| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.4803535402055913, | |
| "global_step": 20000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999999855889116, | |
| "loss": 3.4871, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999999357728057, | |
| "loss": 2.9969, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999998503737748, | |
| "loss": 2.9734, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999997293918308, | |
| "loss": 2.9203, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999995728269912, | |
| "loss": 2.7908, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999993806792782, | |
| "loss": 2.7973, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.000499999152948719, | |
| "loss": 2.8511, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999988896353463, | |
| "loss": 2.8423, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999985907391973, | |
| "loss": 2.8411, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999982562603146, | |
| "loss": 2.838, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.000499997886198746, | |
| "loss": 2.8751, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999974805545439, | |
| "loss": 2.7868, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999970393277663, | |
| "loss": 2.8003, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999965625184758, | |
| "loss": 2.8343, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999960501267404, | |
| "loss": 2.8053, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999955021526329, | |
| "loss": 2.7828, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999949185962313, | |
| "loss": 2.7822, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999942994576189, | |
| "loss": 2.7426, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999936447368836, | |
| "loss": 2.7808, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0004999929544341185, | |
| "loss": 2.7522, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999922285494221, | |
| "loss": 2.8008, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999914670828975, | |
| "loss": 2.748, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999906700346533, | |
| "loss": 2.783, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999898374048027, | |
| "loss": 2.7879, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999889691934643, | |
| "loss": 2.696, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999880654007619, | |
| "loss": 2.7354, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999871260268238, | |
| "loss": 2.7336, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999861510717839, | |
| "loss": 2.7077, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999851405357809, | |
| "loss": 2.7408, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999840944189586, | |
| "loss": 2.7115, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000499983012721466, | |
| "loss": 2.6973, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999818954434569, | |
| "loss": 2.6644, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999807425850907, | |
| "loss": 2.7442, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000499979554146531, | |
| "loss": 2.7007, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999783301279471, | |
| "loss": 2.7284, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999770705295135, | |
| "loss": 2.6649, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999757753514091, | |
| "loss": 2.7269, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999744445938185, | |
| "loss": 2.64, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999730782569309, | |
| "loss": 2.6886, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000499971676340941, | |
| "loss": 2.6911, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999702388460482, | |
| "loss": 2.6814, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999687657724571, | |
| "loss": 2.7235, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999672571203774, | |
| "loss": 2.6898, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999657128900239, | |
| "loss": 2.6222, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999641330816162, | |
| "loss": 2.6827, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999625176953794, | |
| "loss": 2.6634, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999608667315432, | |
| "loss": 2.6976, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999591801903428, | |
| "loss": 2.6856, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000499957458072018, | |
| "loss": 2.624, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000499955700376814, | |
| "loss": 2.6693, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999539071049812, | |
| "loss": 2.7058, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999520782567746, | |
| "loss": 2.6878, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999502138324544, | |
| "loss": 2.6674, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999483138322863, | |
| "loss": 2.634, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999463782565405, | |
| "loss": 2.6242, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999444071054925, | |
| "loss": 2.636, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999424003794229, | |
| "loss": 2.6926, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999403580786175, | |
| "loss": 2.6564, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999382802033666, | |
| "loss": 2.6474, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999361667539663, | |
| "loss": 2.6462, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999340177307172, | |
| "loss": 2.6069, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0004999318331339252, | |
| "loss": 2.632, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999296129639014, | |
| "loss": 2.6533, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999273572209616, | |
| "loss": 2.6524, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000499925065905427, | |
| "loss": 2.6229, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999227390176237, | |
| "loss": 2.6587, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999203765578828, | |
| "loss": 2.6175, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999179785265406, | |
| "loss": 2.6214, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999155449239384, | |
| "loss": 2.6338, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999130757504227, | |
| "loss": 2.6364, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999105710063449, | |
| "loss": 2.6509, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999080306920613, | |
| "loss": 2.6465, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999054548079338, | |
| "loss": 2.6452, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999028433543286, | |
| "loss": 2.6858, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004999001963316179, | |
| "loss": 2.6624, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998975137401781, | |
| "loss": 2.5747, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000499894795580391, | |
| "loss": 2.6239, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998920418526438, | |
| "loss": 2.612, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998892525573282, | |
| "loss": 2.6312, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998864276948413, | |
| "loss": 2.6321, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000499883567265585, | |
| "loss": 2.6128, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998806712699667, | |
| "loss": 2.6299, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998777397083982, | |
| "loss": 2.6444, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998747725812971, | |
| "loss": 2.6058, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998717698890857, | |
| "loss": 2.5646, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998687316321911, | |
| "loss": 2.6089, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998656578110461, | |
| "loss": 2.6257, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998625484260881, | |
| "loss": 2.6794, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998594034777594, | |
| "loss": 2.6585, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998562229665079, | |
| "loss": 2.6161, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998530068927862, | |
| "loss": 2.5684, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000499849755257052, | |
| "loss": 2.6284, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998464680597682, | |
| "loss": 2.5965, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998431453014028, | |
| "loss": 2.5825, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998397869824284, | |
| "loss": 2.6056, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998363931033231, | |
| "loss": 2.5434, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998329636645702, | |
| "loss": 2.6535, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998294986666576, | |
| "loss": 2.7084, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998259981100785, | |
| "loss": 2.6648, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998224619953312, | |
| "loss": 2.6241, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000499818890322919, | |
| "loss": 2.5672, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998152830933501, | |
| "loss": 2.6968, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998116403071381, | |
| "loss": 2.5603, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004998079619648014, | |
| "loss": 2.6031, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004998042480668638, | |
| "loss": 2.5779, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004998004986138534, | |
| "loss": 2.6792, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997967136063043, | |
| "loss": 2.6038, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997928930447551, | |
| "loss": 2.6469, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997890369297496, | |
| "loss": 2.62, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997851452618364, | |
| "loss": 2.6517, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997812180415697, | |
| "loss": 2.6313, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997772552695085, | |
| "loss": 2.62, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997732569462167, | |
| "loss": 2.6205, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997692230722633, | |
| "loss": 2.5727, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997651536482227, | |
| "loss": 2.5626, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997610486746738, | |
| "loss": 2.6196, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997569081522012, | |
| "loss": 2.6219, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997527320813939, | |
| "loss": 2.6443, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997485204628466, | |
| "loss": 2.6027, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997442732971584, | |
| "loss": 2.6495, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997399905849342, | |
| "loss": 2.6099, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997356723267833, | |
| "loss": 2.6074, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997313185233204, | |
| "loss": 2.6089, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997269291751651, | |
| "loss": 2.6539, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997225042829423, | |
| "loss": 2.6044, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997180438472816, | |
| "loss": 2.5768, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997135478688181, | |
| "loss": 2.5801, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997090163481916, | |
| "loss": 2.6145, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004997044492860471, | |
| "loss": 2.6338, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996998466830345, | |
| "loss": 2.6048, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996952085398091, | |
| "loss": 2.6023, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000499690534857031, | |
| "loss": 2.6037, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996858256353653, | |
| "loss": 2.5876, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996810808754824, | |
| "loss": 2.5932, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996763005780576, | |
| "loss": 2.6014, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996714847437712, | |
| "loss": 2.6211, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996666333733088, | |
| "loss": 2.6615, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996617464673609, | |
| "loss": 2.6057, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996568240266228, | |
| "loss": 2.5722, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996518660517954, | |
| "loss": 2.5588, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996468725435843, | |
| "loss": 2.612, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996418435027002, | |
| "loss": 2.5926, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996367789298589, | |
| "loss": 2.5635, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996316788257812, | |
| "loss": 2.5991, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004996265431911932, | |
| "loss": 2.6328, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004996213720268257, | |
| "loss": 2.5715, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004996161653334147, | |
| "loss": 2.5659, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004996109231117013, | |
| "loss": 2.5627, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004996056453624317, | |
| "loss": 2.5869, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004996003320863571, | |
| "loss": 2.6017, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995949832842337, | |
| "loss": 2.532, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995895989568227, | |
| "loss": 2.6129, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995841791048906, | |
| "loss": 2.6216, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995787237292089, | |
| "loss": 2.5878, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995732328305538, | |
| "loss": 2.5592, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000499567706409707, | |
| "loss": 2.637, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995621444674552, | |
| "loss": 2.571, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995565470045898, | |
| "loss": 2.5604, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995509140219076, | |
| "loss": 2.6172, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995452455202105, | |
| "loss": 2.6257, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995395415003051, | |
| "loss": 2.5337, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995338019630033, | |
| "loss": 2.5631, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000499528026909122, | |
| "loss": 2.5511, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995222163394834, | |
| "loss": 2.609, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995163702549142, | |
| "loss": 2.5627, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995104886562466, | |
| "loss": 2.5877, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004995045715443178, | |
| "loss": 2.5274, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994986189199701, | |
| "loss": 2.545, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994926307840505, | |
| "loss": 2.5648, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994866071374115, | |
| "loss": 2.5424, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994805479809103, | |
| "loss": 2.5759, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994744533154094, | |
| "loss": 2.6146, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994683231417762, | |
| "loss": 2.589, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994621574608833, | |
| "loss": 2.6288, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994559562736083, | |
| "loss": 2.5964, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994497195808336, | |
| "loss": 2.5697, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994434473834472, | |
| "loss": 2.5815, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994371396823416, | |
| "loss": 2.5621, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994307964784147, | |
| "loss": 2.5517, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994244177725693, | |
| "loss": 2.6157, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994180035657132, | |
| "loss": 2.5651, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994115538587595, | |
| "loss": 2.5868, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004994050686526261, | |
| "loss": 2.6103, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004993985479482362, | |
| "loss": 2.6364, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004993919917465178, | |
| "loss": 2.6651, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0004993854000484038, | |
| "loss": 2.5453, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000499378772854833, | |
| "loss": 2.5657, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000499372110166748, | |
| "loss": 2.5375, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993654119850975, | |
| "loss": 2.5654, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993586783108349, | |
| "loss": 2.6298, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993519091449184, | |
| "loss": 2.5405, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993451044883116, | |
| "loss": 2.6205, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993382643419831, | |
| "loss": 2.5851, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993313887069062, | |
| "loss": 2.5528, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993244775840596, | |
| "loss": 2.5741, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993175309744273, | |
| "loss": 2.5482, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993105488789976, | |
| "loss": 2.5643, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004993035312987645, | |
| "loss": 2.5748, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992964782347269, | |
| "loss": 2.5976, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992893896878884, | |
| "loss": 2.5567, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000499282265659258, | |
| "loss": 2.516, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00049927510614985, | |
| "loss": 2.5344, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992679111606829, | |
| "loss": 2.6098, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992606806927813, | |
| "loss": 2.567, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992534147471738, | |
| "loss": 2.5139, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992461133248951, | |
| "loss": 2.6271, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992387764269841, | |
| "loss": 2.6127, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992314040544851, | |
| "loss": 2.5665, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992239962084475, | |
| "loss": 2.5802, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992165528899256, | |
| "loss": 2.5765, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000499209074099979, | |
| "loss": 2.5625, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004992015598396718, | |
| "loss": 2.5349, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991940101100739, | |
| "loss": 2.5717, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991864249122598, | |
| "loss": 2.5886, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991788042473088, | |
| "loss": 2.5742, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000499171148116306, | |
| "loss": 2.5869, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991634565203407, | |
| "loss": 2.5709, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000499155729460508, | |
| "loss": 2.6118, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991479669379076, | |
| "loss": 2.5576, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991401689536442, | |
| "loss": 2.5306, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991323355088279, | |
| "loss": 2.5807, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991244666045735, | |
| "loss": 2.5359, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991165622420011, | |
| "loss": 2.5796, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991086224222358, | |
| "loss": 2.5182, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004991006471464075, | |
| "loss": 2.6564, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004990926364156513, | |
| "loss": 2.638, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004990845902311077, | |
| "loss": 2.5452, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004990765085939216, | |
| "loss": 2.5622, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990683915052436, | |
| "loss": 2.5423, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990602389662287, | |
| "loss": 2.5891, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990520509780374, | |
| "loss": 2.5968, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990438275418351, | |
| "loss": 2.5594, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990355686587923, | |
| "loss": 2.5458, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990272743300843, | |
| "loss": 2.5304, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000499018944556892, | |
| "loss": 2.5505, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990105793404008, | |
| "loss": 2.6207, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004990021786818013, | |
| "loss": 2.6038, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000498993742582289, | |
| "loss": 2.5597, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989852710430651, | |
| "loss": 2.5344, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000498976764065335, | |
| "loss": 2.6021, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989682216503097, | |
| "loss": 2.5924, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989596437992049, | |
| "loss": 2.5742, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989510305132416, | |
| "loss": 2.5709, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989423817936458, | |
| "loss": 2.6258, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989336976416484, | |
| "loss": 2.5664, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989249780584853, | |
| "loss": 2.5551, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989162230453978, | |
| "loss": 2.5816, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004989074326036319, | |
| "loss": 2.5314, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988986067344388, | |
| "loss": 2.6238, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988897454390747, | |
| "loss": 2.6155, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988808487188009, | |
| "loss": 2.6048, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988719165748835, | |
| "loss": 2.5906, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000498862949008594, | |
| "loss": 2.5826, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988539460212088, | |
| "loss": 2.5686, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988449076140091, | |
| "loss": 2.5948, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988358337882816, | |
| "loss": 2.5445, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988267245453175, | |
| "loss": 2.5417, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988175798864137, | |
| "loss": 2.5712, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004988083998128714, | |
| "loss": 2.5913, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987991843259975, | |
| "loss": 2.5529, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987899334271035, | |
| "loss": 2.5581, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987806471175063, | |
| "loss": 2.5334, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987713253985273, | |
| "loss": 2.5911, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987619682714936, | |
| "loss": 2.5962, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987525757377368, | |
| "loss": 2.5038, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987431477985938, | |
| "loss": 2.5732, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987336844554066, | |
| "loss": 2.5667, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987241857095221, | |
| "loss": 2.5632, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987146515622921, | |
| "loss": 2.545, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004987050820150739, | |
| "loss": 2.5234, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986954770692293, | |
| "loss": 2.5294, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986858367261256, | |
| "loss": 2.6312, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986761609871347, | |
| "loss": 2.5496, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986664498536339, | |
| "loss": 2.5203, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986567033270055, | |
| "loss": 2.6495, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986469214086364, | |
| "loss": 2.592, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986371040999192, | |
| "loss": 2.532, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986272514022512, | |
| "loss": 2.591, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986173633170346, | |
| "loss": 2.5353, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004986074398456769, | |
| "loss": 2.565, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985974809895903, | |
| "loss": 2.6095, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985874867501927, | |
| "loss": 2.5576, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985774571289062, | |
| "loss": 2.5559, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985673921271584, | |
| "loss": 2.5174, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985572917463821, | |
| "loss": 2.5949, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985471559880146, | |
| "loss": 2.5327, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985369848534988, | |
| "loss": 2.6106, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985267783442823, | |
| "loss": 2.6212, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985165364618176, | |
| "loss": 2.6182, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004985062592075627, | |
| "loss": 2.5897, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984959465829804, | |
| "loss": 2.5322, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984855985895382, | |
| "loss": 2.4541, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984752152287093, | |
| "loss": 2.5546, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984647965019714, | |
| "loss": 2.5689, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984543424108075, | |
| "loss": 2.5218, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984438529567054, | |
| "loss": 2.6078, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984333281411582, | |
| "loss": 2.5947, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000498422767965664, | |
| "loss": 2.5687, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984121724317256, | |
| "loss": 2.5415, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004984015415408514, | |
| "loss": 2.5362, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983908752945542, | |
| "loss": 2.521, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983801736943524, | |
| "loss": 2.5902, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000498369436741769, | |
| "loss": 2.565, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983586644383322, | |
| "loss": 2.5585, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983478567855754, | |
| "loss": 2.5787, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983370137850369, | |
| "loss": 2.5266, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983261354382598, | |
| "loss": 2.4995, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983152217467924, | |
| "loss": 2.5792, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004983042727121883, | |
| "loss": 2.5772, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004982932883360058, | |
| "loss": 2.4929, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004982822686198083, | |
| "loss": 2.508, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0004982712135651643, | |
| "loss": 2.5573, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004982601231736472, | |
| "loss": 2.6541, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004982489974468355, | |
| "loss": 2.4863, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000498237836386313, | |
| "loss": 2.5646, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004982266399936678, | |
| "loss": 2.5657, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000498215408270494, | |
| "loss": 2.5708, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00049820414121839, | |
| "loss": 2.5032, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004981928388389594, | |
| "loss": 2.5778, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000498181501133811, | |
| "loss": 2.5989, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004981701281045585, | |
| "loss": 2.5047, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004981587197528205, | |
| "loss": 2.5285, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000498147276080221, | |
| "loss": 2.5799, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004981357970883886, | |
| "loss": 2.5478, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004981242827789572, | |
| "loss": 2.5617, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004981138897052764, | |
| "loss": 2.5479, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004981023082969261, | |
| "loss": 2.5556, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004980906915757434, | |
| "loss": 2.5612, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004980790395433817, | |
| "loss": 2.4746, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004980673522014993, | |
| "loss": 2.5154, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00049805562955176, | |
| "loss": 2.5829, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004980438715958318, | |
| "loss": 2.5546, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004980320783353888, | |
| "loss": 2.6171, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004980202497721091, | |
| "loss": 2.5568, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004980083859076766, | |
| "loss": 2.6162, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979964867437797, | |
| "loss": 2.5808, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979845522821121, | |
| "loss": 2.5665, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979725825243724, | |
| "loss": 2.5723, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979605774722644, | |
| "loss": 2.5039, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979485371274967, | |
| "loss": 2.5566, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979364614917831, | |
| "loss": 2.5481, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979243505668421, | |
| "loss": 2.545, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004979122043543979, | |
| "loss": 2.5013, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000497900022856179, | |
| "loss": 2.5336, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978878060739191, | |
| "loss": 2.5577, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978755540093573, | |
| "loss": 2.5775, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978632666642374, | |
| "loss": 2.5011, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978509440403083, | |
| "loss": 2.5834, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978385861393237, | |
| "loss": 2.5143, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978261929630427, | |
| "loss": 2.574, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978137645132294, | |
| "loss": 2.5587, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004978013007916523, | |
| "loss": 2.5767, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0004977888018000858, | |
| "loss": 2.5945, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004977762675403087, | |
| "loss": 2.5834, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004977636980141051, | |
| "loss": 2.6018, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004977510932232639, | |
| "loss": 2.5104, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004977384531695794, | |
| "loss": 2.4966, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004977257778548506, | |
| "loss": 2.5948, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004977130672808815, | |
| "loss": 2.5277, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004977003214494813, | |
| "loss": 2.5563, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004976875403624641, | |
| "loss": 2.5506, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004976747240216491, | |
| "loss": 2.5865, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004976618724288604, | |
| "loss": 2.583, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004976489855859272, | |
| "loss": 2.5752, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004976360634946838, | |
| "loss": 2.5702, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004976231061569694, | |
| "loss": 2.6195, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004976101135746281, | |
| "loss": 2.5673, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975970857495093, | |
| "loss": 2.5204, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975840226834672, | |
| "loss": 2.5061, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975709243783612, | |
| "loss": 2.538, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975577908360554, | |
| "loss": 2.536, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975446220584195, | |
| "loss": 2.554, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975314180473274, | |
| "loss": 2.5227, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975181788046585, | |
| "loss": 2.5565, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004975049043322976, | |
| "loss": 2.5408, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004974915946321335, | |
| "loss": 2.5228, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000497478249706061, | |
| "loss": 2.6042, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004974648695559794, | |
| "loss": 2.5345, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000497451454183793, | |
| "loss": 2.5326, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004974380035914114, | |
| "loss": 2.5002, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004974245177807489, | |
| "loss": 2.5726, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004974109967537251, | |
| "loss": 2.5553, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973974405122644, | |
| "loss": 2.5504, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973838490582962, | |
| "loss": 2.5844, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973702223937552, | |
| "loss": 2.5305, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973565605205807, | |
| "loss": 2.5156, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973428634407174, | |
| "loss": 2.5365, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973291311561146, | |
| "loss": 2.519, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973153636687271, | |
| "loss": 2.5673, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004973015609805142, | |
| "loss": 2.5224, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004972877230934407, | |
| "loss": 2.4995, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000497273850009476, | |
| "loss": 2.5978, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004972599417305948, | |
| "loss": 2.5886, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004972459982587765, | |
| "loss": 2.6197, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004972320195960059, | |
| "loss": 2.5968, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004972180057442726, | |
| "loss": 2.5245, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004972039567055711, | |
| "loss": 2.5286, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004971898724819011, | |
| "loss": 2.5388, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004971757530752671, | |
| "loss": 2.5169, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004971615984876789, | |
| "loss": 2.5923, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004971474087211511, | |
| "loss": 2.5335, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004971331837777033, | |
| "loss": 2.5831, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004971189236593603, | |
| "loss": 2.5268, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004971046283681515, | |
| "loss": 2.5427, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004970902979061119, | |
| "loss": 2.5611, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.000497075932275281, | |
| "loss": 2.6082, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004970615314777035, | |
| "loss": 2.5046, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004970470955154291, | |
| "loss": 2.5643, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004970326243905126, | |
| "loss": 2.5789, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004970181181050134, | |
| "loss": 2.4757, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004970035766609965, | |
| "loss": 2.4911, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004969890000605315, | |
| "loss": 2.5294, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004969743883056932, | |
| "loss": 2.4996, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004969597413985612, | |
| "loss": 2.5432, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004969450593412202, | |
| "loss": 2.5731, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004969303421357601, | |
| "loss": 2.6248, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004969155897842754, | |
| "loss": 2.5902, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004969008022888659, | |
| "loss": 2.4852, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004968859796516365, | |
| "loss": 2.4911, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004968711218746967, | |
| "loss": 2.5582, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004968562289601613, | |
| "loss": 2.5825, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004968413009101502, | |
| "loss": 2.5981, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.000496826337726788, | |
| "loss": 2.5896, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004968113394122044, | |
| "loss": 2.5474, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004967963059685341, | |
| "loss": 2.5784, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004967812373979169, | |
| "loss": 2.5398, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004967661337024976, | |
| "loss": 2.5089, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004967509948844258, | |
| "loss": 2.519, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004967358209458564, | |
| "loss": 2.5411, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004967206118889491, | |
| "loss": 2.5732, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004967053677158686, | |
| "loss": 2.5164, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004966900884287845, | |
| "loss": 2.5663, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004966747740298717, | |
| "loss": 2.5126, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004966594245213098, | |
| "loss": 2.5704, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004966440399052836, | |
| "loss": 2.5488, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004966286201839828, | |
| "loss": 2.5444, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0004966131653596021, | |
| "loss": 2.5005, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004965976754343413, | |
| "loss": 2.5497, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000496582150410405, | |
| "loss": 2.6044, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004965665902900029, | |
| "loss": 2.6103, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004965509950753498, | |
| "loss": 2.4993, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004965353647686653, | |
| "loss": 2.4838, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004965196993721743, | |
| "loss": 2.4869, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000496503998888106, | |
| "loss": 2.5232, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004964882633186955, | |
| "loss": 2.514, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004964724926661825, | |
| "loss": 2.5312, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004964566869328115, | |
| "loss": 2.578, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004964408461208319, | |
| "loss": 2.557, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004964249702324989, | |
| "loss": 2.4945, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004964090592700718, | |
| "loss": 2.5046, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004963931132358153, | |
| "loss": 2.5283, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000496377132131999, | |
| "loss": 2.5303, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004963611159608975, | |
| "loss": 2.4897, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004963450647247906, | |
| "loss": 2.5238, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004963289784259627, | |
| "loss": 2.5362, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004963128570667034, | |
| "loss": 2.5355, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004962967006493074, | |
| "loss": 2.5849, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004962805091760741, | |
| "loss": 2.5662, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004962642826493082, | |
| "loss": 2.4892, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004962480210713193, | |
| "loss": 2.5777, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004962317244444217, | |
| "loss": 2.5342, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004962153927709353, | |
| "loss": 2.5293, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004961990260531843, | |
| "loss": 2.5612, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004961826242934983, | |
| "loss": 2.4989, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004961661874942117, | |
| "loss": 2.557, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004961497156576641, | |
| "loss": 2.542, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004961332087862, | |
| "loss": 2.5551, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004961166668821687, | |
| "loss": 2.5445, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004961000899479247, | |
| "loss": 2.5366, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004960834779858275, | |
| "loss": 2.5445, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004960668309982415, | |
| "loss": 2.5115, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000496050148987536, | |
| "loss": 2.562, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004960334319560854, | |
| "loss": 2.5237, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004960166799062691, | |
| "loss": 2.5659, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004959998928404715, | |
| "loss": 2.5704, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004959830707610819, | |
| "loss": 2.5654, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004959662136704945, | |
| "loss": 2.557, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0004959493215711087, | |
| "loss": 2.4897, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004959323944653288, | |
| "loss": 2.572, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000495915432355564, | |
| "loss": 2.5766, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004958984352442286, | |
| "loss": 2.5121, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004958814031337419, | |
| "loss": 2.4934, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004958643360265279, | |
| "loss": 2.6128, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000495847233925016, | |
| "loss": 2.5246, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004958300968316402, | |
| "loss": 2.5577, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004958129247488397, | |
| "loss": 2.484, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004957957176790587, | |
| "loss": 2.5616, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004957784756247463, | |
| "loss": 2.4936, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004957611985883566, | |
| "loss": 2.58, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004957438865723486, | |
| "loss": 2.5083, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004957265395791865, | |
| "loss": 2.4819, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000495709157611339, | |
| "loss": 2.5151, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004956917406712805, | |
| "loss": 2.4644, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004956742887614897, | |
| "loss": 2.5624, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004956568018844507, | |
| "loss": 2.5754, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004956392800426525, | |
| "loss": 2.5833, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004956217232385888, | |
| "loss": 2.5116, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004956041314747587, | |
| "loss": 2.4878, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004955865047536658, | |
| "loss": 2.627, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004955688430778193, | |
| "loss": 2.4976, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004955511464497329, | |
| "loss": 2.5709, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004955334148719252, | |
| "loss": 2.5895, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004955156483469201, | |
| "loss": 2.5231, | |
| "step": 5030 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004954978468772463, | |
| "loss": 2.5312, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004954800104654375, | |
| "loss": 2.5598, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004954621391140326, | |
| "loss": 2.5421, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000495444232825575, | |
| "loss": 2.5199, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004954262916026135, | |
| "loss": 2.4944, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004954083154477015, | |
| "loss": 2.5115, | |
| "step": 5090 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004953903043633977, | |
| "loss": 2.5126, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004953722583522658, | |
| "loss": 2.446, | |
| "step": 5110 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004953541774168742, | |
| "loss": 2.6025, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004953360615597963, | |
| "loss": 2.5208, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004953179107836105, | |
| "loss": 2.5167, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004952997250909006, | |
| "loss": 2.5042, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004952815044842546, | |
| "loss": 2.4997, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000495263248966266, | |
| "loss": 2.5308, | |
| "step": 5170 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004952449585395334, | |
| "loss": 2.5414, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004952266332066597, | |
| "loss": 2.5362, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004952082729702534, | |
| "loss": 2.5335, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004951898778329277, | |
| "loss": 2.6089, | |
| "step": 5210 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004951714477973008, | |
| "loss": 2.5703, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.000495152982865996, | |
| "loss": 2.5374, | |
| "step": 5230 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004951344830416412, | |
| "loss": 2.5378, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004951159483268698, | |
| "loss": 2.5335, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004950973787243196, | |
| "loss": 2.5149, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004950787742366339, | |
| "loss": 2.5265, | |
| "step": 5270 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004950601348664605, | |
| "loss": 2.5606, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004950414606164525, | |
| "loss": 2.5707, | |
| "step": 5290 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004950227514892678, | |
| "loss": 2.5363, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004950040074875692, | |
| "loss": 2.5186, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004949852286140247, | |
| "loss": 2.5089, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004949664148713071, | |
| "loss": 2.5773, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004949475662620942, | |
| "loss": 2.5507, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004949286827890687, | |
| "loss": 2.4891, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004949097644549185, | |
| "loss": 2.5419, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004948908112623361, | |
| "loss": 2.5012, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004948718232140191, | |
| "loss": 2.57, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004948528003126702, | |
| "loss": 2.5826, | |
| "step": 5390 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.000494833742560997, | |
| "loss": 2.5257, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004948146499617119, | |
| "loss": 2.5108, | |
| "step": 5410 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004947955225175325, | |
| "loss": 2.542, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004947763602311812, | |
| "loss": 2.5763, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004947571631053854, | |
| "loss": 2.5096, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004947379311428774, | |
| "loss": 2.5886, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004947186643463945, | |
| "loss": 2.4859, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004946993627186792, | |
| "loss": 2.5634, | |
| "step": 5470 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004946800262624786, | |
| "loss": 2.5464, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004946606549805448, | |
| "loss": 2.5726, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.000494641248875635, | |
| "loss": 2.5298, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004946218079505114, | |
| "loss": 2.5529, | |
| "step": 5510 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.000494602332207941, | |
| "loss": 2.5434, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004945828216506956, | |
| "loss": 2.5649, | |
| "step": 5530 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004945632762815526, | |
| "loss": 2.5616, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004945436961032937, | |
| "loss": 2.5637, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004945240811187057, | |
| "loss": 2.5155, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004945044313305805, | |
| "loss": 2.4839, | |
| "step": 5570 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004944847467417149, | |
| "loss": 2.5336, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004944650273549108, | |
| "loss": 2.6058, | |
| "step": 5590 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004944452731729746, | |
| "loss": 2.5009, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0004944254841987182, | |
| "loss": 2.5141, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.000494405660434958, | |
| "loss": 2.588, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004943858018845157, | |
| "loss": 2.5345, | |
| "step": 5630 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004943659085502177, | |
| "loss": 2.4981, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004943459804348955, | |
| "loss": 2.4701, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004943260175413855, | |
| "loss": 2.4917, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.000494306019872529, | |
| "loss": 2.5416, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004942859874311724, | |
| "loss": 2.4962, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004942659202201669, | |
| "loss": 2.4898, | |
| "step": 5690 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004942458182423688, | |
| "loss": 2.5407, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004942256815006391, | |
| "loss": 2.523, | |
| "step": 5710 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004942055099978438, | |
| "loss": 2.5788, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004941853037368543, | |
| "loss": 2.5076, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004941650627205463, | |
| "loss": 2.6023, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004941447869518009, | |
| "loss": 2.4776, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004941244764335038, | |
| "loss": 2.5034, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004941041311685461, | |
| "loss": 2.566, | |
| "step": 5770 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004940837511598233, | |
| "loss": 2.5236, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004940633364102364, | |
| "loss": 2.5121, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004940428869226908, | |
| "loss": 2.5473, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004940224027000971, | |
| "loss": 2.5438, | |
| "step": 5810 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004940018837453711, | |
| "loss": 2.5331, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004939813300614333, | |
| "loss": 2.5063, | |
| "step": 5830 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004939607416512089, | |
| "loss": 2.4819, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004939401185176283, | |
| "loss": 2.5368, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004939194606636271, | |
| "loss": 2.5501, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004938987680921454, | |
| "loss": 2.5764, | |
| "step": 5870 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004938780408061283, | |
| "loss": 2.5278, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004938572788085262, | |
| "loss": 2.5153, | |
| "step": 5890 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.000493836482102294, | |
| "loss": 2.5262, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004938156506903916, | |
| "loss": 2.4982, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004937947845757843, | |
| "loss": 2.5665, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004937738837614419, | |
| "loss": 2.5907, | |
| "step": 5930 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004937529482503391, | |
| "loss": 2.5652, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004937319780454559, | |
| "loss": 2.5768, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004937109731497769, | |
| "loss": 2.5628, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004936899335662917, | |
| "loss": 2.5346, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004936688592979951, | |
| "loss": 2.6023, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004936477503478865, | |
| "loss": 2.6156, | |
| "step": 5990 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004936266067189705, | |
| "loss": 2.5852, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004936054284142563, | |
| "loss": 2.6043, | |
| "step": 6010 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004935842154367583, | |
| "loss": 2.5359, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0004935629677894959, | |
| "loss": 2.6143, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004935416854754933, | |
| "loss": 2.5196, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004935203684977795, | |
| "loss": 2.5343, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004934990168593887, | |
| "loss": 2.5528, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00049347763056336, | |
| "loss": 2.5252, | |
| "step": 6070 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004934562096127371, | |
| "loss": 2.572, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000493434754010569, | |
| "loss": 2.5583, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004934132637599097, | |
| "loss": 2.5042, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004933917388638177, | |
| "loss": 2.5118, | |
| "step": 6110 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004933701793253568, | |
| "loss": 2.534, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004933485851475955, | |
| "loss": 2.5335, | |
| "step": 6130 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004933269563336075, | |
| "loss": 2.5667, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004933052928864709, | |
| "loss": 2.5293, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004932835948092696, | |
| "loss": 2.5083, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004932618621050917, | |
| "loss": 2.5234, | |
| "step": 6170 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004932400947770304, | |
| "loss": 2.5295, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000493218292828184, | |
| "loss": 2.5481, | |
| "step": 6190 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004931964562616554, | |
| "loss": 2.5662, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004931745850805528, | |
| "loss": 2.4865, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004931526792879892, | |
| "loss": 2.5432, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004931307388870824, | |
| "loss": 2.4741, | |
| "step": 6230 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004931087638809553, | |
| "loss": 2.5365, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004930867542727354, | |
| "loss": 2.5459, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004930647100655556, | |
| "loss": 2.498, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004930426312625536, | |
| "loss": 2.5288, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004930205178668716, | |
| "loss": 2.551, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004929983698816572, | |
| "loss": 2.5706, | |
| "step": 6290 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004929761873100628, | |
| "loss": 2.5562, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004929539701552456, | |
| "loss": 2.5513, | |
| "step": 6310 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004929317184203678, | |
| "loss": 2.5474, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004929094321085967, | |
| "loss": 2.5221, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004928871112231041, | |
| "loss": 2.575, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004928647557670672, | |
| "loss": 2.5772, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004928423657436676, | |
| "loss": 2.523, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004928199411560924, | |
| "loss": 2.5617, | |
| "step": 6370 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004927974820075332, | |
| "loss": 2.5132, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004927749883011866, | |
| "loss": 2.5358, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004927524600402543, | |
| "loss": 2.5094, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004927298972279428, | |
| "loss": 2.5867, | |
| "step": 6410 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004927072998674633, | |
| "loss": 2.4521, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004926846679620323, | |
| "loss": 2.5832, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004926620015148711, | |
| "loss": 2.5095, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004926393005292057, | |
| "loss": 2.606, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004926165650082671, | |
| "loss": 2.5507, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004925937949552914, | |
| "loss": 2.539, | |
| "step": 6470 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004925709903735196, | |
| "loss": 2.474, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004925481512661975, | |
| "loss": 2.5245, | |
| "step": 6490 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004925252776365755, | |
| "loss": 2.5624, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004925023694879097, | |
| "loss": 2.5657, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004924794268234603, | |
| "loss": 2.5445, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004924564496464931, | |
| "loss": 2.4908, | |
| "step": 6530 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004924334379602781, | |
| "loss": 2.5565, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004924103917680909, | |
| "loss": 2.6322, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004923873110732115, | |
| "loss": 2.5302, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004923641958789252, | |
| "loss": 2.5552, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000492341046188522, | |
| "loss": 2.5387, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004923178620052966, | |
| "loss": 2.5912, | |
| "step": 6590 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004922946433325491, | |
| "loss": 2.5186, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004922713901735842, | |
| "loss": 2.6154, | |
| "step": 6610 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004922481025317115, | |
| "loss": 2.499, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004922247804102456, | |
| "loss": 2.5363, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004922037610236175, | |
| "loss": 2.5905, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004921803734000738, | |
| "loss": 2.5299, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004921569513065769, | |
| "loss": 2.5492, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004921334947464604, | |
| "loss": 2.508, | |
| "step": 6670 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004921100037230632, | |
| "loss": 2.4594, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004920864782397285, | |
| "loss": 2.5114, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004920629182998049, | |
| "loss": 2.5465, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004920393239066459, | |
| "loss": 2.5416, | |
| "step": 6710 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004920156950636093, | |
| "loss": 2.5362, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004919920317740584, | |
| "loss": 2.4794, | |
| "step": 6730 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004919683340413616, | |
| "loss": 2.4981, | |
| "step": 6740 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004919446018688913, | |
| "loss": 2.5181, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004919208352600256, | |
| "loss": 2.5417, | |
| "step": 6760 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004918970342181473, | |
| "loss": 2.5154, | |
| "step": 6770 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004918731987466439, | |
| "loss": 2.5431, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004918493288489081, | |
| "loss": 2.5907, | |
| "step": 6790 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004918254245283372, | |
| "loss": 2.5224, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004918014857883335, | |
| "loss": 2.5333, | |
| "step": 6810 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004917775126323045, | |
| "loss": 2.574, | |
| "step": 6820 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004917535050636622, | |
| "loss": 2.5394, | |
| "step": 6830 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004917294630858235, | |
| "loss": 2.4887, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0004917053867022105, | |
| "loss": 2.5104, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00049168127591625, | |
| "loss": 2.468, | |
| "step": 6860 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004916571307313737, | |
| "loss": 2.5628, | |
| "step": 6870 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004916329511510183, | |
| "loss": 2.5458, | |
| "step": 6880 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004916087371786251, | |
| "loss": 2.4946, | |
| "step": 6890 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004915844888176409, | |
| "loss": 2.5286, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004915602060715168, | |
| "loss": 2.5253, | |
| "step": 6910 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004915358889437088, | |
| "loss": 2.4748, | |
| "step": 6920 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004915115374376784, | |
| "loss": 2.5094, | |
| "step": 6930 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004914871515568913, | |
| "loss": 2.5571, | |
| "step": 6940 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004914627313048185, | |
| "loss": 2.5289, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004914382766849357, | |
| "loss": 2.5304, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004914137877007237, | |
| "loss": 2.4569, | |
| "step": 6970 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004913892643556678, | |
| "loss": 2.537, | |
| "step": 6980 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004913647066532589, | |
| "loss": 2.518, | |
| "step": 6990 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004913401145969919, | |
| "loss": 2.5473, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004913154881903673, | |
| "loss": 2.5467, | |
| "step": 7010 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004912908274368902, | |
| "loss": 2.5051, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004912661323400703, | |
| "loss": 2.5434, | |
| "step": 7030 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004912414029034229, | |
| "loss": 2.5079, | |
| "step": 7040 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004912166391304676, | |
| "loss": 2.5231, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004911918410247289, | |
| "loss": 2.5327, | |
| "step": 7060 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004911670085897367, | |
| "loss": 2.4631, | |
| "step": 7070 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004911421418290253, | |
| "loss": 2.4736, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.000491117240746134, | |
| "loss": 2.5329, | |
| "step": 7090 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.000491092305344607, | |
| "loss": 2.5367, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004910673356279934, | |
| "loss": 2.4862, | |
| "step": 7110 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004910423315998472, | |
| "loss": 2.5973, | |
| "step": 7120 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004910172932637274, | |
| "loss": 2.476, | |
| "step": 7130 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004909922206231974, | |
| "loss": 2.5796, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004909671136818262, | |
| "loss": 2.5209, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004909419724431872, | |
| "loss": 2.5255, | |
| "step": 7160 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004909167969108588, | |
| "loss": 2.528, | |
| "step": 7170 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004908915870884241, | |
| "loss": 2.5318, | |
| "step": 7180 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004908663429794715, | |
| "loss": 2.5298, | |
| "step": 7190 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.000490841064587594, | |
| "loss": 2.5611, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004908157519163893, | |
| "loss": 2.5744, | |
| "step": 7210 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004907904049694605, | |
| "loss": 2.4975, | |
| "step": 7220 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.000490765023750415, | |
| "loss": 2.5416, | |
| "step": 7230 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004907396082628656, | |
| "loss": 2.5316, | |
| "step": 7240 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004907141585104296, | |
| "loss": 2.4884, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004906886744967293, | |
| "loss": 2.4595, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004906631562253918, | |
| "loss": 2.6201, | |
| "step": 7270 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0004906376037000494, | |
| "loss": 2.5497, | |
| "step": 7280 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004906120169243389, | |
| "loss": 2.5548, | |
| "step": 7290 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000490586395901902, | |
| "loss": 2.5192, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004905607406363856, | |
| "loss": 2.4508, | |
| "step": 7310 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000490535051131441, | |
| "loss": 2.5351, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000490509327390725, | |
| "loss": 2.5056, | |
| "step": 7330 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004904835694178986, | |
| "loss": 2.5102, | |
| "step": 7340 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004904577772166281, | |
| "loss": 2.5897, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004904319507905844, | |
| "loss": 2.5587, | |
| "step": 7360 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004904060901434436, | |
| "loss": 2.5699, | |
| "step": 7370 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004903801952788865, | |
| "loss": 2.4632, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004903542662005986, | |
| "loss": 2.5195, | |
| "step": 7390 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004903309007804499, | |
| "loss": 2.4926, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004903049067062451, | |
| "loss": 2.5151, | |
| "step": 7410 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004902788784290256, | |
| "loss": 2.5064, | |
| "step": 7420 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004902528159524959, | |
| "loss": 2.5318, | |
| "step": 7430 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004902267192803657, | |
| "loss": 2.4508, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004902005884163492, | |
| "loss": 2.5051, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004901744233641659, | |
| "loss": 2.574, | |
| "step": 7460 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004901482241275396, | |
| "loss": 2.5474, | |
| "step": 7470 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004901219907101995, | |
| "loss": 2.5259, | |
| "step": 7480 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004900957231158793, | |
| "loss": 2.5176, | |
| "step": 7490 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000490069421348318, | |
| "loss": 2.4825, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004900430854112589, | |
| "loss": 2.5205, | |
| "step": 7510 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004900167153084504, | |
| "loss": 2.5239, | |
| "step": 7520 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004899903110436461, | |
| "loss": 2.5319, | |
| "step": 7530 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004899638726206039, | |
| "loss": 2.548, | |
| "step": 7540 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000489937400043087, | |
| "loss": 2.5479, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000489910893314863, | |
| "loss": 2.5116, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000489884352439705, | |
| "loss": 2.5425, | |
| "step": 7570 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004898577774213905, | |
| "loss": 2.5161, | |
| "step": 7580 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004898311682637018, | |
| "loss": 2.4576, | |
| "step": 7590 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004898045249704264, | |
| "loss": 2.4789, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004897778475453565, | |
| "loss": 2.548, | |
| "step": 7610 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004897511359922891, | |
| "loss": 2.4826, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000489724390315026, | |
| "loss": 2.5211, | |
| "step": 7630 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004896976105173742, | |
| "loss": 2.5603, | |
| "step": 7640 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000489670796603145, | |
| "loss": 2.5257, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004896439485761551, | |
| "loss": 2.5226, | |
| "step": 7660 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004896170664402257, | |
| "loss": 2.5378, | |
| "step": 7670 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004895901501991831, | |
| "loss": 2.5309, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004895631998568583, | |
| "loss": 2.5318, | |
| "step": 7690 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004895362154170871, | |
| "loss": 2.5586, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004895091968837105, | |
| "loss": 2.5026, | |
| "step": 7710 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004894821442605736, | |
| "loss": 2.5712, | |
| "step": 7720 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004894550575515273, | |
| "loss": 2.5526, | |
| "step": 7730 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004894279367604268, | |
| "loss": 2.5771, | |
| "step": 7740 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004894007818911321, | |
| "loss": 2.5358, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004893735929475083, | |
| "loss": 2.5122, | |
| "step": 7760 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004893463699334254, | |
| "loss": 2.5698, | |
| "step": 7770 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004893191128527577, | |
| "loss": 2.5597, | |
| "step": 7780 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004892918217093852, | |
| "loss": 2.5425, | |
| "step": 7790 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004892644965071919, | |
| "loss": 2.596, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004892371372500674, | |
| "loss": 2.5331, | |
| "step": 7810 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004892097439419057, | |
| "loss": 2.5161, | |
| "step": 7820 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004891823165866055, | |
| "loss": 2.5049, | |
| "step": 7830 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004891548551880706, | |
| "loss": 2.5163, | |
| "step": 7840 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004891273597502101, | |
| "loss": 2.5393, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004890998302769369, | |
| "loss": 2.5318, | |
| "step": 7860 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004890722667721697, | |
| "loss": 2.4846, | |
| "step": 7870 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004890446692398315, | |
| "loss": 2.5587, | |
| "step": 7880 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004890170376838503, | |
| "loss": 2.4969, | |
| "step": 7890 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.000488989372108159, | |
| "loss": 2.5095, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004889616725166951, | |
| "loss": 2.4844, | |
| "step": 7910 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004889339389134016, | |
| "loss": 2.5708, | |
| "step": 7920 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004889061713022253, | |
| "loss": 2.5844, | |
| "step": 7930 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004888783696871188, | |
| "loss": 2.5585, | |
| "step": 7940 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004888505340720391, | |
| "loss": 2.5356, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004888226644609478, | |
| "loss": 2.4999, | |
| "step": 7960 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.000488794760857812, | |
| "loss": 2.5001, | |
| "step": 7970 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004887668232666032, | |
| "loss": 2.5503, | |
| "step": 7980 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004887388516912976, | |
| "loss": 2.5694, | |
| "step": 7990 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004887108461358767, | |
| "loss": 2.5327, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004886828066043263, | |
| "loss": 2.5373, | |
| "step": 8010 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004886547331006375, | |
| "loss": 2.5168, | |
| "step": 8020 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004886266256288062, | |
| "loss": 2.5754, | |
| "step": 8030 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004885984841928326, | |
| "loss": 2.5211, | |
| "step": 8040 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004885703087967224, | |
| "loss": 2.5467, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004885420994444858, | |
| "loss": 2.5561, | |
| "step": 8060 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004885138561401377, | |
| "loss": 2.5086, | |
| "step": 8070 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004884855788876984, | |
| "loss": 2.5144, | |
| "step": 8080 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004884572676911923, | |
| "loss": 2.4998, | |
| "step": 8090 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004884289225546491, | |
| "loss": 2.4899, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0004884005434821032, | |
| "loss": 2.5413, | |
| "step": 8110 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.000488372130477594, | |
| "loss": 2.4236, | |
| "step": 8120 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004883436835451654, | |
| "loss": 2.5394, | |
| "step": 8130 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004883152026888664, | |
| "loss": 2.5144, | |
| "step": 8140 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004882866879127505, | |
| "loss": 2.5319, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048825813922087646, | |
| "loss": 2.4748, | |
| "step": 8160 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004882295566173077, | |
| "loss": 2.5425, | |
| "step": 8170 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048820094010611236, | |
| "loss": 2.4608, | |
| "step": 8180 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004881722896913635, | |
| "loss": 2.5557, | |
| "step": 8190 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004881436053771389, | |
| "loss": 2.5616, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004881148871675214, | |
| "loss": 2.5322, | |
| "step": 8210 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004880861350665983, | |
| "loss": 2.5417, | |
| "step": 8220 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004880573490784622, | |
| "loss": 2.5991, | |
| "step": 8230 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048802852920721, | |
| "loss": 2.429, | |
| "step": 8240 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004879996754569439, | |
| "loss": 2.513, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048797078783177053, | |
| "loss": 2.5191, | |
| "step": 8260 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048794186633580164, | |
| "loss": 2.5401, | |
| "step": 8270 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048791291097315364, | |
| "loss": 2.5169, | |
| "step": 8280 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004878839217479478, | |
| "loss": 2.5629, | |
| "step": 8290 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048785489866431014, | |
| "loss": 2.4349, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048782584172637165, | |
| "loss": 2.5095, | |
| "step": 8310 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048779675093826803, | |
| "loss": 2.4796, | |
| "step": 8320 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048776762630413985, | |
| "loss": 2.5365, | |
| "step": 8330 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048773846782813234, | |
| "loss": 2.5191, | |
| "step": 8340 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048770927551439593, | |
| "loss": 2.5618, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048768004936708534, | |
| "loss": 2.5195, | |
| "step": 8360 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004876507893903605, | |
| "loss": 2.5409, | |
| "step": 8370 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048762149558838607, | |
| "loss": 2.4873, | |
| "step": 8380 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048759216796533144, | |
| "loss": 2.5492, | |
| "step": 8390 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004875628065253709, | |
| "loss": 2.5033, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048753341127268334, | |
| "loss": 2.5711, | |
| "step": 8410 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048750398221145293, | |
| "loss": 2.538, | |
| "step": 8420 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004874745193458682, | |
| "loss": 2.5147, | |
| "step": 8430 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048744502268012257, | |
| "loss": 2.4645, | |
| "step": 8440 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004874154922184145, | |
| "loss": 2.568, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004873859279649471, | |
| "loss": 2.5086, | |
| "step": 8460 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004873563299239281, | |
| "loss": 2.4851, | |
| "step": 8470 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048732669809957056, | |
| "loss": 2.5279, | |
| "step": 8480 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004872970324960918, | |
| "loss": 2.5428, | |
| "step": 8490 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0004872673331177142, | |
| "loss": 2.5061, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048723759996866504, | |
| "loss": 2.5048, | |
| "step": 8510 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048720783305317616, | |
| "loss": 2.4591, | |
| "step": 8520 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00048717803237548446, | |
| "loss": 2.5629, | |
| "step": 8530 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048714819793983143, | |
| "loss": 2.4768, | |
| "step": 8540 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048711832975046346, | |
| "loss": 2.5034, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048708842781163176, | |
| "loss": 2.5183, | |
| "step": 8560 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004870584921275923, | |
| "loss": 2.518, | |
| "step": 8570 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.000487028522702606, | |
| "loss": 2.5804, | |
| "step": 8580 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004869985195409383, | |
| "loss": 2.5195, | |
| "step": 8590 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004869684826468597, | |
| "loss": 2.4545, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004869384120246454, | |
| "loss": 2.4323, | |
| "step": 8610 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004869083076785753, | |
| "loss": 2.4878, | |
| "step": 8620 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048687816961293437, | |
| "loss": 2.5676, | |
| "step": 8630 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048684799783201205, | |
| "loss": 2.4564, | |
| "step": 8640 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048681779234010283, | |
| "loss": 2.5636, | |
| "step": 8650 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048678755314150593, | |
| "loss": 2.5314, | |
| "step": 8660 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048675728024052527, | |
| "loss": 2.5547, | |
| "step": 8670 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048672697364146974, | |
| "loss": 2.5618, | |
| "step": 8680 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004866966333486528, | |
| "loss": 2.5123, | |
| "step": 8690 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048666625936639297, | |
| "loss": 2.5665, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048663585169901336, | |
| "loss": 2.5149, | |
| "step": 8710 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048660541035084205, | |
| "loss": 2.5337, | |
| "step": 8720 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004865749353262116, | |
| "loss": 2.4607, | |
| "step": 8730 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048654442662945973, | |
| "loss": 2.5306, | |
| "step": 8740 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004865138842649288, | |
| "loss": 2.4574, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048648330823696586, | |
| "loss": 2.4983, | |
| "step": 8760 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048645269854992294, | |
| "loss": 2.5543, | |
| "step": 8770 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048642205520815676, | |
| "loss": 2.583, | |
| "step": 8780 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048639137821602883, | |
| "loss": 2.4791, | |
| "step": 8790 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004863606675779054, | |
| "loss": 2.542, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004863299232981576, | |
| "loss": 2.4648, | |
| "step": 8810 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004862991453811614, | |
| "loss": 2.4428, | |
| "step": 8820 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048626833383129744, | |
| "loss": 2.5686, | |
| "step": 8830 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048623748865295104, | |
| "loss": 2.5727, | |
| "step": 8840 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048620660985051263, | |
| "loss": 2.5765, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048617569742837713, | |
| "loss": 2.4668, | |
| "step": 8860 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048614475139094446, | |
| "loss": 2.5309, | |
| "step": 8870 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004861137717426191, | |
| "loss": 2.573, | |
| "step": 8880 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004860827584878105, | |
| "loss": 2.513, | |
| "step": 8890 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004860517116309329, | |
| "loss": 2.5, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004860206311764052, | |
| "loss": 2.5337, | |
| "step": 8910 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048598951712865103, | |
| "loss": 2.5854, | |
| "step": 8920 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048595836949209896, | |
| "loss": 2.5265, | |
| "step": 8930 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00048592718827118243, | |
| "loss": 2.5244, | |
| "step": 8940 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004858959734703393, | |
| "loss": 2.5271, | |
| "step": 8950 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048586472509401263, | |
| "loss": 2.5153, | |
| "step": 8960 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048583344314664993, | |
| "loss": 2.4984, | |
| "step": 8970 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004858021276327037, | |
| "loss": 2.5511, | |
| "step": 8980 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000485770778556631, | |
| "loss": 2.5409, | |
| "step": 8990 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004857393959228939, | |
| "loss": 2.5153, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004857079797359591, | |
| "loss": 2.593, | |
| "step": 9010 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004856765300002982, | |
| "loss": 2.5274, | |
| "step": 9020 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004856450467203875, | |
| "loss": 2.521, | |
| "step": 9030 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004856135299007078, | |
| "loss": 2.5454, | |
| "step": 9040 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004855819795457453, | |
| "loss": 2.4956, | |
| "step": 9050 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004855503956599904, | |
| "loss": 2.4448, | |
| "step": 9060 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004855187782479386, | |
| "loss": 2.5293, | |
| "step": 9070 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048548712731409, | |
| "loss": 2.5492, | |
| "step": 9080 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004854554428629496, | |
| "loss": 2.5481, | |
| "step": 9090 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000485423724899027, | |
| "loss": 2.4744, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004853919734268367, | |
| "loss": 2.489, | |
| "step": 9110 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000485360188450898, | |
| "loss": 2.5009, | |
| "step": 9120 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048532836997573493, | |
| "loss": 2.5212, | |
| "step": 9130 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048529651800587617, | |
| "loss": 2.5175, | |
| "step": 9140 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004852646325458553, | |
| "loss": 2.5256, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048523271360021065, | |
| "loss": 2.4885, | |
| "step": 9160 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048520076117348533, | |
| "loss": 2.5278, | |
| "step": 9170 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048516877527022704, | |
| "loss": 2.5603, | |
| "step": 9180 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048513675589498864, | |
| "loss": 2.5113, | |
| "step": 9190 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048510470305232733, | |
| "loss": 2.4844, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048507261674680523, | |
| "loss": 2.4827, | |
| "step": 9210 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004850404969829894, | |
| "loss": 2.5388, | |
| "step": 9220 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004850083437654512, | |
| "loss": 2.5186, | |
| "step": 9230 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004849761570987674, | |
| "loss": 2.5464, | |
| "step": 9240 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048494393698751904, | |
| "loss": 2.57, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048491168343629197, | |
| "loss": 2.5314, | |
| "step": 9260 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000484879396449677, | |
| "loss": 2.4585, | |
| "step": 9270 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048484707603226957, | |
| "loss": 2.5167, | |
| "step": 9280 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004848147221886699, | |
| "loss": 2.528, | |
| "step": 9290 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000484782334923483, | |
| "loss": 2.4785, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048474991424131853, | |
| "loss": 2.5506, | |
| "step": 9310 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004847174601467911, | |
| "loss": 2.5288, | |
| "step": 9320 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004846849726445198, | |
| "loss": 2.4625, | |
| "step": 9330 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00048465245173912876, | |
| "loss": 2.5387, | |
| "step": 9340 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004846198974352467, | |
| "loss": 2.5236, | |
| "step": 9350 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0004845873097375071, | |
| "loss": 2.6822, | |
| "step": 9360 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048455468865054826, | |
| "loss": 2.5064, | |
| "step": 9370 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048452203417901317, | |
| "loss": 2.4583, | |
| "step": 9380 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004844893463275496, | |
| "loss": 2.493, | |
| "step": 9390 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004844566251008101, | |
| "loss": 2.5302, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004844238705034519, | |
| "loss": 2.5663, | |
| "step": 9410 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048439108254013714, | |
| "loss": 2.4936, | |
| "step": 9420 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004843582612155324, | |
| "loss": 2.4725, | |
| "step": 9430 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004843254065343093, | |
| "loss": 2.5412, | |
| "step": 9440 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004842925185011441, | |
| "loss": 2.5752, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048425959712071774, | |
| "loss": 2.5842, | |
| "step": 9460 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048422664239771606, | |
| "loss": 2.5258, | |
| "step": 9470 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004841936543368296, | |
| "loss": 2.5472, | |
| "step": 9480 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048416063294275347, | |
| "loss": 2.4784, | |
| "step": 9490 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004841275782201878, | |
| "loss": 2.5275, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048409449017383725, | |
| "loss": 2.5766, | |
| "step": 9510 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048406136880841126, | |
| "loss": 2.4899, | |
| "step": 9520 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004840282141286242, | |
| "loss": 2.5405, | |
| "step": 9530 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048399502613919484, | |
| "loss": 2.5145, | |
| "step": 9540 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048396180484484706, | |
| "loss": 2.5267, | |
| "step": 9550 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048392855025030914, | |
| "loss": 2.5416, | |
| "step": 9560 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048389526236031444, | |
| "loss": 2.5031, | |
| "step": 9570 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004838619411796007, | |
| "loss": 2.486, | |
| "step": 9580 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048382858671291073, | |
| "loss": 2.4852, | |
| "step": 9590 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004837951989649919, | |
| "loss": 2.5473, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004837617779405963, | |
| "loss": 2.5644, | |
| "step": 9610 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048372832364448084, | |
| "loss": 2.4929, | |
| "step": 9620 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004836948360814071, | |
| "loss": 2.5748, | |
| "step": 9630 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048366131525614137, | |
| "loss": 2.4606, | |
| "step": 9640 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048362776117345487, | |
| "loss": 2.5153, | |
| "step": 9650 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004835941738381233, | |
| "loss": 2.5266, | |
| "step": 9660 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004835605532549272, | |
| "loss": 2.5345, | |
| "step": 9670 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048352689942865196, | |
| "loss": 2.5751, | |
| "step": 9680 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004834932123640875, | |
| "loss": 2.5734, | |
| "step": 9690 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004834594920660286, | |
| "loss": 2.5082, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048342573853927455, | |
| "loss": 2.4498, | |
| "step": 9710 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048339195178862985, | |
| "loss": 2.5814, | |
| "step": 9720 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004833581318189032, | |
| "loss": 2.5311, | |
| "step": 9730 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048332427863490837, | |
| "loss": 2.5387, | |
| "step": 9740 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048329039224146364, | |
| "loss": 2.529, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004832564726433922, | |
| "loss": 2.5404, | |
| "step": 9760 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00048322251984552183, | |
| "loss": 2.5415, | |
| "step": 9770 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0004831885338526852, | |
| "loss": 2.4912, | |
| "step": 9780 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004831545146697195, | |
| "loss": 2.4882, | |
| "step": 9790 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004831204623014667, | |
| "loss": 2.5481, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048308637675277365, | |
| "loss": 2.5279, | |
| "step": 9810 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048305225802849173, | |
| "loss": 2.5243, | |
| "step": 9820 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004830181061334771, | |
| "loss": 2.5333, | |
| "step": 9830 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004829839210725908, | |
| "loss": 2.5383, | |
| "step": 9840 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048294970285069826, | |
| "loss": 2.5704, | |
| "step": 9850 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004829154514726699, | |
| "loss": 2.5403, | |
| "step": 9860 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004828811669433808, | |
| "loss": 2.535, | |
| "step": 9870 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004828468492677107, | |
| "loss": 2.4692, | |
| "step": 9880 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048281249845054413, | |
| "loss": 2.5308, | |
| "step": 9890 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004827781144967703, | |
| "loss": 2.4852, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004827436974112831, | |
| "loss": 2.4735, | |
| "step": 9910 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004827092471989811, | |
| "loss": 2.5586, | |
| "step": 9920 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048267476386476793, | |
| "loss": 2.5126, | |
| "step": 9930 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004826402474135514, | |
| "loss": 2.5112, | |
| "step": 9940 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004826056978502443, | |
| "loss": 2.5092, | |
| "step": 9950 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004825711151797643, | |
| "loss": 2.5596, | |
| "step": 9960 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004825364994070335, | |
| "loss": 2.5572, | |
| "step": 9970 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048250185053697893, | |
| "loss": 2.5777, | |
| "step": 9980 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048246716857453203, | |
| "loss": 2.536, | |
| "step": 9990 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004824324535246293, | |
| "loss": 2.5377, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048239770539221173, | |
| "loss": 2.5726, | |
| "step": 10010 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004823629241822251, | |
| "loss": 2.5471, | |
| "step": 10020 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048232810989961986, | |
| "loss": 2.5314, | |
| "step": 10030 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048229326254935125, | |
| "loss": 2.4723, | |
| "step": 10040 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004822583821363791, | |
| "loss": 2.5341, | |
| "step": 10050 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048222346866566795, | |
| "loss": 2.5401, | |
| "step": 10060 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004821885221421872, | |
| "loss": 2.5061, | |
| "step": 10070 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004821535425709108, | |
| "loss": 2.5217, | |
| "step": 10080 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004821185299568175, | |
| "loss": 2.5295, | |
| "step": 10090 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004820834843048906, | |
| "loss": 2.4793, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048204840562011833, | |
| "loss": 2.5249, | |
| "step": 10110 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004820132939074934, | |
| "loss": 2.494, | |
| "step": 10120 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004819781491720133, | |
| "loss": 2.5209, | |
| "step": 10130 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048194297141868043, | |
| "loss": 2.5347, | |
| "step": 10140 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048190776065250155, | |
| "loss": 2.5497, | |
| "step": 10150 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004818725168784882, | |
| "loss": 2.5709, | |
| "step": 10160 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048183724010165686, | |
| "loss": 2.5479, | |
| "step": 10170 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004818019303270285, | |
| "loss": 2.4623, | |
| "step": 10180 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048176658755962867, | |
| "loss": 2.5178, | |
| "step": 10190 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00048173121180448796, | |
| "loss": 2.5248, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004816958030666413, | |
| "loss": 2.5025, | |
| "step": 10210 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004816603613511286, | |
| "loss": 2.5385, | |
| "step": 10220 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004816248866629943, | |
| "loss": 2.5499, | |
| "step": 10230 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004815893790072876, | |
| "loss": 2.5252, | |
| "step": 10240 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004815538383890623, | |
| "loss": 2.5239, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.000481518264813377, | |
| "loss": 2.5169, | |
| "step": 10260 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004814826582852949, | |
| "loss": 2.5451, | |
| "step": 10270 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048144701880988407, | |
| "loss": 2.4921, | |
| "step": 10280 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.000481411346392217, | |
| "loss": 2.519, | |
| "step": 10290 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048137564103737106, | |
| "loss": 2.5891, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004813399027504283, | |
| "loss": 2.4593, | |
| "step": 10310 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004813041315364753, | |
| "loss": 2.5655, | |
| "step": 10320 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048126832740060356, | |
| "loss": 2.5011, | |
| "step": 10330 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.000481232490347909, | |
| "loss": 2.468, | |
| "step": 10340 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048119662038349255, | |
| "loss": 2.5139, | |
| "step": 10350 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004811607175124595, | |
| "loss": 2.4483, | |
| "step": 10360 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048112478173992, | |
| "loss": 2.5341, | |
| "step": 10370 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048108881307098897, | |
| "loss": 2.4655, | |
| "step": 10380 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048105281151078575, | |
| "loss": 2.5499, | |
| "step": 10390 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004810167770644345, | |
| "loss": 2.5231, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004809807097370642, | |
| "loss": 2.5663, | |
| "step": 10410 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048094460953380827, | |
| "loss": 2.4315, | |
| "step": 10420 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004809084764598049, | |
| "loss": 2.5684, | |
| "step": 10430 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048087231052019704, | |
| "loss": 2.5453, | |
| "step": 10440 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048083611172013217, | |
| "loss": 2.5265, | |
| "step": 10450 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004807998800647626, | |
| "loss": 2.4804, | |
| "step": 10460 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004807636155592453, | |
| "loss": 2.5568, | |
| "step": 10470 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004807273182087417, | |
| "loss": 2.5621, | |
| "step": 10480 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048069098801841815, | |
| "loss": 2.4805, | |
| "step": 10490 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048065462499344557, | |
| "loss": 2.5492, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048061822913899964, | |
| "loss": 2.4595, | |
| "step": 10510 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004805818004602606, | |
| "loss": 2.5164, | |
| "step": 10520 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004805453389624134, | |
| "loss": 2.4818, | |
| "step": 10530 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004805088446506476, | |
| "loss": 2.5265, | |
| "step": 10540 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004804723175301576, | |
| "loss": 2.4116, | |
| "step": 10550 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048043575760614246, | |
| "loss": 2.5084, | |
| "step": 10560 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004803991648838055, | |
| "loss": 2.489, | |
| "step": 10570 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048036253936835534, | |
| "loss": 2.5187, | |
| "step": 10580 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004803258810650048, | |
| "loss": 2.5113, | |
| "step": 10590 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00048028918997897153, | |
| "loss": 2.498, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0004802524661154779, | |
| "loss": 2.5112, | |
| "step": 10610 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00048021570947975086, | |
| "loss": 2.4791, | |
| "step": 10620 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.000480178920077022, | |
| "loss": 2.5017, | |
| "step": 10630 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004801420979125276, | |
| "loss": 2.4354, | |
| "step": 10640 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004801052429915087, | |
| "loss": 2.5091, | |
| "step": 10650 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00048006835531921096, | |
| "loss": 2.4623, | |
| "step": 10660 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004800314349008845, | |
| "loss": 2.5214, | |
| "step": 10670 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004799944817417844, | |
| "loss": 2.566, | |
| "step": 10680 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004799574958471703, | |
| "loss": 2.5156, | |
| "step": 10690 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004799204772223064, | |
| "loss": 2.4196, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004798834258724615, | |
| "loss": 2.5032, | |
| "step": 10710 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047984634180290935, | |
| "loss": 2.5608, | |
| "step": 10720 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047980922501892823, | |
| "loss": 2.5363, | |
| "step": 10730 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047977207552580095, | |
| "loss": 2.5091, | |
| "step": 10740 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.000479734893328815, | |
| "loss": 2.5386, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004796976784332626, | |
| "loss": 2.5165, | |
| "step": 10760 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047966043084444077, | |
| "loss": 2.5297, | |
| "step": 10770 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047962315056765094, | |
| "loss": 2.4516, | |
| "step": 10780 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004795858376081992, | |
| "loss": 2.5011, | |
| "step": 10790 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047954849197139645, | |
| "loss": 2.4533, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047951111366255805, | |
| "loss": 2.5183, | |
| "step": 10810 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004794737026870043, | |
| "loss": 2.4444, | |
| "step": 10820 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047943625905005985, | |
| "loss": 2.5066, | |
| "step": 10830 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004793987827570542, | |
| "loss": 2.4947, | |
| "step": 10840 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004793612738133213, | |
| "loss": 2.4873, | |
| "step": 10850 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047932373222419987, | |
| "loss": 2.54, | |
| "step": 10860 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047928615799503333, | |
| "loss": 2.5693, | |
| "step": 10870 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004792485511311697, | |
| "loss": 2.56, | |
| "step": 10880 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004792109116379616, | |
| "loss": 2.4267, | |
| "step": 10890 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047917323952076634, | |
| "loss": 2.4614, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047913553478494587, | |
| "loss": 2.4743, | |
| "step": 10910 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004790977974358667, | |
| "loss": 2.4688, | |
| "step": 10920 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004790600274789001, | |
| "loss": 2.4796, | |
| "step": 10930 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047902222491942196, | |
| "loss": 2.4828, | |
| "step": 10940 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004789843897628128, | |
| "loss": 2.6211, | |
| "step": 10950 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047894652201445765, | |
| "loss": 2.591, | |
| "step": 10960 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047890862167974637, | |
| "loss": 2.486, | |
| "step": 10970 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047887068876407334, | |
| "loss": 2.4526, | |
| "step": 10980 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004788327232728378, | |
| "loss": 2.5296, | |
| "step": 10990 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004787947252114432, | |
| "loss": 2.5839, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004787566945852979, | |
| "loss": 2.5249, | |
| "step": 11010 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00047871863139981497, | |
| "loss": 2.5182, | |
| "step": 11020 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.000478680535660412, | |
| "loss": 2.5672, | |
| "step": 11030 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004786424073725112, | |
| "loss": 2.5092, | |
| "step": 11040 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047860424654153945, | |
| "loss": 2.4808, | |
| "step": 11050 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004785660531729281, | |
| "loss": 2.5133, | |
| "step": 11060 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004785278272721135, | |
| "loss": 2.4587, | |
| "step": 11070 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047848956884453633, | |
| "loss": 2.5353, | |
| "step": 11080 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004784512778956419, | |
| "loss": 2.5249, | |
| "step": 11090 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004784129544308804, | |
| "loss": 2.5099, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004783745984557063, | |
| "loss": 2.5517, | |
| "step": 11110 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004783362099755789, | |
| "loss": 2.5725, | |
| "step": 11120 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004782977889959622, | |
| "loss": 2.4226, | |
| "step": 11130 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004782593355223246, | |
| "loss": 2.5053, | |
| "step": 11140 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004782208495601393, | |
| "loss": 2.4869, | |
| "step": 11150 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004781823311148841, | |
| "loss": 2.4909, | |
| "step": 11160 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004781437801920414, | |
| "loss": 2.5832, | |
| "step": 11170 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004781051967970982, | |
| "loss": 2.5066, | |
| "step": 11180 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004780665809355461, | |
| "loss": 2.5261, | |
| "step": 11190 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047802793261288146, | |
| "loss": 2.5102, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004779892518346051, | |
| "loss": 2.5159, | |
| "step": 11210 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047795053860622255, | |
| "loss": 2.4716, | |
| "step": 11220 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004779117929332439, | |
| "loss": 2.5471, | |
| "step": 11230 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047787301482118385, | |
| "loss": 2.5389, | |
| "step": 11240 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004778342042755619, | |
| "loss": 2.4909, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004777953613019018, | |
| "loss": 2.5056, | |
| "step": 11260 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004777564859057324, | |
| "loss": 2.4714, | |
| "step": 11270 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047771757809258667, | |
| "loss": 2.5186, | |
| "step": 11280 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004776786378680026, | |
| "loss": 2.5064, | |
| "step": 11290 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047763966523752243, | |
| "loss": 2.4884, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047760066020669347, | |
| "loss": 2.528, | |
| "step": 11310 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004775616227810671, | |
| "loss": 2.5058, | |
| "step": 11320 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047752255296619977, | |
| "loss": 2.5311, | |
| "step": 11330 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004774834507676523, | |
| "loss": 2.5295, | |
| "step": 11340 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047744431619099016, | |
| "loss": 2.5185, | |
| "step": 11350 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047740514924178335, | |
| "loss": 2.5345, | |
| "step": 11360 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047736594992560676, | |
| "loss": 2.5326, | |
| "step": 11370 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047732671824803963, | |
| "loss": 2.5773, | |
| "step": 11380 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004772874542146658, | |
| "loss": 2.5523, | |
| "step": 11390 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047724815783107383, | |
| "loss": 2.4708, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047720882910285686, | |
| "loss": 2.5284, | |
| "step": 11410 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047716946803561254, | |
| "loss": 2.5492, | |
| "step": 11420 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004771300746349434, | |
| "loss": 2.5243, | |
| "step": 11430 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00047709064890645613, | |
| "loss": 2.5171, | |
| "step": 11440 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047705119085576244, | |
| "loss": 2.5814, | |
| "step": 11450 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004770117004884783, | |
| "loss": 2.5789, | |
| "step": 11460 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047697217781022454, | |
| "loss": 2.5091, | |
| "step": 11470 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047693262282662654, | |
| "loss": 2.4775, | |
| "step": 11480 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047689303554331416, | |
| "loss": 2.477, | |
| "step": 11490 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047685341596592193, | |
| "loss": 2.5398, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004768137641000889, | |
| "loss": 2.4787, | |
| "step": 11510 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047677407995145904, | |
| "loss": 2.5357, | |
| "step": 11520 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047673436352568035, | |
| "loss": 2.4794, | |
| "step": 11530 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047669461482840596, | |
| "loss": 2.4977, | |
| "step": 11540 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004766548338652933, | |
| "loss": 2.5189, | |
| "step": 11550 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004766150206420044, | |
| "loss": 2.5602, | |
| "step": 11560 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.000476575175164206, | |
| "loss": 2.5019, | |
| "step": 11570 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047653529743756946, | |
| "loss": 2.4963, | |
| "step": 11580 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047649538746777055, | |
| "loss": 2.472, | |
| "step": 11590 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004764554452604897, | |
| "loss": 2.4669, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.000476415470821412, | |
| "loss": 2.4812, | |
| "step": 11610 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004763754641562271, | |
| "loss": 2.4627, | |
| "step": 11620 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047633542527062914, | |
| "loss": 2.4374, | |
| "step": 11630 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.000476295354170317, | |
| "loss": 2.5043, | |
| "step": 11640 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047625525086099406, | |
| "loss": 2.4857, | |
| "step": 11650 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004762151153483683, | |
| "loss": 2.5738, | |
| "step": 11660 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004761749476381522, | |
| "loss": 2.4797, | |
| "step": 11670 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004761387691747432, | |
| "loss": 2.4641, | |
| "step": 11680 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004760985403048601, | |
| "loss": 2.488, | |
| "step": 11690 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047605827925397906, | |
| "loss": 2.496, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004760179860278305, | |
| "loss": 2.532, | |
| "step": 11710 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047597766063214934, | |
| "loss": 2.5034, | |
| "step": 11720 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047593730307267536, | |
| "loss": 2.5029, | |
| "step": 11730 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004758969133551525, | |
| "loss": 2.4898, | |
| "step": 11740 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004758605351190015, | |
| "loss": 2.5262, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004758200843170276, | |
| "loss": 2.5455, | |
| "step": 11760 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004757796013736888, | |
| "loss": 2.5617, | |
| "step": 11770 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004757390862947472, | |
| "loss": 2.4517, | |
| "step": 11780 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047569853908596937, | |
| "loss": 2.44, | |
| "step": 11790 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047565795975312646, | |
| "loss": 2.4874, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004756173483019942, | |
| "loss": 2.5236, | |
| "step": 11810 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004755767047383529, | |
| "loss": 2.5021, | |
| "step": 11820 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047553602906798737, | |
| "loss": 2.5279, | |
| "step": 11830 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004754953212966871, | |
| "loss": 2.4573, | |
| "step": 11840 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00047545458143024606, | |
| "loss": 2.4706, | |
| "step": 11850 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0004754138094744629, | |
| "loss": 2.4952, | |
| "step": 11860 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004753730054351407, | |
| "loss": 2.5838, | |
| "step": 11870 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047533216931808713, | |
| "loss": 2.5231, | |
| "step": 11880 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047529130112911456, | |
| "loss": 2.5052, | |
| "step": 11890 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047525040087403983, | |
| "loss": 2.4431, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047520946855868424, | |
| "loss": 2.5268, | |
| "step": 11910 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047516850418887385, | |
| "loss": 2.5503, | |
| "step": 11920 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004751275077704392, | |
| "loss": 2.5117, | |
| "step": 11930 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047508647930921525, | |
| "loss": 2.5273, | |
| "step": 11940 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004750454188110418, | |
| "loss": 2.5186, | |
| "step": 11950 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047500432628176306, | |
| "loss": 2.5413, | |
| "step": 11960 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047496320172722765, | |
| "loss": 2.5696, | |
| "step": 11970 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047492204515328906, | |
| "loss": 2.5141, | |
| "step": 11980 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047488085656580516, | |
| "loss": 2.5649, | |
| "step": 11990 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004748396359706383, | |
| "loss": 2.5253, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047479838337365543, | |
| "loss": 2.5452, | |
| "step": 12010 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004747570987807283, | |
| "loss": 2.5055, | |
| "step": 12020 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004747157821977329, | |
| "loss": 2.471, | |
| "step": 12030 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047467443363054994, | |
| "loss": 2.5355, | |
| "step": 12040 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004746330530850645, | |
| "loss": 2.4778, | |
| "step": 12050 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047459164056716665, | |
| "loss": 2.5041, | |
| "step": 12060 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004745501960827503, | |
| "loss": 2.4646, | |
| "step": 12070 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047450871963771466, | |
| "loss": 2.5086, | |
| "step": 12080 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004744672112379629, | |
| "loss": 2.4898, | |
| "step": 12090 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004744256708894031, | |
| "loss": 2.5131, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047438409859794784, | |
| "loss": 2.5167, | |
| "step": 12110 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047434249436951407, | |
| "loss": 2.5112, | |
| "step": 12120 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004743008582100233, | |
| "loss": 2.4866, | |
| "step": 12130 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004742591901254019, | |
| "loss": 2.5069, | |
| "step": 12140 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047421749012158045, | |
| "loss": 2.5768, | |
| "step": 12150 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047417575820449416, | |
| "loss": 2.5358, | |
| "step": 12160 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004741339943800828, | |
| "loss": 2.575, | |
| "step": 12170 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004740921986542908, | |
| "loss": 2.5548, | |
| "step": 12180 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047405037103306685, | |
| "loss": 2.5418, | |
| "step": 12190 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047400851152236444, | |
| "loss": 2.4854, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047396662012814164, | |
| "loss": 2.4689, | |
| "step": 12210 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004739246968563606, | |
| "loss": 2.5099, | |
| "step": 12220 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004738827417129886, | |
| "loss": 2.526, | |
| "step": 12230 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047384075470399716, | |
| "loss": 2.5019, | |
| "step": 12240 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047379873583536225, | |
| "loss": 2.4612, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047375668511306457, | |
| "loss": 2.5453, | |
| "step": 12260 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0004737146025430893, | |
| "loss": 2.4714, | |
| "step": 12270 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00047367248813142603, | |
| "loss": 2.5144, | |
| "step": 12280 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004736303418840691, | |
| "loss": 2.5303, | |
| "step": 12290 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004735881638070171, | |
| "loss": 2.4992, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004735459539062734, | |
| "loss": 2.449, | |
| "step": 12310 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047350371218784576, | |
| "loss": 2.4509, | |
| "step": 12320 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047346143865774663, | |
| "loss": 2.4978, | |
| "step": 12330 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047341913332199276, | |
| "loss": 2.5282, | |
| "step": 12340 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004733767961866056, | |
| "loss": 2.5608, | |
| "step": 12350 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047333442725761113, | |
| "loss": 2.5331, | |
| "step": 12360 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047329202654103965, | |
| "loss": 2.5249, | |
| "step": 12370 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047324959404292615, | |
| "loss": 2.5865, | |
| "step": 12380 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004732071297693102, | |
| "loss": 2.476, | |
| "step": 12390 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004731646337262358, | |
| "loss": 2.5163, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047312210591975146, | |
| "loss": 2.5657, | |
| "step": 12410 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004730795463559103, | |
| "loss": 2.4798, | |
| "step": 12420 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004730369550407697, | |
| "loss": 2.4615, | |
| "step": 12430 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.000472994331980392, | |
| "loss": 2.5564, | |
| "step": 12440 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004729516771808437, | |
| "loss": 2.5585, | |
| "step": 12450 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047290899064819593, | |
| "loss": 2.5662, | |
| "step": 12460 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004728662723885243, | |
| "loss": 2.575, | |
| "step": 12470 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004728235224079091, | |
| "loss": 2.4801, | |
| "step": 12480 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004727807407124349, | |
| "loss": 2.5476, | |
| "step": 12490 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047273792730819094, | |
| "loss": 2.5157, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004726950822012709, | |
| "loss": 2.4381, | |
| "step": 12510 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.000472652205397773, | |
| "loss": 2.5186, | |
| "step": 12520 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004726092969038, | |
| "loss": 2.4887, | |
| "step": 12530 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004725663567254591, | |
| "loss": 2.5257, | |
| "step": 12540 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.000472523384868862, | |
| "loss": 2.5407, | |
| "step": 12550 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047248038134012517, | |
| "loss": 2.4588, | |
| "step": 12560 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004724373461453691, | |
| "loss": 2.4755, | |
| "step": 12570 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004723942792907193, | |
| "loss": 2.4989, | |
| "step": 12580 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004723511807823055, | |
| "loss": 2.5007, | |
| "step": 12590 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004723080506262618, | |
| "loss": 2.5052, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004722648888287272, | |
| "loss": 2.5264, | |
| "step": 12610 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004722216953958449, | |
| "loss": 2.5253, | |
| "step": 12620 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004721784703337628, | |
| "loss": 2.5408, | |
| "step": 12630 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.000472135213648633, | |
| "loss": 2.5209, | |
| "step": 12640 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00047209192534661244, | |
| "loss": 2.4705, | |
| "step": 12650 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004720486054338624, | |
| "loss": 2.4932, | |
| "step": 12660 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004720052539165487, | |
| "loss": 2.5126, | |
| "step": 12670 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004719618708008415, | |
| "loss": 2.4261, | |
| "step": 12680 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004719184560929158, | |
| "loss": 2.4705, | |
| "step": 12690 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047187500979895084, | |
| "loss": 2.4616, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004718315319251303, | |
| "loss": 2.4082, | |
| "step": 12710 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047178802247764247, | |
| "loss": 2.4784, | |
| "step": 12720 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047174448146268023, | |
| "loss": 2.4978, | |
| "step": 12730 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004717009088864408, | |
| "loss": 2.5025, | |
| "step": 12740 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004716573047551259, | |
| "loss": 2.4929, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047161366907494184, | |
| "loss": 2.4521, | |
| "step": 12760 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004715700018520993, | |
| "loss": 2.5522, | |
| "step": 12770 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047152630309281365, | |
| "loss": 2.56, | |
| "step": 12780 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004714825728033044, | |
| "loss": 2.5355, | |
| "step": 12790 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004714388109897959, | |
| "loss": 2.4948, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004713950176585169, | |
| "loss": 2.4317, | |
| "step": 12810 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047135119281570037, | |
| "loss": 2.5288, | |
| "step": 12820 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004713073364675842, | |
| "loss": 2.5371, | |
| "step": 12830 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004712634486204104, | |
| "loss": 2.5313, | |
| "step": 12840 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004712195292804256, | |
| "loss": 2.5361, | |
| "step": 12850 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.000471175578453881, | |
| "loss": 2.4454, | |
| "step": 12860 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047113159614703214, | |
| "loss": 2.5771, | |
| "step": 12870 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004710875823661391, | |
| "loss": 2.4706, | |
| "step": 12880 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004710435371174665, | |
| "loss": 2.5341, | |
| "step": 12890 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004709994604072833, | |
| "loss": 2.495, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.000470955352241863, | |
| "loss": 2.4172, | |
| "step": 12910 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004709112126274838, | |
| "loss": 2.5732, | |
| "step": 12920 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047086704157042785, | |
| "loss": 2.5008, | |
| "step": 12930 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004708228390769823, | |
| "loss": 2.5286, | |
| "step": 12940 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004707786051534385, | |
| "loss": 2.4573, | |
| "step": 12950 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004707343398060924, | |
| "loss": 2.5211, | |
| "step": 12960 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004706900430412443, | |
| "loss": 2.5025, | |
| "step": 12970 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047064571486519905, | |
| "loss": 2.5758, | |
| "step": 12980 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047060135528426594, | |
| "loss": 2.4291, | |
| "step": 12990 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047055696430475883, | |
| "loss": 2.5559, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004705125419329958, | |
| "loss": 2.5652, | |
| "step": 13010 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004704680881752997, | |
| "loss": 2.4955, | |
| "step": 13020 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047042360303799777, | |
| "loss": 2.5347, | |
| "step": 13030 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047037908652742145, | |
| "loss": 2.5307, | |
| "step": 13040 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047033453864990705, | |
| "loss": 2.4466, | |
| "step": 13050 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.000470289959411795, | |
| "loss": 2.4972, | |
| "step": 13060 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004702453488194305, | |
| "loss": 2.4343, | |
| "step": 13070 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047020070687916283, | |
| "loss": 2.5234, | |
| "step": 13080 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00047015603359734615, | |
| "loss": 2.5255, | |
| "step": 13090 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004701113289803388, | |
| "loss": 2.5133, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0004700665930345037, | |
| "loss": 2.5708, | |
| "step": 13110 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00047002182576620817, | |
| "loss": 2.5532, | |
| "step": 13120 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000469977027181824, | |
| "loss": 2.5189, | |
| "step": 13130 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004699321972877275, | |
| "loss": 2.5097, | |
| "step": 13140 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046988733609029936, | |
| "loss": 2.5624, | |
| "step": 13150 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046984244359592474, | |
| "loss": 2.5266, | |
| "step": 13160 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004697975198109933, | |
| "loss": 2.4798, | |
| "step": 13170 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004697525647418991, | |
| "loss": 2.5493, | |
| "step": 13180 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004697075783950406, | |
| "loss": 2.5079, | |
| "step": 13190 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004696625607768209, | |
| "loss": 2.4943, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046961751189364745, | |
| "loss": 2.5205, | |
| "step": 13210 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046957243175193197, | |
| "loss": 2.5775, | |
| "step": 13220 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000469527320358091, | |
| "loss": 2.4763, | |
| "step": 13230 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004694821777185452, | |
| "loss": 2.504, | |
| "step": 13240 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004694370038397198, | |
| "loss": 2.5577, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046939179872804447, | |
| "loss": 2.5632, | |
| "step": 13260 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046934656238995344, | |
| "loss": 2.4655, | |
| "step": 13270 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046930129483188513, | |
| "loss": 2.4743, | |
| "step": 13280 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046925599606028267, | |
| "loss": 2.5772, | |
| "step": 13290 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004692106660815935, | |
| "loss": 2.5073, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004691653049022694, | |
| "loss": 2.5086, | |
| "step": 13310 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004691199125287668, | |
| "loss": 2.5596, | |
| "step": 13320 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046907448896754645, | |
| "loss": 2.4635, | |
| "step": 13330 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004690290342250736, | |
| "loss": 2.4651, | |
| "step": 13340 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046898354830781784, | |
| "loss": 2.4664, | |
| "step": 13350 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004689380312222533, | |
| "loss": 2.5494, | |
| "step": 13360 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046889248297485856, | |
| "loss": 2.4976, | |
| "step": 13370 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046884690357211646, | |
| "loss": 2.5401, | |
| "step": 13380 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046880129302051456, | |
| "loss": 2.5556, | |
| "step": 13390 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004687556513265445, | |
| "loss": 2.5154, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004687099784967026, | |
| "loss": 2.5211, | |
| "step": 13410 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004686642745374896, | |
| "loss": 2.4928, | |
| "step": 13420 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046861853945541064, | |
| "loss": 2.534, | |
| "step": 13430 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004685727732569752, | |
| "loss": 2.5647, | |
| "step": 13440 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046852697594869743, | |
| "loss": 2.4762, | |
| "step": 13450 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046848114753709547, | |
| "loss": 2.4843, | |
| "step": 13460 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004684352880286923, | |
| "loss": 2.5751, | |
| "step": 13470 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004683893974300153, | |
| "loss": 2.547, | |
| "step": 13480 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046834347574759597, | |
| "loss": 2.5196, | |
| "step": 13490 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0004682975229879705, | |
| "loss": 2.5452, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046825153915767937, | |
| "loss": 2.5336, | |
| "step": 13510 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046820552426326764, | |
| "loss": 2.4998, | |
| "step": 13520 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00046815947831128457, | |
| "loss": 2.5035, | |
| "step": 13530 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046811340130828417, | |
| "loss": 2.5393, | |
| "step": 13540 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046806729326082444, | |
| "loss": 2.4767, | |
| "step": 13550 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046802115417546807, | |
| "loss": 2.5106, | |
| "step": 13560 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004679749840587821, | |
| "loss": 2.5557, | |
| "step": 13570 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004679287829173382, | |
| "loss": 2.5133, | |
| "step": 13580 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046788255075771204, | |
| "loss": 2.5481, | |
| "step": 13590 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046783628758648394, | |
| "loss": 2.513, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004677899934102387, | |
| "loss": 2.4759, | |
| "step": 13610 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004677436682355654, | |
| "loss": 2.5025, | |
| "step": 13620 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046769731206905765, | |
| "loss": 2.4778, | |
| "step": 13630 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004676509249173133, | |
| "loss": 2.4753, | |
| "step": 13640 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004676045067869348, | |
| "loss": 2.4744, | |
| "step": 13650 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046755805768452884, | |
| "loss": 2.5376, | |
| "step": 13660 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046751157761670673, | |
| "loss": 2.4891, | |
| "step": 13670 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004674650665900839, | |
| "loss": 2.519, | |
| "step": 13680 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046741852461128046, | |
| "loss": 2.4974, | |
| "step": 13690 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004673719516869208, | |
| "loss": 2.4557, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046732534782363365, | |
| "loss": 2.4654, | |
| "step": 13710 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046727871302805235, | |
| "loss": 2.4869, | |
| "step": 13720 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004672320473068144, | |
| "loss": 2.4943, | |
| "step": 13730 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004671853506665618, | |
| "loss": 2.5723, | |
| "step": 13740 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046713862311394106, | |
| "loss": 2.4123, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004670918646556029, | |
| "loss": 2.5217, | |
| "step": 13760 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004670450752982026, | |
| "loss": 2.4512, | |
| "step": 13770 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046699825504839976, | |
| "loss": 2.4829, | |
| "step": 13780 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004669514039128584, | |
| "loss": 2.5121, | |
| "step": 13790 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004669045218982468, | |
| "loss": 2.5034, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004668576090112379, | |
| "loss": 2.4972, | |
| "step": 13810 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004668106652585089, | |
| "loss": 2.4663, | |
| "step": 13820 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046676369064674127, | |
| "loss": 2.4418, | |
| "step": 13830 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046671668518262114, | |
| "loss": 2.4717, | |
| "step": 13840 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046666964887283867, | |
| "loss": 2.4793, | |
| "step": 13850 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004666225817240888, | |
| "loss": 2.502, | |
| "step": 13860 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004665754837430706, | |
| "loss": 2.4623, | |
| "step": 13870 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046652835493648766, | |
| "loss": 2.5331, | |
| "step": 13880 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004664811953110478, | |
| "loss": 2.505, | |
| "step": 13890 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046643400487346356, | |
| "loss": 2.5111, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046638678363045133, | |
| "loss": 2.5439, | |
| "step": 13910 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004663395315887324, | |
| "loss": 2.5534, | |
| "step": 13920 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0004662922487550322, | |
| "loss": 2.4713, | |
| "step": 13930 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00046624493513608054, | |
| "loss": 2.558, | |
| "step": 13940 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046619759073861163, | |
| "loss": 2.4227, | |
| "step": 13950 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004661502155693641, | |
| "loss": 2.4859, | |
| "step": 13960 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000466102809635081, | |
| "loss": 2.5517, | |
| "step": 13970 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046605537294250966, | |
| "loss": 2.5445, | |
| "step": 13980 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004660079054984018, | |
| "loss": 2.4981, | |
| "step": 13990 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004659604073095136, | |
| "loss": 2.561, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004659128783826055, | |
| "loss": 2.5353, | |
| "step": 14010 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004658653187244424, | |
| "loss": 2.4962, | |
| "step": 14020 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046581772834179355, | |
| "loss": 2.5608, | |
| "step": 14030 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046577010724143254, | |
| "loss": 2.533, | |
| "step": 14040 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004657224554301374, | |
| "loss": 2.5174, | |
| "step": 14050 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004656747729146905, | |
| "loss": 2.4456, | |
| "step": 14060 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046562705970187857, | |
| "loss": 2.5037, | |
| "step": 14070 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046557931579849275, | |
| "loss": 2.5486, | |
| "step": 14080 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046553154121132845, | |
| "loss": 2.5336, | |
| "step": 14090 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004654837359471855, | |
| "loss": 2.4941, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046543590001286815, | |
| "loss": 2.494, | |
| "step": 14110 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004653880334151849, | |
| "loss": 2.4672, | |
| "step": 14120 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046534013616094885, | |
| "loss": 2.575, | |
| "step": 14130 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046529220825697716, | |
| "loss": 2.5397, | |
| "step": 14140 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004652442497100915, | |
| "loss": 2.5255, | |
| "step": 14150 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046519626052711797, | |
| "loss": 2.486, | |
| "step": 14160 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004651482407148868, | |
| "loss": 2.4949, | |
| "step": 14170 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000465100190280233, | |
| "loss": 2.4742, | |
| "step": 14180 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046505210922999536, | |
| "loss": 2.4961, | |
| "step": 14190 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004650039975710175, | |
| "loss": 2.4995, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004649558553101473, | |
| "loss": 2.5423, | |
| "step": 14210 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046490768245423687, | |
| "loss": 2.4906, | |
| "step": 14220 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004648594790101426, | |
| "loss": 2.466, | |
| "step": 14230 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004648112449847256, | |
| "loss": 2.4881, | |
| "step": 14240 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000464762980384851, | |
| "loss": 2.508, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004647146852173883, | |
| "loss": 2.4449, | |
| "step": 14260 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046466635948921154, | |
| "loss": 2.4947, | |
| "step": 14270 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000464618003207199, | |
| "loss": 2.4884, | |
| "step": 14280 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004645696163782333, | |
| "loss": 2.5429, | |
| "step": 14290 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046452119900920145, | |
| "loss": 2.4876, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004644727511069946, | |
| "loss": 2.5007, | |
| "step": 14310 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004644242726785087, | |
| "loss": 2.4788, | |
| "step": 14320 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004643757637306436, | |
| "loss": 2.48, | |
| "step": 14330 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046432722427030366, | |
| "loss": 2.4772, | |
| "step": 14340 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00046427865430439765, | |
| "loss": 2.5244, | |
| "step": 14350 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004642300538398386, | |
| "loss": 2.4827, | |
| "step": 14360 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046418142288354386, | |
| "loss": 2.4751, | |
| "step": 14370 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004641327614424352, | |
| "loss": 2.5109, | |
| "step": 14380 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046408406952343873, | |
| "loss": 2.5147, | |
| "step": 14390 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046403534713348473, | |
| "loss": 2.4317, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004639865942795081, | |
| "loss": 2.524, | |
| "step": 14410 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046393781096844766, | |
| "loss": 2.4863, | |
| "step": 14420 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046388899720724714, | |
| "loss": 2.484, | |
| "step": 14430 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046384015300285407, | |
| "loss": 2.5052, | |
| "step": 14440 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046379127836222066, | |
| "loss": 2.4663, | |
| "step": 14450 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046374237329230327, | |
| "loss": 2.5435, | |
| "step": 14460 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046369343780006264, | |
| "loss": 2.5309, | |
| "step": 14470 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046364447189246383, | |
| "loss": 2.5446, | |
| "step": 14480 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046359547557647617, | |
| "loss": 2.4443, | |
| "step": 14490 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004635464488590736, | |
| "loss": 2.4253, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.000463497391747234, | |
| "loss": 2.4991, | |
| "step": 14510 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004634483042479398, | |
| "loss": 2.5398, | |
| "step": 14520 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046339918636817775, | |
| "loss": 2.5436, | |
| "step": 14530 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004633500381149388, | |
| "loss": 2.5029, | |
| "step": 14540 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046330085949521836, | |
| "loss": 2.511, | |
| "step": 14550 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046325165051601614, | |
| "loss": 2.5527, | |
| "step": 14560 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046320241118433616, | |
| "loss": 2.4902, | |
| "step": 14570 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046315314150718656, | |
| "loss": 2.4793, | |
| "step": 14580 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004631038414915801, | |
| "loss": 2.4142, | |
| "step": 14590 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046305451114453383, | |
| "loss": 2.5277, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004630051504730689, | |
| "loss": 2.4951, | |
| "step": 14610 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046295575948421096, | |
| "loss": 2.5458, | |
| "step": 14620 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046290633818498985, | |
| "loss": 2.5105, | |
| "step": 14630 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046285688658243983, | |
| "loss": 2.5751, | |
| "step": 14640 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046280740468359943, | |
| "loss": 2.4752, | |
| "step": 14650 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046275789249551147, | |
| "loss": 2.5784, | |
| "step": 14660 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004627083500252232, | |
| "loss": 2.4736, | |
| "step": 14670 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.000462658777279786, | |
| "loss": 2.5245, | |
| "step": 14680 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004626091742662556, | |
| "loss": 2.5082, | |
| "step": 14690 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004625595409916922, | |
| "loss": 2.5135, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046250987746316013, | |
| "loss": 2.4688, | |
| "step": 14710 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046246018368772816, | |
| "loss": 2.4956, | |
| "step": 14720 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004624104596724691, | |
| "loss": 2.5033, | |
| "step": 14730 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004623607054244604, | |
| "loss": 2.5287, | |
| "step": 14740 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00046231092095078363, | |
| "loss": 2.4656, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004622611062585248, | |
| "loss": 2.4735, | |
| "step": 14760 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.000462211261354774, | |
| "loss": 2.4303, | |
| "step": 14770 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0004621613862466257, | |
| "loss": 2.4452, | |
| "step": 14780 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046211148094117894, | |
| "loss": 2.4627, | |
| "step": 14790 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004620615454455366, | |
| "loss": 2.4603, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004620115797668062, | |
| "loss": 2.4903, | |
| "step": 14810 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004619615839120993, | |
| "loss": 2.5384, | |
| "step": 14820 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004619115578885321, | |
| "loss": 2.5015, | |
| "step": 14830 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046186150170322485, | |
| "loss": 2.4949, | |
| "step": 14840 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.000461811415363302, | |
| "loss": 2.475, | |
| "step": 14850 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046176129887589245, | |
| "loss": 2.5515, | |
| "step": 14860 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004617111522481295, | |
| "loss": 2.5006, | |
| "step": 14870 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046166097548715055, | |
| "loss": 2.4779, | |
| "step": 14880 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004616107686000972, | |
| "loss": 2.4585, | |
| "step": 14890 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004615605315941157, | |
| "loss": 2.4562, | |
| "step": 14900 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004615102644763563, | |
| "loss": 2.5066, | |
| "step": 14910 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046145996725397353, | |
| "loss": 2.5082, | |
| "step": 14920 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004614096399341264, | |
| "loss": 2.4652, | |
| "step": 14930 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046135928252397795, | |
| "loss": 2.4965, | |
| "step": 14940 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004613088950306957, | |
| "loss": 2.4779, | |
| "step": 14950 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046125847746145143, | |
| "loss": 2.4911, | |
| "step": 14960 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046120802982342114, | |
| "loss": 2.4493, | |
| "step": 14970 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004611575521237852, | |
| "loss": 2.4772, | |
| "step": 14980 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046110704436972793, | |
| "loss": 2.5292, | |
| "step": 14990 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046105650656843845, | |
| "loss": 2.5273, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046100593872710983, | |
| "loss": 2.5077, | |
| "step": 15010 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004609553408529394, | |
| "loss": 2.5803, | |
| "step": 15020 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004609047129531289, | |
| "loss": 2.4742, | |
| "step": 15030 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004608540550348843, | |
| "loss": 2.5024, | |
| "step": 15040 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004608033671054158, | |
| "loss": 2.509, | |
| "step": 15050 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004607526491719379, | |
| "loss": 2.4795, | |
| "step": 15060 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046070190124166945, | |
| "loss": 2.4776, | |
| "step": 15070 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004606511233218334, | |
| "loss": 2.4679, | |
| "step": 15080 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046060031541965706, | |
| "loss": 2.4802, | |
| "step": 15090 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004605494775423721, | |
| "loss": 2.5409, | |
| "step": 15100 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046049860969721426, | |
| "loss": 2.4953, | |
| "step": 15110 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046044771189142364, | |
| "loss": 2.5197, | |
| "step": 15120 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046039678413224466, | |
| "loss": 2.464, | |
| "step": 15130 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.000460345826426926, | |
| "loss": 2.5532, | |
| "step": 15140 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004602948387827205, | |
| "loss": 2.4582, | |
| "step": 15150 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046024382120688536, | |
| "loss": 2.4632, | |
| "step": 15160 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0004601927737066819, | |
| "loss": 2.5226, | |
| "step": 15170 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046014169628937594, | |
| "loss": 2.4251, | |
| "step": 15180 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00046009058896223735, | |
| "loss": 2.5279, | |
| "step": 15190 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004600394517325403, | |
| "loss": 2.4982, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045998828460756325, | |
| "loss": 2.5032, | |
| "step": 15210 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045993708759458897, | |
| "loss": 2.4597, | |
| "step": 15220 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004598858607009043, | |
| "loss": 2.4905, | |
| "step": 15230 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004598346039338006, | |
| "loss": 2.5193, | |
| "step": 15240 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004597833173005732, | |
| "loss": 2.525, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045973200080852195, | |
| "loss": 2.5088, | |
| "step": 15260 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004596806544649507, | |
| "loss": 2.511, | |
| "step": 15270 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045962927827716767, | |
| "loss": 2.5765, | |
| "step": 15280 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004595778722524854, | |
| "loss": 2.5656, | |
| "step": 15290 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045952643639822054, | |
| "loss": 2.5377, | |
| "step": 15300 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045948011863113975, | |
| "loss": 2.4865, | |
| "step": 15310 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045942862612084073, | |
| "loss": 2.4938, | |
| "step": 15320 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004593771038022016, | |
| "loss": 2.5417, | |
| "step": 15330 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004593255516825556, | |
| "loss": 2.4263, | |
| "step": 15340 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004592739697692403, | |
| "loss": 2.5083, | |
| "step": 15350 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004592223580695973, | |
| "loss": 2.4863, | |
| "step": 15360 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045917071659097277, | |
| "loss": 2.5124, | |
| "step": 15370 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004591190453407168, | |
| "loss": 2.4654, | |
| "step": 15380 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004590673443261839, | |
| "loss": 2.5327, | |
| "step": 15390 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045901561355473275, | |
| "loss": 2.5557, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004589638530337263, | |
| "loss": 2.488, | |
| "step": 15410 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004589120627705317, | |
| "loss": 2.4689, | |
| "step": 15420 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045886024277252025, | |
| "loss": 2.4934, | |
| "step": 15430 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045880839304706777, | |
| "loss": 2.506, | |
| "step": 15440 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.000458756513601554, | |
| "loss": 2.5255, | |
| "step": 15450 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045870460444336303, | |
| "loss": 2.4558, | |
| "step": 15460 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004586526655798833, | |
| "loss": 2.5462, | |
| "step": 15470 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004586006970185072, | |
| "loss": 2.4778, | |
| "step": 15480 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045854869876663155, | |
| "loss": 2.5155, | |
| "step": 15490 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045849667083165744, | |
| "loss": 2.4781, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045844461322098995, | |
| "loss": 2.4714, | |
| "step": 15510 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045839252594203874, | |
| "loss": 2.4063, | |
| "step": 15520 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045834040900221725, | |
| "loss": 2.4576, | |
| "step": 15530 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004582882624089435, | |
| "loss": 2.4979, | |
| "step": 15540 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045823608616963956, | |
| "loss": 2.4914, | |
| "step": 15550 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045818388029173184, | |
| "loss": 2.4669, | |
| "step": 15560 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004581316447826508, | |
| "loss": 2.5353, | |
| "step": 15570 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004580793796498313, | |
| "loss": 2.4637, | |
| "step": 15580 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00045802708490071223, | |
| "loss": 2.5162, | |
| "step": 15590 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004579747605427369, | |
| "loss": 2.5285, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0004579224065833526, | |
| "loss": 2.4937, | |
| "step": 15610 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045787002303001104, | |
| "loss": 2.5355, | |
| "step": 15620 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045781760989016806, | |
| "loss": 2.5003, | |
| "step": 15630 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045776516717128377, | |
| "loss": 2.5369, | |
| "step": 15640 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045771269488082235, | |
| "loss": 2.5065, | |
| "step": 15650 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004576601930262522, | |
| "loss": 2.4835, | |
| "step": 15660 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004576076616150462, | |
| "loss": 2.4485, | |
| "step": 15670 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045755510065468105, | |
| "loss": 2.5256, | |
| "step": 15680 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004575025101526379, | |
| "loss": 2.5252, | |
| "step": 15690 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045744989011640215, | |
| "loss": 2.486, | |
| "step": 15700 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004573972405534632, | |
| "loss": 2.5022, | |
| "step": 15710 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004573445614713147, | |
| "loss": 2.4672, | |
| "step": 15720 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045729185287745466, | |
| "loss": 2.4849, | |
| "step": 15730 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004572391147793852, | |
| "loss": 2.5141, | |
| "step": 15740 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004571863471846126, | |
| "loss": 2.4942, | |
| "step": 15750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004571335501006473, | |
| "loss": 2.4429, | |
| "step": 15760 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045708072353500395, | |
| "loss": 2.5178, | |
| "step": 15770 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045702786749520164, | |
| "loss": 2.5277, | |
| "step": 15780 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045697498198876344, | |
| "loss": 2.5142, | |
| "step": 15790 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045692206702321636, | |
| "loss": 2.5496, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045686912260609226, | |
| "loss": 2.4695, | |
| "step": 15810 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004568161487449265, | |
| "loss": 2.4697, | |
| "step": 15820 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045676314544725917, | |
| "loss": 2.5158, | |
| "step": 15830 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045671011272063413, | |
| "loss": 2.5271, | |
| "step": 15840 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004566570505725998, | |
| "loss": 2.4846, | |
| "step": 15850 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004566039590107084, | |
| "loss": 2.5556, | |
| "step": 15860 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045655083804251674, | |
| "loss": 2.5378, | |
| "step": 15870 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004564976876755855, | |
| "loss": 2.529, | |
| "step": 15880 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045644450791747975, | |
| "loss": 2.4782, | |
| "step": 15890 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004563912987757686, | |
| "loss": 2.4914, | |
| "step": 15900 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045633806025802544, | |
| "loss": 2.4727, | |
| "step": 15910 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045628479237182776, | |
| "loss": 2.5596, | |
| "step": 15920 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045623149512475725, | |
| "loss": 2.5346, | |
| "step": 15930 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045617816852439985, | |
| "loss": 2.3758, | |
| "step": 15940 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004561248125783457, | |
| "loss": 2.5176, | |
| "step": 15950 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004560714272941888, | |
| "loss": 2.4259, | |
| "step": 15960 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004560180126795278, | |
| "loss": 2.523, | |
| "step": 15970 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045596456874196525, | |
| "loss": 2.5393, | |
| "step": 15980 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004559110954891078, | |
| "loss": 2.4886, | |
| "step": 15990 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045585759292856654, | |
| "loss": 2.5128, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0004558040610679566, | |
| "loss": 2.4944, | |
| "step": 16010 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00045575049991489707, | |
| "loss": 2.5164, | |
| "step": 16020 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004556969094770116, | |
| "loss": 2.4839, | |
| "step": 16030 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045564328976192766, | |
| "loss": 2.5508, | |
| "step": 16040 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045558964077727716, | |
| "loss": 2.4978, | |
| "step": 16050 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045553596253069596, | |
| "loss": 2.4807, | |
| "step": 16060 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004554822550298242, | |
| "loss": 2.511, | |
| "step": 16070 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045542851828230625, | |
| "loss": 2.4616, | |
| "step": 16080 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004553747522957904, | |
| "loss": 2.5188, | |
| "step": 16090 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004553209570779294, | |
| "loss": 2.4695, | |
| "step": 16100 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004552671326363799, | |
| "loss": 2.4628, | |
| "step": 16110 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045521327897880294, | |
| "loss": 2.4024, | |
| "step": 16120 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004551593961128635, | |
| "loss": 2.4988, | |
| "step": 16130 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045510548404623086, | |
| "loss": 2.5268, | |
| "step": 16140 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004550515427865785, | |
| "loss": 2.5067, | |
| "step": 16150 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045499757234158386, | |
| "loss": 2.5406, | |
| "step": 16160 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004549435727189287, | |
| "loss": 2.5094, | |
| "step": 16170 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045488954392629885, | |
| "loss": 2.5511, | |
| "step": 16180 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004548354859713843, | |
| "loss": 2.5027, | |
| "step": 16190 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004547813988618793, | |
| "loss": 2.5352, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045472728260548213, | |
| "loss": 2.4966, | |
| "step": 16210 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004546731372098952, | |
| "loss": 2.5046, | |
| "step": 16220 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045461896268282517, | |
| "loss": 2.4708, | |
| "step": 16230 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045456475903198274, | |
| "loss": 2.5015, | |
| "step": 16240 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004545105262650829, | |
| "loss": 2.5042, | |
| "step": 16250 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004544562643898446, | |
| "loss": 2.5876, | |
| "step": 16260 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045440197341399107, | |
| "loss": 2.4939, | |
| "step": 16270 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045434765334524967, | |
| "loss": 2.5172, | |
| "step": 16280 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045429330419135183, | |
| "loss": 2.5796, | |
| "step": 16290 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045423892596003313, | |
| "loss": 2.5151, | |
| "step": 16300 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004541845186590334, | |
| "loss": 2.5382, | |
| "step": 16310 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045413008229609644, | |
| "loss": 2.4907, | |
| "step": 16320 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045407561687897034, | |
| "loss": 2.4898, | |
| "step": 16330 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004540211224154073, | |
| "loss": 2.5249, | |
| "step": 16340 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004539665989131634, | |
| "loss": 2.4957, | |
| "step": 16350 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004539120463799993, | |
| "loss": 2.5136, | |
| "step": 16360 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004538574648236794, | |
| "loss": 2.4318, | |
| "step": 16370 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004538028542519725, | |
| "loss": 2.4056, | |
| "step": 16380 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045374821467265137, | |
| "loss": 2.4772, | |
| "step": 16390 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004536935460934929, | |
| "loss": 2.4405, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045363884852227823, | |
| "loss": 2.4526, | |
| "step": 16410 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004535841219667926, | |
| "loss": 2.5394, | |
| "step": 16420 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00045352936643482523, | |
| "loss": 2.4858, | |
| "step": 16430 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0004534745819341697, | |
| "loss": 2.5338, | |
| "step": 16440 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045341976847262347, | |
| "loss": 2.5375, | |
| "step": 16450 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004533649260579882, | |
| "loss": 2.5013, | |
| "step": 16460 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045331005469806987, | |
| "loss": 2.4565, | |
| "step": 16470 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004532551544006783, | |
| "loss": 2.5322, | |
| "step": 16480 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045320022517362753, | |
| "loss": 2.5181, | |
| "step": 16490 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004531452670247358, | |
| "loss": 2.4925, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004530902799618254, | |
| "loss": 2.5083, | |
| "step": 16510 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045303526399272266, | |
| "loss": 2.5098, | |
| "step": 16520 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004529802191252581, | |
| "loss": 2.5494, | |
| "step": 16530 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004529251453672665, | |
| "loss": 2.4094, | |
| "step": 16540 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045287004272658645, | |
| "loss": 2.4951, | |
| "step": 16550 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004528149112110608, | |
| "loss": 2.4973, | |
| "step": 16560 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045275975082853667, | |
| "loss": 2.4767, | |
| "step": 16570 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.000452704561586865, | |
| "loss": 2.5196, | |
| "step": 16580 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.000452649343493901, | |
| "loss": 2.5252, | |
| "step": 16590 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045259409655750387, | |
| "loss": 2.4521, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004525388207855372, | |
| "loss": 2.4612, | |
| "step": 16610 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004524835161858683, | |
| "loss": 2.5224, | |
| "step": 16620 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004524281827663689, | |
| "loss": 2.4684, | |
| "step": 16630 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004523728205349146, | |
| "loss": 2.5479, | |
| "step": 16640 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045231742949938526, | |
| "loss": 2.5184, | |
| "step": 16650 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004522620096676648, | |
| "loss": 2.4493, | |
| "step": 16660 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045220656104764126, | |
| "loss": 2.5452, | |
| "step": 16670 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045215108364720654, | |
| "loss": 2.5043, | |
| "step": 16680 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.000452095577474257, | |
| "loss": 2.5281, | |
| "step": 16690 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045204004253669293, | |
| "loss": 2.5114, | |
| "step": 16700 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004519844788424187, | |
| "loss": 2.5105, | |
| "step": 16710 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004519288863993427, | |
| "loss": 2.5229, | |
| "step": 16720 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045187326521537764, | |
| "loss": 2.4916, | |
| "step": 16730 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004518176152984401, | |
| "loss": 2.4385, | |
| "step": 16740 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004517619366564508, | |
| "loss": 2.5153, | |
| "step": 16750 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004517062292973346, | |
| "loss": 2.487, | |
| "step": 16760 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004516504932290205, | |
| "loss": 2.4918, | |
| "step": 16770 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004515947284594414, | |
| "loss": 2.5716, | |
| "step": 16780 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045153893499653443, | |
| "loss": 2.4547, | |
| "step": 16790 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004514831128482408, | |
| "loss": 2.4764, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004514272620225058, | |
| "loss": 2.4639, | |
| "step": 16810 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004513713825272788, | |
| "loss": 2.497, | |
| "step": 16820 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.000451315474370513, | |
| "loss": 2.4834, | |
| "step": 16830 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00045125953756016616, | |
| "loss": 2.5361, | |
| "step": 16840 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004512035721041998, | |
| "loss": 2.4667, | |
| "step": 16850 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004511475780105795, | |
| "loss": 2.4997, | |
| "step": 16860 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045109155528727507, | |
| "loss": 2.5144, | |
| "step": 16870 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045103550394226023, | |
| "loss": 2.4371, | |
| "step": 16880 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.000450979423983513, | |
| "loss": 2.5045, | |
| "step": 16890 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045092331541901533, | |
| "loss": 2.5637, | |
| "step": 16900 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045086717825675303, | |
| "loss": 2.5454, | |
| "step": 16910 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004508110125047165, | |
| "loss": 2.4843, | |
| "step": 16920 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045075481817089967, | |
| "loss": 2.4821, | |
| "step": 16930 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045069859526330097, | |
| "loss": 2.5256, | |
| "step": 16940 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004506423437899226, | |
| "loss": 2.5039, | |
| "step": 16950 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004505860637587709, | |
| "loss": 2.5152, | |
| "step": 16960 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004505297551778563, | |
| "loss": 2.4165, | |
| "step": 16970 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004504734180551935, | |
| "loss": 2.465, | |
| "step": 16980 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004504170523988008, | |
| "loss": 2.487, | |
| "step": 16990 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004503606582167009, | |
| "loss": 2.4871, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004503042355169206, | |
| "loss": 2.4778, | |
| "step": 17010 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045024778430749045, | |
| "loss": 2.4842, | |
| "step": 17020 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004501913045964454, | |
| "loss": 2.3827, | |
| "step": 17030 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004501347963918242, | |
| "loss": 2.4935, | |
| "step": 17040 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00045007825970166986, | |
| "loss": 2.4481, | |
| "step": 17050 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004500216945340293, | |
| "loss": 2.4935, | |
| "step": 17060 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044996510089695353, | |
| "loss": 2.4603, | |
| "step": 17070 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004499084787984976, | |
| "loss": 2.4409, | |
| "step": 17080 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004498518282467208, | |
| "loss": 2.499, | |
| "step": 17090 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044979514924968603, | |
| "loss": 2.4527, | |
| "step": 17100 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044973844181546066, | |
| "loss": 2.558, | |
| "step": 17110 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044968170595211594, | |
| "loss": 2.5074, | |
| "step": 17120 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044962494166772717, | |
| "loss": 2.493, | |
| "step": 17130 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004495681489703738, | |
| "loss": 2.5091, | |
| "step": 17140 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004495113278681391, | |
| "loss": 2.4695, | |
| "step": 17150 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004494544783691105, | |
| "loss": 2.5108, | |
| "step": 17160 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044939760048137966, | |
| "loss": 2.4733, | |
| "step": 17170 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.000449340694213042, | |
| "loss": 2.5299, | |
| "step": 17180 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044928375957219705, | |
| "loss": 2.5455, | |
| "step": 17190 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044922679656694847, | |
| "loss": 2.4885, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044916980520540395, | |
| "loss": 2.528, | |
| "step": 17210 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044911278549567506, | |
| "loss": 2.4547, | |
| "step": 17220 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004490557374458776, | |
| "loss": 2.504, | |
| "step": 17230 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044899866106413126, | |
| "loss": 2.4681, | |
| "step": 17240 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0004489415563585599, | |
| "loss": 2.524, | |
| "step": 17250 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044888442333729125, | |
| "loss": 2.5064, | |
| "step": 17260 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00044882726200845714, | |
| "loss": 2.5344, | |
| "step": 17270 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004487700723801935, | |
| "loss": 2.4999, | |
| "step": 17280 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004487128544606403, | |
| "loss": 2.5362, | |
| "step": 17290 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044865560825794125, | |
| "loss": 2.4864, | |
| "step": 17300 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004485983337802446, | |
| "loss": 2.4981, | |
| "step": 17310 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.000448541031035702, | |
| "loss": 2.5049, | |
| "step": 17320 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044848370003246966, | |
| "loss": 2.5127, | |
| "step": 17330 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004484263407787076, | |
| "loss": 2.4714, | |
| "step": 17340 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044836895328257976, | |
| "loss": 2.4601, | |
| "step": 17350 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004483115375522542, | |
| "loss": 2.5166, | |
| "step": 17360 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004482540935959031, | |
| "loss": 2.5186, | |
| "step": 17370 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044819662142170255, | |
| "loss": 2.4691, | |
| "step": 17380 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044813912103783255, | |
| "loss": 2.4929, | |
| "step": 17390 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004480815924524774, | |
| "loss": 2.5017, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.000448024035673825, | |
| "loss": 2.496, | |
| "step": 17410 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044796645071006774, | |
| "loss": 2.4943, | |
| "step": 17420 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004479088375694017, | |
| "loss": 2.4794, | |
| "step": 17430 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044785119626002714, | |
| "loss": 2.4997, | |
| "step": 17440 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044779352679014806, | |
| "loss": 2.4312, | |
| "step": 17450 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004477358291679727, | |
| "loss": 2.4659, | |
| "step": 17460 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004476781034017134, | |
| "loss": 2.4723, | |
| "step": 17470 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004476203494995864, | |
| "loss": 2.5042, | |
| "step": 17480 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044756256746981174, | |
| "loss": 2.4907, | |
| "step": 17490 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044750475732061365, | |
| "loss": 2.4671, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004474527041510289, | |
| "loss": 2.5526, | |
| "step": 17510 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004473948405975985, | |
| "loss": 2.4511, | |
| "step": 17520 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004473369489486176, | |
| "loss": 2.4619, | |
| "step": 17530 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044727902921232594, | |
| "loss": 2.5007, | |
| "step": 17540 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044722108139696743, | |
| "loss": 2.4786, | |
| "step": 17550 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004471631055107898, | |
| "loss": 2.4501, | |
| "step": 17560 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044710510156204496, | |
| "loss": 2.545, | |
| "step": 17570 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004470470695589887, | |
| "loss": 2.5224, | |
| "step": 17580 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044698900950988063, | |
| "loss": 2.4788, | |
| "step": 17590 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044693092142298485, | |
| "loss": 2.5411, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004468728053065689, | |
| "loss": 2.4314, | |
| "step": 17610 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004468146611689046, | |
| "loss": 2.5196, | |
| "step": 17620 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044675648901826775, | |
| "loss": 2.4928, | |
| "step": 17630 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044669828886293806, | |
| "loss": 2.5184, | |
| "step": 17640 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044664006071119935, | |
| "loss": 2.4901, | |
| "step": 17650 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004465818045713391, | |
| "loss": 2.5385, | |
| "step": 17660 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004465235204516493, | |
| "loss": 2.4778, | |
| "step": 17670 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00044646520836042543, | |
| "loss": 2.5684, | |
| "step": 17680 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0004464068683059672, | |
| "loss": 2.5133, | |
| "step": 17690 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004463485002965783, | |
| "loss": 2.4792, | |
| "step": 17700 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044629010434056625, | |
| "loss": 2.4996, | |
| "step": 17710 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004462316804462428, | |
| "loss": 2.5444, | |
| "step": 17720 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004461732286219234, | |
| "loss": 2.5438, | |
| "step": 17730 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044611474887592764, | |
| "loss": 2.525, | |
| "step": 17740 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044605624121657906, | |
| "loss": 2.4898, | |
| "step": 17750 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004459977056522051, | |
| "loss": 2.5913, | |
| "step": 17760 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044593914219113724, | |
| "loss": 2.4773, | |
| "step": 17770 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044588055084171096, | |
| "loss": 2.5185, | |
| "step": 17780 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004458219316122657, | |
| "loss": 2.5672, | |
| "step": 17790 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004457632845111448, | |
| "loss": 2.4117, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004457046095466955, | |
| "loss": 2.504, | |
| "step": 17810 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004456459067272692, | |
| "loss": 2.4892, | |
| "step": 17820 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044558717606122127, | |
| "loss": 2.4976, | |
| "step": 17830 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044552841755691086, | |
| "loss": 2.5147, | |
| "step": 17840 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004454696312227011, | |
| "loss": 2.4472, | |
| "step": 17850 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044541081706695926, | |
| "loss": 2.4883, | |
| "step": 17860 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044535197509805636, | |
| "loss": 2.5339, | |
| "step": 17870 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004452931053243676, | |
| "loss": 2.4652, | |
| "step": 17880 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044523420775427193, | |
| "loss": 2.5356, | |
| "step": 17890 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004451752823961523, | |
| "loss": 2.4901, | |
| "step": 17900 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004451163292583958, | |
| "loss": 2.5118, | |
| "step": 17910 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004450573483493933, | |
| "loss": 2.5217, | |
| "step": 17920 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004449983396775395, | |
| "loss": 2.4693, | |
| "step": 17930 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004449393032512333, | |
| "loss": 2.4984, | |
| "step": 17940 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044488023907887766, | |
| "loss": 2.4952, | |
| "step": 17950 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000444821147168879, | |
| "loss": 2.5061, | |
| "step": 17960 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000444762027529648, | |
| "loss": 2.4813, | |
| "step": 17970 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004447028801695995, | |
| "loss": 2.4999, | |
| "step": 17980 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004446437050971517, | |
| "loss": 2.4939, | |
| "step": 17990 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004445845023207274, | |
| "loss": 2.4784, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004445252718487529, | |
| "loss": 2.5115, | |
| "step": 18010 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044446601368965853, | |
| "loss": 2.5569, | |
| "step": 18020 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044440672785187876, | |
| "loss": 2.5447, | |
| "step": 18030 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004443474143438516, | |
| "loss": 2.4548, | |
| "step": 18040 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004442880731740195, | |
| "loss": 2.5214, | |
| "step": 18050 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004442287043508285, | |
| "loss": 2.4986, | |
| "step": 18060 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044416930788272865, | |
| "loss": 2.5004, | |
| "step": 18070 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004441098837781739, | |
| "loss": 2.4914, | |
| "step": 18080 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00044405043204562224, | |
| "loss": 2.508, | |
| "step": 18090 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004439909526935356, | |
| "loss": 2.4651, | |
| "step": 18100 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0004439314457303797, | |
| "loss": 2.5433, | |
| "step": 18110 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004438719111646242, | |
| "loss": 2.5142, | |
| "step": 18120 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000443812349004743, | |
| "loss": 2.4248, | |
| "step": 18130 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044375275925921345, | |
| "loss": 2.5325, | |
| "step": 18140 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004436931419365171, | |
| "loss": 2.5463, | |
| "step": 18150 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004436334970451395, | |
| "loss": 2.4654, | |
| "step": 18160 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044357382459356986, | |
| "loss": 2.4433, | |
| "step": 18170 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044351412459030167, | |
| "loss": 2.4985, | |
| "step": 18180 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004434543970438319, | |
| "loss": 2.5186, | |
| "step": 18190 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004433946419626619, | |
| "loss": 2.4886, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004433348593552965, | |
| "loss": 2.4507, | |
| "step": 18210 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044327504923024475, | |
| "loss": 2.4182, | |
| "step": 18220 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044321521159601963, | |
| "loss": 2.5312, | |
| "step": 18230 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044315534646113776, | |
| "loss": 2.4319, | |
| "step": 18240 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044309545383412, | |
| "loss": 2.4488, | |
| "step": 18250 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004430355337234909, | |
| "loss": 2.5598, | |
| "step": 18260 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000442975586137779, | |
| "loss": 2.541, | |
| "step": 18270 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004429156110855167, | |
| "loss": 2.5041, | |
| "step": 18280 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004428556085752404, | |
| "loss": 2.596, | |
| "step": 18290 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004427955786154905, | |
| "loss": 2.4432, | |
| "step": 18300 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004427355212148109, | |
| "loss": 2.5528, | |
| "step": 18310 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044267543638174976, | |
| "loss": 2.5592, | |
| "step": 18320 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044261532412485916, | |
| "loss": 2.5664, | |
| "step": 18330 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044255518445269495, | |
| "loss": 2.4974, | |
| "step": 18340 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044249501737381685, | |
| "loss": 2.5144, | |
| "step": 18350 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004424348228967886, | |
| "loss": 2.4311, | |
| "step": 18360 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044237460103017777, | |
| "loss": 2.4588, | |
| "step": 18370 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044231435178255585, | |
| "loss": 2.49, | |
| "step": 18380 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004422540751624982, | |
| "loss": 2.4677, | |
| "step": 18390 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044219377117858406, | |
| "loss": 2.4693, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044213343983939667, | |
| "loss": 2.5343, | |
| "step": 18410 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044207308115352305, | |
| "loss": 2.4375, | |
| "step": 18420 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004420126951295542, | |
| "loss": 2.4837, | |
| "step": 18430 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000441952281776085, | |
| "loss": 2.4179, | |
| "step": 18440 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004418918411017141, | |
| "loss": 2.446, | |
| "step": 18450 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004418313731150441, | |
| "loss": 2.4394, | |
| "step": 18460 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004417708778246816, | |
| "loss": 2.5379, | |
| "step": 18470 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044171035523923695, | |
| "loss": 2.5142, | |
| "step": 18480 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00044164980536732455, | |
| "loss": 2.5121, | |
| "step": 18490 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004415892282175624, | |
| "loss": 2.4249, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004415286237985726, | |
| "loss": 2.5187, | |
| "step": 18510 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0004414679921189811, | |
| "loss": 2.4819, | |
| "step": 18520 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004414073331874178, | |
| "loss": 2.4917, | |
| "step": 18530 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004413466470125163, | |
| "loss": 2.5667, | |
| "step": 18540 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044128593360291413, | |
| "loss": 2.4779, | |
| "step": 18550 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004412251929672528, | |
| "loss": 2.4459, | |
| "step": 18560 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044116442511417766, | |
| "loss": 2.4911, | |
| "step": 18570 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044110363005233776, | |
| "loss": 2.4189, | |
| "step": 18580 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004410428077903864, | |
| "loss": 2.4442, | |
| "step": 18590 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004409819583369803, | |
| "loss": 2.5281, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044092108170078035, | |
| "loss": 2.4975, | |
| "step": 18610 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044086017789045123, | |
| "loss": 2.4473, | |
| "step": 18620 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004407992469146615, | |
| "loss": 2.5012, | |
| "step": 18630 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044073828878208357, | |
| "loss": 2.4588, | |
| "step": 18640 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044067730350139366, | |
| "loss": 2.5068, | |
| "step": 18650 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000440616291081272, | |
| "loss": 2.4834, | |
| "step": 18660 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044055525153040254, | |
| "loss": 2.4716, | |
| "step": 18670 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004404941848574731, | |
| "loss": 2.419, | |
| "step": 18680 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004404330910711755, | |
| "loss": 2.5014, | |
| "step": 18690 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004403719701802053, | |
| "loss": 2.4721, | |
| "step": 18700 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004403108221932619, | |
| "loss": 2.4931, | |
| "step": 18710 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044024964711904865, | |
| "loss": 2.5473, | |
| "step": 18720 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044018844496627264, | |
| "loss": 2.4623, | |
| "step": 18730 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044012721574364496, | |
| "loss": 2.4448, | |
| "step": 18740 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00044006595945988037, | |
| "loss": 2.5178, | |
| "step": 18750 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004400046761236977, | |
| "loss": 2.5232, | |
| "step": 18760 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043994336574381944, | |
| "loss": 2.5255, | |
| "step": 18770 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043988202832897204, | |
| "loss": 2.5662, | |
| "step": 18780 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004398206638878857, | |
| "loss": 2.4733, | |
| "step": 18790 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004397592724292946, | |
| "loss": 2.4792, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043969785396193666, | |
| "loss": 2.5084, | |
| "step": 18810 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043963640849455364, | |
| "loss": 2.553, | |
| "step": 18820 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004395749360358913, | |
| "loss": 2.5486, | |
| "step": 18830 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000439513436594699, | |
| "loss": 2.4751, | |
| "step": 18840 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004394519101797301, | |
| "loss": 2.4746, | |
| "step": 18850 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004393903567997417, | |
| "loss": 2.5157, | |
| "step": 18860 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043932877646349496, | |
| "loss": 2.4971, | |
| "step": 18870 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004392671691797546, | |
| "loss": 2.5431, | |
| "step": 18880 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004392055349572892, | |
| "loss": 2.5813, | |
| "step": 18890 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0004391438738048715, | |
| "loss": 2.5033, | |
| "step": 18900 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043908218573127767, | |
| "loss": 2.5365, | |
| "step": 18910 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043902047074528795, | |
| "loss": 2.4737, | |
| "step": 18920 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043895872885568634, | |
| "loss": 2.5728, | |
| "step": 18930 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00043889696007126064, | |
| "loss": 2.4635, | |
| "step": 18940 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043883516440080243, | |
| "loss": 2.4192, | |
| "step": 18950 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043877334185310734, | |
| "loss": 2.5424, | |
| "step": 18960 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043871149243697466, | |
| "loss": 2.5656, | |
| "step": 18970 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043864961616120745, | |
| "loss": 2.4394, | |
| "step": 18980 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004385877130346127, | |
| "loss": 2.476, | |
| "step": 18990 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004385257830660012, | |
| "loss": 2.5114, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004384638262641876, | |
| "loss": 2.521, | |
| "step": 19010 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004384018426379902, | |
| "loss": 2.5139, | |
| "step": 19020 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004383398321962313, | |
| "loss": 2.5511, | |
| "step": 19030 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.000438277794947737, | |
| "loss": 2.4438, | |
| "step": 19040 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043821573090133706, | |
| "loss": 2.4152, | |
| "step": 19050 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004381536400658652, | |
| "loss": 2.5017, | |
| "step": 19060 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.000438091522450159, | |
| "loss": 2.5394, | |
| "step": 19070 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043802937806305974, | |
| "loss": 2.5255, | |
| "step": 19080 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004379672069134125, | |
| "loss": 2.4649, | |
| "step": 19090 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043790500901006623, | |
| "loss": 2.4643, | |
| "step": 19100 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043784278436187366, | |
| "loss": 2.4827, | |
| "step": 19110 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043778053297769125, | |
| "loss": 2.5556, | |
| "step": 19120 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004377182548663795, | |
| "loss": 2.5277, | |
| "step": 19130 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004376559500368025, | |
| "loss": 2.4746, | |
| "step": 19140 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004375936184978281, | |
| "loss": 2.5033, | |
| "step": 19150 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043753126025832824, | |
| "loss": 2.4258, | |
| "step": 19160 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043746887532717825, | |
| "loss": 2.4715, | |
| "step": 19170 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004374064637132578, | |
| "loss": 2.5185, | |
| "step": 19180 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004373440254254497, | |
| "loss": 2.4621, | |
| "step": 19190 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004372815604726411, | |
| "loss": 2.5059, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043721906886372275, | |
| "loss": 2.5585, | |
| "step": 19210 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004371565506075892, | |
| "loss": 2.5015, | |
| "step": 19220 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043709400571313863, | |
| "loss": 2.4351, | |
| "step": 19230 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043703143418927335, | |
| "loss": 2.5118, | |
| "step": 19240 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043696883604489914, | |
| "loss": 2.4694, | |
| "step": 19250 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043690621128892573, | |
| "loss": 2.4145, | |
| "step": 19260 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004368435599302666, | |
| "loss": 2.4887, | |
| "step": 19270 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043678088197783907, | |
| "loss": 2.5303, | |
| "step": 19280 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004367181774405642, | |
| "loss": 2.5137, | |
| "step": 19290 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043665544632736687, | |
| "loss": 2.4567, | |
| "step": 19300 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004365926886471756, | |
| "loss": 2.4359, | |
| "step": 19310 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004365299044089229, | |
| "loss": 2.4731, | |
| "step": 19320 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043646709362154493, | |
| "loss": 2.509, | |
| "step": 19330 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00043640425629398163, | |
| "loss": 2.5066, | |
| "step": 19340 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004363413924351768, | |
| "loss": 2.5281, | |
| "step": 19350 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0004362785020540778, | |
| "loss": 2.5036, | |
| "step": 19360 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004362155851596362, | |
| "loss": 2.4796, | |
| "step": 19370 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004361526417608068, | |
| "loss": 2.4491, | |
| "step": 19380 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004360896718665487, | |
| "loss": 2.5537, | |
| "step": 19390 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004360266754858243, | |
| "loss": 2.4436, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043596365262760006, | |
| "loss": 2.5069, | |
| "step": 19410 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004359006033008462, | |
| "loss": 2.4464, | |
| "step": 19420 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004358375275145365, | |
| "loss": 2.462, | |
| "step": 19430 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004357744252776488, | |
| "loss": 2.5064, | |
| "step": 19440 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004357112965991645, | |
| "loss": 2.4771, | |
| "step": 19450 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004356481414880687, | |
| "loss": 2.5665, | |
| "step": 19460 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004355849599533505, | |
| "loss": 2.4984, | |
| "step": 19470 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004355217520040026, | |
| "loss": 2.4885, | |
| "step": 19480 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004354585176490216, | |
| "loss": 2.5118, | |
| "step": 19490 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004353952568974076, | |
| "loss": 2.5722, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043533196975816457, | |
| "loss": 2.4677, | |
| "step": 19510 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004352686562403005, | |
| "loss": 2.5287, | |
| "step": 19520 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004352053163528267, | |
| "loss": 2.4705, | |
| "step": 19530 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043514195010475865, | |
| "loss": 2.5063, | |
| "step": 19540 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043507855750511514, | |
| "loss": 2.4616, | |
| "step": 19550 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004350151385629192, | |
| "loss": 2.4643, | |
| "step": 19560 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043495169328719717, | |
| "loss": 2.5257, | |
| "step": 19570 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004348882216869794, | |
| "loss": 2.5102, | |
| "step": 19580 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004348247237712999, | |
| "loss": 2.5018, | |
| "step": 19590 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004347611995491965, | |
| "loss": 2.4501, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043469764902971057, | |
| "loss": 2.5578, | |
| "step": 19610 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004346340722218874, | |
| "loss": 2.4842, | |
| "step": 19620 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043457046913477623, | |
| "loss": 2.5539, | |
| "step": 19630 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043450683977742943, | |
| "loss": 2.4859, | |
| "step": 19640 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043444318415890374, | |
| "loss": 2.4602, | |
| "step": 19650 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004343795022882593, | |
| "loss": 2.5343, | |
| "step": 19660 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043431579417456, | |
| "loss": 2.5148, | |
| "step": 19670 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043425205982687364, | |
| "loss": 2.4889, | |
| "step": 19680 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004341882992542715, | |
| "loss": 2.5061, | |
| "step": 19690 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043412451246582887, | |
| "loss": 2.44, | |
| "step": 19700 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043406069947062454, | |
| "loss": 2.5135, | |
| "step": 19710 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043399686027774116, | |
| "loss": 2.496, | |
| "step": 19720 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004339329948962651, | |
| "loss": 2.5391, | |
| "step": 19730 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004338691033352863, | |
| "loss": 2.5732, | |
| "step": 19740 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004338051856038988, | |
| "loss": 2.4996, | |
| "step": 19750 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0004337412417111999, | |
| "loss": 2.5548, | |
| "step": 19760 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00043367727166629096, | |
| "loss": 2.5098, | |
| "step": 19770 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043361327547827686, | |
| "loss": 2.5115, | |
| "step": 19780 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043354925315626644, | |
| "loss": 2.5306, | |
| "step": 19790 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0004334852047093719, | |
| "loss": 2.5407, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043342113014670957, | |
| "loss": 2.4587, | |
| "step": 19810 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043335702947739914, | |
| "loss": 2.4496, | |
| "step": 19820 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043329290271056423, | |
| "loss": 2.4215, | |
| "step": 19830 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0004332287498553321, | |
| "loss": 2.4787, | |
| "step": 19840 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0004331645709208338, | |
| "loss": 2.5035, | |
| "step": 19850 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.000433100365916204, | |
| "loss": 2.4389, | |
| "step": 19860 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0004330361348505811, | |
| "loss": 2.5536, | |
| "step": 19870 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043297187773310716, | |
| "loss": 2.5321, | |
| "step": 19880 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043290759457292806, | |
| "loss": 2.5207, | |
| "step": 19890 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043284328537919334, | |
| "loss": 2.4326, | |
| "step": 19900 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043277895016105615, | |
| "loss": 2.4312, | |
| "step": 19910 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0004327145889276736, | |
| "loss": 2.5208, | |
| "step": 19920 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043265020168820626, | |
| "loss": 2.4183, | |
| "step": 19930 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043258578845181846, | |
| "loss": 2.4705, | |
| "step": 19940 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0004325213492276783, | |
| "loss": 2.4667, | |
| "step": 19950 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043245688402495736, | |
| "loss": 2.5512, | |
| "step": 19960 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0004323923928528313, | |
| "loss": 2.4491, | |
| "step": 19970 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043232787572047913, | |
| "loss": 2.4414, | |
| "step": 19980 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043226333263708374, | |
| "loss": 2.5323, | |
| "step": 19990 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00043219876361183166, | |
| "loss": 2.485, | |
| "step": 20000 | |
| } | |
| ], | |
| "max_steps": 83272, | |
| "num_train_epochs": 2, | |
| "total_flos": 1.6500878227229639e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |