, "loss": 1.0437, "step": 176985 }, { "epoch": 0.44, "learning_rate": 5.573053500427803e-05, "loss": 1.0438, "step": 176990 }, { "epoch": 0.44, "learning_rate": 5.572927676279632e-05, "loss": 1.0475, "step": 176995 }, { "epoch": 0.44, "learning_rate": 5.5728018521314615e-05, "loss": 1.0435, "step": 177000 }, { "epoch": 0.44, "learning_rate": 5.572676027983291e-05, "loss": 1.0472, "step": 177005 }, { "epoch": 0.44, "learning_rate": 5.5725502038351205e-05, "loss": 1.0459, "step": 177010 }, { "epoch": 0.44, "learning_rate": 5.57242437968695e-05, "loss": 1.0472, "step": 177015 }, { "epoch": 0.44, "learning_rate": 5.5722985555387795e-05, "loss": 1.0452, "step": 177020 }, { "epoch": 0.44, "learning_rate": 5.572172731390609e-05, "loss": 1.046, "step": 177025 }, { "epoch": 0.44, "learning_rate": 5.5720469072424385e-05, "loss": 1.0481, "step": 177030 }, { "epoch": 0.44, "learning_rate": 5.571921083094268e-05, "loss": 1.0481, "step": 177035 }, { "epoch": 0.44, "learning_rate": 5.5717952589460975e-05, "loss": 1.0494, "step": 177040 }, { "epoch": 0.44, "learning_rate": 5.571669434797927e-05, "loss": 1.0463, "step": 177045 }, { "epoch": 0.44, "learning_rate": 5.5715436106497565e-05, "loss": 1.0426, "step": 177050 }, { "epoch": 0.44, "learning_rate": 5.571417786501586e-05, "loss": 1.0473, "step": 177055 }, { "epoch": 0.44, "learning_rate": 5.5712919623534155e-05, "loss": 1.0457, "step": 177060 }, { "epoch": 0.44, "learning_rate": 5.5711661382052446e-05, "loss": 1.0432, "step": 177065 }, { "epoch": 0.44, "learning_rate": 5.5710403140570745e-05, "loss": 1.0475, "step": 177070 }, { "epoch": 0.44, "learning_rate": 5.5709144899089036e-05, "loss": 1.0461, "step": 177075 }, { "epoch": 0.44, "learning_rate": 5.5707886657607335e-05, "loss": 1.0686, "step": 177080 }, { "epoch": 0.44, "learning_rate": 5.5706628416125626e-05, "loss": 1.0472, "step": 177085 }, { "epoch": 0.44, "learning_rate": 5.5705370174643925e-05, "loss": 1.0452, "step": 177090 }, { "epoch": 0.44, "learning_rate": 5.5704111933162216e-05, "loss": 1.0482, "step": 177095 }, { "epoch": 0.44, "learning_rate": 5.5702853691680514e-05, "loss": 1.0456, "step": 177100 }, { "epoch": 0.44, "learning_rate": 5.5701595450198806e-05, "loss": 1.0447, "step": 177105 }, { "epoch": 0.44, "learning_rate": 5.5700337208717104e-05, "loss": 1.0469, "step": 177110 }, { "epoch": 0.44, "learning_rate": 5.5699078967235396e-05, "loss": 1.0695, "step": 177115 }, { "epoch": 0.44, "learning_rate": 5.569782072575369e-05, "loss": 1.0467, "step": 177120 }, { "epoch": 0.44, "learning_rate": 5.5696562484271986e-05, "loss": 1.0453, "step": 177125 }, { "epoch": 0.44, "learning_rate": 5.569530424279028e-05, "loss": 1.046, "step": 177130 }, { "epoch": 0.44, "learning_rate": 5.5694046001308576e-05, "loss": 1.0477, "step": 177135 }, { "epoch": 0.44, "learning_rate": 5.569278775982687e-05, "loss": 1.0463, "step": 177140 }, { "epoch": 0.44, "learning_rate": 5.5691529518345166e-05, "loss": 1.0478, "step": 177145 }, { "epoch": 0.44, "learning_rate": 5.569027127686346e-05, "loss": 1.0471, "step": 177150 }, { "epoch": 0.44, "learning_rate": 5.5689013035381756e-05, "loss": 1.0441, "step": 177155 }, { "epoch": 0.44, "learning_rate": 5.568775479390005e-05, "loss": 1.0462, "step": 177160 }, { "epoch": 0.44, "learning_rate": 5.5686496552418346e-05, "loss": 1.0498, "step": 177165 }, { "epoch": 0.44, "learning_rate": 5.568523831093664e-05, "loss": 1.0655, "step": 177170 }, { "epoch": 0.44, "learning_rate": 5.5683980069454936e-05, "loss": 1.0468, "step": 177175 }, { "epoch": 0.44, "learning_rate": 5.568272182797323e-05, "loss": 1.0457, "step": 177180 }, { "epoch": 0.44, "learning_rate": 5.568146358649152e-05, "loss": 1.0453, "step": 177185 }, { "epoch": 0.44, "learning_rate": 5.568020534500982e-05, "loss": 1.0416, "step": 177190 }, { "epoch": 0.44, "learning_rate": 5.567894710352811e-05, "loss": 1.0443, "step": 177195 }, { "epoch": 0.44, "learning_rate": 5.567768886204641e-05, "loss": 1.0436, "step": 177200 }, { "epoch": 0.44, "learning_rate": 5.56764306205647e-05, "loss": 1.0431, "step": 177205 }, { "epoch": 0.44, "learning_rate": 5.5675172379083e-05, "loss": 1.0467, "step": 177210 }, { "epoch": 0.44, "learning_rate": 5.567391413760129e-05, "loss": 1.0445, "step": 177215 }, { "epoch": 0.44, "learning_rate": 5.567265589611959e-05, "loss": 1.0479, "step": 177220 }, { "epoch": 0.44, "learning_rate": 5.567139765463788e-05, "loss": 1.0462, "step": 177225 }, { "epoch": 0.44, "learning_rate": 5.567013941315618e-05, "loss": 1.046, "step": 177230 }, { "epoch": 0.44, "learning_rate": 5.566888117167447e-05, "loss": 1.0457, "step": 177235 }, { "epoch": 0.44, "learning_rate": 5.566762293019277e-05, "loss": 1.0459, "step": 177240 }, { "epoch": 0.44, "learning_rate": 5.566636468871106e-05, "loss": 1.0433, "step": 177245 }, { "epoch": 0.44, "learning_rate": 5.566510644722935e-05, "loss": 1.0469, "step": 177250 }, { "epoch": 0.44, "learning_rate": 5.566384820574765e-05, "loss": 1.0462, "step": 177255 }, { "epoch": 0.44, "learning_rate": 5.566258996426594e-05, "loss": 1.0419, "step": 177260 }, { "epoch": 0.44, "learning_rate": 5.566133172278424e-05, "loss": 1.0447, "step": 177265 }, { "epoch": 0.44, "learning_rate": 5.566007348130253e-05, "loss": 1.0479, "step": 177270 }, { "epoch": 0.44, "learning_rate": 5.565881523982083e-05, "loss": 1.0677, "step": 177275 }, { "epoch": 0.45, "learning_rate": 5.565755699833912e-05, "loss": 1.0486, "step": 177280 }, { "epoch": 0.45, "learning_rate": 5.565629875685742e-05, "loss": 1.0458, "step": 177285 }, { "epoch": 0.45, "learning_rate": 5.565504051537571e-05, "loss": 1.0463, "step": 177290 }, { "epoch": 0.45, "learning_rate": 5.565378227389401e-05, "loss": 1.0477, "step": 177295 }, { "epoch": 0.45, "learning_rate": 5.56525240324123e-05, "loss": 1.0468, "step": 177300 }, { "epoch": 0.45, "learning_rate": 5.56512657909306e-05, "loss": 1.0447, "step": 177305 }, { "epoch": 0.45, "learning_rate": 5.565000754944889e-05, "loss": 1.0451, "step": 177310 }, { "epoch": 0.45, "learning_rate": 5.564874930796718e-05, "loss": 1.0412, "step": 177315 }, { "epoch": 0.45, "learning_rate": 5.564749106648548e-05, "loss": 1.0472, "step": 177320 }, { "epoch": 0.45, "learning_rate": 5.564623282500377e-05, "loss": 1.0441, "step": 177325 }, { "epoch": 0.45, "learning_rate": 5.564497458352207e-05, "loss": 1.0673, "step": 177330 }, { "epoch": 0.45, "learning_rate": 5.564371634204036e-05, "loss": 1.0437, "step": 177335 }, { "epoch": 0.45, "learning_rate": 5.564245810055866e-05, "loss": 1.0463, "step": 177340 }, { "epoch": 0.45, "learning_rate": 5.564119985907695e-05, "loss": 1.0454, "step": 177345 }, { "epoch": 0.45, "learning_rate": 5.563994161759525e-05, "loss": 1.0442, "step": 177350 }, { "epoch": 0.45, "learning_rate": 5.563868337611354e-05, "loss": 1.0431, "step": 177355 }, { "epoch": 0.45, "learning_rate": 5.563742513463184e-05, "loss": 1.0559, "step": 177360 }, { "epoch": 0.45, "learning_rate": 5.563616689315013e-05, "loss": 1.0441, "step": 177365 }, { "epoch": 0.45, "learning_rate": 5.563490865166843e-05, "loss": 1.0446, "step": 177370 }, { "epoch": 0.45, "learning_rate": 5.563365041018672e-05, "loss": 1.0468, "step": 177375 }, { "epoch": 0.45, "learning_rate": 5.563239216870501e-05, "loss": 1.0482, "step": 177380 }, { "epoch": 0.45, "learning_rate": 5.563113392722331e-05, "loss": 1.0416, "step": 177385 }, { "epoch": 0.45, "learning_rate": 5.56298756857416e-05, "loss": 1.0471, "step": 177390 }, { "epoch": 0.45, "learning_rate": 5.56286174442599e-05, "loss": 1.0445, "step": 177395 }, { "epoch": 0.45, "learning_rate": 5.562735920277819e-05, "loss": 1.049, "step": 177400 }, { "epoch": 0.45, "learning_rate": 5.562610096129649e-05, "loss": 1.0444, "step": 177405 }, { "epoch": 0.45, "learning_rate": 5.562484271981478e-05, "loss": 1.0456, "step": 177410 }, { "epoch": 0.45, "learning_rate": 5.562358447833308e-05, "loss": 1.0493, "step": 177415 }, { "epoch": 0.45, "learning_rate": 5.562232623685137e-05, "loss": 1.0465, "step": 177420 }, { "epoch": 0.45, "learning_rate": 5.562106799536967e-05, "loss": 1.049, "step": 177425 }, { "epoch": 0.45, "learning_rate": 5.561980975388796e-05, "loss": 1.0458, "step": 177430 }, { "epoch": 0.45, "learning_rate": 5.561855151240626e-05, "loss": 1.0708, "step": 177435 }, { "epoch": 0.45, "learning_rate": 5.561729327092455e-05, "loss": 1.0442, "step": 177440 }, { "epoch": 0.45, "learning_rate": 5.5616035029442845e-05, "loss": 1.0472, "step": 177445 }, { "epoch": 0.45, "learning_rate": 5.561477678796116e-05, "loss": 1.0474, "step": 177450 }, { "epoch": 0.45, "learning_rate": 5.561351854647945e-05, "loss": 1.0435, "step": 177455 }, { "epoch": 0.45, "learning_rate": 5.561226030499774e-05, "loss": 1.0463, "step": 177460 }, { "epoch": 0.45, "learning_rate": 5.561100206351604e-05, "loss": 1.0436, "step": 177465 }, { "epoch": 0.45, "learning_rate": 5.560974382203433e-05, "loss": 1.0595, "step": 177470 }, { "epoch": 0.45, "learning_rate": 5.560848558055263e-05, "loss": 1.0439, "step": 177475 }, { "epoch": 0.45, "learning_rate": 5.560722733907092e-05, "loss": 1.0478, "step": 177480 }, { "epoch": 0.45, "learning_rate": 5.560596909758922e-05, "loss": 1.0439, "step": 177485 }, { "epoch": 0.45, "learning_rate": 5.560471085610751e-05, "loss": 1.0484, "step": 177490 }, { "epoch": 0.45, "learning_rate": 5.560345261462581e-05, "loss": 1.0454, "step": 177495 }, { "epoch": 0.45, "learning_rate": 5.56021943731441e-05, "loss": 1.0462, "step": 177500 }, { "epoch": 0.45, "learning_rate": 5.56009361316624e-05, "loss": 1.0477, "step": 177505 }, { "epoch": 0.45, "learning_rate": 5.559967789018069e-05, "loss": 1.0483, "step": 177510 }, { "epoch": 0.45, "learning_rate": 5.559841964869899e-05, "loss": 1.0448, "step": 177515 }, { "epoch": 0.45, "learning_rate": 5.559716140721728e-05, "loss": 1.0682, "step": 177520 }, { "epoch": 0.45, "learning_rate": 5.559590316573557e-05, "loss": 1.0461, "step": 177525 }, { "epoch": 0.45, "learning_rate": 5.559464492425387e-05, "loss": 1.0455, "step": 177530 }, { "epoch": 0.45, "learning_rate": 5.559338668277216e-05, "loss": 1.0436, "step": 177535 }, { "epoch": 0.45, "learning_rate": 5.559212844129046e-05, "loss": 1.0651, "step": 177540 }, { "epoch": 0.45, "learning_rate": 5.559087019980875e-05, "loss": 1.0459, "step": 177545 }, { "epoch": 0.45, "learning_rate": 5.558961195832705e-05, "loss": 1.044, "step": 177550 }, { "epoch": 0.45, "learning_rate": 5.558835371684534e-05, "loss": 1.042, "step": 177555 }, { "epoch": 0.45, "learning_rate": 5.558709547536364e-05, "loss": 1.0462, "step": 177560 }, { "epoch": 0.45, "learning_rate": 5.558583723388193e-05, "loss": 1.0485, "step": 177565 }, { "epoch": 0.45, "learning_rate": 5.558457899240023e-05, "loss": 1.0646, "step": 177570 }, { "epoch": 0.45, "learning_rate": 5.558332075091852e-05, "loss": 1.0492, "step": 177575 }, { "epoch": 0.45, "learning_rate": 5.558206250943682e-05, "loss": 1.0441, "step": 177580 }, { "epoch": 0.45, "learning_rate": 5.558080426795511e-05, "loss": 1.046, "step": 177585 }, { "epoch": 0.45, "learning_rate": 5.55795460264734e-05, "loss": 1.0438, "step": 177590 }, { "epoch": 0.45, "learning_rate": 5.55782877849917e-05, "loss": 1.0445, "step": 177595 }, { "epoch": 0.45, "learning_rate": 5.557702954350999e-05, "loss": 1.049, "step": 177600 }, { "epoch": 0.45, "learning_rate": 5.557577130202829e-05, "loss": 1.0449, "step": 177605 }, { "epoch": 0.45, "learning_rate": 5.557451306054658e-05, "loss": 1.0488, "step": 177610 }, { "epoch": 0.45, "learning_rate": 5.557325481906488e-05, "loss": 1.0431, "step": 177615 }, { "epoch": 0.45, "learning_rate": 5.557199657758317e-05, "loss": 1.0446, "step": 177620 }, { "epoch": 0.45, "learning_rate": 5.557073833610147e-05, "loss": 1.0461, "step": 177625 }, { "epoch": 0.45, "learning_rate": 5.556948009461976e-05, "loss": 1.0457, "step": 177630 }, { "epoch": 0.45, "learning_rate": 5.556822185313806e-05, "loss": 1.0442, "step": 177635 }, { "epoch": 0.45, "learning_rate": 5.556696361165635e-05, "loss": 1.0447, "step": 177640 }, { "epoch": 0.45, "learning_rate": 5.556570537017465e-05, "loss": 1.046, "step": 177645 }, { "epoch": 0.45, "learning_rate": 5.556444712869294e-05, "loss": 1.0463, "step": 177650 }, { "epoch": 0.45, "learning_rate": 5.5563188887211234e-05, "loss": 1.0456, "step": 177655 }, { "epoch": 0.45, "learning_rate": 5.556193064572953e-05, "loss": 1.0444, "step": 177660 }, { "epoch": 0.45, "learning_rate": 5.5560672404247824e-05, "loss": 1.0431, "step": 177665 }, { "epoch": 0.45, "learning_rate": 5.555941416276612e-05, "loss": 1.0651, "step": 177670 }, { "epoch": 0.45, "learning_rate": 5.5558155921284414e-05, "loss": 1.045, "step": 177675 }, { "epoch": 0.45, "learning_rate": 5.555689767980271e-05, "loss": 1.0656, "step": 177680 }, { "epoch": 0.45, "learning_rate": 5.5555639438321004e-05, "loss": 1.049, "step": 177685 }, { "epoch": 0.45, "learning_rate": 5.55543811968393e-05, "loss": 1.0457, "step": 177690 }, { "epoch": 0.45, "learning_rate": 5.5553122955357594e-05, "loss": 1.0486, "step": 177695 }, { "epoch": 0.45, "learning_rate": 5.555186471387589e-05, "loss": 1.0451, "step": 177700 }, { "epoch": 0.45, "learning_rate": 5.5550606472394184e-05, "loss": 1.042, "step": 177705 }, { "epoch": 0.45, "learning_rate": 5.554934823091248e-05, "loss": 1.0474, "step": 177710 }, { "epoch": 0.45, "learning_rate": 5.5548089989430774e-05, "loss": 1.0455, "step": 177715 }, { "epoch": 0.45, "learning_rate": 5.5546831747949066e-05, "loss": 1.0593, "step": 177720 }, { "epoch": 0.45, "learning_rate": 5.5545573506467364e-05, "loss": 1.0459, "step": 177725 }, { "epoch": 0.45, "learning_rate": 5.5544315264985656e-05, "loss": 1.0449, "step": 177730 }, { "epoch": 0.45, "learning_rate": 5.5543057023503954e-05, "loss": 1.0592, "step": 177735 }, { "epoch": 0.45, "learning_rate": 5.5541798782022246e-05, "loss": 1.0452, "step": 177740 }, { "epoch": 0.45, "learning_rate": 5.5540540540540544e-05, "loss": 1.0447, "step": 177745 }, { "epoch": 0.45, "learning_rate": 5.5539282299058836e-05, "loss": 1.0456, "step": 177750 }, { "epoch": 0.45, "learning_rate": 5.5538024057577134e-05, "loss": 1.0471, "step": 177755 }, { "epoch": 0.45, "learning_rate": 5.5536765816095425e-05, "loss": 1.0454, "step": 177760 }, { "epoch": 0.45, "learning_rate": 5.5535507574613724e-05, "loss": 1.0493, "step": 177765 }, { "epoch": 0.45, "learning_rate": 5.5534249333132015e-05, "loss": 1.0462, "step": 177770 }, { "epoch": 0.45, "learning_rate": 5.553299109165031e-05, "loss": 1.0463, "step": 177775 }, { "epoch": 0.45, "learning_rate": 5.5531732850168605e-05, "loss": 1.0431, "step": 177780 }, { "epoch": 0.45, "learning_rate": 5.55304746086869e-05, "loss": 1.0453, "step": 177785 }, { "epoch": 0.45, "learning_rate": 5.5529216367205195e-05, "loss": 1.045, "step": 177790 }, { "epoch": 0.45, "learning_rate": 5.552795812572349e-05, "loss": 1.0446, "step": 177795 }, { "epoch": 0.45, "learning_rate": 5.5526699884241785e-05, "loss": 1.0486, "step": 177800 }, { "epoch": 0.45, "learning_rate": 5.552544164276008e-05, "loss": 1.0463, "step": 177805 }, { "epoch": 0.45, "learning_rate": 5.5524183401278375e-05, "loss": 1.0477, "step": 177810 }, { "epoch": 0.45, "learning_rate": 5.552292515979667e-05, "loss": 1.0453, "step": 177815 }, { "epoch": 0.45, "learning_rate": 5.5521666918314965e-05, "loss": 1.0457, "step": 177820 }, { "epoch": 0.45, "learning_rate": 5.552040867683326e-05, "loss": 1.0454, "step": 177825 }, { "epoch": 0.45, "learning_rate": 5.5519150435351555e-05, "loss": 1.0449, "step": 177830 }, { "epoch": 0.45, "learning_rate": 5.551789219386985e-05, "loss": 1.0484, "step": 177835 }, { "epoch": 0.45, "learning_rate": 5.551663395238814e-05, "loss": 1.0428, "step": 177840 }, { "epoch": 0.45, "learning_rate": 5.551537571090644e-05, "loss": 1.0448, "step": 177845 }, { "epoch": 0.45, "learning_rate": 5.551411746942473e-05, "loss": 1.0705, "step": 177850 }, { "epoch": 0.45, "learning_rate": 5.551285922794303e-05, "loss": 1.0459, "step": 177855 }, { "epoch": 0.45, "learning_rate": 5.551160098646132e-05, "loss": 1.0444, "step": 177860 }, { "epoch": 0.45, "learning_rate": 5.551034274497962e-05, "loss": 1.0443, "step": 177865 }, { "epoch": 0.45, "learning_rate": 5.550908450349791e-05, "loss": 1.048, "step": 177870 }, { "epoch": 0.45, "learning_rate": 5.550782626201621e-05, "loss": 1.0461, "step": 177875 }, { "epoch": 0.45, "learning_rate": 5.55065680205345e-05, "loss": 1.0673, "step": 177880 }, { "epoch": 0.45, "learning_rate": 5.55053097790528e-05, "loss": 1.0452, "step": 177885 }, { "epoch": 0.45, "learning_rate": 5.550405153757109e-05, "loss": 1.0463, "step": 177890 }, { "epoch": 0.45, "learning_rate": 5.550279329608939e-05, "loss": 1.0461, "step": 177895 }, { "epoch": 0.45, "learning_rate": 5.550153505460768e-05, "loss": 1.0416, "step": 177900 }, { "epoch": 0.45, "learning_rate": 5.550027681312597e-05, "loss": 1.0436, "step": 177905 }, { "epoch": 0.45, "learning_rate": 5.549901857164427e-05, "loss": 1.0445, "step": 177910 }, { "epoch": 0.45, "learning_rate": 5.549776033016256e-05, "loss": 1.064, "step": 177915 }, { "epoch": 0.45, "learning_rate": 5.549650208868086e-05, "loss": 1.0463, "step": 177920 }, { "epoch": 0.45, "learning_rate": 5.549524384719915e-05, "loss": 1.0435, "step": 177925 }, { "epoch": 0.45, "learning_rate": 5.549398560571745e-05, "loss": 1.0443, "step": 177930 }, { "epoch": 0.45, "learning_rate": 5.549272736423574e-05, "loss": 1.0452, "step": 177935 }, { "epoch": 0.45, "learning_rate": 5.549146912275404e-05, "loss": 1.0444, "step": 177940 }, { "epoch": 0.45, "learning_rate": 5.549021088127233e-05, "loss": 1.0454, "step": 177945 }, { "epoch": 0.45, "learning_rate": 5.5488952639790635e-05, "loss": 1.0466, "step": 177950 }, { "epoch": 0.45, "learning_rate": 5.548769439830893e-05, "loss": 1.0438, "step": 177955 }, { "epoch": 0.45, "learning_rate": 5.5486436156827225e-05, "loss": 1.0464, "step": 177960 }, { "epoch": 0.45, "learning_rate": 5.548517791534552e-05, "loss": 1.0424, "step": 177965 }, { "epoch": 0.45, "learning_rate": 5.5483919673863815e-05, "loss": 1.0471, "step": 177970 }, { "epoch": 0.45, "learning_rate": 5.548266143238211e-05, "loss": 1.0473, "step": 177975 }, { "epoch": 0.45, "learning_rate": 5.5481403190900405e-05, "loss": 1.0441, "step": 177980 }, { "epoch": 0.45, "learning_rate": 5.5480144949418696e-05, "loss": 1.044, "step": 177985 }, { "epoch": 0.45, "learning_rate": 5.5478886707936995e-05, "loss": 1.0443, "step": 177990 }, { "epoch": 0.45, "learning_rate": 5.5477628466455286e-05, "loss": 1.046, "step": 177995 }, { "epoch": 0.45, "learning_rate": 5.5476370224973585e-05, "loss": 1.0455, "step": 178000 }, { "epoch": 0.45, "learning_rate": 5.5475111983491876e-05, "loss": 1.0446, "step": 178005 }, { "epoch": 0.45, "learning_rate": 5.5473853742010175e-05, "loss": 1.0453, "step": 178010 }, { "epoch": 0.45, "learning_rate": 5.5472595500528466e-05, "loss": 1.0499, "step": 178015 }, { "epoch": 0.45, "learning_rate": 5.5471337259046765e-05, "loss": 1.0657, "step": 178020 }, { "epoch": 0.45, "learning_rate": 5.5470079017565056e-05, "loss": 1.045, "step": 178025 }, { "epoch": 0.45, "learning_rate": 5.5468820776083355e-05, "loss": 1.0468, "step": 178030 }, { "epoch": 0.45, "learning_rate": 5.5467562534601646e-05, "loss": 1.0447, "step": 178035 }, { "epoch": 0.45, "learning_rate": 5.5466304293119945e-05, "loss": 1.0434, "step": 178040 }, { "epoch": 0.45, "learning_rate": 5.5465046051638236e-05, "loss": 1.0468, "step": 178045 }, { "epoch": 0.45, "learning_rate": 5.546378781015653e-05, "loss": 1.0469, "step": 178050 }, { "epoch": 0.45, "learning_rate": 5.5462529568674826e-05, "loss": 1.0469, "step": 178055 }, { "epoch": 0.45, "learning_rate": 5.546127132719312e-05, "loss": 1.0442, "step": 178060 }, { "epoch": 0.45, "learning_rate": 5.5460013085711416e-05, "loss": 1.0455, "step": 178065 }, { "epoch": 0.45, "learning_rate": 5.545875484422971e-05, "loss": 1.0449, "step": 178070 }, { "epoch": 0.45, "learning_rate": 5.5457496602748006e-05, "loss": 1.0465, "step": 178075 }, { "epoch": 0.45, "learning_rate": 5.54562383612663e-05, "loss": 1.0472, "step": 178080 }, { "epoch": 0.45, "learning_rate": 5.5454980119784596e-05, "loss": 1.049, "step": 178085 }, { "epoch": 0.45, "learning_rate": 5.545372187830289e-05, "loss": 1.0468, "step": 178090 }, { "epoch": 0.45, "learning_rate": 5.5452463636821186e-05, "loss": 1.0426, "step": 178095 }, { "epoch": 0.45, "learning_rate": 5.545120539533948e-05, "loss": 1.0421, "step": 178100 }, { "epoch": 0.45, "learning_rate": 5.5449947153857776e-05, "loss": 1.0454, "step": 178105 }, { "epoch": 0.45, "learning_rate": 5.544868891237607e-05, "loss": 1.0439, "step": 178110 }, { "epoch": 0.45, "learning_rate": 5.544743067089436e-05, "loss": 1.0493, "step": 178115 }, { "epoch": 0.45, "learning_rate": 5.544617242941266e-05, "loss": 1.046, "step": 178120 }, { "epoch": 0.45, "learning_rate": 5.544491418793095e-05, "loss": 1.0469, "step": 178125 }, { "epoch": 0.45, "learning_rate": 5.544365594644925e-05, "loss": 1.0457, "step": 178130 }, { "epoch": 0.45, "learning_rate": 5.544239770496754e-05, "loss": 1.0463, "step": 178135 }, { "epoch": 0.45, "learning_rate": 5.544113946348584e-05, "loss": 1.0456, "step": 178140 }, { "epoch": 0.45, "learning_rate": 5.543988122200413e-05, "loss": 1.0459, "step": 178145 }, { "epoch": 0.45, "learning_rate": 5.543862298052243e-05, "loss": 1.0454, "step": 178150 }, { "epoch": 0.45, "learning_rate": 5.543736473904072e-05, "loss": 1.0634, "step": 178155 }, { "epoch": 0.45, "learning_rate": 5.543610649755902e-05, "loss": 1.0697, "step": 178160 }, { "epoch": 0.45, "learning_rate": 5.543484825607731e-05, "loss": 1.0447, "step": 178165 }, { "epoch": 0.45, "learning_rate": 5.543359001459561e-05, "loss": 1.045, "step": 178170 }, { "epoch": 0.45, "learning_rate": 5.54323317731139e-05, "loss": 1.0425, "step": 178175 }, { "epoch": 0.45, "learning_rate": 5.543107353163219e-05, "loss": 1.0425, "step": 178180 }, { "epoch": 0.45, "learning_rate": 5.542981529015049e-05, "loss": 1.0468, "step": 178185 }, { "epoch": 0.45, "learning_rate": 5.542855704866878e-05, "loss": 1.0446, "step": 178190 }, { "epoch": 0.45, "learning_rate": 5.542729880718708e-05, "loss": 1.0449, "step": 178195 }, { "epoch": 0.45, "learning_rate": 5.542604056570537e-05, "loss": 1.0469, "step": 178200 }, { "epoch": 0.45, "learning_rate": 5.542478232422367e-05, "loss": 1.0442, "step": 178205 }, { "epoch": 0.45, "learning_rate": 5.542352408274196e-05, "loss": 1.0443, "step": 178210 }, { "epoch": 0.45, "learning_rate": 5.542226584126026e-05, "loss": 1.0428, "step": 178215 }, { "epoch": 0.45, "learning_rate": 5.542100759977855e-05, "loss": 1.044, "step": 178220 }, { "epoch": 0.45, "learning_rate": 5.541974935829685e-05, "loss": 1.0407, "step": 178225 }, { "epoch": 0.45, "learning_rate": 5.541849111681514e-05, "loss": 1.0416, "step": 178230 }, { "epoch": 0.45, "learning_rate": 5.541723287533344e-05, "loss": 1.0496, "step": 178235 }, { "epoch": 0.45, "learning_rate": 5.541597463385173e-05, "loss": 1.0461, "step": 178240 }, { "epoch": 0.45, "learning_rate": 5.541471639237002e-05, "loss": 1.0456, "step": 178245 }, { "epoch": 0.45, "learning_rate": 5.541345815088832e-05, "loss": 1.0474, "step": 178250 }, { "epoch": 0.45, "learning_rate": 5.541219990940661e-05, "loss": 1.0455, "step": 178255 }, { "epoch": 0.45, "learning_rate": 5.541094166792491e-05, "loss": 1.0458, "step": 178260 }, { "epoch": 0.45, "learning_rate": 5.54096834264432e-05, "loss": 1.0454, "step": 178265 }, { "epoch": 0.45, "learning_rate": 5.54084251849615e-05, "loss": 1.0445, "step": 178270 }, { "epoch": 0.45, "learning_rate": 5.540716694347979e-05, "loss": 1.0448, "step": 178275 }, { "epoch": 0.45, "learning_rate": 5.540590870199809e-05, "loss": 1.0461, "step": 178280 }, { "epoch": 0.45, "learning_rate": 5.540465046051638e-05, "loss": 1.0417, "step": 178285 }, { "epoch": 0.45, "learning_rate": 5.540339221903468e-05, "loss": 1.0469, "step": 178290 }, { "epoch": 0.45, "learning_rate": 5.540213397755297e-05, "loss": 1.0437, "step": 178295 }, { "epoch": 0.45, "learning_rate": 5.540087573607127e-05, "loss": 1.0439, "step": 178300 }, { "epoch": 0.45, "learning_rate": 5.539961749458956e-05, "loss": 1.0447, "step": 178305 }, { "epoch": 0.45, "learning_rate": 5.5398359253107854e-05, "loss": 1.0461, "step": 178310 }, { "epoch": 0.45, "learning_rate": 5.539710101162615e-05, "loss": 1.0423, "step": 178315 }, { "epoch": 0.45, "learning_rate": 5.5395842770144444e-05, "loss": 1.0486, "step": 178320 }, { "epoch": 0.45, "learning_rate": 5.539458452866274e-05, "loss": 1.0478, "step": 178325 }, { "epoch": 0.45, "learning_rate": 5.5393326287181034e-05, "loss": 1.0449, "step": 178330 }, { "epoch": 0.45, "learning_rate": 5.539206804569933e-05, "loss": 1.0461, "step": 178335 }, { "epoch": 0.45, "learning_rate": 5.5390809804217623e-05, "loss": 1.0425, "step": 178340 }, { "epoch": 0.45, "learning_rate": 5.538955156273592e-05, "loss": 1.0662, "step": 178345 }, { "epoch": 0.45, "learning_rate": 5.5388293321254213e-05, "loss": 1.0443, "step": 178350 }, { "epoch": 0.45, "learning_rate": 5.538703507977251e-05, "loss": 1.0467, "step": 178355 }, { "epoch": 0.45, "learning_rate": 5.5385776838290803e-05, "loss": 1.0432, "step": 178360 }, { "epoch": 0.45, "learning_rate": 5.53845185968091e-05, "loss": 1.0459, "step": 178365 }, { "epoch": 0.45, "learning_rate": 5.5383260355327393e-05, "loss": 1.0462, "step": 178370 }, { "epoch": 0.45, "learning_rate": 5.5382002113845685e-05, "loss": 1.0451, "step": 178375 }, { "epoch": 0.45, "learning_rate": 5.538074387236398e-05, "loss": 1.0446, "step": 178380 }, { "epoch": 0.45, "learning_rate": 5.5379485630882275e-05, "loss": 1.0477, "step": 178385 }, { "epoch": 0.45, "learning_rate": 5.537822738940057e-05, "loss": 1.0434, "step": 178390 }, { "epoch": 0.45, "learning_rate": 5.5376969147918865e-05, "loss": 1.0424, "step": 178395 }, { "epoch": 0.45, "learning_rate": 5.537571090643716e-05, "loss": 1.0447, "step": 178400 }, { "epoch": 0.45, "learning_rate": 5.5374452664955455e-05, "loss": 1.0443, "step": 178405 }, { "epoch": 0.45, "learning_rate": 5.537319442347375e-05, "loss": 1.0426, "step": 178410 }, { "epoch": 0.45, "learning_rate": 5.5371936181992045e-05, "loss": 1.0562, "step": 178415 }, { "epoch": 0.45, "learning_rate": 5.537067794051034e-05, "loss": 1.0437, "step": 178420 }, { "epoch": 0.45, "learning_rate": 5.5369419699028635e-05, "loss": 1.0644, "step": 178425 }, { "epoch": 0.45, "learning_rate": 5.536816145754693e-05, "loss": 1.093, "step": 178430 }, { "epoch": 0.45, "learning_rate": 5.5366903216065225e-05, "loss": 1.047, "step": 178435 }, { "epoch": 0.45, "learning_rate": 5.5365644974583516e-05, "loss": 1.0453, "step": 178440 }, { "epoch": 0.45, "learning_rate": 5.5364386733101815e-05, "loss": 1.0483, "step": 178445 }, { "epoch": 0.45, "learning_rate": 5.536312849162012e-05, "loss": 1.0454, "step": 178450 }, { "epoch": 0.45, "learning_rate": 5.536187025013841e-05, "loss": 1.0431, "step": 178455 }, { "epoch": 0.45, "learning_rate": 5.536061200865671e-05, "loss": 1.0497, "step": 178460 }, { "epoch": 0.45, "learning_rate": 5.5359353767175e-05, "loss": 1.0443, "step": 178465 }, { "epoch": 0.45, "learning_rate": 5.53580955256933e-05, "loss": 1.0432, "step": 178470 }, { "epoch": 0.45, "learning_rate": 5.535683728421159e-05, "loss": 1.0449, "step": 178475 }, { "epoch": 0.45, "learning_rate": 5.535557904272989e-05, "loss": 1.0413, "step": 178480 }, { "epoch": 0.45, "learning_rate": 5.535432080124818e-05, "loss": 1.0473, "step": 178485 }, { "epoch": 0.45, "learning_rate": 5.535306255976648e-05, "loss": 1.0439, "step": 178490 }, { "epoch": 0.45, "learning_rate": 5.535180431828477e-05, "loss": 1.0451, "step": 178495 }, { "epoch": 0.45, "learning_rate": 5.535054607680307e-05, "loss": 1.0437, "step": 178500 }, { "epoch": 0.45, "learning_rate": 5.534928783532136e-05, "loss": 1.0435, "step": 178505 }, { "epoch": 0.45, "learning_rate": 5.534802959383966e-05, "loss": 1.0435, "step": 178510 }, { "epoch": 0.45, "learning_rate": 5.534677135235795e-05, "loss": 1.0696, "step": 178515 }, { "epoch": 0.45, "learning_rate": 5.534551311087624e-05, "loss": 1.0483, "step": 178520 }, { "epoch": 0.45, "learning_rate": 5.534425486939454e-05, "loss": 1.0442, "step": 178525 }, { "epoch": 0.45, "learning_rate": 5.534299662791283e-05, "loss": 1.0465, "step": 178530 }, { "epoch": 0.45, "learning_rate": 5.534173838643113e-05, "loss": 1.0468, "step": 178535 }, { "epoch": 0.45, "learning_rate": 5.534048014494942e-05, "loss": 1.044, "step": 178540 }, { "epoch": 0.45, "learning_rate": 5.533922190346772e-05, "loss": 1.0437, "step": 178545 }, { "epoch": 0.45, "learning_rate": 5.533796366198601e-05, "loss": 1.0455, "step": 178550 }, { "epoch": 0.45, "learning_rate": 5.533670542050431e-05, "loss": 1.0458, "step": 178555 }, { "epoch": 0.45, "learning_rate": 5.53354471790226e-05, "loss": 1.0481, "step": 178560 }, { "epoch": 0.45, "learning_rate": 5.53341889375409e-05, "loss": 1.0467, "step": 178565 }, { "epoch": 0.45, "learning_rate": 5.533293069605919e-05, "loss": 1.0438, "step": 178570 }, { "epoch": 0.45, "learning_rate": 5.533167245457749e-05, "loss": 1.048, "step": 178575 }, { "epoch": 0.45, "learning_rate": 5.533041421309578e-05, "loss": 1.046, "step": 178580 }, { "epoch": 0.45, "learning_rate": 5.5329155971614074e-05, "loss": 1.0468, "step": 178585 }, { "epoch": 0.45, "learning_rate": 5.532789773013237e-05, "loss": 1.0465, "step": 178590 }, { "epoch": 0.45, "learning_rate": 5.5326639488650664e-05, "loss": 1.044, "step": 178595 }, { "epoch": 0.45, "learning_rate": 5.532538124716896e-05, "loss": 1.0464, "step": 178600 }, { "epoch": 0.45, "learning_rate": 5.5324123005687254e-05, "loss": 1.0461, "step": 178605 }, { "epoch": 0.45, "learning_rate": 5.532286476420555e-05, "loss": 1.0447, "step": 178610 }, { "epoch": 0.45, "learning_rate": 5.5321606522723844e-05, "loss": 1.0483, "step": 178615 }, { "epoch": 0.45, "learning_rate": 5.532034828124214e-05, "loss": 1.0444, "step": 178620 }, { "epoch": 0.45, "learning_rate": 5.5319090039760434e-05, "loss": 1.0429, "step": 178625 }, { "epoch": 0.45, "learning_rate": 5.531783179827873e-05, "loss": 1.0452, "step": 178630 }, { "epoch": 0.45, "learning_rate": 5.5316573556797024e-05, "loss": 1.0434, "step": 178635 }, { "epoch": 0.45, "learning_rate": 5.5315315315315316e-05, "loss": 1.047, "step": 178640 }, { "epoch": 0.45, "learning_rate": 5.5314057073833614e-05, "loss": 1.0454, "step": 178645 }, { "epoch": 0.45, "learning_rate": 5.5312798832351906e-05, "loss": 1.0444, "step": 178650 }, { "epoch": 0.45, "learning_rate": 5.5311540590870204e-05, "loss": 1.0665, "step": 178655 }, { "epoch": 0.45, "learning_rate": 5.5310282349388496e-05, "loss": 1.0453, "step": 178660 }, { "epoch": 0.45, "learning_rate": 5.5309024107906794e-05, "loss": 1.047, "step": 178665 }, { "epoch": 0.45, "learning_rate": 5.5307765866425086e-05, "loss": 1.0692, "step": 178670 }, { "epoch": 0.45, "learning_rate": 5.5306507624943384e-05, "loss": 1.0452, "step": 178675 }, { "epoch": 0.45, "learning_rate": 5.5305249383461676e-05, "loss": 1.0441, "step": 178680 }, { "epoch": 0.45, "learning_rate": 5.5303991141979974e-05, "loss": 1.0481, "step": 178685 }, { "epoch": 0.45, "learning_rate": 5.5302732900498266e-05, "loss": 1.0434, "step": 178690 }, { "epoch": 0.45, "learning_rate": 5.5301474659016564e-05, "loss": 1.0449, "step": 178695 }, { "epoch": 0.45, "learning_rate": 5.5300216417534856e-05, "loss": 1.0484, "step": 178700 }, { "epoch": 0.45, "learning_rate": 5.529895817605315e-05, "loss": 1.0471, "step": 178705 }, { "epoch": 0.45, "learning_rate": 5.5297699934571446e-05, "loss": 1.045, "step": 178710 }, { "epoch": 0.45, "learning_rate": 5.529644169308974e-05, "loss": 1.0429, "step": 178715 }, { "epoch": 0.45, "learning_rate": 5.5295183451608036e-05, "loss": 1.0467, "step": 178720 }, { "epoch": 0.45, "learning_rate": 5.529392521012633e-05, "loss": 1.0465, "step": 178725 }, { "epoch": 0.45, "learning_rate": 5.5292666968644626e-05, "loss": 1.0465, "step": 178730 }, { "epoch": 0.45, "learning_rate": 5.529140872716292e-05, "loss": 1.0455, "step": 178735 }, { "epoch": 0.45, "learning_rate": 5.5290150485681216e-05, "loss": 1.047, "step": 178740 }, { "epoch": 0.45, "learning_rate": 5.528889224419951e-05, "loss": 1.0476, "step": 178745 }, { "epoch": 0.45, "learning_rate": 5.5287634002717806e-05, "loss": 1.0452, "step": 178750 }, { "epoch": 0.45, "learning_rate": 5.52863757612361e-05, "loss": 1.0468, "step": 178755 }, { "epoch": 0.45, "learning_rate": 5.5285117519754395e-05, "loss": 1.0477, "step": 178760 }, { "epoch": 0.45, "learning_rate": 5.528385927827269e-05, "loss": 1.0447, "step": 178765 }, { "epoch": 0.45, "learning_rate": 5.528260103679098e-05, "loss": 1.0464, "step": 178770 }, { "epoch": 0.45, "learning_rate": 5.528134279530928e-05, "loss": 1.0468, "step": 178775 }, { "epoch": 0.45, "learning_rate": 5.528008455382757e-05, "loss": 1.0491, "step": 178780 }, { "epoch": 0.45, "learning_rate": 5.527882631234587e-05, "loss": 1.0442, "step": 178785 }, { "epoch": 0.45, "learning_rate": 5.527756807086416e-05, "loss": 1.0454, "step": 178790 }, { "epoch": 0.45, "learning_rate": 5.527630982938246e-05, "loss": 1.0487, "step": 178795 }, { "epoch": 0.45, "learning_rate": 5.527505158790075e-05, "loss": 1.0463, "step": 178800 }, { "epoch": 0.45, "learning_rate": 5.527379334641905e-05, "loss": 1.0442, "step": 178805 }, { "epoch": 0.45, "learning_rate": 5.527253510493734e-05, "loss": 1.0434, "step": 178810 }, { "epoch": 0.45, "learning_rate": 5.527127686345564e-05, "loss": 1.0461, "step": 178815 }, { "epoch": 0.45, "learning_rate": 5.527001862197393e-05, "loss": 1.0469, "step": 178820 }, { "epoch": 0.45, "learning_rate": 5.526876038049223e-05, "loss": 1.045, "step": 178825 }, { "epoch": 0.45, "learning_rate": 5.526750213901052e-05, "loss": 1.0455, "step": 178830 }, { "epoch": 0.45, "learning_rate": 5.526624389752881e-05, "loss": 1.0474, "step": 178835 }, { "epoch": 0.45, "learning_rate": 5.526498565604711e-05, "loss": 1.047, "step": 178840 }, { "epoch": 0.45, "learning_rate": 5.52637274145654e-05, "loss": 1.0456, "step": 178845 }, { "epoch": 0.45, "learning_rate": 5.52624691730837e-05, "loss": 1.0446, "step": 178850 }, { "epoch": 0.45, "learning_rate": 5.526121093160199e-05, "loss": 1.0468, "step": 178855 }, { "epoch": 0.45, "learning_rate": 5.525995269012029e-05, "loss": 1.0412, "step": 178860 }, { "epoch": 0.45, "learning_rate": 5.525869444863858e-05, "loss": 1.0445, "step": 178865 }, { "epoch": 0.45, "learning_rate": 5.525743620715688e-05, "loss": 1.0426, "step": 178870 }, { "epoch": 0.45, "learning_rate": 5.525617796567517e-05, "loss": 1.0453, "step": 178875 }, { "epoch": 0.45, "learning_rate": 5.525491972419347e-05, "loss": 1.0455, "step": 178880 }, { "epoch": 0.45, "learning_rate": 5.525366148271176e-05, "loss": 1.0443, "step": 178885 }, { "epoch": 0.45, "learning_rate": 5.525240324123006e-05, "loss": 1.044, "step": 178890 }, { "epoch": 0.45, "learning_rate": 5.525114499974835e-05, "loss": 1.0436, "step": 178895 }, { "epoch": 0.45, "learning_rate": 5.524988675826664e-05, "loss": 1.0443, "step": 178900 }, { "epoch": 0.45, "learning_rate": 5.524862851678494e-05, "loss": 1.0432, "step": 178905 }, { "epoch": 0.45, "learning_rate": 5.524737027530323e-05, "loss": 1.047, "step": 178910 }, { "epoch": 0.45, "learning_rate": 5.524611203382153e-05, "loss": 1.0469, "step": 178915 }, { "epoch": 0.45, "learning_rate": 5.524485379233982e-05, "loss": 1.0463, "step": 178920 }, { "epoch": 0.45, "learning_rate": 5.524359555085812e-05, "loss": 1.0487, "step": 178925 }, { "epoch": 0.45, "learning_rate": 5.524233730937641e-05, "loss": 1.044, "step": 178930 }, { "epoch": 0.45, "learning_rate": 5.524107906789471e-05, "loss": 1.0445, "step": 178935 }, { "epoch": 0.45, "learning_rate": 5.5239820826413e-05, "loss": 1.0463, "step": 178940 }, { "epoch": 0.45, "learning_rate": 5.5238562584931307e-05, "loss": 1.0417, "step": 178945 }, { "epoch": 0.45, "learning_rate": 5.5237304343449605e-05, "loss": 1.0459, "step": 178950 }, { "epoch": 0.45, "learning_rate": 5.5236046101967897e-05, "loss": 1.0465, "step": 178955 }, { "epoch": 0.45, "learning_rate": 5.5234787860486195e-05, "loss": 1.044, "step": 178960 }, { "epoch": 0.45, "learning_rate": 5.5233529619004487e-05, "loss": 1.0445, "step": 178965 }, { "epoch": 0.45, "learning_rate": 5.5232271377522785e-05, "loss": 1.0449, "step": 178970 }, { "epoch": 0.45, "learning_rate": 5.5231013136041076e-05, "loss": 1.0418, "step": 178975 }, { "epoch": 0.45, "learning_rate": 5.522975489455937e-05, "loss": 1.0473, "step": 178980 }, { "epoch": 0.45, "learning_rate": 5.5228496653077666e-05, "loss": 1.0451, "step": 178985 }, { "epoch": 0.45, "learning_rate": 5.522723841159596e-05, "loss": 1.0463, "step": 178990 }, { "epoch": 0.45, "learning_rate": 5.5225980170114256e-05, "loss": 1.0481, "step": 178995 }, { "epoch": 0.45, "learning_rate": 5.522472192863255e-05, "loss": 1.047, "step": 179000 }, { "epoch": 0.45, "learning_rate": 5.5223463687150846e-05, "loss": 1.0447, "step": 179005 }, { "epoch": 0.45, "learning_rate": 5.522220544566914e-05, "loss": 1.0458, "step": 179010 }, { "epoch": 0.45, "learning_rate": 5.5220947204187436e-05, "loss": 1.0463, "step": 179015 }, { "epoch": 0.45, "learning_rate": 5.521968896270573e-05, "loss": 1.0458, "step": 179020 }, { "epoch": 0.45, "learning_rate": 5.5218430721224026e-05, "loss": 1.0481, "step": 179025 }, { "epoch": 0.45, "learning_rate": 5.521717247974232e-05, "loss": 1.0464, "step": 179030 }, { "epoch": 0.45, "learning_rate": 5.5215914238260616e-05, "loss": 1.0451, "step": 179035 }, { "epoch": 0.45, "learning_rate": 5.521465599677891e-05, "loss": 1.0665, "step": 179040 }, { "epoch": 0.45, "learning_rate": 5.52133977552972e-05, "loss": 1.0471, "step": 179045 }, { "epoch": 0.45, "learning_rate": 5.52121395138155e-05, "loss": 1.0453, "step": 179050 }, { "epoch": 0.45, "learning_rate": 5.521088127233379e-05, "loss": 1.0457, "step": 179055 }, { "epoch": 0.45, "learning_rate": 5.520962303085209e-05, "loss": 1.0432, "step": 179060 }, { "epoch": 0.45, "learning_rate": 5.520836478937038e-05, "loss": 1.0436, "step": 179065 }, { "epoch": 0.45, "learning_rate": 5.520710654788868e-05, "loss": 1.0473, "step": 179070 }, { "epoch": 0.45, "learning_rate": 5.520584830640697e-05, "loss": 1.0476, "step": 179075 }, { "epoch": 0.45, "learning_rate": 5.520459006492527e-05, "loss": 1.044, "step": 179080 }, { "epoch": 0.45, "learning_rate": 5.520333182344356e-05, "loss": 1.0448, "step": 179085 }, { "epoch": 0.45, "learning_rate": 5.520207358196186e-05, "loss": 1.0439, "step": 179090 }, { "epoch": 0.45, "learning_rate": 5.520081534048015e-05, "loss": 1.0444, "step": 179095 }, { "epoch": 0.45, "learning_rate": 5.519955709899845e-05, "loss": 1.0445, "step": 179100 }, { "epoch": 0.45, "learning_rate": 5.519829885751674e-05, "loss": 1.0451, "step": 179105 }, { "epoch": 0.45, "learning_rate": 5.519704061603503e-05, "loss": 1.0453, "step": 179110 }, { "epoch": 0.45, "learning_rate": 5.519578237455333e-05, "loss": 1.0455, "step": 179115 }, { "epoch": 0.45, "learning_rate": 5.519452413307162e-05, "loss": 1.0439, "step": 179120 }, { "epoch": 0.45, "learning_rate": 5.519326589158992e-05, "loss": 1.0475, "step": 179125 }, { "epoch": 0.45, "learning_rate": 5.519200765010821e-05, "loss": 1.046, "step": 179130 }, { "epoch": 0.45, "learning_rate": 5.519074940862651e-05, "loss": 1.043, "step": 179135 }, { "epoch": 0.45, "learning_rate": 5.51894911671448e-05, "loss": 1.0467, "step": 179140 }, { "epoch": 0.45, "learning_rate": 5.51882329256631e-05, "loss": 1.0439, "step": 179145 }, { "epoch": 0.45, "learning_rate": 5.518697468418139e-05, "loss": 1.0446, "step": 179150 }, { "epoch": 0.45, "learning_rate": 5.518571644269969e-05, "loss": 1.046, "step": 179155 }, { "epoch": 0.45, "learning_rate": 5.518445820121798e-05, "loss": 1.0445, "step": 179160 }, { "epoch": 0.45, "learning_rate": 5.518319995973628e-05, "loss": 1.0665, "step": 179165 }, { "epoch": 0.45, "learning_rate": 5.518194171825457e-05, "loss": 1.0463, "step": 179170 }, { "epoch": 0.45, "learning_rate": 5.518068347677286e-05, "loss": 1.0457, "step": 179175 }, { "epoch": 0.45, "learning_rate": 5.517942523529116e-05, "loss": 1.0465, "step": 179180 }, { "epoch": 0.45, "learning_rate": 5.517816699380945e-05, "loss": 1.0462, "step": 179185 }, { "epoch": 0.45, "learning_rate": 5.517690875232775e-05, "loss": 1.0452, "step": 179190 }, { "epoch": 0.45, "learning_rate": 5.517565051084604e-05, "loss": 1.0446, "step": 179195 }, { "epoch": 0.45, "learning_rate": 5.517439226936434e-05, "loss": 1.0478, "step": 179200 }, { "epoch": 0.45, "learning_rate": 5.517313402788263e-05, "loss": 1.0438, "step": 179205 }, { "epoch": 0.45, "learning_rate": 5.517187578640093e-05, "loss": 1.0456, "step": 179210 }, { "epoch": 0.45, "learning_rate": 5.517061754491922e-05, "loss": 1.0448, "step": 179215 }, { "epoch": 0.45, "learning_rate": 5.516935930343752e-05, "loss": 1.0467, "step": 179220 }, { "epoch": 0.45, "learning_rate": 5.516810106195581e-05, "loss": 1.0486, "step": 179225 }, { "epoch": 0.45, "learning_rate": 5.516684282047411e-05, "loss": 1.0428, "step": 179230 }, { "epoch": 0.45, "learning_rate": 5.51655845789924e-05, "loss": 1.0447, "step": 179235 }, { "epoch": 0.45, "learning_rate": 5.5164326337510694e-05, "loss": 1.0415, "step": 179240 }, { "epoch": 0.45, "learning_rate": 5.516306809602899e-05, "loss": 1.0469, "step": 179245 }, { "epoch": 0.45, "learning_rate": 5.5161809854547284e-05, "loss": 1.0447, "step": 179250 }, { "epoch": 0.45, "learning_rate": 5.516055161306558e-05, "loss": 1.0439, "step": 179255 }, { "epoch": 0.45, "learning_rate": 5.5159293371583874e-05, "loss": 1.0669, "step": 179260 }, { "epoch": 0.45, "learning_rate": 5.515803513010217e-05, "loss": 1.0453, "step": 179265 }, { "epoch": 0.45, "learning_rate": 5.5156776888620464e-05, "loss": 1.0443, "step": 179270 }, { "epoch": 0.45, "learning_rate": 5.515551864713876e-05, "loss": 1.0468, "step": 179275 }, { "epoch": 0.45, "learning_rate": 5.5154260405657054e-05, "loss": 1.0456, "step": 179280 }, { "epoch": 0.45, "learning_rate": 5.515300216417535e-05, "loss": 1.0434, "step": 179285 }, { "epoch": 0.45, "learning_rate": 5.5151743922693644e-05, "loss": 1.0451, "step": 179290 }, { "epoch": 0.45, "learning_rate": 5.5150485681211935e-05, "loss": 1.0503, "step": 179295 }, { "epoch": 0.45, "learning_rate": 5.5149227439730234e-05, "loss": 1.0468, "step": 179300 }, { "epoch": 0.45, "learning_rate": 5.5147969198248525e-05, "loss": 1.0455, "step": 179305 }, { "epoch": 0.45, "learning_rate": 5.5146710956766824e-05, "loss": 1.044, "step": 179310 }, { "epoch": 0.45, "learning_rate": 5.5145452715285115e-05, "loss": 1.0425, "step": 179315 }, { "epoch": 0.45, "learning_rate": 5.5144194473803414e-05, "loss": 1.0437, "step": 179320 }, { "epoch": 0.45, "learning_rate": 5.5142936232321705e-05, "loss": 1.0436, "step": 179325 }, { "epoch": 0.45, "learning_rate": 5.5141677990840004e-05, "loss": 1.0472, "step": 179330 }, { "epoch": 0.45, "learning_rate": 5.5140419749358295e-05, "loss": 1.0427, "step": 179335 }, { "epoch": 0.45, "learning_rate": 5.5139161507876593e-05, "loss": 1.0455, "step": 179340 }, { "epoch": 0.45, "learning_rate": 5.5137903266394885e-05, "loss": 1.0459, "step": 179345 }, { "epoch": 0.45, "learning_rate": 5.5136645024913183e-05, "loss": 1.0443, "step": 179350 }, { "epoch": 0.45, "learning_rate": 5.5135386783431475e-05, "loss": 1.0435, "step": 179355 }, { "epoch": 0.45, "learning_rate": 5.513412854194977e-05, "loss": 1.0441, "step": 179360 }, { "epoch": 0.45, "learning_rate": 5.5132870300468065e-05, "loss": 1.0448, "step": 179365 }, { "epoch": 0.45, "learning_rate": 5.513161205898636e-05, "loss": 1.0577, "step": 179370 }, { "epoch": 0.45, "learning_rate": 5.5130353817504655e-05, "loss": 1.0464, "step": 179375 }, { "epoch": 0.45, "learning_rate": 5.5129095576022947e-05, "loss": 1.0445, "step": 179380 }, { "epoch": 0.45, "learning_rate": 5.5127837334541245e-05, "loss": 1.0603, "step": 179385 }, { "epoch": 0.45, "learning_rate": 5.5126579093059537e-05, "loss": 1.0464, "step": 179390 }, { "epoch": 0.45, "learning_rate": 5.5125320851577835e-05, "loss": 1.0457, "step": 179395 }, { "epoch": 0.45, "learning_rate": 5.5124062610096127e-05, "loss": 1.0468, "step": 179400 }, { "epoch": 0.45, "learning_rate": 5.5122804368614425e-05, "loss": 1.0659, "step": 179405 }, { "epoch": 0.45, "learning_rate": 5.5121546127132717e-05, "loss": 1.0461, "step": 179410 }, { "epoch": 0.45, "learning_rate": 5.5120287885651015e-05, "loss": 1.0445, "step": 179415 }, { "epoch": 0.45, "learning_rate": 5.5119029644169306e-05, "loss": 1.0447, "step": 179420 }, { "epoch": 0.45, "learning_rate": 5.51177714026876e-05, "loss": 1.045, "step": 179425 }, { "epoch": 0.45, "learning_rate": 5.5116513161205896e-05, "loss": 1.0455, "step": 179430 }, { "epoch": 0.45, "learning_rate": 5.511525491972419e-05, "loss": 1.0443, "step": 179435 }, { "epoch": 0.45, "learning_rate": 5.5113996678242486e-05, "loss": 1.0452, "step": 179440 }, { "epoch": 0.45, "learning_rate": 5.511273843676079e-05, "loss": 1.0451, "step": 179445 }, { "epoch": 0.45, "learning_rate": 5.511148019527908e-05, "loss": 1.0406, "step": 179450 }, { "epoch": 0.45, "learning_rate": 5.511022195379738e-05, "loss": 1.0434, "step": 179455 }, { "epoch": 0.45, "learning_rate": 5.510896371231567e-05, "loss": 1.045, "step": 179460 }, { "epoch": 0.45, "learning_rate": 5.510770547083397e-05, "loss": 1.0457, "step": 179465 }, { "epoch": 0.45, "learning_rate": 5.510644722935226e-05, "loss": 1.0456, "step": 179470 }, { "epoch": 0.45, "learning_rate": 5.510518898787056e-05, "loss": 1.0461, "step": 179475 }, { "epoch": 0.45, "learning_rate": 5.510393074638885e-05, "loss": 1.0468, "step": 179480 }, { "epoch": 0.45, "learning_rate": 5.510267250490715e-05, "loss": 1.0471, "step": 179485 }, { "epoch": 0.45, "learning_rate": 5.510141426342544e-05, "loss": 1.0464, "step": 179490 }, { "epoch": 0.45, "learning_rate": 5.510015602194374e-05, "loss": 1.0471, "step": 179495 }, { "epoch": 0.45, "learning_rate": 5.509889778046203e-05, "loss": 1.0447, "step": 179500 }, { "epoch": 0.45, "learning_rate": 5.5097639538980325e-05, "loss": 1.0434, "step": 179505 }, { "epoch": 0.45, "learning_rate": 5.509638129749862e-05, "loss": 1.0492, "step": 179510 }, { "epoch": 0.45, "learning_rate": 5.5095123056016915e-05, "loss": 1.0462, "step": 179515 }, { "epoch": 0.45, "learning_rate": 5.509386481453521e-05, "loss": 1.0446, "step": 179520 }, { "epoch": 0.45, "learning_rate": 5.5092606573053505e-05, "loss": 1.0452, "step": 179525 }, { "epoch": 0.45, "learning_rate": 5.50913483315718e-05, "loss": 1.0437, "step": 179530 }, { "epoch": 0.45, "learning_rate": 5.5090090090090095e-05, "loss": 1.0448, "step": 179535 }, { "epoch": 0.45, "learning_rate": 5.508883184860839e-05, "loss": 1.0712, "step": 179540 }, { "epoch": 0.45, "learning_rate": 5.5087573607126685e-05, "loss": 1.0435, "step": 179545 }, { "epoch": 0.45, "learning_rate": 5.508631536564498e-05, "loss": 1.0434, "step": 179550 }, { "epoch": 0.45, "learning_rate": 5.5085057124163274e-05, "loss": 1.0459, "step": 179555 }, { "epoch": 0.45, "learning_rate": 5.508379888268157e-05, "loss": 1.0475, "step": 179560 }, { "epoch": 0.45, "learning_rate": 5.5082540641199864e-05, "loss": 1.046, "step": 179565 }, { "epoch": 0.45, "learning_rate": 5.5081282399718156e-05, "loss": 1.046, "step": 179570 }, { "epoch": 0.45, "learning_rate": 5.5080024158236454e-05, "loss": 1.0465, "step": 179575 }, { "epoch": 0.45, "learning_rate": 5.5078765916754746e-05, "loss": 1.0447, "step": 179580 }, { "epoch": 0.45, "learning_rate": 5.5077507675273044e-05, "loss": 1.0461, "step": 179585 }, { "epoch": 0.45, "learning_rate": 5.5076249433791336e-05, "loss": 1.0416, "step": 179590 }, { "epoch": 0.45, "learning_rate": 5.5074991192309634e-05, "loss": 1.0418, "step": 179595 }, { "epoch": 0.45, "learning_rate": 5.5073732950827926e-05, "loss": 1.0477, "step": 179600 }, { "epoch": 0.45, "learning_rate": 5.5072474709346224e-05, "loss": 1.0428, "step": 179605 }, { "epoch": 0.45, "learning_rate": 5.5071216467864516e-05, "loss": 1.0487, "step": 179610 }, { "epoch": 0.45, "learning_rate": 5.5069958226382814e-05, "loss": 1.0488, "step": 179615 }, { "epoch": 0.45, "learning_rate": 5.5068699984901106e-05, "loss": 1.0458, "step": 179620 }, { "epoch": 0.45, "learning_rate": 5.5067441743419404e-05, "loss": 1.0434, "step": 179625 }, { "epoch": 0.45, "learning_rate": 5.5066183501937696e-05, "loss": 1.0481, "step": 179630 }, { "epoch": 0.45, "learning_rate": 5.506492526045599e-05, "loss": 1.0471, "step": 179635 }, { "epoch": 0.45, "learning_rate": 5.5063667018974286e-05, "loss": 1.0426, "step": 179640 }, { "epoch": 0.45, "learning_rate": 5.506240877749258e-05, "loss": 1.0456, "step": 179645 }, { "epoch": 0.45, "learning_rate": 5.5061150536010876e-05, "loss": 1.0683, "step": 179650 }, { "epoch": 0.45, "learning_rate": 5.505989229452917e-05, "loss": 1.0686, "step": 179655 }, { "epoch": 0.45, "learning_rate": 5.5058634053047466e-05, "loss": 1.0429, "step": 179660 }, { "epoch": 0.45, "learning_rate": 5.505737581156576e-05, "loss": 1.0485, "step": 179665 }, { "epoch": 0.45, "learning_rate": 5.5056117570084056e-05, "loss": 1.044, "step": 179670 }, { "epoch": 0.45, "learning_rate": 5.505485932860235e-05, "loss": 1.0445, "step": 179675 }, { "epoch": 0.45, "learning_rate": 5.5053601087120646e-05, "loss": 1.0452, "step": 179680 }, { "epoch": 0.45, "learning_rate": 5.505234284563894e-05, "loss": 1.0458, "step": 179685 }, { "epoch": 0.45, "learning_rate": 5.5051084604157236e-05, "loss": 1.0462, "step": 179690 }, { "epoch": 0.45, "learning_rate": 5.504982636267553e-05, "loss": 1.046, "step": 179695 }, { "epoch": 0.45, "learning_rate": 5.504856812119382e-05, "loss": 1.0447, "step": 179700 }, { "epoch": 0.45, "learning_rate": 5.504730987971212e-05, "loss": 1.0432, "step": 179705 }, { "epoch": 0.45, "learning_rate": 5.504605163823041e-05, "loss": 1.0464, "step": 179710 }, { "epoch": 0.45, "learning_rate": 5.504479339674871e-05, "loss": 1.0448, "step": 179715 }, { "epoch": 0.45, "learning_rate": 5.5043535155267e-05, "loss": 1.0419, "step": 179720 }, { "epoch": 0.45, "learning_rate": 5.50422769137853e-05, "loss": 1.0462, "step": 179725 }, { "epoch": 0.45, "learning_rate": 5.504101867230359e-05, "loss": 1.0452, "step": 179730 }, { "epoch": 0.45, "learning_rate": 5.503976043082189e-05, "loss": 1.0435, "step": 179735 }, { "epoch": 0.45, "learning_rate": 5.503850218934018e-05, "loss": 1.0451, "step": 179740 }, { "epoch": 0.45, "learning_rate": 5.503724394785848e-05, "loss": 1.0453, "step": 179745 }, { "epoch": 0.45, "learning_rate": 5.503598570637677e-05, "loss": 1.0424, "step": 179750 }, { "epoch": 0.45, "learning_rate": 5.503472746489507e-05, "loss": 1.0469, "step": 179755 }, { "epoch": 0.45, "learning_rate": 5.503346922341336e-05, "loss": 1.0478, "step": 179760 }, { "epoch": 0.45, "learning_rate": 5.503221098193165e-05, "loss": 1.046, "step": 179765 }, { "epoch": 0.45, "learning_rate": 5.503095274044995e-05, "loss": 1.0628, "step": 179770 }, { "epoch": 0.45, "learning_rate": 5.502969449896824e-05, "loss": 1.0453, "step": 179775 }, { "epoch": 0.45, "learning_rate": 5.502843625748654e-05, "loss": 1.0472, "step": 179780 }, { "epoch": 0.45, "learning_rate": 5.502717801600483e-05, "loss": 1.0447, "step": 179785 }, { "epoch": 0.45, "learning_rate": 5.502591977452313e-05, "loss": 1.0458, "step": 179790 }, { "epoch": 0.45, "learning_rate": 5.502466153304142e-05, "loss": 1.0467, "step": 179795 }, { "epoch": 0.45, "learning_rate": 5.502340329155972e-05, "loss": 1.0441, "step": 179800 }, { "epoch": 0.45, "learning_rate": 5.502214505007801e-05, "loss": 1.0459, "step": 179805 }, { "epoch": 0.45, "learning_rate": 5.502088680859631e-05, "loss": 1.0473, "step": 179810 }, { "epoch": 0.45, "learning_rate": 5.50196285671146e-05, "loss": 1.0474, "step": 179815 }, { "epoch": 0.45, "learning_rate": 5.50183703256329e-05, "loss": 1.0468, "step": 179820 }, { "epoch": 0.45, "learning_rate": 5.501711208415119e-05, "loss": 1.041, "step": 179825 }, { "epoch": 0.45, "learning_rate": 5.501585384266948e-05, "loss": 1.0713, "step": 179830 }, { "epoch": 0.45, "learning_rate": 5.501459560118778e-05, "loss": 1.0473, "step": 179835 }, { "epoch": 0.45, "learning_rate": 5.501333735970607e-05, "loss": 1.0465, "step": 179840 }, { "epoch": 0.45, "learning_rate": 5.501207911822437e-05, "loss": 1.0452, "step": 179845 }, { "epoch": 0.45, "learning_rate": 5.501082087674266e-05, "loss": 1.0441, "step": 179850 }, { "epoch": 0.45, "learning_rate": 5.500956263526096e-05, "loss": 1.0462, "step": 179855 }, { "epoch": 0.45, "learning_rate": 5.500830439377925e-05, "loss": 1.0458, "step": 179860 }, { "epoch": 0.45, "learning_rate": 5.500704615229755e-05, "loss": 1.0477, "step": 179865 }, { "epoch": 0.45, "learning_rate": 5.500578791081584e-05, "loss": 1.0457, "step": 179870 }, { "epoch": 0.45, "learning_rate": 5.500452966933414e-05, "loss": 1.0435, "step": 179875 }, { "epoch": 0.45, "learning_rate": 5.500327142785243e-05, "loss": 1.0451, "step": 179880 }, { "epoch": 0.45, "learning_rate": 5.500201318637073e-05, "loss": 1.0459, "step": 179885 }, { "epoch": 0.45, "learning_rate": 5.500075494488902e-05, "loss": 1.048, "step": 179890 }, { "epoch": 0.45, "learning_rate": 5.499949670340731e-05, "loss": 1.044, "step": 179895 }, { "epoch": 0.45, "learning_rate": 5.499823846192561e-05, "loss": 1.0437, "step": 179900 }, { "epoch": 0.45, "learning_rate": 5.49969802204439e-05, "loss": 1.0475, "step": 179905 }, { "epoch": 0.45, "learning_rate": 5.49957219789622e-05, "loss": 1.0443, "step": 179910 }, { "epoch": 0.45, "learning_rate": 5.499446373748049e-05, "loss": 1.045, "step": 179915 }, { "epoch": 0.45, "learning_rate": 5.499320549599879e-05, "loss": 1.0475, "step": 179920 }, { "epoch": 0.45, "learning_rate": 5.499194725451708e-05, "loss": 1.0458, "step": 179925 }, { "epoch": 0.45, "learning_rate": 5.499068901303538e-05, "loss": 1.0444, "step": 179930 }, { "epoch": 0.45, "learning_rate": 5.498943077155367e-05, "loss": 1.0449, "step": 179935 }, { "epoch": 0.45, "learning_rate": 5.498817253007197e-05, "loss": 1.0454, "step": 179940 }, { "epoch": 0.45, "learning_rate": 5.4986914288590277e-05, "loss": 1.0431, "step": 179945 }, { "epoch": 0.45, "learning_rate": 5.498565604710857e-05, "loss": 1.0462, "step": 179950 }, { "epoch": 0.45, "learning_rate": 5.4984397805626867e-05, "loss": 1.0441, "step": 179955 }, { "epoch": 0.45, "learning_rate": 5.498313956414516e-05, "loss": 1.0475, "step": 179960 }, { "epoch": 0.45, "learning_rate": 5.4981881322663457e-05, "loss": 1.0453, "step": 179965 }, { "epoch": 0.45, "learning_rate": 5.498062308118175e-05, "loss": 1.0445, "step": 179970 }, { "epoch": 0.45, "learning_rate": 5.497936483970004e-05, "loss": 1.0462, "step": 179975 }, { "epoch": 0.45, "learning_rate": 5.497810659821834e-05, "loss": 1.0464, "step": 179980 }, { "epoch": 0.45, "learning_rate": 5.497684835673663e-05, "loss": 1.045, "step": 179985 }, { "epoch": 0.45, "learning_rate": 5.497559011525493e-05, "loss": 1.048, "step": 179990 }, { "epoch": 0.45, "learning_rate": 5.497433187377322e-05, "loss": 1.0476, "step": 179995 }, { "epoch": 0.45, "learning_rate": 5.497307363229152e-05, "loss": 1.0456, "step": 180000 }, { "epoch": 0.45, "learning_rate": 5.497181539080981e-05, "loss": 1.0452, "step": 180005 }, { "epoch": 0.45, "learning_rate": 5.497055714932811e-05, "loss": 1.0474, "step": 180010 }, { "epoch": 0.45, "learning_rate": 5.49692989078464e-05, "loss": 1.0463, "step": 180015 }, { "epoch": 0.45, "learning_rate": 5.49680406663647e-05, "loss": 1.0417, "step": 180020 }, { "epoch": 0.45, "learning_rate": 5.496678242488299e-05, "loss": 1.0454, "step": 180025 }, { "epoch": 0.45, "learning_rate": 5.496552418340129e-05, "loss": 1.045, "step": 180030 }, { "epoch": 0.45, "learning_rate": 5.496426594191958e-05, "loss": 1.0428, "step": 180035 }, { "epoch": 0.45, "learning_rate": 5.496300770043787e-05, "loss": 1.044, "step": 180040 }, { "epoch": 0.45, "learning_rate": 5.496174945895617e-05, "loss": 1.0488, "step": 180045 }, { "epoch": 0.45, "learning_rate": 5.496049121747446e-05, "loss": 1.0483, "step": 180050 }, { "epoch": 0.45, "learning_rate": 5.495923297599276e-05, "loss": 1.0455, "step": 180055 }, { "epoch": 0.45, "learning_rate": 5.495797473451105e-05, "loss": 1.0466, "step": 180060 }, { "epoch": 0.45, "learning_rate": 5.495671649302935e-05, "loss": 1.0426, "step": 180065 }, { "epoch": 0.45, "learning_rate": 5.495545825154764e-05, "loss": 1.0424, "step": 180070 }, { "epoch": 0.45, "learning_rate": 5.495420001006594e-05, "loss": 1.0449, "step": 180075 }, { "epoch": 0.45, "learning_rate": 5.495294176858423e-05, "loss": 1.0446, "step": 180080 }, { "epoch": 0.45, "learning_rate": 5.495168352710253e-05, "loss": 1.0458, "step": 180085 }, { "epoch": 0.45, "learning_rate": 5.495042528562082e-05, "loss": 1.0456, "step": 180090 }, { "epoch": 0.45, "learning_rate": 5.494916704413911e-05, "loss": 1.0478, "step": 180095 }, { "epoch": 0.45, "learning_rate": 5.494790880265741e-05, "loss": 1.046, "step": 180100 }, { "epoch": 0.45, "learning_rate": 5.49466505611757e-05, "loss": 1.0436, "step": 180105 }, { "epoch": 0.45, "learning_rate": 5.4945392319694e-05, "loss": 1.0476, "step": 180110 }, { "epoch": 0.45, "learning_rate": 5.494413407821229e-05, "loss": 1.0437, "step": 180115 }, { "epoch": 0.45, "learning_rate": 5.494287583673059e-05, "loss": 1.0471, "step": 180120 }, { "epoch": 0.45, "learning_rate": 5.494161759524888e-05, "loss": 1.0453, "step": 180125 }, { "epoch": 0.45, "learning_rate": 5.494035935376718e-05, "loss": 1.0478, "step": 180130 }, { "epoch": 0.45, "learning_rate": 5.493910111228547e-05, "loss": 1.043, "step": 180135 }, { "epoch": 0.45, "learning_rate": 5.493784287080377e-05, "loss": 1.0452, "step": 180140 }, { "epoch": 0.45, "learning_rate": 5.493658462932206e-05, "loss": 1.0451, "step": 180145 }, { "epoch": 0.45, "learning_rate": 5.493532638784036e-05, "loss": 1.0719, "step": 180150 }, { "epoch": 0.45, "learning_rate": 5.493406814635865e-05, "loss": 1.065, "step": 180155 }, { "epoch": 0.45, "learning_rate": 5.4932809904876944e-05, "loss": 1.0476, "step": 180160 }, { "epoch": 0.45, "learning_rate": 5.493155166339524e-05, "loss": 1.0475, "step": 180165 }, { "epoch": 0.45, "learning_rate": 5.4930293421913534e-05, "loss": 1.0439, "step": 180170 }, { "epoch": 0.45, "learning_rate": 5.492903518043183e-05, "loss": 1.0479, "step": 180175 }, { "epoch": 0.45, "learning_rate": 5.4927776938950124e-05, "loss": 1.0448, "step": 180180 }, { "epoch": 0.45, "learning_rate": 5.492651869746842e-05, "loss": 1.0427, "step": 180185 }, { "epoch": 0.45, "learning_rate": 5.4925260455986714e-05, "loss": 1.0418, "step": 180190 }, { "epoch": 0.45, "learning_rate": 5.492400221450501e-05, "loss": 1.045, "step": 180195 }, { "epoch": 0.45, "learning_rate": 5.4922743973023304e-05, "loss": 1.0464, "step": 180200 }, { "epoch": 0.45, "learning_rate": 5.49214857315416e-05, "loss": 1.0443, "step": 180205 }, { "epoch": 0.45, "learning_rate": 5.4920227490059894e-05, "loss": 1.0454, "step": 180210 }, { "epoch": 0.45, "learning_rate": 5.491896924857819e-05, "loss": 1.0431, "step": 180215 }, { "epoch": 0.45, "learning_rate": 5.4917711007096484e-05, "loss": 1.0468, "step": 180220 }, { "epoch": 0.45, "learning_rate": 5.4916452765614775e-05, "loss": 1.0425, "step": 180225 }, { "epoch": 0.45, "learning_rate": 5.4915194524133074e-05, "loss": 1.0444, "step": 180230 }, { "epoch": 0.45, "learning_rate": 5.4913936282651365e-05, "loss": 1.043, "step": 180235 }, { "epoch": 0.45, "learning_rate": 5.4912678041169664e-05, "loss": 1.0522, "step": 180240 }, { "epoch": 0.45, "learning_rate": 5.4911419799687955e-05, "loss": 1.0419, "step": 180245 }, { "epoch": 0.45, "learning_rate": 5.4910161558206254e-05, "loss": 1.0465, "step": 180250 }, { "epoch": 0.45, "learning_rate": 5.4908903316724545e-05, "loss": 1.0446, "step": 180255 }, { "epoch": 0.45, "learning_rate": 5.4907645075242844e-05, "loss": 1.0436, "step": 180260 }, { "epoch": 0.45, "learning_rate": 5.4906386833761135e-05, "loss": 1.0468, "step": 180265 }, { "epoch": 0.45, "learning_rate": 5.4905128592279434e-05, "loss": 1.0437, "step": 180270 }, { "epoch": 0.45, "learning_rate": 5.4903870350797725e-05, "loss": 1.0485, "step": 180275 }, { "epoch": 0.45, "learning_rate": 5.4902612109316024e-05, "loss": 1.0464, "step": 180280 }, { "epoch": 0.45, "learning_rate": 5.4901353867834315e-05, "loss": 1.0452, "step": 180285 }, { "epoch": 0.45, "learning_rate": 5.490009562635261e-05, "loss": 1.0444, "step": 180290 }, { "epoch": 0.45, "learning_rate": 5.4898837384870905e-05, "loss": 1.0453, "step": 180295 }, { "epoch": 0.45, "learning_rate": 5.48975791433892e-05, "loss": 1.0464, "step": 180300 }, { "epoch": 0.45, "learning_rate": 5.4896320901907495e-05, "loss": 1.0442, "step": 180305 }, { "epoch": 0.45, "learning_rate": 5.489506266042579e-05, "loss": 1.0456, "step": 180310 }, { "epoch": 0.45, "learning_rate": 5.4893804418944085e-05, "loss": 1.0435, "step": 180315 }, { "epoch": 0.45, "learning_rate": 5.489254617746238e-05, "loss": 1.0474, "step": 180320 }, { "epoch": 0.45, "learning_rate": 5.4891287935980675e-05, "loss": 1.0441, "step": 180325 }, { "epoch": 0.45, "learning_rate": 5.489002969449897e-05, "loss": 1.0466, "step": 180330 }, { "epoch": 0.45, "learning_rate": 5.4888771453017265e-05, "loss": 1.0443, "step": 180335 }, { "epoch": 0.45, "learning_rate": 5.488751321153556e-05, "loss": 1.046, "step": 180340 }, { "epoch": 0.45, "learning_rate": 5.4886254970053855e-05, "loss": 1.0446, "step": 180345 }, { "epoch": 0.45, "learning_rate": 5.488499672857215e-05, "loss": 1.0454, "step": 180350 }, { "epoch": 0.45, "learning_rate": 5.488373848709044e-05, "loss": 1.0473, "step": 180355 }, { "epoch": 0.45, "learning_rate": 5.488248024560874e-05, "loss": 1.0475, "step": 180360 }, { "epoch": 0.45, "learning_rate": 5.488122200412703e-05, "loss": 1.0475, "step": 180365 }, { "epoch": 0.45, "learning_rate": 5.487996376264533e-05, "loss": 1.0469, "step": 180370 }, { "epoch": 0.45, "learning_rate": 5.487870552116362e-05, "loss": 1.0435, "step": 180375 }, { "epoch": 0.45, "learning_rate": 5.487744727968192e-05, "loss": 1.0484, "step": 180380 }, { "epoch": 0.45, "learning_rate": 5.487618903820021e-05, "loss": 1.0459, "step": 180385 }, { "epoch": 0.45, "learning_rate": 5.4874930796718507e-05, "loss": 1.043, "step": 180390 }, { "epoch": 0.45, "learning_rate": 5.48736725552368e-05, "loss": 1.0456, "step": 180395 }, { "epoch": 0.45, "learning_rate": 5.4872414313755097e-05, "loss": 1.0449, "step": 180400 }, { "epoch": 0.45, "learning_rate": 5.487115607227339e-05, "loss": 1.0454, "step": 180405 }, { "epoch": 0.45, "learning_rate": 5.4869897830791687e-05, "loss": 1.0454, "step": 180410 }, { "epoch": 0.45, "learning_rate": 5.486863958930998e-05, "loss": 1.047, "step": 180415 }, { "epoch": 0.45, "learning_rate": 5.486738134782827e-05, "loss": 1.0467, "step": 180420 }, { "epoch": 0.45, "learning_rate": 5.486612310634657e-05, "loss": 1.0458, "step": 180425 }, { "epoch": 0.45, "learning_rate": 5.486486486486486e-05, "loss": 1.045, "step": 180430 }, { "epoch": 0.45, "learning_rate": 5.486360662338316e-05, "loss": 1.0454, "step": 180435 }, { "epoch": 0.45, "learning_rate": 5.486234838190145e-05, "loss": 1.0439, "step": 180440 }, { "epoch": 0.45, "learning_rate": 5.4861090140419755e-05, "loss": 1.047, "step": 180445 }, { "epoch": 0.45, "learning_rate": 5.485983189893805e-05, "loss": 1.0484, "step": 180450 }, { "epoch": 0.45, "learning_rate": 5.4858573657456345e-05, "loss": 1.0461, "step": 180455 }, { "epoch": 0.45, "learning_rate": 5.485731541597464e-05, "loss": 1.0435, "step": 180460 }, { "epoch": 0.45, "learning_rate": 5.4856057174492935e-05, "loss": 1.0463, "step": 180465 }, { "epoch": 0.45, "learning_rate": 5.485479893301123e-05, "loss": 1.0432, "step": 180470 }, { "epoch": 0.45, "learning_rate": 5.4853540691529525e-05, "loss": 1.0448, "step": 180475 }, { "epoch": 0.45, "learning_rate": 5.485228245004782e-05, "loss": 1.047, "step": 180480 }, { "epoch": 0.45, "learning_rate": 5.4851024208566115e-05, "loss": 1.0398, "step": 180485 }, { "epoch": 0.45, "learning_rate": 5.484976596708441e-05, "loss": 1.0457, "step": 180490 }, { "epoch": 0.45, "learning_rate": 5.4848507725602705e-05, "loss": 1.0468, "step": 180495 }, { "epoch": 0.45, "learning_rate": 5.4847249484120996e-05, "loss": 1.0453, "step": 180500 }, { "epoch": 0.45, "learning_rate": 5.4845991242639295e-05, "loss": 1.0461, "step": 180505 }, { "epoch": 0.45, "learning_rate": 5.4844733001157586e-05, "loss": 1.0445, "step": 180510 }, { "epoch": 0.45, "learning_rate": 5.4843474759675885e-05, "loss": 1.0439, "step": 180515 }, { "epoch": 0.45, "learning_rate": 5.4842216518194176e-05, "loss": 1.0483, "step": 180520 }, { "epoch": 0.45, "learning_rate": 5.4840958276712475e-05, "loss": 1.0442, "step": 180525 }, { "epoch": 0.45, "learning_rate": 5.4839700035230766e-05, "loss": 1.0452, "step": 180530 }, { "epoch": 0.45, "learning_rate": 5.4838441793749065e-05, "loss": 1.0472, "step": 180535 }, { "epoch": 0.45, "learning_rate": 5.4837183552267356e-05, "loss": 1.044, "step": 180540 }, { "epoch": 0.45, "learning_rate": 5.4835925310785655e-05, "loss": 1.0467, "step": 180545 }, { "epoch": 0.45, "learning_rate": 5.4834667069303946e-05, "loss": 1.0419, "step": 180550 }, { "epoch": 0.45, "learning_rate": 5.4833408827822244e-05, "loss": 1.0452, "step": 180555 }, { "epoch": 0.45, "learning_rate": 5.4832150586340536e-05, "loss": 1.0453, "step": 180560 }, { "epoch": 0.45, "learning_rate": 5.483089234485883e-05, "loss": 1.045, "step": 180565 }, { "epoch": 0.45, "learning_rate": 5.4829634103377126e-05, "loss": 1.0488, "step": 180570 }, { "epoch": 0.45, "learning_rate": 5.482837586189542e-05, "loss": 1.0666, "step": 180575 }, { "epoch": 0.45, "learning_rate": 5.4827117620413716e-05, "loss": 1.0435, "step": 180580 }, { "epoch": 0.45, "learning_rate": 5.482585937893201e-05, "loss": 1.0464, "step": 180585 }, { "epoch": 0.45, "learning_rate": 5.4824601137450306e-05, "loss": 1.0465, "step": 180590 }, { "epoch": 0.45, "learning_rate": 5.48233428959686e-05, "loss": 1.0448, "step": 180595 }, { "epoch": 0.45, "learning_rate": 5.4822084654486896e-05, "loss": 1.0463, "step": 180600 }, { "epoch": 0.45, "learning_rate": 5.482082641300519e-05, "loss": 1.046, "step": 180605 }, { "epoch": 0.45, "learning_rate": 5.4819568171523486e-05, "loss": 1.0471, "step": 180610 }, { "epoch": 0.45, "learning_rate": 5.481830993004178e-05, "loss": 1.0447, "step": 180615 }, { "epoch": 0.45, "learning_rate": 5.4817051688560076e-05, "loss": 1.0464, "step": 180620 }, { "epoch": 0.45, "learning_rate": 5.481579344707837e-05, "loss": 1.0451, "step": 180625 }, { "epoch": 0.45, "learning_rate": 5.481453520559666e-05, "loss": 1.0448, "step": 180630 }, { "epoch": 0.45, "learning_rate": 5.481327696411496e-05, "loss": 1.0453, "step": 180635 }, { "epoch": 0.45, "learning_rate": 5.481201872263325e-05, "loss": 1.0448, "step": 180640 }, { "epoch": 0.45, "learning_rate": 5.481076048115155e-05, "loss": 1.0473, "step": 180645 }, { "epoch": 0.45, "learning_rate": 5.480950223966984e-05, "loss": 1.0478, "step": 180650 }, { "epoch": 0.45, "learning_rate": 5.480824399818814e-05, "loss": 1.0471, "step": 180655 }, { "epoch": 0.45, "learning_rate": 5.480698575670643e-05, "loss": 1.0437, "step": 180660 }, { "epoch": 0.45, "learning_rate": 5.480572751522473e-05, "loss": 1.0478, "step": 180665 }, { "epoch": 0.45, "learning_rate": 5.480446927374302e-05, "loss": 1.0625, "step": 180670 }, { "epoch": 0.45, "learning_rate": 5.480321103226132e-05, "loss": 1.046, "step": 180675 }, { "epoch": 0.45, "learning_rate": 5.480195279077961e-05, "loss": 1.048, "step": 180680 }, { "epoch": 0.45, "learning_rate": 5.480069454929791e-05, "loss": 1.0463, "step": 180685 }, { "epoch": 0.45, "learning_rate": 5.47994363078162e-05, "loss": 1.044, "step": 180690 }, { "epoch": 0.45, "learning_rate": 5.479817806633449e-05, "loss": 1.0455, "step": 180695 }, { "epoch": 0.45, "learning_rate": 5.479691982485279e-05, "loss": 1.047, "step": 180700 }, { "epoch": 0.45, "learning_rate": 5.479566158337108e-05, "loss": 1.0423, "step": 180705 }, { "epoch": 0.45, "learning_rate": 5.479440334188938e-05, "loss": 1.0472, "step": 180710 }, { "epoch": 0.45, "learning_rate": 5.479314510040767e-05, "loss": 1.0449, "step": 180715 }, { "epoch": 0.45, "learning_rate": 5.479188685892597e-05, "loss": 1.0432, "step": 180720 }, { "epoch": 0.45, "learning_rate": 5.479062861744426e-05, "loss": 1.0459, "step": 180725 }, { "epoch": 0.45, "learning_rate": 5.478937037596256e-05, "loss": 1.0468, "step": 180730 }, { "epoch": 0.45, "learning_rate": 5.478811213448085e-05, "loss": 1.0474, "step": 180735 }, { "epoch": 0.45, "learning_rate": 5.478685389299915e-05, "loss": 1.0466, "step": 180740 }, { "epoch": 0.45, "learning_rate": 5.478559565151744e-05, "loss": 1.0471, "step": 180745 }, { "epoch": 0.45, "learning_rate": 5.478433741003574e-05, "loss": 1.0451, "step": 180750 }, { "epoch": 0.45, "learning_rate": 5.478307916855403e-05, "loss": 1.0458, "step": 180755 }, { "epoch": 0.45, "learning_rate": 5.478182092707232e-05, "loss": 1.0473, "step": 180760 }, { "epoch": 0.45, "learning_rate": 5.478056268559062e-05, "loss": 1.0471, "step": 180765 }, { "epoch": 0.45, "learning_rate": 5.477930444410891e-05, "loss": 1.0461, "step": 180770 }, { "epoch": 0.45, "learning_rate": 5.477804620262721e-05, "loss": 1.0441, "step": 180775 }, { "epoch": 0.45, "learning_rate": 5.47767879611455e-05, "loss": 1.0451, "step": 180780 }, { "epoch": 0.45, "learning_rate": 5.47755297196638e-05, "loss": 1.045, "step": 180785 }, { "epoch": 0.45, "learning_rate": 5.477427147818209e-05, "loss": 1.0487, "step": 180790 }, { "epoch": 0.45, "learning_rate": 5.477301323670039e-05, "loss": 1.0465, "step": 180795 }, { "epoch": 0.45, "learning_rate": 5.477175499521868e-05, "loss": 1.0493, "step": 180800 }, { "epoch": 0.45, "learning_rate": 5.477049675373698e-05, "loss": 1.0392, "step": 180805 }, { "epoch": 0.45, "learning_rate": 5.476923851225527e-05, "loss": 1.0648, "step": 180810 }, { "epoch": 0.45, "learning_rate": 5.4767980270773563e-05, "loss": 1.0459, "step": 180815 }, { "epoch": 0.45, "learning_rate": 5.476672202929186e-05, "loss": 1.0462, "step": 180820 }, { "epoch": 0.45, "learning_rate": 5.4765463787810153e-05, "loss": 1.0452, "step": 180825 }, { "epoch": 0.45, "learning_rate": 5.476420554632845e-05, "loss": 1.0471, "step": 180830 }, { "epoch": 0.45, "learning_rate": 5.476294730484674e-05, "loss": 1.0466, "step": 180835 }, { "epoch": 0.45, "learning_rate": 5.476168906336504e-05, "loss": 1.0448, "step": 180840 }, { "epoch": 0.45, "learning_rate": 5.476043082188333e-05, "loss": 1.0493, "step": 180845 }, { "epoch": 0.45, "learning_rate": 5.475917258040163e-05, "loss": 1.046, "step": 180850 }, { "epoch": 0.45, "learning_rate": 5.475791433891992e-05, "loss": 1.0421, "step": 180855 }, { "epoch": 0.45, "learning_rate": 5.475665609743822e-05, "loss": 1.0457, "step": 180860 }, { "epoch": 0.45, "learning_rate": 5.475539785595651e-05, "loss": 1.0443, "step": 180865 }, { "epoch": 0.45, "learning_rate": 5.475413961447481e-05, "loss": 1.0473, "step": 180870 }, { "epoch": 0.45, "learning_rate": 5.47528813729931e-05, "loss": 1.0487, "step": 180875 }, { "epoch": 0.45, "learning_rate": 5.4751623131511395e-05, "loss": 1.0441, "step": 180880 }, { "epoch": 0.45, "learning_rate": 5.475036489002969e-05, "loss": 1.0457, "step": 180885 }, { "epoch": 0.45, "learning_rate": 5.4749106648547985e-05, "loss": 1.064, "step": 180890 }, { "epoch": 0.45, "learning_rate": 5.474784840706628e-05, "loss": 1.0474, "step": 180895 }, { "epoch": 0.45, "learning_rate": 5.4746590165584575e-05, "loss": 1.0438, "step": 180900 }, { "epoch": 0.45, "learning_rate": 5.474533192410287e-05, "loss": 1.0479, "step": 180905 }, { "epoch": 0.45, "learning_rate": 5.4744073682621165e-05, "loss": 1.0438, "step": 180910 }, { "epoch": 0.45, "learning_rate": 5.474281544113946e-05, "loss": 1.047, "step": 180915 }, { "epoch": 0.45, "learning_rate": 5.4741557199657755e-05, "loss": 1.0486, "step": 180920 }, { "epoch": 0.45, "learning_rate": 5.474029895817605e-05, "loss": 1.0488, "step": 180925 }, { "epoch": 0.45, "learning_rate": 5.4739040716694345e-05, "loss": 1.0443, "step": 180930 }, { "epoch": 0.45, "learning_rate": 5.473778247521264e-05, "loss": 1.0448, "step": 180935 }, { "epoch": 0.45, "learning_rate": 5.4736524233730935e-05, "loss": 1.0477, "step": 180940 }, { "epoch": 0.45, "learning_rate": 5.473526599224924e-05, "loss": 1.0462, "step": 180945 }, { "epoch": 0.45, "learning_rate": 5.473400775076754e-05, "loss": 1.0455, "step": 180950 }, { "epoch": 0.45, "learning_rate": 5.473274950928583e-05, "loss": 1.0432, "step": 180955 }, { "epoch": 0.45, "learning_rate": 5.473149126780412e-05, "loss": 1.045, "step": 180960 }, { "epoch": 0.45, "learning_rate": 5.473023302632242e-05, "loss": 1.0455, "step": 180965 }, { "epoch": 0.45, "learning_rate": 5.472897478484071e-05, "loss": 1.0644, "step": 180970 }, { "epoch": 0.45, "learning_rate": 5.472771654335901e-05, "loss": 1.0454, "step": 180975 }, { "epoch": 0.45, "learning_rate": 5.47264583018773e-05, "loss": 1.045, "step": 180980 }, { "epoch": 0.45, "learning_rate": 5.47252000603956e-05, "loss": 1.0436, "step": 180985 }, { "epoch": 0.45, "learning_rate": 5.472394181891389e-05, "loss": 1.0688, "step": 180990 }, { "epoch": 0.45, "learning_rate": 5.472268357743219e-05, "loss": 1.0451, "step": 180995 }, { "epoch": 0.45, "learning_rate": 5.472142533595048e-05, "loss": 1.0483, "step": 181000 }, { "epoch": 0.45, "learning_rate": 5.472016709446878e-05, "loss": 1.0421, "step": 181005 }, { "epoch": 0.45, "learning_rate": 5.471890885298707e-05, "loss": 1.0457, "step": 181010 }, { "epoch": 0.45, "learning_rate": 5.471765061150537e-05, "loss": 1.0457, "step": 181015 }, { "epoch": 0.45, "learning_rate": 5.471639237002366e-05, "loss": 1.064, "step": 181020 }, { "epoch": 0.45, "learning_rate": 5.471513412854195e-05, "loss": 1.044, "step": 181025 }, { "epoch": 0.45, "learning_rate": 5.471387588706025e-05, "loss": 1.0449, "step": 181030 }, { "epoch": 0.45, "learning_rate": 5.471261764557854e-05, "loss": 1.0453, "step": 181035 }, { "epoch": 0.45, "learning_rate": 5.471135940409684e-05, "loss": 1.0684, "step": 181040 }, { "epoch": 0.45, "learning_rate": 5.471010116261513e-05, "loss": 1.0426, "step": 181045 }, { "epoch": 0.45, "learning_rate": 5.470884292113343e-05, "loss": 1.0651, "step": 181050 }, { "epoch": 0.45, "learning_rate": 5.470758467965172e-05, "loss": 1.046, "step": 181055 }, { "epoch": 0.45, "learning_rate": 5.470632643817002e-05, "loss": 1.0428, "step": 181060 }, { "epoch": 0.45, "learning_rate": 5.470506819668831e-05, "loss": 1.045, "step": 181065 }, { "epoch": 0.45, "learning_rate": 5.470380995520661e-05, "loss": 1.0482, "step": 181070 }, { "epoch": 0.45, "learning_rate": 5.47025517137249e-05, "loss": 1.0458, "step": 181075 }, { "epoch": 0.45, "learning_rate": 5.47012934722432e-05, "loss": 1.0469, "step": 181080 }, { "epoch": 0.45, "learning_rate": 5.470003523076149e-05, "loss": 1.0438, "step": 181085 }, { "epoch": 0.45, "learning_rate": 5.4698776989279784e-05, "loss": 1.0454, "step": 181090 }, { "epoch": 0.45, "learning_rate": 5.469751874779808e-05, "loss": 1.0423, "step": 181095 }, { "epoch": 0.45, "learning_rate": 5.4696260506316374e-05, "loss": 1.0465, "step": 181100 }, { "epoch": 0.45, "learning_rate": 5.469500226483467e-05, "loss": 1.047, "step": 181105 }, { "epoch": 0.45, "learning_rate": 5.4693744023352964e-05, "loss": 1.0426, "step": 181110 }, { "epoch": 0.45, "learning_rate": 5.469248578187126e-05, "loss": 1.0446, "step": 181115 }, { "epoch": 0.45, "learning_rate": 5.4691227540389554e-05, "loss": 1.0441, "step": 181120 }, { "epoch": 0.45, "learning_rate": 5.468996929890785e-05, "loss": 1.047, "step": 181125 }, { "epoch": 0.45, "learning_rate": 5.4688711057426144e-05, "loss": 1.0475, "step": 181130 }, { "epoch": 0.45, "learning_rate": 5.468745281594444e-05, "loss": 1.0423, "step": 181135 }, { "epoch": 0.45, "learning_rate": 5.4686194574462734e-05, "loss": 1.0443, "step": 181140 }, { "epoch": 0.45, "learning_rate": 5.468493633298103e-05, "loss": 1.0444, "step": 181145 }, { "epoch": 0.45, "learning_rate": 5.4683678091499324e-05, "loss": 1.0459, "step": 181150 }, { "epoch": 0.45, "learning_rate": 5.4682419850017616e-05, "loss": 1.0456, "step": 181155 }, { "epoch": 0.45, "learning_rate": 5.4681161608535914e-05, "loss": 1.046, "step": 181160 }, { "epoch": 0.45, "learning_rate": 5.4679903367054206e-05, "loss": 1.0427, "step": 181165 }, { "epoch": 0.45, "learning_rate": 5.4678645125572504e-05, "loss": 1.0475, "step": 181170 }, { "epoch": 0.45, "learning_rate": 5.4677386884090796e-05, "loss": 1.0472, "step": 181175 }, { "epoch": 0.45, "learning_rate": 5.4676128642609094e-05, "loss": 1.0461, "step": 181180 }, { "epoch": 0.45, "learning_rate": 5.4674870401127386e-05, "loss": 1.0422, "step": 181185 }, { "epoch": 0.45, "learning_rate": 5.4673612159645684e-05, "loss": 1.0457, "step": 181190 }, { "epoch": 0.45, "learning_rate": 5.4672353918163976e-05, "loss": 1.0662, "step": 181195 }, { "epoch": 0.45, "learning_rate": 5.4671095676682274e-05, "loss": 1.0449, "step": 181200 }, { "epoch": 0.45, "learning_rate": 5.4669837435200566e-05, "loss": 1.0462, "step": 181205 }, { "epoch": 0.45, "learning_rate": 5.4668579193718864e-05, "loss": 1.0446, "step": 181210 }, { "epoch": 0.45, "learning_rate": 5.4667320952237155e-05, "loss": 1.0516, "step": 181215 }, { "epoch": 0.45, "learning_rate": 5.466606271075545e-05, "loss": 1.0507, "step": 181220 }, { "epoch": 0.45, "learning_rate": 5.4664804469273745e-05, "loss": 1.0417, "step": 181225 }, { "epoch": 0.45, "learning_rate": 5.466354622779204e-05, "loss": 1.0423, "step": 181230 }, { "epoch": 0.45, "learning_rate": 5.4662287986310335e-05, "loss": 1.0478, "step": 181235 }, { "epoch": 0.45, "learning_rate": 5.466102974482863e-05, "loss": 1.0458, "step": 181240 }, { "epoch": 0.45, "learning_rate": 5.4659771503346925e-05, "loss": 1.0466, "step": 181245 }, { "epoch": 0.45, "learning_rate": 5.465851326186522e-05, "loss": 1.0448, "step": 181250 }, { "epoch": 0.45, "learning_rate": 5.4657255020383515e-05, "loss": 1.0446, "step": 181255 }, { "epoch": 0.45, "learning_rate": 5.465599677890181e-05, "loss": 1.0479, "step": 181260 }, { "epoch": 0.46, "learning_rate": 5.4654738537420105e-05, "loss": 1.0449, "step": 181265 }, { "epoch": 0.46, "learning_rate": 5.46534802959384e-05, "loss": 1.0441, "step": 181270 }, { "epoch": 0.46, "learning_rate": 5.4652222054456695e-05, "loss": 1.0462, "step": 181275 }, { "epoch": 0.46, "learning_rate": 5.465096381297499e-05, "loss": 1.0438, "step": 181280 }, { "epoch": 0.46, "learning_rate": 5.464970557149328e-05, "loss": 1.0441, "step": 181285 }, { "epoch": 0.46, "learning_rate": 5.464844733001158e-05, "loss": 1.0455, "step": 181290 }, { "epoch": 0.46, "learning_rate": 5.464718908852987e-05, "loss": 1.0479, "step": 181295 }, { "epoch": 0.46, "learning_rate": 5.464593084704817e-05, "loss": 1.0457, "step": 181300 }, { "epoch": 0.46, "learning_rate": 5.464467260556646e-05, "loss": 1.0475, "step": 181305 }, { "epoch": 0.46, "learning_rate": 5.464341436408476e-05, "loss": 1.0454, "step": 181310 }, { "epoch": 0.46, "learning_rate": 5.464215612260305e-05, "loss": 1.0465, "step": 181315 }, { "epoch": 0.46, "learning_rate": 5.464089788112135e-05, "loss": 1.047, "step": 181320 }, { "epoch": 0.46, "learning_rate": 5.463963963963964e-05, "loss": 1.0406, "step": 181325 }, { "epoch": 0.46, "learning_rate": 5.463838139815794e-05, "loss": 1.043, "step": 181330 }, { "epoch": 0.46, "learning_rate": 5.463712315667623e-05, "loss": 1.0468, "step": 181335 }, { "epoch": 0.46, "learning_rate": 5.463586491519453e-05, "loss": 1.0457, "step": 181340 }, { "epoch": 0.46, "learning_rate": 5.463460667371282e-05, "loss": 1.0437, "step": 181345 }, { "epoch": 0.46, "learning_rate": 5.463334843223111e-05, "loss": 1.0445, "step": 181350 }, { "epoch": 0.46, "learning_rate": 5.463209019074941e-05, "loss": 1.0441, "step": 181355 }, { "epoch": 0.46, "learning_rate": 5.46308319492677e-05, "loss": 1.0463, "step": 181360 }, { "epoch": 0.46, "learning_rate": 5.4629573707786e-05, "loss": 1.0475, "step": 181365 }, { "epoch": 0.46, "learning_rate": 5.462831546630429e-05, "loss": 1.045, "step": 181370 }, { "epoch": 0.46, "learning_rate": 5.462705722482259e-05, "loss": 1.048, "step": 181375 }, { "epoch": 0.46, "learning_rate": 5.462579898334088e-05, "loss": 1.0452, "step": 181380 }, { "epoch": 0.46, "learning_rate": 5.462454074185918e-05, "loss": 1.0483, "step": 181385 }, { "epoch": 0.46, "learning_rate": 5.462328250037747e-05, "loss": 1.047, "step": 181390 }, { "epoch": 0.46, "learning_rate": 5.462202425889577e-05, "loss": 1.0467, "step": 181395 }, { "epoch": 0.46, "learning_rate": 5.462076601741406e-05, "loss": 1.0469, "step": 181400 }, { "epoch": 0.46, "learning_rate": 5.461950777593236e-05, "loss": 1.0445, "step": 181405 }, { "epoch": 0.46, "learning_rate": 5.461824953445065e-05, "loss": 1.048, "step": 181410 }, { "epoch": 0.46, "learning_rate": 5.461699129296894e-05, "loss": 1.0458, "step": 181415 }, { "epoch": 0.46, "learning_rate": 5.461573305148724e-05, "loss": 1.0432, "step": 181420 }, { "epoch": 0.46, "learning_rate": 5.461447481000553e-05, "loss": 1.0465, "step": 181425 }, { "epoch": 0.46, "learning_rate": 5.461321656852383e-05, "loss": 1.0464, "step": 181430 }, { "epoch": 0.46, "learning_rate": 5.461195832704212e-05, "loss": 1.045, "step": 181435 }, { "epoch": 0.46, "learning_rate": 5.461070008556042e-05, "loss": 1.0425, "step": 181440 }, { "epoch": 0.46, "learning_rate": 5.4609441844078725e-05, "loss": 1.047, "step": 181445 }, { "epoch": 0.46, "learning_rate": 5.4608183602597016e-05, "loss": 1.0459, "step": 181450 }, { "epoch": 0.46, "learning_rate": 5.4606925361115315e-05, "loss": 1.0448, "step": 181455 }, { "epoch": 0.46, "learning_rate": 5.4605667119633606e-05, "loss": 1.0488, "step": 181460 }, { "epoch": 0.46, "learning_rate": 5.4604408878151905e-05, "loss": 1.0476, "step": 181465 }, { "epoch": 0.46, "learning_rate": 5.4603150636670196e-05, "loss": 1.0431, "step": 181470 }, { "epoch": 0.46, "learning_rate": 5.4601892395188495e-05, "loss": 1.0448, "step": 181475 }, { "epoch": 0.46, "learning_rate": 5.4600634153706786e-05, "loss": 1.0449, "step": 181480 }, { "epoch": 0.46, "learning_rate": 5.4599375912225085e-05, "loss": 1.0452, "step": 181485 }, { "epoch": 0.46, "learning_rate": 5.4598117670743376e-05, "loss": 1.0442, "step": 181490 }, { "epoch": 0.46, "learning_rate": 5.459685942926167e-05, "loss": 1.0438, "step": 181495 }, { "epoch": 0.46, "learning_rate": 5.4595601187779966e-05, "loss": 1.0482, "step": 181500 }, { "epoch": 0.46, "learning_rate": 5.459434294629826e-05, "loss": 1.0496, "step": 181505 }, { "epoch": 0.46, "learning_rate": 5.4593084704816556e-05, "loss": 1.0462, "step": 181510 }, { "epoch": 0.46, "learning_rate": 5.459182646333485e-05, "loss": 1.042, "step": 181515 }, { "epoch": 0.46, "learning_rate": 5.4590568221853146e-05, "loss": 1.0427, "step": 181520 }, { "epoch": 0.46, "learning_rate": 5.458930998037144e-05, "loss": 1.0468, "step": 181525 }, { "epoch": 0.46, "learning_rate": 5.4588051738889736e-05, "loss": 1.0421, "step": 181530 }, { "epoch": 0.46, "learning_rate": 5.458679349740803e-05, "loss": 1.0457, "step": 181535 }, { "epoch": 0.46, "learning_rate": 5.4585535255926326e-05, "loss": 1.0404, "step": 181540 }, { "epoch": 0.46, "learning_rate": 5.458427701444462e-05, "loss": 1.049, "step": 181545 }, { "epoch": 0.46, "learning_rate": 5.4583018772962916e-05, "loss": 1.0473, "step": 181550 }, { "epoch": 0.46, "learning_rate": 5.458176053148121e-05, "loss": 1.0435, "step": 181555 }, { "epoch": 0.46, "learning_rate": 5.45805022899995e-05, "loss": 1.0467, "step": 181560 }, { "epoch": 0.46, "learning_rate": 5.45792440485178e-05, "loss": 1.0473, "step": 181565 }, { "epoch": 0.46, "learning_rate": 5.457798580703609e-05, "loss": 1.0471, "step": 181570 }, { "epoch": 0.46, "learning_rate": 5.457672756555439e-05, "loss": 1.0473, "step": 181575 }, { "epoch": 0.46, "learning_rate": 5.457546932407268e-05, "loss": 1.0656, "step": 181580 }, { "epoch": 0.46, "learning_rate": 5.457421108259098e-05, "loss": 1.0449, "step": 181585 }, { "epoch": 0.46, "learning_rate": 5.457295284110927e-05, "loss": 1.0453, "step": 181590 }, { "epoch": 0.46, "learning_rate": 5.457169459962757e-05, "loss": 1.0441, "step": 181595 }, { "epoch": 0.46, "learning_rate": 5.457043635814586e-05, "loss": 1.0448, "step": 181600 }, { "epoch": 0.46, "learning_rate": 5.456917811666416e-05, "loss": 1.0469, "step": 181605 }, { "epoch": 0.46, "learning_rate": 5.456791987518245e-05, "loss": 1.0412, "step": 181610 }, { "epoch": 0.46, "learning_rate": 5.456666163370074e-05, "loss": 1.0454, "step": 181615 }, { "epoch": 0.46, "learning_rate": 5.456540339221904e-05, "loss": 1.0451, "step": 181620 }, { "epoch": 0.46, "learning_rate": 5.456414515073733e-05, "loss": 1.0434, "step": 181625 }, { "epoch": 0.46, "learning_rate": 5.456288690925563e-05, "loss": 1.048, "step": 181630 }, { "epoch": 0.46, "learning_rate": 5.456162866777392e-05, "loss": 1.0458, "step": 181635 }, { "epoch": 0.46, "learning_rate": 5.456037042629222e-05, "loss": 1.0448, "step": 181640 }, { "epoch": 0.46, "learning_rate": 5.455911218481051e-05, "loss": 1.048, "step": 181645 }, { "epoch": 0.46, "learning_rate": 5.455785394332881e-05, "loss": 1.0435, "step": 181650 }, { "epoch": 0.46, "learning_rate": 5.45565957018471e-05, "loss": 1.0429, "step": 181655 }, { "epoch": 0.46, "learning_rate": 5.45553374603654e-05, "loss": 1.0442, "step": 181660 }, { "epoch": 0.46, "learning_rate": 5.455407921888369e-05, "loss": 1.068, "step": 181665 }, { "epoch": 0.46, "learning_rate": 5.455282097740199e-05, "loss": 1.0483, "step": 181670 }, { "epoch": 0.46, "learning_rate": 5.455156273592028e-05, "loss": 1.0452, "step": 181675 }, { "epoch": 0.46, "learning_rate": 5.455030449443857e-05, "loss": 1.0461, "step": 181680 }, { "epoch": 0.46, "learning_rate": 5.454904625295687e-05, "loss": 1.0442, "step": 181685 }, { "epoch": 0.46, "learning_rate": 5.454778801147516e-05, "loss": 1.0429, "step": 181690 }, { "epoch": 0.46, "learning_rate": 5.454652976999346e-05, "loss": 1.0452, "step": 181695 }, { "epoch": 0.46, "learning_rate": 5.454527152851175e-05, "loss": 1.0465, "step": 181700 }, { "epoch": 0.46, "learning_rate": 5.454401328703005e-05, "loss": 1.0609, "step": 181705 }, { "epoch": 0.46, "learning_rate": 5.454275504554834e-05, "loss": 1.0455, "step": 181710 }, { "epoch": 0.46, "learning_rate": 5.454149680406664e-05, "loss": 1.0461, "step": 181715 }, { "epoch": 0.46, "learning_rate": 5.454023856258493e-05, "loss": 1.045, "step": 181720 }, { "epoch": 0.46, "learning_rate": 5.453898032110323e-05, "loss": 1.0446, "step": 181725 }, { "epoch": 0.46, "learning_rate": 5.453772207962152e-05, "loss": 1.046, "step": 181730 }, { "epoch": 0.46, "learning_rate": 5.453646383813982e-05, "loss": 1.0461, "step": 181735 }, { "epoch": 0.46, "learning_rate": 5.453520559665811e-05, "loss": 1.0435, "step": 181740 }, { "epoch": 0.46, "learning_rate": 5.4533947355176404e-05, "loss": 1.0464, "step": 181745 }, { "epoch": 0.46, "learning_rate": 5.45326891136947e-05, "loss": 1.0473, "step": 181750 }, { "epoch": 0.46, "learning_rate": 5.4531430872212994e-05, "loss": 1.0423, "step": 181755 }, { "epoch": 0.46, "learning_rate": 5.453017263073129e-05, "loss": 1.0443, "step": 181760 }, { "epoch": 0.46, "learning_rate": 5.4528914389249584e-05, "loss": 1.0441, "step": 181765 }, { "epoch": 0.46, "learning_rate": 5.452765614776788e-05, "loss": 1.0449, "step": 181770 }, { "epoch": 0.46, "learning_rate": 5.4526397906286174e-05, "loss": 1.0456, "step": 181775 }, { "epoch": 0.46, "learning_rate": 5.452513966480447e-05, "loss": 1.0458, "step": 181780 }, { "epoch": 0.46, "learning_rate": 5.4523881423322764e-05, "loss": 1.044, "step": 181785 }, { "epoch": 0.46, "learning_rate": 5.452262318184106e-05, "loss": 1.0472, "step": 181790 }, { "epoch": 0.46, "learning_rate": 5.4521364940359353e-05, "loss": 1.0435, "step": 181795 }, { "epoch": 0.46, "learning_rate": 5.452010669887765e-05, "loss": 1.0457, "step": 181800 }, { "epoch": 0.46, "learning_rate": 5.4518848457395943e-05, "loss": 1.0468, "step": 181805 }, { "epoch": 0.46, "learning_rate": 5.4517590215914235e-05, "loss": 1.0451, "step": 181810 }, { "epoch": 0.46, "learning_rate": 5.4516331974432533e-05, "loss": 1.0454, "step": 181815 }, { "epoch": 0.46, "learning_rate": 5.4515073732950825e-05, "loss": 1.0457, "step": 181820 }, { "epoch": 0.46, "learning_rate": 5.4513815491469123e-05, "loss": 1.0427, "step": 181825 }, { "epoch": 0.46, "learning_rate": 5.4512557249987415e-05, "loss": 1.0467, "step": 181830 }, { "epoch": 0.46, "learning_rate": 5.451129900850571e-05, "loss": 1.0466, "step": 181835 }, { "epoch": 0.46, "learning_rate": 5.4510040767024005e-05, "loss": 1.0411, "step": 181840 }, { "epoch": 0.46, "learning_rate": 5.45087825255423e-05, "loss": 1.0461, "step": 181845 }, { "epoch": 0.46, "learning_rate": 5.4507524284060595e-05, "loss": 1.0438, "step": 181850 }, { "epoch": 0.46, "learning_rate": 5.450626604257889e-05, "loss": 1.0458, "step": 181855 }, { "epoch": 0.46, "learning_rate": 5.4505007801097185e-05, "loss": 1.0496, "step": 181860 }, { "epoch": 0.46, "learning_rate": 5.450374955961548e-05, "loss": 1.0693, "step": 181865 }, { "epoch": 0.46, "learning_rate": 5.4502491318133775e-05, "loss": 1.044, "step": 181870 }, { "epoch": 0.46, "learning_rate": 5.4501233076652066e-05, "loss": 1.0452, "step": 181875 }, { "epoch": 0.46, "learning_rate": 5.4499974835170365e-05, "loss": 1.0448, "step": 181880 }, { "epoch": 0.46, "learning_rate": 5.4498716593688656e-05, "loss": 1.0478, "step": 181885 }, { "epoch": 0.46, "learning_rate": 5.4497458352206955e-05, "loss": 1.0663, "step": 181890 }, { "epoch": 0.46, "learning_rate": 5.4496200110725246e-05, "loss": 1.0458, "step": 181895 }, { "epoch": 0.46, "learning_rate": 5.4494941869243545e-05, "loss": 1.0436, "step": 181900 }, { "epoch": 0.46, "learning_rate": 5.4493683627761836e-05, "loss": 1.0449, "step": 181905 }, { "epoch": 0.46, "learning_rate": 5.4492425386280135e-05, "loss": 1.0429, "step": 181910 }, { "epoch": 0.46, "learning_rate": 5.4491167144798426e-05, "loss": 1.0466, "step": 181915 }, { "epoch": 0.46, "learning_rate": 5.4489908903316725e-05, "loss": 1.0441, "step": 181920 }, { "epoch": 0.46, "learning_rate": 5.4488650661835016e-05, "loss": 1.0419, "step": 181925 }, { "epoch": 0.46, "learning_rate": 5.4487392420353315e-05, "loss": 1.0848, "step": 181930 }, { "epoch": 0.46, "learning_rate": 5.4486134178871606e-05, "loss": 1.0432, "step": 181935 }, { "epoch": 0.46, "learning_rate": 5.448487593738991e-05, "loss": 1.0467, "step": 181940 }, { "epoch": 0.46, "learning_rate": 5.448361769590821e-05, "loss": 1.0468, "step": 181945 }, { "epoch": 0.46, "learning_rate": 5.44823594544265e-05, "loss": 1.0471, "step": 181950 }, { "epoch": 0.46, "learning_rate": 5.448110121294479e-05, "loss": 1.0479, "step": 181955 }, { "epoch": 0.46, "learning_rate": 5.447984297146309e-05, "loss": 1.0447, "step": 181960 }, { "epoch": 0.46, "learning_rate": 5.447858472998138e-05, "loss": 1.0453, "step": 181965 }, { "epoch": 0.46, "learning_rate": 5.447732648849968e-05, "loss": 1.0435, "step": 181970 }, { "epoch": 0.46, "learning_rate": 5.447606824701797e-05, "loss": 1.048, "step": 181975 }, { "epoch": 0.46, "learning_rate": 5.447481000553627e-05, "loss": 1.0591, "step": 181980 }, { "epoch": 0.46, "learning_rate": 5.447355176405456e-05, "loss": 1.0456, "step": 181985 }, { "epoch": 0.46, "learning_rate": 5.447229352257286e-05, "loss": 1.0448, "step": 181990 }, { "epoch": 0.46, "learning_rate": 5.447103528109115e-05, "loss": 1.0491, "step": 181995 }, { "epoch": 0.46, "learning_rate": 5.446977703960945e-05, "loss": 1.0711, "step": 182000 }, { "epoch": 0.46, "learning_rate": 5.446851879812774e-05, "loss": 1.0477, "step": 182005 }, { "epoch": 0.46, "learning_rate": 5.446726055664604e-05, "loss": 1.0473, "step": 182010 }, { "epoch": 0.46, "learning_rate": 5.446600231516433e-05, "loss": 1.0462, "step": 182015 }, { "epoch": 0.46, "learning_rate": 5.4464744073682624e-05, "loss": 1.0472, "step": 182020 }, { "epoch": 0.46, "learning_rate": 5.446348583220092e-05, "loss": 1.045, "step": 182025 }, { "epoch": 0.46, "learning_rate": 5.4462227590719214e-05, "loss": 1.0475, "step": 182030 }, { "epoch": 0.46, "learning_rate": 5.446096934923751e-05, "loss": 1.0424, "step": 182035 }, { "epoch": 0.46, "learning_rate": 5.4459711107755804e-05, "loss": 1.0489, "step": 182040 }, { "epoch": 0.46, "learning_rate": 5.44584528662741e-05, "loss": 1.0472, "step": 182045 }, { "epoch": 0.46, "learning_rate": 5.4457194624792394e-05, "loss": 1.0456, "step": 182050 }, { "epoch": 0.46, "learning_rate": 5.445593638331069e-05, "loss": 1.0475, "step": 182055 }, { "epoch": 0.46, "learning_rate": 5.4454678141828984e-05, "loss": 1.0453, "step": 182060 }, { "epoch": 0.46, "learning_rate": 5.445341990034728e-05, "loss": 1.0484, "step": 182065 }, { "epoch": 0.46, "learning_rate": 5.4452161658865574e-05, "loss": 1.063, "step": 182070 }, { "epoch": 0.46, "learning_rate": 5.445090341738387e-05, "loss": 1.0451, "step": 182075 }, { "epoch": 0.46, "learning_rate": 5.4449645175902164e-05, "loss": 1.0456, "step": 182080 }, { "epoch": 0.46, "learning_rate": 5.4448386934420456e-05, "loss": 1.0676, "step": 182085 }, { "epoch": 0.46, "learning_rate": 5.4447128692938754e-05, "loss": 1.0681, "step": 182090 }, { "epoch": 0.46, "learning_rate": 5.4445870451457046e-05, "loss": 1.0447, "step": 182095 }, { "epoch": 0.46, "learning_rate": 5.4444612209975344e-05, "loss": 1.0456, "step": 182100 }, { "epoch": 0.46, "learning_rate": 5.4443353968493636e-05, "loss": 1.0495, "step": 182105 }, { "epoch": 0.46, "learning_rate": 5.4442095727011934e-05, "loss": 1.0443, "step": 182110 }, { "epoch": 0.46, "learning_rate": 5.4440837485530226e-05, "loss": 1.0465, "step": 182115 }, { "epoch": 0.46, "learning_rate": 5.4439579244048524e-05, "loss": 1.0471, "step": 182120 }, { "epoch": 0.46, "learning_rate": 5.4438321002566816e-05, "loss": 1.0477, "step": 182125 }, { "epoch": 0.46, "learning_rate": 5.4437062761085114e-05, "loss": 1.0694, "step": 182130 }, { "epoch": 0.46, "learning_rate": 5.4435804519603406e-05, "loss": 1.048, "step": 182135 }, { "epoch": 0.46, "learning_rate": 5.4434546278121704e-05, "loss": 1.0444, "step": 182140 }, { "epoch": 0.46, "learning_rate": 5.4433288036639996e-05, "loss": 1.0462, "step": 182145 }, { "epoch": 0.46, "learning_rate": 5.443202979515829e-05, "loss": 1.0458, "step": 182150 }, { "epoch": 0.46, "learning_rate": 5.4430771553676586e-05, "loss": 1.0465, "step": 182155 }, { "epoch": 0.46, "learning_rate": 5.442951331219488e-05, "loss": 1.0465, "step": 182160 }, { "epoch": 0.46, "learning_rate": 5.4428255070713176e-05, "loss": 1.0431, "step": 182165 }, { "epoch": 0.46, "learning_rate": 5.442699682923147e-05, "loss": 1.048, "step": 182170 }, { "epoch": 0.46, "learning_rate": 5.4425738587749766e-05, "loss": 1.0463, "step": 182175 }, { "epoch": 0.46, "learning_rate": 5.442448034626806e-05, "loss": 1.0483, "step": 182180 }, { "epoch": 0.46, "learning_rate": 5.4423222104786356e-05, "loss": 1.0446, "step": 182185 }, { "epoch": 0.46, "learning_rate": 5.442196386330465e-05, "loss": 1.0451, "step": 182190 }, { "epoch": 0.46, "learning_rate": 5.4420705621822946e-05, "loss": 1.0423, "step": 182195 }, { "epoch": 0.46, "learning_rate": 5.441944738034124e-05, "loss": 1.0443, "step": 182200 }, { "epoch": 0.46, "learning_rate": 5.4418189138859536e-05, "loss": 1.0475, "step": 182205 }, { "epoch": 0.46, "learning_rate": 5.441693089737783e-05, "loss": 1.0463, "step": 182210 }, { "epoch": 0.46, "learning_rate": 5.441567265589612e-05, "loss": 1.0431, "step": 182215 }, { "epoch": 0.46, "learning_rate": 5.441441441441442e-05, "loss": 1.045, "step": 182220 }, { "epoch": 0.46, "learning_rate": 5.441315617293271e-05, "loss": 1.043, "step": 182225 }, { "epoch": 0.46, "learning_rate": 5.441189793145101e-05, "loss": 1.0488, "step": 182230 }, { "epoch": 0.46, "learning_rate": 5.44106396899693e-05, "loss": 1.0454, "step": 182235 }, { "epoch": 0.46, "learning_rate": 5.44093814484876e-05, "loss": 1.044, "step": 182240 }, { "epoch": 0.46, "learning_rate": 5.440812320700589e-05, "loss": 1.0434, "step": 182245 }, { "epoch": 0.46, "learning_rate": 5.440686496552419e-05, "loss": 1.0436, "step": 182250 }, { "epoch": 0.46, "learning_rate": 5.440560672404248e-05, "loss": 1.0505, "step": 182255 }, { "epoch": 0.46, "learning_rate": 5.440434848256078e-05, "loss": 1.0412, "step": 182260 }, { "epoch": 0.46, "learning_rate": 5.440309024107907e-05, "loss": 1.0437, "step": 182265 }, { "epoch": 0.46, "learning_rate": 5.440183199959736e-05, "loss": 1.0425, "step": 182270 }, { "epoch": 0.46, "learning_rate": 5.440057375811566e-05, "loss": 1.0474, "step": 182275 }, { "epoch": 0.46, "learning_rate": 5.439931551663395e-05, "loss": 1.0438, "step": 182280 }, { "epoch": 0.46, "learning_rate": 5.439805727515225e-05, "loss": 1.0683, "step": 182285 }, { "epoch": 0.46, "learning_rate": 5.439679903367054e-05, "loss": 1.0441, "step": 182290 }, { "epoch": 0.46, "learning_rate": 5.439554079218884e-05, "loss": 1.0447, "step": 182295 }, { "epoch": 0.46, "learning_rate": 5.439428255070713e-05, "loss": 1.0457, "step": 182300 }, { "epoch": 0.46, "learning_rate": 5.439302430922543e-05, "loss": 1.0461, "step": 182305 }, { "epoch": 0.46, "learning_rate": 5.439176606774372e-05, "loss": 1.0428, "step": 182310 }, { "epoch": 0.46, "learning_rate": 5.439050782626202e-05, "loss": 1.046, "step": 182315 }, { "epoch": 0.46, "learning_rate": 5.438924958478031e-05, "loss": 1.0413, "step": 182320 }, { "epoch": 0.46, "learning_rate": 5.438799134329861e-05, "loss": 1.0426, "step": 182325 }, { "epoch": 0.46, "learning_rate": 5.43867331018169e-05, "loss": 1.0464, "step": 182330 }, { "epoch": 0.46, "learning_rate": 5.438547486033519e-05, "loss": 1.0688, "step": 182335 }, { "epoch": 0.46, "learning_rate": 5.438421661885349e-05, "loss": 1.0475, "step": 182340 }, { "epoch": 0.46, "learning_rate": 5.438295837737178e-05, "loss": 1.0444, "step": 182345 }, { "epoch": 0.46, "learning_rate": 5.438170013589008e-05, "loss": 1.0487, "step": 182350 }, { "epoch": 0.46, "learning_rate": 5.438044189440837e-05, "loss": 1.043, "step": 182355 }, { "epoch": 0.46, "learning_rate": 5.437918365292667e-05, "loss": 1.0455, "step": 182360 }, { "epoch": 0.46, "learning_rate": 5.437792541144496e-05, "loss": 1.0429, "step": 182365 }, { "epoch": 0.46, "learning_rate": 5.437666716996326e-05, "loss": 1.0427, "step": 182370 }, { "epoch": 0.46, "learning_rate": 5.437540892848155e-05, "loss": 1.0573, "step": 182375 }, { "epoch": 0.46, "learning_rate": 5.437415068699985e-05, "loss": 1.0442, "step": 182380 }, { "epoch": 0.46, "learning_rate": 5.437289244551814e-05, "loss": 1.0481, "step": 182385 }, { "epoch": 0.46, "learning_rate": 5.437163420403644e-05, "loss": 1.0448, "step": 182390 }, { "epoch": 0.46, "learning_rate": 5.437037596255473e-05, "loss": 1.0425, "step": 182395 }, { "epoch": 0.46, "learning_rate": 5.436911772107302e-05, "loss": 1.0447, "step": 182400 }, { "epoch": 0.46, "learning_rate": 5.436785947959132e-05, "loss": 1.0656, "step": 182405 }, { "epoch": 0.46, "learning_rate": 5.436660123810961e-05, "loss": 1.0454, "step": 182410 }, { "epoch": 0.46, "learning_rate": 5.436534299662791e-05, "loss": 1.0462, "step": 182415 }, { "epoch": 0.46, "learning_rate": 5.43640847551462e-05, "loss": 1.0537, "step": 182420 }, { "epoch": 0.46, "learning_rate": 5.43628265136645e-05, "loss": 1.0462, "step": 182425 }, { "epoch": 0.46, "learning_rate": 5.436156827218279e-05, "loss": 1.0459, "step": 182430 }, { "epoch": 0.46, "learning_rate": 5.436031003070109e-05, "loss": 1.047, "step": 182435 }, { "epoch": 0.46, "learning_rate": 5.4359051789219396e-05, "loss": 1.0463, "step": 182440 }, { "epoch": 0.46, "learning_rate": 5.435779354773769e-05, "loss": 1.0432, "step": 182445 }, { "epoch": 0.46, "learning_rate": 5.4356535306255986e-05, "loss": 1.0454, "step": 182450 }, { "epoch": 0.46, "learning_rate": 5.435527706477428e-05, "loss": 1.0437, "step": 182455 }, { "epoch": 0.46, "learning_rate": 5.4354018823292576e-05, "loss": 1.0459, "step": 182460 }, { "epoch": 0.46, "learning_rate": 5.435276058181087e-05, "loss": 1.0436, "step": 182465 }, { "epoch": 0.46, "learning_rate": 5.4351502340329166e-05, "loss": 1.045, "step": 182470 }, { "epoch": 0.46, "learning_rate": 5.435024409884746e-05, "loss": 1.0475, "step": 182475 }, { "epoch": 0.46, "learning_rate": 5.434898585736575e-05, "loss": 1.0455, "step": 182480 }, { "epoch": 0.46, "learning_rate": 5.434772761588405e-05, "loss": 1.0449, "step": 182485 }, { "epoch": 0.46, "learning_rate": 5.434646937440234e-05, "loss": 1.0467, "step": 182490 }, { "epoch": 0.46, "learning_rate": 5.434521113292064e-05, "loss": 1.0396, "step": 182495 }, { "epoch": 0.46, "learning_rate": 5.434395289143893e-05, "loss": 1.0423, "step": 182500 }, { "epoch": 0.46, "learning_rate": 5.434269464995723e-05, "loss": 1.0475, "step": 182505 }, { "epoch": 0.46, "learning_rate": 5.434143640847552e-05, "loss": 1.0444, "step": 182510 }, { "epoch": 0.46, "learning_rate": 5.434017816699382e-05, "loss": 1.0452, "step": 182515 }, { "epoch": 0.46, "learning_rate": 5.433891992551211e-05, "loss": 1.0435, "step": 182520 }, { "epoch": 0.46, "learning_rate": 5.433766168403041e-05, "loss": 1.0432, "step": 182525 }, { "epoch": 0.46, "learning_rate": 5.43364034425487e-05, "loss": 1.0464, "step": 182530 }, { "epoch": 0.46, "learning_rate": 5.4335145201067e-05, "loss": 1.0671, "step": 182535 }, { "epoch": 0.46, "learning_rate": 5.433388695958529e-05, "loss": 1.0462, "step": 182540 }, { "epoch": 0.46, "learning_rate": 5.433262871810358e-05, "loss": 1.0492, "step": 182545 }, { "epoch": 0.46, "learning_rate": 5.433137047662188e-05, "loss": 1.0462, "step": 182550 }, { "epoch": 0.46, "learning_rate": 5.433011223514017e-05, "loss": 1.0452, "step": 182555 }, { "epoch": 0.46, "learning_rate": 5.432885399365847e-05, "loss": 1.0473, "step": 182560 }, { "epoch": 0.46, "learning_rate": 5.432759575217676e-05, "loss": 1.0465, "step": 182565 }, { "epoch": 0.46, "learning_rate": 5.432633751069506e-05, "loss": 1.0459, "step": 182570 }, { "epoch": 0.46, "learning_rate": 5.432507926921335e-05, "loss": 1.0467, "step": 182575 }, { "epoch": 0.46, "learning_rate": 5.432382102773165e-05, "loss": 1.0486, "step": 182580 }, { "epoch": 0.46, "learning_rate": 5.432256278624994e-05, "loss": 1.067, "step": 182585 }, { "epoch": 0.46, "learning_rate": 5.432130454476824e-05, "loss": 1.0423, "step": 182590 }, { "epoch": 0.46, "learning_rate": 5.432004630328653e-05, "loss": 1.0472, "step": 182595 }, { "epoch": 0.46, "learning_rate": 5.431878806180483e-05, "loss": 1.0433, "step": 182600 }, { "epoch": 0.46, "learning_rate": 5.431752982032312e-05, "loss": 1.0678, "step": 182605 }, { "epoch": 0.46, "learning_rate": 5.431627157884141e-05, "loss": 1.0657, "step": 182610 }, { "epoch": 0.46, "learning_rate": 5.431501333735971e-05, "loss": 1.0423, "step": 182615 }, { "epoch": 0.46, "learning_rate": 5.4313755095878e-05, "loss": 1.0462, "step": 182620 }, { "epoch": 0.46, "learning_rate": 5.43124968543963e-05, "loss": 1.0437, "step": 182625 }, { "epoch": 0.46, "learning_rate": 5.431123861291459e-05, "loss": 1.0466, "step": 182630 }, { "epoch": 0.46, "learning_rate": 5.430998037143289e-05, "loss": 1.0498, "step": 182635 }, { "epoch": 0.46, "learning_rate": 5.430872212995118e-05, "loss": 1.0447, "step": 182640 }, { "epoch": 0.46, "learning_rate": 5.430746388846948e-05, "loss": 1.0447, "step": 182645 }, { "epoch": 0.46, "learning_rate": 5.430620564698777e-05, "loss": 1.0462, "step": 182650 }, { "epoch": 0.46, "learning_rate": 5.430494740550607e-05, "loss": 1.0475, "step": 182655 }, { "epoch": 0.46, "learning_rate": 5.430368916402436e-05, "loss": 1.0447, "step": 182660 }, { "epoch": 0.46, "learning_rate": 5.430243092254266e-05, "loss": 1.0459, "step": 182665 }, { "epoch": 0.46, "learning_rate": 5.430117268106095e-05, "loss": 1.044, "step": 182670 }, { "epoch": 0.46, "learning_rate": 5.4299914439579244e-05, "loss": 1.0463, "step": 182675 }, { "epoch": 0.46, "learning_rate": 5.429865619809754e-05, "loss": 1.0443, "step": 182680 }, { "epoch": 0.46, "learning_rate": 5.4297397956615834e-05, "loss": 1.0464, "step": 182685 }, { "epoch": 0.46, "learning_rate": 5.429613971513413e-05, "loss": 1.0461, "step": 182690 }, { "epoch": 0.46, "learning_rate": 5.4294881473652424e-05, "loss": 1.0451, "step": 182695 }, { "epoch": 0.46, "learning_rate": 5.429362323217072e-05, "loss": 1.0443, "step": 182700 }, { "epoch": 0.46, "learning_rate": 5.4292364990689014e-05, "loss": 1.0469, "step": 182705 }, { "epoch": 0.46, "learning_rate": 5.429110674920731e-05, "loss": 1.0463, "step": 182710 }, { "epoch": 0.46, "learning_rate": 5.4289848507725604e-05, "loss": 1.043, "step": 182715 }, { "epoch": 0.46, "learning_rate": 5.42885902662439e-05, "loss": 1.0514, "step": 182720 }, { "epoch": 0.46, "learning_rate": 5.4287332024762194e-05, "loss": 1.0454, "step": 182725 }, { "epoch": 0.46, "learning_rate": 5.428607378328049e-05, "loss": 1.0458, "step": 182730 }, { "epoch": 0.46, "learning_rate": 5.4284815541798784e-05, "loss": 1.0465, "step": 182735 }, { "epoch": 0.46, "learning_rate": 5.4283557300317075e-05, "loss": 1.0439, "step": 182740 }, { "epoch": 0.46, "learning_rate": 5.4282299058835374e-05, "loss": 1.0461, "step": 182745 }, { "epoch": 0.46, "learning_rate": 5.4281040817353665e-05, "loss": 1.0457, "step": 182750 }, { "epoch": 0.46, "learning_rate": 5.4279782575871964e-05, "loss": 1.0617, "step": 182755 }, { "epoch": 0.46, "learning_rate": 5.4278524334390255e-05, "loss": 1.0464, "step": 182760 }, { "epoch": 0.46, "learning_rate": 5.4277266092908554e-05, "loss": 1.0433, "step": 182765 }, { "epoch": 0.46, "learning_rate": 5.4276007851426845e-05, "loss": 1.0471, "step": 182770 }, { "epoch": 0.46, "learning_rate": 5.4274749609945144e-05, "loss": 1.0445, "step": 182775 }, { "epoch": 0.46, "learning_rate": 5.4273491368463435e-05, "loss": 1.0437, "step": 182780 }, { "epoch": 0.46, "learning_rate": 5.4272233126981734e-05, "loss": 1.0456, "step": 182785 }, { "epoch": 0.46, "learning_rate": 5.4270974885500025e-05, "loss": 1.0453, "step": 182790 }, { "epoch": 0.46, "learning_rate": 5.4269716644018323e-05, "loss": 1.0448, "step": 182795 }, { "epoch": 0.46, "learning_rate": 5.4268458402536615e-05, "loss": 1.0447, "step": 182800 }, { "epoch": 0.46, "learning_rate": 5.426720016105491e-05, "loss": 1.0469, "step": 182805 }, { "epoch": 0.46, "learning_rate": 5.4265941919573205e-05, "loss": 1.0451, "step": 182810 }, { "epoch": 0.46, "learning_rate": 5.42646836780915e-05, "loss": 1.0434, "step": 182815 }, { "epoch": 0.46, "learning_rate": 5.4263425436609795e-05, "loss": 1.0423, "step": 182820 }, { "epoch": 0.46, "learning_rate": 5.426216719512809e-05, "loss": 1.0445, "step": 182825 }, { "epoch": 0.46, "learning_rate": 5.4260908953646385e-05, "loss": 1.0455, "step": 182830 }, { "epoch": 0.46, "learning_rate": 5.4259650712164677e-05, "loss": 1.0453, "step": 182835 }, { "epoch": 0.46, "learning_rate": 5.4258392470682975e-05, "loss": 1.0468, "step": 182840 }, { "epoch": 0.46, "learning_rate": 5.4257134229201267e-05, "loss": 1.0448, "step": 182845 }, { "epoch": 0.46, "learning_rate": 5.4255875987719565e-05, "loss": 1.0485, "step": 182850 }, { "epoch": 0.46, "learning_rate": 5.4254617746237857e-05, "loss": 1.0457, "step": 182855 }, { "epoch": 0.46, "learning_rate": 5.4253359504756155e-05, "loss": 1.0439, "step": 182860 }, { "epoch": 0.46, "learning_rate": 5.4252101263274447e-05, "loss": 1.045, "step": 182865 }, { "epoch": 0.46, "learning_rate": 5.425084302179274e-05, "loss": 1.0457, "step": 182870 }, { "epoch": 0.46, "learning_rate": 5.4249584780311036e-05, "loss": 1.0426, "step": 182875 }, { "epoch": 0.46, "learning_rate": 5.424832653882933e-05, "loss": 1.0461, "step": 182880 }, { "epoch": 0.46, "learning_rate": 5.4247068297347626e-05, "loss": 1.0476, "step": 182885 }, { "epoch": 0.46, "learning_rate": 5.424581005586592e-05, "loss": 1.0458, "step": 182890 }, { "epoch": 0.46, "learning_rate": 5.4244551814384216e-05, "loss": 1.0436, "step": 182895 }, { "epoch": 0.46, "learning_rate": 5.424329357290251e-05, "loss": 1.0446, "step": 182900 }, { "epoch": 0.46, "learning_rate": 5.4242035331420806e-05, "loss": 1.0478, "step": 182905 }, { "epoch": 0.46, "learning_rate": 5.42407770899391e-05, "loss": 1.0448, "step": 182910 }, { "epoch": 0.46, "learning_rate": 5.4239518848457396e-05, "loss": 1.0666, "step": 182915 }, { "epoch": 0.46, "learning_rate": 5.423826060697569e-05, "loss": 1.0444, "step": 182920 }, { "epoch": 0.46, "learning_rate": 5.423700236549398e-05, "loss": 1.0487, "step": 182925 }, { "epoch": 0.46, "learning_rate": 5.423574412401228e-05, "loss": 1.0455, "step": 182930 }, { "epoch": 0.46, "learning_rate": 5.423448588253057e-05, "loss": 1.0436, "step": 182935 }, { "epoch": 0.46, "learning_rate": 5.423322764104888e-05, "loss": 1.0463, "step": 182940 }, { "epoch": 0.46, "learning_rate": 5.423196939956717e-05, "loss": 1.0457, "step": 182945 }, { "epoch": 0.46, "learning_rate": 5.4230711158085465e-05, "loss": 1.0627, "step": 182950 }, { "epoch": 0.46, "learning_rate": 5.422945291660376e-05, "loss": 1.046, "step": 182955 }, { "epoch": 0.46, "learning_rate": 5.4228194675122055e-05, "loss": 1.0466, "step": 182960 }, { "epoch": 0.46, "learning_rate": 5.422693643364035e-05, "loss": 1.0676, "step": 182965 }, { "epoch": 0.46, "learning_rate": 5.4225678192158645e-05, "loss": 1.0439, "step": 182970 }, { "epoch": 0.46, "learning_rate": 5.422441995067694e-05, "loss": 1.0439, "step": 182975 }, { "epoch": 0.46, "learning_rate": 5.4223161709195235e-05, "loss": 1.0687, "step": 182980 }, { "epoch": 0.46, "learning_rate": 5.422190346771353e-05, "loss": 1.0453, "step": 182985 }, { "epoch": 0.46, "learning_rate": 5.4220645226231825e-05, "loss": 1.046, "step": 182990 }, { "epoch": 0.46, "learning_rate": 5.421938698475012e-05, "loss": 1.0431, "step": 182995 }, { "epoch": 0.46, "learning_rate": 5.4218128743268415e-05, "loss": 1.0441, "step": 183000 }, { "epoch": 0.46, "learning_rate": 5.421687050178671e-05, "loss": 1.0445, "step": 183005 }, { "epoch": 0.46, "learning_rate": 5.4215612260305004e-05, "loss": 1.0466, "step": 183010 }, { "epoch": 0.46, "learning_rate": 5.4214354018823296e-05, "loss": 1.067, "step": 183015 }, { "epoch": 0.46, "learning_rate": 5.4213095777341594e-05, "loss": 1.0448, "step": 183020 }, { "epoch": 0.46, "learning_rate": 5.4211837535859886e-05, "loss": 1.0451, "step": 183025 }, { "epoch": 0.46, "learning_rate": 5.4210579294378184e-05, "loss": 1.0431, "step": 183030 }, { "epoch": 0.46, "learning_rate": 5.4209321052896476e-05, "loss": 1.0454, "step": 183035 }, { "epoch": 0.46, "learning_rate": 5.4208062811414774e-05, "loss": 1.0466, "step": 183040 }, { "epoch": 0.46, "learning_rate": 5.4206804569933066e-05, "loss": 1.0439, "step": 183045 }, { "epoch": 0.46, "learning_rate": 5.4205546328451364e-05, "loss": 1.0434, "step": 183050 }, { "epoch": 0.46, "learning_rate": 5.4204288086969656e-05, "loss": 1.046, "step": 183055 }, { "epoch": 0.46, "learning_rate": 5.4203029845487954e-05, "loss": 1.0425, "step": 183060 }, { "epoch": 0.46, "learning_rate": 5.4201771604006246e-05, "loss": 1.0663, "step": 183065 }, { "epoch": 0.46, "learning_rate": 5.4200513362524544e-05, "loss": 1.0472, "step": 183070 }, { "epoch": 0.46, "learning_rate": 5.4199255121042836e-05, "loss": 1.0471, "step": 183075 }, { "epoch": 0.46, "learning_rate": 5.419799687956113e-05, "loss": 1.045, "step": 183080 }, { "epoch": 0.46, "learning_rate": 5.4196738638079426e-05, "loss": 1.0447, "step": 183085 }, { "epoch": 0.46, "learning_rate": 5.419548039659772e-05, "loss": 1.0434, "step": 183090 }, { "epoch": 0.46, "learning_rate": 5.4194222155116016e-05, "loss": 1.0455, "step": 183095 }, { "epoch": 0.46, "learning_rate": 5.419296391363431e-05, "loss": 1.0451, "step": 183100 }, { "epoch": 0.46, "learning_rate": 5.4191705672152606e-05, "loss": 1.0413, "step": 183105 }, { "epoch": 0.46, "learning_rate": 5.41904474306709e-05, "loss": 1.0435, "step": 183110 }, { "epoch": 0.46, "learning_rate": 5.4189189189189196e-05, "loss": 1.0443, "step": 183115 }, { "epoch": 0.46, "learning_rate": 5.418793094770749e-05, "loss": 1.0463, "step": 183120 }, { "epoch": 0.46, "learning_rate": 5.4186672706225786e-05, "loss": 1.062, "step": 183125 }, { "epoch": 0.46, "learning_rate": 5.418541446474408e-05, "loss": 1.0414, "step": 183130 }, { "epoch": 0.46, "learning_rate": 5.418415622326237e-05, "loss": 1.0462, "step": 183135 }, { "epoch": 0.46, "learning_rate": 5.418289798178067e-05, "loss": 1.0433, "step": 183140 }, { "epoch": 0.46, "learning_rate": 5.418163974029896e-05, "loss": 1.0477, "step": 183145 }, { "epoch": 0.46, "learning_rate": 5.418038149881726e-05, "loss": 1.0483, "step": 183150 }, { "epoch": 0.46, "learning_rate": 5.417912325733555e-05, "loss": 1.045, "step": 183155 }, { "epoch": 0.46, "learning_rate": 5.417786501585385e-05, "loss": 1.0435, "step": 183160 }, { "epoch": 0.46, "learning_rate": 5.417660677437214e-05, "loss": 1.0435, "step": 183165 }, { "epoch": 0.46, "learning_rate": 5.417534853289044e-05, "loss": 1.0456, "step": 183170 }, { "epoch": 0.46, "learning_rate": 5.417409029140873e-05, "loss": 1.0466, "step": 183175 }, { "epoch": 0.46, "learning_rate": 5.417283204992703e-05, "loss": 1.0448, "step": 183180 }, { "epoch": 0.46, "learning_rate": 5.417157380844532e-05, "loss": 1.0477, "step": 183185 }, { "epoch": 0.46, "learning_rate": 5.417031556696362e-05, "loss": 1.0443, "step": 183190 }, { "epoch": 0.46, "learning_rate": 5.416905732548191e-05, "loss": 1.0439, "step": 183195 }, { "epoch": 0.46, "learning_rate": 5.41677990840002e-05, "loss": 1.0453, "step": 183200 }, { "epoch": 0.46, "learning_rate": 5.41665408425185e-05, "loss": 1.0442, "step": 183205 }, { "epoch": 0.46, "learning_rate": 5.416528260103679e-05, "loss": 1.0433, "step": 183210 }, { "epoch": 0.46, "learning_rate": 5.416402435955509e-05, "loss": 1.0408, "step": 183215 }, { "epoch": 0.46, "learning_rate": 5.416276611807338e-05, "loss": 1.0458, "step": 183220 }, { "epoch": 0.46, "learning_rate": 5.416150787659168e-05, "loss": 1.045, "step": 183225 }, { "epoch": 0.46, "learning_rate": 5.416024963510997e-05, "loss": 1.0463, "step": 183230 }, { "epoch": 0.46, "learning_rate": 5.415899139362827e-05, "loss": 1.0458, "step": 183235 }, { "epoch": 0.46, "learning_rate": 5.415773315214656e-05, "loss": 1.0442, "step": 183240 }, { "epoch": 0.46, "learning_rate": 5.415647491066486e-05, "loss": 1.046, "step": 183245 }, { "epoch": 0.46, "learning_rate": 5.415521666918315e-05, "loss": 1.0446, "step": 183250 }, { "epoch": 0.46, "learning_rate": 5.415395842770145e-05, "loss": 1.0461, "step": 183255 }, { "epoch": 0.46, "learning_rate": 5.415270018621974e-05, "loss": 1.0481, "step": 183260 }, { "epoch": 0.46, "learning_rate": 5.415144194473803e-05, "loss": 1.0443, "step": 183265 }, { "epoch": 0.46, "learning_rate": 5.415018370325633e-05, "loss": 1.0636, "step": 183270 }, { "epoch": 0.46, "learning_rate": 5.414892546177462e-05, "loss": 1.0481, "step": 183275 }, { "epoch": 0.46, "learning_rate": 5.414766722029292e-05, "loss": 1.0432, "step": 183280 }, { "epoch": 0.46, "learning_rate": 5.414640897881121e-05, "loss": 1.0462, "step": 183285 }, { "epoch": 0.46, "learning_rate": 5.414515073732951e-05, "loss": 1.0733, "step": 183290 }, { "epoch": 0.46, "learning_rate": 5.41438924958478e-05, "loss": 1.0444, "step": 183295 }, { "epoch": 0.46, "learning_rate": 5.41426342543661e-05, "loss": 1.0446, "step": 183300 }, { "epoch": 0.46, "learning_rate": 5.414137601288439e-05, "loss": 1.0459, "step": 183305 }, { "epoch": 0.46, "learning_rate": 5.414011777140269e-05, "loss": 1.0463, "step": 183310 }, { "epoch": 0.46, "learning_rate": 5.413885952992098e-05, "loss": 1.048, "step": 183315 }, { "epoch": 0.46, "learning_rate": 5.413760128843928e-05, "loss": 1.0438, "step": 183320 }, { "epoch": 0.46, "learning_rate": 5.413634304695757e-05, "loss": 1.046, "step": 183325 }, { "epoch": 0.46, "learning_rate": 5.413508480547586e-05, "loss": 1.0502, "step": 183330 }, { "epoch": 0.46, "learning_rate": 5.413382656399416e-05, "loss": 1.0436, "step": 183335 }, { "epoch": 0.46, "learning_rate": 5.413256832251245e-05, "loss": 1.0415, "step": 183340 }, { "epoch": 0.46, "learning_rate": 5.413131008103075e-05, "loss": 1.0447, "step": 183345 }, { "epoch": 0.46, "learning_rate": 5.413005183954904e-05, "loss": 1.0459, "step": 183350 }, { "epoch": 0.46, "learning_rate": 5.412879359806734e-05, "loss": 1.0453, "step": 183355 }, { "epoch": 0.46, "learning_rate": 5.412753535658563e-05, "loss": 1.0456, "step": 183360 }, { "epoch": 0.46, "learning_rate": 5.412627711510393e-05, "loss": 1.0476, "step": 183365 }, { "epoch": 0.46, "learning_rate": 5.412501887362222e-05, "loss": 1.0443, "step": 183370 }, { "epoch": 0.46, "learning_rate": 5.412376063214052e-05, "loss": 1.0461, "step": 183375 }, { "epoch": 0.46, "learning_rate": 5.412250239065881e-05, "loss": 1.0546, "step": 183380 }, { "epoch": 0.46, "learning_rate": 5.412124414917711e-05, "loss": 1.0474, "step": 183385 }, { "epoch": 0.46, "learning_rate": 5.41199859076954e-05, "loss": 1.0494, "step": 183390 }, { "epoch": 0.46, "learning_rate": 5.4118727666213695e-05, "loss": 1.0444, "step": 183395 }, { "epoch": 0.46, "learning_rate": 5.411746942473199e-05, "loss": 1.0469, "step": 183400 }, { "epoch": 0.46, "learning_rate": 5.4116211183250285e-05, "loss": 1.0471, "step": 183405 }, { "epoch": 0.46, "learning_rate": 5.411495294176858e-05, "loss": 1.0464, "step": 183410 }, { "epoch": 0.46, "learning_rate": 5.4113694700286875e-05, "loss": 1.0459, "step": 183415 }, { "epoch": 0.46, "learning_rate": 5.411243645880517e-05, "loss": 1.0471, "step": 183420 }, { "epoch": 0.46, "learning_rate": 5.4111178217323465e-05, "loss": 1.0459, "step": 183425 }, { "epoch": 0.46, "learning_rate": 5.410991997584176e-05, "loss": 1.0471, "step": 183430 }, { "epoch": 0.46, "learning_rate": 5.4108661734360055e-05, "loss": 1.0417, "step": 183435 }, { "epoch": 0.46, "learning_rate": 5.410740349287836e-05, "loss": 1.0442, "step": 183440 }, { "epoch": 0.46, "learning_rate": 5.410614525139666e-05, "loss": 1.0421, "step": 183445 }, { "epoch": 0.46, "learning_rate": 5.410488700991495e-05, "loss": 1.0455, "step": 183450 }, { "epoch": 0.46, "learning_rate": 5.410362876843325e-05, "loss": 1.0464, "step": 183455 }, { "epoch": 0.46, "learning_rate": 5.410237052695154e-05, "loss": 1.0439, "step": 183460 }, { "epoch": 0.46, "learning_rate": 5.410111228546984e-05, "loss": 1.0471, "step": 183465 }, { "epoch": 0.46, "learning_rate": 5.409985404398813e-05, "loss": 1.0457, "step": 183470 }, { "epoch": 0.46, "learning_rate": 5.409859580250642e-05, "loss": 1.0911, "step": 183475 }, { "epoch": 0.46, "learning_rate": 5.409733756102472e-05, "loss": 1.0413, "step": 183480 }, { "epoch": 0.46, "learning_rate": 5.409607931954301e-05, "loss": 1.0455, "step": 183485 }, { "epoch": 0.46, "learning_rate": 5.409482107806131e-05, "loss": 1.0469, "step": 183490 }, { "epoch": 0.46, "learning_rate": 5.40935628365796e-05, "loss": 1.0481, "step": 183495 }, { "epoch": 0.46, "learning_rate": 5.40923045950979e-05, "loss": 1.0458, "step": 183500 }, { "epoch": 0.46, "learning_rate": 5.409104635361619e-05, "loss": 1.0447, "step": 183505 }, { "epoch": 0.46, "learning_rate": 5.408978811213449e-05, "loss": 1.046, "step": 183510 }, { "epoch": 0.46, "learning_rate": 5.408852987065278e-05, "loss": 1.0463, "step": 183515 }, { "epoch": 0.46, "learning_rate": 5.408727162917108e-05, "loss": 1.0438, "step": 183520 }, { "epoch": 0.46, "learning_rate": 5.408601338768937e-05, "loss": 1.0455, "step": 183525 }, { "epoch": 0.46, "learning_rate": 5.408475514620767e-05, "loss": 1.0451, "step": 183530 }, { "epoch": 0.46, "learning_rate": 5.408349690472596e-05, "loss": 1.0439, "step": 183535 }, { "epoch": 0.46, "learning_rate": 5.408223866324425e-05, "loss": 1.0452, "step": 183540 }, { "epoch": 0.46, "learning_rate": 5.408098042176255e-05, "loss": 1.0436, "step": 183545 }, { "epoch": 0.46, "learning_rate": 5.407972218028084e-05, "loss": 1.0513, "step": 183550 }, { "epoch": 0.46, "learning_rate": 5.407846393879914e-05, "loss": 1.0453, "step": 183555 }, { "epoch": 0.46, "learning_rate": 5.407720569731743e-05, "loss": 1.0441, "step": 183560 }, { "epoch": 0.46, "learning_rate": 5.407594745583573e-05, "loss": 1.0447, "step": 183565 }, { "epoch": 0.46, "learning_rate": 5.407468921435402e-05, "loss": 1.0487, "step": 183570 }, { "epoch": 0.46, "learning_rate": 5.407343097287232e-05, "loss": 1.0464, "step": 183575 }, { "epoch": 0.46, "learning_rate": 5.407217273139061e-05, "loss": 1.0471, "step": 183580 }, { "epoch": 0.46, "learning_rate": 5.407091448990891e-05, "loss": 1.0452, "step": 183585 }, { "epoch": 0.46, "learning_rate": 5.40696562484272e-05, "loss": 1.0427, "step": 183590 }, { "epoch": 0.46, "learning_rate": 5.40683980069455e-05, "loss": 1.0438, "step": 183595 }, { "epoch": 0.46, "learning_rate": 5.406713976546379e-05, "loss": 1.0432, "step": 183600 }, { "epoch": 0.46, "learning_rate": 5.4065881523982084e-05, "loss": 1.0471, "step": 183605 }, { "epoch": 0.46, "learning_rate": 5.406462328250038e-05, "loss": 1.0436, "step": 183610 }, { "epoch": 0.46, "learning_rate": 5.4063365041018674e-05, "loss": 1.0465, "step": 183615 }, { "epoch": 0.46, "learning_rate": 5.406210679953697e-05, "loss": 1.0601, "step": 183620 }, { "epoch": 0.46, "learning_rate": 5.4060848558055264e-05, "loss": 1.0479, "step": 183625 }, { "epoch": 0.46, "learning_rate": 5.405959031657356e-05, "loss": 1.0434, "step": 183630 }, { "epoch": 0.46, "learning_rate": 5.4058332075091854e-05, "loss": 1.0444, "step": 183635 }, { "epoch": 0.46, "learning_rate": 5.405707383361015e-05, "loss": 1.0467, "step": 183640 }, { "epoch": 0.46, "learning_rate": 5.4055815592128444e-05, "loss": 1.0417, "step": 183645 }, { "epoch": 0.46, "learning_rate": 5.405455735064674e-05, "loss": 1.0435, "step": 183650 }, { "epoch": 0.46, "learning_rate": 5.4053299109165034e-05, "loss": 1.0446, "step": 183655 }, { "epoch": 0.46, "learning_rate": 5.405204086768333e-05, "loss": 1.0496, "step": 183660 }, { "epoch": 0.46, "learning_rate": 5.4050782626201624e-05, "loss": 1.0444, "step": 183665 }, { "epoch": 0.46, "learning_rate": 5.4049524384719915e-05, "loss": 1.0452, "step": 183670 }, { "epoch": 0.46, "learning_rate": 5.4048266143238214e-05, "loss": 1.0455, "step": 183675 }, { "epoch": 0.46, "learning_rate": 5.4047007901756505e-05, "loss": 1.0437, "step": 183680 }, { "epoch": 0.46, "learning_rate": 5.4045749660274804e-05, "loss": 1.0459, "step": 183685 }, { "epoch": 0.46, "learning_rate": 5.4044491418793095e-05, "loss": 1.0427, "step": 183690 }, { "epoch": 0.46, "learning_rate": 5.4043233177311394e-05, "loss": 1.0445, "step": 183695 }, { "epoch": 0.46, "learning_rate": 5.4041974935829685e-05, "loss": 1.0475, "step": 183700 }, { "epoch": 0.46, "learning_rate": 5.4040716694347984e-05, "loss": 1.0447, "step": 183705 }, { "epoch": 0.46, "learning_rate": 5.4039458452866275e-05, "loss": 1.0461, "step": 183710 }, { "epoch": 0.46, "learning_rate": 5.4038200211384574e-05, "loss": 1.0448, "step": 183715 }, { "epoch": 0.46, "learning_rate": 5.4036941969902865e-05, "loss": 1.0449, "step": 183720 }, { "epoch": 0.46, "learning_rate": 5.4035683728421164e-05, "loss": 1.0467, "step": 183725 }, { "epoch": 0.46, "learning_rate": 5.4034425486939455e-05, "loss": 1.0471, "step": 183730 }, { "epoch": 0.46, "learning_rate": 5.403316724545775e-05, "loss": 1.0465, "step": 183735 }, { "epoch": 0.46, "learning_rate": 5.4031909003976045e-05, "loss": 1.0476, "step": 183740 }, { "epoch": 0.46, "learning_rate": 5.403065076249434e-05, "loss": 1.049, "step": 183745 }, { "epoch": 0.46, "learning_rate": 5.4029392521012635e-05, "loss": 1.0426, "step": 183750 }, { "epoch": 0.46, "learning_rate": 5.402813427953093e-05, "loss": 1.0488, "step": 183755 }, { "epoch": 0.46, "learning_rate": 5.4026876038049225e-05, "loss": 1.0451, "step": 183760 }, { "epoch": 0.46, "learning_rate": 5.402561779656752e-05, "loss": 1.0646, "step": 183765 }, { "epoch": 0.46, "learning_rate": 5.4024359555085815e-05, "loss": 1.0436, "step": 183770 }, { "epoch": 0.46, "learning_rate": 5.402310131360411e-05, "loss": 1.0451, "step": 183775 }, { "epoch": 0.46, "learning_rate": 5.4021843072122405e-05, "loss": 1.0659, "step": 183780 }, { "epoch": 0.46, "learning_rate": 5.40205848306407e-05, "loss": 1.0463, "step": 183785 }, { "epoch": 0.46, "learning_rate": 5.401932658915899e-05, "loss": 1.0482, "step": 183790 }, { "epoch": 0.46, "learning_rate": 5.401806834767729e-05, "loss": 1.0453, "step": 183795 }, { "epoch": 0.46, "learning_rate": 5.401681010619558e-05, "loss": 1.0475, "step": 183800 }, { "epoch": 0.46, "learning_rate": 5.401555186471388e-05, "loss": 1.0488, "step": 183805 }, { "epoch": 0.46, "learning_rate": 5.401429362323217e-05, "loss": 1.044, "step": 183810 }, { "epoch": 0.46, "learning_rate": 5.401303538175047e-05, "loss": 1.0423, "step": 183815 }, { "epoch": 0.46, "learning_rate": 5.401177714026876e-05, "loss": 1.0447, "step": 183820 }, { "epoch": 0.46, "learning_rate": 5.401051889878706e-05, "loss": 1.0645, "step": 183825 }, { "epoch": 0.46, "learning_rate": 5.400926065730535e-05, "loss": 1.0459, "step": 183830 }, { "epoch": 0.46, "learning_rate": 5.400800241582365e-05, "loss": 1.0467, "step": 183835 }, { "epoch": 0.46, "learning_rate": 5.400674417434194e-05, "loss": 1.0457, "step": 183840 }, { "epoch": 0.46, "learning_rate": 5.4005485932860237e-05, "loss": 1.048, "step": 183845 }, { "epoch": 0.46, "learning_rate": 5.400422769137853e-05, "loss": 1.0438, "step": 183850 }, { "epoch": 0.46, "learning_rate": 5.400296944989682e-05, "loss": 1.0441, "step": 183855 }, { "epoch": 0.46, "learning_rate": 5.400171120841512e-05, "loss": 1.0463, "step": 183860 }, { "epoch": 0.46, "learning_rate": 5.400045296693341e-05, "loss": 1.0444, "step": 183865 }, { "epoch": 0.46, "learning_rate": 5.399919472545171e-05, "loss": 1.0473, "step": 183870 }, { "epoch": 0.46, "learning_rate": 5.399793648397e-05, "loss": 1.0438, "step": 183875 }, { "epoch": 0.46, "learning_rate": 5.39966782424883e-05, "loss": 1.0474, "step": 183880 }, { "epoch": 0.46, "learning_rate": 5.399542000100659e-05, "loss": 1.0434, "step": 183885 }, { "epoch": 0.46, "learning_rate": 5.399416175952489e-05, "loss": 1.0452, "step": 183890 }, { "epoch": 0.46, "learning_rate": 5.399290351804318e-05, "loss": 1.0498, "step": 183895 }, { "epoch": 0.46, "learning_rate": 5.399164527656148e-05, "loss": 1.0433, "step": 183900 }, { "epoch": 0.46, "learning_rate": 5.399038703507977e-05, "loss": 1.045, "step": 183905 }, { "epoch": 0.46, "learning_rate": 5.398912879359807e-05, "loss": 1.0428, "step": 183910 }, { "epoch": 0.46, "learning_rate": 5.398787055211636e-05, "loss": 1.044, "step": 183915 }, { "epoch": 0.46, "learning_rate": 5.398661231063465e-05, "loss": 1.0417, "step": 183920 }, { "epoch": 0.46, "learning_rate": 5.398535406915295e-05, "loss": 1.0476, "step": 183925 }, { "epoch": 0.46, "learning_rate": 5.398409582767124e-05, "loss": 1.0444, "step": 183930 }, { "epoch": 0.46, "learning_rate": 5.398283758618954e-05, "loss": 1.0423, "step": 183935 }, { "epoch": 0.46, "learning_rate": 5.3981579344707845e-05, "loss": 1.0443, "step": 183940 }, { "epoch": 0.46, "learning_rate": 5.3980321103226136e-05, "loss": 1.0658, "step": 183945 }, { "epoch": 0.46, "learning_rate": 5.3979062861744435e-05, "loss": 1.0461, "step": 183950 }, { "epoch": 0.46, "learning_rate": 5.3977804620262726e-05, "loss": 1.0468, "step": 183955 }, { "epoch": 0.46, "learning_rate": 5.3976546378781025e-05, "loss": 1.0466, "step": 183960 }, { "epoch": 0.46, "learning_rate": 5.3975288137299316e-05, "loss": 1.0493, "step": 183965 }, { "epoch": 0.46, "learning_rate": 5.3974029895817615e-05, "loss": 1.0455, "step": 183970 }, { "epoch": 0.46, "learning_rate": 5.3972771654335906e-05, "loss": 1.0439, "step": 183975 }, { "epoch": 0.46, "learning_rate": 5.3971513412854205e-05, "loss": 1.0442, "step": 183980 }, { "epoch": 0.46, "learning_rate": 5.3970255171372496e-05, "loss": 1.0454, "step": 183985 }, { "epoch": 0.46, "learning_rate": 5.3968996929890795e-05, "loss": 1.0446, "step": 183990 }, { "epoch": 0.46, "learning_rate": 5.3967738688409086e-05, "loss": 1.0424, "step": 183995 }, { "epoch": 0.46, "learning_rate": 5.396648044692738e-05, "loss": 1.0427, "step": 184000 }, { "epoch": 0.46, "learning_rate": 5.3965222205445676e-05, "loss": 1.0495, "step": 184005 }, { "epoch": 0.46, "learning_rate": 5.396396396396397e-05, "loss": 1.0458, "step": 184010 }, { "epoch": 0.46, "learning_rate": 5.3962705722482266e-05, "loss": 1.0465, "step": 184015 }, { "epoch": 0.46, "learning_rate": 5.396144748100056e-05, "loss": 1.0454, "step": 184020 }, { "epoch": 0.46, "learning_rate": 5.3960189239518856e-05, "loss": 1.0465, "step": 184025 }, { "epoch": 0.46, "learning_rate": 5.395893099803715e-05, "loss": 1.0457, "step": 184030 }, { "epoch": 0.46, "learning_rate": 5.3957672756555446e-05, "loss": 1.047, "step": 184035 }, { "epoch": 0.46, "learning_rate": 5.395641451507374e-05, "loss": 1.0666, "step": 184040 }, { "epoch": 0.46, "learning_rate": 5.3955156273592036e-05, "loss": 1.0449, "step": 184045 }, { "epoch": 0.46, "learning_rate": 5.395389803211033e-05, "loss": 1.0469, "step": 184050 }, { "epoch": 0.46, "learning_rate": 5.3952639790628626e-05, "loss": 1.0461, "step": 184055 }, { "epoch": 0.46, "learning_rate": 5.395138154914692e-05, "loss": 1.0467, "step": 184060 }, { "epoch": 0.46, "learning_rate": 5.395012330766521e-05, "loss": 1.0456, "step": 184065 }, { "epoch": 0.46, "learning_rate": 5.394886506618351e-05, "loss": 1.042, "step": 184070 }, { "epoch": 0.46, "learning_rate": 5.39476068247018e-05, "loss": 1.0441, "step": 184075 }, { "epoch": 0.46, "learning_rate": 5.39463485832201e-05, "loss": 1.0463, "step": 184080 }, { "epoch": 0.46, "learning_rate": 5.394509034173839e-05, "loss": 1.0439, "step": 184085 }, { "epoch": 0.46, "learning_rate": 5.394383210025669e-05, "loss": 1.047, "step": 184090 }, { "epoch": 0.46, "learning_rate": 5.394257385877498e-05, "loss": 1.0448, "step": 184095 }, { "epoch": 0.46, "learning_rate": 5.394131561729328e-05, "loss": 1.0472, "step": 184100 }, { "epoch": 0.46, "learning_rate": 5.394005737581157e-05, "loss": 1.0433, "step": 184105 }, { "epoch": 0.46, "learning_rate": 5.393879913432987e-05, "loss": 1.0478, "step": 184110 }, { "epoch": 0.46, "learning_rate": 5.393754089284816e-05, "loss": 1.044, "step": 184115 }, { "epoch": 0.46, "learning_rate": 5.393628265136646e-05, "loss": 1.0472, "step": 184120 }, { "epoch": 0.46, "learning_rate": 5.393502440988475e-05, "loss": 1.0457, "step": 184125 }, { "epoch": 0.46, "learning_rate": 5.393376616840304e-05, "loss": 1.0485, "step": 184130 }, { "epoch": 0.46, "learning_rate": 5.393250792692134e-05, "loss": 1.0845, "step": 184135 }, { "epoch": 0.46, "learning_rate": 5.393124968543963e-05, "loss": 1.0472, "step": 184140 }, { "epoch": 0.46, "learning_rate": 5.392999144395793e-05, "loss": 1.0486, "step": 184145 }, { "epoch": 0.46, "learning_rate": 5.392873320247622e-05, "loss": 1.0462, "step": 184150 }, { "epoch": 0.46, "learning_rate": 5.392747496099452e-05, "loss": 1.0455, "step": 184155 }, { "epoch": 0.46, "learning_rate": 5.392621671951281e-05, "loss": 1.0474, "step": 184160 }, { "epoch": 0.46, "learning_rate": 5.392495847803111e-05, "loss": 1.047, "step": 184165 }, { "epoch": 0.46, "learning_rate": 5.39237002365494e-05, "loss": 1.0446, "step": 184170 }, { "epoch": 0.46, "learning_rate": 5.39224419950677e-05, "loss": 1.0455, "step": 184175 }, { "epoch": 0.46, "learning_rate": 5.392118375358599e-05, "loss": 1.0448, "step": 184180 }, { "epoch": 0.46, "learning_rate": 5.391992551210429e-05, "loss": 1.0485, "step": 184185 }, { "epoch": 0.46, "learning_rate": 5.391866727062258e-05, "loss": 1.049, "step": 184190 }, { "epoch": 0.46, "learning_rate": 5.391740902914087e-05, "loss": 1.0434, "step": 184195 }, { "epoch": 0.46, "learning_rate": 5.391615078765917e-05, "loss": 1.0456, "step": 184200 }, { "epoch": 0.46, "learning_rate": 5.391489254617746e-05, "loss": 1.0462, "step": 184205 }, { "epoch": 0.46, "learning_rate": 5.391363430469576e-05, "loss": 1.0446, "step": 184210 }, { "epoch": 0.46, "learning_rate": 5.391237606321405e-05, "loss": 1.0461, "step": 184215 }, { "epoch": 0.46, "learning_rate": 5.391111782173235e-05, "loss": 1.0459, "step": 184220 }, { "epoch": 0.46, "learning_rate": 5.390985958025064e-05, "loss": 1.0491, "step": 184225 }, { "epoch": 0.46, "learning_rate": 5.390860133876894e-05, "loss": 1.0424, "step": 184230 }, { "epoch": 0.46, "learning_rate": 5.390734309728723e-05, "loss": 1.0693, "step": 184235 }, { "epoch": 0.46, "learning_rate": 5.390608485580553e-05, "loss": 1.0451, "step": 184240 }, { "epoch": 0.46, "learning_rate": 5.390482661432382e-05, "loss": 1.0482, "step": 184245 }, { "epoch": 0.46, "learning_rate": 5.390356837284212e-05, "loss": 1.0443, "step": 184250 }, { "epoch": 0.46, "learning_rate": 5.390231013136041e-05, "loss": 1.0442, "step": 184255 }, { "epoch": 0.46, "learning_rate": 5.3901051889878703e-05, "loss": 1.0479, "step": 184260 }, { "epoch": 0.46, "learning_rate": 5.3899793648397e-05, "loss": 1.045, "step": 184265 }, { "epoch": 0.46, "learning_rate": 5.3898535406915293e-05, "loss": 1.0449, "step": 184270 }, { "epoch": 0.46, "learning_rate": 5.389727716543359e-05, "loss": 1.0431, "step": 184275 }, { "epoch": 0.46, "learning_rate": 5.3896018923951883e-05, "loss": 1.0448, "step": 184280 }, { "epoch": 0.46, "learning_rate": 5.389476068247018e-05, "loss": 1.0488, "step": 184285 }, { "epoch": 0.46, "learning_rate": 5.389350244098847e-05, "loss": 1.043, "step": 184290 }, { "epoch": 0.46, "learning_rate": 5.389224419950677e-05, "loss": 1.0641, "step": 184295 }, { "epoch": 0.46, "learning_rate": 5.389098595802506e-05, "loss": 1.0431, "step": 184300 }, { "epoch": 0.46, "learning_rate": 5.388972771654336e-05, "loss": 1.0431, "step": 184305 }, { "epoch": 0.46, "learning_rate": 5.388846947506165e-05, "loss": 1.0468, "step": 184310 }, { "epoch": 0.46, "learning_rate": 5.388721123357995e-05, "loss": 1.0429, "step": 184315 }, { "epoch": 0.46, "learning_rate": 5.388595299209824e-05, "loss": 1.0456, "step": 184320 }, { "epoch": 0.46, "learning_rate": 5.3884694750616535e-05, "loss": 1.0465, "step": 184325 }, { "epoch": 0.46, "learning_rate": 5.388343650913483e-05, "loss": 1.0405, "step": 184330 }, { "epoch": 0.46, "learning_rate": 5.3882178267653125e-05, "loss": 1.0456, "step": 184335 }, { "epoch": 0.46, "learning_rate": 5.388092002617142e-05, "loss": 1.0432, "step": 184340 }, { "epoch": 0.46, "learning_rate": 5.3879661784689715e-05, "loss": 1.0471, "step": 184345 }, { "epoch": 0.46, "learning_rate": 5.387840354320801e-05, "loss": 1.0464, "step": 184350 }, { "epoch": 0.46, "learning_rate": 5.3877145301726305e-05, "loss": 1.0466, "step": 184355 }, { "epoch": 0.46, "learning_rate": 5.38758870602446e-05, "loss": 1.0444, "step": 184360 }, { "epoch": 0.46, "learning_rate": 5.3874628818762895e-05, "loss": 1.0448, "step": 184365 }, { "epoch": 0.46, "learning_rate": 5.387337057728119e-05, "loss": 1.0449, "step": 184370 }, { "epoch": 0.46, "learning_rate": 5.3872112335799485e-05, "loss": 1.0443, "step": 184375 }, { "epoch": 0.46, "learning_rate": 5.387085409431778e-05, "loss": 1.0464, "step": 184380 }, { "epoch": 0.46, "learning_rate": 5.3869595852836075e-05, "loss": 1.0642, "step": 184385 }, { "epoch": 0.46, "learning_rate": 5.3868337611354366e-05, "loss": 1.0437, "step": 184390 }, { "epoch": 0.46, "learning_rate": 5.3867079369872665e-05, "loss": 1.0453, "step": 184395 }, { "epoch": 0.46, "learning_rate": 5.3865821128390956e-05, "loss": 1.0465, "step": 184400 }, { "epoch": 0.46, "learning_rate": 5.3864562886909255e-05, "loss": 1.0463, "step": 184405 }, { "epoch": 0.46, "learning_rate": 5.3863304645427546e-05, "loss": 1.0447, "step": 184410 }, { "epoch": 0.46, "learning_rate": 5.3862046403945845e-05, "loss": 1.0469, "step": 184415 }, { "epoch": 0.46, "learning_rate": 5.3860788162464136e-05, "loss": 1.0425, "step": 184420 }, { "epoch": 0.46, "learning_rate": 5.3859529920982435e-05, "loss": 1.0475, "step": 184425 }, { "epoch": 0.46, "learning_rate": 5.3858271679500726e-05, "loss": 1.0465, "step": 184430 }, { "epoch": 0.46, "learning_rate": 5.3857013438019025e-05, "loss": 1.0449, "step": 184435 }, { "epoch": 0.46, "learning_rate": 5.385575519653733e-05, "loss": 1.0448, "step": 184440 }, { "epoch": 0.46, "learning_rate": 5.385449695505562e-05, "loss": 1.0402, "step": 184445 }, { "epoch": 0.46, "learning_rate": 5.385323871357392e-05, "loss": 1.0482, "step": 184450 }, { "epoch": 0.46, "learning_rate": 5.385198047209221e-05, "loss": 1.0443, "step": 184455 }, { "epoch": 0.46, "learning_rate": 5.385072223061051e-05, "loss": 1.0449, "step": 184460 }, { "epoch": 0.46, "learning_rate": 5.38494639891288e-05, "loss": 1.0425, "step": 184465 }, { "epoch": 0.46, "learning_rate": 5.384820574764709e-05, "loss": 1.046, "step": 184470 }, { "epoch": 0.46, "learning_rate": 5.384694750616539e-05, "loss": 1.0435, "step": 184475 }, { "epoch": 0.46, "learning_rate": 5.384568926468368e-05, "loss": 1.041, "step": 184480 }, { "epoch": 0.46, "learning_rate": 5.384443102320198e-05, "loss": 1.0466, "step": 184485 }, { "epoch": 0.46, "learning_rate": 5.384317278172027e-05, "loss": 1.0461, "step": 184490 }, { "epoch": 0.46, "learning_rate": 5.384191454023857e-05, "loss": 1.0427, "step": 184495 }, { "epoch": 0.46, "learning_rate": 5.384065629875686e-05, "loss": 1.0451, "step": 184500 }, { "epoch": 0.46, "learning_rate": 5.383939805727516e-05, "loss": 1.0454, "step": 184505 }, { "epoch": 0.46, "learning_rate": 5.383813981579345e-05, "loss": 1.0464, "step": 184510 }, { "epoch": 0.46, "learning_rate": 5.383688157431175e-05, "loss": 1.0438, "step": 184515 }, { "epoch": 0.46, "learning_rate": 5.383562333283004e-05, "loss": 1.0463, "step": 184520 }, { "epoch": 0.46, "learning_rate": 5.383436509134834e-05, "loss": 1.045, "step": 184525 }, { "epoch": 0.46, "learning_rate": 5.383310684986663e-05, "loss": 1.0457, "step": 184530 }, { "epoch": 0.46, "learning_rate": 5.3831848608384924e-05, "loss": 1.0456, "step": 184535 }, { "epoch": 0.46, "learning_rate": 5.383059036690322e-05, "loss": 1.0477, "step": 184540 }, { "epoch": 0.46, "learning_rate": 5.3829332125421514e-05, "loss": 1.0484, "step": 184545 }, { "epoch": 0.46, "learning_rate": 5.382807388393981e-05, "loss": 1.0452, "step": 184550 }, { "epoch": 0.46, "learning_rate": 5.3826815642458104e-05, "loss": 1.0474, "step": 184555 }, { "epoch": 0.46, "learning_rate": 5.38255574009764e-05, "loss": 1.044, "step": 184560 }, { "epoch": 0.46, "learning_rate": 5.3824299159494694e-05, "loss": 1.0464, "step": 184565 }, { "epoch": 0.46, "learning_rate": 5.382304091801299e-05, "loss": 1.0431, "step": 184570 }, { "epoch": 0.46, "learning_rate": 5.3821782676531284e-05, "loss": 1.0464, "step": 184575 }, { "epoch": 0.46, "learning_rate": 5.382052443504958e-05, "loss": 1.0438, "step": 184580 }, { "epoch": 0.46, "learning_rate": 5.3819266193567874e-05, "loss": 1.0477, "step": 184585 }, { "epoch": 0.46, "learning_rate": 5.3818007952086166e-05, "loss": 1.0464, "step": 184590 }, { "epoch": 0.46, "learning_rate": 5.3816749710604464e-05, "loss": 1.0436, "step": 184595 }, { "epoch": 0.46, "learning_rate": 5.3815491469122756e-05, "loss": 1.0467, "step": 184600 }, { "epoch": 0.46, "learning_rate": 5.3814233227641054e-05, "loss": 1.0447, "step": 184605 }, { "epoch": 0.46, "learning_rate": 5.3812974986159346e-05, "loss": 1.0439, "step": 184610 }, { "epoch": 0.46, "learning_rate": 5.3811716744677644e-05, "loss": 1.0438, "step": 184615 }, { "epoch": 0.46, "learning_rate": 5.3810458503195936e-05, "loss": 1.0436, "step": 184620 }, { "epoch": 0.46, "learning_rate": 5.3809200261714234e-05, "loss": 1.0449, "step": 184625 }, { "epoch": 0.46, "learning_rate": 5.3807942020232526e-05, "loss": 1.0615, "step": 184630 }, { "epoch": 0.46, "learning_rate": 5.3806683778750824e-05, "loss": 1.0438, "step": 184635 }, { "epoch": 0.46, "learning_rate": 5.3805425537269116e-05, "loss": 1.046, "step": 184640 }, { "epoch": 0.46, "learning_rate": 5.3804167295787414e-05, "loss": 1.0458, "step": 184645 }, { "epoch": 0.46, "learning_rate": 5.3802909054305706e-05, "loss": 1.0464, "step": 184650 }, { "epoch": 0.46, "learning_rate": 5.3801650812824e-05, "loss": 1.0438, "step": 184655 }, { "epoch": 0.46, "learning_rate": 5.3800392571342296e-05, "loss": 1.0469, "step": 184660 }, { "epoch": 0.46, "learning_rate": 5.379913432986059e-05, "loss": 1.0444, "step": 184665 }, { "epoch": 0.46, "learning_rate": 5.3797876088378885e-05, "loss": 1.0452, "step": 184670 }, { "epoch": 0.46, "learning_rate": 5.379661784689718e-05, "loss": 1.0479, "step": 184675 }, { "epoch": 0.46, "learning_rate": 5.3795359605415475e-05, "loss": 1.0434, "step": 184680 }, { "epoch": 0.46, "learning_rate": 5.379410136393377e-05, "loss": 1.0699, "step": 184685 }, { "epoch": 0.46, "learning_rate": 5.3792843122452065e-05, "loss": 1.0463, "step": 184690 }, { "epoch": 0.46, "learning_rate": 5.379158488097036e-05, "loss": 1.0469, "step": 184695 }, { "epoch": 0.46, "learning_rate": 5.3790326639488655e-05, "loss": 1.0446, "step": 184700 }, { "epoch": 0.46, "learning_rate": 5.378906839800695e-05, "loss": 1.0444, "step": 184705 }, { "epoch": 0.46, "learning_rate": 5.3787810156525245e-05, "loss": 1.049, "step": 184710 }, { "epoch": 0.46, "learning_rate": 5.378655191504354e-05, "loss": 1.0448, "step": 184715 }, { "epoch": 0.46, "learning_rate": 5.378529367356183e-05, "loss": 1.046, "step": 184720 }, { "epoch": 0.46, "learning_rate": 5.378403543208013e-05, "loss": 1.0446, "step": 184725 }, { "epoch": 0.46, "learning_rate": 5.378277719059842e-05, "loss": 1.0478, "step": 184730 }, { "epoch": 0.46, "learning_rate": 5.378151894911672e-05, "loss": 1.0424, "step": 184735 }, { "epoch": 0.46, "learning_rate": 5.378026070763501e-05, "loss": 1.0446, "step": 184740 }, { "epoch": 0.46, "learning_rate": 5.377900246615331e-05, "loss": 1.0486, "step": 184745 }, { "epoch": 0.46, "learning_rate": 5.37777442246716e-05, "loss": 1.0432, "step": 184750 }, { "epoch": 0.46, "learning_rate": 5.37764859831899e-05, "loss": 1.0454, "step": 184755 }, { "epoch": 0.46, "learning_rate": 5.377522774170819e-05, "loss": 1.0457, "step": 184760 }, { "epoch": 0.46, "learning_rate": 5.377396950022649e-05, "loss": 1.0434, "step": 184765 }, { "epoch": 0.46, "learning_rate": 5.377271125874478e-05, "loss": 1.0451, "step": 184770 }, { "epoch": 0.46, "learning_rate": 5.377145301726308e-05, "loss": 1.048, "step": 184775 }, { "epoch": 0.46, "learning_rate": 5.377019477578137e-05, "loss": 1.0445, "step": 184780 }, { "epoch": 0.46, "learning_rate": 5.376893653429966e-05, "loss": 1.043, "step": 184785 }, { "epoch": 0.46, "learning_rate": 5.376767829281796e-05, "loss": 1.0471, "step": 184790 }, { "epoch": 0.46, "learning_rate": 5.376642005133625e-05, "loss": 1.0466, "step": 184795 }, { "epoch": 0.46, "learning_rate": 5.376516180985455e-05, "loss": 1.0477, "step": 184800 }, { "epoch": 0.46, "learning_rate": 5.376390356837284e-05, "loss": 1.0482, "step": 184805 }, { "epoch": 0.46, "learning_rate": 5.376264532689114e-05, "loss": 1.0432, "step": 184810 }, { "epoch": 0.46, "learning_rate": 5.376138708540943e-05, "loss": 1.0443, "step": 184815 }, { "epoch": 0.46, "learning_rate": 5.376012884392773e-05, "loss": 1.0446, "step": 184820 }, { "epoch": 0.46, "learning_rate": 5.375887060244602e-05, "loss": 1.045, "step": 184825 }, { "epoch": 0.46, "learning_rate": 5.375761236096432e-05, "loss": 1.0464, "step": 184830 }, { "epoch": 0.46, "learning_rate": 5.375635411948261e-05, "loss": 1.0467, "step": 184835 }, { "epoch": 0.46, "learning_rate": 5.375509587800091e-05, "loss": 1.0437, "step": 184840 }, { "epoch": 0.46, "learning_rate": 5.37538376365192e-05, "loss": 1.044, "step": 184845 }, { "epoch": 0.46, "learning_rate": 5.375257939503749e-05, "loss": 1.0676, "step": 184850 }, { "epoch": 0.46, "learning_rate": 5.375132115355579e-05, "loss": 1.0486, "step": 184855 }, { "epoch": 0.46, "learning_rate": 5.375006291207408e-05, "loss": 1.0503, "step": 184860 }, { "epoch": 0.46, "learning_rate": 5.374880467059238e-05, "loss": 1.0497, "step": 184865 }, { "epoch": 0.46, "learning_rate": 5.374754642911067e-05, "loss": 1.0476, "step": 184870 }, { "epoch": 0.46, "learning_rate": 5.374628818762897e-05, "loss": 1.0451, "step": 184875 }, { "epoch": 0.46, "learning_rate": 5.374502994614726e-05, "loss": 1.046, "step": 184880 }, { "epoch": 0.46, "learning_rate": 5.374377170466556e-05, "loss": 1.0434, "step": 184885 }, { "epoch": 0.46, "learning_rate": 5.374251346318385e-05, "loss": 1.0452, "step": 184890 }, { "epoch": 0.46, "learning_rate": 5.374125522170215e-05, "loss": 1.0456, "step": 184895 }, { "epoch": 0.46, "learning_rate": 5.373999698022044e-05, "loss": 1.0453, "step": 184900 }, { "epoch": 0.46, "learning_rate": 5.373873873873874e-05, "loss": 1.0453, "step": 184905 }, { "epoch": 0.46, "learning_rate": 5.373748049725703e-05, "loss": 1.0451, "step": 184910 }, { "epoch": 0.46, "learning_rate": 5.373622225577532e-05, "loss": 1.0662, "step": 184915 }, { "epoch": 0.46, "learning_rate": 5.373496401429362e-05, "loss": 1.0474, "step": 184920 }, { "epoch": 0.46, "learning_rate": 5.373370577281191e-05, "loss": 1.0693, "step": 184925 }, { "epoch": 0.46, "learning_rate": 5.373244753133021e-05, "loss": 1.0462, "step": 184930 }, { "epoch": 0.46, "learning_rate": 5.37311892898485e-05, "loss": 1.0445, "step": 184935 }, { "epoch": 0.46, "learning_rate": 5.372993104836681e-05, "loss": 1.0467, "step": 184940 }, { "epoch": 0.46, "learning_rate": 5.3728672806885106e-05, "loss": 1.0468, "step": 184945 }, { "epoch": 0.46, "learning_rate": 5.37274145654034e-05, "loss": 1.0421, "step": 184950 }, { "epoch": 0.46, "learning_rate": 5.3726156323921696e-05, "loss": 1.0429, "step": 184955 }, { "epoch": 0.46, "learning_rate": 5.372489808243999e-05, "loss": 1.0473, "step": 184960 }, { "epoch": 0.46, "learning_rate": 5.3723639840958286e-05, "loss": 1.0434, "step": 184965 }, { "epoch": 0.46, "learning_rate": 5.372238159947658e-05, "loss": 1.0493, "step": 184970 }, { "epoch": 0.46, "learning_rate": 5.3721123357994876e-05, "loss": 1.0457, "step": 184975 }, { "epoch": 0.46, "learning_rate": 5.371986511651317e-05, "loss": 1.0451, "step": 184980 }, { "epoch": 0.46, "learning_rate": 5.3718606875031466e-05, "loss": 1.0443, "step": 184985 }, { "epoch": 0.46, "learning_rate": 5.371734863354976e-05, "loss": 1.0441, "step": 184990 }, { "epoch": 0.46, "learning_rate": 5.371609039206805e-05, "loss": 1.0483, "step": 184995 }, { "epoch": 0.46, "learning_rate": 5.371483215058635e-05, "loss": 1.0712, "step": 185000 }, { "epoch": 0.46, "learning_rate": 5.371357390910464e-05, "loss": 1.0455, "step": 185005 }, { "epoch": 0.46, "learning_rate": 5.371231566762294e-05, "loss": 1.0454, "step": 185010 }, { "epoch": 0.46, "learning_rate": 5.371105742614123e-05, "loss": 1.0457, "step": 185015 }, { "epoch": 0.46, "learning_rate": 5.370979918465953e-05, "loss": 1.0453, "step": 185020 }, { "epoch": 0.46, "learning_rate": 5.370854094317782e-05, "loss": 1.0468, "step": 185025 }, { "epoch": 0.46, "learning_rate": 5.370728270169612e-05, "loss": 1.0454, "step": 185030 }, { "epoch": 0.46, "learning_rate": 5.370602446021441e-05, "loss": 1.0439, "step": 185035 }, { "epoch": 0.46, "learning_rate": 5.370476621873271e-05, "loss": 1.0431, "step": 185040 }, { "epoch": 0.46, "learning_rate": 5.3703507977251e-05, "loss": 1.0453, "step": 185045 }, { "epoch": 0.46, "learning_rate": 5.37022497357693e-05, "loss": 1.0462, "step": 185050 }, { "epoch": 0.46, "learning_rate": 5.370099149428759e-05, "loss": 1.0445, "step": 185055 }, { "epoch": 0.46, "learning_rate": 5.369973325280588e-05, "loss": 1.0475, "step": 185060 }, { "epoch": 0.46, "learning_rate": 5.369847501132418e-05, "loss": 1.0441, "step": 185065 }, { "epoch": 0.46, "learning_rate": 5.369721676984247e-05, "loss": 1.045, "step": 185070 }, { "epoch": 0.46, "learning_rate": 5.369595852836077e-05, "loss": 1.0456, "step": 185075 }, { "epoch": 0.46, "learning_rate": 5.369470028687906e-05, "loss": 1.0456, "step": 185080 }, { "epoch": 0.46, "learning_rate": 5.369344204539736e-05, "loss": 1.0443, "step": 185085 }, { "epoch": 0.46, "learning_rate": 5.369218380391565e-05, "loss": 1.0468, "step": 185090 }, { "epoch": 0.46, "learning_rate": 5.369092556243395e-05, "loss": 1.0469, "step": 185095 }, { "epoch": 0.46, "learning_rate": 5.368966732095224e-05, "loss": 1.0474, "step": 185100 }, { "epoch": 0.46, "learning_rate": 5.368840907947054e-05, "loss": 1.0438, "step": 185105 }, { "epoch": 0.46, "learning_rate": 5.368715083798883e-05, "loss": 1.0437, "step": 185110 }, { "epoch": 0.46, "learning_rate": 5.368589259650713e-05, "loss": 1.0454, "step": 185115 }, { "epoch": 0.46, "learning_rate": 5.368463435502542e-05, "loss": 1.0681, "step": 185120 }, { "epoch": 0.46, "learning_rate": 5.368337611354371e-05, "loss": 1.0428, "step": 185125 }, { "epoch": 0.46, "learning_rate": 5.368211787206201e-05, "loss": 1.0446, "step": 185130 }, { "epoch": 0.46, "learning_rate": 5.36808596305803e-05, "loss": 1.0449, "step": 185135 }, { "epoch": 0.46, "learning_rate": 5.36796013890986e-05, "loss": 1.0444, "step": 185140 }, { "epoch": 0.46, "learning_rate": 5.367834314761689e-05, "loss": 1.0498, "step": 185145 }, { "epoch": 0.46, "learning_rate": 5.367708490613519e-05, "loss": 1.0466, "step": 185150 }, { "epoch": 0.46, "learning_rate": 5.367582666465348e-05, "loss": 1.0468, "step": 185155 }, { "epoch": 0.46, "learning_rate": 5.367456842317178e-05, "loss": 1.0456, "step": 185160 }, { "epoch": 0.46, "learning_rate": 5.367331018169007e-05, "loss": 1.0463, "step": 185165 }, { "epoch": 0.46, "learning_rate": 5.367205194020837e-05, "loss": 1.0693, "step": 185170 }, { "epoch": 0.46, "learning_rate": 5.367079369872666e-05, "loss": 1.0452, "step": 185175 }, { "epoch": 0.46, "learning_rate": 5.366953545724496e-05, "loss": 1.0453, "step": 185180 }, { "epoch": 0.46, "learning_rate": 5.366827721576325e-05, "loss": 1.0439, "step": 185185 }, { "epoch": 0.46, "learning_rate": 5.3667018974281544e-05, "loss": 1.0447, "step": 185190 }, { "epoch": 0.46, "learning_rate": 5.366576073279984e-05, "loss": 1.0473, "step": 185195 }, { "epoch": 0.46, "learning_rate": 5.3664502491318134e-05, "loss": 1.0434, "step": 185200 }, { "epoch": 0.46, "learning_rate": 5.366324424983643e-05, "loss": 1.0441, "step": 185205 }, { "epoch": 0.46, "learning_rate": 5.3661986008354724e-05, "loss": 1.045, "step": 185210 }, { "epoch": 0.46, "learning_rate": 5.366072776687302e-05, "loss": 1.0436, "step": 185215 }, { "epoch": 0.46, "learning_rate": 5.3659469525391314e-05, "loss": 1.0433, "step": 185220 }, { "epoch": 0.46, "learning_rate": 5.365821128390961e-05, "loss": 1.0426, "step": 185225 }, { "epoch": 0.46, "learning_rate": 5.3656953042427904e-05, "loss": 1.0406, "step": 185230 }, { "epoch": 0.46, "learning_rate": 5.36556948009462e-05, "loss": 1.0423, "step": 185235 }, { "epoch": 0.46, "learning_rate": 5.3654436559464494e-05, "loss": 1.0445, "step": 185240 }, { "epoch": 0.46, "learning_rate": 5.3653178317982785e-05, "loss": 1.05, "step": 185245 }, { "epoch": 0.47, "learning_rate": 5.3651920076501083e-05, "loss": 1.0424, "step": 185250 }, { "epoch": 0.47, "learning_rate": 5.3650661835019375e-05, "loss": 1.0457, "step": 185255 }, { "epoch": 0.47, "learning_rate": 5.3649403593537673e-05, "loss": 1.0431, "step": 185260 }, { "epoch": 0.47, "learning_rate": 5.3648145352055965e-05, "loss": 1.0434, "step": 185265 }, { "epoch": 0.47, "learning_rate": 5.3646887110574263e-05, "loss": 1.0455, "step": 185270 }, { "epoch": 0.47, "learning_rate": 5.3645628869092555e-05, "loss": 1.0475, "step": 185275 }, { "epoch": 0.47, "learning_rate": 5.3644370627610853e-05, "loss": 1.0452, "step": 185280 }, { "epoch": 0.47, "learning_rate": 5.3643112386129145e-05, "loss": 1.0444, "step": 185285 }, { "epoch": 0.47, "learning_rate": 5.364185414464744e-05, "loss": 1.0431, "step": 185290 }, { "epoch": 0.47, "learning_rate": 5.3640595903165735e-05, "loss": 1.0443, "step": 185295 }, { "epoch": 0.47, "learning_rate": 5.363933766168403e-05, "loss": 1.0433, "step": 185300 }, { "epoch": 0.47, "learning_rate": 5.3638079420202325e-05, "loss": 1.0478, "step": 185305 }, { "epoch": 0.47, "learning_rate": 5.3636821178720617e-05, "loss": 1.0444, "step": 185310 }, { "epoch": 0.47, "learning_rate": 5.3635562937238915e-05, "loss": 1.0429, "step": 185315 }, { "epoch": 0.47, "learning_rate": 5.3634304695757207e-05, "loss": 1.0449, "step": 185320 }, { "epoch": 0.47, "learning_rate": 5.3633046454275505e-05, "loss": 1.045, "step": 185325 }, { "epoch": 0.47, "learning_rate": 5.3631788212793796e-05, "loss": 1.0428, "step": 185330 }, { "epoch": 0.47, "learning_rate": 5.3630529971312095e-05, "loss": 1.0578, "step": 185335 }, { "epoch": 0.47, "learning_rate": 5.3629271729830386e-05, "loss": 1.0462, "step": 185340 }, { "epoch": 0.47, "learning_rate": 5.3628013488348685e-05, "loss": 1.0464, "step": 185345 }, { "epoch": 0.47, "learning_rate": 5.3626755246866976e-05, "loss": 1.0426, "step": 185350 }, { "epoch": 0.47, "learning_rate": 5.3625497005385275e-05, "loss": 1.0421, "step": 185355 }, { "epoch": 0.47, "learning_rate": 5.3624238763903566e-05, "loss": 1.0469, "step": 185360 }, { "epoch": 0.47, "learning_rate": 5.3622980522421865e-05, "loss": 1.0457, "step": 185365 }, { "epoch": 0.47, "learning_rate": 5.3621722280940156e-05, "loss": 1.0474, "step": 185370 }, { "epoch": 0.47, "learning_rate": 5.362046403945845e-05, "loss": 1.0451, "step": 185375 }, { "epoch": 0.47, "learning_rate": 5.3619205797976746e-05, "loss": 1.0451, "step": 185380 }, { "epoch": 0.47, "learning_rate": 5.361794755649504e-05, "loss": 1.0458, "step": 185385 }, { "epoch": 0.47, "learning_rate": 5.3616689315013336e-05, "loss": 1.0482, "step": 185390 }, { "epoch": 0.47, "learning_rate": 5.361543107353163e-05, "loss": 1.0431, "step": 185395 }, { "epoch": 0.47, "learning_rate": 5.3614172832049926e-05, "loss": 1.0463, "step": 185400 }, { "epoch": 0.47, "learning_rate": 5.361291459056822e-05, "loss": 1.0498, "step": 185405 }, { "epoch": 0.47, "learning_rate": 5.3611656349086516e-05, "loss": 1.0426, "step": 185410 }, { "epoch": 0.47, "learning_rate": 5.361039810760481e-05, "loss": 1.0482, "step": 185415 }, { "epoch": 0.47, "learning_rate": 5.3609139866123106e-05, "loss": 1.0437, "step": 185420 }, { "epoch": 0.47, "learning_rate": 5.36078816246414e-05, "loss": 1.0434, "step": 185425 }, { "epoch": 0.47, "learning_rate": 5.3606623383159696e-05, "loss": 1.0443, "step": 185430 }, { "epoch": 0.47, "learning_rate": 5.3605365141678e-05, "loss": 1.0471, "step": 185435 }, { "epoch": 0.47, "learning_rate": 5.360410690019629e-05, "loss": 1.0436, "step": 185440 }, { "epoch": 0.47, "learning_rate": 5.360284865871459e-05, "loss": 1.0465, "step": 185445 }, { "epoch": 0.47, "learning_rate": 5.360159041723288e-05, "loss": 1.0441, "step": 185450 }, { "epoch": 0.47, "learning_rate": 5.3600332175751175e-05, "loss": 1.0455, "step": 185455 }, { "epoch": 0.47, "learning_rate": 5.359907393426947e-05, "loss": 1.0442, "step": 185460 }, { "epoch": 0.47, "learning_rate": 5.3597815692787764e-05, "loss": 1.0486, "step": 185465 }, { "epoch": 0.47, "learning_rate": 5.359655745130606e-05, "loss": 1.043, "step": 185470 }, { "epoch": 0.47, "learning_rate": 5.3595299209824354e-05, "loss": 1.0497, "step": 185475 }, { "epoch": 0.47, "learning_rate": 5.359404096834265e-05, "loss": 1.0428, "step": 185480 }, { "epoch": 0.47, "learning_rate": 5.3592782726860944e-05, "loss": 1.0439, "step": 185485 }, { "epoch": 0.47, "learning_rate": 5.359152448537924e-05, "loss": 1.0461, "step": 185490 }, { "epoch": 0.47, "learning_rate": 5.3590266243897534e-05, "loss": 1.0442, "step": 185495 }, { "epoch": 0.47, "learning_rate": 5.358900800241583e-05, "loss": 1.0485, "step": 185500 }, { "epoch": 0.47, "learning_rate": 5.3587749760934124e-05, "loss": 1.0448, "step": 185505 }, { "epoch": 0.47, "learning_rate": 5.358649151945242e-05, "loss": 1.0471, "step": 185510 }, { "epoch": 0.47, "learning_rate": 5.3585233277970714e-05, "loss": 1.0473, "step": 185515 }, { "epoch": 0.47, "learning_rate": 5.3583975036489006e-05, "loss": 1.0452, "step": 185520 }, { "epoch": 0.47, "learning_rate": 5.3582716795007304e-05, "loss": 1.0444, "step": 185525 }, { "epoch": 0.47, "learning_rate": 5.3581458553525596e-05, "loss": 1.0466, "step": 185530 }, { "epoch": 0.47, "learning_rate": 5.3580200312043894e-05, "loss": 1.047, "step": 185535 }, { "epoch": 0.47, "learning_rate": 5.3578942070562186e-05, "loss": 1.0437, "step": 185540 }, { "epoch": 0.47, "learning_rate": 5.3577683829080484e-05, "loss": 1.0454, "step": 185545 }, { "epoch": 0.47, "learning_rate": 5.3576425587598776e-05, "loss": 1.0463, "step": 185550 }, { "epoch": 0.47, "learning_rate": 5.3575167346117074e-05, "loss": 1.0449, "step": 185555 }, { "epoch": 0.47, "learning_rate": 5.3573909104635366e-05, "loss": 1.0454, "step": 185560 }, { "epoch": 0.47, "learning_rate": 5.3572650863153664e-05, "loss": 1.043, "step": 185565 }, { "epoch": 0.47, "learning_rate": 5.3571392621671956e-05, "loss": 1.0445, "step": 185570 }, { "epoch": 0.47, "learning_rate": 5.3570134380190254e-05, "loss": 1.0452, "step": 185575 }, { "epoch": 0.47, "learning_rate": 5.3568876138708546e-05, "loss": 1.045, "step": 185580 }, { "epoch": 0.47, "learning_rate": 5.356761789722684e-05, "loss": 1.048, "step": 185585 }, { "epoch": 0.47, "learning_rate": 5.3566359655745136e-05, "loss": 1.0459, "step": 185590 }, { "epoch": 0.47, "learning_rate": 5.356510141426343e-05, "loss": 1.0471, "step": 185595 }, { "epoch": 0.47, "learning_rate": 5.3563843172781726e-05, "loss": 1.0669, "step": 185600 }, { "epoch": 0.47, "learning_rate": 5.356258493130002e-05, "loss": 1.0457, "step": 185605 }, { "epoch": 0.47, "learning_rate": 5.3561326689818316e-05, "loss": 1.0448, "step": 185610 }, { "epoch": 0.47, "learning_rate": 5.356006844833661e-05, "loss": 1.0468, "step": 185615 }, { "epoch": 0.47, "learning_rate": 5.3558810206854906e-05, "loss": 1.0452, "step": 185620 }, { "epoch": 0.47, "learning_rate": 5.35575519653732e-05, "loss": 1.0442, "step": 185625 }, { "epoch": 0.47, "learning_rate": 5.3556293723891496e-05, "loss": 1.0461, "step": 185630 }, { "epoch": 0.47, "learning_rate": 5.355503548240979e-05, "loss": 1.0418, "step": 185635 }, { "epoch": 0.47, "learning_rate": 5.3553777240928086e-05, "loss": 1.0457, "step": 185640 }, { "epoch": 0.47, "learning_rate": 5.355251899944638e-05, "loss": 1.0461, "step": 185645 }, { "epoch": 0.47, "learning_rate": 5.355126075796467e-05, "loss": 1.0417, "step": 185650 }, { "epoch": 0.47, "learning_rate": 5.355000251648297e-05, "loss": 1.0453, "step": 185655 }, { "epoch": 0.47, "learning_rate": 5.354874427500126e-05, "loss": 1.0465, "step": 185660 }, { "epoch": 0.47, "learning_rate": 5.354748603351956e-05, "loss": 1.0462, "step": 185665 }, { "epoch": 0.47, "learning_rate": 5.354622779203785e-05, "loss": 1.0472, "step": 185670 }, { "epoch": 0.47, "learning_rate": 5.354496955055615e-05, "loss": 1.0455, "step": 185675 }, { "epoch": 0.47, "learning_rate": 5.354371130907444e-05, "loss": 1.0444, "step": 185680 }, { "epoch": 0.47, "learning_rate": 5.354245306759274e-05, "loss": 1.0457, "step": 185685 }, { "epoch": 0.47, "learning_rate": 5.354119482611103e-05, "loss": 1.0464, "step": 185690 }, { "epoch": 0.47, "learning_rate": 5.353993658462933e-05, "loss": 1.0463, "step": 185695 }, { "epoch": 0.47, "learning_rate": 5.353867834314762e-05, "loss": 1.0466, "step": 185700 }, { "epoch": 0.47, "learning_rate": 5.353742010166592e-05, "loss": 1.0475, "step": 185705 }, { "epoch": 0.47, "learning_rate": 5.353616186018421e-05, "loss": 1.0477, "step": 185710 }, { "epoch": 0.47, "learning_rate": 5.35349036187025e-05, "loss": 1.0429, "step": 185715 }, { "epoch": 0.47, "learning_rate": 5.35336453772208e-05, "loss": 1.0459, "step": 185720 }, { "epoch": 0.47, "learning_rate": 5.353238713573909e-05, "loss": 1.0476, "step": 185725 }, { "epoch": 0.47, "learning_rate": 5.353112889425739e-05, "loss": 1.0414, "step": 185730 }, { "epoch": 0.47, "learning_rate": 5.352987065277568e-05, "loss": 1.0461, "step": 185735 }, { "epoch": 0.47, "learning_rate": 5.352861241129398e-05, "loss": 1.0468, "step": 185740 }, { "epoch": 0.47, "learning_rate": 5.352735416981227e-05, "loss": 1.0442, "step": 185745 }, { "epoch": 0.47, "learning_rate": 5.352609592833057e-05, "loss": 1.046, "step": 185750 }, { "epoch": 0.47, "learning_rate": 5.352483768684886e-05, "loss": 1.0463, "step": 185755 }, { "epoch": 0.47, "learning_rate": 5.352357944536716e-05, "loss": 1.0441, "step": 185760 }, { "epoch": 0.47, "learning_rate": 5.352232120388545e-05, "loss": 1.0473, "step": 185765 }, { "epoch": 0.47, "learning_rate": 5.352106296240375e-05, "loss": 1.052, "step": 185770 }, { "epoch": 0.47, "learning_rate": 5.351980472092204e-05, "loss": 1.0431, "step": 185775 }, { "epoch": 0.47, "learning_rate": 5.351854647944033e-05, "loss": 1.0427, "step": 185780 }, { "epoch": 0.47, "learning_rate": 5.351728823795863e-05, "loss": 1.0425, "step": 185785 }, { "epoch": 0.47, "learning_rate": 5.351602999647692e-05, "loss": 1.0435, "step": 185790 }, { "epoch": 0.47, "learning_rate": 5.351477175499522e-05, "loss": 1.0489, "step": 185795 }, { "epoch": 0.47, "learning_rate": 5.351351351351351e-05, "loss": 1.0463, "step": 185800 }, { "epoch": 0.47, "learning_rate": 5.351225527203181e-05, "loss": 1.0438, "step": 185805 }, { "epoch": 0.47, "learning_rate": 5.35109970305501e-05, "loss": 1.0466, "step": 185810 }, { "epoch": 0.47, "learning_rate": 5.35097387890684e-05, "loss": 1.0434, "step": 185815 }, { "epoch": 0.47, "learning_rate": 5.350848054758669e-05, "loss": 1.0445, "step": 185820 }, { "epoch": 0.47, "learning_rate": 5.350722230610499e-05, "loss": 1.0463, "step": 185825 }, { "epoch": 0.47, "learning_rate": 5.350596406462328e-05, "loss": 1.0439, "step": 185830 }, { "epoch": 0.47, "learning_rate": 5.350470582314158e-05, "loss": 1.0442, "step": 185835 }, { "epoch": 0.47, "learning_rate": 5.350344758165987e-05, "loss": 1.045, "step": 185840 }, { "epoch": 0.47, "learning_rate": 5.350218934017816e-05, "loss": 1.043, "step": 185845 }, { "epoch": 0.47, "learning_rate": 5.350093109869646e-05, "loss": 1.0459, "step": 185850 }, { "epoch": 0.47, "learning_rate": 5.349967285721475e-05, "loss": 1.0458, "step": 185855 }, { "epoch": 0.47, "learning_rate": 5.349841461573305e-05, "loss": 1.0448, "step": 185860 }, { "epoch": 0.47, "learning_rate": 5.349715637425134e-05, "loss": 1.0446, "step": 185865 }, { "epoch": 0.47, "learning_rate": 5.349589813276964e-05, "loss": 1.0466, "step": 185870 }, { "epoch": 0.47, "learning_rate": 5.349463989128793e-05, "loss": 1.0429, "step": 185875 }, { "epoch": 0.47, "learning_rate": 5.349338164980623e-05, "loss": 1.0448, "step": 185880 }, { "epoch": 0.47, "learning_rate": 5.349212340832452e-05, "loss": 1.0453, "step": 185885 }, { "epoch": 0.47, "learning_rate": 5.349086516684282e-05, "loss": 1.0657, "step": 185890 }, { "epoch": 0.47, "learning_rate": 5.348960692536111e-05, "loss": 1.0471, "step": 185895 }, { "epoch": 0.47, "learning_rate": 5.3488348683879405e-05, "loss": 1.0459, "step": 185900 }, { "epoch": 0.47, "learning_rate": 5.34870904423977e-05, "loss": 1.0461, "step": 185905 }, { "epoch": 0.47, "learning_rate": 5.3485832200915994e-05, "loss": 1.0466, "step": 185910 }, { "epoch": 0.47, "learning_rate": 5.348457395943429e-05, "loss": 1.0471, "step": 185915 }, { "epoch": 0.47, "learning_rate": 5.3483315717952584e-05, "loss": 1.0434, "step": 185920 }, { "epoch": 0.47, "learning_rate": 5.348205747647088e-05, "loss": 1.0472, "step": 185925 }, { "epoch": 0.47, "learning_rate": 5.3480799234989174e-05, "loss": 1.0459, "step": 185930 }, { "epoch": 0.47, "learning_rate": 5.347954099350748e-05, "loss": 1.092, "step": 185935 }, { "epoch": 0.47, "learning_rate": 5.347828275202578e-05, "loss": 1.0479, "step": 185940 }, { "epoch": 0.47, "learning_rate": 5.347702451054407e-05, "loss": 1.0423, "step": 185945 }, { "epoch": 0.47, "learning_rate": 5.347576626906237e-05, "loss": 1.0466, "step": 185950 }, { "epoch": 0.47, "learning_rate": 5.347450802758066e-05, "loss": 1.0452, "step": 185955 }, { "epoch": 0.47, "learning_rate": 5.347324978609896e-05, "loss": 1.046, "step": 185960 }, { "epoch": 0.47, "learning_rate": 5.347199154461725e-05, "loss": 1.0478, "step": 185965 }, { "epoch": 0.47, "learning_rate": 5.347073330313555e-05, "loss": 1.0449, "step": 185970 }, { "epoch": 0.47, "learning_rate": 5.346947506165384e-05, "loss": 1.0433, "step": 185975 }, { "epoch": 0.47, "learning_rate": 5.346821682017214e-05, "loss": 1.0675, "step": 185980 }, { "epoch": 0.47, "learning_rate": 5.346695857869043e-05, "loss": 1.0461, "step": 185985 }, { "epoch": 0.47, "learning_rate": 5.346570033720872e-05, "loss": 1.0456, "step": 185990 }, { "epoch": 0.47, "learning_rate": 5.346444209572702e-05, "loss": 1.0455, "step": 185995 }, { "epoch": 0.47, "learning_rate": 5.346318385424531e-05, "loss": 1.0442, "step": 186000 }, { "epoch": 0.47, "learning_rate": 5.346192561276361e-05, "loss": 1.0448, "step": 186005 }, { "epoch": 0.47, "learning_rate": 5.34606673712819e-05, "loss": 1.0435, "step": 186010 }, { "epoch": 0.47, "learning_rate": 5.34594091298002e-05, "loss": 1.0445, "step": 186015 }, { "epoch": 0.47, "learning_rate": 5.345815088831849e-05, "loss": 1.0438, "step": 186020 }, { "epoch": 0.47, "learning_rate": 5.345689264683679e-05, "loss": 1.0689, "step": 186025 }, { "epoch": 0.47, "learning_rate": 5.345563440535508e-05, "loss": 1.0455, "step": 186030 }, { "epoch": 0.47, "learning_rate": 5.345437616387338e-05, "loss": 1.043, "step": 186035 }, { "epoch": 0.47, "learning_rate": 5.345311792239167e-05, "loss": 1.045, "step": 186040 }, { "epoch": 0.47, "learning_rate": 5.345185968090997e-05, "loss": 1.0467, "step": 186045 }, { "epoch": 0.47, "learning_rate": 5.345060143942826e-05, "loss": 1.0455, "step": 186050 }, { "epoch": 0.47, "learning_rate": 5.344934319794655e-05, "loss": 1.0455, "step": 186055 }, { "epoch": 0.47, "learning_rate": 5.344808495646485e-05, "loss": 1.0472, "step": 186060 }, { "epoch": 0.47, "learning_rate": 5.344682671498314e-05, "loss": 1.0437, "step": 186065 }, { "epoch": 0.47, "learning_rate": 5.344556847350144e-05, "loss": 1.046, "step": 186070 }, { "epoch": 0.47, "learning_rate": 5.344431023201973e-05, "loss": 1.0444, "step": 186075 }, { "epoch": 0.47, "learning_rate": 5.344305199053803e-05, "loss": 1.0466, "step": 186080 }, { "epoch": 0.47, "learning_rate": 5.344179374905632e-05, "loss": 1.0484, "step": 186085 }, { "epoch": 0.47, "learning_rate": 5.344053550757462e-05, "loss": 1.0461, "step": 186090 }, { "epoch": 0.47, "learning_rate": 5.343927726609291e-05, "loss": 1.0466, "step": 186095 }, { "epoch": 0.47, "learning_rate": 5.343801902461121e-05, "loss": 1.0431, "step": 186100 }, { "epoch": 0.47, "learning_rate": 5.34367607831295e-05, "loss": 1.0468, "step": 186105 }, { "epoch": 0.47, "learning_rate": 5.3435502541647794e-05, "loss": 1.0434, "step": 186110 }, { "epoch": 0.47, "learning_rate": 5.343424430016609e-05, "loss": 1.0442, "step": 186115 }, { "epoch": 0.47, "learning_rate": 5.3432986058684384e-05, "loss": 1.0437, "step": 186120 }, { "epoch": 0.47, "learning_rate": 5.343172781720268e-05, "loss": 1.0424, "step": 186125 }, { "epoch": 0.47, "learning_rate": 5.3430469575720974e-05, "loss": 1.046, "step": 186130 }, { "epoch": 0.47, "learning_rate": 5.342921133423927e-05, "loss": 1.0429, "step": 186135 }, { "epoch": 0.47, "learning_rate": 5.3427953092757564e-05, "loss": 1.0462, "step": 186140 }, { "epoch": 0.47, "learning_rate": 5.342669485127586e-05, "loss": 1.0486, "step": 186145 }, { "epoch": 0.47, "learning_rate": 5.3425436609794154e-05, "loss": 1.0481, "step": 186150 }, { "epoch": 0.47, "learning_rate": 5.342417836831245e-05, "loss": 1.0443, "step": 186155 }, { "epoch": 0.47, "learning_rate": 5.3422920126830744e-05, "loss": 1.0451, "step": 186160 }, { "epoch": 0.47, "learning_rate": 5.342166188534904e-05, "loss": 1.0461, "step": 186165 }, { "epoch": 0.47, "learning_rate": 5.3420403643867334e-05, "loss": 1.0462, "step": 186170 }, { "epoch": 0.47, "learning_rate": 5.3419145402385625e-05, "loss": 1.045, "step": 186175 }, { "epoch": 0.47, "learning_rate": 5.3417887160903924e-05, "loss": 1.0429, "step": 186180 }, { "epoch": 0.47, "learning_rate": 5.3416628919422215e-05, "loss": 1.0437, "step": 186185 }, { "epoch": 0.47, "learning_rate": 5.3415370677940514e-05, "loss": 1.0456, "step": 186190 }, { "epoch": 0.47, "learning_rate": 5.3414112436458805e-05, "loss": 1.0443, "step": 186195 }, { "epoch": 0.47, "learning_rate": 5.3412854194977104e-05, "loss": 1.0419, "step": 186200 }, { "epoch": 0.47, "learning_rate": 5.3411595953495395e-05, "loss": 1.0463, "step": 186205 }, { "epoch": 0.47, "learning_rate": 5.3410337712013694e-05, "loss": 1.0466, "step": 186210 }, { "epoch": 0.47, "learning_rate": 5.3409079470531985e-05, "loss": 1.0446, "step": 186215 }, { "epoch": 0.47, "learning_rate": 5.3407821229050284e-05, "loss": 1.0459, "step": 186220 }, { "epoch": 0.47, "learning_rate": 5.3406562987568575e-05, "loss": 1.0463, "step": 186225 }, { "epoch": 0.47, "learning_rate": 5.3405304746086874e-05, "loss": 1.0463, "step": 186230 }, { "epoch": 0.47, "learning_rate": 5.3404046504605165e-05, "loss": 1.0449, "step": 186235 }, { "epoch": 0.47, "learning_rate": 5.340278826312346e-05, "loss": 1.0724, "step": 186240 }, { "epoch": 0.47, "learning_rate": 5.3401530021641755e-05, "loss": 1.0458, "step": 186245 }, { "epoch": 0.47, "learning_rate": 5.340027178016005e-05, "loss": 1.0454, "step": 186250 }, { "epoch": 0.47, "learning_rate": 5.3399013538678345e-05, "loss": 1.0437, "step": 186255 }, { "epoch": 0.47, "learning_rate": 5.339775529719664e-05, "loss": 1.0432, "step": 186260 }, { "epoch": 0.47, "learning_rate": 5.3396497055714935e-05, "loss": 1.0453, "step": 186265 }, { "epoch": 0.47, "learning_rate": 5.339523881423323e-05, "loss": 1.043, "step": 186270 }, { "epoch": 0.47, "learning_rate": 5.3393980572751525e-05, "loss": 1.0469, "step": 186275 }, { "epoch": 0.47, "learning_rate": 5.339272233126982e-05, "loss": 1.0454, "step": 186280 }, { "epoch": 0.47, "learning_rate": 5.3391464089788115e-05, "loss": 1.0425, "step": 186285 }, { "epoch": 0.47, "learning_rate": 5.3390205848306407e-05, "loss": 1.044, "step": 186290 }, { "epoch": 0.47, "learning_rate": 5.3388947606824705e-05, "loss": 1.0459, "step": 186295 }, { "epoch": 0.47, "learning_rate": 5.3387689365342997e-05, "loss": 1.0474, "step": 186300 }, { "epoch": 0.47, "learning_rate": 5.338643112386129e-05, "loss": 1.0446, "step": 186305 }, { "epoch": 0.47, "learning_rate": 5.3385172882379587e-05, "loss": 1.0469, "step": 186310 }, { "epoch": 0.47, "learning_rate": 5.338391464089788e-05, "loss": 1.0454, "step": 186315 }, { "epoch": 0.47, "learning_rate": 5.3382656399416177e-05, "loss": 1.0878, "step": 186320 }, { "epoch": 0.47, "learning_rate": 5.338139815793447e-05, "loss": 1.0451, "step": 186325 }, { "epoch": 0.47, "learning_rate": 5.3380139916452766e-05, "loss": 1.0613, "step": 186330 }, { "epoch": 0.47, "learning_rate": 5.337888167497106e-05, "loss": 1.0425, "step": 186335 }, { "epoch": 0.47, "learning_rate": 5.3377623433489356e-05, "loss": 1.0437, "step": 186340 }, { "epoch": 0.47, "learning_rate": 5.337636519200765e-05, "loss": 1.045, "step": 186345 }, { "epoch": 0.47, "learning_rate": 5.3375106950525946e-05, "loss": 1.0449, "step": 186350 }, { "epoch": 0.47, "learning_rate": 5.337384870904424e-05, "loss": 1.0673, "step": 186355 }, { "epoch": 0.47, "learning_rate": 5.3372590467562536e-05, "loss": 1.0452, "step": 186360 }, { "epoch": 0.47, "learning_rate": 5.337133222608083e-05, "loss": 1.0471, "step": 186365 }, { "epoch": 0.47, "learning_rate": 5.337007398459912e-05, "loss": 1.0449, "step": 186370 }, { "epoch": 0.47, "learning_rate": 5.336881574311742e-05, "loss": 1.0491, "step": 186375 }, { "epoch": 0.47, "learning_rate": 5.336755750163571e-05, "loss": 1.0418, "step": 186380 }, { "epoch": 0.47, "learning_rate": 5.336629926015401e-05, "loss": 1.0463, "step": 186385 }, { "epoch": 0.47, "learning_rate": 5.33650410186723e-05, "loss": 1.0439, "step": 186390 }, { "epoch": 0.47, "learning_rate": 5.33637827771906e-05, "loss": 1.0452, "step": 186395 }, { "epoch": 0.47, "learning_rate": 5.336252453570889e-05, "loss": 1.0454, "step": 186400 }, { "epoch": 0.47, "learning_rate": 5.336126629422719e-05, "loss": 1.0451, "step": 186405 }, { "epoch": 0.47, "learning_rate": 5.336000805274548e-05, "loss": 1.0449, "step": 186410 }, { "epoch": 0.47, "learning_rate": 5.335874981126378e-05, "loss": 1.0449, "step": 186415 }, { "epoch": 0.47, "learning_rate": 5.335749156978207e-05, "loss": 1.044, "step": 186420 }, { "epoch": 0.47, "learning_rate": 5.335623332830037e-05, "loss": 1.0477, "step": 186425 }, { "epoch": 0.47, "learning_rate": 5.335497508681866e-05, "loss": 1.0441, "step": 186430 }, { "epoch": 0.47, "learning_rate": 5.3353716845336965e-05, "loss": 1.0444, "step": 186435 }, { "epoch": 0.47, "learning_rate": 5.335245860385526e-05, "loss": 1.0439, "step": 186440 }, { "epoch": 0.47, "learning_rate": 5.3351200362373555e-05, "loss": 1.0447, "step": 186445 }, { "epoch": 0.47, "learning_rate": 5.3349942120891846e-05, "loss": 1.046, "step": 186450 }, { "epoch": 0.47, "learning_rate": 5.3348683879410145e-05, "loss": 1.0455, "step": 186455 }, { "epoch": 0.47, "learning_rate": 5.3347425637928436e-05, "loss": 1.0423, "step": 186460 }, { "epoch": 0.47, "learning_rate": 5.3346167396446734e-05, "loss": 1.0436, "step": 186465 }, { "epoch": 0.47, "learning_rate": 5.3344909154965026e-05, "loss": 1.0446, "step": 186470 }, { "epoch": 0.47, "learning_rate": 5.3343650913483324e-05, "loss": 1.0428, "step": 186475 }, { "epoch": 0.47, "learning_rate": 5.3342392672001616e-05, "loss": 1.0468, "step": 186480 }, { "epoch": 0.47, "learning_rate": 5.3341134430519914e-05, "loss": 1.0455, "step": 186485 }, { "epoch": 0.47, "learning_rate": 5.3339876189038206e-05, "loss": 1.0472, "step": 186490 }, { "epoch": 0.47, "learning_rate": 5.3338617947556504e-05, "loss": 1.0448, "step": 186495 }, { "epoch": 0.47, "learning_rate": 5.3337359706074796e-05, "loss": 1.0459, "step": 186500 }, { "epoch": 0.47, "learning_rate": 5.3336101464593094e-05, "loss": 1.0427, "step": 186505 }, { "epoch": 0.47, "learning_rate": 5.3334843223111386e-05, "loss": 1.0448, "step": 186510 }, { "epoch": 0.47, "learning_rate": 5.333358498162968e-05, "loss": 1.046, "step": 186515 }, { "epoch": 0.47, "learning_rate": 5.3332326740147976e-05, "loss": 1.0444, "step": 186520 }, { "epoch": 0.47, "learning_rate": 5.333106849866627e-05, "loss": 1.0452, "step": 186525 }, { "epoch": 0.47, "learning_rate": 5.3329810257184566e-05, "loss": 1.0525, "step": 186530 }, { "epoch": 0.47, "learning_rate": 5.332855201570286e-05, "loss": 1.0458, "step": 186535 }, { "epoch": 0.47, "learning_rate": 5.3327293774221156e-05, "loss": 1.0693, "step": 186540 }, { "epoch": 0.47, "learning_rate": 5.332603553273945e-05, "loss": 1.0434, "step": 186545 }, { "epoch": 0.47, "learning_rate": 5.3324777291257746e-05, "loss": 1.0667, "step": 186550 }, { "epoch": 0.47, "learning_rate": 5.332351904977604e-05, "loss": 1.0435, "step": 186555 }, { "epoch": 0.47, "learning_rate": 5.3322260808294336e-05, "loss": 1.042, "step": 186560 }, { "epoch": 0.47, "learning_rate": 5.332100256681263e-05, "loss": 1.0454, "step": 186565 }, { "epoch": 0.47, "learning_rate": 5.3319744325330926e-05, "loss": 1.0411, "step": 186570 }, { "epoch": 0.47, "learning_rate": 5.331848608384922e-05, "loss": 1.0438, "step": 186575 }, { "epoch": 0.47, "learning_rate": 5.331722784236751e-05, "loss": 1.0426, "step": 186580 }, { "epoch": 0.47, "learning_rate": 5.331596960088581e-05, "loss": 1.0462, "step": 186585 }, { "epoch": 0.47, "learning_rate": 5.33147113594041e-05, "loss": 1.0688, "step": 186590 }, { "epoch": 0.47, "learning_rate": 5.33134531179224e-05, "loss": 1.0447, "step": 186595 }, { "epoch": 0.47, "learning_rate": 5.331219487644069e-05, "loss": 1.0465, "step": 186600 }, { "epoch": 0.47, "learning_rate": 5.331093663495899e-05, "loss": 1.0485, "step": 186605 }, { "epoch": 0.47, "learning_rate": 5.330967839347728e-05, "loss": 1.0485, "step": 186610 }, { "epoch": 0.47, "learning_rate": 5.330842015199558e-05, "loss": 1.0643, "step": 186615 }, { "epoch": 0.47, "learning_rate": 5.330716191051387e-05, "loss": 1.044, "step": 186620 }, { "epoch": 0.47, "learning_rate": 5.330590366903217e-05, "loss": 1.0704, "step": 186625 }, { "epoch": 0.47, "learning_rate": 5.330464542755046e-05, "loss": 1.046, "step": 186630 }, { "epoch": 0.47, "learning_rate": 5.330338718606876e-05, "loss": 1.0454, "step": 186635 }, { "epoch": 0.47, "learning_rate": 5.330212894458705e-05, "loss": 1.0472, "step": 186640 }, { "epoch": 0.47, "learning_rate": 5.330087070310534e-05, "loss": 1.0482, "step": 186645 }, { "epoch": 0.47, "learning_rate": 5.329961246162364e-05, "loss": 1.0441, "step": 186650 }, { "epoch": 0.47, "learning_rate": 5.329835422014193e-05, "loss": 1.0451, "step": 186655 }, { "epoch": 0.47, "learning_rate": 5.329709597866023e-05, "loss": 1.0713, "step": 186660 }, { "epoch": 0.47, "learning_rate": 5.329583773717852e-05, "loss": 1.0443, "step": 186665 }, { "epoch": 0.47, "learning_rate": 5.329457949569682e-05, "loss": 1.0483, "step": 186670 }, { "epoch": 0.47, "learning_rate": 5.329332125421511e-05, "loss": 1.0476, "step": 186675 }, { "epoch": 0.47, "learning_rate": 5.329206301273341e-05, "loss": 1.0468, "step": 186680 }, { "epoch": 0.47, "learning_rate": 5.32908047712517e-05, "loss": 1.0446, "step": 186685 }, { "epoch": 0.47, "learning_rate": 5.328954652977e-05, "loss": 1.0426, "step": 186690 }, { "epoch": 0.47, "learning_rate": 5.328828828828829e-05, "loss": 1.0475, "step": 186695 }, { "epoch": 0.47, "learning_rate": 5.328703004680659e-05, "loss": 1.044, "step": 186700 }, { "epoch": 0.47, "learning_rate": 5.328577180532488e-05, "loss": 1.0479, "step": 186705 }, { "epoch": 0.47, "learning_rate": 5.328451356384317e-05, "loss": 1.0462, "step": 186710 }, { "epoch": 0.47, "learning_rate": 5.328325532236147e-05, "loss": 1.0448, "step": 186715 }, { "epoch": 0.47, "learning_rate": 5.328199708087976e-05, "loss": 1.0433, "step": 186720 }, { "epoch": 0.47, "learning_rate": 5.328073883939806e-05, "loss": 1.0438, "step": 186725 }, { "epoch": 0.47, "learning_rate": 5.327948059791635e-05, "loss": 1.0432, "step": 186730 }, { "epoch": 0.47, "learning_rate": 5.327822235643465e-05, "loss": 1.0451, "step": 186735 }, { "epoch": 0.47, "learning_rate": 5.327696411495294e-05, "loss": 1.0463, "step": 186740 }, { "epoch": 0.47, "learning_rate": 5.327570587347124e-05, "loss": 1.0457, "step": 186745 }, { "epoch": 0.47, "learning_rate": 5.327444763198953e-05, "loss": 1.0456, "step": 186750 }, { "epoch": 0.47, "learning_rate": 5.327318939050783e-05, "loss": 1.0453, "step": 186755 }, { "epoch": 0.47, "learning_rate": 5.327193114902612e-05, "loss": 1.0471, "step": 186760 }, { "epoch": 0.47, "learning_rate": 5.327067290754441e-05, "loss": 1.0443, "step": 186765 }, { "epoch": 0.47, "learning_rate": 5.326941466606271e-05, "loss": 1.0465, "step": 186770 }, { "epoch": 0.47, "learning_rate": 5.3268156424581e-05, "loss": 1.0448, "step": 186775 }, { "epoch": 0.47, "learning_rate": 5.32668981830993e-05, "loss": 1.0451, "step": 186780 }, { "epoch": 0.47, "learning_rate": 5.326563994161759e-05, "loss": 1.0456, "step": 186785 }, { "epoch": 0.47, "learning_rate": 5.326438170013589e-05, "loss": 1.0457, "step": 186790 }, { "epoch": 0.47, "learning_rate": 5.326312345865418e-05, "loss": 1.0448, "step": 186795 }, { "epoch": 0.47, "learning_rate": 5.326186521717248e-05, "loss": 1.0452, "step": 186800 }, { "epoch": 0.47, "learning_rate": 5.326060697569077e-05, "loss": 1.0475, "step": 186805 }, { "epoch": 0.47, "learning_rate": 5.325934873420907e-05, "loss": 1.0473, "step": 186810 }, { "epoch": 0.47, "learning_rate": 5.325809049272736e-05, "loss": 1.0442, "step": 186815 }, { "epoch": 0.47, "learning_rate": 5.325683225124566e-05, "loss": 1.046, "step": 186820 }, { "epoch": 0.47, "learning_rate": 5.325557400976395e-05, "loss": 1.0459, "step": 186825 }, { "epoch": 0.47, "learning_rate": 5.3254315768282245e-05, "loss": 1.0445, "step": 186830 }, { "epoch": 0.47, "learning_rate": 5.325305752680054e-05, "loss": 1.0444, "step": 186835 }, { "epoch": 0.47, "learning_rate": 5.3251799285318835e-05, "loss": 1.0459, "step": 186840 }, { "epoch": 0.47, "learning_rate": 5.325054104383713e-05, "loss": 1.0462, "step": 186845 }, { "epoch": 0.47, "learning_rate": 5.3249282802355425e-05, "loss": 1.0461, "step": 186850 }, { "epoch": 0.47, "learning_rate": 5.324802456087372e-05, "loss": 1.0445, "step": 186855 }, { "epoch": 0.47, "learning_rate": 5.3246766319392015e-05, "loss": 1.0413, "step": 186860 }, { "epoch": 0.47, "learning_rate": 5.324550807791031e-05, "loss": 1.0454, "step": 186865 }, { "epoch": 0.47, "learning_rate": 5.3244249836428605e-05, "loss": 1.0457, "step": 186870 }, { "epoch": 0.47, "learning_rate": 5.32429915949469e-05, "loss": 1.0473, "step": 186875 }, { "epoch": 0.47, "learning_rate": 5.3241733353465195e-05, "loss": 1.0434, "step": 186880 }, { "epoch": 0.47, "learning_rate": 5.324047511198349e-05, "loss": 1.0455, "step": 186885 }, { "epoch": 0.47, "learning_rate": 5.3239216870501785e-05, "loss": 1.0475, "step": 186890 }, { "epoch": 0.47, "learning_rate": 5.3237958629020076e-05, "loss": 1.0467, "step": 186895 }, { "epoch": 0.47, "learning_rate": 5.3236700387538375e-05, "loss": 1.0433, "step": 186900 }, { "epoch": 0.47, "learning_rate": 5.3235442146056666e-05, "loss": 1.0442, "step": 186905 }, { "epoch": 0.47, "learning_rate": 5.3234183904574964e-05, "loss": 1.0436, "step": 186910 }, { "epoch": 0.47, "learning_rate": 5.3232925663093256e-05, "loss": 1.047, "step": 186915 }, { "epoch": 0.47, "learning_rate": 5.3231667421611554e-05, "loss": 1.0453, "step": 186920 }, { "epoch": 0.47, "learning_rate": 5.3230409180129846e-05, "loss": 1.047, "step": 186925 }, { "epoch": 0.47, "learning_rate": 5.3229150938648144e-05, "loss": 1.0423, "step": 186930 }, { "epoch": 0.47, "learning_rate": 5.322789269716645e-05, "loss": 1.0461, "step": 186935 }, { "epoch": 0.47, "learning_rate": 5.322663445568474e-05, "loss": 1.0448, "step": 186940 }, { "epoch": 0.47, "learning_rate": 5.322537621420304e-05, "loss": 1.0417, "step": 186945 }, { "epoch": 0.47, "learning_rate": 5.322411797272133e-05, "loss": 1.0452, "step": 186950 }, { "epoch": 0.47, "learning_rate": 5.322285973123963e-05, "loss": 1.0437, "step": 186955 }, { "epoch": 0.47, "learning_rate": 5.322160148975792e-05, "loss": 1.0493, "step": 186960 }, { "epoch": 0.47, "learning_rate": 5.322034324827622e-05, "loss": 1.0453, "step": 186965 }, { "epoch": 0.47, "learning_rate": 5.321908500679451e-05, "loss": 1.0419, "step": 186970 }, { "epoch": 0.47, "learning_rate": 5.32178267653128e-05, "loss": 1.0458, "step": 186975 }, { "epoch": 0.47, "learning_rate": 5.32165685238311e-05, "loss": 1.0448, "step": 186980 }, { "epoch": 0.47, "learning_rate": 5.321531028234939e-05, "loss": 1.0444, "step": 186985 }, { "epoch": 0.47, "learning_rate": 5.321405204086769e-05, "loss": 1.0432, "step": 186990 }, { "epoch": 0.47, "learning_rate": 5.321279379938598e-05, "loss": 1.0461, "step": 186995 }, { "epoch": 0.47, "learning_rate": 5.321153555790428e-05, "loss": 1.0451, "step": 187000 }, { "epoch": 0.47, "learning_rate": 5.321027731642257e-05, "loss": 1.0464, "step": 187005 }, { "epoch": 0.47, "learning_rate": 5.320901907494087e-05, "loss": 1.0436, "step": 187010 }, { "epoch": 0.47, "learning_rate": 5.320776083345916e-05, "loss": 1.0467, "step": 187015 }, { "epoch": 0.47, "learning_rate": 5.320650259197746e-05, "loss": 1.0426, "step": 187020 }, { "epoch": 0.47, "learning_rate": 5.320524435049575e-05, "loss": 1.0464, "step": 187025 }, { "epoch": 0.47, "learning_rate": 5.320398610901405e-05, "loss": 1.045, "step": 187030 }, { "epoch": 0.47, "learning_rate": 5.320272786753234e-05, "loss": 1.0683, "step": 187035 }, { "epoch": 0.47, "learning_rate": 5.3201469626050634e-05, "loss": 1.0428, "step": 187040 }, { "epoch": 0.47, "learning_rate": 5.320021138456893e-05, "loss": 1.042, "step": 187045 }, { "epoch": 0.47, "learning_rate": 5.3198953143087224e-05, "loss": 1.0436, "step": 187050 }, { "epoch": 0.47, "learning_rate": 5.319769490160552e-05, "loss": 1.0437, "step": 187055 }, { "epoch": 0.47, "learning_rate": 5.3196436660123814e-05, "loss": 1.0429, "step": 187060 }, { "epoch": 0.47, "learning_rate": 5.319517841864211e-05, "loss": 1.0483, "step": 187065 }, { "epoch": 0.47, "learning_rate": 5.3193920177160404e-05, "loss": 1.0451, "step": 187070 }, { "epoch": 0.47, "learning_rate": 5.31926619356787e-05, "loss": 1.0458, "step": 187075 }, { "epoch": 0.47, "learning_rate": 5.3191403694196994e-05, "loss": 1.0431, "step": 187080 }, { "epoch": 0.47, "learning_rate": 5.319014545271529e-05, "loss": 1.0485, "step": 187085 }, { "epoch": 0.47, "learning_rate": 5.3188887211233584e-05, "loss": 1.0429, "step": 187090 }, { "epoch": 0.47, "learning_rate": 5.318762896975188e-05, "loss": 1.0454, "step": 187095 }, { "epoch": 0.47, "learning_rate": 5.3186370728270174e-05, "loss": 1.0472, "step": 187100 }, { "epoch": 0.47, "learning_rate": 5.3185112486788466e-05, "loss": 1.0451, "step": 187105 }, { "epoch": 0.47, "learning_rate": 5.3183854245306764e-05, "loss": 1.0478, "step": 187110 }, { "epoch": 0.47, "learning_rate": 5.3182596003825056e-05, "loss": 1.0442, "step": 187115 }, { "epoch": 0.47, "learning_rate": 5.3181337762343354e-05, "loss": 1.0657, "step": 187120 }, { "epoch": 0.47, "learning_rate": 5.3180079520861645e-05, "loss": 1.0465, "step": 187125 }, { "epoch": 0.47, "learning_rate": 5.3178821279379944e-05, "loss": 1.0428, "step": 187130 }, { "epoch": 0.47, "learning_rate": 5.3177563037898235e-05, "loss": 1.0475, "step": 187135 }, { "epoch": 0.47, "learning_rate": 5.3176304796416534e-05, "loss": 1.0446, "step": 187140 }, { "epoch": 0.47, "learning_rate": 5.3175046554934825e-05, "loss": 1.0463, "step": 187145 }, { "epoch": 0.47, "learning_rate": 5.3173788313453124e-05, "loss": 1.0455, "step": 187150 }, { "epoch": 0.47, "learning_rate": 5.3172530071971415e-05, "loss": 1.0438, "step": 187155 }, { "epoch": 0.47, "learning_rate": 5.3171271830489714e-05, "loss": 1.045, "step": 187160 }, { "epoch": 0.47, "learning_rate": 5.3170013589008005e-05, "loss": 1.0463, "step": 187165 }, { "epoch": 0.47, "learning_rate": 5.31687553475263e-05, "loss": 1.0701, "step": 187170 }, { "epoch": 0.47, "learning_rate": 5.3167497106044595e-05, "loss": 1.0452, "step": 187175 }, { "epoch": 0.47, "learning_rate": 5.316623886456289e-05, "loss": 1.0431, "step": 187180 }, { "epoch": 0.47, "learning_rate": 5.3164980623081185e-05, "loss": 1.0655, "step": 187185 }, { "epoch": 0.47, "learning_rate": 5.316372238159948e-05, "loss": 1.0443, "step": 187190 }, { "epoch": 0.47, "learning_rate": 5.3162464140117775e-05, "loss": 1.0451, "step": 187195 }, { "epoch": 0.47, "learning_rate": 5.316120589863607e-05, "loss": 1.0456, "step": 187200 }, { "epoch": 0.47, "learning_rate": 5.3159947657154365e-05, "loss": 1.0436, "step": 187205 }, { "epoch": 0.47, "learning_rate": 5.315868941567266e-05, "loss": 1.0474, "step": 187210 }, { "epoch": 0.47, "learning_rate": 5.3157431174190955e-05, "loss": 1.0467, "step": 187215 }, { "epoch": 0.47, "learning_rate": 5.315617293270925e-05, "loss": 1.0424, "step": 187220 }, { "epoch": 0.47, "learning_rate": 5.3154914691227545e-05, "loss": 1.0457, "step": 187225 }, { "epoch": 0.47, "learning_rate": 5.315365644974584e-05, "loss": 1.0446, "step": 187230 }, { "epoch": 0.47, "learning_rate": 5.315239820826413e-05, "loss": 1.0473, "step": 187235 }, { "epoch": 0.47, "learning_rate": 5.315113996678243e-05, "loss": 1.0459, "step": 187240 }, { "epoch": 0.47, "learning_rate": 5.314988172530072e-05, "loss": 1.0431, "step": 187245 }, { "epoch": 0.47, "learning_rate": 5.314862348381902e-05, "loss": 1.0443, "step": 187250 }, { "epoch": 0.47, "learning_rate": 5.314736524233731e-05, "loss": 1.0477, "step": 187255 }, { "epoch": 0.47, "learning_rate": 5.314610700085561e-05, "loss": 1.0447, "step": 187260 }, { "epoch": 0.47, "learning_rate": 5.31448487593739e-05, "loss": 1.0468, "step": 187265 }, { "epoch": 0.47, "learning_rate": 5.31435905178922e-05, "loss": 1.0473, "step": 187270 }, { "epoch": 0.47, "learning_rate": 5.314233227641049e-05, "loss": 1.0437, "step": 187275 }, { "epoch": 0.47, "learning_rate": 5.314107403492879e-05, "loss": 1.0431, "step": 187280 }, { "epoch": 0.47, "learning_rate": 5.313981579344708e-05, "loss": 1.0665, "step": 187285 }, { "epoch": 0.47, "learning_rate": 5.313855755196538e-05, "loss": 1.043, "step": 187290 }, { "epoch": 0.47, "learning_rate": 5.313729931048367e-05, "loss": 1.0441, "step": 187295 }, { "epoch": 0.47, "learning_rate": 5.313604106900196e-05, "loss": 1.0442, "step": 187300 }, { "epoch": 0.47, "learning_rate": 5.313478282752026e-05, "loss": 1.045, "step": 187305 }, { "epoch": 0.47, "learning_rate": 5.313352458603855e-05, "loss": 1.0435, "step": 187310 }, { "epoch": 0.47, "learning_rate": 5.313226634455685e-05, "loss": 1.0453, "step": 187315 }, { "epoch": 0.47, "learning_rate": 5.313100810307514e-05, "loss": 1.0458, "step": 187320 }, { "epoch": 0.47, "learning_rate": 5.312974986159344e-05, "loss": 1.0455, "step": 187325 }, { "epoch": 0.47, "learning_rate": 5.312849162011173e-05, "loss": 1.0464, "step": 187330 }, { "epoch": 0.47, "learning_rate": 5.312723337863003e-05, "loss": 1.0468, "step": 187335 }, { "epoch": 0.47, "learning_rate": 5.312597513714832e-05, "loss": 1.0441, "step": 187340 }, { "epoch": 0.47, "learning_rate": 5.312471689566662e-05, "loss": 1.0469, "step": 187345 }, { "epoch": 0.47, "learning_rate": 5.312345865418491e-05, "loss": 1.0581, "step": 187350 }, { "epoch": 0.47, "learning_rate": 5.312220041270321e-05, "loss": 1.0465, "step": 187355 }, { "epoch": 0.47, "learning_rate": 5.31209421712215e-05, "loss": 1.045, "step": 187360 }, { "epoch": 0.47, "learning_rate": 5.311968392973979e-05, "loss": 1.0476, "step": 187365 }, { "epoch": 0.47, "learning_rate": 5.311842568825809e-05, "loss": 1.0468, "step": 187370 }, { "epoch": 0.47, "learning_rate": 5.311716744677638e-05, "loss": 1.0411, "step": 187375 }, { "epoch": 0.47, "learning_rate": 5.311590920529468e-05, "loss": 1.045, "step": 187380 }, { "epoch": 0.47, "learning_rate": 5.311465096381297e-05, "loss": 1.0467, "step": 187385 }, { "epoch": 0.47, "learning_rate": 5.311339272233127e-05, "loss": 1.0475, "step": 187390 }, { "epoch": 0.47, "learning_rate": 5.311213448084956e-05, "loss": 1.0431, "step": 187395 }, { "epoch": 0.47, "learning_rate": 5.311087623936786e-05, "loss": 1.0511, "step": 187400 }, { "epoch": 0.47, "learning_rate": 5.310961799788615e-05, "loss": 1.0455, "step": 187405 }, { "epoch": 0.47, "learning_rate": 5.310835975640445e-05, "loss": 1.0491, "step": 187410 }, { "epoch": 0.47, "learning_rate": 5.310710151492274e-05, "loss": 1.0474, "step": 187415 }, { "epoch": 0.47, "learning_rate": 5.310584327344103e-05, "loss": 1.0452, "step": 187420 }, { "epoch": 0.47, "learning_rate": 5.310458503195933e-05, "loss": 1.0456, "step": 187425 }, { "epoch": 0.47, "learning_rate": 5.310332679047762e-05, "loss": 1.044, "step": 187430 }, { "epoch": 0.47, "learning_rate": 5.3102068548995935e-05, "loss": 1.059, "step": 187435 }, { "epoch": 0.47, "learning_rate": 5.3100810307514226e-05, "loss": 1.0443, "step": 187440 }, { "epoch": 0.47, "learning_rate": 5.309955206603252e-05, "loss": 1.0467, "step": 187445 }, { "epoch": 0.47, "learning_rate": 5.3098293824550816e-05, "loss": 1.0914, "step": 187450 }, { "epoch": 0.47, "learning_rate": 5.309703558306911e-05, "loss": 1.0678, "step": 187455 }, { "epoch": 0.47, "learning_rate": 5.3095777341587406e-05, "loss": 1.0449, "step": 187460 }, { "epoch": 0.47, "learning_rate": 5.30945191001057e-05, "loss": 1.0411, "step": 187465 }, { "epoch": 0.47, "learning_rate": 5.3093260858623996e-05, "loss": 1.0471, "step": 187470 }, { "epoch": 0.47, "learning_rate": 5.309200261714229e-05, "loss": 1.045, "step": 187475 }, { "epoch": 0.47, "learning_rate": 5.3090744375660586e-05, "loss": 1.0447, "step": 187480 }, { "epoch": 0.47, "learning_rate": 5.308948613417888e-05, "loss": 1.0442, "step": 187485 }, { "epoch": 0.47, "learning_rate": 5.3088227892697176e-05, "loss": 1.0453, "step": 187490 }, { "epoch": 0.47, "learning_rate": 5.308696965121547e-05, "loss": 1.0475, "step": 187495 }, { "epoch": 0.47, "learning_rate": 5.3085711409733766e-05, "loss": 1.0458, "step": 187500 }, { "epoch": 0.47, "learning_rate": 5.308445316825206e-05, "loss": 1.0456, "step": 187505 }, { "epoch": 0.47, "learning_rate": 5.308319492677035e-05, "loss": 1.065, "step": 187510 }, { "epoch": 0.47, "learning_rate": 5.308193668528865e-05, "loss": 1.0476, "step": 187515 }, { "epoch": 0.47, "learning_rate": 5.308067844380694e-05, "loss": 1.0456, "step": 187520 }, { "epoch": 0.47, "learning_rate": 5.307942020232524e-05, "loss": 1.0472, "step": 187525 }, { "epoch": 0.47, "learning_rate": 5.307816196084353e-05, "loss": 1.045, "step": 187530 }, { "epoch": 0.47, "learning_rate": 5.307690371936183e-05, "loss": 1.0455, "step": 187535 }, { "epoch": 0.47, "learning_rate": 5.307564547788012e-05, "loss": 1.0467, "step": 187540 }, { "epoch": 0.47, "learning_rate": 5.307438723639842e-05, "loss": 1.046, "step": 187545 }, { "epoch": 0.47, "learning_rate": 5.307312899491671e-05, "loss": 1.0452, "step": 187550 }, { "epoch": 0.47, "learning_rate": 5.307187075343501e-05, "loss": 1.0443, "step": 187555 }, { "epoch": 0.47, "learning_rate": 5.30706125119533e-05, "loss": 1.0451, "step": 187560 }, { "epoch": 0.47, "learning_rate": 5.306935427047159e-05, "loss": 1.061, "step": 187565 }, { "epoch": 0.47, "learning_rate": 5.306809602898989e-05, "loss": 1.0419, "step": 187570 }, { "epoch": 0.47, "learning_rate": 5.306683778750818e-05, "loss": 1.0435, "step": 187575 }, { "epoch": 0.47, "learning_rate": 5.306557954602648e-05, "loss": 1.0569, "step": 187580 }, { "epoch": 0.47, "learning_rate": 5.306432130454477e-05, "loss": 1.0467, "step": 187585 }, { "epoch": 0.47, "learning_rate": 5.306306306306307e-05, "loss": 1.0469, "step": 187590 }, { "epoch": 0.47, "learning_rate": 5.306180482158136e-05, "loss": 1.0695, "step": 187595 }, { "epoch": 0.47, "learning_rate": 5.306054658009966e-05, "loss": 1.0472, "step": 187600 }, { "epoch": 0.47, "learning_rate": 5.305928833861795e-05, "loss": 1.0467, "step": 187605 }, { "epoch": 0.47, "learning_rate": 5.305803009713625e-05, "loss": 1.0454, "step": 187610 }, { "epoch": 0.47, "learning_rate": 5.305677185565454e-05, "loss": 1.0484, "step": 187615 }, { "epoch": 0.47, "learning_rate": 5.305551361417284e-05, "loss": 1.0462, "step": 187620 }, { "epoch": 0.47, "learning_rate": 5.305425537269113e-05, "loss": 1.0428, "step": 187625 }, { "epoch": 0.47, "learning_rate": 5.305299713120942e-05, "loss": 1.0462, "step": 187630 }, { "epoch": 0.47, "learning_rate": 5.305173888972772e-05, "loss": 1.0426, "step": 187635 }, { "epoch": 0.47, "learning_rate": 5.305048064824601e-05, "loss": 1.0424, "step": 187640 }, { "epoch": 0.47, "learning_rate": 5.304922240676431e-05, "loss": 1.0471, "step": 187645 }, { "epoch": 0.47, "learning_rate": 5.30479641652826e-05, "loss": 1.0668, "step": 187650 }, { "epoch": 0.47, "learning_rate": 5.30467059238009e-05, "loss": 1.0459, "step": 187655 }, { "epoch": 0.47, "learning_rate": 5.304544768231919e-05, "loss": 1.0409, "step": 187660 }, { "epoch": 0.47, "learning_rate": 5.304418944083749e-05, "loss": 1.0449, "step": 187665 }, { "epoch": 0.47, "learning_rate": 5.304293119935578e-05, "loss": 1.0447, "step": 187670 }, { "epoch": 0.47, "learning_rate": 5.304167295787408e-05, "loss": 1.0445, "step": 187675 }, { "epoch": 0.47, "learning_rate": 5.304041471639237e-05, "loss": 1.0464, "step": 187680 }, { "epoch": 0.47, "learning_rate": 5.303915647491067e-05, "loss": 1.0475, "step": 187685 }, { "epoch": 0.47, "learning_rate": 5.303789823342896e-05, "loss": 1.0456, "step": 187690 }, { "epoch": 0.47, "learning_rate": 5.3036639991947254e-05, "loss": 1.0446, "step": 187695 }, { "epoch": 0.47, "learning_rate": 5.303538175046555e-05, "loss": 1.0445, "step": 187700 }, { "epoch": 0.47, "learning_rate": 5.3034123508983843e-05, "loss": 1.0474, "step": 187705 }, { "epoch": 0.47, "learning_rate": 5.303286526750214e-05, "loss": 1.0445, "step": 187710 }, { "epoch": 0.47, "learning_rate": 5.3031607026020433e-05, "loss": 1.0469, "step": 187715 }, { "epoch": 0.47, "learning_rate": 5.303034878453873e-05, "loss": 1.0482, "step": 187720 }, { "epoch": 0.47, "learning_rate": 5.3029090543057023e-05, "loss": 1.0477, "step": 187725 }, { "epoch": 0.47, "learning_rate": 5.302783230157532e-05, "loss": 1.0449, "step": 187730 }, { "epoch": 0.47, "learning_rate": 5.3026574060093613e-05, "loss": 1.0442, "step": 187735 }, { "epoch": 0.47, "learning_rate": 5.302531581861191e-05, "loss": 1.0493, "step": 187740 }, { "epoch": 0.47, "learning_rate": 5.30240575771302e-05, "loss": 1.0448, "step": 187745 }, { "epoch": 0.47, "learning_rate": 5.30227993356485e-05, "loss": 1.0469, "step": 187750 }, { "epoch": 0.47, "learning_rate": 5.302154109416679e-05, "loss": 1.0437, "step": 187755 }, { "epoch": 0.47, "learning_rate": 5.3020282852685085e-05, "loss": 1.048, "step": 187760 }, { "epoch": 0.47, "learning_rate": 5.301902461120338e-05, "loss": 1.0495, "step": 187765 }, { "epoch": 0.47, "learning_rate": 5.3017766369721675e-05, "loss": 1.044, "step": 187770 }, { "epoch": 0.47, "learning_rate": 5.301650812823997e-05, "loss": 1.0468, "step": 187775 }, { "epoch": 0.47, "learning_rate": 5.3015249886758265e-05, "loss": 1.0468, "step": 187780 }, { "epoch": 0.47, "learning_rate": 5.301399164527656e-05, "loss": 1.0437, "step": 187785 }, { "epoch": 0.47, "learning_rate": 5.3012733403794855e-05, "loss": 1.0448, "step": 187790 }, { "epoch": 0.47, "learning_rate": 5.301147516231315e-05, "loss": 1.04, "step": 187795 }, { "epoch": 0.47, "learning_rate": 5.3010216920831445e-05, "loss": 1.0453, "step": 187800 }, { "epoch": 0.47, "learning_rate": 5.300895867934974e-05, "loss": 1.0451, "step": 187805 }, { "epoch": 0.47, "learning_rate": 5.3007700437868035e-05, "loss": 1.0471, "step": 187810 }, { "epoch": 0.47, "learning_rate": 5.300644219638633e-05, "loss": 1.0471, "step": 187815 }, { "epoch": 0.47, "learning_rate": 5.3005183954904625e-05, "loss": 1.0467, "step": 187820 }, { "epoch": 0.47, "learning_rate": 5.3003925713422916e-05, "loss": 1.045, "step": 187825 }, { "epoch": 0.47, "learning_rate": 5.3002667471941215e-05, "loss": 1.0442, "step": 187830 }, { "epoch": 0.47, "learning_rate": 5.3001409230459506e-05, "loss": 1.0434, "step": 187835 }, { "epoch": 0.47, "learning_rate": 5.3000150988977805e-05, "loss": 1.0579, "step": 187840 }, { "epoch": 0.47, "learning_rate": 5.2998892747496096e-05, "loss": 1.0465, "step": 187845 }, { "epoch": 0.47, "learning_rate": 5.2997634506014395e-05, "loss": 1.0413, "step": 187850 }, { "epoch": 0.47, "learning_rate": 5.2996376264532686e-05, "loss": 1.0446, "step": 187855 }, { "epoch": 0.47, "learning_rate": 5.2995118023050985e-05, "loss": 1.0415, "step": 187860 }, { "epoch": 0.47, "learning_rate": 5.2993859781569276e-05, "loss": 1.042, "step": 187865 }, { "epoch": 0.47, "learning_rate": 5.2992601540087575e-05, "loss": 1.043, "step": 187870 }, { "epoch": 0.47, "learning_rate": 5.2991343298605866e-05, "loss": 1.0448, "step": 187875 }, { "epoch": 0.47, "learning_rate": 5.2990085057124165e-05, "loss": 1.045, "step": 187880 }, { "epoch": 0.47, "learning_rate": 5.2988826815642456e-05, "loss": 1.0457, "step": 187885 }, { "epoch": 0.47, "learning_rate": 5.298756857416075e-05, "loss": 1.0454, "step": 187890 }, { "epoch": 0.47, "learning_rate": 5.2986310332679046e-05, "loss": 1.0449, "step": 187895 }, { "epoch": 0.47, "learning_rate": 5.298505209119734e-05, "loss": 1.0438, "step": 187900 }, { "epoch": 0.47, "learning_rate": 5.2983793849715636e-05, "loss": 1.05, "step": 187905 }, { "epoch": 0.47, "learning_rate": 5.298253560823393e-05, "loss": 1.0478, "step": 187910 }, { "epoch": 0.47, "learning_rate": 5.2981277366752226e-05, "loss": 1.0487, "step": 187915 }, { "epoch": 0.47, "learning_rate": 5.298001912527052e-05, "loss": 1.0465, "step": 187920 }, { "epoch": 0.47, "learning_rate": 5.2978760883788816e-05, "loss": 1.0429, "step": 187925 }, { "epoch": 0.47, "learning_rate": 5.297750264230711e-05, "loss": 1.0449, "step": 187930 }, { "epoch": 0.47, "learning_rate": 5.297624440082541e-05, "loss": 1.0465, "step": 187935 }, { "epoch": 0.47, "learning_rate": 5.297498615934371e-05, "loss": 1.0644, "step": 187940 }, { "epoch": 0.47, "learning_rate": 5.2973727917862e-05, "loss": 1.0431, "step": 187945 }, { "epoch": 0.47, "learning_rate": 5.29724696763803e-05, "loss": 1.0446, "step": 187950 }, { "epoch": 0.47, "learning_rate": 5.297121143489859e-05, "loss": 1.0443, "step": 187955 }, { "epoch": 0.47, "learning_rate": 5.296995319341689e-05, "loss": 1.0476, "step": 187960 }, { "epoch": 0.47, "learning_rate": 5.296869495193518e-05, "loss": 1.0472, "step": 187965 }, { "epoch": 0.47, "learning_rate": 5.2967436710453474e-05, "loss": 1.0641, "step": 187970 }, { "epoch": 0.47, "learning_rate": 5.296617846897177e-05, "loss": 1.0433, "step": 187975 }, { "epoch": 0.47, "learning_rate": 5.2964920227490064e-05, "loss": 1.0451, "step": 187980 }, { "epoch": 0.47, "learning_rate": 5.296366198600836e-05, "loss": 1.0428, "step": 187985 }, { "epoch": 0.47, "learning_rate": 5.2962403744526654e-05, "loss": 1.0441, "step": 187990 }, { "epoch": 0.47, "learning_rate": 5.296114550304495e-05, "loss": 1.0445, "step": 187995 }, { "epoch": 0.47, "learning_rate": 5.2959887261563244e-05, "loss": 1.0436, "step": 188000 }, { "epoch": 0.47, "learning_rate": 5.295862902008154e-05, "loss": 1.0458, "step": 188005 }, { "epoch": 0.47, "learning_rate": 5.2957370778599834e-05, "loss": 1.0439, "step": 188010 }, { "epoch": 0.47, "learning_rate": 5.295611253711813e-05, "loss": 1.0461, "step": 188015 }, { "epoch": 0.47, "learning_rate": 5.2954854295636424e-05, "loss": 1.0448, "step": 188020 }, { "epoch": 0.47, "learning_rate": 5.295359605415472e-05, "loss": 1.044, "step": 188025 }, { "epoch": 0.47, "learning_rate": 5.2952337812673014e-05, "loss": 1.0426, "step": 188030 }, { "epoch": 0.47, "learning_rate": 5.2951079571191306e-05, "loss": 1.0435, "step": 188035 }, { "epoch": 0.47, "learning_rate": 5.2949821329709604e-05, "loss": 1.0457, "step": 188040 }, { "epoch": 0.47, "learning_rate": 5.2948563088227896e-05, "loss": 1.0473, "step": 188045 }, { "epoch": 0.47, "learning_rate": 5.2947304846746194e-05, "loss": 1.044, "step": 188050 }, { "epoch": 0.47, "learning_rate": 5.2946046605264486e-05, "loss": 1.0588, "step": 188055 }, { "epoch": 0.47, "learning_rate": 5.2944788363782784e-05, "loss": 1.0437, "step": 188060 }, { "epoch": 0.47, "learning_rate": 5.2943530122301076e-05, "loss": 1.0444, "step": 188065 }, { "epoch": 0.47, "learning_rate": 5.2942271880819374e-05, "loss": 1.0453, "step": 188070 }, { "epoch": 0.47, "learning_rate": 5.2941013639337666e-05, "loss": 1.0439, "step": 188075 }, { "epoch": 0.47, "learning_rate": 5.2939755397855964e-05, "loss": 1.0452, "step": 188080 }, { "epoch": 0.47, "learning_rate": 5.2938497156374256e-05, "loss": 1.0459, "step": 188085 }, { "epoch": 0.47, "learning_rate": 5.2937238914892554e-05, "loss": 1.0475, "step": 188090 }, { "epoch": 0.47, "learning_rate": 5.2935980673410846e-05, "loss": 1.0482, "step": 188095 }, { "epoch": 0.47, "learning_rate": 5.293472243192914e-05, "loss": 1.0419, "step": 188100 }, { "epoch": 0.47, "learning_rate": 5.2933464190447436e-05, "loss": 1.0443, "step": 188105 }, { "epoch": 0.47, "learning_rate": 5.293220594896573e-05, "loss": 1.0458, "step": 188110 }, { "epoch": 0.47, "learning_rate": 5.2930947707484026e-05, "loss": 1.0467, "step": 188115 }, { "epoch": 0.47, "learning_rate": 5.292968946600232e-05, "loss": 1.0442, "step": 188120 }, { "epoch": 0.47, "learning_rate": 5.2928431224520615e-05, "loss": 1.0453, "step": 188125 }, { "epoch": 0.47, "learning_rate": 5.292717298303891e-05, "loss": 1.0416, "step": 188130 }, { "epoch": 0.47, "learning_rate": 5.2925914741557205e-05, "loss": 1.0454, "step": 188135 }, { "epoch": 0.47, "learning_rate": 5.29246565000755e-05, "loss": 1.0456, "step": 188140 }, { "epoch": 0.47, "learning_rate": 5.2923398258593795e-05, "loss": 1.0441, "step": 188145 }, { "epoch": 0.47, "learning_rate": 5.292214001711209e-05, "loss": 1.0425, "step": 188150 }, { "epoch": 0.47, "learning_rate": 5.2920881775630385e-05, "loss": 1.0469, "step": 188155 }, { "epoch": 0.47, "learning_rate": 5.291962353414868e-05, "loss": 1.048, "step": 188160 }, { "epoch": 0.47, "learning_rate": 5.291836529266697e-05, "loss": 1.0424, "step": 188165 }, { "epoch": 0.47, "learning_rate": 5.291710705118527e-05, "loss": 1.0426, "step": 188170 }, { "epoch": 0.47, "learning_rate": 5.291584880970356e-05, "loss": 1.0444, "step": 188175 }, { "epoch": 0.47, "learning_rate": 5.291459056822186e-05, "loss": 1.0439, "step": 188180 }, { "epoch": 0.47, "learning_rate": 5.291333232674015e-05, "loss": 1.0439, "step": 188185 }, { "epoch": 0.47, "learning_rate": 5.291207408525845e-05, "loss": 1.045, "step": 188190 }, { "epoch": 0.47, "learning_rate": 5.291081584377674e-05, "loss": 1.0442, "step": 188195 }, { "epoch": 0.47, "learning_rate": 5.290955760229504e-05, "loss": 1.0462, "step": 188200 }, { "epoch": 0.47, "learning_rate": 5.290829936081333e-05, "loss": 1.0455, "step": 188205 }, { "epoch": 0.47, "learning_rate": 5.290704111933163e-05, "loss": 1.0427, "step": 188210 }, { "epoch": 0.47, "learning_rate": 5.290578287784992e-05, "loss": 1.0446, "step": 188215 }, { "epoch": 0.47, "learning_rate": 5.290452463636821e-05, "loss": 1.0496, "step": 188220 }, { "epoch": 0.47, "learning_rate": 5.290326639488651e-05, "loss": 1.0431, "step": 188225 }, { "epoch": 0.47, "learning_rate": 5.29020081534048e-05, "loss": 1.0579, "step": 188230 }, { "epoch": 0.47, "learning_rate": 5.29007499119231e-05, "loss": 1.0439, "step": 188235 }, { "epoch": 0.47, "learning_rate": 5.289949167044139e-05, "loss": 1.0451, "step": 188240 }, { "epoch": 0.47, "learning_rate": 5.289823342895969e-05, "loss": 1.0451, "step": 188245 }, { "epoch": 0.47, "learning_rate": 5.289697518747798e-05, "loss": 1.0437, "step": 188250 }, { "epoch": 0.47, "learning_rate": 5.289571694599628e-05, "loss": 1.0466, "step": 188255 }, { "epoch": 0.47, "learning_rate": 5.289445870451457e-05, "loss": 1.0465, "step": 188260 }, { "epoch": 0.47, "learning_rate": 5.289320046303287e-05, "loss": 1.0443, "step": 188265 }, { "epoch": 0.47, "learning_rate": 5.289194222155116e-05, "loss": 1.0451, "step": 188270 }, { "epoch": 0.47, "learning_rate": 5.289068398006946e-05, "loss": 1.0461, "step": 188275 }, { "epoch": 0.47, "learning_rate": 5.288942573858775e-05, "loss": 1.0468, "step": 188280 }, { "epoch": 0.47, "learning_rate": 5.288816749710604e-05, "loss": 1.0451, "step": 188285 }, { "epoch": 0.47, "learning_rate": 5.288690925562434e-05, "loss": 1.047, "step": 188290 }, { "epoch": 0.47, "learning_rate": 5.288565101414263e-05, "loss": 1.0484, "step": 188295 }, { "epoch": 0.47, "learning_rate": 5.288439277266093e-05, "loss": 1.0463, "step": 188300 }, { "epoch": 0.47, "learning_rate": 5.288313453117922e-05, "loss": 1.0452, "step": 188305 }, { "epoch": 0.47, "learning_rate": 5.288187628969752e-05, "loss": 1.0417, "step": 188310 }, { "epoch": 0.47, "learning_rate": 5.288061804821581e-05, "loss": 1.0424, "step": 188315 }, { "epoch": 0.47, "learning_rate": 5.287935980673411e-05, "loss": 1.0495, "step": 188320 }, { "epoch": 0.47, "learning_rate": 5.28781015652524e-05, "loss": 1.0451, "step": 188325 }, { "epoch": 0.47, "learning_rate": 5.28768433237707e-05, "loss": 1.0425, "step": 188330 }, { "epoch": 0.47, "learning_rate": 5.287558508228899e-05, "loss": 1.0679, "step": 188335 }, { "epoch": 0.47, "learning_rate": 5.287432684080729e-05, "loss": 1.0452, "step": 188340 }, { "epoch": 0.47, "learning_rate": 5.287306859932558e-05, "loss": 1.0447, "step": 188345 }, { "epoch": 0.47, "learning_rate": 5.287181035784387e-05, "loss": 1.0422, "step": 188350 }, { "epoch": 0.47, "learning_rate": 5.287055211636217e-05, "loss": 1.0668, "step": 188355 }, { "epoch": 0.47, "learning_rate": 5.286929387488046e-05, "loss": 1.0459, "step": 188360 }, { "epoch": 0.47, "learning_rate": 5.2868287281695106e-05, "loss": 1.0452, "step": 188365 }, { "epoch": 0.47, "learning_rate": 5.28670290402134e-05, "loss": 1.0467, "step": 188370 }, { "epoch": 0.47, "learning_rate": 5.2865770798731696e-05, "loss": 1.0473, "step": 188375 }, { "epoch": 0.47, "learning_rate": 5.286451255724999e-05, "loss": 1.0477, "step": 188380 }, { "epoch": 0.47, "learning_rate": 5.2863254315768286e-05, "loss": 1.0432, "step": 188385 }, { "epoch": 0.47, "learning_rate": 5.286199607428658e-05, "loss": 1.0446, "step": 188390 }, { "epoch": 0.47, "learning_rate": 5.2860737832804876e-05, "loss": 1.0442, "step": 188395 }, { "epoch": 0.47, "learning_rate": 5.285947959132317e-05, "loss": 1.0449, "step": 188400 }, { "epoch": 0.47, "learning_rate": 5.2858221349841466e-05, "loss": 1.0461, "step": 188405 }, { "epoch": 0.47, "learning_rate": 5.285696310835976e-05, "loss": 1.0471, "step": 188410 }, { "epoch": 0.47, "learning_rate": 5.2855704866878055e-05, "loss": 1.0437, "step": 188415 }, { "epoch": 0.47, "learning_rate": 5.285444662539635e-05, "loss": 1.0448, "step": 188420 }, { "epoch": 0.47, "learning_rate": 5.2853188383914645e-05, "loss": 1.0462, "step": 188425 }, { "epoch": 0.47, "learning_rate": 5.285193014243294e-05, "loss": 1.044, "step": 188430 }, { "epoch": 0.47, "learning_rate": 5.285067190095123e-05, "loss": 1.0725, "step": 188435 }, { "epoch": 0.47, "learning_rate": 5.284941365946953e-05, "loss": 1.0468, "step": 188440 }, { "epoch": 0.47, "learning_rate": 5.284815541798782e-05, "loss": 1.046, "step": 188445 }, { "epoch": 0.47, "learning_rate": 5.284689717650612e-05, "loss": 1.0424, "step": 188450 }, { "epoch": 0.47, "learning_rate": 5.284563893502441e-05, "loss": 1.0462, "step": 188455 }, { "epoch": 0.47, "learning_rate": 5.284438069354271e-05, "loss": 1.0417, "step": 188460 }, { "epoch": 0.47, "learning_rate": 5.2843122452061e-05, "loss": 1.0457, "step": 188465 }, { "epoch": 0.47, "learning_rate": 5.28418642105793e-05, "loss": 1.0469, "step": 188470 }, { "epoch": 0.47, "learning_rate": 5.284060596909759e-05, "loss": 1.0461, "step": 188475 }, { "epoch": 0.47, "learning_rate": 5.283934772761589e-05, "loss": 1.0474, "step": 188480 }, { "epoch": 0.47, "learning_rate": 5.283808948613418e-05, "loss": 1.0437, "step": 188485 }, { "epoch": 0.47, "learning_rate": 5.283683124465248e-05, "loss": 1.0464, "step": 188490 }, { "epoch": 0.47, "learning_rate": 5.283557300317077e-05, "loss": 1.0443, "step": 188495 }, { "epoch": 0.47, "learning_rate": 5.283431476168906e-05, "loss": 1.0471, "step": 188500 }, { "epoch": 0.47, "learning_rate": 5.283305652020736e-05, "loss": 1.0441, "step": 188505 }, { "epoch": 0.47, "learning_rate": 5.283179827872565e-05, "loss": 1.0449, "step": 188510 }, { "epoch": 0.47, "learning_rate": 5.283054003724395e-05, "loss": 1.0427, "step": 188515 }, { "epoch": 0.47, "learning_rate": 5.282928179576224e-05, "loss": 1.0434, "step": 188520 }, { "epoch": 0.47, "learning_rate": 5.282802355428054e-05, "loss": 1.0434, "step": 188525 }, { "epoch": 0.47, "learning_rate": 5.282676531279883e-05, "loss": 1.0449, "step": 188530 }, { "epoch": 0.47, "learning_rate": 5.282550707131713e-05, "loss": 1.0417, "step": 188535 }, { "epoch": 0.47, "learning_rate": 5.282424882983542e-05, "loss": 1.0671, "step": 188540 }, { "epoch": 0.47, "learning_rate": 5.282299058835372e-05, "loss": 1.0455, "step": 188545 }, { "epoch": 0.47, "learning_rate": 5.282173234687201e-05, "loss": 1.0456, "step": 188550 }, { "epoch": 0.47, "learning_rate": 5.282047410539031e-05, "loss": 1.048, "step": 188555 }, { "epoch": 0.47, "learning_rate": 5.28192158639086e-05, "loss": 1.0484, "step": 188560 }, { "epoch": 0.47, "learning_rate": 5.281795762242689e-05, "loss": 1.0599, "step": 188565 }, { "epoch": 0.47, "learning_rate": 5.281669938094519e-05, "loss": 1.0465, "step": 188570 }, { "epoch": 0.47, "learning_rate": 5.281544113946348e-05, "loss": 1.0462, "step": 188575 }, { "epoch": 0.47, "learning_rate": 5.281418289798178e-05, "loss": 1.0432, "step": 188580 }, { "epoch": 0.47, "learning_rate": 5.281292465650007e-05, "loss": 1.0474, "step": 188585 }, { "epoch": 0.47, "learning_rate": 5.281166641501837e-05, "loss": 1.0454, "step": 188590 }, { "epoch": 0.47, "learning_rate": 5.281040817353666e-05, "loss": 1.0455, "step": 188595 }, { "epoch": 0.47, "learning_rate": 5.280914993205496e-05, "loss": 1.0449, "step": 188600 }, { "epoch": 0.47, "learning_rate": 5.280789169057325e-05, "loss": 1.0452, "step": 188605 }, { "epoch": 0.47, "learning_rate": 5.280663344909155e-05, "loss": 1.0466, "step": 188610 }, { "epoch": 0.47, "learning_rate": 5.280537520760984e-05, "loss": 1.0416, "step": 188615 }, { "epoch": 0.47, "learning_rate": 5.280411696612814e-05, "loss": 1.0426, "step": 188620 }, { "epoch": 0.47, "learning_rate": 5.280285872464643e-05, "loss": 1.0465, "step": 188625 }, { "epoch": 0.47, "learning_rate": 5.280160048316472e-05, "loss": 1.0458, "step": 188630 }, { "epoch": 0.47, "learning_rate": 5.2800342241683035e-05, "loss": 1.0688, "step": 188635 }, { "epoch": 0.47, "learning_rate": 5.2799084000201326e-05, "loss": 1.041, "step": 188640 }, { "epoch": 0.47, "learning_rate": 5.279782575871962e-05, "loss": 1.0428, "step": 188645 }, { "epoch": 0.47, "learning_rate": 5.2796567517237916e-05, "loss": 1.046, "step": 188650 }, { "epoch": 0.47, "learning_rate": 5.279530927575621e-05, "loss": 1.0467, "step": 188655 }, { "epoch": 0.47, "learning_rate": 5.2794051034274506e-05, "loss": 1.046, "step": 188660 }, { "epoch": 0.47, "learning_rate": 5.27927927927928e-05, "loss": 1.0437, "step": 188665 }, { "epoch": 0.47, "learning_rate": 5.2791534551311096e-05, "loss": 1.0666, "step": 188670 }, { "epoch": 0.47, "learning_rate": 5.279027630982939e-05, "loss": 1.0419, "step": 188675 }, { "epoch": 0.47, "learning_rate": 5.2789018068347686e-05, "loss": 1.0752, "step": 188680 }, { "epoch": 0.47, "learning_rate": 5.278775982686598e-05, "loss": 1.0412, "step": 188685 }, { "epoch": 0.47, "learning_rate": 5.2786501585384276e-05, "loss": 1.0455, "step": 188690 }, { "epoch": 0.47, "learning_rate": 5.278524334390257e-05, "loss": 1.0493, "step": 188695 }, { "epoch": 0.47, "learning_rate": 5.2783985102420866e-05, "loss": 1.048, "step": 188700 }, { "epoch": 0.47, "learning_rate": 5.278272686093916e-05, "loss": 1.0432, "step": 188705 }, { "epoch": 0.47, "learning_rate": 5.278146861945745e-05, "loss": 1.0426, "step": 188710 }, { "epoch": 0.47, "learning_rate": 5.278021037797575e-05, "loss": 1.0454, "step": 188715 }, { "epoch": 0.47, "learning_rate": 5.277895213649404e-05, "loss": 1.0467, "step": 188720 }, { "epoch": 0.47, "learning_rate": 5.277769389501234e-05, "loss": 1.0465, "step": 188725 }, { "epoch": 0.47, "learning_rate": 5.277643565353063e-05, "loss": 1.0441, "step": 188730 }, { "epoch": 0.47, "learning_rate": 5.277517741204893e-05, "loss": 1.043, "step": 188735 }, { "epoch": 0.47, "learning_rate": 5.277391917056722e-05, "loss": 1.0492, "step": 188740 }, { "epoch": 0.47, "learning_rate": 5.277266092908552e-05, "loss": 1.0692, "step": 188745 }, { "epoch": 0.47, "learning_rate": 5.277140268760381e-05, "loss": 1.0423, "step": 188750 }, { "epoch": 0.47, "learning_rate": 5.277014444612211e-05, "loss": 1.0447, "step": 188755 }, { "epoch": 0.47, "learning_rate": 5.27688862046404e-05, "loss": 1.0442, "step": 188760 }, { "epoch": 0.47, "learning_rate": 5.27676279631587e-05, "loss": 1.044, "step": 188765 }, { "epoch": 0.47, "learning_rate": 5.276636972167699e-05, "loss": 1.0426, "step": 188770 }, { "epoch": 0.47, "learning_rate": 5.276511148019528e-05, "loss": 1.0459, "step": 188775 }, { "epoch": 0.47, "learning_rate": 5.276385323871358e-05, "loss": 1.0595, "step": 188780 }, { "epoch": 0.47, "learning_rate": 5.276259499723187e-05, "loss": 1.044, "step": 188785 }, { "epoch": 0.47, "learning_rate": 5.276133675575017e-05, "loss": 1.0453, "step": 188790 }, { "epoch": 0.47, "learning_rate": 5.276007851426846e-05, "loss": 1.044, "step": 188795 }, { "epoch": 0.47, "learning_rate": 5.275882027278676e-05, "loss": 1.044, "step": 188800 }, { "epoch": 0.47, "learning_rate": 5.275756203130505e-05, "loss": 1.0489, "step": 188805 }, { "epoch": 0.47, "learning_rate": 5.275630378982335e-05, "loss": 1.0432, "step": 188810 }, { "epoch": 0.47, "learning_rate": 5.275504554834164e-05, "loss": 1.0437, "step": 188815 }, { "epoch": 0.47, "learning_rate": 5.275378730685994e-05, "loss": 1.0484, "step": 188820 }, { "epoch": 0.47, "learning_rate": 5.275252906537823e-05, "loss": 1.0442, "step": 188825 }, { "epoch": 0.47, "learning_rate": 5.275127082389653e-05, "loss": 1.0683, "step": 188830 }, { "epoch": 0.47, "learning_rate": 5.275001258241482e-05, "loss": 1.0462, "step": 188835 }, { "epoch": 0.47, "learning_rate": 5.274875434093311e-05, "loss": 1.0457, "step": 188840 }, { "epoch": 0.47, "learning_rate": 5.274749609945141e-05, "loss": 1.044, "step": 188845 }, { "epoch": 0.47, "learning_rate": 5.27462378579697e-05, "loss": 1.0462, "step": 188850 }, { "epoch": 0.47, "learning_rate": 5.2744979616488e-05, "loss": 1.042, "step": 188855 }, { "epoch": 0.47, "learning_rate": 5.274372137500629e-05, "loss": 1.0453, "step": 188860 }, { "epoch": 0.47, "learning_rate": 5.274246313352459e-05, "loss": 1.0441, "step": 188865 }, { "epoch": 0.47, "learning_rate": 5.274120489204288e-05, "loss": 1.0717, "step": 188870 }, { "epoch": 0.47, "learning_rate": 5.273994665056118e-05, "loss": 1.0471, "step": 188875 }, { "epoch": 0.47, "learning_rate": 5.273868840907947e-05, "loss": 1.0415, "step": 188880 }, { "epoch": 0.47, "learning_rate": 5.273743016759777e-05, "loss": 1.0424, "step": 188885 }, { "epoch": 0.47, "learning_rate": 5.273617192611606e-05, "loss": 1.0465, "step": 188890 }, { "epoch": 0.47, "learning_rate": 5.273491368463436e-05, "loss": 1.0421, "step": 188895 }, { "epoch": 0.47, "learning_rate": 5.273365544315265e-05, "loss": 1.0434, "step": 188900 }, { "epoch": 0.47, "learning_rate": 5.2732397201670944e-05, "loss": 1.0447, "step": 188905 }, { "epoch": 0.47, "learning_rate": 5.273113896018924e-05, "loss": 1.0449, "step": 188910 }, { "epoch": 0.47, "learning_rate": 5.2729880718707534e-05, "loss": 1.0456, "step": 188915 }, { "epoch": 0.47, "learning_rate": 5.272862247722583e-05, "loss": 1.0441, "step": 188920 }, { "epoch": 0.47, "learning_rate": 5.2727364235744124e-05, "loss": 1.0444, "step": 188925 }, { "epoch": 0.47, "learning_rate": 5.272610599426242e-05, "loss": 1.0442, "step": 188930 }, { "epoch": 0.47, "learning_rate": 5.2724847752780714e-05, "loss": 1.0437, "step": 188935 }, { "epoch": 0.47, "learning_rate": 5.272358951129901e-05, "loss": 1.0465, "step": 188940 }, { "epoch": 0.47, "learning_rate": 5.2722331269817304e-05, "loss": 1.0447, "step": 188945 }, { "epoch": 0.47, "learning_rate": 5.27210730283356e-05, "loss": 1.0473, "step": 188950 }, { "epoch": 0.47, "learning_rate": 5.2719814786853894e-05, "loss": 1.0437, "step": 188955 }, { "epoch": 0.47, "learning_rate": 5.2718556545372185e-05, "loss": 1.0464, "step": 188960 }, { "epoch": 0.47, "learning_rate": 5.2717298303890484e-05, "loss": 1.0459, "step": 188965 }, { "epoch": 0.47, "learning_rate": 5.2716040062408775e-05, "loss": 1.0457, "step": 188970 }, { "epoch": 0.47, "learning_rate": 5.2714781820927074e-05, "loss": 1.045, "step": 188975 }, { "epoch": 0.47, "learning_rate": 5.2713523579445365e-05, "loss": 1.044, "step": 188980 }, { "epoch": 0.47, "learning_rate": 5.271251698626001e-05, "loss": 1.0442, "step": 188985 }, { "epoch": 0.47, "learning_rate": 5.27112587447783e-05, "loss": 1.0424, "step": 188990 }, { "epoch": 0.47, "learning_rate": 5.27100005032966e-05, "loss": 1.0394, "step": 188995 }, { "epoch": 0.47, "learning_rate": 5.270874226181489e-05, "loss": 1.0476, "step": 189000 }, { "epoch": 0.47, "learning_rate": 5.270748402033319e-05, "loss": 1.0459, "step": 189005 }, { "epoch": 0.47, "learning_rate": 5.270622577885148e-05, "loss": 1.044, "step": 189010 }, { "epoch": 0.47, "learning_rate": 5.270496753736978e-05, "loss": 1.0438, "step": 189015 }, { "epoch": 0.47, "learning_rate": 5.270370929588807e-05, "loss": 1.0446, "step": 189020 }, { "epoch": 0.47, "learning_rate": 5.270245105440637e-05, "loss": 1.0453, "step": 189025 }, { "epoch": 0.47, "learning_rate": 5.270119281292466e-05, "loss": 1.046, "step": 189030 }, { "epoch": 0.47, "learning_rate": 5.269993457144296e-05, "loss": 1.0437, "step": 189035 }, { "epoch": 0.47, "learning_rate": 5.269867632996125e-05, "loss": 1.0437, "step": 189040 }, { "epoch": 0.47, "learning_rate": 5.269741808847954e-05, "loss": 1.0447, "step": 189045 }, { "epoch": 0.47, "learning_rate": 5.269615984699784e-05, "loss": 1.0439, "step": 189050 }, { "epoch": 0.47, "learning_rate": 5.269490160551613e-05, "loss": 1.0474, "step": 189055 }, { "epoch": 0.47, "learning_rate": 5.269364336403443e-05, "loss": 1.0462, "step": 189060 }, { "epoch": 0.47, "learning_rate": 5.269238512255272e-05, "loss": 1.0471, "step": 189065 }, { "epoch": 0.47, "learning_rate": 5.269112688107102e-05, "loss": 1.0445, "step": 189070 }, { "epoch": 0.47, "learning_rate": 5.268986863958931e-05, "loss": 1.0439, "step": 189075 }, { "epoch": 0.47, "learning_rate": 5.268861039810761e-05, "loss": 1.0426, "step": 189080 }, { "epoch": 0.47, "learning_rate": 5.26873521566259e-05, "loss": 1.0687, "step": 189085 }, { "epoch": 0.47, "learning_rate": 5.26860939151442e-05, "loss": 1.0459, "step": 189090 }, { "epoch": 0.47, "learning_rate": 5.268483567366249e-05, "loss": 1.0474, "step": 189095 }, { "epoch": 0.47, "learning_rate": 5.268357743218079e-05, "loss": 1.0438, "step": 189100 }, { "epoch": 0.47, "learning_rate": 5.268231919069908e-05, "loss": 1.0461, "step": 189105 }, { "epoch": 0.47, "learning_rate": 5.268106094921737e-05, "loss": 1.0465, "step": 189110 }, { "epoch": 0.47, "learning_rate": 5.267980270773567e-05, "loss": 1.045, "step": 189115 }, { "epoch": 0.47, "learning_rate": 5.267854446625396e-05, "loss": 1.0467, "step": 189120 }, { "epoch": 0.47, "learning_rate": 5.267728622477226e-05, "loss": 1.0451, "step": 189125 }, { "epoch": 0.47, "learning_rate": 5.267602798329055e-05, "loss": 1.0443, "step": 189130 }, { "epoch": 0.47, "learning_rate": 5.267476974180885e-05, "loss": 1.0429, "step": 189135 }, { "epoch": 0.47, "learning_rate": 5.267351150032714e-05, "loss": 1.0442, "step": 189140 }, { "epoch": 0.47, "learning_rate": 5.267225325884544e-05, "loss": 1.0465, "step": 189145 }, { "epoch": 0.47, "learning_rate": 5.267099501736373e-05, "loss": 1.0475, "step": 189150 }, { "epoch": 0.47, "learning_rate": 5.266973677588203e-05, "loss": 1.0455, "step": 189155 }, { "epoch": 0.47, "learning_rate": 5.266847853440032e-05, "loss": 1.047, "step": 189160 }, { "epoch": 0.47, "learning_rate": 5.266722029291862e-05, "loss": 1.0467, "step": 189165 }, { "epoch": 0.47, "learning_rate": 5.266596205143691e-05, "loss": 1.0445, "step": 189170 }, { "epoch": 0.47, "learning_rate": 5.2664703809955204e-05, "loss": 1.0472, "step": 189175 }, { "epoch": 0.47, "learning_rate": 5.26634455684735e-05, "loss": 1.0455, "step": 189180 }, { "epoch": 0.47, "learning_rate": 5.2662187326991794e-05, "loss": 1.0458, "step": 189185 }, { "epoch": 0.47, "learning_rate": 5.266092908551009e-05, "loss": 1.0431, "step": 189190 }, { "epoch": 0.47, "learning_rate": 5.2659670844028384e-05, "loss": 1.0456, "step": 189195 }, { "epoch": 0.47, "learning_rate": 5.265841260254668e-05, "loss": 1.0476, "step": 189200 }, { "epoch": 0.47, "learning_rate": 5.2657154361064974e-05, "loss": 1.0429, "step": 189205 }, { "epoch": 0.47, "learning_rate": 5.265589611958327e-05, "loss": 1.0452, "step": 189210 }, { "epoch": 0.47, "learning_rate": 5.2654637878101564e-05, "loss": 1.046, "step": 189215 }, { "epoch": 0.47, "learning_rate": 5.265337963661986e-05, "loss": 1.0436, "step": 189220 }, { "epoch": 0.47, "learning_rate": 5.2652121395138154e-05, "loss": 1.0424, "step": 189225 }, { "epoch": 0.47, "learning_rate": 5.265086315365645e-05, "loss": 1.0454, "step": 189230 }, { "epoch": 0.48, "learning_rate": 5.2649604912174744e-05, "loss": 1.0431, "step": 189235 }, { "epoch": 0.48, "learning_rate": 5.2648346670693035e-05, "loss": 1.0447, "step": 189240 }, { "epoch": 0.48, "learning_rate": 5.2647088429211334e-05, "loss": 1.0454, "step": 189245 }, { "epoch": 0.48, "learning_rate": 5.2645830187729625e-05, "loss": 1.0427, "step": 189250 }, { "epoch": 0.48, "learning_rate": 5.2644571946247924e-05, "loss": 1.0472, "step": 189255 }, { "epoch": 0.48, "learning_rate": 5.2643313704766215e-05, "loss": 1.0472, "step": 189260 }, { "epoch": 0.48, "learning_rate": 5.2642055463284514e-05, "loss": 1.0416, "step": 189265 }, { "epoch": 0.48, "learning_rate": 5.2640797221802805e-05, "loss": 1.0413, "step": 189270 }, { "epoch": 0.48, "learning_rate": 5.2639538980321104e-05, "loss": 1.0423, "step": 189275 }, { "epoch": 0.48, "learning_rate": 5.2638280738839395e-05, "loss": 1.045, "step": 189280 }, { "epoch": 0.48, "learning_rate": 5.2637022497357693e-05, "loss": 1.0447, "step": 189285 }, { "epoch": 0.48, "learning_rate": 5.2635764255875985e-05, "loss": 1.0435, "step": 189290 }, { "epoch": 0.48, "learning_rate": 5.2634506014394283e-05, "loss": 1.0422, "step": 189295 }, { "epoch": 0.48, "learning_rate": 5.2633247772912575e-05, "loss": 1.0463, "step": 189300 }, { "epoch": 0.48, "learning_rate": 5.263198953143087e-05, "loss": 1.0476, "step": 189305 }, { "epoch": 0.48, "learning_rate": 5.2630731289949165e-05, "loss": 1.0473, "step": 189310 }, { "epoch": 0.48, "learning_rate": 5.262947304846746e-05, "loss": 1.0439, "step": 189315 }, { "epoch": 0.48, "learning_rate": 5.2628214806985755e-05, "loss": 1.0428, "step": 189320 }, { "epoch": 0.48, "learning_rate": 5.2626956565504047e-05, "loss": 1.0469, "step": 189325 }, { "epoch": 0.48, "learning_rate": 5.2625698324022345e-05, "loss": 1.045, "step": 189330 }, { "epoch": 0.48, "learning_rate": 5.262444008254065e-05, "loss": 1.0483, "step": 189335 }, { "epoch": 0.48, "learning_rate": 5.262318184105894e-05, "loss": 1.0434, "step": 189340 }, { "epoch": 0.48, "learning_rate": 5.262192359957724e-05, "loss": 1.0463, "step": 189345 }, { "epoch": 0.48, "learning_rate": 5.262066535809553e-05, "loss": 1.0459, "step": 189350 }, { "epoch": 0.48, "learning_rate": 5.261940711661383e-05, "loss": 1.0435, "step": 189355 }, { "epoch": 0.48, "learning_rate": 5.261814887513212e-05, "loss": 1.0417, "step": 189360 }, { "epoch": 0.48, "learning_rate": 5.261689063365042e-05, "loss": 1.0483, "step": 189365 }, { "epoch": 0.48, "learning_rate": 5.261563239216871e-05, "loss": 1.0488, "step": 189370 }, { "epoch": 0.48, "learning_rate": 5.261437415068701e-05, "loss": 1.0452, "step": 189375 }, { "epoch": 0.48, "learning_rate": 5.26131159092053e-05, "loss": 1.0464, "step": 189380 }, { "epoch": 0.48, "learning_rate": 5.261185766772359e-05, "loss": 1.0451, "step": 189385 }, { "epoch": 0.48, "learning_rate": 5.261059942624189e-05, "loss": 1.0448, "step": 189390 }, { "epoch": 0.48, "learning_rate": 5.260934118476018e-05, "loss": 1.0433, "step": 189395 }, { "epoch": 0.48, "learning_rate": 5.260808294327848e-05, "loss": 1.0445, "step": 189400 }, { "epoch": 0.48, "learning_rate": 5.260682470179677e-05, "loss": 1.0438, "step": 189405 }, { "epoch": 0.48, "learning_rate": 5.260556646031507e-05, "loss": 1.0427, "step": 189410 }, { "epoch": 0.48, "learning_rate": 5.260430821883336e-05, "loss": 1.0474, "step": 189415 }, { "epoch": 0.48, "learning_rate": 5.260304997735166e-05, "loss": 1.0444, "step": 189420 }, { "epoch": 0.48, "learning_rate": 5.260179173586995e-05, "loss": 1.0405, "step": 189425 }, { "epoch": 0.48, "learning_rate": 5.260053349438825e-05, "loss": 1.045, "step": 189430 }, { "epoch": 0.48, "learning_rate": 5.259927525290654e-05, "loss": 1.0448, "step": 189435 }, { "epoch": 0.48, "learning_rate": 5.259801701142484e-05, "loss": 1.043, "step": 189440 }, { "epoch": 0.48, "learning_rate": 5.259675876994313e-05, "loss": 1.0431, "step": 189445 }, { "epoch": 0.48, "learning_rate": 5.2595500528461425e-05, "loss": 1.0467, "step": 189450 }, { "epoch": 0.48, "learning_rate": 5.259424228697972e-05, "loss": 1.044, "step": 189455 }, { "epoch": 0.48, "learning_rate": 5.2592984045498015e-05, "loss": 1.0665, "step": 189460 }, { "epoch": 0.48, "learning_rate": 5.259172580401631e-05, "loss": 1.0448, "step": 189465 }, { "epoch": 0.48, "learning_rate": 5.2590467562534605e-05, "loss": 1.0476, "step": 189470 }, { "epoch": 0.48, "learning_rate": 5.25892093210529e-05, "loss": 1.0462, "step": 189475 }, { "epoch": 0.48, "learning_rate": 5.2587951079571195e-05, "loss": 1.0447, "step": 189480 }, { "epoch": 0.48, "learning_rate": 5.258669283808949e-05, "loss": 1.0469, "step": 189485 }, { "epoch": 0.48, "learning_rate": 5.2585434596607785e-05, "loss": 1.0432, "step": 189490 }, { "epoch": 0.48, "learning_rate": 5.258417635512608e-05, "loss": 1.0471, "step": 189495 }, { "epoch": 0.48, "learning_rate": 5.2582918113644374e-05, "loss": 1.0438, "step": 189500 }, { "epoch": 0.48, "learning_rate": 5.258165987216267e-05, "loss": 1.0644, "step": 189505 }, { "epoch": 0.48, "learning_rate": 5.2580401630680964e-05, "loss": 1.0443, "step": 189510 }, { "epoch": 0.48, "learning_rate": 5.2579143389199256e-05, "loss": 1.0426, "step": 189515 }, { "epoch": 0.48, "learning_rate": 5.2577885147717554e-05, "loss": 1.0457, "step": 189520 }, { "epoch": 0.48, "learning_rate": 5.2576626906235846e-05, "loss": 1.0428, "step": 189525 }, { "epoch": 0.48, "learning_rate": 5.2575368664754144e-05, "loss": 1.065, "step": 189530 }, { "epoch": 0.48, "learning_rate": 5.2574110423272436e-05, "loss": 1.0452, "step": 189535 }, { "epoch": 0.48, "learning_rate": 5.2572852181790734e-05, "loss": 1.045, "step": 189540 }, { "epoch": 0.48, "learning_rate": 5.2571593940309026e-05, "loss": 1.0433, "step": 189545 }, { "epoch": 0.48, "learning_rate": 5.2570335698827324e-05, "loss": 1.0442, "step": 189550 }, { "epoch": 0.48, "learning_rate": 5.2569077457345616e-05, "loss": 1.0489, "step": 189555 }, { "epoch": 0.48, "learning_rate": 5.2567819215863914e-05, "loss": 1.0439, "step": 189560 }, { "epoch": 0.48, "learning_rate": 5.2566560974382206e-05, "loss": 1.0463, "step": 189565 }, { "epoch": 0.48, "learning_rate": 5.2565302732900504e-05, "loss": 1.0479, "step": 189570 }, { "epoch": 0.48, "learning_rate": 5.2564044491418796e-05, "loss": 1.0466, "step": 189575 }, { "epoch": 0.48, "learning_rate": 5.256278624993709e-05, "loss": 1.0477, "step": 189580 }, { "epoch": 0.48, "learning_rate": 5.2561528008455386e-05, "loss": 1.0457, "step": 189585 }, { "epoch": 0.48, "learning_rate": 5.256026976697368e-05, "loss": 1.0452, "step": 189590 }, { "epoch": 0.48, "learning_rate": 5.2559011525491976e-05, "loss": 1.0456, "step": 189595 }, { "epoch": 0.48, "learning_rate": 5.255775328401027e-05, "loss": 1.0419, "step": 189600 }, { "epoch": 0.48, "learning_rate": 5.2556495042528566e-05, "loss": 1.0444, "step": 189605 }, { "epoch": 0.48, "learning_rate": 5.255523680104686e-05, "loss": 1.0479, "step": 189610 }, { "epoch": 0.48, "learning_rate": 5.2553978559565156e-05, "loss": 1.0465, "step": 189615 }, { "epoch": 0.48, "learning_rate": 5.255272031808345e-05, "loss": 1.0447, "step": 189620 }, { "epoch": 0.48, "learning_rate": 5.2551462076601746e-05, "loss": 1.0727, "step": 189625 }, { "epoch": 0.48, "learning_rate": 5.255020383512004e-05, "loss": 1.0458, "step": 189630 }, { "epoch": 0.48, "learning_rate": 5.2548945593638336e-05, "loss": 1.0478, "step": 189635 }, { "epoch": 0.48, "learning_rate": 5.254768735215663e-05, "loss": 1.0449, "step": 189640 }, { "epoch": 0.48, "learning_rate": 5.254642911067492e-05, "loss": 1.0455, "step": 189645 }, { "epoch": 0.48, "learning_rate": 5.254517086919322e-05, "loss": 1.0459, "step": 189650 }, { "epoch": 0.48, "learning_rate": 5.254391262771151e-05, "loss": 1.0443, "step": 189655 }, { "epoch": 0.48, "learning_rate": 5.254265438622981e-05, "loss": 1.0413, "step": 189660 }, { "epoch": 0.48, "learning_rate": 5.25413961447481e-05, "loss": 1.0441, "step": 189665 }, { "epoch": 0.48, "learning_rate": 5.25401379032664e-05, "loss": 1.0439, "step": 189670 }, { "epoch": 0.48, "learning_rate": 5.253887966178469e-05, "loss": 1.0475, "step": 189675 }, { "epoch": 0.48, "learning_rate": 5.253762142030299e-05, "loss": 1.0474, "step": 189680 }, { "epoch": 0.48, "learning_rate": 5.253636317882128e-05, "loss": 1.049, "step": 189685 }, { "epoch": 0.48, "learning_rate": 5.253510493733958e-05, "loss": 1.048, "step": 189690 }, { "epoch": 0.48, "learning_rate": 5.253384669585787e-05, "loss": 1.0429, "step": 189695 }, { "epoch": 0.48, "learning_rate": 5.253258845437616e-05, "loss": 1.0423, "step": 189700 }, { "epoch": 0.48, "learning_rate": 5.253133021289446e-05, "loss": 1.0458, "step": 189705 }, { "epoch": 0.48, "learning_rate": 5.253007197141275e-05, "loss": 1.044, "step": 189710 }, { "epoch": 0.48, "learning_rate": 5.252881372993105e-05, "loss": 1.0422, "step": 189715 }, { "epoch": 0.48, "learning_rate": 5.252755548844934e-05, "loss": 1.0487, "step": 189720 }, { "epoch": 0.48, "learning_rate": 5.252629724696764e-05, "loss": 1.0478, "step": 189725 }, { "epoch": 0.48, "learning_rate": 5.252503900548593e-05, "loss": 1.0481, "step": 189730 }, { "epoch": 0.48, "learning_rate": 5.252378076400423e-05, "loss": 1.0453, "step": 189735 }, { "epoch": 0.48, "learning_rate": 5.2522774170818865e-05, "loss": 1.044, "step": 189740 }, { "epoch": 0.48, "learning_rate": 5.252151592933716e-05, "loss": 1.0455, "step": 189745 }, { "epoch": 0.48, "learning_rate": 5.2520257687855455e-05, "loss": 1.0454, "step": 189750 }, { "epoch": 0.48, "learning_rate": 5.251899944637375e-05, "loss": 1.0456, "step": 189755 }, { "epoch": 0.48, "learning_rate": 5.2517741204892045e-05, "loss": 1.0435, "step": 189760 }, { "epoch": 0.48, "learning_rate": 5.251648296341034e-05, "loss": 1.0457, "step": 189765 }, { "epoch": 0.48, "learning_rate": 5.2515224721928635e-05, "loss": 1.0454, "step": 189770 }, { "epoch": 0.48, "learning_rate": 5.251396648044693e-05, "loss": 1.0429, "step": 189775 }, { "epoch": 0.48, "learning_rate": 5.2512708238965225e-05, "loss": 1.0452, "step": 189780 }, { "epoch": 0.48, "learning_rate": 5.2511449997483516e-05, "loss": 1.0451, "step": 189785 }, { "epoch": 0.48, "learning_rate": 5.2510191756001814e-05, "loss": 1.0456, "step": 189790 }, { "epoch": 0.48, "learning_rate": 5.2508933514520106e-05, "loss": 1.0451, "step": 189795 }, { "epoch": 0.48, "learning_rate": 5.2507675273038404e-05, "loss": 1.0454, "step": 189800 }, { "epoch": 0.48, "learning_rate": 5.2506417031556696e-05, "loss": 1.0449, "step": 189805 }, { "epoch": 0.48, "learning_rate": 5.2505158790074994e-05, "loss": 1.0438, "step": 189810 }, { "epoch": 0.48, "learning_rate": 5.2503900548593286e-05, "loss": 1.0446, "step": 189815 }, { "epoch": 0.48, "learning_rate": 5.2502642307111584e-05, "loss": 1.0452, "step": 189820 }, { "epoch": 0.48, "learning_rate": 5.2501384065629876e-05, "loss": 1.0457, "step": 189825 }, { "epoch": 0.48, "learning_rate": 5.2500125824148174e-05, "loss": 1.0459, "step": 189830 }, { "epoch": 0.48, "learning_rate": 5.2498867582666466e-05, "loss": 1.0468, "step": 189835 }, { "epoch": 0.48, "learning_rate": 5.2497609341184764e-05, "loss": 1.0433, "step": 189840 }, { "epoch": 0.48, "learning_rate": 5.2496351099703056e-05, "loss": 1.0464, "step": 189845 }, { "epoch": 0.48, "learning_rate": 5.249509285822135e-05, "loss": 1.0443, "step": 189850 }, { "epoch": 0.48, "learning_rate": 5.249408626503599e-05, "loss": 1.0619, "step": 189855 }, { "epoch": 0.48, "learning_rate": 5.249282802355429e-05, "loss": 1.0433, "step": 189860 }, { "epoch": 0.48, "learning_rate": 5.249156978207258e-05, "loss": 1.0456, "step": 189865 }, { "epoch": 0.48, "learning_rate": 5.249031154059087e-05, "loss": 1.0457, "step": 189870 }, { "epoch": 0.48, "learning_rate": 5.248905329910917e-05, "loss": 1.0449, "step": 189875 }, { "epoch": 0.48, "learning_rate": 5.248779505762746e-05, "loss": 1.0447, "step": 189880 }, { "epoch": 0.48, "learning_rate": 5.248653681614576e-05, "loss": 1.042, "step": 189885 }, { "epoch": 0.48, "learning_rate": 5.248527857466405e-05, "loss": 1.0467, "step": 189890 }, { "epoch": 0.48, "learning_rate": 5.248402033318235e-05, "loss": 1.0434, "step": 189895 }, { "epoch": 0.48, "learning_rate": 5.248276209170064e-05, "loss": 1.0658, "step": 189900 }, { "epoch": 0.48, "learning_rate": 5.248150385021894e-05, "loss": 1.0449, "step": 189905 }, { "epoch": 0.48, "learning_rate": 5.248024560873723e-05, "loss": 1.0441, "step": 189910 }, { "epoch": 0.48, "learning_rate": 5.247898736725553e-05, "loss": 1.0436, "step": 189915 }, { "epoch": 0.48, "learning_rate": 5.247772912577382e-05, "loss": 1.0445, "step": 189920 }, { "epoch": 0.48, "learning_rate": 5.247647088429212e-05, "loss": 1.0459, "step": 189925 }, { "epoch": 0.48, "learning_rate": 5.247521264281041e-05, "loss": 1.0434, "step": 189930 }, { "epoch": 0.48, "learning_rate": 5.24739544013287e-05, "loss": 1.072, "step": 189935 }, { "epoch": 0.48, "learning_rate": 5.2472696159847e-05, "loss": 1.0452, "step": 189940 }, { "epoch": 0.48, "learning_rate": 5.247143791836529e-05, "loss": 1.0449, "step": 189945 }, { "epoch": 0.48, "learning_rate": 5.247017967688359e-05, "loss": 1.0435, "step": 189950 }, { "epoch": 0.48, "learning_rate": 5.246892143540188e-05, "loss": 1.0416, "step": 189955 }, { "epoch": 0.48, "learning_rate": 5.246766319392018e-05, "loss": 1.0433, "step": 189960 }, { "epoch": 0.48, "learning_rate": 5.246640495243847e-05, "loss": 1.0462, "step": 189965 }, { "epoch": 0.48, "learning_rate": 5.246514671095677e-05, "loss": 1.0452, "step": 189970 }, { "epoch": 0.48, "learning_rate": 5.246388846947506e-05, "loss": 1.0456, "step": 189975 }, { "epoch": 0.48, "learning_rate": 5.246263022799336e-05, "loss": 1.044, "step": 189980 }, { "epoch": 0.48, "learning_rate": 5.246137198651165e-05, "loss": 1.0458, "step": 189985 }, { "epoch": 0.48, "learning_rate": 5.246011374502995e-05, "loss": 1.0438, "step": 189990 }, { "epoch": 0.48, "learning_rate": 5.245885550354824e-05, "loss": 1.0684, "step": 189995 }, { "epoch": 0.48, "learning_rate": 5.2457597262066535e-05, "loss": 1.0625, "step": 190000 }, { "epoch": 0.48, "learning_rate": 5.245633902058483e-05, "loss": 1.0462, "step": 190005 }, { "epoch": 0.48, "learning_rate": 5.2455080779103125e-05, "loss": 1.0446, "step": 190010 }, { "epoch": 0.48, "learning_rate": 5.245382253762142e-05, "loss": 1.0469, "step": 190015 }, { "epoch": 0.48, "learning_rate": 5.2452564296139715e-05, "loss": 1.0428, "step": 190020 }, { "epoch": 0.48, "learning_rate": 5.245130605465801e-05, "loss": 1.044, "step": 190025 }, { "epoch": 0.48, "learning_rate": 5.2450047813176305e-05, "loss": 1.0442, "step": 190030 }, { "epoch": 0.48, "learning_rate": 5.24487895716946e-05, "loss": 1.0464, "step": 190035 }, { "epoch": 0.48, "learning_rate": 5.2447531330212895e-05, "loss": 1.0445, "step": 190040 }, { "epoch": 0.48, "learning_rate": 5.244627308873119e-05, "loss": 1.0453, "step": 190045 }, { "epoch": 0.48, "learning_rate": 5.2445014847249485e-05, "loss": 1.0442, "step": 190050 }, { "epoch": 0.48, "learning_rate": 5.244375660576778e-05, "loss": 1.0452, "step": 190055 }, { "epoch": 0.48, "learning_rate": 5.2442498364286075e-05, "loss": 1.0435, "step": 190060 }, { "epoch": 0.48, "learning_rate": 5.2441240122804366e-05, "loss": 1.0443, "step": 190065 }, { "epoch": 0.48, "learning_rate": 5.2439981881322665e-05, "loss": 1.044, "step": 190070 }, { "epoch": 0.48, "learning_rate": 5.2438723639840956e-05, "loss": 1.0439, "step": 190075 }, { "epoch": 0.48, "learning_rate": 5.2437465398359254e-05, "loss": 1.0429, "step": 190080 }, { "epoch": 0.48, "learning_rate": 5.2436207156877546e-05, "loss": 1.0435, "step": 190085 }, { "epoch": 0.48, "learning_rate": 5.2434948915395844e-05, "loss": 1.0452, "step": 190090 }, { "epoch": 0.48, "learning_rate": 5.2433690673914136e-05, "loss": 1.0471, "step": 190095 }, { "epoch": 0.48, "learning_rate": 5.2432432432432434e-05, "loss": 1.0447, "step": 190100 }, { "epoch": 0.48, "learning_rate": 5.2431174190950726e-05, "loss": 1.0456, "step": 190105 }, { "epoch": 0.48, "learning_rate": 5.2429915949469024e-05, "loss": 1.0475, "step": 190110 }, { "epoch": 0.48, "learning_rate": 5.2428657707987316e-05, "loss": 1.046, "step": 190115 }, { "epoch": 0.48, "learning_rate": 5.2427399466505614e-05, "loss": 1.0428, "step": 190120 }, { "epoch": 0.48, "learning_rate": 5.2426141225023906e-05, "loss": 1.0455, "step": 190125 }, { "epoch": 0.48, "learning_rate": 5.24248829835422e-05, "loss": 1.0443, "step": 190130 }, { "epoch": 0.48, "learning_rate": 5.2423624742060496e-05, "loss": 1.0435, "step": 190135 }, { "epoch": 0.48, "learning_rate": 5.242236650057879e-05, "loss": 1.0448, "step": 190140 }, { "epoch": 0.48, "learning_rate": 5.2421108259097086e-05, "loss": 1.045, "step": 190145 }, { "epoch": 0.48, "learning_rate": 5.241985001761538e-05, "loss": 1.0418, "step": 190150 }, { "epoch": 0.48, "learning_rate": 5.2418591776133676e-05, "loss": 1.0454, "step": 190155 }, { "epoch": 0.48, "learning_rate": 5.241733353465197e-05, "loss": 1.0432, "step": 190160 }, { "epoch": 0.48, "learning_rate": 5.2416075293170266e-05, "loss": 1.0423, "step": 190165 }, { "epoch": 0.48, "learning_rate": 5.241481705168856e-05, "loss": 1.0447, "step": 190170 }, { "epoch": 0.48, "learning_rate": 5.2413558810206856e-05, "loss": 1.0453, "step": 190175 }, { "epoch": 0.48, "learning_rate": 5.241230056872515e-05, "loss": 1.0462, "step": 190180 }, { "epoch": 0.48, "learning_rate": 5.2411042327243446e-05, "loss": 1.0463, "step": 190185 }, { "epoch": 0.48, "learning_rate": 5.240978408576174e-05, "loss": 1.0451, "step": 190190 }, { "epoch": 0.48, "learning_rate": 5.240852584428003e-05, "loss": 1.0453, "step": 190195 }, { "epoch": 0.48, "learning_rate": 5.240726760279833e-05, "loss": 1.0467, "step": 190200 }, { "epoch": 0.48, "learning_rate": 5.240600936131662e-05, "loss": 1.0457, "step": 190205 }, { "epoch": 0.48, "learning_rate": 5.240475111983492e-05, "loss": 1.0498, "step": 190210 }, { "epoch": 0.48, "learning_rate": 5.240349287835321e-05, "loss": 1.0449, "step": 190215 }, { "epoch": 0.48, "learning_rate": 5.240223463687151e-05, "loss": 1.044, "step": 190220 }, { "epoch": 0.48, "learning_rate": 5.24009763953898e-05, "loss": 1.0465, "step": 190225 }, { "epoch": 0.48, "learning_rate": 5.23997181539081e-05, "loss": 1.0478, "step": 190230 }, { "epoch": 0.48, "learning_rate": 5.23984599124264e-05, "loss": 1.0467, "step": 190235 }, { "epoch": 0.48, "learning_rate": 5.2397201670944694e-05, "loss": 1.0473, "step": 190240 }, { "epoch": 0.48, "learning_rate": 5.239594342946299e-05, "loss": 1.0425, "step": 190245 }, { "epoch": 0.48, "learning_rate": 5.2394685187981284e-05, "loss": 1.0442, "step": 190250 }, { "epoch": 0.48, "learning_rate": 5.239342694649958e-05, "loss": 1.0442, "step": 190255 }, { "epoch": 0.48, "learning_rate": 5.2392168705017874e-05, "loss": 1.0419, "step": 190260 }, { "epoch": 0.48, "learning_rate": 5.239091046353617e-05, "loss": 1.0453, "step": 190265 }, { "epoch": 0.48, "learning_rate": 5.2389652222054464e-05, "loss": 1.0449, "step": 190270 }, { "epoch": 0.48, "learning_rate": 5.2388393980572756e-05, "loss": 1.0451, "step": 190275 }, { "epoch": 0.48, "learning_rate": 5.2387135739091054e-05, "loss": 1.0474, "step": 190280 }, { "epoch": 0.48, "learning_rate": 5.2385877497609346e-05, "loss": 1.0417, "step": 190285 }, { "epoch": 0.48, "learning_rate": 5.2384619256127644e-05, "loss": 1.046, "step": 190290 }, { "epoch": 0.48, "learning_rate": 5.2383361014645935e-05, "loss": 1.0488, "step": 190295 }, { "epoch": 0.48, "learning_rate": 5.2382102773164234e-05, "loss": 1.0464, "step": 190300 }, { "epoch": 0.48, "learning_rate": 5.2380844531682525e-05, "loss": 1.0421, "step": 190305 }, { "epoch": 0.48, "learning_rate": 5.2379586290200824e-05, "loss": 1.044, "step": 190310 }, { "epoch": 0.48, "learning_rate": 5.2378328048719115e-05, "loss": 1.0449, "step": 190315 }, { "epoch": 0.48, "learning_rate": 5.2377069807237414e-05, "loss": 1.0461, "step": 190320 }, { "epoch": 0.48, "learning_rate": 5.2375811565755705e-05, "loss": 1.0459, "step": 190325 }, { "epoch": 0.48, "learning_rate": 5.2374553324274004e-05, "loss": 1.0451, "step": 190330 }, { "epoch": 0.48, "learning_rate": 5.2373295082792295e-05, "loss": 1.0474, "step": 190335 }, { "epoch": 0.48, "learning_rate": 5.237203684131059e-05, "loss": 1.0448, "step": 190340 }, { "epoch": 0.48, "learning_rate": 5.2370778599828885e-05, "loss": 1.0439, "step": 190345 }, { "epoch": 0.48, "learning_rate": 5.236952035834718e-05, "loss": 1.0436, "step": 190350 }, { "epoch": 0.48, "learning_rate": 5.2368262116865475e-05, "loss": 1.0443, "step": 190355 }, { "epoch": 0.48, "learning_rate": 5.236700387538377e-05, "loss": 1.0468, "step": 190360 }, { "epoch": 0.48, "learning_rate": 5.2365745633902065e-05, "loss": 1.0407, "step": 190365 }, { "epoch": 0.48, "learning_rate": 5.236448739242036e-05, "loss": 1.0435, "step": 190370 }, { "epoch": 0.48, "learning_rate": 5.2363229150938655e-05, "loss": 1.0449, "step": 190375 }, { "epoch": 0.48, "learning_rate": 5.236197090945695e-05, "loss": 1.0447, "step": 190380 }, { "epoch": 0.48, "learning_rate": 5.2360712667975245e-05, "loss": 1.0447, "step": 190385 }, { "epoch": 0.48, "learning_rate": 5.235945442649354e-05, "loss": 1.046, "step": 190390 }, { "epoch": 0.48, "learning_rate": 5.2358196185011835e-05, "loss": 1.0426, "step": 190395 }, { "epoch": 0.48, "learning_rate": 5.235693794353013e-05, "loss": 1.0457, "step": 190400 }, { "epoch": 0.48, "learning_rate": 5.235567970204842e-05, "loss": 1.0432, "step": 190405 }, { "epoch": 0.48, "learning_rate": 5.235442146056672e-05, "loss": 1.0458, "step": 190410 }, { "epoch": 0.48, "learning_rate": 5.235316321908501e-05, "loss": 1.046, "step": 190415 }, { "epoch": 0.48, "learning_rate": 5.235190497760331e-05, "loss": 1.0437, "step": 190420 }, { "epoch": 0.48, "learning_rate": 5.23506467361216e-05, "loss": 1.0452, "step": 190425 }, { "epoch": 0.48, "learning_rate": 5.23493884946399e-05, "loss": 1.0453, "step": 190430 }, { "epoch": 0.48, "learning_rate": 5.234813025315819e-05, "loss": 1.0451, "step": 190435 }, { "epoch": 0.48, "learning_rate": 5.234687201167649e-05, "loss": 1.0425, "step": 190440 }, { "epoch": 0.48, "learning_rate": 5.234561377019478e-05, "loss": 1.0459, "step": 190445 }, { "epoch": 0.48, "learning_rate": 5.234435552871308e-05, "loss": 1.0442, "step": 190450 }, { "epoch": 0.48, "learning_rate": 5.234309728723137e-05, "loss": 1.0474, "step": 190455 }, { "epoch": 0.48, "learning_rate": 5.234183904574967e-05, "loss": 1.0419, "step": 190460 }, { "epoch": 0.48, "learning_rate": 5.234058080426796e-05, "loss": 1.0477, "step": 190465 }, { "epoch": 0.48, "learning_rate": 5.233932256278625e-05, "loss": 1.046, "step": 190470 }, { "epoch": 0.48, "learning_rate": 5.233806432130455e-05, "loss": 1.0428, "step": 190475 }, { "epoch": 0.48, "learning_rate": 5.233680607982284e-05, "loss": 1.0458, "step": 190480 }, { "epoch": 0.48, "learning_rate": 5.233554783834114e-05, "loss": 1.0457, "step": 190485 }, { "epoch": 0.48, "learning_rate": 5.233428959685943e-05, "loss": 1.0659, "step": 190490 }, { "epoch": 0.48, "learning_rate": 5.233303135537773e-05, "loss": 1.0419, "step": 190495 }, { "epoch": 0.48, "learning_rate": 5.233177311389602e-05, "loss": 1.046, "step": 190500 }, { "epoch": 0.48, "learning_rate": 5.233051487241432e-05, "loss": 1.0447, "step": 190505 }, { "epoch": 0.48, "learning_rate": 5.232925663093261e-05, "loss": 1.0438, "step": 190510 }, { "epoch": 0.48, "learning_rate": 5.232799838945091e-05, "loss": 1.0444, "step": 190515 }, { "epoch": 0.48, "learning_rate": 5.23267401479692e-05, "loss": 1.0445, "step": 190520 }, { "epoch": 0.48, "learning_rate": 5.232548190648749e-05, "loss": 1.0446, "step": 190525 }, { "epoch": 0.48, "learning_rate": 5.232422366500579e-05, "loss": 1.0425, "step": 190530 }, { "epoch": 0.48, "learning_rate": 5.232296542352408e-05, "loss": 1.0471, "step": 190535 }, { "epoch": 0.48, "learning_rate": 5.232170718204238e-05, "loss": 1.0469, "step": 190540 }, { "epoch": 0.48, "learning_rate": 5.232044894056067e-05, "loss": 1.0449, "step": 190545 }, { "epoch": 0.48, "learning_rate": 5.231919069907897e-05, "loss": 1.0472, "step": 190550 }, { "epoch": 0.48, "learning_rate": 5.231793245759726e-05, "loss": 1.0453, "step": 190555 }, { "epoch": 0.48, "learning_rate": 5.231667421611556e-05, "loss": 1.0446, "step": 190560 }, { "epoch": 0.48, "learning_rate": 5.231541597463385e-05, "loss": 1.0717, "step": 190565 }, { "epoch": 0.48, "learning_rate": 5.231415773315215e-05, "loss": 1.0448, "step": 190570 }, { "epoch": 0.48, "learning_rate": 5.231289949167044e-05, "loss": 1.0437, "step": 190575 }, { "epoch": 0.48, "learning_rate": 5.231164125018874e-05, "loss": 1.0478, "step": 190580 }, { "epoch": 0.48, "learning_rate": 5.231038300870703e-05, "loss": 1.0461, "step": 190585 }, { "epoch": 0.48, "learning_rate": 5.230912476722532e-05, "loss": 1.0436, "step": 190590 }, { "epoch": 0.48, "learning_rate": 5.230786652574362e-05, "loss": 1.0455, "step": 190595 }, { "epoch": 0.48, "learning_rate": 5.230660828426191e-05, "loss": 1.0468, "step": 190600 }, { "epoch": 0.48, "learning_rate": 5.230535004278021e-05, "loss": 1.0447, "step": 190605 }, { "epoch": 0.48, "learning_rate": 5.23040918012985e-05, "loss": 1.0464, "step": 190610 }, { "epoch": 0.48, "learning_rate": 5.23028335598168e-05, "loss": 1.0447, "step": 190615 }, { "epoch": 0.48, "learning_rate": 5.230157531833509e-05, "loss": 1.0442, "step": 190620 }, { "epoch": 0.48, "learning_rate": 5.230031707685339e-05, "loss": 1.0437, "step": 190625 }, { "epoch": 0.48, "learning_rate": 5.229905883537168e-05, "loss": 1.042, "step": 190630 }, { "epoch": 0.48, "learning_rate": 5.229780059388998e-05, "loss": 1.0435, "step": 190635 }, { "epoch": 0.48, "learning_rate": 5.229654235240827e-05, "loss": 1.0469, "step": 190640 }, { "epoch": 0.48, "learning_rate": 5.229528411092657e-05, "loss": 1.0493, "step": 190645 }, { "epoch": 0.48, "learning_rate": 5.229402586944486e-05, "loss": 1.0425, "step": 190650 }, { "epoch": 0.48, "learning_rate": 5.2292767627963154e-05, "loss": 1.0427, "step": 190655 }, { "epoch": 0.48, "learning_rate": 5.229150938648145e-05, "loss": 1.0453, "step": 190660 }, { "epoch": 0.48, "learning_rate": 5.2290251144999744e-05, "loss": 1.044, "step": 190665 }, { "epoch": 0.48, "learning_rate": 5.228899290351804e-05, "loss": 1.0434, "step": 190670 }, { "epoch": 0.48, "learning_rate": 5.2287734662036334e-05, "loss": 1.0451, "step": 190675 }, { "epoch": 0.48, "learning_rate": 5.228647642055463e-05, "loss": 1.0417, "step": 190680 }, { "epoch": 0.48, "learning_rate": 5.2285218179072924e-05, "loss": 1.0471, "step": 190685 }, { "epoch": 0.48, "learning_rate": 5.228395993759122e-05, "loss": 1.0451, "step": 190690 }, { "epoch": 0.48, "learning_rate": 5.2282701696109514e-05, "loss": 1.0515, "step": 190695 }, { "epoch": 0.48, "learning_rate": 5.228144345462781e-05, "loss": 1.048, "step": 190700 }, { "epoch": 0.48, "learning_rate": 5.2280185213146104e-05, "loss": 1.0456, "step": 190705 }, { "epoch": 0.48, "learning_rate": 5.22789269716644e-05, "loss": 1.045, "step": 190710 }, { "epoch": 0.48, "learning_rate": 5.2277668730182694e-05, "loss": 1.0432, "step": 190715 }, { "epoch": 0.48, "learning_rate": 5.2276410488700986e-05, "loss": 1.0441, "step": 190720 }, { "epoch": 0.48, "learning_rate": 5.2275152247219284e-05, "loss": 1.0444, "step": 190725 }, { "epoch": 0.48, "learning_rate": 5.2273894005737576e-05, "loss": 1.0462, "step": 190730 }, { "epoch": 0.48, "learning_rate": 5.227263576425588e-05, "loss": 1.0413, "step": 190735 }, { "epoch": 0.48, "learning_rate": 5.227137752277418e-05, "loss": 1.0432, "step": 190740 }, { "epoch": 0.48, "learning_rate": 5.227011928129247e-05, "loss": 1.0462, "step": 190745 }, { "epoch": 0.48, "learning_rate": 5.226886103981077e-05, "loss": 1.045, "step": 190750 }, { "epoch": 0.48, "learning_rate": 5.226760279832906e-05, "loss": 1.0465, "step": 190755 }, { "epoch": 0.48, "learning_rate": 5.226634455684736e-05, "loss": 1.0448, "step": 190760 }, { "epoch": 0.48, "learning_rate": 5.226508631536565e-05, "loss": 1.0462, "step": 190765 }, { "epoch": 0.48, "learning_rate": 5.226382807388395e-05, "loss": 1.0474, "step": 190770 }, { "epoch": 0.48, "learning_rate": 5.226256983240224e-05, "loss": 1.043, "step": 190775 }, { "epoch": 0.48, "learning_rate": 5.226131159092054e-05, "loss": 1.0447, "step": 190780 }, { "epoch": 0.48, "learning_rate": 5.226005334943883e-05, "loss": 1.0441, "step": 190785 }, { "epoch": 0.48, "learning_rate": 5.225879510795713e-05, "loss": 1.0449, "step": 190790 }, { "epoch": 0.48, "learning_rate": 5.225753686647542e-05, "loss": 1.0467, "step": 190795 }, { "epoch": 0.48, "learning_rate": 5.225627862499371e-05, "loss": 1.0473, "step": 190800 }, { "epoch": 0.48, "learning_rate": 5.225502038351201e-05, "loss": 1.0461, "step": 190805 }, { "epoch": 0.48, "learning_rate": 5.22537621420303e-05, "loss": 1.0471, "step": 190810 }, { "epoch": 0.48, "learning_rate": 5.22525039005486e-05, "loss": 1.046, "step": 190815 }, { "epoch": 0.48, "learning_rate": 5.225124565906689e-05, "loss": 1.0449, "step": 190820 }, { "epoch": 0.48, "learning_rate": 5.224998741758519e-05, "loss": 1.0465, "step": 190825 }, { "epoch": 0.48, "learning_rate": 5.224872917610348e-05, "loss": 1.0421, "step": 190830 }, { "epoch": 0.48, "learning_rate": 5.224747093462178e-05, "loss": 1.0452, "step": 190835 }, { "epoch": 0.48, "learning_rate": 5.224621269314007e-05, "loss": 1.0475, "step": 190840 }, { "epoch": 0.48, "learning_rate": 5.224495445165837e-05, "loss": 1.0447, "step": 190845 }, { "epoch": 0.48, "learning_rate": 5.224369621017666e-05, "loss": 1.043, "step": 190850 }, { "epoch": 0.48, "learning_rate": 5.224243796869496e-05, "loss": 1.0441, "step": 190855 }, { "epoch": 0.48, "learning_rate": 5.224117972721325e-05, "loss": 1.0456, "step": 190860 }, { "epoch": 0.48, "learning_rate": 5.2239921485731544e-05, "loss": 1.0451, "step": 190865 }, { "epoch": 0.48, "learning_rate": 5.223866324424984e-05, "loss": 1.0456, "step": 190870 }, { "epoch": 0.48, "learning_rate": 5.2237405002768133e-05, "loss": 1.0448, "step": 190875 }, { "epoch": 0.48, "learning_rate": 5.223614676128643e-05, "loss": 1.0422, "step": 190880 }, { "epoch": 0.48, "learning_rate": 5.2234888519804723e-05, "loss": 1.0487, "step": 190885 }, { "epoch": 0.48, "learning_rate": 5.223363027832302e-05, "loss": 1.0469, "step": 190890 }, { "epoch": 0.48, "learning_rate": 5.2232372036841313e-05, "loss": 1.0472, "step": 190895 }, { "epoch": 0.48, "learning_rate": 5.223111379535961e-05, "loss": 1.0442, "step": 190900 }, { "epoch": 0.48, "learning_rate": 5.2229855553877903e-05, "loss": 1.0461, "step": 190905 }, { "epoch": 0.48, "learning_rate": 5.22285973123962e-05, "loss": 1.0432, "step": 190910 }, { "epoch": 0.48, "learning_rate": 5.222733907091449e-05, "loss": 1.0431, "step": 190915 }, { "epoch": 0.48, "learning_rate": 5.222608082943279e-05, "loss": 1.0449, "step": 190920 }, { "epoch": 0.48, "learning_rate": 5.222482258795108e-05, "loss": 1.0461, "step": 190925 }, { "epoch": 0.48, "learning_rate": 5.2223564346469375e-05, "loss": 1.0419, "step": 190930 }, { "epoch": 0.48, "learning_rate": 5.222230610498767e-05, "loss": 1.0477, "step": 190935 }, { "epoch": 0.48, "learning_rate": 5.2221047863505965e-05, "loss": 1.0573, "step": 190940 }, { "epoch": 0.48, "learning_rate": 5.221978962202426e-05, "loss": 1.0436, "step": 190945 }, { "epoch": 0.48, "learning_rate": 5.2218531380542555e-05, "loss": 1.0429, "step": 190950 }, { "epoch": 0.48, "learning_rate": 5.221727313906085e-05, "loss": 1.044, "step": 190955 }, { "epoch": 0.48, "learning_rate": 5.2216014897579145e-05, "loss": 1.045, "step": 190960 }, { "epoch": 0.48, "learning_rate": 5.221475665609744e-05, "loss": 1.0458, "step": 190965 }, { "epoch": 0.48, "learning_rate": 5.2213498414615735e-05, "loss": 1.0414, "step": 190970 }, { "epoch": 0.48, "learning_rate": 5.221224017313403e-05, "loss": 1.0457, "step": 190975 }, { "epoch": 0.48, "learning_rate": 5.2210981931652325e-05, "loss": 1.0453, "step": 190980 }, { "epoch": 0.48, "learning_rate": 5.220972369017062e-05, "loss": 1.0422, "step": 190985 }, { "epoch": 0.48, "learning_rate": 5.2208465448688915e-05, "loss": 1.0457, "step": 190990 }, { "epoch": 0.48, "learning_rate": 5.2207207207207206e-05, "loss": 1.0704, "step": 190995 }, { "epoch": 0.48, "learning_rate": 5.2205948965725505e-05, "loss": 1.0478, "step": 191000 }, { "epoch": 0.48, "learning_rate": 5.2204690724243796e-05, "loss": 1.0497, "step": 191005 }, { "epoch": 0.48, "learning_rate": 5.2203432482762095e-05, "loss": 1.0449, "step": 191010 }, { "epoch": 0.48, "learning_rate": 5.2202174241280386e-05, "loss": 1.0425, "step": 191015 }, { "epoch": 0.48, "learning_rate": 5.2200915999798685e-05, "loss": 1.0463, "step": 191020 }, { "epoch": 0.48, "learning_rate": 5.2199657758316976e-05, "loss": 1.0439, "step": 191025 }, { "epoch": 0.48, "learning_rate": 5.2198399516835275e-05, "loss": 1.0414, "step": 191030 }, { "epoch": 0.48, "learning_rate": 5.2197141275353566e-05, "loss": 1.045, "step": 191035 }, { "epoch": 0.48, "learning_rate": 5.2195883033871865e-05, "loss": 1.0456, "step": 191040 }, { "epoch": 0.48, "learning_rate": 5.2194624792390156e-05, "loss": 1.0448, "step": 191045 }, { "epoch": 0.48, "learning_rate": 5.2193366550908455e-05, "loss": 1.0423, "step": 191050 }, { "epoch": 0.48, "learning_rate": 5.2192108309426746e-05, "loss": 1.0459, "step": 191055 }, { "epoch": 0.48, "learning_rate": 5.219085006794504e-05, "loss": 1.0455, "step": 191060 }, { "epoch": 0.48, "learning_rate": 5.2189591826463336e-05, "loss": 1.0465, "step": 191065 }, { "epoch": 0.48, "learning_rate": 5.218833358498163e-05, "loss": 1.0458, "step": 191070 }, { "epoch": 0.48, "learning_rate": 5.2187075343499926e-05, "loss": 1.0459, "step": 191075 }, { "epoch": 0.48, "learning_rate": 5.218581710201822e-05, "loss": 1.047, "step": 191080 }, { "epoch": 0.48, "learning_rate": 5.2184558860536516e-05, "loss": 1.0448, "step": 191085 }, { "epoch": 0.48, "learning_rate": 5.218330061905481e-05, "loss": 1.0428, "step": 191090 }, { "epoch": 0.48, "learning_rate": 5.2182042377573106e-05, "loss": 1.0443, "step": 191095 }, { "epoch": 0.48, "learning_rate": 5.21807841360914e-05, "loss": 1.0494, "step": 191100 }, { "epoch": 0.48, "learning_rate": 5.2179525894609696e-05, "loss": 1.0437, "step": 191105 }, { "epoch": 0.48, "learning_rate": 5.217826765312799e-05, "loss": 1.0458, "step": 191110 }, { "epoch": 0.48, "learning_rate": 5.2177009411646286e-05, "loss": 1.0459, "step": 191115 }, { "epoch": 0.48, "learning_rate": 5.217575117016458e-05, "loss": 1.0456, "step": 191120 }, { "epoch": 0.48, "learning_rate": 5.217449292868287e-05, "loss": 1.0444, "step": 191125 }, { "epoch": 0.48, "learning_rate": 5.217323468720117e-05, "loss": 1.0435, "step": 191130 }, { "epoch": 0.48, "learning_rate": 5.217197644571946e-05, "loss": 1.0437, "step": 191135 }, { "epoch": 0.48, "learning_rate": 5.217071820423776e-05, "loss": 1.0421, "step": 191140 }, { "epoch": 0.48, "learning_rate": 5.216945996275605e-05, "loss": 1.0444, "step": 191145 }, { "epoch": 0.48, "learning_rate": 5.216820172127435e-05, "loss": 1.0461, "step": 191150 }, { "epoch": 0.48, "learning_rate": 5.216694347979264e-05, "loss": 1.0471, "step": 191155 }, { "epoch": 0.48, "learning_rate": 5.216568523831094e-05, "loss": 1.0454, "step": 191160 }, { "epoch": 0.48, "learning_rate": 5.216442699682923e-05, "loss": 1.0448, "step": 191165 }, { "epoch": 0.48, "learning_rate": 5.216316875534753e-05, "loss": 1.0474, "step": 191170 }, { "epoch": 0.48, "learning_rate": 5.216191051386582e-05, "loss": 1.0459, "step": 191175 }, { "epoch": 0.48, "learning_rate": 5.216065227238411e-05, "loss": 1.0449, "step": 191180 }, { "epoch": 0.48, "learning_rate": 5.215939403090241e-05, "loss": 1.0446, "step": 191185 }, { "epoch": 0.48, "learning_rate": 5.21581357894207e-05, "loss": 1.0478, "step": 191190 }, { "epoch": 0.48, "learning_rate": 5.2156877547939e-05, "loss": 1.0473, "step": 191195 }, { "epoch": 0.48, "learning_rate": 5.215561930645729e-05, "loss": 1.0457, "step": 191200 }, { "epoch": 0.48, "learning_rate": 5.215436106497559e-05, "loss": 1.0457, "step": 191205 }, { "epoch": 0.48, "learning_rate": 5.215310282349388e-05, "loss": 1.043, "step": 191210 }, { "epoch": 0.48, "learning_rate": 5.215184458201218e-05, "loss": 1.0461, "step": 191215 }, { "epoch": 0.48, "learning_rate": 5.215058634053047e-05, "loss": 1.0452, "step": 191220 }, { "epoch": 0.48, "learning_rate": 5.214932809904877e-05, "loss": 1.0878, "step": 191225 }, { "epoch": 0.48, "learning_rate": 5.2148069857567074e-05, "loss": 1.0428, "step": 191230 }, { "epoch": 0.48, "learning_rate": 5.2146811616085366e-05, "loss": 1.0432, "step": 191235 }, { "epoch": 0.48, "learning_rate": 5.2145553374603664e-05, "loss": 1.0438, "step": 191240 }, { "epoch": 0.48, "learning_rate": 5.2144295133121956e-05, "loss": 1.0478, "step": 191245 }, { "epoch": 0.48, "learning_rate": 5.2143036891640254e-05, "loss": 1.0487, "step": 191250 }, { "epoch": 0.48, "learning_rate": 5.2141778650158546e-05, "loss": 1.0468, "step": 191255 }, { "epoch": 0.48, "learning_rate": 5.2140520408676844e-05, "loss": 1.0475, "step": 191260 }, { "epoch": 0.48, "learning_rate": 5.2139262167195136e-05, "loss": 1.0453, "step": 191265 }, { "epoch": 0.48, "learning_rate": 5.213800392571343e-05, "loss": 1.0454, "step": 191270 }, { "epoch": 0.48, "learning_rate": 5.2136745684231726e-05, "loss": 1.0455, "step": 191275 }, { "epoch": 0.48, "learning_rate": 5.213548744275002e-05, "loss": 1.0436, "step": 191280 }, { "epoch": 0.48, "learning_rate": 5.2134229201268316e-05, "loss": 1.0444, "step": 191285 }, { "epoch": 0.48, "learning_rate": 5.213297095978661e-05, "loss": 1.0441, "step": 191290 }, { "epoch": 0.48, "learning_rate": 5.2131712718304905e-05, "loss": 1.0469, "step": 191295 }, { "epoch": 0.48, "learning_rate": 5.21304544768232e-05, "loss": 1.0419, "step": 191300 }, { "epoch": 0.48, "learning_rate": 5.2129196235341495e-05, "loss": 1.0461, "step": 191305 }, { "epoch": 0.48, "learning_rate": 5.212793799385979e-05, "loss": 1.0452, "step": 191310 }, { "epoch": 0.48, "learning_rate": 5.2126679752378085e-05, "loss": 1.0463, "step": 191315 }, { "epoch": 0.48, "learning_rate": 5.212542151089638e-05, "loss": 1.0444, "step": 191320 }, { "epoch": 0.48, "learning_rate": 5.212416326941467e-05, "loss": 1.0463, "step": 191325 }, { "epoch": 0.48, "learning_rate": 5.212290502793297e-05, "loss": 1.0453, "step": 191330 }, { "epoch": 0.48, "learning_rate": 5.212164678645126e-05, "loss": 1.0466, "step": 191335 }, { "epoch": 0.48, "learning_rate": 5.212038854496956e-05, "loss": 1.0449, "step": 191340 }, { "epoch": 0.48, "learning_rate": 5.211913030348785e-05, "loss": 1.044, "step": 191345 }, { "epoch": 0.48, "learning_rate": 5.211787206200615e-05, "loss": 1.0433, "step": 191350 }, { "epoch": 0.48, "learning_rate": 5.211661382052444e-05, "loss": 1.0442, "step": 191355 }, { "epoch": 0.48, "learning_rate": 5.211535557904274e-05, "loss": 1.0478, "step": 191360 }, { "epoch": 0.48, "learning_rate": 5.211409733756103e-05, "loss": 1.0438, "step": 191365 }, { "epoch": 0.48, "learning_rate": 5.211283909607933e-05, "loss": 1.0462, "step": 191370 }, { "epoch": 0.48, "learning_rate": 5.211158085459762e-05, "loss": 1.045, "step": 191375 }, { "epoch": 0.48, "learning_rate": 5.211032261311592e-05, "loss": 1.0436, "step": 191380 }, { "epoch": 0.48, "learning_rate": 5.210906437163421e-05, "loss": 1.0462, "step": 191385 }, { "epoch": 0.48, "learning_rate": 5.21078061301525e-05, "loss": 1.0458, "step": 191390 }, { "epoch": 0.48, "learning_rate": 5.21065478886708e-05, "loss": 1.0424, "step": 191395 }, { "epoch": 0.48, "learning_rate": 5.210528964718909e-05, "loss": 1.0453, "step": 191400 }, { "epoch": 0.48, "learning_rate": 5.210403140570739e-05, "loss": 1.0474, "step": 191405 }, { "epoch": 0.48, "learning_rate": 5.210277316422568e-05, "loss": 1.0457, "step": 191410 }, { "epoch": 0.48, "learning_rate": 5.210151492274398e-05, "loss": 1.0456, "step": 191415 }, { "epoch": 0.48, "learning_rate": 5.210025668126227e-05, "loss": 1.0666, "step": 191420 }, { "epoch": 0.48, "learning_rate": 5.209899843978057e-05, "loss": 1.0438, "step": 191425 }, { "epoch": 0.48, "learning_rate": 5.209774019829886e-05, "loss": 1.0436, "step": 191430 }, { "epoch": 0.48, "learning_rate": 5.209648195681716e-05, "loss": 1.0438, "step": 191435 }, { "epoch": 0.48, "learning_rate": 5.209522371533545e-05, "loss": 1.0586, "step": 191440 }, { "epoch": 0.48, "learning_rate": 5.209396547385375e-05, "loss": 1.0416, "step": 191445 }, { "epoch": 0.48, "learning_rate": 5.209270723237204e-05, "loss": 1.0433, "step": 191450 }, { "epoch": 0.48, "learning_rate": 5.209144899089033e-05, "loss": 1.0485, "step": 191455 }, { "epoch": 0.48, "learning_rate": 5.209019074940863e-05, "loss": 1.0463, "step": 191460 }, { "epoch": 0.48, "learning_rate": 5.208893250792692e-05, "loss": 1.0466, "step": 191465 }, { "epoch": 0.48, "learning_rate": 5.208767426644522e-05, "loss": 1.0436, "step": 191470 }, { "epoch": 0.48, "learning_rate": 5.208641602496351e-05, "loss": 1.0446, "step": 191475 }, { "epoch": 0.48, "learning_rate": 5.208515778348181e-05, "loss": 1.044, "step": 191480 }, { "epoch": 0.48, "learning_rate": 5.20838995420001e-05, "loss": 1.0467, "step": 191485 }, { "epoch": 0.48, "learning_rate": 5.20826413005184e-05, "loss": 1.0451, "step": 191490 }, { "epoch": 0.48, "learning_rate": 5.208138305903669e-05, "loss": 1.0447, "step": 191495 }, { "epoch": 0.48, "learning_rate": 5.208012481755499e-05, "loss": 1.0449, "step": 191500 }, { "epoch": 0.48, "learning_rate": 5.207886657607328e-05, "loss": 1.0422, "step": 191505 }, { "epoch": 0.48, "learning_rate": 5.207760833459158e-05, "loss": 1.0464, "step": 191510 }, { "epoch": 0.48, "learning_rate": 5.207635009310987e-05, "loss": 1.0463, "step": 191515 }, { "epoch": 0.48, "learning_rate": 5.207509185162816e-05, "loss": 1.0479, "step": 191520 }, { "epoch": 0.48, "learning_rate": 5.207383361014646e-05, "loss": 1.0471, "step": 191525 }, { "epoch": 0.48, "learning_rate": 5.207257536866475e-05, "loss": 1.0439, "step": 191530 }, { "epoch": 0.48, "learning_rate": 5.207131712718305e-05, "loss": 1.0423, "step": 191535 }, { "epoch": 0.48, "learning_rate": 5.207005888570134e-05, "loss": 1.0469, "step": 191540 }, { "epoch": 0.48, "learning_rate": 5.206880064421964e-05, "loss": 1.046, "step": 191545 }, { "epoch": 0.48, "learning_rate": 5.206754240273793e-05, "loss": 1.0458, "step": 191550 }, { "epoch": 0.48, "learning_rate": 5.206628416125623e-05, "loss": 1.0681, "step": 191555 }, { "epoch": 0.48, "learning_rate": 5.206502591977452e-05, "loss": 1.0488, "step": 191560 }, { "epoch": 0.48, "learning_rate": 5.206376767829282e-05, "loss": 1.0476, "step": 191565 }, { "epoch": 0.48, "learning_rate": 5.206250943681111e-05, "loss": 1.0439, "step": 191570 }, { "epoch": 0.48, "learning_rate": 5.206125119532941e-05, "loss": 1.0474, "step": 191575 }, { "epoch": 0.48, "learning_rate": 5.20599929538477e-05, "loss": 1.0447, "step": 191580 }, { "epoch": 0.48, "learning_rate": 5.2058734712365994e-05, "loss": 1.0445, "step": 191585 }, { "epoch": 0.48, "learning_rate": 5.205747647088429e-05, "loss": 1.0447, "step": 191590 }, { "epoch": 0.48, "learning_rate": 5.2056218229402584e-05, "loss": 1.044, "step": 191595 }, { "epoch": 0.48, "learning_rate": 5.205495998792088e-05, "loss": 1.049, "step": 191600 }, { "epoch": 0.48, "learning_rate": 5.2053701746439174e-05, "loss": 1.045, "step": 191605 }, { "epoch": 0.48, "learning_rate": 5.205244350495747e-05, "loss": 1.0461, "step": 191610 }, { "epoch": 0.48, "learning_rate": 5.2051185263475764e-05, "loss": 1.0474, "step": 191615 }, { "epoch": 0.48, "learning_rate": 5.204992702199406e-05, "loss": 1.0458, "step": 191620 }, { "epoch": 0.48, "learning_rate": 5.2048668780512354e-05, "loss": 1.0441, "step": 191625 }, { "epoch": 0.48, "learning_rate": 5.204741053903065e-05, "loss": 1.047, "step": 191630 }, { "epoch": 0.48, "learning_rate": 5.2046152297548944e-05, "loss": 1.0427, "step": 191635 }, { "epoch": 0.48, "learning_rate": 5.204489405606724e-05, "loss": 1.045, "step": 191640 }, { "epoch": 0.48, "learning_rate": 5.2043635814585534e-05, "loss": 1.0474, "step": 191645 }, { "epoch": 0.48, "learning_rate": 5.2042377573103826e-05, "loss": 1.0467, "step": 191650 }, { "epoch": 0.48, "learning_rate": 5.2041119331622124e-05, "loss": 1.0422, "step": 191655 }, { "epoch": 0.48, "learning_rate": 5.2039861090140416e-05, "loss": 1.043, "step": 191660 }, { "epoch": 0.48, "learning_rate": 5.2038602848658714e-05, "loss": 1.046, "step": 191665 }, { "epoch": 0.48, "learning_rate": 5.2037344607177006e-05, "loss": 1.0478, "step": 191670 }, { "epoch": 0.48, "learning_rate": 5.2036086365695304e-05, "loss": 1.048, "step": 191675 }, { "epoch": 0.48, "learning_rate": 5.2034828124213596e-05, "loss": 1.0436, "step": 191680 }, { "epoch": 0.48, "learning_rate": 5.2033569882731894e-05, "loss": 1.0441, "step": 191685 }, { "epoch": 0.48, "learning_rate": 5.2032311641250186e-05, "loss": 1.0444, "step": 191690 }, { "epoch": 0.48, "learning_rate": 5.2031053399768484e-05, "loss": 1.0428, "step": 191695 }, { "epoch": 0.48, "learning_rate": 5.2029795158286776e-05, "loss": 1.0606, "step": 191700 }, { "epoch": 0.48, "learning_rate": 5.2028536916805074e-05, "loss": 1.043, "step": 191705 }, { "epoch": 0.48, "learning_rate": 5.2027278675323366e-05, "loss": 1.0438, "step": 191710 }, { "epoch": 0.48, "learning_rate": 5.202602043384166e-05, "loss": 1.0433, "step": 191715 }, { "epoch": 0.48, "learning_rate": 5.2024762192359956e-05, "loss": 1.0464, "step": 191720 }, { "epoch": 0.48, "learning_rate": 5.202350395087825e-05, "loss": 1.0448, "step": 191725 }, { "epoch": 0.48, "learning_rate": 5.202224570939655e-05, "loss": 1.0424, "step": 191730 }, { "epoch": 0.48, "learning_rate": 5.202098746791485e-05, "loss": 1.0456, "step": 191735 }, { "epoch": 0.48, "learning_rate": 5.201972922643314e-05, "loss": 1.0423, "step": 191740 }, { "epoch": 0.48, "learning_rate": 5.201847098495144e-05, "loss": 1.0463, "step": 191745 }, { "epoch": 0.48, "learning_rate": 5.201721274346973e-05, "loss": 1.0451, "step": 191750 }, { "epoch": 0.48, "learning_rate": 5.201595450198803e-05, "loss": 1.0441, "step": 191755 }, { "epoch": 0.48, "learning_rate": 5.201469626050632e-05, "loss": 1.0439, "step": 191760 }, { "epoch": 0.48, "learning_rate": 5.201343801902462e-05, "loss": 1.0473, "step": 191765 }, { "epoch": 0.48, "learning_rate": 5.201217977754291e-05, "loss": 1.0436, "step": 191770 }, { "epoch": 0.48, "learning_rate": 5.201092153606121e-05, "loss": 1.0453, "step": 191775 }, { "epoch": 0.48, "learning_rate": 5.20096632945795e-05, "loss": 1.0425, "step": 191780 }, { "epoch": 0.48, "learning_rate": 5.20084050530978e-05, "loss": 1.0459, "step": 191785 }, { "epoch": 0.48, "learning_rate": 5.200714681161609e-05, "loss": 1.0452, "step": 191790 }, { "epoch": 0.48, "learning_rate": 5.2005888570134384e-05, "loss": 1.0452, "step": 191795 }, { "epoch": 0.48, "learning_rate": 5.200463032865268e-05, "loss": 1.0462, "step": 191800 }, { "epoch": 0.48, "learning_rate": 5.2003372087170974e-05, "loss": 1.045, "step": 191805 }, { "epoch": 0.48, "learning_rate": 5.200211384568927e-05, "loss": 1.0428, "step": 191810 }, { "epoch": 0.48, "learning_rate": 5.2000855604207564e-05, "loss": 1.0443, "step": 191815 }, { "epoch": 0.48, "learning_rate": 5.199959736272586e-05, "loss": 1.0691, "step": 191820 }, { "epoch": 0.48, "learning_rate": 5.1998339121244154e-05, "loss": 1.0482, "step": 191825 }, { "epoch": 0.48, "learning_rate": 5.199708087976245e-05, "loss": 1.0431, "step": 191830 }, { "epoch": 0.48, "learning_rate": 5.1995822638280744e-05, "loss": 1.044, "step": 191835 }, { "epoch": 0.48, "learning_rate": 5.199456439679904e-05, "loss": 1.0451, "step": 191840 }, { "epoch": 0.48, "learning_rate": 5.1993306155317334e-05, "loss": 1.0447, "step": 191845 }, { "epoch": 0.48, "learning_rate": 5.199204791383563e-05, "loss": 1.0479, "step": 191850 }, { "epoch": 0.48, "learning_rate": 5.1990789672353924e-05, "loss": 1.0439, "step": 191855 }, { "epoch": 0.48, "learning_rate": 5.1989531430872215e-05, "loss": 1.0442, "step": 191860 }, { "epoch": 0.48, "learning_rate": 5.1988273189390514e-05, "loss": 1.0447, "step": 191865 }, { "epoch": 0.48, "learning_rate": 5.1987014947908805e-05, "loss": 1.0481, "step": 191870 }, { "epoch": 0.48, "learning_rate": 5.1985756706427103e-05, "loss": 1.0415, "step": 191875 }, { "epoch": 0.48, "learning_rate": 5.1984498464945395e-05, "loss": 1.0446, "step": 191880 }, { "epoch": 0.48, "learning_rate": 5.1983240223463693e-05, "loss": 1.0468, "step": 191885 }, { "epoch": 0.48, "learning_rate": 5.1981981981981985e-05, "loss": 1.0442, "step": 191890 }, { "epoch": 0.48, "learning_rate": 5.1980723740500283e-05, "loss": 1.0461, "step": 191895 }, { "epoch": 0.48, "learning_rate": 5.1979465499018575e-05, "loss": 1.046, "step": 191900 }, { "epoch": 0.48, "learning_rate": 5.1978207257536873e-05, "loss": 1.0415, "step": 191905 }, { "epoch": 0.48, "learning_rate": 5.1976949016055165e-05, "loss": 1.0466, "step": 191910 }, { "epoch": 0.48, "learning_rate": 5.197569077457346e-05, "loss": 1.0432, "step": 191915 }, { "epoch": 0.48, "learning_rate": 5.1974432533091755e-05, "loss": 1.0434, "step": 191920 }, { "epoch": 0.48, "learning_rate": 5.1973174291610047e-05, "loss": 1.0644, "step": 191925 }, { "epoch": 0.48, "learning_rate": 5.1971916050128345e-05, "loss": 1.0671, "step": 191930 }, { "epoch": 0.48, "learning_rate": 5.1970657808646637e-05, "loss": 1.0459, "step": 191935 }, { "epoch": 0.48, "learning_rate": 5.1969399567164935e-05, "loss": 1.0472, "step": 191940 }, { "epoch": 0.48, "learning_rate": 5.1968141325683227e-05, "loss": 1.0431, "step": 191945 }, { "epoch": 0.48, "learning_rate": 5.1966883084201525e-05, "loss": 1.0433, "step": 191950 }, { "epoch": 0.48, "learning_rate": 5.1965624842719816e-05, "loss": 1.0475, "step": 191955 }, { "epoch": 0.48, "learning_rate": 5.1964366601238115e-05, "loss": 1.0438, "step": 191960 }, { "epoch": 0.48, "learning_rate": 5.1963108359756406e-05, "loss": 1.0454, "step": 191965 }, { "epoch": 0.48, "learning_rate": 5.1961850118274705e-05, "loss": 1.046, "step": 191970 }, { "epoch": 0.48, "learning_rate": 5.1960591876792996e-05, "loss": 1.0453, "step": 191975 }, { "epoch": 0.48, "learning_rate": 5.195933363531129e-05, "loss": 1.0466, "step": 191980 }, { "epoch": 0.48, "learning_rate": 5.1958075393829586e-05, "loss": 1.044, "step": 191985 }, { "epoch": 0.48, "learning_rate": 5.195681715234788e-05, "loss": 1.044, "step": 191990 }, { "epoch": 0.48, "learning_rate": 5.1955558910866176e-05, "loss": 1.0464, "step": 191995 }, { "epoch": 0.48, "learning_rate": 5.195430066938447e-05, "loss": 1.0464, "step": 192000 }, { "epoch": 0.48, "learning_rate": 5.1953042427902766e-05, "loss": 1.0458, "step": 192005 }, { "epoch": 0.48, "learning_rate": 5.195178418642106e-05, "loss": 1.0434, "step": 192010 }, { "epoch": 0.48, "learning_rate": 5.1950525944939356e-05, "loss": 1.0478, "step": 192015 }, { "epoch": 0.48, "learning_rate": 5.194926770345765e-05, "loss": 1.0451, "step": 192020 }, { "epoch": 0.48, "learning_rate": 5.1948009461975946e-05, "loss": 1.0455, "step": 192025 }, { "epoch": 0.48, "learning_rate": 5.194675122049424e-05, "loss": 1.0652, "step": 192030 }, { "epoch": 0.48, "learning_rate": 5.1945492979012536e-05, "loss": 1.0449, "step": 192035 }, { "epoch": 0.48, "learning_rate": 5.194423473753083e-05, "loss": 1.0424, "step": 192040 }, { "epoch": 0.48, "learning_rate": 5.194297649604912e-05, "loss": 1.045, "step": 192045 }, { "epoch": 0.48, "learning_rate": 5.194171825456742e-05, "loss": 1.0459, "step": 192050 }, { "epoch": 0.48, "learning_rate": 5.194046001308571e-05, "loss": 1.0441, "step": 192055 }, { "epoch": 0.48, "learning_rate": 5.193920177160401e-05, "loss": 1.0447, "step": 192060 }, { "epoch": 0.48, "learning_rate": 5.19379435301223e-05, "loss": 1.0436, "step": 192065 }, { "epoch": 0.48, "learning_rate": 5.19366852886406e-05, "loss": 1.0454, "step": 192070 }, { "epoch": 0.48, "learning_rate": 5.193542704715889e-05, "loss": 1.0422, "step": 192075 }, { "epoch": 0.48, "learning_rate": 5.193416880567719e-05, "loss": 1.0449, "step": 192080 }, { "epoch": 0.48, "learning_rate": 5.193291056419548e-05, "loss": 1.0458, "step": 192085 }, { "epoch": 0.48, "learning_rate": 5.193165232271378e-05, "loss": 1.0468, "step": 192090 }, { "epoch": 0.48, "learning_rate": 5.193039408123207e-05, "loss": 1.0462, "step": 192095 }, { "epoch": 0.48, "learning_rate": 5.192913583975037e-05, "loss": 1.0452, "step": 192100 }, { "epoch": 0.48, "learning_rate": 5.192787759826866e-05, "loss": 1.0459, "step": 192105 }, { "epoch": 0.48, "learning_rate": 5.192661935678695e-05, "loss": 1.0442, "step": 192110 }, { "epoch": 0.48, "learning_rate": 5.192536111530525e-05, "loss": 1.0451, "step": 192115 }, { "epoch": 0.48, "learning_rate": 5.192410287382354e-05, "loss": 1.0456, "step": 192120 }, { "epoch": 0.48, "learning_rate": 5.192284463234184e-05, "loss": 1.046, "step": 192125 }, { "epoch": 0.48, "learning_rate": 5.192158639086013e-05, "loss": 1.0418, "step": 192130 }, { "epoch": 0.48, "learning_rate": 5.192032814937843e-05, "loss": 1.0449, "step": 192135 }, { "epoch": 0.48, "learning_rate": 5.191906990789672e-05, "loss": 1.0482, "step": 192140 }, { "epoch": 0.48, "learning_rate": 5.191781166641502e-05, "loss": 1.0458, "step": 192145 }, { "epoch": 0.48, "learning_rate": 5.191655342493331e-05, "loss": 1.0465, "step": 192150 }, { "epoch": 0.48, "learning_rate": 5.191529518345161e-05, "loss": 1.0437, "step": 192155 }, { "epoch": 0.48, "learning_rate": 5.19140369419699e-05, "loss": 1.0453, "step": 192160 }, { "epoch": 0.48, "learning_rate": 5.19127787004882e-05, "loss": 1.0444, "step": 192165 }, { "epoch": 0.48, "learning_rate": 5.191152045900649e-05, "loss": 1.0481, "step": 192170 }, { "epoch": 0.48, "learning_rate": 5.191026221752478e-05, "loss": 1.0419, "step": 192175 }, { "epoch": 0.48, "learning_rate": 5.190900397604308e-05, "loss": 1.0449, "step": 192180 }, { "epoch": 0.48, "learning_rate": 5.190774573456137e-05, "loss": 1.0435, "step": 192185 }, { "epoch": 0.48, "learning_rate": 5.190648749307967e-05, "loss": 1.0407, "step": 192190 }, { "epoch": 0.48, "learning_rate": 5.190522925159796e-05, "loss": 1.0468, "step": 192195 }, { "epoch": 0.48, "learning_rate": 5.190397101011626e-05, "loss": 1.0452, "step": 192200 }, { "epoch": 0.48, "learning_rate": 5.190271276863455e-05, "loss": 1.0604, "step": 192205 }, { "epoch": 0.48, "learning_rate": 5.190145452715285e-05, "loss": 1.0442, "step": 192210 }, { "epoch": 0.48, "learning_rate": 5.190019628567114e-05, "loss": 1.0479, "step": 192215 }, { "epoch": 0.48, "learning_rate": 5.189893804418944e-05, "loss": 1.0436, "step": 192220 }, { "epoch": 0.48, "learning_rate": 5.189767980270773e-05, "loss": 1.0441, "step": 192225 }, { "epoch": 0.48, "learning_rate": 5.189642156122604e-05, "loss": 1.0421, "step": 192230 }, { "epoch": 0.48, "learning_rate": 5.1895163319744336e-05, "loss": 1.0503, "step": 192235 }, { "epoch": 0.48, "learning_rate": 5.189390507826263e-05, "loss": 1.0446, "step": 192240 }, { "epoch": 0.48, "learning_rate": 5.1892646836780926e-05, "loss": 1.0445, "step": 192245 }, { "epoch": 0.48, "learning_rate": 5.189138859529922e-05, "loss": 1.0435, "step": 192250 }, { "epoch": 0.48, "learning_rate": 5.189013035381751e-05, "loss": 1.0603, "step": 192255 }, { "epoch": 0.48, "learning_rate": 5.188887211233581e-05, "loss": 1.0434, "step": 192260 }, { "epoch": 0.48, "learning_rate": 5.18876138708541e-05, "loss": 1.0421, "step": 192265 }, { "epoch": 0.48, "learning_rate": 5.18863556293724e-05, "loss": 1.0483, "step": 192270 }, { "epoch": 0.48, "learning_rate": 5.188509738789069e-05, "loss": 1.0474, "step": 192275 }, { "epoch": 0.48, "learning_rate": 5.188383914640899e-05, "loss": 1.0452, "step": 192280 }, { "epoch": 0.48, "learning_rate": 5.188258090492728e-05, "loss": 1.0475, "step": 192285 }, { "epoch": 0.48, "learning_rate": 5.188132266344558e-05, "loss": 1.0449, "step": 192290 }, { "epoch": 0.48, "learning_rate": 5.188006442196387e-05, "loss": 1.0445, "step": 192295 }, { "epoch": 0.48, "learning_rate": 5.187880618048217e-05, "loss": 1.0465, "step": 192300 }, { "epoch": 0.48, "learning_rate": 5.187754793900046e-05, "loss": 1.0453, "step": 192305 }, { "epoch": 0.48, "learning_rate": 5.187628969751876e-05, "loss": 1.0442, "step": 192310 }, { "epoch": 0.48, "learning_rate": 5.187503145603705e-05, "loss": 1.042, "step": 192315 }, { "epoch": 0.48, "learning_rate": 5.187377321455534e-05, "loss": 1.0462, "step": 192320 }, { "epoch": 0.48, "learning_rate": 5.187251497307364e-05, "loss": 1.0449, "step": 192325 }, { "epoch": 0.48, "learning_rate": 5.187125673159193e-05, "loss": 1.0441, "step": 192330 }, { "epoch": 0.48, "learning_rate": 5.186999849011023e-05, "loss": 1.0458, "step": 192335 }, { "epoch": 0.48, "learning_rate": 5.186874024862852e-05, "loss": 1.0445, "step": 192340 }, { "epoch": 0.48, "learning_rate": 5.186748200714682e-05, "loss": 1.0443, "step": 192345 }, { "epoch": 0.48, "learning_rate": 5.186622376566511e-05, "loss": 1.0474, "step": 192350 }, { "epoch": 0.48, "learning_rate": 5.186496552418341e-05, "loss": 1.0418, "step": 192355 }, { "epoch": 0.48, "learning_rate": 5.18637072827017e-05, "loss": 1.0462, "step": 192360 }, { "epoch": 0.48, "learning_rate": 5.186244904122e-05, "loss": 1.0453, "step": 192365 }, { "epoch": 0.48, "learning_rate": 5.186119079973829e-05, "loss": 1.048, "step": 192370 }, { "epoch": 0.48, "learning_rate": 5.185993255825659e-05, "loss": 1.0418, "step": 192375 }, { "epoch": 0.48, "learning_rate": 5.185867431677488e-05, "loss": 1.0448, "step": 192380 }, { "epoch": 0.48, "learning_rate": 5.185741607529317e-05, "loss": 1.0442, "step": 192385 }, { "epoch": 0.48, "learning_rate": 5.185615783381147e-05, "loss": 1.0436, "step": 192390 }, { "epoch": 0.48, "learning_rate": 5.185489959232976e-05, "loss": 1.0448, "step": 192395 }, { "epoch": 0.48, "learning_rate": 5.185364135084806e-05, "loss": 1.047, "step": 192400 }, { "epoch": 0.48, "learning_rate": 5.185238310936635e-05, "loss": 1.046, "step": 192405 }, { "epoch": 0.48, "learning_rate": 5.185112486788465e-05, "loss": 1.0467, "step": 192410 }, { "epoch": 0.48, "learning_rate": 5.184986662640294e-05, "loss": 1.044, "step": 192415 }, { "epoch": 0.48, "learning_rate": 5.184860838492124e-05, "loss": 1.0459, "step": 192420 }, { "epoch": 0.48, "learning_rate": 5.184735014343953e-05, "loss": 1.0444, "step": 192425 }, { "epoch": 0.48, "learning_rate": 5.184609190195783e-05, "loss": 1.0438, "step": 192430 }, { "epoch": 0.48, "learning_rate": 5.184483366047612e-05, "loss": 1.0441, "step": 192435 }, { "epoch": 0.48, "learning_rate": 5.184357541899442e-05, "loss": 1.0424, "step": 192440 }, { "epoch": 0.48, "learning_rate": 5.184231717751271e-05, "loss": 1.0448, "step": 192445 }, { "epoch": 0.48, "learning_rate": 5.1841058936031e-05, "loss": 1.0451, "step": 192450 }, { "epoch": 0.48, "learning_rate": 5.18398006945493e-05, "loss": 1.0454, "step": 192455 }, { "epoch": 0.48, "learning_rate": 5.183854245306759e-05, "loss": 1.0442, "step": 192460 }, { "epoch": 0.48, "learning_rate": 5.183728421158589e-05, "loss": 1.0447, "step": 192465 }, { "epoch": 0.48, "learning_rate": 5.183602597010418e-05, "loss": 1.0482, "step": 192470 }, { "epoch": 0.48, "learning_rate": 5.183476772862248e-05, "loss": 1.0439, "step": 192475 }, { "epoch": 0.48, "learning_rate": 5.183350948714077e-05, "loss": 1.0459, "step": 192480 }, { "epoch": 0.48, "learning_rate": 5.183225124565907e-05, "loss": 1.0448, "step": 192485 }, { "epoch": 0.48, "learning_rate": 5.183099300417736e-05, "loss": 1.0466, "step": 192490 }, { "epoch": 0.48, "learning_rate": 5.182973476269566e-05, "loss": 1.0424, "step": 192495 }, { "epoch": 0.48, "learning_rate": 5.182847652121395e-05, "loss": 1.045, "step": 192500 }, { "epoch": 0.48, "learning_rate": 5.182721827973225e-05, "loss": 1.0425, "step": 192505 }, { "epoch": 0.48, "learning_rate": 5.182596003825054e-05, "loss": 1.0415, "step": 192510 }, { "epoch": 0.48, "learning_rate": 5.1824701796768835e-05, "loss": 1.0449, "step": 192515 }, { "epoch": 0.48, "learning_rate": 5.182344355528713e-05, "loss": 1.0467, "step": 192520 }, { "epoch": 0.48, "learning_rate": 5.1822185313805425e-05, "loss": 1.0479, "step": 192525 }, { "epoch": 0.48, "learning_rate": 5.182092707232372e-05, "loss": 1.047, "step": 192530 }, { "epoch": 0.48, "learning_rate": 5.1819668830842014e-05, "loss": 1.0458, "step": 192535 }, { "epoch": 0.48, "learning_rate": 5.181841058936031e-05, "loss": 1.0444, "step": 192540 }, { "epoch": 0.48, "learning_rate": 5.1817152347878604e-05, "loss": 1.0448, "step": 192545 }, { "epoch": 0.48, "learning_rate": 5.18158941063969e-05, "loss": 1.0482, "step": 192550 }, { "epoch": 0.48, "learning_rate": 5.1814635864915194e-05, "loss": 1.04, "step": 192555 }, { "epoch": 0.48, "learning_rate": 5.181337762343349e-05, "loss": 1.0472, "step": 192560 }, { "epoch": 0.48, "learning_rate": 5.1812119381951784e-05, "loss": 1.0438, "step": 192565 }, { "epoch": 0.48, "learning_rate": 5.181086114047008e-05, "loss": 1.0465, "step": 192570 }, { "epoch": 0.48, "learning_rate": 5.1809602898988374e-05, "loss": 1.0455, "step": 192575 }, { "epoch": 0.48, "learning_rate": 5.1808344657506666e-05, "loss": 1.0475, "step": 192580 }, { "epoch": 0.48, "learning_rate": 5.1807086416024964e-05, "loss": 1.0656, "step": 192585 }, { "epoch": 0.48, "learning_rate": 5.1805828174543256e-05, "loss": 1.0439, "step": 192590 }, { "epoch": 0.48, "learning_rate": 5.1804569933061554e-05, "loss": 1.0456, "step": 192595 }, { "epoch": 0.48, "learning_rate": 5.1803311691579846e-05, "loss": 1.0473, "step": 192600 }, { "epoch": 0.48, "learning_rate": 5.1802053450098144e-05, "loss": 1.0475, "step": 192605 }, { "epoch": 0.48, "learning_rate": 5.1800795208616436e-05, "loss": 1.0449, "step": 192610 }, { "epoch": 0.48, "learning_rate": 5.1799536967134734e-05, "loss": 1.0429, "step": 192615 }, { "epoch": 0.48, "learning_rate": 5.1798278725653026e-05, "loss": 1.0466, "step": 192620 }, { "epoch": 0.48, "learning_rate": 5.1797020484171324e-05, "loss": 1.0443, "step": 192625 }, { "epoch": 0.48, "learning_rate": 5.1795762242689616e-05, "loss": 1.0468, "step": 192630 }, { "epoch": 0.48, "learning_rate": 5.179450400120791e-05, "loss": 1.0432, "step": 192635 }, { "epoch": 0.48, "learning_rate": 5.1793245759726206e-05, "loss": 1.0483, "step": 192640 }, { "epoch": 0.48, "learning_rate": 5.17919875182445e-05, "loss": 1.0445, "step": 192645 }, { "epoch": 0.48, "learning_rate": 5.1790729276762796e-05, "loss": 1.0421, "step": 192650 }, { "epoch": 0.48, "learning_rate": 5.178947103528109e-05, "loss": 1.042, "step": 192655 }, { "epoch": 0.48, "learning_rate": 5.1788212793799386e-05, "loss": 1.0456, "step": 192660 }, { "epoch": 0.48, "learning_rate": 5.178695455231768e-05, "loss": 1.0449, "step": 192665 }, { "epoch": 0.48, "learning_rate": 5.1785696310835976e-05, "loss": 1.0447, "step": 192670 }, { "epoch": 0.48, "learning_rate": 5.178443806935427e-05, "loss": 1.0462, "step": 192675 }, { "epoch": 0.48, "learning_rate": 5.1783179827872566e-05, "loss": 1.0464, "step": 192680 }, { "epoch": 0.48, "learning_rate": 5.178192158639086e-05, "loss": 1.044, "step": 192685 }, { "epoch": 0.48, "learning_rate": 5.1780663344909156e-05, "loss": 1.0431, "step": 192690 }, { "epoch": 0.48, "learning_rate": 5.177940510342745e-05, "loss": 1.0468, "step": 192695 }, { "epoch": 0.48, "learning_rate": 5.177814686194574e-05, "loss": 1.0425, "step": 192700 }, { "epoch": 0.48, "learning_rate": 5.177688862046404e-05, "loss": 1.048, "step": 192705 }, { "epoch": 0.48, "learning_rate": 5.177563037898233e-05, "loss": 1.0454, "step": 192710 }, { "epoch": 0.48, "learning_rate": 5.177437213750063e-05, "loss": 1.0452, "step": 192715 }, { "epoch": 0.48, "learning_rate": 5.177311389601892e-05, "loss": 1.0453, "step": 192720 }, { "epoch": 0.48, "learning_rate": 5.177185565453722e-05, "loss": 1.0433, "step": 192725 }, { "epoch": 0.48, "learning_rate": 5.177059741305552e-05, "loss": 1.0426, "step": 192730 }, { "epoch": 0.48, "learning_rate": 5.1769339171573814e-05, "loss": 1.0423, "step": 192735 }, { "epoch": 0.48, "learning_rate": 5.176808093009211e-05, "loss": 1.0456, "step": 192740 }, { "epoch": 0.48, "learning_rate": 5.1766822688610404e-05, "loss": 1.0447, "step": 192745 }, { "epoch": 0.48, "learning_rate": 5.17655644471287e-05, "loss": 1.0704, "step": 192750 }, { "epoch": 0.48, "learning_rate": 5.1764306205646994e-05, "loss": 1.0458, "step": 192755 }, { "epoch": 0.48, "learning_rate": 5.176304796416529e-05, "loss": 1.0429, "step": 192760 }, { "epoch": 0.48, "learning_rate": 5.1761789722683584e-05, "loss": 1.0449, "step": 192765 }, { "epoch": 0.48, "learning_rate": 5.176053148120188e-05, "loss": 1.0462, "step": 192770 }, { "epoch": 0.48, "learning_rate": 5.1759273239720174e-05, "loss": 1.0445, "step": 192775 }, { "epoch": 0.48, "learning_rate": 5.175801499823847e-05, "loss": 1.0455, "step": 192780 }, { "epoch": 0.48, "learning_rate": 5.1756756756756764e-05, "loss": 1.0462, "step": 192785 }, { "epoch": 0.48, "learning_rate": 5.1755498515275055e-05, "loss": 1.0459, "step": 192790 }, { "epoch": 0.48, "learning_rate": 5.1754240273793354e-05, "loss": 1.0435, "step": 192795 }, { "epoch": 0.48, "learning_rate": 5.1752982032311645e-05, "loss": 1.0457, "step": 192800 }, { "epoch": 0.48, "learning_rate": 5.1751723790829944e-05, "loss": 1.0483, "step": 192805 }, { "epoch": 0.48, "learning_rate": 5.1750465549348235e-05, "loss": 1.0449, "step": 192810 }, { "epoch": 0.48, "learning_rate": 5.1749207307866534e-05, "loss": 1.0407, "step": 192815 }, { "epoch": 0.48, "learning_rate": 5.1747949066384825e-05, "loss": 1.0441, "step": 192820 }, { "epoch": 0.48, "learning_rate": 5.1746690824903124e-05, "loss": 1.044, "step": 192825 }, { "epoch": 0.48, "learning_rate": 5.1745432583421415e-05, "loss": 1.0665, "step": 192830 }, { "epoch": 0.48, "learning_rate": 5.1744174341939714e-05, "loss": 1.0512, "step": 192835 }, { "epoch": 0.48, "learning_rate": 5.1742916100458005e-05, "loss": 1.0416, "step": 192840 }, { "epoch": 0.48, "learning_rate": 5.17416578589763e-05, "loss": 1.0456, "step": 192845 }, { "epoch": 0.48, "learning_rate": 5.1740399617494595e-05, "loss": 1.0423, "step": 192850 }, { "epoch": 0.48, "learning_rate": 5.173914137601289e-05, "loss": 1.0428, "step": 192855 }, { "epoch": 0.48, "learning_rate": 5.1737883134531185e-05, "loss": 1.0469, "step": 192860 }, { "epoch": 0.48, "learning_rate": 5.173662489304948e-05, "loss": 1.0475, "step": 192865 }, { "epoch": 0.48, "learning_rate": 5.1735366651567775e-05, "loss": 1.0427, "step": 192870 }, { "epoch": 0.48, "learning_rate": 5.173410841008607e-05, "loss": 1.0473, "step": 192875 }, { "epoch": 0.48, "learning_rate": 5.1732850168604365e-05, "loss": 1.0437, "step": 192880 }, { "epoch": 0.48, "learning_rate": 5.173159192712266e-05, "loss": 1.0465, "step": 192885 }, { "epoch": 0.48, "learning_rate": 5.1730333685640955e-05, "loss": 1.046, "step": 192890 }, { "epoch": 0.48, "learning_rate": 5.172907544415925e-05, "loss": 1.0454, "step": 192895 }, { "epoch": 0.48, "learning_rate": 5.1727817202677545e-05, "loss": 1.0454, "step": 192900 }, { "epoch": 0.48, "learning_rate": 5.172655896119584e-05, "loss": 1.044, "step": 192905 }, { "epoch": 0.48, "learning_rate": 5.172530071971413e-05, "loss": 1.0853, "step": 192910 }, { "epoch": 0.48, "learning_rate": 5.1724042478232427e-05, "loss": 1.0446, "step": 192915 }, { "epoch": 0.48, "learning_rate": 5.172278423675072e-05, "loss": 1.0466, "step": 192920 }, { "epoch": 0.48, "learning_rate": 5.1721525995269017e-05, "loss": 1.0433, "step": 192925 }, { "epoch": 0.48, "learning_rate": 5.172026775378731e-05, "loss": 1.0477, "step": 192930 }, { "epoch": 0.48, "learning_rate": 5.1719009512305607e-05, "loss": 1.06, "step": 192935 }, { "epoch": 0.48, "learning_rate": 5.17177512708239e-05, "loss": 1.0461, "step": 192940 }, { "epoch": 0.48, "learning_rate": 5.1716493029342197e-05, "loss": 1.0425, "step": 192945 }, { "epoch": 0.48, "learning_rate": 5.171523478786049e-05, "loss": 1.0429, "step": 192950 }, { "epoch": 0.48, "learning_rate": 5.1713976546378786e-05, "loss": 1.038, "step": 192955 }, { "epoch": 0.48, "learning_rate": 5.171271830489708e-05, "loss": 1.0467, "step": 192960 }, { "epoch": 0.48, "learning_rate": 5.1711460063415376e-05, "loss": 1.0438, "step": 192965 }, { "epoch": 0.48, "learning_rate": 5.171020182193367e-05, "loss": 1.043, "step": 192970 }, { "epoch": 0.48, "learning_rate": 5.170894358045196e-05, "loss": 1.0439, "step": 192975 }, { "epoch": 0.48, "learning_rate": 5.170768533897026e-05, "loss": 1.0444, "step": 192980 }, { "epoch": 0.48, "learning_rate": 5.170642709748855e-05, "loss": 1.0462, "step": 192985 }, { "epoch": 0.48, "learning_rate": 5.170516885600685e-05, "loss": 1.0489, "step": 192990 }, { "epoch": 0.48, "learning_rate": 5.170391061452514e-05, "loss": 1.0464, "step": 192995 }, { "epoch": 0.48, "learning_rate": 5.170265237304344e-05, "loss": 1.046, "step": 193000 }, { "epoch": 0.48, "learning_rate": 5.170139413156173e-05, "loss": 1.0459, "step": 193005 }, { "epoch": 0.48, "learning_rate": 5.170013589008003e-05, "loss": 1.0456, "step": 193010 }, { "epoch": 0.48, "learning_rate": 5.169887764859832e-05, "loss": 1.0442, "step": 193015 }, { "epoch": 0.48, "learning_rate": 5.169761940711662e-05, "loss": 1.0467, "step": 193020 }, { "epoch": 0.48, "learning_rate": 5.169636116563491e-05, "loss": 1.0456, "step": 193025 }, { "epoch": 0.48, "learning_rate": 5.169510292415321e-05, "loss": 1.0463, "step": 193030 }, { "epoch": 0.48, "learning_rate": 5.16938446826715e-05, "loss": 1.0453, "step": 193035 }, { "epoch": 0.48, "learning_rate": 5.169258644118979e-05, "loss": 1.0443, "step": 193040 }, { "epoch": 0.48, "learning_rate": 5.169132819970809e-05, "loss": 1.0451, "step": 193045 }, { "epoch": 0.48, "learning_rate": 5.169006995822638e-05, "loss": 1.0424, "step": 193050 }, { "epoch": 0.48, "learning_rate": 5.168881171674468e-05, "loss": 1.0445, "step": 193055 }, { "epoch": 0.48, "learning_rate": 5.168755347526297e-05, "loss": 1.043, "step": 193060 }, { "epoch": 0.48, "learning_rate": 5.168629523378127e-05, "loss": 1.0477, "step": 193065 }, { "epoch": 0.48, "learning_rate": 5.168503699229956e-05, "loss": 1.0429, "step": 193070 }, { "epoch": 0.48, "learning_rate": 5.168377875081786e-05, "loss": 1.0448, "step": 193075 }, { "epoch": 0.48, "learning_rate": 5.168252050933615e-05, "loss": 1.0452, "step": 193080 }, { "epoch": 0.48, "learning_rate": 5.168126226785445e-05, "loss": 1.0413, "step": 193085 }, { "epoch": 0.48, "learning_rate": 5.168000402637274e-05, "loss": 1.0435, "step": 193090 }, { "epoch": 0.48, "learning_rate": 5.167874578489104e-05, "loss": 1.0446, "step": 193095 }, { "epoch": 0.48, "learning_rate": 5.167748754340933e-05, "loss": 1.045, "step": 193100 }, { "epoch": 0.48, "learning_rate": 5.167622930192762e-05, "loss": 1.0448, "step": 193105 }, { "epoch": 0.48, "learning_rate": 5.167497106044592e-05, "loss": 1.0448, "step": 193110 }, { "epoch": 0.48, "learning_rate": 5.167371281896421e-05, "loss": 1.0476, "step": 193115 }, { "epoch": 0.48, "learning_rate": 5.167245457748251e-05, "loss": 1.0451, "step": 193120 }, { "epoch": 0.48, "learning_rate": 5.16711963360008e-05, "loss": 1.0443, "step": 193125 }, { "epoch": 0.48, "learning_rate": 5.16699380945191e-05, "loss": 1.0429, "step": 193130 }, { "epoch": 0.48, "learning_rate": 5.166867985303739e-05, "loss": 1.0426, "step": 193135 }, { "epoch": 0.48, "learning_rate": 5.166742161155569e-05, "loss": 1.0471, "step": 193140 }, { "epoch": 0.48, "learning_rate": 5.166616337007398e-05, "loss": 1.0436, "step": 193145 }, { "epoch": 0.48, "learning_rate": 5.166490512859228e-05, "loss": 1.0462, "step": 193150 }, { "epoch": 0.48, "learning_rate": 5.166364688711057e-05, "loss": 1.0469, "step": 193155 }, { "epoch": 0.48, "learning_rate": 5.166238864562887e-05, "loss": 1.0447, "step": 193160 }, { "epoch": 0.48, "learning_rate": 5.166113040414716e-05, "loss": 1.0456, "step": 193165 }, { "epoch": 0.48, "learning_rate": 5.1659872162665454e-05, "loss": 1.0461, "step": 193170 }, { "epoch": 0.48, "learning_rate": 5.165861392118375e-05, "loss": 1.0453, "step": 193175 }, { "epoch": 0.48, "learning_rate": 5.1657355679702044e-05, "loss": 1.0459, "step": 193180 }, { "epoch": 0.48, "learning_rate": 5.165609743822034e-05, "loss": 1.0438, "step": 193185 }, { "epoch": 0.48, "learning_rate": 5.1654839196738634e-05, "loss": 1.0426, "step": 193190 }, { "epoch": 0.48, "learning_rate": 5.165358095525693e-05, "loss": 1.0436, "step": 193195 }, { "epoch": 0.48, "learning_rate": 5.1652322713775224e-05, "loss": 1.0452, "step": 193200 }, { "epoch": 0.48, "learning_rate": 5.165106447229352e-05, "loss": 1.0444, "step": 193205 }, { "epoch": 0.48, "learning_rate": 5.1649806230811814e-05, "loss": 1.0441, "step": 193210 }, { "epoch": 0.49, "learning_rate": 5.164854798933011e-05, "loss": 1.0447, "step": 193215 }, { "epoch": 0.49, "learning_rate": 5.1647289747848404e-05, "loss": 1.0484, "step": 193220 }, { "epoch": 0.49, "learning_rate": 5.16460315063667e-05, "loss": 1.0426, "step": 193225 }, { "epoch": 0.49, "learning_rate": 5.164477326488501e-05, "loss": 1.0461, "step": 193230 }, { "epoch": 0.49, "learning_rate": 5.16435150234033e-05, "loss": 1.0443, "step": 193235 }, { "epoch": 0.49, "learning_rate": 5.16422567819216e-05, "loss": 1.0665, "step": 193240 }, { "epoch": 0.49, "learning_rate": 5.164099854043989e-05, "loss": 1.0451, "step": 193245 }, { "epoch": 0.49, "learning_rate": 5.163974029895818e-05, "loss": 1.0453, "step": 193250 }, { "epoch": 0.49, "learning_rate": 5.163848205747648e-05, "loss": 1.0454, "step": 193255 }, { "epoch": 0.49, "learning_rate": 5.163722381599477e-05, "loss": 1.0448, "step": 193260 }, { "epoch": 0.49, "learning_rate": 5.163596557451307e-05, "loss": 1.0433, "step": 193265 }, { "epoch": 0.49, "learning_rate": 5.163470733303136e-05, "loss": 1.0434, "step": 193270 }, { "epoch": 0.49, "learning_rate": 5.163344909154966e-05, "loss": 1.0442, "step": 193275 }, { "epoch": 0.49, "learning_rate": 5.163219085006795e-05, "loss": 1.0469, "step": 193280 }, { "epoch": 0.49, "learning_rate": 5.163093260858625e-05, "loss": 1.046, "step": 193285 }, { "epoch": 0.49, "learning_rate": 5.162967436710454e-05, "loss": 1.0444, "step": 193290 }, { "epoch": 0.49, "learning_rate": 5.162841612562284e-05, "loss": 1.0475, "step": 193295 }, { "epoch": 0.49, "learning_rate": 5.162715788414113e-05, "loss": 1.0712, "step": 193300 }, { "epoch": 0.49, "learning_rate": 5.162589964265943e-05, "loss": 1.0433, "step": 193305 }, { "epoch": 0.49, "learning_rate": 5.162464140117772e-05, "loss": 1.0427, "step": 193310 }, { "epoch": 0.49, "learning_rate": 5.162338315969601e-05, "loss": 1.046, "step": 193315 }, { "epoch": 0.49, "learning_rate": 5.162212491821431e-05, "loss": 1.0414, "step": 193320 }, { "epoch": 0.49, "learning_rate": 5.16208666767326e-05, "loss": 1.0398, "step": 193325 }, { "epoch": 0.49, "learning_rate": 5.16196084352509e-05, "loss": 1.0446, "step": 193330 }, { "epoch": 0.49, "learning_rate": 5.161835019376919e-05, "loss": 1.0455, "step": 193335 }, { "epoch": 0.49, "learning_rate": 5.161709195228749e-05, "loss": 1.0452, "step": 193340 }, { "epoch": 0.49, "learning_rate": 5.161583371080578e-05, "loss": 1.0473, "step": 193345 }, { "epoch": 0.49, "learning_rate": 5.161457546932408e-05, "loss": 1.0403, "step": 193350 }, { "epoch": 0.49, "learning_rate": 5.161331722784237e-05, "loss": 1.0463, "step": 193355 }, { "epoch": 0.49, "learning_rate": 5.161205898636067e-05, "loss": 1.0417, "step": 193360 }, { "epoch": 0.49, "learning_rate": 5.161080074487896e-05, "loss": 1.0441, "step": 193365 }, { "epoch": 0.49, "learning_rate": 5.160954250339726e-05, "loss": 1.0466, "step": 193370 }, { "epoch": 0.49, "learning_rate": 5.160828426191555e-05, "loss": 1.0447, "step": 193375 }, { "epoch": 0.49, "learning_rate": 5.160702602043384e-05, "loss": 1.0451, "step": 193380 }, { "epoch": 0.49, "learning_rate": 5.160576777895214e-05, "loss": 1.0456, "step": 193385 }, { "epoch": 0.49, "learning_rate": 5.160450953747043e-05, "loss": 1.0451, "step": 193390 }, { "epoch": 0.49, "learning_rate": 5.160325129598873e-05, "loss": 1.0444, "step": 193395 }, { "epoch": 0.49, "learning_rate": 5.160199305450702e-05, "loss": 1.045, "step": 193400 }, { "epoch": 0.49, "learning_rate": 5.160073481302532e-05, "loss": 1.047, "step": 193405 }, { "epoch": 0.49, "learning_rate": 5.159947657154361e-05, "loss": 1.0438, "step": 193410 }, { "epoch": 0.49, "learning_rate": 5.159821833006191e-05, "loss": 1.0464, "step": 193415 }, { "epoch": 0.49, "learning_rate": 5.15969600885802e-05, "loss": 1.0468, "step": 193420 }, { "epoch": 0.49, "learning_rate": 5.15957018470985e-05, "loss": 1.0485, "step": 193425 }, { "epoch": 0.49, "learning_rate": 5.159444360561679e-05, "loss": 1.0458, "step": 193430 }, { "epoch": 0.49, "learning_rate": 5.159318536413509e-05, "loss": 1.066, "step": 193435 }, { "epoch": 0.49, "learning_rate": 5.159192712265338e-05, "loss": 1.0467, "step": 193440 }, { "epoch": 0.49, "learning_rate": 5.1590668881171675e-05, "loss": 1.0443, "step": 193445 }, { "epoch": 0.49, "learning_rate": 5.158941063968997e-05, "loss": 1.0439, "step": 193450 }, { "epoch": 0.49, "learning_rate": 5.1588152398208265e-05, "loss": 1.0445, "step": 193455 }, { "epoch": 0.49, "learning_rate": 5.158689415672656e-05, "loss": 1.0446, "step": 193460 }, { "epoch": 0.49, "learning_rate": 5.1585635915244855e-05, "loss": 1.0425, "step": 193465 }, { "epoch": 0.49, "learning_rate": 5.158437767376315e-05, "loss": 1.0442, "step": 193470 }, { "epoch": 0.49, "learning_rate": 5.1583119432281445e-05, "loss": 1.0465, "step": 193475 }, { "epoch": 0.49, "learning_rate": 5.158186119079974e-05, "loss": 1.0474, "step": 193480 }, { "epoch": 0.49, "learning_rate": 5.1580602949318035e-05, "loss": 1.0477, "step": 193485 }, { "epoch": 0.49, "learning_rate": 5.157934470783633e-05, "loss": 1.0433, "step": 193490 }, { "epoch": 0.49, "learning_rate": 5.1578086466354625e-05, "loss": 1.0457, "step": 193495 }, { "epoch": 0.49, "learning_rate": 5.1576828224872916e-05, "loss": 1.0484, "step": 193500 }, { "epoch": 0.49, "learning_rate": 5.1575569983391215e-05, "loss": 1.0425, "step": 193505 }, { "epoch": 0.49, "learning_rate": 5.1574311741909506e-05, "loss": 1.0465, "step": 193510 }, { "epoch": 0.49, "learning_rate": 5.1573053500427805e-05, "loss": 1.0442, "step": 193515 }, { "epoch": 0.49, "learning_rate": 5.1571795258946096e-05, "loss": 1.0463, "step": 193520 }, { "epoch": 0.49, "learning_rate": 5.1570537017464395e-05, "loss": 1.0445, "step": 193525 }, { "epoch": 0.49, "learning_rate": 5.1569278775982686e-05, "loss": 1.0429, "step": 193530 }, { "epoch": 0.49, "learning_rate": 5.1568020534500984e-05, "loss": 1.0453, "step": 193535 }, { "epoch": 0.49, "learning_rate": 5.1566762293019276e-05, "loss": 1.0428, "step": 193540 }, { "epoch": 0.49, "learning_rate": 5.1565504051537574e-05, "loss": 1.045, "step": 193545 }, { "epoch": 0.49, "learning_rate": 5.1564245810055866e-05, "loss": 1.0426, "step": 193550 }, { "epoch": 0.49, "learning_rate": 5.1562987568574164e-05, "loss": 1.0427, "step": 193555 }, { "epoch": 0.49, "learning_rate": 5.1561729327092456e-05, "loss": 1.0436, "step": 193560 }, { "epoch": 0.49, "learning_rate": 5.156047108561075e-05, "loss": 1.0429, "step": 193565 }, { "epoch": 0.49, "learning_rate": 5.1559212844129046e-05, "loss": 1.0468, "step": 193570 }, { "epoch": 0.49, "learning_rate": 5.155795460264734e-05, "loss": 1.0466, "step": 193575 }, { "epoch": 0.49, "learning_rate": 5.1556696361165636e-05, "loss": 1.043, "step": 193580 }, { "epoch": 0.49, "learning_rate": 5.155543811968393e-05, "loss": 1.0432, "step": 193585 }, { "epoch": 0.49, "learning_rate": 5.1554179878202226e-05, "loss": 1.0477, "step": 193590 }, { "epoch": 0.49, "learning_rate": 5.155292163672052e-05, "loss": 1.0477, "step": 193595 }, { "epoch": 0.49, "learning_rate": 5.1551663395238816e-05, "loss": 1.0454, "step": 193600 }, { "epoch": 0.49, "learning_rate": 5.155040515375711e-05, "loss": 1.0461, "step": 193605 }, { "epoch": 0.49, "learning_rate": 5.1549146912275406e-05, "loss": 1.0467, "step": 193610 }, { "epoch": 0.49, "learning_rate": 5.15478886707937e-05, "loss": 1.0408, "step": 193615 }, { "epoch": 0.49, "learning_rate": 5.1546630429311996e-05, "loss": 1.0427, "step": 193620 }, { "epoch": 0.49, "learning_rate": 5.154537218783029e-05, "loss": 1.0443, "step": 193625 }, { "epoch": 0.49, "learning_rate": 5.154411394634858e-05, "loss": 1.0443, "step": 193630 }, { "epoch": 0.49, "learning_rate": 5.154285570486688e-05, "loss": 1.046, "step": 193635 }, { "epoch": 0.49, "learning_rate": 5.154159746338517e-05, "loss": 1.0431, "step": 193640 }, { "epoch": 0.49, "learning_rate": 5.154033922190347e-05, "loss": 1.0419, "step": 193645 }, { "epoch": 0.49, "learning_rate": 5.153908098042176e-05, "loss": 1.0442, "step": 193650 }, { "epoch": 0.49, "learning_rate": 5.153782273894006e-05, "loss": 1.0436, "step": 193655 }, { "epoch": 0.49, "learning_rate": 5.153656449745835e-05, "loss": 1.0453, "step": 193660 }, { "epoch": 0.49, "learning_rate": 5.153530625597665e-05, "loss": 1.045, "step": 193665 }, { "epoch": 0.49, "learning_rate": 5.153404801449494e-05, "loss": 1.0439, "step": 193670 }, { "epoch": 0.49, "learning_rate": 5.153278977301324e-05, "loss": 1.0499, "step": 193675 }, { "epoch": 0.49, "learning_rate": 5.153153153153153e-05, "loss": 1.0453, "step": 193680 }, { "epoch": 0.49, "learning_rate": 5.153027329004983e-05, "loss": 1.0476, "step": 193685 }, { "epoch": 0.49, "learning_rate": 5.152901504856812e-05, "loss": 1.045, "step": 193690 }, { "epoch": 0.49, "learning_rate": 5.152775680708641e-05, "loss": 1.0452, "step": 193695 }, { "epoch": 0.49, "learning_rate": 5.152649856560471e-05, "loss": 1.0442, "step": 193700 }, { "epoch": 0.49, "learning_rate": 5.1525240324123e-05, "loss": 1.0472, "step": 193705 }, { "epoch": 0.49, "learning_rate": 5.15239820826413e-05, "loss": 1.0479, "step": 193710 }, { "epoch": 0.49, "learning_rate": 5.152272384115959e-05, "loss": 1.0455, "step": 193715 }, { "epoch": 0.49, "learning_rate": 5.152146559967789e-05, "loss": 1.0439, "step": 193720 }, { "epoch": 0.49, "learning_rate": 5.152020735819618e-05, "loss": 1.0474, "step": 193725 }, { "epoch": 0.49, "learning_rate": 5.1518949116714486e-05, "loss": 1.0447, "step": 193730 }, { "epoch": 0.49, "learning_rate": 5.1517690875232784e-05, "loss": 1.045, "step": 193735 }, { "epoch": 0.49, "learning_rate": 5.1516432633751076e-05, "loss": 1.0456, "step": 193740 }, { "epoch": 0.49, "learning_rate": 5.1515174392269374e-05, "loss": 1.0459, "step": 193745 }, { "epoch": 0.49, "learning_rate": 5.1513916150787665e-05, "loss": 1.0443, "step": 193750 }, { "epoch": 0.49, "learning_rate": 5.1512657909305964e-05, "loss": 1.047, "step": 193755 }, { "epoch": 0.49, "learning_rate": 5.1511399667824255e-05, "loss": 1.0449, "step": 193760 }, { "epoch": 0.49, "learning_rate": 5.1510141426342554e-05, "loss": 1.0452, "step": 193765 }, { "epoch": 0.49, "learning_rate": 5.1508883184860845e-05, "loss": 1.0482, "step": 193770 }, { "epoch": 0.49, "learning_rate": 5.150762494337914e-05, "loss": 1.0481, "step": 193775 }, { "epoch": 0.49, "learning_rate": 5.1506366701897435e-05, "loss": 1.0455, "step": 193780 }, { "epoch": 0.49, "learning_rate": 5.150510846041573e-05, "loss": 1.0433, "step": 193785 }, { "epoch": 0.49, "learning_rate": 5.1503850218934025e-05, "loss": 1.0476, "step": 193790 }, { "epoch": 0.49, "learning_rate": 5.150259197745232e-05, "loss": 1.0458, "step": 193795 }, { "epoch": 0.49, "learning_rate": 5.1501333735970615e-05, "loss": 1.0457, "step": 193800 }, { "epoch": 0.49, "learning_rate": 5.150007549448891e-05, "loss": 1.0456, "step": 193805 }, { "epoch": 0.49, "learning_rate": 5.1498817253007205e-05, "loss": 1.0442, "step": 193810 }, { "epoch": 0.49, "learning_rate": 5.14975590115255e-05, "loss": 1.0456, "step": 193815 }, { "epoch": 0.49, "learning_rate": 5.1496300770043795e-05, "loss": 1.0477, "step": 193820 }, { "epoch": 0.49, "learning_rate": 5.149504252856209e-05, "loss": 1.0434, "step": 193825 }, { "epoch": 0.49, "learning_rate": 5.1493784287080385e-05, "loss": 1.0463, "step": 193830 }, { "epoch": 0.49, "learning_rate": 5.149252604559868e-05, "loss": 1.0455, "step": 193835 }, { "epoch": 0.49, "learning_rate": 5.149126780411697e-05, "loss": 1.044, "step": 193840 }, { "epoch": 0.49, "learning_rate": 5.149000956263527e-05, "loss": 1.0448, "step": 193845 }, { "epoch": 0.49, "learning_rate": 5.148875132115356e-05, "loss": 1.0436, "step": 193850 }, { "epoch": 0.49, "learning_rate": 5.148749307967186e-05, "loss": 1.0428, "step": 193855 }, { "epoch": 0.49, "learning_rate": 5.148623483819015e-05, "loss": 1.0446, "step": 193860 }, { "epoch": 0.49, "learning_rate": 5.148497659670845e-05, "loss": 1.0437, "step": 193865 }, { "epoch": 0.49, "learning_rate": 5.148371835522674e-05, "loss": 1.0434, "step": 193870 }, { "epoch": 0.49, "learning_rate": 5.148246011374504e-05, "loss": 1.0439, "step": 193875 }, { "epoch": 0.49, "learning_rate": 5.148120187226333e-05, "loss": 1.0423, "step": 193880 }, { "epoch": 0.49, "learning_rate": 5.147994363078163e-05, "loss": 1.045, "step": 193885 }, { "epoch": 0.49, "learning_rate": 5.147868538929992e-05, "loss": 1.0455, "step": 193890 }, { "epoch": 0.49, "learning_rate": 5.147742714781822e-05, "loss": 1.0474, "step": 193895 }, { "epoch": 0.49, "learning_rate": 5.147616890633651e-05, "loss": 1.0451, "step": 193900 }, { "epoch": 0.49, "learning_rate": 5.14749106648548e-05, "loss": 1.0427, "step": 193905 }, { "epoch": 0.49, "learning_rate": 5.14736524233731e-05, "loss": 1.0445, "step": 193910 }, { "epoch": 0.49, "learning_rate": 5.147239418189139e-05, "loss": 1.0477, "step": 193915 }, { "epoch": 0.49, "learning_rate": 5.147113594040969e-05, "loss": 1.0424, "step": 193920 }, { "epoch": 0.49, "learning_rate": 5.146987769892798e-05, "loss": 1.0468, "step": 193925 }, { "epoch": 0.49, "learning_rate": 5.146861945744628e-05, "loss": 1.0462, "step": 193930 }, { "epoch": 0.49, "learning_rate": 5.146736121596457e-05, "loss": 1.0436, "step": 193935 }, { "epoch": 0.49, "learning_rate": 5.146610297448287e-05, "loss": 1.046, "step": 193940 }, { "epoch": 0.49, "learning_rate": 5.146484473300116e-05, "loss": 1.0698, "step": 193945 }, { "epoch": 0.49, "learning_rate": 5.146358649151946e-05, "loss": 1.0459, "step": 193950 }, { "epoch": 0.49, "learning_rate": 5.146232825003775e-05, "loss": 1.0462, "step": 193955 }, { "epoch": 0.49, "learning_rate": 5.146107000855605e-05, "loss": 1.0467, "step": 193960 }, { "epoch": 0.49, "learning_rate": 5.145981176707434e-05, "loss": 1.046, "step": 193965 }, { "epoch": 0.49, "learning_rate": 5.145855352559263e-05, "loss": 1.0481, "step": 193970 }, { "epoch": 0.49, "learning_rate": 5.145729528411093e-05, "loss": 1.0425, "step": 193975 }, { "epoch": 0.49, "learning_rate": 5.145603704262922e-05, "loss": 1.0459, "step": 193980 }, { "epoch": 0.49, "learning_rate": 5.145477880114752e-05, "loss": 1.0458, "step": 193985 }, { "epoch": 0.49, "learning_rate": 5.145352055966581e-05, "loss": 1.0443, "step": 193990 }, { "epoch": 0.49, "learning_rate": 5.145226231818411e-05, "loss": 1.0431, "step": 193995 }, { "epoch": 0.49, "learning_rate": 5.14510040767024e-05, "loss": 1.0437, "step": 194000 }, { "epoch": 0.49, "learning_rate": 5.14497458352207e-05, "loss": 1.0445, "step": 194005 }, { "epoch": 0.49, "learning_rate": 5.144848759373899e-05, "loss": 1.0434, "step": 194010 }, { "epoch": 0.49, "learning_rate": 5.144722935225729e-05, "loss": 1.0454, "step": 194015 }, { "epoch": 0.49, "learning_rate": 5.144597111077558e-05, "loss": 1.0453, "step": 194020 }, { "epoch": 0.49, "learning_rate": 5.144471286929388e-05, "loss": 1.0429, "step": 194025 }, { "epoch": 0.49, "learning_rate": 5.144345462781217e-05, "loss": 1.0464, "step": 194030 }, { "epoch": 0.49, "learning_rate": 5.144219638633046e-05, "loss": 1.0467, "step": 194035 }, { "epoch": 0.49, "learning_rate": 5.144093814484876e-05, "loss": 1.046, "step": 194040 }, { "epoch": 0.49, "learning_rate": 5.143967990336705e-05, "loss": 1.043, "step": 194045 }, { "epoch": 0.49, "learning_rate": 5.143842166188535e-05, "loss": 1.0414, "step": 194050 }, { "epoch": 0.49, "learning_rate": 5.143716342040364e-05, "loss": 1.0453, "step": 194055 }, { "epoch": 0.49, "learning_rate": 5.143590517892194e-05, "loss": 1.0431, "step": 194060 }, { "epoch": 0.49, "learning_rate": 5.143464693744023e-05, "loss": 1.0464, "step": 194065 }, { "epoch": 0.49, "learning_rate": 5.143338869595853e-05, "loss": 1.0436, "step": 194070 }, { "epoch": 0.49, "learning_rate": 5.143213045447682e-05, "loss": 1.048, "step": 194075 }, { "epoch": 0.49, "learning_rate": 5.143087221299512e-05, "loss": 1.041, "step": 194080 }, { "epoch": 0.49, "learning_rate": 5.142961397151341e-05, "loss": 1.0476, "step": 194085 }, { "epoch": 0.49, "learning_rate": 5.142835573003171e-05, "loss": 1.0444, "step": 194090 }, { "epoch": 0.49, "learning_rate": 5.142709748855e-05, "loss": 1.0433, "step": 194095 }, { "epoch": 0.49, "learning_rate": 5.1425839247068294e-05, "loss": 1.0453, "step": 194100 }, { "epoch": 0.49, "learning_rate": 5.142458100558659e-05, "loss": 1.0424, "step": 194105 }, { "epoch": 0.49, "learning_rate": 5.1423322764104884e-05, "loss": 1.0401, "step": 194110 }, { "epoch": 0.49, "learning_rate": 5.142206452262318e-05, "loss": 1.0462, "step": 194115 }, { "epoch": 0.49, "learning_rate": 5.1420806281141474e-05, "loss": 1.0461, "step": 194120 }, { "epoch": 0.49, "learning_rate": 5.141954803965977e-05, "loss": 1.0479, "step": 194125 }, { "epoch": 0.49, "learning_rate": 5.1418289798178064e-05, "loss": 1.0462, "step": 194130 }, { "epoch": 0.49, "learning_rate": 5.141703155669636e-05, "loss": 1.0441, "step": 194135 }, { "epoch": 0.49, "learning_rate": 5.1415773315214654e-05, "loss": 1.0412, "step": 194140 }, { "epoch": 0.49, "learning_rate": 5.141451507373295e-05, "loss": 1.0424, "step": 194145 }, { "epoch": 0.49, "learning_rate": 5.1413256832251244e-05, "loss": 1.0449, "step": 194150 }, { "epoch": 0.49, "learning_rate": 5.1411998590769536e-05, "loss": 1.0443, "step": 194155 }, { "epoch": 0.49, "learning_rate": 5.1410740349287834e-05, "loss": 1.043, "step": 194160 }, { "epoch": 0.49, "learning_rate": 5.1409482107806126e-05, "loss": 1.0438, "step": 194165 }, { "epoch": 0.49, "learning_rate": 5.1408223866324424e-05, "loss": 1.046, "step": 194170 }, { "epoch": 0.49, "learning_rate": 5.1406965624842716e-05, "loss": 1.0434, "step": 194175 }, { "epoch": 0.49, "learning_rate": 5.1405707383361014e-05, "loss": 1.0635, "step": 194180 }, { "epoch": 0.49, "learning_rate": 5.1404449141879306e-05, "loss": 1.0458, "step": 194185 }, { "epoch": 0.49, "learning_rate": 5.1403190900397604e-05, "loss": 1.0448, "step": 194190 }, { "epoch": 0.49, "learning_rate": 5.1401932658915895e-05, "loss": 1.0439, "step": 194195 }, { "epoch": 0.49, "learning_rate": 5.1400674417434194e-05, "loss": 1.0436, "step": 194200 }, { "epoch": 0.49, "learning_rate": 5.1399416175952485e-05, "loss": 1.0435, "step": 194205 }, { "epoch": 0.49, "learning_rate": 5.1398157934470784e-05, "loss": 1.047, "step": 194210 }, { "epoch": 0.49, "learning_rate": 5.1396899692989075e-05, "loss": 1.0461, "step": 194215 }, { "epoch": 0.49, "learning_rate": 5.139564145150737e-05, "loss": 1.0438, "step": 194220 }, { "epoch": 0.49, "learning_rate": 5.139438321002568e-05, "loss": 1.045, "step": 194225 }, { "epoch": 0.49, "learning_rate": 5.139312496854397e-05, "loss": 1.0411, "step": 194230 }, { "epoch": 0.49, "learning_rate": 5.139186672706227e-05, "loss": 1.0462, "step": 194235 }, { "epoch": 0.49, "learning_rate": 5.139060848558056e-05, "loss": 1.0451, "step": 194240 }, { "epoch": 0.49, "learning_rate": 5.138935024409885e-05, "loss": 1.0451, "step": 194245 }, { "epoch": 0.49, "learning_rate": 5.138809200261715e-05, "loss": 1.0461, "step": 194250 }, { "epoch": 0.49, "learning_rate": 5.138683376113544e-05, "loss": 1.0422, "step": 194255 }, { "epoch": 0.49, "learning_rate": 5.138557551965374e-05, "loss": 1.0453, "step": 194260 }, { "epoch": 0.49, "learning_rate": 5.138431727817203e-05, "loss": 1.0462, "step": 194265 }, { "epoch": 0.49, "learning_rate": 5.138305903669033e-05, "loss": 1.0499, "step": 194270 }, { "epoch": 0.49, "learning_rate": 5.138180079520862e-05, "loss": 1.0481, "step": 194275 }, { "epoch": 0.49, "learning_rate": 5.138054255372692e-05, "loss": 1.0452, "step": 194280 }, { "epoch": 0.49, "learning_rate": 5.137928431224521e-05, "loss": 1.0458, "step": 194285 }, { "epoch": 0.49, "learning_rate": 5.137802607076351e-05, "loss": 1.0433, "step": 194290 }, { "epoch": 0.49, "learning_rate": 5.13767678292818e-05, "loss": 1.0444, "step": 194295 }, { "epoch": 0.49, "learning_rate": 5.1375509587800094e-05, "loss": 1.0446, "step": 194300 }, { "epoch": 0.49, "learning_rate": 5.137425134631839e-05, "loss": 1.0445, "step": 194305 }, { "epoch": 0.49, "learning_rate": 5.1372993104836684e-05, "loss": 1.0443, "step": 194310 }, { "epoch": 0.49, "learning_rate": 5.137173486335498e-05, "loss": 1.0429, "step": 194315 }, { "epoch": 0.49, "learning_rate": 5.1370476621873274e-05, "loss": 1.0472, "step": 194320 }, { "epoch": 0.49, "learning_rate": 5.136921838039157e-05, "loss": 1.0637, "step": 194325 }, { "epoch": 0.49, "learning_rate": 5.1367960138909863e-05, "loss": 1.0475, "step": 194330 }, { "epoch": 0.49, "learning_rate": 5.136670189742816e-05, "loss": 1.043, "step": 194335 }, { "epoch": 0.49, "learning_rate": 5.1365443655946453e-05, "loss": 1.0464, "step": 194340 }, { "epoch": 0.49, "learning_rate": 5.136418541446475e-05, "loss": 1.0455, "step": 194345 }, { "epoch": 0.49, "learning_rate": 5.1362927172983043e-05, "loss": 1.0426, "step": 194350 }, { "epoch": 0.49, "learning_rate": 5.136166893150134e-05, "loss": 1.06, "step": 194355 }, { "epoch": 0.49, "learning_rate": 5.1360410690019633e-05, "loss": 1.0468, "step": 194360 }, { "epoch": 0.49, "learning_rate": 5.1359152448537925e-05, "loss": 1.0466, "step": 194365 }, { "epoch": 0.49, "learning_rate": 5.135789420705622e-05, "loss": 1.0417, "step": 194370 }, { "epoch": 0.49, "learning_rate": 5.1356635965574515e-05, "loss": 1.0438, "step": 194375 }, { "epoch": 0.49, "learning_rate": 5.135537772409281e-05, "loss": 1.0449, "step": 194380 }, { "epoch": 0.49, "learning_rate": 5.1354119482611105e-05, "loss": 1.0425, "step": 194385 }, { "epoch": 0.49, "learning_rate": 5.13528612411294e-05, "loss": 1.0456, "step": 194390 }, { "epoch": 0.49, "learning_rate": 5.1351602999647695e-05, "loss": 1.0462, "step": 194395 }, { "epoch": 0.49, "learning_rate": 5.135034475816599e-05, "loss": 1.07, "step": 194400 }, { "epoch": 0.49, "learning_rate": 5.1349086516684285e-05, "loss": 1.045, "step": 194405 }, { "epoch": 0.49, "learning_rate": 5.134782827520258e-05, "loss": 1.0429, "step": 194410 }, { "epoch": 0.49, "learning_rate": 5.1346570033720875e-05, "loss": 1.0459, "step": 194415 }, { "epoch": 0.49, "learning_rate": 5.134531179223917e-05, "loss": 1.0452, "step": 194420 }, { "epoch": 0.49, "learning_rate": 5.1344053550757465e-05, "loss": 1.0423, "step": 194425 }, { "epoch": 0.49, "learning_rate": 5.1342795309275756e-05, "loss": 1.0462, "step": 194430 }, { "epoch": 0.49, "learning_rate": 5.1341537067794055e-05, "loss": 1.0467, "step": 194435 }, { "epoch": 0.49, "learning_rate": 5.1340278826312346e-05, "loss": 1.0497, "step": 194440 }, { "epoch": 0.49, "learning_rate": 5.1339020584830645e-05, "loss": 1.0458, "step": 194445 }, { "epoch": 0.49, "learning_rate": 5.1337762343348936e-05, "loss": 1.0459, "step": 194450 }, { "epoch": 0.49, "learning_rate": 5.1336504101867235e-05, "loss": 1.0429, "step": 194455 }, { "epoch": 0.49, "learning_rate": 5.1335245860385526e-05, "loss": 1.0479, "step": 194460 }, { "epoch": 0.49, "learning_rate": 5.1333987618903825e-05, "loss": 1.0453, "step": 194465 }, { "epoch": 0.49, "learning_rate": 5.1332729377422116e-05, "loss": 1.0459, "step": 194470 }, { "epoch": 0.49, "learning_rate": 5.1331471135940415e-05, "loss": 1.0475, "step": 194475 }, { "epoch": 0.49, "learning_rate": 5.1330212894458706e-05, "loss": 1.0468, "step": 194480 }, { "epoch": 0.49, "learning_rate": 5.1328954652977005e-05, "loss": 1.0447, "step": 194485 }, { "epoch": 0.49, "learning_rate": 5.1327696411495296e-05, "loss": 1.0453, "step": 194490 }, { "epoch": 0.49, "learning_rate": 5.132643817001359e-05, "loss": 1.0484, "step": 194495 }, { "epoch": 0.49, "learning_rate": 5.1325179928531886e-05, "loss": 1.0441, "step": 194500 }, { "epoch": 0.49, "learning_rate": 5.132392168705018e-05, "loss": 1.0459, "step": 194505 }, { "epoch": 0.49, "learning_rate": 5.1322663445568476e-05, "loss": 1.0416, "step": 194510 }, { "epoch": 0.49, "learning_rate": 5.132140520408677e-05, "loss": 1.0426, "step": 194515 }, { "epoch": 0.49, "learning_rate": 5.1320146962605066e-05, "loss": 1.0422, "step": 194520 }, { "epoch": 0.49, "learning_rate": 5.131888872112336e-05, "loss": 1.0458, "step": 194525 }, { "epoch": 0.49, "learning_rate": 5.1317630479641656e-05, "loss": 1.0426, "step": 194530 }, { "epoch": 0.49, "learning_rate": 5.131637223815995e-05, "loss": 1.0445, "step": 194535 }, { "epoch": 0.49, "learning_rate": 5.1315113996678246e-05, "loss": 1.0474, "step": 194540 }, { "epoch": 0.49, "learning_rate": 5.131385575519654e-05, "loss": 1.0453, "step": 194545 }, { "epoch": 0.49, "learning_rate": 5.1312597513714836e-05, "loss": 1.043, "step": 194550 }, { "epoch": 0.49, "learning_rate": 5.131133927223313e-05, "loss": 1.0447, "step": 194555 }, { "epoch": 0.49, "learning_rate": 5.131008103075142e-05, "loss": 1.0467, "step": 194560 }, { "epoch": 0.49, "learning_rate": 5.130882278926972e-05, "loss": 1.0723, "step": 194565 }, { "epoch": 0.49, "learning_rate": 5.130756454778801e-05, "loss": 1.0439, "step": 194570 }, { "epoch": 0.49, "learning_rate": 5.130630630630631e-05, "loss": 1.0459, "step": 194575 }, { "epoch": 0.49, "learning_rate": 5.13050480648246e-05, "loss": 1.0451, "step": 194580 }, { "epoch": 0.49, "learning_rate": 5.13037898233429e-05, "loss": 1.0476, "step": 194585 }, { "epoch": 0.49, "learning_rate": 5.130253158186119e-05, "loss": 1.0669, "step": 194590 }, { "epoch": 0.49, "learning_rate": 5.130127334037949e-05, "loss": 1.0445, "step": 194595 }, { "epoch": 0.49, "learning_rate": 5.130001509889778e-05, "loss": 1.0429, "step": 194600 }, { "epoch": 0.49, "learning_rate": 5.129875685741608e-05, "loss": 1.0453, "step": 194605 }, { "epoch": 0.49, "learning_rate": 5.129749861593437e-05, "loss": 1.0463, "step": 194610 }, { "epoch": 0.49, "learning_rate": 5.129624037445267e-05, "loss": 1.0426, "step": 194615 }, { "epoch": 0.49, "learning_rate": 5.129498213297096e-05, "loss": 1.0428, "step": 194620 }, { "epoch": 0.49, "learning_rate": 5.129372389148925e-05, "loss": 1.0453, "step": 194625 }, { "epoch": 0.49, "learning_rate": 5.129246565000755e-05, "loss": 1.0432, "step": 194630 }, { "epoch": 0.49, "learning_rate": 5.129120740852584e-05, "loss": 1.0449, "step": 194635 }, { "epoch": 0.49, "learning_rate": 5.128994916704414e-05, "loss": 1.0445, "step": 194640 }, { "epoch": 0.49, "learning_rate": 5.128869092556243e-05, "loss": 1.0574, "step": 194645 }, { "epoch": 0.49, "learning_rate": 5.128743268408073e-05, "loss": 1.0446, "step": 194650 }, { "epoch": 0.49, "learning_rate": 5.128617444259902e-05, "loss": 1.0709, "step": 194655 }, { "epoch": 0.49, "learning_rate": 5.128491620111732e-05, "loss": 1.0465, "step": 194660 }, { "epoch": 0.49, "learning_rate": 5.128365795963561e-05, "loss": 1.0454, "step": 194665 }, { "epoch": 0.49, "learning_rate": 5.128239971815391e-05, "loss": 1.0437, "step": 194670 }, { "epoch": 0.49, "learning_rate": 5.12811414766722e-05, "loss": 1.0457, "step": 194675 }, { "epoch": 0.49, "learning_rate": 5.12798832351905e-05, "loss": 1.0458, "step": 194680 }, { "epoch": 0.49, "learning_rate": 5.127862499370879e-05, "loss": 1.0471, "step": 194685 }, { "epoch": 0.49, "learning_rate": 5.127736675222708e-05, "loss": 1.0458, "step": 194690 }, { "epoch": 0.49, "learning_rate": 5.127610851074538e-05, "loss": 1.0449, "step": 194695 }, { "epoch": 0.49, "learning_rate": 5.127485026926367e-05, "loss": 1.0482, "step": 194700 }, { "epoch": 0.49, "learning_rate": 5.127359202778197e-05, "loss": 1.0463, "step": 194705 }, { "epoch": 0.49, "learning_rate": 5.127233378630026e-05, "loss": 1.0459, "step": 194710 }, { "epoch": 0.49, "learning_rate": 5.127107554481856e-05, "loss": 1.0422, "step": 194715 }, { "epoch": 0.49, "learning_rate": 5.126981730333685e-05, "loss": 1.0733, "step": 194720 }, { "epoch": 0.49, "learning_rate": 5.126855906185516e-05, "loss": 1.0455, "step": 194725 }, { "epoch": 0.49, "learning_rate": 5.1267300820373456e-05, "loss": 1.0461, "step": 194730 }, { "epoch": 0.49, "learning_rate": 5.126604257889175e-05, "loss": 1.0457, "step": 194735 }, { "epoch": 0.49, "learning_rate": 5.1264784337410046e-05, "loss": 1.0465, "step": 194740 }, { "epoch": 0.49, "learning_rate": 5.126352609592834e-05, "loss": 1.0428, "step": 194745 }, { "epoch": 0.49, "learning_rate": 5.1262267854446635e-05, "loss": 1.0447, "step": 194750 }, { "epoch": 0.49, "learning_rate": 5.126100961296493e-05, "loss": 1.047, "step": 194755 }, { "epoch": 0.49, "learning_rate": 5.1259751371483225e-05, "loss": 1.0425, "step": 194760 }, { "epoch": 0.49, "learning_rate": 5.125849313000152e-05, "loss": 1.0421, "step": 194765 }, { "epoch": 0.49, "learning_rate": 5.125723488851981e-05, "loss": 1.044, "step": 194770 }, { "epoch": 0.49, "learning_rate": 5.125597664703811e-05, "loss": 1.0446, "step": 194775 }, { "epoch": 0.49, "learning_rate": 5.12547184055564e-05, "loss": 1.0445, "step": 194780 }, { "epoch": 0.49, "learning_rate": 5.12534601640747e-05, "loss": 1.0455, "step": 194785 }, { "epoch": 0.49, "learning_rate": 5.125220192259299e-05, "loss": 1.0437, "step": 194790 }, { "epoch": 0.49, "learning_rate": 5.125094368111129e-05, "loss": 1.0478, "step": 194795 }, { "epoch": 0.49, "learning_rate": 5.124968543962958e-05, "loss": 1.0432, "step": 194800 }, { "epoch": 0.49, "learning_rate": 5.124842719814788e-05, "loss": 1.0469, "step": 194805 }, { "epoch": 0.49, "learning_rate": 5.124716895666617e-05, "loss": 1.0444, "step": 194810 }, { "epoch": 0.49, "learning_rate": 5.124591071518447e-05, "loss": 1.0449, "step": 194815 }, { "epoch": 0.49, "learning_rate": 5.124465247370276e-05, "loss": 1.0468, "step": 194820 }, { "epoch": 0.49, "learning_rate": 5.124339423222106e-05, "loss": 1.0418, "step": 194825 }, { "epoch": 0.49, "learning_rate": 5.124213599073935e-05, "loss": 1.0459, "step": 194830 }, { "epoch": 0.49, "learning_rate": 5.124087774925764e-05, "loss": 1.0416, "step": 194835 }, { "epoch": 0.49, "learning_rate": 5.123961950777594e-05, "loss": 1.0431, "step": 194840 }, { "epoch": 0.49, "learning_rate": 5.123836126629423e-05, "loss": 1.0443, "step": 194845 }, { "epoch": 0.49, "learning_rate": 5.123710302481253e-05, "loss": 1.0456, "step": 194850 }, { "epoch": 0.49, "learning_rate": 5.123584478333082e-05, "loss": 1.0424, "step": 194855 }, { "epoch": 0.49, "learning_rate": 5.123458654184912e-05, "loss": 1.0456, "step": 194860 }, { "epoch": 0.49, "learning_rate": 5.123332830036741e-05, "loss": 1.044, "step": 194865 }, { "epoch": 0.49, "learning_rate": 5.123207005888571e-05, "loss": 1.044, "step": 194870 }, { "epoch": 0.49, "learning_rate": 5.1230811817404e-05, "loss": 1.0458, "step": 194875 }, { "epoch": 0.49, "learning_rate": 5.12295535759223e-05, "loss": 1.0478, "step": 194880 }, { "epoch": 0.49, "learning_rate": 5.122829533444059e-05, "loss": 1.0469, "step": 194885 }, { "epoch": 0.49, "learning_rate": 5.122703709295889e-05, "loss": 1.0451, "step": 194890 }, { "epoch": 0.49, "learning_rate": 5.122577885147718e-05, "loss": 1.0426, "step": 194895 }, { "epoch": 0.49, "learning_rate": 5.122452060999547e-05, "loss": 1.0465, "step": 194900 }, { "epoch": 0.49, "learning_rate": 5.122326236851377e-05, "loss": 1.0446, "step": 194905 }, { "epoch": 0.49, "learning_rate": 5.122200412703206e-05, "loss": 1.0453, "step": 194910 }, { "epoch": 0.49, "learning_rate": 5.122074588555036e-05, "loss": 1.0455, "step": 194915 }, { "epoch": 0.49, "learning_rate": 5.121948764406865e-05, "loss": 1.0439, "step": 194920 }, { "epoch": 0.49, "learning_rate": 5.121822940258695e-05, "loss": 1.045, "step": 194925 }, { "epoch": 0.49, "learning_rate": 5.121697116110524e-05, "loss": 1.0449, "step": 194930 }, { "epoch": 0.49, "learning_rate": 5.121571291962354e-05, "loss": 1.0455, "step": 194935 }, { "epoch": 0.49, "learning_rate": 5.121445467814183e-05, "loss": 1.043, "step": 194940 }, { "epoch": 0.49, "learning_rate": 5.121319643666013e-05, "loss": 1.045, "step": 194945 }, { "epoch": 0.49, "learning_rate": 5.121193819517842e-05, "loss": 1.0445, "step": 194950 }, { "epoch": 0.49, "learning_rate": 5.121067995369671e-05, "loss": 1.045, "step": 194955 }, { "epoch": 0.49, "learning_rate": 5.120942171221501e-05, "loss": 1.0452, "step": 194960 }, { "epoch": 0.49, "learning_rate": 5.12081634707333e-05, "loss": 1.0435, "step": 194965 }, { "epoch": 0.49, "learning_rate": 5.12069052292516e-05, "loss": 1.0454, "step": 194970 }, { "epoch": 0.49, "learning_rate": 5.120564698776989e-05, "loss": 1.0452, "step": 194975 }, { "epoch": 0.49, "learning_rate": 5.120438874628819e-05, "loss": 1.0417, "step": 194980 }, { "epoch": 0.49, "learning_rate": 5.120313050480648e-05, "loss": 1.0451, "step": 194985 }, { "epoch": 0.49, "learning_rate": 5.120187226332478e-05, "loss": 1.0456, "step": 194990 }, { "epoch": 0.49, "learning_rate": 5.120061402184307e-05, "loss": 1.045, "step": 194995 }, { "epoch": 0.49, "learning_rate": 5.119935578036137e-05, "loss": 1.0444, "step": 195000 }, { "epoch": 0.49, "learning_rate": 5.119809753887966e-05, "loss": 1.0431, "step": 195005 }, { "epoch": 0.49, "learning_rate": 5.119683929739796e-05, "loss": 1.0472, "step": 195010 }, { "epoch": 0.49, "learning_rate": 5.119558105591625e-05, "loss": 1.0427, "step": 195015 }, { "epoch": 0.49, "learning_rate": 5.1194322814434544e-05, "loss": 1.0443, "step": 195020 }, { "epoch": 0.49, "learning_rate": 5.119306457295284e-05, "loss": 1.0473, "step": 195025 }, { "epoch": 0.49, "learning_rate": 5.1191806331471134e-05, "loss": 1.0441, "step": 195030 }, { "epoch": 0.49, "learning_rate": 5.119054808998943e-05, "loss": 1.0421, "step": 195035 }, { "epoch": 0.49, "learning_rate": 5.1189289848507724e-05, "loss": 1.0475, "step": 195040 }, { "epoch": 0.49, "learning_rate": 5.118803160702602e-05, "loss": 1.0481, "step": 195045 }, { "epoch": 0.49, "learning_rate": 5.1186773365544314e-05, "loss": 1.0438, "step": 195050 }, { "epoch": 0.49, "learning_rate": 5.118551512406261e-05, "loss": 1.0468, "step": 195055 }, { "epoch": 0.49, "learning_rate": 5.1184256882580904e-05, "loss": 1.0459, "step": 195060 }, { "epoch": 0.49, "learning_rate": 5.11829986410992e-05, "loss": 1.0434, "step": 195065 }, { "epoch": 0.49, "learning_rate": 5.1181740399617494e-05, "loss": 1.0441, "step": 195070 }, { "epoch": 0.49, "learning_rate": 5.118048215813579e-05, "loss": 1.0422, "step": 195075 }, { "epoch": 0.49, "learning_rate": 5.1179223916654084e-05, "loss": 1.046, "step": 195080 }, { "epoch": 0.49, "learning_rate": 5.1177965675172376e-05, "loss": 1.0468, "step": 195085 }, { "epoch": 0.49, "learning_rate": 5.1176707433690674e-05, "loss": 1.0452, "step": 195090 }, { "epoch": 0.49, "learning_rate": 5.1175449192208966e-05, "loss": 1.0482, "step": 195095 }, { "epoch": 0.49, "learning_rate": 5.1174190950727264e-05, "loss": 1.0456, "step": 195100 }, { "epoch": 0.49, "learning_rate": 5.1172932709245556e-05, "loss": 1.0454, "step": 195105 }, { "epoch": 0.49, "learning_rate": 5.1171674467763854e-05, "loss": 1.0457, "step": 195110 }, { "epoch": 0.49, "learning_rate": 5.1170416226282146e-05, "loss": 1.0421, "step": 195115 }, { "epoch": 0.49, "learning_rate": 5.1169157984800444e-05, "loss": 1.0423, "step": 195120 }, { "epoch": 0.49, "learning_rate": 5.1167899743318736e-05, "loss": 1.0439, "step": 195125 }, { "epoch": 0.49, "learning_rate": 5.1166641501837034e-05, "loss": 1.044, "step": 195130 }, { "epoch": 0.49, "learning_rate": 5.1165383260355326e-05, "loss": 1.0467, "step": 195135 }, { "epoch": 0.49, "learning_rate": 5.1164125018873624e-05, "loss": 1.0441, "step": 195140 }, { "epoch": 0.49, "learning_rate": 5.1162866777391916e-05, "loss": 1.0456, "step": 195145 }, { "epoch": 0.49, "learning_rate": 5.116160853591021e-05, "loss": 1.0426, "step": 195150 }, { "epoch": 0.49, "learning_rate": 5.1160350294428506e-05, "loss": 1.0442, "step": 195155 }, { "epoch": 0.49, "learning_rate": 5.11590920529468e-05, "loss": 1.0455, "step": 195160 }, { "epoch": 0.49, "learning_rate": 5.1157833811465096e-05, "loss": 1.0422, "step": 195165 }, { "epoch": 0.49, "learning_rate": 5.115657556998339e-05, "loss": 1.0423, "step": 195170 }, { "epoch": 0.49, "learning_rate": 5.1155317328501686e-05, "loss": 1.0457, "step": 195175 }, { "epoch": 0.49, "learning_rate": 5.115405908701998e-05, "loss": 1.0446, "step": 195180 }, { "epoch": 0.49, "learning_rate": 5.1152800845538276e-05, "loss": 1.0432, "step": 195185 }, { "epoch": 0.49, "learning_rate": 5.115154260405657e-05, "loss": 1.0435, "step": 195190 }, { "epoch": 0.49, "learning_rate": 5.1150284362574865e-05, "loss": 1.0447, "step": 195195 }, { "epoch": 0.49, "learning_rate": 5.114902612109316e-05, "loss": 1.043, "step": 195200 }, { "epoch": 0.49, "learning_rate": 5.1147767879611455e-05, "loss": 1.0716, "step": 195205 }, { "epoch": 0.49, "learning_rate": 5.114650963812975e-05, "loss": 1.0444, "step": 195210 }, { "epoch": 0.49, "learning_rate": 5.114525139664804e-05, "loss": 1.0492, "step": 195215 }, { "epoch": 0.49, "learning_rate": 5.114399315516634e-05, "loss": 1.0463, "step": 195220 }, { "epoch": 0.49, "learning_rate": 5.114273491368464e-05, "loss": 1.0454, "step": 195225 }, { "epoch": 0.49, "learning_rate": 5.1141476672202934e-05, "loss": 1.0448, "step": 195230 }, { "epoch": 0.49, "learning_rate": 5.114021843072123e-05, "loss": 1.0453, "step": 195235 }, { "epoch": 0.49, "learning_rate": 5.1138960189239524e-05, "loss": 1.0447, "step": 195240 }, { "epoch": 0.49, "learning_rate": 5.113770194775782e-05, "loss": 1.0458, "step": 195245 }, { "epoch": 0.49, "learning_rate": 5.1136443706276114e-05, "loss": 1.0458, "step": 195250 }, { "epoch": 0.49, "learning_rate": 5.113518546479441e-05, "loss": 1.0458, "step": 195255 }, { "epoch": 0.49, "learning_rate": 5.1133927223312704e-05, "loss": 1.0427, "step": 195260 }, { "epoch": 0.49, "learning_rate": 5.1132668981831e-05, "loss": 1.0657, "step": 195265 }, { "epoch": 0.49, "learning_rate": 5.1131410740349294e-05, "loss": 1.0471, "step": 195270 }, { "epoch": 0.49, "learning_rate": 5.113015249886759e-05, "loss": 1.0432, "step": 195275 }, { "epoch": 0.49, "learning_rate": 5.1128894257385884e-05, "loss": 1.0455, "step": 195280 }, { "epoch": 0.49, "learning_rate": 5.112763601590418e-05, "loss": 1.0434, "step": 195285 }, { "epoch": 0.49, "learning_rate": 5.1126377774422474e-05, "loss": 1.0447, "step": 195290 }, { "epoch": 0.49, "learning_rate": 5.1125119532940765e-05, "loss": 1.0433, "step": 195295 }, { "epoch": 0.49, "learning_rate": 5.1123861291459064e-05, "loss": 1.0444, "step": 195300 }, { "epoch": 0.49, "learning_rate": 5.1122603049977355e-05, "loss": 1.0459, "step": 195305 }, { "epoch": 0.49, "learning_rate": 5.1121344808495654e-05, "loss": 1.0425, "step": 195310 }, { "epoch": 0.49, "learning_rate": 5.1120086567013945e-05, "loss": 1.0454, "step": 195315 }, { "epoch": 0.49, "learning_rate": 5.1118828325532244e-05, "loss": 1.0412, "step": 195320 }, { "epoch": 0.49, "learning_rate": 5.1117570084050535e-05, "loss": 1.0435, "step": 195325 }, { "epoch": 0.49, "learning_rate": 5.1116311842568833e-05, "loss": 1.0472, "step": 195330 }, { "epoch": 0.49, "learning_rate": 5.1115053601087125e-05, "loss": 1.0469, "step": 195335 }, { "epoch": 0.49, "learning_rate": 5.1113795359605423e-05, "loss": 1.0445, "step": 195340 }, { "epoch": 0.49, "learning_rate": 5.1112537118123715e-05, "loss": 1.0442, "step": 195345 }, { "epoch": 0.49, "learning_rate": 5.1111278876642013e-05, "loss": 1.0412, "step": 195350 }, { "epoch": 0.49, "learning_rate": 5.1110020635160305e-05, "loss": 1.0445, "step": 195355 }, { "epoch": 0.49, "learning_rate": 5.11087623936786e-05, "loss": 1.0434, "step": 195360 }, { "epoch": 0.49, "learning_rate": 5.1107504152196895e-05, "loss": 1.0431, "step": 195365 }, { "epoch": 0.49, "learning_rate": 5.1106245910715187e-05, "loss": 1.0455, "step": 195370 }, { "epoch": 0.49, "learning_rate": 5.1104987669233485e-05, "loss": 1.0689, "step": 195375 }, { "epoch": 0.49, "learning_rate": 5.1103729427751777e-05, "loss": 1.0458, "step": 195380 }, { "epoch": 0.49, "learning_rate": 5.1102471186270075e-05, "loss": 1.0475, "step": 195385 }, { "epoch": 0.49, "learning_rate": 5.1101212944788367e-05, "loss": 1.0468, "step": 195390 }, { "epoch": 0.49, "learning_rate": 5.1099954703306665e-05, "loss": 1.0445, "step": 195395 }, { "epoch": 0.49, "learning_rate": 5.1098696461824957e-05, "loss": 1.0454, "step": 195400 }, { "epoch": 0.49, "learning_rate": 5.1097438220343255e-05, "loss": 1.0446, "step": 195405 }, { "epoch": 0.49, "learning_rate": 5.1096179978861546e-05, "loss": 1.0439, "step": 195410 }, { "epoch": 0.49, "learning_rate": 5.1094921737379845e-05, "loss": 1.0645, "step": 195415 }, { "epoch": 0.49, "learning_rate": 5.1093663495898136e-05, "loss": 1.0454, "step": 195420 }, { "epoch": 0.49, "learning_rate": 5.109240525441643e-05, "loss": 1.0449, "step": 195425 }, { "epoch": 0.49, "learning_rate": 5.1091147012934726e-05, "loss": 1.0447, "step": 195430 }, { "epoch": 0.49, "learning_rate": 5.108988877145302e-05, "loss": 1.0461, "step": 195435 }, { "epoch": 0.49, "learning_rate": 5.1088630529971316e-05, "loss": 1.0458, "step": 195440 }, { "epoch": 0.49, "learning_rate": 5.108737228848961e-05, "loss": 1.0445, "step": 195445 }, { "epoch": 0.49, "learning_rate": 5.1086114047007906e-05, "loss": 1.0439, "step": 195450 }, { "epoch": 0.49, "learning_rate": 5.10848558055262e-05, "loss": 1.0441, "step": 195455 }, { "epoch": 0.49, "learning_rate": 5.1083597564044496e-05, "loss": 1.0448, "step": 195460 }, { "epoch": 0.49, "learning_rate": 5.108233932256279e-05, "loss": 1.0432, "step": 195465 }, { "epoch": 0.49, "learning_rate": 5.1081081081081086e-05, "loss": 1.0427, "step": 195470 }, { "epoch": 0.49, "learning_rate": 5.107982283959938e-05, "loss": 1.0445, "step": 195475 }, { "epoch": 0.49, "learning_rate": 5.1078564598117676e-05, "loss": 1.0439, "step": 195480 }, { "epoch": 0.49, "learning_rate": 5.107730635663597e-05, "loss": 1.0685, "step": 195485 }, { "epoch": 0.49, "learning_rate": 5.107604811515426e-05, "loss": 1.0493, "step": 195490 }, { "epoch": 0.49, "learning_rate": 5.107478987367256e-05, "loss": 1.0456, "step": 195495 }, { "epoch": 0.49, "learning_rate": 5.107353163219085e-05, "loss": 1.0485, "step": 195500 }, { "epoch": 0.49, "learning_rate": 5.107227339070915e-05, "loss": 1.0456, "step": 195505 }, { "epoch": 0.49, "learning_rate": 5.107101514922744e-05, "loss": 1.0447, "step": 195510 }, { "epoch": 0.49, "learning_rate": 5.106975690774574e-05, "loss": 1.0455, "step": 195515 }, { "epoch": 0.49, "learning_rate": 5.106849866626403e-05, "loss": 1.0456, "step": 195520 }, { "epoch": 0.49, "learning_rate": 5.106724042478233e-05, "loss": 1.0454, "step": 195525 }, { "epoch": 0.49, "learning_rate": 5.106598218330062e-05, "loss": 1.0456, "step": 195530 }, { "epoch": 0.49, "learning_rate": 5.106472394181892e-05, "loss": 1.0438, "step": 195535 }, { "epoch": 0.49, "learning_rate": 5.106346570033721e-05, "loss": 1.0443, "step": 195540 }, { "epoch": 0.49, "learning_rate": 5.106220745885551e-05, "loss": 1.047, "step": 195545 }, { "epoch": 0.49, "learning_rate": 5.10609492173738e-05, "loss": 1.0447, "step": 195550 }, { "epoch": 0.49, "learning_rate": 5.105969097589209e-05, "loss": 1.0447, "step": 195555 }, { "epoch": 0.49, "learning_rate": 5.105843273441039e-05, "loss": 1.0459, "step": 195560 }, { "epoch": 0.49, "learning_rate": 5.105717449292868e-05, "loss": 1.0467, "step": 195565 }, { "epoch": 0.49, "learning_rate": 5.105591625144698e-05, "loss": 1.0437, "step": 195570 }, { "epoch": 0.49, "learning_rate": 5.105465800996527e-05, "loss": 1.0464, "step": 195575 }, { "epoch": 0.49, "learning_rate": 5.105339976848357e-05, "loss": 1.0435, "step": 195580 }, { "epoch": 0.49, "learning_rate": 5.105214152700186e-05, "loss": 1.045, "step": 195585 }, { "epoch": 0.49, "learning_rate": 5.105088328552016e-05, "loss": 1.0437, "step": 195590 }, { "epoch": 0.49, "learning_rate": 5.104962504403845e-05, "loss": 1.0437, "step": 195595 }, { "epoch": 0.49, "learning_rate": 5.104836680255675e-05, "loss": 1.044, "step": 195600 }, { "epoch": 0.49, "learning_rate": 5.104710856107504e-05, "loss": 1.0595, "step": 195605 }, { "epoch": 0.49, "learning_rate": 5.104585031959334e-05, "loss": 1.0461, "step": 195610 }, { "epoch": 0.49, "learning_rate": 5.104459207811163e-05, "loss": 1.0448, "step": 195615 }, { "epoch": 0.49, "learning_rate": 5.104333383662992e-05, "loss": 1.0445, "step": 195620 }, { "epoch": 0.49, "learning_rate": 5.104207559514822e-05, "loss": 1.0436, "step": 195625 }, { "epoch": 0.49, "learning_rate": 5.104081735366651e-05, "loss": 1.046, "step": 195630 }, { "epoch": 0.49, "learning_rate": 5.103955911218481e-05, "loss": 1.0453, "step": 195635 }, { "epoch": 0.49, "learning_rate": 5.10383008707031e-05, "loss": 1.0451, "step": 195640 }, { "epoch": 0.49, "learning_rate": 5.10370426292214e-05, "loss": 1.0481, "step": 195645 }, { "epoch": 0.49, "learning_rate": 5.103578438773969e-05, "loss": 1.0462, "step": 195650 }, { "epoch": 0.49, "learning_rate": 5.103452614625799e-05, "loss": 1.0453, "step": 195655 }, { "epoch": 0.49, "learning_rate": 5.103326790477628e-05, "loss": 1.0421, "step": 195660 }, { "epoch": 0.49, "learning_rate": 5.103200966329458e-05, "loss": 1.0461, "step": 195665 }, { "epoch": 0.49, "learning_rate": 5.103075142181287e-05, "loss": 1.0444, "step": 195670 }, { "epoch": 0.49, "learning_rate": 5.1029493180331164e-05, "loss": 1.0448, "step": 195675 }, { "epoch": 0.49, "learning_rate": 5.102823493884946e-05, "loss": 1.0706, "step": 195680 }, { "epoch": 0.49, "learning_rate": 5.1026976697367754e-05, "loss": 1.0449, "step": 195685 }, { "epoch": 0.49, "learning_rate": 5.102571845588605e-05, "loss": 1.0445, "step": 195690 }, { "epoch": 0.49, "learning_rate": 5.1024460214404344e-05, "loss": 1.0436, "step": 195695 }, { "epoch": 0.49, "learning_rate": 5.102320197292264e-05, "loss": 1.0443, "step": 195700 }, { "epoch": 0.49, "learning_rate": 5.1021943731440934e-05, "loss": 1.049, "step": 195705 }, { "epoch": 0.49, "learning_rate": 5.102068548995923e-05, "loss": 1.0428, "step": 195710 }, { "epoch": 0.49, "learning_rate": 5.1019427248477524e-05, "loss": 1.0439, "step": 195715 }, { "epoch": 0.49, "learning_rate": 5.101816900699582e-05, "loss": 1.0423, "step": 195720 }, { "epoch": 0.49, "learning_rate": 5.101691076551413e-05, "loss": 1.0461, "step": 195725 }, { "epoch": 0.49, "learning_rate": 5.101565252403242e-05, "loss": 1.0448, "step": 195730 }, { "epoch": 0.49, "learning_rate": 5.101439428255072e-05, "loss": 1.0422, "step": 195735 }, { "epoch": 0.49, "learning_rate": 5.101313604106901e-05, "loss": 1.0443, "step": 195740 }, { "epoch": 0.49, "learning_rate": 5.101187779958731e-05, "loss": 1.0468, "step": 195745 }, { "epoch": 0.49, "learning_rate": 5.10106195581056e-05, "loss": 1.0436, "step": 195750 }, { "epoch": 0.49, "learning_rate": 5.10093613166239e-05, "loss": 1.0473, "step": 195755 }, { "epoch": 0.49, "learning_rate": 5.100810307514219e-05, "loss": 1.0446, "step": 195760 }, { "epoch": 0.49, "learning_rate": 5.100684483366048e-05, "loss": 1.0472, "step": 195765 }, { "epoch": 0.49, "learning_rate": 5.100558659217878e-05, "loss": 1.0453, "step": 195770 }, { "epoch": 0.49, "learning_rate": 5.100432835069707e-05, "loss": 1.0447, "step": 195775 }, { "epoch": 0.49, "learning_rate": 5.100307010921537e-05, "loss": 1.0471, "step": 195780 }, { "epoch": 0.49, "learning_rate": 5.100181186773366e-05, "loss": 1.0462, "step": 195785 }, { "epoch": 0.49, "learning_rate": 5.100055362625196e-05, "loss": 1.047, "step": 195790 }, { "epoch": 0.49, "learning_rate": 5.099929538477025e-05, "loss": 1.0428, "step": 195795 }, { "epoch": 0.49, "learning_rate": 5.099803714328855e-05, "loss": 1.0408, "step": 195800 }, { "epoch": 0.49, "learning_rate": 5.099677890180684e-05, "loss": 1.0434, "step": 195805 }, { "epoch": 0.49, "learning_rate": 5.099552066032514e-05, "loss": 1.0434, "step": 195810 }, { "epoch": 0.49, "learning_rate": 5.099426241884343e-05, "loss": 1.0468, "step": 195815 }, { "epoch": 0.49, "learning_rate": 5.099300417736172e-05, "loss": 1.0429, "step": 195820 }, { "epoch": 0.49, "learning_rate": 5.099174593588002e-05, "loss": 1.0442, "step": 195825 }, { "epoch": 0.49, "learning_rate": 5.099048769439831e-05, "loss": 1.0447, "step": 195830 }, { "epoch": 0.49, "learning_rate": 5.098922945291661e-05, "loss": 1.0445, "step": 195835 }, { "epoch": 0.49, "learning_rate": 5.09879712114349e-05, "loss": 1.0449, "step": 195840 }, { "epoch": 0.49, "learning_rate": 5.09867129699532e-05, "loss": 1.0428, "step": 195845 }, { "epoch": 0.49, "learning_rate": 5.098545472847149e-05, "loss": 1.0461, "step": 195850 }, { "epoch": 0.49, "learning_rate": 5.098419648698979e-05, "loss": 1.0452, "step": 195855 }, { "epoch": 0.49, "learning_rate": 5.098293824550808e-05, "loss": 1.0442, "step": 195860 }, { "epoch": 0.49, "learning_rate": 5.098168000402638e-05, "loss": 1.0464, "step": 195865 }, { "epoch": 0.49, "learning_rate": 5.098042176254467e-05, "loss": 1.0432, "step": 195870 }, { "epoch": 0.49, "learning_rate": 5.097916352106297e-05, "loss": 1.0451, "step": 195875 }, { "epoch": 0.49, "learning_rate": 5.097790527958126e-05, "loss": 1.0439, "step": 195880 }, { "epoch": 0.49, "learning_rate": 5.097664703809955e-05, "loss": 1.0463, "step": 195885 }, { "epoch": 0.49, "learning_rate": 5.097538879661785e-05, "loss": 1.0445, "step": 195890 }, { "epoch": 0.49, "learning_rate": 5.097413055513614e-05, "loss": 1.0481, "step": 195895 }, { "epoch": 0.49, "learning_rate": 5.097287231365444e-05, "loss": 1.0477, "step": 195900 }, { "epoch": 0.49, "learning_rate": 5.097161407217273e-05, "loss": 1.0451, "step": 195905 }, { "epoch": 0.49, "learning_rate": 5.097035583069103e-05, "loss": 1.0432, "step": 195910 }, { "epoch": 0.49, "learning_rate": 5.096909758920932e-05, "loss": 1.0433, "step": 195915 }, { "epoch": 0.49, "learning_rate": 5.096783934772762e-05, "loss": 1.0674, "step": 195920 }, { "epoch": 0.49, "learning_rate": 5.096658110624591e-05, "loss": 1.0476, "step": 195925 }, { "epoch": 0.49, "learning_rate": 5.096532286476421e-05, "loss": 1.0449, "step": 195930 }, { "epoch": 0.49, "learning_rate": 5.09640646232825e-05, "loss": 1.0448, "step": 195935 }, { "epoch": 0.49, "learning_rate": 5.09628063818008e-05, "loss": 1.0437, "step": 195940 }, { "epoch": 0.49, "learning_rate": 5.096154814031909e-05, "loss": 1.0456, "step": 195945 }, { "epoch": 0.49, "learning_rate": 5.0960289898837385e-05, "loss": 1.0442, "step": 195950 }, { "epoch": 0.49, "learning_rate": 5.095903165735568e-05, "loss": 1.0427, "step": 195955 }, { "epoch": 0.49, "learning_rate": 5.0957773415873975e-05, "loss": 1.0411, "step": 195960 }, { "epoch": 0.49, "learning_rate": 5.095651517439227e-05, "loss": 1.0448, "step": 195965 }, { "epoch": 0.49, "learning_rate": 5.0955256932910565e-05, "loss": 1.0462, "step": 195970 }, { "epoch": 0.49, "learning_rate": 5.095399869142886e-05, "loss": 1.0466, "step": 195975 }, { "epoch": 0.49, "learning_rate": 5.0952740449947155e-05, "loss": 1.0452, "step": 195980 }, { "epoch": 0.49, "learning_rate": 5.095148220846545e-05, "loss": 1.0693, "step": 195985 }, { "epoch": 0.49, "learning_rate": 5.0950223966983744e-05, "loss": 1.0427, "step": 195990 }, { "epoch": 0.49, "learning_rate": 5.094896572550204e-05, "loss": 1.045, "step": 195995 }, { "epoch": 0.49, "learning_rate": 5.0947707484020334e-05, "loss": 1.043, "step": 196000 }, { "epoch": 0.49, "learning_rate": 5.094644924253863e-05, "loss": 1.0682, "step": 196005 }, { "epoch": 0.49, "learning_rate": 5.0945191001056924e-05, "loss": 1.0452, "step": 196010 }, { "epoch": 0.49, "learning_rate": 5.0943932759575216e-05, "loss": 1.0463, "step": 196015 }, { "epoch": 0.49, "learning_rate": 5.0942674518093514e-05, "loss": 1.0478, "step": 196020 }, { "epoch": 0.49, "learning_rate": 5.0941416276611806e-05, "loss": 1.0457, "step": 196025 }, { "epoch": 0.49, "learning_rate": 5.0940158035130104e-05, "loss": 1.0439, "step": 196030 }, { "epoch": 0.49, "learning_rate": 5.0938899793648396e-05, "loss": 1.0486, "step": 196035 }, { "epoch": 0.49, "learning_rate": 5.0937641552166694e-05, "loss": 1.0435, "step": 196040 }, { "epoch": 0.49, "learning_rate": 5.0936383310684986e-05, "loss": 1.0417, "step": 196045 }, { "epoch": 0.49, "learning_rate": 5.0935125069203284e-05, "loss": 1.0454, "step": 196050 }, { "epoch": 0.49, "learning_rate": 5.0933866827721576e-05, "loss": 1.0414, "step": 196055 }, { "epoch": 0.49, "learning_rate": 5.0932608586239874e-05, "loss": 1.0455, "step": 196060 }, { "epoch": 0.49, "learning_rate": 5.0931350344758166e-05, "loss": 1.0442, "step": 196065 }, { "epoch": 0.49, "learning_rate": 5.0930092103276464e-05, "loss": 1.0444, "step": 196070 }, { "epoch": 0.49, "learning_rate": 5.0928833861794756e-05, "loss": 1.0464, "step": 196075 }, { "epoch": 0.49, "learning_rate": 5.092757562031305e-05, "loss": 1.0431, "step": 196080 }, { "epoch": 0.49, "learning_rate": 5.0926317378831346e-05, "loss": 1.0482, "step": 196085 }, { "epoch": 0.49, "learning_rate": 5.092505913734964e-05, "loss": 1.0459, "step": 196090 }, { "epoch": 0.49, "learning_rate": 5.0923800895867936e-05, "loss": 1.0445, "step": 196095 }, { "epoch": 0.49, "learning_rate": 5.092254265438623e-05, "loss": 1.0442, "step": 196100 }, { "epoch": 0.49, "learning_rate": 5.0921284412904526e-05, "loss": 1.0439, "step": 196105 }, { "epoch": 0.49, "learning_rate": 5.092002617142282e-05, "loss": 1.0433, "step": 196110 }, { "epoch": 0.49, "learning_rate": 5.0918767929941116e-05, "loss": 1.0468, "step": 196115 }, { "epoch": 0.49, "learning_rate": 5.091750968845941e-05, "loss": 1.0475, "step": 196120 }, { "epoch": 0.49, "learning_rate": 5.0916251446977706e-05, "loss": 1.0435, "step": 196125 }, { "epoch": 0.49, "learning_rate": 5.0914993205496e-05, "loss": 1.0468, "step": 196130 }, { "epoch": 0.49, "learning_rate": 5.0913734964014296e-05, "loss": 1.0445, "step": 196135 }, { "epoch": 0.49, "learning_rate": 5.091247672253259e-05, "loss": 1.044, "step": 196140 }, { "epoch": 0.49, "learning_rate": 5.091121848105088e-05, "loss": 1.0442, "step": 196145 }, { "epoch": 0.49, "learning_rate": 5.090996023956918e-05, "loss": 1.0428, "step": 196150 }, { "epoch": 0.49, "learning_rate": 5.090870199808747e-05, "loss": 1.0464, "step": 196155 }, { "epoch": 0.49, "learning_rate": 5.090744375660577e-05, "loss": 1.0457, "step": 196160 }, { "epoch": 0.49, "learning_rate": 5.090618551512406e-05, "loss": 1.0447, "step": 196165 }, { "epoch": 0.49, "learning_rate": 5.090492727364236e-05, "loss": 1.0447, "step": 196170 }, { "epoch": 0.49, "learning_rate": 5.090366903216065e-05, "loss": 1.0434, "step": 196175 }, { "epoch": 0.49, "learning_rate": 5.090241079067895e-05, "loss": 1.045, "step": 196180 }, { "epoch": 0.49, "learning_rate": 5.090115254919724e-05, "loss": 1.0419, "step": 196185 }, { "epoch": 0.49, "learning_rate": 5.089989430771554e-05, "loss": 1.0444, "step": 196190 }, { "epoch": 0.49, "learning_rate": 5.089863606623383e-05, "loss": 1.0455, "step": 196195 }, { "epoch": 0.49, "learning_rate": 5.089737782475213e-05, "loss": 1.0449, "step": 196200 }, { "epoch": 0.49, "learning_rate": 5.089611958327042e-05, "loss": 1.045, "step": 196205 }, { "epoch": 0.49, "learning_rate": 5.089486134178871e-05, "loss": 1.0428, "step": 196210 }, { "epoch": 0.49, "learning_rate": 5.089360310030701e-05, "loss": 1.0456, "step": 196215 }, { "epoch": 0.49, "learning_rate": 5.08923448588253e-05, "loss": 1.0437, "step": 196220 }, { "epoch": 0.49, "learning_rate": 5.0891086617343605e-05, "loss": 1.0455, "step": 196225 }, { "epoch": 0.49, "learning_rate": 5.0889828375861904e-05, "loss": 1.0458, "step": 196230 }, { "epoch": 0.49, "learning_rate": 5.0888570134380195e-05, "loss": 1.0442, "step": 196235 }, { "epoch": 0.49, "learning_rate": 5.0887311892898494e-05, "loss": 1.0443, "step": 196240 }, { "epoch": 0.49, "learning_rate": 5.0886053651416785e-05, "loss": 1.0458, "step": 196245 }, { "epoch": 0.49, "learning_rate": 5.0884795409935084e-05, "loss": 1.0453, "step": 196250 }, { "epoch": 0.49, "learning_rate": 5.0883537168453375e-05, "loss": 1.0449, "step": 196255 }, { "epoch": 0.49, "learning_rate": 5.0882278926971674e-05, "loss": 1.0475, "step": 196260 }, { "epoch": 0.49, "learning_rate": 5.0881020685489965e-05, "loss": 1.0465, "step": 196265 }, { "epoch": 0.49, "learning_rate": 5.0879762444008264e-05, "loss": 1.0435, "step": 196270 }, { "epoch": 0.49, "learning_rate": 5.0878504202526555e-05, "loss": 1.0449, "step": 196275 }, { "epoch": 0.49, "learning_rate": 5.0877245961044854e-05, "loss": 1.0479, "step": 196280 }, { "epoch": 0.49, "learning_rate": 5.0875987719563145e-05, "loss": 1.0452, "step": 196285 }, { "epoch": 0.49, "learning_rate": 5.087472947808144e-05, "loss": 1.0443, "step": 196290 }, { "epoch": 0.49, "learning_rate": 5.0873471236599735e-05, "loss": 1.0429, "step": 196295 }, { "epoch": 0.49, "learning_rate": 5.087221299511803e-05, "loss": 1.0476, "step": 196300 }, { "epoch": 0.49, "learning_rate": 5.0870954753636325e-05, "loss": 1.0434, "step": 196305 }, { "epoch": 0.49, "learning_rate": 5.086969651215462e-05, "loss": 1.0444, "step": 196310 }, { "epoch": 0.49, "learning_rate": 5.0868438270672915e-05, "loss": 1.0456, "step": 196315 }, { "epoch": 0.49, "learning_rate": 5.086718002919121e-05, "loss": 1.0457, "step": 196320 }, { "epoch": 0.49, "learning_rate": 5.0865921787709505e-05, "loss": 1.045, "step": 196325 }, { "epoch": 0.49, "learning_rate": 5.08646635462278e-05, "loss": 1.0449, "step": 196330 }, { "epoch": 0.49, "learning_rate": 5.0863405304746095e-05, "loss": 1.0437, "step": 196335 }, { "epoch": 0.49, "learning_rate": 5.086214706326439e-05, "loss": 1.0431, "step": 196340 }, { "epoch": 0.49, "learning_rate": 5.0860888821782685e-05, "loss": 1.0453, "step": 196345 }, { "epoch": 0.49, "learning_rate": 5.085963058030098e-05, "loss": 1.0425, "step": 196350 }, { "epoch": 0.49, "learning_rate": 5.085837233881927e-05, "loss": 1.0488, "step": 196355 }, { "epoch": 0.49, "learning_rate": 5.085711409733757e-05, "loss": 1.0447, "step": 196360 }, { "epoch": 0.49, "learning_rate": 5.085585585585586e-05, "loss": 1.0475, "step": 196365 }, { "epoch": 0.49, "learning_rate": 5.0854597614374157e-05, "loss": 1.0475, "step": 196370 }, { "epoch": 0.49, "learning_rate": 5.085333937289245e-05, "loss": 1.0426, "step": 196375 }, { "epoch": 0.49, "learning_rate": 5.0852081131410747e-05, "loss": 1.0443, "step": 196380 }, { "epoch": 0.49, "learning_rate": 5.085082288992904e-05, "loss": 1.0444, "step": 196385 }, { "epoch": 0.49, "learning_rate": 5.0849564648447337e-05, "loss": 1.0452, "step": 196390 }, { "epoch": 0.49, "learning_rate": 5.084830640696563e-05, "loss": 1.0444, "step": 196395 }, { "epoch": 0.49, "learning_rate": 5.0847048165483927e-05, "loss": 1.0436, "step": 196400 }, { "epoch": 0.49, "learning_rate": 5.084578992400222e-05, "loss": 1.0435, "step": 196405 }, { "epoch": 0.49, "learning_rate": 5.0844531682520516e-05, "loss": 1.0621, "step": 196410 }, { "epoch": 0.49, "learning_rate": 5.084327344103881e-05, "loss": 1.0433, "step": 196415 }, { "epoch": 0.49, "learning_rate": 5.08420151995571e-05, "loss": 1.045, "step": 196420 }, { "epoch": 0.49, "learning_rate": 5.08407569580754e-05, "loss": 1.0447, "step": 196425 }, { "epoch": 0.49, "learning_rate": 5.083949871659369e-05, "loss": 1.048, "step": 196430 }, { "epoch": 0.49, "learning_rate": 5.083824047511199e-05, "loss": 1.043, "step": 196435 }, { "epoch": 0.49, "learning_rate": 5.083698223363028e-05, "loss": 1.0472, "step": 196440 }, { "epoch": 0.49, "learning_rate": 5.083572399214858e-05, "loss": 1.0474, "step": 196445 }, { "epoch": 0.49, "learning_rate": 5.083446575066687e-05, "loss": 1.0489, "step": 196450 }, { "epoch": 0.49, "learning_rate": 5.083320750918517e-05, "loss": 1.0454, "step": 196455 }, { "epoch": 0.49, "learning_rate": 5.083194926770346e-05, "loss": 1.0455, "step": 196460 }, { "epoch": 0.49, "learning_rate": 5.083069102622176e-05, "loss": 1.0464, "step": 196465 }, { "epoch": 0.49, "learning_rate": 5.082943278474005e-05, "loss": 1.0405, "step": 196470 }, { "epoch": 0.49, "learning_rate": 5.082817454325834e-05, "loss": 1.0467, "step": 196475 }, { "epoch": 0.49, "learning_rate": 5.082691630177664e-05, "loss": 1.0452, "step": 196480 }, { "epoch": 0.49, "learning_rate": 5.082565806029493e-05, "loss": 1.0428, "step": 196485 }, { "epoch": 0.49, "learning_rate": 5.082439981881323e-05, "loss": 1.0434, "step": 196490 }, { "epoch": 0.49, "learning_rate": 5.082314157733152e-05, "loss": 1.0451, "step": 196495 }, { "epoch": 0.49, "learning_rate": 5.082188333584982e-05, "loss": 1.0625, "step": 196500 }, { "epoch": 0.49, "learning_rate": 5.082062509436811e-05, "loss": 1.0445, "step": 196505 }, { "epoch": 0.49, "learning_rate": 5.081936685288641e-05, "loss": 1.047, "step": 196510 }, { "epoch": 0.49, "learning_rate": 5.08181086114047e-05, "loss": 1.0449, "step": 196515 }, { "epoch": 0.49, "learning_rate": 5.0816850369923e-05, "loss": 1.0452, "step": 196520 }, { "epoch": 0.49, "learning_rate": 5.081559212844129e-05, "loss": 1.0443, "step": 196525 }, { "epoch": 0.49, "learning_rate": 5.081433388695959e-05, "loss": 1.046, "step": 196530 }, { "epoch": 0.49, "learning_rate": 5.081307564547788e-05, "loss": 1.0423, "step": 196535 }, { "epoch": 0.49, "learning_rate": 5.081181740399617e-05, "loss": 1.0452, "step": 196540 }, { "epoch": 0.49, "learning_rate": 5.081055916251447e-05, "loss": 1.0478, "step": 196545 }, { "epoch": 0.49, "learning_rate": 5.080930092103276e-05, "loss": 1.0431, "step": 196550 }, { "epoch": 0.49, "learning_rate": 5.080804267955106e-05, "loss": 1.0452, "step": 196555 }, { "epoch": 0.49, "learning_rate": 5.080678443806935e-05, "loss": 1.0449, "step": 196560 }, { "epoch": 0.49, "learning_rate": 5.080552619658765e-05, "loss": 1.046, "step": 196565 }, { "epoch": 0.49, "learning_rate": 5.080426795510594e-05, "loss": 1.0478, "step": 196570 }, { "epoch": 0.49, "learning_rate": 5.080300971362424e-05, "loss": 1.0463, "step": 196575 }, { "epoch": 0.49, "learning_rate": 5.080175147214253e-05, "loss": 1.0433, "step": 196580 }, { "epoch": 0.49, "learning_rate": 5.080049323066083e-05, "loss": 1.0434, "step": 196585 }, { "epoch": 0.49, "learning_rate": 5.079923498917912e-05, "loss": 1.0459, "step": 196590 }, { "epoch": 0.49, "learning_rate": 5.079797674769742e-05, "loss": 1.0468, "step": 196595 }, { "epoch": 0.49, "learning_rate": 5.079671850621571e-05, "loss": 1.0435, "step": 196600 }, { "epoch": 0.49, "learning_rate": 5.0795460264734004e-05, "loss": 1.0438, "step": 196605 }, { "epoch": 0.49, "learning_rate": 5.07942020232523e-05, "loss": 1.043, "step": 196610 }, { "epoch": 0.49, "learning_rate": 5.0792943781770594e-05, "loss": 1.0452, "step": 196615 }, { "epoch": 0.49, "learning_rate": 5.079168554028889e-05, "loss": 1.0492, "step": 196620 }, { "epoch": 0.49, "learning_rate": 5.0790427298807184e-05, "loss": 1.0479, "step": 196625 }, { "epoch": 0.49, "learning_rate": 5.078916905732548e-05, "loss": 1.0452, "step": 196630 }, { "epoch": 0.49, "learning_rate": 5.0787910815843774e-05, "loss": 1.0462, "step": 196635 }, { "epoch": 0.49, "learning_rate": 5.078665257436207e-05, "loss": 1.0444, "step": 196640 }, { "epoch": 0.49, "learning_rate": 5.0785394332880364e-05, "loss": 1.046, "step": 196645 }, { "epoch": 0.49, "learning_rate": 5.078413609139866e-05, "loss": 1.044, "step": 196650 }, { "epoch": 0.49, "learning_rate": 5.0782877849916954e-05, "loss": 1.045, "step": 196655 }, { "epoch": 0.49, "learning_rate": 5.078161960843525e-05, "loss": 1.0442, "step": 196660 }, { "epoch": 0.49, "learning_rate": 5.0780361366953544e-05, "loss": 1.0455, "step": 196665 }, { "epoch": 0.49, "learning_rate": 5.0779103125471835e-05, "loss": 1.0466, "step": 196670 }, { "epoch": 0.49, "learning_rate": 5.0777844883990134e-05, "loss": 1.0441, "step": 196675 }, { "epoch": 0.49, "learning_rate": 5.0776586642508425e-05, "loss": 1.0405, "step": 196680 }, { "epoch": 0.49, "learning_rate": 5.0775328401026724e-05, "loss": 1.0472, "step": 196685 }, { "epoch": 0.49, "learning_rate": 5.0774070159545015e-05, "loss": 1.0427, "step": 196690 }, { "epoch": 0.49, "learning_rate": 5.0772811918063314e-05, "loss": 1.0425, "step": 196695 }, { "epoch": 0.49, "learning_rate": 5.0771553676581605e-05, "loss": 1.0442, "step": 196700 }, { "epoch": 0.49, "learning_rate": 5.0770295435099904e-05, "loss": 1.0651, "step": 196705 }, { "epoch": 0.49, "learning_rate": 5.0769037193618195e-05, "loss": 1.0439, "step": 196710 }, { "epoch": 0.49, "learning_rate": 5.0767778952136494e-05, "loss": 1.0397, "step": 196715 }, { "epoch": 0.49, "learning_rate": 5.0766520710654785e-05, "loss": 1.0467, "step": 196720 }, { "epoch": 0.49, "learning_rate": 5.076526246917309e-05, "loss": 1.0457, "step": 196725 }, { "epoch": 0.49, "learning_rate": 5.076400422769139e-05, "loss": 1.047, "step": 196730 }, { "epoch": 0.49, "learning_rate": 5.076274598620968e-05, "loss": 1.0428, "step": 196735 }, { "epoch": 0.49, "learning_rate": 5.076148774472798e-05, "loss": 1.0468, "step": 196740 }, { "epoch": 0.49, "learning_rate": 5.076022950324627e-05, "loss": 1.0474, "step": 196745 }, { "epoch": 0.49, "learning_rate": 5.075897126176456e-05, "loss": 1.0455, "step": 196750 }, { "epoch": 0.49, "learning_rate": 5.075771302028286e-05, "loss": 1.0494, "step": 196755 }, { "epoch": 0.49, "learning_rate": 5.075645477880115e-05, "loss": 1.047, "step": 196760 }, { "epoch": 0.49, "learning_rate": 5.075519653731945e-05, "loss": 1.0465, "step": 196765 }, { "epoch": 0.49, "learning_rate": 5.075393829583774e-05, "loss": 1.0429, "step": 196770 }, { "epoch": 0.49, "learning_rate": 5.075268005435604e-05, "loss": 1.0456, "step": 196775 }, { "epoch": 0.49, "learning_rate": 5.075142181287433e-05, "loss": 1.0455, "step": 196780 }, { "epoch": 0.49, "learning_rate": 5.075016357139263e-05, "loss": 1.0444, "step": 196785 }, { "epoch": 0.49, "learning_rate": 5.074890532991092e-05, "loss": 1.044, "step": 196790 }, { "epoch": 0.49, "learning_rate": 5.074764708842922e-05, "loss": 1.0447, "step": 196795 }, { "epoch": 0.49, "learning_rate": 5.074638884694751e-05, "loss": 1.0468, "step": 196800 }, { "epoch": 0.49, "learning_rate": 5.074513060546581e-05, "loss": 1.045, "step": 196805 }, { "epoch": 0.49, "learning_rate": 5.07438723639841e-05, "loss": 1.0428, "step": 196810 }, { "epoch": 0.49, "learning_rate": 5.074261412250239e-05, "loss": 1.0447, "step": 196815 }, { "epoch": 0.49, "learning_rate": 5.074135588102069e-05, "loss": 1.0435, "step": 196820 }, { "epoch": 0.49, "learning_rate": 5.074009763953898e-05, "loss": 1.0445, "step": 196825 }, { "epoch": 0.49, "learning_rate": 5.073883939805728e-05, "loss": 1.0453, "step": 196830 }, { "epoch": 0.49, "learning_rate": 5.073758115657557e-05, "loss": 1.0623, "step": 196835 }, { "epoch": 0.49, "learning_rate": 5.073632291509387e-05, "loss": 1.0443, "step": 196840 }, { "epoch": 0.49, "learning_rate": 5.073506467361216e-05, "loss": 1.0424, "step": 196845 }, { "epoch": 0.49, "learning_rate": 5.073380643213046e-05, "loss": 1.0437, "step": 196850 }, { "epoch": 0.49, "learning_rate": 5.073254819064875e-05, "loss": 1.0467, "step": 196855 }, { "epoch": 0.49, "learning_rate": 5.073128994916705e-05, "loss": 1.0464, "step": 196860 }, { "epoch": 0.49, "learning_rate": 5.073003170768534e-05, "loss": 1.0438, "step": 196865 }, { "epoch": 0.49, "learning_rate": 5.072877346620364e-05, "loss": 1.0433, "step": 196870 }, { "epoch": 0.49, "learning_rate": 5.072751522472193e-05, "loss": 1.0461, "step": 196875 }, { "epoch": 0.49, "learning_rate": 5.0726256983240225e-05, "loss": 1.0469, "step": 196880 }, { "epoch": 0.49, "learning_rate": 5.072499874175852e-05, "loss": 1.042, "step": 196885 }, { "epoch": 0.49, "learning_rate": 5.0723740500276815e-05, "loss": 1.0457, "step": 196890 }, { "epoch": 0.49, "learning_rate": 5.072248225879511e-05, "loss": 1.0412, "step": 196895 }, { "epoch": 0.49, "learning_rate": 5.0721224017313405e-05, "loss": 1.0449, "step": 196900 }, { "epoch": 0.49, "learning_rate": 5.07199657758317e-05, "loss": 1.0438, "step": 196905 }, { "epoch": 0.49, "learning_rate": 5.0718707534349995e-05, "loss": 1.0432, "step": 196910 }, { "epoch": 0.49, "learning_rate": 5.071744929286829e-05, "loss": 1.0457, "step": 196915 }, { "epoch": 0.49, "learning_rate": 5.0716191051386585e-05, "loss": 1.0433, "step": 196920 }, { "epoch": 0.49, "learning_rate": 5.071493280990488e-05, "loss": 1.0442, "step": 196925 }, { "epoch": 0.49, "learning_rate": 5.0713674568423175e-05, "loss": 1.0417, "step": 196930 }, { "epoch": 0.49, "learning_rate": 5.071241632694147e-05, "loss": 1.0439, "step": 196935 }, { "epoch": 0.49, "learning_rate": 5.0711158085459765e-05, "loss": 1.0411, "step": 196940 }, { "epoch": 0.49, "learning_rate": 5.0709899843978056e-05, "loss": 1.0401, "step": 196945 }, { "epoch": 0.49, "learning_rate": 5.0708641602496355e-05, "loss": 1.0467, "step": 196950 }, { "epoch": 0.49, "learning_rate": 5.0707383361014646e-05, "loss": 1.0466, "step": 196955 }, { "epoch": 0.49, "learning_rate": 5.0706125119532945e-05, "loss": 1.0443, "step": 196960 }, { "epoch": 0.49, "learning_rate": 5.0704866878051236e-05, "loss": 1.0697, "step": 196965 }, { "epoch": 0.49, "learning_rate": 5.0703608636569535e-05, "loss": 1.0445, "step": 196970 }, { "epoch": 0.49, "learning_rate": 5.0702350395087826e-05, "loss": 1.0419, "step": 196975 }, { "epoch": 0.49, "learning_rate": 5.0701092153606125e-05, "loss": 1.044, "step": 196980 }, { "epoch": 0.49, "learning_rate": 5.0699833912124416e-05, "loss": 1.0454, "step": 196985 }, { "epoch": 0.49, "learning_rate": 5.0698575670642714e-05, "loss": 1.0447, "step": 196990 }, { "epoch": 0.49, "learning_rate": 5.0697317429161006e-05, "loss": 1.047, "step": 196995 }, { "epoch": 0.49, "learning_rate": 5.0696059187679304e-05, "loss": 1.0444, "step": 197000 }, { "epoch": 0.49, "learning_rate": 5.0694800946197596e-05, "loss": 1.0439, "step": 197005 }, { "epoch": 0.49, "learning_rate": 5.069354270471589e-05, "loss": 1.0452, "step": 197010 }, { "epoch": 0.49, "learning_rate": 5.0692284463234186e-05, "loss": 1.0471, "step": 197015 }, { "epoch": 0.49, "learning_rate": 5.069102622175248e-05, "loss": 1.0408, "step": 197020 }, { "epoch": 0.49, "learning_rate": 5.0689767980270776e-05, "loss": 1.043, "step": 197025 }, { "epoch": 0.49, "learning_rate": 5.068850973878907e-05, "loss": 1.044, "step": 197030 }, { "epoch": 0.49, "learning_rate": 5.0687251497307366e-05, "loss": 1.0439, "step": 197035 }, { "epoch": 0.49, "learning_rate": 5.068599325582566e-05, "loss": 1.0413, "step": 197040 }, { "epoch": 0.49, "learning_rate": 5.0684735014343956e-05, "loss": 1.0466, "step": 197045 }, { "epoch": 0.49, "learning_rate": 5.068347677286225e-05, "loss": 1.0448, "step": 197050 }, { "epoch": 0.49, "learning_rate": 5.0682218531380546e-05, "loss": 1.0442, "step": 197055 }, { "epoch": 0.49, "learning_rate": 5.068096028989884e-05, "loss": 1.0466, "step": 197060 }, { "epoch": 0.49, "learning_rate": 5.0679702048417136e-05, "loss": 1.0478, "step": 197065 }, { "epoch": 0.49, "learning_rate": 5.067844380693543e-05, "loss": 1.0425, "step": 197070 }, { "epoch": 0.49, "learning_rate": 5.067718556545372e-05, "loss": 1.0463, "step": 197075 }, { "epoch": 0.49, "learning_rate": 5.067592732397202e-05, "loss": 1.0444, "step": 197080 }, { "epoch": 0.49, "learning_rate": 5.067466908249031e-05, "loss": 1.0686, "step": 197085 }, { "epoch": 0.49, "learning_rate": 5.067341084100861e-05, "loss": 1.0462, "step": 197090 }, { "epoch": 0.49, "learning_rate": 5.06721525995269e-05, "loss": 1.0453, "step": 197095 }, { "epoch": 0.49, "learning_rate": 5.06708943580452e-05, "loss": 1.0435, "step": 197100 }, { "epoch": 0.49, "learning_rate": 5.066963611656349e-05, "loss": 1.0467, "step": 197105 }, { "epoch": 0.49, "learning_rate": 5.066837787508179e-05, "loss": 1.0424, "step": 197110 }, { "epoch": 0.49, "learning_rate": 5.066711963360008e-05, "loss": 1.0417, "step": 197115 }, { "epoch": 0.49, "learning_rate": 5.066586139211838e-05, "loss": 1.0458, "step": 197120 }, { "epoch": 0.49, "learning_rate": 5.066460315063667e-05, "loss": 1.0424, "step": 197125 }, { "epoch": 0.49, "learning_rate": 5.066334490915496e-05, "loss": 1.0447, "step": 197130 }, { "epoch": 0.49, "learning_rate": 5.066208666767326e-05, "loss": 1.0626, "step": 197135 }, { "epoch": 0.49, "learning_rate": 5.066082842619155e-05, "loss": 1.0466, "step": 197140 }, { "epoch": 0.49, "learning_rate": 5.065957018470985e-05, "loss": 1.0443, "step": 197145 }, { "epoch": 0.49, "learning_rate": 5.065831194322814e-05, "loss": 1.0429, "step": 197150 }, { "epoch": 0.49, "learning_rate": 5.065705370174644e-05, "loss": 1.0408, "step": 197155 }, { "epoch": 0.49, "learning_rate": 5.065579546026473e-05, "loss": 1.0416, "step": 197160 }, { "epoch": 0.49, "learning_rate": 5.065453721878303e-05, "loss": 1.0432, "step": 197165 }, { "epoch": 0.49, "learning_rate": 5.065327897730132e-05, "loss": 1.0456, "step": 197170 }, { "epoch": 0.49, "learning_rate": 5.065202073581962e-05, "loss": 1.0453, "step": 197175 }, { "epoch": 0.49, "learning_rate": 5.065076249433791e-05, "loss": 1.0445, "step": 197180 }, { "epoch": 0.49, "learning_rate": 5.064950425285621e-05, "loss": 1.0472, "step": 197185 }, { "epoch": 0.49, "learning_rate": 5.06482460113745e-05, "loss": 1.0446, "step": 197190 }, { "epoch": 0.49, "learning_rate": 5.064698776989279e-05, "loss": 1.0413, "step": 197195 }, { "epoch": 0.5, "learning_rate": 5.064572952841109e-05, "loss": 1.0472, "step": 197200 }, { "epoch": 0.5, "learning_rate": 5.064447128692938e-05, "loss": 1.0452, "step": 197205 }, { "epoch": 0.5, "learning_rate": 5.064321304544768e-05, "loss": 1.0683, "step": 197210 }, { "epoch": 0.5, "learning_rate": 5.064195480396597e-05, "loss": 1.0442, "step": 197215 }, { "epoch": 0.5, "learning_rate": 5.064069656248428e-05, "loss": 1.0451, "step": 197220 }, { "epoch": 0.5, "learning_rate": 5.0639438321002575e-05, "loss": 1.0443, "step": 197225 }, { "epoch": 0.5, "learning_rate": 5.063818007952087e-05, "loss": 1.0448, "step": 197230 }, { "epoch": 0.5, "learning_rate": 5.0636921838039165e-05, "loss": 1.0447, "step": 197235 }, { "epoch": 0.5, "learning_rate": 5.063566359655746e-05, "loss": 1.046, "step": 197240 }, { "epoch": 0.5, "learning_rate": 5.0634405355075755e-05, "loss": 1.0448, "step": 197245 }, { "epoch": 0.5, "learning_rate": 5.063314711359405e-05, "loss": 1.0449, "step": 197250 }, { "epoch": 0.5, "learning_rate": 5.0631888872112345e-05, "loss": 1.0431, "step": 197255 }, { "epoch": 0.5, "learning_rate": 5.063063063063064e-05, "loss": 1.042, "step": 197260 }, { "epoch": 0.5, "learning_rate": 5.0629372389148935e-05, "loss": 1.0429, "step": 197265 }, { "epoch": 0.5, "learning_rate": 5.062811414766723e-05, "loss": 1.0418, "step": 197270 }, { "epoch": 0.5, "learning_rate": 5.062685590618552e-05, "loss": 1.0455, "step": 197275 }, { "epoch": 0.5, "learning_rate": 5.062559766470382e-05, "loss": 1.0441, "step": 197280 }, { "epoch": 0.5, "learning_rate": 5.062433942322211e-05, "loss": 1.0445, "step": 197285 }, { "epoch": 0.5, "learning_rate": 5.062308118174041e-05, "loss": 1.0482, "step": 197290 }, { "epoch": 0.5, "learning_rate": 5.06218229402587e-05, "loss": 1.0454, "step": 197295 }, { "epoch": 0.5, "learning_rate": 5.0620564698777e-05, "loss": 1.0447, "step": 197300 }, { "epoch": 0.5, "learning_rate": 5.061930645729529e-05, "loss": 1.0897, "step": 197305 }, { "epoch": 0.5, "learning_rate": 5.061804821581359e-05, "loss": 1.0459, "step": 197310 }, { "epoch": 0.5, "learning_rate": 5.061678997433188e-05, "loss": 1.0469, "step": 197315 }, { "epoch": 0.5, "learning_rate": 5.061553173285018e-05, "loss": 1.0452, "step": 197320 }, { "epoch": 0.5, "learning_rate": 5.061427349136847e-05, "loss": 1.0481, "step": 197325 }, { "epoch": 0.5, "learning_rate": 5.061301524988677e-05, "loss": 1.0437, "step": 197330 }, { "epoch": 0.5, "learning_rate": 5.061175700840506e-05, "loss": 1.0448, "step": 197335 }, { "epoch": 0.5, "learning_rate": 5.061049876692335e-05, "loss": 1.0434, "step": 197340 }, { "epoch": 0.5, "learning_rate": 5.060924052544165e-05, "loss": 1.0455, "step": 197345 }, { "epoch": 0.5, "learning_rate": 5.060798228395994e-05, "loss": 1.044, "step": 197350 }, { "epoch": 0.5, "learning_rate": 5.060672404247824e-05, "loss": 1.0458, "step": 197355 }, { "epoch": 0.5, "learning_rate": 5.060546580099653e-05, "loss": 1.0448, "step": 197360 }, { "epoch": 0.5, "learning_rate": 5.060420755951483e-05, "loss": 1.0473, "step": 197365 }, { "epoch": 0.5, "learning_rate": 5.060294931803312e-05, "loss": 1.0436, "step": 197370 }, { "epoch": 0.5, "learning_rate": 5.060169107655142e-05, "loss": 1.0435, "step": 197375 }, { "epoch": 0.5, "learning_rate": 5.060043283506971e-05, "loss": 1.0439, "step": 197380 }, { "epoch": 0.5, "learning_rate": 5.059917459358801e-05, "loss": 1.0464, "step": 197385 }, { "epoch": 0.5, "learning_rate": 5.05979163521063e-05, "loss": 1.0428, "step": 197390 }, { "epoch": 0.5, "learning_rate": 5.05966581106246e-05, "loss": 1.0467, "step": 197395 }, { "epoch": 0.5, "learning_rate": 5.059539986914289e-05, "loss": 1.0413, "step": 197400 }, { "epoch": 0.5, "learning_rate": 5.059414162766118e-05, "loss": 1.0476, "step": 197405 }, { "epoch": 0.5, "learning_rate": 5.059288338617948e-05, "loss": 1.045, "step": 197410 }, { "epoch": 0.5, "learning_rate": 5.059162514469777e-05, "loss": 1.0454, "step": 197415 }, { "epoch": 0.5, "learning_rate": 5.059036690321607e-05, "loss": 1.0431, "step": 197420 }, { "epoch": 0.5, "learning_rate": 5.058910866173436e-05, "loss": 1.0426, "step": 197425 }, { "epoch": 0.5, "learning_rate": 5.058785042025266e-05, "loss": 1.0459, "step": 197430 }, { "epoch": 0.5, "learning_rate": 5.058659217877095e-05, "loss": 1.043, "step": 197435 }, { "epoch": 0.5, "learning_rate": 5.058533393728925e-05, "loss": 1.0478, "step": 197440 }, { "epoch": 0.5, "learning_rate": 5.058407569580754e-05, "loss": 1.0452, "step": 197445 }, { "epoch": 0.5, "learning_rate": 5.058281745432584e-05, "loss": 1.0456, "step": 197450 }, { "epoch": 0.5, "learning_rate": 5.058155921284413e-05, "loss": 1.0416, "step": 197455 }, { "epoch": 0.5, "learning_rate": 5.058030097136243e-05, "loss": 1.0429, "step": 197460 }, { "epoch": 0.5, "learning_rate": 5.057904272988072e-05, "loss": 1.0413, "step": 197465 }, { "epoch": 0.5, "learning_rate": 5.057778448839901e-05, "loss": 1.0487, "step": 197470 }, { "epoch": 0.5, "learning_rate": 5.057652624691731e-05, "loss": 1.044, "step": 197475 }, { "epoch": 0.5, "learning_rate": 5.05752680054356e-05, "loss": 1.0486, "step": 197480 }, { "epoch": 0.5, "learning_rate": 5.05740097639539e-05, "loss": 1.0445, "step": 197485 }, { "epoch": 0.5, "learning_rate": 5.057275152247219e-05, "loss": 1.0434, "step": 197490 }, { "epoch": 0.5, "learning_rate": 5.057149328099049e-05, "loss": 1.0475, "step": 197495 }, { "epoch": 0.5, "learning_rate": 5.057023503950878e-05, "loss": 1.0426, "step": 197500 }, { "epoch": 0.5, "learning_rate": 5.056897679802708e-05, "loss": 1.0424, "step": 197505 }, { "epoch": 0.5, "learning_rate": 5.056771855654537e-05, "loss": 1.0441, "step": 197510 }, { "epoch": 0.5, "learning_rate": 5.056646031506367e-05, "loss": 1.0439, "step": 197515 }, { "epoch": 0.5, "learning_rate": 5.056520207358196e-05, "loss": 1.0419, "step": 197520 }, { "epoch": 0.5, "learning_rate": 5.056394383210026e-05, "loss": 1.0455, "step": 197525 }, { "epoch": 0.5, "learning_rate": 5.056268559061855e-05, "loss": 1.0445, "step": 197530 }, { "epoch": 0.5, "learning_rate": 5.0561427349136844e-05, "loss": 1.044, "step": 197535 }, { "epoch": 0.5, "learning_rate": 5.056016910765514e-05, "loss": 1.0464, "step": 197540 }, { "epoch": 0.5, "learning_rate": 5.0558910866173434e-05, "loss": 1.0438, "step": 197545 }, { "epoch": 0.5, "learning_rate": 5.055765262469173e-05, "loss": 1.045, "step": 197550 }, { "epoch": 0.5, "learning_rate": 5.0556394383210024e-05, "loss": 1.0456, "step": 197555 }, { "epoch": 0.5, "learning_rate": 5.055513614172832e-05, "loss": 1.0433, "step": 197560 }, { "epoch": 0.5, "learning_rate": 5.0553877900246614e-05, "loss": 1.0422, "step": 197565 }, { "epoch": 0.5, "learning_rate": 5.055261965876491e-05, "loss": 1.042, "step": 197570 }, { "epoch": 0.5, "learning_rate": 5.0551361417283204e-05, "loss": 1.0439, "step": 197575 }, { "epoch": 0.5, "learning_rate": 5.05501031758015e-05, "loss": 1.0447, "step": 197580 }, { "epoch": 0.5, "learning_rate": 5.0548844934319794e-05, "loss": 1.046, "step": 197585 }, { "epoch": 0.5, "learning_rate": 5.054758669283809e-05, "loss": 1.0436, "step": 197590 }, { "epoch": 0.5, "learning_rate": 5.0546328451356384e-05, "loss": 1.0455, "step": 197595 }, { "epoch": 0.5, "learning_rate": 5.0545070209874676e-05, "loss": 1.0588, "step": 197600 }, { "epoch": 0.5, "learning_rate": 5.0543811968392974e-05, "loss": 1.0434, "step": 197605 }, { "epoch": 0.5, "learning_rate": 5.0542553726911266e-05, "loss": 1.0451, "step": 197610 }, { "epoch": 0.5, "learning_rate": 5.0541295485429564e-05, "loss": 1.0449, "step": 197615 }, { "epoch": 0.5, "learning_rate": 5.0540037243947856e-05, "loss": 1.0424, "step": 197620 }, { "epoch": 0.5, "learning_rate": 5.0538779002466154e-05, "loss": 1.0475, "step": 197625 }, { "epoch": 0.5, "learning_rate": 5.0537520760984446e-05, "loss": 1.0684, "step": 197630 }, { "epoch": 0.5, "learning_rate": 5.0536262519502744e-05, "loss": 1.0419, "step": 197635 }, { "epoch": 0.5, "learning_rate": 5.0535004278021036e-05, "loss": 1.0466, "step": 197640 }, { "epoch": 0.5, "learning_rate": 5.0533746036539334e-05, "loss": 1.0444, "step": 197645 }, { "epoch": 0.5, "learning_rate": 5.0532487795057625e-05, "loss": 1.0446, "step": 197650 }, { "epoch": 0.5, "learning_rate": 5.0531229553575924e-05, "loss": 1.0445, "step": 197655 }, { "epoch": 0.5, "learning_rate": 5.0529971312094215e-05, "loss": 1.0455, "step": 197660 }, { "epoch": 0.5, "learning_rate": 5.052871307061251e-05, "loss": 1.0461, "step": 197665 }, { "epoch": 0.5, "learning_rate": 5.0527454829130805e-05, "loss": 1.0453, "step": 197670 }, { "epoch": 0.5, "learning_rate": 5.05261965876491e-05, "loss": 1.0436, "step": 197675 }, { "epoch": 0.5, "learning_rate": 5.0524938346167395e-05, "loss": 1.0453, "step": 197680 }, { "epoch": 0.5, "learning_rate": 5.052368010468569e-05, "loss": 1.0452, "step": 197685 }, { "epoch": 0.5, "learning_rate": 5.0522421863203985e-05, "loss": 1.0452, "step": 197690 }, { "epoch": 0.5, "learning_rate": 5.052116362172228e-05, "loss": 1.0444, "step": 197695 }, { "epoch": 0.5, "learning_rate": 5.0519905380240575e-05, "loss": 1.0444, "step": 197700 }, { "epoch": 0.5, "learning_rate": 5.051864713875887e-05, "loss": 1.0439, "step": 197705 }, { "epoch": 0.5, "learning_rate": 5.0517388897277165e-05, "loss": 1.0438, "step": 197710 }, { "epoch": 0.5, "learning_rate": 5.051613065579546e-05, "loss": 1.043, "step": 197715 }, { "epoch": 0.5, "learning_rate": 5.051487241431376e-05, "loss": 1.0463, "step": 197720 }, { "epoch": 0.5, "learning_rate": 5.051361417283206e-05, "loss": 1.0455, "step": 197725 }, { "epoch": 0.5, "learning_rate": 5.051235593135035e-05, "loss": 1.0463, "step": 197730 }, { "epoch": 0.5, "learning_rate": 5.051109768986865e-05, "loss": 1.0466, "step": 197735 }, { "epoch": 0.5, "learning_rate": 5.050983944838694e-05, "loss": 1.0438, "step": 197740 }, { "epoch": 0.5, "learning_rate": 5.0508581206905234e-05, "loss": 1.0467, "step": 197745 }, { "epoch": 0.5, "learning_rate": 5.050732296542353e-05, "loss": 1.0422, "step": 197750 }, { "epoch": 0.5, "learning_rate": 5.0506064723941824e-05, "loss": 1.0486, "step": 197755 }, { "epoch": 0.5, "learning_rate": 5.050480648246012e-05, "loss": 1.0435, "step": 197760 }, { "epoch": 0.5, "learning_rate": 5.0503548240978414e-05, "loss": 1.0405, "step": 197765 }, { "epoch": 0.5, "learning_rate": 5.050228999949671e-05, "loss": 1.0456, "step": 197770 }, { "epoch": 0.5, "learning_rate": 5.0501031758015004e-05, "loss": 1.0612, "step": 197775 }, { "epoch": 0.5, "learning_rate": 5.04997735165333e-05, "loss": 1.0421, "step": 197780 }, { "epoch": 0.5, "learning_rate": 5.0498515275051593e-05, "loss": 1.0468, "step": 197785 }, { "epoch": 0.5, "learning_rate": 5.049725703356989e-05, "loss": 1.042, "step": 197790 }, { "epoch": 0.5, "learning_rate": 5.0495998792088183e-05, "loss": 1.0473, "step": 197795 }, { "epoch": 0.5, "learning_rate": 5.049474055060648e-05, "loss": 1.0436, "step": 197800 }, { "epoch": 0.5, "learning_rate": 5.0493482309124773e-05, "loss": 1.0421, "step": 197805 }, { "epoch": 0.5, "learning_rate": 5.0492224067643065e-05, "loss": 1.0415, "step": 197810 }, { "epoch": 0.5, "learning_rate": 5.0490965826161363e-05, "loss": 1.0415, "step": 197815 }, { "epoch": 0.5, "learning_rate": 5.0489707584679655e-05, "loss": 1.046, "step": 197820 }, { "epoch": 0.5, "learning_rate": 5.048844934319795e-05, "loss": 1.0444, "step": 197825 }, { "epoch": 0.5, "learning_rate": 5.0487191101716245e-05, "loss": 1.0487, "step": 197830 }, { "epoch": 0.5, "learning_rate": 5.048593286023454e-05, "loss": 1.0463, "step": 197835 }, { "epoch": 0.5, "learning_rate": 5.0484674618752835e-05, "loss": 1.0441, "step": 197840 }, { "epoch": 0.5, "learning_rate": 5.048341637727113e-05, "loss": 1.0441, "step": 197845 }, { "epoch": 0.5, "learning_rate": 5.0482158135789425e-05, "loss": 1.0442, "step": 197850 }, { "epoch": 0.5, "learning_rate": 5.048089989430772e-05, "loss": 1.0469, "step": 197855 }, { "epoch": 0.5, "learning_rate": 5.0479641652826015e-05, "loss": 1.0456, "step": 197860 }, { "epoch": 0.5, "learning_rate": 5.047838341134431e-05, "loss": 1.0433, "step": 197865 }, { "epoch": 0.5, "learning_rate": 5.0477125169862605e-05, "loss": 1.0444, "step": 197870 }, { "epoch": 0.5, "learning_rate": 5.0475866928380896e-05, "loss": 1.0412, "step": 197875 }, { "epoch": 0.5, "learning_rate": 5.0474608686899195e-05, "loss": 1.0448, "step": 197880 }, { "epoch": 0.5, "learning_rate": 5.0473350445417486e-05, "loss": 1.0636, "step": 197885 }, { "epoch": 0.5, "learning_rate": 5.0472092203935785e-05, "loss": 1.0427, "step": 197890 }, { "epoch": 0.5, "learning_rate": 5.0470833962454076e-05, "loss": 1.043, "step": 197895 }, { "epoch": 0.5, "learning_rate": 5.0469575720972375e-05, "loss": 1.047, "step": 197900 }, { "epoch": 0.5, "learning_rate": 5.0468317479490666e-05, "loss": 1.0465, "step": 197905 }, { "epoch": 0.5, "learning_rate": 5.0467059238008965e-05, "loss": 1.0446, "step": 197910 }, { "epoch": 0.5, "learning_rate": 5.0465800996527256e-05, "loss": 1.0456, "step": 197915 }, { "epoch": 0.5, "learning_rate": 5.0464542755045555e-05, "loss": 1.0445, "step": 197920 }, { "epoch": 0.5, "learning_rate": 5.0463284513563846e-05, "loss": 1.0477, "step": 197925 }, { "epoch": 0.5, "learning_rate": 5.046202627208214e-05, "loss": 1.0435, "step": 197930 }, { "epoch": 0.5, "learning_rate": 5.0460768030600436e-05, "loss": 1.0452, "step": 197935 }, { "epoch": 0.5, "learning_rate": 5.045950978911873e-05, "loss": 1.0481, "step": 197940 }, { "epoch": 0.5, "learning_rate": 5.0458251547637026e-05, "loss": 1.0434, "step": 197945 }, { "epoch": 0.5, "learning_rate": 5.045699330615532e-05, "loss": 1.0654, "step": 197950 }, { "epoch": 0.5, "learning_rate": 5.0455735064673616e-05, "loss": 1.0445, "step": 197955 }, { "epoch": 0.5, "learning_rate": 5.045472847148825e-05, "loss": 1.0437, "step": 197960 }, { "epoch": 0.5, "learning_rate": 5.045347023000655e-05, "loss": 1.0642, "step": 197965 }, { "epoch": 0.5, "learning_rate": 5.045221198852484e-05, "loss": 1.0452, "step": 197970 }, { "epoch": 0.5, "learning_rate": 5.045095374704314e-05, "loss": 1.0472, "step": 197975 }, { "epoch": 0.5, "learning_rate": 5.044969550556143e-05, "loss": 1.043, "step": 197980 }, { "epoch": 0.5, "learning_rate": 5.044843726407973e-05, "loss": 1.046, "step": 197985 }, { "epoch": 0.5, "learning_rate": 5.044717902259802e-05, "loss": 1.0448, "step": 197990 }, { "epoch": 0.5, "learning_rate": 5.044592078111632e-05, "loss": 1.0431, "step": 197995 }, { "epoch": 0.5, "learning_rate": 5.044466253963461e-05, "loss": 1.0441, "step": 198000 }, { "epoch": 0.5, "learning_rate": 5.044340429815291e-05, "loss": 1.0383, "step": 198005 }, { "epoch": 0.5, "learning_rate": 5.04421460566712e-05, "loss": 1.0445, "step": 198010 }, { "epoch": 0.5, "learning_rate": 5.0440887815189494e-05, "loss": 1.0457, "step": 198015 }, { "epoch": 0.5, "learning_rate": 5.043962957370779e-05, "loss": 1.0655, "step": 198020 }, { "epoch": 0.5, "learning_rate": 5.0438371332226084e-05, "loss": 1.0438, "step": 198025 }, { "epoch": 0.5, "learning_rate": 5.043711309074438e-05, "loss": 1.0434, "step": 198030 }, { "epoch": 0.5, "learning_rate": 5.0435854849262674e-05, "loss": 1.0438, "step": 198035 }, { "epoch": 0.5, "learning_rate": 5.043459660778097e-05, "loss": 1.0449, "step": 198040 }, { "epoch": 0.5, "learning_rate": 5.0433338366299264e-05, "loss": 1.0478, "step": 198045 }, { "epoch": 0.5, "learning_rate": 5.043208012481756e-05, "loss": 1.0453, "step": 198050 }, { "epoch": 0.5, "learning_rate": 5.0430821883335854e-05, "loss": 1.0438, "step": 198055 }, { "epoch": 0.5, "learning_rate": 5.042956364185415e-05, "loss": 1.0457, "step": 198060 }, { "epoch": 0.5, "learning_rate": 5.0428305400372444e-05, "loss": 1.0433, "step": 198065 }, { "epoch": 0.5, "learning_rate": 5.042704715889074e-05, "loss": 1.0442, "step": 198070 }, { "epoch": 0.5, "learning_rate": 5.0425788917409033e-05, "loss": 1.0451, "step": 198075 }, { "epoch": 0.5, "learning_rate": 5.0424530675927325e-05, "loss": 1.0438, "step": 198080 }, { "epoch": 0.5, "learning_rate": 5.0423272434445623e-05, "loss": 1.0461, "step": 198085 }, { "epoch": 0.5, "learning_rate": 5.0422014192963915e-05, "loss": 1.048, "step": 198090 }, { "epoch": 0.5, "learning_rate": 5.0420755951482213e-05, "loss": 1.0424, "step": 198095 }, { "epoch": 0.5, "learning_rate": 5.0419497710000505e-05, "loss": 1.0446, "step": 198100 }, { "epoch": 0.5, "learning_rate": 5.0418239468518803e-05, "loss": 1.0483, "step": 198105 }, { "epoch": 0.5, "learning_rate": 5.0416981227037095e-05, "loss": 1.0437, "step": 198110 }, { "epoch": 0.5, "learning_rate": 5.041572298555539e-05, "loss": 1.0457, "step": 198115 }, { "epoch": 0.5, "learning_rate": 5.0414464744073685e-05, "loss": 1.0435, "step": 198120 }, { "epoch": 0.5, "learning_rate": 5.041320650259198e-05, "loss": 1.0471, "step": 198125 }, { "epoch": 0.5, "learning_rate": 5.0411948261110275e-05, "loss": 1.0445, "step": 198130 }, { "epoch": 0.5, "learning_rate": 5.041069001962857e-05, "loss": 1.0444, "step": 198135 }, { "epoch": 0.5, "learning_rate": 5.0409431778146865e-05, "loss": 1.0459, "step": 198140 }, { "epoch": 0.5, "learning_rate": 5.0408173536665157e-05, "loss": 1.0462, "step": 198145 }, { "epoch": 0.5, "learning_rate": 5.0406915295183455e-05, "loss": 1.0447, "step": 198150 }, { "epoch": 0.5, "learning_rate": 5.0405657053701746e-05, "loss": 1.0428, "step": 198155 }, { "epoch": 0.5, "learning_rate": 5.0404398812220045e-05, "loss": 1.044, "step": 198160 }, { "epoch": 0.5, "learning_rate": 5.0403140570738336e-05, "loss": 1.0454, "step": 198165 }, { "epoch": 0.5, "learning_rate": 5.0401882329256635e-05, "loss": 1.0458, "step": 198170 }, { "epoch": 0.5, "learning_rate": 5.0400624087774926e-05, "loss": 1.0459, "step": 198175 }, { "epoch": 0.5, "learning_rate": 5.0399365846293225e-05, "loss": 1.0442, "step": 198180 }, { "epoch": 0.5, "learning_rate": 5.0398107604811516e-05, "loss": 1.046, "step": 198185 }, { "epoch": 0.5, "learning_rate": 5.0396849363329815e-05, "loss": 1.0472, "step": 198190 }, { "epoch": 0.5, "learning_rate": 5.0395591121848106e-05, "loss": 1.0689, "step": 198195 }, { "epoch": 0.5, "learning_rate": 5.0394332880366405e-05, "loss": 1.0446, "step": 198200 }, { "epoch": 0.5, "learning_rate": 5.0393074638884696e-05, "loss": 1.0665, "step": 198205 }, { "epoch": 0.5, "learning_rate": 5.039181639740299e-05, "loss": 1.0425, "step": 198210 }, { "epoch": 0.5, "learning_rate": 5.0390558155921286e-05, "loss": 1.044, "step": 198215 }, { "epoch": 0.5, "learning_rate": 5.038929991443958e-05, "loss": 1.068, "step": 198220 }, { "epoch": 0.5, "learning_rate": 5.0388041672957876e-05, "loss": 1.0457, "step": 198225 }, { "epoch": 0.5, "learning_rate": 5.038678343147617e-05, "loss": 1.0452, "step": 198230 }, { "epoch": 0.5, "learning_rate": 5.0385525189994466e-05, "loss": 1.0448, "step": 198235 }, { "epoch": 0.5, "learning_rate": 5.038426694851276e-05, "loss": 1.0436, "step": 198240 }, { "epoch": 0.5, "learning_rate": 5.0383008707031056e-05, "loss": 1.044, "step": 198245 }, { "epoch": 0.5, "learning_rate": 5.038175046554935e-05, "loss": 1.0452, "step": 198250 }, { "epoch": 0.5, "learning_rate": 5.0380492224067646e-05, "loss": 1.0439, "step": 198255 }, { "epoch": 0.5, "learning_rate": 5.037923398258594e-05, "loss": 1.0453, "step": 198260 }, { "epoch": 0.5, "learning_rate": 5.0377975741104236e-05, "loss": 1.0435, "step": 198265 }, { "epoch": 0.5, "learning_rate": 5.037671749962253e-05, "loss": 1.0429, "step": 198270 }, { "epoch": 0.5, "learning_rate": 5.037545925814082e-05, "loss": 1.0424, "step": 198275 }, { "epoch": 0.5, "learning_rate": 5.037420101665912e-05, "loss": 1.0453, "step": 198280 }, { "epoch": 0.5, "learning_rate": 5.037294277517741e-05, "loss": 1.0433, "step": 198285 }, { "epoch": 0.5, "learning_rate": 5.037168453369571e-05, "loss": 1.0469, "step": 198290 }, { "epoch": 0.5, "learning_rate": 5.0370426292214e-05, "loss": 1.042, "step": 198295 }, { "epoch": 0.5, "learning_rate": 5.03691680507323e-05, "loss": 1.045, "step": 198300 }, { "epoch": 0.5, "learning_rate": 5.036790980925059e-05, "loss": 1.044, "step": 198305 }, { "epoch": 0.5, "learning_rate": 5.036665156776889e-05, "loss": 1.0455, "step": 198310 }, { "epoch": 0.5, "learning_rate": 5.036539332628718e-05, "loss": 1.0427, "step": 198315 }, { "epoch": 0.5, "learning_rate": 5.036413508480548e-05, "loss": 1.0461, "step": 198320 }, { "epoch": 0.5, "learning_rate": 5.036287684332377e-05, "loss": 1.0442, "step": 198325 }, { "epoch": 0.5, "learning_rate": 5.036161860184207e-05, "loss": 1.0463, "step": 198330 }, { "epoch": 0.5, "learning_rate": 5.036036036036036e-05, "loss": 1.0422, "step": 198335 }, { "epoch": 0.5, "learning_rate": 5.035910211887865e-05, "loss": 1.0455, "step": 198340 }, { "epoch": 0.5, "learning_rate": 5.035784387739695e-05, "loss": 1.045, "step": 198345 }, { "epoch": 0.5, "learning_rate": 5.035658563591524e-05, "loss": 1.0472, "step": 198350 }, { "epoch": 0.5, "learning_rate": 5.035532739443354e-05, "loss": 1.0464, "step": 198355 }, { "epoch": 0.5, "learning_rate": 5.035406915295183e-05, "loss": 1.0424, "step": 198360 }, { "epoch": 0.5, "learning_rate": 5.035281091147013e-05, "loss": 1.0458, "step": 198365 }, { "epoch": 0.5, "learning_rate": 5.035155266998842e-05, "loss": 1.0448, "step": 198370 }, { "epoch": 0.5, "learning_rate": 5.035029442850672e-05, "loss": 1.0442, "step": 198375 }, { "epoch": 0.5, "learning_rate": 5.034903618702501e-05, "loss": 1.0415, "step": 198380 }, { "epoch": 0.5, "learning_rate": 5.034777794554331e-05, "loss": 1.0441, "step": 198385 }, { "epoch": 0.5, "learning_rate": 5.03465197040616e-05, "loss": 1.0441, "step": 198390 }, { "epoch": 0.5, "learning_rate": 5.03452614625799e-05, "loss": 1.045, "step": 198395 }, { "epoch": 0.5, "learning_rate": 5.034400322109819e-05, "loss": 1.045, "step": 198400 }, { "epoch": 0.5, "learning_rate": 5.034274497961648e-05, "loss": 1.0453, "step": 198405 }, { "epoch": 0.5, "learning_rate": 5.034148673813478e-05, "loss": 1.0427, "step": 198410 }, { "epoch": 0.5, "learning_rate": 5.034022849665307e-05, "loss": 1.0468, "step": 198415 }, { "epoch": 0.5, "learning_rate": 5.033897025517138e-05, "loss": 1.0437, "step": 198420 }, { "epoch": 0.5, "learning_rate": 5.0337712013689676e-05, "loss": 1.0435, "step": 198425 }, { "epoch": 0.5, "learning_rate": 5.033645377220797e-05, "loss": 1.0439, "step": 198430 }, { "epoch": 0.5, "learning_rate": 5.0335195530726266e-05, "loss": 1.0657, "step": 198435 }, { "epoch": 0.5, "learning_rate": 5.033393728924456e-05, "loss": 1.0653, "step": 198440 }, { "epoch": 0.5, "learning_rate": 5.0332679047762856e-05, "loss": 1.0458, "step": 198445 }, { "epoch": 0.5, "learning_rate": 5.033142080628115e-05, "loss": 1.046, "step": 198450 }, { "epoch": 0.5, "learning_rate": 5.0330162564799446e-05, "loss": 1.0433, "step": 198455 }, { "epoch": 0.5, "learning_rate": 5.032890432331774e-05, "loss": 1.0439, "step": 198460 }, { "epoch": 0.5, "learning_rate": 5.0327646081836036e-05, "loss": 1.0484, "step": 198465 }, { "epoch": 0.5, "learning_rate": 5.032638784035433e-05, "loss": 1.0471, "step": 198470 }, { "epoch": 0.5, "learning_rate": 5.0325129598872626e-05, "loss": 1.0444, "step": 198475 }, { "epoch": 0.5, "learning_rate": 5.032387135739092e-05, "loss": 1.0446, "step": 198480 }, { "epoch": 0.5, "learning_rate": 5.032261311590921e-05, "loss": 1.0438, "step": 198485 }, { "epoch": 0.5, "learning_rate": 5.032135487442751e-05, "loss": 1.0467, "step": 198490 }, { "epoch": 0.5, "learning_rate": 5.03200966329458e-05, "loss": 1.046, "step": 198495 }, { "epoch": 0.5, "learning_rate": 5.03188383914641e-05, "loss": 1.0456, "step": 198500 }, { "epoch": 0.5, "learning_rate": 5.031758014998239e-05, "loss": 1.0456, "step": 198505 }, { "epoch": 0.5, "learning_rate": 5.031632190850069e-05, "loss": 1.0444, "step": 198510 }, { "epoch": 0.5, "learning_rate": 5.031506366701898e-05, "loss": 1.0454, "step": 198515 }, { "epoch": 0.5, "learning_rate": 5.031380542553728e-05, "loss": 1.0455, "step": 198520 }, { "epoch": 0.5, "learning_rate": 5.031254718405557e-05, "loss": 1.0448, "step": 198525 }, { "epoch": 0.5, "learning_rate": 5.031128894257387e-05, "loss": 1.0442, "step": 198530 }, { "epoch": 0.5, "learning_rate": 5.031003070109216e-05, "loss": 1.0438, "step": 198535 }, { "epoch": 0.5, "learning_rate": 5.030877245961046e-05, "loss": 1.0447, "step": 198540 }, { "epoch": 0.5, "learning_rate": 5.030751421812875e-05, "loss": 1.0464, "step": 198545 }, { "epoch": 0.5, "learning_rate": 5.030625597664704e-05, "loss": 1.044, "step": 198550 }, { "epoch": 0.5, "learning_rate": 5.030499773516534e-05, "loss": 1.0467, "step": 198555 }, { "epoch": 0.5, "learning_rate": 5.030373949368363e-05, "loss": 1.0433, "step": 198560 }, { "epoch": 0.5, "learning_rate": 5.030248125220193e-05, "loss": 1.0444, "step": 198565 }, { "epoch": 0.5, "learning_rate": 5.030122301072022e-05, "loss": 1.0434, "step": 198570 }, { "epoch": 0.5, "learning_rate": 5.029996476923852e-05, "loss": 1.0446, "step": 198575 }, { "epoch": 0.5, "learning_rate": 5.029870652775681e-05, "loss": 1.0436, "step": 198580 }, { "epoch": 0.5, "learning_rate": 5.029744828627511e-05, "loss": 1.0458, "step": 198585 }, { "epoch": 0.5, "learning_rate": 5.02961900447934e-05, "loss": 1.0423, "step": 198590 }, { "epoch": 0.5, "learning_rate": 5.02949318033117e-05, "loss": 1.0471, "step": 198595 }, { "epoch": 0.5, "learning_rate": 5.029367356182999e-05, "loss": 1.0427, "step": 198600 }, { "epoch": 0.5, "learning_rate": 5.029241532034829e-05, "loss": 1.0402, "step": 198605 }, { "epoch": 0.5, "learning_rate": 5.029115707886658e-05, "loss": 1.0411, "step": 198610 }, { "epoch": 0.5, "learning_rate": 5.028989883738487e-05, "loss": 1.0464, "step": 198615 }, { "epoch": 0.5, "learning_rate": 5.028864059590317e-05, "loss": 1.0455, "step": 198620 }, { "epoch": 0.5, "learning_rate": 5.028738235442146e-05, "loss": 1.0414, "step": 198625 }, { "epoch": 0.5, "learning_rate": 5.028612411293976e-05, "loss": 1.0438, "step": 198630 }, { "epoch": 0.5, "learning_rate": 5.028486587145805e-05, "loss": 1.0432, "step": 198635 }, { "epoch": 0.5, "learning_rate": 5.028360762997635e-05, "loss": 1.0437, "step": 198640 }, { "epoch": 0.5, "learning_rate": 5.028234938849464e-05, "loss": 1.045, "step": 198645 }, { "epoch": 0.5, "learning_rate": 5.028109114701294e-05, "loss": 1.0434, "step": 198650 }, { "epoch": 0.5, "learning_rate": 5.027983290553123e-05, "loss": 1.0911, "step": 198655 }, { "epoch": 0.5, "learning_rate": 5.027857466404953e-05, "loss": 1.0688, "step": 198660 }, { "epoch": 0.5, "learning_rate": 5.027731642256782e-05, "loss": 1.0462, "step": 198665 }, { "epoch": 0.5, "learning_rate": 5.027605818108611e-05, "loss": 1.0441, "step": 198670 }, { "epoch": 0.5, "learning_rate": 5.027479993960441e-05, "loss": 1.0468, "step": 198675 }, { "epoch": 0.5, "learning_rate": 5.02735416981227e-05, "loss": 1.0455, "step": 198680 }, { "epoch": 0.5, "learning_rate": 5.0272283456641e-05, "loss": 1.0467, "step": 198685 }, { "epoch": 0.5, "learning_rate": 5.027102521515929e-05, "loss": 1.044, "step": 198690 }, { "epoch": 0.5, "learning_rate": 5.026976697367759e-05, "loss": 1.0474, "step": 198695 }, { "epoch": 0.5, "learning_rate": 5.026850873219588e-05, "loss": 1.0465, "step": 198700 }, { "epoch": 0.5, "learning_rate": 5.026725049071418e-05, "loss": 1.0459, "step": 198705 }, { "epoch": 0.5, "learning_rate": 5.026599224923247e-05, "loss": 1.0431, "step": 198710 }, { "epoch": 0.5, "learning_rate": 5.026473400775077e-05, "loss": 1.0466, "step": 198715 }, { "epoch": 0.5, "learning_rate": 5.026347576626906e-05, "loss": 1.0456, "step": 198720 }, { "epoch": 0.5, "learning_rate": 5.026221752478736e-05, "loss": 1.0443, "step": 198725 }, { "epoch": 0.5, "learning_rate": 5.026095928330565e-05, "loss": 1.0447, "step": 198730 }, { "epoch": 0.5, "learning_rate": 5.0259701041823944e-05, "loss": 1.044, "step": 198735 }, { "epoch": 0.5, "learning_rate": 5.025844280034224e-05, "loss": 1.0431, "step": 198740 }, { "epoch": 0.5, "learning_rate": 5.0257184558860534e-05, "loss": 1.0434, "step": 198745 }, { "epoch": 0.5, "learning_rate": 5.025592631737883e-05, "loss": 1.044, "step": 198750 }, { "epoch": 0.5, "learning_rate": 5.0254668075897124e-05, "loss": 1.0403, "step": 198755 }, { "epoch": 0.5, "learning_rate": 5.025340983441542e-05, "loss": 1.044, "step": 198760 }, { "epoch": 0.5, "learning_rate": 5.0252151592933714e-05, "loss": 1.0457, "step": 198765 }, { "epoch": 0.5, "learning_rate": 5.025089335145201e-05, "loss": 1.0463, "step": 198770 }, { "epoch": 0.5, "learning_rate": 5.0249635109970304e-05, "loss": 1.0441, "step": 198775 }, { "epoch": 0.5, "learning_rate": 5.02483768684886e-05, "loss": 1.0469, "step": 198780 }, { "epoch": 0.5, "learning_rate": 5.0247118627006894e-05, "loss": 1.067, "step": 198785 }, { "epoch": 0.5, "learning_rate": 5.024586038552519e-05, "loss": 1.0416, "step": 198790 }, { "epoch": 0.5, "learning_rate": 5.0244602144043484e-05, "loss": 1.06, "step": 198795 }, { "epoch": 0.5, "learning_rate": 5.0243343902561776e-05, "loss": 1.0465, "step": 198800 }, { "epoch": 0.5, "learning_rate": 5.0242085661080074e-05, "loss": 1.0435, "step": 198805 }, { "epoch": 0.5, "learning_rate": 5.0240827419598366e-05, "loss": 1.0424, "step": 198810 }, { "epoch": 0.5, "learning_rate": 5.0239569178116664e-05, "loss": 1.0432, "step": 198815 }, { "epoch": 0.5, "learning_rate": 5.0238310936634956e-05, "loss": 1.0447, "step": 198820 }, { "epoch": 0.5, "learning_rate": 5.0237052695153254e-05, "loss": 1.0412, "step": 198825 }, { "epoch": 0.5, "learning_rate": 5.0235794453671546e-05, "loss": 1.0444, "step": 198830 }, { "epoch": 0.5, "learning_rate": 5.0234536212189844e-05, "loss": 1.0445, "step": 198835 }, { "epoch": 0.5, "learning_rate": 5.0233277970708136e-05, "loss": 1.0461, "step": 198840 }, { "epoch": 0.5, "learning_rate": 5.0232019729226434e-05, "loss": 1.0464, "step": 198845 }, { "epoch": 0.5, "learning_rate": 5.0230761487744726e-05, "loss": 1.0768, "step": 198850 }, { "epoch": 0.5, "learning_rate": 5.0229503246263024e-05, "loss": 1.0432, "step": 198855 }, { "epoch": 0.5, "learning_rate": 5.0228245004781316e-05, "loss": 1.0455, "step": 198860 }, { "epoch": 0.5, "learning_rate": 5.022698676329961e-05, "loss": 1.0434, "step": 198865 }, { "epoch": 0.5, "learning_rate": 5.0225728521817906e-05, "loss": 1.0438, "step": 198870 }, { "epoch": 0.5, "learning_rate": 5.02244702803362e-05, "loss": 1.0619, "step": 198875 }, { "epoch": 0.5, "learning_rate": 5.0223212038854496e-05, "loss": 1.043, "step": 198880 }, { "epoch": 0.5, "learning_rate": 5.022195379737279e-05, "loss": 1.0441, "step": 198885 }, { "epoch": 0.5, "learning_rate": 5.0220695555891086e-05, "loss": 1.0464, "step": 198890 }, { "epoch": 0.5, "learning_rate": 5.021943731440938e-05, "loss": 1.042, "step": 198895 }, { "epoch": 0.5, "learning_rate": 5.0218179072927676e-05, "loss": 1.0463, "step": 198900 }, { "epoch": 0.5, "learning_rate": 5.021692083144597e-05, "loss": 1.0442, "step": 198905 }, { "epoch": 0.5, "learning_rate": 5.0215662589964266e-05, "loss": 1.0466, "step": 198910 }, { "epoch": 0.5, "learning_rate": 5.021440434848256e-05, "loss": 1.0466, "step": 198915 }, { "epoch": 0.5, "learning_rate": 5.021314610700086e-05, "loss": 1.0411, "step": 198920 }, { "epoch": 0.5, "learning_rate": 5.021188786551916e-05, "loss": 1.0443, "step": 198925 }, { "epoch": 0.5, "learning_rate": 5.021062962403745e-05, "loss": 1.042, "step": 198930 }, { "epoch": 0.5, "learning_rate": 5.020937138255575e-05, "loss": 1.0436, "step": 198935 }, { "epoch": 0.5, "learning_rate": 5.020811314107404e-05, "loss": 1.0454, "step": 198940 }, { "epoch": 0.5, "learning_rate": 5.0206854899592334e-05, "loss": 1.043, "step": 198945 }, { "epoch": 0.5, "learning_rate": 5.020559665811063e-05, "loss": 1.0428, "step": 198950 }, { "epoch": 0.5, "learning_rate": 5.0204338416628924e-05, "loss": 1.0447, "step": 198955 }, { "epoch": 0.5, "learning_rate": 5.020308017514722e-05, "loss": 1.048, "step": 198960 }, { "epoch": 0.5, "learning_rate": 5.0201821933665514e-05, "loss": 1.069, "step": 198965 }, { "epoch": 0.5, "learning_rate": 5.020056369218381e-05, "loss": 1.0461, "step": 198970 }, { "epoch": 0.5, "learning_rate": 5.0199305450702104e-05, "loss": 1.0446, "step": 198975 }, { "epoch": 0.5, "learning_rate": 5.01980472092204e-05, "loss": 1.0456, "step": 198980 }, { "epoch": 0.5, "learning_rate": 5.0196788967738694e-05, "loss": 1.043, "step": 198985 }, { "epoch": 0.5, "learning_rate": 5.019553072625699e-05, "loss": 1.0435, "step": 198990 }, { "epoch": 0.5, "learning_rate": 5.0194272484775284e-05, "loss": 1.0455, "step": 198995 }, { "epoch": 0.5, "learning_rate": 5.019301424329358e-05, "loss": 1.0463, "step": 199000 }, { "epoch": 0.5, "learning_rate": 5.0191756001811874e-05, "loss": 1.0464, "step": 199005 }, { "epoch": 0.5, "learning_rate": 5.0190497760330165e-05, "loss": 1.0435, "step": 199010 }, { "epoch": 0.5, "learning_rate": 5.0189239518848464e-05, "loss": 1.0438, "step": 199015 }, { "epoch": 0.5, "learning_rate": 5.0187981277366755e-05, "loss": 1.0459, "step": 199020 }, { "epoch": 0.5, "learning_rate": 5.0186723035885054e-05, "loss": 1.0458, "step": 199025 }, { "epoch": 0.5, "learning_rate": 5.0185464794403345e-05, "loss": 1.0467, "step": 199030 }, { "epoch": 0.5, "learning_rate": 5.0184206552921644e-05, "loss": 1.0457, "step": 199035 }, { "epoch": 0.5, "learning_rate": 5.0182948311439935e-05, "loss": 1.0448, "step": 199040 }, { "epoch": 0.5, "learning_rate": 5.0181690069958234e-05, "loss": 1.043, "step": 199045 }, { "epoch": 0.5, "learning_rate": 5.0180431828476525e-05, "loss": 1.0465, "step": 199050 }, { "epoch": 0.5, "learning_rate": 5.0179173586994824e-05, "loss": 1.045, "step": 199055 }, { "epoch": 0.5, "learning_rate": 5.0177915345513115e-05, "loss": 1.0444, "step": 199060 }, { "epoch": 0.5, "learning_rate": 5.0176657104031414e-05, "loss": 1.0467, "step": 199065 }, { "epoch": 0.5, "learning_rate": 5.0175398862549705e-05, "loss": 1.0438, "step": 199070 }, { "epoch": 0.5, "learning_rate": 5.0174140621068e-05, "loss": 1.045, "step": 199075 }, { "epoch": 0.5, "learning_rate": 5.0172882379586295e-05, "loss": 1.0662, "step": 199080 }, { "epoch": 0.5, "learning_rate": 5.017162413810459e-05, "loss": 1.0442, "step": 199085 }, { "epoch": 0.5, "learning_rate": 5.0170365896622885e-05, "loss": 1.0474, "step": 199090 }, { "epoch": 0.5, "learning_rate": 5.016910765514118e-05, "loss": 1.0434, "step": 199095 }, { "epoch": 0.5, "learning_rate": 5.0167849413659475e-05, "loss": 1.0646, "step": 199100 }, { "epoch": 0.5, "learning_rate": 5.016659117217777e-05, "loss": 1.0449, "step": 199105 }, { "epoch": 0.5, "learning_rate": 5.0165332930696065e-05, "loss": 1.0434, "step": 199110 }, { "epoch": 0.5, "learning_rate": 5.0164074689214357e-05, "loss": 1.0435, "step": 199115 }, { "epoch": 0.5, "learning_rate": 5.0162816447732655e-05, "loss": 1.0465, "step": 199120 }, { "epoch": 0.5, "learning_rate": 5.0161558206250947e-05, "loss": 1.043, "step": 199125 }, { "epoch": 0.5, "learning_rate": 5.0160299964769245e-05, "loss": 1.0434, "step": 199130 }, { "epoch": 0.5, "learning_rate": 5.0159041723287537e-05, "loss": 1.0436, "step": 199135 }, { "epoch": 0.5, "learning_rate": 5.015778348180583e-05, "loss": 1.0405, "step": 199140 }, { "epoch": 0.5, "learning_rate": 5.0156525240324127e-05, "loss": 1.0454, "step": 199145 }, { "epoch": 0.5, "learning_rate": 5.015526699884242e-05, "loss": 1.0428, "step": 199150 }, { "epoch": 0.5, "learning_rate": 5.0154008757360716e-05, "loss": 1.0463, "step": 199155 }, { "epoch": 0.5, "learning_rate": 5.015275051587901e-05, "loss": 1.0452, "step": 199160 }, { "epoch": 0.5, "learning_rate": 5.0151492274397306e-05, "loss": 1.0482, "step": 199165 }, { "epoch": 0.5, "learning_rate": 5.01502340329156e-05, "loss": 1.0464, "step": 199170 }, { "epoch": 0.5, "learning_rate": 5.0148975791433896e-05, "loss": 1.044, "step": 199175 }, { "epoch": 0.5, "learning_rate": 5.014771754995219e-05, "loss": 1.0451, "step": 199180 }, { "epoch": 0.5, "learning_rate": 5.0146459308470486e-05, "loss": 1.0684, "step": 199185 }, { "epoch": 0.5, "learning_rate": 5.014520106698878e-05, "loss": 1.0458, "step": 199190 }, { "epoch": 0.5, "learning_rate": 5.0143942825507076e-05, "loss": 1.0437, "step": 199195 }, { "epoch": 0.5, "learning_rate": 5.014268458402537e-05, "loss": 1.0466, "step": 199200 }, { "epoch": 0.5, "learning_rate": 5.014142634254366e-05, "loss": 1.0452, "step": 199205 }, { "epoch": 0.5, "learning_rate": 5.014016810106196e-05, "loss": 1.0461, "step": 199210 }, { "epoch": 0.5, "learning_rate": 5.013890985958025e-05, "loss": 1.0442, "step": 199215 }, { "epoch": 0.5, "learning_rate": 5.013765161809855e-05, "loss": 1.0468, "step": 199220 }, { "epoch": 0.5, "learning_rate": 5.013639337661684e-05, "loss": 1.0439, "step": 199225 }, { "epoch": 0.5, "learning_rate": 5.013513513513514e-05, "loss": 1.0431, "step": 199230 }, { "epoch": 0.5, "learning_rate": 5.013387689365343e-05, "loss": 1.0468, "step": 199235 }, { "epoch": 0.5, "learning_rate": 5.013261865217173e-05, "loss": 1.0412, "step": 199240 }, { "epoch": 0.5, "learning_rate": 5.013136041069002e-05, "loss": 1.046, "step": 199245 }, { "epoch": 0.5, "learning_rate": 5.013010216920832e-05, "loss": 1.0419, "step": 199250 }, { "epoch": 0.5, "learning_rate": 5.012884392772661e-05, "loss": 1.047, "step": 199255 }, { "epoch": 0.5, "learning_rate": 5.012758568624491e-05, "loss": 1.0456, "step": 199260 }, { "epoch": 0.5, "learning_rate": 5.01263274447632e-05, "loss": 1.0464, "step": 199265 }, { "epoch": 0.5, "learning_rate": 5.012506920328149e-05, "loss": 1.0419, "step": 199270 }, { "epoch": 0.5, "learning_rate": 5.012381096179979e-05, "loss": 1.0439, "step": 199275 }, { "epoch": 0.5, "learning_rate": 5.012255272031808e-05, "loss": 1.0471, "step": 199280 }, { "epoch": 0.5, "learning_rate": 5.012129447883638e-05, "loss": 1.0432, "step": 199285 }, { "epoch": 0.5, "learning_rate": 5.012003623735467e-05, "loss": 1.0469, "step": 199290 }, { "epoch": 0.5, "learning_rate": 5.011877799587297e-05, "loss": 1.0474, "step": 199295 }, { "epoch": 0.5, "learning_rate": 5.011751975439126e-05, "loss": 1.0468, "step": 199300 }, { "epoch": 0.5, "learning_rate": 5.011626151290956e-05, "loss": 1.045, "step": 199305 }, { "epoch": 0.5, "learning_rate": 5.011500327142785e-05, "loss": 1.0429, "step": 199310 }, { "epoch": 0.5, "learning_rate": 5.011374502994615e-05, "loss": 1.0637, "step": 199315 }, { "epoch": 0.5, "learning_rate": 5.011248678846444e-05, "loss": 1.0439, "step": 199320 }, { "epoch": 0.5, "learning_rate": 5.011122854698273e-05, "loss": 1.0431, "step": 199325 }, { "epoch": 0.5, "learning_rate": 5.010997030550103e-05, "loss": 1.0456, "step": 199330 }, { "epoch": 0.5, "learning_rate": 5.010871206401932e-05, "loss": 1.0422, "step": 199335 }, { "epoch": 0.5, "learning_rate": 5.010745382253762e-05, "loss": 1.0415, "step": 199340 }, { "epoch": 0.5, "learning_rate": 5.010619558105591e-05, "loss": 1.0412, "step": 199345 }, { "epoch": 0.5, "learning_rate": 5.010493733957421e-05, "loss": 1.0449, "step": 199350 }, { "epoch": 0.5, "learning_rate": 5.01036790980925e-05, "loss": 1.046, "step": 199355 }, { "epoch": 0.5, "learning_rate": 5.01024208566108e-05, "loss": 1.0424, "step": 199360 }, { "epoch": 0.5, "learning_rate": 5.010116261512909e-05, "loss": 1.0427, "step": 199365 }, { "epoch": 0.5, "learning_rate": 5.009990437364739e-05, "loss": 1.0441, "step": 199370 }, { "epoch": 0.5, "learning_rate": 5.009864613216568e-05, "loss": 1.0443, "step": 199375 }, { "epoch": 0.5, "learning_rate": 5.009738789068398e-05, "loss": 1.0453, "step": 199380 }, { "epoch": 0.5, "learning_rate": 5.009612964920227e-05, "loss": 1.0478, "step": 199385 }, { "epoch": 0.5, "learning_rate": 5.0094871407720564e-05, "loss": 1.0418, "step": 199390 }, { "epoch": 0.5, "learning_rate": 5.009361316623886e-05, "loss": 1.0414, "step": 199395 }, { "epoch": 0.5, "learning_rate": 5.0092354924757154e-05, "loss": 1.0475, "step": 199400 }, { "epoch": 0.5, "learning_rate": 5.009109668327545e-05, "loss": 1.0417, "step": 199405 }, { "epoch": 0.5, "learning_rate": 5.0089838441793744e-05, "loss": 1.0436, "step": 199410 }, { "epoch": 0.5, "learning_rate": 5.008858020031204e-05, "loss": 1.0636, "step": 199415 }, { "epoch": 0.5, "learning_rate": 5.008732195883035e-05, "loss": 1.0665, "step": 199420 }, { "epoch": 0.5, "learning_rate": 5.008606371734864e-05, "loss": 1.0414, "step": 199425 }, { "epoch": 0.5, "learning_rate": 5.008480547586694e-05, "loss": 1.0472, "step": 199430 }, { "epoch": 0.5, "learning_rate": 5.008354723438523e-05, "loss": 1.0449, "step": 199435 }, { "epoch": 0.5, "learning_rate": 5.008228899290353e-05, "loss": 1.0475, "step": 199440 }, { "epoch": 0.5, "learning_rate": 5.008103075142182e-05, "loss": 1.0429, "step": 199445 }, { "epoch": 0.5, "learning_rate": 5.007977250994012e-05, "loss": 1.0416, "step": 199450 }, { "epoch": 0.5, "learning_rate": 5.007851426845841e-05, "loss": 1.0447, "step": 199455 }, { "epoch": 0.5, "learning_rate": 5.007725602697671e-05, "loss": 1.044, "step": 199460 }, { "epoch": 0.5, "learning_rate": 5.0075997785495e-05, "loss": 1.0447, "step": 199465 }, { "epoch": 0.5, "learning_rate": 5.00747395440133e-05, "loss": 1.0447, "step": 199470 }, { "epoch": 0.5, "learning_rate": 5.007348130253159e-05, "loss": 1.0452, "step": 199475 }, { "epoch": 0.5, "learning_rate": 5.007222306104988e-05, "loss": 1.0464, "step": 199480 }, { "epoch": 0.5, "learning_rate": 5.007096481956818e-05, "loss": 1.0471, "step": 199485 }, { "epoch": 0.5, "learning_rate": 5.006970657808647e-05, "loss": 1.044, "step": 199490 }, { "epoch": 0.5, "learning_rate": 5.006844833660477e-05, "loss": 1.0441, "step": 199495 }, { "epoch": 0.5, "learning_rate": 5.006719009512306e-05, "loss": 1.0426, "step": 199500 }, { "epoch": 0.5, "learning_rate": 5.006593185364136e-05, "loss": 1.0395, "step": 199505 }, { "epoch": 0.5, "learning_rate": 5.006467361215965e-05, "loss": 1.0627, "step": 199510 }, { "epoch": 0.5, "learning_rate": 5.006341537067795e-05, "loss": 1.0462, "step": 199515 }, { "epoch": 0.5, "learning_rate": 5.006215712919624e-05, "loss": 1.0456, "step": 199520 }, { "epoch": 0.5, "learning_rate": 5.006089888771454e-05, "loss": 1.0459, "step": 199525 }, { "epoch": 0.5, "learning_rate": 5.005964064623283e-05, "loss": 1.0428, "step": 199530 }, { "epoch": 0.5, "learning_rate": 5.005838240475112e-05, "loss": 1.0452, "step": 199535 }, { "epoch": 0.5, "learning_rate": 5.005712416326942e-05, "loss": 1.044, "step": 199540 }, { "epoch": 0.5, "learning_rate": 5.005586592178771e-05, "loss": 1.0432, "step": 199545 }, { "epoch": 0.5, "learning_rate": 5.005460768030601e-05, "loss": 1.0403, "step": 199550 }, { "epoch": 0.5, "learning_rate": 5.00533494388243e-05, "loss": 1.0434, "step": 199555 }, { "epoch": 0.5, "learning_rate": 5.00520911973426e-05, "loss": 1.0444, "step": 199560 }, { "epoch": 0.5, "learning_rate": 5.005083295586089e-05, "loss": 1.0448, "step": 199565 }, { "epoch": 0.5, "learning_rate": 5.004957471437919e-05, "loss": 1.0434, "step": 199570 }, { "epoch": 0.5, "learning_rate": 5.004831647289748e-05, "loss": 1.0417, "step": 199575 }, { "epoch": 0.5, "learning_rate": 5.004705823141578e-05, "loss": 1.0435, "step": 199580 }, { "epoch": 0.5, "learning_rate": 5.004579998993407e-05, "loss": 1.0446, "step": 199585 }, { "epoch": 0.5, "learning_rate": 5.004454174845237e-05, "loss": 1.0448, "step": 199590 }, { "epoch": 0.5, "learning_rate": 5.004328350697066e-05, "loss": 1.0448, "step": 199595 }, { "epoch": 0.5, "learning_rate": 5.004202526548895e-05, "loss": 1.0455, "step": 199600 }, { "epoch": 0.5, "learning_rate": 5.004076702400725e-05, "loss": 1.0469, "step": 199605 }, { "epoch": 0.5, "learning_rate": 5.003950878252554e-05, "loss": 1.045, "step": 199610 }, { "epoch": 0.5, "learning_rate": 5.003825054104384e-05, "loss": 1.0454, "step": 199615 }, { "epoch": 0.5, "learning_rate": 5.003699229956213e-05, "loss": 1.0437, "step": 199620 }, { "epoch": 0.5, "learning_rate": 5.003573405808043e-05, "loss": 1.0443, "step": 199625 }, { "epoch": 0.5, "learning_rate": 5.003447581659872e-05, "loss": 1.0435, "step": 199630 }, { "epoch": 0.5, "learning_rate": 5.003321757511702e-05, "loss": 1.0461, "step": 199635 }, { "epoch": 0.5, "learning_rate": 5.003195933363531e-05, "loss": 1.0464, "step": 199640 }, { "epoch": 0.5, "learning_rate": 5.003070109215361e-05, "loss": 1.0448, "step": 199645 }, { "epoch": 0.5, "learning_rate": 5.00294428506719e-05, "loss": 1.0467, "step": 199650 }, { "epoch": 0.5, "learning_rate": 5.00281846091902e-05, "loss": 1.044, "step": 199655 }, { "epoch": 0.5, "learning_rate": 5.002692636770849e-05, "loss": 1.0439, "step": 199660 }, { "epoch": 0.5, "learning_rate": 5.0025668126226785e-05, "loss": 1.0431, "step": 199665 }, { "epoch": 0.5, "learning_rate": 5.002440988474508e-05, "loss": 1.0648, "step": 199670 }, { "epoch": 0.5, "learning_rate": 5.0023151643263375e-05, "loss": 1.0418, "step": 199675 }, { "epoch": 0.5, "learning_rate": 5.002189340178167e-05, "loss": 1.0438, "step": 199680 }, { "epoch": 0.5, "learning_rate": 5.0020635160299965e-05, "loss": 1.0446, "step": 199685 }, { "epoch": 0.5, "learning_rate": 5.001937691881826e-05, "loss": 1.0424, "step": 199690 }, { "epoch": 0.5, "learning_rate": 5.0018118677336555e-05, "loss": 1.045, "step": 199695 }, { "epoch": 0.5, "learning_rate": 5.001686043585485e-05, "loss": 1.0497, "step": 199700 }, { "epoch": 0.5, "learning_rate": 5.0015602194373145e-05, "loss": 1.048, "step": 199705 }, { "epoch": 0.5, "learning_rate": 5.001434395289144e-05, "loss": 1.0462, "step": 199710 }, { "epoch": 0.5, "learning_rate": 5.0013085711409735e-05, "loss": 1.0443, "step": 199715 }, { "epoch": 0.5, "learning_rate": 5.001182746992803e-05, "loss": 1.0444, "step": 199720 }, { "epoch": 0.5, "learning_rate": 5.0010569228446325e-05, "loss": 1.0647, "step": 199725 }, { "epoch": 0.5, "learning_rate": 5.0009310986964616e-05, "loss": 1.0442, "step": 199730 }, { "epoch": 0.5, "learning_rate": 5.0008052745482914e-05, "loss": 1.0449, "step": 199735 }, { "epoch": 0.5, "learning_rate": 5.0006794504001206e-05, "loss": 1.0479, "step": 199740 }, { "epoch": 0.5, "learning_rate": 5.0005536262519504e-05, "loss": 1.0403, "step": 199745 }, { "epoch": 0.5, "learning_rate": 5.0004278021037796e-05, "loss": 1.0438, "step": 199750 }, { "epoch": 0.5, "learning_rate": 5.0003019779556094e-05, "loss": 1.0663, "step": 199755 }, { "epoch": 0.5, "learning_rate": 5.0001761538074386e-05, "loss": 1.0471, "step": 199760 }, { "epoch": 0.5, "learning_rate": 5.0000503296592684e-05, "loss": 1.042, "step": 199765 }, { "epoch": 0.5, "learning_rate": 4.9999245055110976e-05, "loss": 1.0495, "step": 199770 }, { "epoch": 0.5, "learning_rate": 4.9997986813629274e-05, "loss": 1.0422, "step": 199775 }, { "epoch": 0.5, "learning_rate": 4.9996728572147566e-05, "loss": 1.044, "step": 199780 }, { "epoch": 0.5, "learning_rate": 4.9995470330665864e-05, "loss": 1.041, "step": 199785 }, { "epoch": 0.5, "learning_rate": 4.9994212089184156e-05, "loss": 1.0449, "step": 199790 }, { "epoch": 0.5, "learning_rate": 4.9992953847702454e-05, "loss": 1.042, "step": 199795 }, { "epoch": 0.5, "learning_rate": 4.999169560622075e-05, "loss": 1.0839, "step": 199800 }, { "epoch": 0.5, "learning_rate": 4.9990437364739044e-05, "loss": 1.0453, "step": 199805 }, { "epoch": 0.5, "learning_rate": 4.998917912325734e-05, "loss": 1.0449, "step": 199810 }, { "epoch": 0.5, "learning_rate": 4.998817253007197e-05, "loss": 1.071, "step": 199815 }, { "epoch": 0.5, "learning_rate": 4.998716593688661e-05, "loss": 1.0429, "step": 199820 }, { "epoch": 0.5, "learning_rate": 4.9985907695404906e-05, "loss": 1.0426, "step": 199825 }, { "epoch": 0.5, "learning_rate": 4.99846494539232e-05, "loss": 1.0439, "step": 199830 }, { "epoch": 0.5, "learning_rate": 4.9983391212441496e-05, "loss": 1.0403, "step": 199835 }, { "epoch": 0.5, "learning_rate": 4.998213297095979e-05, "loss": 1.0412, "step": 199840 }, { "epoch": 0.5, "learning_rate": 4.9980874729478086e-05, "loss": 1.0464, "step": 199845 }, { "epoch": 0.5, "learning_rate": 4.997961648799638e-05, "loss": 1.0456, "step": 199850 }, { "epoch": 0.5, "learning_rate": 4.997835824651467e-05, "loss": 1.0449, "step": 199855 }, { "epoch": 0.5, "learning_rate": 4.997710000503297e-05, "loss": 1.0481, "step": 199860 }, { "epoch": 0.5, "learning_rate": 4.997584176355126e-05, "loss": 1.0426, "step": 199865 }, { "epoch": 0.5, "learning_rate": 4.997458352206956e-05, "loss": 1.0434, "step": 199870 }, { "epoch": 0.5, "learning_rate": 4.997332528058785e-05, "loss": 1.0438, "step": 199875 }, { "epoch": 0.5, "learning_rate": 4.997206703910615e-05, "loss": 1.045, "step": 199880 }, { "epoch": 0.5, "learning_rate": 4.997080879762444e-05, "loss": 1.0464, "step": 199885 }, { "epoch": 0.5, "learning_rate": 4.996955055614274e-05, "loss": 1.0426, "step": 199890 }, { "epoch": 0.5, "learning_rate": 4.996829231466103e-05, "loss": 1.0429, "step": 199895 }, { "epoch": 0.5, "learning_rate": 4.996703407317933e-05, "loss": 1.0438, "step": 199900 }, { "epoch": 0.5, "learning_rate": 4.996577583169762e-05, "loss": 1.0443, "step": 199905 }, { "epoch": 0.5, "learning_rate": 4.996451759021592e-05, "loss": 1.0472, "step": 199910 }, { "epoch": 0.5, "learning_rate": 4.996325934873421e-05, "loss": 1.045, "step": 199915 }, { "epoch": 0.5, "learning_rate": 4.99620011072525e-05, "loss": 1.046, "step": 199920 }, { "epoch": 0.5, "learning_rate": 4.99607428657708e-05, "loss": 1.0488, "step": 199925 }, { "epoch": 0.5, "learning_rate": 4.995948462428909e-05, "loss": 1.0439, "step": 199930 }, { "epoch": 0.5, "learning_rate": 4.995822638280739e-05, "loss": 1.0472, "step": 199935 }, { "epoch": 0.5, "learning_rate": 4.995696814132568e-05, "loss": 1.0448, "step": 199940 }, { "epoch": 0.5, "learning_rate": 4.9955709899843986e-05, "loss": 1.0456, "step": 199945 }, { "epoch": 0.5, "learning_rate": 4.995445165836228e-05, "loss": 1.0445, "step": 199950 }, { "epoch": 0.5, "learning_rate": 4.9953193416880576e-05, "loss": 1.0433, "step": 199955 }, { "epoch": 0.5, "learning_rate": 4.995193517539887e-05, "loss": 1.0707, "step": 199960 }, { "epoch": 0.5, "learning_rate": 4.995067693391716e-05, "loss": 1.0442, "step": 199965 }, { "epoch": 0.5, "learning_rate": 4.994941869243546e-05, "loss": 1.0447, "step": 199970 }, { "epoch": 0.5, "learning_rate": 4.994816045095375e-05, "loss": 1.0442, "step": 199975 }, { "epoch": 0.5, "learning_rate": 4.994690220947205e-05, "loss": 1.0439, "step": 199980 }, { "epoch": 0.5, "learning_rate": 4.994564396799034e-05, "loss": 1.0423, "step": 199985 }, { "epoch": 0.5, "learning_rate": 4.994438572650864e-05, "loss": 1.0467, "step": 199990 }, { "epoch": 0.5, "learning_rate": 4.994312748502693e-05, "loss": 1.0432, "step": 199995 }, { "epoch": 0.5, "learning_rate": 4.994186924354523e-05, "loss": 1.045, "step": 200000 }, { "epoch": 0.5, "learning_rate": 4.994061100206352e-05, "loss": 1.0445, "step": 200005 }, { "epoch": 0.5, "learning_rate": 4.993935276058182e-05, "loss": 1.0423, "step": 200010 }, { "epoch": 0.5, "learning_rate": 4.993809451910011e-05, "loss": 1.0464, "step": 200015 }, { "epoch": 0.5, "learning_rate": 4.993683627761841e-05, "loss": 1.0441, "step": 200020 }, { "epoch": 0.5, "learning_rate": 4.99355780361367e-05, "loss": 1.0421, "step": 200025 }, { "epoch": 0.5, "learning_rate": 4.993431979465499e-05, "loss": 1.0455, "step": 200030 }, { "epoch": 0.5, "learning_rate": 4.993306155317329e-05, "loss": 1.0423, "step": 200035 }, { "epoch": 0.5, "learning_rate": 4.993180331169158e-05, "loss": 1.0463, "step": 200040 }, { "epoch": 0.5, "learning_rate": 4.993054507020988e-05, "loss": 1.0419, "step": 200045 }, { "epoch": 0.5, "learning_rate": 4.992928682872817e-05, "loss": 1.045, "step": 200050 }, { "epoch": 0.5, "learning_rate": 4.992802858724647e-05, "loss": 1.0456, "step": 200055 }, { "epoch": 0.5, "learning_rate": 4.992677034576476e-05, "loss": 1.0429, "step": 200060 }, { "epoch": 0.5, "learning_rate": 4.992551210428306e-05, "loss": 1.0446, "step": 200065 }, { "epoch": 0.5, "learning_rate": 4.992425386280135e-05, "loss": 1.0415, "step": 200070 }, { "epoch": 0.5, "learning_rate": 4.992299562131965e-05, "loss": 1.0452, "step": 200075 }, { "epoch": 0.5, "learning_rate": 4.992173737983794e-05, "loss": 1.0437, "step": 200080 }, { "epoch": 0.5, "learning_rate": 4.992047913835624e-05, "loss": 1.0427, "step": 200085 }, { "epoch": 0.5, "learning_rate": 4.991922089687453e-05, "loss": 1.0414, "step": 200090 }, { "epoch": 0.5, "learning_rate": 4.991796265539282e-05, "loss": 1.0462, "step": 200095 }, { "epoch": 0.5, "learning_rate": 4.991670441391112e-05, "loss": 1.0446, "step": 200100 }, { "epoch": 0.5, "learning_rate": 4.991544617242941e-05, "loss": 1.0476, "step": 200105 }, { "epoch": 0.5, "learning_rate": 4.991418793094771e-05, "loss": 1.0462, "step": 200110 }, { "epoch": 0.5, "learning_rate": 4.9912929689466e-05, "loss": 1.0458, "step": 200115 }, { "epoch": 0.5, "learning_rate": 4.99116714479843e-05, "loss": 1.0452, "step": 200120 }, { "epoch": 0.5, "learning_rate": 4.991041320650259e-05, "loss": 1.0455, "step": 200125 }, { "epoch": 0.5, "learning_rate": 4.990915496502089e-05, "loss": 1.0649, "step": 200130 }, { "epoch": 0.5, "learning_rate": 4.990789672353918e-05, "loss": 1.0462, "step": 200135 }, { "epoch": 0.5, "learning_rate": 4.990663848205748e-05, "loss": 1.0442, "step": 200140 }, { "epoch": 0.5, "learning_rate": 4.990538024057577e-05, "loss": 1.0454, "step": 200145 }, { "epoch": 0.5, "learning_rate": 4.990412199909406e-05, "loss": 1.0427, "step": 200150 }, { "epoch": 0.5, "learning_rate": 4.990286375761236e-05, "loss": 1.0451, "step": 200155 }, { "epoch": 0.5, "learning_rate": 4.990160551613065e-05, "loss": 1.0465, "step": 200160 }, { "epoch": 0.5, "learning_rate": 4.990034727464895e-05, "loss": 1.047, "step": 200165 }, { "epoch": 0.5, "learning_rate": 4.989908903316724e-05, "loss": 1.0622, "step": 200170 }, { "epoch": 0.5, "learning_rate": 4.989783079168554e-05, "loss": 1.0422, "step": 200175 }, { "epoch": 0.5, "learning_rate": 4.989657255020383e-05, "loss": 1.0477, "step": 200180 }, { "epoch": 0.5, "learning_rate": 4.989531430872213e-05, "loss": 1.0472, "step": 200185 }, { "epoch": 0.5, "learning_rate": 4.989405606724042e-05, "loss": 1.0427, "step": 200190 }, { "epoch": 0.5, "learning_rate": 4.989279782575872e-05, "loss": 1.0446, "step": 200195 }, { "epoch": 0.5, "learning_rate": 4.989153958427702e-05, "loss": 1.0418, "step": 200200 }, { "epoch": 0.5, "learning_rate": 4.989028134279531e-05, "loss": 1.0443, "step": 200205 }, { "epoch": 0.5, "learning_rate": 4.988902310131361e-05, "loss": 1.0428, "step": 200210 }, { "epoch": 0.5, "learning_rate": 4.98877648598319e-05, "loss": 1.046, "step": 200215 }, { "epoch": 0.5, "learning_rate": 4.98865066183502e-05, "loss": 1.0435, "step": 200220 }, { "epoch": 0.5, "learning_rate": 4.988524837686849e-05, "loss": 1.0446, "step": 200225 }, { "epoch": 0.5, "learning_rate": 4.988399013538679e-05, "loss": 1.0459, "step": 200230 }, { "epoch": 0.5, "learning_rate": 4.988273189390508e-05, "loss": 1.0456, "step": 200235 }, { "epoch": 0.5, "learning_rate": 4.988147365242338e-05, "loss": 1.0666, "step": 200240 }, { "epoch": 0.5, "learning_rate": 4.988021541094167e-05, "loss": 1.043, "step": 200245 }, { "epoch": 0.5, "learning_rate": 4.987895716945997e-05, "loss": 1.0421, "step": 200250 }, { "epoch": 0.5, "learning_rate": 4.987769892797826e-05, "loss": 1.0419, "step": 200255 }, { "epoch": 0.5, "learning_rate": 4.987644068649655e-05, "loss": 1.0459, "step": 200260 }, { "epoch": 0.5, "learning_rate": 4.987518244501485e-05, "loss": 1.0437, "step": 200265 }, { "epoch": 0.5, "learning_rate": 4.987392420353314e-05, "loss": 1.0457, "step": 200270 }, { "epoch": 0.5, "learning_rate": 4.987266596205144e-05, "loss": 1.0446, "step": 200275 }, { "epoch": 0.5, "learning_rate": 4.987140772056973e-05, "loss": 1.0446, "step": 200280 }, { "epoch": 0.5, "learning_rate": 4.987014947908803e-05, "loss": 1.0484, "step": 200285 }, { "epoch": 0.5, "learning_rate": 4.986889123760632e-05, "loss": 1.0444, "step": 200290 }, { "epoch": 0.5, "learning_rate": 4.986763299612462e-05, "loss": 1.0483, "step": 200295 }, { "epoch": 0.5, "learning_rate": 4.986637475464291e-05, "loss": 1.0443, "step": 200300 }, { "epoch": 0.5, "learning_rate": 4.986511651316121e-05, "loss": 1.0442, "step": 200305 }, { "epoch": 0.5, "learning_rate": 4.98638582716795e-05, "loss": 1.0444, "step": 200310 }, { "epoch": 0.5, "learning_rate": 4.98626000301978e-05, "loss": 1.042, "step": 200315 }, { "epoch": 0.5, "learning_rate": 4.986134178871609e-05, "loss": 1.0466, "step": 200320 }, { "epoch": 0.5, "learning_rate": 4.9860083547234384e-05, "loss": 1.0449, "step": 200325 }, { "epoch": 0.5, "learning_rate": 4.985882530575268e-05, "loss": 1.0434, "step": 200330 }, { "epoch": 0.5, "learning_rate": 4.9857567064270974e-05, "loss": 1.0428, "step": 200335 }, { "epoch": 0.5, "learning_rate": 4.985630882278927e-05, "loss": 1.0426, "step": 200340 }, { "epoch": 0.5, "learning_rate": 4.9855050581307564e-05, "loss": 1.0458, "step": 200345 }, { "epoch": 0.5, "learning_rate": 4.985379233982586e-05, "loss": 1.0444, "step": 200350 }, { "epoch": 0.5, "learning_rate": 4.9852534098344154e-05, "loss": 1.0434, "step": 200355 }, { "epoch": 0.5, "learning_rate": 4.985127585686245e-05, "loss": 1.0456, "step": 200360 }, { "epoch": 0.5, "learning_rate": 4.9850017615380744e-05, "loss": 1.0446, "step": 200365 }, { "epoch": 0.5, "learning_rate": 4.984875937389904e-05, "loss": 1.0465, "step": 200370 }, { "epoch": 0.5, "learning_rate": 4.9847501132417334e-05, "loss": 1.0436, "step": 200375 }, { "epoch": 0.5, "learning_rate": 4.984624289093563e-05, "loss": 1.0421, "step": 200380 }, { "epoch": 0.5, "learning_rate": 4.9844984649453924e-05, "loss": 1.0406, "step": 200385 }, { "epoch": 0.5, "learning_rate": 4.9843726407972216e-05, "loss": 1.0442, "step": 200390 }, { "epoch": 0.5, "learning_rate": 4.9842468166490514e-05, "loss": 1.0428, "step": 200395 }, { "epoch": 0.5, "learning_rate": 4.9841209925008806e-05, "loss": 1.0425, "step": 200400 }, { "epoch": 0.5, "learning_rate": 4.9839951683527104e-05, "loss": 1.0452, "step": 200405 }, { "epoch": 0.5, "learning_rate": 4.9838693442045396e-05, "loss": 1.045, "step": 200410 }, { "epoch": 0.5, "learning_rate": 4.9837435200563694e-05, "loss": 1.0455, "step": 200415 }, { "epoch": 0.5, "learning_rate": 4.9836176959081986e-05, "loss": 1.0465, "step": 200420 }, { "epoch": 0.5, "learning_rate": 4.9834918717600284e-05, "loss": 1.0464, "step": 200425 }, { "epoch": 0.5, "learning_rate": 4.9833660476118576e-05, "loss": 1.0464, "step": 200430 }, { "epoch": 0.5, "learning_rate": 4.9832402234636874e-05, "loss": 1.0646, "step": 200435 }, { "epoch": 0.5, "learning_rate": 4.9831143993155166e-05, "loss": 1.0465, "step": 200440 }, { "epoch": 0.5, "learning_rate": 4.9829885751673464e-05, "loss": 1.0459, "step": 200445 }, { "epoch": 0.5, "learning_rate": 4.982862751019176e-05, "loss": 1.0464, "step": 200450 }, { "epoch": 0.5, "learning_rate": 4.9827369268710054e-05, "loss": 1.0418, "step": 200455 }, { "epoch": 0.5, "learning_rate": 4.982611102722835e-05, "loss": 1.043, "step": 200460 }, { "epoch": 0.5, "learning_rate": 4.9824852785746644e-05, "loss": 1.0427, "step": 200465 }, { "epoch": 0.5, "learning_rate": 4.982359454426494e-05, "loss": 1.0443, "step": 200470 }, { "epoch": 0.5, "learning_rate": 4.9822336302783234e-05, "loss": 1.0444, "step": 200475 }, { "epoch": 0.5, "learning_rate": 4.982107806130153e-05, "loss": 1.0434, "step": 200480 }, { "epoch": 0.5, "learning_rate": 4.9819819819819824e-05, "loss": 1.0439, "step": 200485 }, { "epoch": 0.5, "learning_rate": 4.9818561578338116e-05, "loss": 1.043, "step": 200490 }, { "epoch": 0.5, "learning_rate": 4.9817303336856414e-05, "loss": 1.0461, "step": 200495 }, { "epoch": 0.5, "learning_rate": 4.9816045095374706e-05, "loss": 1.0457, "step": 200500 }, { "epoch": 0.5, "learning_rate": 4.9814786853893004e-05, "loss": 1.0456, "step": 200505 }, { "epoch": 0.5, "learning_rate": 4.9813528612411296e-05, "loss": 1.0451, "step": 200510 }, { "epoch": 0.5, "learning_rate": 4.9812270370929594e-05, "loss": 1.0456, "step": 200515 }, { "epoch": 0.5, "learning_rate": 4.9811012129447886e-05, "loss": 1.0446, "step": 200520 }, { "epoch": 0.5, "learning_rate": 4.9809753887966184e-05, "loss": 1.0458, "step": 200525 }, { "epoch": 0.5, "learning_rate": 4.9808495646484475e-05, "loss": 1.0441, "step": 200530 }, { "epoch": 0.5, "learning_rate": 4.9807237405002774e-05, "loss": 1.0485, "step": 200535 }, { "epoch": 0.5, "learning_rate": 4.9805979163521065e-05, "loss": 1.0426, "step": 200540 }, { "epoch": 0.5, "learning_rate": 4.9804720922039364e-05, "loss": 1.0457, "step": 200545 }, { "epoch": 0.5, "learning_rate": 4.9803462680557655e-05, "loss": 1.0462, "step": 200550 }, { "epoch": 0.5, "learning_rate": 4.980220443907595e-05, "loss": 1.0454, "step": 200555 }, { "epoch": 0.5, "learning_rate": 4.9800946197594245e-05, "loss": 1.0433, "step": 200560 }, { "epoch": 0.5, "learning_rate": 4.979968795611254e-05, "loss": 1.0449, "step": 200565 }, { "epoch": 0.5, "learning_rate": 4.9798429714630835e-05, "loss": 1.0441, "step": 200570 }, { "epoch": 0.5, "learning_rate": 4.979717147314913e-05, "loss": 1.0458, "step": 200575 }, { "epoch": 0.5, "learning_rate": 4.9795913231667425e-05, "loss": 1.043, "step": 200580 }, { "epoch": 0.5, "learning_rate": 4.979465499018572e-05, "loss": 1.0441, "step": 200585 }, { "epoch": 0.5, "learning_rate": 4.9793396748704015e-05, "loss": 1.0428, "step": 200590 }, { "epoch": 0.5, "learning_rate": 4.979213850722231e-05, "loss": 1.0424, "step": 200595 }, { "epoch": 0.5, "learning_rate": 4.9790880265740605e-05, "loss": 1.0488, "step": 200600 }, { "epoch": 0.5, "learning_rate": 4.97896220242589e-05, "loss": 1.0434, "step": 200605 }, { "epoch": 0.5, "learning_rate": 4.9788363782777195e-05, "loss": 1.0445, "step": 200610 }, { "epoch": 0.5, "learning_rate": 4.978710554129549e-05, "loss": 1.043, "step": 200615 }, { "epoch": 0.5, "learning_rate": 4.978584729981378e-05, "loss": 1.0429, "step": 200620 }, { "epoch": 0.5, "learning_rate": 4.978458905833208e-05, "loss": 1.0464, "step": 200625 }, { "epoch": 0.5, "learning_rate": 4.978333081685037e-05, "loss": 1.0451, "step": 200630 }, { "epoch": 0.5, "learning_rate": 4.978207257536867e-05, "loss": 1.0426, "step": 200635 }, { "epoch": 0.5, "learning_rate": 4.978081433388696e-05, "loss": 1.0644, "step": 200640 }, { "epoch": 0.5, "learning_rate": 4.977955609240526e-05, "loss": 1.0446, "step": 200645 }, { "epoch": 0.5, "learning_rate": 4.977829785092355e-05, "loss": 1.046, "step": 200650 }, { "epoch": 0.5, "learning_rate": 4.977703960944185e-05, "loss": 1.0441, "step": 200655 }, { "epoch": 0.5, "learning_rate": 4.977578136796014e-05, "loss": 1.0452, "step": 200660 }, { "epoch": 0.5, "learning_rate": 4.977452312647844e-05, "loss": 1.0459, "step": 200665 }, { "epoch": 0.5, "learning_rate": 4.977326488499673e-05, "loss": 1.0473, "step": 200670 }, { "epoch": 0.5, "learning_rate": 4.977200664351503e-05, "loss": 1.0431, "step": 200675 }, { "epoch": 0.5, "learning_rate": 4.977074840203332e-05, "loss": 1.0441, "step": 200680 }, { "epoch": 0.5, "learning_rate": 4.976949016055161e-05, "loss": 1.0441, "step": 200685 }, { "epoch": 0.5, "learning_rate": 4.976823191906991e-05, "loss": 1.0449, "step": 200690 }, { "epoch": 0.5, "learning_rate": 4.976697367758821e-05, "loss": 1.0437, "step": 200695 }, { "epoch": 0.5, "learning_rate": 4.9765715436106505e-05, "loss": 1.0435, "step": 200700 }, { "epoch": 0.5, "learning_rate": 4.9764457194624797e-05, "loss": 1.0415, "step": 200705 }, { "epoch": 0.5, "learning_rate": 4.9763198953143095e-05, "loss": 1.0446, "step": 200710 }, { "epoch": 0.5, "learning_rate": 4.9761940711661387e-05, "loss": 1.0394, "step": 200715 }, { "epoch": 0.5, "learning_rate": 4.976068247017968e-05, "loss": 1.0442, "step": 200720 }, { "epoch": 0.5, "learning_rate": 4.9759424228697977e-05, "loss": 1.0443, "step": 200725 }, { "epoch": 0.5, "learning_rate": 4.975816598721627e-05, "loss": 1.0442, "step": 200730 }, { "epoch": 0.5, "learning_rate": 4.9756907745734567e-05, "loss": 1.0399, "step": 200735 }, { "epoch": 0.5, "learning_rate": 4.975564950425286e-05, "loss": 1.0464, "step": 200740 }, { "epoch": 0.5, "learning_rate": 4.9754391262771156e-05, "loss": 1.0435, "step": 200745 }, { "epoch": 0.5, "learning_rate": 4.975313302128945e-05, "loss": 1.0449, "step": 200750 }, { "epoch": 0.5, "learning_rate": 4.9751874779807746e-05, "loss": 1.0418, "step": 200755 }, { "epoch": 0.5, "learning_rate": 4.975061653832604e-05, "loss": 1.0473, "step": 200760 }, { "epoch": 0.5, "learning_rate": 4.9749358296844336e-05, "loss": 1.0452, "step": 200765 }, { "epoch": 0.5, "learning_rate": 4.974810005536263e-05, "loss": 1.0453, "step": 200770 }, { "epoch": 0.5, "learning_rate": 4.9746841813880926e-05, "loss": 1.0421, "step": 200775 }, { "epoch": 0.5, "learning_rate": 4.974558357239922e-05, "loss": 1.0481, "step": 200780 }, { "epoch": 0.5, "learning_rate": 4.974432533091751e-05, "loss": 1.0436, "step": 200785 }, { "epoch": 0.5, "learning_rate": 4.974306708943581e-05, "loss": 1.0415, "step": 200790 }, { "epoch": 0.5, "learning_rate": 4.97418088479541e-05, "loss": 1.0455, "step": 200795 }, { "epoch": 0.5, "learning_rate": 4.97405506064724e-05, "loss": 1.0575, "step": 200800 }, { "epoch": 0.5, "learning_rate": 4.973929236499069e-05, "loss": 1.0467, "step": 200805 }, { "epoch": 0.5, "learning_rate": 4.973803412350899e-05, "loss": 1.0431, "step": 200810 }, { "epoch": 0.5, "learning_rate": 4.973677588202728e-05, "loss": 1.0473, "step": 200815 }, { "epoch": 0.5, "learning_rate": 4.973551764054558e-05, "loss": 1.0417, "step": 200820 }, { "epoch": 0.5, "learning_rate": 4.973425939906387e-05, "loss": 1.0466, "step": 200825 }, { "epoch": 0.5, "learning_rate": 4.973300115758217e-05, "loss": 1.0492, "step": 200830 }, { "epoch": 0.5, "learning_rate": 4.973174291610046e-05, "loss": 1.0482, "step": 200835 }, { "epoch": 0.5, "learning_rate": 4.973048467461876e-05, "loss": 1.043, "step": 200840 }, { "epoch": 0.5, "learning_rate": 4.972922643313705e-05, "loss": 1.0416, "step": 200845 }, { "epoch": 0.5, "learning_rate": 4.972796819165534e-05, "loss": 1.0414, "step": 200850 }, { "epoch": 0.5, "learning_rate": 4.972670995017364e-05, "loss": 1.0451, "step": 200855 }, { "epoch": 0.5, "learning_rate": 4.972545170869193e-05, "loss": 1.0906, "step": 200860 }, { "epoch": 0.5, "learning_rate": 4.972419346721023e-05, "loss": 1.0434, "step": 200865 }, { "epoch": 0.5, "learning_rate": 4.972293522572852e-05, "loss": 1.044, "step": 200870 }, { "epoch": 0.5, "learning_rate": 4.972167698424682e-05, "loss": 1.0448, "step": 200875 }, { "epoch": 0.5, "learning_rate": 4.972041874276511e-05, "loss": 1.0468, "step": 200880 }, { "epoch": 0.5, "learning_rate": 4.971916050128341e-05, "loss": 1.0682, "step": 200885 }, { "epoch": 0.5, "learning_rate": 4.97179022598017e-05, "loss": 1.044, "step": 200890 }, { "epoch": 0.5, "learning_rate": 4.971664401832e-05, "loss": 1.0474, "step": 200895 }, { "epoch": 0.5, "learning_rate": 4.971538577683829e-05, "loss": 1.0459, "step": 200900 }, { "epoch": 0.5, "learning_rate": 4.971412753535659e-05, "loss": 1.0434, "step": 200905 }, { "epoch": 0.5, "learning_rate": 4.971286929387488e-05, "loss": 1.0457, "step": 200910 }, { "epoch": 0.5, "learning_rate": 4.971161105239317e-05, "loss": 1.0432, "step": 200915 }, { "epoch": 0.5, "learning_rate": 4.971035281091147e-05, "loss": 1.0463, "step": 200920 }, { "epoch": 0.5, "learning_rate": 4.970909456942976e-05, "loss": 1.0442, "step": 200925 }, { "epoch": 0.5, "learning_rate": 4.970783632794806e-05, "loss": 1.0453, "step": 200930 }, { "epoch": 0.5, "learning_rate": 4.970657808646635e-05, "loss": 1.0449, "step": 200935 }, { "epoch": 0.5, "learning_rate": 4.970531984498465e-05, "loss": 1.0432, "step": 200940 }, { "epoch": 0.5, "learning_rate": 4.970406160350295e-05, "loss": 1.0463, "step": 200945 }, { "epoch": 0.5, "learning_rate": 4.970280336202124e-05, "loss": 1.0424, "step": 200950 }, { "epoch": 0.5, "learning_rate": 4.970154512053954e-05, "loss": 1.0434, "step": 200955 }, { "epoch": 0.5, "learning_rate": 4.970028687905783e-05, "loss": 1.0645, "step": 200960 }, { "epoch": 0.5, "learning_rate": 4.969902863757613e-05, "loss": 1.0444, "step": 200965 }, { "epoch": 0.5, "learning_rate": 4.969777039609442e-05, "loss": 1.0602, "step": 200970 }, { "epoch": 0.5, "learning_rate": 4.969651215461272e-05, "loss": 1.0455, "step": 200975 }, { "epoch": 0.5, "learning_rate": 4.969525391313101e-05, "loss": 1.0489, "step": 200980 }, { "epoch": 0.5, "learning_rate": 4.969399567164931e-05, "loss": 1.0416, "step": 200985 }, { "epoch": 0.5, "learning_rate": 4.96927374301676e-05, "loss": 1.0469, "step": 200990 }, { "epoch": 0.5, "learning_rate": 4.96914791886859e-05, "loss": 1.047, "step": 200995 }, { "epoch": 0.5, "learning_rate": 4.969022094720419e-05, "loss": 1.0447, "step": 201000 }, { "epoch": 0.5, "learning_rate": 4.968896270572249e-05, "loss": 1.0442, "step": 201005 }, { "epoch": 0.5, "learning_rate": 4.968770446424078e-05, "loss": 1.043, "step": 201010 }, { "epoch": 0.5, "learning_rate": 4.968644622275907e-05, "loss": 1.0448, "step": 201015 }, { "epoch": 0.5, "learning_rate": 4.968518798127737e-05, "loss": 1.0467, "step": 201020 }, { "epoch": 0.5, "learning_rate": 4.968392973979566e-05, "loss": 1.0442, "step": 201025 }, { "epoch": 0.5, "learning_rate": 4.968267149831396e-05, "loss": 1.0473, "step": 201030 }, { "epoch": 0.5, "learning_rate": 4.968141325683225e-05, "loss": 1.0458, "step": 201035 }, { "epoch": 0.5, "learning_rate": 4.968015501535055e-05, "loss": 1.0442, "step": 201040 }, { "epoch": 0.5, "learning_rate": 4.967889677386884e-05, "loss": 1.0441, "step": 201045 }, { "epoch": 0.5, "learning_rate": 4.967763853238714e-05, "loss": 1.0445, "step": 201050 }, { "epoch": 0.5, "learning_rate": 4.967638029090543e-05, "loss": 1.0448, "step": 201055 }, { "epoch": 0.5, "learning_rate": 4.967512204942373e-05, "loss": 1.0646, "step": 201060 }, { "epoch": 0.5, "learning_rate": 4.967386380794202e-05, "loss": 1.041, "step": 201065 }, { "epoch": 0.5, "learning_rate": 4.967260556646032e-05, "loss": 1.0441, "step": 201070 }, { "epoch": 0.5, "learning_rate": 4.967134732497861e-05, "loss": 1.0482, "step": 201075 }, { "epoch": 0.5, "learning_rate": 4.9670089083496904e-05, "loss": 1.0457, "step": 201080 }, { "epoch": 0.5, "learning_rate": 4.96688308420152e-05, "loss": 1.0435, "step": 201085 }, { "epoch": 0.5, "learning_rate": 4.9667572600533494e-05, "loss": 1.0451, "step": 201090 }, { "epoch": 0.5, "learning_rate": 4.966631435905179e-05, "loss": 1.04, "step": 201095 }, { "epoch": 0.5, "learning_rate": 4.9665056117570084e-05, "loss": 1.0457, "step": 201100 }, { "epoch": 0.5, "learning_rate": 4.966379787608838e-05, "loss": 1.0453, "step": 201105 }, { "epoch": 0.5, "learning_rate": 4.9662539634606673e-05, "loss": 1.0414, "step": 201110 }, { "epoch": 0.5, "learning_rate": 4.966128139312497e-05, "loss": 1.0447, "step": 201115 }, { "epoch": 0.5, "learning_rate": 4.9660023151643263e-05, "loss": 1.0469, "step": 201120 }, { "epoch": 0.5, "learning_rate": 4.965876491016156e-05, "loss": 1.0443, "step": 201125 }, { "epoch": 0.5, "learning_rate": 4.9657506668679853e-05, "loss": 1.0448, "step": 201130 }, { "epoch": 0.5, "learning_rate": 4.965624842719815e-05, "loss": 1.0437, "step": 201135 }, { "epoch": 0.5, "learning_rate": 4.9654990185716443e-05, "loss": 1.0428, "step": 201140 }, { "epoch": 0.5, "learning_rate": 4.9653731944234735e-05, "loss": 1.0435, "step": 201145 }, { "epoch": 0.5, "learning_rate": 4.965247370275303e-05, "loss": 1.046, "step": 201150 }, { "epoch": 0.5, "learning_rate": 4.9651215461271325e-05, "loss": 1.0416, "step": 201155 }, { "epoch": 0.5, "learning_rate": 4.964995721978962e-05, "loss": 1.046, "step": 201160 }, { "epoch": 0.5, "learning_rate": 4.9648698978307915e-05, "loss": 1.0444, "step": 201165 }, { "epoch": 0.5, "learning_rate": 4.964744073682621e-05, "loss": 1.0428, "step": 201170 }, { "epoch": 0.5, "learning_rate": 4.9646182495344505e-05, "loss": 1.0648, "step": 201175 }, { "epoch": 0.5, "learning_rate": 4.96449242538628e-05, "loss": 1.0469, "step": 201180 }, { "epoch": 0.51, "learning_rate": 4.9643666012381095e-05, "loss": 1.0439, "step": 201185 }, { "epoch": 0.51, "learning_rate": 4.964240777089939e-05, "loss": 1.0473, "step": 201190 }, { "epoch": 0.51, "learning_rate": 4.964114952941769e-05, "loss": 1.0472, "step": 201195 }, { "epoch": 0.51, "learning_rate": 4.963989128793598e-05, "loss": 1.0422, "step": 201200 }, { "epoch": 0.51, "learning_rate": 4.963863304645428e-05, "loss": 1.0423, "step": 201205 }, { "epoch": 0.51, "learning_rate": 4.963737480497257e-05, "loss": 1.0467, "step": 201210 }, { "epoch": 0.51, "learning_rate": 4.963611656349087e-05, "loss": 1.0451, "step": 201215 }, { "epoch": 0.51, "learning_rate": 4.963485832200916e-05, "loss": 1.0429, "step": 201220 }, { "epoch": 0.51, "learning_rate": 4.963360008052746e-05, "loss": 1.0429, "step": 201225 }, { "epoch": 0.51, "learning_rate": 4.963234183904575e-05, "loss": 1.046, "step": 201230 }, { "epoch": 0.51, "learning_rate": 4.963108359756405e-05, "loss": 1.0429, "step": 201235 }, { "epoch": 0.51, "learning_rate": 4.962982535608234e-05, "loss": 1.0461, "step": 201240 }, { "epoch": 0.51, "learning_rate": 4.962856711460064e-05, "loss": 1.045, "step": 201245 }, { "epoch": 0.51, "learning_rate": 4.962730887311893e-05, "loss": 1.0478, "step": 201250 }, { "epoch": 0.51, "learning_rate": 4.9626050631637225e-05, "loss": 1.0458, "step": 201255 }, { "epoch": 0.51, "learning_rate": 4.962479239015552e-05, "loss": 1.0466, "step": 201260 }, { "epoch": 0.51, "learning_rate": 4.9623534148673815e-05, "loss": 1.045, "step": 201265 }, { "epoch": 0.51, "learning_rate": 4.962227590719211e-05, "loss": 1.0455, "step": 201270 }, { "epoch": 0.51, "learning_rate": 4.9621017665710405e-05, "loss": 1.047, "step": 201275 }, { "epoch": 0.51, "learning_rate": 4.96197594242287e-05, "loss": 1.0503, "step": 201280 }, { "epoch": 0.51, "learning_rate": 4.9618501182746995e-05, "loss": 1.0448, "step": 201285 }, { "epoch": 0.51, "learning_rate": 4.961724294126529e-05, "loss": 1.0425, "step": 201290 }, { "epoch": 0.51, "learning_rate": 4.9615984699783585e-05, "loss": 1.0477, "step": 201295 }, { "epoch": 0.51, "learning_rate": 4.961472645830188e-05, "loss": 1.0438, "step": 201300 }, { "epoch": 0.51, "learning_rate": 4.9613468216820175e-05, "loss": 1.0464, "step": 201305 }, { "epoch": 0.51, "learning_rate": 4.9612209975338466e-05, "loss": 1.0435, "step": 201310 }, { "epoch": 0.51, "learning_rate": 4.9610951733856765e-05, "loss": 1.044, "step": 201315 }, { "epoch": 0.51, "learning_rate": 4.9609693492375056e-05, "loss": 1.0455, "step": 201320 }, { "epoch": 0.51, "learning_rate": 4.9608435250893354e-05, "loss": 1.0457, "step": 201325 }, { "epoch": 0.51, "learning_rate": 4.9607177009411646e-05, "loss": 1.0477, "step": 201330 }, { "epoch": 0.51, "learning_rate": 4.9605918767929944e-05, "loss": 1.0453, "step": 201335 }, { "epoch": 0.51, "learning_rate": 4.9604660526448236e-05, "loss": 1.0682, "step": 201340 }, { "epoch": 0.51, "learning_rate": 4.9603402284966534e-05, "loss": 1.0428, "step": 201345 }, { "epoch": 0.51, "learning_rate": 4.9602144043484826e-05, "loss": 1.0428, "step": 201350 }, { "epoch": 0.51, "learning_rate": 4.9600885802003124e-05, "loss": 1.0434, "step": 201355 }, { "epoch": 0.51, "learning_rate": 4.9599627560521416e-05, "loss": 1.043, "step": 201360 }, { "epoch": 0.51, "learning_rate": 4.9598369319039714e-05, "loss": 1.0485, "step": 201365 }, { "epoch": 0.51, "learning_rate": 4.9597111077558006e-05, "loss": 1.0464, "step": 201370 }, { "epoch": 0.51, "learning_rate": 4.95958528360763e-05, "loss": 1.0464, "step": 201375 }, { "epoch": 0.51, "learning_rate": 4.9594594594594596e-05, "loss": 1.0459, "step": 201380 }, { "epoch": 0.51, "learning_rate": 4.959333635311289e-05, "loss": 1.0431, "step": 201385 }, { "epoch": 0.51, "learning_rate": 4.9592078111631186e-05, "loss": 1.0445, "step": 201390 }, { "epoch": 0.51, "learning_rate": 4.959081987014948e-05, "loss": 1.0441, "step": 201395 }, { "epoch": 0.51, "learning_rate": 4.9589561628667776e-05, "loss": 1.0436, "step": 201400 }, { "epoch": 0.51, "learning_rate": 4.958830338718607e-05, "loss": 1.0455, "step": 201405 }, { "epoch": 0.51, "learning_rate": 4.9587045145704366e-05, "loss": 1.0459, "step": 201410 }, { "epoch": 0.51, "learning_rate": 4.958578690422266e-05, "loss": 1.0475, "step": 201415 }, { "epoch": 0.51, "learning_rate": 4.9584528662740956e-05, "loss": 1.0461, "step": 201420 }, { "epoch": 0.51, "learning_rate": 4.958327042125925e-05, "loss": 1.0429, "step": 201425 }, { "epoch": 0.51, "learning_rate": 4.9582012179777546e-05, "loss": 1.0442, "step": 201430 }, { "epoch": 0.51, "learning_rate": 4.958075393829584e-05, "loss": 1.0446, "step": 201435 }, { "epoch": 0.51, "learning_rate": 4.957949569681413e-05, "loss": 1.0449, "step": 201440 }, { "epoch": 0.51, "learning_rate": 4.9578237455332434e-05, "loss": 1.0433, "step": 201445 }, { "epoch": 0.51, "learning_rate": 4.9576979213850726e-05, "loss": 1.045, "step": 201450 }, { "epoch": 0.51, "learning_rate": 4.9575720972369024e-05, "loss": 1.0445, "step": 201455 }, { "epoch": 0.51, "learning_rate": 4.9574462730887316e-05, "loss": 1.0461, "step": 201460 }, { "epoch": 0.51, "learning_rate": 4.9573204489405614e-05, "loss": 1.0435, "step": 201465 }, { "epoch": 0.51, "learning_rate": 4.9571946247923906e-05, "loss": 1.0457, "step": 201470 }, { "epoch": 0.51, "learning_rate": 4.9570688006442204e-05, "loss": 1.045, "step": 201475 }, { "epoch": 0.51, "learning_rate": 4.9569429764960496e-05, "loss": 1.0428, "step": 201480 }, { "epoch": 0.51, "learning_rate": 4.956817152347879e-05, "loss": 1.0448, "step": 201485 }, { "epoch": 0.51, "learning_rate": 4.9566913281997086e-05, "loss": 1.0424, "step": 201490 }, { "epoch": 0.51, "learning_rate": 4.956565504051538e-05, "loss": 1.0428, "step": 201495 }, { "epoch": 0.51, "learning_rate": 4.9564396799033676e-05, "loss": 1.0437, "step": 201500 }, { "epoch": 0.51, "learning_rate": 4.956313855755197e-05, "loss": 1.0437, "step": 201505 }, { "epoch": 0.51, "learning_rate": 4.9561880316070266e-05, "loss": 1.0703, "step": 201510 }, { "epoch": 0.51, "learning_rate": 4.956062207458856e-05, "loss": 1.0471, "step": 201515 }, { "epoch": 0.51, "learning_rate": 4.9559363833106856e-05, "loss": 1.0457, "step": 201520 }, { "epoch": 0.51, "learning_rate": 4.955810559162515e-05, "loss": 1.0444, "step": 201525 }, { "epoch": 0.51, "learning_rate": 4.9556847350143445e-05, "loss": 1.0449, "step": 201530 }, { "epoch": 0.51, "learning_rate": 4.955558910866174e-05, "loss": 1.0441, "step": 201535 }, { "epoch": 0.51, "learning_rate": 4.9554330867180035e-05, "loss": 1.0431, "step": 201540 }, { "epoch": 0.51, "learning_rate": 4.955307262569833e-05, "loss": 1.0446, "step": 201545 }, { "epoch": 0.51, "learning_rate": 4.955181438421662e-05, "loss": 1.0463, "step": 201550 }, { "epoch": 0.51, "learning_rate": 4.955055614273492e-05, "loss": 1.0477, "step": 201555 }, { "epoch": 0.51, "learning_rate": 4.954929790125321e-05, "loss": 1.0719, "step": 201560 }, { "epoch": 0.51, "learning_rate": 4.954803965977151e-05, "loss": 1.0453, "step": 201565 }, { "epoch": 0.51, "learning_rate": 4.95467814182898e-05, "loss": 1.0484, "step": 201570 }, { "epoch": 0.51, "learning_rate": 4.95455231768081e-05, "loss": 1.044, "step": 201575 }, { "epoch": 0.51, "learning_rate": 4.954426493532639e-05, "loss": 1.0422, "step": 201580 }, { "epoch": 0.51, "learning_rate": 4.954300669384469e-05, "loss": 1.0413, "step": 201585 }, { "epoch": 0.51, "learning_rate": 4.954174845236298e-05, "loss": 1.0678, "step": 201590 }, { "epoch": 0.51, "learning_rate": 4.954049021088128e-05, "loss": 1.045, "step": 201595 }, { "epoch": 0.51, "learning_rate": 4.953923196939957e-05, "loss": 1.0443, "step": 201600 }, { "epoch": 0.51, "learning_rate": 4.953797372791787e-05, "loss": 1.0454, "step": 201605 }, { "epoch": 0.51, "learning_rate": 4.953671548643616e-05, "loss": 1.0435, "step": 201610 }, { "epoch": 0.51, "learning_rate": 4.953545724495445e-05, "loss": 1.0424, "step": 201615 }, { "epoch": 0.51, "learning_rate": 4.953419900347275e-05, "loss": 1.0461, "step": 201620 }, { "epoch": 0.51, "learning_rate": 4.953294076199104e-05, "loss": 1.0436, "step": 201625 }, { "epoch": 0.51, "learning_rate": 4.953168252050934e-05, "loss": 1.0417, "step": 201630 }, { "epoch": 0.51, "learning_rate": 4.953042427902763e-05, "loss": 1.0433, "step": 201635 }, { "epoch": 0.51, "learning_rate": 4.952916603754593e-05, "loss": 1.0453, "step": 201640 }, { "epoch": 0.51, "learning_rate": 4.952790779606422e-05, "loss": 1.0473, "step": 201645 }, { "epoch": 0.51, "learning_rate": 4.952664955458252e-05, "loss": 1.0449, "step": 201650 }, { "epoch": 0.51, "learning_rate": 4.952539131310081e-05, "loss": 1.0434, "step": 201655 }, { "epoch": 0.51, "learning_rate": 4.952413307161911e-05, "loss": 1.0464, "step": 201660 }, { "epoch": 0.51, "learning_rate": 4.95228748301374e-05, "loss": 1.0433, "step": 201665 }, { "epoch": 0.51, "learning_rate": 4.952161658865569e-05, "loss": 1.0472, "step": 201670 }, { "epoch": 0.51, "learning_rate": 4.952035834717399e-05, "loss": 1.0419, "step": 201675 }, { "epoch": 0.51, "learning_rate": 4.951910010569228e-05, "loss": 1.0417, "step": 201680 }, { "epoch": 0.51, "learning_rate": 4.951784186421058e-05, "loss": 1.0419, "step": 201685 }, { "epoch": 0.51, "learning_rate": 4.951658362272887e-05, "loss": 1.0443, "step": 201690 }, { "epoch": 0.51, "learning_rate": 4.951532538124718e-05, "loss": 1.0434, "step": 201695 }, { "epoch": 0.51, "learning_rate": 4.951406713976547e-05, "loss": 1.0448, "step": 201700 }, { "epoch": 0.51, "learning_rate": 4.9512808898283767e-05, "loss": 1.0424, "step": 201705 }, { "epoch": 0.51, "learning_rate": 4.951155065680206e-05, "loss": 1.0456, "step": 201710 }, { "epoch": 0.51, "learning_rate": 4.951029241532035e-05, "loss": 1.0616, "step": 201715 }, { "epoch": 0.51, "learning_rate": 4.950903417383865e-05, "loss": 1.0427, "step": 201720 }, { "epoch": 0.51, "learning_rate": 4.950777593235694e-05, "loss": 1.0433, "step": 201725 }, { "epoch": 0.51, "learning_rate": 4.950651769087524e-05, "loss": 1.0445, "step": 201730 }, { "epoch": 0.51, "learning_rate": 4.950525944939353e-05, "loss": 1.045, "step": 201735 }, { "epoch": 0.51, "learning_rate": 4.950400120791183e-05, "loss": 1.0448, "step": 201740 }, { "epoch": 0.51, "learning_rate": 4.950274296643012e-05, "loss": 1.0441, "step": 201745 }, { "epoch": 0.51, "learning_rate": 4.950148472494842e-05, "loss": 1.0417, "step": 201750 }, { "epoch": 0.51, "learning_rate": 4.950022648346671e-05, "loss": 1.0682, "step": 201755 }, { "epoch": 0.51, "learning_rate": 4.949896824198501e-05, "loss": 1.0456, "step": 201760 }, { "epoch": 0.51, "learning_rate": 4.94977100005033e-05, "loss": 1.0447, "step": 201765 }, { "epoch": 0.51, "learning_rate": 4.94964517590216e-05, "loss": 1.0422, "step": 201770 }, { "epoch": 0.51, "learning_rate": 4.949519351753989e-05, "loss": 1.0444, "step": 201775 }, { "epoch": 0.51, "learning_rate": 4.949393527605818e-05, "loss": 1.0453, "step": 201780 }, { "epoch": 0.51, "learning_rate": 4.949267703457648e-05, "loss": 1.0467, "step": 201785 }, { "epoch": 0.51, "learning_rate": 4.949141879309477e-05, "loss": 1.0422, "step": 201790 }, { "epoch": 0.51, "learning_rate": 4.949016055161307e-05, "loss": 1.0415, "step": 201795 }, { "epoch": 0.51, "learning_rate": 4.948890231013136e-05, "loss": 1.0443, "step": 201800 }, { "epoch": 0.51, "learning_rate": 4.948764406864966e-05, "loss": 1.0432, "step": 201805 }, { "epoch": 0.51, "learning_rate": 4.948638582716795e-05, "loss": 1.0437, "step": 201810 }, { "epoch": 0.51, "learning_rate": 4.948512758568625e-05, "loss": 1.0462, "step": 201815 }, { "epoch": 0.51, "learning_rate": 4.948386934420454e-05, "loss": 1.0445, "step": 201820 }, { "epoch": 0.51, "learning_rate": 4.948261110272284e-05, "loss": 1.0443, "step": 201825 }, { "epoch": 0.51, "learning_rate": 4.948135286124113e-05, "loss": 1.0448, "step": 201830 }, { "epoch": 0.51, "learning_rate": 4.948009461975943e-05, "loss": 1.0454, "step": 201835 }, { "epoch": 0.51, "learning_rate": 4.947883637827772e-05, "loss": 1.0433, "step": 201840 }, { "epoch": 0.51, "learning_rate": 4.947757813679601e-05, "loss": 1.0451, "step": 201845 }, { "epoch": 0.51, "learning_rate": 4.947631989531431e-05, "loss": 1.0458, "step": 201850 }, { "epoch": 0.51, "learning_rate": 4.94750616538326e-05, "loss": 1.0431, "step": 201855 }, { "epoch": 0.51, "learning_rate": 4.94738034123509e-05, "loss": 1.0458, "step": 201860 }, { "epoch": 0.51, "learning_rate": 4.947254517086919e-05, "loss": 1.0451, "step": 201865 }, { "epoch": 0.51, "learning_rate": 4.947128692938749e-05, "loss": 1.0619, "step": 201870 }, { "epoch": 0.51, "learning_rate": 4.947002868790578e-05, "loss": 1.0438, "step": 201875 }, { "epoch": 0.51, "learning_rate": 4.946877044642408e-05, "loss": 1.0443, "step": 201880 }, { "epoch": 0.51, "learning_rate": 4.946751220494237e-05, "loss": 1.0453, "step": 201885 }, { "epoch": 0.51, "learning_rate": 4.946625396346067e-05, "loss": 1.0467, "step": 201890 }, { "epoch": 0.51, "learning_rate": 4.946499572197896e-05, "loss": 1.0444, "step": 201895 }, { "epoch": 0.51, "learning_rate": 4.946373748049726e-05, "loss": 1.0437, "step": 201900 }, { "epoch": 0.51, "learning_rate": 4.946247923901555e-05, "loss": 1.046, "step": 201905 }, { "epoch": 0.51, "learning_rate": 4.9461220997533844e-05, "loss": 1.0455, "step": 201910 }, { "epoch": 0.51, "learning_rate": 4.945996275605214e-05, "loss": 1.0469, "step": 201915 }, { "epoch": 0.51, "learning_rate": 4.9458704514570434e-05, "loss": 1.0454, "step": 201920 }, { "epoch": 0.51, "learning_rate": 4.945744627308873e-05, "loss": 1.0439, "step": 201925 }, { "epoch": 0.51, "learning_rate": 4.9456188031607024e-05, "loss": 1.045, "step": 201930 }, { "epoch": 0.51, "learning_rate": 4.945492979012532e-05, "loss": 1.0464, "step": 201935 }, { "epoch": 0.51, "learning_rate": 4.9453671548643614e-05, "loss": 1.0465, "step": 201940 }, { "epoch": 0.51, "learning_rate": 4.945241330716191e-05, "loss": 1.0444, "step": 201945 }, { "epoch": 0.51, "learning_rate": 4.945115506568021e-05, "loss": 1.0446, "step": 201950 }, { "epoch": 0.51, "learning_rate": 4.94498968241985e-05, "loss": 1.0444, "step": 201955 }, { "epoch": 0.51, "learning_rate": 4.94486385827168e-05, "loss": 1.0423, "step": 201960 }, { "epoch": 0.51, "learning_rate": 4.944738034123509e-05, "loss": 1.0409, "step": 201965 }, { "epoch": 0.51, "learning_rate": 4.944612209975339e-05, "loss": 1.0436, "step": 201970 }, { "epoch": 0.51, "learning_rate": 4.944486385827168e-05, "loss": 1.0464, "step": 201975 }, { "epoch": 0.51, "learning_rate": 4.944360561678998e-05, "loss": 1.0664, "step": 201980 }, { "epoch": 0.51, "learning_rate": 4.944234737530827e-05, "loss": 1.0455, "step": 201985 }, { "epoch": 0.51, "learning_rate": 4.944108913382657e-05, "loss": 1.0415, "step": 201990 }, { "epoch": 0.51, "learning_rate": 4.943983089234486e-05, "loss": 1.0473, "step": 201995 }, { "epoch": 0.51, "learning_rate": 4.943857265086316e-05, "loss": 1.0426, "step": 202000 }, { "epoch": 0.51, "learning_rate": 4.943731440938145e-05, "loss": 1.0468, "step": 202005 }, { "epoch": 0.51, "learning_rate": 4.9436056167899744e-05, "loss": 1.0477, "step": 202010 }, { "epoch": 0.51, "learning_rate": 4.943479792641804e-05, "loss": 1.0424, "step": 202015 }, { "epoch": 0.51, "learning_rate": 4.9433539684936334e-05, "loss": 1.047, "step": 202020 }, { "epoch": 0.51, "learning_rate": 4.943228144345463e-05, "loss": 1.0438, "step": 202025 }, { "epoch": 0.51, "learning_rate": 4.9431023201972924e-05, "loss": 1.0444, "step": 202030 }, { "epoch": 0.51, "learning_rate": 4.942976496049122e-05, "loss": 1.0467, "step": 202035 }, { "epoch": 0.51, "learning_rate": 4.9428506719009514e-05, "loss": 1.0459, "step": 202040 }, { "epoch": 0.51, "learning_rate": 4.942724847752781e-05, "loss": 1.0773, "step": 202045 }, { "epoch": 0.51, "learning_rate": 4.9425990236046104e-05, "loss": 1.0435, "step": 202050 }, { "epoch": 0.51, "learning_rate": 4.94247319945644e-05, "loss": 1.0438, "step": 202055 }, { "epoch": 0.51, "learning_rate": 4.9423473753082694e-05, "loss": 1.0461, "step": 202060 }, { "epoch": 0.51, "learning_rate": 4.942221551160099e-05, "loss": 1.0457, "step": 202065 }, { "epoch": 0.51, "learning_rate": 4.9420957270119284e-05, "loss": 1.0448, "step": 202070 }, { "epoch": 0.51, "learning_rate": 4.9419699028637575e-05, "loss": 1.0465, "step": 202075 }, { "epoch": 0.51, "learning_rate": 4.9418440787155874e-05, "loss": 1.0422, "step": 202080 }, { "epoch": 0.51, "learning_rate": 4.9417182545674165e-05, "loss": 1.0451, "step": 202085 }, { "epoch": 0.51, "learning_rate": 4.9415924304192464e-05, "loss": 1.0484, "step": 202090 }, { "epoch": 0.51, "learning_rate": 4.9414666062710755e-05, "loss": 1.0464, "step": 202095 }, { "epoch": 0.51, "learning_rate": 4.9413407821229054e-05, "loss": 1.041, "step": 202100 }, { "epoch": 0.51, "learning_rate": 4.9412149579747345e-05, "loss": 1.0448, "step": 202105 }, { "epoch": 0.51, "learning_rate": 4.9410891338265643e-05, "loss": 1.0432, "step": 202110 }, { "epoch": 0.51, "learning_rate": 4.9409633096783935e-05, "loss": 1.0453, "step": 202115 }, { "epoch": 0.51, "learning_rate": 4.9408374855302233e-05, "loss": 1.0448, "step": 202120 }, { "epoch": 0.51, "learning_rate": 4.9407116613820525e-05, "loss": 1.0465, "step": 202125 }, { "epoch": 0.51, "learning_rate": 4.9405858372338823e-05, "loss": 1.0418, "step": 202130 }, { "epoch": 0.51, "learning_rate": 4.9404600130857115e-05, "loss": 1.0441, "step": 202135 }, { "epoch": 0.51, "learning_rate": 4.940334188937541e-05, "loss": 1.0472, "step": 202140 }, { "epoch": 0.51, "learning_rate": 4.9402083647893705e-05, "loss": 1.0459, "step": 202145 }, { "epoch": 0.51, "learning_rate": 4.9400825406411997e-05, "loss": 1.0461, "step": 202150 }, { "epoch": 0.51, "learning_rate": 4.9399567164930295e-05, "loss": 1.0442, "step": 202155 }, { "epoch": 0.51, "learning_rate": 4.9398308923448587e-05, "loss": 1.0458, "step": 202160 }, { "epoch": 0.51, "learning_rate": 4.9397050681966885e-05, "loss": 1.0422, "step": 202165 }, { "epoch": 0.51, "learning_rate": 4.9395792440485177e-05, "loss": 1.0603, "step": 202170 }, { "epoch": 0.51, "learning_rate": 4.9394534199003475e-05, "loss": 1.0454, "step": 202175 }, { "epoch": 0.51, "learning_rate": 4.9393275957521767e-05, "loss": 1.0463, "step": 202180 }, { "epoch": 0.51, "learning_rate": 4.9392017716040065e-05, "loss": 1.0456, "step": 202185 }, { "epoch": 0.51, "learning_rate": 4.9390759474558356e-05, "loss": 1.0488, "step": 202190 }, { "epoch": 0.51, "learning_rate": 4.9389501233076655e-05, "loss": 1.0413, "step": 202195 }, { "epoch": 0.51, "learning_rate": 4.938824299159495e-05, "loss": 1.0458, "step": 202200 }, { "epoch": 0.51, "learning_rate": 4.9386984750113245e-05, "loss": 1.0417, "step": 202205 }, { "epoch": 0.51, "learning_rate": 4.938572650863154e-05, "loss": 1.0656, "step": 202210 }, { "epoch": 0.51, "learning_rate": 4.9384468267149835e-05, "loss": 1.0461, "step": 202215 }, { "epoch": 0.51, "learning_rate": 4.938321002566813e-05, "loss": 1.0494, "step": 202220 }, { "epoch": 0.51, "learning_rate": 4.9381951784186425e-05, "loss": 1.0425, "step": 202225 }, { "epoch": 0.51, "learning_rate": 4.938069354270472e-05, "loss": 1.0445, "step": 202230 }, { "epoch": 0.51, "learning_rate": 4.9379435301223015e-05, "loss": 1.0459, "step": 202235 }, { "epoch": 0.51, "learning_rate": 4.9378177059741306e-05, "loss": 1.0431, "step": 202240 }, { "epoch": 0.51, "learning_rate": 4.9376918818259605e-05, "loss": 1.0459, "step": 202245 }, { "epoch": 0.51, "learning_rate": 4.9375660576777896e-05, "loss": 1.0443, "step": 202250 }, { "epoch": 0.51, "learning_rate": 4.9374402335296195e-05, "loss": 1.043, "step": 202255 }, { "epoch": 0.51, "learning_rate": 4.9373144093814486e-05, "loss": 1.0463, "step": 202260 }, { "epoch": 0.51, "learning_rate": 4.9371885852332785e-05, "loss": 1.0468, "step": 202265 }, { "epoch": 0.51, "learning_rate": 4.9370627610851076e-05, "loss": 1.0436, "step": 202270 }, { "epoch": 0.51, "learning_rate": 4.9369369369369375e-05, "loss": 1.0453, "step": 202275 }, { "epoch": 0.51, "learning_rate": 4.9368111127887666e-05, "loss": 1.0545, "step": 202280 }, { "epoch": 0.51, "learning_rate": 4.9366852886405965e-05, "loss": 1.0441, "step": 202285 }, { "epoch": 0.51, "learning_rate": 4.9365594644924256e-05, "loss": 1.0463, "step": 202290 }, { "epoch": 0.51, "learning_rate": 4.9364336403442555e-05, "loss": 1.0435, "step": 202295 }, { "epoch": 0.51, "learning_rate": 4.9363078161960846e-05, "loss": 1.0464, "step": 202300 }, { "epoch": 0.51, "learning_rate": 4.936181992047914e-05, "loss": 1.0433, "step": 202305 }, { "epoch": 0.51, "learning_rate": 4.9360561678997436e-05, "loss": 1.0464, "step": 202310 }, { "epoch": 0.51, "learning_rate": 4.935930343751573e-05, "loss": 1.0453, "step": 202315 }, { "epoch": 0.51, "learning_rate": 4.9358045196034026e-05, "loss": 1.0477, "step": 202320 }, { "epoch": 0.51, "learning_rate": 4.935678695455232e-05, "loss": 1.0457, "step": 202325 }, { "epoch": 0.51, "learning_rate": 4.9355528713070616e-05, "loss": 1.0465, "step": 202330 }, { "epoch": 0.51, "learning_rate": 4.935427047158891e-05, "loss": 1.046, "step": 202335 }, { "epoch": 0.51, "learning_rate": 4.9353012230107206e-05, "loss": 1.0437, "step": 202340 }, { "epoch": 0.51, "learning_rate": 4.93517539886255e-05, "loss": 1.0468, "step": 202345 }, { "epoch": 0.51, "learning_rate": 4.9350495747143796e-05, "loss": 1.0447, "step": 202350 }, { "epoch": 0.51, "learning_rate": 4.934923750566209e-05, "loss": 1.0461, "step": 202355 }, { "epoch": 0.51, "learning_rate": 4.9347979264180386e-05, "loss": 1.0453, "step": 202360 }, { "epoch": 0.51, "learning_rate": 4.934672102269868e-05, "loss": 1.0475, "step": 202365 }, { "epoch": 0.51, "learning_rate": 4.934546278121697e-05, "loss": 1.0438, "step": 202370 }, { "epoch": 0.51, "learning_rate": 4.934420453973527e-05, "loss": 1.0446, "step": 202375 }, { "epoch": 0.51, "learning_rate": 4.934294629825356e-05, "loss": 1.0452, "step": 202380 }, { "epoch": 0.51, "learning_rate": 4.934168805677186e-05, "loss": 1.0456, "step": 202385 }, { "epoch": 0.51, "learning_rate": 4.934042981529015e-05, "loss": 1.0457, "step": 202390 }, { "epoch": 0.51, "learning_rate": 4.933917157380845e-05, "loss": 1.0474, "step": 202395 }, { "epoch": 0.51, "learning_rate": 4.933791333232674e-05, "loss": 1.0411, "step": 202400 }, { "epoch": 0.51, "learning_rate": 4.933665509084504e-05, "loss": 1.0461, "step": 202405 }, { "epoch": 0.51, "learning_rate": 4.933539684936333e-05, "loss": 1.0441, "step": 202410 }, { "epoch": 0.51, "learning_rate": 4.933413860788163e-05, "loss": 1.0607, "step": 202415 }, { "epoch": 0.51, "learning_rate": 4.933288036639992e-05, "loss": 1.0431, "step": 202420 }, { "epoch": 0.51, "learning_rate": 4.933162212491822e-05, "loss": 1.0457, "step": 202425 }, { "epoch": 0.51, "learning_rate": 4.933036388343651e-05, "loss": 1.0424, "step": 202430 }, { "epoch": 0.51, "learning_rate": 4.93291056419548e-05, "loss": 1.0471, "step": 202435 }, { "epoch": 0.51, "learning_rate": 4.9327847400473106e-05, "loss": 1.0493, "step": 202440 }, { "epoch": 0.51, "learning_rate": 4.93265891589914e-05, "loss": 1.043, "step": 202445 }, { "epoch": 0.51, "learning_rate": 4.9325330917509696e-05, "loss": 1.0698, "step": 202450 }, { "epoch": 0.51, "learning_rate": 4.932407267602799e-05, "loss": 1.0418, "step": 202455 }, { "epoch": 0.51, "learning_rate": 4.9322814434546286e-05, "loss": 1.0436, "step": 202460 }, { "epoch": 0.51, "learning_rate": 4.932155619306458e-05, "loss": 1.0421, "step": 202465 }, { "epoch": 0.51, "learning_rate": 4.932029795158287e-05, "loss": 1.0453, "step": 202470 }, { "epoch": 0.51, "learning_rate": 4.931903971010117e-05, "loss": 1.0412, "step": 202475 }, { "epoch": 0.51, "learning_rate": 4.931778146861946e-05, "loss": 1.0468, "step": 202480 }, { "epoch": 0.51, "learning_rate": 4.931652322713776e-05, "loss": 1.0441, "step": 202485 }, { "epoch": 0.51, "learning_rate": 4.931526498565605e-05, "loss": 1.0428, "step": 202490 }, { "epoch": 0.51, "learning_rate": 4.931400674417435e-05, "loss": 1.0422, "step": 202495 }, { "epoch": 0.51, "learning_rate": 4.931274850269264e-05, "loss": 1.0515, "step": 202500 }, { "epoch": 0.51, "learning_rate": 4.931149026121094e-05, "loss": 1.0477, "step": 202505 }, { "epoch": 0.51, "learning_rate": 4.931023201972923e-05, "loss": 1.0444, "step": 202510 }, { "epoch": 0.51, "learning_rate": 4.930897377824753e-05, "loss": 1.0437, "step": 202515 }, { "epoch": 0.51, "learning_rate": 4.930771553676582e-05, "loss": 1.0444, "step": 202520 }, { "epoch": 0.51, "learning_rate": 4.930645729528412e-05, "loss": 1.0639, "step": 202525 }, { "epoch": 0.51, "learning_rate": 4.930519905380241e-05, "loss": 1.0437, "step": 202530 }, { "epoch": 0.51, "learning_rate": 4.93039408123207e-05, "loss": 1.0467, "step": 202535 }, { "epoch": 0.51, "learning_rate": 4.9302682570839e-05, "loss": 1.0432, "step": 202540 }, { "epoch": 0.51, "learning_rate": 4.930142432935729e-05, "loss": 1.0432, "step": 202545 }, { "epoch": 0.51, "learning_rate": 4.930016608787559e-05, "loss": 1.0443, "step": 202550 }, { "epoch": 0.51, "learning_rate": 4.929890784639388e-05, "loss": 1.0436, "step": 202555 }, { "epoch": 0.51, "learning_rate": 4.929764960491218e-05, "loss": 1.0447, "step": 202560 }, { "epoch": 0.51, "learning_rate": 4.929639136343047e-05, "loss": 1.0476, "step": 202565 }, { "epoch": 0.51, "learning_rate": 4.929513312194877e-05, "loss": 1.0443, "step": 202570 }, { "epoch": 0.51, "learning_rate": 4.929387488046706e-05, "loss": 1.0446, "step": 202575 }, { "epoch": 0.51, "learning_rate": 4.929261663898536e-05, "loss": 1.0431, "step": 202580 }, { "epoch": 0.51, "learning_rate": 4.929135839750365e-05, "loss": 1.0433, "step": 202585 }, { "epoch": 0.51, "learning_rate": 4.929010015602195e-05, "loss": 1.0432, "step": 202590 }, { "epoch": 0.51, "learning_rate": 4.928884191454024e-05, "loss": 1.0467, "step": 202595 }, { "epoch": 0.51, "learning_rate": 4.928758367305853e-05, "loss": 1.0483, "step": 202600 }, { "epoch": 0.51, "learning_rate": 4.928632543157683e-05, "loss": 1.0423, "step": 202605 }, { "epoch": 0.51, "learning_rate": 4.928506719009512e-05, "loss": 1.046, "step": 202610 }, { "epoch": 0.51, "learning_rate": 4.928380894861342e-05, "loss": 1.0452, "step": 202615 }, { "epoch": 0.51, "learning_rate": 4.928255070713171e-05, "loss": 1.0433, "step": 202620 }, { "epoch": 0.51, "learning_rate": 4.928129246565001e-05, "loss": 1.0462, "step": 202625 }, { "epoch": 0.51, "learning_rate": 4.92800342241683e-05, "loss": 1.0451, "step": 202630 }, { "epoch": 0.51, "learning_rate": 4.92787759826866e-05, "loss": 1.0667, "step": 202635 }, { "epoch": 0.51, "learning_rate": 4.927751774120489e-05, "loss": 1.0416, "step": 202640 }, { "epoch": 0.51, "learning_rate": 4.927625949972319e-05, "loss": 1.0477, "step": 202645 }, { "epoch": 0.51, "learning_rate": 4.927500125824148e-05, "loss": 1.047, "step": 202650 }, { "epoch": 0.51, "learning_rate": 4.927374301675978e-05, "loss": 1.0436, "step": 202655 }, { "epoch": 0.51, "learning_rate": 4.927248477527807e-05, "loss": 1.0425, "step": 202660 }, { "epoch": 0.51, "learning_rate": 4.927122653379636e-05, "loss": 1.0464, "step": 202665 }, { "epoch": 0.51, "learning_rate": 4.926996829231466e-05, "loss": 1.0437, "step": 202670 }, { "epoch": 0.51, "learning_rate": 4.926871005083295e-05, "loss": 1.047, "step": 202675 }, { "epoch": 0.51, "learning_rate": 4.926745180935125e-05, "loss": 1.0718, "step": 202680 }, { "epoch": 0.51, "learning_rate": 4.926619356786954e-05, "loss": 1.0465, "step": 202685 }, { "epoch": 0.51, "learning_rate": 4.926493532638785e-05, "loss": 1.046, "step": 202690 }, { "epoch": 0.51, "learning_rate": 4.926367708490614e-05, "loss": 1.0453, "step": 202695 }, { "epoch": 0.51, "learning_rate": 4.926241884342444e-05, "loss": 1.0452, "step": 202700 }, { "epoch": 0.51, "learning_rate": 4.926116060194273e-05, "loss": 1.044, "step": 202705 }, { "epoch": 0.51, "learning_rate": 4.925990236046102e-05, "loss": 1.0451, "step": 202710 }, { "epoch": 0.51, "learning_rate": 4.925864411897932e-05, "loss": 1.0416, "step": 202715 }, { "epoch": 0.51, "learning_rate": 4.925738587749761e-05, "loss": 1.067, "step": 202720 }, { "epoch": 0.51, "learning_rate": 4.925612763601591e-05, "loss": 1.047, "step": 202725 }, { "epoch": 0.51, "learning_rate": 4.92548693945342e-05, "loss": 1.0436, "step": 202730 }, { "epoch": 0.51, "learning_rate": 4.92536111530525e-05, "loss": 1.047, "step": 202735 }, { "epoch": 0.51, "learning_rate": 4.925235291157079e-05, "loss": 1.0448, "step": 202740 }, { "epoch": 0.51, "learning_rate": 4.925109467008909e-05, "loss": 1.0465, "step": 202745 }, { "epoch": 0.51, "learning_rate": 4.924983642860738e-05, "loss": 1.0603, "step": 202750 }, { "epoch": 0.51, "learning_rate": 4.924857818712568e-05, "loss": 1.0431, "step": 202755 }, { "epoch": 0.51, "learning_rate": 4.924731994564397e-05, "loss": 1.0419, "step": 202760 }, { "epoch": 0.51, "learning_rate": 4.924606170416227e-05, "loss": 1.0417, "step": 202765 }, { "epoch": 0.51, "learning_rate": 4.924480346268056e-05, "loss": 1.0446, "step": 202770 }, { "epoch": 0.51, "learning_rate": 4.924354522119885e-05, "loss": 1.0435, "step": 202775 }, { "epoch": 0.51, "learning_rate": 4.924228697971715e-05, "loss": 1.0451, "step": 202780 }, { "epoch": 0.51, "learning_rate": 4.924102873823544e-05, "loss": 1.0444, "step": 202785 }, { "epoch": 0.51, "learning_rate": 4.923977049675374e-05, "loss": 1.0449, "step": 202790 }, { "epoch": 0.51, "learning_rate": 4.923851225527203e-05, "loss": 1.0426, "step": 202795 }, { "epoch": 0.51, "learning_rate": 4.923725401379033e-05, "loss": 1.044, "step": 202800 }, { "epoch": 0.51, "learning_rate": 4.923599577230862e-05, "loss": 1.0444, "step": 202805 }, { "epoch": 0.51, "learning_rate": 4.923473753082692e-05, "loss": 1.043, "step": 202810 }, { "epoch": 0.51, "learning_rate": 4.923347928934521e-05, "loss": 1.0419, "step": 202815 }, { "epoch": 0.51, "learning_rate": 4.923222104786351e-05, "loss": 1.043, "step": 202820 }, { "epoch": 0.51, "learning_rate": 4.92309628063818e-05, "loss": 1.0442, "step": 202825 }, { "epoch": 0.51, "learning_rate": 4.9229704564900094e-05, "loss": 1.0448, "step": 202830 }, { "epoch": 0.51, "learning_rate": 4.922844632341839e-05, "loss": 1.0509, "step": 202835 }, { "epoch": 0.51, "learning_rate": 4.9227188081936684e-05, "loss": 1.0424, "step": 202840 }, { "epoch": 0.51, "learning_rate": 4.922592984045498e-05, "loss": 1.0429, "step": 202845 }, { "epoch": 0.51, "learning_rate": 4.9224671598973274e-05, "loss": 1.0426, "step": 202850 }, { "epoch": 0.51, "learning_rate": 4.922341335749157e-05, "loss": 1.0432, "step": 202855 }, { "epoch": 0.51, "learning_rate": 4.9222155116009864e-05, "loss": 1.0427, "step": 202860 }, { "epoch": 0.51, "learning_rate": 4.922089687452816e-05, "loss": 1.0458, "step": 202865 }, { "epoch": 0.51, "learning_rate": 4.9219638633046454e-05, "loss": 1.043, "step": 202870 }, { "epoch": 0.51, "learning_rate": 4.921838039156475e-05, "loss": 1.0448, "step": 202875 }, { "epoch": 0.51, "learning_rate": 4.9217122150083044e-05, "loss": 1.044, "step": 202880 }, { "epoch": 0.51, "learning_rate": 4.921586390860134e-05, "loss": 1.0481, "step": 202885 }, { "epoch": 0.51, "learning_rate": 4.9214605667119634e-05, "loss": 1.045, "step": 202890 }, { "epoch": 0.51, "learning_rate": 4.9213347425637926e-05, "loss": 1.0461, "step": 202895 }, { "epoch": 0.51, "learning_rate": 4.9212089184156224e-05, "loss": 1.0433, "step": 202900 }, { "epoch": 0.51, "learning_rate": 4.9210830942674516e-05, "loss": 1.0437, "step": 202905 }, { "epoch": 0.51, "learning_rate": 4.9209572701192814e-05, "loss": 1.0412, "step": 202910 }, { "epoch": 0.51, "learning_rate": 4.9208314459711106e-05, "loss": 1.0463, "step": 202915 }, { "epoch": 0.51, "learning_rate": 4.9207056218229404e-05, "loss": 1.0486, "step": 202920 }, { "epoch": 0.51, "learning_rate": 4.9205797976747696e-05, "loss": 1.0476, "step": 202925 }, { "epoch": 0.51, "learning_rate": 4.9204539735265994e-05, "loss": 1.0443, "step": 202930 }, { "epoch": 0.51, "learning_rate": 4.9203281493784286e-05, "loss": 1.0421, "step": 202935 }, { "epoch": 0.51, "learning_rate": 4.9202023252302584e-05, "loss": 1.0447, "step": 202940 }, { "epoch": 0.51, "learning_rate": 4.920076501082088e-05, "loss": 1.0439, "step": 202945 }, { "epoch": 0.51, "learning_rate": 4.9199506769339174e-05, "loss": 1.0443, "step": 202950 }, { "epoch": 0.51, "learning_rate": 4.919824852785747e-05, "loss": 1.0475, "step": 202955 }, { "epoch": 0.51, "learning_rate": 4.9196990286375764e-05, "loss": 1.0436, "step": 202960 }, { "epoch": 0.51, "learning_rate": 4.919573204489406e-05, "loss": 1.0455, "step": 202965 }, { "epoch": 0.51, "learning_rate": 4.9194473803412354e-05, "loss": 1.045, "step": 202970 }, { "epoch": 0.51, "learning_rate": 4.919321556193065e-05, "loss": 1.0453, "step": 202975 }, { "epoch": 0.51, "learning_rate": 4.9191957320448944e-05, "loss": 1.0447, "step": 202980 }, { "epoch": 0.51, "learning_rate": 4.919069907896724e-05, "loss": 1.0474, "step": 202985 }, { "epoch": 0.51, "learning_rate": 4.9189440837485534e-05, "loss": 1.0447, "step": 202990 }, { "epoch": 0.51, "learning_rate": 4.918818259600383e-05, "loss": 1.044, "step": 202995 }, { "epoch": 0.51, "learning_rate": 4.9186924354522124e-05, "loss": 1.0441, "step": 203000 }, { "epoch": 0.51, "learning_rate": 4.9185666113040415e-05, "loss": 1.0441, "step": 203005 }, { "epoch": 0.51, "learning_rate": 4.9184407871558714e-05, "loss": 1.0427, "step": 203010 }, { "epoch": 0.51, "learning_rate": 4.9183149630077005e-05, "loss": 1.0467, "step": 203015 }, { "epoch": 0.51, "learning_rate": 4.9181891388595304e-05, "loss": 1.046, "step": 203020 }, { "epoch": 0.51, "learning_rate": 4.9180633147113595e-05, "loss": 1.0443, "step": 203025 }, { "epoch": 0.51, "learning_rate": 4.9179374905631894e-05, "loss": 1.0718, "step": 203030 }, { "epoch": 0.51, "learning_rate": 4.9178116664150185e-05, "loss": 1.043, "step": 203035 }, { "epoch": 0.51, "learning_rate": 4.9176858422668484e-05, "loss": 1.0464, "step": 203040 }, { "epoch": 0.51, "learning_rate": 4.9175600181186775e-05, "loss": 1.0443, "step": 203045 }, { "epoch": 0.51, "learning_rate": 4.9174341939705074e-05, "loss": 1.0424, "step": 203050 }, { "epoch": 0.51, "learning_rate": 4.9173083698223365e-05, "loss": 1.0437, "step": 203055 }, { "epoch": 0.51, "learning_rate": 4.9171825456741664e-05, "loss": 1.0463, "step": 203060 }, { "epoch": 0.51, "learning_rate": 4.9170567215259955e-05, "loss": 1.0468, "step": 203065 }, { "epoch": 0.51, "learning_rate": 4.916930897377825e-05, "loss": 1.0418, "step": 203070 }, { "epoch": 0.51, "learning_rate": 4.9168050732296545e-05, "loss": 1.0659, "step": 203075 }, { "epoch": 0.51, "learning_rate": 4.916679249081484e-05, "loss": 1.0455, "step": 203080 }, { "epoch": 0.51, "learning_rate": 4.9165534249333135e-05, "loss": 1.0427, "step": 203085 }, { "epoch": 0.51, "learning_rate": 4.916427600785143e-05, "loss": 1.0466, "step": 203090 }, { "epoch": 0.51, "learning_rate": 4.9163017766369725e-05, "loss": 1.0435, "step": 203095 }, { "epoch": 0.51, "learning_rate": 4.916175952488802e-05, "loss": 1.0422, "step": 203100 }, { "epoch": 0.51, "learning_rate": 4.9160501283406315e-05, "loss": 1.0457, "step": 203105 }, { "epoch": 0.51, "learning_rate": 4.915924304192461e-05, "loss": 1.044, "step": 203110 }, { "epoch": 0.51, "learning_rate": 4.9157984800442905e-05, "loss": 1.0464, "step": 203115 }, { "epoch": 0.51, "learning_rate": 4.91567265589612e-05, "loss": 1.0451, "step": 203120 }, { "epoch": 0.51, "learning_rate": 4.915546831747949e-05, "loss": 1.0427, "step": 203125 }, { "epoch": 0.51, "learning_rate": 4.915421007599779e-05, "loss": 1.0471, "step": 203130 }, { "epoch": 0.51, "learning_rate": 4.915295183451608e-05, "loss": 1.0488, "step": 203135 }, { "epoch": 0.51, "learning_rate": 4.915169359303438e-05, "loss": 1.0451, "step": 203140 }, { "epoch": 0.51, "learning_rate": 4.915043535155267e-05, "loss": 1.0445, "step": 203145 }, { "epoch": 0.51, "learning_rate": 4.9149177110070967e-05, "loss": 1.0461, "step": 203150 }, { "epoch": 0.51, "learning_rate": 4.914791886858926e-05, "loss": 1.0421, "step": 203155 }, { "epoch": 0.51, "learning_rate": 4.9146660627107557e-05, "loss": 1.0463, "step": 203160 }, { "epoch": 0.51, "learning_rate": 4.914540238562585e-05, "loss": 1.0451, "step": 203165 }, { "epoch": 0.51, "learning_rate": 4.9144144144144147e-05, "loss": 1.0441, "step": 203170 }, { "epoch": 0.51, "learning_rate": 4.914288590266244e-05, "loss": 1.046, "step": 203175 }, { "epoch": 0.51, "learning_rate": 4.9141627661180737e-05, "loss": 1.0446, "step": 203180 }, { "epoch": 0.51, "learning_rate": 4.914036941969903e-05, "loss": 1.0449, "step": 203185 }, { "epoch": 0.51, "learning_rate": 4.9139111178217326e-05, "loss": 1.0459, "step": 203190 }, { "epoch": 0.51, "learning_rate": 4.9137852936735625e-05, "loss": 1.041, "step": 203195 }, { "epoch": 0.51, "learning_rate": 4.9136594695253916e-05, "loss": 1.0441, "step": 203200 }, { "epoch": 0.51, "learning_rate": 4.9135336453772215e-05, "loss": 1.0439, "step": 203205 }, { "epoch": 0.51, "learning_rate": 4.9134078212290506e-05, "loss": 1.0472, "step": 203210 }, { "epoch": 0.51, "learning_rate": 4.9132819970808805e-05, "loss": 1.0485, "step": 203215 }, { "epoch": 0.51, "learning_rate": 4.9131561729327096e-05, "loss": 1.045, "step": 203220 }, { "epoch": 0.51, "learning_rate": 4.9130303487845395e-05, "loss": 1.0476, "step": 203225 }, { "epoch": 0.51, "learning_rate": 4.9129045246363686e-05, "loss": 1.0438, "step": 203230 }, { "epoch": 0.51, "learning_rate": 4.912778700488198e-05, "loss": 1.0451, "step": 203235 }, { "epoch": 0.51, "learning_rate": 4.9126528763400276e-05, "loss": 1.0575, "step": 203240 }, { "epoch": 0.51, "learning_rate": 4.912527052191857e-05, "loss": 1.0899, "step": 203245 }, { "epoch": 0.51, "learning_rate": 4.9124012280436866e-05, "loss": 1.0438, "step": 203250 }, { "epoch": 0.51, "learning_rate": 4.912275403895516e-05, "loss": 1.0445, "step": 203255 }, { "epoch": 0.51, "learning_rate": 4.9121495797473456e-05, "loss": 1.044, "step": 203260 }, { "epoch": 0.51, "learning_rate": 4.912023755599175e-05, "loss": 1.0432, "step": 203265 }, { "epoch": 0.51, "learning_rate": 4.9118979314510046e-05, "loss": 1.0483, "step": 203270 }, { "epoch": 0.51, "learning_rate": 4.911772107302834e-05, "loss": 1.0432, "step": 203275 }, { "epoch": 0.51, "learning_rate": 4.9116462831546636e-05, "loss": 1.0461, "step": 203280 }, { "epoch": 0.51, "learning_rate": 4.911520459006493e-05, "loss": 1.0452, "step": 203285 }, { "epoch": 0.51, "learning_rate": 4.9113946348583226e-05, "loss": 1.0432, "step": 203290 }, { "epoch": 0.51, "learning_rate": 4.911268810710152e-05, "loss": 1.0438, "step": 203295 }, { "epoch": 0.51, "learning_rate": 4.911142986561981e-05, "loss": 1.0436, "step": 203300 }, { "epoch": 0.51, "learning_rate": 4.911017162413811e-05, "loss": 1.0463, "step": 203305 }, { "epoch": 0.51, "learning_rate": 4.91089133826564e-05, "loss": 1.0418, "step": 203310 }, { "epoch": 0.51, "learning_rate": 4.91076551411747e-05, "loss": 1.0419, "step": 203315 }, { "epoch": 0.51, "learning_rate": 4.910639689969299e-05, "loss": 1.0452, "step": 203320 }, { "epoch": 0.51, "learning_rate": 4.910513865821129e-05, "loss": 1.0415, "step": 203325 }, { "epoch": 0.51, "learning_rate": 4.910388041672958e-05, "loss": 1.0436, "step": 203330 }, { "epoch": 0.51, "learning_rate": 4.910262217524788e-05, "loss": 1.0432, "step": 203335 }, { "epoch": 0.51, "learning_rate": 4.910136393376617e-05, "loss": 1.045, "step": 203340 }, { "epoch": 0.51, "learning_rate": 4.910010569228447e-05, "loss": 1.048, "step": 203345 }, { "epoch": 0.51, "learning_rate": 4.909884745080276e-05, "loss": 1.0432, "step": 203350 }, { "epoch": 0.51, "learning_rate": 4.909758920932106e-05, "loss": 1.0449, "step": 203355 }, { "epoch": 0.51, "learning_rate": 4.909633096783935e-05, "loss": 1.0433, "step": 203360 }, { "epoch": 0.51, "learning_rate": 4.909507272635764e-05, "loss": 1.0648, "step": 203365 }, { "epoch": 0.51, "learning_rate": 4.909381448487594e-05, "loss": 1.046, "step": 203370 }, { "epoch": 0.51, "learning_rate": 4.909255624339423e-05, "loss": 1.0467, "step": 203375 }, { "epoch": 0.51, "learning_rate": 4.909129800191253e-05, "loss": 1.0455, "step": 203380 }, { "epoch": 0.51, "learning_rate": 4.909003976043082e-05, "loss": 1.0438, "step": 203385 }, { "epoch": 0.51, "learning_rate": 4.908878151894912e-05, "loss": 1.0469, "step": 203390 }, { "epoch": 0.51, "learning_rate": 4.908752327746741e-05, "loss": 1.042, "step": 203395 }, { "epoch": 0.51, "learning_rate": 4.908626503598571e-05, "loss": 1.046, "step": 203400 }, { "epoch": 0.51, "learning_rate": 4.9085006794504e-05, "loss": 1.0453, "step": 203405 }, { "epoch": 0.51, "learning_rate": 4.90837485530223e-05, "loss": 1.0465, "step": 203410 }, { "epoch": 0.51, "learning_rate": 4.908249031154059e-05, "loss": 1.0432, "step": 203415 }, { "epoch": 0.51, "learning_rate": 4.908123207005889e-05, "loss": 1.0453, "step": 203420 }, { "epoch": 0.51, "learning_rate": 4.907997382857718e-05, "loss": 1.0442, "step": 203425 }, { "epoch": 0.51, "learning_rate": 4.907871558709547e-05, "loss": 1.0464, "step": 203430 }, { "epoch": 0.51, "learning_rate": 4.907745734561377e-05, "loss": 1.0412, "step": 203435 }, { "epoch": 0.51, "learning_rate": 4.907619910413207e-05, "loss": 1.0419, "step": 203440 }, { "epoch": 0.51, "learning_rate": 4.907494086265037e-05, "loss": 1.0493, "step": 203445 }, { "epoch": 0.51, "learning_rate": 4.907368262116866e-05, "loss": 1.0455, "step": 203450 }, { "epoch": 0.51, "learning_rate": 4.907242437968696e-05, "loss": 1.0452, "step": 203455 }, { "epoch": 0.51, "learning_rate": 4.907116613820525e-05, "loss": 1.0454, "step": 203460 }, { "epoch": 0.51, "learning_rate": 4.906990789672354e-05, "loss": 1.0461, "step": 203465 }, { "epoch": 0.51, "learning_rate": 4.906864965524184e-05, "loss": 1.0571, "step": 203470 }, { "epoch": 0.51, "learning_rate": 4.906739141376013e-05, "loss": 1.0443, "step": 203475 }, { "epoch": 0.51, "learning_rate": 4.906613317227843e-05, "loss": 1.0443, "step": 203480 }, { "epoch": 0.51, "learning_rate": 4.906487493079672e-05, "loss": 1.044, "step": 203485 }, { "epoch": 0.51, "learning_rate": 4.906361668931502e-05, "loss": 1.047, "step": 203490 }, { "epoch": 0.51, "learning_rate": 4.906235844783331e-05, "loss": 1.0463, "step": 203495 }, { "epoch": 0.51, "learning_rate": 4.906110020635161e-05, "loss": 1.0458, "step": 203500 }, { "epoch": 0.51, "learning_rate": 4.90598419648699e-05, "loss": 1.0457, "step": 203505 }, { "epoch": 0.51, "learning_rate": 4.90585837233882e-05, "loss": 1.0438, "step": 203510 }, { "epoch": 0.51, "learning_rate": 4.905732548190649e-05, "loss": 1.0466, "step": 203515 }, { "epoch": 0.51, "learning_rate": 4.905606724042479e-05, "loss": 1.0441, "step": 203520 }, { "epoch": 0.51, "learning_rate": 4.905480899894308e-05, "loss": 1.0443, "step": 203525 }, { "epoch": 0.51, "learning_rate": 4.905355075746137e-05, "loss": 1.046, "step": 203530 }, { "epoch": 0.51, "learning_rate": 4.905229251597967e-05, "loss": 1.045, "step": 203535 }, { "epoch": 0.51, "learning_rate": 4.905103427449796e-05, "loss": 1.0466, "step": 203540 }, { "epoch": 0.51, "learning_rate": 4.904977603301626e-05, "loss": 1.0442, "step": 203545 }, { "epoch": 0.51, "learning_rate": 4.904851779153455e-05, "loss": 1.0446, "step": 203550 }, { "epoch": 0.51, "learning_rate": 4.904725955005285e-05, "loss": 1.0451, "step": 203555 }, { "epoch": 0.51, "learning_rate": 4.904600130857114e-05, "loss": 1.0418, "step": 203560 }, { "epoch": 0.51, "learning_rate": 4.904474306708944e-05, "loss": 1.0469, "step": 203565 }, { "epoch": 0.51, "learning_rate": 4.904348482560773e-05, "loss": 1.0411, "step": 203570 }, { "epoch": 0.51, "learning_rate": 4.904222658412603e-05, "loss": 1.0432, "step": 203575 }, { "epoch": 0.51, "learning_rate": 4.904096834264432e-05, "loss": 1.0439, "step": 203580 }, { "epoch": 0.51, "learning_rate": 4.903971010116262e-05, "loss": 1.0477, "step": 203585 }, { "epoch": 0.51, "learning_rate": 4.903845185968091e-05, "loss": 1.0433, "step": 203590 }, { "epoch": 0.51, "learning_rate": 4.90371936181992e-05, "loss": 1.0397, "step": 203595 }, { "epoch": 0.51, "learning_rate": 4.90359353767175e-05, "loss": 1.0439, "step": 203600 }, { "epoch": 0.51, "learning_rate": 4.903467713523579e-05, "loss": 1.066, "step": 203605 }, { "epoch": 0.51, "learning_rate": 4.903341889375409e-05, "loss": 1.0619, "step": 203610 }, { "epoch": 0.51, "learning_rate": 4.903216065227238e-05, "loss": 1.0434, "step": 203615 }, { "epoch": 0.51, "learning_rate": 4.903090241079068e-05, "loss": 1.0433, "step": 203620 }, { "epoch": 0.51, "learning_rate": 4.902964416930897e-05, "loss": 1.0425, "step": 203625 }, { "epoch": 0.51, "learning_rate": 4.902838592782727e-05, "loss": 1.0458, "step": 203630 }, { "epoch": 0.51, "learning_rate": 4.902712768634556e-05, "loss": 1.0444, "step": 203635 }, { "epoch": 0.51, "learning_rate": 4.902586944486386e-05, "loss": 1.0443, "step": 203640 }, { "epoch": 0.51, "learning_rate": 4.902461120338215e-05, "loss": 1.0441, "step": 203645 }, { "epoch": 0.51, "learning_rate": 4.902335296190045e-05, "loss": 1.0435, "step": 203650 }, { "epoch": 0.51, "learning_rate": 4.902209472041874e-05, "loss": 1.0434, "step": 203655 }, { "epoch": 0.51, "learning_rate": 4.9020836478937035e-05, "loss": 1.0434, "step": 203660 }, { "epoch": 0.51, "learning_rate": 4.901957823745533e-05, "loss": 1.07, "step": 203665 }, { "epoch": 0.51, "learning_rate": 4.9018319995973625e-05, "loss": 1.0439, "step": 203670 }, { "epoch": 0.51, "learning_rate": 4.901706175449192e-05, "loss": 1.0707, "step": 203675 }, { "epoch": 0.51, "learning_rate": 4.9015803513010215e-05, "loss": 1.0451, "step": 203680 }, { "epoch": 0.51, "learning_rate": 4.901454527152851e-05, "loss": 1.0437, "step": 203685 }, { "epoch": 0.51, "learning_rate": 4.901328703004681e-05, "loss": 1.0412, "step": 203690 }, { "epoch": 0.51, "learning_rate": 4.90120287885651e-05, "loss": 1.0438, "step": 203695 }, { "epoch": 0.51, "learning_rate": 4.90107705470834e-05, "loss": 1.042, "step": 203700 }, { "epoch": 0.51, "learning_rate": 4.900951230560169e-05, "loss": 1.0447, "step": 203705 }, { "epoch": 0.51, "learning_rate": 4.900825406411999e-05, "loss": 1.0469, "step": 203710 }, { "epoch": 0.51, "learning_rate": 4.900699582263828e-05, "loss": 1.0401, "step": 203715 }, { "epoch": 0.51, "learning_rate": 4.900573758115658e-05, "loss": 1.0437, "step": 203720 }, { "epoch": 0.51, "learning_rate": 4.900447933967487e-05, "loss": 1.0419, "step": 203725 }, { "epoch": 0.51, "learning_rate": 4.900322109819317e-05, "loss": 1.0414, "step": 203730 }, { "epoch": 0.51, "learning_rate": 4.900196285671146e-05, "loss": 1.0468, "step": 203735 }, { "epoch": 0.51, "learning_rate": 4.900070461522976e-05, "loss": 1.046, "step": 203740 }, { "epoch": 0.51, "learning_rate": 4.899944637374805e-05, "loss": 1.0422, "step": 203745 }, { "epoch": 0.51, "learning_rate": 4.899818813226635e-05, "loss": 1.0449, "step": 203750 }, { "epoch": 0.51, "learning_rate": 4.899692989078464e-05, "loss": 1.0453, "step": 203755 }, { "epoch": 0.51, "learning_rate": 4.8995671649302935e-05, "loss": 1.0443, "step": 203760 }, { "epoch": 0.51, "learning_rate": 4.899441340782123e-05, "loss": 1.0453, "step": 203765 }, { "epoch": 0.51, "learning_rate": 4.8993155166339524e-05, "loss": 1.045, "step": 203770 }, { "epoch": 0.51, "learning_rate": 4.899189692485782e-05, "loss": 1.0439, "step": 203775 }, { "epoch": 0.51, "learning_rate": 4.8990638683376114e-05, "loss": 1.0462, "step": 203780 }, { "epoch": 0.51, "learning_rate": 4.898938044189441e-05, "loss": 1.0452, "step": 203785 }, { "epoch": 0.51, "learning_rate": 4.8988122200412704e-05, "loss": 1.0459, "step": 203790 }, { "epoch": 0.51, "learning_rate": 4.8986863958931e-05, "loss": 1.0468, "step": 203795 }, { "epoch": 0.51, "learning_rate": 4.8985605717449294e-05, "loss": 1.0466, "step": 203800 }, { "epoch": 0.51, "learning_rate": 4.898434747596759e-05, "loss": 1.0468, "step": 203805 }, { "epoch": 0.51, "learning_rate": 4.8983089234485884e-05, "loss": 1.0441, "step": 203810 }, { "epoch": 0.51, "learning_rate": 4.898183099300418e-05, "loss": 1.0422, "step": 203815 }, { "epoch": 0.51, "learning_rate": 4.8980572751522474e-05, "loss": 1.0455, "step": 203820 }, { "epoch": 0.51, "learning_rate": 4.8979314510040766e-05, "loss": 1.0427, "step": 203825 }, { "epoch": 0.51, "learning_rate": 4.8978056268559064e-05, "loss": 1.0469, "step": 203830 }, { "epoch": 0.51, "learning_rate": 4.8976798027077356e-05, "loss": 1.0471, "step": 203835 }, { "epoch": 0.51, "learning_rate": 4.8975539785595654e-05, "loss": 1.0641, "step": 203840 }, { "epoch": 0.51, "learning_rate": 4.8974281544113946e-05, "loss": 1.0457, "step": 203845 }, { "epoch": 0.51, "learning_rate": 4.8973023302632244e-05, "loss": 1.0407, "step": 203850 }, { "epoch": 0.51, "learning_rate": 4.8971765061150536e-05, "loss": 1.0443, "step": 203855 }, { "epoch": 0.51, "learning_rate": 4.897075846796517e-05, "loss": 1.0432, "step": 203860 }, { "epoch": 0.51, "learning_rate": 4.8969500226483463e-05, "loss": 1.0434, "step": 203865 }, { "epoch": 0.51, "learning_rate": 4.896824198500176e-05, "loss": 1.0435, "step": 203870 }, { "epoch": 0.51, "learning_rate": 4.8966983743520053e-05, "loss": 1.0454, "step": 203875 }, { "epoch": 0.51, "learning_rate": 4.896572550203835e-05, "loss": 1.0407, "step": 203880 }, { "epoch": 0.51, "learning_rate": 4.8964467260556643e-05, "loss": 1.043, "step": 203885 }, { "epoch": 0.51, "learning_rate": 4.896320901907495e-05, "loss": 1.0439, "step": 203890 }, { "epoch": 0.51, "learning_rate": 4.896195077759324e-05, "loss": 1.0437, "step": 203895 }, { "epoch": 0.51, "learning_rate": 4.896069253611154e-05, "loss": 1.0414, "step": 203900 }, { "epoch": 0.51, "learning_rate": 4.895943429462983e-05, "loss": 1.0434, "step": 203905 }, { "epoch": 0.51, "learning_rate": 4.895817605314812e-05, "loss": 1.0432, "step": 203910 }, { "epoch": 0.51, "learning_rate": 4.895691781166642e-05, "loss": 1.0449, "step": 203915 }, { "epoch": 0.51, "learning_rate": 4.895565957018471e-05, "loss": 1.0458, "step": 203920 }, { "epoch": 0.51, "learning_rate": 4.895440132870301e-05, "loss": 1.0416, "step": 203925 }, { "epoch": 0.51, "learning_rate": 4.89531430872213e-05, "loss": 1.0434, "step": 203930 }, { "epoch": 0.51, "learning_rate": 4.89518848457396e-05, "loss": 1.0471, "step": 203935 }, { "epoch": 0.51, "learning_rate": 4.895062660425789e-05, "loss": 1.0462, "step": 203940 }, { "epoch": 0.51, "learning_rate": 4.894936836277619e-05, "loss": 1.0417, "step": 203945 }, { "epoch": 0.51, "learning_rate": 4.894811012129448e-05, "loss": 1.0457, "step": 203950 }, { "epoch": 0.51, "learning_rate": 4.894685187981278e-05, "loss": 1.0452, "step": 203955 }, { "epoch": 0.51, "learning_rate": 4.894559363833107e-05, "loss": 1.0461, "step": 203960 }, { "epoch": 0.51, "learning_rate": 4.894433539684937e-05, "loss": 1.042, "step": 203965 }, { "epoch": 0.51, "learning_rate": 4.894307715536766e-05, "loss": 1.0446, "step": 203970 }, { "epoch": 0.51, "learning_rate": 4.894181891388595e-05, "loss": 1.0432, "step": 203975 }, { "epoch": 0.51, "learning_rate": 4.894056067240425e-05, "loss": 1.042, "step": 203980 }, { "epoch": 0.51, "learning_rate": 4.893930243092254e-05, "loss": 1.0424, "step": 203985 }, { "epoch": 0.51, "learning_rate": 4.893804418944084e-05, "loss": 1.0694, "step": 203990 }, { "epoch": 0.51, "learning_rate": 4.893678594795913e-05, "loss": 1.0454, "step": 203995 }, { "epoch": 0.51, "learning_rate": 4.893552770647743e-05, "loss": 1.0463, "step": 204000 }, { "epoch": 0.51, "learning_rate": 4.893426946499572e-05, "loss": 1.0468, "step": 204005 }, { "epoch": 0.51, "learning_rate": 4.893301122351402e-05, "loss": 1.0423, "step": 204010 }, { "epoch": 0.51, "learning_rate": 4.893175298203231e-05, "loss": 1.0452, "step": 204015 }, { "epoch": 0.51, "learning_rate": 4.893049474055061e-05, "loss": 1.0442, "step": 204020 }, { "epoch": 0.51, "learning_rate": 4.89292364990689e-05, "loss": 1.0469, "step": 204025 }, { "epoch": 0.51, "learning_rate": 4.89279782575872e-05, "loss": 1.0442, "step": 204030 }, { "epoch": 0.51, "learning_rate": 4.892672001610549e-05, "loss": 1.0452, "step": 204035 }, { "epoch": 0.51, "learning_rate": 4.8925461774623785e-05, "loss": 1.0445, "step": 204040 }, { "epoch": 0.51, "learning_rate": 4.892420353314208e-05, "loss": 1.0408, "step": 204045 }, { "epoch": 0.51, "learning_rate": 4.8922945291660375e-05, "loss": 1.0444, "step": 204050 }, { "epoch": 0.51, "learning_rate": 4.892168705017867e-05, "loss": 1.0466, "step": 204055 }, { "epoch": 0.51, "learning_rate": 4.8920428808696964e-05, "loss": 1.0452, "step": 204060 }, { "epoch": 0.51, "learning_rate": 4.891917056721526e-05, "loss": 1.0415, "step": 204065 }, { "epoch": 0.51, "learning_rate": 4.8917912325733554e-05, "loss": 1.0447, "step": 204070 }, { "epoch": 0.51, "learning_rate": 4.891665408425185e-05, "loss": 1.0441, "step": 204075 }, { "epoch": 0.51, "learning_rate": 4.8915395842770144e-05, "loss": 1.0422, "step": 204080 }, { "epoch": 0.51, "learning_rate": 4.891413760128844e-05, "loss": 1.0445, "step": 204085 }, { "epoch": 0.51, "learning_rate": 4.8912879359806734e-05, "loss": 1.044, "step": 204090 }, { "epoch": 0.51, "learning_rate": 4.891162111832503e-05, "loss": 1.0466, "step": 204095 }, { "epoch": 0.51, "learning_rate": 4.8910362876843324e-05, "loss": 1.0455, "step": 204100 }, { "epoch": 0.51, "learning_rate": 4.8909104635361616e-05, "loss": 1.0464, "step": 204105 }, { "epoch": 0.51, "learning_rate": 4.8907846393879914e-05, "loss": 1.044, "step": 204110 }, { "epoch": 0.51, "learning_rate": 4.8906588152398206e-05, "loss": 1.0449, "step": 204115 }, { "epoch": 0.51, "learning_rate": 4.8905329910916504e-05, "loss": 1.0455, "step": 204120 }, { "epoch": 0.51, "learning_rate": 4.8904071669434796e-05, "loss": 1.046, "step": 204125 }, { "epoch": 0.51, "learning_rate": 4.8902813427953094e-05, "loss": 1.0449, "step": 204130 }, { "epoch": 0.51, "learning_rate": 4.8901555186471386e-05, "loss": 1.0441, "step": 204135 }, { "epoch": 0.51, "learning_rate": 4.8900296944989684e-05, "loss": 1.0438, "step": 204140 }, { "epoch": 0.51, "learning_rate": 4.889903870350798e-05, "loss": 1.0449, "step": 204145 }, { "epoch": 0.51, "learning_rate": 4.8897780462026274e-05, "loss": 1.0446, "step": 204150 }, { "epoch": 0.51, "learning_rate": 4.889652222054457e-05, "loss": 1.0449, "step": 204155 }, { "epoch": 0.51, "learning_rate": 4.8895263979062864e-05, "loss": 1.0409, "step": 204160 }, { "epoch": 0.51, "learning_rate": 4.889400573758116e-05, "loss": 1.044, "step": 204165 }, { "epoch": 0.51, "learning_rate": 4.8892747496099454e-05, "loss": 1.0438, "step": 204170 }, { "epoch": 0.51, "learning_rate": 4.889148925461775e-05, "loss": 1.0436, "step": 204175 }, { "epoch": 0.51, "learning_rate": 4.8890231013136044e-05, "loss": 1.0453, "step": 204180 }, { "epoch": 0.51, "learning_rate": 4.888897277165434e-05, "loss": 1.0441, "step": 204185 }, { "epoch": 0.51, "learning_rate": 4.8887714530172634e-05, "loss": 1.0453, "step": 204190 }, { "epoch": 0.51, "learning_rate": 4.888645628869093e-05, "loss": 1.0447, "step": 204195 }, { "epoch": 0.51, "learning_rate": 4.8885198047209224e-05, "loss": 1.0466, "step": 204200 }, { "epoch": 0.51, "learning_rate": 4.8883939805727516e-05, "loss": 1.0584, "step": 204205 }, { "epoch": 0.51, "learning_rate": 4.8882681564245814e-05, "loss": 1.0414, "step": 204210 }, { "epoch": 0.51, "learning_rate": 4.8881423322764106e-05, "loss": 1.0438, "step": 204215 }, { "epoch": 0.51, "learning_rate": 4.8880165081282404e-05, "loss": 1.0406, "step": 204220 }, { "epoch": 0.51, "learning_rate": 4.8878906839800696e-05, "loss": 1.043, "step": 204225 }, { "epoch": 0.51, "learning_rate": 4.8877648598318994e-05, "loss": 1.0448, "step": 204230 }, { "epoch": 0.51, "learning_rate": 4.8876390356837286e-05, "loss": 1.0456, "step": 204235 }, { "epoch": 0.51, "learning_rate": 4.8875132115355584e-05, "loss": 1.0458, "step": 204240 }, { "epoch": 0.51, "learning_rate": 4.8873873873873876e-05, "loss": 1.0405, "step": 204245 }, { "epoch": 0.51, "learning_rate": 4.8872615632392174e-05, "loss": 1.0469, "step": 204250 }, { "epoch": 0.51, "learning_rate": 4.8871357390910466e-05, "loss": 1.0454, "step": 204255 }, { "epoch": 0.51, "learning_rate": 4.8870099149428764e-05, "loss": 1.0483, "step": 204260 }, { "epoch": 0.51, "learning_rate": 4.8868840907947056e-05, "loss": 1.0461, "step": 204265 }, { "epoch": 0.51, "learning_rate": 4.886758266646535e-05, "loss": 1.0428, "step": 204270 }, { "epoch": 0.51, "learning_rate": 4.8866324424983645e-05, "loss": 1.0423, "step": 204275 }, { "epoch": 0.51, "learning_rate": 4.886506618350194e-05, "loss": 1.0443, "step": 204280 }, { "epoch": 0.51, "learning_rate": 4.8863807942020235e-05, "loss": 1.0435, "step": 204285 }, { "epoch": 0.51, "learning_rate": 4.886254970053853e-05, "loss": 1.044, "step": 204290 }, { "epoch": 0.51, "learning_rate": 4.8861291459056825e-05, "loss": 1.0658, "step": 204295 }, { "epoch": 0.51, "learning_rate": 4.886003321757512e-05, "loss": 1.0454, "step": 204300 }, { "epoch": 0.51, "learning_rate": 4.8858774976093415e-05, "loss": 1.045, "step": 204305 }, { "epoch": 0.51, "learning_rate": 4.885751673461171e-05, "loss": 1.0434, "step": 204310 }, { "epoch": 0.51, "learning_rate": 4.8856258493130005e-05, "loss": 1.0458, "step": 204315 }, { "epoch": 0.51, "learning_rate": 4.88550002516483e-05, "loss": 1.0444, "step": 204320 }, { "epoch": 0.51, "learning_rate": 4.8853742010166595e-05, "loss": 1.0408, "step": 204325 }, { "epoch": 0.51, "learning_rate": 4.885248376868489e-05, "loss": 1.0431, "step": 204330 }, { "epoch": 0.51, "learning_rate": 4.885122552720318e-05, "loss": 1.0433, "step": 204335 }, { "epoch": 0.51, "learning_rate": 4.884996728572148e-05, "loss": 1.0432, "step": 204340 }, { "epoch": 0.51, "learning_rate": 4.884870904423977e-05, "loss": 1.0446, "step": 204345 }, { "epoch": 0.51, "learning_rate": 4.884745080275807e-05, "loss": 1.0467, "step": 204350 }, { "epoch": 0.51, "learning_rate": 4.884619256127636e-05, "loss": 1.046, "step": 204355 }, { "epoch": 0.51, "learning_rate": 4.884493431979466e-05, "loss": 1.048, "step": 204360 }, { "epoch": 0.51, "learning_rate": 4.884367607831295e-05, "loss": 1.0661, "step": 204365 }, { "epoch": 0.51, "learning_rate": 4.884241783683125e-05, "loss": 1.0427, "step": 204370 }, { "epoch": 0.51, "learning_rate": 4.884115959534954e-05, "loss": 1.0461, "step": 204375 }, { "epoch": 0.51, "learning_rate": 4.883990135386784e-05, "loss": 1.0488, "step": 204380 }, { "epoch": 0.51, "learning_rate": 4.883864311238613e-05, "loss": 1.0456, "step": 204385 }, { "epoch": 0.51, "learning_rate": 4.883738487090443e-05, "loss": 1.0444, "step": 204390 }, { "epoch": 0.51, "learning_rate": 4.8836126629422725e-05, "loss": 1.0414, "step": 204395 }, { "epoch": 0.51, "learning_rate": 4.883486838794102e-05, "loss": 1.0391, "step": 204400 }, { "epoch": 0.51, "learning_rate": 4.8833610146459315e-05, "loss": 1.0425, "step": 204405 }, { "epoch": 0.51, "learning_rate": 4.883235190497761e-05, "loss": 1.0399, "step": 204410 }, { "epoch": 0.51, "learning_rate": 4.8831093663495905e-05, "loss": 1.0464, "step": 204415 }, { "epoch": 0.51, "learning_rate": 4.88298354220142e-05, "loss": 1.0487, "step": 204420 }, { "epoch": 0.51, "learning_rate": 4.8828577180532495e-05, "loss": 1.0484, "step": 204425 }, { "epoch": 0.51, "learning_rate": 4.882731893905079e-05, "loss": 1.0423, "step": 204430 }, { "epoch": 0.51, "learning_rate": 4.882606069756908e-05, "loss": 1.0432, "step": 204435 }, { "epoch": 0.51, "learning_rate": 4.882480245608738e-05, "loss": 1.0439, "step": 204440 }, { "epoch": 0.51, "learning_rate": 4.882354421460567e-05, "loss": 1.0678, "step": 204445 }, { "epoch": 0.51, "learning_rate": 4.8822285973123967e-05, "loss": 1.0427, "step": 204450 }, { "epoch": 0.51, "learning_rate": 4.882102773164226e-05, "loss": 1.0425, "step": 204455 }, { "epoch": 0.51, "learning_rate": 4.8819769490160557e-05, "loss": 1.0447, "step": 204460 }, { "epoch": 0.51, "learning_rate": 4.881851124867885e-05, "loss": 1.0416, "step": 204465 }, { "epoch": 0.51, "learning_rate": 4.8817253007197147e-05, "loss": 1.0618, "step": 204470 }, { "epoch": 0.51, "learning_rate": 4.881599476571544e-05, "loss": 1.0442, "step": 204475 }, { "epoch": 0.51, "learning_rate": 4.8814736524233737e-05, "loss": 1.0439, "step": 204480 }, { "epoch": 0.51, "learning_rate": 4.881347828275203e-05, "loss": 1.0416, "step": 204485 }, { "epoch": 0.51, "learning_rate": 4.8812220041270326e-05, "loss": 1.0418, "step": 204490 }, { "epoch": 0.51, "learning_rate": 4.881096179978862e-05, "loss": 1.0437, "step": 204495 }, { "epoch": 0.51, "learning_rate": 4.880970355830691e-05, "loss": 1.0486, "step": 204500 }, { "epoch": 0.51, "learning_rate": 4.880844531682521e-05, "loss": 1.0464, "step": 204505 }, { "epoch": 0.51, "learning_rate": 4.88071870753435e-05, "loss": 1.0427, "step": 204510 }, { "epoch": 0.51, "learning_rate": 4.88059288338618e-05, "loss": 1.0395, "step": 204515 }, { "epoch": 0.51, "learning_rate": 4.880467059238009e-05, "loss": 1.0411, "step": 204520 }, { "epoch": 0.51, "learning_rate": 4.880341235089839e-05, "loss": 1.0454, "step": 204525 }, { "epoch": 0.51, "learning_rate": 4.880215410941668e-05, "loss": 1.0452, "step": 204530 }, { "epoch": 0.51, "learning_rate": 4.880089586793498e-05, "loss": 1.0463, "step": 204535 }, { "epoch": 0.51, "learning_rate": 4.879963762645327e-05, "loss": 1.0629, "step": 204540 }, { "epoch": 0.51, "learning_rate": 4.879837938497157e-05, "loss": 1.0432, "step": 204545 }, { "epoch": 0.51, "learning_rate": 4.879712114348986e-05, "loss": 1.0452, "step": 204550 }, { "epoch": 0.51, "learning_rate": 4.879586290200816e-05, "loss": 1.047, "step": 204555 }, { "epoch": 0.51, "learning_rate": 4.879460466052645e-05, "loss": 1.0448, "step": 204560 }, { "epoch": 0.51, "learning_rate": 4.879334641904474e-05, "loss": 1.0624, "step": 204565 }, { "epoch": 0.51, "learning_rate": 4.879208817756304e-05, "loss": 1.0428, "step": 204570 }, { "epoch": 0.51, "learning_rate": 4.879082993608133e-05, "loss": 1.044, "step": 204575 }, { "epoch": 0.51, "learning_rate": 4.878957169459963e-05, "loss": 1.0462, "step": 204580 }, { "epoch": 0.51, "learning_rate": 4.878831345311792e-05, "loss": 1.0456, "step": 204585 }, { "epoch": 0.51, "learning_rate": 4.878705521163622e-05, "loss": 1.0452, "step": 204590 }, { "epoch": 0.51, "learning_rate": 4.878579697015451e-05, "loss": 1.0422, "step": 204595 }, { "epoch": 0.51, "learning_rate": 4.878453872867281e-05, "loss": 1.0445, "step": 204600 }, { "epoch": 0.51, "learning_rate": 4.87832804871911e-05, "loss": 1.0434, "step": 204605 }, { "epoch": 0.51, "learning_rate": 4.87820222457094e-05, "loss": 1.0451, "step": 204610 }, { "epoch": 0.51, "learning_rate": 4.878076400422769e-05, "loss": 1.0458, "step": 204615 }, { "epoch": 0.51, "learning_rate": 4.877950576274599e-05, "loss": 1.045, "step": 204620 }, { "epoch": 0.51, "learning_rate": 4.877824752126428e-05, "loss": 1.0447, "step": 204625 }, { "epoch": 0.51, "learning_rate": 4.877698927978257e-05, "loss": 1.0414, "step": 204630 }, { "epoch": 0.51, "learning_rate": 4.877573103830087e-05, "loss": 1.0428, "step": 204635 }, { "epoch": 0.51, "learning_rate": 4.877447279681917e-05, "loss": 1.0679, "step": 204640 }, { "epoch": 0.51, "learning_rate": 4.877321455533747e-05, "loss": 1.0435, "step": 204645 }, { "epoch": 0.51, "learning_rate": 4.877195631385576e-05, "loss": 1.0476, "step": 204650 }, { "epoch": 0.51, "learning_rate": 4.877069807237406e-05, "loss": 1.0457, "step": 204655 }, { "epoch": 0.51, "learning_rate": 4.876943983089235e-05, "loss": 1.0438, "step": 204660 }, { "epoch": 0.51, "learning_rate": 4.876818158941064e-05, "loss": 1.0448, "step": 204665 }, { "epoch": 0.51, "learning_rate": 4.876692334792894e-05, "loss": 1.0439, "step": 204670 }, { "epoch": 0.51, "learning_rate": 4.876566510644723e-05, "loss": 1.046, "step": 204675 }, { "epoch": 0.51, "learning_rate": 4.876440686496553e-05, "loss": 1.0446, "step": 204680 }, { "epoch": 0.51, "learning_rate": 4.876314862348382e-05, "loss": 1.0418, "step": 204685 }, { "epoch": 0.51, "learning_rate": 4.876189038200212e-05, "loss": 1.0652, "step": 204690 }, { "epoch": 0.51, "learning_rate": 4.876063214052041e-05, "loss": 1.0417, "step": 204695 }, { "epoch": 0.51, "learning_rate": 4.875937389903871e-05, "loss": 1.0459, "step": 204700 }, { "epoch": 0.51, "learning_rate": 4.8758115657557e-05, "loss": 1.0449, "step": 204705 }, { "epoch": 0.51, "learning_rate": 4.87568574160753e-05, "loss": 1.0447, "step": 204710 }, { "epoch": 0.51, "learning_rate": 4.875559917459359e-05, "loss": 1.0434, "step": 204715 }, { "epoch": 0.51, "learning_rate": 4.875434093311189e-05, "loss": 1.0432, "step": 204720 }, { "epoch": 0.51, "learning_rate": 4.875308269163018e-05, "loss": 1.046, "step": 204725 }, { "epoch": 0.51, "learning_rate": 4.875182445014847e-05, "loss": 1.0442, "step": 204730 }, { "epoch": 0.51, "learning_rate": 4.875056620866677e-05, "loss": 1.0628, "step": 204735 }, { "epoch": 0.51, "learning_rate": 4.874930796718506e-05, "loss": 1.0461, "step": 204740 }, { "epoch": 0.51, "learning_rate": 4.874804972570336e-05, "loss": 1.0456, "step": 204745 }, { "epoch": 0.51, "learning_rate": 4.874679148422165e-05, "loss": 1.0432, "step": 204750 }, { "epoch": 0.51, "learning_rate": 4.874553324273995e-05, "loss": 1.046, "step": 204755 }, { "epoch": 0.51, "learning_rate": 4.874427500125824e-05, "loss": 1.0445, "step": 204760 }, { "epoch": 0.51, "learning_rate": 4.874301675977654e-05, "loss": 1.0432, "step": 204765 }, { "epoch": 0.51, "learning_rate": 4.874175851829483e-05, "loss": 1.0487, "step": 204770 }, { "epoch": 0.51, "learning_rate": 4.874050027681313e-05, "loss": 1.045, "step": 204775 }, { "epoch": 0.51, "learning_rate": 4.873924203533142e-05, "loss": 1.0441, "step": 204780 }, { "epoch": 0.51, "learning_rate": 4.873798379384972e-05, "loss": 1.0436, "step": 204785 }, { "epoch": 0.51, "learning_rate": 4.873672555236801e-05, "loss": 1.0441, "step": 204790 }, { "epoch": 0.51, "learning_rate": 4.8735467310886304e-05, "loss": 1.0455, "step": 204795 }, { "epoch": 0.51, "learning_rate": 4.87342090694046e-05, "loss": 1.0468, "step": 204800 }, { "epoch": 0.51, "learning_rate": 4.8732950827922894e-05, "loss": 1.0412, "step": 204805 }, { "epoch": 0.51, "learning_rate": 4.873169258644119e-05, "loss": 1.0422, "step": 204810 }, { "epoch": 0.51, "learning_rate": 4.8730434344959484e-05, "loss": 1.0431, "step": 204815 }, { "epoch": 0.51, "learning_rate": 4.872917610347778e-05, "loss": 1.0423, "step": 204820 }, { "epoch": 0.51, "learning_rate": 4.8727917861996074e-05, "loss": 1.0444, "step": 204825 }, { "epoch": 0.51, "learning_rate": 4.872665962051437e-05, "loss": 1.0467, "step": 204830 }, { "epoch": 0.51, "learning_rate": 4.8725401379032664e-05, "loss": 1.065, "step": 204835 }, { "epoch": 0.51, "learning_rate": 4.872414313755096e-05, "loss": 1.0461, "step": 204840 }, { "epoch": 0.51, "learning_rate": 4.8722884896069254e-05, "loss": 1.0438, "step": 204845 }, { "epoch": 0.51, "learning_rate": 4.872162665458755e-05, "loss": 1.0409, "step": 204850 }, { "epoch": 0.51, "learning_rate": 4.8720368413105843e-05, "loss": 1.0447, "step": 204855 }, { "epoch": 0.51, "learning_rate": 4.8719110171624135e-05, "loss": 1.0728, "step": 204860 }, { "epoch": 0.51, "learning_rate": 4.8717851930142433e-05, "loss": 1.0441, "step": 204865 }, { "epoch": 0.51, "learning_rate": 4.8716593688660725e-05, "loss": 1.0456, "step": 204870 }, { "epoch": 0.51, "learning_rate": 4.8715335447179023e-05, "loss": 1.0455, "step": 204875 }, { "epoch": 0.51, "learning_rate": 4.8714077205697315e-05, "loss": 1.0462, "step": 204880 }, { "epoch": 0.51, "learning_rate": 4.8712818964215613e-05, "loss": 1.0407, "step": 204885 }, { "epoch": 0.51, "learning_rate": 4.871156072273391e-05, "loss": 1.0468, "step": 204890 }, { "epoch": 0.51, "learning_rate": 4.871030248125221e-05, "loss": 1.042, "step": 204895 }, { "epoch": 0.51, "learning_rate": 4.87090442397705e-05, "loss": 1.0426, "step": 204900 }, { "epoch": 0.51, "learning_rate": 4.870778599828879e-05, "loss": 1.0441, "step": 204905 }, { "epoch": 0.51, "learning_rate": 4.870652775680709e-05, "loss": 1.0442, "step": 204910 }, { "epoch": 0.51, "learning_rate": 4.870526951532538e-05, "loss": 1.0464, "step": 204915 }, { "epoch": 0.51, "learning_rate": 4.870401127384368e-05, "loss": 1.0431, "step": 204920 }, { "epoch": 0.51, "learning_rate": 4.870275303236197e-05, "loss": 1.044, "step": 204925 }, { "epoch": 0.51, "learning_rate": 4.870149479088027e-05, "loss": 1.0434, "step": 204930 }, { "epoch": 0.51, "learning_rate": 4.870023654939856e-05, "loss": 1.0441, "step": 204935 }, { "epoch": 0.51, "learning_rate": 4.869897830791686e-05, "loss": 1.0448, "step": 204940 }, { "epoch": 0.51, "learning_rate": 4.869772006643515e-05, "loss": 1.0418, "step": 204945 }, { "epoch": 0.51, "learning_rate": 4.869646182495345e-05, "loss": 1.045, "step": 204950 }, { "epoch": 0.51, "learning_rate": 4.869520358347174e-05, "loss": 1.0432, "step": 204955 }, { "epoch": 0.51, "learning_rate": 4.869394534199004e-05, "loss": 1.0443, "step": 204960 }, { "epoch": 0.51, "learning_rate": 4.869268710050833e-05, "loss": 1.0652, "step": 204965 }, { "epoch": 0.51, "learning_rate": 4.8691428859026625e-05, "loss": 1.0457, "step": 204970 }, { "epoch": 0.51, "learning_rate": 4.869017061754492e-05, "loss": 1.0606, "step": 204975 }, { "epoch": 0.51, "learning_rate": 4.8688912376063215e-05, "loss": 1.0441, "step": 204980 }, { "epoch": 0.51, "learning_rate": 4.868765413458151e-05, "loss": 1.0423, "step": 204985 }, { "epoch": 0.51, "learning_rate": 4.8686395893099805e-05, "loss": 1.0469, "step": 204990 }, { "epoch": 0.51, "learning_rate": 4.86851376516181e-05, "loss": 1.0419, "step": 204995 }, { "epoch": 0.51, "learning_rate": 4.8683879410136395e-05, "loss": 1.0475, "step": 205000 }, { "epoch": 0.51, "learning_rate": 4.868262116865469e-05, "loss": 1.0438, "step": 205005 }, { "epoch": 0.51, "learning_rate": 4.8681362927172985e-05, "loss": 1.044, "step": 205010 }, { "epoch": 0.51, "learning_rate": 4.868010468569128e-05, "loss": 1.0446, "step": 205015 }, { "epoch": 0.51, "learning_rate": 4.8678846444209575e-05, "loss": 1.0453, "step": 205020 }, { "epoch": 0.51, "learning_rate": 4.8677588202727866e-05, "loss": 1.0423, "step": 205025 }, { "epoch": 0.51, "learning_rate": 4.8676329961246165e-05, "loss": 1.0449, "step": 205030 }, { "epoch": 0.51, "learning_rate": 4.8675071719764456e-05, "loss": 1.0462, "step": 205035 }, { "epoch": 0.51, "learning_rate": 4.8673813478282755e-05, "loss": 1.0459, "step": 205040 }, { "epoch": 0.51, "learning_rate": 4.8672555236801046e-05, "loss": 1.046, "step": 205045 }, { "epoch": 0.51, "learning_rate": 4.8671296995319345e-05, "loss": 1.0458, "step": 205050 }, { "epoch": 0.51, "learning_rate": 4.8670038753837636e-05, "loss": 1.0456, "step": 205055 }, { "epoch": 0.51, "learning_rate": 4.8668780512355935e-05, "loss": 1.0444, "step": 205060 }, { "epoch": 0.51, "learning_rate": 4.8667522270874226e-05, "loss": 1.0458, "step": 205065 }, { "epoch": 0.51, "learning_rate": 4.8666264029392524e-05, "loss": 1.0467, "step": 205070 }, { "epoch": 0.51, "learning_rate": 4.8665005787910816e-05, "loss": 1.0444, "step": 205075 }, { "epoch": 0.51, "learning_rate": 4.8663747546429114e-05, "loss": 1.0438, "step": 205080 }, { "epoch": 0.51, "learning_rate": 4.8662489304947406e-05, "loss": 1.0424, "step": 205085 }, { "epoch": 0.51, "learning_rate": 4.86612310634657e-05, "loss": 1.0441, "step": 205090 }, { "epoch": 0.51, "learning_rate": 4.8659972821983996e-05, "loss": 1.0437, "step": 205095 }, { "epoch": 0.51, "learning_rate": 4.865871458050229e-05, "loss": 1.0461, "step": 205100 }, { "epoch": 0.51, "learning_rate": 4.8657456339020586e-05, "loss": 1.0454, "step": 205105 }, { "epoch": 0.51, "learning_rate": 4.865619809753888e-05, "loss": 1.0434, "step": 205110 }, { "epoch": 0.51, "learning_rate": 4.8654939856057176e-05, "loss": 1.0651, "step": 205115 }, { "epoch": 0.51, "learning_rate": 4.865368161457547e-05, "loss": 1.0695, "step": 205120 }, { "epoch": 0.51, "learning_rate": 4.8652423373093766e-05, "loss": 1.044, "step": 205125 }, { "epoch": 0.51, "learning_rate": 4.865116513161206e-05, "loss": 1.0463, "step": 205130 }, { "epoch": 0.51, "learning_rate": 4.8649906890130356e-05, "loss": 1.0652, "step": 205135 }, { "epoch": 0.51, "learning_rate": 4.8648648648648654e-05, "loss": 1.042, "step": 205140 }, { "epoch": 0.51, "learning_rate": 4.8647390407166946e-05, "loss": 1.0439, "step": 205145 }, { "epoch": 0.51, "learning_rate": 4.8646132165685244e-05, "loss": 1.0477, "step": 205150 }, { "epoch": 0.51, "learning_rate": 4.8644873924203536e-05, "loss": 1.066, "step": 205155 }, { "epoch": 0.51, "learning_rate": 4.8643615682721834e-05, "loss": 1.0432, "step": 205160 }, { "epoch": 0.51, "learning_rate": 4.8642357441240126e-05, "loss": 1.0466, "step": 205165 }, { "epoch": 0.52, "learning_rate": 4.8641099199758424e-05, "loss": 1.0442, "step": 205170 }, { "epoch": 0.52, "learning_rate": 4.8639840958276716e-05, "loss": 1.0662, "step": 205175 }, { "epoch": 0.52, "learning_rate": 4.8638582716795014e-05, "loss": 1.0463, "step": 205180 }, { "epoch": 0.52, "learning_rate": 4.8637324475313306e-05, "loss": 1.0454, "step": 205185 }, { "epoch": 0.52, "learning_rate": 4.8636066233831604e-05, "loss": 1.0462, "step": 205190 }, { "epoch": 0.52, "learning_rate": 4.8634807992349896e-05, "loss": 1.0416, "step": 205195 }, { "epoch": 0.52, "learning_rate": 4.863354975086819e-05, "loss": 1.0454, "step": 205200 }, { "epoch": 0.52, "learning_rate": 4.8632291509386486e-05, "loss": 1.0468, "step": 205205 }, { "epoch": 0.52, "learning_rate": 4.863103326790478e-05, "loss": 1.0448, "step": 205210 }, { "epoch": 0.52, "learning_rate": 4.8629775026423076e-05, "loss": 1.0669, "step": 205215 }, { "epoch": 0.52, "learning_rate": 4.862851678494137e-05, "loss": 1.0417, "step": 205220 }, { "epoch": 0.52, "learning_rate": 4.8627258543459666e-05, "loss": 1.0427, "step": 205225 }, { "epoch": 0.52, "learning_rate": 4.862600030197796e-05, "loss": 1.0445, "step": 205230 }, { "epoch": 0.52, "learning_rate": 4.8624742060496256e-05, "loss": 1.0438, "step": 205235 }, { "epoch": 0.52, "learning_rate": 4.862348381901455e-05, "loss": 1.0483, "step": 205240 }, { "epoch": 0.52, "learning_rate": 4.8622225577532846e-05, "loss": 1.0436, "step": 205245 }, { "epoch": 0.52, "learning_rate": 4.862096733605114e-05, "loss": 1.0473, "step": 205250 }, { "epoch": 0.52, "learning_rate": 4.8619709094569436e-05, "loss": 1.0429, "step": 205255 }, { "epoch": 0.52, "learning_rate": 4.861845085308773e-05, "loss": 1.0432, "step": 205260 }, { "epoch": 0.52, "learning_rate": 4.861719261160602e-05, "loss": 1.0468, "step": 205265 }, { "epoch": 0.52, "learning_rate": 4.861593437012432e-05, "loss": 1.0437, "step": 205270 }, { "epoch": 0.52, "learning_rate": 4.861467612864261e-05, "loss": 1.0418, "step": 205275 }, { "epoch": 0.52, "learning_rate": 4.861341788716091e-05, "loss": 1.0449, "step": 205280 }, { "epoch": 0.52, "learning_rate": 4.86121596456792e-05, "loss": 1.0423, "step": 205285 }, { "epoch": 0.52, "learning_rate": 4.86109014041975e-05, "loss": 1.0484, "step": 205290 }, { "epoch": 0.52, "learning_rate": 4.860964316271579e-05, "loss": 1.0459, "step": 205295 }, { "epoch": 0.52, "learning_rate": 4.860838492123409e-05, "loss": 1.0454, "step": 205300 }, { "epoch": 0.52, "learning_rate": 4.860712667975238e-05, "loss": 1.0627, "step": 205305 }, { "epoch": 0.52, "learning_rate": 4.860586843827068e-05, "loss": 1.0421, "step": 205310 }, { "epoch": 0.52, "learning_rate": 4.860461019678897e-05, "loss": 1.0451, "step": 205315 }, { "epoch": 0.52, "learning_rate": 4.860335195530727e-05, "loss": 1.045, "step": 205320 }, { "epoch": 0.52, "learning_rate": 4.860209371382556e-05, "loss": 1.0691, "step": 205325 }, { "epoch": 0.52, "learning_rate": 4.860083547234385e-05, "loss": 1.0408, "step": 205330 }, { "epoch": 0.52, "learning_rate": 4.859957723086215e-05, "loss": 1.0646, "step": 205335 }, { "epoch": 0.52, "learning_rate": 4.859831898938044e-05, "loss": 1.0691, "step": 205340 }, { "epoch": 0.52, "learning_rate": 4.859706074789874e-05, "loss": 1.0444, "step": 205345 }, { "epoch": 0.52, "learning_rate": 4.859580250641703e-05, "loss": 1.0494, "step": 205350 }, { "epoch": 0.52, "learning_rate": 4.859454426493533e-05, "loss": 1.046, "step": 205355 }, { "epoch": 0.52, "learning_rate": 4.859328602345362e-05, "loss": 1.0422, "step": 205360 }, { "epoch": 0.52, "learning_rate": 4.859202778197192e-05, "loss": 1.045, "step": 205365 }, { "epoch": 0.52, "learning_rate": 4.859076954049021e-05, "loss": 1.0454, "step": 205370 }, { "epoch": 0.52, "learning_rate": 4.858951129900851e-05, "loss": 1.0444, "step": 205375 }, { "epoch": 0.52, "learning_rate": 4.85882530575268e-05, "loss": 1.0446, "step": 205380 }, { "epoch": 0.52, "learning_rate": 4.858699481604509e-05, "loss": 1.045, "step": 205385 }, { "epoch": 0.52, "learning_rate": 4.85857365745634e-05, "loss": 1.0444, "step": 205390 }, { "epoch": 0.52, "learning_rate": 4.858447833308169e-05, "loss": 1.0392, "step": 205395 }, { "epoch": 0.52, "learning_rate": 4.858322009159999e-05, "loss": 1.0446, "step": 205400 }, { "epoch": 0.52, "learning_rate": 4.858196185011828e-05, "loss": 1.0437, "step": 205405 }, { "epoch": 0.52, "learning_rate": 4.858070360863658e-05, "loss": 1.0459, "step": 205410 }, { "epoch": 0.52, "learning_rate": 4.857944536715487e-05, "loss": 1.0455, "step": 205415 }, { "epoch": 0.52, "learning_rate": 4.857818712567317e-05, "loss": 1.0451, "step": 205420 }, { "epoch": 0.52, "learning_rate": 4.857692888419146e-05, "loss": 1.0437, "step": 205425 }, { "epoch": 0.52, "learning_rate": 4.857567064270975e-05, "loss": 1.0433, "step": 205430 }, { "epoch": 0.52, "learning_rate": 4.857441240122805e-05, "loss": 1.0457, "step": 205435 }, { "epoch": 0.52, "learning_rate": 4.857315415974634e-05, "loss": 1.046, "step": 205440 }, { "epoch": 0.52, "learning_rate": 4.857189591826464e-05, "loss": 1.0429, "step": 205445 }, { "epoch": 0.52, "learning_rate": 4.857063767678293e-05, "loss": 1.0456, "step": 205450 }, { "epoch": 0.52, "learning_rate": 4.856937943530123e-05, "loss": 1.045, "step": 205455 }, { "epoch": 0.52, "learning_rate": 4.856812119381952e-05, "loss": 1.0439, "step": 205460 }, { "epoch": 0.52, "learning_rate": 4.856686295233782e-05, "loss": 1.0435, "step": 205465 }, { "epoch": 0.52, "learning_rate": 4.856560471085611e-05, "loss": 1.046, "step": 205470 }, { "epoch": 0.52, "learning_rate": 4.856434646937441e-05, "loss": 1.0453, "step": 205475 }, { "epoch": 0.52, "learning_rate": 4.85630882278927e-05, "loss": 1.0461, "step": 205480 }, { "epoch": 0.52, "learning_rate": 4.8561829986411e-05, "loss": 1.0466, "step": 205485 }, { "epoch": 0.52, "learning_rate": 4.856057174492929e-05, "loss": 1.0445, "step": 205490 }, { "epoch": 0.52, "learning_rate": 4.855931350344758e-05, "loss": 1.0458, "step": 205495 }, { "epoch": 0.52, "learning_rate": 4.855805526196588e-05, "loss": 1.0478, "step": 205500 }, { "epoch": 0.52, "learning_rate": 4.855679702048417e-05, "loss": 1.0642, "step": 205505 }, { "epoch": 0.52, "learning_rate": 4.855553877900247e-05, "loss": 1.0448, "step": 205510 }, { "epoch": 0.52, "learning_rate": 4.855428053752076e-05, "loss": 1.0426, "step": 205515 }, { "epoch": 0.52, "learning_rate": 4.855302229603906e-05, "loss": 1.0452, "step": 205520 }, { "epoch": 0.52, "learning_rate": 4.855176405455735e-05, "loss": 1.0465, "step": 205525 }, { "epoch": 0.52, "learning_rate": 4.855050581307565e-05, "loss": 1.0429, "step": 205530 }, { "epoch": 0.52, "learning_rate": 4.854924757159394e-05, "loss": 1.0428, "step": 205535 }, { "epoch": 0.52, "learning_rate": 4.854798933011224e-05, "loss": 1.045, "step": 205540 }, { "epoch": 0.52, "learning_rate": 4.854673108863053e-05, "loss": 1.044, "step": 205545 }, { "epoch": 0.52, "learning_rate": 4.854547284714883e-05, "loss": 1.046, "step": 205550 }, { "epoch": 0.52, "learning_rate": 4.854421460566712e-05, "loss": 1.0454, "step": 205555 }, { "epoch": 0.52, "learning_rate": 4.854295636418541e-05, "loss": 1.0622, "step": 205560 }, { "epoch": 0.52, "learning_rate": 4.854169812270371e-05, "loss": 1.0412, "step": 205565 }, { "epoch": 0.52, "learning_rate": 4.8540439881222e-05, "loss": 1.0464, "step": 205570 }, { "epoch": 0.52, "learning_rate": 4.85391816397403e-05, "loss": 1.0436, "step": 205575 }, { "epoch": 0.52, "learning_rate": 4.853792339825859e-05, "loss": 1.0424, "step": 205580 }, { "epoch": 0.52, "learning_rate": 4.853666515677689e-05, "loss": 1.0447, "step": 205585 }, { "epoch": 0.52, "learning_rate": 4.853540691529518e-05, "loss": 1.0447, "step": 205590 }, { "epoch": 0.52, "learning_rate": 4.853414867381348e-05, "loss": 1.046, "step": 205595 }, { "epoch": 0.52, "learning_rate": 4.853289043233177e-05, "loss": 1.0471, "step": 205600 }, { "epoch": 0.52, "learning_rate": 4.853163219085007e-05, "loss": 1.0411, "step": 205605 }, { "epoch": 0.52, "learning_rate": 4.853037394936836e-05, "loss": 1.0448, "step": 205610 }, { "epoch": 0.52, "learning_rate": 4.852911570788666e-05, "loss": 1.0427, "step": 205615 }, { "epoch": 0.52, "learning_rate": 4.852785746640495e-05, "loss": 1.0453, "step": 205620 }, { "epoch": 0.52, "learning_rate": 4.8526599224923244e-05, "loss": 1.0428, "step": 205625 }, { "epoch": 0.52, "learning_rate": 4.852534098344154e-05, "loss": 1.0435, "step": 205630 }, { "epoch": 0.52, "learning_rate": 4.8524082741959834e-05, "loss": 1.044, "step": 205635 }, { "epoch": 0.52, "learning_rate": 4.852282450047814e-05, "loss": 1.0412, "step": 205640 }, { "epoch": 0.52, "learning_rate": 4.852156625899643e-05, "loss": 1.0454, "step": 205645 }, { "epoch": 0.52, "learning_rate": 4.852030801751473e-05, "loss": 1.0423, "step": 205650 }, { "epoch": 0.52, "learning_rate": 4.851904977603302e-05, "loss": 1.0447, "step": 205655 }, { "epoch": 0.52, "learning_rate": 4.851779153455131e-05, "loss": 1.0408, "step": 205660 }, { "epoch": 0.52, "learning_rate": 4.851653329306961e-05, "loss": 1.0404, "step": 205665 }, { "epoch": 0.52, "learning_rate": 4.85152750515879e-05, "loss": 1.0417, "step": 205670 }, { "epoch": 0.52, "learning_rate": 4.85140168101062e-05, "loss": 1.0448, "step": 205675 }, { "epoch": 0.52, "learning_rate": 4.851275856862449e-05, "loss": 1.0446, "step": 205680 }, { "epoch": 0.52, "learning_rate": 4.851150032714279e-05, "loss": 1.0456, "step": 205685 }, { "epoch": 0.52, "learning_rate": 4.851024208566108e-05, "loss": 1.044, "step": 205690 }, { "epoch": 0.52, "learning_rate": 4.850898384417938e-05, "loss": 1.045, "step": 205695 }, { "epoch": 0.52, "learning_rate": 4.850772560269767e-05, "loss": 1.0451, "step": 205700 }, { "epoch": 0.52, "learning_rate": 4.850646736121597e-05, "loss": 1.0442, "step": 205705 }, { "epoch": 0.52, "learning_rate": 4.850520911973426e-05, "loss": 1.0454, "step": 205710 }, { "epoch": 0.52, "learning_rate": 4.850395087825256e-05, "loss": 1.0476, "step": 205715 }, { "epoch": 0.52, "learning_rate": 4.850269263677085e-05, "loss": 1.0437, "step": 205720 }, { "epoch": 0.52, "learning_rate": 4.8501434395289144e-05, "loss": 1.044, "step": 205725 }, { "epoch": 0.52, "learning_rate": 4.850017615380744e-05, "loss": 1.0441, "step": 205730 }, { "epoch": 0.52, "learning_rate": 4.8498917912325734e-05, "loss": 1.0915, "step": 205735 }, { "epoch": 0.52, "learning_rate": 4.849765967084403e-05, "loss": 1.0426, "step": 205740 }, { "epoch": 0.52, "learning_rate": 4.8496401429362324e-05, "loss": 1.0446, "step": 205745 }, { "epoch": 0.52, "learning_rate": 4.849514318788062e-05, "loss": 1.0493, "step": 205750 }, { "epoch": 0.52, "learning_rate": 4.8493884946398914e-05, "loss": 1.0435, "step": 205755 }, { "epoch": 0.52, "learning_rate": 4.849262670491721e-05, "loss": 1.0463, "step": 205760 }, { "epoch": 0.52, "learning_rate": 4.8491368463435504e-05, "loss": 1.0432, "step": 205765 }, { "epoch": 0.52, "learning_rate": 4.84901102219538e-05, "loss": 1.0431, "step": 205770 }, { "epoch": 0.52, "learning_rate": 4.8488851980472094e-05, "loss": 1.0432, "step": 205775 }, { "epoch": 0.52, "learning_rate": 4.848759373899039e-05, "loss": 1.0428, "step": 205780 }, { "epoch": 0.52, "learning_rate": 4.8486335497508684e-05, "loss": 1.0442, "step": 205785 }, { "epoch": 0.52, "learning_rate": 4.8485077256026975e-05, "loss": 1.0445, "step": 205790 }, { "epoch": 0.52, "learning_rate": 4.8483819014545274e-05, "loss": 1.0448, "step": 205795 }, { "epoch": 0.52, "learning_rate": 4.8482560773063565e-05, "loss": 1.0439, "step": 205800 }, { "epoch": 0.52, "learning_rate": 4.8481302531581864e-05, "loss": 1.0634, "step": 205805 }, { "epoch": 0.52, "learning_rate": 4.8480044290100155e-05, "loss": 1.0427, "step": 205810 }, { "epoch": 0.52, "learning_rate": 4.8478786048618454e-05, "loss": 1.0444, "step": 205815 }, { "epoch": 0.52, "learning_rate": 4.8477527807136745e-05, "loss": 1.0427, "step": 205820 }, { "epoch": 0.52, "learning_rate": 4.8476269565655044e-05, "loss": 1.0443, "step": 205825 }, { "epoch": 0.52, "learning_rate": 4.8475011324173335e-05, "loss": 1.0402, "step": 205830 }, { "epoch": 0.52, "learning_rate": 4.8473753082691634e-05, "loss": 1.0466, "step": 205835 }, { "epoch": 0.52, "learning_rate": 4.8472494841209925e-05, "loss": 1.0452, "step": 205840 }, { "epoch": 0.52, "learning_rate": 4.8471236599728224e-05, "loss": 1.0655, "step": 205845 }, { "epoch": 0.52, "learning_rate": 4.8469978358246515e-05, "loss": 1.0461, "step": 205850 }, { "epoch": 0.52, "learning_rate": 4.846872011676481e-05, "loss": 1.0416, "step": 205855 }, { "epoch": 0.52, "learning_rate": 4.8467461875283105e-05, "loss": 1.0483, "step": 205860 }, { "epoch": 0.52, "learning_rate": 4.84662036338014e-05, "loss": 1.0453, "step": 205865 }, { "epoch": 0.52, "learning_rate": 4.8464945392319695e-05, "loss": 1.0428, "step": 205870 }, { "epoch": 0.52, "learning_rate": 4.846368715083799e-05, "loss": 1.0437, "step": 205875 }, { "epoch": 0.52, "learning_rate": 4.8462428909356285e-05, "loss": 1.0454, "step": 205880 }, { "epoch": 0.52, "learning_rate": 4.846117066787458e-05, "loss": 1.041, "step": 205885 }, { "epoch": 0.52, "learning_rate": 4.8459912426392875e-05, "loss": 1.0424, "step": 205890 }, { "epoch": 0.52, "learning_rate": 4.845865418491117e-05, "loss": 1.0451, "step": 205895 }, { "epoch": 0.52, "learning_rate": 4.8457395943429465e-05, "loss": 1.0456, "step": 205900 }, { "epoch": 0.52, "learning_rate": 4.845613770194776e-05, "loss": 1.0646, "step": 205905 }, { "epoch": 0.52, "learning_rate": 4.8454879460466055e-05, "loss": 1.0445, "step": 205910 }, { "epoch": 0.52, "learning_rate": 4.845362121898435e-05, "loss": 1.0444, "step": 205915 }, { "epoch": 0.52, "learning_rate": 4.8452362977502645e-05, "loss": 1.0478, "step": 205920 }, { "epoch": 0.52, "learning_rate": 4.845110473602094e-05, "loss": 1.0459, "step": 205925 }, { "epoch": 0.52, "learning_rate": 4.8449846494539235e-05, "loss": 1.0469, "step": 205930 }, { "epoch": 0.52, "learning_rate": 4.844858825305753e-05, "loss": 1.0565, "step": 205935 }, { "epoch": 0.52, "learning_rate": 4.8447330011575825e-05, "loss": 1.0469, "step": 205940 }, { "epoch": 0.52, "learning_rate": 4.844607177009412e-05, "loss": 1.0676, "step": 205945 }, { "epoch": 0.52, "learning_rate": 4.8444813528612415e-05, "loss": 1.0467, "step": 205950 }, { "epoch": 0.52, "learning_rate": 4.8443555287130706e-05, "loss": 1.0429, "step": 205955 }, { "epoch": 0.52, "learning_rate": 4.8442297045649005e-05, "loss": 1.045, "step": 205960 }, { "epoch": 0.52, "learning_rate": 4.8441038804167296e-05, "loss": 1.0448, "step": 205965 }, { "epoch": 0.52, "learning_rate": 4.8439780562685595e-05, "loss": 1.0466, "step": 205970 }, { "epoch": 0.52, "learning_rate": 4.8438522321203886e-05, "loss": 1.0449, "step": 205975 }, { "epoch": 0.52, "learning_rate": 4.8437264079722185e-05, "loss": 1.0435, "step": 205980 }, { "epoch": 0.52, "learning_rate": 4.8436005838240476e-05, "loss": 1.0447, "step": 205985 }, { "epoch": 0.52, "learning_rate": 4.8434747596758775e-05, "loss": 1.0435, "step": 205990 }, { "epoch": 0.52, "learning_rate": 4.8433489355277066e-05, "loss": 1.0452, "step": 205995 }, { "epoch": 0.52, "learning_rate": 4.8432231113795365e-05, "loss": 1.0443, "step": 206000 }, { "epoch": 0.52, "learning_rate": 4.8430972872313656e-05, "loss": 1.0464, "step": 206005 }, { "epoch": 0.52, "learning_rate": 4.8429714630831955e-05, "loss": 1.0451, "step": 206010 }, { "epoch": 0.52, "learning_rate": 4.8428456389350246e-05, "loss": 1.0479, "step": 206015 }, { "epoch": 0.52, "learning_rate": 4.842719814786854e-05, "loss": 1.0449, "step": 206020 }, { "epoch": 0.52, "learning_rate": 4.8425939906386836e-05, "loss": 1.0419, "step": 206025 }, { "epoch": 0.52, "learning_rate": 4.842468166490513e-05, "loss": 1.0455, "step": 206030 }, { "epoch": 0.52, "learning_rate": 4.8423423423423426e-05, "loss": 1.0452, "step": 206035 }, { "epoch": 0.52, "learning_rate": 4.842216518194172e-05, "loss": 1.0445, "step": 206040 }, { "epoch": 0.52, "learning_rate": 4.8420906940460016e-05, "loss": 1.0452, "step": 206045 }, { "epoch": 0.52, "learning_rate": 4.841964869897831e-05, "loss": 1.0431, "step": 206050 }, { "epoch": 0.52, "learning_rate": 4.8418390457496606e-05, "loss": 1.0428, "step": 206055 }, { "epoch": 0.52, "learning_rate": 4.84171322160149e-05, "loss": 1.0483, "step": 206060 }, { "epoch": 0.52, "learning_rate": 4.8415873974533196e-05, "loss": 1.0595, "step": 206065 }, { "epoch": 0.52, "learning_rate": 4.841461573305149e-05, "loss": 1.0442, "step": 206070 }, { "epoch": 0.52, "learning_rate": 4.8413357491569786e-05, "loss": 1.0462, "step": 206075 }, { "epoch": 0.52, "learning_rate": 4.841209925008808e-05, "loss": 1.0419, "step": 206080 }, { "epoch": 0.52, "learning_rate": 4.841084100860637e-05, "loss": 1.0476, "step": 206085 }, { "epoch": 0.52, "learning_rate": 4.840958276712467e-05, "loss": 1.0422, "step": 206090 }, { "epoch": 0.52, "learning_rate": 4.840832452564296e-05, "loss": 1.0447, "step": 206095 }, { "epoch": 0.52, "learning_rate": 4.840706628416126e-05, "loss": 1.0412, "step": 206100 }, { "epoch": 0.52, "learning_rate": 4.840580804267955e-05, "loss": 1.0441, "step": 206105 }, { "epoch": 0.52, "learning_rate": 4.840454980119785e-05, "loss": 1.045, "step": 206110 }, { "epoch": 0.52, "learning_rate": 4.840329155971614e-05, "loss": 1.0457, "step": 206115 }, { "epoch": 0.52, "learning_rate": 4.840203331823444e-05, "loss": 1.0432, "step": 206120 }, { "epoch": 0.52, "learning_rate": 4.840077507675273e-05, "loss": 1.0448, "step": 206125 }, { "epoch": 0.52, "learning_rate": 4.839951683527103e-05, "loss": 1.0478, "step": 206130 }, { "epoch": 0.52, "learning_rate": 4.839825859378932e-05, "loss": 1.0465, "step": 206135 }, { "epoch": 0.52, "learning_rate": 4.839700035230762e-05, "loss": 1.0469, "step": 206140 }, { "epoch": 0.52, "learning_rate": 4.8395742110825916e-05, "loss": 1.0643, "step": 206145 }, { "epoch": 0.52, "learning_rate": 4.839448386934421e-05, "loss": 1.0438, "step": 206150 }, { "epoch": 0.52, "learning_rate": 4.8393225627862506e-05, "loss": 1.0417, "step": 206155 }, { "epoch": 0.52, "learning_rate": 4.83919673863808e-05, "loss": 1.0398, "step": 206160 }, { "epoch": 0.52, "learning_rate": 4.8390709144899096e-05, "loss": 1.0461, "step": 206165 }, { "epoch": 0.52, "learning_rate": 4.838945090341739e-05, "loss": 1.0458, "step": 206170 }, { "epoch": 0.52, "learning_rate": 4.8388192661935686e-05, "loss": 1.0443, "step": 206175 }, { "epoch": 0.52, "learning_rate": 4.838693442045398e-05, "loss": 1.0421, "step": 206180 }, { "epoch": 0.52, "learning_rate": 4.838567617897227e-05, "loss": 1.0472, "step": 206185 }, { "epoch": 0.52, "learning_rate": 4.838441793749057e-05, "loss": 1.0434, "step": 206190 }, { "epoch": 0.52, "learning_rate": 4.838315969600886e-05, "loss": 1.0471, "step": 206195 }, { "epoch": 0.52, "learning_rate": 4.838190145452716e-05, "loss": 1.0426, "step": 206200 }, { "epoch": 0.52, "learning_rate": 4.838064321304545e-05, "loss": 1.0488, "step": 206205 }, { "epoch": 0.52, "learning_rate": 4.837938497156375e-05, "loss": 1.0466, "step": 206210 }, { "epoch": 0.52, "learning_rate": 4.837812673008204e-05, "loss": 1.0701, "step": 206215 }, { "epoch": 0.52, "learning_rate": 4.837686848860034e-05, "loss": 1.0431, "step": 206220 }, { "epoch": 0.52, "learning_rate": 4.837561024711863e-05, "loss": 1.0463, "step": 206225 }, { "epoch": 0.52, "learning_rate": 4.837435200563693e-05, "loss": 1.0457, "step": 206230 }, { "epoch": 0.52, "learning_rate": 4.837309376415522e-05, "loss": 1.0667, "step": 206235 }, { "epoch": 0.52, "learning_rate": 4.837183552267352e-05, "loss": 1.0423, "step": 206240 }, { "epoch": 0.52, "learning_rate": 4.837057728119181e-05, "loss": 1.0454, "step": 206245 }, { "epoch": 0.52, "learning_rate": 4.83693190397101e-05, "loss": 1.0449, "step": 206250 }, { "epoch": 0.52, "learning_rate": 4.83680607982284e-05, "loss": 1.0443, "step": 206255 }, { "epoch": 0.52, "learning_rate": 4.836680255674669e-05, "loss": 1.0416, "step": 206260 }, { "epoch": 0.52, "learning_rate": 4.836554431526499e-05, "loss": 1.0449, "step": 206265 }, { "epoch": 0.52, "learning_rate": 4.836428607378328e-05, "loss": 1.0451, "step": 206270 }, { "epoch": 0.52, "learning_rate": 4.836302783230158e-05, "loss": 1.0435, "step": 206275 }, { "epoch": 0.52, "learning_rate": 4.836176959081987e-05, "loss": 1.0452, "step": 206280 }, { "epoch": 0.52, "learning_rate": 4.836051134933817e-05, "loss": 1.0514, "step": 206285 }, { "epoch": 0.52, "learning_rate": 4.835925310785646e-05, "loss": 1.0425, "step": 206290 }, { "epoch": 0.52, "learning_rate": 4.835799486637476e-05, "loss": 1.0424, "step": 206295 }, { "epoch": 0.52, "learning_rate": 4.835673662489305e-05, "loss": 1.0441, "step": 206300 }, { "epoch": 0.52, "learning_rate": 4.835547838341135e-05, "loss": 1.0421, "step": 206305 }, { "epoch": 0.52, "learning_rate": 4.835422014192964e-05, "loss": 1.0439, "step": 206310 }, { "epoch": 0.52, "learning_rate": 4.835296190044793e-05, "loss": 1.0448, "step": 206315 }, { "epoch": 0.52, "learning_rate": 4.835170365896623e-05, "loss": 1.0428, "step": 206320 }, { "epoch": 0.52, "learning_rate": 4.835044541748452e-05, "loss": 1.0403, "step": 206325 }, { "epoch": 0.52, "learning_rate": 4.834918717600282e-05, "loss": 1.046, "step": 206330 }, { "epoch": 0.52, "learning_rate": 4.834792893452111e-05, "loss": 1.0415, "step": 206335 }, { "epoch": 0.52, "learning_rate": 4.834667069303941e-05, "loss": 1.045, "step": 206340 }, { "epoch": 0.52, "learning_rate": 4.83454124515577e-05, "loss": 1.0428, "step": 206345 }, { "epoch": 0.52, "learning_rate": 4.8344154210076e-05, "loss": 1.0443, "step": 206350 }, { "epoch": 0.52, "learning_rate": 4.834289596859429e-05, "loss": 1.0415, "step": 206355 }, { "epoch": 0.52, "learning_rate": 4.834163772711259e-05, "loss": 1.0443, "step": 206360 }, { "epoch": 0.52, "learning_rate": 4.834037948563088e-05, "loss": 1.0424, "step": 206365 }, { "epoch": 0.52, "learning_rate": 4.833912124414918e-05, "loss": 1.0429, "step": 206370 }, { "epoch": 0.52, "learning_rate": 4.833786300266747e-05, "loss": 1.0453, "step": 206375 }, { "epoch": 0.52, "learning_rate": 4.833660476118576e-05, "loss": 1.0449, "step": 206380 }, { "epoch": 0.52, "learning_rate": 4.833534651970407e-05, "loss": 1.0436, "step": 206385 }, { "epoch": 0.52, "learning_rate": 4.833408827822236e-05, "loss": 1.0442, "step": 206390 }, { "epoch": 0.52, "learning_rate": 4.833283003674066e-05, "loss": 1.046, "step": 206395 }, { "epoch": 0.52, "learning_rate": 4.833157179525895e-05, "loss": 1.0434, "step": 206400 }, { "epoch": 0.52, "learning_rate": 4.833031355377725e-05, "loss": 1.0465, "step": 206405 }, { "epoch": 0.52, "learning_rate": 4.832905531229554e-05, "loss": 1.0452, "step": 206410 }, { "epoch": 0.52, "learning_rate": 4.832779707081384e-05, "loss": 1.0404, "step": 206415 }, { "epoch": 0.52, "learning_rate": 4.832653882933213e-05, "loss": 1.044, "step": 206420 }, { "epoch": 0.52, "learning_rate": 4.832528058785042e-05, "loss": 1.0465, "step": 206425 }, { "epoch": 0.52, "learning_rate": 4.832402234636872e-05, "loss": 1.0432, "step": 206430 }, { "epoch": 0.52, "learning_rate": 4.832276410488701e-05, "loss": 1.0435, "step": 206435 }, { "epoch": 0.52, "learning_rate": 4.832150586340531e-05, "loss": 1.0437, "step": 206440 }, { "epoch": 0.52, "learning_rate": 4.83202476219236e-05, "loss": 1.044, "step": 206445 }, { "epoch": 0.52, "learning_rate": 4.83189893804419e-05, "loss": 1.0455, "step": 206450 }, { "epoch": 0.52, "learning_rate": 4.831773113896019e-05, "loss": 1.0444, "step": 206455 }, { "epoch": 0.52, "learning_rate": 4.831647289747849e-05, "loss": 1.0426, "step": 206460 }, { "epoch": 0.52, "learning_rate": 4.831521465599678e-05, "loss": 1.0806, "step": 206465 }, { "epoch": 0.52, "learning_rate": 4.831395641451508e-05, "loss": 1.0437, "step": 206470 }, { "epoch": 0.52, "learning_rate": 4.831269817303337e-05, "loss": 1.0452, "step": 206475 }, { "epoch": 0.52, "learning_rate": 4.831143993155167e-05, "loss": 1.0433, "step": 206480 }, { "epoch": 0.52, "learning_rate": 4.831018169006996e-05, "loss": 1.0432, "step": 206485 }, { "epoch": 0.52, "learning_rate": 4.830892344858825e-05, "loss": 1.0487, "step": 206490 }, { "epoch": 0.52, "learning_rate": 4.830766520710655e-05, "loss": 1.045, "step": 206495 }, { "epoch": 0.52, "learning_rate": 4.830640696562484e-05, "loss": 1.0462, "step": 206500 }, { "epoch": 0.52, "learning_rate": 4.830514872414314e-05, "loss": 1.0449, "step": 206505 }, { "epoch": 0.52, "learning_rate": 4.830389048266143e-05, "loss": 1.0439, "step": 206510 }, { "epoch": 0.52, "learning_rate": 4.830263224117973e-05, "loss": 1.047, "step": 206515 }, { "epoch": 0.52, "learning_rate": 4.830137399969802e-05, "loss": 1.0441, "step": 206520 }, { "epoch": 0.52, "learning_rate": 4.830011575821632e-05, "loss": 1.0433, "step": 206525 }, { "epoch": 0.52, "learning_rate": 4.829885751673461e-05, "loss": 1.0429, "step": 206530 }, { "epoch": 0.52, "learning_rate": 4.829759927525291e-05, "loss": 1.0465, "step": 206535 }, { "epoch": 0.52, "learning_rate": 4.82963410337712e-05, "loss": 1.0454, "step": 206540 }, { "epoch": 0.52, "learning_rate": 4.8295082792289494e-05, "loss": 1.0423, "step": 206545 }, { "epoch": 0.52, "learning_rate": 4.829382455080779e-05, "loss": 1.0432, "step": 206550 }, { "epoch": 0.52, "learning_rate": 4.8292566309326084e-05, "loss": 1.0453, "step": 206555 }, { "epoch": 0.52, "learning_rate": 4.829130806784438e-05, "loss": 1.0414, "step": 206560 }, { "epoch": 0.52, "learning_rate": 4.8290049826362674e-05, "loss": 1.0467, "step": 206565 }, { "epoch": 0.52, "learning_rate": 4.828879158488097e-05, "loss": 1.0473, "step": 206570 }, { "epoch": 0.52, "learning_rate": 4.8287533343399264e-05, "loss": 1.0434, "step": 206575 }, { "epoch": 0.52, "learning_rate": 4.828627510191756e-05, "loss": 1.0427, "step": 206580 }, { "epoch": 0.52, "learning_rate": 4.8285016860435854e-05, "loss": 1.0437, "step": 206585 }, { "epoch": 0.52, "learning_rate": 4.828375861895415e-05, "loss": 1.0442, "step": 206590 }, { "epoch": 0.52, "learning_rate": 4.8282500377472444e-05, "loss": 1.0433, "step": 206595 }, { "epoch": 0.52, "learning_rate": 4.828124213599074e-05, "loss": 1.0458, "step": 206600 }, { "epoch": 0.52, "learning_rate": 4.8279983894509034e-05, "loss": 1.0458, "step": 206605 }, { "epoch": 0.52, "learning_rate": 4.8278725653027326e-05, "loss": 1.0421, "step": 206610 }, { "epoch": 0.52, "learning_rate": 4.8277467411545624e-05, "loss": 1.0449, "step": 206615 }, { "epoch": 0.52, "learning_rate": 4.8276209170063916e-05, "loss": 1.0469, "step": 206620 }, { "epoch": 0.52, "learning_rate": 4.8274950928582214e-05, "loss": 1.0458, "step": 206625 }, { "epoch": 0.52, "learning_rate": 4.8273692687100506e-05, "loss": 1.0453, "step": 206630 }, { "epoch": 0.52, "learning_rate": 4.827243444561881e-05, "loss": 1.0464, "step": 206635 }, { "epoch": 0.52, "learning_rate": 4.82711762041371e-05, "loss": 1.0446, "step": 206640 }, { "epoch": 0.52, "learning_rate": 4.82699179626554e-05, "loss": 1.0443, "step": 206645 }, { "epoch": 0.52, "learning_rate": 4.826865972117369e-05, "loss": 1.0416, "step": 206650 }, { "epoch": 0.52, "learning_rate": 4.8267401479691984e-05, "loss": 1.0449, "step": 206655 }, { "epoch": 0.52, "learning_rate": 4.826614323821028e-05, "loss": 1.0421, "step": 206660 }, { "epoch": 0.52, "learning_rate": 4.8264884996728574e-05, "loss": 1.0444, "step": 206665 }, { "epoch": 0.52, "learning_rate": 4.826362675524687e-05, "loss": 1.043, "step": 206670 }, { "epoch": 0.52, "learning_rate": 4.8262368513765164e-05, "loss": 1.0429, "step": 206675 }, { "epoch": 0.52, "learning_rate": 4.826111027228346e-05, "loss": 1.0456, "step": 206680 }, { "epoch": 0.52, "learning_rate": 4.8259852030801754e-05, "loss": 1.045, "step": 206685 }, { "epoch": 0.52, "learning_rate": 4.825859378932005e-05, "loss": 1.0445, "step": 206690 }, { "epoch": 0.52, "learning_rate": 4.8257335547838344e-05, "loss": 1.044, "step": 206695 }, { "epoch": 0.52, "learning_rate": 4.825607730635664e-05, "loss": 1.0429, "step": 206700 }, { "epoch": 0.52, "learning_rate": 4.8254819064874934e-05, "loss": 1.0438, "step": 206705 }, { "epoch": 0.52, "learning_rate": 4.825356082339323e-05, "loss": 1.0418, "step": 206710 }, { "epoch": 0.52, "learning_rate": 4.8252302581911524e-05, "loss": 1.0419, "step": 206715 }, { "epoch": 0.52, "learning_rate": 4.8251044340429816e-05, "loss": 1.0457, "step": 206720 }, { "epoch": 0.52, "learning_rate": 4.8249786098948114e-05, "loss": 1.0472, "step": 206725 }, { "epoch": 0.52, "learning_rate": 4.8248527857466405e-05, "loss": 1.0438, "step": 206730 }, { "epoch": 0.52, "learning_rate": 4.8247269615984704e-05, "loss": 1.0641, "step": 206735 }, { "epoch": 0.52, "learning_rate": 4.8246011374502995e-05, "loss": 1.0622, "step": 206740 }, { "epoch": 0.52, "learning_rate": 4.8244753133021294e-05, "loss": 1.0443, "step": 206745 }, { "epoch": 0.52, "learning_rate": 4.8243494891539585e-05, "loss": 1.0472, "step": 206750 }, { "epoch": 0.52, "learning_rate": 4.8242236650057884e-05, "loss": 1.048, "step": 206755 }, { "epoch": 0.52, "learning_rate": 4.8240978408576175e-05, "loss": 1.0418, "step": 206760 }, { "epoch": 0.52, "learning_rate": 4.8239720167094474e-05, "loss": 1.0479, "step": 206765 }, { "epoch": 0.52, "learning_rate": 4.8238461925612765e-05, "loss": 1.0418, "step": 206770 }, { "epoch": 0.52, "learning_rate": 4.8237203684131064e-05, "loss": 1.0471, "step": 206775 }, { "epoch": 0.52, "learning_rate": 4.8235945442649355e-05, "loss": 1.0455, "step": 206780 }, { "epoch": 0.52, "learning_rate": 4.823468720116765e-05, "loss": 1.0445, "step": 206785 }, { "epoch": 0.52, "learning_rate": 4.8233428959685945e-05, "loss": 1.045, "step": 206790 }, { "epoch": 0.52, "learning_rate": 4.823217071820424e-05, "loss": 1.044, "step": 206795 }, { "epoch": 0.52, "learning_rate": 4.8230912476722535e-05, "loss": 1.0444, "step": 206800 }, { "epoch": 0.52, "learning_rate": 4.822965423524083e-05, "loss": 1.0457, "step": 206805 }, { "epoch": 0.52, "learning_rate": 4.8228395993759125e-05, "loss": 1.0451, "step": 206810 }, { "epoch": 0.52, "learning_rate": 4.822713775227742e-05, "loss": 1.0438, "step": 206815 }, { "epoch": 0.52, "learning_rate": 4.8225879510795715e-05, "loss": 1.0438, "step": 206820 }, { "epoch": 0.52, "learning_rate": 4.822462126931401e-05, "loss": 1.0445, "step": 206825 }, { "epoch": 0.52, "learning_rate": 4.8223363027832305e-05, "loss": 1.0458, "step": 206830 }, { "epoch": 0.52, "learning_rate": 4.82221047863506e-05, "loss": 1.0449, "step": 206835 }, { "epoch": 0.52, "learning_rate": 4.822084654486889e-05, "loss": 1.0475, "step": 206840 }, { "epoch": 0.52, "learning_rate": 4.821958830338719e-05, "loss": 1.046, "step": 206845 }, { "epoch": 0.52, "learning_rate": 4.821833006190548e-05, "loss": 1.0446, "step": 206850 }, { "epoch": 0.52, "learning_rate": 4.821707182042378e-05, "loss": 1.0427, "step": 206855 }, { "epoch": 0.52, "learning_rate": 4.821581357894207e-05, "loss": 1.0462, "step": 206860 }, { "epoch": 0.52, "learning_rate": 4.821455533746037e-05, "loss": 1.0471, "step": 206865 }, { "epoch": 0.52, "learning_rate": 4.821329709597866e-05, "loss": 1.0419, "step": 206870 }, { "epoch": 0.52, "learning_rate": 4.821203885449696e-05, "loss": 1.0457, "step": 206875 }, { "epoch": 0.52, "learning_rate": 4.821078061301525e-05, "loss": 1.046, "step": 206880 }, { "epoch": 0.52, "learning_rate": 4.820952237153355e-05, "loss": 1.0471, "step": 206885 }, { "epoch": 0.52, "learning_rate": 4.8208264130051845e-05, "loss": 1.0465, "step": 206890 }, { "epoch": 0.52, "learning_rate": 4.820700588857014e-05, "loss": 1.0501, "step": 206895 }, { "epoch": 0.52, "learning_rate": 4.8205747647088435e-05, "loss": 1.0469, "step": 206900 }, { "epoch": 0.52, "learning_rate": 4.8204489405606727e-05, "loss": 1.0608, "step": 206905 }, { "epoch": 0.52, "learning_rate": 4.8203231164125025e-05, "loss": 1.0475, "step": 206910 }, { "epoch": 0.52, "learning_rate": 4.8201972922643317e-05, "loss": 1.0472, "step": 206915 }, { "epoch": 0.52, "learning_rate": 4.8200714681161615e-05, "loss": 1.043, "step": 206920 }, { "epoch": 0.52, "learning_rate": 4.8199456439679907e-05, "loss": 1.0449, "step": 206925 }, { "epoch": 0.52, "learning_rate": 4.8198198198198205e-05, "loss": 1.0456, "step": 206930 }, { "epoch": 0.52, "learning_rate": 4.8196939956716497e-05, "loss": 1.0454, "step": 206935 }, { "epoch": 0.52, "learning_rate": 4.8195681715234795e-05, "loss": 1.0426, "step": 206940 }, { "epoch": 0.52, "learning_rate": 4.8194423473753086e-05, "loss": 1.0466, "step": 206945 }, { "epoch": 0.52, "learning_rate": 4.819316523227138e-05, "loss": 1.0438, "step": 206950 }, { "epoch": 0.52, "learning_rate": 4.8191906990789676e-05, "loss": 1.046, "step": 206955 }, { "epoch": 0.52, "learning_rate": 4.819064874930797e-05, "loss": 1.0439, "step": 206960 }, { "epoch": 0.52, "learning_rate": 4.8189390507826266e-05, "loss": 1.0479, "step": 206965 }, { "epoch": 0.52, "learning_rate": 4.818813226634456e-05, "loss": 1.042, "step": 206970 }, { "epoch": 0.52, "learning_rate": 4.8186874024862856e-05, "loss": 1.047, "step": 206975 }, { "epoch": 0.52, "learning_rate": 4.818561578338115e-05, "loss": 1.0434, "step": 206980 }, { "epoch": 0.52, "learning_rate": 4.8184357541899446e-05, "loss": 1.0433, "step": 206985 }, { "epoch": 0.52, "learning_rate": 4.818309930041774e-05, "loss": 1.0571, "step": 206990 }, { "epoch": 0.52, "learning_rate": 4.8181841058936036e-05, "loss": 1.041, "step": 206995 }, { "epoch": 0.52, "learning_rate": 4.818058281745433e-05, "loss": 1.0447, "step": 207000 }, { "epoch": 0.52, "learning_rate": 4.8179324575972626e-05, "loss": 1.0421, "step": 207005 }, { "epoch": 0.52, "learning_rate": 4.817806633449092e-05, "loss": 1.0444, "step": 207010 }, { "epoch": 0.52, "learning_rate": 4.817680809300921e-05, "loss": 1.0435, "step": 207015 }, { "epoch": 0.52, "learning_rate": 4.817554985152751e-05, "loss": 1.0433, "step": 207020 }, { "epoch": 0.52, "learning_rate": 4.81742916100458e-05, "loss": 1.0452, "step": 207025 }, { "epoch": 0.52, "learning_rate": 4.81730333685641e-05, "loss": 1.0447, "step": 207030 }, { "epoch": 0.52, "learning_rate": 4.817177512708239e-05, "loss": 1.0453, "step": 207035 }, { "epoch": 0.52, "learning_rate": 4.817051688560069e-05, "loss": 1.0449, "step": 207040 }, { "epoch": 0.52, "learning_rate": 4.816925864411898e-05, "loss": 1.0429, "step": 207045 }, { "epoch": 0.52, "learning_rate": 4.816800040263728e-05, "loss": 1.0433, "step": 207050 }, { "epoch": 0.52, "learning_rate": 4.816674216115557e-05, "loss": 1.0443, "step": 207055 }, { "epoch": 0.52, "learning_rate": 4.816548391967387e-05, "loss": 1.0447, "step": 207060 }, { "epoch": 0.52, "learning_rate": 4.816422567819216e-05, "loss": 1.0437, "step": 207065 }, { "epoch": 0.52, "learning_rate": 4.816296743671046e-05, "loss": 1.0473, "step": 207070 }, { "epoch": 0.52, "learning_rate": 4.816170919522875e-05, "loss": 1.0455, "step": 207075 }, { "epoch": 0.52, "learning_rate": 4.816045095374704e-05, "loss": 1.046, "step": 207080 }, { "epoch": 0.52, "learning_rate": 4.815919271226534e-05, "loss": 1.0438, "step": 207085 }, { "epoch": 0.52, "learning_rate": 4.815793447078363e-05, "loss": 1.0415, "step": 207090 }, { "epoch": 0.52, "learning_rate": 4.815667622930193e-05, "loss": 1.0445, "step": 207095 }, { "epoch": 0.52, "learning_rate": 4.815541798782022e-05, "loss": 1.0474, "step": 207100 }, { "epoch": 0.52, "learning_rate": 4.815415974633852e-05, "loss": 1.0418, "step": 207105 }, { "epoch": 0.52, "learning_rate": 4.815290150485681e-05, "loss": 1.0659, "step": 207110 }, { "epoch": 0.52, "learning_rate": 4.815164326337511e-05, "loss": 1.0455, "step": 207115 }, { "epoch": 0.52, "learning_rate": 4.81503850218934e-05, "loss": 1.0445, "step": 207120 }, { "epoch": 0.52, "learning_rate": 4.81491267804117e-05, "loss": 1.0473, "step": 207125 }, { "epoch": 0.52, "learning_rate": 4.814786853892999e-05, "loss": 1.0467, "step": 207130 }, { "epoch": 0.52, "learning_rate": 4.814661029744829e-05, "loss": 1.0453, "step": 207135 }, { "epoch": 0.52, "learning_rate": 4.814535205596659e-05, "loss": 1.0488, "step": 207140 }, { "epoch": 0.52, "learning_rate": 4.814409381448488e-05, "loss": 1.0457, "step": 207145 }, { "epoch": 0.52, "learning_rate": 4.814283557300318e-05, "loss": 1.0458, "step": 207150 }, { "epoch": 0.52, "learning_rate": 4.814157733152147e-05, "loss": 1.0458, "step": 207155 }, { "epoch": 0.52, "learning_rate": 4.814031909003977e-05, "loss": 1.0431, "step": 207160 }, { "epoch": 0.52, "learning_rate": 4.813906084855806e-05, "loss": 1.046, "step": 207165 }, { "epoch": 0.52, "learning_rate": 4.813780260707636e-05, "loss": 1.0477, "step": 207170 }, { "epoch": 0.52, "learning_rate": 4.813654436559465e-05, "loss": 1.045, "step": 207175 }, { "epoch": 0.52, "learning_rate": 4.813528612411294e-05, "loss": 1.0416, "step": 207180 }, { "epoch": 0.52, "learning_rate": 4.813402788263124e-05, "loss": 1.0475, "step": 207185 }, { "epoch": 0.52, "learning_rate": 4.813276964114953e-05, "loss": 1.0403, "step": 207190 }, { "epoch": 0.52, "learning_rate": 4.813151139966783e-05, "loss": 1.0455, "step": 207195 }, { "epoch": 0.52, "learning_rate": 4.813025315818612e-05, "loss": 1.0453, "step": 207200 }, { "epoch": 0.52, "learning_rate": 4.812899491670442e-05, "loss": 1.0464, "step": 207205 }, { "epoch": 0.52, "learning_rate": 4.812773667522271e-05, "loss": 1.0483, "step": 207210 }, { "epoch": 0.52, "learning_rate": 4.812647843374101e-05, "loss": 1.0442, "step": 207215 }, { "epoch": 0.52, "learning_rate": 4.81252201922593e-05, "loss": 1.0471, "step": 207220 }, { "epoch": 0.52, "learning_rate": 4.81239619507776e-05, "loss": 1.0445, "step": 207225 }, { "epoch": 0.52, "learning_rate": 4.812270370929589e-05, "loss": 1.0465, "step": 207230 }, { "epoch": 0.52, "learning_rate": 4.812144546781419e-05, "loss": 1.0432, "step": 207235 }, { "epoch": 0.52, "learning_rate": 4.812018722633248e-05, "loss": 1.0427, "step": 207240 }, { "epoch": 0.52, "learning_rate": 4.811892898485077e-05, "loss": 1.0448, "step": 207245 }, { "epoch": 0.52, "learning_rate": 4.811767074336907e-05, "loss": 1.0462, "step": 207250 }, { "epoch": 0.52, "learning_rate": 4.811641250188736e-05, "loss": 1.0421, "step": 207255 }, { "epoch": 0.52, "learning_rate": 4.811515426040566e-05, "loss": 1.0445, "step": 207260 }, { "epoch": 0.52, "learning_rate": 4.811389601892395e-05, "loss": 1.0445, "step": 207265 }, { "epoch": 0.52, "learning_rate": 4.811263777744225e-05, "loss": 1.0448, "step": 207270 }, { "epoch": 0.52, "learning_rate": 4.811137953596054e-05, "loss": 1.0418, "step": 207275 }, { "epoch": 0.52, "learning_rate": 4.811012129447884e-05, "loss": 1.0419, "step": 207280 }, { "epoch": 0.52, "learning_rate": 4.810886305299713e-05, "loss": 1.0447, "step": 207285 }, { "epoch": 0.52, "learning_rate": 4.810760481151543e-05, "loss": 1.0467, "step": 207290 }, { "epoch": 0.52, "learning_rate": 4.810634657003372e-05, "loss": 1.0475, "step": 207295 }, { "epoch": 0.52, "learning_rate": 4.810508832855202e-05, "loss": 1.0436, "step": 207300 }, { "epoch": 0.52, "learning_rate": 4.810383008707031e-05, "loss": 1.0452, "step": 207305 }, { "epoch": 0.52, "learning_rate": 4.8102571845588603e-05, "loss": 1.0403, "step": 207310 }, { "epoch": 0.52, "learning_rate": 4.81013136041069e-05, "loss": 1.0421, "step": 207315 }, { "epoch": 0.52, "learning_rate": 4.8100055362625193e-05, "loss": 1.0443, "step": 207320 }, { "epoch": 0.52, "learning_rate": 4.809879712114349e-05, "loss": 1.0455, "step": 207325 }, { "epoch": 0.52, "learning_rate": 4.8097538879661783e-05, "loss": 1.0432, "step": 207330 }, { "epoch": 0.52, "learning_rate": 4.809628063818008e-05, "loss": 1.0479, "step": 207335 }, { "epoch": 0.52, "learning_rate": 4.8095022396698373e-05, "loss": 1.0458, "step": 207340 }, { "epoch": 0.52, "learning_rate": 4.809376415521667e-05, "loss": 1.0452, "step": 207345 }, { "epoch": 0.52, "learning_rate": 4.809250591373496e-05, "loss": 1.0434, "step": 207350 }, { "epoch": 0.52, "learning_rate": 4.809124767225326e-05, "loss": 1.0448, "step": 207355 }, { "epoch": 0.52, "learning_rate": 4.808998943077155e-05, "loss": 1.0473, "step": 207360 }, { "epoch": 0.52, "learning_rate": 4.808873118928985e-05, "loss": 1.0427, "step": 207365 }, { "epoch": 0.52, "learning_rate": 4.808747294780814e-05, "loss": 1.0421, "step": 207370 }, { "epoch": 0.52, "learning_rate": 4.8086214706326435e-05, "loss": 1.0437, "step": 207375 }, { "epoch": 0.52, "learning_rate": 4.808495646484473e-05, "loss": 1.0416, "step": 207380 }, { "epoch": 0.52, "learning_rate": 4.808369822336303e-05, "loss": 1.0443, "step": 207385 }, { "epoch": 0.52, "learning_rate": 4.808243998188133e-05, "loss": 1.0457, "step": 207390 }, { "epoch": 0.52, "learning_rate": 4.808118174039962e-05, "loss": 1.043, "step": 207395 }, { "epoch": 0.52, "learning_rate": 4.807992349891792e-05, "loss": 1.0454, "step": 207400 }, { "epoch": 0.52, "learning_rate": 4.807866525743621e-05, "loss": 1.0447, "step": 207405 }, { "epoch": 0.52, "learning_rate": 4.80774070159545e-05, "loss": 1.0452, "step": 207410 }, { "epoch": 0.52, "learning_rate": 4.80761487744728e-05, "loss": 1.0457, "step": 207415 }, { "epoch": 0.52, "learning_rate": 4.807489053299109e-05, "loss": 1.0418, "step": 207420 }, { "epoch": 0.52, "learning_rate": 4.807363229150939e-05, "loss": 1.0451, "step": 207425 }, { "epoch": 0.52, "learning_rate": 4.807237405002768e-05, "loss": 1.0446, "step": 207430 }, { "epoch": 0.52, "learning_rate": 4.807111580854598e-05, "loss": 1.0649, "step": 207435 }, { "epoch": 0.52, "learning_rate": 4.806985756706427e-05, "loss": 1.0724, "step": 207440 }, { "epoch": 0.52, "learning_rate": 4.806859932558257e-05, "loss": 1.0422, "step": 207445 }, { "epoch": 0.52, "learning_rate": 4.806734108410086e-05, "loss": 1.0442, "step": 207450 }, { "epoch": 0.52, "learning_rate": 4.806608284261916e-05, "loss": 1.0439, "step": 207455 }, { "epoch": 0.52, "learning_rate": 4.806482460113745e-05, "loss": 1.0441, "step": 207460 }, { "epoch": 0.52, "learning_rate": 4.806356635965575e-05, "loss": 1.045, "step": 207465 }, { "epoch": 0.52, "learning_rate": 4.806230811817404e-05, "loss": 1.0467, "step": 207470 }, { "epoch": 0.52, "learning_rate": 4.8061049876692335e-05, "loss": 1.045, "step": 207475 }, { "epoch": 0.52, "learning_rate": 4.805979163521063e-05, "loss": 1.0456, "step": 207480 }, { "epoch": 0.52, "learning_rate": 4.8058533393728925e-05, "loss": 1.0439, "step": 207485 }, { "epoch": 0.52, "learning_rate": 4.805727515224722e-05, "loss": 1.0445, "step": 207490 }, { "epoch": 0.52, "learning_rate": 4.8056016910765515e-05, "loss": 1.0466, "step": 207495 }, { "epoch": 0.52, "learning_rate": 4.805475866928381e-05, "loss": 1.0462, "step": 207500 }, { "epoch": 0.52, "learning_rate": 4.8053500427802105e-05, "loss": 1.0925, "step": 207505 }, { "epoch": 0.52, "learning_rate": 4.80522421863204e-05, "loss": 1.044, "step": 207510 }, { "epoch": 0.52, "learning_rate": 4.8050983944838694e-05, "loss": 1.0426, "step": 207515 }, { "epoch": 0.52, "learning_rate": 4.804972570335699e-05, "loss": 1.0444, "step": 207520 }, { "epoch": 0.52, "learning_rate": 4.8048467461875284e-05, "loss": 1.0414, "step": 207525 }, { "epoch": 0.52, "learning_rate": 4.804720922039358e-05, "loss": 1.0431, "step": 207530 }, { "epoch": 0.52, "learning_rate": 4.8045950978911874e-05, "loss": 1.0453, "step": 207535 }, { "epoch": 0.52, "learning_rate": 4.8044692737430166e-05, "loss": 1.0458, "step": 207540 }, { "epoch": 0.52, "learning_rate": 4.8043434495948464e-05, "loss": 1.0433, "step": 207545 }, { "epoch": 0.52, "learning_rate": 4.8042176254466756e-05, "loss": 1.0429, "step": 207550 }, { "epoch": 0.52, "learning_rate": 4.8040918012985054e-05, "loss": 1.0457, "step": 207555 }, { "epoch": 0.52, "learning_rate": 4.8039659771503346e-05, "loss": 1.0442, "step": 207560 }, { "epoch": 0.52, "learning_rate": 4.8038401530021644e-05, "loss": 1.045, "step": 207565 }, { "epoch": 0.52, "learning_rate": 4.8037143288539936e-05, "loss": 1.0434, "step": 207570 }, { "epoch": 0.52, "learning_rate": 4.8035885047058234e-05, "loss": 1.0467, "step": 207575 }, { "epoch": 0.52, "learning_rate": 4.8034626805576526e-05, "loss": 1.0445, "step": 207580 }, { "epoch": 0.52, "learning_rate": 4.8033368564094824e-05, "loss": 1.0444, "step": 207585 }, { "epoch": 0.52, "learning_rate": 4.8032110322613116e-05, "loss": 1.0488, "step": 207590 }, { "epoch": 0.52, "learning_rate": 4.8030852081131414e-05, "loss": 1.0564, "step": 207595 }, { "epoch": 0.52, "learning_rate": 4.8029593839649706e-05, "loss": 1.0468, "step": 207600 }, { "epoch": 0.52, "learning_rate": 4.8028335598168e-05, "loss": 1.0454, "step": 207605 }, { "epoch": 0.52, "learning_rate": 4.8027077356686296e-05, "loss": 1.0443, "step": 207610 }, { "epoch": 0.52, "learning_rate": 4.802581911520459e-05, "loss": 1.0472, "step": 207615 }, { "epoch": 0.52, "learning_rate": 4.8024560873722886e-05, "loss": 1.0447, "step": 207620 }, { "epoch": 0.52, "learning_rate": 4.802330263224118e-05, "loss": 1.043, "step": 207625 }, { "epoch": 0.52, "learning_rate": 4.8022044390759476e-05, "loss": 1.0455, "step": 207630 }, { "epoch": 0.52, "learning_rate": 4.8020786149277774e-05, "loss": 1.0444, "step": 207635 }, { "epoch": 0.52, "learning_rate": 4.8019527907796066e-05, "loss": 1.0449, "step": 207640 }, { "epoch": 0.52, "learning_rate": 4.8018269666314364e-05, "loss": 1.0432, "step": 207645 }, { "epoch": 0.52, "learning_rate": 4.8017011424832656e-05, "loss": 1.0438, "step": 207650 }, { "epoch": 0.52, "learning_rate": 4.8015753183350954e-05, "loss": 1.0439, "step": 207655 }, { "epoch": 0.52, "learning_rate": 4.8014494941869246e-05, "loss": 1.0446, "step": 207660 }, { "epoch": 0.52, "learning_rate": 4.8013236700387544e-05, "loss": 1.0472, "step": 207665 }, { "epoch": 0.52, "learning_rate": 4.8011978458905836e-05, "loss": 1.0419, "step": 207670 }, { "epoch": 0.52, "learning_rate": 4.8010720217424134e-05, "loss": 1.0426, "step": 207675 }, { "epoch": 0.52, "learning_rate": 4.8009461975942426e-05, "loss": 1.0465, "step": 207680 }, { "epoch": 0.52, "learning_rate": 4.8008203734460724e-05, "loss": 1.0435, "step": 207685 }, { "epoch": 0.52, "learning_rate": 4.8006945492979016e-05, "loss": 1.0449, "step": 207690 }, { "epoch": 0.52, "learning_rate": 4.8005687251497314e-05, "loss": 1.0452, "step": 207695 }, { "epoch": 0.52, "learning_rate": 4.8004429010015606e-05, "loss": 1.0436, "step": 207700 }, { "epoch": 0.52, "learning_rate": 4.80031707685339e-05, "loss": 1.045, "step": 207705 }, { "epoch": 0.52, "learning_rate": 4.8001912527052196e-05, "loss": 1.0445, "step": 207710 }, { "epoch": 0.52, "learning_rate": 4.800065428557049e-05, "loss": 1.0433, "step": 207715 }, { "epoch": 0.52, "learning_rate": 4.7999396044088786e-05, "loss": 1.0461, "step": 207720 }, { "epoch": 0.52, "learning_rate": 4.799813780260708e-05, "loss": 1.0457, "step": 207725 }, { "epoch": 0.52, "learning_rate": 4.7996879561125375e-05, "loss": 1.0472, "step": 207730 }, { "epoch": 0.52, "learning_rate": 4.799562131964367e-05, "loss": 1.0457, "step": 207735 }, { "epoch": 0.52, "learning_rate": 4.7994363078161965e-05, "loss": 1.044, "step": 207740 }, { "epoch": 0.52, "learning_rate": 4.799310483668026e-05, "loss": 1.0402, "step": 207745 }, { "epoch": 0.52, "learning_rate": 4.7991846595198555e-05, "loss": 1.0421, "step": 207750 }, { "epoch": 0.52, "learning_rate": 4.799058835371685e-05, "loss": 1.0446, "step": 207755 }, { "epoch": 0.52, "learning_rate": 4.7989330112235145e-05, "loss": 1.0477, "step": 207760 }, { "epoch": 0.52, "learning_rate": 4.798807187075344e-05, "loss": 1.0441, "step": 207765 }, { "epoch": 0.52, "learning_rate": 4.798681362927173e-05, "loss": 1.0444, "step": 207770 }, { "epoch": 0.52, "learning_rate": 4.798555538779003e-05, "loss": 1.0437, "step": 207775 }, { "epoch": 0.52, "learning_rate": 4.798429714630832e-05, "loss": 1.0441, "step": 207780 }, { "epoch": 0.52, "learning_rate": 4.798303890482662e-05, "loss": 1.0415, "step": 207785 }, { "epoch": 0.52, "learning_rate": 4.798178066334491e-05, "loss": 1.041, "step": 207790 }, { "epoch": 0.52, "learning_rate": 4.798052242186321e-05, "loss": 1.0404, "step": 207795 }, { "epoch": 0.52, "learning_rate": 4.79792641803815e-05, "loss": 1.0458, "step": 207800 }, { "epoch": 0.52, "learning_rate": 4.79780059388998e-05, "loss": 1.0421, "step": 207805 }, { "epoch": 0.52, "learning_rate": 4.797674769741809e-05, "loss": 1.0434, "step": 207810 }, { "epoch": 0.52, "learning_rate": 4.797548945593639e-05, "loss": 1.0438, "step": 207815 }, { "epoch": 0.52, "learning_rate": 4.797423121445468e-05, "loss": 1.0464, "step": 207820 }, { "epoch": 0.52, "learning_rate": 4.797297297297298e-05, "loss": 1.0468, "step": 207825 }, { "epoch": 0.52, "learning_rate": 4.797171473149127e-05, "loss": 1.0427, "step": 207830 }, { "epoch": 0.52, "learning_rate": 4.797045649000956e-05, "loss": 1.0449, "step": 207835 }, { "epoch": 0.52, "learning_rate": 4.796919824852786e-05, "loss": 1.0435, "step": 207840 }, { "epoch": 0.52, "learning_rate": 4.796794000704615e-05, "loss": 1.0458, "step": 207845 }, { "epoch": 0.52, "learning_rate": 4.796668176556445e-05, "loss": 1.0437, "step": 207850 }, { "epoch": 0.52, "learning_rate": 4.796542352408274e-05, "loss": 1.0409, "step": 207855 }, { "epoch": 0.52, "learning_rate": 4.796416528260104e-05, "loss": 1.0444, "step": 207860 }, { "epoch": 0.52, "learning_rate": 4.796290704111933e-05, "loss": 1.044, "step": 207865 }, { "epoch": 0.52, "learning_rate": 4.796164879963763e-05, "loss": 1.0442, "step": 207870 }, { "epoch": 0.52, "learning_rate": 4.796039055815592e-05, "loss": 1.0485, "step": 207875 }, { "epoch": 0.52, "learning_rate": 4.795913231667422e-05, "loss": 1.0434, "step": 207880 }, { "epoch": 0.52, "learning_rate": 4.795787407519252e-05, "loss": 1.0451, "step": 207885 }, { "epoch": 0.52, "learning_rate": 4.795661583371081e-05, "loss": 1.0418, "step": 207890 }, { "epoch": 0.52, "learning_rate": 4.795535759222911e-05, "loss": 1.0464, "step": 207895 }, { "epoch": 0.52, "learning_rate": 4.79540993507474e-05, "loss": 1.0427, "step": 207900 }, { "epoch": 0.52, "learning_rate": 4.7952841109265697e-05, "loss": 1.0438, "step": 207905 }, { "epoch": 0.52, "learning_rate": 4.795158286778399e-05, "loss": 1.0422, "step": 207910 }, { "epoch": 0.52, "learning_rate": 4.7950324626302287e-05, "loss": 1.0457, "step": 207915 }, { "epoch": 0.52, "learning_rate": 4.794906638482058e-05, "loss": 1.0478, "step": 207920 }, { "epoch": 0.52, "learning_rate": 4.7947808143338877e-05, "loss": 1.047, "step": 207925 }, { "epoch": 0.52, "learning_rate": 4.794654990185717e-05, "loss": 1.0442, "step": 207930 }, { "epoch": 0.52, "learning_rate": 4.7945291660375467e-05, "loss": 1.0451, "step": 207935 }, { "epoch": 0.52, "learning_rate": 4.794403341889376e-05, "loss": 1.0453, "step": 207940 }, { "epoch": 0.52, "learning_rate": 4.794277517741205e-05, "loss": 1.046, "step": 207945 }, { "epoch": 0.52, "learning_rate": 4.794151693593035e-05, "loss": 1.0454, "step": 207950 }, { "epoch": 0.52, "learning_rate": 4.794025869444864e-05, "loss": 1.0464, "step": 207955 }, { "epoch": 0.52, "learning_rate": 4.793900045296694e-05, "loss": 1.0427, "step": 207960 }, { "epoch": 0.52, "learning_rate": 4.793774221148523e-05, "loss": 1.0426, "step": 207965 }, { "epoch": 0.52, "learning_rate": 4.793648397000353e-05, "loss": 1.046, "step": 207970 }, { "epoch": 0.52, "learning_rate": 4.793522572852182e-05, "loss": 1.044, "step": 207975 }, { "epoch": 0.52, "learning_rate": 4.793396748704012e-05, "loss": 1.0432, "step": 207980 }, { "epoch": 0.52, "learning_rate": 4.793270924555841e-05, "loss": 1.0459, "step": 207985 }, { "epoch": 0.52, "learning_rate": 4.793145100407671e-05, "loss": 1.0446, "step": 207990 }, { "epoch": 0.52, "learning_rate": 4.7930192762595e-05, "loss": 1.0434, "step": 207995 }, { "epoch": 0.52, "learning_rate": 4.792893452111329e-05, "loss": 1.0454, "step": 208000 }, { "epoch": 0.52, "learning_rate": 4.792767627963159e-05, "loss": 1.0435, "step": 208005 }, { "epoch": 0.52, "learning_rate": 4.792641803814988e-05, "loss": 1.0443, "step": 208010 }, { "epoch": 0.52, "learning_rate": 4.792515979666818e-05, "loss": 1.0465, "step": 208015 }, { "epoch": 0.52, "learning_rate": 4.792390155518647e-05, "loss": 1.0438, "step": 208020 }, { "epoch": 0.52, "learning_rate": 4.792289496200111e-05, "loss": 1.0478, "step": 208025 }, { "epoch": 0.52, "learning_rate": 4.7921636720519405e-05, "loss": 1.0459, "step": 208030 }, { "epoch": 0.52, "learning_rate": 4.79203784790377e-05, "loss": 1.0456, "step": 208035 }, { "epoch": 0.52, "learning_rate": 4.7919120237555995e-05, "loss": 1.043, "step": 208040 }, { "epoch": 0.52, "learning_rate": 4.791786199607429e-05, "loss": 1.0432, "step": 208045 }, { "epoch": 0.52, "learning_rate": 4.791660375459258e-05, "loss": 1.0421, "step": 208050 }, { "epoch": 0.52, "learning_rate": 4.791534551311088e-05, "loss": 1.042, "step": 208055 }, { "epoch": 0.52, "learning_rate": 4.791408727162917e-05, "loss": 1.0432, "step": 208060 }, { "epoch": 0.52, "learning_rate": 4.791282903014747e-05, "loss": 1.042, "step": 208065 }, { "epoch": 0.52, "learning_rate": 4.791157078866576e-05, "loss": 1.0445, "step": 208070 }, { "epoch": 0.52, "learning_rate": 4.791031254718406e-05, "loss": 1.042, "step": 208075 }, { "epoch": 0.52, "learning_rate": 4.790905430570235e-05, "loss": 1.0455, "step": 208080 }, { "epoch": 0.52, "learning_rate": 4.790779606422065e-05, "loss": 1.0457, "step": 208085 }, { "epoch": 0.52, "learning_rate": 4.7906537822738945e-05, "loss": 1.0445, "step": 208090 }, { "epoch": 0.52, "learning_rate": 4.790527958125724e-05, "loss": 1.0472, "step": 208095 }, { "epoch": 0.52, "learning_rate": 4.7904021339775535e-05, "loss": 1.0454, "step": 208100 }, { "epoch": 0.52, "learning_rate": 4.790276309829383e-05, "loss": 1.0427, "step": 208105 }, { "epoch": 0.52, "learning_rate": 4.7901504856812125e-05, "loss": 1.0426, "step": 208110 }, { "epoch": 0.52, "learning_rate": 4.790024661533042e-05, "loss": 1.0452, "step": 208115 }, { "epoch": 0.52, "learning_rate": 4.7898988373848715e-05, "loss": 1.043, "step": 208120 }, { "epoch": 0.52, "learning_rate": 4.789773013236701e-05, "loss": 1.0466, "step": 208125 }, { "epoch": 0.52, "learning_rate": 4.7896471890885305e-05, "loss": 1.0438, "step": 208130 }, { "epoch": 0.52, "learning_rate": 4.78952136494036e-05, "loss": 1.0641, "step": 208135 }, { "epoch": 0.52, "learning_rate": 4.7893955407921895e-05, "loss": 1.0458, "step": 208140 }, { "epoch": 0.52, "learning_rate": 4.789269716644019e-05, "loss": 1.0439, "step": 208145 }, { "epoch": 0.52, "learning_rate": 4.789143892495848e-05, "loss": 1.0448, "step": 208150 }, { "epoch": 0.52, "learning_rate": 4.789018068347678e-05, "loss": 1.0432, "step": 208155 }, { "epoch": 0.52, "learning_rate": 4.788892244199507e-05, "loss": 1.0459, "step": 208160 }, { "epoch": 0.52, "learning_rate": 4.788766420051337e-05, "loss": 1.0429, "step": 208165 }, { "epoch": 0.52, "learning_rate": 4.788640595903166e-05, "loss": 1.0457, "step": 208170 }, { "epoch": 0.52, "learning_rate": 4.788514771754996e-05, "loss": 1.045, "step": 208175 }, { "epoch": 0.52, "learning_rate": 4.788388947606825e-05, "loss": 1.043, "step": 208180 }, { "epoch": 0.52, "learning_rate": 4.788263123458655e-05, "loss": 1.0659, "step": 208185 }, { "epoch": 0.52, "learning_rate": 4.788137299310484e-05, "loss": 1.0432, "step": 208190 }, { "epoch": 0.52, "learning_rate": 4.7880114751623137e-05, "loss": 1.0407, "step": 208195 }, { "epoch": 0.52, "learning_rate": 4.787885651014143e-05, "loss": 1.0454, "step": 208200 }, { "epoch": 0.52, "learning_rate": 4.7877598268659727e-05, "loss": 1.0443, "step": 208205 }, { "epoch": 0.52, "learning_rate": 4.787634002717802e-05, "loss": 1.0426, "step": 208210 }, { "epoch": 0.52, "learning_rate": 4.787508178569631e-05, "loss": 1.0468, "step": 208215 }, { "epoch": 0.52, "learning_rate": 4.787382354421461e-05, "loss": 1.0446, "step": 208220 }, { "epoch": 0.52, "learning_rate": 4.78725653027329e-05, "loss": 1.0449, "step": 208225 }, { "epoch": 0.52, "learning_rate": 4.78713070612512e-05, "loss": 1.0432, "step": 208230 }, { "epoch": 0.52, "learning_rate": 4.787004881976949e-05, "loss": 1.0456, "step": 208235 }, { "epoch": 0.52, "learning_rate": 4.786879057828779e-05, "loss": 1.0433, "step": 208240 }, { "epoch": 0.52, "learning_rate": 4.786753233680608e-05, "loss": 1.0422, "step": 208245 }, { "epoch": 0.52, "learning_rate": 4.786627409532438e-05, "loss": 1.0432, "step": 208250 }, { "epoch": 0.52, "learning_rate": 4.786501585384267e-05, "loss": 1.0417, "step": 208255 }, { "epoch": 0.52, "learning_rate": 4.786375761236097e-05, "loss": 1.0672, "step": 208260 }, { "epoch": 0.52, "learning_rate": 4.786249937087926e-05, "loss": 1.0472, "step": 208265 }, { "epoch": 0.52, "learning_rate": 4.786124112939756e-05, "loss": 1.0455, "step": 208270 }, { "epoch": 0.52, "learning_rate": 4.785998288791585e-05, "loss": 1.0446, "step": 208275 }, { "epoch": 0.52, "learning_rate": 4.785872464643414e-05, "loss": 1.0447, "step": 208280 }, { "epoch": 0.52, "learning_rate": 4.785746640495244e-05, "loss": 1.0451, "step": 208285 }, { "epoch": 0.52, "learning_rate": 4.785620816347073e-05, "loss": 1.0411, "step": 208290 }, { "epoch": 0.52, "learning_rate": 4.785494992198903e-05, "loss": 1.0651, "step": 208295 }, { "epoch": 0.52, "learning_rate": 4.785369168050732e-05, "loss": 1.0442, "step": 208300 }, { "epoch": 0.52, "learning_rate": 4.785243343902562e-05, "loss": 1.0414, "step": 208305 }, { "epoch": 0.52, "learning_rate": 4.785117519754391e-05, "loss": 1.045, "step": 208310 }, { "epoch": 0.52, "learning_rate": 4.784991695606221e-05, "loss": 1.0435, "step": 208315 }, { "epoch": 0.52, "learning_rate": 4.78486587145805e-05, "loss": 1.0438, "step": 208320 }, { "epoch": 0.52, "learning_rate": 4.78474004730988e-05, "loss": 1.044, "step": 208325 }, { "epoch": 0.52, "learning_rate": 4.784614223161709e-05, "loss": 1.0443, "step": 208330 }, { "epoch": 0.52, "learning_rate": 4.784488399013539e-05, "loss": 1.0451, "step": 208335 }, { "epoch": 0.52, "learning_rate": 4.784362574865369e-05, "loss": 1.0444, "step": 208340 }, { "epoch": 0.52, "learning_rate": 4.784236750717198e-05, "loss": 1.0476, "step": 208345 }, { "epoch": 0.52, "learning_rate": 4.784110926569028e-05, "loss": 1.0433, "step": 208350 }, { "epoch": 0.52, "learning_rate": 4.783985102420857e-05, "loss": 1.0404, "step": 208355 }, { "epoch": 0.52, "learning_rate": 4.783859278272687e-05, "loss": 1.0435, "step": 208360 }, { "epoch": 0.52, "learning_rate": 4.783733454124516e-05, "loss": 1.0451, "step": 208365 }, { "epoch": 0.52, "learning_rate": 4.783607629976346e-05, "loss": 1.0462, "step": 208370 }, { "epoch": 0.52, "learning_rate": 4.783481805828175e-05, "loss": 1.0456, "step": 208375 }, { "epoch": 0.52, "learning_rate": 4.783355981680004e-05, "loss": 1.044, "step": 208380 }, { "epoch": 0.52, "learning_rate": 4.783230157531834e-05, "loss": 1.0438, "step": 208385 }, { "epoch": 0.52, "learning_rate": 4.783104333383663e-05, "loss": 1.0429, "step": 208390 }, { "epoch": 0.52, "learning_rate": 4.782978509235493e-05, "loss": 1.0423, "step": 208395 }, { "epoch": 0.52, "learning_rate": 4.782852685087322e-05, "loss": 1.0459, "step": 208400 }, { "epoch": 0.52, "learning_rate": 4.782726860939152e-05, "loss": 1.046, "step": 208405 }, { "epoch": 0.52, "learning_rate": 4.782601036790981e-05, "loss": 1.0427, "step": 208410 }, { "epoch": 0.52, "learning_rate": 4.782475212642811e-05, "loss": 1.0411, "step": 208415 }, { "epoch": 0.52, "learning_rate": 4.78234938849464e-05, "loss": 1.0409, "step": 208420 }, { "epoch": 0.52, "learning_rate": 4.78222356434647e-05, "loss": 1.0438, "step": 208425 }, { "epoch": 0.52, "learning_rate": 4.782097740198299e-05, "loss": 1.052, "step": 208430 }, { "epoch": 0.52, "learning_rate": 4.781971916050129e-05, "loss": 1.0451, "step": 208435 }, { "epoch": 0.52, "learning_rate": 4.781846091901958e-05, "loss": 1.0444, "step": 208440 }, { "epoch": 0.52, "learning_rate": 4.781720267753787e-05, "loss": 1.0401, "step": 208445 }, { "epoch": 0.52, "learning_rate": 4.781594443605617e-05, "loss": 1.0427, "step": 208450 }, { "epoch": 0.52, "learning_rate": 4.781468619457446e-05, "loss": 1.0615, "step": 208455 }, { "epoch": 0.52, "learning_rate": 4.781342795309276e-05, "loss": 1.0471, "step": 208460 }, { "epoch": 0.52, "learning_rate": 4.781216971161105e-05, "loss": 1.0465, "step": 208465 }, { "epoch": 0.52, "learning_rate": 4.781091147012935e-05, "loss": 1.0455, "step": 208470 }, { "epoch": 0.52, "learning_rate": 4.780965322864764e-05, "loss": 1.0438, "step": 208475 }, { "epoch": 0.52, "learning_rate": 4.780839498716594e-05, "loss": 1.0408, "step": 208480 }, { "epoch": 0.52, "learning_rate": 4.780713674568423e-05, "loss": 1.0404, "step": 208485 }, { "epoch": 0.52, "learning_rate": 4.780587850420253e-05, "loss": 1.0455, "step": 208490 }, { "epoch": 0.52, "learning_rate": 4.780462026272082e-05, "loss": 1.0449, "step": 208495 }, { "epoch": 0.52, "learning_rate": 4.780336202123912e-05, "loss": 1.0408, "step": 208500 }, { "epoch": 0.52, "learning_rate": 4.780210377975741e-05, "loss": 1.068, "step": 208505 }, { "epoch": 0.52, "learning_rate": 4.7800845538275704e-05, "loss": 1.0462, "step": 208510 }, { "epoch": 0.52, "learning_rate": 4.7799587296794e-05, "loss": 1.0458, "step": 208515 }, { "epoch": 0.52, "learning_rate": 4.7798329055312294e-05, "loss": 1.0426, "step": 208520 }, { "epoch": 0.52, "learning_rate": 4.779707081383059e-05, "loss": 1.0428, "step": 208525 }, { "epoch": 0.52, "learning_rate": 4.7795812572348884e-05, "loss": 1.0432, "step": 208530 }, { "epoch": 0.52, "learning_rate": 4.779455433086718e-05, "loss": 1.0459, "step": 208535 }, { "epoch": 0.52, "learning_rate": 4.7793296089385474e-05, "loss": 1.044, "step": 208540 }, { "epoch": 0.52, "learning_rate": 4.779203784790377e-05, "loss": 1.0454, "step": 208545 }, { "epoch": 0.52, "learning_rate": 4.7790779606422064e-05, "loss": 1.0453, "step": 208550 }, { "epoch": 0.52, "learning_rate": 4.778952136494036e-05, "loss": 1.0457, "step": 208555 }, { "epoch": 0.52, "learning_rate": 4.7788263123458654e-05, "loss": 1.0658, "step": 208560 }, { "epoch": 0.52, "learning_rate": 4.778700488197695e-05, "loss": 1.0686, "step": 208565 }, { "epoch": 0.52, "learning_rate": 4.7785746640495244e-05, "loss": 1.0447, "step": 208570 }, { "epoch": 0.52, "learning_rate": 4.7784488399013535e-05, "loss": 1.043, "step": 208575 }, { "epoch": 0.52, "learning_rate": 4.7783230157531834e-05, "loss": 1.042, "step": 208580 }, { "epoch": 0.52, "learning_rate": 4.778197191605013e-05, "loss": 1.0441, "step": 208585 }, { "epoch": 0.52, "learning_rate": 4.778071367456843e-05, "loss": 1.0424, "step": 208590 }, { "epoch": 0.52, "learning_rate": 4.777945543308672e-05, "loss": 1.0427, "step": 208595 }, { "epoch": 0.52, "learning_rate": 4.777819719160502e-05, "loss": 1.0422, "step": 208600 }, { "epoch": 0.52, "learning_rate": 4.777693895012331e-05, "loss": 1.0427, "step": 208605 }, { "epoch": 0.52, "learning_rate": 4.777568070864161e-05, "loss": 1.0442, "step": 208610 }, { "epoch": 0.52, "learning_rate": 4.77744224671599e-05, "loss": 1.0436, "step": 208615 }, { "epoch": 0.52, "learning_rate": 4.7773164225678193e-05, "loss": 1.0399, "step": 208620 }, { "epoch": 0.52, "learning_rate": 4.777190598419649e-05, "loss": 1.0476, "step": 208625 }, { "epoch": 0.52, "learning_rate": 4.7770647742714783e-05, "loss": 1.0433, "step": 208630 }, { "epoch": 0.52, "learning_rate": 4.776938950123308e-05, "loss": 1.0464, "step": 208635 }, { "epoch": 0.52, "learning_rate": 4.776813125975137e-05, "loss": 1.0455, "step": 208640 }, { "epoch": 0.52, "learning_rate": 4.776687301826967e-05, "loss": 1.0612, "step": 208645 }, { "epoch": 0.52, "learning_rate": 4.776561477678796e-05, "loss": 1.044, "step": 208650 }, { "epoch": 0.52, "learning_rate": 4.776435653530626e-05, "loss": 1.0439, "step": 208655 }, { "epoch": 0.52, "learning_rate": 4.776309829382455e-05, "loss": 1.0462, "step": 208660 }, { "epoch": 0.52, "learning_rate": 4.776184005234285e-05, "loss": 1.0431, "step": 208665 }, { "epoch": 0.52, "learning_rate": 4.776058181086114e-05, "loss": 1.046, "step": 208670 }, { "epoch": 0.52, "learning_rate": 4.775932356937944e-05, "loss": 1.0409, "step": 208675 }, { "epoch": 0.52, "learning_rate": 4.775806532789773e-05, "loss": 1.0434, "step": 208680 }, { "epoch": 0.52, "learning_rate": 4.7756807086416025e-05, "loss": 1.0445, "step": 208685 }, { "epoch": 0.52, "learning_rate": 4.775554884493432e-05, "loss": 1.0435, "step": 208690 }, { "epoch": 0.52, "learning_rate": 4.7754290603452615e-05, "loss": 1.0454, "step": 208695 }, { "epoch": 0.52, "learning_rate": 4.775303236197091e-05, "loss": 1.0427, "step": 208700 }, { "epoch": 0.52, "learning_rate": 4.7751774120489205e-05, "loss": 1.0444, "step": 208705 }, { "epoch": 0.52, "learning_rate": 4.77505158790075e-05, "loss": 1.0439, "step": 208710 }, { "epoch": 0.52, "learning_rate": 4.7749257637525795e-05, "loss": 1.0473, "step": 208715 }, { "epoch": 0.52, "learning_rate": 4.774799939604409e-05, "loss": 1.0501, "step": 208720 }, { "epoch": 0.52, "learning_rate": 4.7746741154562385e-05, "loss": 1.0443, "step": 208725 }, { "epoch": 0.52, "learning_rate": 4.774548291308068e-05, "loss": 1.0454, "step": 208730 }, { "epoch": 0.52, "learning_rate": 4.7744224671598975e-05, "loss": 1.0467, "step": 208735 }, { "epoch": 0.52, "learning_rate": 4.7742966430117266e-05, "loss": 1.0438, "step": 208740 }, { "epoch": 0.52, "learning_rate": 4.7741708188635565e-05, "loss": 1.0425, "step": 208745 }, { "epoch": 0.52, "learning_rate": 4.7740449947153856e-05, "loss": 1.0461, "step": 208750 }, { "epoch": 0.52, "learning_rate": 4.7739191705672155e-05, "loss": 1.0439, "step": 208755 }, { "epoch": 0.52, "learning_rate": 4.7737933464190446e-05, "loss": 1.0445, "step": 208760 }, { "epoch": 0.52, "learning_rate": 4.7736675222708745e-05, "loss": 1.0413, "step": 208765 }, { "epoch": 0.52, "learning_rate": 4.7735416981227036e-05, "loss": 1.0454, "step": 208770 }, { "epoch": 0.52, "learning_rate": 4.7734158739745335e-05, "loss": 1.0437, "step": 208775 }, { "epoch": 0.52, "learning_rate": 4.7732900498263626e-05, "loss": 1.0461, "step": 208780 }, { "epoch": 0.52, "learning_rate": 4.7731642256781925e-05, "loss": 1.0429, "step": 208785 }, { "epoch": 0.52, "learning_rate": 4.7730384015300216e-05, "loss": 1.0433, "step": 208790 }, { "epoch": 0.52, "learning_rate": 4.7729125773818515e-05, "loss": 1.0425, "step": 208795 }, { "epoch": 0.52, "learning_rate": 4.7727867532336806e-05, "loss": 1.043, "step": 208800 }, { "epoch": 0.52, "learning_rate": 4.77266092908551e-05, "loss": 1.0452, "step": 208805 }, { "epoch": 0.52, "learning_rate": 4.7725351049373396e-05, "loss": 1.0467, "step": 208810 }, { "epoch": 0.52, "learning_rate": 4.772409280789169e-05, "loss": 1.0441, "step": 208815 }, { "epoch": 0.52, "learning_rate": 4.7722834566409986e-05, "loss": 1.0458, "step": 208820 }, { "epoch": 0.52, "learning_rate": 4.772157632492828e-05, "loss": 1.0428, "step": 208825 }, { "epoch": 0.52, "learning_rate": 4.7720318083446576e-05, "loss": 1.0422, "step": 208830 }, { "epoch": 0.52, "learning_rate": 4.7719059841964874e-05, "loss": 1.0462, "step": 208835 }, { "epoch": 0.52, "learning_rate": 4.771780160048317e-05, "loss": 1.0441, "step": 208840 }, { "epoch": 0.52, "learning_rate": 4.7716543359001464e-05, "loss": 1.0437, "step": 208845 }, { "epoch": 0.52, "learning_rate": 4.7715285117519756e-05, "loss": 1.0452, "step": 208850 }, { "epoch": 0.52, "learning_rate": 4.7714026876038054e-05, "loss": 1.0444, "step": 208855 }, { "epoch": 0.52, "learning_rate": 4.7712768634556346e-05, "loss": 1.0606, "step": 208860 }, { "epoch": 0.52, "learning_rate": 4.7711510393074644e-05, "loss": 1.0417, "step": 208865 }, { "epoch": 0.52, "learning_rate": 4.7710252151592936e-05, "loss": 1.0401, "step": 208870 }, { "epoch": 0.52, "learning_rate": 4.7708993910111234e-05, "loss": 1.0423, "step": 208875 }, { "epoch": 0.52, "learning_rate": 4.7707735668629526e-05, "loss": 1.0462, "step": 208880 }, { "epoch": 0.52, "learning_rate": 4.7706477427147824e-05, "loss": 1.0455, "step": 208885 }, { "epoch": 0.52, "learning_rate": 4.7705219185666116e-05, "loss": 1.0446, "step": 208890 }, { "epoch": 0.52, "learning_rate": 4.7703960944184414e-05, "loss": 1.0451, "step": 208895 }, { "epoch": 0.52, "learning_rate": 4.7702702702702706e-05, "loss": 1.0413, "step": 208900 }, { "epoch": 0.52, "learning_rate": 4.7701444461221004e-05, "loss": 1.0449, "step": 208905 }, { "epoch": 0.52, "learning_rate": 4.7700186219739296e-05, "loss": 1.0462, "step": 208910 }, { "epoch": 0.52, "learning_rate": 4.769892797825759e-05, "loss": 1.0427, "step": 208915 }, { "epoch": 0.52, "learning_rate": 4.7697669736775886e-05, "loss": 1.043, "step": 208920 }, { "epoch": 0.52, "learning_rate": 4.769641149529418e-05, "loss": 1.0457, "step": 208925 }, { "epoch": 0.52, "learning_rate": 4.7695153253812476e-05, "loss": 1.045, "step": 208930 }, { "epoch": 0.52, "learning_rate": 4.769389501233077e-05, "loss": 1.0457, "step": 208935 }, { "epoch": 0.52, "learning_rate": 4.7692636770849066e-05, "loss": 1.0432, "step": 208940 }, { "epoch": 0.52, "learning_rate": 4.769137852936736e-05, "loss": 1.0419, "step": 208945 }, { "epoch": 0.52, "learning_rate": 4.7690120287885656e-05, "loss": 1.0441, "step": 208950 }, { "epoch": 0.52, "learning_rate": 4.768886204640395e-05, "loss": 1.0419, "step": 208955 }, { "epoch": 0.52, "learning_rate": 4.7687603804922246e-05, "loss": 1.0453, "step": 208960 }, { "epoch": 0.52, "learning_rate": 4.768634556344054e-05, "loss": 1.0444, "step": 208965 }, { "epoch": 0.52, "learning_rate": 4.7685087321958836e-05, "loss": 1.0445, "step": 208970 }, { "epoch": 0.52, "learning_rate": 4.768382908047713e-05, "loss": 1.0453, "step": 208975 }, { "epoch": 0.52, "learning_rate": 4.768257083899542e-05, "loss": 1.0447, "step": 208980 }, { "epoch": 0.52, "learning_rate": 4.768131259751372e-05, "loss": 1.0457, "step": 208985 }, { "epoch": 0.52, "learning_rate": 4.768005435603201e-05, "loss": 1.0441, "step": 208990 }, { "epoch": 0.52, "learning_rate": 4.767879611455031e-05, "loss": 1.0458, "step": 208995 }, { "epoch": 0.52, "learning_rate": 4.76775378730686e-05, "loss": 1.044, "step": 209000 }, { "epoch": 0.52, "learning_rate": 4.76762796315869e-05, "loss": 1.0433, "step": 209005 }, { "epoch": 0.52, "learning_rate": 4.767502139010519e-05, "loss": 1.0451, "step": 209010 }, { "epoch": 0.52, "learning_rate": 4.767376314862349e-05, "loss": 1.0424, "step": 209015 }, { "epoch": 0.52, "learning_rate": 4.767250490714178e-05, "loss": 1.0473, "step": 209020 }, { "epoch": 0.52, "learning_rate": 4.767124666566008e-05, "loss": 1.0704, "step": 209025 }, { "epoch": 0.52, "learning_rate": 4.766998842417837e-05, "loss": 1.0458, "step": 209030 }, { "epoch": 0.52, "learning_rate": 4.766873018269667e-05, "loss": 1.0428, "step": 209035 }, { "epoch": 0.52, "learning_rate": 4.766747194121496e-05, "loss": 1.042, "step": 209040 }, { "epoch": 0.52, "learning_rate": 4.766621369973325e-05, "loss": 1.0439, "step": 209045 }, { "epoch": 0.52, "learning_rate": 4.766495545825155e-05, "loss": 1.0431, "step": 209050 }, { "epoch": 0.52, "learning_rate": 4.766369721676984e-05, "loss": 1.0426, "step": 209055 }, { "epoch": 0.52, "learning_rate": 4.766243897528814e-05, "loss": 1.0631, "step": 209060 }, { "epoch": 0.52, "learning_rate": 4.766118073380643e-05, "loss": 1.0425, "step": 209065 }, { "epoch": 0.52, "learning_rate": 4.765992249232473e-05, "loss": 1.0417, "step": 209070 }, { "epoch": 0.52, "learning_rate": 4.765866425084302e-05, "loss": 1.0443, "step": 209075 }, { "epoch": 0.52, "learning_rate": 4.765740600936132e-05, "loss": 1.0463, "step": 209080 }, { "epoch": 0.52, "learning_rate": 4.765614776787962e-05, "loss": 1.0661, "step": 209085 }, { "epoch": 0.52, "learning_rate": 4.765488952639791e-05, "loss": 1.0432, "step": 209090 }, { "epoch": 0.52, "learning_rate": 4.765363128491621e-05, "loss": 1.0435, "step": 209095 }, { "epoch": 0.52, "learning_rate": 4.76523730434345e-05, "loss": 1.0462, "step": 209100 }, { "epoch": 0.52, "learning_rate": 4.76511148019528e-05, "loss": 1.0429, "step": 209105 }, { "epoch": 0.52, "learning_rate": 4.764985656047109e-05, "loss": 1.0498, "step": 209110 }, { "epoch": 0.52, "learning_rate": 4.764859831898939e-05, "loss": 1.0477, "step": 209115 }, { "epoch": 0.52, "learning_rate": 4.764734007750768e-05, "loss": 1.0442, "step": 209120 }, { "epoch": 0.52, "learning_rate": 4.764608183602598e-05, "loss": 1.0409, "step": 209125 }, { "epoch": 0.52, "learning_rate": 4.764482359454427e-05, "loss": 1.0432, "step": 209130 }, { "epoch": 0.52, "learning_rate": 4.764356535306257e-05, "loss": 1.0437, "step": 209135 }, { "epoch": 0.52, "learning_rate": 4.764230711158086e-05, "loss": 1.039, "step": 209140 }, { "epoch": 0.52, "learning_rate": 4.764104887009915e-05, "loss": 1.0667, "step": 209145 }, { "epoch": 0.53, "learning_rate": 4.763979062861745e-05, "loss": 1.0451, "step": 209150 }, { "epoch": 0.53, "learning_rate": 4.763853238713574e-05, "loss": 1.0435, "step": 209155 }, { "epoch": 0.53, "learning_rate": 4.763727414565404e-05, "loss": 1.0451, "step": 209160 }, { "epoch": 0.53, "learning_rate": 4.763601590417233e-05, "loss": 1.0448, "step": 209165 }, { "epoch": 0.53, "learning_rate": 4.763475766269063e-05, "loss": 1.0415, "step": 209170 }, { "epoch": 0.53, "learning_rate": 4.763349942120892e-05, "loss": 1.0451, "step": 209175 }, { "epoch": 0.53, "learning_rate": 4.763224117972722e-05, "loss": 1.0416, "step": 209180 }, { "epoch": 0.53, "learning_rate": 4.763098293824551e-05, "loss": 1.0455, "step": 209185 }, { "epoch": 0.53, "learning_rate": 4.762972469676381e-05, "loss": 1.0445, "step": 209190 }, { "epoch": 0.53, "learning_rate": 4.76284664552821e-05, "loss": 1.0458, "step": 209195 }, { "epoch": 0.53, "learning_rate": 4.76272082138004e-05, "loss": 1.0447, "step": 209200 }, { "epoch": 0.53, "learning_rate": 4.762594997231869e-05, "loss": 1.0419, "step": 209205 }, { "epoch": 0.53, "learning_rate": 4.762469173083698e-05, "loss": 1.0433, "step": 209210 }, { "epoch": 0.53, "learning_rate": 4.762343348935528e-05, "loss": 1.0435, "step": 209215 }, { "epoch": 0.53, "learning_rate": 4.762217524787357e-05, "loss": 1.046, "step": 209220 }, { "epoch": 0.53, "learning_rate": 4.762091700639187e-05, "loss": 1.0489, "step": 209225 }, { "epoch": 0.53, "learning_rate": 4.761965876491016e-05, "loss": 1.0407, "step": 209230 }, { "epoch": 0.53, "learning_rate": 4.761840052342846e-05, "loss": 1.0457, "step": 209235 }, { "epoch": 0.53, "learning_rate": 4.761714228194675e-05, "loss": 1.0448, "step": 209240 }, { "epoch": 0.53, "learning_rate": 4.761588404046505e-05, "loss": 1.0465, "step": 209245 }, { "epoch": 0.53, "learning_rate": 4.761462579898334e-05, "loss": 1.0709, "step": 209250 }, { "epoch": 0.53, "learning_rate": 4.761336755750164e-05, "loss": 1.0463, "step": 209255 }, { "epoch": 0.53, "learning_rate": 4.761210931601993e-05, "loss": 1.0447, "step": 209260 }, { "epoch": 0.53, "learning_rate": 4.761085107453823e-05, "loss": 1.0433, "step": 209265 }, { "epoch": 0.53, "learning_rate": 4.760959283305652e-05, "loss": 1.0451, "step": 209270 }, { "epoch": 0.53, "learning_rate": 4.760833459157481e-05, "loss": 1.0455, "step": 209275 }, { "epoch": 0.53, "learning_rate": 4.760707635009311e-05, "loss": 1.0439, "step": 209280 }, { "epoch": 0.53, "learning_rate": 4.76058181086114e-05, "loss": 1.0424, "step": 209285 }, { "epoch": 0.53, "learning_rate": 4.76045598671297e-05, "loss": 1.0461, "step": 209290 }, { "epoch": 0.53, "learning_rate": 4.760330162564799e-05, "loss": 1.0458, "step": 209295 }, { "epoch": 0.53, "learning_rate": 4.760204338416629e-05, "loss": 1.045, "step": 209300 }, { "epoch": 0.53, "learning_rate": 4.760078514268458e-05, "loss": 1.0462, "step": 209305 }, { "epoch": 0.53, "learning_rate": 4.759952690120288e-05, "loss": 1.0451, "step": 209310 }, { "epoch": 0.53, "learning_rate": 4.759826865972117e-05, "loss": 1.0453, "step": 209315 }, { "epoch": 0.53, "learning_rate": 4.759701041823947e-05, "loss": 1.0432, "step": 209320 }, { "epoch": 0.53, "learning_rate": 4.759575217675776e-05, "loss": 1.067, "step": 209325 }, { "epoch": 0.53, "learning_rate": 4.759449393527606e-05, "loss": 1.0429, "step": 209330 }, { "epoch": 0.53, "learning_rate": 4.759323569379436e-05, "loss": 1.045, "step": 209335 }, { "epoch": 0.53, "learning_rate": 4.759197745231265e-05, "loss": 1.0425, "step": 209340 }, { "epoch": 0.53, "learning_rate": 4.759071921083095e-05, "loss": 1.0442, "step": 209345 }, { "epoch": 0.53, "learning_rate": 4.758946096934924e-05, "loss": 1.0452, "step": 209350 }, { "epoch": 0.53, "learning_rate": 4.758820272786754e-05, "loss": 1.0412, "step": 209355 }, { "epoch": 0.53, "learning_rate": 4.758694448638583e-05, "loss": 1.0461, "step": 209360 }, { "epoch": 0.53, "learning_rate": 4.758568624490413e-05, "loss": 1.0411, "step": 209365 }, { "epoch": 0.53, "learning_rate": 4.758442800342242e-05, "loss": 1.0472, "step": 209370 }, { "epoch": 0.53, "learning_rate": 4.758316976194071e-05, "loss": 1.0697, "step": 209375 }, { "epoch": 0.53, "learning_rate": 4.758191152045901e-05, "loss": 1.0456, "step": 209380 }, { "epoch": 0.53, "learning_rate": 4.75806532789773e-05, "loss": 1.0463, "step": 209385 }, { "epoch": 0.53, "learning_rate": 4.75793950374956e-05, "loss": 1.0465, "step": 209390 }, { "epoch": 0.53, "learning_rate": 4.757813679601389e-05, "loss": 1.045, "step": 209395 }, { "epoch": 0.53, "learning_rate": 4.757687855453219e-05, "loss": 1.0448, "step": 209400 }, { "epoch": 0.53, "learning_rate": 4.757562031305048e-05, "loss": 1.0426, "step": 209405 }, { "epoch": 0.53, "learning_rate": 4.757436207156878e-05, "loss": 1.0437, "step": 209410 }, { "epoch": 0.53, "learning_rate": 4.757310383008707e-05, "loss": 1.0421, "step": 209415 }, { "epoch": 0.53, "learning_rate": 4.757184558860537e-05, "loss": 1.0444, "step": 209420 }, { "epoch": 0.53, "learning_rate": 4.757058734712366e-05, "loss": 1.0434, "step": 209425 }, { "epoch": 0.53, "learning_rate": 4.756932910564196e-05, "loss": 1.0419, "step": 209430 }, { "epoch": 0.53, "learning_rate": 4.756807086416025e-05, "loss": 1.0449, "step": 209435 }, { "epoch": 0.53, "learning_rate": 4.7566812622678544e-05, "loss": 1.0438, "step": 209440 }, { "epoch": 0.53, "learning_rate": 4.756555438119684e-05, "loss": 1.0456, "step": 209445 }, { "epoch": 0.53, "learning_rate": 4.7564296139715134e-05, "loss": 1.045, "step": 209450 }, { "epoch": 0.53, "learning_rate": 4.756303789823343e-05, "loss": 1.0405, "step": 209455 }, { "epoch": 0.53, "learning_rate": 4.7561779656751724e-05, "loss": 1.0437, "step": 209460 }, { "epoch": 0.53, "learning_rate": 4.756052141527002e-05, "loss": 1.0458, "step": 209465 }, { "epoch": 0.53, "learning_rate": 4.7559263173788314e-05, "loss": 1.0449, "step": 209470 }, { "epoch": 0.53, "learning_rate": 4.755800493230661e-05, "loss": 1.0432, "step": 209475 }, { "epoch": 0.53, "learning_rate": 4.7556746690824904e-05, "loss": 1.046, "step": 209480 }, { "epoch": 0.53, "learning_rate": 4.75554884493432e-05, "loss": 1.0451, "step": 209485 }, { "epoch": 0.53, "learning_rate": 4.7554230207861494e-05, "loss": 1.0433, "step": 209490 }, { "epoch": 0.53, "learning_rate": 4.755297196637979e-05, "loss": 1.0425, "step": 209495 }, { "epoch": 0.53, "learning_rate": 4.7551713724898084e-05, "loss": 1.0441, "step": 209500 }, { "epoch": 0.53, "learning_rate": 4.7550455483416375e-05, "loss": 1.0416, "step": 209505 }, { "epoch": 0.53, "learning_rate": 4.7549197241934674e-05, "loss": 1.0423, "step": 209510 }, { "epoch": 0.53, "learning_rate": 4.7547939000452965e-05, "loss": 1.0444, "step": 209515 }, { "epoch": 0.53, "learning_rate": 4.7546680758971264e-05, "loss": 1.0405, "step": 209520 }, { "epoch": 0.53, "learning_rate": 4.7545422517489555e-05, "loss": 1.0432, "step": 209525 }, { "epoch": 0.53, "learning_rate": 4.7544164276007854e-05, "loss": 1.0459, "step": 209530 }, { "epoch": 0.53, "learning_rate": 4.7542906034526145e-05, "loss": 1.046, "step": 209535 }, { "epoch": 0.53, "learning_rate": 4.7541647793044444e-05, "loss": 1.0454, "step": 209540 }, { "epoch": 0.53, "learning_rate": 4.7540389551562735e-05, "loss": 1.0448, "step": 209545 }, { "epoch": 0.53, "learning_rate": 4.7539131310081034e-05, "loss": 1.0442, "step": 209550 }, { "epoch": 0.53, "learning_rate": 4.7537873068599325e-05, "loss": 1.0492, "step": 209555 }, { "epoch": 0.53, "learning_rate": 4.7536614827117624e-05, "loss": 1.0433, "step": 209560 }, { "epoch": 0.53, "learning_rate": 4.7535356585635915e-05, "loss": 1.0434, "step": 209565 }, { "epoch": 0.53, "learning_rate": 4.753409834415421e-05, "loss": 1.0462, "step": 209570 }, { "epoch": 0.53, "learning_rate": 4.7532840102672505e-05, "loss": 1.0447, "step": 209575 }, { "epoch": 0.53, "learning_rate": 4.75315818611908e-05, "loss": 1.0724, "step": 209580 }, { "epoch": 0.53, "learning_rate": 4.75303236197091e-05, "loss": 1.0445, "step": 209585 }, { "epoch": 0.53, "learning_rate": 4.7529065378227394e-05, "loss": 1.0412, "step": 209590 }, { "epoch": 0.53, "learning_rate": 4.752780713674569e-05, "loss": 1.0419, "step": 209595 }, { "epoch": 0.53, "learning_rate": 4.7526548895263984e-05, "loss": 1.0735, "step": 209600 }, { "epoch": 0.53, "learning_rate": 4.7525290653782275e-05, "loss": 1.0436, "step": 209605 }, { "epoch": 0.53, "learning_rate": 4.7524032412300573e-05, "loss": 1.0422, "step": 209610 }, { "epoch": 0.53, "learning_rate": 4.7522774170818865e-05, "loss": 1.0413, "step": 209615 }, { "epoch": 0.53, "learning_rate": 4.7521515929337163e-05, "loss": 1.0447, "step": 209620 }, { "epoch": 0.53, "learning_rate": 4.7520257687855455e-05, "loss": 1.0455, "step": 209625 }, { "epoch": 0.53, "learning_rate": 4.7518999446373753e-05, "loss": 1.0424, "step": 209630 }, { "epoch": 0.53, "learning_rate": 4.7517741204892045e-05, "loss": 1.1041, "step": 209635 }, { "epoch": 0.53, "learning_rate": 4.751648296341034e-05, "loss": 1.045, "step": 209640 }, { "epoch": 0.53, "learning_rate": 4.7515224721928635e-05, "loss": 1.0451, "step": 209645 }, { "epoch": 0.53, "learning_rate": 4.751396648044693e-05, "loss": 1.0456, "step": 209650 }, { "epoch": 0.53, "learning_rate": 4.7512708238965225e-05, "loss": 1.0431, "step": 209655 }, { "epoch": 0.53, "learning_rate": 4.751144999748352e-05, "loss": 1.0452, "step": 209660 }, { "epoch": 0.53, "learning_rate": 4.7510191756001815e-05, "loss": 1.045, "step": 209665 }, { "epoch": 0.53, "learning_rate": 4.7508933514520107e-05, "loss": 1.0432, "step": 209670 }, { "epoch": 0.53, "learning_rate": 4.7507675273038405e-05, "loss": 1.0427, "step": 209675 }, { "epoch": 0.53, "learning_rate": 4.7506417031556696e-05, "loss": 1.0426, "step": 209680 }, { "epoch": 0.53, "learning_rate": 4.7505158790074995e-05, "loss": 1.0459, "step": 209685 }, { "epoch": 0.53, "learning_rate": 4.7503900548593286e-05, "loss": 1.0654, "step": 209690 }, { "epoch": 0.53, "learning_rate": 4.7502642307111585e-05, "loss": 1.0412, "step": 209695 }, { "epoch": 0.53, "learning_rate": 4.7501384065629876e-05, "loss": 1.0411, "step": 209700 }, { "epoch": 0.53, "learning_rate": 4.7500125824148175e-05, "loss": 1.0447, "step": 209705 }, { "epoch": 0.53, "learning_rate": 4.7498867582666466e-05, "loss": 1.0436, "step": 209710 }, { "epoch": 0.53, "learning_rate": 4.7497609341184765e-05, "loss": 1.0462, "step": 209715 }, { "epoch": 0.53, "learning_rate": 4.7496351099703056e-05, "loss": 1.0431, "step": 209720 }, { "epoch": 0.53, "learning_rate": 4.7495092858221355e-05, "loss": 1.0441, "step": 209725 }, { "epoch": 0.53, "learning_rate": 4.7493834616739646e-05, "loss": 1.0439, "step": 209730 }, { "epoch": 0.53, "learning_rate": 4.749257637525794e-05, "loss": 1.0479, "step": 209735 }, { "epoch": 0.53, "learning_rate": 4.7491318133776236e-05, "loss": 1.0453, "step": 209740 }, { "epoch": 0.53, "learning_rate": 4.749005989229453e-05, "loss": 1.0436, "step": 209745 }, { "epoch": 0.53, "learning_rate": 4.7488801650812826e-05, "loss": 1.043, "step": 209750 }, { "epoch": 0.53, "learning_rate": 4.748754340933112e-05, "loss": 1.0436, "step": 209755 }, { "epoch": 0.53, "learning_rate": 4.7486285167849416e-05, "loss": 1.0429, "step": 209760 }, { "epoch": 0.53, "learning_rate": 4.748502692636771e-05, "loss": 1.0423, "step": 209765 }, { "epoch": 0.53, "learning_rate": 4.7483768684886006e-05, "loss": 1.0438, "step": 209770 }, { "epoch": 0.53, "learning_rate": 4.74825104434043e-05, "loss": 1.0441, "step": 209775 }, { "epoch": 0.53, "learning_rate": 4.7481252201922596e-05, "loss": 1.0452, "step": 209780 }, { "epoch": 0.53, "learning_rate": 4.747999396044089e-05, "loss": 1.0444, "step": 209785 }, { "epoch": 0.53, "learning_rate": 4.7478735718959186e-05, "loss": 1.0415, "step": 209790 }, { "epoch": 0.53, "learning_rate": 4.747747747747748e-05, "loss": 1.0461, "step": 209795 }, { "epoch": 0.53, "learning_rate": 4.747621923599577e-05, "loss": 1.0468, "step": 209800 }, { "epoch": 0.53, "learning_rate": 4.747496099451407e-05, "loss": 1.0446, "step": 209805 }, { "epoch": 0.53, "learning_rate": 4.747370275303236e-05, "loss": 1.0391, "step": 209810 }, { "epoch": 0.53, "learning_rate": 4.747244451155066e-05, "loss": 1.0661, "step": 209815 }, { "epoch": 0.53, "learning_rate": 4.747118627006895e-05, "loss": 1.0475, "step": 209820 }, { "epoch": 0.53, "learning_rate": 4.746992802858725e-05, "loss": 1.0439, "step": 209825 }, { "epoch": 0.53, "learning_rate": 4.746866978710554e-05, "loss": 1.0442, "step": 209830 }, { "epoch": 0.53, "learning_rate": 4.7467411545623844e-05, "loss": 1.0465, "step": 209835 }, { "epoch": 0.53, "learning_rate": 4.7466153304142136e-05, "loss": 1.0445, "step": 209840 }, { "epoch": 0.53, "learning_rate": 4.746489506266043e-05, "loss": 1.0431, "step": 209845 }, { "epoch": 0.53, "learning_rate": 4.7463636821178726e-05, "loss": 1.0457, "step": 209850 }, { "epoch": 0.53, "learning_rate": 4.746237857969702e-05, "loss": 1.0473, "step": 209855 }, { "epoch": 0.53, "learning_rate": 4.7461120338215316e-05, "loss": 1.0471, "step": 209860 }, { "epoch": 0.53, "learning_rate": 4.745986209673361e-05, "loss": 1.045, "step": 209865 }, { "epoch": 0.53, "learning_rate": 4.7458603855251906e-05, "loss": 1.0433, "step": 209870 }, { "epoch": 0.53, "learning_rate": 4.74573456137702e-05, "loss": 1.0466, "step": 209875 }, { "epoch": 0.53, "learning_rate": 4.7456087372288496e-05, "loss": 1.0457, "step": 209880 }, { "epoch": 0.53, "learning_rate": 4.745482913080679e-05, "loss": 1.0432, "step": 209885 }, { "epoch": 0.53, "learning_rate": 4.7453570889325086e-05, "loss": 1.0439, "step": 209890 }, { "epoch": 0.53, "learning_rate": 4.745231264784338e-05, "loss": 1.0445, "step": 209895 }, { "epoch": 0.53, "learning_rate": 4.745105440636167e-05, "loss": 1.0433, "step": 209900 }, { "epoch": 0.53, "learning_rate": 4.744979616487997e-05, "loss": 1.0427, "step": 209905 }, { "epoch": 0.53, "learning_rate": 4.744853792339826e-05, "loss": 1.0449, "step": 209910 }, { "epoch": 0.53, "learning_rate": 4.744727968191656e-05, "loss": 1.0434, "step": 209915 }, { "epoch": 0.53, "learning_rate": 4.744602144043485e-05, "loss": 1.0431, "step": 209920 }, { "epoch": 0.53, "learning_rate": 4.744476319895315e-05, "loss": 1.0407, "step": 209925 }, { "epoch": 0.53, "learning_rate": 4.744350495747144e-05, "loss": 1.046, "step": 209930 }, { "epoch": 0.53, "learning_rate": 4.744224671598974e-05, "loss": 1.0416, "step": 209935 }, { "epoch": 0.53, "learning_rate": 4.744124012280437e-05, "loss": 1.0451, "step": 209940 }, { "epoch": 0.53, "learning_rate": 4.7439981881322665e-05, "loss": 1.0479, "step": 209945 }, { "epoch": 0.53, "learning_rate": 4.7438723639840957e-05, "loss": 1.046, "step": 209950 }, { "epoch": 0.53, "learning_rate": 4.7437465398359255e-05, "loss": 1.0442, "step": 209955 }, { "epoch": 0.53, "learning_rate": 4.7436207156877547e-05, "loss": 1.0427, "step": 209960 }, { "epoch": 0.53, "learning_rate": 4.7434948915395845e-05, "loss": 1.0415, "step": 209965 }, { "epoch": 0.53, "learning_rate": 4.7433690673914136e-05, "loss": 1.045, "step": 209970 }, { "epoch": 0.53, "learning_rate": 4.7432432432432435e-05, "loss": 1.0399, "step": 209975 }, { "epoch": 0.53, "learning_rate": 4.7431174190950726e-05, "loss": 1.0444, "step": 209980 }, { "epoch": 0.53, "learning_rate": 4.7429915949469025e-05, "loss": 1.0447, "step": 209985 }, { "epoch": 0.53, "learning_rate": 4.7428657707987316e-05, "loss": 1.0441, "step": 209990 }, { "epoch": 0.53, "learning_rate": 4.7427399466505615e-05, "loss": 1.0419, "step": 209995 }, { "epoch": 0.53, "learning_rate": 4.7426141225023906e-05, "loss": 1.0464, "step": 210000 }, { "epoch": 0.53, "learning_rate": 4.7424882983542205e-05, "loss": 1.0449, "step": 210005 }, { "epoch": 0.53, "learning_rate": 4.7423624742060496e-05, "loss": 1.0447, "step": 210010 }, { "epoch": 0.53, "learning_rate": 4.742236650057879e-05, "loss": 1.0472, "step": 210015 }, { "epoch": 0.53, "learning_rate": 4.7421108259097086e-05, "loss": 1.0556, "step": 210020 }, { "epoch": 0.53, "learning_rate": 4.741985001761538e-05, "loss": 1.0676, "step": 210025 }, { "epoch": 0.53, "learning_rate": 4.7418591776133676e-05, "loss": 1.0432, "step": 210030 }, { "epoch": 0.53, "learning_rate": 4.7417333534651975e-05, "loss": 1.0439, "step": 210035 }, { "epoch": 0.53, "learning_rate": 4.741607529317027e-05, "loss": 1.0423, "step": 210040 }, { "epoch": 0.53, "learning_rate": 4.7414817051688565e-05, "loss": 1.0418, "step": 210045 }, { "epoch": 0.53, "learning_rate": 4.7413558810206856e-05, "loss": 1.0459, "step": 210050 }, { "epoch": 0.53, "learning_rate": 4.7412300568725155e-05, "loss": 1.0484, "step": 210055 }, { "epoch": 0.53, "learning_rate": 4.7411042327243446e-05, "loss": 1.044, "step": 210060 }, { "epoch": 0.53, "learning_rate": 4.7409784085761745e-05, "loss": 1.0422, "step": 210065 }, { "epoch": 0.53, "learning_rate": 4.7408525844280036e-05, "loss": 1.0419, "step": 210070 }, { "epoch": 0.53, "learning_rate": 4.7407267602798335e-05, "loss": 1.0478, "step": 210075 }, { "epoch": 0.53, "learning_rate": 4.7406009361316626e-05, "loss": 1.0404, "step": 210080 }, { "epoch": 0.53, "learning_rate": 4.7404751119834925e-05, "loss": 1.0448, "step": 210085 }, { "epoch": 0.53, "learning_rate": 4.7403492878353216e-05, "loss": 1.0632, "step": 210090 }, { "epoch": 0.53, "learning_rate": 4.7402234636871515e-05, "loss": 1.0449, "step": 210095 }, { "epoch": 0.53, "learning_rate": 4.7400976395389806e-05, "loss": 1.046, "step": 210100 }, { "epoch": 0.53, "learning_rate": 4.7399718153908104e-05, "loss": 1.0434, "step": 210105 }, { "epoch": 0.53, "learning_rate": 4.7398459912426396e-05, "loss": 1.0469, "step": 210110 }, { "epoch": 0.53, "learning_rate": 4.739720167094469e-05, "loss": 1.0421, "step": 210115 }, { "epoch": 0.53, "learning_rate": 4.7395943429462986e-05, "loss": 1.0473, "step": 210120 }, { "epoch": 0.53, "learning_rate": 4.739468518798128e-05, "loss": 1.0451, "step": 210125 }, { "epoch": 0.53, "learning_rate": 4.7393426946499576e-05, "loss": 1.0464, "step": 210130 }, { "epoch": 0.53, "learning_rate": 4.739216870501787e-05, "loss": 1.0486, "step": 210135 }, { "epoch": 0.53, "learning_rate": 4.7390910463536166e-05, "loss": 1.0439, "step": 210140 }, { "epoch": 0.53, "learning_rate": 4.738965222205446e-05, "loss": 1.0415, "step": 210145 }, { "epoch": 0.53, "learning_rate": 4.7388393980572756e-05, "loss": 1.0432, "step": 210150 }, { "epoch": 0.53, "learning_rate": 4.738713573909105e-05, "loss": 1.0485, "step": 210155 }, { "epoch": 0.53, "learning_rate": 4.7385877497609346e-05, "loss": 1.044, "step": 210160 }, { "epoch": 0.53, "learning_rate": 4.738461925612764e-05, "loss": 1.0454, "step": 210165 }, { "epoch": 0.53, "learning_rate": 4.7383361014645936e-05, "loss": 1.0437, "step": 210170 }, { "epoch": 0.53, "learning_rate": 4.738210277316423e-05, "loss": 1.0443, "step": 210175 }, { "epoch": 0.53, "learning_rate": 4.738084453168252e-05, "loss": 1.0431, "step": 210180 }, { "epoch": 0.53, "learning_rate": 4.737958629020082e-05, "loss": 1.0429, "step": 210185 }, { "epoch": 0.53, "learning_rate": 4.737832804871911e-05, "loss": 1.0454, "step": 210190 }, { "epoch": 0.53, "learning_rate": 4.737706980723741e-05, "loss": 1.0435, "step": 210195 }, { "epoch": 0.53, "learning_rate": 4.73758115657557e-05, "loss": 1.0475, "step": 210200 }, { "epoch": 0.53, "learning_rate": 4.7374553324274e-05, "loss": 1.044, "step": 210205 }, { "epoch": 0.53, "learning_rate": 4.737329508279229e-05, "loss": 1.0421, "step": 210210 }, { "epoch": 0.53, "learning_rate": 4.737203684131059e-05, "loss": 1.0415, "step": 210215 }, { "epoch": 0.53, "learning_rate": 4.737077859982888e-05, "loss": 1.0432, "step": 210220 }, { "epoch": 0.53, "learning_rate": 4.736952035834718e-05, "loss": 1.0432, "step": 210225 }, { "epoch": 0.53, "learning_rate": 4.736826211686547e-05, "loss": 1.0413, "step": 210230 }, { "epoch": 0.53, "learning_rate": 4.736700387538377e-05, "loss": 1.0432, "step": 210235 }, { "epoch": 0.53, "learning_rate": 4.736574563390206e-05, "loss": 1.0446, "step": 210240 }, { "epoch": 0.53, "learning_rate": 4.736448739242035e-05, "loss": 1.0429, "step": 210245 }, { "epoch": 0.53, "learning_rate": 4.736322915093865e-05, "loss": 1.0432, "step": 210250 }, { "epoch": 0.53, "learning_rate": 4.736197090945694e-05, "loss": 1.0442, "step": 210255 }, { "epoch": 0.53, "learning_rate": 4.736071266797524e-05, "loss": 1.0408, "step": 210260 }, { "epoch": 0.53, "learning_rate": 4.735945442649353e-05, "loss": 1.0438, "step": 210265 }, { "epoch": 0.53, "learning_rate": 4.735819618501183e-05, "loss": 1.0458, "step": 210270 }, { "epoch": 0.53, "learning_rate": 4.735693794353012e-05, "loss": 1.0416, "step": 210275 }, { "epoch": 0.53, "learning_rate": 4.735567970204842e-05, "loss": 1.0421, "step": 210280 }, { "epoch": 0.53, "learning_rate": 4.735442146056672e-05, "loss": 1.0439, "step": 210285 }, { "epoch": 0.53, "learning_rate": 4.735316321908501e-05, "loss": 1.0442, "step": 210290 }, { "epoch": 0.53, "learning_rate": 4.735190497760331e-05, "loss": 1.0457, "step": 210295 }, { "epoch": 0.53, "learning_rate": 4.73506467361216e-05, "loss": 1.0457, "step": 210300 }, { "epoch": 0.53, "learning_rate": 4.73493884946399e-05, "loss": 1.0396, "step": 210305 }, { "epoch": 0.53, "learning_rate": 4.734813025315819e-05, "loss": 1.0464, "step": 210310 }, { "epoch": 0.53, "learning_rate": 4.734687201167649e-05, "loss": 1.0469, "step": 210315 }, { "epoch": 0.53, "learning_rate": 4.734561377019478e-05, "loss": 1.065, "step": 210320 }, { "epoch": 0.53, "learning_rate": 4.734435552871308e-05, "loss": 1.0467, "step": 210325 }, { "epoch": 0.53, "learning_rate": 4.734309728723137e-05, "loss": 1.0437, "step": 210330 }, { "epoch": 0.53, "learning_rate": 4.734183904574967e-05, "loss": 1.0444, "step": 210335 }, { "epoch": 0.53, "learning_rate": 4.734058080426796e-05, "loss": 1.0448, "step": 210340 }, { "epoch": 0.53, "learning_rate": 4.733932256278625e-05, "loss": 1.0447, "step": 210345 }, { "epoch": 0.53, "learning_rate": 4.733806432130455e-05, "loss": 1.0445, "step": 210350 }, { "epoch": 0.53, "learning_rate": 4.733680607982284e-05, "loss": 1.0446, "step": 210355 }, { "epoch": 0.53, "learning_rate": 4.733554783834114e-05, "loss": 1.0458, "step": 210360 }, { "epoch": 0.53, "learning_rate": 4.733428959685943e-05, "loss": 1.0458, "step": 210365 }, { "epoch": 0.53, "learning_rate": 4.733303135537773e-05, "loss": 1.0459, "step": 210370 }, { "epoch": 0.53, "learning_rate": 4.733177311389602e-05, "loss": 1.0439, "step": 210375 }, { "epoch": 0.53, "learning_rate": 4.733051487241432e-05, "loss": 1.0434, "step": 210380 }, { "epoch": 0.53, "learning_rate": 4.732925663093261e-05, "loss": 1.0466, "step": 210385 }, { "epoch": 0.53, "learning_rate": 4.732799838945091e-05, "loss": 1.0412, "step": 210390 }, { "epoch": 0.53, "learning_rate": 4.73267401479692e-05, "loss": 1.046, "step": 210395 }, { "epoch": 0.53, "learning_rate": 4.73254819064875e-05, "loss": 1.0432, "step": 210400 }, { "epoch": 0.53, "learning_rate": 4.732422366500579e-05, "loss": 1.0427, "step": 210405 }, { "epoch": 0.53, "learning_rate": 4.732296542352408e-05, "loss": 1.0416, "step": 210410 }, { "epoch": 0.53, "learning_rate": 4.732170718204238e-05, "loss": 1.0429, "step": 210415 }, { "epoch": 0.53, "learning_rate": 4.732044894056067e-05, "loss": 1.0439, "step": 210420 }, { "epoch": 0.53, "learning_rate": 4.731919069907897e-05, "loss": 1.0404, "step": 210425 }, { "epoch": 0.53, "learning_rate": 4.731793245759726e-05, "loss": 1.0442, "step": 210430 }, { "epoch": 0.53, "learning_rate": 4.731667421611556e-05, "loss": 1.0453, "step": 210435 }, { "epoch": 0.53, "learning_rate": 4.731541597463385e-05, "loss": 1.0449, "step": 210440 }, { "epoch": 0.53, "learning_rate": 4.731415773315215e-05, "loss": 1.044, "step": 210445 }, { "epoch": 0.53, "learning_rate": 4.731289949167044e-05, "loss": 1.0472, "step": 210450 }, { "epoch": 0.53, "learning_rate": 4.731164125018874e-05, "loss": 1.0454, "step": 210455 }, { "epoch": 0.53, "learning_rate": 4.731038300870703e-05, "loss": 1.0434, "step": 210460 }, { "epoch": 0.53, "learning_rate": 4.730912476722533e-05, "loss": 1.0433, "step": 210465 }, { "epoch": 0.53, "learning_rate": 4.730786652574362e-05, "loss": 1.0421, "step": 210470 }, { "epoch": 0.53, "learning_rate": 4.730660828426191e-05, "loss": 1.0432, "step": 210475 }, { "epoch": 0.53, "learning_rate": 4.730535004278021e-05, "loss": 1.0438, "step": 210480 }, { "epoch": 0.53, "learning_rate": 4.73040918012985e-05, "loss": 1.0438, "step": 210485 }, { "epoch": 0.53, "learning_rate": 4.73028335598168e-05, "loss": 1.0452, "step": 210490 }, { "epoch": 0.53, "learning_rate": 4.730157531833509e-05, "loss": 1.0444, "step": 210495 }, { "epoch": 0.53, "learning_rate": 4.730031707685339e-05, "loss": 1.0451, "step": 210500 }, { "epoch": 0.53, "learning_rate": 4.729905883537168e-05, "loss": 1.044, "step": 210505 }, { "epoch": 0.53, "learning_rate": 4.729780059388998e-05, "loss": 1.0447, "step": 210510 }, { "epoch": 0.53, "learning_rate": 4.729654235240827e-05, "loss": 1.0464, "step": 210515 }, { "epoch": 0.53, "learning_rate": 4.729528411092657e-05, "loss": 1.0458, "step": 210520 }, { "epoch": 0.53, "learning_rate": 4.729402586944486e-05, "loss": 1.0444, "step": 210525 }, { "epoch": 0.53, "learning_rate": 4.729276762796316e-05, "loss": 1.0419, "step": 210530 }, { "epoch": 0.53, "learning_rate": 4.729150938648146e-05, "loss": 1.0457, "step": 210535 }, { "epoch": 0.53, "learning_rate": 4.729025114499975e-05, "loss": 1.0468, "step": 210540 }, { "epoch": 0.53, "learning_rate": 4.728899290351805e-05, "loss": 1.0479, "step": 210545 }, { "epoch": 0.53, "learning_rate": 4.728773466203634e-05, "loss": 1.044, "step": 210550 }, { "epoch": 0.53, "learning_rate": 4.728647642055464e-05, "loss": 1.0454, "step": 210555 }, { "epoch": 0.53, "learning_rate": 4.728521817907293e-05, "loss": 1.0434, "step": 210560 }, { "epoch": 0.53, "learning_rate": 4.728395993759123e-05, "loss": 1.0435, "step": 210565 }, { "epoch": 0.53, "learning_rate": 4.728270169610952e-05, "loss": 1.0473, "step": 210570 }, { "epoch": 0.53, "learning_rate": 4.728144345462782e-05, "loss": 1.0452, "step": 210575 }, { "epoch": 0.53, "learning_rate": 4.728018521314611e-05, "loss": 1.0443, "step": 210580 }, { "epoch": 0.53, "learning_rate": 4.72789269716644e-05, "loss": 1.0432, "step": 210585 }, { "epoch": 0.53, "learning_rate": 4.72776687301827e-05, "loss": 1.0458, "step": 210590 }, { "epoch": 0.53, "learning_rate": 4.727641048870099e-05, "loss": 1.0457, "step": 210595 }, { "epoch": 0.53, "learning_rate": 4.727515224721929e-05, "loss": 1.0423, "step": 210600 }, { "epoch": 0.53, "learning_rate": 4.727389400573758e-05, "loss": 1.0467, "step": 210605 }, { "epoch": 0.53, "learning_rate": 4.727263576425588e-05, "loss": 1.048, "step": 210610 }, { "epoch": 0.53, "learning_rate": 4.727137752277417e-05, "loss": 1.043, "step": 210615 }, { "epoch": 0.53, "learning_rate": 4.727011928129247e-05, "loss": 1.0433, "step": 210620 }, { "epoch": 0.53, "learning_rate": 4.726886103981076e-05, "loss": 1.0444, "step": 210625 }, { "epoch": 0.53, "learning_rate": 4.726760279832906e-05, "loss": 1.0457, "step": 210630 }, { "epoch": 0.53, "learning_rate": 4.726634455684735e-05, "loss": 1.0434, "step": 210635 }, { "epoch": 0.53, "learning_rate": 4.7265086315365644e-05, "loss": 1.046, "step": 210640 }, { "epoch": 0.53, "learning_rate": 4.726382807388394e-05, "loss": 1.0463, "step": 210645 }, { "epoch": 0.53, "learning_rate": 4.7262569832402234e-05, "loss": 1.0448, "step": 210650 }, { "epoch": 0.53, "learning_rate": 4.726131159092053e-05, "loss": 1.0414, "step": 210655 }, { "epoch": 0.53, "learning_rate": 4.7260053349438824e-05, "loss": 1.0433, "step": 210660 }, { "epoch": 0.53, "learning_rate": 4.725879510795712e-05, "loss": 1.0445, "step": 210665 }, { "epoch": 0.53, "learning_rate": 4.7257536866475414e-05, "loss": 1.0447, "step": 210670 }, { "epoch": 0.53, "learning_rate": 4.725627862499371e-05, "loss": 1.0435, "step": 210675 }, { "epoch": 0.53, "learning_rate": 4.7255020383512004e-05, "loss": 1.0452, "step": 210680 }, { "epoch": 0.53, "learning_rate": 4.72537621420303e-05, "loss": 1.0497, "step": 210685 }, { "epoch": 0.53, "learning_rate": 4.7252503900548594e-05, "loss": 1.0454, "step": 210690 }, { "epoch": 0.53, "learning_rate": 4.725124565906689e-05, "loss": 1.0448, "step": 210695 }, { "epoch": 0.53, "learning_rate": 4.7249987417585184e-05, "loss": 1.045, "step": 210700 }, { "epoch": 0.53, "learning_rate": 4.7248729176103476e-05, "loss": 1.0504, "step": 210705 }, { "epoch": 0.53, "learning_rate": 4.7247470934621774e-05, "loss": 1.0445, "step": 210710 }, { "epoch": 0.53, "learning_rate": 4.7246212693140066e-05, "loss": 1.0448, "step": 210715 }, { "epoch": 0.53, "learning_rate": 4.7244954451658364e-05, "loss": 1.0416, "step": 210720 }, { "epoch": 0.53, "learning_rate": 4.7243696210176656e-05, "loss": 1.0433, "step": 210725 }, { "epoch": 0.53, "learning_rate": 4.7242437968694954e-05, "loss": 1.0436, "step": 210730 }, { "epoch": 0.53, "learning_rate": 4.7241179727213246e-05, "loss": 1.0431, "step": 210735 }, { "epoch": 0.53, "learning_rate": 4.7239921485731544e-05, "loss": 1.0419, "step": 210740 }, { "epoch": 0.53, "learning_rate": 4.7238663244249836e-05, "loss": 1.0427, "step": 210745 }, { "epoch": 0.53, "learning_rate": 4.7237405002768134e-05, "loss": 1.0438, "step": 210750 }, { "epoch": 0.53, "learning_rate": 4.7236146761286426e-05, "loss": 1.0419, "step": 210755 }, { "epoch": 0.53, "learning_rate": 4.7234888519804724e-05, "loss": 1.0423, "step": 210760 }, { "epoch": 0.53, "learning_rate": 4.7233630278323015e-05, "loss": 1.0447, "step": 210765 }, { "epoch": 0.53, "learning_rate": 4.723237203684131e-05, "loss": 1.0439, "step": 210770 }, { "epoch": 0.53, "learning_rate": 4.7231113795359605e-05, "loss": 1.0436, "step": 210775 }, { "epoch": 0.53, "learning_rate": 4.7229855553877904e-05, "loss": 1.0437, "step": 210780 }, { "epoch": 0.53, "learning_rate": 4.72285973123962e-05, "loss": 1.0439, "step": 210785 }, { "epoch": 0.53, "learning_rate": 4.7227339070914494e-05, "loss": 1.0409, "step": 210790 }, { "epoch": 0.53, "learning_rate": 4.722608082943279e-05, "loss": 1.0422, "step": 210795 }, { "epoch": 0.53, "learning_rate": 4.7224822587951084e-05, "loss": 1.046, "step": 210800 }, { "epoch": 0.53, "learning_rate": 4.722356434646938e-05, "loss": 1.0442, "step": 210805 }, { "epoch": 0.53, "learning_rate": 4.7222306104987674e-05, "loss": 1.0886, "step": 210810 }, { "epoch": 0.53, "learning_rate": 4.7221047863505965e-05, "loss": 1.0447, "step": 210815 }, { "epoch": 0.53, "learning_rate": 4.7219789622024264e-05, "loss": 1.043, "step": 210820 }, { "epoch": 0.53, "learning_rate": 4.7218531380542555e-05, "loss": 1.0436, "step": 210825 }, { "epoch": 0.53, "learning_rate": 4.7217273139060854e-05, "loss": 1.065, "step": 210830 }, { "epoch": 0.53, "learning_rate": 4.7216014897579145e-05, "loss": 1.0434, "step": 210835 }, { "epoch": 0.53, "learning_rate": 4.7214756656097444e-05, "loss": 1.0443, "step": 210840 }, { "epoch": 0.53, "learning_rate": 4.7213498414615735e-05, "loss": 1.0478, "step": 210845 }, { "epoch": 0.53, "learning_rate": 4.7212240173134034e-05, "loss": 1.0435, "step": 210850 }, { "epoch": 0.53, "learning_rate": 4.7210981931652325e-05, "loss": 1.0443, "step": 210855 }, { "epoch": 0.53, "learning_rate": 4.7209723690170624e-05, "loss": 1.0453, "step": 210860 }, { "epoch": 0.53, "learning_rate": 4.7208465448688915e-05, "loss": 1.0442, "step": 210865 }, { "epoch": 0.53, "learning_rate": 4.7207207207207214e-05, "loss": 1.0651, "step": 210870 }, { "epoch": 0.53, "learning_rate": 4.7205948965725505e-05, "loss": 1.0467, "step": 210875 }, { "epoch": 0.53, "learning_rate": 4.72046907242438e-05, "loss": 1.0462, "step": 210880 }, { "epoch": 0.53, "learning_rate": 4.7203432482762095e-05, "loss": 1.0641, "step": 210885 }, { "epoch": 0.53, "learning_rate": 4.720217424128039e-05, "loss": 1.0445, "step": 210890 }, { "epoch": 0.53, "learning_rate": 4.7200915999798685e-05, "loss": 1.0452, "step": 210895 }, { "epoch": 0.53, "learning_rate": 4.719965775831698e-05, "loss": 1.046, "step": 210900 }, { "epoch": 0.53, "learning_rate": 4.7198399516835275e-05, "loss": 1.0434, "step": 210905 }, { "epoch": 0.53, "learning_rate": 4.719714127535357e-05, "loss": 1.0465, "step": 210910 }, { "epoch": 0.53, "learning_rate": 4.7195883033871865e-05, "loss": 1.0456, "step": 210915 }, { "epoch": 0.53, "learning_rate": 4.719462479239016e-05, "loss": 1.0452, "step": 210920 }, { "epoch": 0.53, "learning_rate": 4.7193366550908455e-05, "loss": 1.0453, "step": 210925 }, { "epoch": 0.53, "learning_rate": 4.719210830942675e-05, "loss": 1.0644, "step": 210930 }, { "epoch": 0.53, "learning_rate": 4.719085006794504e-05, "loss": 1.0429, "step": 210935 }, { "epoch": 0.53, "learning_rate": 4.7189591826463337e-05, "loss": 1.0453, "step": 210940 }, { "epoch": 0.53, "learning_rate": 4.718833358498163e-05, "loss": 1.0426, "step": 210945 }, { "epoch": 0.53, "learning_rate": 4.7187075343499927e-05, "loss": 1.0431, "step": 210950 }, { "epoch": 0.53, "learning_rate": 4.718581710201822e-05, "loss": 1.0425, "step": 210955 }, { "epoch": 0.53, "learning_rate": 4.7184558860536517e-05, "loss": 1.045, "step": 210960 }, { "epoch": 0.53, "learning_rate": 4.718330061905481e-05, "loss": 1.0441, "step": 210965 }, { "epoch": 0.53, "learning_rate": 4.7182042377573107e-05, "loss": 1.0436, "step": 210970 }, { "epoch": 0.53, "learning_rate": 4.71807841360914e-05, "loss": 1.0433, "step": 210975 }, { "epoch": 0.53, "learning_rate": 4.7179525894609696e-05, "loss": 1.0446, "step": 210980 }, { "epoch": 0.53, "learning_rate": 4.717826765312799e-05, "loss": 1.0444, "step": 210985 }, { "epoch": 0.53, "learning_rate": 4.7177009411646286e-05, "loss": 1.0449, "step": 210990 }, { "epoch": 0.53, "learning_rate": 4.717575117016458e-05, "loss": 1.0463, "step": 210995 }, { "epoch": 0.53, "learning_rate": 4.717449292868287e-05, "loss": 1.0443, "step": 211000 }, { "epoch": 0.53, "learning_rate": 4.717323468720117e-05, "loss": 1.0428, "step": 211005 }, { "epoch": 0.53, "learning_rate": 4.717197644571946e-05, "loss": 1.0448, "step": 211010 }, { "epoch": 0.53, "learning_rate": 4.717071820423776e-05, "loss": 1.043, "step": 211015 }, { "epoch": 0.53, "learning_rate": 4.716945996275605e-05, "loss": 1.0458, "step": 211020 }, { "epoch": 0.53, "learning_rate": 4.716820172127435e-05, "loss": 1.0426, "step": 211025 }, { "epoch": 0.53, "learning_rate": 4.7166943479792646e-05, "loss": 1.0458, "step": 211030 }, { "epoch": 0.53, "learning_rate": 4.7165685238310945e-05, "loss": 1.0452, "step": 211035 }, { "epoch": 0.53, "learning_rate": 4.7164426996829236e-05, "loss": 1.066, "step": 211040 }, { "epoch": 0.53, "learning_rate": 4.716316875534753e-05, "loss": 1.0442, "step": 211045 }, { "epoch": 0.53, "learning_rate": 4.7161910513865826e-05, "loss": 1.0419, "step": 211050 }, { "epoch": 0.53, "learning_rate": 4.716065227238412e-05, "loss": 1.0445, "step": 211055 }, { "epoch": 0.53, "learning_rate": 4.7159394030902416e-05, "loss": 1.0442, "step": 211060 }, { "epoch": 0.53, "learning_rate": 4.715813578942071e-05, "loss": 1.0459, "step": 211065 }, { "epoch": 0.53, "learning_rate": 4.7156877547939006e-05, "loss": 1.0449, "step": 211070 }, { "epoch": 0.53, "learning_rate": 4.71556193064573e-05, "loss": 1.0463, "step": 211075 }, { "epoch": 0.53, "learning_rate": 4.7154361064975596e-05, "loss": 1.0423, "step": 211080 }, { "epoch": 0.53, "learning_rate": 4.715310282349389e-05, "loss": 1.0544, "step": 211085 }, { "epoch": 0.53, "learning_rate": 4.7151844582012186e-05, "loss": 1.0435, "step": 211090 }, { "epoch": 0.53, "learning_rate": 4.715058634053048e-05, "loss": 1.0443, "step": 211095 }, { "epoch": 0.53, "learning_rate": 4.7149328099048776e-05, "loss": 1.0446, "step": 211100 }, { "epoch": 0.53, "learning_rate": 4.714806985756707e-05, "loss": 1.0416, "step": 211105 }, { "epoch": 0.53, "learning_rate": 4.714681161608536e-05, "loss": 1.0447, "step": 211110 }, { "epoch": 0.53, "learning_rate": 4.714555337460366e-05, "loss": 1.0468, "step": 211115 }, { "epoch": 0.53, "learning_rate": 4.714429513312195e-05, "loss": 1.0449, "step": 211120 }, { "epoch": 0.53, "learning_rate": 4.714303689164025e-05, "loss": 1.0418, "step": 211125 }, { "epoch": 0.53, "learning_rate": 4.714177865015854e-05, "loss": 1.0449, "step": 211130 }, { "epoch": 0.53, "learning_rate": 4.714052040867684e-05, "loss": 1.0466, "step": 211135 }, { "epoch": 0.53, "learning_rate": 4.713926216719513e-05, "loss": 1.0449, "step": 211140 }, { "epoch": 0.53, "learning_rate": 4.713800392571343e-05, "loss": 1.0444, "step": 211145 }, { "epoch": 0.53, "learning_rate": 4.713674568423172e-05, "loss": 1.0449, "step": 211150 }, { "epoch": 0.53, "learning_rate": 4.713548744275002e-05, "loss": 1.0465, "step": 211155 }, { "epoch": 0.53, "learning_rate": 4.713422920126831e-05, "loss": 1.0415, "step": 211160 }, { "epoch": 0.53, "learning_rate": 4.713297095978661e-05, "loss": 1.0438, "step": 211165 }, { "epoch": 0.53, "learning_rate": 4.71317127183049e-05, "loss": 1.0451, "step": 211170 }, { "epoch": 0.53, "learning_rate": 4.713045447682319e-05, "loss": 1.0443, "step": 211175 }, { "epoch": 0.53, "learning_rate": 4.712919623534149e-05, "loss": 1.0554, "step": 211180 }, { "epoch": 0.53, "learning_rate": 4.712793799385978e-05, "loss": 1.0445, "step": 211185 }, { "epoch": 0.53, "learning_rate": 4.712667975237808e-05, "loss": 1.0418, "step": 211190 }, { "epoch": 0.53, "learning_rate": 4.712542151089637e-05, "loss": 1.058, "step": 211195 }, { "epoch": 0.53, "learning_rate": 4.712416326941467e-05, "loss": 1.0457, "step": 211200 }, { "epoch": 0.53, "learning_rate": 4.712290502793296e-05, "loss": 1.0443, "step": 211205 }, { "epoch": 0.53, "learning_rate": 4.712164678645126e-05, "loss": 1.0438, "step": 211210 }, { "epoch": 0.53, "learning_rate": 4.712038854496955e-05, "loss": 1.0425, "step": 211215 }, { "epoch": 0.53, "learning_rate": 4.711913030348785e-05, "loss": 1.0429, "step": 211220 }, { "epoch": 0.53, "learning_rate": 4.711787206200614e-05, "loss": 1.0428, "step": 211225 }, { "epoch": 0.53, "learning_rate": 4.711661382052444e-05, "loss": 1.0441, "step": 211230 }, { "epoch": 0.53, "learning_rate": 4.711535557904273e-05, "loss": 1.0451, "step": 211235 }, { "epoch": 0.53, "learning_rate": 4.711409733756102e-05, "loss": 1.0431, "step": 211240 }, { "epoch": 0.53, "learning_rate": 4.711283909607932e-05, "loss": 1.0668, "step": 211245 }, { "epoch": 0.53, "learning_rate": 4.711158085459761e-05, "loss": 1.045, "step": 211250 }, { "epoch": 0.53, "learning_rate": 4.711032261311591e-05, "loss": 1.0438, "step": 211255 }, { "epoch": 0.53, "learning_rate": 4.71090643716342e-05, "loss": 1.0444, "step": 211260 }, { "epoch": 0.53, "learning_rate": 4.71078061301525e-05, "loss": 1.0411, "step": 211265 }, { "epoch": 0.53, "learning_rate": 4.710654788867079e-05, "loss": 1.0455, "step": 211270 }, { "epoch": 0.53, "learning_rate": 4.710528964718909e-05, "loss": 1.0443, "step": 211275 }, { "epoch": 0.53, "learning_rate": 4.710403140570739e-05, "loss": 1.044, "step": 211280 }, { "epoch": 0.53, "learning_rate": 4.710277316422568e-05, "loss": 1.043, "step": 211285 }, { "epoch": 0.53, "learning_rate": 4.710151492274398e-05, "loss": 1.0434, "step": 211290 }, { "epoch": 0.53, "learning_rate": 4.710025668126227e-05, "loss": 1.0436, "step": 211295 }, { "epoch": 0.53, "learning_rate": 4.709899843978057e-05, "loss": 1.0406, "step": 211300 }, { "epoch": 0.53, "learning_rate": 4.709774019829886e-05, "loss": 1.0418, "step": 211305 }, { "epoch": 0.53, "learning_rate": 4.709648195681716e-05, "loss": 1.0422, "step": 211310 }, { "epoch": 0.53, "learning_rate": 4.709522371533545e-05, "loss": 1.0446, "step": 211315 }, { "epoch": 0.53, "learning_rate": 4.709396547385375e-05, "loss": 1.0464, "step": 211320 }, { "epoch": 0.53, "learning_rate": 4.709270723237204e-05, "loss": 1.0434, "step": 211325 }, { "epoch": 0.53, "learning_rate": 4.709144899089034e-05, "loss": 1.0483, "step": 211330 }, { "epoch": 0.53, "learning_rate": 4.709019074940863e-05, "loss": 1.0417, "step": 211335 }, { "epoch": 0.53, "learning_rate": 4.708893250792692e-05, "loss": 1.0436, "step": 211340 }, { "epoch": 0.53, "learning_rate": 4.708767426644522e-05, "loss": 1.0436, "step": 211345 }, { "epoch": 0.53, "learning_rate": 4.708641602496351e-05, "loss": 1.0419, "step": 211350 }, { "epoch": 0.53, "learning_rate": 4.708515778348181e-05, "loss": 1.0436, "step": 211355 }, { "epoch": 0.53, "learning_rate": 4.70838995420001e-05, "loss": 1.0464, "step": 211360 }, { "epoch": 0.53, "learning_rate": 4.70826413005184e-05, "loss": 1.0443, "step": 211365 }, { "epoch": 0.53, "learning_rate": 4.708138305903669e-05, "loss": 1.0447, "step": 211370 }, { "epoch": 0.53, "learning_rate": 4.708012481755499e-05, "loss": 1.0436, "step": 211375 }, { "epoch": 0.53, "learning_rate": 4.707886657607328e-05, "loss": 1.0472, "step": 211380 }, { "epoch": 0.53, "learning_rate": 4.707760833459158e-05, "loss": 1.0409, "step": 211385 }, { "epoch": 0.53, "learning_rate": 4.707635009310987e-05, "loss": 1.0433, "step": 211390 }, { "epoch": 0.53, "learning_rate": 4.707509185162817e-05, "loss": 1.0449, "step": 211395 }, { "epoch": 0.53, "learning_rate": 4.707383361014646e-05, "loss": 1.0435, "step": 211400 }, { "epoch": 0.53, "learning_rate": 4.707257536866475e-05, "loss": 1.045, "step": 211405 }, { "epoch": 0.53, "learning_rate": 4.707131712718305e-05, "loss": 1.0436, "step": 211410 }, { "epoch": 0.53, "learning_rate": 4.707005888570134e-05, "loss": 1.0441, "step": 211415 }, { "epoch": 0.53, "learning_rate": 4.706880064421964e-05, "loss": 1.0442, "step": 211420 }, { "epoch": 0.53, "learning_rate": 4.706754240273793e-05, "loss": 1.0412, "step": 211425 }, { "epoch": 0.53, "learning_rate": 4.706628416125623e-05, "loss": 1.0447, "step": 211430 }, { "epoch": 0.53, "learning_rate": 4.706502591977452e-05, "loss": 1.0442, "step": 211435 }, { "epoch": 0.53, "learning_rate": 4.706376767829282e-05, "loss": 1.0435, "step": 211440 }, { "epoch": 0.53, "learning_rate": 4.706250943681111e-05, "loss": 1.0439, "step": 211445 }, { "epoch": 0.53, "learning_rate": 4.706125119532941e-05, "loss": 1.0667, "step": 211450 }, { "epoch": 0.53, "learning_rate": 4.70599929538477e-05, "loss": 1.0425, "step": 211455 }, { "epoch": 0.53, "learning_rate": 4.7058734712366e-05, "loss": 1.0476, "step": 211460 }, { "epoch": 0.53, "learning_rate": 4.705747647088429e-05, "loss": 1.0446, "step": 211465 }, { "epoch": 0.53, "learning_rate": 4.7056218229402585e-05, "loss": 1.0451, "step": 211470 }, { "epoch": 0.53, "learning_rate": 4.705495998792088e-05, "loss": 1.0427, "step": 211475 }, { "epoch": 0.53, "learning_rate": 4.7053701746439175e-05, "loss": 1.0434, "step": 211480 }, { "epoch": 0.53, "learning_rate": 4.705244350495747e-05, "loss": 1.0457, "step": 211485 }, { "epoch": 0.53, "learning_rate": 4.7051185263475765e-05, "loss": 1.0444, "step": 211490 }, { "epoch": 0.53, "learning_rate": 4.704992702199406e-05, "loss": 1.042, "step": 211495 }, { "epoch": 0.53, "learning_rate": 4.7048668780512355e-05, "loss": 1.0412, "step": 211500 }, { "epoch": 0.53, "learning_rate": 4.704741053903065e-05, "loss": 1.044, "step": 211505 }, { "epoch": 0.53, "learning_rate": 4.7046152297548945e-05, "loss": 1.0441, "step": 211510 }, { "epoch": 0.53, "learning_rate": 4.704489405606724e-05, "loss": 1.0647, "step": 211515 }, { "epoch": 0.53, "learning_rate": 4.7043635814585535e-05, "loss": 1.0452, "step": 211520 }, { "epoch": 0.53, "learning_rate": 4.704237757310383e-05, "loss": 1.0458, "step": 211525 }, { "epoch": 0.53, "learning_rate": 4.704111933162213e-05, "loss": 1.0429, "step": 211530 }, { "epoch": 0.53, "learning_rate": 4.703986109014042e-05, "loss": 1.0633, "step": 211535 }, { "epoch": 0.53, "learning_rate": 4.703860284865872e-05, "loss": 1.0453, "step": 211540 }, { "epoch": 0.53, "learning_rate": 4.703734460717701e-05, "loss": 1.0448, "step": 211545 }, { "epoch": 0.53, "learning_rate": 4.703608636569531e-05, "loss": 1.0442, "step": 211550 }, { "epoch": 0.53, "learning_rate": 4.70348281242136e-05, "loss": 1.0416, "step": 211555 }, { "epoch": 0.53, "learning_rate": 4.70335698827319e-05, "loss": 1.0608, "step": 211560 }, { "epoch": 0.53, "learning_rate": 4.703231164125019e-05, "loss": 1.0447, "step": 211565 }, { "epoch": 0.53, "learning_rate": 4.7031053399768484e-05, "loss": 1.0441, "step": 211570 }, { "epoch": 0.53, "learning_rate": 4.702979515828678e-05, "loss": 1.043, "step": 211575 }, { "epoch": 0.53, "learning_rate": 4.7028536916805074e-05, "loss": 1.0433, "step": 211580 }, { "epoch": 0.53, "learning_rate": 4.702727867532337e-05, "loss": 1.0463, "step": 211585 }, { "epoch": 0.53, "learning_rate": 4.7026020433841664e-05, "loss": 1.0444, "step": 211590 }, { "epoch": 0.53, "learning_rate": 4.702476219235996e-05, "loss": 1.0446, "step": 211595 }, { "epoch": 0.53, "learning_rate": 4.7023503950878254e-05, "loss": 1.0411, "step": 211600 }, { "epoch": 0.53, "learning_rate": 4.702224570939655e-05, "loss": 1.043, "step": 211605 }, { "epoch": 0.53, "learning_rate": 4.7020987467914844e-05, "loss": 1.0464, "step": 211610 }, { "epoch": 0.53, "learning_rate": 4.701972922643314e-05, "loss": 1.0433, "step": 211615 }, { "epoch": 0.53, "learning_rate": 4.7018470984951434e-05, "loss": 1.0445, "step": 211620 }, { "epoch": 0.53, "learning_rate": 4.701721274346973e-05, "loss": 1.044, "step": 211625 }, { "epoch": 0.53, "learning_rate": 4.7015954501988024e-05, "loss": 1.043, "step": 211630 }, { "epoch": 0.53, "learning_rate": 4.7014696260506316e-05, "loss": 1.0457, "step": 211635 }, { "epoch": 0.53, "learning_rate": 4.7013438019024614e-05, "loss": 1.0438, "step": 211640 }, { "epoch": 0.53, "learning_rate": 4.7012179777542906e-05, "loss": 1.0427, "step": 211645 }, { "epoch": 0.53, "learning_rate": 4.7010921536061204e-05, "loss": 1.0465, "step": 211650 }, { "epoch": 0.53, "learning_rate": 4.7009663294579496e-05, "loss": 1.0543, "step": 211655 }, { "epoch": 0.53, "learning_rate": 4.7008405053097794e-05, "loss": 1.042, "step": 211660 }, { "epoch": 0.53, "learning_rate": 4.7007146811616086e-05, "loss": 1.0432, "step": 211665 }, { "epoch": 0.53, "learning_rate": 4.7005888570134384e-05, "loss": 1.0409, "step": 211670 }, { "epoch": 0.53, "learning_rate": 4.7004630328652676e-05, "loss": 1.0458, "step": 211675 }, { "epoch": 0.53, "learning_rate": 4.7003372087170974e-05, "loss": 1.0593, "step": 211680 }, { "epoch": 0.53, "learning_rate": 4.7002113845689266e-05, "loss": 1.0449, "step": 211685 }, { "epoch": 0.53, "learning_rate": 4.7000855604207564e-05, "loss": 1.0478, "step": 211690 }, { "epoch": 0.53, "learning_rate": 4.6999597362725856e-05, "loss": 1.0427, "step": 211695 }, { "epoch": 0.53, "learning_rate": 4.699833912124415e-05, "loss": 1.0485, "step": 211700 }, { "epoch": 0.53, "learning_rate": 4.6997080879762446e-05, "loss": 1.0439, "step": 211705 }, { "epoch": 0.53, "learning_rate": 4.699582263828074e-05, "loss": 1.0445, "step": 211710 }, { "epoch": 0.53, "learning_rate": 4.6994564396799036e-05, "loss": 1.0429, "step": 211715 }, { "epoch": 0.53, "learning_rate": 4.699330615531733e-05, "loss": 1.0438, "step": 211720 }, { "epoch": 0.53, "learning_rate": 4.6992047913835626e-05, "loss": 1.0475, "step": 211725 }, { "epoch": 0.53, "learning_rate": 4.699078967235392e-05, "loss": 1.0433, "step": 211730 }, { "epoch": 0.53, "learning_rate": 4.6989531430872216e-05, "loss": 1.0428, "step": 211735 }, { "epoch": 0.53, "learning_rate": 4.698827318939051e-05, "loss": 1.0406, "step": 211740 }, { "epoch": 0.53, "learning_rate": 4.6987014947908806e-05, "loss": 1.0451, "step": 211745 }, { "epoch": 0.53, "learning_rate": 4.69857567064271e-05, "loss": 1.0435, "step": 211750 }, { "epoch": 0.53, "learning_rate": 4.6984498464945396e-05, "loss": 1.046, "step": 211755 }, { "epoch": 0.53, "learning_rate": 4.698324022346369e-05, "loss": 1.0465, "step": 211760 }, { "epoch": 0.53, "learning_rate": 4.698198198198198e-05, "loss": 1.0437, "step": 211765 }, { "epoch": 0.53, "learning_rate": 4.698072374050028e-05, "loss": 1.0416, "step": 211770 }, { "epoch": 0.53, "learning_rate": 4.697946549901857e-05, "loss": 1.0423, "step": 211775 }, { "epoch": 0.53, "learning_rate": 4.6978207257536874e-05, "loss": 1.0425, "step": 211780 }, { "epoch": 0.53, "learning_rate": 4.6976949016055165e-05, "loss": 1.0654, "step": 211785 }, { "epoch": 0.53, "learning_rate": 4.6975690774573464e-05, "loss": 1.0415, "step": 211790 }, { "epoch": 0.53, "learning_rate": 4.6974432533091755e-05, "loss": 1.0489, "step": 211795 }, { "epoch": 0.53, "learning_rate": 4.697317429161005e-05, "loss": 1.0441, "step": 211800 }, { "epoch": 0.53, "learning_rate": 4.6971916050128345e-05, "loss": 1.0431, "step": 211805 }, { "epoch": 0.53, "learning_rate": 4.697065780864664e-05, "loss": 1.0494, "step": 211810 }, { "epoch": 0.53, "learning_rate": 4.6969399567164935e-05, "loss": 1.0441, "step": 211815 }, { "epoch": 0.53, "learning_rate": 4.696814132568323e-05, "loss": 1.0404, "step": 211820 }, { "epoch": 0.53, "learning_rate": 4.6966883084201525e-05, "loss": 1.0436, "step": 211825 }, { "epoch": 0.53, "learning_rate": 4.696562484271982e-05, "loss": 1.0429, "step": 211830 }, { "epoch": 0.53, "learning_rate": 4.6964366601238115e-05, "loss": 1.0436, "step": 211835 }, { "epoch": 0.53, "learning_rate": 4.696310835975641e-05, "loss": 1.0449, "step": 211840 }, { "epoch": 0.53, "learning_rate": 4.6961850118274705e-05, "loss": 1.0475, "step": 211845 }, { "epoch": 0.53, "learning_rate": 4.6960591876793e-05, "loss": 1.0455, "step": 211850 }, { "epoch": 0.53, "learning_rate": 4.6959333635311295e-05, "loss": 1.0423, "step": 211855 }, { "epoch": 0.53, "learning_rate": 4.695807539382959e-05, "loss": 1.0462, "step": 211860 }, { "epoch": 0.53, "learning_rate": 4.695681715234788e-05, "loss": 1.0453, "step": 211865 }, { "epoch": 0.53, "learning_rate": 4.695555891086618e-05, "loss": 1.0924, "step": 211870 }, { "epoch": 0.53, "learning_rate": 4.695430066938447e-05, "loss": 1.0477, "step": 211875 }, { "epoch": 0.53, "learning_rate": 4.695304242790277e-05, "loss": 1.0464, "step": 211880 }, { "epoch": 0.53, "learning_rate": 4.695178418642106e-05, "loss": 1.0455, "step": 211885 }, { "epoch": 0.53, "learning_rate": 4.695052594493936e-05, "loss": 1.0453, "step": 211890 }, { "epoch": 0.53, "learning_rate": 4.694926770345765e-05, "loss": 1.044, "step": 211895 }, { "epoch": 0.53, "learning_rate": 4.694800946197595e-05, "loss": 1.0421, "step": 211900 }, { "epoch": 0.53, "learning_rate": 4.694675122049424e-05, "loss": 1.0482, "step": 211905 }, { "epoch": 0.53, "learning_rate": 4.694549297901254e-05, "loss": 1.0446, "step": 211910 }, { "epoch": 0.53, "learning_rate": 4.694423473753083e-05, "loss": 1.0428, "step": 211915 }, { "epoch": 0.53, "learning_rate": 4.694297649604913e-05, "loss": 1.0476, "step": 211920 }, { "epoch": 0.53, "learning_rate": 4.694171825456742e-05, "loss": 1.0437, "step": 211925 }, { "epoch": 0.53, "learning_rate": 4.694046001308571e-05, "loss": 1.0446, "step": 211930 }, { "epoch": 0.53, "learning_rate": 4.693920177160401e-05, "loss": 1.0444, "step": 211935 }, { "epoch": 0.53, "learning_rate": 4.69379435301223e-05, "loss": 1.0447, "step": 211940 }, { "epoch": 0.53, "learning_rate": 4.69366852886406e-05, "loss": 1.0442, "step": 211945 }, { "epoch": 0.53, "learning_rate": 4.693542704715889e-05, "loss": 1.0446, "step": 211950 }, { "epoch": 0.53, "learning_rate": 4.693416880567719e-05, "loss": 1.0459, "step": 211955 }, { "epoch": 0.53, "learning_rate": 4.693291056419548e-05, "loss": 1.0443, "step": 211960 }, { "epoch": 0.53, "learning_rate": 4.693165232271378e-05, "loss": 1.0445, "step": 211965 }, { "epoch": 0.53, "learning_rate": 4.693039408123207e-05, "loss": 1.0454, "step": 211970 }, { "epoch": 0.53, "learning_rate": 4.692913583975037e-05, "loss": 1.0449, "step": 211975 }, { "epoch": 0.53, "learning_rate": 4.692787759826866e-05, "loss": 1.0431, "step": 211980 }, { "epoch": 0.53, "learning_rate": 4.692661935678696e-05, "loss": 1.0457, "step": 211985 }, { "epoch": 0.53, "learning_rate": 4.692536111530525e-05, "loss": 1.0449, "step": 211990 }, { "epoch": 0.53, "learning_rate": 4.692410287382354e-05, "loss": 1.0425, "step": 211995 }, { "epoch": 0.53, "learning_rate": 4.692284463234184e-05, "loss": 1.0438, "step": 212000 }, { "epoch": 0.53, "learning_rate": 4.692158639086013e-05, "loss": 1.0457, "step": 212005 }, { "epoch": 0.53, "learning_rate": 4.692032814937843e-05, "loss": 1.0434, "step": 212010 }, { "epoch": 0.53, "learning_rate": 4.691906990789672e-05, "loss": 1.0425, "step": 212015 }, { "epoch": 0.53, "learning_rate": 4.691781166641502e-05, "loss": 1.0463, "step": 212020 }, { "epoch": 0.53, "learning_rate": 4.691655342493331e-05, "loss": 1.0479, "step": 212025 }, { "epoch": 0.53, "learning_rate": 4.6915295183451616e-05, "loss": 1.0462, "step": 212030 }, { "epoch": 0.53, "learning_rate": 4.691403694196991e-05, "loss": 1.042, "step": 212035 }, { "epoch": 0.53, "learning_rate": 4.69127787004882e-05, "loss": 1.0784, "step": 212040 }, { "epoch": 0.53, "learning_rate": 4.69115204590065e-05, "loss": 1.0455, "step": 212045 }, { "epoch": 0.53, "learning_rate": 4.691026221752479e-05, "loss": 1.0431, "step": 212050 }, { "epoch": 0.53, "learning_rate": 4.690900397604309e-05, "loss": 1.0467, "step": 212055 }, { "epoch": 0.53, "learning_rate": 4.690774573456138e-05, "loss": 1.0406, "step": 212060 }, { "epoch": 0.53, "learning_rate": 4.690648749307968e-05, "loss": 1.0429, "step": 212065 }, { "epoch": 0.53, "learning_rate": 4.690522925159797e-05, "loss": 1.0453, "step": 212070 }, { "epoch": 0.53, "learning_rate": 4.690397101011627e-05, "loss": 1.0453, "step": 212075 }, { "epoch": 0.53, "learning_rate": 4.690271276863456e-05, "loss": 1.0446, "step": 212080 }, { "epoch": 0.53, "learning_rate": 4.690145452715286e-05, "loss": 1.0422, "step": 212085 }, { "epoch": 0.53, "learning_rate": 4.690019628567115e-05, "loss": 1.0442, "step": 212090 }, { "epoch": 0.53, "learning_rate": 4.689893804418944e-05, "loss": 1.0635, "step": 212095 }, { "epoch": 0.53, "learning_rate": 4.689767980270774e-05, "loss": 1.0429, "step": 212100 }, { "epoch": 0.53, "learning_rate": 4.689642156122603e-05, "loss": 1.0419, "step": 212105 }, { "epoch": 0.53, "learning_rate": 4.689516331974433e-05, "loss": 1.042, "step": 212110 }, { "epoch": 0.53, "learning_rate": 4.689390507826262e-05, "loss": 1.043, "step": 212115 }, { "epoch": 0.53, "learning_rate": 4.689264683678092e-05, "loss": 1.0438, "step": 212120 }, { "epoch": 0.53, "learning_rate": 4.689138859529921e-05, "loss": 1.0447, "step": 212125 }, { "epoch": 0.53, "learning_rate": 4.689013035381751e-05, "loss": 1.0437, "step": 212130 }, { "epoch": 0.53, "learning_rate": 4.68888721123358e-05, "loss": 1.0452, "step": 212135 }, { "epoch": 0.53, "learning_rate": 4.68876138708541e-05, "loss": 1.0446, "step": 212140 }, { "epoch": 0.53, "learning_rate": 4.688635562937239e-05, "loss": 1.047, "step": 212145 }, { "epoch": 0.53, "learning_rate": 4.688509738789069e-05, "loss": 1.0447, "step": 212150 }, { "epoch": 0.53, "learning_rate": 4.688383914640898e-05, "loss": 1.064, "step": 212155 }, { "epoch": 0.53, "learning_rate": 4.688258090492727e-05, "loss": 1.0466, "step": 212160 }, { "epoch": 0.53, "learning_rate": 4.688132266344557e-05, "loss": 1.0458, "step": 212165 }, { "epoch": 0.53, "learning_rate": 4.688006442196386e-05, "loss": 1.0441, "step": 212170 }, { "epoch": 0.53, "learning_rate": 4.687880618048216e-05, "loss": 1.0451, "step": 212175 }, { "epoch": 0.53, "learning_rate": 4.687754793900045e-05, "loss": 1.0412, "step": 212180 }, { "epoch": 0.53, "learning_rate": 4.687628969751875e-05, "loss": 1.0468, "step": 212185 }, { "epoch": 0.53, "learning_rate": 4.687503145603704e-05, "loss": 1.0691, "step": 212190 }, { "epoch": 0.53, "learning_rate": 4.687377321455534e-05, "loss": 1.0432, "step": 212195 }, { "epoch": 0.53, "learning_rate": 4.687251497307363e-05, "loss": 1.0399, "step": 212200 }, { "epoch": 0.53, "learning_rate": 4.687125673159193e-05, "loss": 1.0592, "step": 212205 }, { "epoch": 0.53, "learning_rate": 4.686999849011022e-05, "loss": 1.0497, "step": 212210 }, { "epoch": 0.53, "learning_rate": 4.686874024862852e-05, "loss": 1.0427, "step": 212215 }, { "epoch": 0.53, "learning_rate": 4.686748200714681e-05, "loss": 1.0443, "step": 212220 }, { "epoch": 0.53, "learning_rate": 4.6866223765665104e-05, "loss": 1.0483, "step": 212225 }, { "epoch": 0.53, "learning_rate": 4.68649655241834e-05, "loss": 1.0438, "step": 212230 }, { "epoch": 0.53, "learning_rate": 4.6863707282701694e-05, "loss": 1.0438, "step": 212235 }, { "epoch": 0.53, "learning_rate": 4.686244904121999e-05, "loss": 1.0429, "step": 212240 }, { "epoch": 0.53, "learning_rate": 4.6861190799738284e-05, "loss": 1.0442, "step": 212245 }, { "epoch": 0.53, "learning_rate": 4.685993255825658e-05, "loss": 1.042, "step": 212250 }, { "epoch": 0.53, "learning_rate": 4.6858674316774874e-05, "loss": 1.0459, "step": 212255 }, { "epoch": 0.53, "learning_rate": 4.685741607529317e-05, "loss": 1.0427, "step": 212260 }, { "epoch": 0.53, "learning_rate": 4.6856157833811464e-05, "loss": 1.042, "step": 212265 }, { "epoch": 0.53, "learning_rate": 4.685489959232976e-05, "loss": 1.0449, "step": 212270 }, { "epoch": 0.53, "learning_rate": 4.6853641350848054e-05, "loss": 1.0446, "step": 212275 }, { "epoch": 0.53, "learning_rate": 4.685238310936635e-05, "loss": 1.0464, "step": 212280 }, { "epoch": 0.53, "learning_rate": 4.685112486788465e-05, "loss": 1.0461, "step": 212285 }, { "epoch": 0.53, "learning_rate": 4.684986662640294e-05, "loss": 1.0445, "step": 212290 }, { "epoch": 0.53, "learning_rate": 4.684860838492124e-05, "loss": 1.0411, "step": 212295 }, { "epoch": 0.53, "learning_rate": 4.684735014343953e-05, "loss": 1.0437, "step": 212300 }, { "epoch": 0.53, "learning_rate": 4.684609190195783e-05, "loss": 1.0432, "step": 212305 }, { "epoch": 0.53, "learning_rate": 4.684483366047612e-05, "loss": 1.041, "step": 212310 }, { "epoch": 0.53, "learning_rate": 4.684357541899442e-05, "loss": 1.0447, "step": 212315 }, { "epoch": 0.53, "learning_rate": 4.684231717751271e-05, "loss": 1.0654, "step": 212320 }, { "epoch": 0.53, "learning_rate": 4.684105893603101e-05, "loss": 1.0438, "step": 212325 }, { "epoch": 0.53, "learning_rate": 4.68398006945493e-05, "loss": 1.0437, "step": 212330 }, { "epoch": 0.53, "learning_rate": 4.6838542453067594e-05, "loss": 1.0418, "step": 212335 }, { "epoch": 0.53, "learning_rate": 4.683728421158589e-05, "loss": 1.0417, "step": 212340 }, { "epoch": 0.53, "learning_rate": 4.6836025970104183e-05, "loss": 1.0454, "step": 212345 }, { "epoch": 0.53, "learning_rate": 4.683476772862248e-05, "loss": 1.0444, "step": 212350 }, { "epoch": 0.53, "learning_rate": 4.6833509487140773e-05, "loss": 1.0422, "step": 212355 }, { "epoch": 0.53, "learning_rate": 4.683225124565907e-05, "loss": 1.0481, "step": 212360 }, { "epoch": 0.53, "learning_rate": 4.6830993004177363e-05, "loss": 1.0468, "step": 212365 }, { "epoch": 0.53, "learning_rate": 4.682973476269566e-05, "loss": 1.0433, "step": 212370 }, { "epoch": 0.53, "learning_rate": 4.6828476521213953e-05, "loss": 1.04, "step": 212375 }, { "epoch": 0.53, "learning_rate": 4.682721827973225e-05, "loss": 1.0668, "step": 212380 }, { "epoch": 0.53, "learning_rate": 4.682596003825054e-05, "loss": 1.0446, "step": 212385 }, { "epoch": 0.53, "learning_rate": 4.682470179676884e-05, "loss": 1.0419, "step": 212390 }, { "epoch": 0.53, "learning_rate": 4.682344355528713e-05, "loss": 1.0433, "step": 212395 }, { "epoch": 0.53, "learning_rate": 4.6822185313805425e-05, "loss": 1.0444, "step": 212400 }, { "epoch": 0.53, "learning_rate": 4.682092707232372e-05, "loss": 1.0665, "step": 212405 }, { "epoch": 0.53, "learning_rate": 4.6819668830842015e-05, "loss": 1.0444, "step": 212410 }, { "epoch": 0.53, "learning_rate": 4.681841058936031e-05, "loss": 1.0697, "step": 212415 }, { "epoch": 0.53, "learning_rate": 4.6817152347878605e-05, "loss": 1.0444, "step": 212420 }, { "epoch": 0.53, "learning_rate": 4.68158941063969e-05, "loss": 1.0441, "step": 212425 }, { "epoch": 0.53, "learning_rate": 4.6814635864915195e-05, "loss": 1.0429, "step": 212430 }, { "epoch": 0.53, "learning_rate": 4.681337762343349e-05, "loss": 1.045, "step": 212435 }, { "epoch": 0.53, "learning_rate": 4.6812119381951785e-05, "loss": 1.0424, "step": 212440 }, { "epoch": 0.53, "learning_rate": 4.681086114047008e-05, "loss": 1.0431, "step": 212445 }, { "epoch": 0.53, "learning_rate": 4.6809602898988375e-05, "loss": 1.067, "step": 212450 }, { "epoch": 0.53, "learning_rate": 4.6808344657506666e-05, "loss": 1.0476, "step": 212455 }, { "epoch": 0.53, "learning_rate": 4.6807086416024965e-05, "loss": 1.0419, "step": 212460 }, { "epoch": 0.53, "learning_rate": 4.6805828174543256e-05, "loss": 1.068, "step": 212465 }, { "epoch": 0.53, "learning_rate": 4.6804569933061555e-05, "loss": 1.0431, "step": 212470 }, { "epoch": 0.53, "learning_rate": 4.6803311691579846e-05, "loss": 1.0421, "step": 212475 }, { "epoch": 0.53, "learning_rate": 4.6802053450098145e-05, "loss": 1.041, "step": 212480 }, { "epoch": 0.53, "learning_rate": 4.6800795208616436e-05, "loss": 1.0451, "step": 212485 }, { "epoch": 0.53, "learning_rate": 4.6799536967134735e-05, "loss": 1.043, "step": 212490 }, { "epoch": 0.53, "learning_rate": 4.6798278725653026e-05, "loss": 1.0413, "step": 212495 }, { "epoch": 0.53, "learning_rate": 4.6797020484171325e-05, "loss": 1.0448, "step": 212500 }, { "epoch": 0.53, "learning_rate": 4.6795762242689616e-05, "loss": 1.044, "step": 212505 }, { "epoch": 0.53, "learning_rate": 4.6794504001207915e-05, "loss": 1.0434, "step": 212510 }, { "epoch": 0.53, "learning_rate": 4.6793245759726206e-05, "loss": 1.0436, "step": 212515 }, { "epoch": 0.53, "learning_rate": 4.67919875182445e-05, "loss": 1.0456, "step": 212520 }, { "epoch": 0.53, "learning_rate": 4.6790729276762796e-05, "loss": 1.0442, "step": 212525 }, { "epoch": 0.53, "learning_rate": 4.6789471035281095e-05, "loss": 1.0413, "step": 212530 }, { "epoch": 0.53, "learning_rate": 4.678821279379939e-05, "loss": 1.0429, "step": 212535 }, { "epoch": 0.53, "learning_rate": 4.6786954552317685e-05, "loss": 1.0459, "step": 212540 }, { "epoch": 0.53, "learning_rate": 4.678569631083598e-05, "loss": 1.0411, "step": 212545 }, { "epoch": 0.53, "learning_rate": 4.6784438069354275e-05, "loss": 1.0449, "step": 212550 }, { "epoch": 0.53, "learning_rate": 4.678317982787257e-05, "loss": 1.0424, "step": 212555 }, { "epoch": 0.53, "learning_rate": 4.6781921586390864e-05, "loss": 1.0405, "step": 212560 }, { "epoch": 0.53, "learning_rate": 4.6780663344909156e-05, "loss": 1.0425, "step": 212565 }, { "epoch": 0.53, "learning_rate": 4.6779405103427454e-05, "loss": 1.0466, "step": 212570 }, { "epoch": 0.53, "learning_rate": 4.6778146861945746e-05, "loss": 1.0464, "step": 212575 }, { "epoch": 0.53, "learning_rate": 4.6776888620464044e-05, "loss": 1.0459, "step": 212580 }, { "epoch": 0.53, "learning_rate": 4.6775630378982336e-05, "loss": 1.0452, "step": 212585 }, { "epoch": 0.53, "learning_rate": 4.6774372137500634e-05, "loss": 1.0459, "step": 212590 }, { "epoch": 0.53, "learning_rate": 4.6773113896018926e-05, "loss": 1.0456, "step": 212595 }, { "epoch": 0.53, "learning_rate": 4.6771855654537224e-05, "loss": 1.0437, "step": 212600 }, { "epoch": 0.53, "learning_rate": 4.6770597413055516e-05, "loss": 1.0449, "step": 212605 }, { "epoch": 0.53, "learning_rate": 4.6769339171573814e-05, "loss": 1.0432, "step": 212610 }, { "epoch": 0.53, "learning_rate": 4.6768080930092106e-05, "loss": 1.0465, "step": 212615 }, { "epoch": 0.53, "learning_rate": 4.6766822688610404e-05, "loss": 1.0462, "step": 212620 }, { "epoch": 0.53, "learning_rate": 4.6765564447128696e-05, "loss": 1.0422, "step": 212625 }, { "epoch": 0.53, "learning_rate": 4.676430620564699e-05, "loss": 1.0399, "step": 212630 }, { "epoch": 0.53, "learning_rate": 4.6763047964165286e-05, "loss": 1.0412, "step": 212635 }, { "epoch": 0.53, "learning_rate": 4.676178972268358e-05, "loss": 1.0433, "step": 212640 }, { "epoch": 0.53, "learning_rate": 4.6760531481201876e-05, "loss": 1.0473, "step": 212645 }, { "epoch": 0.53, "learning_rate": 4.675927323972017e-05, "loss": 1.0474, "step": 212650 }, { "epoch": 0.53, "learning_rate": 4.6758014998238466e-05, "loss": 1.0449, "step": 212655 }, { "epoch": 0.53, "learning_rate": 4.675675675675676e-05, "loss": 1.043, "step": 212660 }, { "epoch": 0.53, "learning_rate": 4.6755498515275056e-05, "loss": 1.042, "step": 212665 }, { "epoch": 0.53, "learning_rate": 4.675424027379335e-05, "loss": 1.0461, "step": 212670 }, { "epoch": 0.53, "learning_rate": 4.6752982032311646e-05, "loss": 1.0431, "step": 212675 }, { "epoch": 0.53, "learning_rate": 4.675172379082994e-05, "loss": 1.0451, "step": 212680 }, { "epoch": 0.53, "learning_rate": 4.6750465549348236e-05, "loss": 1.0443, "step": 212685 }, { "epoch": 0.53, "learning_rate": 4.674920730786653e-05, "loss": 1.0442, "step": 212690 }, { "epoch": 0.53, "learning_rate": 4.674794906638482e-05, "loss": 1.0433, "step": 212695 }, { "epoch": 0.53, "learning_rate": 4.674669082490312e-05, "loss": 1.0449, "step": 212700 }, { "epoch": 0.53, "learning_rate": 4.674543258342141e-05, "loss": 1.0458, "step": 212705 }, { "epoch": 0.53, "learning_rate": 4.674417434193971e-05, "loss": 1.0475, "step": 212710 }, { "epoch": 0.53, "learning_rate": 4.6742916100458e-05, "loss": 1.044, "step": 212715 }, { "epoch": 0.53, "learning_rate": 4.67416578589763e-05, "loss": 1.042, "step": 212720 }, { "epoch": 0.53, "learning_rate": 4.674039961749459e-05, "loss": 1.0437, "step": 212725 }, { "epoch": 0.53, "learning_rate": 4.673914137601289e-05, "loss": 1.0441, "step": 212730 }, { "epoch": 0.53, "learning_rate": 4.673788313453118e-05, "loss": 1.067, "step": 212735 }, { "epoch": 0.53, "learning_rate": 4.673662489304948e-05, "loss": 1.0421, "step": 212740 }, { "epoch": 0.53, "learning_rate": 4.673536665156777e-05, "loss": 1.0465, "step": 212745 }, { "epoch": 0.53, "learning_rate": 4.673410841008607e-05, "loss": 1.047, "step": 212750 }, { "epoch": 0.53, "learning_rate": 4.673285016860436e-05, "loss": 1.0454, "step": 212755 }, { "epoch": 0.53, "learning_rate": 4.673159192712265e-05, "loss": 1.0433, "step": 212760 }, { "epoch": 0.53, "learning_rate": 4.673033368564095e-05, "loss": 1.0453, "step": 212765 }, { "epoch": 0.53, "learning_rate": 4.672907544415924e-05, "loss": 1.0428, "step": 212770 }, { "epoch": 0.53, "learning_rate": 4.672781720267754e-05, "loss": 1.0409, "step": 212775 }, { "epoch": 0.53, "learning_rate": 4.672655896119584e-05, "loss": 1.0422, "step": 212780 }, { "epoch": 0.53, "learning_rate": 4.6725300719714135e-05, "loss": 1.0435, "step": 212785 }, { "epoch": 0.53, "learning_rate": 4.672404247823243e-05, "loss": 1.046, "step": 212790 }, { "epoch": 0.53, "learning_rate": 4.672278423675072e-05, "loss": 1.0456, "step": 212795 }, { "epoch": 0.53, "learning_rate": 4.672152599526902e-05, "loss": 1.0471, "step": 212800 }, { "epoch": 0.53, "learning_rate": 4.672026775378731e-05, "loss": 1.0451, "step": 212805 }, { "epoch": 0.53, "learning_rate": 4.671900951230561e-05, "loss": 1.044, "step": 212810 }, { "epoch": 0.53, "learning_rate": 4.67177512708239e-05, "loss": 1.0433, "step": 212815 }, { "epoch": 0.53, "learning_rate": 4.67164930293422e-05, "loss": 1.0683, "step": 212820 }, { "epoch": 0.53, "learning_rate": 4.671523478786049e-05, "loss": 1.0455, "step": 212825 }, { "epoch": 0.53, "learning_rate": 4.671397654637879e-05, "loss": 1.043, "step": 212830 }, { "epoch": 0.53, "learning_rate": 4.671271830489708e-05, "loss": 1.0475, "step": 212835 }, { "epoch": 0.53, "learning_rate": 4.671146006341538e-05, "loss": 1.0439, "step": 212840 }, { "epoch": 0.53, "learning_rate": 4.671020182193367e-05, "loss": 1.0459, "step": 212845 }, { "epoch": 0.53, "learning_rate": 4.670894358045197e-05, "loss": 1.0457, "step": 212850 }, { "epoch": 0.53, "learning_rate": 4.670768533897026e-05, "loss": 1.0422, "step": 212855 }, { "epoch": 0.53, "learning_rate": 4.670642709748855e-05, "loss": 1.0464, "step": 212860 }, { "epoch": 0.53, "learning_rate": 4.670516885600685e-05, "loss": 1.0428, "step": 212865 }, { "epoch": 0.53, "learning_rate": 4.670391061452514e-05, "loss": 1.0438, "step": 212870 }, { "epoch": 0.53, "learning_rate": 4.670265237304344e-05, "loss": 1.0439, "step": 212875 }, { "epoch": 0.53, "learning_rate": 4.670139413156173e-05, "loss": 1.0422, "step": 212880 }, { "epoch": 0.53, "learning_rate": 4.670013589008003e-05, "loss": 1.0433, "step": 212885 }, { "epoch": 0.53, "learning_rate": 4.669887764859832e-05, "loss": 1.0447, "step": 212890 }, { "epoch": 0.53, "learning_rate": 4.669761940711662e-05, "loss": 1.0458, "step": 212895 }, { "epoch": 0.53, "learning_rate": 4.669636116563491e-05, "loss": 1.0422, "step": 212900 }, { "epoch": 0.53, "learning_rate": 4.669510292415321e-05, "loss": 1.0402, "step": 212905 }, { "epoch": 0.53, "learning_rate": 4.66938446826715e-05, "loss": 1.0461, "step": 212910 }, { "epoch": 0.53, "learning_rate": 4.66925864411898e-05, "loss": 1.0423, "step": 212915 }, { "epoch": 0.53, "learning_rate": 4.669132819970809e-05, "loss": 1.044, "step": 212920 }, { "epoch": 0.53, "learning_rate": 4.669006995822638e-05, "loss": 1.0462, "step": 212925 }, { "epoch": 0.53, "learning_rate": 4.668881171674468e-05, "loss": 1.0487, "step": 212930 }, { "epoch": 0.53, "learning_rate": 4.668755347526297e-05, "loss": 1.0495, "step": 212935 }, { "epoch": 0.53, "learning_rate": 4.668629523378127e-05, "loss": 1.0453, "step": 212940 }, { "epoch": 0.53, "learning_rate": 4.668503699229956e-05, "loss": 1.0458, "step": 212945 }, { "epoch": 0.53, "learning_rate": 4.668377875081786e-05, "loss": 1.0438, "step": 212950 }, { "epoch": 0.53, "learning_rate": 4.668252050933615e-05, "loss": 1.0445, "step": 212955 }, { "epoch": 0.53, "learning_rate": 4.668126226785445e-05, "loss": 1.0451, "step": 212960 }, { "epoch": 0.53, "learning_rate": 4.668000402637274e-05, "loss": 1.0468, "step": 212965 }, { "epoch": 0.53, "learning_rate": 4.667874578489104e-05, "loss": 1.045, "step": 212970 }, { "epoch": 0.53, "learning_rate": 4.667748754340933e-05, "loss": 1.0432, "step": 212975 }, { "epoch": 0.53, "learning_rate": 4.667622930192763e-05, "loss": 1.0452, "step": 212980 }, { "epoch": 0.53, "learning_rate": 4.667497106044592e-05, "loss": 1.0453, "step": 212985 }, { "epoch": 0.53, "learning_rate": 4.667371281896421e-05, "loss": 1.0407, "step": 212990 }, { "epoch": 0.53, "learning_rate": 4.667245457748251e-05, "loss": 1.0424, "step": 212995 }, { "epoch": 0.53, "learning_rate": 4.66711963360008e-05, "loss": 1.0468, "step": 213000 }, { "epoch": 0.53, "learning_rate": 4.66699380945191e-05, "loss": 1.044, "step": 213005 }, { "epoch": 0.53, "learning_rate": 4.666867985303739e-05, "loss": 1.0451, "step": 213010 }, { "epoch": 0.53, "learning_rate": 4.666742161155569e-05, "loss": 1.0441, "step": 213015 }, { "epoch": 0.53, "learning_rate": 4.666616337007398e-05, "loss": 1.0436, "step": 213020 }, { "epoch": 0.53, "learning_rate": 4.666490512859228e-05, "loss": 1.045, "step": 213025 }, { "epoch": 0.53, "learning_rate": 4.666364688711058e-05, "loss": 1.0435, "step": 213030 }, { "epoch": 0.53, "learning_rate": 4.666238864562887e-05, "loss": 1.0459, "step": 213035 }, { "epoch": 0.53, "learning_rate": 4.666113040414717e-05, "loss": 1.0453, "step": 213040 }, { "epoch": 0.53, "learning_rate": 4.665987216266546e-05, "loss": 1.0442, "step": 213045 }, { "epoch": 0.53, "learning_rate": 4.665861392118376e-05, "loss": 1.0432, "step": 213050 }, { "epoch": 0.53, "learning_rate": 4.665735567970205e-05, "loss": 1.0446, "step": 213055 }, { "epoch": 0.53, "learning_rate": 4.665609743822035e-05, "loss": 1.0451, "step": 213060 }, { "epoch": 0.53, "learning_rate": 4.665483919673864e-05, "loss": 1.0444, "step": 213065 }, { "epoch": 0.53, "learning_rate": 4.665358095525694e-05, "loss": 1.0435, "step": 213070 }, { "epoch": 0.53, "learning_rate": 4.665232271377523e-05, "loss": 1.0457, "step": 213075 }, { "epoch": 0.53, "learning_rate": 4.665106447229353e-05, "loss": 1.0433, "step": 213080 }, { "epoch": 0.53, "learning_rate": 4.664980623081182e-05, "loss": 1.041, "step": 213085 }, { "epoch": 0.53, "learning_rate": 4.664854798933011e-05, "loss": 1.0431, "step": 213090 }, { "epoch": 0.53, "learning_rate": 4.664728974784841e-05, "loss": 1.0431, "step": 213095 }, { "epoch": 0.53, "learning_rate": 4.66460315063667e-05, "loss": 1.0426, "step": 213100 }, { "epoch": 0.53, "learning_rate": 4.6644773264885e-05, "loss": 1.0432, "step": 213105 }, { "epoch": 0.53, "learning_rate": 4.664351502340329e-05, "loss": 1.0463, "step": 213110 }, { "epoch": 0.53, "learning_rate": 4.664225678192159e-05, "loss": 1.045, "step": 213115 }, { "epoch": 0.53, "learning_rate": 4.664099854043988e-05, "loss": 1.0653, "step": 213120 }, { "epoch": 0.53, "learning_rate": 4.663974029895818e-05, "loss": 1.0504, "step": 213125 }, { "epoch": 0.53, "learning_rate": 4.663848205747647e-05, "loss": 1.0434, "step": 213130 }, { "epoch": 0.54, "learning_rate": 4.663722381599477e-05, "loss": 1.0442, "step": 213135 }, { "epoch": 0.54, "learning_rate": 4.663596557451306e-05, "loss": 1.0461, "step": 213140 }, { "epoch": 0.54, "learning_rate": 4.663470733303136e-05, "loss": 1.043, "step": 213145 }, { "epoch": 0.54, "learning_rate": 4.663344909154965e-05, "loss": 1.0455, "step": 213150 }, { "epoch": 0.54, "learning_rate": 4.6632190850067944e-05, "loss": 1.0441, "step": 213155 }, { "epoch": 0.54, "learning_rate": 4.663093260858624e-05, "loss": 1.0458, "step": 213160 }, { "epoch": 0.54, "learning_rate": 4.6629674367104534e-05, "loss": 1.0424, "step": 213165 }, { "epoch": 0.54, "learning_rate": 4.662841612562283e-05, "loss": 1.0428, "step": 213170 }, { "epoch": 0.54, "learning_rate": 4.6627157884141124e-05, "loss": 1.0435, "step": 213175 }, { "epoch": 0.54, "learning_rate": 4.662589964265942e-05, "loss": 1.0454, "step": 213180 }, { "epoch": 0.54, "learning_rate": 4.6624641401177714e-05, "loss": 1.0472, "step": 213185 }, { "epoch": 0.54, "learning_rate": 4.662338315969601e-05, "loss": 1.0445, "step": 213190 }, { "epoch": 0.54, "learning_rate": 4.6622124918214304e-05, "loss": 1.0459, "step": 213195 }, { "epoch": 0.54, "learning_rate": 4.66208666767326e-05, "loss": 1.0689, "step": 213200 }, { "epoch": 0.54, "learning_rate": 4.6619608435250894e-05, "loss": 1.0465, "step": 213205 }, { "epoch": 0.54, "learning_rate": 4.661835019376919e-05, "loss": 1.0436, "step": 213210 }, { "epoch": 0.54, "learning_rate": 4.6617091952287484e-05, "loss": 1.0448, "step": 213215 }, { "epoch": 0.54, "learning_rate": 4.6615833710805775e-05, "loss": 1.0644, "step": 213220 }, { "epoch": 0.54, "learning_rate": 4.6614575469324074e-05, "loss": 1.0402, "step": 213225 }, { "epoch": 0.54, "learning_rate": 4.6613317227842365e-05, "loss": 1.0473, "step": 213230 }, { "epoch": 0.54, "learning_rate": 4.6612058986360664e-05, "loss": 1.0855, "step": 213235 }, { "epoch": 0.54, "learning_rate": 4.6610800744878955e-05, "loss": 1.0461, "step": 213240 }, { "epoch": 0.54, "learning_rate": 4.6609542503397254e-05, "loss": 1.044, "step": 213245 }, { "epoch": 0.54, "learning_rate": 4.6608284261915545e-05, "loss": 1.0495, "step": 213250 }, { "epoch": 0.54, "learning_rate": 4.6607026020433844e-05, "loss": 1.0444, "step": 213255 }, { "epoch": 0.54, "learning_rate": 4.6605767778952135e-05, "loss": 1.0445, "step": 213260 }, { "epoch": 0.54, "learning_rate": 4.6604509537470434e-05, "loss": 1.0463, "step": 213265 }, { "epoch": 0.54, "learning_rate": 4.6603251295988725e-05, "loss": 1.0442, "step": 213270 }, { "epoch": 0.54, "learning_rate": 4.6601993054507024e-05, "loss": 1.0457, "step": 213275 }, { "epoch": 0.54, "learning_rate": 4.660073481302532e-05, "loss": 1.0441, "step": 213280 }, { "epoch": 0.54, "learning_rate": 4.6599476571543614e-05, "loss": 1.0458, "step": 213285 }, { "epoch": 0.54, "learning_rate": 4.659821833006191e-05, "loss": 1.045, "step": 213290 }, { "epoch": 0.54, "learning_rate": 4.6596960088580204e-05, "loss": 1.0411, "step": 213295 }, { "epoch": 0.54, "learning_rate": 4.65957018470985e-05, "loss": 1.0469, "step": 213300 }, { "epoch": 0.54, "learning_rate": 4.6594443605616794e-05, "loss": 1.0423, "step": 213305 }, { "epoch": 0.54, "learning_rate": 4.659318536413509e-05, "loss": 1.0458, "step": 213310 }, { "epoch": 0.54, "learning_rate": 4.6591927122653384e-05, "loss": 1.0448, "step": 213315 }, { "epoch": 0.54, "learning_rate": 4.6590668881171675e-05, "loss": 1.0436, "step": 213320 }, { "epoch": 0.54, "learning_rate": 4.6589410639689974e-05, "loss": 1.0452, "step": 213325 }, { "epoch": 0.54, "learning_rate": 4.6588152398208265e-05, "loss": 1.0446, "step": 213330 }, { "epoch": 0.54, "learning_rate": 4.6586894156726564e-05, "loss": 1.0676, "step": 213335 }, { "epoch": 0.54, "learning_rate": 4.6585635915244855e-05, "loss": 1.0447, "step": 213340 }, { "epoch": 0.54, "learning_rate": 4.6584377673763154e-05, "loss": 1.0461, "step": 213345 }, { "epoch": 0.54, "learning_rate": 4.6583119432281445e-05, "loss": 1.0443, "step": 213350 }, { "epoch": 0.54, "learning_rate": 4.6581861190799743e-05, "loss": 1.046, "step": 213355 }, { "epoch": 0.54, "learning_rate": 4.6580602949318035e-05, "loss": 1.0496, "step": 213360 }, { "epoch": 0.54, "learning_rate": 4.6579344707836333e-05, "loss": 1.0448, "step": 213365 }, { "epoch": 0.54, "learning_rate": 4.6578086466354625e-05, "loss": 1.0445, "step": 213370 }, { "epoch": 0.54, "learning_rate": 4.6576828224872923e-05, "loss": 1.047, "step": 213375 }, { "epoch": 0.54, "learning_rate": 4.6575569983391215e-05, "loss": 1.0398, "step": 213380 }, { "epoch": 0.54, "learning_rate": 4.657431174190951e-05, "loss": 1.0424, "step": 213385 }, { "epoch": 0.54, "learning_rate": 4.6573053500427805e-05, "loss": 1.0436, "step": 213390 }, { "epoch": 0.54, "learning_rate": 4.6571795258946097e-05, "loss": 1.0422, "step": 213395 }, { "epoch": 0.54, "learning_rate": 4.6570537017464395e-05, "loss": 1.0444, "step": 213400 }, { "epoch": 0.54, "learning_rate": 4.6569278775982687e-05, "loss": 1.0469, "step": 213405 }, { "epoch": 0.54, "learning_rate": 4.6568020534500985e-05, "loss": 1.0692, "step": 213410 }, { "epoch": 0.54, "learning_rate": 4.6566762293019277e-05, "loss": 1.0452, "step": 213415 }, { "epoch": 0.54, "learning_rate": 4.6565504051537575e-05, "loss": 1.0462, "step": 213420 }, { "epoch": 0.54, "learning_rate": 4.6564245810055866e-05, "loss": 1.0436, "step": 213425 }, { "epoch": 0.54, "learning_rate": 4.6562987568574165e-05, "loss": 1.0435, "step": 213430 }, { "epoch": 0.54, "learning_rate": 4.6561729327092456e-05, "loss": 1.0556, "step": 213435 }, { "epoch": 0.54, "learning_rate": 4.6560471085610755e-05, "loss": 1.0441, "step": 213440 }, { "epoch": 0.54, "learning_rate": 4.6559212844129046e-05, "loss": 1.0459, "step": 213445 }, { "epoch": 0.54, "learning_rate": 4.655795460264734e-05, "loss": 1.0458, "step": 213450 }, { "epoch": 0.54, "learning_rate": 4.6556696361165636e-05, "loss": 1.043, "step": 213455 }, { "epoch": 0.54, "learning_rate": 4.655543811968393e-05, "loss": 1.0439, "step": 213460 }, { "epoch": 0.54, "learning_rate": 4.6554179878202226e-05, "loss": 1.0452, "step": 213465 }, { "epoch": 0.54, "learning_rate": 4.655292163672052e-05, "loss": 1.0426, "step": 213470 }, { "epoch": 0.54, "learning_rate": 4.6551663395238816e-05, "loss": 1.0448, "step": 213475 }, { "epoch": 0.54, "learning_rate": 4.655040515375711e-05, "loss": 1.0457, "step": 213480 }, { "epoch": 0.54, "learning_rate": 4.6549146912275406e-05, "loss": 1.0603, "step": 213485 }, { "epoch": 0.54, "learning_rate": 4.65478886707937e-05, "loss": 1.0454, "step": 213490 }, { "epoch": 0.54, "learning_rate": 4.6546630429311996e-05, "loss": 1.0461, "step": 213495 }, { "epoch": 0.54, "learning_rate": 4.654537218783029e-05, "loss": 1.0451, "step": 213500 }, { "epoch": 0.54, "learning_rate": 4.6544113946348586e-05, "loss": 1.0423, "step": 213505 }, { "epoch": 0.54, "learning_rate": 4.654285570486688e-05, "loss": 1.0426, "step": 213510 }, { "epoch": 0.54, "learning_rate": 4.654159746338517e-05, "loss": 1.0422, "step": 213515 }, { "epoch": 0.54, "learning_rate": 4.654033922190347e-05, "loss": 1.0453, "step": 213520 }, { "epoch": 0.54, "learning_rate": 4.653908098042176e-05, "loss": 1.0441, "step": 213525 }, { "epoch": 0.54, "learning_rate": 4.6537822738940065e-05, "loss": 1.0481, "step": 213530 }, { "epoch": 0.54, "learning_rate": 4.6536564497458356e-05, "loss": 1.0449, "step": 213535 }, { "epoch": 0.54, "learning_rate": 4.6535306255976655e-05, "loss": 1.0439, "step": 213540 }, { "epoch": 0.54, "learning_rate": 4.6534048014494946e-05, "loss": 1.0434, "step": 213545 }, { "epoch": 0.54, "learning_rate": 4.6532789773013245e-05, "loss": 1.0635, "step": 213550 }, { "epoch": 0.54, "learning_rate": 4.6531531531531536e-05, "loss": 1.0459, "step": 213555 }, { "epoch": 0.54, "learning_rate": 4.653027329004983e-05, "loss": 1.0439, "step": 213560 }, { "epoch": 0.54, "learning_rate": 4.6529015048568126e-05, "loss": 1.0441, "step": 213565 }, { "epoch": 0.54, "learning_rate": 4.652775680708642e-05, "loss": 1.0435, "step": 213570 }, { "epoch": 0.54, "learning_rate": 4.6526498565604716e-05, "loss": 1.0457, "step": 213575 }, { "epoch": 0.54, "learning_rate": 4.652524032412301e-05, "loss": 1.0453, "step": 213580 }, { "epoch": 0.54, "learning_rate": 4.6523982082641306e-05, "loss": 1.0428, "step": 213585 }, { "epoch": 0.54, "learning_rate": 4.65227238411596e-05, "loss": 1.0439, "step": 213590 }, { "epoch": 0.54, "learning_rate": 4.6521465599677896e-05, "loss": 1.0415, "step": 213595 }, { "epoch": 0.54, "learning_rate": 4.652020735819619e-05, "loss": 1.0453, "step": 213600 }, { "epoch": 0.54, "learning_rate": 4.6518949116714486e-05, "loss": 1.0438, "step": 213605 }, { "epoch": 0.54, "learning_rate": 4.651769087523278e-05, "loss": 1.0438, "step": 213610 }, { "epoch": 0.54, "learning_rate": 4.651643263375107e-05, "loss": 1.0666, "step": 213615 }, { "epoch": 0.54, "learning_rate": 4.651517439226937e-05, "loss": 1.0441, "step": 213620 }, { "epoch": 0.54, "learning_rate": 4.651391615078766e-05, "loss": 1.0462, "step": 213625 }, { "epoch": 0.54, "learning_rate": 4.651265790930596e-05, "loss": 1.0466, "step": 213630 }, { "epoch": 0.54, "learning_rate": 4.651139966782425e-05, "loss": 1.0443, "step": 213635 }, { "epoch": 0.54, "learning_rate": 4.651014142634255e-05, "loss": 1.0437, "step": 213640 }, { "epoch": 0.54, "learning_rate": 4.650888318486084e-05, "loss": 1.0469, "step": 213645 }, { "epoch": 0.54, "learning_rate": 4.650762494337914e-05, "loss": 1.0453, "step": 213650 }, { "epoch": 0.54, "learning_rate": 4.650636670189743e-05, "loss": 1.0425, "step": 213655 }, { "epoch": 0.54, "learning_rate": 4.650510846041573e-05, "loss": 1.0457, "step": 213660 }, { "epoch": 0.54, "learning_rate": 4.650385021893402e-05, "loss": 1.0435, "step": 213665 }, { "epoch": 0.54, "learning_rate": 4.650259197745232e-05, "loss": 1.0436, "step": 213670 }, { "epoch": 0.54, "learning_rate": 4.650133373597061e-05, "loss": 1.0442, "step": 213675 }, { "epoch": 0.54, "learning_rate": 4.65000754944889e-05, "loss": 1.0458, "step": 213680 }, { "epoch": 0.54, "learning_rate": 4.64988172530072e-05, "loss": 1.0424, "step": 213685 }, { "epoch": 0.54, "learning_rate": 4.649755901152549e-05, "loss": 1.045, "step": 213690 }, { "epoch": 0.54, "learning_rate": 4.649630077004379e-05, "loss": 1.0438, "step": 213695 }, { "epoch": 0.54, "learning_rate": 4.649504252856208e-05, "loss": 1.0452, "step": 213700 }, { "epoch": 0.54, "learning_rate": 4.649378428708038e-05, "loss": 1.0449, "step": 213705 }, { "epoch": 0.54, "learning_rate": 4.649252604559867e-05, "loss": 1.0434, "step": 213710 }, { "epoch": 0.54, "learning_rate": 4.649126780411697e-05, "loss": 1.0453, "step": 213715 }, { "epoch": 0.54, "learning_rate": 4.649000956263526e-05, "loss": 1.0419, "step": 213720 }, { "epoch": 0.54, "learning_rate": 4.648875132115356e-05, "loss": 1.0416, "step": 213725 }, { "epoch": 0.54, "learning_rate": 4.648749307967185e-05, "loss": 1.044, "step": 213730 }, { "epoch": 0.54, "learning_rate": 4.648623483819015e-05, "loss": 1.0675, "step": 213735 }, { "epoch": 0.54, "learning_rate": 4.648497659670844e-05, "loss": 1.0448, "step": 213740 }, { "epoch": 0.54, "learning_rate": 4.648371835522673e-05, "loss": 1.0427, "step": 213745 }, { "epoch": 0.54, "learning_rate": 4.648246011374503e-05, "loss": 1.0423, "step": 213750 }, { "epoch": 0.54, "learning_rate": 4.648120187226332e-05, "loss": 1.0443, "step": 213755 }, { "epoch": 0.54, "learning_rate": 4.647994363078162e-05, "loss": 1.0459, "step": 213760 }, { "epoch": 0.54, "learning_rate": 4.647868538929991e-05, "loss": 1.0448, "step": 213765 }, { "epoch": 0.54, "learning_rate": 4.647742714781821e-05, "loss": 1.0454, "step": 213770 }, { "epoch": 0.54, "learning_rate": 4.64761689063365e-05, "loss": 1.0441, "step": 213775 }, { "epoch": 0.54, "learning_rate": 4.647491066485481e-05, "loss": 1.0427, "step": 213780 }, { "epoch": 0.54, "learning_rate": 4.64736524233731e-05, "loss": 1.0422, "step": 213785 }, { "epoch": 0.54, "learning_rate": 4.647239418189139e-05, "loss": 1.0433, "step": 213790 }, { "epoch": 0.54, "learning_rate": 4.647113594040969e-05, "loss": 1.0441, "step": 213795 }, { "epoch": 0.54, "learning_rate": 4.646987769892798e-05, "loss": 1.0459, "step": 213800 }, { "epoch": 0.54, "learning_rate": 4.646861945744628e-05, "loss": 1.0455, "step": 213805 }, { "epoch": 0.54, "learning_rate": 4.646736121596457e-05, "loss": 1.0471, "step": 213810 }, { "epoch": 0.54, "learning_rate": 4.6466354622779206e-05, "loss": 1.043, "step": 213815 }, { "epoch": 0.54, "learning_rate": 4.6465096381297505e-05, "loss": 1.043, "step": 213820 }, { "epoch": 0.54, "learning_rate": 4.6463838139815796e-05, "loss": 1.0424, "step": 213825 }, { "epoch": 0.54, "learning_rate": 4.646257989833409e-05, "loss": 1.0433, "step": 213830 }, { "epoch": 0.54, "learning_rate": 4.6461321656852386e-05, "loss": 1.0472, "step": 213835 }, { "epoch": 0.54, "learning_rate": 4.646006341537068e-05, "loss": 1.0613, "step": 213840 }, { "epoch": 0.54, "learning_rate": 4.6458805173888976e-05, "loss": 1.0444, "step": 213845 }, { "epoch": 0.54, "learning_rate": 4.645754693240727e-05, "loss": 1.0446, "step": 213850 }, { "epoch": 0.54, "learning_rate": 4.6456288690925566e-05, "loss": 1.0435, "step": 213855 }, { "epoch": 0.54, "learning_rate": 4.645503044944386e-05, "loss": 1.0447, "step": 213860 }, { "epoch": 0.54, "learning_rate": 4.6453772207962156e-05, "loss": 1.0449, "step": 213865 }, { "epoch": 0.54, "learning_rate": 4.645251396648045e-05, "loss": 1.0663, "step": 213870 }, { "epoch": 0.54, "learning_rate": 4.6451255724998746e-05, "loss": 1.043, "step": 213875 }, { "epoch": 0.54, "learning_rate": 4.644999748351704e-05, "loss": 1.0458, "step": 213880 }, { "epoch": 0.54, "learning_rate": 4.6448739242035336e-05, "loss": 1.0439, "step": 213885 }, { "epoch": 0.54, "learning_rate": 4.644748100055363e-05, "loss": 1.0439, "step": 213890 }, { "epoch": 0.54, "learning_rate": 4.644622275907192e-05, "loss": 1.0439, "step": 213895 }, { "epoch": 0.54, "learning_rate": 4.644496451759022e-05, "loss": 1.0444, "step": 213900 }, { "epoch": 0.54, "learning_rate": 4.644370627610851e-05, "loss": 1.0432, "step": 213905 }, { "epoch": 0.54, "learning_rate": 4.644244803462681e-05, "loss": 1.0412, "step": 213910 }, { "epoch": 0.54, "learning_rate": 4.64411897931451e-05, "loss": 1.0437, "step": 213915 }, { "epoch": 0.54, "learning_rate": 4.64399315516634e-05, "loss": 1.0433, "step": 213920 }, { "epoch": 0.54, "learning_rate": 4.643867331018169e-05, "loss": 1.0449, "step": 213925 }, { "epoch": 0.54, "learning_rate": 4.643741506869999e-05, "loss": 1.0671, "step": 213930 }, { "epoch": 0.54, "learning_rate": 4.643615682721828e-05, "loss": 1.0455, "step": 213935 }, { "epoch": 0.54, "learning_rate": 4.643489858573658e-05, "loss": 1.0421, "step": 213940 }, { "epoch": 0.54, "learning_rate": 4.643364034425487e-05, "loss": 1.042, "step": 213945 }, { "epoch": 0.54, "learning_rate": 4.643238210277317e-05, "loss": 1.0663, "step": 213950 }, { "epoch": 0.54, "learning_rate": 4.643112386129146e-05, "loss": 1.0427, "step": 213955 }, { "epoch": 0.54, "learning_rate": 4.642986561980975e-05, "loss": 1.0448, "step": 213960 }, { "epoch": 0.54, "learning_rate": 4.642860737832805e-05, "loss": 1.0455, "step": 213965 }, { "epoch": 0.54, "learning_rate": 4.642734913684634e-05, "loss": 1.0412, "step": 213970 }, { "epoch": 0.54, "learning_rate": 4.642609089536464e-05, "loss": 1.0433, "step": 213975 }, { "epoch": 0.54, "learning_rate": 4.642483265388294e-05, "loss": 1.0431, "step": 213980 }, { "epoch": 0.54, "learning_rate": 4.6423574412401236e-05, "loss": 1.0468, "step": 213985 }, { "epoch": 0.54, "learning_rate": 4.642231617091953e-05, "loss": 1.0425, "step": 213990 }, { "epoch": 0.54, "learning_rate": 4.642105792943782e-05, "loss": 1.0704, "step": 213995 }, { "epoch": 0.54, "learning_rate": 4.641979968795612e-05, "loss": 1.0454, "step": 214000 } ], "logging_steps": 5, "max_steps": 398380, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 1000, "total_flos": 0.0, "train_batch_size": 2, "trial_name": null, "trial_params": null }