diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,135424 @@
+{
+  "best_metric": 0.567031216286806,
+  "best_model_checkpoint": "../checkpoints/pretrain_clean/sandy-star-569_exp7_custom_rassp_neims/checkpoint-224000",
+  "epoch": 0.9999553591357528,
+  "global_step": 224000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 5.6312,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 4.6821,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-06,
+      "loss": 3.6321,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 3.023,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-06,
+      "loss": 2.6799,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-06,
+      "loss": 2.4379,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5000000000000004e-06,
+      "loss": 2.1893,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.9779,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5e-06,
+      "loss": 1.8255,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-06,
+      "loss": 1.748,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 1.6764,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-06,
+      "loss": 1.6328,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 1.598,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 1.5572,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.5e-06,
+      "loss": 1.5327,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.5113,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.500000000000002e-06,
+      "loss": 1.4871,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-06,
+      "loss": 1.4737,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.5e-06,
+      "loss": 1.4602,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 1.4283,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.05e-05,
+      "loss": 1.4138,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 1.398,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1500000000000002e-05,
+      "loss": 1.3849,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-05,
+      "loss": 1.3664,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.25e-05,
+      "loss": 1.3479,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 1.3367,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 1.3338,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 1.3033,
+      "step": 280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.45e-05,
+      "loss": 1.2978,
+      "step": 290
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-05,
+      "loss": 1.2837,
+      "step": 300
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.55e-05,
+      "loss": 1.2658,
+      "step": 310
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 1.259,
+      "step": 320
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.65e-05,
+      "loss": 1.2587,
+      "step": 330
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 1.2363,
+      "step": 340
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.75e-05,
+      "loss": 1.2299,
+      "step": 350
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-05,
+      "loss": 1.2127,
+      "step": 360
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.85e-05,
+      "loss": 1.1884,
+      "step": 370
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9e-05,
+      "loss": 1.2035,
+      "step": 380
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9500000000000003e-05,
+      "loss": 1.1833,
+      "step": 390
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 1.1625,
+      "step": 400
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.05e-05,
+      "loss": 1.1529,
+      "step": 410
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1e-05,
+      "loss": 1.152,
+      "step": 420
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.15e-05,
+      "loss": 1.1415,
+      "step": 430
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 1.1204,
+      "step": 440
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.25e-05,
+      "loss": 1.1233,
+      "step": 450
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 1.1146,
+      "step": 460
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.35e-05,
+      "loss": 1.1,
+      "step": 470
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4e-05,
+      "loss": 1.0915,
+      "step": 480
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.45e-05,
+      "loss": 1.0861,
+      "step": 490
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-05,
+      "loss": 1.0895,
+      "step": 500
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5500000000000003e-05,
+      "loss": 1.0562,
+      "step": 510
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 1.0528,
+      "step": 520
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6500000000000004e-05,
+      "loss": 1.0552,
+      "step": 530
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 1.0509,
+      "step": 540
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7500000000000004e-05,
+      "loss": 1.0406,
+      "step": 550
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 1.0405,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 1.0507,
+      "step": 570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9e-05,
+      "loss": 1.0223,
+      "step": 580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.95e-05,
+      "loss": 1.0203,
+      "step": 590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-05,
+      "loss": 1.022,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.05e-05,
+      "loss": 1.0133,
+      "step": 610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1e-05,
+      "loss": 1.011,
+      "step": 620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.15e-05,
+      "loss": 0.997,
+      "step": 630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.9955,
+      "step": 640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 1.0008,
+      "step": 650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3e-05,
+      "loss": 0.9762,
+      "step": 660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.35e-05,
+      "loss": 0.9747,
+      "step": 670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.9655,
+      "step": 680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.45e-05,
+      "loss": 0.951,
+      "step": 690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5e-05,
+      "loss": 0.9574,
+      "step": 700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.55e-05,
+      "loss": 0.958,
+      "step": 710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6e-05,
+      "loss": 0.9536,
+      "step": 720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.65e-05,
+      "loss": 0.9435,
+      "step": 730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7e-05,
+      "loss": 0.9277,
+      "step": 740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 0.9369,
+      "step": 750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8e-05,
+      "loss": 0.9304,
+      "step": 760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.85e-05,
+      "loss": 0.929,
+      "step": 770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 0.9181,
+      "step": 780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9500000000000005e-05,
+      "loss": 0.9128,
+      "step": 790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 0.9203,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.05e-05,
+      "loss": 0.9045,
+      "step": 810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1e-05,
+      "loss": 0.9146,
+      "step": 820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.15e-05,
+      "loss": 0.8995,
+      "step": 830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2e-05,
+      "loss": 0.8941,
+      "step": 840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.25e-05,
+      "loss": 0.882,
+      "step": 850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3e-05,
+      "loss": 0.8762,
+      "step": 860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.35e-05,
+      "loss": 0.873,
+      "step": 870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.8696,
+      "step": 880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4500000000000004e-05,
+      "loss": 0.8757,
+      "step": 890
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5e-05,
+      "loss": 0.8679,
+      "step": 900
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.55e-05,
+      "loss": 0.8856,
+      "step": 910
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.8639,
+      "step": 920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6500000000000005e-05,
+      "loss": 0.8649,
+      "step": 930
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7e-05,
+      "loss": 0.854,
+      "step": 940
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.75e-05,
+      "loss": 0.85,
+      "step": 950
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8e-05,
+      "loss": 0.8576,
+      "step": 960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.85e-05,
+      "loss": 0.851,
+      "step": 970
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9e-05,
+      "loss": 0.8442,
+      "step": 980
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 0.8519,
+      "step": 990
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-05,
+      "loss": 0.823,
+      "step": 1000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.999549590127015e-05,
+      "loss": 0.8245,
+      "step": 1010
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.999099180254032e-05,
+      "loss": 0.8233,
+      "step": 1020
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.998648770381047e-05,
+      "loss": 0.8105,
+      "step": 1030
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9981983605080626e-05,
+      "loss": 0.8243,
+      "step": 1040
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.997747950635078e-05,
+      "loss": 0.8143,
+      "step": 1050
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9972975407620934e-05,
+      "loss": 0.8189,
+      "step": 1060
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.996847130889109e-05,
+      "loss": 0.8096,
+      "step": 1070
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.996396721016125e-05,
+      "loss": 0.8259,
+      "step": 1080
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9959463111431406e-05,
+      "loss": 0.8043,
+      "step": 1090
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9954959012701564e-05,
+      "loss": 0.8135,
+      "step": 1100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9950454913971714e-05,
+      "loss": 0.8117,
+      "step": 1110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.994595081524187e-05,
+      "loss": 0.8017,
+      "step": 1120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.994144671651203e-05,
+      "loss": 0.7817,
+      "step": 1130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.993694261778219e-05,
+      "loss": 0.7965,
+      "step": 1140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.993243851905234e-05,
+      "loss": 0.799,
+      "step": 1150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9927934420322495e-05,
+      "loss": 0.7889,
+      "step": 1160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.992343032159265e-05,
+      "loss": 0.7837,
+      "step": 1170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.99189262228628e-05,
+      "loss": 0.7745,
+      "step": 1180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.991442212413297e-05,
+      "loss": 0.7818,
+      "step": 1190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.990991802540312e-05,
+      "loss": 0.777,
+      "step": 1200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.990541392667327e-05,
+      "loss": 0.7802,
+      "step": 1210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.990090982794343e-05,
+      "loss": 0.7723,
+      "step": 1220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9896405729213584e-05,
+      "loss": 0.7667,
+      "step": 1230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.989190163048375e-05,
+      "loss": 0.7629,
+      "step": 1240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.98873975317539e-05,
+      "loss": 0.7658,
+      "step": 1250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.988289343302405e-05,
+      "loss": 0.7675,
+      "step": 1260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9878389334294214e-05,
+      "loss": 0.7783,
+      "step": 1270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9873885235564365e-05,
+      "loss": 0.7725,
+      "step": 1280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.986938113683452e-05,
+      "loss": 0.7466,
+      "step": 1290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.986487703810468e-05,
+      "loss": 0.7584,
+      "step": 1300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.986037293937483e-05,
+      "loss": 0.7501,
+      "step": 1310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.985586884064499e-05,
+      "loss": 0.7441,
+      "step": 1320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9851364741915145e-05,
+      "loss": 0.7623,
+      "step": 1330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.98468606431853e-05,
+      "loss": 0.7599,
+      "step": 1340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9842356544455454e-05,
+      "loss": 0.7447,
+      "step": 1350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.983785244572561e-05,
+      "loss": 0.7466,
+      "step": 1360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.983334834699577e-05,
+      "loss": 0.7459,
+      "step": 1370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9828844248265926e-05,
+      "loss": 0.7451,
+      "step": 1380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9824340149536084e-05,
+      "loss": 0.7386,
+      "step": 1390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9819836050806234e-05,
+      "loss": 0.7357,
+      "step": 1400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.981533195207639e-05,
+      "loss": 0.7353,
+      "step": 1410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.981082785334655e-05,
+      "loss": 0.7355,
+      "step": 1420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.98063237546167e-05,
+      "loss": 0.7362,
+      "step": 1430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9801819655886864e-05,
+      "loss": 0.7297,
+      "step": 1440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9797315557157015e-05,
+      "loss": 0.7255,
+      "step": 1450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9792811458427166e-05,
+      "loss": 0.7226,
+      "step": 1460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.978830735969733e-05,
+      "loss": 0.7212,
+      "step": 1470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.978380326096748e-05,
+      "loss": 0.7315,
+      "step": 1480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.977929916223764e-05,
+      "loss": 0.7288,
+      "step": 1490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9774795063507796e-05,
+      "loss": 0.7242,
+      "step": 1500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9770290964777946e-05,
+      "loss": 0.7262,
+      "step": 1510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.976578686604811e-05,
+      "loss": 0.7144,
+      "step": 1520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.976128276731826e-05,
+      "loss": 0.7137,
+      "step": 1530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.975677866858842e-05,
+      "loss": 0.7135,
+      "step": 1540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9752274569858576e-05,
+      "loss": 0.7088,
+      "step": 1550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.974777047112873e-05,
+      "loss": 0.7179,
+      "step": 1560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9743266372398885e-05,
+      "loss": 0.7158,
+      "step": 1570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.973876227366904e-05,
+      "loss": 0.7069,
+      "step": 1580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.97342581749392e-05,
+      "loss": 0.7045,
+      "step": 1590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.972975407620935e-05,
+      "loss": 0.7086,
+      "step": 1600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.972524997747951e-05,
+      "loss": 0.7045,
+      "step": 1610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9720745878749665e-05,
+      "loss": 0.6964,
+      "step": 1620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9716241780019816e-05,
+      "loss": 0.6968,
+      "step": 1630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.971173768128998e-05,
+      "loss": 0.7022,
+      "step": 1640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.970723358256013e-05,
+      "loss": 0.7158,
+      "step": 1650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.970272948383029e-05,
+      "loss": 0.7046,
+      "step": 1660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9698225385100446e-05,
+      "loss": 0.7001,
+      "step": 1670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.96937212863706e-05,
+      "loss": 0.6951,
+      "step": 1680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9689217187640754e-05,
+      "loss": 0.687,
+      "step": 1690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.968471308891091e-05,
+      "loss": 0.692,
+      "step": 1700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.968020899018106e-05,
+      "loss": 0.7096,
+      "step": 1710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.967570489145123e-05,
+      "loss": 0.6935,
+      "step": 1720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.967120079272138e-05,
+      "loss": 0.6947,
+      "step": 1730
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9666696693991535e-05,
+      "loss": 0.6895,
+      "step": 1740
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.966219259526169e-05,
+      "loss": 0.6882,
+      "step": 1750
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.965768849653184e-05,
+      "loss": 0.6756,
+      "step": 1760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9653184397802e-05,
+      "loss": 0.6963,
+      "step": 1770
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.964868029907216e-05,
+      "loss": 0.6819,
+      "step": 1780
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9644176200342316e-05,
+      "loss": 0.6813,
+      "step": 1790
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.963967210161247e-05,
+      "loss": 0.6769,
+      "step": 1800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9635168002882624e-05,
+      "loss": 0.6751,
+      "step": 1810
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.963066390415278e-05,
+      "loss": 0.6791,
+      "step": 1820
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.962615980542294e-05,
+      "loss": 0.6782,
+      "step": 1830
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9621655706693096e-05,
+      "loss": 0.6866,
+      "step": 1840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.961715160796325e-05,
+      "loss": 0.6788,
+      "step": 1850
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9612647509233404e-05,
+      "loss": 0.6751,
+      "step": 1860
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.960814341050356e-05,
+      "loss": 0.6736,
+      "step": 1870
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.960363931177371e-05,
+      "loss": 0.6721,
+      "step": 1880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.959913521304388e-05,
+      "loss": 0.6709,
+      "step": 1890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.959463111431403e-05,
+      "loss": 0.6756,
+      "step": 1900
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.959012701558418e-05,
+      "loss": 0.6746,
+      "step": 1910
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.958562291685434e-05,
+      "loss": 0.6728,
+      "step": 1920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.958111881812449e-05,
+      "loss": 0.6707,
+      "step": 1930
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.957661471939465e-05,
+      "loss": 0.671,
+      "step": 1940
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.957211062066481e-05,
+      "loss": 0.666,
+      "step": 1950
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.956760652193496e-05,
+      "loss": 0.6593,
+      "step": 1960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.956310242320512e-05,
+      "loss": 0.6597,
+      "step": 1970
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9558598324475274e-05,
+      "loss": 0.661,
+      "step": 1980
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.955409422574543e-05,
+      "loss": 0.6622,
+      "step": 1990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.954959012701559e-05,
+      "loss": 0.6585,
+      "step": 2000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.954508602828574e-05,
+      "loss": 0.657,
+      "step": 2010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.95405819295559e-05,
+      "loss": 0.6502,
+      "step": 2020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9536077830826055e-05,
+      "loss": 0.6568,
+      "step": 2030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.953157373209621e-05,
+      "loss": 0.6657,
+      "step": 2040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.952706963336636e-05,
+      "loss": 0.6577,
+      "step": 2050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.952256553463652e-05,
+      "loss": 0.6433,
+      "step": 2060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.951806143590668e-05,
+      "loss": 0.6408,
+      "step": 2070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9513557337176835e-05,
+      "loss": 0.645,
+      "step": 2080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.950905323844699e-05,
+      "loss": 0.6593,
+      "step": 2090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9504549139717144e-05,
+      "loss": 0.6411,
+      "step": 2100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.95000450409873e-05,
+      "loss": 0.6464,
+      "step": 2110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.949554094225746e-05,
+      "loss": 0.6513,
+      "step": 2120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.949103684352761e-05,
+      "loss": 0.6556,
+      "step": 2130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.948653274479777e-05,
+      "loss": 0.6494,
+      "step": 2140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9482028646067924e-05,
+      "loss": 0.6472,
+      "step": 2150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9477524547338075e-05,
+      "loss": 0.6541,
+      "step": 2160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.947302044860824e-05,
+      "loss": 0.6434,
+      "step": 2170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.946851634987839e-05,
+      "loss": 0.6395,
+      "step": 2180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.946401225114855e-05,
+      "loss": 0.6425,
+      "step": 2190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9459508152418705e-05,
+      "loss": 0.6401,
+      "step": 2200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9455004053688856e-05,
+      "loss": 0.6413,
+      "step": 2210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.945049995495902e-05,
+      "loss": 0.6343,
+      "step": 2220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.944599585622917e-05,
+      "loss": 0.6376,
+      "step": 2230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.944149175749933e-05,
+      "loss": 0.6449,
+      "step": 2240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9436987658769486e-05,
+      "loss": 0.6488,
+      "step": 2250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9432483560039636e-05,
+      "loss": 0.6449,
+      "step": 2260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9427979461309794e-05,
+      "loss": 0.6362,
+      "step": 2270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.942347536257995e-05,
+      "loss": 0.6282,
+      "step": 2280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.941897126385011e-05,
+      "loss": 0.6325,
+      "step": 2290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.941446716512026e-05,
+      "loss": 0.6248,
+      "step": 2300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.940996306639042e-05,
+      "loss": 0.6321,
+      "step": 2310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9405458967660575e-05,
+      "loss": 0.6337,
+      "step": 2320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9400954868930725e-05,
+      "loss": 0.6247,
+      "step": 2330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.939645077020088e-05,
+      "loss": 0.6413,
+      "step": 2340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.939194667147104e-05,
+      "loss": 0.6293,
+      "step": 2350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.93874425727412e-05,
+      "loss": 0.6231,
+      "step": 2360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9382938474011355e-05,
+      "loss": 0.6164,
+      "step": 2370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9378434375281506e-05,
+      "loss": 0.624,
+      "step": 2380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9373930276551663e-05,
+      "loss": 0.6209,
+      "step": 2390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.936942617782182e-05,
+      "loss": 0.6242,
+      "step": 2400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.936492207909197e-05,
+      "loss": 0.6156,
+      "step": 2410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9360417980362136e-05,
+      "loss": 0.6301,
+      "step": 2420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9355913881632287e-05,
+      "loss": 0.6267,
+      "step": 2430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9351409782902444e-05,
+      "loss": 0.6251,
+      "step": 2440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.93469056841726e-05,
+      "loss": 0.6307,
+      "step": 2450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.934240158544275e-05,
+      "loss": 0.6141,
+      "step": 2460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.933789748671291e-05,
+      "loss": 0.6151,
+      "step": 2470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.933339338798307e-05,
+      "loss": 0.6142,
+      "step": 2480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9328889289253225e-05,
+      "loss": 0.6157,
+      "step": 2490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.932438519052338e-05,
+      "loss": 0.6208,
+      "step": 2500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.931988109179353e-05,
+      "loss": 0.6123,
+      "step": 2510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.931537699306369e-05,
+      "loss": 0.6286,
+      "step": 2520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.931087289433385e-05,
+      "loss": 0.6165,
+      "step": 2530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9306368795604005e-05,
+      "loss": 0.6264,
+      "step": 2540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9301864696874156e-05,
+      "loss": 0.6146,
+      "step": 2550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9297360598144314e-05,
+      "loss": 0.6107,
+      "step": 2560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.929285649941447e-05,
+      "loss": 0.6246,
+      "step": 2570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.928835240068462e-05,
+      "loss": 0.6186,
+      "step": 2580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.928384830195478e-05,
+      "loss": 0.6052,
+      "step": 2590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.927934420322494e-05,
+      "loss": 0.6133,
+      "step": 2600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.927484010449509e-05,
+      "loss": 0.6051,
+      "step": 2610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.927033600576525e-05,
+      "loss": 0.6103,
+      "step": 2620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.92658319070354e-05,
+      "loss": 0.6126,
+      "step": 2630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.926132780830556e-05,
+      "loss": 0.6001,
+      "step": 2640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.925682370957572e-05,
+      "loss": 0.6016,
+      "step": 2650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.925231961084587e-05,
+      "loss": 0.5969,
+      "step": 2660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.924781551211603e-05,
+      "loss": 0.6054,
+      "step": 2670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.924331141338618e-05,
+      "loss": 0.607,
+      "step": 2680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.923880731465634e-05,
+      "loss": 0.6077,
+      "step": 2690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.92343032159265e-05,
+      "loss": 0.6062,
+      "step": 2700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.922979911719665e-05,
+      "loss": 0.614,
+      "step": 2710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9225295018466806e-05,
+      "loss": 0.6075,
+      "step": 2720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9220790919736964e-05,
+      "loss": 0.5958,
+      "step": 2730
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.921628682100712e-05,
+      "loss": 0.6007,
+      "step": 2740
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.921178272227727e-05,
+      "loss": 0.6013,
+      "step": 2750
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.920727862354743e-05,
+      "loss": 0.6001,
+      "step": 2760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.920277452481759e-05,
+      "loss": 0.596,
+      "step": 2770
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9198270426087745e-05,
+      "loss": 0.5937,
+      "step": 2780
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9193766327357895e-05,
+      "loss": 0.5946,
+      "step": 2790
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.918926222862805e-05,
+      "loss": 0.5972,
+      "step": 2800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.918475812989821e-05,
+      "loss": 0.5965,
+      "step": 2810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.918025403116837e-05,
+      "loss": 0.5984,
+      "step": 2820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.917574993243852e-05,
+      "loss": 0.6022,
+      "step": 2830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9171245833708676e-05,
+      "loss": 0.598,
+      "step": 2840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9166741734978834e-05,
+      "loss": 0.5907,
+      "step": 2850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9162237636248984e-05,
+      "loss": 0.587,
+      "step": 2860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.915773353751915e-05,
+      "loss": 0.593,
+      "step": 2870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.91532294387893e-05,
+      "loss": 0.5956,
+      "step": 2880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.914872534005946e-05,
+      "loss": 0.5963,
+      "step": 2890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9144221241329614e-05,
+      "loss": 0.5877,
+      "step": 2900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9139717142599765e-05,
+      "loss": 0.59,
+      "step": 2910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.913521304386993e-05,
+      "loss": 0.5918,
+      "step": 2920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.913070894514008e-05,
+      "loss": 0.5927,
+      "step": 2930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.912620484641024e-05,
+      "loss": 0.5965,
+      "step": 2940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9121700747680395e-05,
+      "loss": 0.587,
+      "step": 2950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9117196648950546e-05,
+      "loss": 0.5781,
+      "step": 2960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.91126925502207e-05,
+      "loss": 0.5836,
+      "step": 2970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.910818845149086e-05,
+      "loss": 0.5858,
+      "step": 2980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.910368435276101e-05,
+      "loss": 0.589,
+      "step": 2990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.909918025403117e-05,
+      "loss": 0.591,
+      "step": 3000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9094676155301326e-05,
+      "loss": 0.5819,
+      "step": 3010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9090172056571484e-05,
+      "loss": 0.5861,
+      "step": 3020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9085667957841634e-05,
+      "loss": 0.5859,
+      "step": 3030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.908116385911179e-05,
+      "loss": 0.5843,
+      "step": 3040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.907665976038195e-05,
+      "loss": 0.5909,
+      "step": 3050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.907215566165211e-05,
+      "loss": 0.5877,
+      "step": 3060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9067651562922264e-05,
+      "loss": 0.5849,
+      "step": 3070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9063147464192415e-05,
+      "loss": 0.5802,
+      "step": 3080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.905864336546257e-05,
+      "loss": 0.5763,
+      "step": 3090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.905413926673273e-05,
+      "loss": 0.5846,
+      "step": 3100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.904963516800288e-05,
+      "loss": 0.573,
+      "step": 3110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9045131069273045e-05,
+      "loss": 0.575,
+      "step": 3120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9040626970543196e-05,
+      "loss": 0.5837,
+      "step": 3130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.903612287181335e-05,
+      "loss": 0.5814,
+      "step": 3140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.903161877308351e-05,
+      "loss": 0.5766,
+      "step": 3150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.902711467435366e-05,
+      "loss": 0.5823,
+      "step": 3160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.902261057562382e-05,
+      "loss": 0.5753,
+      "step": 3170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9018106476893977e-05,
+      "loss": 0.5861,
+      "step": 3180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9013602378164134e-05,
+      "loss": 0.578,
+      "step": 3190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.900909827943429e-05,
+      "loss": 0.5695,
+      "step": 3200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.900459418070444e-05,
+      "loss": 0.5804,
+      "step": 3210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.90000900819746e-05,
+      "loss": 0.5748,
+      "step": 3220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.899558598324476e-05,
+      "loss": 0.5747,
+      "step": 3230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.899108188451491e-05,
+      "loss": 0.5843,
+      "step": 3240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8986577785785065e-05,
+      "loss": 0.5747,
+      "step": 3250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.898207368705522e-05,
+      "loss": 0.5733,
+      "step": 3260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.897756958832538e-05,
+      "loss": 0.567,
+      "step": 3270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.897306548959553e-05,
+      "loss": 0.5765,
+      "step": 3280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.896856139086569e-05,
+      "loss": 0.5722,
+      "step": 3290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8964057292135846e-05,
+      "loss": 0.5716,
+      "step": 3300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8959553193406e-05,
+      "loss": 0.572,
+      "step": 3310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.895504909467616e-05,
+      "loss": 0.5678,
+      "step": 3320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.895054499594631e-05,
+      "loss": 0.5619,
+      "step": 3330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.894604089721647e-05,
+      "loss": 0.5623,
+      "step": 3340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.894153679848663e-05,
+      "loss": 0.565,
+      "step": 3350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.893703269975678e-05,
+      "loss": 0.5696,
+      "step": 3360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.893252860102694e-05,
+      "loss": 0.5626,
+      "step": 3370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.892802450229709e-05,
+      "loss": 0.5641,
+      "step": 3380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.892352040356725e-05,
+      "loss": 0.5678,
+      "step": 3390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.891901630483741e-05,
+      "loss": 0.5631,
+      "step": 3400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.891451220610756e-05,
+      "loss": 0.5699,
+      "step": 3410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8910008107377716e-05,
+      "loss": 0.5677,
+      "step": 3420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.890550400864787e-05,
+      "loss": 0.5635,
+      "step": 3430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8900999909918024e-05,
+      "loss": 0.5675,
+      "step": 3440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.889649581118818e-05,
+      "loss": 0.5553,
+      "step": 3450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.889199171245834e-05,
+      "loss": 0.5655,
+      "step": 3460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8887487613728496e-05,
+      "loss": 0.5568,
+      "step": 3470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8882983514998654e-05,
+      "loss": 0.5627,
+      "step": 3480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8878479416268805e-05,
+      "loss": 0.5623,
+      "step": 3490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.887397531753896e-05,
+      "loss": 0.5674,
+      "step": 3500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.886947121880912e-05,
+      "loss": 0.558,
+      "step": 3510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.886496712007928e-05,
+      "loss": 0.5593,
+      "step": 3520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.886046302134943e-05,
+      "loss": 0.5604,
+      "step": 3530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8855958922619585e-05,
+      "loss": 0.5658,
+      "step": 3540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.885145482388974e-05,
+      "loss": 0.5564,
+      "step": 3550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8846950725159893e-05,
+      "loss": 0.5589,
+      "step": 3560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.884244662643006e-05,
+      "loss": 0.555,
+      "step": 3570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.883794252770021e-05,
+      "loss": 0.5544,
+      "step": 3580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8833438428970366e-05,
+      "loss": 0.5582,
+      "step": 3590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8828934330240523e-05,
+      "loss": 0.5516,
+      "step": 3600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8824430231510674e-05,
+      "loss": 0.5566,
+      "step": 3610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.881992613278084e-05,
+      "loss": 0.5517,
+      "step": 3620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.881542203405099e-05,
+      "loss": 0.5518,
+      "step": 3630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.881091793532114e-05,
+      "loss": 0.5511,
+      "step": 3640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8806413836591304e-05,
+      "loss": 0.5618,
+      "step": 3650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8801909737861455e-05,
+      "loss": 0.5454,
+      "step": 3660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.879740563913161e-05,
+      "loss": 0.5564,
+      "step": 3670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.879290154040177e-05,
+      "loss": 0.558,
+      "step": 3680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.878839744167192e-05,
+      "loss": 0.5519,
+      "step": 3690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.878389334294208e-05,
+      "loss": 0.5461,
+      "step": 3700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8779389244212236e-05,
+      "loss": 0.5558,
+      "step": 3710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.877488514548239e-05,
+      "loss": 0.5593,
+      "step": 3720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8770381046752544e-05,
+      "loss": 0.5586,
+      "step": 3730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.87658769480227e-05,
+      "loss": 0.5455,
+      "step": 3740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.876137284929286e-05,
+      "loss": 0.5522,
+      "step": 3750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8756868750563016e-05,
+      "loss": 0.553,
+      "step": 3760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8752364651833174e-05,
+      "loss": 0.5501,
+      "step": 3770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8747860553103324e-05,
+      "loss": 0.5448,
+      "step": 3780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.874335645437348e-05,
+      "loss": 0.5454,
+      "step": 3790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.873885235564364e-05,
+      "loss": 0.5429,
+      "step": 3800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.873434825691379e-05,
+      "loss": 0.5503,
+      "step": 3810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8729844158183954e-05,
+      "loss": 0.555,
+      "step": 3820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8725340059454105e-05,
+      "loss": 0.5426,
+      "step": 3830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.872083596072426e-05,
+      "loss": 0.5435,
+      "step": 3840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.871633186199442e-05,
+      "loss": 0.5513,
+      "step": 3850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.871182776326457e-05,
+      "loss": 0.5425,
+      "step": 3860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.870732366453473e-05,
+      "loss": 0.5486,
+      "step": 3870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8702819565804886e-05,
+      "loss": 0.5487,
+      "step": 3880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8698315467075037e-05,
+      "loss": 0.5357,
+      "step": 3890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.86938113683452e-05,
+      "loss": 0.551,
+      "step": 3900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.868930726961535e-05,
+      "loss": 0.5425,
+      "step": 3910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.868480317088551e-05,
+      "loss": 0.5472,
+      "step": 3920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8680299072155666e-05,
+      "loss": 0.5333,
+      "step": 3930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.867579497342582e-05,
+      "loss": 0.543,
+      "step": 3940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8671290874695975e-05,
+      "loss": 0.5498,
+      "step": 3950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.866678677596613e-05,
+      "loss": 0.5499,
+      "step": 3960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.866228267723629e-05,
+      "loss": 0.5519,
+      "step": 3970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.865777857850644e-05,
+      "loss": 0.5457,
+      "step": 3980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.86532744797766e-05,
+      "loss": 0.546,
+      "step": 3990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8648770381046755e-05,
+      "loss": 0.5387,
+      "step": 4000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8644266282316906e-05,
+      "loss": 0.5373,
+      "step": 4010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.863976218358707e-05,
+      "loss": 0.5405,
+      "step": 4020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.863525808485722e-05,
+      "loss": 0.5395,
+      "step": 4030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.863075398612738e-05,
+      "loss": 0.5483,
+      "step": 4040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8626249887397536e-05,
+      "loss": 0.5434,
+      "step": 4050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.862174578866769e-05,
+      "loss": 0.5366,
+      "step": 4060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.861724168993785e-05,
+      "loss": 0.5329,
+      "step": 4070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8612737591208e-05,
+      "loss": 0.5478,
+      "step": 4080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.860823349247815e-05,
+      "loss": 0.5339,
+      "step": 4090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.860372939374832e-05,
+      "loss": 0.5342,
+      "step": 4100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.859922529501847e-05,
+      "loss": 0.5387,
+      "step": 4110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8594721196288625e-05,
+      "loss": 0.5368,
+      "step": 4120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.859021709755878e-05,
+      "loss": 0.5337,
+      "step": 4130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.858571299882893e-05,
+      "loss": 0.5409,
+      "step": 4140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.858120890009909e-05,
+      "loss": 0.5447,
+      "step": 4150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.857670480136925e-05,
+      "loss": 0.5323,
+      "step": 4160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8572200702639406e-05,
+      "loss": 0.5305,
+      "step": 4170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.856769660390956e-05,
+      "loss": 0.5379,
+      "step": 4180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8563192505179714e-05,
+      "loss": 0.5368,
+      "step": 4190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.855868840644987e-05,
+      "loss": 0.5368,
+      "step": 4200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.855418430772003e-05,
+      "loss": 0.531,
+      "step": 4210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8549680208990186e-05,
+      "loss": 0.5291,
+      "step": 4220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.854517611026034e-05,
+      "loss": 0.5398,
+      "step": 4230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8540672011530495e-05,
+      "loss": 0.5346,
+      "step": 4240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.853616791280065e-05,
+      "loss": 0.529,
+      "step": 4250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.85316638140708e-05,
+      "loss": 0.5343,
+      "step": 4260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.852715971534097e-05,
+      "loss": 0.5276,
+      "step": 4270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.852265561661112e-05,
+      "loss": 0.5326,
+      "step": 4280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.851815151788127e-05,
+      "loss": 0.5272,
+      "step": 4290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.851364741915143e-05,
+      "loss": 0.5221,
+      "step": 4300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8509143320421583e-05,
+      "loss": 0.5332,
+      "step": 4310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.850463922169175e-05,
+      "loss": 0.5353,
+      "step": 4320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.85001351229619e-05,
+      "loss": 0.5251,
+      "step": 4330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.849563102423205e-05,
+      "loss": 0.525,
+      "step": 4340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8491126925502213e-05,
+      "loss": 0.5233,
+      "step": 4350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8486622826772364e-05,
+      "loss": 0.5301,
+      "step": 4360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.848211872804252e-05,
+      "loss": 0.5247,
+      "step": 4370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.847761462931268e-05,
+      "loss": 0.5301,
+      "step": 4380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.847311053058283e-05,
+      "loss": 0.5239,
+      "step": 4390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.846860643185299e-05,
+      "loss": 0.5225,
+      "step": 4400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8464102333123145e-05,
+      "loss": 0.5249,
+      "step": 4410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.84595982343933e-05,
+      "loss": 0.5241,
+      "step": 4420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.845509413566345e-05,
+      "loss": 0.5242,
+      "step": 4430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.845059003693361e-05,
+      "loss": 0.519,
+      "step": 4440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.844608593820377e-05,
+      "loss": 0.5223,
+      "step": 4450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8441581839473925e-05,
+      "loss": 0.5246,
+      "step": 4460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.843707774074408e-05,
+      "loss": 0.5242,
+      "step": 4470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8432573642014234e-05,
+      "loss": 0.5208,
+      "step": 4480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.842806954328439e-05,
+      "loss": 0.5207,
+      "step": 4490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.842356544455455e-05,
+      "loss": 0.5179,
+      "step": 4500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.84190613458247e-05,
+      "loss": 0.5216,
+      "step": 4510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8414557247094864e-05,
+      "loss": 0.5266,
+      "step": 4520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8410053148365014e-05,
+      "loss": 0.5256,
+      "step": 4530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8405549049635165e-05,
+      "loss": 0.5296,
+      "step": 4540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.840104495090533e-05,
+      "loss": 0.5135,
+      "step": 4550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.839654085217548e-05,
+      "loss": 0.5172,
+      "step": 4560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.839203675344564e-05,
+      "loss": 0.5212,
+      "step": 4570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8387532654715795e-05,
+      "loss": 0.5257,
+      "step": 4580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8383028555985946e-05,
+      "loss": 0.5268,
+      "step": 4590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.837852445725611e-05,
+      "loss": 0.5204,
+      "step": 4600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.837402035852626e-05,
+      "loss": 0.5172,
+      "step": 4610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.836951625979642e-05,
+      "loss": 0.509,
+      "step": 4620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8365012161066576e-05,
+      "loss": 0.5138,
+      "step": 4630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8360508062336726e-05,
+      "loss": 0.5104,
+      "step": 4640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8356003963606884e-05,
+      "loss": 0.5066,
+      "step": 4650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.835149986487704e-05,
+      "loss": 0.5159,
+      "step": 4660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.83469957661472e-05,
+      "loss": 0.515,
+      "step": 4670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.834249166741735e-05,
+      "loss": 0.524,
+      "step": 4680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.833798756868751e-05,
+      "loss": 0.5194,
+      "step": 4690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8333483469957665e-05,
+      "loss": 0.5196,
+      "step": 4700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8328979371227815e-05,
+      "loss": 0.5163,
+      "step": 4710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.832447527249798e-05,
+      "loss": 0.509,
+      "step": 4720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.831997117376813e-05,
+      "loss": 0.5138,
+      "step": 4730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.831546707503829e-05,
+      "loss": 0.5125,
+      "step": 4740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8310962976308445e-05,
+      "loss": 0.5086,
+      "step": 4750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8306458877578596e-05,
+      "loss": 0.5187,
+      "step": 4760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.830195477884876e-05,
+      "loss": 0.5172,
+      "step": 4770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.829745068011891e-05,
+      "loss": 0.5107,
+      "step": 4780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.829294658138906e-05,
+      "loss": 0.5124,
+      "step": 4790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8288442482659226e-05,
+      "loss": 0.5193,
+      "step": 4800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.828393838392938e-05,
+      "loss": 0.5234,
+      "step": 4810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8279434285199534e-05,
+      "loss": 0.5183,
+      "step": 4820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.827493018646969e-05,
+      "loss": 0.5062,
+      "step": 4830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.827042608773984e-05,
+      "loss": 0.5161,
+      "step": 4840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.826592198901e-05,
+      "loss": 0.5056,
+      "step": 4850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.826141789028016e-05,
+      "loss": 0.5121,
+      "step": 4860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8256913791550315e-05,
+      "loss": 0.505,
+      "step": 4870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.825240969282047e-05,
+      "loss": 0.5041,
+      "step": 4880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.824790559409062e-05,
+      "loss": 0.5174,
+      "step": 4890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.824340149536078e-05,
+      "loss": 0.5112,
+      "step": 4900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.823889739663094e-05,
+      "loss": 0.5041,
+      "step": 4910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8234393297901096e-05,
+      "loss": 0.5107,
+      "step": 4920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8229889199171246e-05,
+      "loss": 0.5069,
+      "step": 4930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8225385100441404e-05,
+      "loss": 0.5033,
+      "step": 4940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.822088100171156e-05,
+      "loss": 0.5079,
+      "step": 4950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.821637690298171e-05,
+      "loss": 0.5112,
+      "step": 4960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.8211872804251876e-05,
+      "loss": 0.5072,
+      "step": 4970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.820736870552203e-05,
+      "loss": 0.5161,
+      "step": 4980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.820286460679218e-05,
+      "loss": 0.5092,
+      "step": 4990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.819836050806234e-05,
+      "loss": 0.5129,
+      "step": 5000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.819385640933249e-05,
+      "loss": 0.503,
+      "step": 5010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.818935231060265e-05,
+      "loss": 0.5073,
+      "step": 5020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.818484821187281e-05,
+      "loss": 0.509,
+      "step": 5030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.818034411314296e-05,
+      "loss": 0.5086,
+      "step": 5040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.817584001441312e-05,
+      "loss": 0.5002,
+      "step": 5050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.817133591568327e-05,
+      "loss": 0.5097,
+      "step": 5060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.816683181695343e-05,
+      "loss": 0.5133,
+      "step": 5070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.816232771822359e-05,
+      "loss": 0.5051,
+      "step": 5080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.815782361949374e-05,
+      "loss": 0.4998,
+      "step": 5090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8153319520763897e-05,
+      "loss": 0.5067,
+      "step": 5100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8148815422034054e-05,
+      "loss": 0.5022,
+      "step": 5110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.814431132330421e-05,
+      "loss": 0.4996,
+      "step": 5120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.813980722457436e-05,
+      "loss": 0.5028,
+      "step": 5130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.813530312584452e-05,
+      "loss": 0.5042,
+      "step": 5140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.813079902711468e-05,
+      "loss": 0.503,
+      "step": 5150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8126294928384835e-05,
+      "loss": 0.5028,
+      "step": 5160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.812179082965499e-05,
+      "loss": 0.5074,
+      "step": 5170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.811728673092514e-05,
+      "loss": 0.4989,
+      "step": 5180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.81127826321953e-05,
+      "loss": 0.4971,
+      "step": 5190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.810827853346546e-05,
+      "loss": 0.4975,
+      "step": 5200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.810377443473561e-05,
+      "loss": 0.5005,
+      "step": 5210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8099270336005766e-05,
+      "loss": 0.4984,
+      "step": 5220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8094766237275924e-05,
+      "loss": 0.5091,
+      "step": 5230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8090262138546074e-05,
+      "loss": 0.5023,
+      "step": 5240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.808575803981624e-05,
+      "loss": 0.5034,
+      "step": 5250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.808125394108639e-05,
+      "loss": 0.5039,
+      "step": 5260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.807674984235655e-05,
+      "loss": 0.4982,
+      "step": 5270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8072245743626704e-05,
+      "loss": 0.4987,
+      "step": 5280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8067741644896855e-05,
+      "loss": 0.5001,
+      "step": 5290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.806323754616702e-05,
+      "loss": 0.4977,
+      "step": 5300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.805873344743717e-05,
+      "loss": 0.4986,
+      "step": 5310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.805422934870733e-05,
+      "loss": 0.5021,
+      "step": 5320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8049725249977485e-05,
+      "loss": 0.4976,
+      "step": 5330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8045221151247636e-05,
+      "loss": 0.4944,
+      "step": 5340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.804071705251779e-05,
+      "loss": 0.4973,
+      "step": 5350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.803621295378795e-05,
+      "loss": 0.4967,
+      "step": 5360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.803170885505811e-05,
+      "loss": 0.4922,
+      "step": 5370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.802720475632826e-05,
+      "loss": 0.4996,
+      "step": 5380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8022700657598416e-05,
+      "loss": 0.4939,
+      "step": 5390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8018196558868574e-05,
+      "loss": 0.4895,
+      "step": 5400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8013692460138725e-05,
+      "loss": 0.4889,
+      "step": 5410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.800918836140889e-05,
+      "loss": 0.5032,
+      "step": 5420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.800468426267904e-05,
+      "loss": 0.4973,
+      "step": 5430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.80001801639492e-05,
+      "loss": 0.5013,
+      "step": 5440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7995676065219355e-05,
+      "loss": 0.4948,
+      "step": 5450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7991171966489505e-05,
+      "loss": 0.4958,
+      "step": 5460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.798666786775966e-05,
+      "loss": 0.4929,
+      "step": 5470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.798216376902982e-05,
+      "loss": 0.4822,
+      "step": 5480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.797765967029997e-05,
+      "loss": 0.4921,
+      "step": 5490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7973155571570135e-05,
+      "loss": 0.4982,
+      "step": 5500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7968651472840286e-05,
+      "loss": 0.4989,
+      "step": 5510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7964147374110443e-05,
+      "loss": 0.487,
+      "step": 5520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.79596432753806e-05,
+      "loss": 0.4991,
+      "step": 5530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.795513917665075e-05,
+      "loss": 0.4882,
+      "step": 5540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.795063507792091e-05,
+      "loss": 0.4846,
+      "step": 5550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.794613097919107e-05,
+      "loss": 0.4884,
+      "step": 5560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7941626880461224e-05,
+      "loss": 0.4951,
+      "step": 5570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.793712278173138e-05,
+      "loss": 0.4937,
+      "step": 5580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.793261868300153e-05,
+      "loss": 0.4953,
+      "step": 5590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.792811458427169e-05,
+      "loss": 0.4963,
+      "step": 5600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.792361048554185e-05,
+      "loss": 0.4959,
+      "step": 5610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7919106386812005e-05,
+      "loss": 0.4988,
+      "step": 5620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7914602288082156e-05,
+      "loss": 0.4915,
+      "step": 5630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.791009818935231e-05,
+      "loss": 0.4896,
+      "step": 5640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.790559409062247e-05,
+      "loss": 0.4882,
+      "step": 5650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.790108999189262e-05,
+      "loss": 0.4927,
+      "step": 5660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.789658589316278e-05,
+      "loss": 0.4839,
+      "step": 5670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7892081794432936e-05,
+      "loss": 0.486,
+      "step": 5680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.788757769570309e-05,
+      "loss": 0.487,
+      "step": 5690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.788307359697325e-05,
+      "loss": 0.4852,
+      "step": 5700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.78785694982434e-05,
+      "loss": 0.4874,
+      "step": 5710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.787406539951356e-05,
+      "loss": 0.4999,
+      "step": 5720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.786956130078372e-05,
+      "loss": 0.4855,
+      "step": 5730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.786505720205387e-05,
+      "loss": 0.4927,
+      "step": 5740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.786055310332403e-05,
+      "loss": 0.4927,
+      "step": 5750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.785604900459418e-05,
+      "loss": 0.4917,
+      "step": 5760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.785154490586434e-05,
+      "loss": 0.4843,
+      "step": 5770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.78470408071345e-05,
+      "loss": 0.4861,
+      "step": 5780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.784253670840465e-05,
+      "loss": 0.4865,
+      "step": 5790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7838032609674806e-05,
+      "loss": 0.4785,
+      "step": 5800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.783352851094496e-05,
+      "loss": 0.486,
+      "step": 5810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.782902441221512e-05,
+      "loss": 0.4827,
+      "step": 5820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.782452031348527e-05,
+      "loss": 0.4852,
+      "step": 5830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.782001621475543e-05,
+      "loss": 0.4819,
+      "step": 5840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7815512116025587e-05,
+      "loss": 0.4837,
+      "step": 5850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.781100801729574e-05,
+      "loss": 0.488,
+      "step": 5860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7806503918565895e-05,
+      "loss": 0.4852,
+      "step": 5870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.780199981983605e-05,
+      "loss": 0.4879,
+      "step": 5880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.779749572110621e-05,
+      "loss": 0.489,
+      "step": 5890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.779299162237637e-05,
+      "loss": 0.4785,
+      "step": 5900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.778848752364652e-05,
+      "loss": 0.4834,
+      "step": 5910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7783983424916675e-05,
+      "loss": 0.4749,
+      "step": 5920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.777947932618683e-05,
+      "loss": 0.4841,
+      "step": 5930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7774975227456984e-05,
+      "loss": 0.4821,
+      "step": 5940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.777047112872715e-05,
+      "loss": 0.4818,
+      "step": 5950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.77659670299973e-05,
+      "loss": 0.4799,
+      "step": 5960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7761462931267456e-05,
+      "loss": 0.4845,
+      "step": 5970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7756958832537614e-05,
+      "loss": 0.4808,
+      "step": 5980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7752454733807764e-05,
+      "loss": 0.4817,
+      "step": 5990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.774795063507792e-05,
+      "loss": 0.4811,
+      "step": 6000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.774344653634808e-05,
+      "loss": 0.4869,
+      "step": 6010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.773894243761824e-05,
+      "loss": 0.4907,
+      "step": 6020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7734438338888394e-05,
+      "loss": 0.4819,
+      "step": 6030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7729934240158545e-05,
+      "loss": 0.4809,
+      "step": 6040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.77254301414287e-05,
+      "loss": 0.4842,
+      "step": 6050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.772092604269886e-05,
+      "loss": 0.481,
+      "step": 6060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.771642194396902e-05,
+      "loss": 0.4821,
+      "step": 6070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.771191784523917e-05,
+      "loss": 0.4771,
+      "step": 6080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7707413746509326e-05,
+      "loss": 0.4883,
+      "step": 6090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.770290964777948e-05,
+      "loss": 0.4789,
+      "step": 6100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7698405549049634e-05,
+      "loss": 0.4801,
+      "step": 6110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.769390145031979e-05,
+      "loss": 0.4812,
+      "step": 6120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.768939735158995e-05,
+      "loss": 0.4795,
+      "step": 6130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.76848932528601e-05,
+      "loss": 0.4765,
+      "step": 6140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7680389154130264e-05,
+      "loss": 0.4835,
+      "step": 6150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.7675885055400415e-05,
+      "loss": 0.4804,
+      "step": 6160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.767138095667057e-05,
+      "loss": 0.4768,
+      "step": 6170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.766687685794073e-05,
+      "loss": 0.4778,
+      "step": 6180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.766237275921088e-05,
+      "loss": 0.4706,
+      "step": 6190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7657868660481045e-05,
+      "loss": 0.4827,
+      "step": 6200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7653364561751195e-05,
+      "loss": 0.4738,
+      "step": 6210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.764886046302135e-05,
+      "loss": 0.4757,
+      "step": 6220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.764435636429151e-05,
+      "loss": 0.4799,
+      "step": 6230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.763985226556166e-05,
+      "loss": 0.4713,
+      "step": 6240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.763534816683182e-05,
+      "loss": 0.4699,
+      "step": 6250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7630844068101976e-05,
+      "loss": 0.4725,
+      "step": 6260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7626339969372133e-05,
+      "loss": 0.4737,
+      "step": 6270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7621835870642284e-05,
+      "loss": 0.4824,
+      "step": 6280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.761733177191244e-05,
+      "loss": 0.4754,
+      "step": 6290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.76128276731826e-05,
+      "loss": 0.4784,
+      "step": 6300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7608323574452757e-05,
+      "loss": 0.4798,
+      "step": 6310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.760381947572291e-05,
+      "loss": 0.4668,
+      "step": 6320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7599315376993065e-05,
+      "loss": 0.4731,
+      "step": 6330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.759481127826322e-05,
+      "loss": 0.4688,
+      "step": 6340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.759030717953338e-05,
+      "loss": 0.4738,
+      "step": 6350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.758580308080353e-05,
+      "loss": 0.4716,
+      "step": 6360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.758129898207369e-05,
+      "loss": 0.4722,
+      "step": 6370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7576794883343845e-05,
+      "loss": 0.4708,
+      "step": 6380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7572290784613996e-05,
+      "loss": 0.4769,
+      "step": 6390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.756778668588416e-05,
+      "loss": 0.4841,
+      "step": 6400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.756328258715431e-05,
+      "loss": 0.4659,
+      "step": 6410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.755877848842447e-05,
+      "loss": 0.4707,
+      "step": 6420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7554274389694626e-05,
+      "loss": 0.4651,
+      "step": 6430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.754977029096478e-05,
+      "loss": 0.47,
+      "step": 6440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.754526619223494e-05,
+      "loss": 0.4697,
+      "step": 6450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.754076209350509e-05,
+      "loss": 0.4685,
+      "step": 6460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.753625799477525e-05,
+      "loss": 0.4718,
+      "step": 6470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.753175389604541e-05,
+      "loss": 0.4688,
+      "step": 6480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.752724979731556e-05,
+      "loss": 0.4705,
+      "step": 6490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7522745698585715e-05,
+      "loss": 0.4787,
+      "step": 6500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.751824159985587e-05,
+      "loss": 0.4714,
+      "step": 6510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.751373750112602e-05,
+      "loss": 0.4659,
+      "step": 6520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.750923340239618e-05,
+      "loss": 0.4721,
+      "step": 6530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.750472930366634e-05,
+      "loss": 0.4701,
+      "step": 6540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7500225204936496e-05,
+      "loss": 0.4704,
+      "step": 6550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7495721106206646e-05,
+      "loss": 0.4699,
+      "step": 6560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7491217007476804e-05,
+      "loss": 0.4692,
+      "step": 6570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.748671290874696e-05,
+      "loss": 0.4699,
+      "step": 6580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.748220881001712e-05,
+      "loss": 0.4725,
+      "step": 6590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7477704711287276e-05,
+      "loss": 0.4656,
+      "step": 6600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.747320061255743e-05,
+      "loss": 0.4645,
+      "step": 6610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7468696513827585e-05,
+      "loss": 0.4674,
+      "step": 6620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.746419241509774e-05,
+      "loss": 0.475,
+      "step": 6630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.745968831636789e-05,
+      "loss": 0.4701,
+      "step": 6640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.745518421763806e-05,
+      "loss": 0.471,
+      "step": 6650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.745068011890821e-05,
+      "loss": 0.4634,
+      "step": 6660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7446176020178365e-05,
+      "loss": 0.4669,
+      "step": 6670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.744167192144852e-05,
+      "loss": 0.477,
+      "step": 6680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7437167822718674e-05,
+      "loss": 0.4706,
+      "step": 6690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.743266372398883e-05,
+      "loss": 0.4631,
+      "step": 6700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.742815962525899e-05,
+      "loss": 0.4605,
+      "step": 6710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7423655526529146e-05,
+      "loss": 0.4624,
+      "step": 6720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7419151427799304e-05,
+      "loss": 0.4602,
+      "step": 6730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7414647329069454e-05,
+      "loss": 0.4651,
+      "step": 6740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.741014323033961e-05,
+      "loss": 0.4632,
+      "step": 6750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.740563913160977e-05,
+      "loss": 0.4608,
+      "step": 6760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.740113503287992e-05,
+      "loss": 0.4642,
+      "step": 6770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.739663093415008e-05,
+      "loss": 0.4609,
+      "step": 6780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7392126835420235e-05,
+      "loss": 0.4711,
+      "step": 6790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.738762273669039e-05,
+      "loss": 0.4691,
+      "step": 6800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.738311863796054e-05,
+      "loss": 0.4581,
+      "step": 6810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.73786145392307e-05,
+      "loss": 0.4652,
+      "step": 6820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.737411044050086e-05,
+      "loss": 0.4605,
+      "step": 6830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.736960634177101e-05,
+      "loss": 0.4673,
+      "step": 6840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.736510224304117e-05,
+      "loss": 0.4621,
+      "step": 6850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7360598144311324e-05,
+      "loss": 0.4601,
+      "step": 6860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.735609404558148e-05,
+      "loss": 0.4636,
+      "step": 6870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.735158994685164e-05,
+      "loss": 0.464,
+      "step": 6880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.734708584812179e-05,
+      "loss": 0.4613,
+      "step": 6890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7342581749391954e-05,
+      "loss": 0.4593,
+      "step": 6900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7338077650662104e-05,
+      "loss": 0.4559,
+      "step": 6910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.733357355193226e-05,
+      "loss": 0.4626,
+      "step": 6920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.732906945320242e-05,
+      "loss": 0.4659,
+      "step": 6930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.732456535447257e-05,
+      "loss": 0.4651,
+      "step": 6940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.732006125574273e-05,
+      "loss": 0.4608,
+      "step": 6950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7315557157012885e-05,
+      "loss": 0.4626,
+      "step": 6960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7311053058283036e-05,
+      "loss": 0.4504,
+      "step": 6970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.730654895955319e-05,
+      "loss": 0.4611,
+      "step": 6980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.730204486082335e-05,
+      "loss": 0.4561,
+      "step": 6990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.729754076209351e-05,
+      "loss": 0.4556,
+      "step": 7000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7293036663363666e-05,
+      "loss": 0.459,
+      "step": 7010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7288532564633817e-05,
+      "loss": 0.4649,
+      "step": 7020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7284028465903974e-05,
+      "loss": 0.4576,
+      "step": 7030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.727952436717413e-05,
+      "loss": 0.4645,
+      "step": 7040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.727502026844429e-05,
+      "loss": 0.4677,
+      "step": 7050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.727051616971444e-05,
+      "loss": 0.4592,
+      "step": 7060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.72660120709846e-05,
+      "loss": 0.4596,
+      "step": 7070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7261507972254755e-05,
+      "loss": 0.4563,
+      "step": 7080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7257003873524905e-05,
+      "loss": 0.4632,
+      "step": 7090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.725249977479507e-05,
+      "loss": 0.4574,
+      "step": 7100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.724799567606522e-05,
+      "loss": 0.4519,
+      "step": 7110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.724349157733538e-05,
+      "loss": 0.4637,
+      "step": 7120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7238987478605535e-05,
+      "loss": 0.4606,
+      "step": 7130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7234483379875686e-05,
+      "loss": 0.4589,
+      "step": 7140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.722997928114585e-05,
+      "loss": 0.4564,
+      "step": 7150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7225475182416e-05,
+      "loss": 0.458,
+      "step": 7160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.722097108368615e-05,
+      "loss": 0.4643,
+      "step": 7170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7216466984956316e-05,
+      "loss": 0.4613,
+      "step": 7180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.721196288622647e-05,
+      "loss": 0.4594,
+      "step": 7190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7207458787496624e-05,
+      "loss": 0.4554,
+      "step": 7200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.720295468876678e-05,
+      "loss": 0.4582,
+      "step": 7210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.719845059003693e-05,
+      "loss": 0.4581,
+      "step": 7220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.719394649130709e-05,
+      "loss": 0.4556,
+      "step": 7230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.718944239257725e-05,
+      "loss": 0.4555,
+      "step": 7240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7184938293847405e-05,
+      "loss": 0.4587,
+      "step": 7250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.7180434195117556e-05,
+      "loss": 0.4608,
+      "step": 7260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.717593009638771e-05,
+      "loss": 0.444,
+      "step": 7270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.717142599765787e-05,
+      "loss": 0.4546,
+      "step": 7280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.716692189892803e-05,
+      "loss": 0.4577,
+      "step": 7290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7162417800198186e-05,
+      "loss": 0.4527,
+      "step": 7300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7157913701468336e-05,
+      "loss": 0.4608,
+      "step": 7310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7153409602738494e-05,
+      "loss": 0.4599,
+      "step": 7320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.714890550400865e-05,
+      "loss": 0.4579,
+      "step": 7330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.71444014052788e-05,
+      "loss": 0.4503,
+      "step": 7340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7139897306548966e-05,
+      "loss": 0.4478,
+      "step": 7350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.713539320781912e-05,
+      "loss": 0.4559,
+      "step": 7360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7130889109089275e-05,
+      "loss": 0.4539,
+      "step": 7370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.712638501035943e-05,
+      "loss": 0.4545,
+      "step": 7380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.712188091162958e-05,
+      "loss": 0.452,
+      "step": 7390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.711737681289974e-05,
+      "loss": 0.4511,
+      "step": 7400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.71128727141699e-05,
+      "loss": 0.454,
+      "step": 7410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.710836861544005e-05,
+      "loss": 0.4555,
+      "step": 7420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.710386451671021e-05,
+      "loss": 0.4583,
+      "step": 7430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7099360417980363e-05,
+      "loss": 0.4585,
+      "step": 7440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.709485631925052e-05,
+      "loss": 0.4575,
+      "step": 7450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.709035222052068e-05,
+      "loss": 0.4586,
+      "step": 7460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.708584812179083e-05,
+      "loss": 0.4548,
+      "step": 7470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.708134402306099e-05,
+      "loss": 0.4567,
+      "step": 7480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7076839924331144e-05,
+      "loss": 0.4568,
+      "step": 7490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.70723358256013e-05,
+      "loss": 0.4638,
+      "step": 7500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.706783172687145e-05,
+      "loss": 0.4506,
+      "step": 7510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.706332762814161e-05,
+      "loss": 0.4497,
+      "step": 7520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.705882352941177e-05,
+      "loss": 0.4505,
+      "step": 7530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.705431943068192e-05,
+      "loss": 0.4499,
+      "step": 7540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.704981533195208e-05,
+      "loss": 0.4446,
+      "step": 7550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.704531123322223e-05,
+      "loss": 0.4461,
+      "step": 7560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.704080713449239e-05,
+      "loss": 0.448,
+      "step": 7570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.703630303576255e-05,
+      "loss": 0.4527,
+      "step": 7580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.70317989370327e-05,
+      "loss": 0.4523,
+      "step": 7590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.702729483830286e-05,
+      "loss": 0.4528,
+      "step": 7600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7022790739573014e-05,
+      "loss": 0.4479,
+      "step": 7610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7018286640843164e-05,
+      "loss": 0.4542,
+      "step": 7620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.701378254211333e-05,
+      "loss": 0.4498,
+      "step": 7630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.700927844338348e-05,
+      "loss": 0.4534,
+      "step": 7640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.700477434465364e-05,
+      "loss": 0.4505,
+      "step": 7650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7000270245923794e-05,
+      "loss": 0.445,
+      "step": 7660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6995766147193945e-05,
+      "loss": 0.4439,
+      "step": 7670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.69912620484641e-05,
+      "loss": 0.4478,
+      "step": 7680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.698675794973426e-05,
+      "loss": 0.4442,
+      "step": 7690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.698225385100442e-05,
+      "loss": 0.4467,
+      "step": 7700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6977749752274575e-05,
+      "loss": 0.4457,
+      "step": 7710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6973245653544726e-05,
+      "loss": 0.4461,
+      "step": 7720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.696874155481488e-05,
+      "loss": 0.4501,
+      "step": 7730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.696423745608504e-05,
+      "loss": 0.4508,
+      "step": 7740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.69597333573552e-05,
+      "loss": 0.4473,
+      "step": 7750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.695522925862535e-05,
+      "loss": 0.4528,
+      "step": 7760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6950725159895507e-05,
+      "loss": 0.4446,
+      "step": 7770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6946221061165664e-05,
+      "loss": 0.4406,
+      "step": 7780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6941716962435815e-05,
+      "loss": 0.4438,
+      "step": 7790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.693721286370598e-05,
+      "loss": 0.446,
+      "step": 7800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.693270876497613e-05,
+      "loss": 0.4494,
+      "step": 7810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.692820466624628e-05,
+      "loss": 0.4474,
+      "step": 7820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6923700567516445e-05,
+      "loss": 0.4492,
+      "step": 7830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6919196468786595e-05,
+      "loss": 0.4432,
+      "step": 7840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.691469237005676e-05,
+      "loss": 0.449,
+      "step": 7850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.691018827132691e-05,
+      "loss": 0.4399,
+      "step": 7860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.690568417259706e-05,
+      "loss": 0.4467,
+      "step": 7870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6901180073867225e-05,
+      "loss": 0.4426,
+      "step": 7880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6896675975137376e-05,
+      "loss": 0.447,
+      "step": 7890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6892171876407534e-05,
+      "loss": 0.4474,
+      "step": 7900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.688766777767769e-05,
+      "loss": 0.44,
+      "step": 7910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.688316367894784e-05,
+      "loss": 0.4416,
+      "step": 7920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6878659580218e-05,
+      "loss": 0.4417,
+      "step": 7930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.687415548148816e-05,
+      "loss": 0.4425,
+      "step": 7940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6869651382758314e-05,
+      "loss": 0.4444,
+      "step": 7950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6865147284028465e-05,
+      "loss": 0.4454,
+      "step": 7960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.686064318529862e-05,
+      "loss": 0.4463,
+      "step": 7970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.685613908656878e-05,
+      "loss": 0.4432,
+      "step": 7980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.685163498783894e-05,
+      "loss": 0.4375,
+      "step": 7990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6847130889109095e-05,
+      "loss": 0.4416,
+      "step": 8000
+    },
+    {
+      "epoch": 0.07,
+      "eval_NEIMS_canon_smiles": 0.9101666666666667,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.0007,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.3176162209335286,
+      "eval_NEIMS_exact_mols": 0.0006833333333333333,
+      "eval_NEIMS_exact_smiles": 0.0006,
+      "eval_NEIMS_loss": 0.4383198618888855,
+      "eval_NEIMS_matched_formulas": 0.1968,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.21320042323766042,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.0006833333333333333,
+      "eval_NEIMS_runtime": 742.0808,
+      "eval_NEIMS_samples_per_second": 80.854,
+      "eval_NEIMS_steps_per_second": 1.264,
+      "step": 8000
+    },
+    {
+      "epoch": 0.07,
+      "eval_RASSP_canon_smiles": 0.9061842588573146,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.001402477710622099,
+      "eval_RASSP_daylight_tanimoto_simil": 0.3428026749410063,
+      "eval_RASSP_exact_mols": 0.001402477710622099,
+      "eval_RASSP_exact_smiles": 0.0012856045680702574,
+      "eval_RASSP_loss": 0.41522035002708435,
+      "eval_RASSP_matched_formulas": 0.23663472134103583,
+      "eval_RASSP_morgan_tanimoto_simil": 0.2166757279880561,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.0014191738738437906,
+      "eval_RASSP_runtime": 810.9619,
+      "eval_RASSP_samples_per_second": 73.856,
+      "eval_RASSP_steps_per_second": 1.154,
+      "step": 8000
+    },
+    {
+      "epoch": 0.07,
+      "eval_NIST_canon_smiles": 0.861340809880399,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.0008517585264577492,
+      "eval_NIST_daylight_tanimoto_simil": 0.20700074836664745,
+      "eval_NIST_exact_mols": 0.0008162685878553431,
+      "eval_NIST_exact_smiles": 0.0007807786492529368,
+      "eval_NIST_loss": 1.4345619678497314,
+      "eval_NIST_matched_formulas": 0.02083259395961245,
+      "eval_NIST_morgan_tanimoto_simil": 0.16496379689687352,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.0008162685878553431,
+      "eval_NIST_runtime": 531.7761,
+      "eval_NIST_samples_per_second": 52.987,
+      "eval_NIST_steps_per_second": 0.829,
+      "step": 8000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6842626790379246e-05,
+      "loss": 0.4428,
+      "step": 8010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.68381226916494e-05,
+      "loss": 0.4477,
+      "step": 8020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.683361859291956e-05,
+      "loss": 0.4491,
+      "step": 8030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.682911449418971e-05,
+      "loss": 0.4455,
+      "step": 8040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6824610395459876e-05,
+      "loss": 0.437,
+      "step": 8050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6820106296730026e-05,
+      "loss": 0.4461,
+      "step": 8060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.681560219800018e-05,
+      "loss": 0.4465,
+      "step": 8070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.681109809927034e-05,
+      "loss": 0.4435,
+      "step": 8080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.680659400054049e-05,
+      "loss": 0.4386,
+      "step": 8090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.680208990181065e-05,
+      "loss": 0.4474,
+      "step": 8100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.679758580308081e-05,
+      "loss": 0.449,
+      "step": 8110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.679308170435096e-05,
+      "loss": 0.4357,
+      "step": 8120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.678857760562112e-05,
+      "loss": 0.4312,
+      "step": 8130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.678407350689127e-05,
+      "loss": 0.4414,
+      "step": 8140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.677956940816143e-05,
+      "loss": 0.4379,
+      "step": 8150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.677506530943159e-05,
+      "loss": 0.4408,
+      "step": 8160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.677056121070174e-05,
+      "loss": 0.4408,
+      "step": 8170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6766057111971896e-05,
+      "loss": 0.4368,
+      "step": 8180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6761553013242053e-05,
+      "loss": 0.4357,
+      "step": 8190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.675704891451221e-05,
+      "loss": 0.4416,
+      "step": 8200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.675254481578236e-05,
+      "loss": 0.4376,
+      "step": 8210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.674804071705252e-05,
+      "loss": 0.4423,
+      "step": 8220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.674353661832268e-05,
+      "loss": 0.4399,
+      "step": 8230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.673903251959283e-05,
+      "loss": 0.4386,
+      "step": 8240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.673452842086299e-05,
+      "loss": 0.4469,
+      "step": 8250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.673002432213314e-05,
+      "loss": 0.4382,
+      "step": 8260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.67255202234033e-05,
+      "loss": 0.441,
+      "step": 8270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.672101612467346e-05,
+      "loss": 0.4388,
+      "step": 8280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.671651202594361e-05,
+      "loss": 0.4379,
+      "step": 8290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.671200792721377e-05,
+      "loss": 0.4457,
+      "step": 8300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.670750382848392e-05,
+      "loss": 0.4402,
+      "step": 8310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6702999729754074e-05,
+      "loss": 0.4349,
+      "step": 8320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.669849563102424e-05,
+      "loss": 0.4348,
+      "step": 8330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.669399153229439e-05,
+      "loss": 0.444,
+      "step": 8340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6689487433564546e-05,
+      "loss": 0.4368,
+      "step": 8350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6684983334834704e-05,
+      "loss": 0.4306,
+      "step": 8360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.6680479236104854e-05,
+      "loss": 0.4403,
+      "step": 8370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.667597513737501e-05,
+      "loss": 0.4349,
+      "step": 8380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.667147103864517e-05,
+      "loss": 0.44,
+      "step": 8390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.666696693991533e-05,
+      "loss": 0.4341,
+      "step": 8400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6662462841185484e-05,
+      "loss": 0.4396,
+      "step": 8410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6657958742455635e-05,
+      "loss": 0.4365,
+      "step": 8420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.665345464372579e-05,
+      "loss": 0.4437,
+      "step": 8430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.664895054499595e-05,
+      "loss": 0.4389,
+      "step": 8440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.664444644626611e-05,
+      "loss": 0.4344,
+      "step": 8450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.663994234753626e-05,
+      "loss": 0.436,
+      "step": 8460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6635438248806416e-05,
+      "loss": 0.4299,
+      "step": 8470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.663093415007657e-05,
+      "loss": 0.4394,
+      "step": 8480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6626430051346724e-05,
+      "loss": 0.4458,
+      "step": 8490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.662192595261689e-05,
+      "loss": 0.4399,
+      "step": 8500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.661742185388704e-05,
+      "loss": 0.4387,
+      "step": 8510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.661291775515719e-05,
+      "loss": 0.4362,
+      "step": 8520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6608413656427354e-05,
+      "loss": 0.4389,
+      "step": 8530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6603909557697505e-05,
+      "loss": 0.4324,
+      "step": 8540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.659940545896766e-05,
+      "loss": 0.4322,
+      "step": 8550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.659490136023782e-05,
+      "loss": 0.434,
+      "step": 8560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.659039726150797e-05,
+      "loss": 0.4264,
+      "step": 8570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6585893162778135e-05,
+      "loss": 0.4354,
+      "step": 8580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6581389064048285e-05,
+      "loss": 0.4372,
+      "step": 8590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.657688496531844e-05,
+      "loss": 0.4346,
+      "step": 8600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.65723808665886e-05,
+      "loss": 0.4328,
+      "step": 8610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.656787676785875e-05,
+      "loss": 0.4304,
+      "step": 8620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.656337266912891e-05,
+      "loss": 0.426,
+      "step": 8630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6558868570399066e-05,
+      "loss": 0.4298,
+      "step": 8640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6554364471669224e-05,
+      "loss": 0.4302,
+      "step": 8650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6549860372939374e-05,
+      "loss": 0.437,
+      "step": 8660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.654535627420953e-05,
+      "loss": 0.4311,
+      "step": 8670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.654085217547969e-05,
+      "loss": 0.4265,
+      "step": 8680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.653634807674985e-05,
+      "loss": 0.4339,
+      "step": 8690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6531843978020004e-05,
+      "loss": 0.427,
+      "step": 8700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6527339879290155e-05,
+      "loss": 0.4301,
+      "step": 8710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.652283578056031e-05,
+      "loss": 0.4358,
+      "step": 8720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.651833168183047e-05,
+      "loss": 0.4322,
+      "step": 8730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.651382758310062e-05,
+      "loss": 0.4275,
+      "step": 8740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.650932348437078e-05,
+      "loss": 0.4339,
+      "step": 8750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6504819385640936e-05,
+      "loss": 0.4276,
+      "step": 8760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6500315286911086e-05,
+      "loss": 0.4348,
+      "step": 8770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.649581118818125e-05,
+      "loss": 0.4317,
+      "step": 8780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.64913070894514e-05,
+      "loss": 0.429,
+      "step": 8790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.648680299072156e-05,
+      "loss": 0.43,
+      "step": 8800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6482298891991716e-05,
+      "loss": 0.4343,
+      "step": 8810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.647779479326187e-05,
+      "loss": 0.4364,
+      "step": 8820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.647329069453203e-05,
+      "loss": 0.4291,
+      "step": 8830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.646878659580218e-05,
+      "loss": 0.4254,
+      "step": 8840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.646428249707234e-05,
+      "loss": 0.4275,
+      "step": 8850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.64597783983425e-05,
+      "loss": 0.4303,
+      "step": 8860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.645527429961265e-05,
+      "loss": 0.4389,
+      "step": 8870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6450770200882805e-05,
+      "loss": 0.4288,
+      "step": 8880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.644626610215296e-05,
+      "loss": 0.4287,
+      "step": 8890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.644176200342312e-05,
+      "loss": 0.4321,
+      "step": 8900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.643725790469327e-05,
+      "loss": 0.4184,
+      "step": 8910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.643275380596343e-05,
+      "loss": 0.4287,
+      "step": 8920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6428249707233586e-05,
+      "loss": 0.4309,
+      "step": 8930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6423745608503737e-05,
+      "loss": 0.4304,
+      "step": 8940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.64192415097739e-05,
+      "loss": 0.4299,
+      "step": 8950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.641473741104405e-05,
+      "loss": 0.4312,
+      "step": 8960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.641023331231421e-05,
+      "loss": 0.4333,
+      "step": 8970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6405729213584367e-05,
+      "loss": 0.4306,
+      "step": 8980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.640122511485452e-05,
+      "loss": 0.4217,
+      "step": 8990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6396721016124675e-05,
+      "loss": 0.4264,
+      "step": 9000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.639221691739483e-05,
+      "loss": 0.4296,
+      "step": 9010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.638771281866498e-05,
+      "loss": 0.4258,
+      "step": 9020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.638320871993515e-05,
+      "loss": 0.4363,
+      "step": 9030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.63787046212053e-05,
+      "loss": 0.4255,
+      "step": 9040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6374200522475455e-05,
+      "loss": 0.422,
+      "step": 9050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.636969642374561e-05,
+      "loss": 0.4235,
+      "step": 9060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6365192325015764e-05,
+      "loss": 0.4239,
+      "step": 9070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.636068822628592e-05,
+      "loss": 0.435,
+      "step": 9080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.635618412755608e-05,
+      "loss": 0.4327,
+      "step": 9090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6351680028826236e-05,
+      "loss": 0.4259,
+      "step": 9100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6347175930096394e-05,
+      "loss": 0.43,
+      "step": 9110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6342671831366544e-05,
+      "loss": 0.4256,
+      "step": 9120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.63381677326367e-05,
+      "loss": 0.4322,
+      "step": 9130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.633366363390686e-05,
+      "loss": 0.4324,
+      "step": 9140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.632915953517702e-05,
+      "loss": 0.4259,
+      "step": 9150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.632465543644717e-05,
+      "loss": 0.4284,
+      "step": 9160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6320151337717325e-05,
+      "loss": 0.4245,
+      "step": 9170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.631564723898748e-05,
+      "loss": 0.4294,
+      "step": 9180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.631114314025763e-05,
+      "loss": 0.4311,
+      "step": 9190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.630663904152779e-05,
+      "loss": 0.4211,
+      "step": 9200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.630213494279795e-05,
+      "loss": 0.4243,
+      "step": 9210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.62976308440681e-05,
+      "loss": 0.4251,
+      "step": 9220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.629312674533826e-05,
+      "loss": 0.4349,
+      "step": 9230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6288622646608414e-05,
+      "loss": 0.4237,
+      "step": 9240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.628411854787857e-05,
+      "loss": 0.4172,
+      "step": 9250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.627961444914873e-05,
+      "loss": 0.424,
+      "step": 9260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.627511035041888e-05,
+      "loss": 0.4256,
+      "step": 9270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6270606251689044e-05,
+      "loss": 0.4252,
+      "step": 9280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6266102152959195e-05,
+      "loss": 0.4235,
+      "step": 9290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.626159805422935e-05,
+      "loss": 0.4206,
+      "step": 9300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.625709395549951e-05,
+      "loss": 0.4221,
+      "step": 9310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.625258985676966e-05,
+      "loss": 0.4207,
+      "step": 9320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.624808575803982e-05,
+      "loss": 0.4245,
+      "step": 9330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6243581659309975e-05,
+      "loss": 0.4178,
+      "step": 9340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.623907756058013e-05,
+      "loss": 0.4159,
+      "step": 9350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6234573461850283e-05,
+      "loss": 0.4194,
+      "step": 9360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.623006936312044e-05,
+      "loss": 0.4239,
+      "step": 9370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.62255652643906e-05,
+      "loss": 0.4261,
+      "step": 9380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6221061165660756e-05,
+      "loss": 0.4186,
+      "step": 9390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.621655706693091e-05,
+      "loss": 0.4261,
+      "step": 9400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6212052968201064e-05,
+      "loss": 0.4197,
+      "step": 9410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.620754886947122e-05,
+      "loss": 0.4327,
+      "step": 9420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.620304477074138e-05,
+      "loss": 0.4267,
+      "step": 9430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.619854067201153e-05,
+      "loss": 0.4204,
+      "step": 9440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.619403657328169e-05,
+      "loss": 0.424,
+      "step": 9450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6189532474551845e-05,
+      "loss": 0.4207,
+      "step": 9460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6185028375821996e-05,
+      "loss": 0.4277,
+      "step": 9470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.618052427709216e-05,
+      "loss": 0.4201,
+      "step": 9480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.617602017836231e-05,
+      "loss": 0.4256,
+      "step": 9490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.617151607963247e-05,
+      "loss": 0.423,
+      "step": 9500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6167011980902626e-05,
+      "loss": 0.4301,
+      "step": 9510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.6162507882172776e-05,
+      "loss": 0.4175,
+      "step": 9520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.615800378344294e-05,
+      "loss": 0.4238,
+      "step": 9530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.615349968471309e-05,
+      "loss": 0.4174,
+      "step": 9540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.614899558598325e-05,
+      "loss": 0.4243,
+      "step": 9550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6144491487253406e-05,
+      "loss": 0.4269,
+      "step": 9560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.613998738852356e-05,
+      "loss": 0.4243,
+      "step": 9570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6135483289793714e-05,
+      "loss": 0.4174,
+      "step": 9580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.613097919106387e-05,
+      "loss": 0.4191,
+      "step": 9590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.612647509233403e-05,
+      "loss": 0.4187,
+      "step": 9600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.612197099360418e-05,
+      "loss": 0.4266,
+      "step": 9610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.611746689487434e-05,
+      "loss": 0.4163,
+      "step": 9620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6112962796144495e-05,
+      "loss": 0.4208,
+      "step": 9630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6108458697414646e-05,
+      "loss": 0.421,
+      "step": 9640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.61039545986848e-05,
+      "loss": 0.4181,
+      "step": 9650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.609945049995496e-05,
+      "loss": 0.421,
+      "step": 9660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.609494640122512e-05,
+      "loss": 0.4275,
+      "step": 9670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6090442302495276e-05,
+      "loss": 0.4141,
+      "step": 9680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6085938203765427e-05,
+      "loss": 0.425,
+      "step": 9690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6081434105035584e-05,
+      "loss": 0.4203,
+      "step": 9700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.607693000630574e-05,
+      "loss": 0.418,
+      "step": 9710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.607242590757589e-05,
+      "loss": 0.4144,
+      "step": 9720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6067921808846056e-05,
+      "loss": 0.4161,
+      "step": 9730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.606341771011621e-05,
+      "loss": 0.4209,
+      "step": 9740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6058913611386365e-05,
+      "loss": 0.4216,
+      "step": 9750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.605440951265652e-05,
+      "loss": 0.4177,
+      "step": 9760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.604990541392667e-05,
+      "loss": 0.4215,
+      "step": 9770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.604540131519683e-05,
+      "loss": 0.413,
+      "step": 9780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.604089721646699e-05,
+      "loss": 0.4112,
+      "step": 9790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6036393117737145e-05,
+      "loss": 0.4181,
+      "step": 9800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.60318890190073e-05,
+      "loss": 0.4139,
+      "step": 9810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6027384920277454e-05,
+      "loss": 0.4213,
+      "step": 9820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.602288082154761e-05,
+      "loss": 0.4175,
+      "step": 9830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.601837672281777e-05,
+      "loss": 0.4179,
+      "step": 9840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.601387262408792e-05,
+      "loss": 0.4186,
+      "step": 9850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.600936852535808e-05,
+      "loss": 0.4142,
+      "step": 9860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6004864426628234e-05,
+      "loss": 0.4213,
+      "step": 9870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.600036032789839e-05,
+      "loss": 0.4221,
+      "step": 9880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.599585622916854e-05,
+      "loss": 0.4134,
+      "step": 9890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.59913521304387e-05,
+      "loss": 0.4151,
+      "step": 9900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.598684803170886e-05,
+      "loss": 0.4173,
+      "step": 9910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.598234393297901e-05,
+      "loss": 0.4174,
+      "step": 9920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.597783983424917e-05,
+      "loss": 0.4229,
+      "step": 9930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.597333573551932e-05,
+      "loss": 0.412,
+      "step": 9940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.596883163678948e-05,
+      "loss": 0.4179,
+      "step": 9950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.596432753805964e-05,
+      "loss": 0.4109,
+      "step": 9960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.595982343932979e-05,
+      "loss": 0.4129,
+      "step": 9970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.595531934059995e-05,
+      "loss": 0.4156,
+      "step": 9980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5950815241870104e-05,
+      "loss": 0.4174,
+      "step": 9990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.594631114314026e-05,
+      "loss": 0.4106,
+      "step": 10000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.594180704441042e-05,
+      "loss": 0.4197,
+      "step": 10010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.593730294568057e-05,
+      "loss": 0.4133,
+      "step": 10020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.593279884695073e-05,
+      "loss": 0.4083,
+      "step": 10030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5928294748220885e-05,
+      "loss": 0.4184,
+      "step": 10040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5923790649491035e-05,
+      "loss": 0.4168,
+      "step": 10050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.591973696063418e-05,
+      "loss": 0.4122,
+      "step": 10060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5915232861904337e-05,
+      "loss": 0.4192,
+      "step": 10070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5910728763174494e-05,
+      "loss": 0.4138,
+      "step": 10080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5906224664444645e-05,
+      "loss": 0.4146,
+      "step": 10090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.59017205657148e-05,
+      "loss": 0.4114,
+      "step": 10100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.589721646698496e-05,
+      "loss": 0.4186,
+      "step": 10110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.589271236825512e-05,
+      "loss": 0.4192,
+      "step": 10120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.588820826952527e-05,
+      "loss": 0.4254,
+      "step": 10130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5883704170795425e-05,
+      "loss": 0.4187,
+      "step": 10140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.587920007206558e-05,
+      "loss": 0.4155,
+      "step": 10150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5874695973335734e-05,
+      "loss": 0.4127,
+      "step": 10160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.58701918746059e-05,
+      "loss": 0.4118,
+      "step": 10170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.586568777587605e-05,
+      "loss": 0.4147,
+      "step": 10180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5861183677146206e-05,
+      "loss": 0.4157,
+      "step": 10190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5856679578416364e-05,
+      "loss": 0.4113,
+      "step": 10200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5852175479686514e-05,
+      "loss": 0.4162,
+      "step": 10210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.584767138095667e-05,
+      "loss": 0.4091,
+      "step": 10220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.584316728222683e-05,
+      "loss": 0.4215,
+      "step": 10230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.583866318349698e-05,
+      "loss": 0.4153,
+      "step": 10240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5834159084767144e-05,
+      "loss": 0.4196,
+      "step": 10250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5829654986037295e-05,
+      "loss": 0.4186,
+      "step": 10260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.582515088730745e-05,
+      "loss": 0.4124,
+      "step": 10270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.582064678857761e-05,
+      "loss": 0.4131,
+      "step": 10280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.581614268984776e-05,
+      "loss": 0.4109,
+      "step": 10290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.581163859111792e-05,
+      "loss": 0.4065,
+      "step": 10300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5807134492388076e-05,
+      "loss": 0.411,
+      "step": 10310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.580263039365823e-05,
+      "loss": 0.412,
+      "step": 10320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.579812629492839e-05,
+      "loss": 0.4125,
+      "step": 10330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.579362219619854e-05,
+      "loss": 0.4103,
+      "step": 10340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.57891180974687e-05,
+      "loss": 0.414,
+      "step": 10350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5784613998738856e-05,
+      "loss": 0.41,
+      "step": 10360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5780109900009014e-05,
+      "loss": 0.4108,
+      "step": 10370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5775605801279165e-05,
+      "loss": 0.4198,
+      "step": 10380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.577110170254932e-05,
+      "loss": 0.4196,
+      "step": 10390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.576659760381948e-05,
+      "loss": 0.4072,
+      "step": 10400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.576209350508963e-05,
+      "loss": 0.4127,
+      "step": 10410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.575758940635979e-05,
+      "loss": 0.4103,
+      "step": 10420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5753085307629945e-05,
+      "loss": 0.4104,
+      "step": 10430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5748581208900096e-05,
+      "loss": 0.4116,
+      "step": 10440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.574407711017026e-05,
+      "loss": 0.4097,
+      "step": 10450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.573957301144041e-05,
+      "loss": 0.4162,
+      "step": 10460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.573506891271057e-05,
+      "loss": 0.4091,
+      "step": 10470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5730564813980726e-05,
+      "loss": 0.4086,
+      "step": 10480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.572606071525088e-05,
+      "loss": 0.4175,
+      "step": 10490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.572155661652104e-05,
+      "loss": 0.416,
+      "step": 10500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.571705251779119e-05,
+      "loss": 0.4084,
+      "step": 10510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.571254841906135e-05,
+      "loss": 0.4102,
+      "step": 10520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.570804432033151e-05,
+      "loss": 0.4111,
+      "step": 10530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.570354022160166e-05,
+      "loss": 0.41,
+      "step": 10540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5699036122871815e-05,
+      "loss": 0.4059,
+      "step": 10550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.569453202414197e-05,
+      "loss": 0.4127,
+      "step": 10560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.569002792541213e-05,
+      "loss": 0.4124,
+      "step": 10570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.568552382668228e-05,
+      "loss": 0.4074,
+      "step": 10580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.568101972795244e-05,
+      "loss": 0.4051,
+      "step": 10590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5676515629222596e-05,
+      "loss": 0.4012,
+      "step": 10600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.567201153049275e-05,
+      "loss": 0.3994,
+      "step": 10610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5667507431762904e-05,
+      "loss": 0.4108,
+      "step": 10620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.566300333303306e-05,
+      "loss": 0.4074,
+      "step": 10630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.565849923430322e-05,
+      "loss": 0.403,
+      "step": 10640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5653995135573376e-05,
+      "loss": 0.4043,
+      "step": 10650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.564949103684353e-05,
+      "loss": 0.4067,
+      "step": 10660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5644986938113684e-05,
+      "loss": 0.4113,
+      "step": 10670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.564048283938384e-05,
+      "loss": 0.4099,
+      "step": 10680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.563597874065399e-05,
+      "loss": 0.4096,
+      "step": 10690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.563147464192416e-05,
+      "loss": 0.4099,
+      "step": 10700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.562697054319431e-05,
+      "loss": 0.4147,
+      "step": 10710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5622466444464465e-05,
+      "loss": 0.4083,
+      "step": 10720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.561796234573462e-05,
+      "loss": 0.4085,
+      "step": 10730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.561345824700477e-05,
+      "loss": 0.4028,
+      "step": 10740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.560895414827494e-05,
+      "loss": 0.4065,
+      "step": 10750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.560445004954509e-05,
+      "loss": 0.4139,
+      "step": 10760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5599945950815246e-05,
+      "loss": 0.409,
+      "step": 10770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.55954418520854e-05,
+      "loss": 0.4004,
+      "step": 10780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5590937753355554e-05,
+      "loss": 0.3992,
+      "step": 10790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.558643365462571e-05,
+      "loss": 0.4009,
+      "step": 10800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.558192955589587e-05,
+      "loss": 0.4141,
+      "step": 10810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.557742545716602e-05,
+      "loss": 0.4048,
+      "step": 10820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.557292135843618e-05,
+      "loss": 0.4016,
+      "step": 10830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5568417259706335e-05,
+      "loss": 0.4098,
+      "step": 10840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.556391316097649e-05,
+      "loss": 0.4073,
+      "step": 10850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.555940906224664e-05,
+      "loss": 0.4088,
+      "step": 10860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.55549049635168e-05,
+      "loss": 0.413,
+      "step": 10870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.555040086478696e-05,
+      "loss": 0.4073,
+      "step": 10880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5545896766057115e-05,
+      "loss": 0.4025,
+      "step": 10890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.554139266732727e-05,
+      "loss": 0.4116,
+      "step": 10900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5536888568597424e-05,
+      "loss": 0.3997,
+      "step": 10910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.553238446986758e-05,
+      "loss": 0.4063,
+      "step": 10920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.552788037113774e-05,
+      "loss": 0.4013,
+      "step": 10930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.552337627240789e-05,
+      "loss": 0.4054,
+      "step": 10940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5518872173678054e-05,
+      "loss": 0.4122,
+      "step": 10950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5514368074948204e-05,
+      "loss": 0.3989,
+      "step": 10960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.550986397621836e-05,
+      "loss": 0.4,
+      "step": 10970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.550535987748852e-05,
+      "loss": 0.4072,
+      "step": 10980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.550085577875867e-05,
+      "loss": 0.4066,
+      "step": 10990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.549635168002883e-05,
+      "loss": 0.407,
+      "step": 11000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5491847581298985e-05,
+      "loss": 0.4013,
+      "step": 11010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.548734348256914e-05,
+      "loss": 0.4019,
+      "step": 11020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.54828393838393e-05,
+      "loss": 0.4017,
+      "step": 11030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.547833528510945e-05,
+      "loss": 0.4031,
+      "step": 11040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.547383118637961e-05,
+      "loss": 0.4032,
+      "step": 11050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5469327087649766e-05,
+      "loss": 0.4068,
+      "step": 11060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5464822988919916e-05,
+      "loss": 0.4055,
+      "step": 11070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5460318890190074e-05,
+      "loss": 0.4014,
+      "step": 11080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.545581479146023e-05,
+      "loss": 0.4062,
+      "step": 11090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.545131069273039e-05,
+      "loss": 0.3928,
+      "step": 11100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.544680659400054e-05,
+      "loss": 0.4021,
+      "step": 11110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.54423024952707e-05,
+      "loss": 0.3974,
+      "step": 11120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5437798396540855e-05,
+      "loss": 0.3969,
+      "step": 11130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5433294297811005e-05,
+      "loss": 0.4056,
+      "step": 11140
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.542879019908117e-05,
+      "loss": 0.4166,
+      "step": 11150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.542428610035132e-05,
+      "loss": 0.4086,
+      "step": 11160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.541978200162148e-05,
+      "loss": 0.4073,
+      "step": 11170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5415277902891635e-05,
+      "loss": 0.406,
+      "step": 11180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5410773804161786e-05,
+      "loss": 0.404,
+      "step": 11190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.540626970543195e-05,
+      "loss": 0.4084,
+      "step": 11200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.54017656067021e-05,
+      "loss": 0.3998,
+      "step": 11210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.539726150797226e-05,
+      "loss": 0.4024,
+      "step": 11220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5392757409242416e-05,
+      "loss": 0.401,
+      "step": 11230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5388253310512567e-05,
+      "loss": 0.3996,
+      "step": 11240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5383749211782724e-05,
+      "loss": 0.3962,
+      "step": 11250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.537924511305288e-05,
+      "loss": 0.3998,
+      "step": 11260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.537474101432303e-05,
+      "loss": 0.3991,
+      "step": 11270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.537023691559319e-05,
+      "loss": 0.4011,
+      "step": 11280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.536573281686335e-05,
+      "loss": 0.4039,
+      "step": 11290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5361228718133505e-05,
+      "loss": 0.3988,
+      "step": 11300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.535672461940366e-05,
+      "loss": 0.406,
+      "step": 11310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.535222052067381e-05,
+      "loss": 0.3949,
+      "step": 11320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.534771642194397e-05,
+      "loss": 0.3954,
+      "step": 11330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.534321232321413e-05,
+      "loss": 0.3944,
+      "step": 11340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5338708224484285e-05,
+      "loss": 0.4005,
+      "step": 11350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5334204125754436e-05,
+      "loss": 0.3951,
+      "step": 11360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5329700027024594e-05,
+      "loss": 0.3957,
+      "step": 11370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.532519592829475e-05,
+      "loss": 0.3995,
+      "step": 11380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.53206918295649e-05,
+      "loss": 0.4007,
+      "step": 11390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5316187730835066e-05,
+      "loss": 0.3968,
+      "step": 11400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.531168363210522e-05,
+      "loss": 0.4015,
+      "step": 11410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5307179533375374e-05,
+      "loss": 0.4037,
+      "step": 11420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.530267543464553e-05,
+      "loss": 0.3992,
+      "step": 11430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.529817133591568e-05,
+      "loss": 0.3987,
+      "step": 11440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.529366723718585e-05,
+      "loss": 0.3972,
+      "step": 11450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5289163138456e-05,
+      "loss": 0.4025,
+      "step": 11460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.528465903972615e-05,
+      "loss": 0.3964,
+      "step": 11470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.528015494099631e-05,
+      "loss": 0.4009,
+      "step": 11480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.527565084226646e-05,
+      "loss": 0.3963,
+      "step": 11490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.527114674353662e-05,
+      "loss": 0.3987,
+      "step": 11500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.526664264480678e-05,
+      "loss": 0.3989,
+      "step": 11510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.526213854607693e-05,
+      "loss": 0.397,
+      "step": 11520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5257634447347086e-05,
+      "loss": 0.3993,
+      "step": 11530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5253130348617244e-05,
+      "loss": 0.4045,
+      "step": 11540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.52486262498874e-05,
+      "loss": 0.4007,
+      "step": 11550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.524412215115755e-05,
+      "loss": 0.4013,
+      "step": 11560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.523961805242771e-05,
+      "loss": 0.3975,
+      "step": 11570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.523511395369787e-05,
+      "loss": 0.3981,
+      "step": 11580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5230609854968025e-05,
+      "loss": 0.3975,
+      "step": 11590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.522610575623818e-05,
+      "loss": 0.4031,
+      "step": 11600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.522160165750833e-05,
+      "loss": 0.3976,
+      "step": 11610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.521709755877849e-05,
+      "loss": 0.4013,
+      "step": 11620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.521259346004865e-05,
+      "loss": 0.3989,
+      "step": 11630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.52080893613188e-05,
+      "loss": 0.3949,
+      "step": 11640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.520358526258896e-05,
+      "loss": 0.4069,
+      "step": 11650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5199081163859114e-05,
+      "loss": 0.389,
+      "step": 11660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.519457706512927e-05,
+      "loss": 0.3993,
+      "step": 11670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.519007296639943e-05,
+      "loss": 0.3866,
+      "step": 11680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.518556886766958e-05,
+      "loss": 0.4033,
+      "step": 11690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.518106476893974e-05,
+      "loss": 0.399,
+      "step": 11700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5176560670209894e-05,
+      "loss": 0.393,
+      "step": 11710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5172056571480045e-05,
+      "loss": 0.3923,
+      "step": 11720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.516755247275021e-05,
+      "loss": 0.3892,
+      "step": 11730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.516304837402036e-05,
+      "loss": 0.396,
+      "step": 11740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.515854427529052e-05,
+      "loss": 0.4043,
+      "step": 11750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.5154040176560675e-05,
+      "loss": 0.3955,
+      "step": 11760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5149536077830826e-05,
+      "loss": 0.4022,
+      "step": 11770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.514503197910098e-05,
+      "loss": 0.4031,
+      "step": 11780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.514052788037114e-05,
+      "loss": 0.3997,
+      "step": 11790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.51360237816413e-05,
+      "loss": 0.3949,
+      "step": 11800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.513151968291145e-05,
+      "loss": 0.3939,
+      "step": 11810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5127015584181606e-05,
+      "loss": 0.3911,
+      "step": 11820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5122511485451764e-05,
+      "loss": 0.3927,
+      "step": 11830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5118007386721914e-05,
+      "loss": 0.3907,
+      "step": 11840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.511350328799208e-05,
+      "loss": 0.3871,
+      "step": 11850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.510899918926223e-05,
+      "loss": 0.3999,
+      "step": 11860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.510449509053239e-05,
+      "loss": 0.3958,
+      "step": 11870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5099990991802544e-05,
+      "loss": 0.3957,
+      "step": 11880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5095486893072695e-05,
+      "loss": 0.3987,
+      "step": 11890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.509098279434286e-05,
+      "loss": 0.3993,
+      "step": 11900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.508647869561301e-05,
+      "loss": 0.3859,
+      "step": 11910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.508197459688316e-05,
+      "loss": 0.3927,
+      "step": 11920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5077470498153325e-05,
+      "loss": 0.3943,
+      "step": 11930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5072966399423476e-05,
+      "loss": 0.3943,
+      "step": 11940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.506846230069363e-05,
+      "loss": 0.3894,
+      "step": 11950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.506395820196379e-05,
+      "loss": 0.3881,
+      "step": 11960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.505945410323394e-05,
+      "loss": 0.3908,
+      "step": 11970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.50549500045041e-05,
+      "loss": 0.3999,
+      "step": 11980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5050445905774257e-05,
+      "loss": 0.3927,
+      "step": 11990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5045941807044414e-05,
+      "loss": 0.3952,
+      "step": 12000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.504143770831457e-05,
+      "loss": 0.3894,
+      "step": 12010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.503693360958472e-05,
+      "loss": 0.3979,
+      "step": 12020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.503242951085488e-05,
+      "loss": 0.3965,
+      "step": 12030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.502792541212504e-05,
+      "loss": 0.3978,
+      "step": 12040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5023421313395195e-05,
+      "loss": 0.3971,
+      "step": 12050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.501936762453833e-05,
+      "loss": 0.3899,
+      "step": 12060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.501486352580849e-05,
+      "loss": 0.3998,
+      "step": 12070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.501035942707864e-05,
+      "loss": 0.3976,
+      "step": 12080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.50058553283488e-05,
+      "loss": 0.393,
+      "step": 12090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.5001351229618955e-05,
+      "loss": 0.3926,
+      "step": 12100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.499684713088911e-05,
+      "loss": 0.3937,
+      "step": 12110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.499234303215927e-05,
+      "loss": 0.3834,
+      "step": 12120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.498783893342942e-05,
+      "loss": 0.3944,
+      "step": 12130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.498333483469958e-05,
+      "loss": 0.3928,
+      "step": 12140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4978830735969736e-05,
+      "loss": 0.4,
+      "step": 12150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4974326637239886e-05,
+      "loss": 0.3895,
+      "step": 12160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.496982253851005e-05,
+      "loss": 0.4059,
+      "step": 12170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.49653184397802e-05,
+      "loss": 0.3932,
+      "step": 12180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.496081434105036e-05,
+      "loss": 0.3923,
+      "step": 12190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4956310242320516e-05,
+      "loss": 0.3869,
+      "step": 12200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.495180614359067e-05,
+      "loss": 0.398,
+      "step": 12210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4947302044860824e-05,
+      "loss": 0.3987,
+      "step": 12220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.494279794613098e-05,
+      "loss": 0.3932,
+      "step": 12230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.493829384740114e-05,
+      "loss": 0.3971,
+      "step": 12240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.49337897486713e-05,
+      "loss": 0.3907,
+      "step": 12250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.492928564994145e-05,
+      "loss": 0.391,
+      "step": 12260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4924781551211605e-05,
+      "loss": 0.3926,
+      "step": 12270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.492027745248176e-05,
+      "loss": 0.3855,
+      "step": 12280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.491577335375191e-05,
+      "loss": 0.3925,
+      "step": 12290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.491126925502207e-05,
+      "loss": 0.3927,
+      "step": 12300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.490676515629223e-05,
+      "loss": 0.3919,
+      "step": 12310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4902261057562386e-05,
+      "loss": 0.392,
+      "step": 12320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4897756958832537e-05,
+      "loss": 0.3901,
+      "step": 12330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4893252860102694e-05,
+      "loss": 0.3853,
+      "step": 12340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.488874876137285e-05,
+      "loss": 0.3845,
+      "step": 12350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4884244662643e-05,
+      "loss": 0.3849,
+      "step": 12360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4879740563913167e-05,
+      "loss": 0.3845,
+      "step": 12370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.487523646518332e-05,
+      "loss": 0.3931,
+      "step": 12380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4870732366453475e-05,
+      "loss": 0.3952,
+      "step": 12390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.486622826772363e-05,
+      "loss": 0.3905,
+      "step": 12400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.486172416899378e-05,
+      "loss": 0.3874,
+      "step": 12410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.485722007026395e-05,
+      "loss": 0.3888,
+      "step": 12420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.48527159715341e-05,
+      "loss": 0.3862,
+      "step": 12430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4848211872804255e-05,
+      "loss": 0.3905,
+      "step": 12440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.484370777407441e-05,
+      "loss": 0.3858,
+      "step": 12450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4839203675344564e-05,
+      "loss": 0.3892,
+      "step": 12460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.483469957661472e-05,
+      "loss": 0.3862,
+      "step": 12470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.483019547788488e-05,
+      "loss": 0.3913,
+      "step": 12480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.482569137915503e-05,
+      "loss": 0.3881,
+      "step": 12490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.482118728042519e-05,
+      "loss": 0.3887,
+      "step": 12500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4816683181695344e-05,
+      "loss": 0.3998,
+      "step": 12510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.48121790829655e-05,
+      "loss": 0.3872,
+      "step": 12520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.480767498423566e-05,
+      "loss": 0.3832,
+      "step": 12530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.480317088550581e-05,
+      "loss": 0.3909,
+      "step": 12540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.479866678677597e-05,
+      "loss": 0.3839,
+      "step": 12550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4794162688046125e-05,
+      "loss": 0.3841,
+      "step": 12560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.478965858931628e-05,
+      "loss": 0.3856,
+      "step": 12570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.478515449058643e-05,
+      "loss": 0.3882,
+      "step": 12580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.478065039185659e-05,
+      "loss": 0.3814,
+      "step": 12590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.477614629312675e-05,
+      "loss": 0.3879,
+      "step": 12600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.47716421943969e-05,
+      "loss": 0.3856,
+      "step": 12610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.476713809566706e-05,
+      "loss": 0.3889,
+      "step": 12620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4762633996937214e-05,
+      "loss": 0.394,
+      "step": 12630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.475812989820737e-05,
+      "loss": 0.3823,
+      "step": 12640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.475362579947753e-05,
+      "loss": 0.3898,
+      "step": 12650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.474912170074768e-05,
+      "loss": 0.3932,
+      "step": 12660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4744617602017844e-05,
+      "loss": 0.3882,
+      "step": 12670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4740113503287995e-05,
+      "loss": 0.3881,
+      "step": 12680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4735609404558145e-05,
+      "loss": 0.3901,
+      "step": 12690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.473110530582831e-05,
+      "loss": 0.3834,
+      "step": 12700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.472660120709846e-05,
+      "loss": 0.3892,
+      "step": 12710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.472209710836862e-05,
+      "loss": 0.3877,
+      "step": 12720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4717593009638775e-05,
+      "loss": 0.3842,
+      "step": 12730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4713088910908926e-05,
+      "loss": 0.3932,
+      "step": 12740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4708584812179083e-05,
+      "loss": 0.3888,
+      "step": 12750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.470408071344924e-05,
+      "loss": 0.3854,
+      "step": 12760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.46995766147194e-05,
+      "loss": 0.3887,
+      "step": 12770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.469507251598955e-05,
+      "loss": 0.3969,
+      "step": 12780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.469056841725971e-05,
+      "loss": 0.3828,
+      "step": 12790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4686064318529864e-05,
+      "loss": 0.3808,
+      "step": 12800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.468156021980002e-05,
+      "loss": 0.3869,
+      "step": 12810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.467705612107018e-05,
+      "loss": 0.386,
+      "step": 12820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.467255202234033e-05,
+      "loss": 0.3838,
+      "step": 12830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.466804792361049e-05,
+      "loss": 0.3827,
+      "step": 12840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4663543824880645e-05,
+      "loss": 0.3814,
+      "step": 12850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4659039726150796e-05,
+      "loss": 0.3863,
+      "step": 12860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.465453562742096e-05,
+      "loss": 0.3784,
+      "step": 12870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.465003152869111e-05,
+      "loss": 0.3854,
+      "step": 12880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.464552742996127e-05,
+      "loss": 0.3889,
+      "step": 12890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4641023331231426e-05,
+      "loss": 0.3876,
+      "step": 12900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4636519232501576e-05,
+      "loss": 0.3912,
+      "step": 12910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4632015133771734e-05,
+      "loss": 0.3842,
+      "step": 12920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.462751103504189e-05,
+      "loss": 0.3836,
+      "step": 12930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.462300693631204e-05,
+      "loss": 0.3818,
+      "step": 12940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4618502837582206e-05,
+      "loss": 0.3838,
+      "step": 12950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.461399873885236e-05,
+      "loss": 0.3818,
+      "step": 12960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4609494640122514e-05,
+      "loss": 0.3852,
+      "step": 12970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.460499054139267e-05,
+      "loss": 0.3878,
+      "step": 12980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.460048644266282e-05,
+      "loss": 0.3834,
+      "step": 12990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.459598234393298e-05,
+      "loss": 0.3879,
+      "step": 13000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.459147824520314e-05,
+      "loss": 0.3806,
+      "step": 13010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4586974146473295e-05,
+      "loss": 0.3859,
+      "step": 13020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4582470047743446e-05,
+      "loss": 0.3859,
+      "step": 13030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.45779659490136e-05,
+      "loss": 0.3763,
+      "step": 13040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.457346185028376e-05,
+      "loss": 0.3828,
+      "step": 13050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.456895775155391e-05,
+      "loss": 0.3845,
+      "step": 13060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4564453652824076e-05,
+      "loss": 0.388,
+      "step": 13070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4559949554094227e-05,
+      "loss": 0.3858,
+      "step": 13080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4555445455364384e-05,
+      "loss": 0.3922,
+      "step": 13090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.455094135663454e-05,
+      "loss": 0.3793,
+      "step": 13100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.454643725790469e-05,
+      "loss": 0.3843,
+      "step": 13110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4541933159174856e-05,
+      "loss": 0.3881,
+      "step": 13120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.453742906044501e-05,
+      "loss": 0.387,
+      "step": 13130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.453292496171516e-05,
+      "loss": 0.39,
+      "step": 13140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.452842086298532e-05,
+      "loss": 0.3851,
+      "step": 13150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.452391676425547e-05,
+      "loss": 0.3856,
+      "step": 13160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.451941266552563e-05,
+      "loss": 0.3808,
+      "step": 13170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.451490856679579e-05,
+      "loss": 0.3771,
+      "step": 13180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.451040446806594e-05,
+      "loss": 0.3809,
+      "step": 13190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4505900369336096e-05,
+      "loss": 0.3885,
+      "step": 13200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4501396270606254e-05,
+      "loss": 0.3761,
+      "step": 13210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.449689217187641e-05,
+      "loss": 0.3877,
+      "step": 13220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.449238807314657e-05,
+      "loss": 0.3866,
+      "step": 13230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.448788397441672e-05,
+      "loss": 0.3873,
+      "step": 13240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.448337987568688e-05,
+      "loss": 0.3809,
+      "step": 13250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4478875776957034e-05,
+      "loss": 0.3812,
+      "step": 13260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.447437167822719e-05,
+      "loss": 0.3844,
+      "step": 13270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.446986757949734e-05,
+      "loss": 0.3798,
+      "step": 13280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.44653634807675e-05,
+      "loss": 0.3814,
+      "step": 13290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.446085938203766e-05,
+      "loss": 0.3808,
+      "step": 13300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.445635528330781e-05,
+      "loss": 0.3736,
+      "step": 13310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.445185118457797e-05,
+      "loss": 0.3772,
+      "step": 13320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.444734708584812e-05,
+      "loss": 0.3839,
+      "step": 13330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4442842987118274e-05,
+      "loss": 0.3743,
+      "step": 13340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.443833888838844e-05,
+      "loss": 0.3845,
+      "step": 13350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.443383478965859e-05,
+      "loss": 0.3827,
+      "step": 13360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.442933069092875e-05,
+      "loss": 0.3754,
+      "step": 13370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4424826592198904e-05,
+      "loss": 0.3807,
+      "step": 13380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4420322493469055e-05,
+      "loss": 0.3783,
+      "step": 13390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.441581839473922e-05,
+      "loss": 0.3812,
+      "step": 13400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.441131429600937e-05,
+      "loss": 0.3814,
+      "step": 13410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.440681019727953e-05,
+      "loss": 0.378,
+      "step": 13420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4402306098549685e-05,
+      "loss": 0.3785,
+      "step": 13430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4397801999819835e-05,
+      "loss": 0.3827,
+      "step": 13440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.439329790108999e-05,
+      "loss": 0.3787,
+      "step": 13450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.438879380236015e-05,
+      "loss": 0.3871,
+      "step": 13460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.438428970363031e-05,
+      "loss": 0.3857,
+      "step": 13470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.437978560490046e-05,
+      "loss": 0.3758,
+      "step": 13480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4375281506170616e-05,
+      "loss": 0.3859,
+      "step": 13490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4370777407440773e-05,
+      "loss": 0.382,
+      "step": 13500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.436627330871093e-05,
+      "loss": 0.3843,
+      "step": 13510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.436176920998109e-05,
+      "loss": 0.3798,
+      "step": 13520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.435726511125124e-05,
+      "loss": 0.3803,
+      "step": 13530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.43527610125214e-05,
+      "loss": 0.379,
+      "step": 13540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4348256913791554e-05,
+      "loss": 0.3794,
+      "step": 13550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4343752815061705e-05,
+      "loss": 0.3771,
+      "step": 13560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.433924871633187e-05,
+      "loss": 0.3787,
+      "step": 13570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.433474461760202e-05,
+      "loss": 0.3849,
+      "step": 13580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.433024051887217e-05,
+      "loss": 0.3856,
+      "step": 13590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4325736420142335e-05,
+      "loss": 0.3799,
+      "step": 13600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4321232321412486e-05,
+      "loss": 0.3757,
+      "step": 13610
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.431672822268264e-05,
+      "loss": 0.3806,
+      "step": 13620
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.43122241239528e-05,
+      "loss": 0.3808,
+      "step": 13630
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.430772002522295e-05,
+      "loss": 0.3831,
+      "step": 13640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4303215926493115e-05,
+      "loss": 0.3807,
+      "step": 13650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4298711827763266e-05,
+      "loss": 0.3856,
+      "step": 13660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4294207729033424e-05,
+      "loss": 0.3799,
+      "step": 13670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.428970363030358e-05,
+      "loss": 0.376,
+      "step": 13680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.428519953157373e-05,
+      "loss": 0.3802,
+      "step": 13690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.428069543284389e-05,
+      "loss": 0.3736,
+      "step": 13700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.427619133411405e-05,
+      "loss": 0.3733,
+      "step": 13710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4271687235384204e-05,
+      "loss": 0.384,
+      "step": 13720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4267183136654355e-05,
+      "loss": 0.3802,
+      "step": 13730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.426267903792451e-05,
+      "loss": 0.3796,
+      "step": 13740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.425817493919467e-05,
+      "loss": 0.3759,
+      "step": 13750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.425367084046482e-05,
+      "loss": 0.3734,
+      "step": 13760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4249166741734985e-05,
+      "loss": 0.3796,
+      "step": 13770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4244662643005136e-05,
+      "loss": 0.3843,
+      "step": 13780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.424015854427529e-05,
+      "loss": 0.3821,
+      "step": 13790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.423565444554545e-05,
+      "loss": 0.3774,
+      "step": 13800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.42311503468156e-05,
+      "loss": 0.3688,
+      "step": 13810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4226646248085766e-05,
+      "loss": 0.3778,
+      "step": 13820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4222142149355916e-05,
+      "loss": 0.3673,
+      "step": 13830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.421763805062607e-05,
+      "loss": 0.3784,
+      "step": 13840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.421313395189623e-05,
+      "loss": 0.3785,
+      "step": 13850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.420862985316638e-05,
+      "loss": 0.3729,
+      "step": 13860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.420412575443654e-05,
+      "loss": 0.3801,
+      "step": 13870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.41996216557067e-05,
+      "loss": 0.3755,
+      "step": 13880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.419511755697685e-05,
+      "loss": 0.3764,
+      "step": 13890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4190613458247005e-05,
+      "loss": 0.3748,
+      "step": 13900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.418610935951716e-05,
+      "loss": 0.3778,
+      "step": 13910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.418160526078732e-05,
+      "loss": 0.378,
+      "step": 13920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.417710116205748e-05,
+      "loss": 0.376,
+      "step": 13930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.417259706332763e-05,
+      "loss": 0.3844,
+      "step": 13940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4168092964597786e-05,
+      "loss": 0.3763,
+      "step": 13950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4163588865867944e-05,
+      "loss": 0.3754,
+      "step": 13960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.41590847671381e-05,
+      "loss": 0.3782,
+      "step": 13970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.415458066840825e-05,
+      "loss": 0.3762,
+      "step": 13980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.415007656967841e-05,
+      "loss": 0.3765,
+      "step": 13990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.414557247094857e-05,
+      "loss": 0.3691,
+      "step": 14000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.414106837221872e-05,
+      "loss": 0.375,
+      "step": 14010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.413656427348888e-05,
+      "loss": 0.377,
+      "step": 14020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.413206017475903e-05,
+      "loss": 0.3734,
+      "step": 14030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.412755607602918e-05,
+      "loss": 0.3807,
+      "step": 14040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.412305197729935e-05,
+      "loss": 0.3767,
+      "step": 14050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.41185478785695e-05,
+      "loss": 0.3696,
+      "step": 14060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.411449418971264e-05,
+      "loss": 0.3744,
+      "step": 14070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.410999009098279e-05,
+      "loss": 0.3749,
+      "step": 14080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.410548599225296e-05,
+      "loss": 0.3762,
+      "step": 14090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.410098189352311e-05,
+      "loss": 0.3763,
+      "step": 14100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.409647779479326e-05,
+      "loss": 0.3748,
+      "step": 14110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.409197369606342e-05,
+      "loss": 0.378,
+      "step": 14120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.408746959733357e-05,
+      "loss": 0.3776,
+      "step": 14130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.408296549860373e-05,
+      "loss": 0.376,
+      "step": 14140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.407846139987389e-05,
+      "loss": 0.3743,
+      "step": 14150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.407395730114404e-05,
+      "loss": 0.3679,
+      "step": 14160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.40694532024142e-05,
+      "loss": 0.3715,
+      "step": 14170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.4064949103684354e-05,
+      "loss": 0.3763,
+      "step": 14180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.406044500495451e-05,
+      "loss": 0.3745,
+      "step": 14190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.405594090622467e-05,
+      "loss": 0.3732,
+      "step": 14200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.405143680749482e-05,
+      "loss": 0.3737,
+      "step": 14210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.404693270876498e-05,
+      "loss": 0.3672,
+      "step": 14220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.4042428610035135e-05,
+      "loss": 0.3753,
+      "step": 14230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.403792451130529e-05,
+      "loss": 0.3712,
+      "step": 14240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.403342041257544e-05,
+      "loss": 0.3709,
+      "step": 14250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.40289163138456e-05,
+      "loss": 0.3674,
+      "step": 14260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.402441221511576e-05,
+      "loss": 0.3614,
+      "step": 14270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.401990811638591e-05,
+      "loss": 0.3746,
+      "step": 14280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.401540401765607e-05,
+      "loss": 0.374,
+      "step": 14290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.4010899918926224e-05,
+      "loss": 0.3769,
+      "step": 14300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.400639582019638e-05,
+      "loss": 0.3751,
+      "step": 14310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.400189172146654e-05,
+      "loss": 0.3735,
+      "step": 14320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.399738762273669e-05,
+      "loss": 0.378,
+      "step": 14330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3992883524006854e-05,
+      "loss": 0.3709,
+      "step": 14340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3988379425277004e-05,
+      "loss": 0.3685,
+      "step": 14350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3983875326547155e-05,
+      "loss": 0.3718,
+      "step": 14360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.397937122781732e-05,
+      "loss": 0.3767,
+      "step": 14370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.397486712908747e-05,
+      "loss": 0.368,
+      "step": 14380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.397036303035763e-05,
+      "loss": 0.3698,
+      "step": 14390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3965858931627785e-05,
+      "loss": 0.3747,
+      "step": 14400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3961354832897936e-05,
+      "loss": 0.377,
+      "step": 14410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.395685073416809e-05,
+      "loss": 0.3758,
+      "step": 14420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.395234663543825e-05,
+      "loss": 0.3711,
+      "step": 14430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.394784253670841e-05,
+      "loss": 0.3786,
+      "step": 14440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3943338437978566e-05,
+      "loss": 0.3774,
+      "step": 14450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3938834339248716e-05,
+      "loss": 0.3718,
+      "step": 14460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3934330240518874e-05,
+      "loss": 0.3706,
+      "step": 14470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.392982614178903e-05,
+      "loss": 0.3746,
+      "step": 14480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.392532204305919e-05,
+      "loss": 0.3657,
+      "step": 14490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.392081794432934e-05,
+      "loss": 0.3743,
+      "step": 14500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.39163138455995e-05,
+      "loss": 0.3704,
+      "step": 14510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3911809746869654e-05,
+      "loss": 0.3764,
+      "step": 14520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3907305648139805e-05,
+      "loss": 0.3634,
+      "step": 14530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.390280154940997e-05,
+      "loss": 0.3709,
+      "step": 14540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.389829745068012e-05,
+      "loss": 0.3657,
+      "step": 14550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.389379335195027e-05,
+      "loss": 0.3773,
+      "step": 14560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3889289253220435e-05,
+      "loss": 0.3764,
+      "step": 14570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3884785154490586e-05,
+      "loss": 0.3739,
+      "step": 14580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.388028105576075e-05,
+      "loss": 0.3757,
+      "step": 14590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.38757769570309e-05,
+      "loss": 0.3727,
+      "step": 14600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.387127285830105e-05,
+      "loss": 0.3738,
+      "step": 14610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3866768759571216e-05,
+      "loss": 0.3728,
+      "step": 14620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3862264660841367e-05,
+      "loss": 0.3678,
+      "step": 14630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3857760562111524e-05,
+      "loss": 0.3772,
+      "step": 14640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.385325646338168e-05,
+      "loss": 0.3719,
+      "step": 14650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.384875236465183e-05,
+      "loss": 0.3743,
+      "step": 14660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.384424826592199e-05,
+      "loss": 0.3693,
+      "step": 14670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.383974416719215e-05,
+      "loss": 0.3745,
+      "step": 14680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3835240068462305e-05,
+      "loss": 0.3723,
+      "step": 14690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3830735969732455e-05,
+      "loss": 0.3731,
+      "step": 14700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.382623187100261e-05,
+      "loss": 0.3704,
+      "step": 14710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.382172777227277e-05,
+      "loss": 0.3669,
+      "step": 14720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.381722367354293e-05,
+      "loss": 0.3679,
+      "step": 14730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3812719574813085e-05,
+      "loss": 0.3693,
+      "step": 14740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3808215476083236e-05,
+      "loss": 0.3677,
+      "step": 14750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3803711377353394e-05,
+      "loss": 0.3719,
+      "step": 14760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.379920727862355e-05,
+      "loss": 0.3744,
+      "step": 14770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.37947031798937e-05,
+      "loss": 0.3742,
+      "step": 14780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3790199081163866e-05,
+      "loss": 0.3727,
+      "step": 14790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.378569498243402e-05,
+      "loss": 0.3681,
+      "step": 14800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.378119088370417e-05,
+      "loss": 0.3668,
+      "step": 14810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.377668678497433e-05,
+      "loss": 0.3693,
+      "step": 14820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.377218268624448e-05,
+      "loss": 0.372,
+      "step": 14830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.376767858751464e-05,
+      "loss": 0.371,
+      "step": 14840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.37631744887848e-05,
+      "loss": 0.3661,
+      "step": 14850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.375867039005495e-05,
+      "loss": 0.374,
+      "step": 14860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.375416629132511e-05,
+      "loss": 0.3716,
+      "step": 14870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.374966219259526e-05,
+      "loss": 0.371,
+      "step": 14880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.374515809386542e-05,
+      "loss": 0.3707,
+      "step": 14890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.374065399513558e-05,
+      "loss": 0.3665,
+      "step": 14900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.373614989640573e-05,
+      "loss": 0.3642,
+      "step": 14910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3731645797675886e-05,
+      "loss": 0.3679,
+      "step": 14920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3727141698946044e-05,
+      "loss": 0.3696,
+      "step": 14930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.37226376002162e-05,
+      "loss": 0.3655,
+      "step": 14940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.371813350148635e-05,
+      "loss": 0.3685,
+      "step": 14950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.371362940275651e-05,
+      "loss": 0.376,
+      "step": 14960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.370912530402667e-05,
+      "loss": 0.366,
+      "step": 14970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.370462120529682e-05,
+      "loss": 0.3678,
+      "step": 14980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.370011710656698e-05,
+      "loss": 0.3696,
+      "step": 14990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.369561300783713e-05,
+      "loss": 0.3632,
+      "step": 15000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.369110890910729e-05,
+      "loss": 0.3721,
+      "step": 15010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.368660481037745e-05,
+      "loss": 0.3652,
+      "step": 15020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.36821007116476e-05,
+      "loss": 0.3654,
+      "step": 15030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.367759661291776e-05,
+      "loss": 0.3676,
+      "step": 15040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3673092514187913e-05,
+      "loss": 0.3676,
+      "step": 15050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3668588415458064e-05,
+      "loss": 0.3736,
+      "step": 15060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.366408431672823e-05,
+      "loss": 0.3684,
+      "step": 15070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.365958021799838e-05,
+      "loss": 0.3724,
+      "step": 15080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.365507611926854e-05,
+      "loss": 0.3657,
+      "step": 15090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3650572020538694e-05,
+      "loss": 0.3694,
+      "step": 15100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3646067921808845e-05,
+      "loss": 0.3665,
+      "step": 15110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3641563823079e-05,
+      "loss": 0.3739,
+      "step": 15120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.363705972434916e-05,
+      "loss": 0.3717,
+      "step": 15130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.363255562561932e-05,
+      "loss": 0.3765,
+      "step": 15140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3628051526889475e-05,
+      "loss": 0.3655,
+      "step": 15150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3623547428159626e-05,
+      "loss": 0.3705,
+      "step": 15160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.361904332942978e-05,
+      "loss": 0.3707,
+      "step": 15170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.361453923069994e-05,
+      "loss": 0.3655,
+      "step": 15180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.36100351319701e-05,
+      "loss": 0.3667,
+      "step": 15190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.360553103324025e-05,
+      "loss": 0.3657,
+      "step": 15200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3601026934510406e-05,
+      "loss": 0.3656,
+      "step": 15210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3596522835780564e-05,
+      "loss": 0.3607,
+      "step": 15220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3592018737050714e-05,
+      "loss": 0.3678,
+      "step": 15230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.358751463832088e-05,
+      "loss": 0.3698,
+      "step": 15240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.358301053959103e-05,
+      "loss": 0.3654,
+      "step": 15250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.357850644086118e-05,
+      "loss": 0.3688,
+      "step": 15260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3574002342131344e-05,
+      "loss": 0.368,
+      "step": 15270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3569498243401495e-05,
+      "loss": 0.3703,
+      "step": 15280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.356499414467165e-05,
+      "loss": 0.3677,
+      "step": 15290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.356049004594181e-05,
+      "loss": 0.3656,
+      "step": 15300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.355598594721196e-05,
+      "loss": 0.3666,
+      "step": 15310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3551481848482125e-05,
+      "loss": 0.3701,
+      "step": 15320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3546977749752276e-05,
+      "loss": 0.3717,
+      "step": 15330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.354247365102243e-05,
+      "loss": 0.3707,
+      "step": 15340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.353796955229259e-05,
+      "loss": 0.3654,
+      "step": 15350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.353346545356274e-05,
+      "loss": 0.372,
+      "step": 15360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.35289613548329e-05,
+      "loss": 0.3656,
+      "step": 15370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3524457256103057e-05,
+      "loss": 0.3654,
+      "step": 15380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3519953157373214e-05,
+      "loss": 0.373,
+      "step": 15390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3515449058643365e-05,
+      "loss": 0.3712,
+      "step": 15400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.351094495991352e-05,
+      "loss": 0.3634,
+      "step": 15410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.350644086118368e-05,
+      "loss": 0.3684,
+      "step": 15420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.350193676245384e-05,
+      "loss": 0.3648,
+      "step": 15430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3497432663723995e-05,
+      "loss": 0.3659,
+      "step": 15440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3492928564994145e-05,
+      "loss": 0.3631,
+      "step": 15450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.34884244662643e-05,
+      "loss": 0.3682,
+      "step": 15460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.348392036753446e-05,
+      "loss": 0.3658,
+      "step": 15470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.347941626880461e-05,
+      "loss": 0.368,
+      "step": 15480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.347491217007477e-05,
+      "loss": 0.363,
+      "step": 15490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3470408071344926e-05,
+      "loss": 0.3585,
+      "step": 15500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.346590397261508e-05,
+      "loss": 0.3592,
+      "step": 15510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.346139987388524e-05,
+      "loss": 0.3607,
+      "step": 15520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.345689577515539e-05,
+      "loss": 0.3598,
+      "step": 15530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.345239167642555e-05,
+      "loss": 0.3657,
+      "step": 15540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.344788757769571e-05,
+      "loss": 0.3659,
+      "step": 15550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.344338347896586e-05,
+      "loss": 0.3633,
+      "step": 15560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.343887938023602e-05,
+      "loss": 0.3667,
+      "step": 15570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.343437528150617e-05,
+      "loss": 0.3681,
+      "step": 15580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.342987118277633e-05,
+      "loss": 0.3605,
+      "step": 15590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.342536708404649e-05,
+      "loss": 0.3753,
+      "step": 15600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.342086298531664e-05,
+      "loss": 0.3653,
+      "step": 15610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3416358886586796e-05,
+      "loss": 0.3636,
+      "step": 15620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.341185478785695e-05,
+      "loss": 0.3582,
+      "step": 15630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.340735068912711e-05,
+      "loss": 0.3688,
+      "step": 15640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.340284659039726e-05,
+      "loss": 0.3642,
+      "step": 15650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.339834249166742e-05,
+      "loss": 0.358,
+      "step": 15660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3393838392937576e-05,
+      "loss": 0.3671,
+      "step": 15670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.338933429420773e-05,
+      "loss": 0.3653,
+      "step": 15680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.338483019547789e-05,
+      "loss": 0.3646,
+      "step": 15690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.338032609674804e-05,
+      "loss": 0.3643,
+      "step": 15700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.33758219980182e-05,
+      "loss": 0.3583,
+      "step": 15710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.337131789928836e-05,
+      "loss": 0.359,
+      "step": 15720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.336681380055851e-05,
+      "loss": 0.3677,
+      "step": 15730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3362309701828665e-05,
+      "loss": 0.3614,
+      "step": 15740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.335780560309882e-05,
+      "loss": 0.3623,
+      "step": 15750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3353301504368973e-05,
+      "loss": 0.3551,
+      "step": 15760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.334879740563914e-05,
+      "loss": 0.3628,
+      "step": 15770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.334429330690929e-05,
+      "loss": 0.3689,
+      "step": 15780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3339789208179446e-05,
+      "loss": 0.3594,
+      "step": 15790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3335285109449603e-05,
+      "loss": 0.3628,
+      "step": 15800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3330781010719754e-05,
+      "loss": 0.3657,
+      "step": 15810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.332627691198991e-05,
+      "loss": 0.3619,
+      "step": 15820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.332177281326007e-05,
+      "loss": 0.3698,
+      "step": 15830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.331726871453023e-05,
+      "loss": 0.3647,
+      "step": 15840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3312764615800384e-05,
+      "loss": 0.3661,
+      "step": 15850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3308260517070535e-05,
+      "loss": 0.3647,
+      "step": 15860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.330375641834069e-05,
+      "loss": 0.3623,
+      "step": 15870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.329925231961085e-05,
+      "loss": 0.3679,
+      "step": 15880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.329474822088101e-05,
+      "loss": 0.3595,
+      "step": 15890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.329024412215116e-05,
+      "loss": 0.361,
+      "step": 15900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3285740023421316e-05,
+      "loss": 0.3593,
+      "step": 15910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.328123592469147e-05,
+      "loss": 0.3678,
+      "step": 15920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3276731825961624e-05,
+      "loss": 0.3646,
+      "step": 15930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.327222772723178e-05,
+      "loss": 0.3672,
+      "step": 15940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.326772362850194e-05,
+      "loss": 0.3612,
+      "step": 15950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.326321952977209e-05,
+      "loss": 0.3609,
+      "step": 15960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3258715431042254e-05,
+      "loss": 0.3561,
+      "step": 15970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3254211332312404e-05,
+      "loss": 0.3553,
+      "step": 15980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.324970723358256e-05,
+      "loss": 0.3646,
+      "step": 15990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.324520313485272e-05,
+      "loss": 0.3612,
+      "step": 16000
+    },
+    {
+      "epoch": 0.14,
+      "eval_NEIMS_canon_smiles": 0.91085,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.0039,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.3440219888393381,
+      "eval_NEIMS_exact_mols": 0.0038833333333333333,
+      "eval_NEIMS_exact_smiles": 0.0037166666666666667,
+      "eval_NEIMS_loss": 0.36472371220588684,
+      "eval_NEIMS_matched_formulas": 0.3067,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.24112814002003777,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.003933333333333333,
+      "eval_NEIMS_runtime": 723.6863,
+      "eval_NEIMS_samples_per_second": 82.909,
+      "eval_NEIMS_steps_per_second": 1.296,
+      "step": 16000
+    },
+    {
+      "epoch": 0.14,
+      "eval_RASSP_canon_smiles": 0.9057501586135506,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.006995692389888804,
+      "eval_RASSP_daylight_tanimoto_simil": 0.3861519831986421,
+      "eval_RASSP_exact_mols": 0.006945603900223729,
+      "eval_RASSP_exact_smiles": 0.006578288309346512,
+      "eval_RASSP_loss": 0.3267086148262024,
+      "eval_RASSP_matched_formulas": 0.39488095635622933,
+      "eval_RASSP_morgan_tanimoto_simil": 0.25511754811398485,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.006978996226667112,
+      "eval_RASSP_runtime": 828.7523,
+      "eval_RASSP_samples_per_second": 72.27,
+      "eval_RASSP_steps_per_second": 1.129,
+      "step": 16000
+    },
+    {
+      "epoch": 0.14,
+      "eval_NIST_canon_smiles": 0.8548461511161586,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.002555275579373248,
+      "eval_NIST_daylight_tanimoto_simil": 0.21976448765447698,
+      "eval_NIST_exact_mols": 0.002271356070553998,
+      "eval_NIST_exact_smiles": 0.002093906377541967,
+      "eval_NIST_loss": 1.5112570524215698,
+      "eval_NIST_matched_formulas": 0.03275721333002094,
+      "eval_NIST_morgan_tanimoto_simil": 0.1761940699405261,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.002448805763566029,
+      "eval_NIST_runtime": 455.2328,
+      "eval_NIST_samples_per_second": 61.896,
+      "eval_NIST_steps_per_second": 0.969,
+      "step": 16000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.324069903612287e-05,
+      "loss": 0.3649,
+      "step": 16010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3236194937393034e-05,
+      "loss": 0.3641,
+      "step": 16020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3231690838663185e-05,
+      "loss": 0.358,
+      "step": 16030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.322718673993334e-05,
+      "loss": 0.359,
+      "step": 16040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.32226826412035e-05,
+      "loss": 0.3644,
+      "step": 16050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.321817854247365e-05,
+      "loss": 0.3683,
+      "step": 16060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.321367444374381e-05,
+      "loss": 0.3613,
+      "step": 16070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3209620754886945e-05,
+      "loss": 0.3691,
+      "step": 16080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.320511665615711e-05,
+      "loss": 0.3645,
+      "step": 16090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.320061255742726e-05,
+      "loss": 0.3535,
+      "step": 16100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.319610845869742e-05,
+      "loss": 0.359,
+      "step": 16110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3191604359967575e-05,
+      "loss": 0.3661,
+      "step": 16120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3187100261237726e-05,
+      "loss": 0.3574,
+      "step": 16130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3182596162507883e-05,
+      "loss": 0.3624,
+      "step": 16140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.317809206377804e-05,
+      "loss": 0.3632,
+      "step": 16150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.31735879650482e-05,
+      "loss": 0.3618,
+      "step": 16160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.316908386631835e-05,
+      "loss": 0.3676,
+      "step": 16170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.316457976758851e-05,
+      "loss": 0.362,
+      "step": 16180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3160075668858664e-05,
+      "loss": 0.3617,
+      "step": 16190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3155571570128815e-05,
+      "loss": 0.3557,
+      "step": 16200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.315106747139898e-05,
+      "loss": 0.3608,
+      "step": 16210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.314656337266913e-05,
+      "loss": 0.3637,
+      "step": 16220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.314205927393929e-05,
+      "loss": 0.3556,
+      "step": 16230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3137555175209445e-05,
+      "loss": 0.3638,
+      "step": 16240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3133051076479596e-05,
+      "loss": 0.3564,
+      "step": 16250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.312854697774975e-05,
+      "loss": 0.3611,
+      "step": 16260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.312404287901991e-05,
+      "loss": 0.3588,
+      "step": 16270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.311953878029006e-05,
+      "loss": 0.3539,
+      "step": 16280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3115034681560226e-05,
+      "loss": 0.359,
+      "step": 16290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3110530582830376e-05,
+      "loss": 0.3583,
+      "step": 16300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3106026484100534e-05,
+      "loss": 0.3643,
+      "step": 16310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.310152238537069e-05,
+      "loss": 0.3579,
+      "step": 16320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.309701828664084e-05,
+      "loss": 0.3635,
+      "step": 16330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3092514187911e-05,
+      "loss": 0.3603,
+      "step": 16340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.308801008918116e-05,
+      "loss": 0.3601,
+      "step": 16350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3083505990451314e-05,
+      "loss": 0.359,
+      "step": 16360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.307900189172147e-05,
+      "loss": 0.359,
+      "step": 16370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.307449779299162e-05,
+      "loss": 0.3657,
+      "step": 16380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.306999369426178e-05,
+      "loss": 0.3587,
+      "step": 16390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.306548959553194e-05,
+      "loss": 0.3602,
+      "step": 16400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3060985496802095e-05,
+      "loss": 0.361,
+      "step": 16410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3056481398072246e-05,
+      "loss": 0.3638,
+      "step": 16420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.30519772993424e-05,
+      "loss": 0.3648,
+      "step": 16430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.304747320061256e-05,
+      "loss": 0.3533,
+      "step": 16440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.304296910188271e-05,
+      "loss": 0.3564,
+      "step": 16450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3038465003152876e-05,
+      "loss": 0.3616,
+      "step": 16460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3033960904423026e-05,
+      "loss": 0.3596,
+      "step": 16470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.302945680569318e-05,
+      "loss": 0.3579,
+      "step": 16480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.302495270696334e-05,
+      "loss": 0.3562,
+      "step": 16490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.302044860823349e-05,
+      "loss": 0.3583,
+      "step": 16500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.301594450950365e-05,
+      "loss": 0.3562,
+      "step": 16510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.301144041077381e-05,
+      "loss": 0.3561,
+      "step": 16520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.300693631204396e-05,
+      "loss": 0.3529,
+      "step": 16530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.300243221331412e-05,
+      "loss": 0.3531,
+      "step": 16540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.299792811458427e-05,
+      "loss": 0.3581,
+      "step": 16550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.299342401585443e-05,
+      "loss": 0.3566,
+      "step": 16560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.298891991712459e-05,
+      "loss": 0.3619,
+      "step": 16570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.298441581839474e-05,
+      "loss": 0.3625,
+      "step": 16580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2979911719664896e-05,
+      "loss": 0.3538,
+      "step": 16590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2975407620935054e-05,
+      "loss": 0.3573,
+      "step": 16600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.297090352220521e-05,
+      "loss": 0.3588,
+      "step": 16610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.296639942347536e-05,
+      "loss": 0.3622,
+      "step": 16620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.296189532474552e-05,
+      "loss": 0.3578,
+      "step": 16630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.295739122601568e-05,
+      "loss": 0.3637,
+      "step": 16640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2952887127285834e-05,
+      "loss": 0.3577,
+      "step": 16650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.294838302855599e-05,
+      "loss": 0.3579,
+      "step": 16660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.294387892982614e-05,
+      "loss": 0.3601,
+      "step": 16670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.29393748310963e-05,
+      "loss": 0.3551,
+      "step": 16680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.293487073236646e-05,
+      "loss": 0.3608,
+      "step": 16690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.293036663363661e-05,
+      "loss": 0.3514,
+      "step": 16700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2925862534906766e-05,
+      "loss": 0.3493,
+      "step": 16710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.292135843617692e-05,
+      "loss": 0.3605,
+      "step": 16720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2916854337447074e-05,
+      "loss": 0.3541,
+      "step": 16730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.291235023871724e-05,
+      "loss": 0.3551,
+      "step": 16740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.290784613998739e-05,
+      "loss": 0.3555,
+      "step": 16750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2903342041257546e-05,
+      "loss": 0.3547,
+      "step": 16760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2898837942527704e-05,
+      "loss": 0.3633,
+      "step": 16770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2894333843797855e-05,
+      "loss": 0.356,
+      "step": 16780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.288982974506802e-05,
+      "loss": 0.3537,
+      "step": 16790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.288532564633817e-05,
+      "loss": 0.3549,
+      "step": 16800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.288082154760833e-05,
+      "loss": 0.3536,
+      "step": 16810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2876317448878485e-05,
+      "loss": 0.3616,
+      "step": 16820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2871813350148635e-05,
+      "loss": 0.3591,
+      "step": 16830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.286730925141879e-05,
+      "loss": 0.3602,
+      "step": 16840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.286280515268895e-05,
+      "loss": 0.3541,
+      "step": 16850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.285830105395911e-05,
+      "loss": 0.3476,
+      "step": 16860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.285379695522926e-05,
+      "loss": 0.3537,
+      "step": 16870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2849292856499416e-05,
+      "loss": 0.3562,
+      "step": 16880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2844788757769573e-05,
+      "loss": 0.3559,
+      "step": 16890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2840284659039724e-05,
+      "loss": 0.3565,
+      "step": 16900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.283578056030988e-05,
+      "loss": 0.3485,
+      "step": 16910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.283127646158004e-05,
+      "loss": 0.3528,
+      "step": 16920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2826772362850197e-05,
+      "loss": 0.3576,
+      "step": 16930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2822268264120354e-05,
+      "loss": 0.3571,
+      "step": 16940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2817764165390505e-05,
+      "loss": 0.3599,
+      "step": 16950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.281326006666066e-05,
+      "loss": 0.3554,
+      "step": 16960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.280875596793082e-05,
+      "loss": 0.3571,
+      "step": 16970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.280425186920097e-05,
+      "loss": 0.3595,
+      "step": 16980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2799747770471135e-05,
+      "loss": 0.3558,
+      "step": 16990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2795243671741285e-05,
+      "loss": 0.358,
+      "step": 17000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.279073957301144e-05,
+      "loss": 0.3592,
+      "step": 17010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.27862354742816e-05,
+      "loss": 0.3611,
+      "step": 17020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.278173137555175e-05,
+      "loss": 0.3566,
+      "step": 17030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.277722727682191e-05,
+      "loss": 0.3543,
+      "step": 17040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2772723178092066e-05,
+      "loss": 0.3492,
+      "step": 17050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2768219079362224e-05,
+      "loss": 0.35,
+      "step": 17060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.276371498063238e-05,
+      "loss": 0.352,
+      "step": 17070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.275921088190253e-05,
+      "loss": 0.3558,
+      "step": 17080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.275470678317269e-05,
+      "loss": 0.3529,
+      "step": 17090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.275020268444285e-05,
+      "loss": 0.3526,
+      "step": 17100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2745698585713004e-05,
+      "loss": 0.3574,
+      "step": 17110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2741194486983155e-05,
+      "loss": 0.3556,
+      "step": 17120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.273669038825331e-05,
+      "loss": 0.3532,
+      "step": 17130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.273218628952347e-05,
+      "loss": 0.3551,
+      "step": 17140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.272768219079362e-05,
+      "loss": 0.3486,
+      "step": 17150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.272317809206378e-05,
+      "loss": 0.351,
+      "step": 17160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2718673993333936e-05,
+      "loss": 0.3512,
+      "step": 17170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2714169894604086e-05,
+      "loss": 0.3546,
+      "step": 17180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.270966579587425e-05,
+      "loss": 0.3574,
+      "step": 17190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.27051616971444e-05,
+      "loss": 0.3508,
+      "step": 17200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.270065759841456e-05,
+      "loss": 0.356,
+      "step": 17210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2696153499684716e-05,
+      "loss": 0.3444,
+      "step": 17220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.269164940095487e-05,
+      "loss": 0.3564,
+      "step": 17230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.268714530222503e-05,
+      "loss": 0.3607,
+      "step": 17240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.268264120349518e-05,
+      "loss": 0.3625,
+      "step": 17250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.267813710476534e-05,
+      "loss": 0.3566,
+      "step": 17260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.26736330060355e-05,
+      "loss": 0.3533,
+      "step": 17270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.266912890730565e-05,
+      "loss": 0.3546,
+      "step": 17280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2664624808575805e-05,
+      "loss": 0.3535,
+      "step": 17290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.266012070984596e-05,
+      "loss": 0.3532,
+      "step": 17300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.265561661111612e-05,
+      "loss": 0.349,
+      "step": 17310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.265111251238627e-05,
+      "loss": 0.352,
+      "step": 17320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.264660841365643e-05,
+      "loss": 0.3522,
+      "step": 17330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2642104314926586e-05,
+      "loss": 0.3479,
+      "step": 17340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2637600216196744e-05,
+      "loss": 0.353,
+      "step": 17350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2633096117466894e-05,
+      "loss": 0.3529,
+      "step": 17360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.262859201873705e-05,
+      "loss": 0.3499,
+      "step": 17370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.262408792000721e-05,
+      "loss": 0.3464,
+      "step": 17380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.261958382127737e-05,
+      "loss": 0.3499,
+      "step": 17390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.261507972254752e-05,
+      "loss": 0.3528,
+      "step": 17400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2610575623817675e-05,
+      "loss": 0.3546,
+      "step": 17410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.260607152508783e-05,
+      "loss": 0.3502,
+      "step": 17420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.260156742635798e-05,
+      "loss": 0.3458,
+      "step": 17430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.259706332762815e-05,
+      "loss": 0.3537,
+      "step": 17440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.25925592288983e-05,
+      "loss": 0.355,
+      "step": 17450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2588055130168456e-05,
+      "loss": 0.3547,
+      "step": 17460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.258355103143861e-05,
+      "loss": 0.3454,
+      "step": 17470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2579046932708764e-05,
+      "loss": 0.3506,
+      "step": 17480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.257454283397893e-05,
+      "loss": 0.3536,
+      "step": 17490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.257003873524908e-05,
+      "loss": 0.3516,
+      "step": 17500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2565534636519236e-05,
+      "loss": 0.3497,
+      "step": 17510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2561030537789394e-05,
+      "loss": 0.3481,
+      "step": 17520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2556526439059544e-05,
+      "loss": 0.3528,
+      "step": 17530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.25520223403297e-05,
+      "loss": 0.346,
+      "step": 17540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.254751824159986e-05,
+      "loss": 0.3546,
+      "step": 17550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.254301414287001e-05,
+      "loss": 0.3465,
+      "step": 17560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.253851004414017e-05,
+      "loss": 0.3476,
+      "step": 17570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2534005945410325e-05,
+      "loss": 0.3551,
+      "step": 17580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.252950184668048e-05,
+      "loss": 0.3521,
+      "step": 17590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.252499774795063e-05,
+      "loss": 0.3497,
+      "step": 17600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.252049364922079e-05,
+      "loss": 0.3491,
+      "step": 17610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.251598955049095e-05,
+      "loss": 0.3463,
+      "step": 17620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2511485451761106e-05,
+      "loss": 0.3498,
+      "step": 17630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.250698135303126e-05,
+      "loss": 0.3523,
+      "step": 17640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2502477254301414e-05,
+      "loss": 0.3508,
+      "step": 17650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.249797315557157e-05,
+      "loss": 0.3499,
+      "step": 17660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.249346905684173e-05,
+      "loss": 0.3488,
+      "step": 17670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.248896495811188e-05,
+      "loss": 0.3502,
+      "step": 17680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2484460859382044e-05,
+      "loss": 0.3427,
+      "step": 17690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2479956760652195e-05,
+      "loss": 0.352,
+      "step": 17700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.247545266192235e-05,
+      "loss": 0.3532,
+      "step": 17710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.247094856319251e-05,
+      "loss": 0.3536,
+      "step": 17720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.246644446446266e-05,
+      "loss": 0.3445,
+      "step": 17730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.246194036573282e-05,
+      "loss": 0.3509,
+      "step": 17740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2457436267002975e-05,
+      "loss": 0.3493,
+      "step": 17750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.245293216827313e-05,
+      "loss": 0.3479,
+      "step": 17760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.244842806954329e-05,
+      "loss": 0.3554,
+      "step": 17770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.244392397081344e-05,
+      "loss": 0.3456,
+      "step": 17780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.24394198720836e-05,
+      "loss": 0.3501,
+      "step": 17790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2434915773353756e-05,
+      "loss": 0.3466,
+      "step": 17800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.243041167462391e-05,
+      "loss": 0.3477,
+      "step": 17810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2425907575894064e-05,
+      "loss": 0.3534,
+      "step": 17820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.242140347716422e-05,
+      "loss": 0.3512,
+      "step": 17830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.241689937843438e-05,
+      "loss": 0.3498,
+      "step": 17840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.241239527970453e-05,
+      "loss": 0.3527,
+      "step": 17850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.240789118097469e-05,
+      "loss": 0.3481,
+      "step": 17860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2403387082244845e-05,
+      "loss": 0.343,
+      "step": 17870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2398882983514996e-05,
+      "loss": 0.3509,
+      "step": 17880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.239437888478516e-05,
+      "loss": 0.3467,
+      "step": 17890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.238987478605531e-05,
+      "loss": 0.3457,
+      "step": 17900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.238537068732547e-05,
+      "loss": 0.3491,
+      "step": 17910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2380866588595626e-05,
+      "loss": 0.3453,
+      "step": 17920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2376362489865776e-05,
+      "loss": 0.3426,
+      "step": 17930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.237185839113594e-05,
+      "loss": 0.3409,
+      "step": 17940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.236735429240609e-05,
+      "loss": 0.3478,
+      "step": 17950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.236285019367625e-05,
+      "loss": 0.3462,
+      "step": 17960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2358346094946406e-05,
+      "loss": 0.3543,
+      "step": 17970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.235384199621656e-05,
+      "loss": 0.3456,
+      "step": 17980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2349337897486715e-05,
+      "loss": 0.3535,
+      "step": 17990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.234483379875687e-05,
+      "loss": 0.3459,
+      "step": 18000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.234032970002702e-05,
+      "loss": 0.3531,
+      "step": 18010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.233582560129718e-05,
+      "loss": 0.3519,
+      "step": 18020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.233132150256734e-05,
+      "loss": 0.3497,
+      "step": 18030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2326817403837495e-05,
+      "loss": 0.3478,
+      "step": 18040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.232231330510765e-05,
+      "loss": 0.3473,
+      "step": 18050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2317809206377803e-05,
+      "loss": 0.3449,
+      "step": 18060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.231330510764796e-05,
+      "loss": 0.3544,
+      "step": 18070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.230880100891812e-05,
+      "loss": 0.346,
+      "step": 18080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2304296910188276e-05,
+      "loss": 0.3526,
+      "step": 18090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.229979281145843e-05,
+      "loss": 0.3572,
+      "step": 18100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2295288712728584e-05,
+      "loss": 0.3499,
+      "step": 18110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.229078461399874e-05,
+      "loss": 0.3527,
+      "step": 18120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.228628051526889e-05,
+      "loss": 0.353,
+      "step": 18130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.228177641653906e-05,
+      "loss": 0.3465,
+      "step": 18140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.227727231780921e-05,
+      "loss": 0.3451,
+      "step": 18150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2272768219079365e-05,
+      "loss": 0.3461,
+      "step": 18160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.226826412034952e-05,
+      "loss": 0.3461,
+      "step": 18170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.226376002161967e-05,
+      "loss": 0.3388,
+      "step": 18180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.225970633276282e-05,
+      "loss": 0.3459,
+      "step": 18190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.225520223403297e-05,
+      "loss": 0.3471,
+      "step": 18200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.225069813530313e-05,
+      "loss": 0.3539,
+      "step": 18210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.224619403657328e-05,
+      "loss": 0.3486,
+      "step": 18220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.224168993784344e-05,
+      "loss": 0.3535,
+      "step": 18230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.22371858391136e-05,
+      "loss": 0.3473,
+      "step": 18240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.223268174038375e-05,
+      "loss": 0.3449,
+      "step": 18250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2228177641653906e-05,
+      "loss": 0.3474,
+      "step": 18260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.222367354292406e-05,
+      "loss": 0.3438,
+      "step": 18270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.221916944419422e-05,
+      "loss": 0.3456,
+      "step": 18280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.221466534546438e-05,
+      "loss": 0.3511,
+      "step": 18290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.221016124673453e-05,
+      "loss": 0.3421,
+      "step": 18300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2205657148004686e-05,
+      "loss": 0.3423,
+      "step": 18310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2201153049274844e-05,
+      "loss": 0.3488,
+      "step": 18320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2196648950545e-05,
+      "loss": 0.3443,
+      "step": 18330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.219214485181515e-05,
+      "loss": 0.3533,
+      "step": 18340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.218764075308531e-05,
+      "loss": 0.3534,
+      "step": 18350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.218313665435547e-05,
+      "loss": 0.3464,
+      "step": 18360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.217863255562562e-05,
+      "loss": 0.3519,
+      "step": 18370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2174128456895775e-05,
+      "loss": 0.3433,
+      "step": 18380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.216962435816593e-05,
+      "loss": 0.3522,
+      "step": 18390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2165120259436084e-05,
+      "loss": 0.3488,
+      "step": 18400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.216061616070625e-05,
+      "loss": 0.3439,
+      "step": 18410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.21561120619764e-05,
+      "loss": 0.3445,
+      "step": 18420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2151607963246556e-05,
+      "loss": 0.3459,
+      "step": 18430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2147103864516713e-05,
+      "loss": 0.3392,
+      "step": 18440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2142599765786864e-05,
+      "loss": 0.3451,
+      "step": 18450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.213809566705703e-05,
+      "loss": 0.3451,
+      "step": 18460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.213359156832718e-05,
+      "loss": 0.3495,
+      "step": 18470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.212908746959734e-05,
+      "loss": 0.3505,
+      "step": 18480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2124583370867494e-05,
+      "loss": 0.3428,
+      "step": 18490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2120079272137645e-05,
+      "loss": 0.3421,
+      "step": 18500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.21155751734078e-05,
+      "loss": 0.3459,
+      "step": 18510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.211107107467796e-05,
+      "loss": 0.3419,
+      "step": 18520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.210656697594812e-05,
+      "loss": 0.3473,
+      "step": 18530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.210206287721827e-05,
+      "loss": 0.3486,
+      "step": 18540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2097558778488426e-05,
+      "loss": 0.3456,
+      "step": 18550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.209305467975858e-05,
+      "loss": 0.3498,
+      "step": 18560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.208855058102874e-05,
+      "loss": 0.3464,
+      "step": 18570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.208404648229889e-05,
+      "loss": 0.3515,
+      "step": 18580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.207954238356905e-05,
+      "loss": 0.3425,
+      "step": 18590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2075038284839206e-05,
+      "loss": 0.3448,
+      "step": 18600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2070534186109364e-05,
+      "loss": 0.3483,
+      "step": 18610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2066030087379514e-05,
+      "loss": 0.345,
+      "step": 18620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.206152598864967e-05,
+      "loss": 0.3406,
+      "step": 18630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.205702188991983e-05,
+      "loss": 0.3412,
+      "step": 18640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.205251779118998e-05,
+      "loss": 0.3438,
+      "step": 18650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2048013692460144e-05,
+      "loss": 0.3501,
+      "step": 18660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2043509593730295e-05,
+      "loss": 0.3467,
+      "step": 18670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.203900549500045e-05,
+      "loss": 0.3425,
+      "step": 18680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.203450139627061e-05,
+      "loss": 0.347,
+      "step": 18690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.202999729754076e-05,
+      "loss": 0.3384,
+      "step": 18700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2025493198810925e-05,
+      "loss": 0.3425,
+      "step": 18710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.2020989100081076e-05,
+      "loss": 0.3451,
+      "step": 18720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.201648500135123e-05,
+      "loss": 0.3497,
+      "step": 18730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.201198090262139e-05,
+      "loss": 0.345,
+      "step": 18740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.200747680389154e-05,
+      "loss": 0.3435,
+      "step": 18750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.20029727051617e-05,
+      "loss": 0.3491,
+      "step": 18760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1998468606431857e-05,
+      "loss": 0.3386,
+      "step": 18770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.199396450770201e-05,
+      "loss": 0.3443,
+      "step": 18780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1989460408972165e-05,
+      "loss": 0.3468,
+      "step": 18790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.198495631024232e-05,
+      "loss": 0.3455,
+      "step": 18800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.198045221151248e-05,
+      "loss": 0.3462,
+      "step": 18810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.197594811278263e-05,
+      "loss": 0.3425,
+      "step": 18820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.197144401405279e-05,
+      "loss": 0.3493,
+      "step": 18830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1966939915322945e-05,
+      "loss": 0.3464,
+      "step": 18840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.19624358165931e-05,
+      "loss": 0.3433,
+      "step": 18850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.195793171786326e-05,
+      "loss": 0.3377,
+      "step": 18860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.195342761913341e-05,
+      "loss": 0.3458,
+      "step": 18870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.194892352040357e-05,
+      "loss": 0.3449,
+      "step": 18880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1944419421673726e-05,
+      "loss": 0.3462,
+      "step": 18890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.193991532294388e-05,
+      "loss": 0.3382,
+      "step": 18900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.193541122421404e-05,
+      "loss": 0.3397,
+      "step": 18910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.193090712548419e-05,
+      "loss": 0.3355,
+      "step": 18920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.192640302675435e-05,
+      "loss": 0.3346,
+      "step": 18930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.192189892802451e-05,
+      "loss": 0.3478,
+      "step": 18940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.191739482929466e-05,
+      "loss": 0.3475,
+      "step": 18950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1912890730564815e-05,
+      "loss": 0.347,
+      "step": 18960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.190838663183497e-05,
+      "loss": 0.3373,
+      "step": 18970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.190388253310512e-05,
+      "loss": 0.3435,
+      "step": 18980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.189937843437529e-05,
+      "loss": 0.3369,
+      "step": 18990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.189487433564544e-05,
+      "loss": 0.3411,
+      "step": 19000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1890370236915596e-05,
+      "loss": 0.3469,
+      "step": 19010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.188586613818575e-05,
+      "loss": 0.3417,
+      "step": 19020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1881362039455904e-05,
+      "loss": 0.3467,
+      "step": 19030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.187685794072606e-05,
+      "loss": 0.3469,
+      "step": 19040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.187235384199622e-05,
+      "loss": 0.3398,
+      "step": 19050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1867849743266376e-05,
+      "loss": 0.3401,
+      "step": 19060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.186334564453653e-05,
+      "loss": 0.344,
+      "step": 19070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1858841545806685e-05,
+      "loss": 0.3411,
+      "step": 19080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.185433744707684e-05,
+      "loss": 0.345,
+      "step": 19090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.184983334834699e-05,
+      "loss": 0.3425,
+      "step": 19100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.184532924961716e-05,
+      "loss": 0.3407,
+      "step": 19110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.184082515088731e-05,
+      "loss": 0.3441,
+      "step": 19120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1836321052157465e-05,
+      "loss": 0.3392,
+      "step": 19130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.183181695342762e-05,
+      "loss": 0.3422,
+      "step": 19140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1827312854697773e-05,
+      "loss": 0.3522,
+      "step": 19150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.182280875596794e-05,
+      "loss": 0.3417,
+      "step": 19160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.181830465723809e-05,
+      "loss": 0.3505,
+      "step": 19170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1813800558508246e-05,
+      "loss": 0.341,
+      "step": 19180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1809296459778403e-05,
+      "loss": 0.3373,
+      "step": 19190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1804792361048554e-05,
+      "loss": 0.3381,
+      "step": 19200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.180028826231871e-05,
+      "loss": 0.3337,
+      "step": 19210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.179578416358887e-05,
+      "loss": 0.3393,
+      "step": 19220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.179128006485902e-05,
+      "loss": 0.3395,
+      "step": 19230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.178677596612918e-05,
+      "loss": 0.3418,
+      "step": 19240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1782271867399335e-05,
+      "loss": 0.343,
+      "step": 19250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.177776776866949e-05,
+      "loss": 0.3453,
+      "step": 19260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.177326366993965e-05,
+      "loss": 0.3394,
+      "step": 19270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.17687595712098e-05,
+      "loss": 0.3354,
+      "step": 19280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.176425547247996e-05,
+      "loss": 0.3425,
+      "step": 19290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1759751373750116e-05,
+      "loss": 0.348,
+      "step": 19300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.175524727502027e-05,
+      "loss": 0.3341,
+      "step": 19310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1750743176290424e-05,
+      "loss": 0.3415,
+      "step": 19320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.174623907756058e-05,
+      "loss": 0.3448,
+      "step": 19330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.174173497883074e-05,
+      "loss": 0.3501,
+      "step": 19340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.173723088010089e-05,
+      "loss": 0.3401,
+      "step": 19350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1732726781371054e-05,
+      "loss": 0.3403,
+      "step": 19360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1728222682641204e-05,
+      "loss": 0.3443,
+      "step": 19370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.172371858391136e-05,
+      "loss": 0.3462,
+      "step": 19380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.171921448518152e-05,
+      "loss": 0.3381,
+      "step": 19390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.171471038645167e-05,
+      "loss": 0.3439,
+      "step": 19400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1710206287721834e-05,
+      "loss": 0.3393,
+      "step": 19410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1705702188991985e-05,
+      "loss": 0.3398,
+      "step": 19420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1701198090262136e-05,
+      "loss": 0.3362,
+      "step": 19430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.16966939915323e-05,
+      "loss": 0.3359,
+      "step": 19440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.169218989280245e-05,
+      "loss": 0.3372,
+      "step": 19450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.168768579407261e-05,
+      "loss": 0.3334,
+      "step": 19460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1683181695342766e-05,
+      "loss": 0.3383,
+      "step": 19470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1678677596612916e-05,
+      "loss": 0.3442,
+      "step": 19480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1674173497883074e-05,
+      "loss": 0.3448,
+      "step": 19490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.166966939915323e-05,
+      "loss": 0.3337,
+      "step": 19500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.166516530042339e-05,
+      "loss": 0.3366,
+      "step": 19510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.166066120169354e-05,
+      "loss": 0.3402,
+      "step": 19520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.16561571029637e-05,
+      "loss": 0.3349,
+      "step": 19530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1651653004233855e-05,
+      "loss": 0.3357,
+      "step": 19540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.164714890550401e-05,
+      "loss": 0.3397,
+      "step": 19550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.164264480677417e-05,
+      "loss": 0.3419,
+      "step": 19560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.163814070804432e-05,
+      "loss": 0.3438,
+      "step": 19570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.163363660931448e-05,
+      "loss": 0.3433,
+      "step": 19580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1629132510584635e-05,
+      "loss": 0.3399,
+      "step": 19590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.1624628411854786e-05,
+      "loss": 0.3427,
+      "step": 19600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.162012431312495e-05,
+      "loss": 0.3392,
+      "step": 19610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.16156202143951e-05,
+      "loss": 0.3397,
+      "step": 19620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.161111611566525e-05,
+      "loss": 0.3364,
+      "step": 19630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1606612016935416e-05,
+      "loss": 0.3429,
+      "step": 19640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.160210791820557e-05,
+      "loss": 0.3411,
+      "step": 19650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1597603819475724e-05,
+      "loss": 0.3406,
+      "step": 19660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.159309972074588e-05,
+      "loss": 0.3349,
+      "step": 19670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.158859562201603e-05,
+      "loss": 0.3444,
+      "step": 19680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.15840915232862e-05,
+      "loss": 0.3412,
+      "step": 19690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.157958742455635e-05,
+      "loss": 0.3374,
+      "step": 19700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1575083325826505e-05,
+      "loss": 0.3392,
+      "step": 19710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.157057922709666e-05,
+      "loss": 0.34,
+      "step": 19720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.156607512836681e-05,
+      "loss": 0.3364,
+      "step": 19730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.156157102963697e-05,
+      "loss": 0.3397,
+      "step": 19740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.155706693090713e-05,
+      "loss": 0.3408,
+      "step": 19750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1552562832177286e-05,
+      "loss": 0.3405,
+      "step": 19760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1548058733447436e-05,
+      "loss": 0.3363,
+      "step": 19770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1543554634717594e-05,
+      "loss": 0.3384,
+      "step": 19780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.153905053598775e-05,
+      "loss": 0.3324,
+      "step": 19790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.15345464372579e-05,
+      "loss": 0.3366,
+      "step": 19800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1530042338528066e-05,
+      "loss": 0.3342,
+      "step": 19810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.152553823979822e-05,
+      "loss": 0.3398,
+      "step": 19820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1521034141068374e-05,
+      "loss": 0.3394,
+      "step": 19830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.151653004233853e-05,
+      "loss": 0.3384,
+      "step": 19840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.151202594360868e-05,
+      "loss": 0.3428,
+      "step": 19850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.150752184487885e-05,
+      "loss": 0.3329,
+      "step": 19860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1503017746149e-05,
+      "loss": 0.333,
+      "step": 19870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.149851364741915e-05,
+      "loss": 0.3353,
+      "step": 19880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.149400954868931e-05,
+      "loss": 0.3391,
+      "step": 19890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1489505449959463e-05,
+      "loss": 0.3406,
+      "step": 19900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.148500135122962e-05,
+      "loss": 0.3365,
+      "step": 19910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.148049725249978e-05,
+      "loss": 0.3371,
+      "step": 19920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.147599315376993e-05,
+      "loss": 0.3346,
+      "step": 19930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1471489055040087e-05,
+      "loss": 0.3373,
+      "step": 19940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1466984956310244e-05,
+      "loss": 0.339,
+      "step": 19950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.14624808575804e-05,
+      "loss": 0.3465,
+      "step": 19960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.145797675885056e-05,
+      "loss": 0.3341,
+      "step": 19970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.145347266012071e-05,
+      "loss": 0.3412,
+      "step": 19980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.144896856139087e-05,
+      "loss": 0.3432,
+      "step": 19990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1444464462661025e-05,
+      "loss": 0.3411,
+      "step": 20000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.143996036393118e-05,
+      "loss": 0.3474,
+      "step": 20010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.143545626520133e-05,
+      "loss": 0.3394,
+      "step": 20020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.143095216647149e-05,
+      "loss": 0.3452,
+      "step": 20030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.142644806774165e-05,
+      "loss": 0.3327,
+      "step": 20040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.14219439690118e-05,
+      "loss": 0.3392,
+      "step": 20050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.141743987028196e-05,
+      "loss": 0.3358,
+      "step": 20060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1412935771552114e-05,
+      "loss": 0.336,
+      "step": 20070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1408431672822264e-05,
+      "loss": 0.335,
+      "step": 20080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.140392757409243e-05,
+      "loss": 0.342,
+      "step": 20090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.139942347536258e-05,
+      "loss": 0.338,
+      "step": 20100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1394919376632744e-05,
+      "loss": 0.3317,
+      "step": 20110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1390415277902894e-05,
+      "loss": 0.3402,
+      "step": 20120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1385911179173045e-05,
+      "loss": 0.3419,
+      "step": 20130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.138140708044321e-05,
+      "loss": 0.3359,
+      "step": 20140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.137690298171336e-05,
+      "loss": 0.3329,
+      "step": 20150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.137239888298352e-05,
+      "loss": 0.3337,
+      "step": 20160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1367894784253675e-05,
+      "loss": 0.3338,
+      "step": 20170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1363390685523826e-05,
+      "loss": 0.3343,
+      "step": 20180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.135888658679398e-05,
+      "loss": 0.3347,
+      "step": 20190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.135483289793712e-05,
+      "loss": 0.3349,
+      "step": 20200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1350328799207284e-05,
+      "loss": 0.3325,
+      "step": 20210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1345824700477435e-05,
+      "loss": 0.3339,
+      "step": 20220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.134132060174759e-05,
+      "loss": 0.3342,
+      "step": 20230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.133681650301775e-05,
+      "loss": 0.3338,
+      "step": 20240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.13323124042879e-05,
+      "loss": 0.3314,
+      "step": 20250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.132780830555806e-05,
+      "loss": 0.3375,
+      "step": 20260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1323304206828216e-05,
+      "loss": 0.33,
+      "step": 20270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1318800108098373e-05,
+      "loss": 0.3387,
+      "step": 20280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1314296009368524e-05,
+      "loss": 0.3341,
+      "step": 20290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.130979191063868e-05,
+      "loss": 0.3338,
+      "step": 20300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.130528781190884e-05,
+      "loss": 0.3311,
+      "step": 20310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.130078371317899e-05,
+      "loss": 0.3425,
+      "step": 20320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1296279614449154e-05,
+      "loss": 0.3391,
+      "step": 20330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1291775515719305e-05,
+      "loss": 0.3381,
+      "step": 20340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.128727141698946e-05,
+      "loss": 0.3338,
+      "step": 20350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.128276731825962e-05,
+      "loss": 0.3356,
+      "step": 20360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.127826321952977e-05,
+      "loss": 0.3335,
+      "step": 20370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1273759120799935e-05,
+      "loss": 0.3394,
+      "step": 20380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1269255022070085e-05,
+      "loss": 0.3367,
+      "step": 20390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.126475092334024e-05,
+      "loss": 0.3345,
+      "step": 20400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.12602468246104e-05,
+      "loss": 0.3312,
+      "step": 20410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.125574272588055e-05,
+      "loss": 0.3371,
+      "step": 20420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.125123862715071e-05,
+      "loss": 0.3389,
+      "step": 20430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1246734528420866e-05,
+      "loss": 0.3319,
+      "step": 20440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.124223042969102e-05,
+      "loss": 0.3469,
+      "step": 20450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1237726330961174e-05,
+      "loss": 0.3356,
+      "step": 20460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.123322223223133e-05,
+      "loss": 0.3391,
+      "step": 20470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.122871813350149e-05,
+      "loss": 0.334,
+      "step": 20480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.122421403477165e-05,
+      "loss": 0.3292,
+      "step": 20490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.12197099360418e-05,
+      "loss": 0.3373,
+      "step": 20500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1215205837311955e-05,
+      "loss": 0.3405,
+      "step": 20510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.121070173858211e-05,
+      "loss": 0.3374,
+      "step": 20520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.120619763985227e-05,
+      "loss": 0.3315,
+      "step": 20530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.120169354112242e-05,
+      "loss": 0.336,
+      "step": 20540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.119718944239258e-05,
+      "loss": 0.3368,
+      "step": 20550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1192685343662736e-05,
+      "loss": 0.3298,
+      "step": 20560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1188181244932886e-05,
+      "loss": 0.3328,
+      "step": 20570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.118367714620305e-05,
+      "loss": 0.3318,
+      "step": 20580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.11791730474732e-05,
+      "loss": 0.3331,
+      "step": 20590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.117466894874336e-05,
+      "loss": 0.3307,
+      "step": 20600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1170164850013516e-05,
+      "loss": 0.3398,
+      "step": 20610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.116566075128367e-05,
+      "loss": 0.3381,
+      "step": 20620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.116115665255383e-05,
+      "loss": 0.3316,
+      "step": 20630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.115665255382398e-05,
+      "loss": 0.3366,
+      "step": 20640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.115214845509413e-05,
+      "loss": 0.3357,
+      "step": 20650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.11476443563643e-05,
+      "loss": 0.3376,
+      "step": 20660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.114314025763445e-05,
+      "loss": 0.3354,
+      "step": 20670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1138636158904605e-05,
+      "loss": 0.3334,
+      "step": 20680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.113413206017476e-05,
+      "loss": 0.3374,
+      "step": 20690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.1129627961444914e-05,
+      "loss": 0.3329,
+      "step": 20700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.112512386271507e-05,
+      "loss": 0.3349,
+      "step": 20710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.112061976398523e-05,
+      "loss": 0.3265,
+      "step": 20720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.1116115665255386e-05,
+      "loss": 0.3322,
+      "step": 20730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.111161156652554e-05,
+      "loss": 0.333,
+      "step": 20740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.1107107467795694e-05,
+      "loss": 0.3322,
+      "step": 20750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.110260336906585e-05,
+      "loss": 0.3298,
+      "step": 20760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.109809927033601e-05,
+      "loss": 0.3301,
+      "step": 20770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.109359517160617e-05,
+      "loss": 0.3334,
+      "step": 20780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.108909107287632e-05,
+      "loss": 0.3305,
+      "step": 20790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.1084586974146475e-05,
+      "loss": 0.3309,
+      "step": 20800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.108008287541663e-05,
+      "loss": 0.3269,
+      "step": 20810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.107557877668678e-05,
+      "loss": 0.3296,
+      "step": 20820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.107107467795695e-05,
+      "loss": 0.3337,
+      "step": 20830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.10665705792271e-05,
+      "loss": 0.3282,
+      "step": 20840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.106206648049725e-05,
+      "loss": 0.33,
+      "step": 20850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.105756238176741e-05,
+      "loss": 0.3382,
+      "step": 20860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.1053058283037564e-05,
+      "loss": 0.333,
+      "step": 20870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.104855418430772e-05,
+      "loss": 0.3343,
+      "step": 20880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.104405008557788e-05,
+      "loss": 0.3348,
+      "step": 20890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.103954598684803e-05,
+      "loss": 0.337,
+      "step": 20900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.1035041888118194e-05,
+      "loss": 0.3347,
+      "step": 20910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.1030537789388344e-05,
+      "loss": 0.3388,
+      "step": 20920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.10260336906585e-05,
+      "loss": 0.3309,
+      "step": 20930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.102152959192866e-05,
+      "loss": 0.3343,
+      "step": 20940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.101702549319881e-05,
+      "loss": 0.3313,
+      "step": 20950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.101252139446897e-05,
+      "loss": 0.3338,
+      "step": 20960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.1008017295739125e-05,
+      "loss": 0.3335,
+      "step": 20970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.100351319700928e-05,
+      "loss": 0.3333,
+      "step": 20980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.099900909827943e-05,
+      "loss": 0.3254,
+      "step": 20990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.099450499954959e-05,
+      "loss": 0.3358,
+      "step": 21000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.099000090081975e-05,
+      "loss": 0.3255,
+      "step": 21010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.09854968020899e-05,
+      "loss": 0.3249,
+      "step": 21020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.098099270336006e-05,
+      "loss": 0.3354,
+      "step": 21030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0976488604630214e-05,
+      "loss": 0.3305,
+      "step": 21040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.097198450590037e-05,
+      "loss": 0.3289,
+      "step": 21050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.096748040717053e-05,
+      "loss": 0.329,
+      "step": 21060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.096297630844068e-05,
+      "loss": 0.3291,
+      "step": 21070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0958472209710844e-05,
+      "loss": 0.33,
+      "step": 21080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0953968110980995e-05,
+      "loss": 0.3335,
+      "step": 21090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0949464012251145e-05,
+      "loss": 0.3314,
+      "step": 21100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.094495991352131e-05,
+      "loss": 0.3327,
+      "step": 21110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.094045581479146e-05,
+      "loss": 0.3264,
+      "step": 21120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.093595171606162e-05,
+      "loss": 0.3355,
+      "step": 21130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0931447617331775e-05,
+      "loss": 0.3264,
+      "step": 21140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0926943518601926e-05,
+      "loss": 0.3351,
+      "step": 21150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0922439419872084e-05,
+      "loss": 0.3299,
+      "step": 21160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.091793532114224e-05,
+      "loss": 0.3281,
+      "step": 21170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.09134312224124e-05,
+      "loss": 0.3278,
+      "step": 21180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0908927123682556e-05,
+      "loss": 0.3398,
+      "step": 21190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.090442302495271e-05,
+      "loss": 0.3311,
+      "step": 21200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0899918926222864e-05,
+      "loss": 0.3279,
+      "step": 21210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.089541482749302e-05,
+      "loss": 0.3308,
+      "step": 21220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.089091072876318e-05,
+      "loss": 0.3236,
+      "step": 21230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.088640663003333e-05,
+      "loss": 0.3353,
+      "step": 21240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.088190253130349e-05,
+      "loss": 0.3264,
+      "step": 21250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0877398432573645e-05,
+      "loss": 0.3267,
+      "step": 21260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0872894333843796e-05,
+      "loss": 0.3282,
+      "step": 21270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.086839023511396e-05,
+      "loss": 0.3275,
+      "step": 21280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.086388613638411e-05,
+      "loss": 0.3282,
+      "step": 21290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.085938203765426e-05,
+      "loss": 0.3346,
+      "step": 21300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0854877938924426e-05,
+      "loss": 0.3345,
+      "step": 21310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0850373840194576e-05,
+      "loss": 0.3285,
+      "step": 21320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.084586974146474e-05,
+      "loss": 0.3358,
+      "step": 21330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.084136564273489e-05,
+      "loss": 0.3315,
+      "step": 21340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.083686154400504e-05,
+      "loss": 0.3299,
+      "step": 21350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0832357445275206e-05,
+      "loss": 0.3239,
+      "step": 21360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.082785334654536e-05,
+      "loss": 0.3294,
+      "step": 21370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0823349247815515e-05,
+      "loss": 0.3318,
+      "step": 21380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.081884514908567e-05,
+      "loss": 0.3263,
+      "step": 21390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.081434105035582e-05,
+      "loss": 0.3332,
+      "step": 21400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.080983695162598e-05,
+      "loss": 0.3352,
+      "step": 21410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.080533285289614e-05,
+      "loss": 0.3323,
+      "step": 21420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0800828754166295e-05,
+      "loss": 0.3281,
+      "step": 21430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0796324655436446e-05,
+      "loss": 0.3353,
+      "step": 21440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0791820556706603e-05,
+      "loss": 0.3335,
+      "step": 21450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.078731645797676e-05,
+      "loss": 0.3312,
+      "step": 21460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.078281235924692e-05,
+      "loss": 0.3226,
+      "step": 21470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0778308260517076e-05,
+      "loss": 0.332,
+      "step": 21480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.077380416178723e-05,
+      "loss": 0.3284,
+      "step": 21490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0769300063057384e-05,
+      "loss": 0.3249,
+      "step": 21500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.076479596432754e-05,
+      "loss": 0.3286,
+      "step": 21510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.076029186559769e-05,
+      "loss": 0.3251,
+      "step": 21520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.075578776686786e-05,
+      "loss": 0.3265,
+      "step": 21530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.075128366813801e-05,
+      "loss": 0.3316,
+      "step": 21540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.074677956940816e-05,
+      "loss": 0.3282,
+      "step": 21550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.074227547067832e-05,
+      "loss": 0.3281,
+      "step": 21560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.073777137194847e-05,
+      "loss": 0.331,
+      "step": 21570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.073326727321863e-05,
+      "loss": 0.3293,
+      "step": 21580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.072876317448879e-05,
+      "loss": 0.3263,
+      "step": 21590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.072425907575894e-05,
+      "loss": 0.3311,
+      "step": 21600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.07197549770291e-05,
+      "loss": 0.3242,
+      "step": 21610
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0715250878299254e-05,
+      "loss": 0.3263,
+      "step": 21620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.071074677956941e-05,
+      "loss": 0.334,
+      "step": 21630
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.070624268083957e-05,
+      "loss": 0.3334,
+      "step": 21640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.070173858210972e-05,
+      "loss": 0.331,
+      "step": 21650
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.069723448337988e-05,
+      "loss": 0.3278,
+      "step": 21660
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0692730384650034e-05,
+      "loss": 0.3329,
+      "step": 21670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.068822628592019e-05,
+      "loss": 0.329,
+      "step": 21680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.068372218719034e-05,
+      "loss": 0.3315,
+      "step": 21690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.06792180884605e-05,
+      "loss": 0.3225,
+      "step": 21700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.067471398973066e-05,
+      "loss": 0.3292,
+      "step": 21710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.067020989100081e-05,
+      "loss": 0.332,
+      "step": 21720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.066570579227097e-05,
+      "loss": 0.3335,
+      "step": 21730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.066120169354112e-05,
+      "loss": 0.3238,
+      "step": 21740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.065669759481128e-05,
+      "loss": 0.3307,
+      "step": 21750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.065219349608144e-05,
+      "loss": 0.3233,
+      "step": 21760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.064768939735159e-05,
+      "loss": 0.328,
+      "step": 21770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0643185298621746e-05,
+      "loss": 0.3307,
+      "step": 21780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0638681199891904e-05,
+      "loss": 0.3299,
+      "step": 21790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0634177101162055e-05,
+      "loss": 0.3338,
+      "step": 21800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.062967300243222e-05,
+      "loss": 0.3227,
+      "step": 21810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.062516890370237e-05,
+      "loss": 0.3267,
+      "step": 21820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.062066480497253e-05,
+      "loss": 0.3248,
+      "step": 21830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.0616160706242685e-05,
+      "loss": 0.3272,
+      "step": 21840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0611656607512835e-05,
+      "loss": 0.3301,
+      "step": 21850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.060715250878299e-05,
+      "loss": 0.3314,
+      "step": 21860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.060264841005315e-05,
+      "loss": 0.3268,
+      "step": 21870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.059814431132331e-05,
+      "loss": 0.3316,
+      "step": 21880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0593640212593465e-05,
+      "loss": 0.3232,
+      "step": 21890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0589136113863616e-05,
+      "loss": 0.3289,
+      "step": 21900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0584632015133774e-05,
+      "loss": 0.3308,
+      "step": 21910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.058012791640393e-05,
+      "loss": 0.3298,
+      "step": 21920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.057562381767409e-05,
+      "loss": 0.3251,
+      "step": 21930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.057111971894424e-05,
+      "loss": 0.3257,
+      "step": 21940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.05666156202144e-05,
+      "loss": 0.3273,
+      "step": 21950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0562111521484554e-05,
+      "loss": 0.3267,
+      "step": 21960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0557607422754705e-05,
+      "loss": 0.3248,
+      "step": 21970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.055310332402487e-05,
+      "loss": 0.3237,
+      "step": 21980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.054859922529502e-05,
+      "loss": 0.3269,
+      "step": 21990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.054409512656517e-05,
+      "loss": 0.3223,
+      "step": 22000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0539591027835335e-05,
+      "loss": 0.3288,
+      "step": 22010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0535086929105486e-05,
+      "loss": 0.3247,
+      "step": 22020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.053058283037564e-05,
+      "loss": 0.3266,
+      "step": 22030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.05260787316458e-05,
+      "loss": 0.3303,
+      "step": 22040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.052157463291595e-05,
+      "loss": 0.3217,
+      "step": 22050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0517070534186116e-05,
+      "loss": 0.3264,
+      "step": 22060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0512566435456266e-05,
+      "loss": 0.3226,
+      "step": 22070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0508062336726424e-05,
+      "loss": 0.327,
+      "step": 22080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.050355823799658e-05,
+      "loss": 0.3259,
+      "step": 22090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.049905413926673e-05,
+      "loss": 0.3271,
+      "step": 22100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.049455004053689e-05,
+      "loss": 0.318,
+      "step": 22110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.049004594180705e-05,
+      "loss": 0.3322,
+      "step": 22120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0485541843077205e-05,
+      "loss": 0.3242,
+      "step": 22130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0481037744347355e-05,
+      "loss": 0.3297,
+      "step": 22140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.047653364561751e-05,
+      "loss": 0.3311,
+      "step": 22150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.047202954688767e-05,
+      "loss": 0.3224,
+      "step": 22160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.046752544815783e-05,
+      "loss": 0.3266,
+      "step": 22170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0463021349427985e-05,
+      "loss": 0.3144,
+      "step": 22180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0458517250698136e-05,
+      "loss": 0.3331,
+      "step": 22190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0454013151968293e-05,
+      "loss": 0.3235,
+      "step": 22200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.044950905323845e-05,
+      "loss": 0.3265,
+      "step": 22210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.044545536438159e-05,
+      "loss": 0.3343,
+      "step": 22220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0440951265651745e-05,
+      "loss": 0.332,
+      "step": 22230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.04364471669219e-05,
+      "loss": 0.322,
+      "step": 22240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.043194306819206e-05,
+      "loss": 0.3236,
+      "step": 22250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.042743896946221e-05,
+      "loss": 0.3323,
+      "step": 22260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.042293487073237e-05,
+      "loss": 0.3214,
+      "step": 22270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0418430772002526e-05,
+      "loss": 0.3247,
+      "step": 22280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.041392667327268e-05,
+      "loss": 0.3228,
+      "step": 22290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.040942257454284e-05,
+      "loss": 0.3202,
+      "step": 22300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.040491847581299e-05,
+      "loss": 0.3258,
+      "step": 22310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.040041437708314e-05,
+      "loss": 0.3278,
+      "step": 22320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.039591027835331e-05,
+      "loss": 0.3249,
+      "step": 22330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.039140617962346e-05,
+      "loss": 0.3299,
+      "step": 22340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0386902080893615e-05,
+      "loss": 0.331,
+      "step": 22350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.038239798216377e-05,
+      "loss": 0.3334,
+      "step": 22360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.037789388343392e-05,
+      "loss": 0.3193,
+      "step": 22370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.037338978470409e-05,
+      "loss": 0.3308,
+      "step": 22380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.036888568597424e-05,
+      "loss": 0.3264,
+      "step": 22390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0364381587244396e-05,
+      "loss": 0.3233,
+      "step": 22400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.035987748851455e-05,
+      "loss": 0.3292,
+      "step": 22410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0355373389784704e-05,
+      "loss": 0.3215,
+      "step": 22420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.035086929105486e-05,
+      "loss": 0.3234,
+      "step": 22430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.034636519232502e-05,
+      "loss": 0.3274,
+      "step": 22440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0341861093595176e-05,
+      "loss": 0.3207,
+      "step": 22450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.033735699486533e-05,
+      "loss": 0.3342,
+      "step": 22460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0332852896135485e-05,
+      "loss": 0.3247,
+      "step": 22470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.032834879740564e-05,
+      "loss": 0.3293,
+      "step": 22480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.032384469867579e-05,
+      "loss": 0.3278,
+      "step": 22490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.031934059994596e-05,
+      "loss": 0.3196,
+      "step": 22500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.031483650121611e-05,
+      "loss": 0.3278,
+      "step": 22510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0310332402486265e-05,
+      "loss": 0.3225,
+      "step": 22520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.030582830375642e-05,
+      "loss": 0.3291,
+      "step": 22530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0301324205026573e-05,
+      "loss": 0.3239,
+      "step": 22540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.029682010629674e-05,
+      "loss": 0.3267,
+      "step": 22550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.029231600756689e-05,
+      "loss": 0.3242,
+      "step": 22560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.028781190883704e-05,
+      "loss": 0.3183,
+      "step": 22570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0283307810107203e-05,
+      "loss": 0.3207,
+      "step": 22580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0278803711377354e-05,
+      "loss": 0.3222,
+      "step": 22590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.027429961264751e-05,
+      "loss": 0.3251,
+      "step": 22600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.026979551391767e-05,
+      "loss": 0.3291,
+      "step": 22610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.026529141518782e-05,
+      "loss": 0.3279,
+      "step": 22620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.026078731645798e-05,
+      "loss": 0.325,
+      "step": 22630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0256283217728135e-05,
+      "loss": 0.3203,
+      "step": 22640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.025177911899829e-05,
+      "loss": 0.3268,
+      "step": 22650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.024727502026845e-05,
+      "loss": 0.3249,
+      "step": 22660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.02427709215386e-05,
+      "loss": 0.3309,
+      "step": 22670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.023826682280876e-05,
+      "loss": 0.3212,
+      "step": 22680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0233762724078915e-05,
+      "loss": 0.3226,
+      "step": 22690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.022925862534907e-05,
+      "loss": 0.3264,
+      "step": 22700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0224754526619224e-05,
+      "loss": 0.335,
+      "step": 22710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.022025042788938e-05,
+      "loss": 0.3205,
+      "step": 22720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.021574632915954e-05,
+      "loss": 0.3239,
+      "step": 22730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.021124223042969e-05,
+      "loss": 0.3248,
+      "step": 22740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0206738131699854e-05,
+      "loss": 0.3204,
+      "step": 22750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0202234032970004e-05,
+      "loss": 0.3269,
+      "step": 22760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0197729934240155e-05,
+      "loss": 0.3226,
+      "step": 22770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.019322583551032e-05,
+      "loss": 0.3194,
+      "step": 22780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.018872173678047e-05,
+      "loss": 0.3319,
+      "step": 22790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.018421763805063e-05,
+      "loss": 0.323,
+      "step": 22800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0179713539320785e-05,
+      "loss": 0.3278,
+      "step": 22810
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0175209440590936e-05,
+      "loss": 0.3213,
+      "step": 22820
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.01707053418611e-05,
+      "loss": 0.3211,
+      "step": 22830
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.016620124313125e-05,
+      "loss": 0.32,
+      "step": 22840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.016169714440141e-05,
+      "loss": 0.3232,
+      "step": 22850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0157193045671566e-05,
+      "loss": 0.3282,
+      "step": 22860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0152688946941716e-05,
+      "loss": 0.3264,
+      "step": 22870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0148184848211874e-05,
+      "loss": 0.3207,
+      "step": 22880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.014368074948203e-05,
+      "loss": 0.3221,
+      "step": 22890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.013917665075219e-05,
+      "loss": 0.3235,
+      "step": 22900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.013467255202234e-05,
+      "loss": 0.3242,
+      "step": 22910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.01301684532925e-05,
+      "loss": 0.3273,
+      "step": 22920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0125664354562655e-05,
+      "loss": 0.3245,
+      "step": 22930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0121160255832805e-05,
+      "loss": 0.3207,
+      "step": 22940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.011665615710297e-05,
+      "loss": 0.3178,
+      "step": 22950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.011215205837312e-05,
+      "loss": 0.3209,
+      "step": 22960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.010764795964328e-05,
+      "loss": 0.3248,
+      "step": 22970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0103143860913435e-05,
+      "loss": 0.3171,
+      "step": 22980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0098639762183586e-05,
+      "loss": 0.3194,
+      "step": 22990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0094135663453744e-05,
+      "loss": 0.32,
+      "step": 23000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.00896315647239e-05,
+      "loss": 0.3213,
+      "step": 23010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.008512746599405e-05,
+      "loss": 0.322,
+      "step": 23020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0080623367264216e-05,
+      "loss": 0.3218,
+      "step": 23030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.007611926853437e-05,
+      "loss": 0.3196,
+      "step": 23040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0071615169804524e-05,
+      "loss": 0.3249,
+      "step": 23050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.006711107107468e-05,
+      "loss": 0.3237,
+      "step": 23060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.006260697234483e-05,
+      "loss": 0.318,
+      "step": 23070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.005810287361499e-05,
+      "loss": 0.3252,
+      "step": 23080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.005359877488515e-05,
+      "loss": 0.3189,
+      "step": 23090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0049094676155305e-05,
+      "loss": 0.3199,
+      "step": 23100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.004459057742546e-05,
+      "loss": 0.3179,
+      "step": 23110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.004008647869561e-05,
+      "loss": 0.3216,
+      "step": 23120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.003558237996577e-05,
+      "loss": 0.3203,
+      "step": 23130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.003107828123593e-05,
+      "loss": 0.3196,
+      "step": 23140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0026574182506086e-05,
+      "loss": 0.3151,
+      "step": 23150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0022070083776236e-05,
+      "loss": 0.3217,
+      "step": 23160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0017565985046394e-05,
+      "loss": 0.322,
+      "step": 23170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.001306188631655e-05,
+      "loss": 0.3177,
+      "step": 23180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.00085577875867e-05,
+      "loss": 0.323,
+      "step": 23190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.0004053688856866e-05,
+      "loss": 0.3198,
+      "step": 23200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.999954959012702e-05,
+      "loss": 0.3226,
+      "step": 23210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.999504549139717e-05,
+      "loss": 0.3186,
+      "step": 23220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.999054139266733e-05,
+      "loss": 0.3177,
+      "step": 23230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.998603729393748e-05,
+      "loss": 0.3241,
+      "step": 23240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.998153319520764e-05,
+      "loss": 0.3232,
+      "step": 23250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.99770290964778e-05,
+      "loss": 0.3224,
+      "step": 23260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.997252499774795e-05,
+      "loss": 0.3214,
+      "step": 23270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.996802089901811e-05,
+      "loss": 0.3302,
+      "step": 23280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.996351680028826e-05,
+      "loss": 0.3181,
+      "step": 23290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.995901270155842e-05,
+      "loss": 0.315,
+      "step": 23300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.995450860282858e-05,
+      "loss": 0.3283,
+      "step": 23310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.995000450409873e-05,
+      "loss": 0.3227,
+      "step": 23320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9945500405368887e-05,
+      "loss": 0.3239,
+      "step": 23330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9940996306639044e-05,
+      "loss": 0.3177,
+      "step": 23340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.99364922079092e-05,
+      "loss": 0.322,
+      "step": 23350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.993198810917935e-05,
+      "loss": 0.3237,
+      "step": 23360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.992748401044951e-05,
+      "loss": 0.3199,
+      "step": 23370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.992297991171967e-05,
+      "loss": 0.3228,
+      "step": 23380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9918475812989825e-05,
+      "loss": 0.3209,
+      "step": 23390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.991397171425998e-05,
+      "loss": 0.324,
+      "step": 23400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.990946761553013e-05,
+      "loss": 0.3147,
+      "step": 23410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.990496351680029e-05,
+      "loss": 0.3206,
+      "step": 23420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.990045941807045e-05,
+      "loss": 0.3182,
+      "step": 23430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.98959553193406e-05,
+      "loss": 0.3221,
+      "step": 23440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9891451220610756e-05,
+      "loss": 0.3219,
+      "step": 23450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9886947121880914e-05,
+      "loss": 0.3221,
+      "step": 23460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9882443023151064e-05,
+      "loss": 0.3232,
+      "step": 23470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.987793892442123e-05,
+      "loss": 0.3204,
+      "step": 23480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.987343482569138e-05,
+      "loss": 0.3179,
+      "step": 23490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.986893072696154e-05,
+      "loss": 0.3212,
+      "step": 23500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9864426628231694e-05,
+      "loss": 0.3177,
+      "step": 23510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9859922529501845e-05,
+      "loss": 0.3218,
+      "step": 23520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.985541843077201e-05,
+      "loss": 0.3161,
+      "step": 23530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.985091433204216e-05,
+      "loss": 0.3164,
+      "step": 23540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.984641023331232e-05,
+      "loss": 0.3243,
+      "step": 23550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9841906134582475e-05,
+      "loss": 0.321,
+      "step": 23560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9837402035852626e-05,
+      "loss": 0.3203,
+      "step": 23570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.983289793712278e-05,
+      "loss": 0.3165,
+      "step": 23580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.982839383839294e-05,
+      "loss": 0.3108,
+      "step": 23590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.98238897396631e-05,
+      "loss": 0.3153,
+      "step": 23600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.981938564093325e-05,
+      "loss": 0.3238,
+      "step": 23610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9814881542203406e-05,
+      "loss": 0.3163,
+      "step": 23620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9810377443473564e-05,
+      "loss": 0.3164,
+      "step": 23630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9805873344743715e-05,
+      "loss": 0.3169,
+      "step": 23640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.980136924601387e-05,
+      "loss": 0.3191,
+      "step": 23650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.979686514728403e-05,
+      "loss": 0.3203,
+      "step": 23660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.979236104855419e-05,
+      "loss": 0.3167,
+      "step": 23670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9787856949824345e-05,
+      "loss": 0.3174,
+      "step": 23680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9783352851094495e-05,
+      "loss": 0.3174,
+      "step": 23690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.977884875236465e-05,
+      "loss": 0.3175,
+      "step": 23700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.977434465363481e-05,
+      "loss": 0.3232,
+      "step": 23710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.976984055490496e-05,
+      "loss": 0.3225,
+      "step": 23720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9765336456175125e-05,
+      "loss": 0.3248,
+      "step": 23730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9760832357445276e-05,
+      "loss": 0.3256,
+      "step": 23740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9756328258715433e-05,
+      "loss": 0.3223,
+      "step": 23750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.975182415998559e-05,
+      "loss": 0.3194,
+      "step": 23760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.974732006125574e-05,
+      "loss": 0.3192,
+      "step": 23770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.97428159625259e-05,
+      "loss": 0.3152,
+      "step": 23780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.973831186379606e-05,
+      "loss": 0.3165,
+      "step": 23790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9733807765066214e-05,
+      "loss": 0.3174,
+      "step": 23800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.972930366633637e-05,
+      "loss": 0.3163,
+      "step": 23810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.972479956760652e-05,
+      "loss": 0.3199,
+      "step": 23820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.972029546887668e-05,
+      "loss": 0.3178,
+      "step": 23830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.971579137014684e-05,
+      "loss": 0.3179,
+      "step": 23840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9711287271416995e-05,
+      "loss": 0.3149,
+      "step": 23850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9706783172687146e-05,
+      "loss": 0.3227,
+      "step": 23860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.97022790739573e-05,
+      "loss": 0.3264,
+      "step": 23870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.969777497522746e-05,
+      "loss": 0.3189,
+      "step": 23880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.969327087649761e-05,
+      "loss": 0.318,
+      "step": 23890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.968876677776777e-05,
+      "loss": 0.3143,
+      "step": 23900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9684262679037926e-05,
+      "loss": 0.3224,
+      "step": 23910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.967975858030808e-05,
+      "loss": 0.323,
+      "step": 23920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.967525448157824e-05,
+      "loss": 0.3143,
+      "step": 23930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.967075038284839e-05,
+      "loss": 0.3161,
+      "step": 23940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.966624628411855e-05,
+      "loss": 0.3206,
+      "step": 23950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.966174218538871e-05,
+      "loss": 0.3204,
+      "step": 23960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.965723808665886e-05,
+      "loss": 0.3105,
+      "step": 23970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.965273398792902e-05,
+      "loss": 0.315,
+      "step": 23980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.964822988919917e-05,
+      "loss": 0.3221,
+      "step": 23990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.964372579046933e-05,
+      "loss": 0.3177,
+      "step": 24000
+    },
+    {
+      "epoch": 0.21,
+      "eval_NEIMS_canon_smiles": 0.9376666666666666,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.007333333333333333,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.3685876987924181,
+      "eval_NEIMS_exact_mols": 0.007316666666666667,
+      "eval_NEIMS_exact_smiles": 0.007,
+      "eval_NEIMS_loss": 0.32910066843032837,
+      "eval_NEIMS_matched_formulas": 0.34065,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.26411947676764336,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.007366666666666666,
+      "eval_NEIMS_runtime": 718.8898,
+      "eval_NEIMS_samples_per_second": 83.462,
+      "eval_NEIMS_steps_per_second": 1.305,
+      "step": 24000
+    },
+    {
+      "epoch": 0.21,
+      "eval_RASSP_canon_smiles": 0.9262029585601229,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.016562593915918123,
+      "eval_RASSP_daylight_tanimoto_simil": 0.4282559377231019,
+      "eval_RASSP_exact_mols": 0.01652920158947474,
+      "eval_RASSP_exact_smiles": 0.01564430493872508,
+      "eval_RASSP_loss": 0.2825922966003418,
+      "eval_RASSP_matched_formulas": 0.4687781747754366,
+      "eval_RASSP_morgan_tanimoto_simil": 0.2914407359801654,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.016562593915918123,
+      "eval_RASSP_runtime": 825.2247,
+      "eval_RASSP_samples_per_second": 72.579,
+      "eval_RASSP_steps_per_second": 1.134,
+      "step": 24000
+    },
+    {
+      "epoch": 0.21,
+      "eval_NIST_canon_smiles": 0.9001313127728289,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.002768215210987685,
+      "eval_NIST_daylight_tanimoto_simil": 0.23418501426051763,
+      "eval_NIST_exact_mols": 0.0026972353337828725,
+      "eval_NIST_exact_smiles": 0.0025907655179756538,
+      "eval_NIST_loss": 1.490809440612793,
+      "eval_NIST_matched_formulas": 0.03726443553252653,
+      "eval_NIST_morgan_tanimoto_simil": 0.1899051613413322,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.002768215210987685,
+      "eval_NIST_runtime": 446.8802,
+      "eval_NIST_samples_per_second": 63.053,
+      "eval_NIST_steps_per_second": 0.987,
+      "step": 24000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.963922169173949e-05,
+      "loss": 0.316,
+      "step": 24010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.963471759300964e-05,
+      "loss": 0.319,
+      "step": 24020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9630213494279796e-05,
+      "loss": 0.3164,
+      "step": 24030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.962570939554995e-05,
+      "loss": 0.3254,
+      "step": 24040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.962120529682011e-05,
+      "loss": 0.3206,
+      "step": 24050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.961670119809026e-05,
+      "loss": 0.3253,
+      "step": 24060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.961219709936042e-05,
+      "loss": 0.3162,
+      "step": 24070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.9607693000630577e-05,
+      "loss": 0.3183,
+      "step": 24080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9603188901900734e-05,
+      "loss": 0.3215,
+      "step": 24090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9598684803170885e-05,
+      "loss": 0.3141,
+      "step": 24100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.959418070444104e-05,
+      "loss": 0.3205,
+      "step": 24110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.95896766057112e-05,
+      "loss": 0.3134,
+      "step": 24120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.958517250698136e-05,
+      "loss": 0.3169,
+      "step": 24130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.958066840825151e-05,
+      "loss": 0.3232,
+      "step": 24140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9576164309521665e-05,
+      "loss": 0.3156,
+      "step": 24150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.957166021079182e-05,
+      "loss": 0.3207,
+      "step": 24160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9567156112061974e-05,
+      "loss": 0.3196,
+      "step": 24170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.956265201333214e-05,
+      "loss": 0.3154,
+      "step": 24180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.955814791460229e-05,
+      "loss": 0.318,
+      "step": 24190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9553643815872446e-05,
+      "loss": 0.315,
+      "step": 24200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9549139717142604e-05,
+      "loss": 0.3078,
+      "step": 24210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9544635618412754e-05,
+      "loss": 0.3179,
+      "step": 24220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.954013151968292e-05,
+      "loss": 0.316,
+      "step": 24230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.953562742095307e-05,
+      "loss": 0.3144,
+      "step": 24240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.953112332222323e-05,
+      "loss": 0.3149,
+      "step": 24250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9526619223493384e-05,
+      "loss": 0.3248,
+      "step": 24260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9522115124763535e-05,
+      "loss": 0.3198,
+      "step": 24270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.951761102603369e-05,
+      "loss": 0.313,
+      "step": 24280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.951355733717683e-05,
+      "loss": 0.3137,
+      "step": 24290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9509053238446994e-05,
+      "loss": 0.3215,
+      "step": 24300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9504549139717144e-05,
+      "loss": 0.3178,
+      "step": 24310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.95000450409873e-05,
+      "loss": 0.3239,
+      "step": 24320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.949554094225746e-05,
+      "loss": 0.3224,
+      "step": 24330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.949103684352761e-05,
+      "loss": 0.314,
+      "step": 24340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.948653274479777e-05,
+      "loss": 0.3155,
+      "step": 24350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9482028646067925e-05,
+      "loss": 0.3202,
+      "step": 24360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.947752454733808e-05,
+      "loss": 0.3158,
+      "step": 24370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.947302044860823e-05,
+      "loss": 0.3103,
+      "step": 24380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.946851634987839e-05,
+      "loss": 0.3179,
+      "step": 24390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.946401225114855e-05,
+      "loss": 0.3134,
+      "step": 24400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.94595081524187e-05,
+      "loss": 0.3134,
+      "step": 24410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9455004053688857e-05,
+      "loss": 0.3138,
+      "step": 24420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9450499954959014e-05,
+      "loss": 0.3162,
+      "step": 24430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.944599585622917e-05,
+      "loss": 0.321,
+      "step": 24440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.944149175749933e-05,
+      "loss": 0.3236,
+      "step": 24450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.943698765876948e-05,
+      "loss": 0.3186,
+      "step": 24460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.943248356003964e-05,
+      "loss": 0.3133,
+      "step": 24470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9427979461309795e-05,
+      "loss": 0.3145,
+      "step": 24480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9423475362579945e-05,
+      "loss": 0.3145,
+      "step": 24490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.941897126385011e-05,
+      "loss": 0.3196,
+      "step": 24500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.941446716512026e-05,
+      "loss": 0.3101,
+      "step": 24510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.940996306639042e-05,
+      "loss": 0.3182,
+      "step": 24520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9405458967660575e-05,
+      "loss": 0.3143,
+      "step": 24530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9400954868930726e-05,
+      "loss": 0.312,
+      "step": 24540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9396450770200884e-05,
+      "loss": 0.3163,
+      "step": 24550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.939194667147104e-05,
+      "loss": 0.3113,
+      "step": 24560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.93874425727412e-05,
+      "loss": 0.3151,
+      "step": 24570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9382938474011356e-05,
+      "loss": 0.3056,
+      "step": 24580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.937843437528151e-05,
+      "loss": 0.3195,
+      "step": 24590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9373930276551664e-05,
+      "loss": 0.3125,
+      "step": 24600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.936942617782182e-05,
+      "loss": 0.3172,
+      "step": 24610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.936492207909198e-05,
+      "loss": 0.3199,
+      "step": 24620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.936041798036213e-05,
+      "loss": 0.3148,
+      "step": 24630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.935591388163229e-05,
+      "loss": 0.3155,
+      "step": 24640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9351409782902445e-05,
+      "loss": 0.3131,
+      "step": 24650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9346905684172596e-05,
+      "loss": 0.3094,
+      "step": 24660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.934240158544275e-05,
+      "loss": 0.3178,
+      "step": 24670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.933789748671291e-05,
+      "loss": 0.3211,
+      "step": 24680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.933339338798306e-05,
+      "loss": 0.3188,
+      "step": 24690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9328889289253226e-05,
+      "loss": 0.3162,
+      "step": 24700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9324385190523376e-05,
+      "loss": 0.3115,
+      "step": 24710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9319881091793534e-05,
+      "loss": 0.3129,
+      "step": 24720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.931537699306369e-05,
+      "loss": 0.3114,
+      "step": 24730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.931087289433384e-05,
+      "loss": 0.3133,
+      "step": 24740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9306368795604006e-05,
+      "loss": 0.3105,
+      "step": 24750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.930186469687416e-05,
+      "loss": 0.3108,
+      "step": 24760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9297360598144315e-05,
+      "loss": 0.3145,
+      "step": 24770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.929285649941447e-05,
+      "loss": 0.3121,
+      "step": 24780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.928835240068462e-05,
+      "loss": 0.3154,
+      "step": 24790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.928384830195478e-05,
+      "loss": 0.3132,
+      "step": 24800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.927934420322494e-05,
+      "loss": 0.3159,
+      "step": 24810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9274840104495095e-05,
+      "loss": 0.3214,
+      "step": 24820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9270336005765246e-05,
+      "loss": 0.3137,
+      "step": 24830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9265831907035403e-05,
+      "loss": 0.309,
+      "step": 24840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.926132780830556e-05,
+      "loss": 0.3141,
+      "step": 24850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.925682370957572e-05,
+      "loss": 0.3146,
+      "step": 24860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.925231961084587e-05,
+      "loss": 0.3129,
+      "step": 24870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.924781551211603e-05,
+      "loss": 0.3145,
+      "step": 24880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9243311413386184e-05,
+      "loss": 0.3125,
+      "step": 24890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.923880731465634e-05,
+      "loss": 0.3172,
+      "step": 24900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.923430321592649e-05,
+      "loss": 0.3169,
+      "step": 24910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.922979911719665e-05,
+      "loss": 0.3151,
+      "step": 24920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.922529501846681e-05,
+      "loss": 0.311,
+      "step": 24930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.922079091973696e-05,
+      "loss": 0.312,
+      "step": 24940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.921628682100712e-05,
+      "loss": 0.3155,
+      "step": 24950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.921178272227727e-05,
+      "loss": 0.3092,
+      "step": 24960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.920727862354743e-05,
+      "loss": 0.3161,
+      "step": 24970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.920277452481759e-05,
+      "loss": 0.3059,
+      "step": 24980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.919827042608774e-05,
+      "loss": 0.3134,
+      "step": 24990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.91937663273579e-05,
+      "loss": 0.3161,
+      "step": 25000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9189262228628054e-05,
+      "loss": 0.3102,
+      "step": 25010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.918475812989821e-05,
+      "loss": 0.3051,
+      "step": 25020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.918025403116837e-05,
+      "loss": 0.3113,
+      "step": 25030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.917574993243852e-05,
+      "loss": 0.3104,
+      "step": 25040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.917124583370868e-05,
+      "loss": 0.3073,
+      "step": 25050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9166741734978834e-05,
+      "loss": 0.3092,
+      "step": 25060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9162237636248985e-05,
+      "loss": 0.3188,
+      "step": 25070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.915773353751914e-05,
+      "loss": 0.3166,
+      "step": 25080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.91532294387893e-05,
+      "loss": 0.3134,
+      "step": 25090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.914872534005946e-05,
+      "loss": 0.3108,
+      "step": 25100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.914422124132961e-05,
+      "loss": 0.3153,
+      "step": 25110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9139717142599766e-05,
+      "loss": 0.3111,
+      "step": 25120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.913521304386992e-05,
+      "loss": 0.3162,
+      "step": 25130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.913070894514008e-05,
+      "loss": 0.3134,
+      "step": 25140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.912620484641024e-05,
+      "loss": 0.3127,
+      "step": 25150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.912170074768039e-05,
+      "loss": 0.3114,
+      "step": 25160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9117196648950546e-05,
+      "loss": 0.3103,
+      "step": 25170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9112692550220704e-05,
+      "loss": 0.3133,
+      "step": 25180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9108188451490855e-05,
+      "loss": 0.3172,
+      "step": 25190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.910368435276102e-05,
+      "loss": 0.3117,
+      "step": 25200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.909918025403117e-05,
+      "loss": 0.3221,
+      "step": 25210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.909467615530133e-05,
+      "loss": 0.3146,
+      "step": 25220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9090172056571485e-05,
+      "loss": 0.3124,
+      "step": 25230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9085667957841635e-05,
+      "loss": 0.3141,
+      "step": 25240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.908116385911179e-05,
+      "loss": 0.3078,
+      "step": 25250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.907665976038195e-05,
+      "loss": 0.3115,
+      "step": 25260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.907215566165211e-05,
+      "loss": 0.3131,
+      "step": 25270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.906765156292226e-05,
+      "loss": 0.3094,
+      "step": 25280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9063147464192416e-05,
+      "loss": 0.3101,
+      "step": 25290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9058643365462574e-05,
+      "loss": 0.315,
+      "step": 25300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.905413926673273e-05,
+      "loss": 0.3082,
+      "step": 25310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.904963516800288e-05,
+      "loss": 0.3115,
+      "step": 25320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.904513106927304e-05,
+      "loss": 0.3115,
+      "step": 25330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.90406269705432e-05,
+      "loss": 0.3106,
+      "step": 25340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9036122871813354e-05,
+      "loss": 0.3141,
+      "step": 25350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9031618773083505e-05,
+      "loss": 0.3097,
+      "step": 25360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.902711467435366e-05,
+      "loss": 0.3074,
+      "step": 25370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.902261057562382e-05,
+      "loss": 0.31,
+      "step": 25380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.901810647689397e-05,
+      "loss": 0.3189,
+      "step": 25390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9013602378164135e-05,
+      "loss": 0.3078,
+      "step": 25400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9009098279434286e-05,
+      "loss": 0.3092,
+      "step": 25410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.900459418070444e-05,
+      "loss": 0.315,
+      "step": 25420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.90000900819746e-05,
+      "loss": 0.3127,
+      "step": 25430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.899558598324475e-05,
+      "loss": 0.3092,
+      "step": 25440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8991081884514916e-05,
+      "loss": 0.3105,
+      "step": 25450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8986577785785066e-05,
+      "loss": 0.3093,
+      "step": 25460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8982073687055224e-05,
+      "loss": 0.3113,
+      "step": 25470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.897756958832538e-05,
+      "loss": 0.3123,
+      "step": 25480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.897306548959553e-05,
+      "loss": 0.3169,
+      "step": 25490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.896856139086569e-05,
+      "loss": 0.3172,
+      "step": 25500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.896405729213585e-05,
+      "loss": 0.309,
+      "step": 25510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8959553193406e-05,
+      "loss": 0.3057,
+      "step": 25520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8955049094676155e-05,
+      "loss": 0.3108,
+      "step": 25530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.895054499594631e-05,
+      "loss": 0.315,
+      "step": 25540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.894604089721647e-05,
+      "loss": 0.3135,
+      "step": 25550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.894153679848662e-05,
+      "loss": 0.3102,
+      "step": 25560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.893703269975678e-05,
+      "loss": 0.3095,
+      "step": 25570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8932528601026936e-05,
+      "loss": 0.3139,
+      "step": 25580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.892802450229709e-05,
+      "loss": 0.3063,
+      "step": 25590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.892352040356725e-05,
+      "loss": 0.3091,
+      "step": 25600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.89190163048374e-05,
+      "loss": 0.3127,
+      "step": 25610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.891451220610756e-05,
+      "loss": 0.3073,
+      "step": 25620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8910008107377717e-05,
+      "loss": 0.3055,
+      "step": 25630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.890550400864787e-05,
+      "loss": 0.3154,
+      "step": 25640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.890099990991803e-05,
+      "loss": 0.3124,
+      "step": 25650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.889649581118818e-05,
+      "loss": 0.3184,
+      "step": 25660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.889199171245834e-05,
+      "loss": 0.3109,
+      "step": 25670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.88874876137285e-05,
+      "loss": 0.3122,
+      "step": 25680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.888298351499865e-05,
+      "loss": 0.3136,
+      "step": 25690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8878479416268805e-05,
+      "loss": 0.3128,
+      "step": 25700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.887397531753896e-05,
+      "loss": 0.3196,
+      "step": 25710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8869471218809114e-05,
+      "loss": 0.3121,
+      "step": 25720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.886496712007928e-05,
+      "loss": 0.3063,
+      "step": 25730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.886046302134943e-05,
+      "loss": 0.3109,
+      "step": 25740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8855958922619586e-05,
+      "loss": 0.3075,
+      "step": 25750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8851454823889744e-05,
+      "loss": 0.3034,
+      "step": 25760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8846950725159894e-05,
+      "loss": 0.3074,
+      "step": 25770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.884244662643005e-05,
+      "loss": 0.3093,
+      "step": 25780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.883794252770021e-05,
+      "loss": 0.3067,
+      "step": 25790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.883343842897037e-05,
+      "loss": 0.2994,
+      "step": 25800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.882893433024052e-05,
+      "loss": 0.3055,
+      "step": 25810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8824430231510675e-05,
+      "loss": 0.3072,
+      "step": 25820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.881992613278083e-05,
+      "loss": 0.3123,
+      "step": 25830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.881542203405098e-05,
+      "loss": 0.31,
+      "step": 25840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.881091793532115e-05,
+      "loss": 0.308,
+      "step": 25850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.88064138365913e-05,
+      "loss": 0.3071,
+      "step": 25860
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8801909737861456e-05,
+      "loss": 0.3164,
+      "step": 25870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.879740563913161e-05,
+      "loss": 0.3041,
+      "step": 25880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8792901540401764e-05,
+      "loss": 0.313,
+      "step": 25890
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.878839744167193e-05,
+      "loss": 0.3088,
+      "step": 25900
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.878389334294208e-05,
+      "loss": 0.3098,
+      "step": 25910
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8779389244212236e-05,
+      "loss": 0.3083,
+      "step": 25920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8774885145482394e-05,
+      "loss": 0.3033,
+      "step": 25930
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8770381046752545e-05,
+      "loss": 0.3109,
+      "step": 25940
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.87658769480227e-05,
+      "loss": 0.3105,
+      "step": 25950
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.876137284929286e-05,
+      "loss": 0.3096,
+      "step": 25960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.875686875056301e-05,
+      "loss": 0.3016,
+      "step": 25970
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.875236465183317e-05,
+      "loss": 0.3023,
+      "step": 25980
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8747860553103325e-05,
+      "loss": 0.3146,
+      "step": 25990
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.874335645437348e-05,
+      "loss": 0.3102,
+      "step": 26000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.873885235564364e-05,
+      "loss": 0.3155,
+      "step": 26010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.873434825691379e-05,
+      "loss": 0.3086,
+      "step": 26020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.872984415818395e-05,
+      "loss": 0.3138,
+      "step": 26030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8725340059454106e-05,
+      "loss": 0.3133,
+      "step": 26040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8720835960724263e-05,
+      "loss": 0.3084,
+      "step": 26050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8716331861994414e-05,
+      "loss": 0.3169,
+      "step": 26060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.871182776326457e-05,
+      "loss": 0.31,
+      "step": 26070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.870732366453473e-05,
+      "loss": 0.3094,
+      "step": 26080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.870281956580488e-05,
+      "loss": 0.3052,
+      "step": 26090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8698315467075044e-05,
+      "loss": 0.3107,
+      "step": 26100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8693811368345195e-05,
+      "loss": 0.3095,
+      "step": 26110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.868930726961535e-05,
+      "loss": 0.3055,
+      "step": 26120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.868480317088551e-05,
+      "loss": 0.31,
+      "step": 26130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.868029907215566e-05,
+      "loss": 0.3091,
+      "step": 26140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8675794973425825e-05,
+      "loss": 0.3082,
+      "step": 26150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8671290874695976e-05,
+      "loss": 0.309,
+      "step": 26160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8666786775966126e-05,
+      "loss": 0.3068,
+      "step": 26170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.866228267723629e-05,
+      "loss": 0.3072,
+      "step": 26180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.865777857850644e-05,
+      "loss": 0.3077,
+      "step": 26190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.86532744797766e-05,
+      "loss": 0.3074,
+      "step": 26200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8648770381046756e-05,
+      "loss": 0.3059,
+      "step": 26210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.864426628231691e-05,
+      "loss": 0.3079,
+      "step": 26220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8639762183587064e-05,
+      "loss": 0.3095,
+      "step": 26230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.863525808485722e-05,
+      "loss": 0.3059,
+      "step": 26240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.863075398612738e-05,
+      "loss": 0.3079,
+      "step": 26250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.862624988739753e-05,
+      "loss": 0.3115,
+      "step": 26260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.862174578866769e-05,
+      "loss": 0.3133,
+      "step": 26270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8617241689937845e-05,
+      "loss": 0.3111,
+      "step": 26280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.8612737591208e-05,
+      "loss": 0.3107,
+      "step": 26290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.860823349247816e-05,
+      "loss": 0.3084,
+      "step": 26300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.860372939374831e-05,
+      "loss": 0.3124,
+      "step": 26310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.859922529501847e-05,
+      "loss": 0.3085,
+      "step": 26320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8594721196288626e-05,
+      "loss": 0.3092,
+      "step": 26330
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8590217097558777e-05,
+      "loss": 0.306,
+      "step": 26340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.858571299882894e-05,
+      "loss": 0.311,
+      "step": 26350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.858120890009909e-05,
+      "loss": 0.3113,
+      "step": 26360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.857670480136924e-05,
+      "loss": 0.3109,
+      "step": 26370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8572200702639407e-05,
+      "loss": 0.3101,
+      "step": 26380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.856769660390956e-05,
+      "loss": 0.3131,
+      "step": 26390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8563192505179715e-05,
+      "loss": 0.3084,
+      "step": 26400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.855868840644987e-05,
+      "loss": 0.3035,
+      "step": 26410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.855418430772002e-05,
+      "loss": 0.3051,
+      "step": 26420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.854968020899019e-05,
+      "loss": 0.3102,
+      "step": 26430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.854517611026034e-05,
+      "loss": 0.3084,
+      "step": 26440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8540672011530495e-05,
+      "loss": 0.3219,
+      "step": 26450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.853616791280065e-05,
+      "loss": 0.3117,
+      "step": 26460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8531663814070804e-05,
+      "loss": 0.3022,
+      "step": 26470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.852761012521395e-05,
+      "loss": 0.3083,
+      "step": 26480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.85231060264841e-05,
+      "loss": 0.31,
+      "step": 26490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.851860192775426e-05,
+      "loss": 0.3057,
+      "step": 26500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.851409782902441e-05,
+      "loss": 0.307,
+      "step": 26510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.850959373029457e-05,
+      "loss": 0.3074,
+      "step": 26520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.850508963156473e-05,
+      "loss": 0.3055,
+      "step": 26530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.850058553283488e-05,
+      "loss": 0.3101,
+      "step": 26540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8496081434105036e-05,
+      "loss": 0.3131,
+      "step": 26550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8491577335375194e-05,
+      "loss": 0.3098,
+      "step": 26560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.848707323664535e-05,
+      "loss": 0.3093,
+      "step": 26570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.84825691379155e-05,
+      "loss": 0.3122,
+      "step": 26580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.847806503918566e-05,
+      "loss": 0.3041,
+      "step": 26590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.847356094045582e-05,
+      "loss": 0.3022,
+      "step": 26600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.846905684172597e-05,
+      "loss": 0.3074,
+      "step": 26610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.846455274299613e-05,
+      "loss": 0.3117,
+      "step": 26620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.846004864426628e-05,
+      "loss": 0.3013,
+      "step": 26630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.845554454553644e-05,
+      "loss": 0.307,
+      "step": 26640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.84510404468066e-05,
+      "loss": 0.3117,
+      "step": 26650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.844653634807675e-05,
+      "loss": 0.3129,
+      "step": 26660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.844203224934691e-05,
+      "loss": 0.3017,
+      "step": 26670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.843752815061706e-05,
+      "loss": 0.3027,
+      "step": 26680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.843302405188722e-05,
+      "loss": 0.3004,
+      "step": 26690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.842851995315738e-05,
+      "loss": 0.3088,
+      "step": 26700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.842401585442753e-05,
+      "loss": 0.304,
+      "step": 26710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8419511755697687e-05,
+      "loss": 0.3101,
+      "step": 26720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8415007656967844e-05,
+      "loss": 0.3071,
+      "step": 26730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8410503558237995e-05,
+      "loss": 0.306,
+      "step": 26740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.840599945950815e-05,
+      "loss": 0.3052,
+      "step": 26750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.840149536077831e-05,
+      "loss": 0.31,
+      "step": 26760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.839699126204847e-05,
+      "loss": 0.3043,
+      "step": 26770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8392487163318625e-05,
+      "loss": 0.3106,
+      "step": 26780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8387983064588775e-05,
+      "loss": 0.3052,
+      "step": 26790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.838347896585893e-05,
+      "loss": 0.2995,
+      "step": 26800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.837897486712909e-05,
+      "loss": 0.3132,
+      "step": 26810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.837447076839925e-05,
+      "loss": 0.3027,
+      "step": 26820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.83699666696694e-05,
+      "loss": 0.3015,
+      "step": 26830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8365462570939556e-05,
+      "loss": 0.3066,
+      "step": 26840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8360958472209714e-05,
+      "loss": 0.304,
+      "step": 26850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8356454373479864e-05,
+      "loss": 0.3109,
+      "step": 26860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.835195027475003e-05,
+      "loss": 0.3013,
+      "step": 26870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.834744617602018e-05,
+      "loss": 0.3091,
+      "step": 26880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.834294207729034e-05,
+      "loss": 0.3025,
+      "step": 26890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8338437978560494e-05,
+      "loss": 0.3101,
+      "step": 26900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8333933879830645e-05,
+      "loss": 0.3027,
+      "step": 26910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.832942978110081e-05,
+      "loss": 0.3069,
+      "step": 26920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.832492568237096e-05,
+      "loss": 0.3054,
+      "step": 26930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.832042158364111e-05,
+      "loss": 0.3134,
+      "step": 26940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8315917484911275e-05,
+      "loss": 0.3082,
+      "step": 26950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8311413386181426e-05,
+      "loss": 0.302,
+      "step": 26960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.830690928745158e-05,
+      "loss": 0.3031,
+      "step": 26970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.830240518872174e-05,
+      "loss": 0.3091,
+      "step": 26980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.829790108999189e-05,
+      "loss": 0.3073,
+      "step": 26990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.829339699126205e-05,
+      "loss": 0.3036,
+      "step": 27000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8288892892532206e-05,
+      "loss": 0.3055,
+      "step": 27010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8284388793802364e-05,
+      "loss": 0.3068,
+      "step": 27020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8279884695072515e-05,
+      "loss": 0.3074,
+      "step": 27030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.827538059634267e-05,
+      "loss": 0.3071,
+      "step": 27040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.827087649761283e-05,
+      "loss": 0.3066,
+      "step": 27050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.826637239888299e-05,
+      "loss": 0.307,
+      "step": 27060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8261868300153145e-05,
+      "loss": 0.3037,
+      "step": 27070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8257364201423295e-05,
+      "loss": 0.3058,
+      "step": 27080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.825286010269345e-05,
+      "loss": 0.3124,
+      "step": 27090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.824835600396361e-05,
+      "loss": 0.3119,
+      "step": 27100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.824385190523376e-05,
+      "loss": 0.3111,
+      "step": 27110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8239347806503925e-05,
+      "loss": 0.306,
+      "step": 27120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8234843707774076e-05,
+      "loss": 0.3035,
+      "step": 27130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.823033960904423e-05,
+      "loss": 0.305,
+      "step": 27140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.822583551031439e-05,
+      "loss": 0.3047,
+      "step": 27150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.822133141158454e-05,
+      "loss": 0.3051,
+      "step": 27160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.82168273128547e-05,
+      "loss": 0.3003,
+      "step": 27170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.821232321412486e-05,
+      "loss": 0.306,
+      "step": 27180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.820781911539501e-05,
+      "loss": 0.2992,
+      "step": 27190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.820331501666517e-05,
+      "loss": 0.2994,
+      "step": 27200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.819881091793532e-05,
+      "loss": 0.2994,
+      "step": 27210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.819430681920548e-05,
+      "loss": 0.304,
+      "step": 27220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.818980272047564e-05,
+      "loss": 0.3006,
+      "step": 27230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.818529862174579e-05,
+      "loss": 0.301,
+      "step": 27240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8180794523015946e-05,
+      "loss": 0.3035,
+      "step": 27250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.81762904242861e-05,
+      "loss": 0.3076,
+      "step": 27260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.817178632555626e-05,
+      "loss": 0.3085,
+      "step": 27270
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.816728222682641e-05,
+      "loss": 0.3026,
+      "step": 27280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.816277812809657e-05,
+      "loss": 0.3156,
+      "step": 27290
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8158274029366726e-05,
+      "loss": 0.3162,
+      "step": 27300
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.815376993063688e-05,
+      "loss": 0.3032,
+      "step": 27310
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.814926583190704e-05,
+      "loss": 0.301,
+      "step": 27320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.814476173317719e-05,
+      "loss": 0.3062,
+      "step": 27330
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.814025763444735e-05,
+      "loss": 0.3056,
+      "step": 27340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.813575353571751e-05,
+      "loss": 0.3089,
+      "step": 27350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.813124943698766e-05,
+      "loss": 0.3042,
+      "step": 27360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.812674533825782e-05,
+      "loss": 0.3117,
+      "step": 27370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.812224123952797e-05,
+      "loss": 0.3164,
+      "step": 27380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.811773714079812e-05,
+      "loss": 0.308,
+      "step": 27390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.811323304206829e-05,
+      "loss": 0.3097,
+      "step": 27400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.810872894333844e-05,
+      "loss": 0.3051,
+      "step": 27410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8104224844608596e-05,
+      "loss": 0.3082,
+      "step": 27420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.809972074587875e-05,
+      "loss": 0.3054,
+      "step": 27430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.8095216647148904e-05,
+      "loss": 0.3041,
+      "step": 27440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.809071254841906e-05,
+      "loss": 0.3059,
+      "step": 27450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.808620844968922e-05,
+      "loss": 0.3093,
+      "step": 27460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8081704350959376e-05,
+      "loss": 0.3053,
+      "step": 27470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8077200252229534e-05,
+      "loss": 0.3015,
+      "step": 27480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8072696153499685e-05,
+      "loss": 0.3031,
+      "step": 27490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.806819205476984e-05,
+      "loss": 0.3068,
+      "step": 27500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.806368795604e-05,
+      "loss": 0.304,
+      "step": 27510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.805918385731016e-05,
+      "loss": 0.3024,
+      "step": 27520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.805467975858031e-05,
+      "loss": 0.3055,
+      "step": 27530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8050175659850465e-05,
+      "loss": 0.3051,
+      "step": 27540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.804567156112062e-05,
+      "loss": 0.2969,
+      "step": 27550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8041167462390774e-05,
+      "loss": 0.3,
+      "step": 27560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.803666336366094e-05,
+      "loss": 0.3044,
+      "step": 27570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.803215926493109e-05,
+      "loss": 0.3019,
+      "step": 27580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.802765516620124e-05,
+      "loss": 0.2947,
+      "step": 27590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8023151067471404e-05,
+      "loss": 0.3017,
+      "step": 27600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8018646968741554e-05,
+      "loss": 0.3059,
+      "step": 27610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.801414287001171e-05,
+      "loss": 0.3012,
+      "step": 27620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.800963877128187e-05,
+      "loss": 0.3039,
+      "step": 27630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.800513467255202e-05,
+      "loss": 0.2997,
+      "step": 27640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.8000630573822184e-05,
+      "loss": 0.3026,
+      "step": 27650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7996126475092335e-05,
+      "loss": 0.2984,
+      "step": 27660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.799162237636249e-05,
+      "loss": 0.2984,
+      "step": 27670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.798711827763265e-05,
+      "loss": 0.3042,
+      "step": 27680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.79826141789028e-05,
+      "loss": 0.3026,
+      "step": 27690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.797811008017296e-05,
+      "loss": 0.3016,
+      "step": 27700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7973605981443116e-05,
+      "loss": 0.3067,
+      "step": 27710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.796910188271327e-05,
+      "loss": 0.299,
+      "step": 27720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7964597783983424e-05,
+      "loss": 0.2985,
+      "step": 27730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.796009368525358e-05,
+      "loss": 0.3074,
+      "step": 27740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.795558958652374e-05,
+      "loss": 0.3078,
+      "step": 27750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.795108548779389e-05,
+      "loss": 0.3028,
+      "step": 27760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7946581389064054e-05,
+      "loss": 0.3047,
+      "step": 27770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7942077290334205e-05,
+      "loss": 0.3051,
+      "step": 27780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.793757319160436e-05,
+      "loss": 0.3047,
+      "step": 27790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.793306909287452e-05,
+      "loss": 0.3057,
+      "step": 27800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.792856499414467e-05,
+      "loss": 0.3055,
+      "step": 27810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7924060895414835e-05,
+      "loss": 0.3033,
+      "step": 27820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7919556796684985e-05,
+      "loss": 0.2986,
+      "step": 27830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7915052697955136e-05,
+      "loss": 0.305,
+      "step": 27840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.79105485992253e-05,
+      "loss": 0.304,
+      "step": 27850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.790604450049545e-05,
+      "loss": 0.3072,
+      "step": 27860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.790154040176561e-05,
+      "loss": 0.3019,
+      "step": 27870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7897036303035766e-05,
+      "loss": 0.3012,
+      "step": 27880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.789253220430592e-05,
+      "loss": 0.308,
+      "step": 27890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7888028105576074e-05,
+      "loss": 0.2982,
+      "step": 27900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.788352400684623e-05,
+      "loss": 0.3063,
+      "step": 27910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.787901990811639e-05,
+      "loss": 0.306,
+      "step": 27920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7874515809386547e-05,
+      "loss": 0.2977,
+      "step": 27930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.78700117106567e-05,
+      "loss": 0.302,
+      "step": 27940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7865507611926855e-05,
+      "loss": 0.3005,
+      "step": 27950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.786100351319701e-05,
+      "loss": 0.2951,
+      "step": 27960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.785649941446717e-05,
+      "loss": 0.3032,
+      "step": 27970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.785199531573732e-05,
+      "loss": 0.2984,
+      "step": 27980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.784749121700748e-05,
+      "loss": 0.3032,
+      "step": 27990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7842987118277635e-05,
+      "loss": 0.3019,
+      "step": 28000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7838483019547786e-05,
+      "loss": 0.2992,
+      "step": 28010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.783397892081795e-05,
+      "loss": 0.3086,
+      "step": 28020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.78294748220881e-05,
+      "loss": 0.3027,
+      "step": 28030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.782497072335825e-05,
+      "loss": 0.2993,
+      "step": 28040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7820466624628416e-05,
+      "loss": 0.3017,
+      "step": 28050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.781596252589857e-05,
+      "loss": 0.3064,
+      "step": 28060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.781145842716873e-05,
+      "loss": 0.2979,
+      "step": 28070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.780695432843888e-05,
+      "loss": 0.3044,
+      "step": 28080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.780245022970903e-05,
+      "loss": 0.2986,
+      "step": 28090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.77979461309792e-05,
+      "loss": 0.302,
+      "step": 28100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.779344203224935e-05,
+      "loss": 0.2977,
+      "step": 28110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7788937933519505e-05,
+      "loss": 0.3007,
+      "step": 28120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.778443383478966e-05,
+      "loss": 0.2982,
+      "step": 28130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.777992973605981e-05,
+      "loss": 0.3004,
+      "step": 28140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.777542563732997e-05,
+      "loss": 0.3065,
+      "step": 28150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.777092153860013e-05,
+      "loss": 0.2995,
+      "step": 28160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7766417439870286e-05,
+      "loss": 0.3072,
+      "step": 28170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7761913341140436e-05,
+      "loss": 0.2927,
+      "step": 28180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7757409242410594e-05,
+      "loss": 0.2973,
+      "step": 28190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.775290514368075e-05,
+      "loss": 0.309,
+      "step": 28200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.774840104495091e-05,
+      "loss": 0.3007,
+      "step": 28210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7743896946221066e-05,
+      "loss": 0.2991,
+      "step": 28220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.773939284749122e-05,
+      "loss": 0.3015,
+      "step": 28230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7734888748761375e-05,
+      "loss": 0.3016,
+      "step": 28240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.773038465003153e-05,
+      "loss": 0.2989,
+      "step": 28250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.772588055130168e-05,
+      "loss": 0.3005,
+      "step": 28260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.772137645257185e-05,
+      "loss": 0.3074,
+      "step": 28270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7716872353842e-05,
+      "loss": 0.3059,
+      "step": 28280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.771236825511215e-05,
+      "loss": 0.2996,
+      "step": 28290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.770786415638231e-05,
+      "loss": 0.2966,
+      "step": 28300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7703360057652464e-05,
+      "loss": 0.2949,
+      "step": 28310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.769885595892262e-05,
+      "loss": 0.3018,
+      "step": 28320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.769435186019278e-05,
+      "loss": 0.2984,
+      "step": 28330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.768984776146293e-05,
+      "loss": 0.2961,
+      "step": 28340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7685343662733094e-05,
+      "loss": 0.3022,
+      "step": 28350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7680839564003244e-05,
+      "loss": 0.2991,
+      "step": 28360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.76763354652734e-05,
+      "loss": 0.3025,
+      "step": 28370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.767183136654356e-05,
+      "loss": 0.2986,
+      "step": 28380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.766732726781371e-05,
+      "loss": 0.2995,
+      "step": 28390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.766282316908387e-05,
+      "loss": 0.302,
+      "step": 28400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7658319070354025e-05,
+      "loss": 0.3019,
+      "step": 28410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.765381497162418e-05,
+      "loss": 0.3003,
+      "step": 28420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.764931087289433e-05,
+      "loss": 0.3024,
+      "step": 28430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.764480677416449e-05,
+      "loss": 0.3004,
+      "step": 28440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.764030267543465e-05,
+      "loss": 0.298,
+      "step": 28450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.76357985767048e-05,
+      "loss": 0.3017,
+      "step": 28460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.763129447797496e-05,
+      "loss": 0.3083,
+      "step": 28470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7626790379245114e-05,
+      "loss": 0.2995,
+      "step": 28480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.762273669038826e-05,
+      "loss": 0.301,
+      "step": 28490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.761823259165841e-05,
+      "loss": 0.304,
+      "step": 28500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7613728492928566e-05,
+      "loss": 0.2993,
+      "step": 28510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.760922439419872e-05,
+      "loss": 0.3017,
+      "step": 28520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7604720295468874e-05,
+      "loss": 0.3005,
+      "step": 28530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.760021619673904e-05,
+      "loss": 0.3062,
+      "step": 28540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.759571209800919e-05,
+      "loss": 0.298,
+      "step": 28550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.7591207999279346e-05,
+      "loss": 0.2957,
+      "step": 28560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7586703900549504e-05,
+      "loss": 0.2995,
+      "step": 28570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7582199801819655e-05,
+      "loss": 0.2983,
+      "step": 28580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.757769570308982e-05,
+      "loss": 0.2955,
+      "step": 28590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.757319160435997e-05,
+      "loss": 0.3003,
+      "step": 28600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.756868750563012e-05,
+      "loss": 0.2939,
+      "step": 28610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7564183406900285e-05,
+      "loss": 0.2977,
+      "step": 28620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7559679308170435e-05,
+      "loss": 0.2964,
+      "step": 28630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.755517520944059e-05,
+      "loss": 0.2995,
+      "step": 28640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.755067111071075e-05,
+      "loss": 0.2977,
+      "step": 28650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.75461670119809e-05,
+      "loss": 0.2973,
+      "step": 28660
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.754166291325106e-05,
+      "loss": 0.2948,
+      "step": 28670
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7537158814521216e-05,
+      "loss": 0.2978,
+      "step": 28680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7532654715791374e-05,
+      "loss": 0.3037,
+      "step": 28690
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.752815061706153e-05,
+      "loss": 0.3042,
+      "step": 28700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.752364651833168e-05,
+      "loss": 0.3008,
+      "step": 28710
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.751914241960184e-05,
+      "loss": 0.2954,
+      "step": 28720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7514638320872e-05,
+      "loss": 0.3005,
+      "step": 28730
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7510134222142154e-05,
+      "loss": 0.2979,
+      "step": 28740
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7505630123412305e-05,
+      "loss": 0.2992,
+      "step": 28750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.750112602468246e-05,
+      "loss": 0.3048,
+      "step": 28760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.749662192595262e-05,
+      "loss": 0.3022,
+      "step": 28770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.749211782722277e-05,
+      "loss": 0.302,
+      "step": 28780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7487613728492935e-05,
+      "loss": 0.2938,
+      "step": 28790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7483109629763086e-05,
+      "loss": 0.3022,
+      "step": 28800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7478605531033236e-05,
+      "loss": 0.2986,
+      "step": 28810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.74741014323034e-05,
+      "loss": 0.3007,
+      "step": 28820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.746959733357355e-05,
+      "loss": 0.3019,
+      "step": 28830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7465093234843716e-05,
+      "loss": 0.2875,
+      "step": 28840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7460589136113866e-05,
+      "loss": 0.2984,
+      "step": 28850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.745608503738402e-05,
+      "loss": 0.2969,
+      "step": 28860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.745158093865418e-05,
+      "loss": 0.297,
+      "step": 28870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.744707683992433e-05,
+      "loss": 0.3046,
+      "step": 28880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.744257274119449e-05,
+      "loss": 0.301,
+      "step": 28890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.743806864246465e-05,
+      "loss": 0.2969,
+      "step": 28900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.74335645437348e-05,
+      "loss": 0.3009,
+      "step": 28910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7429060445004955e-05,
+      "loss": 0.301,
+      "step": 28920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.742455634627511e-05,
+      "loss": 0.3037,
+      "step": 28930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.742005224754527e-05,
+      "loss": 0.2984,
+      "step": 28940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.741554814881542e-05,
+      "loss": 0.3053,
+      "step": 28950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.741104405008558e-05,
+      "loss": 0.3004,
+      "step": 28960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7406539951355736e-05,
+      "loss": 0.3031,
+      "step": 28970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.740203585262589e-05,
+      "loss": 0.2959,
+      "step": 28980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.739753175389605e-05,
+      "loss": 0.2991,
+      "step": 28990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.73930276551662e-05,
+      "loss": 0.3042,
+      "step": 29000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.738852355643636e-05,
+      "loss": 0.3054,
+      "step": 29010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7384019457706517e-05,
+      "loss": 0.2965,
+      "step": 29020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.737951535897667e-05,
+      "loss": 0.2966,
+      "step": 29030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.737501126024683e-05,
+      "loss": 0.302,
+      "step": 29040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.737050716151698e-05,
+      "loss": 0.2984,
+      "step": 29050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.736600306278713e-05,
+      "loss": 0.2974,
+      "step": 29060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.73614989640573e-05,
+      "loss": 0.3003,
+      "step": 29070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.735699486532745e-05,
+      "loss": 0.2981,
+      "step": 29080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7352490766597605e-05,
+      "loss": 0.2944,
+      "step": 29090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.734798666786776e-05,
+      "loss": 0.3002,
+      "step": 29100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7343482569137914e-05,
+      "loss": 0.3017,
+      "step": 29110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.733897847040808e-05,
+      "loss": 0.3055,
+      "step": 29120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.733447437167823e-05,
+      "loss": 0.2986,
+      "step": 29130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7329970272948386e-05,
+      "loss": 0.299,
+      "step": 29140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7325466174218544e-05,
+      "loss": 0.2984,
+      "step": 29150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7320962075488694e-05,
+      "loss": 0.2983,
+      "step": 29160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.731645797675885e-05,
+      "loss": 0.2975,
+      "step": 29170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.731195387802901e-05,
+      "loss": 0.3002,
+      "step": 29180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.730744977929917e-05,
+      "loss": 0.3029,
+      "step": 29190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.730294568056932e-05,
+      "loss": 0.3,
+      "step": 29200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7298441581839475e-05,
+      "loss": 0.2993,
+      "step": 29210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.729393748310963e-05,
+      "loss": 0.2968,
+      "step": 29220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.728943338437978e-05,
+      "loss": 0.301,
+      "step": 29230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.728492928564995e-05,
+      "loss": 0.2964,
+      "step": 29240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.72804251869201e-05,
+      "loss": 0.2982,
+      "step": 29250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7275921088190256e-05,
+      "loss": 0.299,
+      "step": 29260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.727141698946041e-05,
+      "loss": 0.2979,
+      "step": 29270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7266912890730564e-05,
+      "loss": 0.2995,
+      "step": 29280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.726240879200072e-05,
+      "loss": 0.3018,
+      "step": 29290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.725790469327088e-05,
+      "loss": 0.2987,
+      "step": 29300
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.725340059454103e-05,
+      "loss": 0.3027,
+      "step": 29310
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7248896495811194e-05,
+      "loss": 0.2963,
+      "step": 29320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7244392397081345e-05,
+      "loss": 0.2952,
+      "step": 29330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.72398882983515e-05,
+      "loss": 0.301,
+      "step": 29340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.723538419962166e-05,
+      "loss": 0.2942,
+      "step": 29350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.723088010089181e-05,
+      "loss": 0.2967,
+      "step": 29360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.722637600216197e-05,
+      "loss": 0.3007,
+      "step": 29370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7221871903432125e-05,
+      "loss": 0.2967,
+      "step": 29380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.721736780470228e-05,
+      "loss": 0.3014,
+      "step": 29390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.721286370597244e-05,
+      "loss": 0.2965,
+      "step": 29400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.720835960724259e-05,
+      "loss": 0.2985,
+      "step": 29410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.720385550851275e-05,
+      "loss": 0.2991,
+      "step": 29420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7199351409782906e-05,
+      "loss": 0.3002,
+      "step": 29430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7194847311053063e-05,
+      "loss": 0.295,
+      "step": 29440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7190343212323214e-05,
+      "loss": 0.2948,
+      "step": 29450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.718583911359337e-05,
+      "loss": 0.2941,
+      "step": 29460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.718133501486353e-05,
+      "loss": 0.2955,
+      "step": 29470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.717683091613368e-05,
+      "loss": 0.298,
+      "step": 29480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7172326817403844e-05,
+      "loss": 0.3002,
+      "step": 29490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7167822718673995e-05,
+      "loss": 0.301,
+      "step": 29500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7163318619944146e-05,
+      "loss": 0.2965,
+      "step": 29510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.715881452121431e-05,
+      "loss": 0.2918,
+      "step": 29520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.715431042248446e-05,
+      "loss": 0.294,
+      "step": 29530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.714980632375462e-05,
+      "loss": 0.2971,
+      "step": 29540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7145302225024776e-05,
+      "loss": 0.3026,
+      "step": 29550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7140798126294926e-05,
+      "loss": 0.2967,
+      "step": 29560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.713629402756509e-05,
+      "loss": 0.2961,
+      "step": 29570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.713178992883524e-05,
+      "loss": 0.2931,
+      "step": 29580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.71272858301054e-05,
+      "loss": 0.2893,
+      "step": 29590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7122781731375556e-05,
+      "loss": 0.2935,
+      "step": 29600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.711827763264571e-05,
+      "loss": 0.2948,
+      "step": 29610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7113773533915864e-05,
+      "loss": 0.2977,
+      "step": 29620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.710926943518602e-05,
+      "loss": 0.2969,
+      "step": 29630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.710476533645618e-05,
+      "loss": 0.2935,
+      "step": 29640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.710026123772633e-05,
+      "loss": 0.2971,
+      "step": 29650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.709575713899649e-05,
+      "loss": 0.2965,
+      "step": 29660
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.7091253040266645e-05,
+      "loss": 0.2974,
+      "step": 29670
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.70867489415368e-05,
+      "loss": 0.2955,
+      "step": 29680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.708224484280696e-05,
+      "loss": 0.2936,
+      "step": 29690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.707774074407711e-05,
+      "loss": 0.2889,
+      "step": 29700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.707323664534727e-05,
+      "loss": 0.2993,
+      "step": 29710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.7068732546617426e-05,
+      "loss": 0.2978,
+      "step": 29720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.7064228447887577e-05,
+      "loss": 0.2974,
+      "step": 29730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.7059724349157734e-05,
+      "loss": 0.2934,
+      "step": 29740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.705522025042789e-05,
+      "loss": 0.298,
+      "step": 29750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.705071615169804e-05,
+      "loss": 0.2998,
+      "step": 29760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.7046212052968207e-05,
+      "loss": 0.2922,
+      "step": 29770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.704170795423836e-05,
+      "loss": 0.2964,
+      "step": 29780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.7037203855508515e-05,
+      "loss": 0.2987,
+      "step": 29790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.703269975677867e-05,
+      "loss": 0.2915,
+      "step": 29800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.702819565804882e-05,
+      "loss": 0.2934,
+      "step": 29810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.702369155931899e-05,
+      "loss": 0.2939,
+      "step": 29820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.701918746058914e-05,
+      "loss": 0.2989,
+      "step": 29830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.7014683361859295e-05,
+      "loss": 0.2954,
+      "step": 29840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.701017926312945e-05,
+      "loss": 0.3008,
+      "step": 29850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.7005675164399604e-05,
+      "loss": 0.2944,
+      "step": 29860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.700117106566976e-05,
+      "loss": 0.2966,
+      "step": 29870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.699666696693992e-05,
+      "loss": 0.3034,
+      "step": 29880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6992162868210076e-05,
+      "loss": 0.2922,
+      "step": 29890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.698765876948023e-05,
+      "loss": 0.2986,
+      "step": 29900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6983154670750384e-05,
+      "loss": 0.2954,
+      "step": 29910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.697865057202054e-05,
+      "loss": 0.3012,
+      "step": 29920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.697414647329069e-05,
+      "loss": 0.298,
+      "step": 29930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.696964237456085e-05,
+      "loss": 0.2923,
+      "step": 29940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.696513827583101e-05,
+      "loss": 0.2914,
+      "step": 29950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6960634177101165e-05,
+      "loss": 0.2955,
+      "step": 29960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.695613007837132e-05,
+      "loss": 0.293,
+      "step": 29970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.695162597964147e-05,
+      "loss": 0.296,
+      "step": 29980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.694712188091163e-05,
+      "loss": 0.2911,
+      "step": 29990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.694261778218179e-05,
+      "loss": 0.293,
+      "step": 30000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.693811368345194e-05,
+      "loss": 0.2924,
+      "step": 30010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.69336095847221e-05,
+      "loss": 0.2981,
+      "step": 30020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6929105485992254e-05,
+      "loss": 0.2941,
+      "step": 30030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.692460138726241e-05,
+      "loss": 0.2966,
+      "step": 30040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.692009728853257e-05,
+      "loss": 0.2936,
+      "step": 30050
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.691559318980272e-05,
+      "loss": 0.2976,
+      "step": 30060
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.691108909107288e-05,
+      "loss": 0.2909,
+      "step": 30070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6906584992343035e-05,
+      "loss": 0.2923,
+      "step": 30080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.690208089361319e-05,
+      "loss": 0.3011,
+      "step": 30090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.689757679488334e-05,
+      "loss": 0.2997,
+      "step": 30100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.68930726961535e-05,
+      "loss": 0.294,
+      "step": 30110
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.688856859742366e-05,
+      "loss": 0.2996,
+      "step": 30120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6884064498693815e-05,
+      "loss": 0.2955,
+      "step": 30130
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.687956039996397e-05,
+      "loss": 0.2948,
+      "step": 30140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6875056301234123e-05,
+      "loss": 0.2935,
+      "step": 30150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.687055220250428e-05,
+      "loss": 0.2919,
+      "step": 30160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.686604810377444e-05,
+      "loss": 0.2947,
+      "step": 30170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.686154400504459e-05,
+      "loss": 0.2941,
+      "step": 30180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.685703990631475e-05,
+      "loss": 0.2917,
+      "step": 30190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6852535807584904e-05,
+      "loss": 0.3018,
+      "step": 30200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6848031708855055e-05,
+      "loss": 0.2991,
+      "step": 30210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.684352761012522e-05,
+      "loss": 0.3017,
+      "step": 30220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.683902351139537e-05,
+      "loss": 0.2922,
+      "step": 30230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.683451941266553e-05,
+      "loss": 0.2959,
+      "step": 30240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6830015313935685e-05,
+      "loss": 0.299,
+      "step": 30250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6825511215205836e-05,
+      "loss": 0.2922,
+      "step": 30260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6821007116476e-05,
+      "loss": 0.2964,
+      "step": 30270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.681650301774615e-05,
+      "loss": 0.2933,
+      "step": 30280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.681199891901631e-05,
+      "loss": 0.2932,
+      "step": 30290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6807494820286465e-05,
+      "loss": 0.2955,
+      "step": 30300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6802990721556616e-05,
+      "loss": 0.2892,
+      "step": 30310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6798486622826774e-05,
+      "loss": 0.2935,
+      "step": 30320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.679398252409693e-05,
+      "loss": 0.293,
+      "step": 30330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.678947842536709e-05,
+      "loss": 0.2949,
+      "step": 30340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.678497432663724e-05,
+      "loss": 0.2955,
+      "step": 30350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.67804702279074e-05,
+      "loss": 0.2949,
+      "step": 30360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6775966129177554e-05,
+      "loss": 0.2944,
+      "step": 30370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6771462030447705e-05,
+      "loss": 0.2923,
+      "step": 30380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.676695793171786e-05,
+      "loss": 0.2942,
+      "step": 30390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.676245383298802e-05,
+      "loss": 0.2915,
+      "step": 30400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.675794973425818e-05,
+      "loss": 0.2969,
+      "step": 30410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6753445635528335e-05,
+      "loss": 0.294,
+      "step": 30420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6748941536798486e-05,
+      "loss": 0.2883,
+      "step": 30430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.674443743806864e-05,
+      "loss": 0.3013,
+      "step": 30440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.67399333393388e-05,
+      "loss": 0.2976,
+      "step": 30450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.673542924060895e-05,
+      "loss": 0.2963,
+      "step": 30460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6730925141879116e-05,
+      "loss": 0.2938,
+      "step": 30470
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6726421043149266e-05,
+      "loss": 0.2953,
+      "step": 30480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6721916944419424e-05,
+      "loss": 0.2895,
+      "step": 30490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.671741284568958e-05,
+      "loss": 0.2942,
+      "step": 30500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.671290874695973e-05,
+      "loss": 0.2949,
+      "step": 30510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.670840464822989e-05,
+      "loss": 0.2915,
+      "step": 30520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.670390054950005e-05,
+      "loss": 0.2946,
+      "step": 30530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6699396450770205e-05,
+      "loss": 0.2855,
+      "step": 30540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.669489235204036e-05,
+      "loss": 0.2949,
+      "step": 30550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.669038825331051e-05,
+      "loss": 0.2922,
+      "step": 30560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.668588415458067e-05,
+      "loss": 0.2893,
+      "step": 30570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.668138005585083e-05,
+      "loss": 0.291,
+      "step": 30580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.667687595712098e-05,
+      "loss": 0.2941,
+      "step": 30590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6672371858391136e-05,
+      "loss": 0.2945,
+      "step": 30600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6667867759661294e-05,
+      "loss": 0.2936,
+      "step": 30610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.666381407080444e-05,
+      "loss": 0.2943,
+      "step": 30620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.665930997207459e-05,
+      "loss": 0.2903,
+      "step": 30630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6654805873344746e-05,
+      "loss": 0.2898,
+      "step": 30640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.66503017746149e-05,
+      "loss": 0.2909,
+      "step": 30650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.664579767588506e-05,
+      "loss": 0.2886,
+      "step": 30660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.664129357715521e-05,
+      "loss": 0.2982,
+      "step": 30670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.663678947842537e-05,
+      "loss": 0.2881,
+      "step": 30680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6632285379695526e-05,
+      "loss": 0.2962,
+      "step": 30690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.662778128096568e-05,
+      "loss": 0.3,
+      "step": 30700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.662327718223584e-05,
+      "loss": 0.297,
+      "step": 30710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.661877308350599e-05,
+      "loss": 0.2901,
+      "step": 30720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.661426898477614e-05,
+      "loss": 0.2934,
+      "step": 30730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.660976488604631e-05,
+      "loss": 0.2908,
+      "step": 30740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.660526078731646e-05,
+      "loss": 0.2942,
+      "step": 30750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.6600756688586615e-05,
+      "loss": 0.2953,
+      "step": 30760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.659625258985677e-05,
+      "loss": 0.2934,
+      "step": 30770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.659174849112692e-05,
+      "loss": 0.2983,
+      "step": 30780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.658724439239709e-05,
+      "loss": 0.2921,
+      "step": 30790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.658274029366724e-05,
+      "loss": 0.2957,
+      "step": 30800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6578236194937396e-05,
+      "loss": 0.2949,
+      "step": 30810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.657373209620755e-05,
+      "loss": 0.2985,
+      "step": 30820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6569227997477704e-05,
+      "loss": 0.2879,
+      "step": 30830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.656472389874786e-05,
+      "loss": 0.2935,
+      "step": 30840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.656021980001802e-05,
+      "loss": 0.2872,
+      "step": 30850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6555715701288176e-05,
+      "loss": 0.2912,
+      "step": 30860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.655121160255833e-05,
+      "loss": 0.2935,
+      "step": 30870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6546707503828485e-05,
+      "loss": 0.2927,
+      "step": 30880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.654220340509864e-05,
+      "loss": 0.2874,
+      "step": 30890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.65376993063688e-05,
+      "loss": 0.2957,
+      "step": 30900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.653319520763896e-05,
+      "loss": 0.293,
+      "step": 30910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.652869110890911e-05,
+      "loss": 0.291,
+      "step": 30920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6524187010179265e-05,
+      "loss": 0.2901,
+      "step": 30930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.651968291144942e-05,
+      "loss": 0.2944,
+      "step": 30940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6515178812719574e-05,
+      "loss": 0.2892,
+      "step": 30950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.651067471398973e-05,
+      "loss": 0.2932,
+      "step": 30960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.650617061525989e-05,
+      "loss": 0.2929,
+      "step": 30970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.650166651653004e-05,
+      "loss": 0.295,
+      "step": 30980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6497162417800204e-05,
+      "loss": 0.2918,
+      "step": 30990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6492658319070354e-05,
+      "loss": 0.2891,
+      "step": 31000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.648815422034051e-05,
+      "loss": 0.3002,
+      "step": 31010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.648365012161067e-05,
+      "loss": 0.2921,
+      "step": 31020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.647914602288082e-05,
+      "loss": 0.294,
+      "step": 31030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6474641924150984e-05,
+      "loss": 0.2926,
+      "step": 31040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6470137825421135e-05,
+      "loss": 0.2951,
+      "step": 31050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.646563372669129e-05,
+      "loss": 0.2945,
+      "step": 31060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.646112962796145e-05,
+      "loss": 0.29,
+      "step": 31070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.64566255292316e-05,
+      "loss": 0.2904,
+      "step": 31080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.645212143050176e-05,
+      "loss": 0.2939,
+      "step": 31090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6447617331771916e-05,
+      "loss": 0.2898,
+      "step": 31100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.644311323304207e-05,
+      "loss": 0.2861,
+      "step": 31110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6438609134312224e-05,
+      "loss": 0.2936,
+      "step": 31120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.643410503558238e-05,
+      "loss": 0.2976,
+      "step": 31130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.642960093685254e-05,
+      "loss": 0.2924,
+      "step": 31140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.642509683812269e-05,
+      "loss": 0.2887,
+      "step": 31150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.642059273939285e-05,
+      "loss": 0.2937,
+      "step": 31160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6416088640663005e-05,
+      "loss": 0.2904,
+      "step": 31170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.641158454193316e-05,
+      "loss": 0.2886,
+      "step": 31180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.640708044320332e-05,
+      "loss": 0.2891,
+      "step": 31190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.640257634447347e-05,
+      "loss": 0.2857,
+      "step": 31200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.639807224574363e-05,
+      "loss": 0.2909,
+      "step": 31210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6393568147013785e-05,
+      "loss": 0.2946,
+      "step": 31220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6389064048283936e-05,
+      "loss": 0.2931,
+      "step": 31230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.63845599495541e-05,
+      "loss": 0.292,
+      "step": 31240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.638005585082425e-05,
+      "loss": 0.2962,
+      "step": 31250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.637555175209441e-05,
+      "loss": 0.2841,
+      "step": 31260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6371047653364566e-05,
+      "loss": 0.288,
+      "step": 31270
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6366543554634717e-05,
+      "loss": 0.2933,
+      "step": 31280
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6362039455904874e-05,
+      "loss": 0.2895,
+      "step": 31290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.635753535717503e-05,
+      "loss": 0.2961,
+      "step": 31300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.635303125844519e-05,
+      "loss": 0.2878,
+      "step": 31310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6348527159715347e-05,
+      "loss": 0.2873,
+      "step": 31320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.63440230609855e-05,
+      "loss": 0.2844,
+      "step": 31330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6339518962255655e-05,
+      "loss": 0.2921,
+      "step": 31340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.633501486352581e-05,
+      "loss": 0.2878,
+      "step": 31350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.633051076479597e-05,
+      "loss": 0.2854,
+      "step": 31360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.632600666606612e-05,
+      "loss": 0.2901,
+      "step": 31370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.632150256733628e-05,
+      "loss": 0.2895,
+      "step": 31380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6316998468606435e-05,
+      "loss": 0.2957,
+      "step": 31390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6312494369876586e-05,
+      "loss": 0.2887,
+      "step": 31400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6307990271146744e-05,
+      "loss": 0.2966,
+      "step": 31410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.63034861724169e-05,
+      "loss": 0.2929,
+      "step": 31420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.629898207368705e-05,
+      "loss": 0.2962,
+      "step": 31430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6294477974957216e-05,
+      "loss": 0.2917,
+      "step": 31440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.628997387622737e-05,
+      "loss": 0.297,
+      "step": 31450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6285469777497524e-05,
+      "loss": 0.294,
+      "step": 31460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.628096567876768e-05,
+      "loss": 0.2918,
+      "step": 31470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.627646158003783e-05,
+      "loss": 0.2915,
+      "step": 31480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6271957481308e-05,
+      "loss": 0.2916,
+      "step": 31490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.626745338257815e-05,
+      "loss": 0.2889,
+      "step": 31500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6262949283848305e-05,
+      "loss": 0.2878,
+      "step": 31510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.625844518511846e-05,
+      "loss": 0.2914,
+      "step": 31520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.625394108638861e-05,
+      "loss": 0.2859,
+      "step": 31530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.624943698765877e-05,
+      "loss": 0.2899,
+      "step": 31540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.624493288892893e-05,
+      "loss": 0.2966,
+      "step": 31550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6240428790199086e-05,
+      "loss": 0.2929,
+      "step": 31560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6235924691469236e-05,
+      "loss": 0.2945,
+      "step": 31570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6231420592739394e-05,
+      "loss": 0.2948,
+      "step": 31580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.622691649400955e-05,
+      "loss": 0.2896,
+      "step": 31590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.622241239527971e-05,
+      "loss": 0.2864,
+      "step": 31600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.621790829654986e-05,
+      "loss": 0.2946,
+      "step": 31610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.621340419782002e-05,
+      "loss": 0.2874,
+      "step": 31620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6208900099090175e-05,
+      "loss": 0.2925,
+      "step": 31630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.620439600036033e-05,
+      "loss": 0.2897,
+      "step": 31640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.619989190163048e-05,
+      "loss": 0.2892,
+      "step": 31650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.619538780290064e-05,
+      "loss": 0.2853,
+      "step": 31660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.61908837041708e-05,
+      "loss": 0.2839,
+      "step": 31670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.618637960544095e-05,
+      "loss": 0.2929,
+      "step": 31680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.618187550671111e-05,
+      "loss": 0.2882,
+      "step": 31690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6177371407981264e-05,
+      "loss": 0.2884,
+      "step": 31700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.617286730925142e-05,
+      "loss": 0.2843,
+      "step": 31710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.616836321052158e-05,
+      "loss": 0.2918,
+      "step": 31720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.616385911179173e-05,
+      "loss": 0.2898,
+      "step": 31730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6159355013061893e-05,
+      "loss": 0.2911,
+      "step": 31740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6154850914332044e-05,
+      "loss": 0.2954,
+      "step": 31750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.61503468156022e-05,
+      "loss": 0.2897,
+      "step": 31760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.614584271687236e-05,
+      "loss": 0.2873,
+      "step": 31770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.614133861814251e-05,
+      "loss": 0.2896,
+      "step": 31780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.613683451941267e-05,
+      "loss": 0.295,
+      "step": 31790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6132330420682825e-05,
+      "loss": 0.2931,
+      "step": 31800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6127826321952976e-05,
+      "loss": 0.287,
+      "step": 31810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.612332222322313e-05,
+      "loss": 0.2945,
+      "step": 31820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.611881812449329e-05,
+      "loss": 0.2906,
+      "step": 31830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.611431402576345e-05,
+      "loss": 0.2914,
+      "step": 31840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.61098099270336e-05,
+      "loss": 0.292,
+      "step": 31850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6105305828303756e-05,
+      "loss": 0.2822,
+      "step": 31860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6100801729573914e-05,
+      "loss": 0.2907,
+      "step": 31870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.609629763084407e-05,
+      "loss": 0.2936,
+      "step": 31880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.609179353211423e-05,
+      "loss": 0.2847,
+      "step": 31890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.608728943338438e-05,
+      "loss": 0.294,
+      "step": 31900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.608278533465454e-05,
+      "loss": 0.2895,
+      "step": 31910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.6078281235924694e-05,
+      "loss": 0.2894,
+      "step": 31920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.6073777137194845e-05,
+      "loss": 0.292,
+      "step": 31930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.606927303846501e-05,
+      "loss": 0.2872,
+      "step": 31940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.606476893973516e-05,
+      "loss": 0.2915,
+      "step": 31950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.606026484100532e-05,
+      "loss": 0.2866,
+      "step": 31960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.6055760742275475e-05,
+      "loss": 0.2839,
+      "step": 31970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.6051256643545626e-05,
+      "loss": 0.2876,
+      "step": 31980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.604675254481578e-05,
+      "loss": 0.3028,
+      "step": 31990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.604224844608594e-05,
+      "loss": 0.2826,
+      "step": 32000
+    },
+    {
+      "epoch": 0.29,
+      "eval_NEIMS_canon_smiles": 0.9411666666666667,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.012166666666666666,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.38288233226711305,
+      "eval_NEIMS_exact_mols": 0.012083333333333333,
+      "eval_NEIMS_exact_smiles": 0.011716666666666667,
+      "eval_NEIMS_loss": 0.3008743226528168,
+      "eval_NEIMS_matched_formulas": 0.40558333333333335,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.2822131343249727,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.0122,
+      "eval_NEIMS_runtime": 704.6298,
+      "eval_NEIMS_samples_per_second": 85.151,
+      "eval_NEIMS_steps_per_second": 1.331,
+      "step": 32000
+    },
+    {
+      "epoch": 0.29,
+      "eval_RASSP_canon_smiles": 0.9329315123384646,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.030520586369252346,
+      "eval_RASSP_daylight_tanimoto_simil": 0.4583983323036867,
+      "eval_RASSP_exact_mols": 0.030286840084148663,
+      "eval_RASSP_exact_smiles": 0.029318462617290545,
+      "eval_RASSP_loss": 0.2483576387166977,
+      "eval_RASSP_matched_formulas": 0.5404881958126022,
+      "eval_RASSP_morgan_tanimoto_simil": 0.3222687479375559,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.03033692857381374,
+      "eval_RASSP_runtime": 812.2405,
+      "eval_RASSP_samples_per_second": 73.739,
+      "eval_RASSP_steps_per_second": 1.152,
+      "step": 32000
+    },
+    {
+      "epoch": 0.29,
+      "eval_NIST_canon_smiles": 0.9088618376690208,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.0056783901763849945,
+      "eval_NIST_daylight_tanimoto_simil": 0.24459893187079448,
+      "eval_NIST_exact_mols": 0.005217020974553714,
+      "eval_NIST_exact_smiles": 0.0050040813429392764,
+      "eval_NIST_loss": 1.4709786176681519,
+      "eval_NIST_matched_formulas": 0.05358980728963339,
+      "eval_NIST_morgan_tanimoto_simil": 0.20517589302071992,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.005642900237782589,
+      "eval_NIST_runtime": 444.8617,
+      "eval_NIST_samples_per_second": 63.339,
+      "eval_NIST_steps_per_second": 0.991,
+      "step": 32000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.603774434735609e-05,
+      "loss": 0.2906,
+      "step": 32010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.6033240248626256e-05,
+      "loss": 0.2851,
+      "step": 32020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.6028736149896407e-05,
+      "loss": 0.2971,
+      "step": 32030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.6024232051166564e-05,
+      "loss": 0.2919,
+      "step": 32040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.601972795243672e-05,
+      "loss": 0.2938,
+      "step": 32050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.601522385370687e-05,
+      "loss": 0.2904,
+      "step": 32060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.601071975497703e-05,
+      "loss": 0.2867,
+      "step": 32070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.600621565624719e-05,
+      "loss": 0.2932,
+      "step": 32080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.6001711557517345e-05,
+      "loss": 0.2901,
+      "step": 32090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5997207458787495e-05,
+      "loss": 0.2906,
+      "step": 32100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.599270336005765e-05,
+      "loss": 0.2919,
+      "step": 32110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.598819926132781e-05,
+      "loss": 0.2897,
+      "step": 32120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.598369516259796e-05,
+      "loss": 0.2912,
+      "step": 32130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5979191063868125e-05,
+      "loss": 0.286,
+      "step": 32140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5974686965138276e-05,
+      "loss": 0.2879,
+      "step": 32150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5970182866408434e-05,
+      "loss": 0.2889,
+      "step": 32160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.596567876767859e-05,
+      "loss": 0.2891,
+      "step": 32170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.596117466894874e-05,
+      "loss": 0.2906,
+      "step": 32180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5956670570218906e-05,
+      "loss": 0.2903,
+      "step": 32190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.595216647148906e-05,
+      "loss": 0.291,
+      "step": 32200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5947662372759214e-05,
+      "loss": 0.2891,
+      "step": 32210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.594315827402937e-05,
+      "loss": 0.2858,
+      "step": 32220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.593865417529952e-05,
+      "loss": 0.2933,
+      "step": 32230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.593415007656968e-05,
+      "loss": 0.2993,
+      "step": 32240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.592964597783984e-05,
+      "loss": 0.2897,
+      "step": 32250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.592514187910999e-05,
+      "loss": 0.2859,
+      "step": 32260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5920637780380146e-05,
+      "loss": 0.2867,
+      "step": 32270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.59161336816503e-05,
+      "loss": 0.286,
+      "step": 32280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.591162958292046e-05,
+      "loss": 0.2904,
+      "step": 32290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.590712548419062e-05,
+      "loss": 0.2969,
+      "step": 32300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.590262138546077e-05,
+      "loss": 0.2862,
+      "step": 32310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5898117286730926e-05,
+      "loss": 0.2867,
+      "step": 32320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5893613188001084e-05,
+      "loss": 0.287,
+      "step": 32330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.588910908927124e-05,
+      "loss": 0.2889,
+      "step": 32340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.588460499054139e-05,
+      "loss": 0.2811,
+      "step": 32350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.588010089181155e-05,
+      "loss": 0.2884,
+      "step": 32360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.587559679308171e-05,
+      "loss": 0.2812,
+      "step": 32370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.587109269435186e-05,
+      "loss": 0.2902,
+      "step": 32380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.586658859562202e-05,
+      "loss": 0.2843,
+      "step": 32390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.586208449689217e-05,
+      "loss": 0.2866,
+      "step": 32400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.585758039816233e-05,
+      "loss": 0.2995,
+      "step": 32410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.585307629943249e-05,
+      "loss": 0.2898,
+      "step": 32420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.584857220070264e-05,
+      "loss": 0.2919,
+      "step": 32430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5844068101972796e-05,
+      "loss": 0.28,
+      "step": 32440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5839564003242953e-05,
+      "loss": 0.2842,
+      "step": 32450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5835059904513104e-05,
+      "loss": 0.2846,
+      "step": 32460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.583055580578327e-05,
+      "loss": 0.2839,
+      "step": 32470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.582605170705342e-05,
+      "loss": 0.2851,
+      "step": 32480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.582154760832358e-05,
+      "loss": 0.2871,
+      "step": 32490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5817043509593734e-05,
+      "loss": 0.2894,
+      "step": 32500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5812539410863885e-05,
+      "loss": 0.29,
+      "step": 32510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.580803531213404e-05,
+      "loss": 0.2848,
+      "step": 32520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.58035312134042e-05,
+      "loss": 0.2892,
+      "step": 32530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.579902711467436e-05,
+      "loss": 0.2847,
+      "step": 32540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.579452301594451e-05,
+      "loss": 0.2855,
+      "step": 32550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5790018917214666e-05,
+      "loss": 0.2871,
+      "step": 32560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.578551481848482e-05,
+      "loss": 0.2843,
+      "step": 32570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5781010719754974e-05,
+      "loss": 0.2928,
+      "step": 32580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.577650662102514e-05,
+      "loss": 0.2858,
+      "step": 32590
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.577200252229529e-05,
+      "loss": 0.2815,
+      "step": 32600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5767498423565446e-05,
+      "loss": 0.2849,
+      "step": 32610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5762994324835604e-05,
+      "loss": 0.2895,
+      "step": 32620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5758490226105754e-05,
+      "loss": 0.297,
+      "step": 32630
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.575398612737592e-05,
+      "loss": 0.2834,
+      "step": 32640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.574948202864607e-05,
+      "loss": 0.2869,
+      "step": 32650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.574497792991622e-05,
+      "loss": 0.284,
+      "step": 32660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5740473831186384e-05,
+      "loss": 0.2849,
+      "step": 32670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5735969732456535e-05,
+      "loss": 0.2882,
+      "step": 32680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.573146563372669e-05,
+      "loss": 0.2853,
+      "step": 32690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.572696153499685e-05,
+      "loss": 0.2881,
+      "step": 32700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5722457436267e-05,
+      "loss": 0.2873,
+      "step": 32710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.571795333753716e-05,
+      "loss": 0.2824,
+      "step": 32720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5713449238807316e-05,
+      "loss": 0.2881,
+      "step": 32730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.570894514007747e-05,
+      "loss": 0.2848,
+      "step": 32740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.570444104134763e-05,
+      "loss": 0.2891,
+      "step": 32750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.569993694261778e-05,
+      "loss": 0.2955,
+      "step": 32760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.569543284388794e-05,
+      "loss": 0.2858,
+      "step": 32770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.569137915503108e-05,
+      "loss": 0.2878,
+      "step": 32780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5686875056301233e-05,
+      "loss": 0.2857,
+      "step": 32790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.568237095757139e-05,
+      "loss": 0.2917,
+      "step": 32800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.567786685884155e-05,
+      "loss": 0.2838,
+      "step": 32810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5673362760111706e-05,
+      "loss": 0.286,
+      "step": 32820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.566885866138186e-05,
+      "loss": 0.2888,
+      "step": 32830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5664354562652014e-05,
+      "loss": 0.2912,
+      "step": 32840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.565985046392217e-05,
+      "loss": 0.2881,
+      "step": 32850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.565534636519233e-05,
+      "loss": 0.2865,
+      "step": 32860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.565084226646248e-05,
+      "loss": 0.2845,
+      "step": 32870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.564633816773264e-05,
+      "loss": 0.2764,
+      "step": 32880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5641834069002795e-05,
+      "loss": 0.2899,
+      "step": 32890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5637329970272946e-05,
+      "loss": 0.285,
+      "step": 32900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.563282587154311e-05,
+      "loss": 0.2981,
+      "step": 32910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.562832177281326e-05,
+      "loss": 0.2895,
+      "step": 32920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.562381767408342e-05,
+      "loss": 0.2915,
+      "step": 32930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5619313575353576e-05,
+      "loss": 0.2888,
+      "step": 32940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5614809476623726e-05,
+      "loss": 0.2845,
+      "step": 32950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.561030537789389e-05,
+      "loss": 0.2864,
+      "step": 32960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.560580127916404e-05,
+      "loss": 0.2822,
+      "step": 32970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.56012971804342e-05,
+      "loss": 0.2878,
+      "step": 32980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5596793081704356e-05,
+      "loss": 0.2885,
+      "step": 32990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.559228898297451e-05,
+      "loss": 0.2878,
+      "step": 33000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.5587784884244664e-05,
+      "loss": 0.2888,
+      "step": 33010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.558328078551482e-05,
+      "loss": 0.2936,
+      "step": 33020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.557877668678497e-05,
+      "loss": 0.2899,
+      "step": 33030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.557427258805513e-05,
+      "loss": 0.2813,
+      "step": 33040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.556976848932529e-05,
+      "loss": 0.2856,
+      "step": 33050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5565264390595445e-05,
+      "loss": 0.2833,
+      "step": 33060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5560760291865596e-05,
+      "loss": 0.2853,
+      "step": 33070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.555625619313575e-05,
+      "loss": 0.2811,
+      "step": 33080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.555175209440591e-05,
+      "loss": 0.2805,
+      "step": 33090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.554724799567607e-05,
+      "loss": 0.2813,
+      "step": 33100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5542743896946226e-05,
+      "loss": 0.2842,
+      "step": 33110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5538239798216377e-05,
+      "loss": 0.2865,
+      "step": 33120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5533735699486534e-05,
+      "loss": 0.2855,
+      "step": 33130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.552923160075669e-05,
+      "loss": 0.2828,
+      "step": 33140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.552472750202684e-05,
+      "loss": 0.2878,
+      "step": 33150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5520223403297006e-05,
+      "loss": 0.2787,
+      "step": 33160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.551571930456716e-05,
+      "loss": 0.2881,
+      "step": 33170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5511215205837315e-05,
+      "loss": 0.2871,
+      "step": 33180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.550671110710747e-05,
+      "loss": 0.2844,
+      "step": 33190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.550220700837762e-05,
+      "loss": 0.2877,
+      "step": 33200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.549770290964778e-05,
+      "loss": 0.285,
+      "step": 33210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.549319881091794e-05,
+      "loss": 0.2815,
+      "step": 33220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.548869471218809e-05,
+      "loss": 0.2851,
+      "step": 33230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.548419061345825e-05,
+      "loss": 0.2889,
+      "step": 33240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5479686514728404e-05,
+      "loss": 0.2844,
+      "step": 33250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.547518241599856e-05,
+      "loss": 0.2892,
+      "step": 33260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.547067831726872e-05,
+      "loss": 0.2921,
+      "step": 33270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.546617421853887e-05,
+      "loss": 0.282,
+      "step": 33280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.546167011980903e-05,
+      "loss": 0.2814,
+      "step": 33290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5457166021079184e-05,
+      "loss": 0.2869,
+      "step": 33300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.545266192234934e-05,
+      "loss": 0.2854,
+      "step": 33310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.544815782361949e-05,
+      "loss": 0.2838,
+      "step": 33320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.544365372488965e-05,
+      "loss": 0.2858,
+      "step": 33330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.543914962615981e-05,
+      "loss": 0.2838,
+      "step": 33340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.543464552742996e-05,
+      "loss": 0.2882,
+      "step": 33350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.543014142870012e-05,
+      "loss": 0.288,
+      "step": 33360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.542563732997027e-05,
+      "loss": 0.2827,
+      "step": 33370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.542113323124043e-05,
+      "loss": 0.286,
+      "step": 33380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.541662913251059e-05,
+      "loss": 0.29,
+      "step": 33390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.541212503378074e-05,
+      "loss": 0.2862,
+      "step": 33400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.54076209350509e-05,
+      "loss": 0.2893,
+      "step": 33410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5403116836321054e-05,
+      "loss": 0.2861,
+      "step": 33420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.539861273759121e-05,
+      "loss": 0.2923,
+      "step": 33430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.539410863886137e-05,
+      "loss": 0.285,
+      "step": 33440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.538960454013152e-05,
+      "loss": 0.2808,
+      "step": 33450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.538510044140168e-05,
+      "loss": 0.28,
+      "step": 33460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5380596342671835e-05,
+      "loss": 0.2806,
+      "step": 33470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5376092243941985e-05,
+      "loss": 0.284,
+      "step": 33480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.537158814521214e-05,
+      "loss": 0.2913,
+      "step": 33490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.53670840464823e-05,
+      "loss": 0.2922,
+      "step": 33500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.536257994775246e-05,
+      "loss": 0.2915,
+      "step": 33510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5358075849022615e-05,
+      "loss": 0.2854,
+      "step": 33520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5353571750292766e-05,
+      "loss": 0.2891,
+      "step": 33530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5349067651562923e-05,
+      "loss": 0.2834,
+      "step": 33540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.534456355283308e-05,
+      "loss": 0.2729,
+      "step": 33550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.534005945410324e-05,
+      "loss": 0.2831,
+      "step": 33560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.533555535537339e-05,
+      "loss": 0.2817,
+      "step": 33570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.533105125664355e-05,
+      "loss": 0.2866,
+      "step": 33580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5326547157913704e-05,
+      "loss": 0.2811,
+      "step": 33590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5322043059183855e-05,
+      "loss": 0.2828,
+      "step": 33600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.531753896045402e-05,
+      "loss": 0.2829,
+      "step": 33610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.531303486172417e-05,
+      "loss": 0.2918,
+      "step": 33620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.530853076299433e-05,
+      "loss": 0.2813,
+      "step": 33630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5304026664264485e-05,
+      "loss": 0.2855,
+      "step": 33640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5299522565534636e-05,
+      "loss": 0.2851,
+      "step": 33650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.52950184668048e-05,
+      "loss": 0.2825,
+      "step": 33660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.529051436807495e-05,
+      "loss": 0.2864,
+      "step": 33670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.52860102693451e-05,
+      "loss": 0.2778,
+      "step": 33680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5281506170615265e-05,
+      "loss": 0.2894,
+      "step": 33690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5277002071885416e-05,
+      "loss": 0.2876,
+      "step": 33700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5272497973155574e-05,
+      "loss": 0.2903,
+      "step": 33710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.526799387442573e-05,
+      "loss": 0.2809,
+      "step": 33720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.526348977569588e-05,
+      "loss": 0.288,
+      "step": 33730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.525898567696604e-05,
+      "loss": 0.2861,
+      "step": 33740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.52544815782362e-05,
+      "loss": 0.2756,
+      "step": 33750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5249977479506354e-05,
+      "loss": 0.2884,
+      "step": 33760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5245473380776505e-05,
+      "loss": 0.2808,
+      "step": 33770
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.524096928204666e-05,
+      "loss": 0.2801,
+      "step": 33780
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.523646518331682e-05,
+      "loss": 0.2804,
+      "step": 33790
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.523196108458698e-05,
+      "loss": 0.2873,
+      "step": 33800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5227456985857135e-05,
+      "loss": 0.2888,
+      "step": 33810
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5222952887127286e-05,
+      "loss": 0.2883,
+      "step": 33820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.521844878839744e-05,
+      "loss": 0.2808,
+      "step": 33830
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.52139446896676e-05,
+      "loss": 0.286,
+      "step": 33840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.520944059093775e-05,
+      "loss": 0.2767,
+      "step": 33850
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5204936492207916e-05,
+      "loss": 0.2799,
+      "step": 33860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5200432393478066e-05,
+      "loss": 0.2838,
+      "step": 33870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.519592829474822e-05,
+      "loss": 0.2813,
+      "step": 33880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.519142419601838e-05,
+      "loss": 0.2816,
+      "step": 33890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.518692009728853e-05,
+      "loss": 0.2851,
+      "step": 33900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.518241599855869e-05,
+      "loss": 0.2822,
+      "step": 33910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.517791189982885e-05,
+      "loss": 0.2821,
+      "step": 33920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5173407801099e-05,
+      "loss": 0.2748,
+      "step": 33930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.516890370236916e-05,
+      "loss": 0.2796,
+      "step": 33940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.516439960363931e-05,
+      "loss": 0.2854,
+      "step": 33950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.515989550490947e-05,
+      "loss": 0.2827,
+      "step": 33960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.515539140617963e-05,
+      "loss": 0.2811,
+      "step": 33970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.515088730744978e-05,
+      "loss": 0.281,
+      "step": 33980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5146383208719936e-05,
+      "loss": 0.2787,
+      "step": 33990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5141879109990094e-05,
+      "loss": 0.2809,
+      "step": 34000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.513737501126025e-05,
+      "loss": 0.2846,
+      "step": 34010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.51328709125304e-05,
+      "loss": 0.2836,
+      "step": 34020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.512836681380056e-05,
+      "loss": 0.2829,
+      "step": 34030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.512386271507072e-05,
+      "loss": 0.2848,
+      "step": 34040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.511935861634087e-05,
+      "loss": 0.2798,
+      "step": 34050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.511485451761103e-05,
+      "loss": 0.2766,
+      "step": 34060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.511035041888118e-05,
+      "loss": 0.2861,
+      "step": 34070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.510584632015134e-05,
+      "loss": 0.2845,
+      "step": 34080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.51013422214215e-05,
+      "loss": 0.2807,
+      "step": 34090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.509683812269165e-05,
+      "loss": 0.283,
+      "step": 34100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.509233402396181e-05,
+      "loss": 0.2881,
+      "step": 34110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.508782992523196e-05,
+      "loss": 0.2798,
+      "step": 34120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5083325826502114e-05,
+      "loss": 0.2852,
+      "step": 34130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.507882172777228e-05,
+      "loss": 0.2818,
+      "step": 34140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.507431762904243e-05,
+      "loss": 0.2877,
+      "step": 34150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5069813530312586e-05,
+      "loss": 0.2882,
+      "step": 34160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.5065309431582744e-05,
+      "loss": 0.2833,
+      "step": 34170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.5060805332852895e-05,
+      "loss": 0.2817,
+      "step": 34180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.505630123412305e-05,
+      "loss": 0.2797,
+      "step": 34190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.505179713539321e-05,
+      "loss": 0.279,
+      "step": 34200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.504729303666337e-05,
+      "loss": 0.2801,
+      "step": 34210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.5042788937933524e-05,
+      "loss": 0.2823,
+      "step": 34220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.5038284839203675e-05,
+      "loss": 0.2801,
+      "step": 34230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.503378074047383e-05,
+      "loss": 0.2867,
+      "step": 34240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.502927664174399e-05,
+      "loss": 0.2824,
+      "step": 34250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.502477254301415e-05,
+      "loss": 0.2864,
+      "step": 34260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.50202684442843e-05,
+      "loss": 0.2802,
+      "step": 34270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.5015764345554456e-05,
+      "loss": 0.2849,
+      "step": 34280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.501126024682461e-05,
+      "loss": 0.2757,
+      "step": 34290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.5006756148094764e-05,
+      "loss": 0.2887,
+      "step": 34300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.500225204936493e-05,
+      "loss": 0.2787,
+      "step": 34310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.499774795063508e-05,
+      "loss": 0.2827,
+      "step": 34320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.499324385190523e-05,
+      "loss": 0.274,
+      "step": 34330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4988739753175394e-05,
+      "loss": 0.2813,
+      "step": 34340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4984235654445545e-05,
+      "loss": 0.2819,
+      "step": 34350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.497973155571571e-05,
+      "loss": 0.2845,
+      "step": 34360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.497522745698586e-05,
+      "loss": 0.284,
+      "step": 34370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.497072335825601e-05,
+      "loss": 0.2822,
+      "step": 34380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4966219259526175e-05,
+      "loss": 0.272,
+      "step": 34390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4961715160796325e-05,
+      "loss": 0.2826,
+      "step": 34400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.495721106206648e-05,
+      "loss": 0.2823,
+      "step": 34410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.495270696333664e-05,
+      "loss": 0.2816,
+      "step": 34420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.494820286460679e-05,
+      "loss": 0.2737,
+      "step": 34430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.494369876587695e-05,
+      "loss": 0.2789,
+      "step": 34440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4939194667147106e-05,
+      "loss": 0.2819,
+      "step": 34450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4934690568417264e-05,
+      "loss": 0.2806,
+      "step": 34460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4930186469687414e-05,
+      "loss": 0.2761,
+      "step": 34470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.492568237095757e-05,
+      "loss": 0.2848,
+      "step": 34480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.492117827222773e-05,
+      "loss": 0.2777,
+      "step": 34490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.491667417349789e-05,
+      "loss": 0.2794,
+      "step": 34500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4912170074768044e-05,
+      "loss": 0.2795,
+      "step": 34510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4907665976038195e-05,
+      "loss": 0.2871,
+      "step": 34520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.490316187730835e-05,
+      "loss": 0.2897,
+      "step": 34530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.489865777857851e-05,
+      "loss": 0.2817,
+      "step": 34540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.489415367984866e-05,
+      "loss": 0.2783,
+      "step": 34550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4889649581118825e-05,
+      "loss": 0.2833,
+      "step": 34560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4885145482388976e-05,
+      "loss": 0.2789,
+      "step": 34570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4880641383659126e-05,
+      "loss": 0.2755,
+      "step": 34580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.487613728492929e-05,
+      "loss": 0.279,
+      "step": 34590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.487163318619944e-05,
+      "loss": 0.2857,
+      "step": 34600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.48671290874696e-05,
+      "loss": 0.2799,
+      "step": 34610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4862624988739756e-05,
+      "loss": 0.2788,
+      "step": 34620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.485812089000991e-05,
+      "loss": 0.2752,
+      "step": 34630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.485361679128007e-05,
+      "loss": 0.2792,
+      "step": 34640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.484911269255022e-05,
+      "loss": 0.2807,
+      "step": 34650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.484460859382038e-05,
+      "loss": 0.2785,
+      "step": 34660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.484010449509054e-05,
+      "loss": 0.2795,
+      "step": 34670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.483560039636069e-05,
+      "loss": 0.2741,
+      "step": 34680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4831096297630845e-05,
+      "loss": 0.2756,
+      "step": 34690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4826592198901e-05,
+      "loss": 0.2748,
+      "step": 34700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.482208810017116e-05,
+      "loss": 0.2817,
+      "step": 34710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.481758400144131e-05,
+      "loss": 0.2794,
+      "step": 34720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.481307990271147e-05,
+      "loss": 0.2841,
+      "step": 34730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4808575803981626e-05,
+      "loss": 0.2796,
+      "step": 34740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.480407170525178e-05,
+      "loss": 0.2751,
+      "step": 34750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.479956760652194e-05,
+      "loss": 0.2836,
+      "step": 34760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.479506350779209e-05,
+      "loss": 0.2786,
+      "step": 34770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.479055940906225e-05,
+      "loss": 0.2777,
+      "step": 34780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.478605531033241e-05,
+      "loss": 0.2783,
+      "step": 34790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.478155121160256e-05,
+      "loss": 0.2855,
+      "step": 34800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4777047112872715e-05,
+      "loss": 0.2856,
+      "step": 34810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.477254301414287e-05,
+      "loss": 0.2816,
+      "step": 34820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.476803891541302e-05,
+      "loss": 0.2778,
+      "step": 34830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.476353481668319e-05,
+      "loss": 0.2795,
+      "step": 34840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.475903071795334e-05,
+      "loss": 0.2813,
+      "step": 34850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4754526619223496e-05,
+      "loss": 0.2787,
+      "step": 34860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.475002252049365e-05,
+      "loss": 0.2863,
+      "step": 34870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4745518421763804e-05,
+      "loss": 0.2768,
+      "step": 34880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.474101432303396e-05,
+      "loss": 0.2761,
+      "step": 34890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.473651022430412e-05,
+      "loss": 0.2817,
+      "step": 34900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4732006125574276e-05,
+      "loss": 0.2739,
+      "step": 34910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.472750202684443e-05,
+      "loss": 0.2827,
+      "step": 34920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4722997928114584e-05,
+      "loss": 0.2784,
+      "step": 34930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.471849382938474e-05,
+      "loss": 0.2801,
+      "step": 34940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.47139897306549e-05,
+      "loss": 0.2819,
+      "step": 34950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.470948563192506e-05,
+      "loss": 0.28,
+      "step": 34960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.470498153319521e-05,
+      "loss": 0.2792,
+      "step": 34970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4700477434465365e-05,
+      "loss": 0.2852,
+      "step": 34980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.469597333573552e-05,
+      "loss": 0.2842,
+      "step": 34990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.469146923700567e-05,
+      "loss": 0.2798,
+      "step": 35000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.468696513827584e-05,
+      "loss": 0.2825,
+      "step": 35010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.468246103954599e-05,
+      "loss": 0.2879,
+      "step": 35020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.467795694081614e-05,
+      "loss": 0.2776,
+      "step": 35030
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.46734528420863e-05,
+      "loss": 0.2774,
+      "step": 35040
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4668948743356454e-05,
+      "loss": 0.2794,
+      "step": 35050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.466444464462661e-05,
+      "loss": 0.2759,
+      "step": 35060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.465994054589677e-05,
+      "loss": 0.279,
+      "step": 35070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.465543644716692e-05,
+      "loss": 0.2828,
+      "step": 35080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4650932348437084e-05,
+      "loss": 0.2756,
+      "step": 35090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4646428249707235e-05,
+      "loss": 0.2726,
+      "step": 35100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.464192415097739e-05,
+      "loss": 0.2828,
+      "step": 35110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.463742005224755e-05,
+      "loss": 0.2784,
+      "step": 35120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.46329159535177e-05,
+      "loss": 0.2858,
+      "step": 35130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4628862264660844e-05,
+      "loss": 0.2844,
+      "step": 35140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4624358165930995e-05,
+      "loss": 0.2861,
+      "step": 35150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.461985406720116e-05,
+      "loss": 0.2826,
+      "step": 35160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.461534996847131e-05,
+      "loss": 0.2808,
+      "step": 35170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.461084586974147e-05,
+      "loss": 0.2796,
+      "step": 35180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4606341771011625e-05,
+      "loss": 0.2879,
+      "step": 35190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4601837672281776e-05,
+      "loss": 0.2807,
+      "step": 35200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.459733357355193e-05,
+      "loss": 0.2783,
+      "step": 35210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.459282947482209e-05,
+      "loss": 0.2783,
+      "step": 35220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.458832537609225e-05,
+      "loss": 0.277,
+      "step": 35230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.45838212773624e-05,
+      "loss": 0.283,
+      "step": 35240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4579317178632556e-05,
+      "loss": 0.2809,
+      "step": 35250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4574813079902714e-05,
+      "loss": 0.2848,
+      "step": 35260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.4570308981172864e-05,
+      "loss": 0.272,
+      "step": 35270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.456580488244303e-05,
+      "loss": 0.2791,
+      "step": 35280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.456130078371318e-05,
+      "loss": 0.2823,
+      "step": 35290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.455679668498334e-05,
+      "loss": 0.2871,
+      "step": 35300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4552292586253494e-05,
+      "loss": 0.2779,
+      "step": 35310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4547788487523645e-05,
+      "loss": 0.2822,
+      "step": 35320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.454328438879381e-05,
+      "loss": 0.2824,
+      "step": 35330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.453878029006396e-05,
+      "loss": 0.2815,
+      "step": 35340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.453427619133411e-05,
+      "loss": 0.2781,
+      "step": 35350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4529772092604275e-05,
+      "loss": 0.2874,
+      "step": 35360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4525267993874426e-05,
+      "loss": 0.2812,
+      "step": 35370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.452076389514458e-05,
+      "loss": 0.2793,
+      "step": 35380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.451625979641474e-05,
+      "loss": 0.285,
+      "step": 35390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.451175569768489e-05,
+      "loss": 0.2782,
+      "step": 35400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.450725159895505e-05,
+      "loss": 0.2789,
+      "step": 35410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4502747500225207e-05,
+      "loss": 0.2798,
+      "step": 35420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4498243401495364e-05,
+      "loss": 0.2778,
+      "step": 35430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.449373930276552e-05,
+      "loss": 0.2799,
+      "step": 35440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.448923520403567e-05,
+      "loss": 0.2832,
+      "step": 35450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.448473110530583e-05,
+      "loss": 0.275,
+      "step": 35460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.448022700657599e-05,
+      "loss": 0.2832,
+      "step": 35470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4475722907846145e-05,
+      "loss": 0.2773,
+      "step": 35480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4471218809116295e-05,
+      "loss": 0.2832,
+      "step": 35490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.446671471038645e-05,
+      "loss": 0.2722,
+      "step": 35500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.446221061165661e-05,
+      "loss": 0.278,
+      "step": 35510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.445770651292676e-05,
+      "loss": 0.2772,
+      "step": 35520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4453202414196925e-05,
+      "loss": 0.278,
+      "step": 35530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4448698315467076e-05,
+      "loss": 0.2745,
+      "step": 35540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.444419421673723e-05,
+      "loss": 0.2813,
+      "step": 35550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.443969011800739e-05,
+      "loss": 0.2801,
+      "step": 35560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.443518601927754e-05,
+      "loss": 0.2811,
+      "step": 35570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4430681920547706e-05,
+      "loss": 0.2743,
+      "step": 35580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.442617782181786e-05,
+      "loss": 0.2736,
+      "step": 35590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.442167372308801e-05,
+      "loss": 0.2805,
+      "step": 35600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.441716962435817e-05,
+      "loss": 0.2832,
+      "step": 35610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.441266552562832e-05,
+      "loss": 0.2765,
+      "step": 35620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.440816142689848e-05,
+      "loss": 0.2764,
+      "step": 35630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.440365732816864e-05,
+      "loss": 0.2715,
+      "step": 35640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.439915322943879e-05,
+      "loss": 0.2783,
+      "step": 35650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4394649130708946e-05,
+      "loss": 0.2769,
+      "step": 35660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.43901450319791e-05,
+      "loss": 0.2808,
+      "step": 35670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.438564093324926e-05,
+      "loss": 0.2827,
+      "step": 35680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.438113683451941e-05,
+      "loss": 0.2811,
+      "step": 35690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.437663273578957e-05,
+      "loss": 0.2743,
+      "step": 35700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4372128637059726e-05,
+      "loss": 0.2823,
+      "step": 35710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4367624538329884e-05,
+      "loss": 0.2791,
+      "step": 35720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.436312043960004e-05,
+      "loss": 0.2823,
+      "step": 35730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.435861634087019e-05,
+      "loss": 0.2817,
+      "step": 35740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.435411224214035e-05,
+      "loss": 0.2775,
+      "step": 35750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.434960814341051e-05,
+      "loss": 0.2726,
+      "step": 35760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.434510404468066e-05,
+      "loss": 0.2706,
+      "step": 35770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.434059994595082e-05,
+      "loss": 0.2761,
+      "step": 35780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.433609584722097e-05,
+      "loss": 0.2777,
+      "step": 35790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4331591748491123e-05,
+      "loss": 0.282,
+      "step": 35800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.432708764976129e-05,
+      "loss": 0.2747,
+      "step": 35810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.432258355103144e-05,
+      "loss": 0.2741,
+      "step": 35820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4318079452301596e-05,
+      "loss": 0.2727,
+      "step": 35830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4313575353571753e-05,
+      "loss": 0.2783,
+      "step": 35840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4309071254841904e-05,
+      "loss": 0.2742,
+      "step": 35850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.430456715611207e-05,
+      "loss": 0.276,
+      "step": 35860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.430006305738222e-05,
+      "loss": 0.2771,
+      "step": 35870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.429555895865238e-05,
+      "loss": 0.269,
+      "step": 35880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4291054859922534e-05,
+      "loss": 0.2727,
+      "step": 35890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4286550761192685e-05,
+      "loss": 0.2778,
+      "step": 35900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.428204666246284e-05,
+      "loss": 0.2805,
+      "step": 35910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4277542563733e-05,
+      "loss": 0.278,
+      "step": 35920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.427303846500316e-05,
+      "loss": 0.2841,
+      "step": 35930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.426853436627331e-05,
+      "loss": 0.285,
+      "step": 35940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4264030267543466e-05,
+      "loss": 0.279,
+      "step": 35950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.425952616881362e-05,
+      "loss": 0.2808,
+      "step": 35960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4255022070083774e-05,
+      "loss": 0.2755,
+      "step": 35970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.425051797135394e-05,
+      "loss": 0.2799,
+      "step": 35980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.424601387262409e-05,
+      "loss": 0.2823,
+      "step": 35990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4241509773894246e-05,
+      "loss": 0.2811,
+      "step": 36000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4237005675164404e-05,
+      "loss": 0.2767,
+      "step": 36010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4232501576434554e-05,
+      "loss": 0.2732,
+      "step": 36020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.422799747770471e-05,
+      "loss": 0.2747,
+      "step": 36030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.422349337897487e-05,
+      "loss": 0.2803,
+      "step": 36040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.421898928024502e-05,
+      "loss": 0.2761,
+      "step": 36050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4214485181515184e-05,
+      "loss": 0.2726,
+      "step": 36060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4209981082785335e-05,
+      "loss": 0.2725,
+      "step": 36070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.420547698405549e-05,
+      "loss": 0.2742,
+      "step": 36080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.420097288532565e-05,
+      "loss": 0.2737,
+      "step": 36090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.41964687865958e-05,
+      "loss": 0.2825,
+      "step": 36100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.419196468786596e-05,
+      "loss": 0.2801,
+      "step": 36110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4187460589136116e-05,
+      "loss": 0.2721,
+      "step": 36120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.418295649040627e-05,
+      "loss": 0.2671,
+      "step": 36130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.417845239167643e-05,
+      "loss": 0.2681,
+      "step": 36140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.417394829294658e-05,
+      "loss": 0.2745,
+      "step": 36150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.416944419421674e-05,
+      "loss": 0.2771,
+      "step": 36160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4164940095486896e-05,
+      "loss": 0.2775,
+      "step": 36170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4160435996757054e-05,
+      "loss": 0.2717,
+      "step": 36180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4155931898027205e-05,
+      "loss": 0.279,
+      "step": 36190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.415142779929736e-05,
+      "loss": 0.2799,
+      "step": 36200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.414692370056752e-05,
+      "loss": 0.2778,
+      "step": 36210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.414241960183767e-05,
+      "loss": 0.2767,
+      "step": 36220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4137915503107835e-05,
+      "loss": 0.2808,
+      "step": 36230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4133411404377985e-05,
+      "loss": 0.2708,
+      "step": 36240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4128907305648136e-05,
+      "loss": 0.2771,
+      "step": 36250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.41244032069183e-05,
+      "loss": 0.2743,
+      "step": 36260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.411989910818845e-05,
+      "loss": 0.2733,
+      "step": 36270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.411539500945861e-05,
+      "loss": 0.2769,
+      "step": 36280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4110890910728766e-05,
+      "loss": 0.2762,
+      "step": 36290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.410638681199892e-05,
+      "loss": 0.2779,
+      "step": 36300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.410188271326908e-05,
+      "loss": 0.275,
+      "step": 36310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.409737861453923e-05,
+      "loss": 0.2738,
+      "step": 36320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.409287451580939e-05,
+      "loss": 0.2812,
+      "step": 36330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.408837041707955e-05,
+      "loss": 0.279,
+      "step": 36340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.40838663183497e-05,
+      "loss": 0.2767,
+      "step": 36350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.4079362219619855e-05,
+      "loss": 0.2724,
+      "step": 36360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.407485812089001e-05,
+      "loss": 0.2771,
+      "step": 36370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.407035402216017e-05,
+      "loss": 0.2817,
+      "step": 36380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.406584992343032e-05,
+      "loss": 0.282,
+      "step": 36390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.406134582470048e-05,
+      "loss": 0.2774,
+      "step": 36400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.4056841725970636e-05,
+      "loss": 0.275,
+      "step": 36410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.405233762724079e-05,
+      "loss": 0.2718,
+      "step": 36420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.404783352851095e-05,
+      "loss": 0.2769,
+      "step": 36430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.40433294297811e-05,
+      "loss": 0.2838,
+      "step": 36440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.403882533105126e-05,
+      "loss": 0.2814,
+      "step": 36450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.4034321232321416e-05,
+      "loss": 0.2765,
+      "step": 36460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.402981713359157e-05,
+      "loss": 0.2803,
+      "step": 36470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.4025313034861725e-05,
+      "loss": 0.272,
+      "step": 36480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.402080893613188e-05,
+      "loss": 0.2777,
+      "step": 36490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.401630483740203e-05,
+      "loss": 0.2764,
+      "step": 36500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.40118007386722e-05,
+      "loss": 0.2717,
+      "step": 36510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.400729663994235e-05,
+      "loss": 0.2727,
+      "step": 36520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.4002792541212505e-05,
+      "loss": 0.2808,
+      "step": 36530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.399828844248266e-05,
+      "loss": 0.2771,
+      "step": 36540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3993784343752813e-05,
+      "loss": 0.2756,
+      "step": 36550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.398928024502298e-05,
+      "loss": 0.269,
+      "step": 36560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.398477614629313e-05,
+      "loss": 0.2729,
+      "step": 36570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3980272047563286e-05,
+      "loss": 0.2727,
+      "step": 36580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.397576794883344e-05,
+      "loss": 0.2722,
+      "step": 36590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3971263850103594e-05,
+      "loss": 0.273,
+      "step": 36600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.396675975137375e-05,
+      "loss": 0.2772,
+      "step": 36610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.396225565264391e-05,
+      "loss": 0.275,
+      "step": 36620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3957751553914067e-05,
+      "loss": 0.2743,
+      "step": 36630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.395324745518422e-05,
+      "loss": 0.2682,
+      "step": 36640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3948743356454375e-05,
+      "loss": 0.2743,
+      "step": 36650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.394423925772453e-05,
+      "loss": 0.2751,
+      "step": 36660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.393973515899468e-05,
+      "loss": 0.2749,
+      "step": 36670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.393523106026484e-05,
+      "loss": 0.2708,
+      "step": 36680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3930726961535e-05,
+      "loss": 0.2728,
+      "step": 36690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3926222862805155e-05,
+      "loss": 0.2783,
+      "step": 36700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.392171876407531e-05,
+      "loss": 0.2717,
+      "step": 36710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3917214665345464e-05,
+      "loss": 0.2721,
+      "step": 36720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.391271056661562e-05,
+      "loss": 0.2732,
+      "step": 36730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.390820646788578e-05,
+      "loss": 0.2751,
+      "step": 36740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.390370236915593e-05,
+      "loss": 0.2769,
+      "step": 36750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3899198270426094e-05,
+      "loss": 0.277,
+      "step": 36760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3894694171696244e-05,
+      "loss": 0.2732,
+      "step": 36770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.38901900729664e-05,
+      "loss": 0.2719,
+      "step": 36780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.388568597423656e-05,
+      "loss": 0.2778,
+      "step": 36790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.388118187550671e-05,
+      "loss": 0.2761,
+      "step": 36800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.387667777677687e-05,
+      "loss": 0.2758,
+      "step": 36810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3872173678047025e-05,
+      "loss": 0.2751,
+      "step": 36820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.386766957931718e-05,
+      "loss": 0.2696,
+      "step": 36830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.386316548058734e-05,
+      "loss": 0.2713,
+      "step": 36840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.385866138185749e-05,
+      "loss": 0.2716,
+      "step": 36850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.385415728312765e-05,
+      "loss": 0.274,
+      "step": 36860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3849653184397806e-05,
+      "loss": 0.2737,
+      "step": 36870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.384514908566796e-05,
+      "loss": 0.2766,
+      "step": 36880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3840644986938114e-05,
+      "loss": 0.2766,
+      "step": 36890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.383614088820827e-05,
+      "loss": 0.2749,
+      "step": 36900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.383163678947843e-05,
+      "loss": 0.2711,
+      "step": 36910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.382713269074858e-05,
+      "loss": 0.2766,
+      "step": 36920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.382262859201874e-05,
+      "loss": 0.2701,
+      "step": 36930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3818124493288895e-05,
+      "loss": 0.2788,
+      "step": 36940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3813620394559045e-05,
+      "loss": 0.275,
+      "step": 36950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.380911629582921e-05,
+      "loss": 0.2674,
+      "step": 36960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.380461219709936e-05,
+      "loss": 0.273,
+      "step": 36970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.380010809836952e-05,
+      "loss": 0.2724,
+      "step": 36980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3795603999639675e-05,
+      "loss": 0.2761,
+      "step": 36990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3791099900909826e-05,
+      "loss": 0.2685,
+      "step": 37000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.378659580217999e-05,
+      "loss": 0.2723,
+      "step": 37010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.378209170345014e-05,
+      "loss": 0.279,
+      "step": 37020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.37775876047203e-05,
+      "loss": 0.2738,
+      "step": 37030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3773083505990456e-05,
+      "loss": 0.2683,
+      "step": 37040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.376857940726061e-05,
+      "loss": 0.2773,
+      "step": 37050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3764075308530764e-05,
+      "loss": 0.2732,
+      "step": 37060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.375957120980092e-05,
+      "loss": 0.2767,
+      "step": 37070
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.375506711107108e-05,
+      "loss": 0.2724,
+      "step": 37080
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.375056301234123e-05,
+      "loss": 0.2746,
+      "step": 37090
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.374605891361139e-05,
+      "loss": 0.2753,
+      "step": 37100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3741554814881545e-05,
+      "loss": 0.275,
+      "step": 37110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.37370507161517e-05,
+      "loss": 0.277,
+      "step": 37120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.373254661742185e-05,
+      "loss": 0.2759,
+      "step": 37130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.372804251869201e-05,
+      "loss": 0.2779,
+      "step": 37140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.372353841996217e-05,
+      "loss": 0.2726,
+      "step": 37150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3719034321232326e-05,
+      "loss": 0.2747,
+      "step": 37160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3714530222502476e-05,
+      "loss": 0.272,
+      "step": 37170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3710026123772634e-05,
+      "loss": 0.2729,
+      "step": 37180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.370552202504279e-05,
+      "loss": 0.2722,
+      "step": 37190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.370101792631294e-05,
+      "loss": 0.274,
+      "step": 37200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3696513827583106e-05,
+      "loss": 0.2725,
+      "step": 37210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.369200972885326e-05,
+      "loss": 0.2759,
+      "step": 37220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3687505630123414e-05,
+      "loss": 0.2716,
+      "step": 37230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.368300153139357e-05,
+      "loss": 0.2745,
+      "step": 37240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.367849743266372e-05,
+      "loss": 0.2694,
+      "step": 37250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.367399333393388e-05,
+      "loss": 0.2793,
+      "step": 37260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.366948923520404e-05,
+      "loss": 0.279,
+      "step": 37270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.366543554634718e-05,
+      "loss": 0.2721,
+      "step": 37280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.366093144761733e-05,
+      "loss": 0.2765,
+      "step": 37290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.365642734888749e-05,
+      "loss": 0.2657,
+      "step": 37300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.365192325015765e-05,
+      "loss": 0.2728,
+      "step": 37310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.36474191514278e-05,
+      "loss": 0.2705,
+      "step": 37320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3642915052697955e-05,
+      "loss": 0.2723,
+      "step": 37330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.363841095396811e-05,
+      "loss": 0.2743,
+      "step": 37340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.363390685523827e-05,
+      "loss": 0.2668,
+      "step": 37350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.362940275650843e-05,
+      "loss": 0.2755,
+      "step": 37360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.362489865777858e-05,
+      "loss": 0.28,
+      "step": 37370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3620394559048736e-05,
+      "loss": 0.2691,
+      "step": 37380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3615890460318894e-05,
+      "loss": 0.2724,
+      "step": 37390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.361138636158905e-05,
+      "loss": 0.2654,
+      "step": 37400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.36068822628592e-05,
+      "loss": 0.2687,
+      "step": 37410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.360237816412936e-05,
+      "loss": 0.2716,
+      "step": 37420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.359787406539952e-05,
+      "loss": 0.2682,
+      "step": 37430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.359336996666967e-05,
+      "loss": 0.2699,
+      "step": 37440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3588865867939825e-05,
+      "loss": 0.2704,
+      "step": 37450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.358436176920998e-05,
+      "loss": 0.2728,
+      "step": 37460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.357985767048013e-05,
+      "loss": 0.2751,
+      "step": 37470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.35753535717503e-05,
+      "loss": 0.2766,
+      "step": 37480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.357084947302045e-05,
+      "loss": 0.2757,
+      "step": 37490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3566345374290606e-05,
+      "loss": 0.271,
+      "step": 37500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.356184127556076e-05,
+      "loss": 0.2705,
+      "step": 37510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3557337176830914e-05,
+      "loss": 0.2664,
+      "step": 37520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.355283307810108e-05,
+      "loss": 0.272,
+      "step": 37530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.354832897937123e-05,
+      "loss": 0.2759,
+      "step": 37540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3543824880641386e-05,
+      "loss": 0.2714,
+      "step": 37550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3539320781911544e-05,
+      "loss": 0.2724,
+      "step": 37560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3534816683181694e-05,
+      "loss": 0.2692,
+      "step": 37570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.353031258445185e-05,
+      "loss": 0.2674,
+      "step": 37580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.352580848572201e-05,
+      "loss": 0.27,
+      "step": 37590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.352130438699217e-05,
+      "loss": 0.2723,
+      "step": 37600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.351680028826232e-05,
+      "loss": 0.275,
+      "step": 37610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3512296189532475e-05,
+      "loss": 0.2736,
+      "step": 37620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.350779209080263e-05,
+      "loss": 0.2713,
+      "step": 37630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.350328799207279e-05,
+      "loss": 0.2704,
+      "step": 37640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.349878389334295e-05,
+      "loss": 0.2721,
+      "step": 37650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.34942797946131e-05,
+      "loss": 0.2751,
+      "step": 37660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3489775695883256e-05,
+      "loss": 0.2746,
+      "step": 37670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.348527159715341e-05,
+      "loss": 0.275,
+      "step": 37680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3480767498423564e-05,
+      "loss": 0.2712,
+      "step": 37690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.347626339969372e-05,
+      "loss": 0.2711,
+      "step": 37700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.347175930096388e-05,
+      "loss": 0.2695,
+      "step": 37710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.346725520223403e-05,
+      "loss": 0.2699,
+      "step": 37720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3462751103504194e-05,
+      "loss": 0.2736,
+      "step": 37730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3458247004774345e-05,
+      "loss": 0.2673,
+      "step": 37740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.34537429060445e-05,
+      "loss": 0.2755,
+      "step": 37750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.344923880731466e-05,
+      "loss": 0.2722,
+      "step": 37760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.344473470858481e-05,
+      "loss": 0.2715,
+      "step": 37770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3440230609854975e-05,
+      "loss": 0.2706,
+      "step": 37780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3435726511125125e-05,
+      "loss": 0.2673,
+      "step": 37790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.343122241239528e-05,
+      "loss": 0.2724,
+      "step": 37800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.342671831366544e-05,
+      "loss": 0.2694,
+      "step": 37810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.342221421493559e-05,
+      "loss": 0.2738,
+      "step": 37820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.341771011620575e-05,
+      "loss": 0.2725,
+      "step": 37830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3413206017475906e-05,
+      "loss": 0.268,
+      "step": 37840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3408701918746064e-05,
+      "loss": 0.2664,
+      "step": 37850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3404197820016214e-05,
+      "loss": 0.2756,
+      "step": 37860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.339969372128637e-05,
+      "loss": 0.2723,
+      "step": 37870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.339518962255653e-05,
+      "loss": 0.2693,
+      "step": 37880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.339068552382668e-05,
+      "loss": 0.2737,
+      "step": 37890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.338618142509684e-05,
+      "loss": 0.2741,
+      "step": 37900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3381677326366995e-05,
+      "loss": 0.2747,
+      "step": 37910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.337717322763715e-05,
+      "loss": 0.2704,
+      "step": 37920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.337266912890731e-05,
+      "loss": 0.2728,
+      "step": 37930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.336816503017746e-05,
+      "loss": 0.264,
+      "step": 37940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.336366093144762e-05,
+      "loss": 0.2731,
+      "step": 37950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3359156832717776e-05,
+      "loss": 0.2757,
+      "step": 37960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3354652733987926e-05,
+      "loss": 0.2766,
+      "step": 37970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.335014863525809e-05,
+      "loss": 0.2673,
+      "step": 37980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.334564453652824e-05,
+      "loss": 0.2707,
+      "step": 37990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.33411404377984e-05,
+      "loss": 0.2728,
+      "step": 38000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3336636339068556e-05,
+      "loss": 0.2723,
+      "step": 38010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.333213224033871e-05,
+      "loss": 0.2726,
+      "step": 38020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3327628141608865e-05,
+      "loss": 0.2751,
+      "step": 38030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.332312404287902e-05,
+      "loss": 0.276,
+      "step": 38040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.331861994414918e-05,
+      "loss": 0.2698,
+      "step": 38050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.331411584541934e-05,
+      "loss": 0.2761,
+      "step": 38060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.330961174668949e-05,
+      "loss": 0.268,
+      "step": 38070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3305107647959645e-05,
+      "loss": 0.2706,
+      "step": 38080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.33006035492298e-05,
+      "loss": 0.2757,
+      "step": 38090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3296099450499953e-05,
+      "loss": 0.2726,
+      "step": 38100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.329159535177011e-05,
+      "loss": 0.2782,
+      "step": 38110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.328709125304027e-05,
+      "loss": 0.267,
+      "step": 38120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3282587154310426e-05,
+      "loss": 0.2767,
+      "step": 38130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.327808305558058e-05,
+      "loss": 0.273,
+      "step": 38140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3273578956850734e-05,
+      "loss": 0.2689,
+      "step": 38150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.326907485812089e-05,
+      "loss": 0.2715,
+      "step": 38160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.326457075939104e-05,
+      "loss": 0.2655,
+      "step": 38170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.326006666066121e-05,
+      "loss": 0.2624,
+      "step": 38180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.325556256193136e-05,
+      "loss": 0.272,
+      "step": 38190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3251058463201515e-05,
+      "loss": 0.2696,
+      "step": 38200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.324655436447167e-05,
+      "loss": 0.2729,
+      "step": 38210
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.324205026574182e-05,
+      "loss": 0.2747,
+      "step": 38220
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.323754616701199e-05,
+      "loss": 0.2733,
+      "step": 38230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.323304206828214e-05,
+      "loss": 0.2691,
+      "step": 38240
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3228537969552296e-05,
+      "loss": 0.2702,
+      "step": 38250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.322403387082245e-05,
+      "loss": 0.2755,
+      "step": 38260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3219529772092604e-05,
+      "loss": 0.2695,
+      "step": 38270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.321502567336276e-05,
+      "loss": 0.2651,
+      "step": 38280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.321052157463292e-05,
+      "loss": 0.2704,
+      "step": 38290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3206017475903076e-05,
+      "loss": 0.2673,
+      "step": 38300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.320151337717323e-05,
+      "loss": 0.2664,
+      "step": 38310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3197009278443384e-05,
+      "loss": 0.2719,
+      "step": 38320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.319250517971354e-05,
+      "loss": 0.2705,
+      "step": 38330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.31880010809837e-05,
+      "loss": 0.2696,
+      "step": 38340
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.318349698225385e-05,
+      "loss": 0.2752,
+      "step": 38350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.317899288352401e-05,
+      "loss": 0.2721,
+      "step": 38360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3174488784794165e-05,
+      "loss": 0.2681,
+      "step": 38370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.316998468606432e-05,
+      "loss": 0.2726,
+      "step": 38380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.316548058733447e-05,
+      "loss": 0.2728,
+      "step": 38390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.316097648860463e-05,
+      "loss": 0.2718,
+      "step": 38400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.315647238987479e-05,
+      "loss": 0.2784,
+      "step": 38410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.315196829114494e-05,
+      "loss": 0.2692,
+      "step": 38420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.31474641924151e-05,
+      "loss": 0.2763,
+      "step": 38430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3142960093685254e-05,
+      "loss": 0.2688,
+      "step": 38440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.313845599495541e-05,
+      "loss": 0.2726,
+      "step": 38450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.313395189622557e-05,
+      "loss": 0.2703,
+      "step": 38460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.312944779749572e-05,
+      "loss": 0.2716,
+      "step": 38470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3124943698765884e-05,
+      "loss": 0.2716,
+      "step": 38480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3120439600036035e-05,
+      "loss": 0.2667,
+      "step": 38490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.311593550130619e-05,
+      "loss": 0.2722,
+      "step": 38500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.311143140257635e-05,
+      "loss": 0.274,
+      "step": 38510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.31069273038465e-05,
+      "loss": 0.2733,
+      "step": 38520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.310242320511666e-05,
+      "loss": 0.2709,
+      "step": 38530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3097919106386815e-05,
+      "loss": 0.2674,
+      "step": 38540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3093415007656966e-05,
+      "loss": 0.2689,
+      "step": 38550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3088910908927124e-05,
+      "loss": 0.2742,
+      "step": 38560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.308440681019728e-05,
+      "loss": 0.2697,
+      "step": 38570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.307990271146744e-05,
+      "loss": 0.2743,
+      "step": 38580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.307539861273759e-05,
+      "loss": 0.2667,
+      "step": 38590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.307089451400775e-05,
+      "loss": 0.2682,
+      "step": 38600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.3066390415277904e-05,
+      "loss": 0.266,
+      "step": 38610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.306188631654806e-05,
+      "loss": 0.2738,
+      "step": 38620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.305738221781822e-05,
+      "loss": 0.2726,
+      "step": 38630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.305287811908837e-05,
+      "loss": 0.2673,
+      "step": 38640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.304837402035853e-05,
+      "loss": 0.2718,
+      "step": 38650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.3043869921628685e-05,
+      "loss": 0.2696,
+      "step": 38660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.3039365822898836e-05,
+      "loss": 0.2673,
+      "step": 38670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.3034861724169e-05,
+      "loss": 0.2786,
+      "step": 38680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.303035762543915e-05,
+      "loss": 0.2723,
+      "step": 38690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.302585352670931e-05,
+      "loss": 0.2661,
+      "step": 38700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.3021349427979466e-05,
+      "loss": 0.2711,
+      "step": 38710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.3016845329249616e-05,
+      "loss": 0.2677,
+      "step": 38720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.3012341230519774e-05,
+      "loss": 0.2694,
+      "step": 38730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.300783713178993e-05,
+      "loss": 0.2639,
+      "step": 38740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.300333303306008e-05,
+      "loss": 0.2771,
+      "step": 38750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2998828934330246e-05,
+      "loss": 0.272,
+      "step": 38760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.29943248356004e-05,
+      "loss": 0.2674,
+      "step": 38770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2989820736870555e-05,
+      "loss": 0.2719,
+      "step": 38780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.298531663814071e-05,
+      "loss": 0.2668,
+      "step": 38790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.298081253941086e-05,
+      "loss": 0.2651,
+      "step": 38800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.297630844068102e-05,
+      "loss": 0.27,
+      "step": 38810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.297180434195118e-05,
+      "loss": 0.2672,
+      "step": 38820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2967300243221335e-05,
+      "loss": 0.2696,
+      "step": 38830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2962796144491486e-05,
+      "loss": 0.2735,
+      "step": 38840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2958292045761643e-05,
+      "loss": 0.2758,
+      "step": 38850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.29537879470318e-05,
+      "loss": 0.2663,
+      "step": 38860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.294928384830195e-05,
+      "loss": 0.266,
+      "step": 38870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2944779749572116e-05,
+      "loss": 0.2725,
+      "step": 38880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.294027565084227e-05,
+      "loss": 0.2643,
+      "step": 38890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2935771552112424e-05,
+      "loss": 0.2734,
+      "step": 38900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.293126745338258e-05,
+      "loss": 0.2666,
+      "step": 38910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.292676335465273e-05,
+      "loss": 0.2683,
+      "step": 38920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2922259255922897e-05,
+      "loss": 0.2666,
+      "step": 38930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.291775515719305e-05,
+      "loss": 0.263,
+      "step": 38940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2913251058463205e-05,
+      "loss": 0.2669,
+      "step": 38950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.290874695973336e-05,
+      "loss": 0.2764,
+      "step": 38960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.290424286100351e-05,
+      "loss": 0.2687,
+      "step": 38970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.289973876227367e-05,
+      "loss": 0.2682,
+      "step": 38980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.289523466354383e-05,
+      "loss": 0.268,
+      "step": 38990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.289073056481398e-05,
+      "loss": 0.2644,
+      "step": 39000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2886226466084136e-05,
+      "loss": 0.272,
+      "step": 39010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2881722367354294e-05,
+      "loss": 0.2716,
+      "step": 39020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.287721826862445e-05,
+      "loss": 0.2678,
+      "step": 39030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.287271416989461e-05,
+      "loss": 0.265,
+      "step": 39040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.286821007116476e-05,
+      "loss": 0.2734,
+      "step": 39050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.286370597243492e-05,
+      "loss": 0.2722,
+      "step": 39060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2859201873705074e-05,
+      "loss": 0.2676,
+      "step": 39070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.285469777497523e-05,
+      "loss": 0.2699,
+      "step": 39080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.285019367624538e-05,
+      "loss": 0.2709,
+      "step": 39090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.284568957751554e-05,
+      "loss": 0.2651,
+      "step": 39100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.28411854787857e-05,
+      "loss": 0.2715,
+      "step": 39110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.283668138005585e-05,
+      "loss": 0.2646,
+      "step": 39120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.283217728132601e-05,
+      "loss": 0.2677,
+      "step": 39130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.282767318259616e-05,
+      "loss": 0.2685,
+      "step": 39140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.282316908386632e-05,
+      "loss": 0.2718,
+      "step": 39150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.281866498513648e-05,
+      "loss": 0.2681,
+      "step": 39160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.281416088640663e-05,
+      "loss": 0.2705,
+      "step": 39170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.280965678767679e-05,
+      "loss": 0.2654,
+      "step": 39180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2805152688946944e-05,
+      "loss": 0.2678,
+      "step": 39190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2800648590217095e-05,
+      "loss": 0.2621,
+      "step": 39200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.279614449148726e-05,
+      "loss": 0.268,
+      "step": 39210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.279164039275741e-05,
+      "loss": 0.27,
+      "step": 39220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.278713629402757e-05,
+      "loss": 0.2692,
+      "step": 39230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2782632195297725e-05,
+      "loss": 0.2671,
+      "step": 39240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2778128096567875e-05,
+      "loss": 0.2653,
+      "step": 39250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.277362399783803e-05,
+      "loss": 0.2701,
+      "step": 39260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.276911989910819e-05,
+      "loss": 0.2768,
+      "step": 39270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.276461580037835e-05,
+      "loss": 0.2681,
+      "step": 39280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.27601117016485e-05,
+      "loss": 0.2706,
+      "step": 39290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2755607602918656e-05,
+      "loss": 0.2738,
+      "step": 39300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2751103504188814e-05,
+      "loss": 0.2769,
+      "step": 39310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.274659940545897e-05,
+      "loss": 0.267,
+      "step": 39320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.274209530672913e-05,
+      "loss": 0.271,
+      "step": 39330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2738041617872266e-05,
+      "loss": 0.2714,
+      "step": 39340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.273353751914242e-05,
+      "loss": 0.2716,
+      "step": 39350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2729033420412574e-05,
+      "loss": 0.2691,
+      "step": 39360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.272452932168273e-05,
+      "loss": 0.2631,
+      "step": 39370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.272002522295289e-05,
+      "loss": 0.2721,
+      "step": 39380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.271552112422304e-05,
+      "loss": 0.27,
+      "step": 39390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2711017025493204e-05,
+      "loss": 0.2627,
+      "step": 39400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2706512926763354e-05,
+      "loss": 0.2655,
+      "step": 39410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.270200882803351e-05,
+      "loss": 0.2717,
+      "step": 39420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.269750472930367e-05,
+      "loss": 0.2719,
+      "step": 39430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.269300063057382e-05,
+      "loss": 0.269,
+      "step": 39440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2688496531843984e-05,
+      "loss": 0.2698,
+      "step": 39450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2683992433114135e-05,
+      "loss": 0.2631,
+      "step": 39460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.267948833438429e-05,
+      "loss": 0.265,
+      "step": 39470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.267498423565445e-05,
+      "loss": 0.2674,
+      "step": 39480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.26704801369246e-05,
+      "loss": 0.271,
+      "step": 39490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.266597603819476e-05,
+      "loss": 0.2712,
+      "step": 39500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2661471939464916e-05,
+      "loss": 0.2683,
+      "step": 39510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2656967840735066e-05,
+      "loss": 0.2703,
+      "step": 39520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2652463742005224e-05,
+      "loss": 0.2708,
+      "step": 39530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.264795964327538e-05,
+      "loss": 0.2765,
+      "step": 39540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.264345554454554e-05,
+      "loss": 0.2657,
+      "step": 39550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2638951445815696e-05,
+      "loss": 0.2668,
+      "step": 39560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.263444734708585e-05,
+      "loss": 0.2691,
+      "step": 39570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2629943248356005e-05,
+      "loss": 0.2669,
+      "step": 39580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.262543914962616e-05,
+      "loss": 0.265,
+      "step": 39590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.262093505089632e-05,
+      "loss": 0.2697,
+      "step": 39600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.261643095216647e-05,
+      "loss": 0.2694,
+      "step": 39610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.261192685343663e-05,
+      "loss": 0.2745,
+      "step": 39620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2607422754706785e-05,
+      "loss": 0.267,
+      "step": 39630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2602918655976936e-05,
+      "loss": 0.2679,
+      "step": 39640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.25984145572471e-05,
+      "loss": 0.2672,
+      "step": 39650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.259391045851725e-05,
+      "loss": 0.2621,
+      "step": 39660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.258940635978741e-05,
+      "loss": 0.2673,
+      "step": 39670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2584902261057566e-05,
+      "loss": 0.2713,
+      "step": 39680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.258039816232772e-05,
+      "loss": 0.2723,
+      "step": 39690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.257589406359788e-05,
+      "loss": 0.2679,
+      "step": 39700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.257138996486803e-05,
+      "loss": 0.2661,
+      "step": 39710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.256688586613819e-05,
+      "loss": 0.2634,
+      "step": 39720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.256238176740835e-05,
+      "loss": 0.2627,
+      "step": 39730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.25578776686785e-05,
+      "loss": 0.2652,
+      "step": 39740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2553373569948655e-05,
+      "loss": 0.2738,
+      "step": 39750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.254886947121881e-05,
+      "loss": 0.2694,
+      "step": 39760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.254436537248896e-05,
+      "loss": 0.2702,
+      "step": 39770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.253986127375912e-05,
+      "loss": 0.269,
+      "step": 39780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.253535717502928e-05,
+      "loss": 0.2682,
+      "step": 39790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2530853076299436e-05,
+      "loss": 0.2695,
+      "step": 39800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2526348977569586e-05,
+      "loss": 0.2625,
+      "step": 39810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2521844878839744e-05,
+      "loss": 0.2629,
+      "step": 39820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.25173407801099e-05,
+      "loss": 0.2773,
+      "step": 39830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.251283668138006e-05,
+      "loss": 0.2761,
+      "step": 39840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2508332582650216e-05,
+      "loss": 0.2662,
+      "step": 39850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.250382848392037e-05,
+      "loss": 0.2767,
+      "step": 39860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2499324385190525e-05,
+      "loss": 0.2685,
+      "step": 39870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.249482028646068e-05,
+      "loss": 0.2698,
+      "step": 39880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.249031618773083e-05,
+      "loss": 0.2655,
+      "step": 39890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2485812089001e-05,
+      "loss": 0.2664,
+      "step": 39900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.248130799027115e-05,
+      "loss": 0.2719,
+      "step": 39910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2476803891541305e-05,
+      "loss": 0.2649,
+      "step": 39920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.247229979281146e-05,
+      "loss": 0.2628,
+      "step": 39930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2467795694081613e-05,
+      "loss": 0.2604,
+      "step": 39940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.246329159535177e-05,
+      "loss": 0.273,
+      "step": 39950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.245878749662193e-05,
+      "loss": 0.2672,
+      "step": 39960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.245428339789208e-05,
+      "loss": 0.2728,
+      "step": 39970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.244977929916224e-05,
+      "loss": 0.267,
+      "step": 39980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2445275200432394e-05,
+      "loss": 0.2714,
+      "step": 39990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.244077110170255e-05,
+      "loss": 0.2708,
+      "step": 40000
+    },
+    {
+      "epoch": 0.36,
+      "eval_NEIMS_canon_smiles": 0.9353333333333333,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.017566666666666668,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.39882045391989296,
+      "eval_NEIMS_exact_mols": 0.017516666666666666,
+      "eval_NEIMS_exact_smiles": 0.01685,
+      "eval_NEIMS_loss": 0.28149697184562683,
+      "eval_NEIMS_matched_formulas": 0.4307,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.2970297724918811,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.017616666666666666,
+      "eval_NEIMS_runtime": 716.5006,
+      "eval_NEIMS_samples_per_second": 83.74,
+      "eval_NEIMS_steps_per_second": 1.309,
+      "step": 40000
+    },
+    {
+      "epoch": 0.36,
+      "eval_RASSP_canon_smiles": 0.9259191237853541,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.04429492102714796,
+      "eval_RASSP_daylight_tanimoto_simil": 0.4832393482022256,
+      "eval_RASSP_exact_mols": 0.04409456706848766,
+      "eval_RASSP_exact_smiles": 0.042408254583096806,
+      "eval_RASSP_loss": 0.22775410115718842,
+      "eval_RASSP_matched_formulas": 0.573262764216783,
+      "eval_RASSP_morgan_tanimoto_simil": 0.34593379198899715,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.04417804788459612,
+      "eval_RASSP_runtime": 820.6658,
+      "eval_RASSP_samples_per_second": 72.982,
+      "eval_RASSP_steps_per_second": 1.141,
+      "step": 40000
+    },
+    {
+      "epoch": 0.36,
+      "eval_NIST_canon_smiles": 0.8790502892429997,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.007630336799517337,
+      "eval_NIST_daylight_tanimoto_simil": 0.25814331426059783,
+      "eval_NIST_exact_mols": 0.007239947474890869,
+      "eval_NIST_exact_smiles": 0.006956027966071619,
+      "eval_NIST_loss": 1.4002385139465332,
+      "eval_NIST_matched_formulas": 0.06413031905454804,
+      "eval_NIST_morgan_tanimoto_simil": 0.2147215198195743,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.007559356922312525,
+      "eval_NIST_runtime": 475.4693,
+      "eval_NIST_samples_per_second": 59.261,
+      "eval_NIST_steps_per_second": 0.928,
+      "step": 40000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.243626700297271e-05,
+      "loss": 0.2685,
+      "step": 40010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.243176290424286e-05,
+      "loss": 0.2671,
+      "step": 40020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.242725880551302e-05,
+      "loss": 0.2632,
+      "step": 40030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2422754706783175e-05,
+      "loss": 0.2661,
+      "step": 40040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.241825060805333e-05,
+      "loss": 0.2719,
+      "step": 40050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.241374650932348e-05,
+      "loss": 0.2689,
+      "step": 40060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.240924241059364e-05,
+      "loss": 0.2738,
+      "step": 40070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.24047383118638e-05,
+      "loss": 0.2683,
+      "step": 40080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.240023421313395e-05,
+      "loss": 0.2654,
+      "step": 40090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.239573011440411e-05,
+      "loss": 0.2713,
+      "step": 40100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2391226015674264e-05,
+      "loss": 0.2611,
+      "step": 40110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.238672191694442e-05,
+      "loss": 0.2669,
+      "step": 40120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.238221781821458e-05,
+      "loss": 0.2617,
+      "step": 40130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.237771371948473e-05,
+      "loss": 0.2653,
+      "step": 40140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2373209620754894e-05,
+      "loss": 0.2711,
+      "step": 40150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2368705522025044e-05,
+      "loss": 0.2699,
+      "step": 40160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2364201423295195e-05,
+      "loss": 0.2662,
+      "step": 40170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.235969732456536e-05,
+      "loss": 0.2711,
+      "step": 40180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.235519322583551e-05,
+      "loss": 0.2664,
+      "step": 40190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.235068912710567e-05,
+      "loss": 0.2659,
+      "step": 40200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2346185028375825e-05,
+      "loss": 0.2731,
+      "step": 40210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2341680929645976e-05,
+      "loss": 0.2656,
+      "step": 40220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.233717683091613e-05,
+      "loss": 0.2703,
+      "step": 40230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.233267273218629e-05,
+      "loss": 0.2693,
+      "step": 40240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.232816863345645e-05,
+      "loss": 0.2745,
+      "step": 40250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2323664534726606e-05,
+      "loss": 0.2676,
+      "step": 40260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2319160435996756e-05,
+      "loss": 0.268,
+      "step": 40270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2314656337266914e-05,
+      "loss": 0.2693,
+      "step": 40280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.231015223853707e-05,
+      "loss": 0.2634,
+      "step": 40290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.230564813980723e-05,
+      "loss": 0.266,
+      "step": 40300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.230114404107738e-05,
+      "loss": 0.2642,
+      "step": 40310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.229663994234754e-05,
+      "loss": 0.2651,
+      "step": 40320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2292135843617695e-05,
+      "loss": 0.2663,
+      "step": 40330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2287631744887845e-05,
+      "loss": 0.2676,
+      "step": 40340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.228312764615801e-05,
+      "loss": 0.2646,
+      "step": 40350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.227862354742816e-05,
+      "loss": 0.2649,
+      "step": 40360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.227411944869832e-05,
+      "loss": 0.2682,
+      "step": 40370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2269615349968475e-05,
+      "loss": 0.2672,
+      "step": 40380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2265111251238626e-05,
+      "loss": 0.2634,
+      "step": 40390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.226060715250879e-05,
+      "loss": 0.2747,
+      "step": 40400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.225610305377894e-05,
+      "loss": 0.2677,
+      "step": 40410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.225159895504909e-05,
+      "loss": 0.2672,
+      "step": 40420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2247094856319256e-05,
+      "loss": 0.2642,
+      "step": 40430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.224259075758941e-05,
+      "loss": 0.2702,
+      "step": 40440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2238086658859564e-05,
+      "loss": 0.265,
+      "step": 40450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.223358256012972e-05,
+      "loss": 0.2685,
+      "step": 40460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.222907846139987e-05,
+      "loss": 0.268,
+      "step": 40470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.222457436267003e-05,
+      "loss": 0.2627,
+      "step": 40480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.222007026394019e-05,
+      "loss": 0.2657,
+      "step": 40490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2215566165210345e-05,
+      "loss": 0.2628,
+      "step": 40500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2211062066480496e-05,
+      "loss": 0.2662,
+      "step": 40510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.220655796775065e-05,
+      "loss": 0.2663,
+      "step": 40520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.220205386902081e-05,
+      "loss": 0.2605,
+      "step": 40530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.219754977029097e-05,
+      "loss": 0.2652,
+      "step": 40540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2193045671561126e-05,
+      "loss": 0.2686,
+      "step": 40550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2188541572831276e-05,
+      "loss": 0.2665,
+      "step": 40560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2184037474101434e-05,
+      "loss": 0.267,
+      "step": 40570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.217953337537159e-05,
+      "loss": 0.2659,
+      "step": 40580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.217502927664174e-05,
+      "loss": 0.2583,
+      "step": 40590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2170525177911906e-05,
+      "loss": 0.2675,
+      "step": 40600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.216602107918206e-05,
+      "loss": 0.2613,
+      "step": 40610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.216151698045221e-05,
+      "loss": 0.2596,
+      "step": 40620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.215701288172237e-05,
+      "loss": 0.2678,
+      "step": 40630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.215250878299252e-05,
+      "loss": 0.2656,
+      "step": 40640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.214800468426268e-05,
+      "loss": 0.2681,
+      "step": 40650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.214350058553284e-05,
+      "loss": 0.2667,
+      "step": 40660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.213899648680299e-05,
+      "loss": 0.2585,
+      "step": 40670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.213449238807315e-05,
+      "loss": 0.2694,
+      "step": 40680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.21299882893433e-05,
+      "loss": 0.2621,
+      "step": 40690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.212548419061346e-05,
+      "loss": 0.2666,
+      "step": 40700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.212098009188362e-05,
+      "loss": 0.2602,
+      "step": 40710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.211647599315377e-05,
+      "loss": 0.2652,
+      "step": 40720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2111971894423927e-05,
+      "loss": 0.2665,
+      "step": 40730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2107467795694084e-05,
+      "loss": 0.2643,
+      "step": 40740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.210296369696424e-05,
+      "loss": 0.2622,
+      "step": 40750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.209845959823439e-05,
+      "loss": 0.2618,
+      "step": 40760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.209395549950455e-05,
+      "loss": 0.2694,
+      "step": 40770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.208945140077471e-05,
+      "loss": 0.2638,
+      "step": 40780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.208494730204486e-05,
+      "loss": 0.2623,
+      "step": 40790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.208044320331502e-05,
+      "loss": 0.2674,
+      "step": 40800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.207593910458517e-05,
+      "loss": 0.2717,
+      "step": 40810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.207143500585533e-05,
+      "loss": 0.2671,
+      "step": 40820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.206693090712549e-05,
+      "loss": 0.266,
+      "step": 40830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.206242680839564e-05,
+      "loss": 0.2624,
+      "step": 40840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.20579227096658e-05,
+      "loss": 0.2616,
+      "step": 40850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2053418610935954e-05,
+      "loss": 0.2722,
+      "step": 40860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2048914512206104e-05,
+      "loss": 0.2619,
+      "step": 40870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.204441041347627e-05,
+      "loss": 0.2642,
+      "step": 40880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.203990631474642e-05,
+      "loss": 0.267,
+      "step": 40890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.203540221601658e-05,
+      "loss": 0.2683,
+      "step": 40900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.2030898117286734e-05,
+      "loss": 0.2646,
+      "step": 40910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.2026394018556885e-05,
+      "loss": 0.2635,
+      "step": 40920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.202188991982704e-05,
+      "loss": 0.2689,
+      "step": 40930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.20173858210972e-05,
+      "loss": 0.2656,
+      "step": 40940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.201288172236736e-05,
+      "loss": 0.268,
+      "step": 40950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.2008377623637515e-05,
+      "loss": 0.2719,
+      "step": 40960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.2003873524907666e-05,
+      "loss": 0.2611,
+      "step": 40970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.199936942617782e-05,
+      "loss": 0.2628,
+      "step": 40980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.199486532744798e-05,
+      "loss": 0.2631,
+      "step": 40990
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.199036122871814e-05,
+      "loss": 0.266,
+      "step": 41000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.198585712998829e-05,
+      "loss": 0.2657,
+      "step": 41010
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1981353031258446e-05,
+      "loss": 0.2631,
+      "step": 41020
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1976848932528604e-05,
+      "loss": 0.263,
+      "step": 41030
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1972344833798755e-05,
+      "loss": 0.2643,
+      "step": 41040
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.196784073506892e-05,
+      "loss": 0.266,
+      "step": 41050
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.196333663633907e-05,
+      "loss": 0.2676,
+      "step": 41060
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.195883253760922e-05,
+      "loss": 0.2692,
+      "step": 41070
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1954328438879385e-05,
+      "loss": 0.2642,
+      "step": 41080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1949824340149535e-05,
+      "loss": 0.2628,
+      "step": 41090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.19453202414197e-05,
+      "loss": 0.2608,
+      "step": 41100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.194081614268985e-05,
+      "loss": 0.261,
+      "step": 41110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.193631204396e-05,
+      "loss": 0.2661,
+      "step": 41120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1931807945230165e-05,
+      "loss": 0.2611,
+      "step": 41130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1927303846500316e-05,
+      "loss": 0.263,
+      "step": 41140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1922799747770473e-05,
+      "loss": 0.2615,
+      "step": 41150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.191829564904063e-05,
+      "loss": 0.2655,
+      "step": 41160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.191379155031078e-05,
+      "loss": 0.2692,
+      "step": 41170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.190928745158094e-05,
+      "loss": 0.2632,
+      "step": 41180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.19047833528511e-05,
+      "loss": 0.2671,
+      "step": 41190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1900279254121254e-05,
+      "loss": 0.2636,
+      "step": 41200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1895775155391405e-05,
+      "loss": 0.2579,
+      "step": 41210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.189127105666156e-05,
+      "loss": 0.2599,
+      "step": 41220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.188676695793172e-05,
+      "loss": 0.2624,
+      "step": 41230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.188226285920188e-05,
+      "loss": 0.2624,
+      "step": 41240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1877758760472035e-05,
+      "loss": 0.27,
+      "step": 41250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1873254661742186e-05,
+      "loss": 0.2622,
+      "step": 41260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.186875056301234e-05,
+      "loss": 0.2584,
+      "step": 41270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.18642464642825e-05,
+      "loss": 0.2595,
+      "step": 41280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.185974236555265e-05,
+      "loss": 0.2687,
+      "step": 41290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1855238266822815e-05,
+      "loss": 0.2702,
+      "step": 41300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1850734168092966e-05,
+      "loss": 0.2644,
+      "step": 41310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.184623006936312e-05,
+      "loss": 0.2612,
+      "step": 41320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.184172597063328e-05,
+      "loss": 0.2621,
+      "step": 41330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.183722187190343e-05,
+      "loss": 0.2664,
+      "step": 41340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.183271777317359e-05,
+      "loss": 0.2648,
+      "step": 41350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.182821367444375e-05,
+      "loss": 0.2645,
+      "step": 41360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.18237095757139e-05,
+      "loss": 0.2577,
+      "step": 41370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.181920547698406e-05,
+      "loss": 0.2639,
+      "step": 41380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.181470137825421e-05,
+      "loss": 0.2602,
+      "step": 41390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.181019727952437e-05,
+      "loss": 0.2664,
+      "step": 41400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.180569318079453e-05,
+      "loss": 0.2639,
+      "step": 41410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.180118908206468e-05,
+      "loss": 0.2644,
+      "step": 41420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1796684983334836e-05,
+      "loss": 0.2613,
+      "step": 41430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.179263129447797e-05,
+      "loss": 0.2675,
+      "step": 41440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.178812719574813e-05,
+      "loss": 0.2628,
+      "step": 41450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.178362309701829e-05,
+      "loss": 0.2612,
+      "step": 41460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1779118998288445e-05,
+      "loss": 0.2618,
+      "step": 41470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.17746148995586e-05,
+      "loss": 0.26,
+      "step": 41480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1770110800828753e-05,
+      "loss": 0.2614,
+      "step": 41490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.176560670209891e-05,
+      "loss": 0.2657,
+      "step": 41500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.176110260336907e-05,
+      "loss": 0.2654,
+      "step": 41510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1756598504639226e-05,
+      "loss": 0.2564,
+      "step": 41520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.175209440590938e-05,
+      "loss": 0.2605,
+      "step": 41530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1747590307179534e-05,
+      "loss": 0.2548,
+      "step": 41540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.174308620844969e-05,
+      "loss": 0.2676,
+      "step": 41550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.173858210971984e-05,
+      "loss": 0.2616,
+      "step": 41560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.173407801099001e-05,
+      "loss": 0.264,
+      "step": 41570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.172957391226016e-05,
+      "loss": 0.2634,
+      "step": 41580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1725069813530315e-05,
+      "loss": 0.27,
+      "step": 41590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.172056571480047e-05,
+      "loss": 0.2617,
+      "step": 41600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.171606161607062e-05,
+      "loss": 0.2648,
+      "step": 41610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.171155751734079e-05,
+      "loss": 0.2618,
+      "step": 41620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.170705341861094e-05,
+      "loss": 0.2629,
+      "step": 41630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.170254931988109e-05,
+      "loss": 0.2634,
+      "step": 41640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.169804522115125e-05,
+      "loss": 0.2621,
+      "step": 41650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1693541122421404e-05,
+      "loss": 0.2604,
+      "step": 41660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.168903702369156e-05,
+      "loss": 0.2635,
+      "step": 41670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.168453292496172e-05,
+      "loss": 0.2625,
+      "step": 41680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.168002882623187e-05,
+      "loss": 0.2606,
+      "step": 41690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.167552472750203e-05,
+      "loss": 0.2743,
+      "step": 41700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1671020628772184e-05,
+      "loss": 0.2694,
+      "step": 41710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.166651653004234e-05,
+      "loss": 0.2619,
+      "step": 41720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.166201243131249e-05,
+      "loss": 0.2685,
+      "step": 41730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.165750833258265e-05,
+      "loss": 0.2651,
+      "step": 41740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.165300423385281e-05,
+      "loss": 0.2682,
+      "step": 41750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1648500135122965e-05,
+      "loss": 0.2676,
+      "step": 41760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.164399603639312e-05,
+      "loss": 0.2594,
+      "step": 41770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.163949193766327e-05,
+      "loss": 0.264,
+      "step": 41780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.163498783893343e-05,
+      "loss": 0.2587,
+      "step": 41790
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.163048374020359e-05,
+      "loss": 0.2622,
+      "step": 41800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.162597964147374e-05,
+      "loss": 0.2618,
+      "step": 41810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.16214755427439e-05,
+      "loss": 0.2641,
+      "step": 41820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1616971444014054e-05,
+      "loss": 0.2595,
+      "step": 41830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1612467345284205e-05,
+      "loss": 0.2653,
+      "step": 41840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.160796324655437e-05,
+      "loss": 0.2586,
+      "step": 41850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.160345914782452e-05,
+      "loss": 0.2673,
+      "step": 41860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.159895504909468e-05,
+      "loss": 0.2624,
+      "step": 41870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1594450950364835e-05,
+      "loss": 0.2618,
+      "step": 41880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1589946851634985e-05,
+      "loss": 0.2661,
+      "step": 41890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.158544275290515e-05,
+      "loss": 0.2633,
+      "step": 41900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.15809386541753e-05,
+      "loss": 0.2643,
+      "step": 41910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.157643455544546e-05,
+      "loss": 0.2573,
+      "step": 41920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1571930456715615e-05,
+      "loss": 0.2619,
+      "step": 41930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1567426357985766e-05,
+      "loss": 0.2663,
+      "step": 41940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1562922259255924e-05,
+      "loss": 0.2607,
+      "step": 41950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.155841816052608e-05,
+      "loss": 0.2649,
+      "step": 41960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.155391406179624e-05,
+      "loss": 0.261,
+      "step": 41970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.154940996306639e-05,
+      "loss": 0.2642,
+      "step": 41980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.154490586433655e-05,
+      "loss": 0.2536,
+      "step": 41990
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1540401765606704e-05,
+      "loss": 0.2601,
+      "step": 42000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1535897666876855e-05,
+      "loss": 0.2599,
+      "step": 42010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.153139356814702e-05,
+      "loss": 0.2597,
+      "step": 42020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.152688946941717e-05,
+      "loss": 0.2634,
+      "step": 42030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.152238537068733e-05,
+      "loss": 0.2628,
+      "step": 42040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1517881271957485e-05,
+      "loss": 0.2565,
+      "step": 42050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1513377173227636e-05,
+      "loss": 0.2613,
+      "step": 42060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.15088730744978e-05,
+      "loss": 0.2639,
+      "step": 42070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.150436897576795e-05,
+      "loss": 0.261,
+      "step": 42080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.14998648770381e-05,
+      "loss": 0.2608,
+      "step": 42090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1495360778308266e-05,
+      "loss": 0.2618,
+      "step": 42100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1490856679578416e-05,
+      "loss": 0.2672,
+      "step": 42110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1486352580848574e-05,
+      "loss": 0.2654,
+      "step": 42120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.148184848211873e-05,
+      "loss": 0.2662,
+      "step": 42130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.147734438338888e-05,
+      "loss": 0.2614,
+      "step": 42140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.147284028465904e-05,
+      "loss": 0.2646,
+      "step": 42150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.14683361859292e-05,
+      "loss": 0.2641,
+      "step": 42160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1463832087199355e-05,
+      "loss": 0.263,
+      "step": 42170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.145932798846951e-05,
+      "loss": 0.2633,
+      "step": 42180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.145482388973966e-05,
+      "loss": 0.2688,
+      "step": 42190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.145031979100982e-05,
+      "loss": 0.2616,
+      "step": 42200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.144581569227998e-05,
+      "loss": 0.2652,
+      "step": 42210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1441311593550135e-05,
+      "loss": 0.2593,
+      "step": 42220
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1436807494820286e-05,
+      "loss": 0.2574,
+      "step": 42230
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1432303396090443e-05,
+      "loss": 0.2618,
+      "step": 42240
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.14277992973606e-05,
+      "loss": 0.2631,
+      "step": 42250
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.142329519863075e-05,
+      "loss": 0.2557,
+      "step": 42260
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1418791099900916e-05,
+      "loss": 0.2654,
+      "step": 42270
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1414287001171067e-05,
+      "loss": 0.2625,
+      "step": 42280
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.140978290244122e-05,
+      "loss": 0.2629,
+      "step": 42290
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.140527880371138e-05,
+      "loss": 0.2673,
+      "step": 42300
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.140077470498153e-05,
+      "loss": 0.2611,
+      "step": 42310
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.139627060625169e-05,
+      "loss": 0.2628,
+      "step": 42320
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.139176650752185e-05,
+      "loss": 0.2638,
+      "step": 42330
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1387262408792e-05,
+      "loss": 0.2642,
+      "step": 42340
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.138275831006216e-05,
+      "loss": 0.2572,
+      "step": 42350
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.137825421133231e-05,
+      "loss": 0.2625,
+      "step": 42360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.137375011260247e-05,
+      "loss": 0.2562,
+      "step": 42370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.136924601387263e-05,
+      "loss": 0.2612,
+      "step": 42380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.136474191514278e-05,
+      "loss": 0.2664,
+      "step": 42390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1360237816412936e-05,
+      "loss": 0.2606,
+      "step": 42400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1355733717683094e-05,
+      "loss": 0.2613,
+      "step": 42410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.135122961895325e-05,
+      "loss": 0.2579,
+      "step": 42420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.13467255202234e-05,
+      "loss": 0.2552,
+      "step": 42430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.134222142149356e-05,
+      "loss": 0.2661,
+      "step": 42440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.133771732276372e-05,
+      "loss": 0.2603,
+      "step": 42450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1333213224033874e-05,
+      "loss": 0.2597,
+      "step": 42460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.132870912530403e-05,
+      "loss": 0.2593,
+      "step": 42470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.132420502657418e-05,
+      "loss": 0.2611,
+      "step": 42480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.131970092784434e-05,
+      "loss": 0.2612,
+      "step": 42490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.13151968291145e-05,
+      "loss": 0.2696,
+      "step": 42500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.131069273038465e-05,
+      "loss": 0.2579,
+      "step": 42510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.130618863165481e-05,
+      "loss": 0.2565,
+      "step": 42520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.130168453292496e-05,
+      "loss": 0.2612,
+      "step": 42530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1297180434195114e-05,
+      "loss": 0.2654,
+      "step": 42540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.129267633546528e-05,
+      "loss": 0.2652,
+      "step": 42550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.128817223673543e-05,
+      "loss": 0.258,
+      "step": 42560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1283668138005586e-05,
+      "loss": 0.2623,
+      "step": 42570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1279164039275744e-05,
+      "loss": 0.2582,
+      "step": 42580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1274659940545895e-05,
+      "loss": 0.2598,
+      "step": 42590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.127015584181606e-05,
+      "loss": 0.2621,
+      "step": 42600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.126565174308621e-05,
+      "loss": 0.2654,
+      "step": 42610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.126114764435637e-05,
+      "loss": 0.255,
+      "step": 42620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1256643545626525e-05,
+      "loss": 0.2579,
+      "step": 42630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1252139446896675e-05,
+      "loss": 0.2626,
+      "step": 42640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.124763534816683e-05,
+      "loss": 0.2591,
+      "step": 42650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.124313124943699e-05,
+      "loss": 0.2551,
+      "step": 42660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.123862715070715e-05,
+      "loss": 0.2633,
+      "step": 42670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.12341230519773e-05,
+      "loss": 0.2603,
+      "step": 42680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1229618953247456e-05,
+      "loss": 0.2664,
+      "step": 42690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1225114854517614e-05,
+      "loss": 0.2625,
+      "step": 42700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1220610755787764e-05,
+      "loss": 0.2626,
+      "step": 42710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.121610665705793e-05,
+      "loss": 0.2585,
+      "step": 42720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.121160255832808e-05,
+      "loss": 0.2577,
+      "step": 42730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.120709845959824e-05,
+      "loss": 0.2604,
+      "step": 42740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1202594360868394e-05,
+      "loss": 0.2636,
+      "step": 42750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1198090262138545e-05,
+      "loss": 0.2566,
+      "step": 42760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.11935861634087e-05,
+      "loss": 0.2637,
+      "step": 42770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.118908206467886e-05,
+      "loss": 0.264,
+      "step": 42780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.118457796594901e-05,
+      "loss": 0.2642,
+      "step": 42790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1180073867219175e-05,
+      "loss": 0.2587,
+      "step": 42800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1175569768489326e-05,
+      "loss": 0.2544,
+      "step": 42810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.117106566975948e-05,
+      "loss": 0.2666,
+      "step": 42820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.116656157102964e-05,
+      "loss": 0.2616,
+      "step": 42830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.116205747229979e-05,
+      "loss": 0.2641,
+      "step": 42840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.115755337356995e-05,
+      "loss": 0.259,
+      "step": 42850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1153049274840106e-05,
+      "loss": 0.256,
+      "step": 42860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1148545176110264e-05,
+      "loss": 0.2614,
+      "step": 42870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.114404107738042e-05,
+      "loss": 0.2639,
+      "step": 42880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.113953697865057e-05,
+      "loss": 0.2599,
+      "step": 42890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.113503287992073e-05,
+      "loss": 0.267,
+      "step": 42900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.113052878119089e-05,
+      "loss": 0.2592,
+      "step": 42910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1126024682461044e-05,
+      "loss": 0.2574,
+      "step": 42920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1121520583731195e-05,
+      "loss": 0.253,
+      "step": 42930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.111701648500135e-05,
+      "loss": 0.2628,
+      "step": 42940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.111251238627151e-05,
+      "loss": 0.258,
+      "step": 42950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.110800828754166e-05,
+      "loss": 0.2625,
+      "step": 42960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.110350418881182e-05,
+      "loss": 0.2502,
+      "step": 42970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1099000090081976e-05,
+      "loss": 0.271,
+      "step": 42980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1094495991352127e-05,
+      "loss": 0.2648,
+      "step": 42990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.108999189262229e-05,
+      "loss": 0.2641,
+      "step": 43000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.108548779389244e-05,
+      "loss": 0.2568,
+      "step": 43010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.10809836951626e-05,
+      "loss": 0.2578,
+      "step": 43020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1076479596432757e-05,
+      "loss": 0.2588,
+      "step": 43030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.107197549770291e-05,
+      "loss": 0.2573,
+      "step": 43040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.106747139897307e-05,
+      "loss": 0.2591,
+      "step": 43050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.106296730024322e-05,
+      "loss": 0.2595,
+      "step": 43060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.105846320151338e-05,
+      "loss": 0.2586,
+      "step": 43070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.105395910278354e-05,
+      "loss": 0.2551,
+      "step": 43080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.104945500405369e-05,
+      "loss": 0.256,
+      "step": 43090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1044950905323845e-05,
+      "loss": 0.257,
+      "step": 43100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.1040446806594e-05,
+      "loss": 0.2603,
+      "step": 43110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.103594270786416e-05,
+      "loss": 0.257,
+      "step": 43120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.103143860913431e-05,
+      "loss": 0.2559,
+      "step": 43130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.102693451040447e-05,
+      "loss": 0.2558,
+      "step": 43140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.1022430411674626e-05,
+      "loss": 0.2564,
+      "step": 43150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.1017926312944784e-05,
+      "loss": 0.2581,
+      "step": 43160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.101342221421494e-05,
+      "loss": 0.2599,
+      "step": 43170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.100891811548509e-05,
+      "loss": 0.2548,
+      "step": 43180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.100441401675525e-05,
+      "loss": 0.2552,
+      "step": 43190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.099990991802541e-05,
+      "loss": 0.2596,
+      "step": 43200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.099540581929556e-05,
+      "loss": 0.2649,
+      "step": 43210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0990901720565715e-05,
+      "loss": 0.2539,
+      "step": 43220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.098639762183587e-05,
+      "loss": 0.2609,
+      "step": 43230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.098189352310602e-05,
+      "loss": 0.2575,
+      "step": 43240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.097738942437619e-05,
+      "loss": 0.2597,
+      "step": 43250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.097288532564634e-05,
+      "loss": 0.2651,
+      "step": 43260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0968381226916496e-05,
+      "loss": 0.2581,
+      "step": 43270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.096387712818665e-05,
+      "loss": 0.2572,
+      "step": 43280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0959373029456804e-05,
+      "loss": 0.2656,
+      "step": 43290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.095486893072697e-05,
+      "loss": 0.2566,
+      "step": 43300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.095036483199712e-05,
+      "loss": 0.2594,
+      "step": 43310
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0945860733267276e-05,
+      "loss": 0.2511,
+      "step": 43320
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0941356634537434e-05,
+      "loss": 0.2602,
+      "step": 43330
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0936852535807585e-05,
+      "loss": 0.2637,
+      "step": 43340
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.093234843707774e-05,
+      "loss": 0.2612,
+      "step": 43350
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.09278443383479e-05,
+      "loss": 0.2604,
+      "step": 43360
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.092334023961806e-05,
+      "loss": 0.2579,
+      "step": 43370
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.091883614088821e-05,
+      "loss": 0.2601,
+      "step": 43380
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0914332042158365e-05,
+      "loss": 0.2523,
+      "step": 43390
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.090982794342852e-05,
+      "loss": 0.2578,
+      "step": 43400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0905323844698673e-05,
+      "loss": 0.2561,
+      "step": 43410
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.090081974596883e-05,
+      "loss": 0.2614,
+      "step": 43420
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.089631564723899e-05,
+      "loss": 0.2576,
+      "step": 43430
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0891811548509146e-05,
+      "loss": 0.2555,
+      "step": 43440
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0887307449779303e-05,
+      "loss": 0.2572,
+      "step": 43450
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0882803351049454e-05,
+      "loss": 0.2557,
+      "step": 43460
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.087829925231961e-05,
+      "loss": 0.263,
+      "step": 43470
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.087379515358977e-05,
+      "loss": 0.2595,
+      "step": 43480
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.086929105485992e-05,
+      "loss": 0.2602,
+      "step": 43490
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0864786956130084e-05,
+      "loss": 0.2545,
+      "step": 43500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0860282857400235e-05,
+      "loss": 0.2566,
+      "step": 43510
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.085577875867039e-05,
+      "loss": 0.2547,
+      "step": 43520
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.085127465994055e-05,
+      "loss": 0.2606,
+      "step": 43530
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.08467705612107e-05,
+      "loss": 0.2561,
+      "step": 43540
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.084226646248086e-05,
+      "loss": 0.2509,
+      "step": 43550
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0837762363751016e-05,
+      "loss": 0.2546,
+      "step": 43560
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.083370867489416e-05,
+      "loss": 0.2597,
+      "step": 43570
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.082920457616431e-05,
+      "loss": 0.2612,
+      "step": 43580
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.082470047743447e-05,
+      "loss": 0.2631,
+      "step": 43590
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0820196378704625e-05,
+      "loss": 0.2634,
+      "step": 43600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0815692279974776e-05,
+      "loss": 0.2649,
+      "step": 43610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.081118818124493e-05,
+      "loss": 0.2584,
+      "step": 43620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.080668408251509e-05,
+      "loss": 0.2629,
+      "step": 43630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.080217998378525e-05,
+      "loss": 0.2543,
+      "step": 43640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.07976758850554e-05,
+      "loss": 0.2574,
+      "step": 43650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0793171786325556e-05,
+      "loss": 0.2637,
+      "step": 43660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0788667687595714e-05,
+      "loss": 0.2619,
+      "step": 43670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.078416358886587e-05,
+      "loss": 0.2583,
+      "step": 43680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.077965949013603e-05,
+      "loss": 0.262,
+      "step": 43690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.077515539140618e-05,
+      "loss": 0.2643,
+      "step": 43700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.077065129267634e-05,
+      "loss": 0.2676,
+      "step": 43710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0766147193946495e-05,
+      "loss": 0.2583,
+      "step": 43720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0761643095216645e-05,
+      "loss": 0.2565,
+      "step": 43730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.075713899648681e-05,
+      "loss": 0.2585,
+      "step": 43740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.075263489775696e-05,
+      "loss": 0.2579,
+      "step": 43750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.074813079902711e-05,
+      "loss": 0.2608,
+      "step": 43760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0743626700297275e-05,
+      "loss": 0.2571,
+      "step": 43770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0739122601567426e-05,
+      "loss": 0.2562,
+      "step": 43780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0734618502837583e-05,
+      "loss": 0.2602,
+      "step": 43790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.073011440410774e-05,
+      "loss": 0.2574,
+      "step": 43800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.072561030537789e-05,
+      "loss": 0.2554,
+      "step": 43810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0721106206648056e-05,
+      "loss": 0.2576,
+      "step": 43820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.071660210791821e-05,
+      "loss": 0.2591,
+      "step": 43830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0712098009188364e-05,
+      "loss": 0.2612,
+      "step": 43840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.070759391045852e-05,
+      "loss": 0.253,
+      "step": 43850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.070308981172867e-05,
+      "loss": 0.255,
+      "step": 43860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.069858571299883e-05,
+      "loss": 0.2518,
+      "step": 43870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.069408161426899e-05,
+      "loss": 0.2543,
+      "step": 43880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0689577515539145e-05,
+      "loss": 0.256,
+      "step": 43890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0685073416809296e-05,
+      "loss": 0.2621,
+      "step": 43900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.068056931807945e-05,
+      "loss": 0.2582,
+      "step": 43910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.067606521934961e-05,
+      "loss": 0.2612,
+      "step": 43920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.067156112061976e-05,
+      "loss": 0.2591,
+      "step": 43930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0667057021889926e-05,
+      "loss": 0.2536,
+      "step": 43940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0662552923160076e-05,
+      "loss": 0.2565,
+      "step": 43950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0658048824430234e-05,
+      "loss": 0.2587,
+      "step": 43960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.065354472570039e-05,
+      "loss": 0.2561,
+      "step": 43970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.064904062697054e-05,
+      "loss": 0.2625,
+      "step": 43980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.06445365282407e-05,
+      "loss": 0.2522,
+      "step": 43990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.064003242951086e-05,
+      "loss": 0.26,
+      "step": 44000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.063552833078101e-05,
+      "loss": 0.2627,
+      "step": 44010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.063102423205117e-05,
+      "loss": 0.2606,
+      "step": 44020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.062652013332132e-05,
+      "loss": 0.258,
+      "step": 44030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.062201603459148e-05,
+      "loss": 0.2553,
+      "step": 44040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.061751193586164e-05,
+      "loss": 0.2605,
+      "step": 44050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.061300783713179e-05,
+      "loss": 0.2658,
+      "step": 44060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0608503738401946e-05,
+      "loss": 0.2584,
+      "step": 44070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.06039996396721e-05,
+      "loss": 0.2598,
+      "step": 44080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.059949554094226e-05,
+      "loss": 0.2625,
+      "step": 44090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.059499144221242e-05,
+      "loss": 0.2563,
+      "step": 44100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.059048734348257e-05,
+      "loss": 0.2572,
+      "step": 44110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0585983244752727e-05,
+      "loss": 0.2531,
+      "step": 44120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0581479146022884e-05,
+      "loss": 0.2653,
+      "step": 44130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.057697504729304e-05,
+      "loss": 0.2587,
+      "step": 44140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.057247094856319e-05,
+      "loss": 0.2617,
+      "step": 44150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.056796684983335e-05,
+      "loss": 0.2572,
+      "step": 44160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.056346275110351e-05,
+      "loss": 0.2614,
+      "step": 44170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.055895865237366e-05,
+      "loss": 0.2562,
+      "step": 44180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0554454553643815e-05,
+      "loss": 0.2544,
+      "step": 44190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.054995045491397e-05,
+      "loss": 0.254,
+      "step": 44200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0545446356184124e-05,
+      "loss": 0.2656,
+      "step": 44210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.054094225745429e-05,
+      "loss": 0.2622,
+      "step": 44220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.053643815872444e-05,
+      "loss": 0.2594,
+      "step": 44230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0531934059994596e-05,
+      "loss": 0.2554,
+      "step": 44240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0527429961264754e-05,
+      "loss": 0.2581,
+      "step": 44250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0522925862534904e-05,
+      "loss": 0.2599,
+      "step": 44260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.051842176380507e-05,
+      "loss": 0.2598,
+      "step": 44270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.051391766507522e-05,
+      "loss": 0.2625,
+      "step": 44280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0509413566345373e-05,
+      "loss": 0.2549,
+      "step": 44290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0504909467615534e-05,
+      "loss": 0.2505,
+      "step": 44300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.050040536888569e-05,
+      "loss": 0.2536,
+      "step": 44310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0495901270155842e-05,
+      "loss": 0.2565,
+      "step": 44320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0491397171426e-05,
+      "loss": 0.2578,
+      "step": 44330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0486893072696154e-05,
+      "loss": 0.2571,
+      "step": 44340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0482388973966308e-05,
+      "loss": 0.254,
+      "step": 44350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.047788487523647e-05,
+      "loss": 0.2522,
+      "step": 44360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0473380776506623e-05,
+      "loss": 0.2538,
+      "step": 44370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.046887667777678e-05,
+      "loss": 0.2551,
+      "step": 44380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0464372579046935e-05,
+      "loss": 0.2552,
+      "step": 44390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.045986848031709e-05,
+      "loss": 0.2596,
+      "step": 44400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0455364381587246e-05,
+      "loss": 0.2589,
+      "step": 44410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.04508602828574e-05,
+      "loss": 0.257,
+      "step": 44420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0446356184127555e-05,
+      "loss": 0.2592,
+      "step": 44430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0441852085397715e-05,
+      "loss": 0.2623,
+      "step": 44440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.043734798666787e-05,
+      "loss": 0.2566,
+      "step": 44450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0432843887938024e-05,
+      "loss": 0.2525,
+      "step": 44460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.042833978920818e-05,
+      "loss": 0.2519,
+      "step": 44470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0423835690478335e-05,
+      "loss": 0.2544,
+      "step": 44480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.041933159174849e-05,
+      "loss": 0.2559,
+      "step": 44490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.041482749301865e-05,
+      "loss": 0.2532,
+      "step": 44500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0410323394288804e-05,
+      "loss": 0.2507,
+      "step": 44510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0405819295558962e-05,
+      "loss": 0.2573,
+      "step": 44520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0401315196829116e-05,
+      "loss": 0.259,
+      "step": 44530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.039681109809927e-05,
+      "loss": 0.2537,
+      "step": 44540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.039230699936943e-05,
+      "loss": 0.2567,
+      "step": 44550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0387802900639585e-05,
+      "loss": 0.2591,
+      "step": 44560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.038329880190974e-05,
+      "loss": 0.2497,
+      "step": 44570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0378794703179897e-05,
+      "loss": 0.2546,
+      "step": 44580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.037429060445005e-05,
+      "loss": 0.254,
+      "step": 44590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0369786505720205e-05,
+      "loss": 0.2525,
+      "step": 44600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0365282406990366e-05,
+      "loss": 0.2509,
+      "step": 44610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0360778308260516e-05,
+      "loss": 0.2584,
+      "step": 44620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.035627420953067e-05,
+      "loss": 0.2555,
+      "step": 44630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.035177011080083e-05,
+      "loss": 0.2611,
+      "step": 44640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0347266012070986e-05,
+      "loss": 0.2597,
+      "step": 44650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0342761913341143e-05,
+      "loss": 0.2606,
+      "step": 44660
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0338257814611297e-05,
+      "loss": 0.2568,
+      "step": 44670
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.033375371588145e-05,
+      "loss": 0.2538,
+      "step": 44680
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0329249617151612e-05,
+      "loss": 0.2523,
+      "step": 44690
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0324745518421766e-05,
+      "loss": 0.2632,
+      "step": 44700
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.032024141969192e-05,
+      "loss": 0.2506,
+      "step": 44710
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0315737320962078e-05,
+      "loss": 0.2543,
+      "step": 44720
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0311233222232232e-05,
+      "loss": 0.2617,
+      "step": 44730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0306729123502386e-05,
+      "loss": 0.2553,
+      "step": 44740
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0302225024772547e-05,
+      "loss": 0.2573,
+      "step": 44750
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.02977209260427e-05,
+      "loss": 0.2592,
+      "step": 44760
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0293216827312855e-05,
+      "loss": 0.2573,
+      "step": 44770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0288712728583013e-05,
+      "loss": 0.2604,
+      "step": 44780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0284208629853167e-05,
+      "loss": 0.2627,
+      "step": 44790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0279704531123328e-05,
+      "loss": 0.2511,
+      "step": 44800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.027520043239348e-05,
+      "loss": 0.2567,
+      "step": 44810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0270696333663632e-05,
+      "loss": 0.2532,
+      "step": 44820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0266192234933793e-05,
+      "loss": 0.2601,
+      "step": 44830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0261688136203947e-05,
+      "loss": 0.2556,
+      "step": 44840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.02571840374741e-05,
+      "loss": 0.255,
+      "step": 44850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.025267993874426e-05,
+      "loss": 0.2606,
+      "step": 44860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0248175840014413e-05,
+      "loss": 0.2553,
+      "step": 44870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0243671741284567e-05,
+      "loss": 0.264,
+      "step": 44880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0239167642554728e-05,
+      "loss": 0.2501,
+      "step": 44890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0234663543824882e-05,
+      "loss": 0.2547,
+      "step": 44900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0230159445095036e-05,
+      "loss": 0.257,
+      "step": 44910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0225655346365194e-05,
+      "loss": 0.2523,
+      "step": 44920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0221151247635348e-05,
+      "loss": 0.2581,
+      "step": 44930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.021664714890551e-05,
+      "loss": 0.2573,
+      "step": 44940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0212143050175663e-05,
+      "loss": 0.2548,
+      "step": 44950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0207638951445817e-05,
+      "loss": 0.2515,
+      "step": 44960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0203134852715974e-05,
+      "loss": 0.2509,
+      "step": 44970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.019863075398613e-05,
+      "loss": 0.2632,
+      "step": 44980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0194126655256283e-05,
+      "loss": 0.2587,
+      "step": 44990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0189622556526444e-05,
+      "loss": 0.2616,
+      "step": 45000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0185118457796598e-05,
+      "loss": 0.2576,
+      "step": 45010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0180614359066752e-05,
+      "loss": 0.2537,
+      "step": 45020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.017611026033691e-05,
+      "loss": 0.2533,
+      "step": 45030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0171606161607063e-05,
+      "loss": 0.2632,
+      "step": 45040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0167102062877217e-05,
+      "loss": 0.2595,
+      "step": 45050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0162597964147375e-05,
+      "loss": 0.2484,
+      "step": 45060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.015809386541753e-05,
+      "loss": 0.2612,
+      "step": 45070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.015358976668769e-05,
+      "loss": 0.2493,
+      "step": 45080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0149085667957844e-05,
+      "loss": 0.2552,
+      "step": 45090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0144581569227998e-05,
+      "loss": 0.2526,
+      "step": 45100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0140077470498156e-05,
+      "loss": 0.2511,
+      "step": 45110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.013557337176831e-05,
+      "loss": 0.2514,
+      "step": 45120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0131069273038464e-05,
+      "loss": 0.2566,
+      "step": 45130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0126565174308625e-05,
+      "loss": 0.2556,
+      "step": 45140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.012206107557878e-05,
+      "loss": 0.255,
+      "step": 45150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0117556976848933e-05,
+      "loss": 0.2571,
+      "step": 45160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.011305287811909e-05,
+      "loss": 0.2536,
+      "step": 45170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0108548779389245e-05,
+      "loss": 0.2529,
+      "step": 45180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.01040446806594e-05,
+      "loss": 0.2609,
+      "step": 45190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.009954058192956e-05,
+      "loss": 0.2547,
+      "step": 45200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0095036483199714e-05,
+      "loss": 0.2496,
+      "step": 45210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.009053238446987e-05,
+      "loss": 0.2565,
+      "step": 45220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0086028285740025e-05,
+      "loss": 0.2575,
+      "step": 45230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.008152418701018e-05,
+      "loss": 0.2534,
+      "step": 45240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.007702008828034e-05,
+      "loss": 0.2531,
+      "step": 45250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0072515989550494e-05,
+      "loss": 0.2563,
+      "step": 45260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0068011890820645e-05,
+      "loss": 0.2531,
+      "step": 45270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0063507792090806e-05,
+      "loss": 0.2561,
+      "step": 45280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.005900369336096e-05,
+      "loss": 0.2567,
+      "step": 45290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0054499594631114e-05,
+      "loss": 0.2661,
+      "step": 45300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.004999549590127e-05,
+      "loss": 0.2506,
+      "step": 45310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0045491397171426e-05,
+      "loss": 0.2552,
+      "step": 45320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.004098729844158e-05,
+      "loss": 0.2486,
+      "step": 45330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.003648319971174e-05,
+      "loss": 0.2516,
+      "step": 45340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0031979100981895e-05,
+      "loss": 0.2555,
+      "step": 45350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.0027475002252052e-05,
+      "loss": 0.2571,
+      "step": 45360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.0022970903522206e-05,
+      "loss": 0.2594,
+      "step": 45370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.001846680479236e-05,
+      "loss": 0.2599,
+      "step": 45380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.001396270606252e-05,
+      "loss": 0.2553,
+      "step": 45390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.0009458607332675e-05,
+      "loss": 0.2541,
+      "step": 45400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.000495450860283e-05,
+      "loss": 0.2511,
+      "step": 45410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.0000450409872987e-05,
+      "loss": 0.2586,
+      "step": 45420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.999594631114314e-05,
+      "loss": 0.2558,
+      "step": 45430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9991442212413295e-05,
+      "loss": 0.2526,
+      "step": 45440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9986938113683456e-05,
+      "loss": 0.2617,
+      "step": 45450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.998243401495361e-05,
+      "loss": 0.2539,
+      "step": 45460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.997792991622376e-05,
+      "loss": 0.2532,
+      "step": 45470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9973425817493922e-05,
+      "loss": 0.255,
+      "step": 45480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9968921718764076e-05,
+      "loss": 0.2582,
+      "step": 45490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9964417620034237e-05,
+      "loss": 0.2543,
+      "step": 45500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9959913521304388e-05,
+      "loss": 0.2515,
+      "step": 45510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.995540942257454e-05,
+      "loss": 0.2508,
+      "step": 45520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9950905323844703e-05,
+      "loss": 0.252,
+      "step": 45530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9946401225114857e-05,
+      "loss": 0.2536,
+      "step": 45540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.994189712638501e-05,
+      "loss": 0.2521,
+      "step": 45550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9937393027655168e-05,
+      "loss": 0.2506,
+      "step": 45560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9932888928925322e-05,
+      "loss": 0.2484,
+      "step": 45570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9928384830195476e-05,
+      "loss": 0.254,
+      "step": 45580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9923880731465637e-05,
+      "loss": 0.2548,
+      "step": 45590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.991937663273579e-05,
+      "loss": 0.2582,
+      "step": 45600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9915322943878932e-05,
+      "loss": 0.2542,
+      "step": 45610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9910818845149086e-05,
+      "loss": 0.2524,
+      "step": 45620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9906314746419243e-05,
+      "loss": 0.252,
+      "step": 45630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9901810647689397e-05,
+      "loss": 0.2519,
+      "step": 45640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.989730654895955e-05,
+      "loss": 0.2556,
+      "step": 45650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9892802450229712e-05,
+      "loss": 0.2489,
+      "step": 45660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9888298351499867e-05,
+      "loss": 0.2494,
+      "step": 45670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.988379425277002e-05,
+      "loss": 0.2535,
+      "step": 45680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9879290154040178e-05,
+      "loss": 0.2529,
+      "step": 45690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9874786055310332e-05,
+      "loss": 0.2571,
+      "step": 45700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9870281956580486e-05,
+      "loss": 0.2575,
+      "step": 45710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9865777857850647e-05,
+      "loss": 0.2502,
+      "step": 45720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.98612737591208e-05,
+      "loss": 0.2565,
+      "step": 45730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.985676966039096e-05,
+      "loss": 0.253,
+      "step": 45740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9852265561661113e-05,
+      "loss": 0.2536,
+      "step": 45750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9847761462931267e-05,
+      "loss": 0.2548,
+      "step": 45760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9843257364201428e-05,
+      "loss": 0.2542,
+      "step": 45770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9838753265471582e-05,
+      "loss": 0.2489,
+      "step": 45780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9834249166741736e-05,
+      "loss": 0.2553,
+      "step": 45790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9829745068011894e-05,
+      "loss": 0.2492,
+      "step": 45800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9825240969282048e-05,
+      "loss": 0.2592,
+      "step": 45810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9820736870552202e-05,
+      "loss": 0.2556,
+      "step": 45820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9816232771822363e-05,
+      "loss": 0.2505,
+      "step": 45830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9811728673092513e-05,
+      "loss": 0.2536,
+      "step": 45840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9807224574362668e-05,
+      "loss": 0.2536,
+      "step": 45850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.980272047563283e-05,
+      "loss": 0.2498,
+      "step": 45860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9798216376902983e-05,
+      "loss": 0.2553,
+      "step": 45870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.979371227817314e-05,
+      "loss": 0.2507,
+      "step": 45880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9789208179443294e-05,
+      "loss": 0.258,
+      "step": 45890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9784704080713448e-05,
+      "loss": 0.2508,
+      "step": 45900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.978019998198361e-05,
+      "loss": 0.2564,
+      "step": 45910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9775695883253763e-05,
+      "loss": 0.2463,
+      "step": 45920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9771191784523917e-05,
+      "loss": 0.2572,
+      "step": 45930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9766687685794075e-05,
+      "loss": 0.2565,
+      "step": 45940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.976218358706423e-05,
+      "loss": 0.2526,
+      "step": 45950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9757679488334383e-05,
+      "loss": 0.2531,
+      "step": 45960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9753175389604544e-05,
+      "loss": 0.2537,
+      "step": 45970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9748671290874698e-05,
+      "loss": 0.2552,
+      "step": 45980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9744167192144852e-05,
+      "loss": 0.2521,
+      "step": 45990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.973966309341501e-05,
+      "loss": 0.2578,
+      "step": 46000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9735158994685164e-05,
+      "loss": 0.2612,
+      "step": 46010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9730654895955325e-05,
+      "loss": 0.2527,
+      "step": 46020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.972615079722548e-05,
+      "loss": 0.2494,
+      "step": 46030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.972164669849563e-05,
+      "loss": 0.2568,
+      "step": 46040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.971714259976579e-05,
+      "loss": 0.2589,
+      "step": 46050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9712638501035944e-05,
+      "loss": 0.2602,
+      "step": 46060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.97081344023061e-05,
+      "loss": 0.2554,
+      "step": 46070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9703630303576256e-05,
+      "loss": 0.2575,
+      "step": 46080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.969912620484641e-05,
+      "loss": 0.2556,
+      "step": 46090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9694622106116564e-05,
+      "loss": 0.2567,
+      "step": 46100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9690118007386725e-05,
+      "loss": 0.2507,
+      "step": 46110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.968561390865688e-05,
+      "loss": 0.2607,
+      "step": 46120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9681109809927033e-05,
+      "loss": 0.2498,
+      "step": 46130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.967660571119719e-05,
+      "loss": 0.2473,
+      "step": 46140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9672101612467345e-05,
+      "loss": 0.2465,
+      "step": 46150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9667597513737506e-05,
+      "loss": 0.2503,
+      "step": 46160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.966309341500766e-05,
+      "loss": 0.2539,
+      "step": 46170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9658589316277814e-05,
+      "loss": 0.2529,
+      "step": 46180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.965408521754797e-05,
+      "loss": 0.2534,
+      "step": 46190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9649581118818126e-05,
+      "loss": 0.2585,
+      "step": 46200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.964507702008828e-05,
+      "loss": 0.2541,
+      "step": 46210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.964057292135844e-05,
+      "loss": 0.2567,
+      "step": 46220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9636068822628595e-05,
+      "loss": 0.2486,
+      "step": 46230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9631564723898745e-05,
+      "loss": 0.2463,
+      "step": 46240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9627060625168906e-05,
+      "loss": 0.2487,
+      "step": 46250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.962255652643906e-05,
+      "loss": 0.2512,
+      "step": 46260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9618052427709214e-05,
+      "loss": 0.254,
+      "step": 46270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9613548328979372e-05,
+      "loss": 0.2546,
+      "step": 46280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9609044230249526e-05,
+      "loss": 0.2578,
+      "step": 46290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9604540131519687e-05,
+      "loss": 0.2542,
+      "step": 46300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.960003603278984e-05,
+      "loss": 0.2542,
+      "step": 46310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9595531934059995e-05,
+      "loss": 0.2523,
+      "step": 46320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9591027835330153e-05,
+      "loss": 0.2542,
+      "step": 46330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9586523736600307e-05,
+      "loss": 0.2542,
+      "step": 46340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.958201963787046e-05,
+      "loss": 0.2537,
+      "step": 46350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9577515539140622e-05,
+      "loss": 0.2513,
+      "step": 46360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9573011440410776e-05,
+      "loss": 0.2478,
+      "step": 46370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.956850734168093e-05,
+      "loss": 0.25,
+      "step": 46380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9564003242951087e-05,
+      "loss": 0.2489,
+      "step": 46390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.955949914422124e-05,
+      "loss": 0.2513,
+      "step": 46400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9554995045491396e-05,
+      "loss": 0.2559,
+      "step": 46410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9550490946761557e-05,
+      "loss": 0.262,
+      "step": 46420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.954598684803171e-05,
+      "loss": 0.2477,
+      "step": 46430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9541482749301868e-05,
+      "loss": 0.2522,
+      "step": 46440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9536978650572022e-05,
+      "loss": 0.2456,
+      "step": 46450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9532474551842176e-05,
+      "loss": 0.2481,
+      "step": 46460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9527970453112337e-05,
+      "loss": 0.2543,
+      "step": 46470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.952346635438249e-05,
+      "loss": 0.2541,
+      "step": 46480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9518962255652642e-05,
+      "loss": 0.2558,
+      "step": 46490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9514458156922803e-05,
+      "loss": 0.2555,
+      "step": 46500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9509954058192957e-05,
+      "loss": 0.253,
+      "step": 46510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.950544995946311e-05,
+      "loss": 0.2526,
+      "step": 46520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.950094586073327e-05,
+      "loss": 0.252,
+      "step": 46530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9496441762003423e-05,
+      "loss": 0.2506,
+      "step": 46540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9491937663273577e-05,
+      "loss": 0.2536,
+      "step": 46550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9487433564543738e-05,
+      "loss": 0.2519,
+      "step": 46560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9482929465813892e-05,
+      "loss": 0.2493,
+      "step": 46570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.947842536708405e-05,
+      "loss": 0.2468,
+      "step": 46580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9473921268354203e-05,
+      "loss": 0.2475,
+      "step": 46590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9469417169624358e-05,
+      "loss": 0.2454,
+      "step": 46600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.946491307089452e-05,
+      "loss": 0.2529,
+      "step": 46610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9460408972164672e-05,
+      "loss": 0.2529,
+      "step": 46620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9455904873434827e-05,
+      "loss": 0.2481,
+      "step": 46630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9451400774704984e-05,
+      "loss": 0.2544,
+      "step": 46640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9446896675975138e-05,
+      "loss": 0.2512,
+      "step": 46650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9442392577245292e-05,
+      "loss": 0.2525,
+      "step": 46660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9437888478515453e-05,
+      "loss": 0.2492,
+      "step": 46670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9433384379785607e-05,
+      "loss": 0.2539,
+      "step": 46680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9428880281055758e-05,
+      "loss": 0.2563,
+      "step": 46690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.942437618232592e-05,
+      "loss": 0.2514,
+      "step": 46700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9419872083596073e-05,
+      "loss": 0.2513,
+      "step": 46710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9415367984866234e-05,
+      "loss": 0.2572,
+      "step": 46720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9410863886136385e-05,
+      "loss": 0.2525,
+      "step": 46730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.940635978740654e-05,
+      "loss": 0.2583,
+      "step": 46740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.94018556886767e-05,
+      "loss": 0.2579,
+      "step": 46750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9397351589946854e-05,
+      "loss": 0.2487,
+      "step": 46760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9392847491217008e-05,
+      "loss": 0.2529,
+      "step": 46770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9388343392487165e-05,
+      "loss": 0.2504,
+      "step": 46780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.938383929375732e-05,
+      "loss": 0.2533,
+      "step": 46790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9379335195027473e-05,
+      "loss": 0.2529,
+      "step": 46800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9374831096297634e-05,
+      "loss": 0.2496,
+      "step": 46810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.937032699756779e-05,
+      "loss": 0.2509,
+      "step": 46820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9365822898837943e-05,
+      "loss": 0.2536,
+      "step": 46830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.93613188001081e-05,
+      "loss": 0.251,
+      "step": 46840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9356814701378254e-05,
+      "loss": 0.2547,
+      "step": 46850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9352310602648415e-05,
+      "loss": 0.2521,
+      "step": 46860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.934780650391857e-05,
+      "loss": 0.2502,
+      "step": 46870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9343302405188723e-05,
+      "loss": 0.257,
+      "step": 46880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.933879830645888e-05,
+      "loss": 0.2548,
+      "step": 46890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9334294207729035e-05,
+      "loss": 0.2563,
+      "step": 46900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.932979010899919e-05,
+      "loss": 0.2508,
+      "step": 46910
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.932528601026935e-05,
+      "loss": 0.254,
+      "step": 46920
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.93207819115395e-05,
+      "loss": 0.2491,
+      "step": 46930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9316277812809655e-05,
+      "loss": 0.2503,
+      "step": 46940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9311773714079816e-05,
+      "loss": 0.2557,
+      "step": 46950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.930726961534997e-05,
+      "loss": 0.2509,
+      "step": 46960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9302765516620124e-05,
+      "loss": 0.2493,
+      "step": 46970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.929826141789028e-05,
+      "loss": 0.2531,
+      "step": 46980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9293757319160435e-05,
+      "loss": 0.2554,
+      "step": 46990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9289253220430596e-05,
+      "loss": 0.2488,
+      "step": 47000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.928474912170075e-05,
+      "loss": 0.2502,
+      "step": 47010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9280245022970904e-05,
+      "loss": 0.2508,
+      "step": 47020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9275740924241062e-05,
+      "loss": 0.2499,
+      "step": 47030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9271236825511216e-05,
+      "loss": 0.2501,
+      "step": 47040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.926673272678137e-05,
+      "loss": 0.2553,
+      "step": 47050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.926222862805153e-05,
+      "loss": 0.2494,
+      "step": 47060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9257724529321685e-05,
+      "loss": 0.2568,
+      "step": 47070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.925322043059184e-05,
+      "loss": 0.2468,
+      "step": 47080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9248716331861997e-05,
+      "loss": 0.2499,
+      "step": 47090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.924421223313215e-05,
+      "loss": 0.2491,
+      "step": 47100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9239708134402305e-05,
+      "loss": 0.2491,
+      "step": 47110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9235204035672466e-05,
+      "loss": 0.248,
+      "step": 47120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.923069993694262e-05,
+      "loss": 0.2478,
+      "step": 47130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9226195838212777e-05,
+      "loss": 0.2502,
+      "step": 47140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.922169173948293e-05,
+      "loss": 0.2556,
+      "step": 47150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9217187640753086e-05,
+      "loss": 0.2476,
+      "step": 47160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9212683542023243e-05,
+      "loss": 0.2501,
+      "step": 47170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9208179443293397e-05,
+      "loss": 0.2494,
+      "step": 47180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.920367534456355e-05,
+      "loss": 0.248,
+      "step": 47190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9199171245833712e-05,
+      "loss": 0.2525,
+      "step": 47200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9194667147103866e-05,
+      "loss": 0.2488,
+      "step": 47210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.919016304837402e-05,
+      "loss": 0.2484,
+      "step": 47220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9185658949644178e-05,
+      "loss": 0.2565,
+      "step": 47230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9181154850914332e-05,
+      "loss": 0.2542,
+      "step": 47240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9176650752184486e-05,
+      "loss": 0.2491,
+      "step": 47250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9172146653454647e-05,
+      "loss": 0.2515,
+      "step": 47260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.91676425547248e-05,
+      "loss": 0.2523,
+      "step": 47270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.916313845599496e-05,
+      "loss": 0.2527,
+      "step": 47280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9158634357265113e-05,
+      "loss": 0.252,
+      "step": 47290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9154130258535267e-05,
+      "loss": 0.252,
+      "step": 47300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9149626159805428e-05,
+      "loss": 0.2462,
+      "step": 47310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9145122061075582e-05,
+      "loss": 0.2607,
+      "step": 47320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9140617962345736e-05,
+      "loss": 0.2541,
+      "step": 47330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9136113863615893e-05,
+      "loss": 0.2559,
+      "step": 47340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9131609764886047e-05,
+      "loss": 0.2553,
+      "step": 47350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.91271056661562e-05,
+      "loss": 0.2517,
+      "step": 47360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9122601567426362e-05,
+      "loss": 0.2515,
+      "step": 47370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9118097468696513e-05,
+      "loss": 0.2613,
+      "step": 47380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9113593369966667e-05,
+      "loss": 0.2461,
+      "step": 47390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9109089271236828e-05,
+      "loss": 0.2546,
+      "step": 47400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9104585172506982e-05,
+      "loss": 0.2547,
+      "step": 47410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.910008107377714e-05,
+      "loss": 0.2506,
+      "step": 47420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9095576975047294e-05,
+      "loss": 0.2543,
+      "step": 47430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9091072876317448e-05,
+      "loss": 0.2525,
+      "step": 47440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.908656877758761e-05,
+      "loss": 0.254,
+      "step": 47450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9082064678857763e-05,
+      "loss": 0.2473,
+      "step": 47460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9077560580127917e-05,
+      "loss": 0.2532,
+      "step": 47470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9073056481398075e-05,
+      "loss": 0.2443,
+      "step": 47480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.906855238266823e-05,
+      "loss": 0.256,
+      "step": 47490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9064048283938383e-05,
+      "loss": 0.2521,
+      "step": 47500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9059544185208544e-05,
+      "loss": 0.2485,
+      "step": 47510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9055040086478698e-05,
+      "loss": 0.248,
+      "step": 47520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9050535987748852e-05,
+      "loss": 0.2547,
+      "step": 47530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.904603188901901e-05,
+      "loss": 0.2477,
+      "step": 47540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9041527790289163e-05,
+      "loss": 0.2506,
+      "step": 47550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9037023691559324e-05,
+      "loss": 0.2541,
+      "step": 47560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.903251959282948e-05,
+      "loss": 0.2458,
+      "step": 47570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.902801549409963e-05,
+      "loss": 0.2484,
+      "step": 47580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.902351139536979e-05,
+      "loss": 0.25,
+      "step": 47590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9019007296639944e-05,
+      "loss": 0.2552,
+      "step": 47600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.9014503197910098e-05,
+      "loss": 0.256,
+      "step": 47610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.9009999099180256e-05,
+      "loss": 0.2498,
+      "step": 47620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.9005945410323393e-05,
+      "loss": 0.2575,
+      "step": 47630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.9001441311593554e-05,
+      "loss": 0.2449,
+      "step": 47640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8996937212863708e-05,
+      "loss": 0.2518,
+      "step": 47650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8992433114133865e-05,
+      "loss": 0.2512,
+      "step": 47660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.898792901540402e-05,
+      "loss": 0.2491,
+      "step": 47670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8983424916674173e-05,
+      "loss": 0.249,
+      "step": 47680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8978920817944334e-05,
+      "loss": 0.2455,
+      "step": 47690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8974416719214485e-05,
+      "loss": 0.2513,
+      "step": 47700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.896991262048464e-05,
+      "loss": 0.2474,
+      "step": 47710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.89654085217548e-05,
+      "loss": 0.2468,
+      "step": 47720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8960904423024954e-05,
+      "loss": 0.2453,
+      "step": 47730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8956400324295108e-05,
+      "loss": 0.2478,
+      "step": 47740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8951896225565266e-05,
+      "loss": 0.2472,
+      "step": 47750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.894739212683542e-05,
+      "loss": 0.2451,
+      "step": 47760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8942888028105574e-05,
+      "loss": 0.2423,
+      "step": 47770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8938383929375735e-05,
+      "loss": 0.2517,
+      "step": 47780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.893387983064589e-05,
+      "loss": 0.25,
+      "step": 47790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8929375731916046e-05,
+      "loss": 0.2523,
+      "step": 47800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.89248716331862e-05,
+      "loss": 0.2506,
+      "step": 47810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8920367534456355e-05,
+      "loss": 0.2461,
+      "step": 47820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8915863435726515e-05,
+      "loss": 0.2528,
+      "step": 47830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.891135933699667e-05,
+      "loss": 0.2516,
+      "step": 47840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8906855238266824e-05,
+      "loss": 0.2473,
+      "step": 47850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.890235113953698e-05,
+      "loss": 0.2515,
+      "step": 47860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8897847040807135e-05,
+      "loss": 0.2518,
+      "step": 47870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.889334294207729e-05,
+      "loss": 0.2498,
+      "step": 47880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.888883884334745e-05,
+      "loss": 0.2447,
+      "step": 47890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8884334744617604e-05,
+      "loss": 0.2456,
+      "step": 47900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8879830645887755e-05,
+      "loss": 0.2544,
+      "step": 47910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8875326547157916e-05,
+      "loss": 0.2526,
+      "step": 47920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.887082244842807e-05,
+      "loss": 0.2495,
+      "step": 47930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.886631834969823e-05,
+      "loss": 0.2486,
+      "step": 47940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.886181425096838e-05,
+      "loss": 0.2469,
+      "step": 47950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8857310152238536e-05,
+      "loss": 0.2448,
+      "step": 47960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8852806053508697e-05,
+      "loss": 0.2495,
+      "step": 47970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.884830195477885e-05,
+      "loss": 0.2495,
+      "step": 47980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8843797856049005e-05,
+      "loss": 0.2458,
+      "step": 47990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8839293757319162e-05,
+      "loss": 0.2575,
+      "step": 48000
+    },
+    {
+      "epoch": 0.43,
+      "eval_NEIMS_canon_smiles": 0.9293,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.024266666666666666,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.40677611496886035,
+      "eval_NEIMS_exact_mols": 0.024116666666666668,
+      "eval_NEIMS_exact_smiles": 0.0234,
+      "eval_NEIMS_loss": 0.26602691411972046,
+      "eval_NEIMS_matched_formulas": 0.46668333333333334,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.3061910596751587,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.024233333333333332,
+      "eval_NEIMS_runtime": 713.3854,
+      "eval_NEIMS_samples_per_second": 84.106,
+      "eval_NEIMS_steps_per_second": 1.315,
+      "step": 48000
+    },
+    {
+      "epoch": 0.43,
+      "eval_RASSP_canon_smiles": 0.9167362340134237,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.05900424082545831,
+      "eval_RASSP_daylight_tanimoto_simil": 0.5040080086306112,
+      "eval_RASSP_exact_mols": 0.05872040605068955,
+      "eval_RASSP_exact_smiles": 0.05633285470998765,
+      "eval_RASSP_loss": 0.20886576175689697,
+      "eval_RASSP_matched_formulas": 0.6356062376865796,
+      "eval_RASSP_morgan_tanimoto_simil": 0.36702623477912233,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.058870671519684774,
+      "eval_RASSP_runtime": 820.2508,
+      "eval_RASSP_samples_per_second": 73.019,
+      "eval_RASSP_steps_per_second": 1.141,
+      "step": 48000
+    },
+    {
+      "epoch": 0.43,
+      "eval_NIST_canon_smiles": 0.8894133513149022,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.00816268587855343,
+      "eval_NIST_daylight_tanimoto_simil": 0.25274513159009915,
+      "eval_NIST_exact_mols": 0.007772296553926961,
+      "eval_NIST_exact_smiles": 0.007097987720481243,
+      "eval_NIST_loss": 1.4815818071365356,
+      "eval_NIST_matched_formulas": 0.06746637328317422,
+      "eval_NIST_morgan_tanimoto_simil": 0.20769413146775717,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.008091706001348618,
+      "eval_NIST_runtime": 438.2907,
+      "eval_NIST_samples_per_second": 64.288,
+      "eval_NIST_steps_per_second": 1.006,
+      "step": 48000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8834789658589316e-05,
+      "loss": 0.2475,
+      "step": 48010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.883028555985947e-05,
+      "loss": 0.2552,
+      "step": 48020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.882578146112963e-05,
+      "loss": 0.2526,
+      "step": 48030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8821277362399785e-05,
+      "loss": 0.2513,
+      "step": 48040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.881677326366994e-05,
+      "loss": 0.2527,
+      "step": 48050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8812269164940097e-05,
+      "loss": 0.2494,
+      "step": 48060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.880776506621025e-05,
+      "loss": 0.2504,
+      "step": 48070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8803260967480412e-05,
+      "loss": 0.2498,
+      "step": 48080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8798756868750566e-05,
+      "loss": 0.2507,
+      "step": 48090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.879425277002072e-05,
+      "loss": 0.2455,
+      "step": 48100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8789748671290878e-05,
+      "loss": 0.2529,
+      "step": 48110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8785244572561032e-05,
+      "loss": 0.2489,
+      "step": 48120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8780740473831186e-05,
+      "loss": 0.2521,
+      "step": 48130
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8776236375101347e-05,
+      "loss": 0.2539,
+      "step": 48140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8771732276371498e-05,
+      "loss": 0.2568,
+      "step": 48150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.876722817764165e-05,
+      "loss": 0.2528,
+      "step": 48160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8762724078911813e-05,
+      "loss": 0.2439,
+      "step": 48170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8758219980181967e-05,
+      "loss": 0.256,
+      "step": 48180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.875371588145212e-05,
+      "loss": 0.2497,
+      "step": 48190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8749211782722278e-05,
+      "loss": 0.2548,
+      "step": 48200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8744707683992432e-05,
+      "loss": 0.2486,
+      "step": 48210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8740203585262593e-05,
+      "loss": 0.2548,
+      "step": 48220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8735699486532747e-05,
+      "loss": 0.2475,
+      "step": 48230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.87311953878029e-05,
+      "loss": 0.247,
+      "step": 48240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872669128907306e-05,
+      "loss": 0.2508,
+      "step": 48250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8722187190343213e-05,
+      "loss": 0.2474,
+      "step": 48260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8717683091613367e-05,
+      "loss": 0.2462,
+      "step": 48270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8713178992883528e-05,
+      "loss": 0.2496,
+      "step": 48280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8708674894153682e-05,
+      "loss": 0.2534,
+      "step": 48290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8704170795423836e-05,
+      "loss": 0.2506,
+      "step": 48300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8699666696693994e-05,
+      "loss": 0.2445,
+      "step": 48310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8695162597964148e-05,
+      "loss": 0.2502,
+      "step": 48320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8690658499234302e-05,
+      "loss": 0.2455,
+      "step": 48330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8686154400504463e-05,
+      "loss": 0.248,
+      "step": 48340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8681650301774614e-05,
+      "loss": 0.249,
+      "step": 48350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8677146203044774e-05,
+      "loss": 0.2538,
+      "step": 48360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.867264210431493e-05,
+      "loss": 0.2475,
+      "step": 48370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8668138005585083e-05,
+      "loss": 0.2508,
+      "step": 48380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.866363390685524e-05,
+      "loss": 0.2493,
+      "step": 48390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8659129808125394e-05,
+      "loss": 0.2525,
+      "step": 48400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.865462570939555e-05,
+      "loss": 0.2498,
+      "step": 48410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.865012161066571e-05,
+      "loss": 0.2452,
+      "step": 48420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8645617511935863e-05,
+      "loss": 0.2461,
+      "step": 48430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8641113413206017e-05,
+      "loss": 0.2437,
+      "step": 48440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8636609314476175e-05,
+      "loss": 0.2502,
+      "step": 48450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.863210521574633e-05,
+      "loss": 0.2464,
+      "step": 48460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8627601117016483e-05,
+      "loss": 0.246,
+      "step": 48470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8623097018286644e-05,
+      "loss": 0.2472,
+      "step": 48480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8618592919556798e-05,
+      "loss": 0.2474,
+      "step": 48490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8614088820826956e-05,
+      "loss": 0.2457,
+      "step": 48500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.860958472209711e-05,
+      "loss": 0.2485,
+      "step": 48510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8605080623367264e-05,
+      "loss": 0.2418,
+      "step": 48520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8600576524637425e-05,
+      "loss": 0.2475,
+      "step": 48530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.859607242590758e-05,
+      "loss": 0.249,
+      "step": 48540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8591568327177733e-05,
+      "loss": 0.248,
+      "step": 48550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.858706422844789e-05,
+      "loss": 0.2474,
+      "step": 48560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8582560129718044e-05,
+      "loss": 0.2494,
+      "step": 48570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.85780560309882e-05,
+      "loss": 0.2411,
+      "step": 48580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.857355193225836e-05,
+      "loss": 0.2423,
+      "step": 48590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.856904783352851e-05,
+      "loss": 0.2437,
+      "step": 48600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8564543734798664e-05,
+      "loss": 0.2527,
+      "step": 48610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8560039636068825e-05,
+      "loss": 0.2508,
+      "step": 48620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.855553553733898e-05,
+      "loss": 0.2513,
+      "step": 48630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8551031438609137e-05,
+      "loss": 0.2424,
+      "step": 48640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.854652733987929e-05,
+      "loss": 0.2484,
+      "step": 48650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8542023241149445e-05,
+      "loss": 0.2478,
+      "step": 48660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8537519142419606e-05,
+      "loss": 0.2483,
+      "step": 48670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.853301504368976e-05,
+      "loss": 0.2492,
+      "step": 48680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8528510944959914e-05,
+      "loss": 0.2424,
+      "step": 48690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.852400684623007e-05,
+      "loss": 0.2476,
+      "step": 48700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8519502747500226e-05,
+      "loss": 0.2444,
+      "step": 48710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.851499864877038e-05,
+      "loss": 0.2445,
+      "step": 48720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.851049455004054e-05,
+      "loss": 0.2446,
+      "step": 48730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8505990451310695e-05,
+      "loss": 0.2518,
+      "step": 48740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.850148635258085e-05,
+      "loss": 0.2454,
+      "step": 48750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8496982253851006e-05,
+      "loss": 0.2485,
+      "step": 48760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.849247815512116e-05,
+      "loss": 0.2465,
+      "step": 48770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.848797405639132e-05,
+      "loss": 0.2497,
+      "step": 48780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8483469957661475e-05,
+      "loss": 0.2498,
+      "step": 48790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8478965858931626e-05,
+      "loss": 0.2449,
+      "step": 48800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8474461760201787e-05,
+      "loss": 0.243,
+      "step": 48810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.846995766147194e-05,
+      "loss": 0.2443,
+      "step": 48820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8465453562742095e-05,
+      "loss": 0.2382,
+      "step": 48830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8460949464012253e-05,
+      "loss": 0.2441,
+      "step": 48840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8456445365282407e-05,
+      "loss": 0.248,
+      "step": 48850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.845194126655256e-05,
+      "loss": 0.2521,
+      "step": 48860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8447437167822722e-05,
+      "loss": 0.2486,
+      "step": 48870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8442933069092876e-05,
+      "loss": 0.2523,
+      "step": 48880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.843842897036303e-05,
+      "loss": 0.2494,
+      "step": 48890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8433924871633188e-05,
+      "loss": 0.2528,
+      "step": 48900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.842942077290334e-05,
+      "loss": 0.2474,
+      "step": 48910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8424916674173503e-05,
+      "loss": 0.2463,
+      "step": 48920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8420412575443657e-05,
+      "loss": 0.2451,
+      "step": 48930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.841590847671381e-05,
+      "loss": 0.253,
+      "step": 48940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8411404377983968e-05,
+      "loss": 0.2516,
+      "step": 48950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8406900279254122e-05,
+      "loss": 0.2512,
+      "step": 48960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8402396180524276e-05,
+      "loss": 0.2456,
+      "step": 48970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8397892081794437e-05,
+      "loss": 0.2492,
+      "step": 48980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.839338798306459e-05,
+      "loss": 0.2497,
+      "step": 48990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8388883884334742e-05,
+      "loss": 0.248,
+      "step": 49000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8384379785604903e-05,
+      "loss": 0.2518,
+      "step": 49010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8379875686875057e-05,
+      "loss": 0.2423,
+      "step": 49020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.837537158814521e-05,
+      "loss": 0.2415,
+      "step": 49030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.837086748941537e-05,
+      "loss": 0.2472,
+      "step": 49040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8366363390685523e-05,
+      "loss": 0.2428,
+      "step": 49050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8361859291955684e-05,
+      "loss": 0.2438,
+      "step": 49060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8357355193225838e-05,
+      "loss": 0.2484,
+      "step": 49070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8352851094495992e-05,
+      "loss": 0.2455,
+      "step": 49080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.834834699576615e-05,
+      "loss": 0.2445,
+      "step": 49090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8343842897036303e-05,
+      "loss": 0.2534,
+      "step": 49100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8339338798306458e-05,
+      "loss": 0.2493,
+      "step": 49110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.833483469957662e-05,
+      "loss": 0.2452,
+      "step": 49120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8330330600846773e-05,
+      "loss": 0.2511,
+      "step": 49130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8325826502116927e-05,
+      "loss": 0.2466,
+      "step": 49140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8321322403387084e-05,
+      "loss": 0.246,
+      "step": 49150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8316818304657238e-05,
+      "loss": 0.2466,
+      "step": 49160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8312314205927392e-05,
+      "loss": 0.241,
+      "step": 49170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8307810107197553e-05,
+      "loss": 0.2489,
+      "step": 49180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8303306008467707e-05,
+      "loss": 0.2522,
+      "step": 49190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8298801909737865e-05,
+      "loss": 0.2498,
+      "step": 49200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.829429781100802e-05,
+      "loss": 0.2511,
+      "step": 49210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8289793712278173e-05,
+      "loss": 0.2511,
+      "step": 49220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8285289613548334e-05,
+      "loss": 0.2482,
+      "step": 49230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8280785514818488e-05,
+      "loss": 0.2427,
+      "step": 49240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.827628141608864e-05,
+      "loss": 0.2517,
+      "step": 49250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.82717773173588e-05,
+      "loss": 0.2551,
+      "step": 49260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8267273218628954e-05,
+      "loss": 0.2445,
+      "step": 49270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8262769119899108e-05,
+      "loss": 0.2517,
+      "step": 49280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8258265021169265e-05,
+      "loss": 0.2482,
+      "step": 49290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.825376092243942e-05,
+      "loss": 0.2504,
+      "step": 49300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8249256823709574e-05,
+      "loss": 0.2442,
+      "step": 49310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8244752724979734e-05,
+      "loss": 0.2517,
+      "step": 49320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.824024862624989e-05,
+      "loss": 0.2456,
+      "step": 49330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8235744527520046e-05,
+      "loss": 0.2465,
+      "step": 49340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.82312404287902e-05,
+      "loss": 0.2506,
+      "step": 49350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8226736330060354e-05,
+      "loss": 0.2426,
+      "step": 49360
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8222232231330515e-05,
+      "loss": 0.2477,
+      "step": 49370
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.821772813260067e-05,
+      "loss": 0.2461,
+      "step": 49380
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8213224033870823e-05,
+      "loss": 0.2457,
+      "step": 49390
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.820871993514098e-05,
+      "loss": 0.2501,
+      "step": 49400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8204215836411135e-05,
+      "loss": 0.2479,
+      "step": 49410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.819971173768129e-05,
+      "loss": 0.2487,
+      "step": 49420
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.819520763895145e-05,
+      "loss": 0.2491,
+      "step": 49430
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8190703540221604e-05,
+      "loss": 0.2465,
+      "step": 49440
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8186199441491755e-05,
+      "loss": 0.2461,
+      "step": 49450
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8181695342761916e-05,
+      "loss": 0.2459,
+      "step": 49460
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.817719124403207e-05,
+      "loss": 0.2498,
+      "step": 49470
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.817268714530223e-05,
+      "loss": 0.2468,
+      "step": 49480
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.816818304657238e-05,
+      "loss": 0.2462,
+      "step": 49490
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8163678947842535e-05,
+      "loss": 0.2514,
+      "step": 49500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8159174849112696e-05,
+      "loss": 0.247,
+      "step": 49510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.815467075038285e-05,
+      "loss": 0.2517,
+      "step": 49520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8150166651653004e-05,
+      "loss": 0.2447,
+      "step": 49530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8145662552923162e-05,
+      "loss": 0.2435,
+      "step": 49540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8141158454193316e-05,
+      "loss": 0.2479,
+      "step": 49550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.813665435546347e-05,
+      "loss": 0.2431,
+      "step": 49560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.813215025673363e-05,
+      "loss": 0.2415,
+      "step": 49570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8127646158003785e-05,
+      "loss": 0.2449,
+      "step": 49580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.812314205927394e-05,
+      "loss": 0.2485,
+      "step": 49590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8118637960544097e-05,
+      "loss": 0.2484,
+      "step": 49600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.811413386181425e-05,
+      "loss": 0.2441,
+      "step": 49610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8109629763084412e-05,
+      "loss": 0.2476,
+      "step": 49620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8105125664354566e-05,
+      "loss": 0.2423,
+      "step": 49630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.810062156562472e-05,
+      "loss": 0.2428,
+      "step": 49640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8096117466894877e-05,
+      "loss": 0.247,
+      "step": 49650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.809161336816503e-05,
+      "loss": 0.2453,
+      "step": 49660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8087109269435186e-05,
+      "loss": 0.2451,
+      "step": 49670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8082605170705347e-05,
+      "loss": 0.2408,
+      "step": 49680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8078101071975497e-05,
+      "loss": 0.2495,
+      "step": 49690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.807359697324565e-05,
+      "loss": 0.2464,
+      "step": 49700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8069092874515812e-05,
+      "loss": 0.2507,
+      "step": 49710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8064588775785966e-05,
+      "loss": 0.2462,
+      "step": 49720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.806008467705612e-05,
+      "loss": 0.2459,
+      "step": 49730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8055580578326278e-05,
+      "loss": 0.2459,
+      "step": 49740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8051076479596432e-05,
+      "loss": 0.2466,
+      "step": 49750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8047022790739576e-05,
+      "loss": 0.2453,
+      "step": 49760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.804251869200973e-05,
+      "loss": 0.2538,
+      "step": 49770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8038014593279887e-05,
+      "loss": 0.2517,
+      "step": 49780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.803351049455004e-05,
+      "loss": 0.2493,
+      "step": 49790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8029006395820196e-05,
+      "loss": 0.2488,
+      "step": 49800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8024502297090353e-05,
+      "loss": 0.2478,
+      "step": 49810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8019998198360507e-05,
+      "loss": 0.2422,
+      "step": 49820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.801549409963066e-05,
+      "loss": 0.2423,
+      "step": 49830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8010990000900822e-05,
+      "loss": 0.2487,
+      "step": 49840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8006485902170976e-05,
+      "loss": 0.2446,
+      "step": 49850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8001981803441134e-05,
+      "loss": 0.2468,
+      "step": 49860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7997477704711288e-05,
+      "loss": 0.2429,
+      "step": 49870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7992973605981442e-05,
+      "loss": 0.2481,
+      "step": 49880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7988469507251603e-05,
+      "loss": 0.2484,
+      "step": 49890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7983965408521757e-05,
+      "loss": 0.2466,
+      "step": 49900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.797946130979191e-05,
+      "loss": 0.2472,
+      "step": 49910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.797495721106207e-05,
+      "loss": 0.2442,
+      "step": 49920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7970453112332223e-05,
+      "loss": 0.2428,
+      "step": 49930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7965949013602377e-05,
+      "loss": 0.2408,
+      "step": 49940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7961444914872538e-05,
+      "loss": 0.2505,
+      "step": 49950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7956940816142692e-05,
+      "loss": 0.2488,
+      "step": 49960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7952436717412846e-05,
+      "loss": 0.2406,
+      "step": 49970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7947932618683003e-05,
+      "loss": 0.24,
+      "step": 49980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7943428519953157e-05,
+      "loss": 0.2512,
+      "step": 49990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.793892442122332e-05,
+      "loss": 0.2498,
+      "step": 50000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7934420322493472e-05,
+      "loss": 0.244,
+      "step": 50010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7929916223763623e-05,
+      "loss": 0.2428,
+      "step": 50020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7925412125033784e-05,
+      "loss": 0.247,
+      "step": 50030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7920908026303938e-05,
+      "loss": 0.2462,
+      "step": 50040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7916403927574092e-05,
+      "loss": 0.2402,
+      "step": 50050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.791189982884425e-05,
+      "loss": 0.2447,
+      "step": 50060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7907395730114404e-05,
+      "loss": 0.246,
+      "step": 50070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7902891631384558e-05,
+      "loss": 0.2454,
+      "step": 50080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.789838753265472e-05,
+      "loss": 0.2445,
+      "step": 50090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7893883433924873e-05,
+      "loss": 0.2458,
+      "step": 50100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7889379335195027e-05,
+      "loss": 0.245,
+      "step": 50110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7884875236465185e-05,
+      "loss": 0.2444,
+      "step": 50120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.788037113773534e-05,
+      "loss": 0.247,
+      "step": 50130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.78758670390055e-05,
+      "loss": 0.2455,
+      "step": 50140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7871362940275654e-05,
+      "loss": 0.2447,
+      "step": 50150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7866858841545808e-05,
+      "loss": 0.2501,
+      "step": 50160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7862354742815965e-05,
+      "loss": 0.2477,
+      "step": 50170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.785785064408612e-05,
+      "loss": 0.2501,
+      "step": 50180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7853346545356273e-05,
+      "loss": 0.2464,
+      "step": 50190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7848842446626434e-05,
+      "loss": 0.2469,
+      "step": 50200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.784433834789659e-05,
+      "loss": 0.2513,
+      "step": 50210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.783983424916674e-05,
+      "loss": 0.2479,
+      "step": 50220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.78353301504369e-05,
+      "loss": 0.2495,
+      "step": 50230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7830826051707054e-05,
+      "loss": 0.2428,
+      "step": 50240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7826321952977208e-05,
+      "loss": 0.2437,
+      "step": 50250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7821817854247366e-05,
+      "loss": 0.2412,
+      "step": 50260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.781731375551752e-05,
+      "loss": 0.2527,
+      "step": 50270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.781280965678768e-05,
+      "loss": 0.2458,
+      "step": 50280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7808305558057835e-05,
+      "loss": 0.2469,
+      "step": 50290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.780380145932799e-05,
+      "loss": 0.2423,
+      "step": 50300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7799297360598146e-05,
+      "loss": 0.245,
+      "step": 50310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.77947932618683e-05,
+      "loss": 0.2412,
+      "step": 50320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7790289163138455e-05,
+      "loss": 0.2464,
+      "step": 50330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7785785064408616e-05,
+      "loss": 0.2487,
+      "step": 50340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.778128096567877e-05,
+      "loss": 0.2455,
+      "step": 50350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7776776866948924e-05,
+      "loss": 0.253,
+      "step": 50360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.777227276821908e-05,
+      "loss": 0.2463,
+      "step": 50370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7767768669489235e-05,
+      "loss": 0.2475,
+      "step": 50380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.776326457075939e-05,
+      "loss": 0.2544,
+      "step": 50390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.775876047202955e-05,
+      "loss": 0.2413,
+      "step": 50400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7754256373299704e-05,
+      "loss": 0.2438,
+      "step": 50410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7749752274569862e-05,
+      "loss": 0.2514,
+      "step": 50420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7745248175840016e-05,
+      "loss": 0.2455,
+      "step": 50430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.774074407711017e-05,
+      "loss": 0.2406,
+      "step": 50440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.773623997838033e-05,
+      "loss": 0.2473,
+      "step": 50450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.773173587965048e-05,
+      "loss": 0.242,
+      "step": 50460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7727231780920636e-05,
+      "loss": 0.2427,
+      "step": 50470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7722727682190797e-05,
+      "loss": 0.2445,
+      "step": 50480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.771822358346095e-05,
+      "loss": 0.2554,
+      "step": 50490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7713719484731105e-05,
+      "loss": 0.2495,
+      "step": 50500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7709215386001262e-05,
+      "loss": 0.255,
+      "step": 50510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7704711287271416e-05,
+      "loss": 0.2503,
+      "step": 50520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.770020718854157e-05,
+      "loss": 0.247,
+      "step": 50530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.769570308981173e-05,
+      "loss": 0.2402,
+      "step": 50540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7691198991081886e-05,
+      "loss": 0.2488,
+      "step": 50550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7686694892352043e-05,
+      "loss": 0.2441,
+      "step": 50560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7682190793622197e-05,
+      "loss": 0.2436,
+      "step": 50570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.767768669489235e-05,
+      "loss": 0.2393,
+      "step": 50580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7673182596162512e-05,
+      "loss": 0.2416,
+      "step": 50590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7668678497432666e-05,
+      "loss": 0.2439,
+      "step": 50600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.766417439870282e-05,
+      "loss": 0.2469,
+      "step": 50610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7659670299972978e-05,
+      "loss": 0.2462,
+      "step": 50620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7655166201243132e-05,
+      "loss": 0.2499,
+      "step": 50630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7650662102513286e-05,
+      "loss": 0.2477,
+      "step": 50640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7646158003783447e-05,
+      "loss": 0.2416,
+      "step": 50650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.76416539050536e-05,
+      "loss": 0.2403,
+      "step": 50660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7637149806323752e-05,
+      "loss": 0.2419,
+      "step": 50670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7632645707593913e-05,
+      "loss": 0.2377,
+      "step": 50680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7628141608864067e-05,
+      "loss": 0.2431,
+      "step": 50690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7623637510134228e-05,
+      "loss": 0.2512,
+      "step": 50700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.761913341140438e-05,
+      "loss": 0.2485,
+      "step": 50710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7614629312674532e-05,
+      "loss": 0.2471,
+      "step": 50720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7610125213944693e-05,
+      "loss": 0.2383,
+      "step": 50730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7605621115214847e-05,
+      "loss": 0.2468,
+      "step": 50740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7601117016485e-05,
+      "loss": 0.2462,
+      "step": 50750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.759661291775516e-05,
+      "loss": 0.2386,
+      "step": 50760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7592108819025313e-05,
+      "loss": 0.2524,
+      "step": 50770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7587604720295467e-05,
+      "loss": 0.2463,
+      "step": 50780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7583100621565628e-05,
+      "loss": 0.2376,
+      "step": 50790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7578596522835782e-05,
+      "loss": 0.2437,
+      "step": 50800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7574092424105936e-05,
+      "loss": 0.2489,
+      "step": 50810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7569588325376094e-05,
+      "loss": 0.2472,
+      "step": 50820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7565084226646248e-05,
+      "loss": 0.2424,
+      "step": 50830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.756058012791641e-05,
+      "loss": 0.2454,
+      "step": 50840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7556076029186563e-05,
+      "loss": 0.2479,
+      "step": 50850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7551571930456717e-05,
+      "loss": 0.2445,
+      "step": 50860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7547067831726874e-05,
+      "loss": 0.2399,
+      "step": 50870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.754256373299703e-05,
+      "loss": 0.2342,
+      "step": 50880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7538059634267183e-05,
+      "loss": 0.2426,
+      "step": 50890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7533555535537344e-05,
+      "loss": 0.2456,
+      "step": 50900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7529051436807494e-05,
+      "loss": 0.2426,
+      "step": 50910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.752454733807765e-05,
+      "loss": 0.2416,
+      "step": 50920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.752004323934781e-05,
+      "loss": 0.2484,
+      "step": 50930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7515539140617963e-05,
+      "loss": 0.2399,
+      "step": 50940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7511035041888117e-05,
+      "loss": 0.2432,
+      "step": 50950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7506530943158275e-05,
+      "loss": 0.2501,
+      "step": 50960
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.750202684442843e-05,
+      "loss": 0.2437,
+      "step": 50970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.749752274569859e-05,
+      "loss": 0.249,
+      "step": 50980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7493018646968744e-05,
+      "loss": 0.2444,
+      "step": 50990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7488514548238898e-05,
+      "loss": 0.2453,
+      "step": 51000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7484010449509056e-05,
+      "loss": 0.247,
+      "step": 51010
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.747950635077921e-05,
+      "loss": 0.2405,
+      "step": 51020
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7475002252049364e-05,
+      "loss": 0.2438,
+      "step": 51030
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7470498153319525e-05,
+      "loss": 0.2432,
+      "step": 51040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.746599405458968e-05,
+      "loss": 0.2495,
+      "step": 51050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7461489955859833e-05,
+      "loss": 0.2486,
+      "step": 51060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.745698585712999e-05,
+      "loss": 0.2452,
+      "step": 51070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7452481758400145e-05,
+      "loss": 0.2474,
+      "step": 51080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.74479776596703e-05,
+      "loss": 0.2403,
+      "step": 51090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.744347356094046e-05,
+      "loss": 0.2465,
+      "step": 51100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.743896946221061e-05,
+      "loss": 0.241,
+      "step": 51110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.743446536348077e-05,
+      "loss": 0.2488,
+      "step": 51120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7429961264750925e-05,
+      "loss": 0.2397,
+      "step": 51130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.742545716602108e-05,
+      "loss": 0.2445,
+      "step": 51140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7420953067291237e-05,
+      "loss": 0.2415,
+      "step": 51150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.741644896856139e-05,
+      "loss": 0.2469,
+      "step": 51160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7411944869831545e-05,
+      "loss": 0.2437,
+      "step": 51170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7407440771101706e-05,
+      "loss": 0.2449,
+      "step": 51180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.740293667237186e-05,
+      "loss": 0.2367,
+      "step": 51190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7398432573642014e-05,
+      "loss": 0.2452,
+      "step": 51200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.739392847491217e-05,
+      "loss": 0.2403,
+      "step": 51210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7389424376182326e-05,
+      "loss": 0.2521,
+      "step": 51220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.738492027745248e-05,
+      "loss": 0.2469,
+      "step": 51230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.738041617872264e-05,
+      "loss": 0.2438,
+      "step": 51240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7375912079992795e-05,
+      "loss": 0.2464,
+      "step": 51250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7371407981262952e-05,
+      "loss": 0.2476,
+      "step": 51260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7366903882533106e-05,
+      "loss": 0.2412,
+      "step": 51270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.736239978380326e-05,
+      "loss": 0.2434,
+      "step": 51280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.735789568507342e-05,
+      "loss": 0.2406,
+      "step": 51290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7353391586343576e-05,
+      "loss": 0.2401,
+      "step": 51300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.734888748761373e-05,
+      "loss": 0.2408,
+      "step": 51310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7344383388883887e-05,
+      "loss": 0.2446,
+      "step": 51320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.733987929015404e-05,
+      "loss": 0.2438,
+      "step": 51330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7335375191424195e-05,
+      "loss": 0.2401,
+      "step": 51340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7330871092694356e-05,
+      "loss": 0.2444,
+      "step": 51350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7326366993964507e-05,
+      "loss": 0.2416,
+      "step": 51360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.732186289523466e-05,
+      "loss": 0.2422,
+      "step": 51370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7317358796504822e-05,
+      "loss": 0.2438,
+      "step": 51380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7312854697774976e-05,
+      "loss": 0.2453,
+      "step": 51390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7308350599045133e-05,
+      "loss": 0.2395,
+      "step": 51400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7303846500315288e-05,
+      "loss": 0.2347,
+      "step": 51410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7299342401585442e-05,
+      "loss": 0.2424,
+      "step": 51420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7294838302855603e-05,
+      "loss": 0.2423,
+      "step": 51430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7290334204125757e-05,
+      "loss": 0.243,
+      "step": 51440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.728583010539591e-05,
+      "loss": 0.2458,
+      "step": 51450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7281326006666068e-05,
+      "loss": 0.241,
+      "step": 51460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7276821907936222e-05,
+      "loss": 0.2434,
+      "step": 51470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7272317809206376e-05,
+      "loss": 0.2508,
+      "step": 51480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7267813710476537e-05,
+      "loss": 0.2492,
+      "step": 51490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.726330961174669e-05,
+      "loss": 0.2409,
+      "step": 51500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7258805513016846e-05,
+      "loss": 0.2423,
+      "step": 51510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7254301414287003e-05,
+      "loss": 0.2388,
+      "step": 51520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7249797315557157e-05,
+      "loss": 0.2462,
+      "step": 51530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7245293216827318e-05,
+      "loss": 0.2446,
+      "step": 51540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7240789118097472e-05,
+      "loss": 0.2399,
+      "step": 51550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7236285019367623e-05,
+      "loss": 0.2433,
+      "step": 51560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7231780920637784e-05,
+      "loss": 0.2428,
+      "step": 51570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7227276821907938e-05,
+      "loss": 0.2425,
+      "step": 51580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7222772723178092e-05,
+      "loss": 0.2419,
+      "step": 51590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.721826862444825e-05,
+      "loss": 0.2465,
+      "step": 51600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7213764525718404e-05,
+      "loss": 0.2437,
+      "step": 51610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7209260426988558e-05,
+      "loss": 0.2435,
+      "step": 51620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.720475632825872e-05,
+      "loss": 0.2439,
+      "step": 51630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7200252229528873e-05,
+      "loss": 0.2453,
+      "step": 51640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7195748130799027e-05,
+      "loss": 0.2427,
+      "step": 51650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7191244032069184e-05,
+      "loss": 0.2422,
+      "step": 51660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.718673993333934e-05,
+      "loss": 0.2465,
+      "step": 51670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.71822358346095e-05,
+      "loss": 0.247,
+      "step": 51680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7177731735879653e-05,
+      "loss": 0.2462,
+      "step": 51690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7173227637149807e-05,
+      "loss": 0.2411,
+      "step": 51700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7168723538419965e-05,
+      "loss": 0.2443,
+      "step": 51710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.716421943969012e-05,
+      "loss": 0.2476,
+      "step": 51720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7159715340960273e-05,
+      "loss": 0.2405,
+      "step": 51730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7155211242230434e-05,
+      "loss": 0.2377,
+      "step": 51740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7150707143500588e-05,
+      "loss": 0.2361,
+      "step": 51750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.714620304477074e-05,
+      "loss": 0.2404,
+      "step": 51760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.71416989460409e-05,
+      "loss": 0.2466,
+      "step": 51770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7137194847311054e-05,
+      "loss": 0.243,
+      "step": 51780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7132690748581208e-05,
+      "loss": 0.2433,
+      "step": 51790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7128186649851365e-05,
+      "loss": 0.2459,
+      "step": 51800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.712368255112152e-05,
+      "loss": 0.2355,
+      "step": 51810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.711917845239168e-05,
+      "loss": 0.2403,
+      "step": 51820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7114674353661835e-05,
+      "loss": 0.24,
+      "step": 51830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.711017025493199e-05,
+      "loss": 0.2431,
+      "step": 51840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.710611656607513e-05,
+      "loss": 0.24,
+      "step": 51850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7101612467345283e-05,
+      "loss": 0.2368,
+      "step": 51860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7097108368615444e-05,
+      "loss": 0.2381,
+      "step": 51870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7092604269885598e-05,
+      "loss": 0.2394,
+      "step": 51880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.708810017115575e-05,
+      "loss": 0.2418,
+      "step": 51890
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.708359607242591e-05,
+      "loss": 0.244,
+      "step": 51900
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7079091973696064e-05,
+      "loss": 0.2421,
+      "step": 51910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.707458787496622e-05,
+      "loss": 0.2454,
+      "step": 51920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7070083776236375e-05,
+      "loss": 0.2426,
+      "step": 51930
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.706557967750653e-05,
+      "loss": 0.2381,
+      "step": 51940
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.706107557877669e-05,
+      "loss": 0.2409,
+      "step": 51950
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7056571480046844e-05,
+      "loss": 0.2405,
+      "step": 51960
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7052067381317e-05,
+      "loss": 0.245,
+      "step": 51970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7047563282587156e-05,
+      "loss": 0.2381,
+      "step": 51980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.704305918385731e-05,
+      "loss": 0.2478,
+      "step": 51990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7038555085127464e-05,
+      "loss": 0.2339,
+      "step": 52000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7034050986397625e-05,
+      "loss": 0.2419,
+      "step": 52010
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.702954688766778e-05,
+      "loss": 0.2456,
+      "step": 52020
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7025042788937933e-05,
+      "loss": 0.2426,
+      "step": 52030
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.702053869020809e-05,
+      "loss": 0.2412,
+      "step": 52040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7016034591478245e-05,
+      "loss": 0.2371,
+      "step": 52050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7011530492748406e-05,
+      "loss": 0.2386,
+      "step": 52060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.700702639401856e-05,
+      "loss": 0.2351,
+      "step": 52070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.7002522295288714e-05,
+      "loss": 0.248,
+      "step": 52080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.699801819655887e-05,
+      "loss": 0.2432,
+      "step": 52090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6993514097829026e-05,
+      "loss": 0.2408,
+      "step": 52100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.698900999909918e-05,
+      "loss": 0.2392,
+      "step": 52110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.698450590036934e-05,
+      "loss": 0.2418,
+      "step": 52120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.698000180163949e-05,
+      "loss": 0.2345,
+      "step": 52130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6975497702909645e-05,
+      "loss": 0.2388,
+      "step": 52140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6970993604179806e-05,
+      "loss": 0.2388,
+      "step": 52150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.696648950544996e-05,
+      "loss": 0.2391,
+      "step": 52160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6961985406720115e-05,
+      "loss": 0.2402,
+      "step": 52170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6957481307990272e-05,
+      "loss": 0.2472,
+      "step": 52180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6952977209260426e-05,
+      "loss": 0.2401,
+      "step": 52190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6948473110530587e-05,
+      "loss": 0.2486,
+      "step": 52200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.694396901180074e-05,
+      "loss": 0.2429,
+      "step": 52210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6939464913070895e-05,
+      "loss": 0.2471,
+      "step": 52220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6934960814341053e-05,
+      "loss": 0.2425,
+      "step": 52230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6930456715611207e-05,
+      "loss": 0.2387,
+      "step": 52240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.692595261688136e-05,
+      "loss": 0.2374,
+      "step": 52250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6921448518151522e-05,
+      "loss": 0.2382,
+      "step": 52260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6916944419421676e-05,
+      "loss": 0.2408,
+      "step": 52270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.691244032069183e-05,
+      "loss": 0.2397,
+      "step": 52280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6907936221961988e-05,
+      "loss": 0.2494,
+      "step": 52290
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.690343212323214e-05,
+      "loss": 0.2435,
+      "step": 52300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6898928024502296e-05,
+      "loss": 0.2406,
+      "step": 52310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6894423925772457e-05,
+      "loss": 0.2458,
+      "step": 52320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6889919827042607e-05,
+      "loss": 0.2377,
+      "step": 52330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6885415728312768e-05,
+      "loss": 0.2406,
+      "step": 52340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6880911629582922e-05,
+      "loss": 0.2359,
+      "step": 52350
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6876407530853076e-05,
+      "loss": 0.2342,
+      "step": 52360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6871903432123234e-05,
+      "loss": 0.2432,
+      "step": 52370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6867399333393388e-05,
+      "loss": 0.2434,
+      "step": 52380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6862895234663542e-05,
+      "loss": 0.2509,
+      "step": 52390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6858391135933703e-05,
+      "loss": 0.2428,
+      "step": 52400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6853887037203857e-05,
+      "loss": 0.2395,
+      "step": 52410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.684938293847401e-05,
+      "loss": 0.2457,
+      "step": 52420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.684487883974417e-05,
+      "loss": 0.2408,
+      "step": 52430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6840374741014323e-05,
+      "loss": 0.2429,
+      "step": 52440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6835870642284477e-05,
+      "loss": 0.2382,
+      "step": 52450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6831366543554638e-05,
+      "loss": 0.2398,
+      "step": 52460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6826862444824792e-05,
+      "loss": 0.2405,
+      "step": 52470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.682235834609495e-05,
+      "loss": 0.2396,
+      "step": 52480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6817854247365103e-05,
+      "loss": 0.2473,
+      "step": 52490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6813350148635258e-05,
+      "loss": 0.2394,
+      "step": 52500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.680884604990542e-05,
+      "loss": 0.2329,
+      "step": 52510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6804341951175573e-05,
+      "loss": 0.2362,
+      "step": 52520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6799837852445727e-05,
+      "loss": 0.2387,
+      "step": 52530
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6795333753715884e-05,
+      "loss": 0.2384,
+      "step": 52540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6790829654986038e-05,
+      "loss": 0.2346,
+      "step": 52550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6786325556256192e-05,
+      "loss": 0.2359,
+      "step": 52560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.678182145752635e-05,
+      "loss": 0.2397,
+      "step": 52570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6777317358796504e-05,
+      "loss": 0.2322,
+      "step": 52580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6772813260066658e-05,
+      "loss": 0.2414,
+      "step": 52590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.676830916133682e-05,
+      "loss": 0.2463,
+      "step": 52600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6763805062606973e-05,
+      "loss": 0.2449,
+      "step": 52610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.675930096387713e-05,
+      "loss": 0.2437,
+      "step": 52620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6754796865147285e-05,
+      "loss": 0.244,
+      "step": 52630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.675029276641744e-05,
+      "loss": 0.2445,
+      "step": 52640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.67457886676876e-05,
+      "loss": 0.2408,
+      "step": 52650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6741284568957754e-05,
+      "loss": 0.2369,
+      "step": 52660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6736780470227908e-05,
+      "loss": 0.2418,
+      "step": 52670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6732276371498065e-05,
+      "loss": 0.2407,
+      "step": 52680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.672777227276822e-05,
+      "loss": 0.2385,
+      "step": 52690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6723268174038374e-05,
+      "loss": 0.2414,
+      "step": 52700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6718764075308534e-05,
+      "loss": 0.237,
+      "step": 52710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.671425997657869e-05,
+      "loss": 0.2405,
+      "step": 52720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6709755877848843e-05,
+      "loss": 0.2446,
+      "step": 52730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6705251779119e-05,
+      "loss": 0.2406,
+      "step": 52740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6700747680389154e-05,
+      "loss": 0.2391,
+      "step": 52750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6696243581659315e-05,
+      "loss": 0.2442,
+      "step": 52760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.669173948292947e-05,
+      "loss": 0.244,
+      "step": 52770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.668723538419962e-05,
+      "loss": 0.2453,
+      "step": 52780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.668273128546978e-05,
+      "loss": 0.2369,
+      "step": 52790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6678227186739935e-05,
+      "loss": 0.2345,
+      "step": 52800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.667372308801009e-05,
+      "loss": 0.2474,
+      "step": 52810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6669218989280246e-05,
+      "loss": 0.2407,
+      "step": 52820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.66647148905504e-05,
+      "loss": 0.238,
+      "step": 52830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6660210791820555e-05,
+      "loss": 0.2397,
+      "step": 52840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6655706693090716e-05,
+      "loss": 0.2447,
+      "step": 52850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.665120259436087e-05,
+      "loss": 0.2407,
+      "step": 52860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6646698495631024e-05,
+      "loss": 0.2382,
+      "step": 52870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.664219439690118e-05,
+      "loss": 0.2363,
+      "step": 52880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6637690298171335e-05,
+      "loss": 0.2319,
+      "step": 52890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6633186199441496e-05,
+      "loss": 0.2394,
+      "step": 52900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.662868210071165e-05,
+      "loss": 0.236,
+      "step": 52910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6624178001981804e-05,
+      "loss": 0.2408,
+      "step": 52920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6619673903251962e-05,
+      "loss": 0.2403,
+      "step": 52930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6615169804522116e-05,
+      "loss": 0.2401,
+      "step": 52940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.661066570579227e-05,
+      "loss": 0.2403,
+      "step": 52950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.660616160706243e-05,
+      "loss": 0.2383,
+      "step": 52960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6601657508332585e-05,
+      "loss": 0.2401,
+      "step": 52970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6597153409602736e-05,
+      "loss": 0.2396,
+      "step": 52980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6592649310872897e-05,
+      "loss": 0.2457,
+      "step": 52990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.658814521214305e-05,
+      "loss": 0.2382,
+      "step": 53000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6583641113413205e-05,
+      "loss": 0.2412,
+      "step": 53010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6579137014683362e-05,
+      "loss": 0.2464,
+      "step": 53020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6574632915953517e-05,
+      "loss": 0.2402,
+      "step": 53030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6570128817223677e-05,
+      "loss": 0.2419,
+      "step": 53040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.656562471849383e-05,
+      "loss": 0.2388,
+      "step": 53050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6561120619763986e-05,
+      "loss": 0.2414,
+      "step": 53060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6556616521034143e-05,
+      "loss": 0.2398,
+      "step": 53070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6552112422304297e-05,
+      "loss": 0.2458,
+      "step": 53080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.654760832357445e-05,
+      "loss": 0.2368,
+      "step": 53090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6543104224844612e-05,
+      "loss": 0.2364,
+      "step": 53100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6538600126114766e-05,
+      "loss": 0.2412,
+      "step": 53110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.653409602738492e-05,
+      "loss": 0.2432,
+      "step": 53120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6529591928655078e-05,
+      "loss": 0.2417,
+      "step": 53130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6525087829925232e-05,
+      "loss": 0.2408,
+      "step": 53140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6520583731195386e-05,
+      "loss": 0.2397,
+      "step": 53150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6516079632465547e-05,
+      "loss": 0.2361,
+      "step": 53160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.65115755337357e-05,
+      "loss": 0.2395,
+      "step": 53170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.650707143500586e-05,
+      "loss": 0.2483,
+      "step": 53180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6502567336276013e-05,
+      "loss": 0.2387,
+      "step": 53190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.6498063237546167e-05,
+      "loss": 0.2406,
+      "step": 53200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6493559138816328e-05,
+      "loss": 0.2421,
+      "step": 53210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.648905504008648e-05,
+      "loss": 0.2358,
+      "step": 53220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6484550941356633e-05,
+      "loss": 0.2403,
+      "step": 53230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6480046842626793e-05,
+      "loss": 0.2403,
+      "step": 53240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6475542743896948e-05,
+      "loss": 0.2385,
+      "step": 53250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.64710386451671e-05,
+      "loss": 0.2386,
+      "step": 53260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.646653454643726e-05,
+      "loss": 0.2383,
+      "step": 53270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6462030447707413e-05,
+      "loss": 0.2402,
+      "step": 53280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6457526348977567e-05,
+      "loss": 0.2387,
+      "step": 53290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6453022250247728e-05,
+      "loss": 0.2391,
+      "step": 53300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6448518151517882e-05,
+      "loss": 0.2362,
+      "step": 53310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.644401405278804e-05,
+      "loss": 0.2385,
+      "step": 53320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6439509954058194e-05,
+      "loss": 0.2388,
+      "step": 53330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6435005855328348e-05,
+      "loss": 0.2372,
+      "step": 53340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.643050175659851e-05,
+      "loss": 0.2396,
+      "step": 53350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6425997657868663e-05,
+      "loss": 0.2415,
+      "step": 53360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6421493559138817e-05,
+      "loss": 0.2369,
+      "step": 53370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6416989460408975e-05,
+      "loss": 0.2364,
+      "step": 53380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.641248536167913e-05,
+      "loss": 0.2398,
+      "step": 53390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6407981262949283e-05,
+      "loss": 0.2397,
+      "step": 53400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6403477164219444e-05,
+      "loss": 0.2399,
+      "step": 53410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6398973065489598e-05,
+      "loss": 0.2363,
+      "step": 53420
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.639446896675975e-05,
+      "loss": 0.2409,
+      "step": 53430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.638996486802991e-05,
+      "loss": 0.2383,
+      "step": 53440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6385460769300063e-05,
+      "loss": 0.2393,
+      "step": 53450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6380956670570224e-05,
+      "loss": 0.2402,
+      "step": 53460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6376452571840375e-05,
+      "loss": 0.2369,
+      "step": 53470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.637194847311053e-05,
+      "loss": 0.2427,
+      "step": 53480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.636744437438069e-05,
+      "loss": 0.2387,
+      "step": 53490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6362940275650844e-05,
+      "loss": 0.2414,
+      "step": 53500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6358436176920998e-05,
+      "loss": 0.241,
+      "step": 53510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6353932078191156e-05,
+      "loss": 0.2348,
+      "step": 53520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.634942797946131e-05,
+      "loss": 0.2411,
+      "step": 53530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6344923880731464e-05,
+      "loss": 0.2381,
+      "step": 53540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6340419782001625e-05,
+      "loss": 0.2435,
+      "step": 53550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.633591568327178e-05,
+      "loss": 0.2419,
+      "step": 53560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6331411584541933e-05,
+      "loss": 0.2411,
+      "step": 53570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.632690748581209e-05,
+      "loss": 0.2421,
+      "step": 53580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6322403387082245e-05,
+      "loss": 0.2401,
+      "step": 53590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6317899288352406e-05,
+      "loss": 0.2437,
+      "step": 53600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.631339518962256e-05,
+      "loss": 0.2424,
+      "step": 53610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6308891090892714e-05,
+      "loss": 0.2402,
+      "step": 53620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.630438699216287e-05,
+      "loss": 0.2406,
+      "step": 53630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6299882893433025e-05,
+      "loss": 0.2351,
+      "step": 53640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.629537879470318e-05,
+      "loss": 0.2403,
+      "step": 53650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.629087469597334e-05,
+      "loss": 0.2339,
+      "step": 53660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.628637059724349e-05,
+      "loss": 0.2396,
+      "step": 53670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6281866498513645e-05,
+      "loss": 0.2374,
+      "step": 53680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6277362399783806e-05,
+      "loss": 0.2404,
+      "step": 53690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.627285830105396e-05,
+      "loss": 0.2459,
+      "step": 53700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6268354202324114e-05,
+      "loss": 0.2354,
+      "step": 53710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6263850103594272e-05,
+      "loss": 0.2393,
+      "step": 53720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6259346004864426e-05,
+      "loss": 0.2339,
+      "step": 53730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6254841906134587e-05,
+      "loss": 0.2377,
+      "step": 53740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.625033780740474e-05,
+      "loss": 0.2413,
+      "step": 53750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6245833708674895e-05,
+      "loss": 0.2399,
+      "step": 53760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6241329609945052e-05,
+      "loss": 0.2416,
+      "step": 53770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6236825511215207e-05,
+      "loss": 0.2465,
+      "step": 53780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.623232141248536e-05,
+      "loss": 0.2357,
+      "step": 53790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.622781731375552e-05,
+      "loss": 0.2398,
+      "step": 53800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6223313215025676e-05,
+      "loss": 0.2396,
+      "step": 53810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.621880911629583e-05,
+      "loss": 0.2456,
+      "step": 53820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6214305017565987e-05,
+      "loss": 0.2385,
+      "step": 53830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.620980091883614e-05,
+      "loss": 0.2411,
+      "step": 53840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6205296820106295e-05,
+      "loss": 0.2397,
+      "step": 53850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6200792721376456e-05,
+      "loss": 0.2346,
+      "step": 53860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6196739032519597e-05,
+      "loss": 0.2372,
+      "step": 53870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.619223493378975e-05,
+      "loss": 0.2335,
+      "step": 53880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6187730835059905e-05,
+      "loss": 0.2382,
+      "step": 53890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6183226736330062e-05,
+      "loss": 0.2383,
+      "step": 53900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6178722637600216e-05,
+      "loss": 0.2343,
+      "step": 53910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.617421853887037e-05,
+      "loss": 0.2382,
+      "step": 53920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.616971444014053e-05,
+      "loss": 0.2444,
+      "step": 53930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6165210341410686e-05,
+      "loss": 0.2414,
+      "step": 53940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6160706242680843e-05,
+      "loss": 0.2447,
+      "step": 53950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6156202143950997e-05,
+      "loss": 0.237,
+      "step": 53960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.615169804522115e-05,
+      "loss": 0.2334,
+      "step": 53970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6147193946491312e-05,
+      "loss": 0.2373,
+      "step": 53980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6142689847761466e-05,
+      "loss": 0.2384,
+      "step": 53990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6138185749031617e-05,
+      "loss": 0.2376,
+      "step": 54000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6133681650301778e-05,
+      "loss": 0.2353,
+      "step": 54010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6129177551571932e-05,
+      "loss": 0.2345,
+      "step": 54020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6124673452842086e-05,
+      "loss": 0.2346,
+      "step": 54030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6120169354112244e-05,
+      "loss": 0.2363,
+      "step": 54040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6115665255382398e-05,
+      "loss": 0.2334,
+      "step": 54050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6111161156652552e-05,
+      "loss": 0.2331,
+      "step": 54060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6106657057922713e-05,
+      "loss": 0.2342,
+      "step": 54070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6102152959192867e-05,
+      "loss": 0.2369,
+      "step": 54080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6097648860463024e-05,
+      "loss": 0.2329,
+      "step": 54090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.609314476173318e-05,
+      "loss": 0.2387,
+      "step": 54100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6088640663003332e-05,
+      "loss": 0.2386,
+      "step": 54110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6084136564273493e-05,
+      "loss": 0.2387,
+      "step": 54120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6079632465543647e-05,
+      "loss": 0.2367,
+      "step": 54130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.60751283668138e-05,
+      "loss": 0.2412,
+      "step": 54140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.607062426808396e-05,
+      "loss": 0.2393,
+      "step": 54150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6066120169354113e-05,
+      "loss": 0.2413,
+      "step": 54160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6061616070624267e-05,
+      "loss": 0.2438,
+      "step": 54170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6057111971894428e-05,
+      "loss": 0.239,
+      "step": 54180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6052607873164582e-05,
+      "loss": 0.2395,
+      "step": 54190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6048103774434733e-05,
+      "loss": 0.2372,
+      "step": 54200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6043599675704894e-05,
+      "loss": 0.2456,
+      "step": 54210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6039095576975048e-05,
+      "loss": 0.2383,
+      "step": 54220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6034591478245202e-05,
+      "loss": 0.2386,
+      "step": 54230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.603008737951536e-05,
+      "loss": 0.2366,
+      "step": 54240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6025583280785514e-05,
+      "loss": 0.238,
+      "step": 54250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6021079182055674e-05,
+      "loss": 0.2428,
+      "step": 54260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.601657508332583e-05,
+      "loss": 0.2346,
+      "step": 54270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6012070984595983e-05,
+      "loss": 0.2315,
+      "step": 54280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.600756688586614e-05,
+      "loss": 0.2337,
+      "step": 54290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.6003062787136294e-05,
+      "loss": 0.2443,
+      "step": 54300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.599855868840645e-05,
+      "loss": 0.2374,
+      "step": 54310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.599405458967661e-05,
+      "loss": 0.2413,
+      "step": 54320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5989550490946763e-05,
+      "loss": 0.2386,
+      "step": 54330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5985046392216917e-05,
+      "loss": 0.2321,
+      "step": 54340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5980542293487075e-05,
+      "loss": 0.2301,
+      "step": 54350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.597603819475723e-05,
+      "loss": 0.2381,
+      "step": 54360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5971534096027383e-05,
+      "loss": 0.2375,
+      "step": 54370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5967029997297544e-05,
+      "loss": 0.2411,
+      "step": 54380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5962525898567698e-05,
+      "loss": 0.2361,
+      "step": 54390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5958021799837856e-05,
+      "loss": 0.2347,
+      "step": 54400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.595351770110801e-05,
+      "loss": 0.2364,
+      "step": 54410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5949013602378164e-05,
+      "loss": 0.2353,
+      "step": 54420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5944509503648325e-05,
+      "loss": 0.2433,
+      "step": 54430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5940005404918475e-05,
+      "loss": 0.2431,
+      "step": 54440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.593550130618863e-05,
+      "loss": 0.2371,
+      "step": 54450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.593099720745879e-05,
+      "loss": 0.2424,
+      "step": 54460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5926493108728945e-05,
+      "loss": 0.2342,
+      "step": 54470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.59219890099991e-05,
+      "loss": 0.2407,
+      "step": 54480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5917484911269256e-05,
+      "loss": 0.2364,
+      "step": 54490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.591298081253941e-05,
+      "loss": 0.2364,
+      "step": 54500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5908476713809564e-05,
+      "loss": 0.2355,
+      "step": 54510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5903972615079725e-05,
+      "loss": 0.2373,
+      "step": 54520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.589946851634988e-05,
+      "loss": 0.2377,
+      "step": 54530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5894964417620037e-05,
+      "loss": 0.2361,
+      "step": 54540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.589046031889019e-05,
+      "loss": 0.2391,
+      "step": 54550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5885956220160345e-05,
+      "loss": 0.2391,
+      "step": 54560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5881452121430506e-05,
+      "loss": 0.236,
+      "step": 54570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.587694802270066e-05,
+      "loss": 0.238,
+      "step": 54580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5872443923970814e-05,
+      "loss": 0.2388,
+      "step": 54590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.586793982524097e-05,
+      "loss": 0.237,
+      "step": 54600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5863435726511126e-05,
+      "loss": 0.2362,
+      "step": 54610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.585893162778128e-05,
+      "loss": 0.2384,
+      "step": 54620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.585442752905144e-05,
+      "loss": 0.2344,
+      "step": 54630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5849923430321595e-05,
+      "loss": 0.2368,
+      "step": 54640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5845419331591746e-05,
+      "loss": 0.2426,
+      "step": 54650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5840915232861906e-05,
+      "loss": 0.2394,
+      "step": 54660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.583641113413206e-05,
+      "loss": 0.236,
+      "step": 54670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5831907035402218e-05,
+      "loss": 0.2384,
+      "step": 54680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5827402936672372e-05,
+      "loss": 0.2351,
+      "step": 54690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5822898837942526e-05,
+      "loss": 0.2415,
+      "step": 54700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5818394739212687e-05,
+      "loss": 0.239,
+      "step": 54710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.581389064048284e-05,
+      "loss": 0.2367,
+      "step": 54720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5809386541752995e-05,
+      "loss": 0.2345,
+      "step": 54730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5804882443023153e-05,
+      "loss": 0.2384,
+      "step": 54740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5800378344293307e-05,
+      "loss": 0.2345,
+      "step": 54750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.579587424556346e-05,
+      "loss": 0.2363,
+      "step": 54760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5791370146833622e-05,
+      "loss": 0.2367,
+      "step": 54770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5786866048103776e-05,
+      "loss": 0.239,
+      "step": 54780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.578236194937393e-05,
+      "loss": 0.2374,
+      "step": 54790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5777857850644088e-05,
+      "loss": 0.2384,
+      "step": 54800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.577335375191424e-05,
+      "loss": 0.2393,
+      "step": 54810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5768849653184403e-05,
+      "loss": 0.2341,
+      "step": 54820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5764345554454557e-05,
+      "loss": 0.2343,
+      "step": 54830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.575984145572471e-05,
+      "loss": 0.2383,
+      "step": 54840
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5755337356994868e-05,
+      "loss": 0.2335,
+      "step": 54850
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5750833258265022e-05,
+      "loss": 0.2345,
+      "step": 54860
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5746329159535176e-05,
+      "loss": 0.2412,
+      "step": 54870
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5741825060805337e-05,
+      "loss": 0.2378,
+      "step": 54880
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5737320962075488e-05,
+      "loss": 0.2325,
+      "step": 54890
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5732816863345642e-05,
+      "loss": 0.2403,
+      "step": 54900
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5728312764615803e-05,
+      "loss": 0.2402,
+      "step": 54910
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5723808665885957e-05,
+      "loss": 0.2351,
+      "step": 54920
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.571930456715611e-05,
+      "loss": 0.2435,
+      "step": 54930
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.571480046842627e-05,
+      "loss": 0.2366,
+      "step": 54940
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5710296369696423e-05,
+      "loss": 0.2401,
+      "step": 54950
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5705792270966584e-05,
+      "loss": 0.2353,
+      "step": 54960
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5701288172236738e-05,
+      "loss": 0.236,
+      "step": 54970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5696784073506892e-05,
+      "loss": 0.2374,
+      "step": 54980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.569227997477705e-05,
+      "loss": 0.2401,
+      "step": 54990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5687775876047204e-05,
+      "loss": 0.2339,
+      "step": 55000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5683271777317358e-05,
+      "loss": 0.2373,
+      "step": 55010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.567876767858752e-05,
+      "loss": 0.2341,
+      "step": 55020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5674263579857673e-05,
+      "loss": 0.2293,
+      "step": 55030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5669759481127827e-05,
+      "loss": 0.2342,
+      "step": 55040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5665255382397984e-05,
+      "loss": 0.2331,
+      "step": 55050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.566075128366814e-05,
+      "loss": 0.233,
+      "step": 55060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5656247184938292e-05,
+      "loss": 0.2417,
+      "step": 55070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5651743086208453e-05,
+      "loss": 0.2379,
+      "step": 55080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5647238987478604e-05,
+      "loss": 0.2314,
+      "step": 55090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5642734888748765e-05,
+      "loss": 0.2355,
+      "step": 55100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.563823079001892e-05,
+      "loss": 0.2316,
+      "step": 55110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5633726691289073e-05,
+      "loss": 0.2312,
+      "step": 55120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.562922259255923e-05,
+      "loss": 0.2346,
+      "step": 55130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5624718493829385e-05,
+      "loss": 0.238,
+      "step": 55140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.562021439509954e-05,
+      "loss": 0.2398,
+      "step": 55150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.56157102963697e-05,
+      "loss": 0.233,
+      "step": 55160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5611206197639854e-05,
+      "loss": 0.2318,
+      "step": 55170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5606702098910008e-05,
+      "loss": 0.2343,
+      "step": 55180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5602198000180165e-05,
+      "loss": 0.2343,
+      "step": 55190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.559769390145032e-05,
+      "loss": 0.2296,
+      "step": 55200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5593189802720474e-05,
+      "loss": 0.2325,
+      "step": 55210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5588685703990634e-05,
+      "loss": 0.239,
+      "step": 55220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.558418160526079e-05,
+      "loss": 0.2397,
+      "step": 55230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5579677506530946e-05,
+      "loss": 0.2361,
+      "step": 55240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.55751734078011e-05,
+      "loss": 0.2397,
+      "step": 55250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5570669309071254e-05,
+      "loss": 0.2372,
+      "step": 55260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5566165210341415e-05,
+      "loss": 0.2342,
+      "step": 55270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.556166111161157e-05,
+      "loss": 0.2422,
+      "step": 55280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5557157012881723e-05,
+      "loss": 0.2324,
+      "step": 55290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.555265291415188e-05,
+      "loss": 0.2342,
+      "step": 55300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5548148815422035e-05,
+      "loss": 0.2333,
+      "step": 55310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.554364471669219e-05,
+      "loss": 0.2325,
+      "step": 55320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5539140617962347e-05,
+      "loss": 0.2298,
+      "step": 55330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.55346365192325e-05,
+      "loss": 0.2328,
+      "step": 55340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5530132420502655e-05,
+      "loss": 0.2351,
+      "step": 55350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5525628321772816e-05,
+      "loss": 0.2348,
+      "step": 55360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.552112422304297e-05,
+      "loss": 0.2396,
+      "step": 55370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5516620124313127e-05,
+      "loss": 0.2336,
+      "step": 55380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.551211602558328e-05,
+      "loss": 0.2386,
+      "step": 55390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5507611926853435e-05,
+      "loss": 0.2337,
+      "step": 55400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5503107828123596e-05,
+      "loss": 0.2318,
+      "step": 55410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.549860372939375e-05,
+      "loss": 0.2428,
+      "step": 55420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5494099630663905e-05,
+      "loss": 0.2388,
+      "step": 55430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.5489595531934062e-05,
+      "loss": 0.2388,
+      "step": 55440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5485091433204216e-05,
+      "loss": 0.2395,
+      "step": 55450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.548058733447437e-05,
+      "loss": 0.2369,
+      "step": 55460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.547608323574453e-05,
+      "loss": 0.237,
+      "step": 55470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5471579137014685e-05,
+      "loss": 0.2348,
+      "step": 55480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.546707503828484e-05,
+      "loss": 0.2439,
+      "step": 55490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5462570939554997e-05,
+      "loss": 0.2358,
+      "step": 55500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.545806684082515e-05,
+      "loss": 0.2353,
+      "step": 55510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5453562742095312e-05,
+      "loss": 0.2354,
+      "step": 55520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5449058643365466e-05,
+      "loss": 0.24,
+      "step": 55530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5444554544635617e-05,
+      "loss": 0.2387,
+      "step": 55540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5440050445905778e-05,
+      "loss": 0.2355,
+      "step": 55550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.543554634717593e-05,
+      "loss": 0.2383,
+      "step": 55560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5431042248446086e-05,
+      "loss": 0.2348,
+      "step": 55570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5426538149716243e-05,
+      "loss": 0.2386,
+      "step": 55580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5422034050986397e-05,
+      "loss": 0.2376,
+      "step": 55590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.541752995225655e-05,
+      "loss": 0.2358,
+      "step": 55600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5413025853526712e-05,
+      "loss": 0.236,
+      "step": 55610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5408521754796866e-05,
+      "loss": 0.2368,
+      "step": 55620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.540401765606702e-05,
+      "loss": 0.2334,
+      "step": 55630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5399513557337178e-05,
+      "loss": 0.2367,
+      "step": 55640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5395009458607332e-05,
+      "loss": 0.2333,
+      "step": 55650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5390505359877493e-05,
+      "loss": 0.2418,
+      "step": 55660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5386001261147647e-05,
+      "loss": 0.2375,
+      "step": 55670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.53814971624178e-05,
+      "loss": 0.2308,
+      "step": 55680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.537699306368796e-05,
+      "loss": 0.2386,
+      "step": 55690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5372488964958113e-05,
+      "loss": 0.2306,
+      "step": 55700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5367984866228267e-05,
+      "loss": 0.2351,
+      "step": 55710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5363480767498428e-05,
+      "loss": 0.2342,
+      "step": 55720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5358976668768582e-05,
+      "loss": 0.2399,
+      "step": 55730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5354472570038733e-05,
+      "loss": 0.2355,
+      "step": 55740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5349968471308893e-05,
+      "loss": 0.2383,
+      "step": 55750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5345464372579048e-05,
+      "loss": 0.2373,
+      "step": 55760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.53409602738492e-05,
+      "loss": 0.2342,
+      "step": 55770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.533645617511936e-05,
+      "loss": 0.2312,
+      "step": 55780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5331952076389513e-05,
+      "loss": 0.2356,
+      "step": 55790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5327447977659674e-05,
+      "loss": 0.2362,
+      "step": 55800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5322943878929828e-05,
+      "loss": 0.2376,
+      "step": 55810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5318439780199982e-05,
+      "loss": 0.2349,
+      "step": 55820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.531393568147014e-05,
+      "loss": 0.2383,
+      "step": 55830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5309431582740294e-05,
+      "loss": 0.2326,
+      "step": 55840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5304927484010448e-05,
+      "loss": 0.2319,
+      "step": 55850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.530042338528061e-05,
+      "loss": 0.2418,
+      "step": 55860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5295919286550763e-05,
+      "loss": 0.2361,
+      "step": 55870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5291415187820917e-05,
+      "loss": 0.2325,
+      "step": 55880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5286911089091075e-05,
+      "loss": 0.2343,
+      "step": 55890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.528240699036123e-05,
+      "loss": 0.2359,
+      "step": 55900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5277902891631383e-05,
+      "loss": 0.2339,
+      "step": 55910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5273398792901544e-05,
+      "loss": 0.2308,
+      "step": 55920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5268894694171698e-05,
+      "loss": 0.2309,
+      "step": 55930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5264390595441855e-05,
+      "loss": 0.2408,
+      "step": 55940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.525988649671201e-05,
+      "loss": 0.2331,
+      "step": 55950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5255382397982164e-05,
+      "loss": 0.2384,
+      "step": 55960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5250878299252324e-05,
+      "loss": 0.2311,
+      "step": 55970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5246374200522475e-05,
+      "loss": 0.2345,
+      "step": 55980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.524187010179263e-05,
+      "loss": 0.233,
+      "step": 55990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.523736600306279e-05,
+      "loss": 0.2351,
+      "step": 56000
+    },
+    {
+      "epoch": 0.5,
+      "eval_NEIMS_canon_smiles": 0.9386833333333333,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.0307,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.4244121859887462,
+      "eval_NEIMS_exact_mols": 0.030566666666666666,
+      "eval_NEIMS_exact_smiles": 0.029716666666666666,
+      "eval_NEIMS_loss": 0.2517029643058777,
+      "eval_NEIMS_matched_formulas": 0.5088333333333334,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.3239689600499408,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.03075,
+      "eval_NEIMS_runtime": 713.1519,
+      "eval_NEIMS_samples_per_second": 84.134,
+      "eval_NEIMS_steps_per_second": 1.315,
+      "step": 56000
+    },
+    {
+      "epoch": 0.5,
+      "eval_RASSP_canon_smiles": 0.9284903329214946,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.07466524192740508,
+      "eval_RASSP_daylight_tanimoto_simil": 0.524916680553281,
+      "eval_RASSP_exact_mols": 0.07439810331585801,
+      "eval_RASSP_exact_smiles": 0.07197715964871272,
+      "eval_RASSP_loss": 0.19337597489356995,
+      "eval_RASSP_matched_formulas": 0.6685310715597556,
+      "eval_RASSP_morgan_tanimoto_simil": 0.3899155475340412,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.07453167262163155,
+      "eval_RASSP_runtime": 814.8474,
+      "eval_RASSP_samples_per_second": 73.503,
+      "eval_RASSP_steps_per_second": 1.149,
+      "step": 56000
+    },
+    {
+      "epoch": 0.5,
+      "eval_NIST_canon_smiles": 0.8920041168328778,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.009972672747276147,
+      "eval_NIST_daylight_tanimoto_simil": 0.26198516526580995,
+      "eval_NIST_exact_mols": 0.00958228342264968,
+      "eval_NIST_exact_smiles": 0.00890797458920396,
+      "eval_NIST_loss": 1.4831839799880981,
+      "eval_NIST_matched_formulas": 0.069631259537921,
+      "eval_NIST_morgan_tanimoto_simil": 0.21832936310122283,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.009866202931468929,
+      "eval_NIST_runtime": 465.4465,
+      "eval_NIST_samples_per_second": 60.538,
+      "eval_NIST_steps_per_second": 0.947,
+      "step": 56000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5232861904332944e-05,
+      "loss": 0.2325,
+      "step": 56010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.52283578056031e-05,
+      "loss": 0.2358,
+      "step": 56020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5223853706873256e-05,
+      "loss": 0.2346,
+      "step": 56030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.521934960814341e-05,
+      "loss": 0.2399,
+      "step": 56040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5214845509413564e-05,
+      "loss": 0.2371,
+      "step": 56050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5210341410683725e-05,
+      "loss": 0.239,
+      "step": 56060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.520583731195388e-05,
+      "loss": 0.2405,
+      "step": 56070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5201333213224037e-05,
+      "loss": 0.2391,
+      "step": 56080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.519682911449419e-05,
+      "loss": 0.2389,
+      "step": 56090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5192325015764345e-05,
+      "loss": 0.2392,
+      "step": 56100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5188271326907485e-05,
+      "loss": 0.2382,
+      "step": 56110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.518376722817764e-05,
+      "loss": 0.24,
+      "step": 56120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.51792631294478e-05,
+      "loss": 0.2312,
+      "step": 56130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5174759030717954e-05,
+      "loss": 0.2368,
+      "step": 56140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.517025493198811e-05,
+      "loss": 0.2377,
+      "step": 56150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5165750833258266e-05,
+      "loss": 0.2339,
+      "step": 56160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.516124673452842e-05,
+      "loss": 0.2295,
+      "step": 56170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.515674263579858e-05,
+      "loss": 0.2329,
+      "step": 56180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5152238537068735e-05,
+      "loss": 0.2396,
+      "step": 56190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.514773443833889e-05,
+      "loss": 0.2324,
+      "step": 56200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5143230339609046e-05,
+      "loss": 0.2348,
+      "step": 56210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.51387262408792e-05,
+      "loss": 0.2332,
+      "step": 56220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5134222142149355e-05,
+      "loss": 0.2374,
+      "step": 56230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5129718043419516e-05,
+      "loss": 0.2364,
+      "step": 56240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.512521394468967e-05,
+      "loss": 0.2336,
+      "step": 56250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5120709845959824e-05,
+      "loss": 0.2349,
+      "step": 56260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.511620574722998e-05,
+      "loss": 0.2343,
+      "step": 56270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5111701648500135e-05,
+      "loss": 0.2376,
+      "step": 56280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5107197549770296e-05,
+      "loss": 0.2366,
+      "step": 56290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.510269345104045e-05,
+      "loss": 0.2299,
+      "step": 56300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.50981893523106e-05,
+      "loss": 0.2397,
+      "step": 56310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5093685253580762e-05,
+      "loss": 0.2346,
+      "step": 56320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5089181154850916e-05,
+      "loss": 0.2348,
+      "step": 56330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.508467705612107e-05,
+      "loss": 0.2372,
+      "step": 56340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5080172957391228e-05,
+      "loss": 0.2387,
+      "step": 56350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5075668858661382e-05,
+      "loss": 0.2336,
+      "step": 56360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5071164759931536e-05,
+      "loss": 0.2378,
+      "step": 56370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5066660661201697e-05,
+      "loss": 0.2327,
+      "step": 56380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.506215656247185e-05,
+      "loss": 0.2324,
+      "step": 56390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5057652463742005e-05,
+      "loss": 0.2315,
+      "step": 56400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5053148365012162e-05,
+      "loss": 0.2331,
+      "step": 56410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5048644266282317e-05,
+      "loss": 0.2386,
+      "step": 56420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5044140167552477e-05,
+      "loss": 0.2319,
+      "step": 56430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.503963606882263e-05,
+      "loss": 0.2405,
+      "step": 56440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5035131970092786e-05,
+      "loss": 0.2346,
+      "step": 56450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5030627871362943e-05,
+      "loss": 0.239,
+      "step": 56460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5026123772633097e-05,
+      "loss": 0.2315,
+      "step": 56470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.502161967390325e-05,
+      "loss": 0.2314,
+      "step": 56480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5017115575173412e-05,
+      "loss": 0.2306,
+      "step": 56490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5012611476443566e-05,
+      "loss": 0.2336,
+      "step": 56500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5008107377713717e-05,
+      "loss": 0.2337,
+      "step": 56510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5003603278983878e-05,
+      "loss": 0.2388,
+      "step": 56520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4999099180254032e-05,
+      "loss": 0.2364,
+      "step": 56530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.499459508152419e-05,
+      "loss": 0.2387,
+      "step": 56540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4990090982794344e-05,
+      "loss": 0.2433,
+      "step": 56550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4985586884064498e-05,
+      "loss": 0.2341,
+      "step": 56560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4981082785334655e-05,
+      "loss": 0.2343,
+      "step": 56570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4976578686604813e-05,
+      "loss": 0.2307,
+      "step": 56580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4972074587874967e-05,
+      "loss": 0.2331,
+      "step": 56590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4967570489145124e-05,
+      "loss": 0.2349,
+      "step": 56600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.496306639041528e-05,
+      "loss": 0.2388,
+      "step": 56610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4958562291685436e-05,
+      "loss": 0.2312,
+      "step": 56620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.495405819295559e-05,
+      "loss": 0.2423,
+      "step": 56630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4949554094225747e-05,
+      "loss": 0.2348,
+      "step": 56640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.49450499954959e-05,
+      "loss": 0.2362,
+      "step": 56650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4940545896766056e-05,
+      "loss": 0.2307,
+      "step": 56660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4936041798036213e-05,
+      "loss": 0.2281,
+      "step": 56670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.493153769930637e-05,
+      "loss": 0.2344,
+      "step": 56680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4927033600576528e-05,
+      "loss": 0.2341,
+      "step": 56690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4922529501846682e-05,
+      "loss": 0.2291,
+      "step": 56700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4918025403116836e-05,
+      "loss": 0.2341,
+      "step": 56710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4913521304386994e-05,
+      "loss": 0.2323,
+      "step": 56720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4909017205657148e-05,
+      "loss": 0.2315,
+      "step": 56730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4904513106927305e-05,
+      "loss": 0.2355,
+      "step": 56740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4900009008197463e-05,
+      "loss": 0.2357,
+      "step": 56750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4895504909467617e-05,
+      "loss": 0.2295,
+      "step": 56760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.489100081073777e-05,
+      "loss": 0.237,
+      "step": 56770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.488649671200793e-05,
+      "loss": 0.2307,
+      "step": 56780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4881992613278086e-05,
+      "loss": 0.2285,
+      "step": 56790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.487748851454824e-05,
+      "loss": 0.241,
+      "step": 56800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4872984415818394e-05,
+      "loss": 0.2333,
+      "step": 56810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4868480317088552e-05,
+      "loss": 0.2307,
+      "step": 56820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.486397621835871e-05,
+      "loss": 0.229,
+      "step": 56830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4859472119628863e-05,
+      "loss": 0.2321,
+      "step": 56840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.485496802089902e-05,
+      "loss": 0.2366,
+      "step": 56850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4850463922169175e-05,
+      "loss": 0.2377,
+      "step": 56860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.484595982343933e-05,
+      "loss": 0.236,
+      "step": 56870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4841455724709487e-05,
+      "loss": 0.2339,
+      "step": 56880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4836951625979644e-05,
+      "loss": 0.2277,
+      "step": 56890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4832447527249798e-05,
+      "loss": 0.2309,
+      "step": 56900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4827943428519952e-05,
+      "loss": 0.2367,
+      "step": 56910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.482343932979011e-05,
+      "loss": 0.2382,
+      "step": 56920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4818935231060267e-05,
+      "loss": 0.2414,
+      "step": 56930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.481443113233042e-05,
+      "loss": 0.2368,
+      "step": 56940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.480992703360058e-05,
+      "loss": 0.2402,
+      "step": 56950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4805422934870733e-05,
+      "loss": 0.234,
+      "step": 56960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.480091883614089e-05,
+      "loss": 0.2295,
+      "step": 56970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4796414737411045e-05,
+      "loss": 0.2358,
+      "step": 56980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4791910638681202e-05,
+      "loss": 0.2292,
+      "step": 56990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4787406539951356e-05,
+      "loss": 0.2303,
+      "step": 57000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.478290244122151e-05,
+      "loss": 0.2303,
+      "step": 57010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4778398342491668e-05,
+      "loss": 0.2362,
+      "step": 57020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4773894243761825e-05,
+      "loss": 0.2337,
+      "step": 57030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4769390145031983e-05,
+      "loss": 0.2319,
+      "step": 57040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4764886046302137e-05,
+      "loss": 0.2301,
+      "step": 57050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.476038194757229e-05,
+      "loss": 0.2342,
+      "step": 57060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.475587784884245e-05,
+      "loss": 0.2375,
+      "step": 57070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4751373750112603e-05,
+      "loss": 0.2322,
+      "step": 57080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.474686965138276e-05,
+      "loss": 0.2373,
+      "step": 57090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4742365552652914e-05,
+      "loss": 0.2328,
+      "step": 57100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4737861453923072e-05,
+      "loss": 0.2302,
+      "step": 57110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4733357355193226e-05,
+      "loss": 0.2304,
+      "step": 57120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4728853256463383e-05,
+      "loss": 0.2295,
+      "step": 57130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.472434915773354e-05,
+      "loss": 0.2377,
+      "step": 57140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4719845059003695e-05,
+      "loss": 0.2329,
+      "step": 57150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.471534096027385e-05,
+      "loss": 0.2327,
+      "step": 57160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4710836861544006e-05,
+      "loss": 0.2333,
+      "step": 57170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4706332762814164e-05,
+      "loss": 0.2263,
+      "step": 57180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4701828664084318e-05,
+      "loss": 0.2281,
+      "step": 57190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4697324565354472e-05,
+      "loss": 0.2297,
+      "step": 57200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.469282046662463e-05,
+      "loss": 0.2297,
+      "step": 57210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4688316367894784e-05,
+      "loss": 0.2371,
+      "step": 57220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.468381226916494e-05,
+      "loss": 0.2348,
+      "step": 57230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.46793081704351e-05,
+      "loss": 0.2267,
+      "step": 57240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4674804071705253e-05,
+      "loss": 0.2358,
+      "step": 57250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4670299972975407e-05,
+      "loss": 0.2337,
+      "step": 57260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4665795874245564e-05,
+      "loss": 0.235,
+      "step": 57270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4661291775515722e-05,
+      "loss": 0.234,
+      "step": 57280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4656787676785876e-05,
+      "loss": 0.2347,
+      "step": 57290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.465228357805603e-05,
+      "loss": 0.2381,
+      "step": 57300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4647779479326188e-05,
+      "loss": 0.2268,
+      "step": 57310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4643275380596345e-05,
+      "loss": 0.2366,
+      "step": 57320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.46387712818665e-05,
+      "loss": 0.232,
+      "step": 57330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4634267183136657e-05,
+      "loss": 0.2359,
+      "step": 57340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.462976308440681e-05,
+      "loss": 0.231,
+      "step": 57350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4625258985676965e-05,
+      "loss": 0.2285,
+      "step": 57360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4620754886947122e-05,
+      "loss": 0.2369,
+      "step": 57370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.461625078821728e-05,
+      "loss": 0.2353,
+      "step": 57380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4611746689487437e-05,
+      "loss": 0.2285,
+      "step": 57390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.460724259075759e-05,
+      "loss": 0.2294,
+      "step": 57400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4602738492027746e-05,
+      "loss": 0.229,
+      "step": 57410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4598234393297903e-05,
+      "loss": 0.2272,
+      "step": 57420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4593730294568057e-05,
+      "loss": 0.2402,
+      "step": 57430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4589226195838215e-05,
+      "loss": 0.2276,
+      "step": 57440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.458472209710837e-05,
+      "loss": 0.2383,
+      "step": 57450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4580217998378526e-05,
+      "loss": 0.2363,
+      "step": 57460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.457571389964868e-05,
+      "loss": 0.2347,
+      "step": 57470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4571209800918838e-05,
+      "loss": 0.2323,
+      "step": 57480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4566705702188995e-05,
+      "loss": 0.2306,
+      "step": 57490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.456220160345915e-05,
+      "loss": 0.2344,
+      "step": 57500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4557697504729304e-05,
+      "loss": 0.2379,
+      "step": 57510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.455319340599946e-05,
+      "loss": 0.2337,
+      "step": 57520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.454868930726962e-05,
+      "loss": 0.2317,
+      "step": 57530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4544185208539773e-05,
+      "loss": 0.2298,
+      "step": 57540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4539681109809927e-05,
+      "loss": 0.2341,
+      "step": 57550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4535177011080084e-05,
+      "loss": 0.2374,
+      "step": 57560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.453067291235024e-05,
+      "loss": 0.2273,
+      "step": 57570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4526168813620396e-05,
+      "loss": 0.2321,
+      "step": 57580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4521664714890553e-05,
+      "loss": 0.2311,
+      "step": 57590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4517160616160707e-05,
+      "loss": 0.2281,
+      "step": 57600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.451265651743086e-05,
+      "loss": 0.2358,
+      "step": 57610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.450815241870102e-05,
+      "loss": 0.2319,
+      "step": 57620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4503648319971177e-05,
+      "loss": 0.2359,
+      "step": 57630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.449914422124133e-05,
+      "loss": 0.2363,
+      "step": 57640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4494640122511485e-05,
+      "loss": 0.2313,
+      "step": 57650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4490136023781642e-05,
+      "loss": 0.2283,
+      "step": 57660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.44856319250518e-05,
+      "loss": 0.2366,
+      "step": 57670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4481127826321954e-05,
+      "loss": 0.2348,
+      "step": 57680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.447662372759211e-05,
+      "loss": 0.2312,
+      "step": 57690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4472119628862265e-05,
+      "loss": 0.2283,
+      "step": 57700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.446761553013242e-05,
+      "loss": 0.2288,
+      "step": 57710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4463111431402577e-05,
+      "loss": 0.2367,
+      "step": 57720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4458607332672735e-05,
+      "loss": 0.2355,
+      "step": 57730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4454103233942892e-05,
+      "loss": 0.231,
+      "step": 57740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4449599135213043e-05,
+      "loss": 0.2328,
+      "step": 57750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.44450950364832e-05,
+      "loss": 0.2346,
+      "step": 57760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4440590937753358e-05,
+      "loss": 0.2272,
+      "step": 57770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4436086839023512e-05,
+      "loss": 0.2331,
+      "step": 57780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.443158274029367e-05,
+      "loss": 0.2283,
+      "step": 57790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4427078641563823e-05,
+      "loss": 0.2332,
+      "step": 57800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.442257454283398e-05,
+      "loss": 0.2315,
+      "step": 57810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4418070444104135e-05,
+      "loss": 0.2313,
+      "step": 57820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4413566345374293e-05,
+      "loss": 0.233,
+      "step": 57830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.440906224664445e-05,
+      "loss": 0.2286,
+      "step": 57840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.44045581479146e-05,
+      "loss": 0.235,
+      "step": 57850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4400054049184758e-05,
+      "loss": 0.233,
+      "step": 57860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4395549950454916e-05,
+      "loss": 0.2392,
+      "step": 57870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4391045851725073e-05,
+      "loss": 0.234,
+      "step": 57880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4386541752995227e-05,
+      "loss": 0.2314,
+      "step": 57890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.438203765426538e-05,
+      "loss": 0.228,
+      "step": 57900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.437753355553554e-05,
+      "loss": 0.238,
+      "step": 57910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4373029456805693e-05,
+      "loss": 0.232,
+      "step": 57920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.436852535807585e-05,
+      "loss": 0.2302,
+      "step": 57930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4364021259346008e-05,
+      "loss": 0.2325,
+      "step": 57940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4359517160616162e-05,
+      "loss": 0.2343,
+      "step": 57950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4355013061886316e-05,
+      "loss": 0.2274,
+      "step": 57960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4350508963156474e-05,
+      "loss": 0.2337,
+      "step": 57970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.434600486442663e-05,
+      "loss": 0.2337,
+      "step": 57980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4341500765696785e-05,
+      "loss": 0.227,
+      "step": 57990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.433699666696694e-05,
+      "loss": 0.2301,
+      "step": 58000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4332492568237097e-05,
+      "loss": 0.2306,
+      "step": 58010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4327988469507254e-05,
+      "loss": 0.2339,
+      "step": 58020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.432348437077741e-05,
+      "loss": 0.2328,
+      "step": 58030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4318980272047566e-05,
+      "loss": 0.2264,
+      "step": 58040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.431447617331772e-05,
+      "loss": 0.2316,
+      "step": 58050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4309972074587874e-05,
+      "loss": 0.229,
+      "step": 58060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4305467975858032e-05,
+      "loss": 0.2345,
+      "step": 58070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.430096387712819e-05,
+      "loss": 0.2381,
+      "step": 58080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4296459778398343e-05,
+      "loss": 0.237,
+      "step": 58090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4291955679668497e-05,
+      "loss": 0.232,
+      "step": 58100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4287451580938655e-05,
+      "loss": 0.2275,
+      "step": 58110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4282947482208812e-05,
+      "loss": 0.2328,
+      "step": 58120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4278443383478966e-05,
+      "loss": 0.2326,
+      "step": 58130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4273939284749124e-05,
+      "loss": 0.2267,
+      "step": 58140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4269435186019278e-05,
+      "loss": 0.2404,
+      "step": 58150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4264931087289436e-05,
+      "loss": 0.2271,
+      "step": 58160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.426042698855959e-05,
+      "loss": 0.2329,
+      "step": 58170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4255922889829747e-05,
+      "loss": 0.2328,
+      "step": 58180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4251418791099905e-05,
+      "loss": 0.2275,
+      "step": 58190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4246914692370055e-05,
+      "loss": 0.2317,
+      "step": 58200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.42428610035132e-05,
+      "loss": 0.2408,
+      "step": 58210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4238356904783353e-05,
+      "loss": 0.2254,
+      "step": 58220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4233852806053507e-05,
+      "loss": 0.2297,
+      "step": 58230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4229348707323665e-05,
+      "loss": 0.2272,
+      "step": 58240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4224844608593822e-05,
+      "loss": 0.2295,
+      "step": 58250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.422034050986398e-05,
+      "loss": 0.2314,
+      "step": 58260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4215836411134134e-05,
+      "loss": 0.2268,
+      "step": 58270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4211332312404288e-05,
+      "loss": 0.231,
+      "step": 58280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4206828213674446e-05,
+      "loss": 0.23,
+      "step": 58290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.42023241149446e-05,
+      "loss": 0.2334,
+      "step": 58300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4197820016214757e-05,
+      "loss": 0.2328,
+      "step": 58310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.419331591748491e-05,
+      "loss": 0.2255,
+      "step": 58320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.418881181875507e-05,
+      "loss": 0.2362,
+      "step": 58330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4184307720025223e-05,
+      "loss": 0.2284,
+      "step": 58340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.417980362129538e-05,
+      "loss": 0.232,
+      "step": 58350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4175299522565538e-05,
+      "loss": 0.2322,
+      "step": 58360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4170795423835692e-05,
+      "loss": 0.228,
+      "step": 58370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4166291325105846e-05,
+      "loss": 0.2304,
+      "step": 58380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4161787226376004e-05,
+      "loss": 0.2314,
+      "step": 58390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.415728312764616e-05,
+      "loss": 0.2329,
+      "step": 58400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4152779028916315e-05,
+      "loss": 0.2292,
+      "step": 58410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.414827493018647e-05,
+      "loss": 0.2382,
+      "step": 58420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4143770831456627e-05,
+      "loss": 0.2335,
+      "step": 58430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.413926673272678e-05,
+      "loss": 0.2257,
+      "step": 58440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.413476263399694e-05,
+      "loss": 0.2302,
+      "step": 58450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4130258535267096e-05,
+      "loss": 0.2275,
+      "step": 58460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.412575443653725e-05,
+      "loss": 0.2319,
+      "step": 58470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4121250337807404e-05,
+      "loss": 0.2332,
+      "step": 58480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.411674623907756e-05,
+      "loss": 0.2381,
+      "step": 58490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.411224214034772e-05,
+      "loss": 0.2322,
+      "step": 58500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4107738041617873e-05,
+      "loss": 0.234,
+      "step": 58510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4103233942888027e-05,
+      "loss": 0.2353,
+      "step": 58520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4098729844158185e-05,
+      "loss": 0.2312,
+      "step": 58530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4094225745428342e-05,
+      "loss": 0.2378,
+      "step": 58540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4089721646698496e-05,
+      "loss": 0.2332,
+      "step": 58550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4085217547968654e-05,
+      "loss": 0.2301,
+      "step": 58560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4080713449238808e-05,
+      "loss": 0.227,
+      "step": 58570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4076209350508962e-05,
+      "loss": 0.2372,
+      "step": 58580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.407170525177912e-05,
+      "loss": 0.236,
+      "step": 58590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4067201153049277e-05,
+      "loss": 0.2318,
+      "step": 58600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4062697054319434e-05,
+      "loss": 0.2275,
+      "step": 58610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4058192955589585e-05,
+      "loss": 0.2293,
+      "step": 58620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4053688856859743e-05,
+      "loss": 0.2316,
+      "step": 58630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.40491847581299e-05,
+      "loss": 0.2237,
+      "step": 58640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4044680659400054e-05,
+      "loss": 0.2323,
+      "step": 58650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4040176560670212e-05,
+      "loss": 0.2291,
+      "step": 58660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4035672461940366e-05,
+      "loss": 0.2271,
+      "step": 58670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4031168363210523e-05,
+      "loss": 0.2315,
+      "step": 58680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4026664264480677e-05,
+      "loss": 0.2319,
+      "step": 58690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4022160165750835e-05,
+      "loss": 0.2394,
+      "step": 58700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4017656067020992e-05,
+      "loss": 0.2326,
+      "step": 58710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4013151968291147e-05,
+      "loss": 0.2258,
+      "step": 58720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.40086478695613e-05,
+      "loss": 0.2304,
+      "step": 58730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4004143770831458e-05,
+      "loss": 0.2307,
+      "step": 58740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3999639672101616e-05,
+      "loss": 0.2331,
+      "step": 58750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.399513557337177e-05,
+      "loss": 0.2242,
+      "step": 58760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3990631474641924e-05,
+      "loss": 0.2254,
+      "step": 58770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.398612737591208e-05,
+      "loss": 0.2367,
+      "step": 58780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3981623277182235e-05,
+      "loss": 0.2333,
+      "step": 58790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3977119178452393e-05,
+      "loss": 0.231,
+      "step": 58800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.397261507972255e-05,
+      "loss": 0.2329,
+      "step": 58810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3968110980992705e-05,
+      "loss": 0.2238,
+      "step": 58820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.396360688226286e-05,
+      "loss": 0.23,
+      "step": 58830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3959102783533016e-05,
+      "loss": 0.2312,
+      "step": 58840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3954598684803174e-05,
+      "loss": 0.2237,
+      "step": 58850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3950094586073328e-05,
+      "loss": 0.2304,
+      "step": 58860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3945590487343482e-05,
+      "loss": 0.2306,
+      "step": 58870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.394108638861364e-05,
+      "loss": 0.2325,
+      "step": 58880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3936582289883797e-05,
+      "loss": 0.2336,
+      "step": 58890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.393207819115395e-05,
+      "loss": 0.2303,
+      "step": 58900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.392757409242411e-05,
+      "loss": 0.2285,
+      "step": 58910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3923069993694263e-05,
+      "loss": 0.2304,
+      "step": 58920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3918565894964417e-05,
+      "loss": 0.2332,
+      "step": 58930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3914061796234574e-05,
+      "loss": 0.2371,
+      "step": 58940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.390955769750473e-05,
+      "loss": 0.2348,
+      "step": 58950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.390505359877489e-05,
+      "loss": 0.2353,
+      "step": 58960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.390054950004504e-05,
+      "loss": 0.2293,
+      "step": 58970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3896045401315197e-05,
+      "loss": 0.23,
+      "step": 58980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3891541302585355e-05,
+      "loss": 0.2268,
+      "step": 58990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.388703720385551e-05,
+      "loss": 0.2351,
+      "step": 59000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3882533105125666e-05,
+      "loss": 0.223,
+      "step": 59010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.387802900639582e-05,
+      "loss": 0.2306,
+      "step": 59020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3873524907665978e-05,
+      "loss": 0.2319,
+      "step": 59030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3869020808936132e-05,
+      "loss": 0.2264,
+      "step": 59040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.386451671020629e-05,
+      "loss": 0.2271,
+      "step": 59050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3860012611476447e-05,
+      "loss": 0.2284,
+      "step": 59060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3855508512746598e-05,
+      "loss": 0.2253,
+      "step": 59070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3851004414016755e-05,
+      "loss": 0.2301,
+      "step": 59080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3846500315286913e-05,
+      "loss": 0.2294,
+      "step": 59090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.384199621655707e-05,
+      "loss": 0.2275,
+      "step": 59100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3837492117827224e-05,
+      "loss": 0.2325,
+      "step": 59110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.383298801909738e-05,
+      "loss": 0.2283,
+      "step": 59120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3828483920367536e-05,
+      "loss": 0.2287,
+      "step": 59130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.382397982163769e-05,
+      "loss": 0.2271,
+      "step": 59140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3819475722907848e-05,
+      "loss": 0.2275,
+      "step": 59150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3814971624178005e-05,
+      "loss": 0.2305,
+      "step": 59160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.381046752544816e-05,
+      "loss": 0.2328,
+      "step": 59170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3805963426718313e-05,
+      "loss": 0.235,
+      "step": 59180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.380145932798847e-05,
+      "loss": 0.2405,
+      "step": 59190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3796955229258628e-05,
+      "loss": 0.2346,
+      "step": 59200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3792451130528782e-05,
+      "loss": 0.2316,
+      "step": 59210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3787947031798936e-05,
+      "loss": 0.2326,
+      "step": 59220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3783442933069094e-05,
+      "loss": 0.2343,
+      "step": 59230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.377893883433925e-05,
+      "loss": 0.2336,
+      "step": 59240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3774434735609406e-05,
+      "loss": 0.227,
+      "step": 59250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3769930636879563e-05,
+      "loss": 0.2329,
+      "step": 59260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3765426538149717e-05,
+      "loss": 0.2273,
+      "step": 59270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.376092243941987e-05,
+      "loss": 0.2321,
+      "step": 59280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.375641834069003e-05,
+      "loss": 0.2267,
+      "step": 59290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3751914241960186e-05,
+      "loss": 0.2317,
+      "step": 59300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.374741014323034e-05,
+      "loss": 0.2299,
+      "step": 59310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3742906044500494e-05,
+      "loss": 0.221,
+      "step": 59320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3738401945770652e-05,
+      "loss": 0.2346,
+      "step": 59330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.373389784704081e-05,
+      "loss": 0.2284,
+      "step": 59340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3729393748310964e-05,
+      "loss": 0.2339,
+      "step": 59350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.372488964958112e-05,
+      "loss": 0.227,
+      "step": 59360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3720385550851275e-05,
+      "loss": 0.2219,
+      "step": 59370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3715881452121433e-05,
+      "loss": 0.2254,
+      "step": 59380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3711377353391587e-05,
+      "loss": 0.2319,
+      "step": 59390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3706873254661744e-05,
+      "loss": 0.2274,
+      "step": 59400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.37023691559319e-05,
+      "loss": 0.2223,
+      "step": 59410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3697865057202052e-05,
+      "loss": 0.2271,
+      "step": 59420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.369336095847221e-05,
+      "loss": 0.2306,
+      "step": 59430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3688856859742367e-05,
+      "loss": 0.2302,
+      "step": 59440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3684352761012525e-05,
+      "loss": 0.2299,
+      "step": 59450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.367984866228268e-05,
+      "loss": 0.2316,
+      "step": 59460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3675344563552833e-05,
+      "loss": 0.2283,
+      "step": 59470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.367084046482299e-05,
+      "loss": 0.2299,
+      "step": 59480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3666336366093145e-05,
+      "loss": 0.2287,
+      "step": 59490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3661832267363302e-05,
+      "loss": 0.232,
+      "step": 59500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.365732816863346e-05,
+      "loss": 0.2334,
+      "step": 59510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3652824069903614e-05,
+      "loss": 0.2286,
+      "step": 59520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3648319971173768e-05,
+      "loss": 0.231,
+      "step": 59530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3643815872443925e-05,
+      "loss": 0.236,
+      "step": 59540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3639311773714083e-05,
+      "loss": 0.2355,
+      "step": 59550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3634807674984237e-05,
+      "loss": 0.2317,
+      "step": 59560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.363030357625439e-05,
+      "loss": 0.2321,
+      "step": 59570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.362579947752455e-05,
+      "loss": 0.2249,
+      "step": 59580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3621295378794706e-05,
+      "loss": 0.2276,
+      "step": 59590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.361679128006486e-05,
+      "loss": 0.2346,
+      "step": 59600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3612287181335018e-05,
+      "loss": 0.2249,
+      "step": 59610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3607783082605172e-05,
+      "loss": 0.2255,
+      "step": 59620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3603278983875326e-05,
+      "loss": 0.2308,
+      "step": 59630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3598774885145483e-05,
+      "loss": 0.2305,
+      "step": 59640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.359427078641564e-05,
+      "loss": 0.2278,
+      "step": 59650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3589766687685795e-05,
+      "loss": 0.2257,
+      "step": 59660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.358526258895595e-05,
+      "loss": 0.2254,
+      "step": 59670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3580758490226107e-05,
+      "loss": 0.2316,
+      "step": 59680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3576254391496264e-05,
+      "loss": 0.229,
+      "step": 59690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3571750292766418e-05,
+      "loss": 0.2308,
+      "step": 59700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3567246194036576e-05,
+      "loss": 0.2241,
+      "step": 59710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.356274209530673e-05,
+      "loss": 0.2354,
+      "step": 59720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3558237996576887e-05,
+      "loss": 0.2308,
+      "step": 59730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.355373389784704e-05,
+      "loss": 0.229,
+      "step": 59740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.35492297991172e-05,
+      "loss": 0.2247,
+      "step": 59750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3544725700387353e-05,
+      "loss": 0.2295,
+      "step": 59760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3540221601657507e-05,
+      "loss": 0.2277,
+      "step": 59770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3535717502927665e-05,
+      "loss": 0.2283,
+      "step": 59780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3531213404197822e-05,
+      "loss": 0.2246,
+      "step": 59790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.352670930546798e-05,
+      "loss": 0.2324,
+      "step": 59800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3522205206738134e-05,
+      "loss": 0.2294,
+      "step": 59810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3517701108008288e-05,
+      "loss": 0.225,
+      "step": 59820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3513197009278445e-05,
+      "loss": 0.2303,
+      "step": 59830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.35086929105486e-05,
+      "loss": 0.2275,
+      "step": 59840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3504188811818757e-05,
+      "loss": 0.2332,
+      "step": 59850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.349968471308891e-05,
+      "loss": 0.2239,
+      "step": 59860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.349518061435907e-05,
+      "loss": 0.2279,
+      "step": 59870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3490676515629223e-05,
+      "loss": 0.2348,
+      "step": 59880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.348617241689938e-05,
+      "loss": 0.228,
+      "step": 59890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3481668318169538e-05,
+      "loss": 0.2323,
+      "step": 59900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.347716421943969e-05,
+      "loss": 0.2321,
+      "step": 59910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3472660120709846e-05,
+      "loss": 0.2281,
+      "step": 59920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3468156021980003e-05,
+      "loss": 0.2273,
+      "step": 59930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.346365192325016e-05,
+      "loss": 0.2305,
+      "step": 59940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3459147824520315e-05,
+      "loss": 0.2261,
+      "step": 59950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.345464372579047e-05,
+      "loss": 0.2237,
+      "step": 59960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3450139627060626e-05,
+      "loss": 0.2318,
+      "step": 59970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.344563552833078e-05,
+      "loss": 0.2234,
+      "step": 59980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3441131429600938e-05,
+      "loss": 0.2281,
+      "step": 59990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3436627330871095e-05,
+      "loss": 0.2311,
+      "step": 60000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.343212323214125e-05,
+      "loss": 0.2289,
+      "step": 60010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3427619133411404e-05,
+      "loss": 0.2313,
+      "step": 60020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.342311503468156e-05,
+      "loss": 0.2346,
+      "step": 60030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.341861093595172e-05,
+      "loss": 0.2265,
+      "step": 60040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3414106837221873e-05,
+      "loss": 0.2293,
+      "step": 60050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3409602738492027e-05,
+      "loss": 0.2261,
+      "step": 60060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3405098639762184e-05,
+      "loss": 0.2287,
+      "step": 60070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3400594541032342e-05,
+      "loss": 0.2314,
+      "step": 60080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3396090442302496e-05,
+      "loss": 0.2292,
+      "step": 60090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3391586343572653e-05,
+      "loss": 0.232,
+      "step": 60100
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3387082244842808e-05,
+      "loss": 0.2301,
+      "step": 60110
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.338257814611296e-05,
+      "loss": 0.2302,
+      "step": 60120
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.337807404738312e-05,
+      "loss": 0.2319,
+      "step": 60130
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3373569948653277e-05,
+      "loss": 0.2268,
+      "step": 60140
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3369065849923434e-05,
+      "loss": 0.2303,
+      "step": 60150
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3364561751193588e-05,
+      "loss": 0.2286,
+      "step": 60160
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3360057652463742e-05,
+      "loss": 0.2344,
+      "step": 60170
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.33555535537339e-05,
+      "loss": 0.229,
+      "step": 60180
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3351049455004054e-05,
+      "loss": 0.2342,
+      "step": 60190
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.334654535627421e-05,
+      "loss": 0.2343,
+      "step": 60200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3342041257544366e-05,
+      "loss": 0.233,
+      "step": 60210
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3337537158814523e-05,
+      "loss": 0.2234,
+      "step": 60220
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3333033060084677e-05,
+      "loss": 0.2308,
+      "step": 60230
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3328528961354835e-05,
+      "loss": 0.233,
+      "step": 60240
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3324024862624992e-05,
+      "loss": 0.2285,
+      "step": 60250
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3319520763895146e-05,
+      "loss": 0.2309,
+      "step": 60260
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.33150166651653e-05,
+      "loss": 0.2309,
+      "step": 60270
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3310512566435458e-05,
+      "loss": 0.2315,
+      "step": 60280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3306008467705615e-05,
+      "loss": 0.2292,
+      "step": 60290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.330150436897577e-05,
+      "loss": 0.2324,
+      "step": 60300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3297000270245924e-05,
+      "loss": 0.2313,
+      "step": 60310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.329249617151608e-05,
+      "loss": 0.2288,
+      "step": 60320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3287992072786235e-05,
+      "loss": 0.2318,
+      "step": 60330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3283938383929376e-05,
+      "loss": 0.2316,
+      "step": 60340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3279434285199533e-05,
+      "loss": 0.224,
+      "step": 60350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3274930186469687e-05,
+      "loss": 0.2301,
+      "step": 60360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3270426087739845e-05,
+      "loss": 0.2255,
+      "step": 60370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3265921989010002e-05,
+      "loss": 0.2297,
+      "step": 60380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3261417890280156e-05,
+      "loss": 0.2275,
+      "step": 60390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.325691379155031e-05,
+      "loss": 0.2295,
+      "step": 60400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3252409692820468e-05,
+      "loss": 0.2265,
+      "step": 60410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3247905594090625e-05,
+      "loss": 0.2202,
+      "step": 60420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.324340149536078e-05,
+      "loss": 0.2309,
+      "step": 60430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3238897396630934e-05,
+      "loss": 0.2259,
+      "step": 60440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.323439329790109e-05,
+      "loss": 0.2379,
+      "step": 60450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.322988919917125e-05,
+      "loss": 0.2279,
+      "step": 60460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3225385100441403e-05,
+      "loss": 0.232,
+      "step": 60470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.322088100171156e-05,
+      "loss": 0.2295,
+      "step": 60480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3216376902981714e-05,
+      "loss": 0.223,
+      "step": 60490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3211872804251868e-05,
+      "loss": 0.2314,
+      "step": 60500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3207368705522026e-05,
+      "loss": 0.2254,
+      "step": 60510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3202864606792183e-05,
+      "loss": 0.2295,
+      "step": 60520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3198360508062337e-05,
+      "loss": 0.2368,
+      "step": 60530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.319385640933249e-05,
+      "loss": 0.2285,
+      "step": 60540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.318935231060265e-05,
+      "loss": 0.2266,
+      "step": 60550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3184848211872806e-05,
+      "loss": 0.2253,
+      "step": 60560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.318034411314296e-05,
+      "loss": 0.2275,
+      "step": 60570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3175840014413118e-05,
+      "loss": 0.2264,
+      "step": 60580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3171335915683272e-05,
+      "loss": 0.2227,
+      "step": 60590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.316683181695343e-05,
+      "loss": 0.2243,
+      "step": 60600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3162327718223584e-05,
+      "loss": 0.2235,
+      "step": 60610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.315782361949374e-05,
+      "loss": 0.2265,
+      "step": 60620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3153319520763895e-05,
+      "loss": 0.2293,
+      "step": 60630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.314881542203405e-05,
+      "loss": 0.2288,
+      "step": 60640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3144311323304207e-05,
+      "loss": 0.2289,
+      "step": 60650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3139807224574364e-05,
+      "loss": 0.2289,
+      "step": 60660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3135303125844522e-05,
+      "loss": 0.2284,
+      "step": 60670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3130799027114676e-05,
+      "loss": 0.2266,
+      "step": 60680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.312629492838483e-05,
+      "loss": 0.2227,
+      "step": 60690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3121790829654988e-05,
+      "loss": 0.2286,
+      "step": 60700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3117286730925142e-05,
+      "loss": 0.2285,
+      "step": 60710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.31127826321953e-05,
+      "loss": 0.2244,
+      "step": 60720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3108278533465453e-05,
+      "loss": 0.2222,
+      "step": 60730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.310377443473561e-05,
+      "loss": 0.2258,
+      "step": 60740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3099270336005765e-05,
+      "loss": 0.2296,
+      "step": 60750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3094766237275922e-05,
+      "loss": 0.2314,
+      "step": 60760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.309026213854608e-05,
+      "loss": 0.2284,
+      "step": 60770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3085758039816234e-05,
+      "loss": 0.2268,
+      "step": 60780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3081253941086388e-05,
+      "loss": 0.2255,
+      "step": 60790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3076749842356546e-05,
+      "loss": 0.2257,
+      "step": 60800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3072245743626703e-05,
+      "loss": 0.2209,
+      "step": 60810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3067741644896857e-05,
+      "loss": 0.2315,
+      "step": 60820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3063237546167015e-05,
+      "loss": 0.2316,
+      "step": 60830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.305873344743717e-05,
+      "loss": 0.2332,
+      "step": 60840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3054229348707323e-05,
+      "loss": 0.2286,
+      "step": 60850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.304972524997748e-05,
+      "loss": 0.2251,
+      "step": 60860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3045221151247638e-05,
+      "loss": 0.2235,
+      "step": 60870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3040717052517792e-05,
+      "loss": 0.2322,
+      "step": 60880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3036212953787946e-05,
+      "loss": 0.2298,
+      "step": 60890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3031708855058104e-05,
+      "loss": 0.2302,
+      "step": 60900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.302720475632826e-05,
+      "loss": 0.2257,
+      "step": 60910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3022700657598415e-05,
+      "loss": 0.2247,
+      "step": 60920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3018196558868573e-05,
+      "loss": 0.2338,
+      "step": 60930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3013692460138727e-05,
+      "loss": 0.2327,
+      "step": 60940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3009188361408884e-05,
+      "loss": 0.2357,
+      "step": 60950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.300468426267904e-05,
+      "loss": 0.2304,
+      "step": 60960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3000180163949196e-05,
+      "loss": 0.2295,
+      "step": 60970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.299567606521935e-05,
+      "loss": 0.2303,
+      "step": 60980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2991171966489504e-05,
+      "loss": 0.2295,
+      "step": 60990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.298666786775966e-05,
+      "loss": 0.2267,
+      "step": 61000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.298216376902982e-05,
+      "loss": 0.2271,
+      "step": 61010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2977659670299977e-05,
+      "loss": 0.2311,
+      "step": 61020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.297315557157013e-05,
+      "loss": 0.2303,
+      "step": 61030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2968651472840285e-05,
+      "loss": 0.2372,
+      "step": 61040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2964147374110442e-05,
+      "loss": 0.2304,
+      "step": 61050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2959643275380596e-05,
+      "loss": 0.2257,
+      "step": 61060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2955139176650754e-05,
+      "loss": 0.2237,
+      "step": 61070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2950635077920908e-05,
+      "loss": 0.2292,
+      "step": 61080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2946130979191065e-05,
+      "loss": 0.2257,
+      "step": 61090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.294162688046122e-05,
+      "loss": 0.2276,
+      "step": 61100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2937122781731377e-05,
+      "loss": 0.2287,
+      "step": 61110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2932618683001535e-05,
+      "loss": 0.233,
+      "step": 61120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.292811458427169e-05,
+      "loss": 0.2285,
+      "step": 61130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2923610485541843e-05,
+      "loss": 0.2326,
+      "step": 61140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2919106386812e-05,
+      "loss": 0.2263,
+      "step": 61150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2914602288082158e-05,
+      "loss": 0.2272,
+      "step": 61160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2910098189352312e-05,
+      "loss": 0.2224,
+      "step": 61170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2905594090622466e-05,
+      "loss": 0.2272,
+      "step": 61180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2901089991892623e-05,
+      "loss": 0.2243,
+      "step": 61190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2896585893162778e-05,
+      "loss": 0.2282,
+      "step": 61200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2892081794432935e-05,
+      "loss": 0.2311,
+      "step": 61210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2887577695703093e-05,
+      "loss": 0.225,
+      "step": 61220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2883073596973247e-05,
+      "loss": 0.2235,
+      "step": 61230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.28785694982434e-05,
+      "loss": 0.23,
+      "step": 61240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2874065399513558e-05,
+      "loss": 0.2261,
+      "step": 61250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2869561300783716e-05,
+      "loss": 0.2219,
+      "step": 61260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.286505720205387e-05,
+      "loss": 0.2279,
+      "step": 61270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2860553103324024e-05,
+      "loss": 0.2302,
+      "step": 61280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.285604900459418e-05,
+      "loss": 0.2317,
+      "step": 61290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.285154490586434e-05,
+      "loss": 0.227,
+      "step": 61300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2847040807134493e-05,
+      "loss": 0.2257,
+      "step": 61310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.284253670840465e-05,
+      "loss": 0.2283,
+      "step": 61320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2838032609674805e-05,
+      "loss": 0.2283,
+      "step": 61330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.283352851094496e-05,
+      "loss": 0.2232,
+      "step": 61340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2829024412215116e-05,
+      "loss": 0.2278,
+      "step": 61350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2824520313485274e-05,
+      "loss": 0.2237,
+      "step": 61360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.282001621475543e-05,
+      "loss": 0.2264,
+      "step": 61370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2815512116025582e-05,
+      "loss": 0.2285,
+      "step": 61380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.281100801729574e-05,
+      "loss": 0.2276,
+      "step": 61390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2806503918565897e-05,
+      "loss": 0.2318,
+      "step": 61400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.280199981983605e-05,
+      "loss": 0.227,
+      "step": 61410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.279749572110621e-05,
+      "loss": 0.2342,
+      "step": 61420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2792991622376363e-05,
+      "loss": 0.2297,
+      "step": 61430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.278848752364652e-05,
+      "loss": 0.2283,
+      "step": 61440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2783983424916674e-05,
+      "loss": 0.2218,
+      "step": 61450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.277947932618683e-05,
+      "loss": 0.2289,
+      "step": 61460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.277497522745699e-05,
+      "loss": 0.2333,
+      "step": 61470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2770471128727143e-05,
+      "loss": 0.2334,
+      "step": 61480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2765967029997297e-05,
+      "loss": 0.2282,
+      "step": 61490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2761462931267455e-05,
+      "loss": 0.2311,
+      "step": 61500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2756958832537612e-05,
+      "loss": 0.2278,
+      "step": 61510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2752454733807766e-05,
+      "loss": 0.2276,
+      "step": 61520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.274795063507792e-05,
+      "loss": 0.23,
+      "step": 61530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2743446536348078e-05,
+      "loss": 0.2276,
+      "step": 61540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2738942437618232e-05,
+      "loss": 0.2245,
+      "step": 61550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.273443833888839e-05,
+      "loss": 0.2269,
+      "step": 61560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2729934240158547e-05,
+      "loss": 0.2243,
+      "step": 61570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.27254301414287e-05,
+      "loss": 0.2297,
+      "step": 61580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2720926042698855e-05,
+      "loss": 0.2252,
+      "step": 61590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2716421943969013e-05,
+      "loss": 0.2269,
+      "step": 61600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.271191784523917e-05,
+      "loss": 0.2266,
+      "step": 61610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2707413746509324e-05,
+      "loss": 0.2234,
+      "step": 61620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.270290964777948e-05,
+      "loss": 0.2251,
+      "step": 61630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2698405549049636e-05,
+      "loss": 0.2295,
+      "step": 61640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2693901450319794e-05,
+      "loss": 0.2315,
+      "step": 61650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2689397351589948e-05,
+      "loss": 0.2277,
+      "step": 61660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2684893252860105e-05,
+      "loss": 0.2245,
+      "step": 61670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.268038915413026e-05,
+      "loss": 0.2277,
+      "step": 61680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2675885055400413e-05,
+      "loss": 0.2291,
+      "step": 61690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.267138095667057e-05,
+      "loss": 0.2277,
+      "step": 61700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.266687685794073e-05,
+      "loss": 0.2281,
+      "step": 61710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2662372759210886e-05,
+      "loss": 0.2254,
+      "step": 61720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2657868660481037e-05,
+      "loss": 0.2217,
+      "step": 61730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2653364561751194e-05,
+      "loss": 0.2176,
+      "step": 61740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.264886046302135e-05,
+      "loss": 0.2242,
+      "step": 61750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2644356364291506e-05,
+      "loss": 0.2284,
+      "step": 61760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2639852265561663e-05,
+      "loss": 0.2263,
+      "step": 61770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2635348166831817e-05,
+      "loss": 0.2285,
+      "step": 61780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2630844068101975e-05,
+      "loss": 0.226,
+      "step": 61790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.262633996937213e-05,
+      "loss": 0.2273,
+      "step": 61800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2621835870642286e-05,
+      "loss": 0.226,
+      "step": 61810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2617331771912444e-05,
+      "loss": 0.2266,
+      "step": 61820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2612827673182595e-05,
+      "loss": 0.234,
+      "step": 61830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2608323574452752e-05,
+      "loss": 0.2323,
+      "step": 61840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.260381947572291e-05,
+      "loss": 0.2309,
+      "step": 61850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2599315376993067e-05,
+      "loss": 0.2269,
+      "step": 61860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.259481127826322e-05,
+      "loss": 0.2268,
+      "step": 61870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2590307179533375e-05,
+      "loss": 0.2255,
+      "step": 61880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2585803080803533e-05,
+      "loss": 0.2253,
+      "step": 61890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2581298982073687e-05,
+      "loss": 0.2254,
+      "step": 61900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2576794883343844e-05,
+      "loss": 0.2241,
+      "step": 61910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2572290784614002e-05,
+      "loss": 0.2334,
+      "step": 61920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2567786685884156e-05,
+      "loss": 0.2181,
+      "step": 61930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.256328258715431e-05,
+      "loss": 0.2254,
+      "step": 61940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2558778488424467e-05,
+      "loss": 0.2278,
+      "step": 61950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2554274389694625e-05,
+      "loss": 0.2276,
+      "step": 61960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.254977029096478e-05,
+      "loss": 0.2257,
+      "step": 61970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2545266192234933e-05,
+      "loss": 0.2239,
+      "step": 61980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.254076209350509e-05,
+      "loss": 0.228,
+      "step": 61990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2536257994775248e-05,
+      "loss": 0.2275,
+      "step": 62000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2531753896045402e-05,
+      "loss": 0.2229,
+      "step": 62010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.252724979731556e-05,
+      "loss": 0.225,
+      "step": 62020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2522745698585714e-05,
+      "loss": 0.2233,
+      "step": 62030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2518241599855868e-05,
+      "loss": 0.2191,
+      "step": 62040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2513737501126025e-05,
+      "loss": 0.2275,
+      "step": 62050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2509233402396183e-05,
+      "loss": 0.2312,
+      "step": 62060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2504729303666337e-05,
+      "loss": 0.2282,
+      "step": 62070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.250022520493649e-05,
+      "loss": 0.2323,
+      "step": 62080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.249572110620665e-05,
+      "loss": 0.2246,
+      "step": 62090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2491217007476806e-05,
+      "loss": 0.2262,
+      "step": 62100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.248671290874696e-05,
+      "loss": 0.2256,
+      "step": 62110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2482208810017118e-05,
+      "loss": 0.2248,
+      "step": 62120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2477704711287272e-05,
+      "loss": 0.2234,
+      "step": 62130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.247320061255743e-05,
+      "loss": 0.2209,
+      "step": 62140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2468696513827583e-05,
+      "loss": 0.2311,
+      "step": 62150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.246419241509774e-05,
+      "loss": 0.233,
+      "step": 62160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2459688316367895e-05,
+      "loss": 0.2311,
+      "step": 62170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.245518421763805e-05,
+      "loss": 0.2257,
+      "step": 62180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2450680118908207e-05,
+      "loss": 0.2265,
+      "step": 62190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2446176020178364e-05,
+      "loss": 0.2296,
+      "step": 62200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.244167192144852e-05,
+      "loss": 0.2265,
+      "step": 62210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2437167822718676e-05,
+      "loss": 0.2295,
+      "step": 62220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.243266372398883e-05,
+      "loss": 0.2276,
+      "step": 62230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2428159625258987e-05,
+      "loss": 0.2262,
+      "step": 62240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.242365552652914e-05,
+      "loss": 0.2222,
+      "step": 62250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.24191514277993e-05,
+      "loss": 0.224,
+      "step": 62260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2414647329069456e-05,
+      "loss": 0.2274,
+      "step": 62270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.241014323033961e-05,
+      "loss": 0.2238,
+      "step": 62280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2405639131609765e-05,
+      "loss": 0.2287,
+      "step": 62290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2401135032879922e-05,
+      "loss": 0.2191,
+      "step": 62300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.239663093415008e-05,
+      "loss": 0.222,
+      "step": 62310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2392126835420234e-05,
+      "loss": 0.2296,
+      "step": 62320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2387622736690388e-05,
+      "loss": 0.2246,
+      "step": 62330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2383118637960545e-05,
+      "loss": 0.2242,
+      "step": 62340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2378614539230703e-05,
+      "loss": 0.2326,
+      "step": 62350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2374110440500857e-05,
+      "loss": 0.226,
+      "step": 62360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2369606341771014e-05,
+      "loss": 0.2243,
+      "step": 62370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.236510224304117e-05,
+      "loss": 0.2263,
+      "step": 62380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2360598144311323e-05,
+      "loss": 0.2267,
+      "step": 62390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.235609404558148e-05,
+      "loss": 0.2245,
+      "step": 62400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2351589946851638e-05,
+      "loss": 0.2226,
+      "step": 62410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2347085848121792e-05,
+      "loss": 0.2238,
+      "step": 62420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2342581749391946e-05,
+      "loss": 0.2268,
+      "step": 62430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2338077650662103e-05,
+      "loss": 0.2244,
+      "step": 62440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.233357355193226e-05,
+      "loss": 0.2266,
+      "step": 62450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2329069453202415e-05,
+      "loss": 0.2238,
+      "step": 62460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2324565354472572e-05,
+      "loss": 0.2243,
+      "step": 62470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2320061255742726e-05,
+      "loss": 0.227,
+      "step": 62480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2315557157012884e-05,
+      "loss": 0.2309,
+      "step": 62490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2311053058283038e-05,
+      "loss": 0.223,
+      "step": 62500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2306548959553196e-05,
+      "loss": 0.2284,
+      "step": 62510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.230204486082335e-05,
+      "loss": 0.2209,
+      "step": 62520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2297540762093504e-05,
+      "loss": 0.2311,
+      "step": 62530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.229303666336366e-05,
+      "loss": 0.223,
+      "step": 62540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.228853256463382e-05,
+      "loss": 0.2209,
+      "step": 62550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2284028465903976e-05,
+      "loss": 0.2269,
+      "step": 62560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.227952436717413e-05,
+      "loss": 0.2207,
+      "step": 62570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2275020268444284e-05,
+      "loss": 0.2242,
+      "step": 62580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2270516169714442e-05,
+      "loss": 0.2183,
+      "step": 62590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2266012070984596e-05,
+      "loss": 0.2251,
+      "step": 62600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2261507972254754e-05,
+      "loss": 0.2242,
+      "step": 62610
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2257003873524908e-05,
+      "loss": 0.2147,
+      "step": 62620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2252499774795065e-05,
+      "loss": 0.2274,
+      "step": 62630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.224799567606522e-05,
+      "loss": 0.226,
+      "step": 62640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2243491577335377e-05,
+      "loss": 0.2323,
+      "step": 62650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2238987478605534e-05,
+      "loss": 0.2267,
+      "step": 62660
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.223448337987569e-05,
+      "loss": 0.2247,
+      "step": 62670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2229979281145842e-05,
+      "loss": 0.2207,
+      "step": 62680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2225475182416e-05,
+      "loss": 0.2252,
+      "step": 62690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2220971083686157e-05,
+      "loss": 0.2219,
+      "step": 62700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.221646698495631e-05,
+      "loss": 0.2203,
+      "step": 62710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2211962886226466e-05,
+      "loss": 0.2242,
+      "step": 62720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2207458787496623e-05,
+      "loss": 0.23,
+      "step": 62730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2202954688766777e-05,
+      "loss": 0.2332,
+      "step": 62740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2198450590036935e-05,
+      "loss": 0.2251,
+      "step": 62750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2193946491307092e-05,
+      "loss": 0.2301,
+      "step": 62760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2189442392577246e-05,
+      "loss": 0.2252,
+      "step": 62770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.21849382938474e-05,
+      "loss": 0.2252,
+      "step": 62780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2180434195117558e-05,
+      "loss": 0.2214,
+      "step": 62790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2175930096387715e-05,
+      "loss": 0.2254,
+      "step": 62800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2171876407530852e-05,
+      "loss": 0.222,
+      "step": 62810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.216737230880101e-05,
+      "loss": 0.2282,
+      "step": 62820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2162868210071167e-05,
+      "loss": 0.2312,
+      "step": 62830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.215836411134132e-05,
+      "loss": 0.2355,
+      "step": 62840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2153860012611476e-05,
+      "loss": 0.234,
+      "step": 62850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2149355913881633e-05,
+      "loss": 0.2296,
+      "step": 62860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.214485181515179e-05,
+      "loss": 0.2298,
+      "step": 62870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2140347716421945e-05,
+      "loss": 0.2332,
+      "step": 62880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2135843617692102e-05,
+      "loss": 0.2225,
+      "step": 62890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2131339518962256e-05,
+      "loss": 0.2242,
+      "step": 62900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.212683542023241e-05,
+      "loss": 0.228,
+      "step": 62910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2122331321502568e-05,
+      "loss": 0.2196,
+      "step": 62920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2117827222772725e-05,
+      "loss": 0.2297,
+      "step": 62930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2113323124042883e-05,
+      "loss": 0.2303,
+      "step": 62940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2108819025313034e-05,
+      "loss": 0.2248,
+      "step": 62950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.210431492658319e-05,
+      "loss": 0.227,
+      "step": 62960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.209981082785335e-05,
+      "loss": 0.2197,
+      "step": 62970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2095306729123503e-05,
+      "loss": 0.2262,
+      "step": 62980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.209080263039366e-05,
+      "loss": 0.2224,
+      "step": 62990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2086298531663814e-05,
+      "loss": 0.2204,
+      "step": 63000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2081794432933972e-05,
+      "loss": 0.2254,
+      "step": 63010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2077290334204126e-05,
+      "loss": 0.2252,
+      "step": 63020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2072786235474283e-05,
+      "loss": 0.2265,
+      "step": 63030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.206828213674444e-05,
+      "loss": 0.228,
+      "step": 63040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.206377803801459e-05,
+      "loss": 0.2214,
+      "step": 63050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.205927393928475e-05,
+      "loss": 0.2254,
+      "step": 63060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2054769840554907e-05,
+      "loss": 0.2238,
+      "step": 63070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2050265741825064e-05,
+      "loss": 0.2241,
+      "step": 63080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2045761643095218e-05,
+      "loss": 0.227,
+      "step": 63090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2041257544365372e-05,
+      "loss": 0.2194,
+      "step": 63100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.203675344563553e-05,
+      "loss": 0.231,
+      "step": 63110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2032249346905684e-05,
+      "loss": 0.2217,
+      "step": 63120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.202774524817584e-05,
+      "loss": 0.2268,
+      "step": 63130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2023241149446e-05,
+      "loss": 0.2258,
+      "step": 63140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2018737050716153e-05,
+      "loss": 0.2237,
+      "step": 63150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2014232951986307e-05,
+      "loss": 0.2231,
+      "step": 63160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2009728853256465e-05,
+      "loss": 0.2245,
+      "step": 63170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2005224754526622e-05,
+      "loss": 0.222,
+      "step": 63180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2000720655796776e-05,
+      "loss": 0.2315,
+      "step": 63190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.199621655706693e-05,
+      "loss": 0.2177,
+      "step": 63200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1991712458337088e-05,
+      "loss": 0.2233,
+      "step": 63210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1987208359607245e-05,
+      "loss": 0.2238,
+      "step": 63220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.19827042608774e-05,
+      "loss": 0.2183,
+      "step": 63230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1978200162147557e-05,
+      "loss": 0.2239,
+      "step": 63240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.197369606341771e-05,
+      "loss": 0.2211,
+      "step": 63250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1969191964687865e-05,
+      "loss": 0.2257,
+      "step": 63260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1964687865958023e-05,
+      "loss": 0.221,
+      "step": 63270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.196018376722818e-05,
+      "loss": 0.2248,
+      "step": 63280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1955679668498334e-05,
+      "loss": 0.2208,
+      "step": 63290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1951175569768488e-05,
+      "loss": 0.2209,
+      "step": 63300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1946671471038646e-05,
+      "loss": 0.22,
+      "step": 63310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1942167372308803e-05,
+      "loss": 0.2173,
+      "step": 63320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1937663273578957e-05,
+      "loss": 0.22,
+      "step": 63330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1933159174849115e-05,
+      "loss": 0.221,
+      "step": 63340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.192865507611927e-05,
+      "loss": 0.2218,
+      "step": 63350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1924150977389426e-05,
+      "loss": 0.226,
+      "step": 63360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.191964687865958e-05,
+      "loss": 0.2214,
+      "step": 63370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1915142779929738e-05,
+      "loss": 0.2247,
+      "step": 63380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1910638681199892e-05,
+      "loss": 0.2304,
+      "step": 63390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1906134582470046e-05,
+      "loss": 0.2272,
+      "step": 63400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1901630483740204e-05,
+      "loss": 0.2262,
+      "step": 63410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.189712638501036e-05,
+      "loss": 0.2256,
+      "step": 63420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.189262228628052e-05,
+      "loss": 0.2317,
+      "step": 63430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1888118187550673e-05,
+      "loss": 0.2248,
+      "step": 63440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1883614088820827e-05,
+      "loss": 0.2097,
+      "step": 63450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1879109990090984e-05,
+      "loss": 0.2217,
+      "step": 63460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.187460589136114e-05,
+      "loss": 0.223,
+      "step": 63470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1870101792631296e-05,
+      "loss": 0.2246,
+      "step": 63480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.186559769390145e-05,
+      "loss": 0.2203,
+      "step": 63490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1861093595171608e-05,
+      "loss": 0.2254,
+      "step": 63500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.185658949644176e-05,
+      "loss": 0.2287,
+      "step": 63510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.185208539771192e-05,
+      "loss": 0.2283,
+      "step": 63520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1847581298982077e-05,
+      "loss": 0.2196,
+      "step": 63530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.184307720025223e-05,
+      "loss": 0.2245,
+      "step": 63540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1838573101522385e-05,
+      "loss": 0.2179,
+      "step": 63550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1834069002792542e-05,
+      "loss": 0.2182,
+      "step": 63560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.18295649040627e-05,
+      "loss": 0.2196,
+      "step": 63570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1825060805332854e-05,
+      "loss": 0.2226,
+      "step": 63580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.182055670660301e-05,
+      "loss": 0.2271,
+      "step": 63590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1816052607873166e-05,
+      "loss": 0.2272,
+      "step": 63600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.181154850914332e-05,
+      "loss": 0.2208,
+      "step": 63610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1807044410413477e-05,
+      "loss": 0.2198,
+      "step": 63620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1802540311683635e-05,
+      "loss": 0.2232,
+      "step": 63630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.179803621295379e-05,
+      "loss": 0.2223,
+      "step": 63640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1793532114223943e-05,
+      "loss": 0.2271,
+      "step": 63650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.17890280154941e-05,
+      "loss": 0.2185,
+      "step": 63660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1784523916764258e-05,
+      "loss": 0.2302,
+      "step": 63670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1780019818034412e-05,
+      "loss": 0.2265,
+      "step": 63680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.177551571930457e-05,
+      "loss": 0.2299,
+      "step": 63690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1771011620574724e-05,
+      "loss": 0.2215,
+      "step": 63700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.176650752184488e-05,
+      "loss": 0.2273,
+      "step": 63710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1762003423115035e-05,
+      "loss": 0.226,
+      "step": 63720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1757499324385193e-05,
+      "loss": 0.2232,
+      "step": 63730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1752995225655347e-05,
+      "loss": 0.2217,
+      "step": 63740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.17484911269255e-05,
+      "loss": 0.2162,
+      "step": 63750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.174398702819566e-05,
+      "loss": 0.2234,
+      "step": 63760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1739482929465816e-05,
+      "loss": 0.2185,
+      "step": 63770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1734978830735973e-05,
+      "loss": 0.2179,
+      "step": 63780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1730474732006127e-05,
+      "loss": 0.2227,
+      "step": 63790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.172597063327628e-05,
+      "loss": 0.227,
+      "step": 63800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.172146653454644e-05,
+      "loss": 0.2243,
+      "step": 63810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1716962435816593e-05,
+      "loss": 0.2247,
+      "step": 63820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.171245833708675e-05,
+      "loss": 0.2262,
+      "step": 63830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1707954238356905e-05,
+      "loss": 0.2215,
+      "step": 63840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1703450139627062e-05,
+      "loss": 0.2261,
+      "step": 63850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1698946040897216e-05,
+      "loss": 0.2271,
+      "step": 63860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1694441942167374e-05,
+      "loss": 0.2195,
+      "step": 63870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.168993784343753e-05,
+      "loss": 0.2222,
+      "step": 63880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1685433744707685e-05,
+      "loss": 0.2203,
+      "step": 63890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.168092964597784e-05,
+      "loss": 0.2229,
+      "step": 63900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1676425547247997e-05,
+      "loss": 0.2215,
+      "step": 63910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1671921448518154e-05,
+      "loss": 0.2254,
+      "step": 63920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.166741734978831e-05,
+      "loss": 0.225,
+      "step": 63930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1662913251058463e-05,
+      "loss": 0.2205,
+      "step": 63940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.165840915232862e-05,
+      "loss": 0.2272,
+      "step": 63950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1653905053598774e-05,
+      "loss": 0.2189,
+      "step": 63960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1649400954868932e-05,
+      "loss": 0.2226,
+      "step": 63970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.164489685613909e-05,
+      "loss": 0.2276,
+      "step": 63980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1640392757409243e-05,
+      "loss": 0.2272,
+      "step": 63990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1635888658679397e-05,
+      "loss": 0.2168,
+      "step": 64000
+    },
+    {
+      "epoch": 0.57,
+      "eval_NEIMS_canon_smiles": 0.9400666666666667,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.03925,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.4345703712458914,
+      "eval_NEIMS_exact_mols": 0.03911666666666667,
+      "eval_NEIMS_exact_smiles": 0.037816666666666665,
+      "eval_NEIMS_loss": 0.24026688933372498,
+      "eval_NEIMS_matched_formulas": 0.5316666666666666,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.33559208040238414,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.0393,
+      "eval_NEIMS_runtime": 712.3444,
+      "eval_NEIMS_samples_per_second": 84.229,
+      "eval_NEIMS_steps_per_second": 1.317,
+      "step": 64000
+    },
+    {
+      "epoch": 0.57,
+      "eval_RASSP_canon_smiles": 0.9279059672087354,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.09319798310348282,
+      "eval_RASSP_daylight_tanimoto_simil": 0.5470292273063273,
+      "eval_RASSP_exact_mols": 0.09276388285971883,
+      "eval_RASSP_exact_smiles": 0.08915751160383344,
+      "eval_RASSP_loss": 0.18058153986930847,
+      "eval_RASSP_matched_formulas": 0.7011219821684976,
+      "eval_RASSP_morgan_tanimoto_simil": 0.411355893083842,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.09286405983904898,
+      "eval_RASSP_runtime": 814.2028,
+      "eval_RASSP_samples_per_second": 73.562,
+      "eval_RASSP_steps_per_second": 1.15,
+      "step": 64000
+    },
+    {
+      "epoch": 0.57,
+      "eval_NIST_canon_smiles": 0.8854739681300351,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.009795223054264116,
+      "eval_NIST_daylight_tanimoto_simil": 0.2610283945884308,
+      "eval_NIST_exact_mols": 0.00947581360684246,
+      "eval_NIST_exact_smiles": 0.008801504773396742,
+      "eval_NIST_loss": 1.4489023685455322,
+      "eval_NIST_matched_formulas": 0.07829080455690812,
+      "eval_NIST_morgan_tanimoto_simil": 0.21492784860372788,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.009724243177059303,
+      "eval_NIST_runtime": 455.6201,
+      "eval_NIST_samples_per_second": 61.843,
+      "eval_NIST_steps_per_second": 0.968,
+      "step": 64000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1631384559949555e-05,
+      "loss": 0.222,
+      "step": 64010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1626880461219712e-05,
+      "loss": 0.2292,
+      "step": 64020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1622376362489867e-05,
+      "loss": 0.2222,
+      "step": 64030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.161787226376002e-05,
+      "loss": 0.2226,
+      "step": 64040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1613368165030178e-05,
+      "loss": 0.2268,
+      "step": 64050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1608864066300336e-05,
+      "loss": 0.2209,
+      "step": 64060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.160435996757049e-05,
+      "loss": 0.2247,
+      "step": 64070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1599855868840647e-05,
+      "loss": 0.2219,
+      "step": 64080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.15953517701108e-05,
+      "loss": 0.2211,
+      "step": 64090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1590847671380955e-05,
+      "loss": 0.225,
+      "step": 64100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1586343572651113e-05,
+      "loss": 0.2235,
+      "step": 64110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.158183947392127e-05,
+      "loss": 0.2244,
+      "step": 64120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1577335375191428e-05,
+      "loss": 0.2233,
+      "step": 64130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.157283127646158e-05,
+      "loss": 0.219,
+      "step": 64140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1568327177731736e-05,
+      "loss": 0.2228,
+      "step": 64150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1563823079001894e-05,
+      "loss": 0.2207,
+      "step": 64160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1559318980272048e-05,
+      "loss": 0.2172,
+      "step": 64170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1554814881542205e-05,
+      "loss": 0.2191,
+      "step": 64180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.155031078281236e-05,
+      "loss": 0.2241,
+      "step": 64190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1545806684082517e-05,
+      "loss": 0.2254,
+      "step": 64200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.154130258535267e-05,
+      "loss": 0.2225,
+      "step": 64210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.153679848662283e-05,
+      "loss": 0.225,
+      "step": 64220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1532294387892986e-05,
+      "loss": 0.2251,
+      "step": 64230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.152779028916314e-05,
+      "loss": 0.226,
+      "step": 64240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1523286190433294e-05,
+      "loss": 0.226,
+      "step": 64250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.151878209170345e-05,
+      "loss": 0.2212,
+      "step": 64260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.151427799297361e-05,
+      "loss": 0.2237,
+      "step": 64270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1509773894243763e-05,
+      "loss": 0.2247,
+      "step": 64280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1505269795513917e-05,
+      "loss": 0.2189,
+      "step": 64290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1500765696784075e-05,
+      "loss": 0.2202,
+      "step": 64300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.149626159805423e-05,
+      "loss": 0.2168,
+      "step": 64310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1491757499324386e-05,
+      "loss": 0.2185,
+      "step": 64320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1487253400594544e-05,
+      "loss": 0.2198,
+      "step": 64330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1482749301864698e-05,
+      "loss": 0.22,
+      "step": 64340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1478245203134852e-05,
+      "loss": 0.2188,
+      "step": 64350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.147374110440501e-05,
+      "loss": 0.2311,
+      "step": 64360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1469237005675167e-05,
+      "loss": 0.2191,
+      "step": 64370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.146473290694532e-05,
+      "loss": 0.2254,
+      "step": 64380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1460228808215475e-05,
+      "loss": 0.2126,
+      "step": 64390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1455724709485633e-05,
+      "loss": 0.225,
+      "step": 64400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.145122061075579e-05,
+      "loss": 0.2274,
+      "step": 64410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1446716512025944e-05,
+      "loss": 0.2231,
+      "step": 64420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1442212413296102e-05,
+      "loss": 0.2267,
+      "step": 64430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1437708314566256e-05,
+      "loss": 0.2263,
+      "step": 64440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.143320421583641e-05,
+      "loss": 0.2185,
+      "step": 64450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1428700117106568e-05,
+      "loss": 0.225,
+      "step": 64460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1424196018376725e-05,
+      "loss": 0.2204,
+      "step": 64470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1419691919646883e-05,
+      "loss": 0.2189,
+      "step": 64480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1415187820917033e-05,
+      "loss": 0.2177,
+      "step": 64490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.141068372218719e-05,
+      "loss": 0.2256,
+      "step": 64500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1406179623457348e-05,
+      "loss": 0.2193,
+      "step": 64510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1401675524727502e-05,
+      "loss": 0.2225,
+      "step": 64520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.139717142599766e-05,
+      "loss": 0.2199,
+      "step": 64530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1392667327267814e-05,
+      "loss": 0.2231,
+      "step": 64540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.138816322853797e-05,
+      "loss": 0.2321,
+      "step": 64550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1383659129808126e-05,
+      "loss": 0.2218,
+      "step": 64560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1379155031078283e-05,
+      "loss": 0.2246,
+      "step": 64570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.137465093234844e-05,
+      "loss": 0.2222,
+      "step": 64580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.137014683361859e-05,
+      "loss": 0.224,
+      "step": 64590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.136564273488875e-05,
+      "loss": 0.2265,
+      "step": 64600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1361138636158906e-05,
+      "loss": 0.2225,
+      "step": 64610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1356634537429064e-05,
+      "loss": 0.2211,
+      "step": 64620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1352130438699218e-05,
+      "loss": 0.2208,
+      "step": 64630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1347626339969372e-05,
+      "loss": 0.2257,
+      "step": 64640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.134312224123953e-05,
+      "loss": 0.2245,
+      "step": 64650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1338618142509684e-05,
+      "loss": 0.2209,
+      "step": 64660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.133411404377984e-05,
+      "loss": 0.2259,
+      "step": 64670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.132960994505e-05,
+      "loss": 0.2212,
+      "step": 64680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1325105846320153e-05,
+      "loss": 0.2221,
+      "step": 64690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1320601747590307e-05,
+      "loss": 0.2228,
+      "step": 64700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1316097648860464e-05,
+      "loss": 0.2282,
+      "step": 64710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1311593550130622e-05,
+      "loss": 0.2252,
+      "step": 64720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1307089451400776e-05,
+      "loss": 0.223,
+      "step": 64730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.130258535267093e-05,
+      "loss": 0.2279,
+      "step": 64740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1298081253941087e-05,
+      "loss": 0.2274,
+      "step": 64750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1293577155211245e-05,
+      "loss": 0.2216,
+      "step": 64760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.12890730564814e-05,
+      "loss": 0.2247,
+      "step": 64770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1284568957751556e-05,
+      "loss": 0.2191,
+      "step": 64780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.128006485902171e-05,
+      "loss": 0.2198,
+      "step": 64790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1275560760291865e-05,
+      "loss": 0.2209,
+      "step": 64800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1271056661562022e-05,
+      "loss": 0.2153,
+      "step": 64810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.126655256283218e-05,
+      "loss": 0.2182,
+      "step": 64820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1262498873975317e-05,
+      "loss": 0.2198,
+      "step": 64830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1257994775245474e-05,
+      "loss": 0.2218,
+      "step": 64840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.125349067651563e-05,
+      "loss": 0.2169,
+      "step": 64850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1248986577785786e-05,
+      "loss": 0.2181,
+      "step": 64860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.124448247905594e-05,
+      "loss": 0.2166,
+      "step": 64870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1239978380326097e-05,
+      "loss": 0.2189,
+      "step": 64880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1235474281596255e-05,
+      "loss": 0.2185,
+      "step": 64890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.123097018286641e-05,
+      "loss": 0.22,
+      "step": 64900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1226466084136566e-05,
+      "loss": 0.2239,
+      "step": 64910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.122196198540672e-05,
+      "loss": 0.219,
+      "step": 64920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1217457886676878e-05,
+      "loss": 0.2202,
+      "step": 64930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1212953787947032e-05,
+      "loss": 0.2228,
+      "step": 64940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.120844968921719e-05,
+      "loss": 0.2254,
+      "step": 64950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1203945590487344e-05,
+      "loss": 0.2263,
+      "step": 64960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1199441491757498e-05,
+      "loss": 0.222,
+      "step": 64970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1194937393027655e-05,
+      "loss": 0.2209,
+      "step": 64980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1190433294297813e-05,
+      "loss": 0.2179,
+      "step": 64990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.118592919556797e-05,
+      "loss": 0.2256,
+      "step": 65000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1181425096838124e-05,
+      "loss": 0.2243,
+      "step": 65010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.117692099810828e-05,
+      "loss": 0.2232,
+      "step": 65020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1172416899378436e-05,
+      "loss": 0.2224,
+      "step": 65030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.116791280064859e-05,
+      "loss": 0.2228,
+      "step": 65040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1163408701918748e-05,
+      "loss": 0.2223,
+      "step": 65050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1158904603188902e-05,
+      "loss": 0.2203,
+      "step": 65060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.115440050445906e-05,
+      "loss": 0.2212,
+      "step": 65070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1149896405729213e-05,
+      "loss": 0.2187,
+      "step": 65080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.114539230699937e-05,
+      "loss": 0.2226,
+      "step": 65090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.114088820826953e-05,
+      "loss": 0.2207,
+      "step": 65100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1136384109539682e-05,
+      "loss": 0.224,
+      "step": 65110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1131880010809837e-05,
+      "loss": 0.2266,
+      "step": 65120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1127375912079994e-05,
+      "loss": 0.2252,
+      "step": 65130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.112287181335015e-05,
+      "loss": 0.2225,
+      "step": 65140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1118367714620306e-05,
+      "loss": 0.2203,
+      "step": 65150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.111386361589046e-05,
+      "loss": 0.2206,
+      "step": 65160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1109359517160617e-05,
+      "loss": 0.226,
+      "step": 65170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.110485541843077e-05,
+      "loss": 0.2224,
+      "step": 65180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.110035131970093e-05,
+      "loss": 0.2188,
+      "step": 65190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1095847220971086e-05,
+      "loss": 0.2209,
+      "step": 65200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.109134312224124e-05,
+      "loss": 0.2219,
+      "step": 65210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1086839023511395e-05,
+      "loss": 0.2284,
+      "step": 65220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1082334924781552e-05,
+      "loss": 0.2158,
+      "step": 65230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.107783082605171e-05,
+      "loss": 0.2227,
+      "step": 65240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1073326727321864e-05,
+      "loss": 0.2202,
+      "step": 65250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1068822628592018e-05,
+      "loss": 0.2212,
+      "step": 65260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1064318529862175e-05,
+      "loss": 0.2181,
+      "step": 65270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1059814431132333e-05,
+      "loss": 0.2178,
+      "step": 65280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1055310332402487e-05,
+      "loss": 0.2153,
+      "step": 65290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1050806233672644e-05,
+      "loss": 0.226,
+      "step": 65300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.10463021349428e-05,
+      "loss": 0.2195,
+      "step": 65310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1041798036212952e-05,
+      "loss": 0.2225,
+      "step": 65320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.103729393748311e-05,
+      "loss": 0.2132,
+      "step": 65330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1032789838753267e-05,
+      "loss": 0.22,
+      "step": 65340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1028285740023425e-05,
+      "loss": 0.2203,
+      "step": 65350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1023781641293576e-05,
+      "loss": 0.2179,
+      "step": 65360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1019277542563733e-05,
+      "loss": 0.2249,
+      "step": 65370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.101477344383389e-05,
+      "loss": 0.2201,
+      "step": 65380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1010269345104045e-05,
+      "loss": 0.2176,
+      "step": 65390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1005765246374202e-05,
+      "loss": 0.2183,
+      "step": 65400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1001261147644356e-05,
+      "loss": 0.2245,
+      "step": 65410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0996757048914514e-05,
+      "loss": 0.2246,
+      "step": 65420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0992252950184668e-05,
+      "loss": 0.2214,
+      "step": 65430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0987748851454825e-05,
+      "loss": 0.2177,
+      "step": 65440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0983244752724983e-05,
+      "loss": 0.2202,
+      "step": 65450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0978740653995134e-05,
+      "loss": 0.2198,
+      "step": 65460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.097423655526529e-05,
+      "loss": 0.225,
+      "step": 65470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.096973245653545e-05,
+      "loss": 0.2175,
+      "step": 65480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0965228357805606e-05,
+      "loss": 0.2123,
+      "step": 65490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.096072425907576e-05,
+      "loss": 0.2198,
+      "step": 65500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0956220160345914e-05,
+      "loss": 0.22,
+      "step": 65510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0951716061616072e-05,
+      "loss": 0.2208,
+      "step": 65520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0947211962886226e-05,
+      "loss": 0.2222,
+      "step": 65530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0942707864156383e-05,
+      "loss": 0.2186,
+      "step": 65540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.093820376542654e-05,
+      "loss": 0.2164,
+      "step": 65550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0933699666696695e-05,
+      "loss": 0.221,
+      "step": 65560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.092919556796685e-05,
+      "loss": 0.2204,
+      "step": 65570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0924691469237007e-05,
+      "loss": 0.2295,
+      "step": 65580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0920187370507164e-05,
+      "loss": 0.219,
+      "step": 65590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0915683271777318e-05,
+      "loss": 0.2187,
+      "step": 65600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0911179173047472e-05,
+      "loss": 0.2256,
+      "step": 65610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.090667507431763e-05,
+      "loss": 0.2246,
+      "step": 65620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0902170975587787e-05,
+      "loss": 0.2144,
+      "step": 65630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.089766687685794e-05,
+      "loss": 0.2189,
+      "step": 65640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.08931627781281e-05,
+      "loss": 0.2192,
+      "step": 65650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0888658679398253e-05,
+      "loss": 0.2246,
+      "step": 65660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0884154580668407e-05,
+      "loss": 0.2187,
+      "step": 65670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0879650481938565e-05,
+      "loss": 0.2184,
+      "step": 65680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0875146383208722e-05,
+      "loss": 0.2127,
+      "step": 65690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.087064228447888e-05,
+      "loss": 0.2202,
+      "step": 65700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.086613818574903e-05,
+      "loss": 0.2221,
+      "step": 65710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0861634087019188e-05,
+      "loss": 0.219,
+      "step": 65720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0857129988289345e-05,
+      "loss": 0.2216,
+      "step": 65730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.08526258895595e-05,
+      "loss": 0.2193,
+      "step": 65740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0848121790829657e-05,
+      "loss": 0.2208,
+      "step": 65750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.084361769209981e-05,
+      "loss": 0.2139,
+      "step": 65760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.083911359336997e-05,
+      "loss": 0.2183,
+      "step": 65770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0834609494640123e-05,
+      "loss": 0.2257,
+      "step": 65780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.083010539591028e-05,
+      "loss": 0.222,
+      "step": 65790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0825601297180438e-05,
+      "loss": 0.2174,
+      "step": 65800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0821097198450588e-05,
+      "loss": 0.223,
+      "step": 65810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0816593099720746e-05,
+      "loss": 0.2195,
+      "step": 65820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0812089000990903e-05,
+      "loss": 0.2185,
+      "step": 65830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.080758490226106e-05,
+      "loss": 0.2189,
+      "step": 65840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0803080803531215e-05,
+      "loss": 0.2213,
+      "step": 65850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.079857670480137e-05,
+      "loss": 0.2189,
+      "step": 65860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0794072606071526e-05,
+      "loss": 0.2193,
+      "step": 65870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.078956850734168e-05,
+      "loss": 0.2214,
+      "step": 65880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0785064408611838e-05,
+      "loss": 0.219,
+      "step": 65890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0780560309881996e-05,
+      "loss": 0.2195,
+      "step": 65900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.077605621115215e-05,
+      "loss": 0.2146,
+      "step": 65910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0771552112422304e-05,
+      "loss": 0.2176,
+      "step": 65920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.076704801369246e-05,
+      "loss": 0.2194,
+      "step": 65930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.076254391496262e-05,
+      "loss": 0.2155,
+      "step": 65940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0758039816232773e-05,
+      "loss": 0.2232,
+      "step": 65950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0753535717502927e-05,
+      "loss": 0.217,
+      "step": 65960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0749031618773084e-05,
+      "loss": 0.2162,
+      "step": 65970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0744527520043242e-05,
+      "loss": 0.2248,
+      "step": 65980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0740023421313396e-05,
+      "loss": 0.217,
+      "step": 65990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0735519322583554e-05,
+      "loss": 0.2239,
+      "step": 66000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0731015223853708e-05,
+      "loss": 0.2155,
+      "step": 66010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0726511125123862e-05,
+      "loss": 0.2198,
+      "step": 66020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.072200702639402e-05,
+      "loss": 0.216,
+      "step": 66030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0717502927664177e-05,
+      "loss": 0.2197,
+      "step": 66040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.071299882893433e-05,
+      "loss": 0.2137,
+      "step": 66050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0708494730204485e-05,
+      "loss": 0.2114,
+      "step": 66060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0703990631474642e-05,
+      "loss": 0.2237,
+      "step": 66070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.06994865327448e-05,
+      "loss": 0.2101,
+      "step": 66080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0694982434014954e-05,
+      "loss": 0.2151,
+      "step": 66090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.069047833528511e-05,
+      "loss": 0.2136,
+      "step": 66100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0685974236555266e-05,
+      "loss": 0.2147,
+      "step": 66110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0681470137825423e-05,
+      "loss": 0.2157,
+      "step": 66120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0676966039095577e-05,
+      "loss": 0.2157,
+      "step": 66130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0672461940365735e-05,
+      "loss": 0.2133,
+      "step": 66140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.066795784163589e-05,
+      "loss": 0.2047,
+      "step": 66150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0663453742906043e-05,
+      "loss": 0.2187,
+      "step": 66160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.06589496441762e-05,
+      "loss": 0.216,
+      "step": 66170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0654445545446358e-05,
+      "loss": 0.2133,
+      "step": 66180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0649941446716515e-05,
+      "loss": 0.2154,
+      "step": 66190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.064543734798667e-05,
+      "loss": 0.2174,
+      "step": 66200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0640933249256824e-05,
+      "loss": 0.2121,
+      "step": 66210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.063642915052698e-05,
+      "loss": 0.2147,
+      "step": 66220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0631925051797135e-05,
+      "loss": 0.215,
+      "step": 66230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0627420953067293e-05,
+      "loss": 0.212,
+      "step": 66240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0622916854337447e-05,
+      "loss": 0.2161,
+      "step": 66250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0618412755607604e-05,
+      "loss": 0.214,
+      "step": 66260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.061390865687776e-05,
+      "loss": 0.2104,
+      "step": 66270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0609404558147916e-05,
+      "loss": 0.215,
+      "step": 66280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0604900459418073e-05,
+      "loss": 0.2139,
+      "step": 66290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0600396360688227e-05,
+      "loss": 0.2144,
+      "step": 66300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.059589226195838e-05,
+      "loss": 0.2164,
+      "step": 66310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.059138816322854e-05,
+      "loss": 0.2084,
+      "step": 66320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0586884064498697e-05,
+      "loss": 0.2132,
+      "step": 66330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.058237996576885e-05,
+      "loss": 0.2207,
+      "step": 66340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0577875867039008e-05,
+      "loss": 0.2121,
+      "step": 66350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0573371768309162e-05,
+      "loss": 0.207,
+      "step": 66360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0568867669579316e-05,
+      "loss": 0.2184,
+      "step": 66370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0564363570849474e-05,
+      "loss": 0.2106,
+      "step": 66380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.055985947211963e-05,
+      "loss": 0.208,
+      "step": 66390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0555355373389785e-05,
+      "loss": 0.2147,
+      "step": 66400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.055085127465994e-05,
+      "loss": 0.211,
+      "step": 66410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0546347175930097e-05,
+      "loss": 0.2123,
+      "step": 66420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0541843077200255e-05,
+      "loss": 0.2126,
+      "step": 66430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.053733897847041e-05,
+      "loss": 0.2149,
+      "step": 66440
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0532834879740566e-05,
+      "loss": 0.2162,
+      "step": 66450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.052833078101072e-05,
+      "loss": 0.2055,
+      "step": 66460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0523826682280878e-05,
+      "loss": 0.2109,
+      "step": 66470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0519322583551032e-05,
+      "loss": 0.2093,
+      "step": 66480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.051481848482119e-05,
+      "loss": 0.2152,
+      "step": 66490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0510314386091343e-05,
+      "loss": 0.2177,
+      "step": 66500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0505810287361498e-05,
+      "loss": 0.2163,
+      "step": 66510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0501306188631655e-05,
+      "loss": 0.2166,
+      "step": 66520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0496802089901813e-05,
+      "loss": 0.2152,
+      "step": 66530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.049229799117197e-05,
+      "loss": 0.2149,
+      "step": 66540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0487793892442124e-05,
+      "loss": 0.2102,
+      "step": 66550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0483289793712278e-05,
+      "loss": 0.2124,
+      "step": 66560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0478785694982436e-05,
+      "loss": 0.2197,
+      "step": 66570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.047428159625259e-05,
+      "loss": 0.2146,
+      "step": 66580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0469777497522747e-05,
+      "loss": 0.2134,
+      "step": 66590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.04652733987929e-05,
+      "loss": 0.2102,
+      "step": 66600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.046076930006306e-05,
+      "loss": 0.2108,
+      "step": 66610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0456265201333213e-05,
+      "loss": 0.2088,
+      "step": 66620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.045176110260337e-05,
+      "loss": 0.2113,
+      "step": 66630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0447257003873528e-05,
+      "loss": 0.2073,
+      "step": 66640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0442752905143682e-05,
+      "loss": 0.1995,
+      "step": 66650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0438248806413836e-05,
+      "loss": 0.2111,
+      "step": 66660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0433744707683994e-05,
+      "loss": 0.2135,
+      "step": 66670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.042924060895415e-05,
+      "loss": 0.2124,
+      "step": 66680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0424736510224305e-05,
+      "loss": 0.2062,
+      "step": 66690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.042023241149446e-05,
+      "loss": 0.2099,
+      "step": 66700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0415728312764617e-05,
+      "loss": 0.2109,
+      "step": 66710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.041122421403477e-05,
+      "loss": 0.2098,
+      "step": 66720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.040672011530493e-05,
+      "loss": 0.2077,
+      "step": 66730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0402216016575086e-05,
+      "loss": 0.2084,
+      "step": 66740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.039771191784524e-05,
+      "loss": 0.2184,
+      "step": 66750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0393207819115394e-05,
+      "loss": 0.2064,
+      "step": 66760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.038870372038555e-05,
+      "loss": 0.2169,
+      "step": 66770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.038419962165571e-05,
+      "loss": 0.2108,
+      "step": 66780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0379695522925863e-05,
+      "loss": 0.2088,
+      "step": 66790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0375191424196017e-05,
+      "loss": 0.2067,
+      "step": 66800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0370687325466175e-05,
+      "loss": 0.2071,
+      "step": 66810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0366183226736332e-05,
+      "loss": 0.2207,
+      "step": 66820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0361679128006486e-05,
+      "loss": 0.2173,
+      "step": 66830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0357175029276644e-05,
+      "loss": 0.2154,
+      "step": 66840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0352670930546798e-05,
+      "loss": 0.2093,
+      "step": 66850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0348166831816952e-05,
+      "loss": 0.2149,
+      "step": 66860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.034366273308711e-05,
+      "loss": 0.2121,
+      "step": 66870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.033960904423025e-05,
+      "loss": 0.2107,
+      "step": 66880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0335104945500404e-05,
+      "loss": 0.2075,
+      "step": 66890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.033060084677056e-05,
+      "loss": 0.2097,
+      "step": 66900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.032609674804072e-05,
+      "loss": 0.2191,
+      "step": 66910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0321592649310873e-05,
+      "loss": 0.2081,
+      "step": 66920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0317088550581027e-05,
+      "loss": 0.2028,
+      "step": 66930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0312584451851185e-05,
+      "loss": 0.2102,
+      "step": 66940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0308080353121342e-05,
+      "loss": 0.2018,
+      "step": 66950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0303576254391496e-05,
+      "loss": 0.2092,
+      "step": 66960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0299072155661654e-05,
+      "loss": 0.2031,
+      "step": 66970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0294568056931808e-05,
+      "loss": 0.2103,
+      "step": 66980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0290063958201966e-05,
+      "loss": 0.2088,
+      "step": 66990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.028555985947212e-05,
+      "loss": 0.2127,
+      "step": 67000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0281055760742277e-05,
+      "loss": 0.2084,
+      "step": 67010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0276551662012435e-05,
+      "loss": 0.2085,
+      "step": 67020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0272047563282585e-05,
+      "loss": 0.2067,
+      "step": 67030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0267543464552743e-05,
+      "loss": 0.2027,
+      "step": 67040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.02630393658229e-05,
+      "loss": 0.2138,
+      "step": 67050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0258535267093058e-05,
+      "loss": 0.2037,
+      "step": 67060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0254031168363212e-05,
+      "loss": 0.2141,
+      "step": 67070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0249527069633366e-05,
+      "loss": 0.2182,
+      "step": 67080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0245022970903524e-05,
+      "loss": 0.2003,
+      "step": 67090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0240518872173678e-05,
+      "loss": 0.2066,
+      "step": 67100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0236014773443835e-05,
+      "loss": 0.2074,
+      "step": 67110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0231510674713993e-05,
+      "loss": 0.2019,
+      "step": 67120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0227006575984147e-05,
+      "loss": 0.2053,
+      "step": 67130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.02225024772543e-05,
+      "loss": 0.2058,
+      "step": 67140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0217998378524458e-05,
+      "loss": 0.2131,
+      "step": 67150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0213494279794616e-05,
+      "loss": 0.2077,
+      "step": 67160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.020899018106477e-05,
+      "loss": 0.2139,
+      "step": 67170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0204486082334924e-05,
+      "loss": 0.2076,
+      "step": 67180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.019998198360508e-05,
+      "loss": 0.2148,
+      "step": 67190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.019547788487524e-05,
+      "loss": 0.21,
+      "step": 67200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0190973786145393e-05,
+      "loss": 0.2062,
+      "step": 67210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.018646968741555e-05,
+      "loss": 0.2098,
+      "step": 67220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0181965588685705e-05,
+      "loss": 0.2142,
+      "step": 67230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.017746148995586e-05,
+      "loss": 0.2099,
+      "step": 67240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0172957391226016e-05,
+      "loss": 0.2043,
+      "step": 67250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0168453292496174e-05,
+      "loss": 0.2043,
+      "step": 67260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0163949193766328e-05,
+      "loss": 0.21,
+      "step": 67270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0159445095036482e-05,
+      "loss": 0.2126,
+      "step": 67280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.015494099630664e-05,
+      "loss": 0.2091,
+      "step": 67290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0150436897576797e-05,
+      "loss": 0.211,
+      "step": 67300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.014593279884695e-05,
+      "loss": 0.2097,
+      "step": 67310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.014142870011711e-05,
+      "loss": 0.2112,
+      "step": 67320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0136924601387263e-05,
+      "loss": 0.2115,
+      "step": 67330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.013242050265742e-05,
+      "loss": 0.2059,
+      "step": 67340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0127916403927574e-05,
+      "loss": 0.216,
+      "step": 67350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0123412305197732e-05,
+      "loss": 0.2087,
+      "step": 67360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0118908206467886e-05,
+      "loss": 0.2126,
+      "step": 67370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.011440410773804e-05,
+      "loss": 0.2081,
+      "step": 67380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0109900009008197e-05,
+      "loss": 0.2094,
+      "step": 67390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0105395910278355e-05,
+      "loss": 0.2064,
+      "step": 67400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0100891811548512e-05,
+      "loss": 0.21,
+      "step": 67410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0096387712818667e-05,
+      "loss": 0.2038,
+      "step": 67420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.009188361408882e-05,
+      "loss": 0.204,
+      "step": 67430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0087379515358978e-05,
+      "loss": 0.2116,
+      "step": 67440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0082875416629132e-05,
+      "loss": 0.2093,
+      "step": 67450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.007837131789929e-05,
+      "loss": 0.2115,
+      "step": 67460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0073867219169444e-05,
+      "loss": 0.2106,
+      "step": 67470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.00693631204396e-05,
+      "loss": 0.2118,
+      "step": 67480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0064859021709755e-05,
+      "loss": 0.2102,
+      "step": 67490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0060354922979913e-05,
+      "loss": 0.2151,
+      "step": 67500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.005585082425007e-05,
+      "loss": 0.2161,
+      "step": 67510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0051346725520225e-05,
+      "loss": 0.2106,
+      "step": 67520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.004684262679038e-05,
+      "loss": 0.2095,
+      "step": 67530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0042338528060536e-05,
+      "loss": 0.2071,
+      "step": 67540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0037834429330694e-05,
+      "loss": 0.2123,
+      "step": 67550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0033330330600848e-05,
+      "loss": 0.2101,
+      "step": 67560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0028826231871002e-05,
+      "loss": 0.205,
+      "step": 67570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.002432213314116e-05,
+      "loss": 0.2093,
+      "step": 67580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0019818034411313e-05,
+      "loss": 0.2125,
+      "step": 67590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.001531393568147e-05,
+      "loss": 0.2018,
+      "step": 67600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.001080983695163e-05,
+      "loss": 0.2073,
+      "step": 67610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0006305738221783e-05,
+      "loss": 0.206,
+      "step": 67620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0001801639491937e-05,
+      "loss": 0.2091,
+      "step": 67630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9997297540762094e-05,
+      "loss": 0.2132,
+      "step": 67640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.999279344203225e-05,
+      "loss": 0.2103,
+      "step": 67650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9988289343302406e-05,
+      "loss": 0.2083,
+      "step": 67660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9983785244572563e-05,
+      "loss": 0.2124,
+      "step": 67670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9979281145842717e-05,
+      "loss": 0.209,
+      "step": 67680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9974777047112875e-05,
+      "loss": 0.2055,
+      "step": 67690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.997027294838303e-05,
+      "loss": 0.2077,
+      "step": 67700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9965768849653186e-05,
+      "loss": 0.206,
+      "step": 67710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.996126475092334e-05,
+      "loss": 0.2041,
+      "step": 67720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9956760652193495e-05,
+      "loss": 0.2078,
+      "step": 67730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9952256553463652e-05,
+      "loss": 0.2085,
+      "step": 67740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.994775245473381e-05,
+      "loss": 0.2144,
+      "step": 67750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9943248356003967e-05,
+      "loss": 0.2128,
+      "step": 67760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.993874425727412e-05,
+      "loss": 0.2078,
+      "step": 67770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9934240158544275e-05,
+      "loss": 0.1997,
+      "step": 67780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9929736059814433e-05,
+      "loss": 0.2041,
+      "step": 67790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9925231961084587e-05,
+      "loss": 0.2056,
+      "step": 67800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9920727862354744e-05,
+      "loss": 0.2133,
+      "step": 67810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.99162237636249e-05,
+      "loss": 0.2055,
+      "step": 67820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9911719664895056e-05,
+      "loss": 0.2028,
+      "step": 67830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.990721556616521e-05,
+      "loss": 0.205,
+      "step": 67840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9902711467435368e-05,
+      "loss": 0.2061,
+      "step": 67850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9898207368705525e-05,
+      "loss": 0.2073,
+      "step": 67860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.989370326997568e-05,
+      "loss": 0.2162,
+      "step": 67870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9889199171245833e-05,
+      "loss": 0.2072,
+      "step": 67880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.988469507251599e-05,
+      "loss": 0.2126,
+      "step": 67890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9880190973786148e-05,
+      "loss": 0.2099,
+      "step": 67900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9875686875056302e-05,
+      "loss": 0.2112,
+      "step": 67910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9871182776326456e-05,
+      "loss": 0.2071,
+      "step": 67920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9866678677596614e-05,
+      "loss": 0.2125,
+      "step": 67930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9862174578866768e-05,
+      "loss": 0.2146,
+      "step": 67940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9857670480136926e-05,
+      "loss": 0.2074,
+      "step": 67950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9853166381407083e-05,
+      "loss": 0.2132,
+      "step": 67960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9848662282677237e-05,
+      "loss": 0.2059,
+      "step": 67970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.984415818394739e-05,
+      "loss": 0.2105,
+      "step": 67980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.983965408521755e-05,
+      "loss": 0.21,
+      "step": 67990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9835149986487706e-05,
+      "loss": 0.2111,
+      "step": 68000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.983064588775786e-05,
+      "loss": 0.2115,
+      "step": 68010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9826141789028014e-05,
+      "loss": 0.21,
+      "step": 68020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9821637690298172e-05,
+      "loss": 0.2084,
+      "step": 68030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.981713359156833e-05,
+      "loss": 0.2068,
+      "step": 68040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9812629492838484e-05,
+      "loss": 0.2115,
+      "step": 68050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.980812539410864e-05,
+      "loss": 0.208,
+      "step": 68060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9803621295378795e-05,
+      "loss": 0.208,
+      "step": 68070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.979911719664895e-05,
+      "loss": 0.2111,
+      "step": 68080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9794613097919107e-05,
+      "loss": 0.2136,
+      "step": 68090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9790108999189264e-05,
+      "loss": 0.2036,
+      "step": 68100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.978560490045942e-05,
+      "loss": 0.2081,
+      "step": 68110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9781100801729572e-05,
+      "loss": 0.2102,
+      "step": 68120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.977659670299973e-05,
+      "loss": 0.2098,
+      "step": 68130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9772092604269887e-05,
+      "loss": 0.2075,
+      "step": 68140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.976758850554004e-05,
+      "loss": 0.2073,
+      "step": 68150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.97630844068102e-05,
+      "loss": 0.2071,
+      "step": 68160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9758580308080353e-05,
+      "loss": 0.2126,
+      "step": 68170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.975407620935051e-05,
+      "loss": 0.2076,
+      "step": 68180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9749572110620665e-05,
+      "loss": 0.2108,
+      "step": 68190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9745068011890822e-05,
+      "loss": 0.206,
+      "step": 68200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.974056391316098e-05,
+      "loss": 0.2127,
+      "step": 68210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.973605981443113e-05,
+      "loss": 0.2107,
+      "step": 68220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9731555715701288e-05,
+      "loss": 0.2121,
+      "step": 68230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9727051616971445e-05,
+      "loss": 0.2051,
+      "step": 68240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9722547518241603e-05,
+      "loss": 0.209,
+      "step": 68250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9718043419511757e-05,
+      "loss": 0.2085,
+      "step": 68260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.971353932078191e-05,
+      "loss": 0.2066,
+      "step": 68270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.970903522205207e-05,
+      "loss": 0.2082,
+      "step": 68280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9704531123322223e-05,
+      "loss": 0.2094,
+      "step": 68290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.970002702459238e-05,
+      "loss": 0.2107,
+      "step": 68300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9695522925862538e-05,
+      "loss": 0.2094,
+      "step": 68310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9691018827132692e-05,
+      "loss": 0.2119,
+      "step": 68320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9686514728402846e-05,
+      "loss": 0.2081,
+      "step": 68330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9682010629673003e-05,
+      "loss": 0.2037,
+      "step": 68340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.967750653094316e-05,
+      "loss": 0.2096,
+      "step": 68350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9673002432213315e-05,
+      "loss": 0.2073,
+      "step": 68360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.966849833348347e-05,
+      "loss": 0.2039,
+      "step": 68370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9663994234753627e-05,
+      "loss": 0.2105,
+      "step": 68380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9659490136023784e-05,
+      "loss": 0.2083,
+      "step": 68390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9654986037293938e-05,
+      "loss": 0.2098,
+      "step": 68400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9650481938564096e-05,
+      "loss": 0.2178,
+      "step": 68410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.964597783983425e-05,
+      "loss": 0.2064,
+      "step": 68420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9641473741104404e-05,
+      "loss": 0.2054,
+      "step": 68430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.963696964237456e-05,
+      "loss": 0.2049,
+      "step": 68440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.963246554364472e-05,
+      "loss": 0.2137,
+      "step": 68450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9627961444914876e-05,
+      "loss": 0.2009,
+      "step": 68460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9623457346185027e-05,
+      "loss": 0.2055,
+      "step": 68470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9618953247455185e-05,
+      "loss": 0.2045,
+      "step": 68480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9614449148725342e-05,
+      "loss": 0.2064,
+      "step": 68490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9609945049995496e-05,
+      "loss": 0.2058,
+      "step": 68500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9605440951265654e-05,
+      "loss": 0.2094,
+      "step": 68510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9600936852535808e-05,
+      "loss": 0.2139,
+      "step": 68520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9596432753805965e-05,
+      "loss": 0.2055,
+      "step": 68530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.959192865507612e-05,
+      "loss": 0.2077,
+      "step": 68540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9587424556346277e-05,
+      "loss": 0.2098,
+      "step": 68550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9582920457616434e-05,
+      "loss": 0.211,
+      "step": 68560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9578416358886585e-05,
+      "loss": 0.2075,
+      "step": 68570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9573912260156743e-05,
+      "loss": 0.2088,
+      "step": 68580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.95694081614269e-05,
+      "loss": 0.2033,
+      "step": 68590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9564904062697057e-05,
+      "loss": 0.2068,
+      "step": 68600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.956039996396721e-05,
+      "loss": 0.2063,
+      "step": 68610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9555895865237366e-05,
+      "loss": 0.2049,
+      "step": 68620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9551391766507523e-05,
+      "loss": 0.203,
+      "step": 68630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9546887667777677e-05,
+      "loss": 0.2098,
+      "step": 68640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9542383569047835e-05,
+      "loss": 0.2116,
+      "step": 68650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9537879470317992e-05,
+      "loss": 0.2094,
+      "step": 68660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9533375371588146e-05,
+      "loss": 0.2068,
+      "step": 68670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.95288712728583e-05,
+      "loss": 0.2062,
+      "step": 68680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9524367174128458e-05,
+      "loss": 0.2049,
+      "step": 68690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9519863075398615e-05,
+      "loss": 0.2045,
+      "step": 68700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.951535897666877e-05,
+      "loss": 0.2053,
+      "step": 68710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9510854877938924e-05,
+      "loss": 0.2071,
+      "step": 68720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.950635077920908e-05,
+      "loss": 0.2081,
+      "step": 68730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.950184668047924e-05,
+      "loss": 0.2088,
+      "step": 68740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9497342581749393e-05,
+      "loss": 0.209,
+      "step": 68750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.949283848301955e-05,
+      "loss": 0.2129,
+      "step": 68760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9488334384289704e-05,
+      "loss": 0.2068,
+      "step": 68770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.948383028555986e-05,
+      "loss": 0.2076,
+      "step": 68780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9479326186830016e-05,
+      "loss": 0.2056,
+      "step": 68790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9474822088100173e-05,
+      "loss": 0.2111,
+      "step": 68800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9470317989370328e-05,
+      "loss": 0.1977,
+      "step": 68810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.946581389064048e-05,
+      "loss": 0.2042,
+      "step": 68820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.946130979191064e-05,
+      "loss": 0.2081,
+      "step": 68830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9456805693180797e-05,
+      "loss": 0.2072,
+      "step": 68840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.945230159445095e-05,
+      "loss": 0.2106,
+      "step": 68850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9447797495721108e-05,
+      "loss": 0.207,
+      "step": 68860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9443293396991262e-05,
+      "loss": 0.2069,
+      "step": 68870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.943878929826142e-05,
+      "loss": 0.2055,
+      "step": 68880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9434285199531574e-05,
+      "loss": 0.2061,
+      "step": 68890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.942978110080173e-05,
+      "loss": 0.211,
+      "step": 68900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9425277002071886e-05,
+      "loss": 0.2033,
+      "step": 68910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.942077290334204e-05,
+      "loss": 0.2098,
+      "step": 68920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9416268804612197e-05,
+      "loss": 0.2113,
+      "step": 68930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9411764705882355e-05,
+      "loss": 0.2032,
+      "step": 68940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9407260607152512e-05,
+      "loss": 0.2091,
+      "step": 68950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9402756508422666e-05,
+      "loss": 0.2057,
+      "step": 68960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.939825240969282e-05,
+      "loss": 0.2049,
+      "step": 68970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9393748310962978e-05,
+      "loss": 0.2079,
+      "step": 68980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9389244212233132e-05,
+      "loss": 0.1995,
+      "step": 68990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.938474011350329e-05,
+      "loss": 0.2056,
+      "step": 69000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9380236014773444e-05,
+      "loss": 0.2035,
+      "step": 69010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.93757319160436e-05,
+      "loss": 0.2053,
+      "step": 69020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9371227817313755e-05,
+      "loss": 0.2064,
+      "step": 69030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9366723718583913e-05,
+      "loss": 0.2078,
+      "step": 69040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.936221961985407e-05,
+      "loss": 0.2119,
+      "step": 69050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9357715521124224e-05,
+      "loss": 0.2057,
+      "step": 69060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.935321142239438e-05,
+      "loss": 0.21,
+      "step": 69070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9348707323664536e-05,
+      "loss": 0.2052,
+      "step": 69080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9344203224934693e-05,
+      "loss": 0.1972,
+      "step": 69090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9339699126204847e-05,
+      "loss": 0.2074,
+      "step": 69100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9335195027475005e-05,
+      "loss": 0.2091,
+      "step": 69110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.933069092874516e-05,
+      "loss": 0.2107,
+      "step": 69120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9326186830015313e-05,
+      "loss": 0.2034,
+      "step": 69130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.932168273128547e-05,
+      "loss": 0.2068,
+      "step": 69140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9317178632555628e-05,
+      "loss": 0.2096,
+      "step": 69150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9312674533825782e-05,
+      "loss": 0.209,
+      "step": 69160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9308170435095936e-05,
+      "loss": 0.2097,
+      "step": 69170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9303666336366094e-05,
+      "loss": 0.2115,
+      "step": 69180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.929916223763625e-05,
+      "loss": 0.2073,
+      "step": 69190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9294658138906405e-05,
+      "loss": 0.2011,
+      "step": 69200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9290154040176563e-05,
+      "loss": 0.2031,
+      "step": 69210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9285649941446717e-05,
+      "loss": 0.2005,
+      "step": 69220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9281596252589857e-05,
+      "loss": 0.209,
+      "step": 69230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.927709215386001e-05,
+      "loss": 0.1991,
+      "step": 69240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.927258805513017e-05,
+      "loss": 0.2003,
+      "step": 69250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9268083956400326e-05,
+      "loss": 0.2062,
+      "step": 69260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.926357985767048e-05,
+      "loss": 0.2015,
+      "step": 69270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9259075758940638e-05,
+      "loss": 0.2,
+      "step": 69280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9254571660210792e-05,
+      "loss": 0.2012,
+      "step": 69290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9250067561480946e-05,
+      "loss": 0.2082,
+      "step": 69300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9245563462751104e-05,
+      "loss": 0.209,
+      "step": 69310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.924105936402126e-05,
+      "loss": 0.2056,
+      "step": 69320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.923655526529142e-05,
+      "loss": 0.2109,
+      "step": 69330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.923205116656157e-05,
+      "loss": 0.2087,
+      "step": 69340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9227547067831727e-05,
+      "loss": 0.2086,
+      "step": 69350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9223042969101884e-05,
+      "loss": 0.2065,
+      "step": 69360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.921853887037204e-05,
+      "loss": 0.2087,
+      "step": 69370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9214034771642196e-05,
+      "loss": 0.2105,
+      "step": 69380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.920953067291235e-05,
+      "loss": 0.216,
+      "step": 69390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9205026574182508e-05,
+      "loss": 0.2126,
+      "step": 69400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9200522475452662e-05,
+      "loss": 0.2074,
+      "step": 69410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.919601837672282e-05,
+      "loss": 0.2108,
+      "step": 69420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9191514277992977e-05,
+      "loss": 0.2022,
+      "step": 69430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9187010179263127e-05,
+      "loss": 0.2108,
+      "step": 69440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9182506080533285e-05,
+      "loss": 0.2066,
+      "step": 69450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9178001981803442e-05,
+      "loss": 0.2087,
+      "step": 69460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.91734978830736e-05,
+      "loss": 0.2072,
+      "step": 69470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9168993784343754e-05,
+      "loss": 0.2072,
+      "step": 69480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9164489685613908e-05,
+      "loss": 0.2064,
+      "step": 69490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9159985586884066e-05,
+      "loss": 0.199,
+      "step": 69500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.915548148815422e-05,
+      "loss": 0.206,
+      "step": 69510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9150977389424377e-05,
+      "loss": 0.2086,
+      "step": 69520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9146473290694535e-05,
+      "loss": 0.2044,
+      "step": 69530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.914196919196469e-05,
+      "loss": 0.2038,
+      "step": 69540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9137465093234843e-05,
+      "loss": 0.2044,
+      "step": 69550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9132960994505e-05,
+      "loss": 0.2099,
+      "step": 69560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9128456895775158e-05,
+      "loss": 0.2033,
+      "step": 69570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9123952797045312e-05,
+      "loss": 0.2082,
+      "step": 69580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9119448698315466e-05,
+      "loss": 0.2041,
+      "step": 69590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9114944599585624e-05,
+      "loss": 0.2115,
+      "step": 69600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.911044050085578e-05,
+      "loss": 0.2045,
+      "step": 69610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9105936402125935e-05,
+      "loss": 0.2086,
+      "step": 69620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9101432303396093e-05,
+      "loss": 0.2063,
+      "step": 69630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9096928204666247e-05,
+      "loss": 0.2051,
+      "step": 69640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.90924241059364e-05,
+      "loss": 0.2088,
+      "step": 69650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.908792000720656e-05,
+      "loss": 0.2052,
+      "step": 69660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9083415908476716e-05,
+      "loss": 0.2061,
+      "step": 69670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.907891180974687e-05,
+      "loss": 0.2133,
+      "step": 69680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9074407711017024e-05,
+      "loss": 0.2054,
+      "step": 69690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.906990361228718e-05,
+      "loss": 0.2026,
+      "step": 69700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.906539951355734e-05,
+      "loss": 0.2038,
+      "step": 69710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9060895414827493e-05,
+      "loss": 0.213,
+      "step": 69720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.905639131609765e-05,
+      "loss": 0.202,
+      "step": 69730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9051887217367805e-05,
+      "loss": 0.205,
+      "step": 69740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9047383118637962e-05,
+      "loss": 0.2048,
+      "step": 69750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9042879019908116e-05,
+      "loss": 0.2053,
+      "step": 69760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9038374921178274e-05,
+      "loss": 0.2085,
+      "step": 69770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.903387082244843e-05,
+      "loss": 0.2083,
+      "step": 69780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9029366723718582e-05,
+      "loss": 0.2015,
+      "step": 69790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.902486262498874e-05,
+      "loss": 0.2078,
+      "step": 69800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9020358526258897e-05,
+      "loss": 0.2064,
+      "step": 69810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9015854427529055e-05,
+      "loss": 0.2115,
+      "step": 69820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.901135032879921e-05,
+      "loss": 0.2057,
+      "step": 69830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9006846230069363e-05,
+      "loss": 0.2073,
+      "step": 69840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.900234213133952e-05,
+      "loss": 0.206,
+      "step": 69850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8997838032609674e-05,
+      "loss": 0.202,
+      "step": 69860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8993333933879832e-05,
+      "loss": 0.2055,
+      "step": 69870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.898882983514999e-05,
+      "loss": 0.2029,
+      "step": 69880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8984325736420143e-05,
+      "loss": 0.2025,
+      "step": 69890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8979821637690298e-05,
+      "loss": 0.207,
+      "step": 69900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8975317538960455e-05,
+      "loss": 0.2049,
+      "step": 69910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8970813440230613e-05,
+      "loss": 0.2078,
+      "step": 69920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8966309341500767e-05,
+      "loss": 0.2072,
+      "step": 69930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.896180524277092e-05,
+      "loss": 0.2065,
+      "step": 69940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8957301144041078e-05,
+      "loss": 0.2055,
+      "step": 69950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8952797045311236e-05,
+      "loss": 0.2049,
+      "step": 69960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.894829294658139e-05,
+      "loss": 0.2062,
+      "step": 69970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8943788847851547e-05,
+      "loss": 0.2028,
+      "step": 69980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.89392847491217e-05,
+      "loss": 0.2085,
+      "step": 69990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8934780650391856e-05,
+      "loss": 0.2131,
+      "step": 70000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8930276551662013e-05,
+      "loss": 0.2083,
+      "step": 70010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.892577245293217e-05,
+      "loss": 0.1996,
+      "step": 70020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8921268354202325e-05,
+      "loss": 0.2099,
+      "step": 70030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.891676425547248e-05,
+      "loss": 0.2089,
+      "step": 70040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8912260156742636e-05,
+      "loss": 0.2031,
+      "step": 70050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8907756058012794e-05,
+      "loss": 0.2077,
+      "step": 70060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8903251959282948e-05,
+      "loss": 0.2092,
+      "step": 70070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8898747860553105e-05,
+      "loss": 0.2042,
+      "step": 70080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.889424376182326e-05,
+      "loss": 0.1994,
+      "step": 70090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8889739663093417e-05,
+      "loss": 0.2043,
+      "step": 70100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.888523556436357e-05,
+      "loss": 0.2034,
+      "step": 70110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.888073146563373e-05,
+      "loss": 0.2041,
+      "step": 70120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8876227366903883e-05,
+      "loss": 0.2077,
+      "step": 70130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8871723268174037e-05,
+      "loss": 0.2043,
+      "step": 70140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8867219169444194e-05,
+      "loss": 0.2117,
+      "step": 70150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.886271507071435e-05,
+      "loss": 0.2099,
+      "step": 70160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.885821097198451e-05,
+      "loss": 0.2062,
+      "step": 70170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8853706873254663e-05,
+      "loss": 0.2019,
+      "step": 70180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8849202774524817e-05,
+      "loss": 0.1999,
+      "step": 70190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8844698675794975e-05,
+      "loss": 0.2052,
+      "step": 70200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.884019457706513e-05,
+      "loss": 0.2153,
+      "step": 70210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8835690478335286e-05,
+      "loss": 0.2065,
+      "step": 70220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.883118637960544e-05,
+      "loss": 0.2025,
+      "step": 70230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8826682280875598e-05,
+      "loss": 0.2032,
+      "step": 70240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8822178182145752e-05,
+      "loss": 0.207,
+      "step": 70250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.881767408341591e-05,
+      "loss": 0.198,
+      "step": 70260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8813169984686067e-05,
+      "loss": 0.2099,
+      "step": 70270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.880866588595622e-05,
+      "loss": 0.2051,
+      "step": 70280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8804161787226375e-05,
+      "loss": 0.2064,
+      "step": 70290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8799657688496533e-05,
+      "loss": 0.2052,
+      "step": 70300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.879515358976669e-05,
+      "loss": 0.2046,
+      "step": 70310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8790649491036844e-05,
+      "loss": 0.2057,
+      "step": 70320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8786145392307e-05,
+      "loss": 0.2013,
+      "step": 70330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8781641293577156e-05,
+      "loss": 0.2056,
+      "step": 70340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.877713719484731e-05,
+      "loss": 0.2041,
+      "step": 70350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8772633096117468e-05,
+      "loss": 0.2038,
+      "step": 70360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8768128997387625e-05,
+      "loss": 0.2049,
+      "step": 70370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.876362489865778e-05,
+      "loss": 0.2051,
+      "step": 70380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8759120799927933e-05,
+      "loss": 0.208,
+      "step": 70390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.875461670119809e-05,
+      "loss": 0.2021,
+      "step": 70400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.875011260246825e-05,
+      "loss": 0.2011,
+      "step": 70410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8745608503738402e-05,
+      "loss": 0.2058,
+      "step": 70420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.874110440500856e-05,
+      "loss": 0.206,
+      "step": 70430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8736600306278714e-05,
+      "loss": 0.2028,
+      "step": 70440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.873209620754887e-05,
+      "loss": 0.2046,
+      "step": 70450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8727592108819026e-05,
+      "loss": 0.209,
+      "step": 70460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8723088010089183e-05,
+      "loss": 0.2082,
+      "step": 70470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8718583911359337e-05,
+      "loss": 0.2038,
+      "step": 70480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.871407981262949e-05,
+      "loss": 0.1967,
+      "step": 70490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.870957571389965e-05,
+      "loss": 0.2049,
+      "step": 70500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8705071615169806e-05,
+      "loss": 0.2093,
+      "step": 70510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8700567516439964e-05,
+      "loss": 0.2057,
+      "step": 70520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8696063417710118e-05,
+      "loss": 0.2079,
+      "step": 70530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8691559318980272e-05,
+      "loss": 0.203,
+      "step": 70540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.868705522025043e-05,
+      "loss": 0.2079,
+      "step": 70550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8682551121520584e-05,
+      "loss": 0.2038,
+      "step": 70560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.867804702279074e-05,
+      "loss": 0.2082,
+      "step": 70570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8673542924060895e-05,
+      "loss": 0.2045,
+      "step": 70580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8669038825331053e-05,
+      "loss": 0.2047,
+      "step": 70590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8664534726601207e-05,
+      "loss": 0.2024,
+      "step": 70600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8660030627871364e-05,
+      "loss": 0.2034,
+      "step": 70610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8655526529141522e-05,
+      "loss": 0.2053,
+      "step": 70620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8651022430411676e-05,
+      "loss": 0.2067,
+      "step": 70630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.864651833168183e-05,
+      "loss": 0.2056,
+      "step": 70640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8642014232951987e-05,
+      "loss": 0.204,
+      "step": 70650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8637510134222145e-05,
+      "loss": 0.2061,
+      "step": 70660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.86330060354923e-05,
+      "loss": 0.2045,
+      "step": 70670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8628501936762453e-05,
+      "loss": 0.1991,
+      "step": 70680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.862399783803261e-05,
+      "loss": 0.2045,
+      "step": 70690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8619493739302765e-05,
+      "loss": 0.2029,
+      "step": 70700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8614989640572922e-05,
+      "loss": 0.2098,
+      "step": 70710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.861048554184308e-05,
+      "loss": 0.2064,
+      "step": 70720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8605981443113234e-05,
+      "loss": 0.2056,
+      "step": 70730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8601477344383388e-05,
+      "loss": 0.2044,
+      "step": 70740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8596973245653545e-05,
+      "loss": 0.2031,
+      "step": 70750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8592469146923703e-05,
+      "loss": 0.2033,
+      "step": 70760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8587965048193857e-05,
+      "loss": 0.2077,
+      "step": 70770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.858346094946401e-05,
+      "loss": 0.2041,
+      "step": 70780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.857895685073417e-05,
+      "loss": 0.1976,
+      "step": 70790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8574452752004326e-05,
+      "loss": 0.2035,
+      "step": 70800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.856994865327448e-05,
+      "loss": 0.2056,
+      "step": 70810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8565444554544638e-05,
+      "loss": 0.2057,
+      "step": 70820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8560940455814792e-05,
+      "loss": 0.1986,
+      "step": 70830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8556436357084946e-05,
+      "loss": 0.2086,
+      "step": 70840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8551932258355103e-05,
+      "loss": 0.2028,
+      "step": 70850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.854742815962526e-05,
+      "loss": 0.2006,
+      "step": 70860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.854292406089542e-05,
+      "loss": 0.2071,
+      "step": 70870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.853841996216557e-05,
+      "loss": 0.2059,
+      "step": 70880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8533915863435727e-05,
+      "loss": 0.1988,
+      "step": 70890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8529411764705884e-05,
+      "loss": 0.2052,
+      "step": 70900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8524907665976038e-05,
+      "loss": 0.2011,
+      "step": 70910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8520403567246196e-05,
+      "loss": 0.2031,
+      "step": 70920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.851589946851635e-05,
+      "loss": 0.2122,
+      "step": 70930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8511395369786507e-05,
+      "loss": 0.2038,
+      "step": 70940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.850689127105666e-05,
+      "loss": 0.2036,
+      "step": 70950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.850238717232682e-05,
+      "loss": 0.2065,
+      "step": 70960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8497883073596976e-05,
+      "loss": 0.1976,
+      "step": 70970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8493378974867127e-05,
+      "loss": 0.2023,
+      "step": 70980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8488874876137285e-05,
+      "loss": 0.2051,
+      "step": 70990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8484370777407442e-05,
+      "loss": 0.2092,
+      "step": 71000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.84798666786776e-05,
+      "loss": 0.2041,
+      "step": 71010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8475362579947754e-05,
+      "loss": 0.2019,
+      "step": 71020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8470858481217908e-05,
+      "loss": 0.2051,
+      "step": 71030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8466354382488065e-05,
+      "loss": 0.2068,
+      "step": 71040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.846185028375822e-05,
+      "loss": 0.2008,
+      "step": 71050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8457346185028377e-05,
+      "loss": 0.2085,
+      "step": 71060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8452842086298534e-05,
+      "loss": 0.2039,
+      "step": 71070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.844833798756869e-05,
+      "loss": 0.2038,
+      "step": 71080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8443833888838843e-05,
+      "loss": 0.2086,
+      "step": 71090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8439329790109e-05,
+      "loss": 0.2036,
+      "step": 71100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8434825691379158e-05,
+      "loss": 0.2002,
+      "step": 71110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.843032159264931e-05,
+      "loss": 0.2052,
+      "step": 71120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8425817493919466e-05,
+      "loss": 0.2062,
+      "step": 71130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8421313395189623e-05,
+      "loss": 0.2079,
+      "step": 71140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.841680929645978e-05,
+      "loss": 0.2048,
+      "step": 71150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8412305197729935e-05,
+      "loss": 0.2088,
+      "step": 71160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8407801099000092e-05,
+      "loss": 0.1988,
+      "step": 71170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8403297000270246e-05,
+      "loss": 0.2001,
+      "step": 71180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.83987929015404e-05,
+      "loss": 0.2015,
+      "step": 71190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8394288802810558e-05,
+      "loss": 0.2043,
+      "step": 71200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8389784704080716e-05,
+      "loss": 0.2078,
+      "step": 71210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8385280605350873e-05,
+      "loss": 0.2011,
+      "step": 71220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8380776506621024e-05,
+      "loss": 0.2076,
+      "step": 71230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.837627240789118e-05,
+      "loss": 0.1981,
+      "step": 71240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.837176830916134e-05,
+      "loss": 0.2124,
+      "step": 71250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8367264210431493e-05,
+      "loss": 0.202,
+      "step": 71260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.836276011170165e-05,
+      "loss": 0.2043,
+      "step": 71270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8358256012971804e-05,
+      "loss": 0.2021,
+      "step": 71280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8354202324114945e-05,
+      "loss": 0.2058,
+      "step": 71290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8349698225385102e-05,
+      "loss": 0.2051,
+      "step": 71300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8345194126655256e-05,
+      "loss": 0.1998,
+      "step": 71310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8340690027925414e-05,
+      "loss": 0.202,
+      "step": 71320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8336185929195568e-05,
+      "loss": 0.2016,
+      "step": 71330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8331681830465726e-05,
+      "loss": 0.1981,
+      "step": 71340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.832717773173588e-05,
+      "loss": 0.2146,
+      "step": 71350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8322673633006034e-05,
+      "loss": 0.2121,
+      "step": 71360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.831816953427619e-05,
+      "loss": 0.2049,
+      "step": 71370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.831366543554635e-05,
+      "loss": 0.206,
+      "step": 71380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8309161336816506e-05,
+      "loss": 0.2066,
+      "step": 71390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.830465723808666e-05,
+      "loss": 0.2003,
+      "step": 71400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8300153139356814e-05,
+      "loss": 0.2056,
+      "step": 71410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8295649040626972e-05,
+      "loss": 0.2037,
+      "step": 71420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8291144941897126e-05,
+      "loss": 0.2041,
+      "step": 71430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8286640843167283e-05,
+      "loss": 0.2031,
+      "step": 71440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8282136744437438e-05,
+      "loss": 0.2046,
+      "step": 71450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8277632645707595e-05,
+      "loss": 0.2073,
+      "step": 71460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.827312854697775e-05,
+      "loss": 0.2043,
+      "step": 71470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8268624448247907e-05,
+      "loss": 0.2027,
+      "step": 71480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8264120349518064e-05,
+      "loss": 0.2034,
+      "step": 71490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8259616250788218e-05,
+      "loss": 0.2042,
+      "step": 71500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8255112152058372e-05,
+      "loss": 0.2028,
+      "step": 71510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.825060805332853e-05,
+      "loss": 0.2028,
+      "step": 71520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8246103954598687e-05,
+      "loss": 0.2032,
+      "step": 71530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.824159985586884e-05,
+      "loss": 0.206,
+      "step": 71540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8237095757138996e-05,
+      "loss": 0.201,
+      "step": 71550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8232591658409153e-05,
+      "loss": 0.2026,
+      "step": 71560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8228087559679307e-05,
+      "loss": 0.1997,
+      "step": 71570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8223583460949465e-05,
+      "loss": 0.2023,
+      "step": 71580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8219079362219622e-05,
+      "loss": 0.2014,
+      "step": 71590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8214575263489776e-05,
+      "loss": 0.2066,
+      "step": 71600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.821007116475993e-05,
+      "loss": 0.2047,
+      "step": 71610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8205567066030088e-05,
+      "loss": 0.2035,
+      "step": 71620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8201062967300245e-05,
+      "loss": 0.2061,
+      "step": 71630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.81965588685704e-05,
+      "loss": 0.2003,
+      "step": 71640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8192054769840554e-05,
+      "loss": 0.2095,
+      "step": 71650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.818755067111071e-05,
+      "loss": 0.2049,
+      "step": 71660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.818304657238087e-05,
+      "loss": 0.2023,
+      "step": 71670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8178542473651023e-05,
+      "loss": 0.2072,
+      "step": 71680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.817403837492118e-05,
+      "loss": 0.2007,
+      "step": 71690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8169534276191334e-05,
+      "loss": 0.2038,
+      "step": 71700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.816503017746149e-05,
+      "loss": 0.2041,
+      "step": 71710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8160526078731646e-05,
+      "loss": 0.2086,
+      "step": 71720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8156021980001803e-05,
+      "loss": 0.209,
+      "step": 71730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.815151788127196e-05,
+      "loss": 0.2026,
+      "step": 71740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8147013782542115e-05,
+      "loss": 0.2044,
+      "step": 71750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.814250968381227e-05,
+      "loss": 0.2033,
+      "step": 71760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8138005585082427e-05,
+      "loss": 0.208,
+      "step": 71770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.813350148635258e-05,
+      "loss": 0.2051,
+      "step": 71780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8128997387622738e-05,
+      "loss": 0.2039,
+      "step": 71790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8124493288892892e-05,
+      "loss": 0.2061,
+      "step": 71800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.811998919016305e-05,
+      "loss": 0.2043,
+      "step": 71810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8115485091433204e-05,
+      "loss": 0.2054,
+      "step": 71820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.811098099270336e-05,
+      "loss": 0.2043,
+      "step": 71830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.810647689397352e-05,
+      "loss": 0.2061,
+      "step": 71840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8101972795243673e-05,
+      "loss": 0.2032,
+      "step": 71850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8097468696513827e-05,
+      "loss": 0.2032,
+      "step": 71860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8092964597783985e-05,
+      "loss": 0.2095,
+      "step": 71870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8088460499054142e-05,
+      "loss": 0.202,
+      "step": 71880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8083956400324296e-05,
+      "loss": 0.2027,
+      "step": 71890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.807945230159445e-05,
+      "loss": 0.2013,
+      "step": 71900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8074948202864608e-05,
+      "loss": 0.205,
+      "step": 71910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8070444104134762e-05,
+      "loss": 0.2006,
+      "step": 71920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.806594000540492e-05,
+      "loss": 0.2032,
+      "step": 71930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8061435906675077e-05,
+      "loss": 0.2078,
+      "step": 71940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.805693180794523e-05,
+      "loss": 0.1991,
+      "step": 71950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8052427709215385e-05,
+      "loss": 0.2041,
+      "step": 71960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8047923610485542e-05,
+      "loss": 0.203,
+      "step": 71970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.80434195117557e-05,
+      "loss": 0.1987,
+      "step": 71980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8038915413025854e-05,
+      "loss": 0.2013,
+      "step": 71990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8034411314296008e-05,
+      "loss": 0.2053,
+      "step": 72000
+    },
+    {
+      "epoch": 0.64,
+      "eval_NEIMS_canon_smiles": 0.9437333333333333,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.04493333333333333,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.4432652596162006,
+      "eval_NEIMS_exact_mols": 0.04465,
+      "eval_NEIMS_exact_smiles": 0.04305,
+      "eval_NEIMS_loss": 0.2352963238954544,
+      "eval_NEIMS_matched_formulas": 0.5242666666666667,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.3444219567794077,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.044866666666666666,
+      "eval_NEIMS_runtime": 714.831,
+      "eval_NEIMS_samples_per_second": 83.936,
+      "eval_NEIMS_steps_per_second": 1.312,
+      "step": 72000
+    },
+    {
+      "epoch": 0.64,
+      "eval_RASSP_canon_smiles": 0.933449093398337,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.10435102013557285,
+      "eval_RASSP_daylight_tanimoto_simil": 0.5564372091100196,
+      "eval_RASSP_exact_mols": 0.10396700838147394,
+      "eval_RASSP_exact_smiles": 0.1006277757371356,
+      "eval_RASSP_loss": 0.17497843503952026,
+      "eval_RASSP_matched_formulas": 0.7069155508064247,
+      "eval_RASSP_morgan_tanimoto_simil": 0.42292690292996077,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.10403379303436071,
+      "eval_RASSP_runtime": 814.6418,
+      "eval_RASSP_samples_per_second": 73.522,
+      "eval_RASSP_steps_per_second": 1.149,
+      "step": 72000
+    },
+    {
+      "epoch": 0.64,
+      "eval_NIST_canon_smiles": 0.886858075735529,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.01164069986158924,
+      "eval_NIST_daylight_tanimoto_simil": 0.264120852504151,
+      "eval_NIST_exact_mols": 0.011214820598360365,
+      "eval_NIST_exact_smiles": 0.010292082194697804,
+      "eval_NIST_loss": 1.469948649406433,
+      "eval_NIST_matched_formulas": 0.07598395854775171,
+      "eval_NIST_morgan_tanimoto_simil": 0.21686319882447022,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.01164069986158924,
+      "eval_NIST_runtime": 460.465,
+      "eval_NIST_samples_per_second": 61.192,
+      "eval_NIST_steps_per_second": 0.958,
+      "step": 72000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8029907215566166e-05,
+      "loss": 0.2084,
+      "step": 72010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8025403116836323e-05,
+      "loss": 0.2058,
+      "step": 72020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8020899018106477e-05,
+      "loss": 0.2041,
+      "step": 72030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8016394919376635e-05,
+      "loss": 0.2053,
+      "step": 72040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.801189082064679e-05,
+      "loss": 0.2039,
+      "step": 72050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8007386721916943e-05,
+      "loss": 0.2023,
+      "step": 72060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.80028826231871e-05,
+      "loss": 0.2014,
+      "step": 72070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7998378524457258e-05,
+      "loss": 0.2092,
+      "step": 72080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7993874425727415e-05,
+      "loss": 0.2053,
+      "step": 72090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7989370326997566e-05,
+      "loss": 0.2067,
+      "step": 72100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7984866228267724e-05,
+      "loss": 0.2023,
+      "step": 72110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.798036212953788e-05,
+      "loss": 0.2052,
+      "step": 72120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7975858030808035e-05,
+      "loss": 0.2007,
+      "step": 72130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7971353932078193e-05,
+      "loss": 0.2017,
+      "step": 72140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7966849833348347e-05,
+      "loss": 0.2011,
+      "step": 72150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7962345734618504e-05,
+      "loss": 0.2067,
+      "step": 72160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.795784163588866e-05,
+      "loss": 0.2022,
+      "step": 72170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7953337537158816e-05,
+      "loss": 0.2021,
+      "step": 72180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7948833438428973e-05,
+      "loss": 0.1973,
+      "step": 72190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7944329339699124e-05,
+      "loss": 0.2088,
+      "step": 72200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.793982524096928e-05,
+      "loss": 0.1975,
+      "step": 72210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.793532114223944e-05,
+      "loss": 0.2057,
+      "step": 72220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7930817043509597e-05,
+      "loss": 0.1998,
+      "step": 72230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.792631294477975e-05,
+      "loss": 0.1966,
+      "step": 72240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7921808846049905e-05,
+      "loss": 0.2025,
+      "step": 72250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7917304747320062e-05,
+      "loss": 0.2061,
+      "step": 72260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7912800648590216e-05,
+      "loss": 0.2082,
+      "step": 72270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7908296549860374e-05,
+      "loss": 0.2035,
+      "step": 72280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.790379245113053e-05,
+      "loss": 0.2047,
+      "step": 72290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7899288352400686e-05,
+      "loss": 0.2006,
+      "step": 72300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.789478425367084e-05,
+      "loss": 0.2002,
+      "step": 72310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7890280154940997e-05,
+      "loss": 0.1968,
+      "step": 72320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7885776056211155e-05,
+      "loss": 0.1959,
+      "step": 72330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.788127195748131e-05,
+      "loss": 0.2076,
+      "step": 72340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7876767858751463e-05,
+      "loss": 0.1988,
+      "step": 72350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.787226376002162e-05,
+      "loss": 0.2009,
+      "step": 72360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7867759661291778e-05,
+      "loss": 0.2069,
+      "step": 72370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7863255562561932e-05,
+      "loss": 0.2003,
+      "step": 72380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.785875146383209e-05,
+      "loss": 0.1995,
+      "step": 72390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7854247365102244e-05,
+      "loss": 0.2025,
+      "step": 72400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7849743266372398e-05,
+      "loss": 0.2106,
+      "step": 72410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7845239167642555e-05,
+      "loss": 0.204,
+      "step": 72420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7840735068912713e-05,
+      "loss": 0.2034,
+      "step": 72430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7836230970182867e-05,
+      "loss": 0.2011,
+      "step": 72440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.783172687145302e-05,
+      "loss": 0.2063,
+      "step": 72450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7827222772723178e-05,
+      "loss": 0.2006,
+      "step": 72460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7822718673993336e-05,
+      "loss": 0.2006,
+      "step": 72470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.781821457526349e-05,
+      "loss": 0.2016,
+      "step": 72480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7813710476533647e-05,
+      "loss": 0.1965,
+      "step": 72490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.78092063778038e-05,
+      "loss": 0.204,
+      "step": 72500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.780470227907396e-05,
+      "loss": 0.2028,
+      "step": 72510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7800198180344113e-05,
+      "loss": 0.2069,
+      "step": 72520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.779569408161427e-05,
+      "loss": 0.2,
+      "step": 72530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7791189982884428e-05,
+      "loss": 0.2031,
+      "step": 72540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778668588415458e-05,
+      "loss": 0.205,
+      "step": 72550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7782181785424736e-05,
+      "loss": 0.1961,
+      "step": 72560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7777677686694894e-05,
+      "loss": 0.2051,
+      "step": 72570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777317358796505e-05,
+      "loss": 0.1994,
+      "step": 72580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7768669489235205e-05,
+      "loss": 0.2057,
+      "step": 72590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.776416539050536e-05,
+      "loss": 0.1978,
+      "step": 72600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7759661291775517e-05,
+      "loss": 0.2046,
+      "step": 72610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775515719304567e-05,
+      "loss": 0.2042,
+      "step": 72620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775065309431583e-05,
+      "loss": 0.2043,
+      "step": 72630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7746148995585986e-05,
+      "loss": 0.2043,
+      "step": 72640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.774164489685614e-05,
+      "loss": 0.2081,
+      "step": 72650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7737140798126294e-05,
+      "loss": 0.1993,
+      "step": 72660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7732636699396452e-05,
+      "loss": 0.2054,
+      "step": 72670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.772813260066661e-05,
+      "loss": 0.207,
+      "step": 72680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7723628501936763e-05,
+      "loss": 0.2052,
+      "step": 72690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7719124403206917e-05,
+      "loss": 0.2025,
+      "step": 72700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7714620304477075e-05,
+      "loss": 0.2066,
+      "step": 72710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7710116205747232e-05,
+      "loss": 0.2014,
+      "step": 72720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7705612107017387e-05,
+      "loss": 0.2033,
+      "step": 72730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7701108008287544e-05,
+      "loss": 0.2013,
+      "step": 72740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7696603909557698e-05,
+      "loss": 0.2031,
+      "step": 72750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7692099810827852e-05,
+      "loss": 0.1943,
+      "step": 72760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.768759571209801e-05,
+      "loss": 0.1996,
+      "step": 72770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7683091613368167e-05,
+      "loss": 0.1973,
+      "step": 72780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.767858751463832e-05,
+      "loss": 0.2007,
+      "step": 72790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7674083415908475e-05,
+      "loss": 0.1985,
+      "step": 72800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7669579317178633e-05,
+      "loss": 0.2033,
+      "step": 72810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.766507521844879e-05,
+      "loss": 0.1982,
+      "step": 72820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7660571119718945e-05,
+      "loss": 0.2025,
+      "step": 72830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7656067020989102e-05,
+      "loss": 0.2067,
+      "step": 72840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7651562922259256e-05,
+      "loss": 0.1982,
+      "step": 72850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7647058823529414e-05,
+      "loss": 0.2014,
+      "step": 72860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7642554724799568e-05,
+      "loss": 0.1974,
+      "step": 72870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7638050626069725e-05,
+      "loss": 0.2024,
+      "step": 72880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.763354652733988e-05,
+      "loss": 0.2022,
+      "step": 72890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7629042428610033e-05,
+      "loss": 0.1998,
+      "step": 72900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.762453832988019e-05,
+      "loss": 0.2054,
+      "step": 72910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.762003423115035e-05,
+      "loss": 0.2058,
+      "step": 72920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7615530132420506e-05,
+      "loss": 0.2022,
+      "step": 72930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.761102603369066e-05,
+      "loss": 0.2036,
+      "step": 72940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7606521934960814e-05,
+      "loss": 0.2105,
+      "step": 72950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.760201783623097e-05,
+      "loss": 0.2026,
+      "step": 72960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7597513737501126e-05,
+      "loss": 0.2047,
+      "step": 72970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7593009638771283e-05,
+      "loss": 0.2084,
+      "step": 72980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7588505540041437e-05,
+      "loss": 0.2027,
+      "step": 72990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7584001441311595e-05,
+      "loss": 0.1996,
+      "step": 73000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.757949734258175e-05,
+      "loss": 0.194,
+      "step": 73010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7574993243851906e-05,
+      "loss": 0.2069,
+      "step": 73020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7570489145122064e-05,
+      "loss": 0.2058,
+      "step": 73030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7565985046392218e-05,
+      "loss": 0.2021,
+      "step": 73040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7561480947662372e-05,
+      "loss": 0.2071,
+      "step": 73050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.755697684893253e-05,
+      "loss": 0.2085,
+      "step": 73060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7552472750202687e-05,
+      "loss": 0.2036,
+      "step": 73070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.754796865147284e-05,
+      "loss": 0.2024,
+      "step": 73080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7543464552742995e-05,
+      "loss": 0.2021,
+      "step": 73090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7538960454013153e-05,
+      "loss": 0.1966,
+      "step": 73100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7534456355283307e-05,
+      "loss": 0.2057,
+      "step": 73110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7529952256553464e-05,
+      "loss": 0.2021,
+      "step": 73120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7525448157823622e-05,
+      "loss": 0.1977,
+      "step": 73130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7520944059093776e-05,
+      "loss": 0.2027,
+      "step": 73140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.751643996036393e-05,
+      "loss": 0.1964,
+      "step": 73150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7511935861634088e-05,
+      "loss": 0.1992,
+      "step": 73160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7507431762904245e-05,
+      "loss": 0.2022,
+      "step": 73170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.75029276641744e-05,
+      "loss": 0.1994,
+      "step": 73180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7498423565444557e-05,
+      "loss": 0.2068,
+      "step": 73190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.749391946671471e-05,
+      "loss": 0.2061,
+      "step": 73200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7489415367984868e-05,
+      "loss": 0.1982,
+      "step": 73210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7484911269255022e-05,
+      "loss": 0.2051,
+      "step": 73220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.748040717052518e-05,
+      "loss": 0.2015,
+      "step": 73230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7475903071795334e-05,
+      "loss": 0.2029,
+      "step": 73240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7471398973065488e-05,
+      "loss": 0.2053,
+      "step": 73250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7466894874335646e-05,
+      "loss": 0.2009,
+      "step": 73260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7462390775605803e-05,
+      "loss": 0.2048,
+      "step": 73270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.745788667687596e-05,
+      "loss": 0.2049,
+      "step": 73280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7453382578146115e-05,
+      "loss": 0.1988,
+      "step": 73290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.744887847941627e-05,
+      "loss": 0.2027,
+      "step": 73300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7444374380686426e-05,
+      "loss": 0.2025,
+      "step": 73310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.743987028195658e-05,
+      "loss": 0.2037,
+      "step": 73320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7435366183226738e-05,
+      "loss": 0.2012,
+      "step": 73330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7430862084496892e-05,
+      "loss": 0.199,
+      "step": 73340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.742635798576705e-05,
+      "loss": 0.2058,
+      "step": 73350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7421853887037204e-05,
+      "loss": 0.1993,
+      "step": 73360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.741734978830736e-05,
+      "loss": 0.2064,
+      "step": 73370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.741284568957752e-05,
+      "loss": 0.2022,
+      "step": 73380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7408341590847673e-05,
+      "loss": 0.1979,
+      "step": 73390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7403837492117827e-05,
+      "loss": 0.205,
+      "step": 73400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7399333393387984e-05,
+      "loss": 0.2066,
+      "step": 73410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.739482929465814e-05,
+      "loss": 0.2039,
+      "step": 73420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7390325195928296e-05,
+      "loss": 0.2019,
+      "step": 73430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.738582109719845e-05,
+      "loss": 0.2019,
+      "step": 73440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7381316998468607e-05,
+      "loss": 0.203,
+      "step": 73450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.737681289973876e-05,
+      "loss": 0.2032,
+      "step": 73460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.737230880100892e-05,
+      "loss": 0.202,
+      "step": 73470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7367804702279076e-05,
+      "loss": 0.1998,
+      "step": 73480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.736330060354923e-05,
+      "loss": 0.2062,
+      "step": 73490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7358796504819385e-05,
+      "loss": 0.1978,
+      "step": 73500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7354292406089542e-05,
+      "loss": 0.1992,
+      "step": 73510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.73497883073597e-05,
+      "loss": 0.2015,
+      "step": 73520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7345284208629854e-05,
+      "loss": 0.2028,
+      "step": 73530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7340780109900008e-05,
+      "loss": 0.2015,
+      "step": 73540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7336276011170165e-05,
+      "loss": 0.2047,
+      "step": 73550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7331771912440323e-05,
+      "loss": 0.2035,
+      "step": 73560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7327267813710477e-05,
+      "loss": 0.1995,
+      "step": 73570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7322763714980634e-05,
+      "loss": 0.1976,
+      "step": 73580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.731825961625079e-05,
+      "loss": 0.2022,
+      "step": 73590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7313755517520943e-05,
+      "loss": 0.2007,
+      "step": 73600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7309701828664086e-05,
+      "loss": 0.1989,
+      "step": 73610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.730519772993424e-05,
+      "loss": 0.2015,
+      "step": 73620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7300693631204395e-05,
+      "loss": 0.2038,
+      "step": 73630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7296189532474552e-05,
+      "loss": 0.2028,
+      "step": 73640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.729168543374471e-05,
+      "loss": 0.1958,
+      "step": 73650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7287181335014864e-05,
+      "loss": 0.1953,
+      "step": 73660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7282677236285018e-05,
+      "loss": 0.2058,
+      "step": 73670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7278173137555175e-05,
+      "loss": 0.2025,
+      "step": 73680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7273669038825333e-05,
+      "loss": 0.2087,
+      "step": 73690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7269164940095487e-05,
+      "loss": 0.2089,
+      "step": 73700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7264660841365644e-05,
+      "loss": 0.204,
+      "step": 73710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.72601567426358e-05,
+      "loss": 0.2006,
+      "step": 73720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7255652643905956e-05,
+      "loss": 0.1996,
+      "step": 73730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.725114854517611e-05,
+      "loss": 0.2056,
+      "step": 73740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7246644446446268e-05,
+      "loss": 0.2026,
+      "step": 73750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7242140347716422e-05,
+      "loss": 0.2011,
+      "step": 73760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7237636248986576e-05,
+      "loss": 0.2017,
+      "step": 73770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7233132150256733e-05,
+      "loss": 0.2057,
+      "step": 73780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.722862805152689e-05,
+      "loss": 0.2024,
+      "step": 73790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7224123952797048e-05,
+      "loss": 0.1989,
+      "step": 73800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7219619854067202e-05,
+      "loss": 0.1998,
+      "step": 73810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7215115755337357e-05,
+      "loss": 0.2033,
+      "step": 73820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7210611656607514e-05,
+      "loss": 0.2022,
+      "step": 73830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7206107557877668e-05,
+      "loss": 0.1997,
+      "step": 73840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7201603459147826e-05,
+      "loss": 0.1993,
+      "step": 73850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7197099360417983e-05,
+      "loss": 0.197,
+      "step": 73860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7192595261688137e-05,
+      "loss": 0.2021,
+      "step": 73870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.718809116295829e-05,
+      "loss": 0.2031,
+      "step": 73880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.718358706422845e-05,
+      "loss": 0.2022,
+      "step": 73890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7179082965498606e-05,
+      "loss": 0.2061,
+      "step": 73900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.717457886676876e-05,
+      "loss": 0.1981,
+      "step": 73910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7170074768038914e-05,
+      "loss": 0.2004,
+      "step": 73920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7165570669309072e-05,
+      "loss": 0.1979,
+      "step": 73930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.716106657057923e-05,
+      "loss": 0.2014,
+      "step": 73940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7156562471849384e-05,
+      "loss": 0.2005,
+      "step": 73950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.715205837311954e-05,
+      "loss": 0.195,
+      "step": 73960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7147554274389695e-05,
+      "loss": 0.1979,
+      "step": 73970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.714305017565985e-05,
+      "loss": 0.1981,
+      "step": 73980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7138546076930007e-05,
+      "loss": 0.2027,
+      "step": 73990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7134041978200164e-05,
+      "loss": 0.1983,
+      "step": 74000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.712953787947032e-05,
+      "loss": 0.2008,
+      "step": 74010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7125033780740472e-05,
+      "loss": 0.2052,
+      "step": 74020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.712052968201063e-05,
+      "loss": 0.2042,
+      "step": 74030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7116025583280787e-05,
+      "loss": 0.2025,
+      "step": 74040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.711152148455094e-05,
+      "loss": 0.2051,
+      "step": 74050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.71070173858211e-05,
+      "loss": 0.2032,
+      "step": 74060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7102513287091253e-05,
+      "loss": 0.2002,
+      "step": 74070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.709800918836141e-05,
+      "loss": 0.1983,
+      "step": 74080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7093505089631565e-05,
+      "loss": 0.1913,
+      "step": 74090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7089000990901722e-05,
+      "loss": 0.1961,
+      "step": 74100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7084496892171876e-05,
+      "loss": 0.1987,
+      "step": 74110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.707999279344203e-05,
+      "loss": 0.1977,
+      "step": 74120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7075488694712188e-05,
+      "loss": 0.2027,
+      "step": 74130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7070984595982345e-05,
+      "loss": 0.198,
+      "step": 74140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7066480497252503e-05,
+      "loss": 0.2091,
+      "step": 74150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7061976398522657e-05,
+      "loss": 0.2009,
+      "step": 74160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.705747229979281e-05,
+      "loss": 0.1973,
+      "step": 74170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.705296820106297e-05,
+      "loss": 0.1996,
+      "step": 74180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7048464102333123e-05,
+      "loss": 0.2012,
+      "step": 74190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.704396000360328e-05,
+      "loss": 0.2066,
+      "step": 74200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7039455904873434e-05,
+      "loss": 0.2021,
+      "step": 74210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7034951806143592e-05,
+      "loss": 0.1993,
+      "step": 74220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7030447707413746e-05,
+      "loss": 0.2042,
+      "step": 74230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7025943608683903e-05,
+      "loss": 0.1979,
+      "step": 74240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.702143950995406e-05,
+      "loss": 0.1983,
+      "step": 74250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7016935411224215e-05,
+      "loss": 0.2085,
+      "step": 74260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.701243131249437e-05,
+      "loss": 0.1986,
+      "step": 74270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7007927213764527e-05,
+      "loss": 0.2027,
+      "step": 74280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7003423115034684e-05,
+      "loss": 0.2009,
+      "step": 74290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6998919016304838e-05,
+      "loss": 0.2031,
+      "step": 74300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6994414917574992e-05,
+      "loss": 0.2011,
+      "step": 74310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.698991081884515e-05,
+      "loss": 0.195,
+      "step": 74320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6985406720115304e-05,
+      "loss": 0.1977,
+      "step": 74330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.698090262138546e-05,
+      "loss": 0.2068,
+      "step": 74340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.697639852265562e-05,
+      "loss": 0.1984,
+      "step": 74350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6971894423925773e-05,
+      "loss": 0.2029,
+      "step": 74360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6967390325195927e-05,
+      "loss": 0.2004,
+      "step": 74370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6962886226466085e-05,
+      "loss": 0.1975,
+      "step": 74380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6958382127736242e-05,
+      "loss": 0.1985,
+      "step": 74390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6953878029006396e-05,
+      "loss": 0.1948,
+      "step": 74400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.694937393027655e-05,
+      "loss": 0.201,
+      "step": 74410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6944869831546708e-05,
+      "loss": 0.2034,
+      "step": 74420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6940365732816865e-05,
+      "loss": 0.2022,
+      "step": 74430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.693586163408702e-05,
+      "loss": 0.2001,
+      "step": 74440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6931357535357177e-05,
+      "loss": 0.1964,
+      "step": 74450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.692685343662733e-05,
+      "loss": 0.2002,
+      "step": 74460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6922349337897485e-05,
+      "loss": 0.1971,
+      "step": 74470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6917845239167643e-05,
+      "loss": 0.2027,
+      "step": 74480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.69133411404378e-05,
+      "loss": 0.1949,
+      "step": 74490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6908837041707958e-05,
+      "loss": 0.1979,
+      "step": 74500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.690433294297811e-05,
+      "loss": 0.1945,
+      "step": 74510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6899828844248266e-05,
+      "loss": 0.1991,
+      "step": 74520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6895324745518423e-05,
+      "loss": 0.2016,
+      "step": 74530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6890820646788577e-05,
+      "loss": 0.2038,
+      "step": 74540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6886316548058735e-05,
+      "loss": 0.2038,
+      "step": 74550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.688181244932889e-05,
+      "loss": 0.2088,
+      "step": 74560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6877308350599046e-05,
+      "loss": 0.2001,
+      "step": 74570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.68728042518692e-05,
+      "loss": 0.2027,
+      "step": 74580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6868300153139358e-05,
+      "loss": 0.1981,
+      "step": 74590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6863796054409516e-05,
+      "loss": 0.1987,
+      "step": 74600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.685929195567967e-05,
+      "loss": 0.1975,
+      "step": 74610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6854787856949824e-05,
+      "loss": 0.1984,
+      "step": 74620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.685028375821998e-05,
+      "loss": 0.199,
+      "step": 74630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.684577965949014e-05,
+      "loss": 0.1989,
+      "step": 74640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6841275560760293e-05,
+      "loss": 0.2051,
+      "step": 74650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6836771462030447e-05,
+      "loss": 0.2019,
+      "step": 74660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6832267363300604e-05,
+      "loss": 0.2,
+      "step": 74670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.682776326457076e-05,
+      "loss": 0.1994,
+      "step": 74680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6823259165840916e-05,
+      "loss": 0.1974,
+      "step": 74690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6818755067111074e-05,
+      "loss": 0.1988,
+      "step": 74700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6814250968381228e-05,
+      "loss": 0.2025,
+      "step": 74710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6809746869651382e-05,
+      "loss": 0.1949,
+      "step": 74720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.680524277092154e-05,
+      "loss": 0.2084,
+      "step": 74730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6800738672191697e-05,
+      "loss": 0.2026,
+      "step": 74740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.679623457346185e-05,
+      "loss": 0.2018,
+      "step": 74750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6791730474732005e-05,
+      "loss": 0.2107,
+      "step": 74760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6787226376002162e-05,
+      "loss": 0.2006,
+      "step": 74770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.678272227727232e-05,
+      "loss": 0.2031,
+      "step": 74780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6778218178542474e-05,
+      "loss": 0.1997,
+      "step": 74790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.677371407981263e-05,
+      "loss": 0.2006,
+      "step": 74800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6769209981082786e-05,
+      "loss": 0.191,
+      "step": 74810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.676470588235294e-05,
+      "loss": 0.2007,
+      "step": 74820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6760201783623097e-05,
+      "loss": 0.2006,
+      "step": 74830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6755697684893255e-05,
+      "loss": 0.2018,
+      "step": 74840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6751193586163412e-05,
+      "loss": 0.2012,
+      "step": 74850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6746689487433563e-05,
+      "loss": 0.1942,
+      "step": 74860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.674218538870372e-05,
+      "loss": 0.207,
+      "step": 74870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6737681289973878e-05,
+      "loss": 0.2015,
+      "step": 74880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6733177191244032e-05,
+      "loss": 0.2,
+      "step": 74890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.672867309251419e-05,
+      "loss": 0.1987,
+      "step": 74900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6724168993784344e-05,
+      "loss": 0.1992,
+      "step": 74910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.67196648950545e-05,
+      "loss": 0.1982,
+      "step": 74920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6715160796324655e-05,
+      "loss": 0.2005,
+      "step": 74930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6710656697594813e-05,
+      "loss": 0.2007,
+      "step": 74940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.670615259886497e-05,
+      "loss": 0.2114,
+      "step": 74950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.670164850013512e-05,
+      "loss": 0.1989,
+      "step": 74960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.669714440140528e-05,
+      "loss": 0.2026,
+      "step": 74970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6692640302675436e-05,
+      "loss": 0.2017,
+      "step": 74980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6688136203945593e-05,
+      "loss": 0.1935,
+      "step": 74990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6683632105215747e-05,
+      "loss": 0.2023,
+      "step": 75000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.66791280064859e-05,
+      "loss": 0.1925,
+      "step": 75010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.667462390775606e-05,
+      "loss": 0.1965,
+      "step": 75020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6670119809026213e-05,
+      "loss": 0.201,
+      "step": 75030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.666561571029637e-05,
+      "loss": 0.2037,
+      "step": 75040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6661111611566528e-05,
+      "loss": 0.1995,
+      "step": 75050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6656607512836682e-05,
+      "loss": 0.1936,
+      "step": 75060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6652103414106836e-05,
+      "loss": 0.2037,
+      "step": 75070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6647599315376994e-05,
+      "loss": 0.1988,
+      "step": 75080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.664309521664715e-05,
+      "loss": 0.2044,
+      "step": 75090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6638591117917305e-05,
+      "loss": 0.2012,
+      "step": 75100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.663408701918746e-05,
+      "loss": 0.1941,
+      "step": 75110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6629582920457617e-05,
+      "loss": 0.1968,
+      "step": 75120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6625078821727775e-05,
+      "loss": 0.2006,
+      "step": 75130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.662057472299793e-05,
+      "loss": 0.2007,
+      "step": 75140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6616070624268086e-05,
+      "loss": 0.1999,
+      "step": 75150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.661156652553824e-05,
+      "loss": 0.203,
+      "step": 75160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6607062426808394e-05,
+      "loss": 0.2052,
+      "step": 75170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6602558328078552e-05,
+      "loss": 0.2039,
+      "step": 75180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.659805422934871e-05,
+      "loss": 0.2007,
+      "step": 75190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6593550130618863e-05,
+      "loss": 0.2043,
+      "step": 75200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6589046031889018e-05,
+      "loss": 0.1989,
+      "step": 75210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6584541933159175e-05,
+      "loss": 0.1997,
+      "step": 75220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6580037834429333e-05,
+      "loss": 0.2057,
+      "step": 75230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6575533735699487e-05,
+      "loss": 0.205,
+      "step": 75240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6571029636969644e-05,
+      "loss": 0.1959,
+      "step": 75250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6566525538239798e-05,
+      "loss": 0.1973,
+      "step": 75260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6562021439509956e-05,
+      "loss": 0.2026,
+      "step": 75270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.655751734078011e-05,
+      "loss": 0.1999,
+      "step": 75280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6553013242050267e-05,
+      "loss": 0.2012,
+      "step": 75290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6548509143320425e-05,
+      "loss": 0.1988,
+      "step": 75300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6544005044590576e-05,
+      "loss": 0.2041,
+      "step": 75310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6539500945860733e-05,
+      "loss": 0.2004,
+      "step": 75320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.653499684713089e-05,
+      "loss": 0.2008,
+      "step": 75330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6530492748401048e-05,
+      "loss": 0.1981,
+      "step": 75340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6525988649671202e-05,
+      "loss": 0.2011,
+      "step": 75350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6521484550941356e-05,
+      "loss": 0.2031,
+      "step": 75360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6516980452211514e-05,
+      "loss": 0.2014,
+      "step": 75370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6512476353481668e-05,
+      "loss": 0.2064,
+      "step": 75380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6507972254751825e-05,
+      "loss": 0.2026,
+      "step": 75390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6503468156021983e-05,
+      "loss": 0.2038,
+      "step": 75400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6498964057292137e-05,
+      "loss": 0.1968,
+      "step": 75410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.649445995856229e-05,
+      "loss": 0.2021,
+      "step": 75420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.648995585983245e-05,
+      "loss": 0.1994,
+      "step": 75430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6485451761102606e-05,
+      "loss": 0.1993,
+      "step": 75440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.648094766237276e-05,
+      "loss": 0.1987,
+      "step": 75450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6476443563642914e-05,
+      "loss": 0.1981,
+      "step": 75460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.647193946491307e-05,
+      "loss": 0.2026,
+      "step": 75470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.646743536618323e-05,
+      "loss": 0.1949,
+      "step": 75480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6462931267453383e-05,
+      "loss": 0.1991,
+      "step": 75490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.645842716872354e-05,
+      "loss": 0.1997,
+      "step": 75500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6453923069993695e-05,
+      "loss": 0.2067,
+      "step": 75510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.644941897126385e-05,
+      "loss": 0.203,
+      "step": 75520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6444914872534006e-05,
+      "loss": 0.2006,
+      "step": 75530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6440410773804164e-05,
+      "loss": 0.1985,
+      "step": 75540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6435906675074318e-05,
+      "loss": 0.1952,
+      "step": 75550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6431402576344472e-05,
+      "loss": 0.1979,
+      "step": 75560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.642689847761463e-05,
+      "loss": 0.2012,
+      "step": 75570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6422394378884787e-05,
+      "loss": 0.203,
+      "step": 75580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.641789028015494e-05,
+      "loss": 0.2033,
+      "step": 75590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.64133861814251e-05,
+      "loss": 0.1969,
+      "step": 75600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6408882082695253e-05,
+      "loss": 0.2074,
+      "step": 75610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.640437798396541e-05,
+      "loss": 0.2015,
+      "step": 75620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6399873885235564e-05,
+      "loss": 0.1977,
+      "step": 75630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6395369786505722e-05,
+      "loss": 0.2022,
+      "step": 75640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6390865687775876e-05,
+      "loss": 0.1993,
+      "step": 75650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.638636158904603e-05,
+      "loss": 0.2031,
+      "step": 75660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6381857490316188e-05,
+      "loss": 0.1984,
+      "step": 75670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6377353391586345e-05,
+      "loss": 0.2042,
+      "step": 75680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6372849292856503e-05,
+      "loss": 0.1999,
+      "step": 75690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6368345194126657e-05,
+      "loss": 0.1994,
+      "step": 75700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.636384109539681e-05,
+      "loss": 0.2008,
+      "step": 75710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.635933699666697e-05,
+      "loss": 0.1978,
+      "step": 75720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6354832897937122e-05,
+      "loss": 0.1993,
+      "step": 75730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.635032879920728e-05,
+      "loss": 0.1959,
+      "step": 75740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6345824700477434e-05,
+      "loss": 0.2007,
+      "step": 75750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.634132060174759e-05,
+      "loss": 0.1968,
+      "step": 75760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6336816503017746e-05,
+      "loss": 0.1979,
+      "step": 75770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6332312404287903e-05,
+      "loss": 0.1987,
+      "step": 75780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.632780830555806e-05,
+      "loss": 0.1988,
+      "step": 75790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6323304206828215e-05,
+      "loss": 0.1988,
+      "step": 75800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.631880010809837e-05,
+      "loss": 0.1958,
+      "step": 75810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6314296009368526e-05,
+      "loss": 0.2012,
+      "step": 75820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6309791910638684e-05,
+      "loss": 0.205,
+      "step": 75830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6305287811908838e-05,
+      "loss": 0.1963,
+      "step": 75840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6300783713178992e-05,
+      "loss": 0.1967,
+      "step": 75850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.629627961444915e-05,
+      "loss": 0.2043,
+      "step": 75860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6291775515719304e-05,
+      "loss": 0.1933,
+      "step": 75870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.628727141698946e-05,
+      "loss": 0.2008,
+      "step": 75880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.628276731825962e-05,
+      "loss": 0.2003,
+      "step": 75890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6278263219529773e-05,
+      "loss": 0.1958,
+      "step": 75900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6273759120799927e-05,
+      "loss": 0.2091,
+      "step": 75910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6269255022070084e-05,
+      "loss": 0.195,
+      "step": 75920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6264750923340242e-05,
+      "loss": 0.2077,
+      "step": 75930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6260246824610396e-05,
+      "loss": 0.2017,
+      "step": 75940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6255742725880553e-05,
+      "loss": 0.2018,
+      "step": 75950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6251238627150707e-05,
+      "loss": 0.2026,
+      "step": 75960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6247184938293848e-05,
+      "loss": 0.2025,
+      "step": 75970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6242680839564002e-05,
+      "loss": 0.2017,
+      "step": 75980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.623817674083416e-05,
+      "loss": 0.204,
+      "step": 75990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6233672642104317e-05,
+      "loss": 0.1983,
+      "step": 76000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.622916854337447e-05,
+      "loss": 0.1983,
+      "step": 76010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.622466444464463e-05,
+      "loss": 0.1944,
+      "step": 76020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6220160345914783e-05,
+      "loss": 0.1964,
+      "step": 76030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6215656247184937e-05,
+      "loss": 0.2014,
+      "step": 76040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6211152148455094e-05,
+      "loss": 0.2019,
+      "step": 76050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6206648049725252e-05,
+      "loss": 0.1961,
+      "step": 76060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.620214395099541e-05,
+      "loss": 0.1927,
+      "step": 76070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.619763985226556e-05,
+      "loss": 0.1997,
+      "step": 76080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6193135753535717e-05,
+      "loss": 0.2026,
+      "step": 76090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6188631654805875e-05,
+      "loss": 0.2077,
+      "step": 76100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.618412755607603e-05,
+      "loss": 0.2017,
+      "step": 76110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6179623457346187e-05,
+      "loss": 0.2027,
+      "step": 76120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.617511935861634e-05,
+      "loss": 0.1978,
+      "step": 76130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6170615259886498e-05,
+      "loss": 0.2026,
+      "step": 76140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6166111161156652e-05,
+      "loss": 0.1977,
+      "step": 76150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.616160706242681e-05,
+      "loss": 0.2038,
+      "step": 76160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6157102963696967e-05,
+      "loss": 0.2014,
+      "step": 76170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6152598864967118e-05,
+      "loss": 0.2044,
+      "step": 76180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6148094766237275e-05,
+      "loss": 0.1986,
+      "step": 76190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6143590667507433e-05,
+      "loss": 0.2039,
+      "step": 76200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.613908656877759e-05,
+      "loss": 0.2019,
+      "step": 76210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6134582470047745e-05,
+      "loss": 0.195,
+      "step": 76220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.61300783713179e-05,
+      "loss": 0.2038,
+      "step": 76230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6125574272588056e-05,
+      "loss": 0.2011,
+      "step": 76240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.612107017385821e-05,
+      "loss": 0.1964,
+      "step": 76250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6116566075128368e-05,
+      "loss": 0.1993,
+      "step": 76260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6112061976398525e-05,
+      "loss": 0.1952,
+      "step": 76270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.610755787766868e-05,
+      "loss": 0.1939,
+      "step": 76280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6103053778938833e-05,
+      "loss": 0.2016,
+      "step": 76290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.609854968020899e-05,
+      "loss": 0.2015,
+      "step": 76300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.609404558147915e-05,
+      "loss": 0.2082,
+      "step": 76310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6089541482749302e-05,
+      "loss": 0.1997,
+      "step": 76320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6085037384019457e-05,
+      "loss": 0.2013,
+      "step": 76330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6080533285289614e-05,
+      "loss": 0.2016,
+      "step": 76340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.607602918655977e-05,
+      "loss": 0.197,
+      "step": 76350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6071525087829926e-05,
+      "loss": 0.2021,
+      "step": 76360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6067020989100083e-05,
+      "loss": 0.2006,
+      "step": 76370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6062516890370237e-05,
+      "loss": 0.2003,
+      "step": 76380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.605801279164039e-05,
+      "loss": 0.1924,
+      "step": 76390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.605350869291055e-05,
+      "loss": 0.1976,
+      "step": 76400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6049004594180706e-05,
+      "loss": 0.1982,
+      "step": 76410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.604450049545086e-05,
+      "loss": 0.2025,
+      "step": 76420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6039996396721015e-05,
+      "loss": 0.1956,
+      "step": 76430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6035492297991172e-05,
+      "loss": 0.1958,
+      "step": 76440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.603098819926133e-05,
+      "loss": 0.1979,
+      "step": 76450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6026484100531484e-05,
+      "loss": 0.1984,
+      "step": 76460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.602198000180164e-05,
+      "loss": 0.1979,
+      "step": 76470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6017475903071795e-05,
+      "loss": 0.2057,
+      "step": 76480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6012971804341953e-05,
+      "loss": 0.1972,
+      "step": 76490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6008467705612107e-05,
+      "loss": 0.2013,
+      "step": 76500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6003963606882264e-05,
+      "loss": 0.1985,
+      "step": 76510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.599945950815242e-05,
+      "loss": 0.1988,
+      "step": 76520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5994955409422573e-05,
+      "loss": 0.1987,
+      "step": 76530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.599045131069273e-05,
+      "loss": 0.2057,
+      "step": 76540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5985947211962888e-05,
+      "loss": 0.1912,
+      "step": 76550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5981443113233045e-05,
+      "loss": 0.1931,
+      "step": 76560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.59769390145032e-05,
+      "loss": 0.1919,
+      "step": 76570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5972434915773353e-05,
+      "loss": 0.1987,
+      "step": 76580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.596793081704351e-05,
+      "loss": 0.1992,
+      "step": 76590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5963426718313665e-05,
+      "loss": 0.1977,
+      "step": 76600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5958922619583822e-05,
+      "loss": 0.2,
+      "step": 76610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.595441852085398e-05,
+      "loss": 0.2021,
+      "step": 76620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5949914422124134e-05,
+      "loss": 0.2066,
+      "step": 76630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5945410323394288e-05,
+      "loss": 0.2017,
+      "step": 76640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5940906224664446e-05,
+      "loss": 0.2011,
+      "step": 76650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5936402125934603e-05,
+      "loss": 0.1974,
+      "step": 76660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5931898027204757e-05,
+      "loss": 0.2039,
+      "step": 76670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.592739392847491e-05,
+      "loss": 0.1978,
+      "step": 76680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.592288982974507e-05,
+      "loss": 0.2045,
+      "step": 76690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5918385731015226e-05,
+      "loss": 0.2017,
+      "step": 76700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.591388163228538e-05,
+      "loss": 0.1978,
+      "step": 76710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5909377533555538e-05,
+      "loss": 0.1955,
+      "step": 76720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5904873434825692e-05,
+      "loss": 0.1979,
+      "step": 76730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5900369336095846e-05,
+      "loss": 0.1938,
+      "step": 76740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5895865237366003e-05,
+      "loss": 0.1919,
+      "step": 76750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.589136113863616e-05,
+      "loss": 0.1979,
+      "step": 76760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5886857039906315e-05,
+      "loss": 0.2,
+      "step": 76770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.588235294117647e-05,
+      "loss": 0.2026,
+      "step": 76780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5877848842446627e-05,
+      "loss": 0.2047,
+      "step": 76790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5873344743716784e-05,
+      "loss": 0.1985,
+      "step": 76800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5868840644986938e-05,
+      "loss": 0.2011,
+      "step": 76810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5864336546257096e-05,
+      "loss": 0.2064,
+      "step": 76820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.585983244752725e-05,
+      "loss": 0.1948,
+      "step": 76830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5855328348797407e-05,
+      "loss": 0.1997,
+      "step": 76840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.585082425006756e-05,
+      "loss": 0.1926,
+      "step": 76850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.584632015133772e-05,
+      "loss": 0.1983,
+      "step": 76860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5841816052607873e-05,
+      "loss": 0.202,
+      "step": 76870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5837311953878027e-05,
+      "loss": 0.2017,
+      "step": 76880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5832807855148185e-05,
+      "loss": 0.1983,
+      "step": 76890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5828303756418342e-05,
+      "loss": 0.2003,
+      "step": 76900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.58237996576885e-05,
+      "loss": 0.1973,
+      "step": 76910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5819295558958654e-05,
+      "loss": 0.2013,
+      "step": 76920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5814791460228808e-05,
+      "loss": 0.2059,
+      "step": 76930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5810287361498965e-05,
+      "loss": 0.198,
+      "step": 76940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.580578326276912e-05,
+      "loss": 0.1959,
+      "step": 76950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5801279164039277e-05,
+      "loss": 0.1974,
+      "step": 76960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.579677506530943e-05,
+      "loss": 0.2001,
+      "step": 76970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.579227096657959e-05,
+      "loss": 0.1983,
+      "step": 76980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5787766867849743e-05,
+      "loss": 0.2009,
+      "step": 76990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.57832627691199e-05,
+      "loss": 0.2021,
+      "step": 77000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5778758670390058e-05,
+      "loss": 0.1965,
+      "step": 77010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5774254571660212e-05,
+      "loss": 0.1999,
+      "step": 77020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5769750472930366e-05,
+      "loss": 0.198,
+      "step": 77030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5765246374200523e-05,
+      "loss": 0.2015,
+      "step": 77040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.576074227547068e-05,
+      "loss": 0.2017,
+      "step": 77050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5756238176740835e-05,
+      "loss": 0.1964,
+      "step": 77060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.575173407801099e-05,
+      "loss": 0.1957,
+      "step": 77070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5747229979281147e-05,
+      "loss": 0.1985,
+      "step": 77080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.57427258805513e-05,
+      "loss": 0.1976,
+      "step": 77090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5738221781821458e-05,
+      "loss": 0.2017,
+      "step": 77100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5733717683091616e-05,
+      "loss": 0.2011,
+      "step": 77110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.572921358436177e-05,
+      "loss": 0.2026,
+      "step": 77120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5724709485631924e-05,
+      "loss": 0.2014,
+      "step": 77130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.572020538690208e-05,
+      "loss": 0.1986,
+      "step": 77140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.571570128817224e-05,
+      "loss": 0.1979,
+      "step": 77150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5711197189442393e-05,
+      "loss": 0.1971,
+      "step": 77160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5706693090712547e-05,
+      "loss": 0.1931,
+      "step": 77170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5702188991982705e-05,
+      "loss": 0.1997,
+      "step": 77180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5697684893252862e-05,
+      "loss": 0.1996,
+      "step": 77190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5693180794523016e-05,
+      "loss": 0.1986,
+      "step": 77200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5688676695793174e-05,
+      "loss": 0.201,
+      "step": 77210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5684172597063328e-05,
+      "loss": 0.2022,
+      "step": 77220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5679668498333482e-05,
+      "loss": 0.1973,
+      "step": 77230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.567516439960364e-05,
+      "loss": 0.1987,
+      "step": 77240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5670660300873797e-05,
+      "loss": 0.1967,
+      "step": 77250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5666156202143954e-05,
+      "loss": 0.2023,
+      "step": 77260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.566165210341411e-05,
+      "loss": 0.1935,
+      "step": 77270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5657148004684262e-05,
+      "loss": 0.1937,
+      "step": 77280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.565264390595442e-05,
+      "loss": 0.201,
+      "step": 77290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5648139807224574e-05,
+      "loss": 0.2066,
+      "step": 77300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.564363570849473e-05,
+      "loss": 0.1945,
+      "step": 77310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5639131609764886e-05,
+      "loss": 0.2001,
+      "step": 77320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5634627511035043e-05,
+      "loss": 0.1999,
+      "step": 77330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5630123412305197e-05,
+      "loss": 0.1957,
+      "step": 77340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5625619313575355e-05,
+      "loss": 0.206,
+      "step": 77350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5621115214845512e-05,
+      "loss": 0.2038,
+      "step": 77360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5616611116115666e-05,
+      "loss": 0.1942,
+      "step": 77370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.561210701738582e-05,
+      "loss": 0.1956,
+      "step": 77380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5607602918655978e-05,
+      "loss": 0.1972,
+      "step": 77390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5603098819926135e-05,
+      "loss": 0.2008,
+      "step": 77400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.559859472119629e-05,
+      "loss": 0.1933,
+      "step": 77410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5594090622466444e-05,
+      "loss": 0.2019,
+      "step": 77420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.55895865237366e-05,
+      "loss": 0.2035,
+      "step": 77430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5585082425006755e-05,
+      "loss": 0.1952,
+      "step": 77440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5580578326276913e-05,
+      "loss": 0.2016,
+      "step": 77450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.557607422754707e-05,
+      "loss": 0.2016,
+      "step": 77460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5571570128817224e-05,
+      "loss": 0.1986,
+      "step": 77470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.556706603008738e-05,
+      "loss": 0.1992,
+      "step": 77480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5562561931357536e-05,
+      "loss": 0.196,
+      "step": 77490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5558057832627693e-05,
+      "loss": 0.1959,
+      "step": 77500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5553553733897848e-05,
+      "loss": 0.1958,
+      "step": 77510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5549049635168e-05,
+      "loss": 0.2019,
+      "step": 77520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.554454553643816e-05,
+      "loss": 0.1978,
+      "step": 77530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5540041437708317e-05,
+      "loss": 0.2013,
+      "step": 77540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.553553733897847e-05,
+      "loss": 0.1999,
+      "step": 77550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5531033240248628e-05,
+      "loss": 0.1995,
+      "step": 77560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5526529141518782e-05,
+      "loss": 0.1996,
+      "step": 77570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5522025042788936e-05,
+      "loss": 0.1966,
+      "step": 77580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5517520944059094e-05,
+      "loss": 0.1953,
+      "step": 77590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.551301684532925e-05,
+      "loss": 0.1944,
+      "step": 77600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.550851274659941e-05,
+      "loss": 0.1956,
+      "step": 77610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.550400864786956e-05,
+      "loss": 0.1976,
+      "step": 77620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5499504549139717e-05,
+      "loss": 0.1994,
+      "step": 77630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5495000450409875e-05,
+      "loss": 0.198,
+      "step": 77640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.549049635168003e-05,
+      "loss": 0.1941,
+      "step": 77650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5485992252950186e-05,
+      "loss": 0.1987,
+      "step": 77660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.548148815422034e-05,
+      "loss": 0.1965,
+      "step": 77670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5476984055490498e-05,
+      "loss": 0.1957,
+      "step": 77680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5472479956760652e-05,
+      "loss": 0.1963,
+      "step": 77690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.546797585803081e-05,
+      "loss": 0.2033,
+      "step": 77700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5463471759300967e-05,
+      "loss": 0.2018,
+      "step": 77710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5458967660571118e-05,
+      "loss": 0.2016,
+      "step": 77720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5454463561841275e-05,
+      "loss": 0.1994,
+      "step": 77730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5449959463111433e-05,
+      "loss": 0.1911,
+      "step": 77740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.544545536438159e-05,
+      "loss": 0.1987,
+      "step": 77750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5440951265651744e-05,
+      "loss": 0.1979,
+      "step": 77760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5436447166921898e-05,
+      "loss": 0.2005,
+      "step": 77770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5431943068192056e-05,
+      "loss": 0.1986,
+      "step": 77780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542743896946221e-05,
+      "loss": 0.1935,
+      "step": 77790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5422934870732367e-05,
+      "loss": 0.1955,
+      "step": 77800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5418430772002525e-05,
+      "loss": 0.199,
+      "step": 77810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541392667327268e-05,
+      "loss": 0.1999,
+      "step": 77820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5409422574542833e-05,
+      "loss": 0.1914,
+      "step": 77830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540491847581299e-05,
+      "loss": 0.1949,
+      "step": 77840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5400414377083148e-05,
+      "loss": 0.2007,
+      "step": 77850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5395910278353302e-05,
+      "loss": 0.198,
+      "step": 77860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5391406179623456e-05,
+      "loss": 0.2054,
+      "step": 77870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5386902080893614e-05,
+      "loss": 0.2002,
+      "step": 77880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.538239798216377e-05,
+      "loss": 0.1973,
+      "step": 77890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5377893883433925e-05,
+      "loss": 0.2005,
+      "step": 77900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5373389784704083e-05,
+      "loss": 0.1932,
+      "step": 77910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5368885685974237e-05,
+      "loss": 0.2049,
+      "step": 77920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536438158724439e-05,
+      "loss": 0.1972,
+      "step": 77930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535987748851455e-05,
+      "loss": 0.198,
+      "step": 77940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5355373389784706e-05,
+      "loss": 0.1993,
+      "step": 77950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535086929105486e-05,
+      "loss": 0.1995,
+      "step": 77960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5346365192325014e-05,
+      "loss": 0.1963,
+      "step": 77970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5341861093595172e-05,
+      "loss": 0.2013,
+      "step": 77980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533735699486533e-05,
+      "loss": 0.2029,
+      "step": 77990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5332852896135483e-05,
+      "loss": 0.201,
+      "step": 78000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532834879740564e-05,
+      "loss": 0.1978,
+      "step": 78010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5323844698675795e-05,
+      "loss": 0.1963,
+      "step": 78020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5319340599945952e-05,
+      "loss": 0.1993,
+      "step": 78030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5314836501216107e-05,
+      "loss": 0.1952,
+      "step": 78040
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5310332402486264e-05,
+      "loss": 0.1969,
+      "step": 78050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530582830375642e-05,
+      "loss": 0.1867,
+      "step": 78060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5301324205026572e-05,
+      "loss": 0.201,
+      "step": 78070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.529682010629673e-05,
+      "loss": 0.1909,
+      "step": 78080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5292316007566887e-05,
+      "loss": 0.1954,
+      "step": 78090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5287811908837045e-05,
+      "loss": 0.1958,
+      "step": 78100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.52833078101072e-05,
+      "loss": 0.1907,
+      "step": 78110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5278803711377353e-05,
+      "loss": 0.1948,
+      "step": 78120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.527429961264751e-05,
+      "loss": 0.2002,
+      "step": 78130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5269795513917665e-05,
+      "loss": 0.1932,
+      "step": 78140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5265291415187822e-05,
+      "loss": 0.1985,
+      "step": 78150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.526078731645798e-05,
+      "loss": 0.1998,
+      "step": 78160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5256283217728135e-05,
+      "loss": 0.2015,
+      "step": 78170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5251779118998288e-05,
+      "loss": 0.1963,
+      "step": 78180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5247275020268445e-05,
+      "loss": 0.1953,
+      "step": 78190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5242770921538601e-05,
+      "loss": 0.1994,
+      "step": 78200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5238266822808755e-05,
+      "loss": 0.1969,
+      "step": 78210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5233762724078913e-05,
+      "loss": 0.1972,
+      "step": 78220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5229258625349068e-05,
+      "loss": 0.1994,
+      "step": 78230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5224754526619226e-05,
+      "loss": 0.201,
+      "step": 78240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.522025042788938e-05,
+      "loss": 0.1975,
+      "step": 78250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5215746329159536e-05,
+      "loss": 0.193,
+      "step": 78260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5211242230429693e-05,
+      "loss": 0.1951,
+      "step": 78270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5206738131699846e-05,
+      "loss": 0.1933,
+      "step": 78280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5202234032970003e-05,
+      "loss": 0.1935,
+      "step": 78290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5197729934240159e-05,
+      "loss": 0.202,
+      "step": 78300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5193225835510316e-05,
+      "loss": 0.1986,
+      "step": 78310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.518872173678047e-05,
+      "loss": 0.196,
+      "step": 78320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5184217638050626e-05,
+      "loss": 0.1935,
+      "step": 78330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5179713539320784e-05,
+      "loss": 0.1921,
+      "step": 78340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5175209440590938e-05,
+      "loss": 0.1923,
+      "step": 78350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5170705341861094e-05,
+      "loss": 0.1928,
+      "step": 78360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5166201243131251e-05,
+      "loss": 0.1959,
+      "step": 78370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5161697144401407e-05,
+      "loss": 0.1973,
+      "step": 78380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5157193045671561e-05,
+      "loss": 0.1992,
+      "step": 78390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5152688946941717e-05,
+      "loss": 0.1949,
+      "step": 78400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5148184848211874e-05,
+      "loss": 0.1976,
+      "step": 78410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5143680749482029e-05,
+      "loss": 0.2016,
+      "step": 78420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5139176650752184e-05,
+      "loss": 0.1945,
+      "step": 78430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5134672552022342e-05,
+      "loss": 0.1974,
+      "step": 78440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5130168453292498e-05,
+      "loss": 0.1984,
+      "step": 78450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5125664354562652e-05,
+      "loss": 0.2012,
+      "step": 78460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.512116025583281e-05,
+      "loss": 0.1993,
+      "step": 78470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5116656157102965e-05,
+      "loss": 0.1993,
+      "step": 78480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5112152058373119e-05,
+      "loss": 0.1909,
+      "step": 78490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5107647959643275e-05,
+      "loss": 0.1977,
+      "step": 78500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5103143860913432e-05,
+      "loss": 0.1963,
+      "step": 78510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5098639762183588e-05,
+      "loss": 0.194,
+      "step": 78520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5094135663453742e-05,
+      "loss": 0.1945,
+      "step": 78530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.50896315647239e-05,
+      "loss": 0.1946,
+      "step": 78540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5085127465994056e-05,
+      "loss": 0.2003,
+      "step": 78550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.508062336726421e-05,
+      "loss": 0.1956,
+      "step": 78560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5076119268534367e-05,
+      "loss": 0.1941,
+      "step": 78570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5071615169804523e-05,
+      "loss": 0.1913,
+      "step": 78580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.506711107107468e-05,
+      "loss": 0.1925,
+      "step": 78590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5062606972344833e-05,
+      "loss": 0.1991,
+      "step": 78600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.505810287361499e-05,
+      "loss": 0.1964,
+      "step": 78610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5053598774885146e-05,
+      "loss": 0.1992,
+      "step": 78620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.50490946761553e-05,
+      "loss": 0.1976,
+      "step": 78630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5044590577425458e-05,
+      "loss": 0.1948,
+      "step": 78640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5040086478695614e-05,
+      "loss": 0.1983,
+      "step": 78650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5035582379965771e-05,
+      "loss": 0.2033,
+      "step": 78660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5031078281235925e-05,
+      "loss": 0.1968,
+      "step": 78670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5026574182506081e-05,
+      "loss": 0.1964,
+      "step": 78680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5022070083776238e-05,
+      "loss": 0.2033,
+      "step": 78690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5017565985046393e-05,
+      "loss": 0.1954,
+      "step": 78700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5013061886316548e-05,
+      "loss": 0.1977,
+      "step": 78710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5008557787586706e-05,
+      "loss": 0.1998,
+      "step": 78720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5004053688856862e-05,
+      "loss": 0.1971,
+      "step": 78730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4999549590127016e-05,
+      "loss": 0.1944,
+      "step": 78740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4995045491397172e-05,
+      "loss": 0.1946,
+      "step": 78750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4990541392667329e-05,
+      "loss": 0.1993,
+      "step": 78760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4986037293937483e-05,
+      "loss": 0.205,
+      "step": 78770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4981533195207639e-05,
+      "loss": 0.1982,
+      "step": 78780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4977029096477796e-05,
+      "loss": 0.2039,
+      "step": 78790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4972524997747952e-05,
+      "loss": 0.1975,
+      "step": 78800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4968020899018106e-05,
+      "loss": 0.1914,
+      "step": 78810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4963516800288264e-05,
+      "loss": 0.1885,
+      "step": 78820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.495901270155842e-05,
+      "loss": 0.1951,
+      "step": 78830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4954508602828574e-05,
+      "loss": 0.1999,
+      "step": 78840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.495000450409873e-05,
+      "loss": 0.1997,
+      "step": 78850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4945500405368887e-05,
+      "loss": 0.1911,
+      "step": 78860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4940996306639043e-05,
+      "loss": 0.1893,
+      "step": 78870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4936492207909197e-05,
+      "loss": 0.1958,
+      "step": 78880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4931988109179354e-05,
+      "loss": 0.1967,
+      "step": 78890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.492748401044951e-05,
+      "loss": 0.1936,
+      "step": 78900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4922979911719664e-05,
+      "loss": 0.1938,
+      "step": 78910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4918475812989822e-05,
+      "loss": 0.2027,
+      "step": 78920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4913971714259978e-05,
+      "loss": 0.1995,
+      "step": 78930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4909467615530135e-05,
+      "loss": 0.1912,
+      "step": 78940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4904963516800288e-05,
+      "loss": 0.1981,
+      "step": 78950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4900459418070445e-05,
+      "loss": 0.1905,
+      "step": 78960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.48959553193406e-05,
+      "loss": 0.2011,
+      "step": 78970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4891451220610755e-05,
+      "loss": 0.1986,
+      "step": 78980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4886947121880912e-05,
+      "loss": 0.1984,
+      "step": 78990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4882443023151068e-05,
+      "loss": 0.1966,
+      "step": 79000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4877938924421226e-05,
+      "loss": 0.1984,
+      "step": 79010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.487343482569138e-05,
+      "loss": 0.1995,
+      "step": 79020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4868930726961536e-05,
+      "loss": 0.2027,
+      "step": 79030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4864426628231693e-05,
+      "loss": 0.2015,
+      "step": 79040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4859922529501846e-05,
+      "loss": 0.1963,
+      "step": 79050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4855418430772003e-05,
+      "loss": 0.1956,
+      "step": 79060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4850914332042159e-05,
+      "loss": 0.1967,
+      "step": 79070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4846410233312316e-05,
+      "loss": 0.1898,
+      "step": 79080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.484190613458247e-05,
+      "loss": 0.1953,
+      "step": 79090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4837402035852626e-05,
+      "loss": 0.1947,
+      "step": 79100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4832897937122784e-05,
+      "loss": 0.1946,
+      "step": 79110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4828393838392938e-05,
+      "loss": 0.1985,
+      "step": 79120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4823889739663094e-05,
+      "loss": 0.198,
+      "step": 79130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4819385640933251e-05,
+      "loss": 0.1994,
+      "step": 79140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4814881542203407e-05,
+      "loss": 0.1906,
+      "step": 79150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4810377443473561e-05,
+      "loss": 0.1897,
+      "step": 79160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4805873344743717e-05,
+      "loss": 0.1968,
+      "step": 79170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4801369246013874e-05,
+      "loss": 0.1947,
+      "step": 79180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4796865147284028e-05,
+      "loss": 0.1986,
+      "step": 79190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4792361048554184e-05,
+      "loss": 0.1973,
+      "step": 79200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4787856949824342e-05,
+      "loss": 0.1924,
+      "step": 79210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4783352851094497e-05,
+      "loss": 0.1968,
+      "step": 79220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4778848752364652e-05,
+      "loss": 0.1982,
+      "step": 79230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4774344653634809e-05,
+      "loss": 0.1968,
+      "step": 79240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4769840554904965e-05,
+      "loss": 0.1994,
+      "step": 79250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4765336456175119e-05,
+      "loss": 0.1939,
+      "step": 79260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4760832357445275e-05,
+      "loss": 0.1947,
+      "step": 79270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4756328258715432e-05,
+      "loss": 0.194,
+      "step": 79280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4751824159985588e-05,
+      "loss": 0.1961,
+      "step": 79290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4747320061255742e-05,
+      "loss": 0.1982,
+      "step": 79300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.47428159625259e-05,
+      "loss": 0.1983,
+      "step": 79310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4738311863796055e-05,
+      "loss": 0.1978,
+      "step": 79320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.473380776506621e-05,
+      "loss": 0.1979,
+      "step": 79330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4729303666336367e-05,
+      "loss": 0.1972,
+      "step": 79340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4724799567606523e-05,
+      "loss": 0.195,
+      "step": 79350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.472029546887668e-05,
+      "loss": 0.1944,
+      "step": 79360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4715791370146834e-05,
+      "loss": 0.1923,
+      "step": 79370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.471128727141699e-05,
+      "loss": 0.2019,
+      "step": 79380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4706783172687146e-05,
+      "loss": 0.1903,
+      "step": 79390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.47022790739573e-05,
+      "loss": 0.1936,
+      "step": 79400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4697774975227458e-05,
+      "loss": 0.1984,
+      "step": 79410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4693270876497613e-05,
+      "loss": 0.1967,
+      "step": 79420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4688766777767771e-05,
+      "loss": 0.1951,
+      "step": 79430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4684262679037925e-05,
+      "loss": 0.1956,
+      "step": 79440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.467975858030808e-05,
+      "loss": 0.2013,
+      "step": 79450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4675254481578238e-05,
+      "loss": 0.1961,
+      "step": 79460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4670750382848392e-05,
+      "loss": 0.2032,
+      "step": 79470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4666246284118548e-05,
+      "loss": 0.1986,
+      "step": 79480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4661742185388706e-05,
+      "loss": 0.1946,
+      "step": 79490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4657238086658862e-05,
+      "loss": 0.1957,
+      "step": 79500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4652733987929016e-05,
+      "loss": 0.1951,
+      "step": 79510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4648229889199171e-05,
+      "loss": 0.1933,
+      "step": 79520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4643725790469329e-05,
+      "loss": 0.1928,
+      "step": 79530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4639221691739483e-05,
+      "loss": 0.1944,
+      "step": 79540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4634717593009639e-05,
+      "loss": 0.206,
+      "step": 79550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4630213494279796e-05,
+      "loss": 0.2023,
+      "step": 79560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4625709395549952e-05,
+      "loss": 0.1988,
+      "step": 79570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4621205296820106e-05,
+      "loss": 0.2009,
+      "step": 79580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4616701198090264e-05,
+      "loss": 0.1951,
+      "step": 79590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.461219709936042e-05,
+      "loss": 0.201,
+      "step": 79600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4607693000630574e-05,
+      "loss": 0.1963,
+      "step": 79610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.460318890190073e-05,
+      "loss": 0.1965,
+      "step": 79620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4598684803170887e-05,
+      "loss": 0.1924,
+      "step": 79630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4594180704441043e-05,
+      "loss": 0.1914,
+      "step": 79640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4589676605711197e-05,
+      "loss": 0.1979,
+      "step": 79650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4585172506981354e-05,
+      "loss": 0.1983,
+      "step": 79660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.458066840825151e-05,
+      "loss": 0.1916,
+      "step": 79670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4576164309521664e-05,
+      "loss": 0.1986,
+      "step": 79680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4571660210791822e-05,
+      "loss": 0.191,
+      "step": 79690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4567156112061977e-05,
+      "loss": 0.1946,
+      "step": 79700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4562652013332135e-05,
+      "loss": 0.1937,
+      "step": 79710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4558147914602287e-05,
+      "loss": 0.1931,
+      "step": 79720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4553643815872445e-05,
+      "loss": 0.194,
+      "step": 79730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.45491397171426e-05,
+      "loss": 0.198,
+      "step": 79740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4544635618412755e-05,
+      "loss": 0.1984,
+      "step": 79750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4540131519682912e-05,
+      "loss": 0.1949,
+      "step": 79760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4535627420953068e-05,
+      "loss": 0.1962,
+      "step": 79770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4531123322223226e-05,
+      "loss": 0.1937,
+      "step": 79780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.452661922349338e-05,
+      "loss": 0.1979,
+      "step": 79790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4522115124763535e-05,
+      "loss": 0.1931,
+      "step": 79800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4517611026033693e-05,
+      "loss": 0.1996,
+      "step": 79810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4513106927303845e-05,
+      "loss": 0.1977,
+      "step": 79820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4508602828574003e-05,
+      "loss": 0.1999,
+      "step": 79830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4504098729844159e-05,
+      "loss": 0.1942,
+      "step": 79840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4499594631114316e-05,
+      "loss": 0.1958,
+      "step": 79850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.449509053238447e-05,
+      "loss": 0.1954,
+      "step": 79860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4490586433654626e-05,
+      "loss": 0.1991,
+      "step": 79870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4486082334924784e-05,
+      "loss": 0.1914,
+      "step": 79880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4481578236194938e-05,
+      "loss": 0.1999,
+      "step": 79890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4477074137465093e-05,
+      "loss": 0.1977,
+      "step": 79900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4472570038735251e-05,
+      "loss": 0.198,
+      "step": 79910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4468065940005407e-05,
+      "loss": 0.1973,
+      "step": 79920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4463561841275561e-05,
+      "loss": 0.1968,
+      "step": 79930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4459057742545717e-05,
+      "loss": 0.1987,
+      "step": 79940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4454553643815874e-05,
+      "loss": 0.1967,
+      "step": 79950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4450049545086028e-05,
+      "loss": 0.1979,
+      "step": 79960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4445545446356184e-05,
+      "loss": 0.193,
+      "step": 79970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4441491757499326e-05,
+      "loss": 0.198,
+      "step": 79980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.443698765876948e-05,
+      "loss": 0.1971,
+      "step": 79990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4432483560039636e-05,
+      "loss": 0.1834,
+      "step": 80000
+    },
+    {
+      "epoch": 0.71,
+      "eval_NEIMS_canon_smiles": 0.938,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.04953333333333333,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.44968452256661257,
+      "eval_NEIMS_exact_mols": 0.04933333333333333,
+      "eval_NEIMS_exact_smiles": 0.04761666666666667,
+      "eval_NEIMS_loss": 0.22744505107402802,
+      "eval_NEIMS_matched_formulas": 0.55905,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.3516388449474886,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.04951666666666667,
+      "eval_NEIMS_runtime": 716.2027,
+      "eval_NEIMS_samples_per_second": 83.775,
+      "eval_NEIMS_steps_per_second": 1.31,
+      "step": 80000
+    },
+    {
+      "epoch": 0.71,
+      "eval_RASSP_canon_smiles": 0.9306608341403145,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.12171502988613217,
+      "eval_RASSP_daylight_tanimoto_simil": 0.5728757311086445,
+      "eval_RASSP_exact_mols": 0.1210638795204862,
+      "eval_RASSP_exact_smiles": 0.11730724279560557,
+      "eval_RASSP_loss": 0.16591234505176544,
+      "eval_RASSP_matched_formulas": 0.7293551941763783,
+      "eval_RASSP_morgan_tanimoto_simil": 0.44149813310147595,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.12113066417337295,
+      "eval_RASSP_runtime": 819.4147,
+      "eval_RASSP_samples_per_second": 73.094,
+      "eval_RASSP_steps_per_second": 1.142,
+      "step": 80000
+    },
+    {
+      "epoch": 0.71,
+      "eval_NIST_canon_smiles": 0.8830251623664691,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.012740887958263832,
+      "eval_NIST_daylight_tanimoto_simil": 0.2658825257202243,
+      "eval_NIST_exact_mols": 0.012066579124818113,
+      "eval_NIST_exact_smiles": 0.01107286084395074,
+      "eval_NIST_loss": 1.4652163982391357,
+      "eval_NIST_matched_formulas": 0.08137842921531746,
+      "eval_NIST_morgan_tanimoto_simil": 0.22094277641876273,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.01266990808105902,
+      "eval_NIST_runtime": 497.5527,
+      "eval_NIST_samples_per_second": 56.631,
+      "eval_NIST_steps_per_second": 0.886,
+      "step": 80000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4427979461309794e-05,
+      "loss": 0.1916,
+      "step": 80010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.442347536257995e-05,
+      "loss": 0.1936,
+      "step": 80020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4418971263850103e-05,
+      "loss": 0.1983,
+      "step": 80030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4414467165120261e-05,
+      "loss": 0.1966,
+      "step": 80040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4409963066390417e-05,
+      "loss": 0.1965,
+      "step": 80050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.440545896766057e-05,
+      "loss": 0.197,
+      "step": 80060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4400954868930727e-05,
+      "loss": 0.1922,
+      "step": 80070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4396450770200884e-05,
+      "loss": 0.1991,
+      "step": 80080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.439194667147104e-05,
+      "loss": 0.195,
+      "step": 80090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4387442572741194e-05,
+      "loss": 0.1965,
+      "step": 80100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4382938474011351e-05,
+      "loss": 0.1991,
+      "step": 80110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4378434375281507e-05,
+      "loss": 0.1903,
+      "step": 80120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4373930276551661e-05,
+      "loss": 0.1976,
+      "step": 80130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4369426177821819e-05,
+      "loss": 0.1935,
+      "step": 80140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4364922079091975e-05,
+      "loss": 0.2009,
+      "step": 80150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4360417980362132e-05,
+      "loss": 0.1994,
+      "step": 80160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4355913881632285e-05,
+      "loss": 0.1943,
+      "step": 80170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4351409782902442e-05,
+      "loss": 0.1922,
+      "step": 80180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4346905684172598e-05,
+      "loss": 0.1975,
+      "step": 80190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4342401585442752e-05,
+      "loss": 0.1924,
+      "step": 80200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.433789748671291e-05,
+      "loss": 0.1952,
+      "step": 80210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4333393387983065e-05,
+      "loss": 0.2005,
+      "step": 80220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4328889289253223e-05,
+      "loss": 0.1913,
+      "step": 80230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4324385190523377e-05,
+      "loss": 0.1877,
+      "step": 80240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4319881091793533e-05,
+      "loss": 0.1959,
+      "step": 80250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.431537699306369e-05,
+      "loss": 0.2045,
+      "step": 80260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4310872894333843e-05,
+      "loss": 0.2024,
+      "step": 80270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4306368795604e-05,
+      "loss": 0.1928,
+      "step": 80280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4301864696874156e-05,
+      "loss": 0.1975,
+      "step": 80290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4297360598144313e-05,
+      "loss": 0.1943,
+      "step": 80300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4292856499414467e-05,
+      "loss": 0.1932,
+      "step": 80310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4288352400684623e-05,
+      "loss": 0.1922,
+      "step": 80320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.428384830195478e-05,
+      "loss": 0.1916,
+      "step": 80330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4279344203224935e-05,
+      "loss": 0.194,
+      "step": 80340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.427484010449509e-05,
+      "loss": 0.1947,
+      "step": 80350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4270336005765248e-05,
+      "loss": 0.1938,
+      "step": 80360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4265831907035404e-05,
+      "loss": 0.1992,
+      "step": 80370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4261327808305558e-05,
+      "loss": 0.1995,
+      "step": 80380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4256823709575714e-05,
+      "loss": 0.1941,
+      "step": 80390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4252319610845871e-05,
+      "loss": 0.1897,
+      "step": 80400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4247815512116025e-05,
+      "loss": 0.2022,
+      "step": 80410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4243311413386181e-05,
+      "loss": 0.1969,
+      "step": 80420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4238807314656339e-05,
+      "loss": 0.1914,
+      "step": 80430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4234303215926495e-05,
+      "loss": 0.192,
+      "step": 80440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4229799117196649e-05,
+      "loss": 0.1907,
+      "step": 80450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4225295018466806e-05,
+      "loss": 0.19,
+      "step": 80460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4220790919736962e-05,
+      "loss": 0.2,
+      "step": 80470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4216286821007116e-05,
+      "loss": 0.1979,
+      "step": 80480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4211782722277272e-05,
+      "loss": 0.1969,
+      "step": 80490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.420727862354743e-05,
+      "loss": 0.1936,
+      "step": 80500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4202774524817585e-05,
+      "loss": 0.1972,
+      "step": 80510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.419827042608774e-05,
+      "loss": 0.193,
+      "step": 80520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4193766327357897e-05,
+      "loss": 0.1923,
+      "step": 80530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4189262228628053e-05,
+      "loss": 0.1965,
+      "step": 80540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4184758129898207e-05,
+      "loss": 0.1976,
+      "step": 80550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4180254031168364e-05,
+      "loss": 0.1916,
+      "step": 80560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.417574993243852e-05,
+      "loss": 0.1928,
+      "step": 80570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4171245833708677e-05,
+      "loss": 0.1934,
+      "step": 80580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.416674173497883e-05,
+      "loss": 0.1992,
+      "step": 80590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4162237636248987e-05,
+      "loss": 0.1982,
+      "step": 80600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4157733537519143e-05,
+      "loss": 0.1983,
+      "step": 80610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4153229438789297e-05,
+      "loss": 0.201,
+      "step": 80620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4148725340059455e-05,
+      "loss": 0.202,
+      "step": 80630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.414422124132961e-05,
+      "loss": 0.1968,
+      "step": 80640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4139717142599768e-05,
+      "loss": 0.1951,
+      "step": 80650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4135213043869922e-05,
+      "loss": 0.197,
+      "step": 80660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4130708945140078e-05,
+      "loss": 0.1967,
+      "step": 80670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4126204846410235e-05,
+      "loss": 0.191,
+      "step": 80680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.412170074768039e-05,
+      "loss": 0.1962,
+      "step": 80690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4117196648950545e-05,
+      "loss": 0.1996,
+      "step": 80700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4112692550220701e-05,
+      "loss": 0.1928,
+      "step": 80710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4108188451490859e-05,
+      "loss": 0.1954,
+      "step": 80720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4103684352761013e-05,
+      "loss": 0.1929,
+      "step": 80730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4099180254031168e-05,
+      "loss": 0.195,
+      "step": 80740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4094676155301326e-05,
+      "loss": 0.1948,
+      "step": 80750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.409017205657148e-05,
+      "loss": 0.1894,
+      "step": 80760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4085667957841636e-05,
+      "loss": 0.1953,
+      "step": 80770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4081163859111793e-05,
+      "loss": 0.195,
+      "step": 80780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.407665976038195e-05,
+      "loss": 0.1959,
+      "step": 80790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4072155661652103e-05,
+      "loss": 0.1941,
+      "step": 80800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.406765156292226e-05,
+      "loss": 0.1968,
+      "step": 80810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4063147464192417e-05,
+      "loss": 0.1974,
+      "step": 80820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.405864336546257e-05,
+      "loss": 0.1953,
+      "step": 80830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4054139266732726e-05,
+      "loss": 0.1918,
+      "step": 80840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4049635168002884e-05,
+      "loss": 0.196,
+      "step": 80850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.404513106927304e-05,
+      "loss": 0.1956,
+      "step": 80860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4040626970543194e-05,
+      "loss": 0.191,
+      "step": 80870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4036122871813351e-05,
+      "loss": 0.1939,
+      "step": 80880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4031618773083507e-05,
+      "loss": 0.1931,
+      "step": 80890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4027114674353661e-05,
+      "loss": 0.2011,
+      "step": 80900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4022610575623819e-05,
+      "loss": 0.1894,
+      "step": 80910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4018106476893975e-05,
+      "loss": 0.1949,
+      "step": 80920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4013602378164132e-05,
+      "loss": 0.1943,
+      "step": 80930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4009098279434284e-05,
+      "loss": 0.1925,
+      "step": 80940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4004594180704442e-05,
+      "loss": 0.196,
+      "step": 80950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4000090081974598e-05,
+      "loss": 0.1907,
+      "step": 80960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3995585983244752e-05,
+      "loss": 0.1898,
+      "step": 80970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.399108188451491e-05,
+      "loss": 0.1967,
+      "step": 80980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3986577785785065e-05,
+      "loss": 0.1869,
+      "step": 80990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3982073687055223e-05,
+      "loss": 0.1942,
+      "step": 81000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3977569588325377e-05,
+      "loss": 0.1874,
+      "step": 81010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3973065489595533e-05,
+      "loss": 0.1959,
+      "step": 81020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.396856139086569e-05,
+      "loss": 0.1933,
+      "step": 81030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3964057292135842e-05,
+      "loss": 0.1962,
+      "step": 81040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3959553193406e-05,
+      "loss": 0.1959,
+      "step": 81050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3955049094676156e-05,
+      "loss": 0.1922,
+      "step": 81060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3950544995946313e-05,
+      "loss": 0.195,
+      "step": 81070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3946040897216467e-05,
+      "loss": 0.1994,
+      "step": 81080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3941536798486623e-05,
+      "loss": 0.1922,
+      "step": 81090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.393703269975678e-05,
+      "loss": 0.192,
+      "step": 81100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3932528601026935e-05,
+      "loss": 0.2026,
+      "step": 81110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.392802450229709e-05,
+      "loss": 0.1941,
+      "step": 81120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3923520403567248e-05,
+      "loss": 0.1931,
+      "step": 81130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3919016304837404e-05,
+      "loss": 0.192,
+      "step": 81140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3914512206107558e-05,
+      "loss": 0.197,
+      "step": 81150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3910008107377714e-05,
+      "loss": 0.2011,
+      "step": 81160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3905504008647871e-05,
+      "loss": 0.1933,
+      "step": 81170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3900999909918025e-05,
+      "loss": 0.1957,
+      "step": 81180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3896495811188181e-05,
+      "loss": 0.1995,
+      "step": 81190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3891991712458339e-05,
+      "loss": 0.1938,
+      "step": 81200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3887487613728494e-05,
+      "loss": 0.1946,
+      "step": 81210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3882983514998648e-05,
+      "loss": 0.1967,
+      "step": 81220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3878479416268806e-05,
+      "loss": 0.1951,
+      "step": 81230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3873975317538962e-05,
+      "loss": 0.1953,
+      "step": 81240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3869471218809116e-05,
+      "loss": 0.1942,
+      "step": 81250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3864967120079272e-05,
+      "loss": 0.2024,
+      "step": 81260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.386046302134943e-05,
+      "loss": 0.1981,
+      "step": 81270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3855958922619585e-05,
+      "loss": 0.2014,
+      "step": 81280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3851454823889739e-05,
+      "loss": 0.196,
+      "step": 81290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3846950725159897e-05,
+      "loss": 0.1972,
+      "step": 81300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3842446626430052e-05,
+      "loss": 0.1946,
+      "step": 81310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3837942527700206e-05,
+      "loss": 0.1991,
+      "step": 81320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3833438428970364e-05,
+      "loss": 0.1957,
+      "step": 81330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.382893433024052e-05,
+      "loss": 0.2002,
+      "step": 81340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3824430231510677e-05,
+      "loss": 0.1921,
+      "step": 81350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.381992613278083e-05,
+      "loss": 0.2011,
+      "step": 81360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3815422034050987e-05,
+      "loss": 0.194,
+      "step": 81370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3810917935321143e-05,
+      "loss": 0.1915,
+      "step": 81380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3806413836591297e-05,
+      "loss": 0.1913,
+      "step": 81390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3801909737861455e-05,
+      "loss": 0.1935,
+      "step": 81400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.379740563913161e-05,
+      "loss": 0.1989,
+      "step": 81410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3792901540401768e-05,
+      "loss": 0.1996,
+      "step": 81420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3788397441671922e-05,
+      "loss": 0.1957,
+      "step": 81430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3783893342942078e-05,
+      "loss": 0.1932,
+      "step": 81440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3779389244212235e-05,
+      "loss": 0.1995,
+      "step": 81450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.377488514548239e-05,
+      "loss": 0.1939,
+      "step": 81460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3770381046752545e-05,
+      "loss": 0.1923,
+      "step": 81470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3765876948022703e-05,
+      "loss": 0.1943,
+      "step": 81480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3761372849292858e-05,
+      "loss": 0.1944,
+      "step": 81490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3756868750563013e-05,
+      "loss": 0.1925,
+      "step": 81500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3752364651833168e-05,
+      "loss": 0.1911,
+      "step": 81510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3747860553103326e-05,
+      "loss": 0.1896,
+      "step": 81520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.374335645437348e-05,
+      "loss": 0.1914,
+      "step": 81530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3738852355643636e-05,
+      "loss": 0.1948,
+      "step": 81540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3734348256913793e-05,
+      "loss": 0.1873,
+      "step": 81550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3729844158183949e-05,
+      "loss": 0.1898,
+      "step": 81560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3725340059454103e-05,
+      "loss": 0.1969,
+      "step": 81570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.372083596072426e-05,
+      "loss": 0.1988,
+      "step": 81580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3716331861994416e-05,
+      "loss": 0.1927,
+      "step": 81590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.371182776326457e-05,
+      "loss": 0.1932,
+      "step": 81600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3707323664534726e-05,
+      "loss": 0.1942,
+      "step": 81610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3702819565804884e-05,
+      "loss": 0.1963,
+      "step": 81620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.369831546707504e-05,
+      "loss": 0.1919,
+      "step": 81630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3693811368345194e-05,
+      "loss": 0.1963,
+      "step": 81640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3689307269615351e-05,
+      "loss": 0.1941,
+      "step": 81650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3684803170885507e-05,
+      "loss": 0.1905,
+      "step": 81660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3680299072155661e-05,
+      "loss": 0.1939,
+      "step": 81670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3675794973425819e-05,
+      "loss": 0.1975,
+      "step": 81680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3671290874695974e-05,
+      "loss": 0.1949,
+      "step": 81690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3666786775966132e-05,
+      "loss": 0.1945,
+      "step": 81700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3662282677236284e-05,
+      "loss": 0.1902,
+      "step": 81710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3657778578506442e-05,
+      "loss": 0.1961,
+      "step": 81720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3653274479776598e-05,
+      "loss": 0.1971,
+      "step": 81730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3648770381046752e-05,
+      "loss": 0.194,
+      "step": 81740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.364426628231691e-05,
+      "loss": 0.1955,
+      "step": 81750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3639762183587065e-05,
+      "loss": 0.1975,
+      "step": 81760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3635258084857222e-05,
+      "loss": 0.1992,
+      "step": 81770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3630753986127377e-05,
+      "loss": 0.202,
+      "step": 81780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3626249887397532e-05,
+      "loss": 0.19,
+      "step": 81790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.362174578866769e-05,
+      "loss": 0.1977,
+      "step": 81800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3617241689937842e-05,
+      "loss": 0.1918,
+      "step": 81810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3612737591208e-05,
+      "loss": 0.1949,
+      "step": 81820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3608233492478156e-05,
+      "loss": 0.196,
+      "step": 81830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3603729393748313e-05,
+      "loss": 0.1928,
+      "step": 81840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3599225295018467e-05,
+      "loss": 0.1905,
+      "step": 81850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3594721196288623e-05,
+      "loss": 0.195,
+      "step": 81860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.359021709755878e-05,
+      "loss": 0.1916,
+      "step": 81870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3585712998828935e-05,
+      "loss": 0.1954,
+      "step": 81880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.358120890009909e-05,
+      "loss": 0.1931,
+      "step": 81890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3576704801369248e-05,
+      "loss": 0.1962,
+      "step": 81900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3572200702639404e-05,
+      "loss": 0.1931,
+      "step": 81910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3567696603909558e-05,
+      "loss": 0.1915,
+      "step": 81920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3563192505179714e-05,
+      "loss": 0.1908,
+      "step": 81930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3558688406449871e-05,
+      "loss": 0.1925,
+      "step": 81940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3554184307720025e-05,
+      "loss": 0.1934,
+      "step": 81950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3549680208990181e-05,
+      "loss": 0.1988,
+      "step": 81960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3545176110260338e-05,
+      "loss": 0.1912,
+      "step": 81970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3540672011530494e-05,
+      "loss": 0.1968,
+      "step": 81980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3536618322673633e-05,
+      "loss": 0.1941,
+      "step": 81990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.353211422394379e-05,
+      "loss": 0.1915,
+      "step": 82000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3527610125213946e-05,
+      "loss": 0.1905,
+      "step": 82010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.35231060264841e-05,
+      "loss": 0.1933,
+      "step": 82020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3518601927754256e-05,
+      "loss": 0.1889,
+      "step": 82030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3514097829024414e-05,
+      "loss": 0.1982,
+      "step": 82040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3509593730294568e-05,
+      "loss": 0.191,
+      "step": 82050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3505089631564723e-05,
+      "loss": 0.1961,
+      "step": 82060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3500585532834881e-05,
+      "loss": 0.1952,
+      "step": 82070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3496081434105037e-05,
+      "loss": 0.1939,
+      "step": 82080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3491577335375191e-05,
+      "loss": 0.1924,
+      "step": 82090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3487073236645348e-05,
+      "loss": 0.1921,
+      "step": 82100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3482569137915504e-05,
+      "loss": 0.1974,
+      "step": 82110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3478065039185658e-05,
+      "loss": 0.1915,
+      "step": 82120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3473560940455816e-05,
+      "loss": 0.1941,
+      "step": 82130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3469056841725972e-05,
+      "loss": 0.1902,
+      "step": 82140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3464552742996129e-05,
+      "loss": 0.1926,
+      "step": 82150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3460048644266281e-05,
+      "loss": 0.1965,
+      "step": 82160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3455544545536439e-05,
+      "loss": 0.1903,
+      "step": 82170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3451040446806595e-05,
+      "loss": 0.2005,
+      "step": 82180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3446536348076749e-05,
+      "loss": 0.1936,
+      "step": 82190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3442032249346906e-05,
+      "loss": 0.1948,
+      "step": 82200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3437528150617062e-05,
+      "loss": 0.1958,
+      "step": 82210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.343302405188722e-05,
+      "loss": 0.1987,
+      "step": 82220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3428519953157374e-05,
+      "loss": 0.193,
+      "step": 82230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.342401585442753e-05,
+      "loss": 0.1923,
+      "step": 82240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3419511755697687e-05,
+      "loss": 0.1959,
+      "step": 82250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.341500765696784e-05,
+      "loss": 0.1879,
+      "step": 82260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3410503558237997e-05,
+      "loss": 0.1932,
+      "step": 82270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3405999459508153e-05,
+      "loss": 0.1905,
+      "step": 82280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.340149536077831e-05,
+      "loss": 0.1941,
+      "step": 82290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3396991262048464e-05,
+      "loss": 0.1935,
+      "step": 82300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.339248716331862e-05,
+      "loss": 0.1958,
+      "step": 82310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3387983064588778e-05,
+      "loss": 0.1907,
+      "step": 82320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3383478965858932e-05,
+      "loss": 0.1918,
+      "step": 82330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3378974867129088e-05,
+      "loss": 0.1976,
+      "step": 82340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3374470768399245e-05,
+      "loss": 0.198,
+      "step": 82350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.33699666696694e-05,
+      "loss": 0.1953,
+      "step": 82360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3365462570939555e-05,
+      "loss": 0.1963,
+      "step": 82370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.336095847220971e-05,
+      "loss": 0.1929,
+      "step": 82380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3356454373479868e-05,
+      "loss": 0.1956,
+      "step": 82390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3351950274750022e-05,
+      "loss": 0.1948,
+      "step": 82400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3347446176020178e-05,
+      "loss": 0.1962,
+      "step": 82410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3342942077290336e-05,
+      "loss": 0.1974,
+      "step": 82420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3338437978560491e-05,
+      "loss": 0.1941,
+      "step": 82430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3333933879830646e-05,
+      "loss": 0.1897,
+      "step": 82440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3329429781100803e-05,
+      "loss": 0.1935,
+      "step": 82450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3324925682370959e-05,
+      "loss": 0.1929,
+      "step": 82460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3320421583641113e-05,
+      "loss": 0.1916,
+      "step": 82470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3315917484911269e-05,
+      "loss": 0.1886,
+      "step": 82480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3311413386181426e-05,
+      "loss": 0.1924,
+      "step": 82490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3306909287451582e-05,
+      "loss": 0.196,
+      "step": 82500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3302405188721736e-05,
+      "loss": 0.1934,
+      "step": 82510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3297901089991894e-05,
+      "loss": 0.1917,
+      "step": 82520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.329339699126205e-05,
+      "loss": 0.193,
+      "step": 82530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3288892892532203e-05,
+      "loss": 0.1892,
+      "step": 82540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3284388793802361e-05,
+      "loss": 0.1957,
+      "step": 82550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3279884695072517e-05,
+      "loss": 0.1945,
+      "step": 82560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3275380596342674e-05,
+      "loss": 0.199,
+      "step": 82570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3270876497612827e-05,
+      "loss": 0.1932,
+      "step": 82580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3266372398882984e-05,
+      "loss": 0.1939,
+      "step": 82590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.326186830015314e-05,
+      "loss": 0.1942,
+      "step": 82600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3257364201423294e-05,
+      "loss": 0.1925,
+      "step": 82610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3252860102693452e-05,
+      "loss": 0.1915,
+      "step": 82620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3248356003963607e-05,
+      "loss": 0.1933,
+      "step": 82630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3243851905233765e-05,
+      "loss": 0.197,
+      "step": 82640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3239347806503919e-05,
+      "loss": 0.1938,
+      "step": 82650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3234843707774075e-05,
+      "loss": 0.1928,
+      "step": 82660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3230339609044232e-05,
+      "loss": 0.1905,
+      "step": 82670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3225835510314385e-05,
+      "loss": 0.1908,
+      "step": 82680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3221331411584542e-05,
+      "loss": 0.1925,
+      "step": 82690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3216827312854698e-05,
+      "loss": 0.193,
+      "step": 82700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3212323214124855e-05,
+      "loss": 0.1926,
+      "step": 82710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.320781911539501e-05,
+      "loss": 0.1907,
+      "step": 82720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3203315016665165e-05,
+      "loss": 0.1895,
+      "step": 82730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3198810917935323e-05,
+      "loss": 0.1941,
+      "step": 82740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3194306819205477e-05,
+      "loss": 0.1934,
+      "step": 82750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3189802720475633e-05,
+      "loss": 0.1932,
+      "step": 82760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.318529862174579e-05,
+      "loss": 0.1876,
+      "step": 82770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3180794523015946e-05,
+      "loss": 0.1993,
+      "step": 82780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.31762904242861e-05,
+      "loss": 0.1927,
+      "step": 82790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3171786325556258e-05,
+      "loss": 0.1932,
+      "step": 82800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3167282226826413e-05,
+      "loss": 0.1932,
+      "step": 82810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3162778128096568e-05,
+      "loss": 0.1905,
+      "step": 82820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3158274029366723e-05,
+      "loss": 0.1949,
+      "step": 82830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.315376993063688e-05,
+      "loss": 0.1981,
+      "step": 82840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3149265831907037e-05,
+      "loss": 0.1904,
+      "step": 82850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.314476173317719e-05,
+      "loss": 0.1948,
+      "step": 82860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3140257634447348e-05,
+      "loss": 0.1927,
+      "step": 82870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3135753535717504e-05,
+      "loss": 0.1926,
+      "step": 82880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3131249436987658e-05,
+      "loss": 0.192,
+      "step": 82890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3126745338257816e-05,
+      "loss": 0.1896,
+      "step": 82900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3122241239527971e-05,
+      "loss": 0.1936,
+      "step": 82910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3117737140798129e-05,
+      "loss": 0.1925,
+      "step": 82920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3113233042068281e-05,
+      "loss": 0.1939,
+      "step": 82930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3108728943338439e-05,
+      "loss": 0.196,
+      "step": 82940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3104224844608595e-05,
+      "loss": 0.1997,
+      "step": 82950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3099720745878749e-05,
+      "loss": 0.1874,
+      "step": 82960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3095216647148906e-05,
+      "loss": 0.1929,
+      "step": 82970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3090712548419062e-05,
+      "loss": 0.1956,
+      "step": 82980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.308620844968922e-05,
+      "loss": 0.1914,
+      "step": 82990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3081704350959374e-05,
+      "loss": 0.1964,
+      "step": 83000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.307720025222953e-05,
+      "loss": 0.1871,
+      "step": 83010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3072696153499687e-05,
+      "loss": 0.1959,
+      "step": 83020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.306819205476984e-05,
+      "loss": 0.1933,
+      "step": 83030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3063687956039997e-05,
+      "loss": 0.1976,
+      "step": 83040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3059183857310153e-05,
+      "loss": 0.1914,
+      "step": 83050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.305467975858031e-05,
+      "loss": 0.197,
+      "step": 83060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3050175659850464e-05,
+      "loss": 0.193,
+      "step": 83070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.304567156112062e-05,
+      "loss": 0.1967,
+      "step": 83080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3041167462390777e-05,
+      "loss": 0.194,
+      "step": 83090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3036663363660932e-05,
+      "loss": 0.1909,
+      "step": 83100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3032159264931087e-05,
+      "loss": 0.1903,
+      "step": 83110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3027655166201245e-05,
+      "loss": 0.1918,
+      "step": 83120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.30231510674714e-05,
+      "loss": 0.1944,
+      "step": 83130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3018646968741555e-05,
+      "loss": 0.1951,
+      "step": 83140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.301414287001171e-05,
+      "loss": 0.1916,
+      "step": 83150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3009638771281868e-05,
+      "loss": 0.1912,
+      "step": 83160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3005134672552022e-05,
+      "loss": 0.1981,
+      "step": 83170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3000630573822178e-05,
+      "loss": 0.1948,
+      "step": 83180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2996126475092335e-05,
+      "loss": 0.1916,
+      "step": 83190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2991622376362491e-05,
+      "loss": 0.1925,
+      "step": 83200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2987118277632645e-05,
+      "loss": 0.1909,
+      "step": 83210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2982614178902803e-05,
+      "loss": 0.1997,
+      "step": 83220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2978110080172959e-05,
+      "loss": 0.1972,
+      "step": 83230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2973605981443113e-05,
+      "loss": 0.1948,
+      "step": 83240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2969101882713269e-05,
+      "loss": 0.1889,
+      "step": 83250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2964597783983426e-05,
+      "loss": 0.1916,
+      "step": 83260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2960093685253582e-05,
+      "loss": 0.1899,
+      "step": 83270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2955589586523736e-05,
+      "loss": 0.1959,
+      "step": 83280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2951085487793893e-05,
+      "loss": 0.1906,
+      "step": 83290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.294658138906405e-05,
+      "loss": 0.1933,
+      "step": 83300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2942077290334203e-05,
+      "loss": 0.2009,
+      "step": 83310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.293757319160436e-05,
+      "loss": 0.1842,
+      "step": 83320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2933069092874517e-05,
+      "loss": 0.1923,
+      "step": 83330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2928564994144674e-05,
+      "loss": 0.1849,
+      "step": 83340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2924060895414827e-05,
+      "loss": 0.1974,
+      "step": 83350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2919556796684984e-05,
+      "loss": 0.1946,
+      "step": 83360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.291505269795514e-05,
+      "loss": 0.194,
+      "step": 83370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2910548599225294e-05,
+      "loss": 0.1954,
+      "step": 83380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2906044500495451e-05,
+      "loss": 0.1929,
+      "step": 83390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2901540401765607e-05,
+      "loss": 0.1899,
+      "step": 83400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2897036303035765e-05,
+      "loss": 0.1926,
+      "step": 83410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2892532204305919e-05,
+      "loss": 0.1994,
+      "step": 83420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2888028105576075e-05,
+      "loss": 0.1921,
+      "step": 83430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2883524006846232e-05,
+      "loss": 0.1918,
+      "step": 83440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2879019908116386e-05,
+      "loss": 0.1944,
+      "step": 83450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2874515809386542e-05,
+      "loss": 0.1889,
+      "step": 83460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2870011710656698e-05,
+      "loss": 0.1935,
+      "step": 83470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2865507611926855e-05,
+      "loss": 0.1906,
+      "step": 83480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.286100351319701e-05,
+      "loss": 0.1958,
+      "step": 83490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2856499414467165e-05,
+      "loss": 0.1912,
+      "step": 83500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2851995315737323e-05,
+      "loss": 0.1918,
+      "step": 83510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2847491217007477e-05,
+      "loss": 0.1938,
+      "step": 83520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2842987118277633e-05,
+      "loss": 0.1888,
+      "step": 83530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.283848301954779e-05,
+      "loss": 0.1936,
+      "step": 83540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2833978920817946e-05,
+      "loss": 0.1915,
+      "step": 83550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.28294748220881e-05,
+      "loss": 0.1955,
+      "step": 83560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2824970723358257e-05,
+      "loss": 0.1875,
+      "step": 83570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2820466624628413e-05,
+      "loss": 0.1923,
+      "step": 83580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2815962525898567e-05,
+      "loss": 0.1926,
+      "step": 83590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2811458427168723e-05,
+      "loss": 0.1921,
+      "step": 83600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.280695432843888e-05,
+      "loss": 0.1905,
+      "step": 83610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2802450229709036e-05,
+      "loss": 0.1966,
+      "step": 83620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.279794613097919e-05,
+      "loss": 0.1967,
+      "step": 83630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2793442032249348e-05,
+      "loss": 0.189,
+      "step": 83640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2788937933519504e-05,
+      "loss": 0.1951,
+      "step": 83650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2784433834789658e-05,
+      "loss": 0.1907,
+      "step": 83660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2779929736059815e-05,
+      "loss": 0.1933,
+      "step": 83670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2775425637329971e-05,
+      "loss": 0.19,
+      "step": 83680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2770921538600129e-05,
+      "loss": 0.1933,
+      "step": 83690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2766417439870281e-05,
+      "loss": 0.1861,
+      "step": 83700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2761913341140439e-05,
+      "loss": 0.1925,
+      "step": 83710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2757409242410594e-05,
+      "loss": 0.1967,
+      "step": 83720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2752905143680749e-05,
+      "loss": 0.1971,
+      "step": 83730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2748401044950906e-05,
+      "loss": 0.1927,
+      "step": 83740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2743896946221062e-05,
+      "loss": 0.1874,
+      "step": 83750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.273939284749122e-05,
+      "loss": 0.1965,
+      "step": 83760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2734888748761373e-05,
+      "loss": 0.192,
+      "step": 83770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.273038465003153e-05,
+      "loss": 0.1877,
+      "step": 83780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2725880551301687e-05,
+      "loss": 0.1918,
+      "step": 83790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2721376452571839e-05,
+      "loss": 0.1946,
+      "step": 83800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2716872353841997e-05,
+      "loss": 0.1963,
+      "step": 83810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2712368255112152e-05,
+      "loss": 0.1871,
+      "step": 83820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.270786415638231e-05,
+      "loss": 0.1866,
+      "step": 83830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2703360057652464e-05,
+      "loss": 0.193,
+      "step": 83840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.269885595892262e-05,
+      "loss": 0.1905,
+      "step": 83850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2694351860192777e-05,
+      "loss": 0.1949,
+      "step": 83860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2689847761462931e-05,
+      "loss": 0.192,
+      "step": 83870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2685343662733087e-05,
+      "loss": 0.1989,
+      "step": 83880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2680839564003245e-05,
+      "loss": 0.1927,
+      "step": 83890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.26763354652734e-05,
+      "loss": 0.1882,
+      "step": 83900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2671831366543555e-05,
+      "loss": 0.1902,
+      "step": 83910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.266732726781371e-05,
+      "loss": 0.1952,
+      "step": 83920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2662823169083868e-05,
+      "loss": 0.1939,
+      "step": 83930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2658319070354022e-05,
+      "loss": 0.1907,
+      "step": 83940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2653814971624178e-05,
+      "loss": 0.1954,
+      "step": 83950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2649310872894335e-05,
+      "loss": 0.1926,
+      "step": 83960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2644806774164491e-05,
+      "loss": 0.181,
+      "step": 83970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2640302675434645e-05,
+      "loss": 0.1945,
+      "step": 83980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2635798576704803e-05,
+      "loss": 0.1905,
+      "step": 83990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2631294477974958e-05,
+      "loss": 0.1937,
+      "step": 84000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2627240789118097e-05,
+      "loss": 0.1903,
+      "step": 84010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2622736690388253e-05,
+      "loss": 0.1987,
+      "step": 84020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.261823259165841e-05,
+      "loss": 0.1952,
+      "step": 84030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2613728492928565e-05,
+      "loss": 0.1883,
+      "step": 84040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.260922439419872e-05,
+      "loss": 0.1874,
+      "step": 84050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2604720295468878e-05,
+      "loss": 0.1896,
+      "step": 84060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2600216196739034e-05,
+      "loss": 0.1892,
+      "step": 84070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2595712098009188e-05,
+      "loss": 0.1989,
+      "step": 84080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2591207999279345e-05,
+      "loss": 0.1914,
+      "step": 84090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2586703900549501e-05,
+      "loss": 0.1963,
+      "step": 84100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2582199801819655e-05,
+      "loss": 0.1949,
+      "step": 84110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2577695703089813e-05,
+      "loss": 0.1892,
+      "step": 84120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2573191604359968e-05,
+      "loss": 0.1896,
+      "step": 84130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2568687505630124e-05,
+      "loss": 0.1943,
+      "step": 84140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2564183406900278e-05,
+      "loss": 0.1869,
+      "step": 84150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2559679308170436e-05,
+      "loss": 0.189,
+      "step": 84160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2555175209440592e-05,
+      "loss": 0.1936,
+      "step": 84170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2550671110710746e-05,
+      "loss": 0.1927,
+      "step": 84180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2546167011980903e-05,
+      "loss": 0.1932,
+      "step": 84190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2541662913251059e-05,
+      "loss": 0.1927,
+      "step": 84200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2537158814521217e-05,
+      "loss": 0.1934,
+      "step": 84210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.253265471579137e-05,
+      "loss": 0.1879,
+      "step": 84220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2528150617061526e-05,
+      "loss": 0.1904,
+      "step": 84230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2523646518331684e-05,
+      "loss": 0.1886,
+      "step": 84240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2519142419601836e-05,
+      "loss": 0.197,
+      "step": 84250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2514638320871994e-05,
+      "loss": 0.189,
+      "step": 84260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.251013422214215e-05,
+      "loss": 0.1885,
+      "step": 84270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2505630123412307e-05,
+      "loss": 0.1891,
+      "step": 84280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2501126024682461e-05,
+      "loss": 0.1948,
+      "step": 84290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2496621925952617e-05,
+      "loss": 0.1931,
+      "step": 84300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2492117827222773e-05,
+      "loss": 0.1927,
+      "step": 84310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2487613728492929e-05,
+      "loss": 0.1928,
+      "step": 84320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2483109629763084e-05,
+      "loss": 0.1923,
+      "step": 84330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2478605531033242e-05,
+      "loss": 0.1911,
+      "step": 84340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2474101432303396e-05,
+      "loss": 0.189,
+      "step": 84350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2469597333573554e-05,
+      "loss": 0.19,
+      "step": 84360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2465093234843708e-05,
+      "loss": 0.1934,
+      "step": 84370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2460589136113863e-05,
+      "loss": 0.1963,
+      "step": 84380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2456085037384021e-05,
+      "loss": 0.1964,
+      "step": 84390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2451580938654175e-05,
+      "loss": 0.1922,
+      "step": 84400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2447076839924332e-05,
+      "loss": 0.186,
+      "step": 84410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2442572741194487e-05,
+      "loss": 0.201,
+      "step": 84420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2438068642464644e-05,
+      "loss": 0.185,
+      "step": 84430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.24335645437348e-05,
+      "loss": 0.1925,
+      "step": 84440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2429060445004954e-05,
+      "loss": 0.1921,
+      "step": 84450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2424556346275111e-05,
+      "loss": 0.1921,
+      "step": 84460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2420052247545266e-05,
+      "loss": 0.1881,
+      "step": 84470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2415548148815423e-05,
+      "loss": 0.1828,
+      "step": 84480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2411044050085579e-05,
+      "loss": 0.1935,
+      "step": 84490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2406539951355735e-05,
+      "loss": 0.1898,
+      "step": 84500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.240203585262589e-05,
+      "loss": 0.1902,
+      "step": 84510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2397531753896045e-05,
+      "loss": 0.1896,
+      "step": 84520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2393027655166202e-05,
+      "loss": 0.1914,
+      "step": 84530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2388523556436358e-05,
+      "loss": 0.1923,
+      "step": 84540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2384019457706514e-05,
+      "loss": 0.1913,
+      "step": 84550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.237951535897667e-05,
+      "loss": 0.1922,
+      "step": 84560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2375011260246825e-05,
+      "loss": 0.194,
+      "step": 84570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2370507161516981e-05,
+      "loss": 0.1895,
+      "step": 84580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2366003062787137e-05,
+      "loss": 0.194,
+      "step": 84590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2361498964057293e-05,
+      "loss": 0.1936,
+      "step": 84600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2356994865327448e-05,
+      "loss": 0.1886,
+      "step": 84610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2352490766597604e-05,
+      "loss": 0.191,
+      "step": 84620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.234798666786776e-05,
+      "loss": 0.1909,
+      "step": 84630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2343482569137916e-05,
+      "loss": 0.1916,
+      "step": 84640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2338978470408072e-05,
+      "loss": 0.1857,
+      "step": 84650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2334474371678227e-05,
+      "loss": 0.1868,
+      "step": 84660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2329970272948383e-05,
+      "loss": 0.1867,
+      "step": 84670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2325466174218539e-05,
+      "loss": 0.1946,
+      "step": 84680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2320962075488695e-05,
+      "loss": 0.1916,
+      "step": 84690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.231645797675885e-05,
+      "loss": 0.1883,
+      "step": 84700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2311953878029008e-05,
+      "loss": 0.1882,
+      "step": 84710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2307449779299162e-05,
+      "loss": 0.1916,
+      "step": 84720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2302945680569318e-05,
+      "loss": 0.1886,
+      "step": 84730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2298441581839474e-05,
+      "loss": 0.1903,
+      "step": 84740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.229393748310963e-05,
+      "loss": 0.1939,
+      "step": 84750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2289433384379787e-05,
+      "loss": 0.1975,
+      "step": 84760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2284929285649941e-05,
+      "loss": 0.1845,
+      "step": 84770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2280425186920099e-05,
+      "loss": 0.1916,
+      "step": 84780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2275921088190253e-05,
+      "loss": 0.1884,
+      "step": 84790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2271416989460409e-05,
+      "loss": 0.1898,
+      "step": 84800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2266912890730566e-05,
+      "loss": 0.1882,
+      "step": 84810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.226240879200072e-05,
+      "loss": 0.1935,
+      "step": 84820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2257904693270878e-05,
+      "loss": 0.1802,
+      "step": 84830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2253400594541034e-05,
+      "loss": 0.1891,
+      "step": 84840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.224889649581119e-05,
+      "loss": 0.1981,
+      "step": 84850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2244392397081345e-05,
+      "loss": 0.1916,
+      "step": 84860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.22398882983515e-05,
+      "loss": 0.1958,
+      "step": 84870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2235384199621657e-05,
+      "loss": 0.1869,
+      "step": 84880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2230880100891813e-05,
+      "loss": 0.1859,
+      "step": 84890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2226376002161968e-05,
+      "loss": 0.1847,
+      "step": 84900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2221871903432124e-05,
+      "loss": 0.1975,
+      "step": 84910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.221736780470228e-05,
+      "loss": 0.1886,
+      "step": 84920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2212863705972436e-05,
+      "loss": 0.191,
+      "step": 84930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2208359607242591e-05,
+      "loss": 0.1937,
+      "step": 84940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2203855508512747e-05,
+      "loss": 0.186,
+      "step": 84950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2199351409782903e-05,
+      "loss": 0.1857,
+      "step": 84960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2194847311053059e-05,
+      "loss": 0.1889,
+      "step": 84970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2190343212323215e-05,
+      "loss": 0.1928,
+      "step": 84980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.218583911359337e-05,
+      "loss": 0.1902,
+      "step": 84990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2181335014863526e-05,
+      "loss": 0.1923,
+      "step": 85000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2176830916133682e-05,
+      "loss": 0.1923,
+      "step": 85010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2172326817403838e-05,
+      "loss": 0.1896,
+      "step": 85020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2167822718673994e-05,
+      "loss": 0.1889,
+      "step": 85030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.216331861994415e-05,
+      "loss": 0.1941,
+      "step": 85040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2158814521214305e-05,
+      "loss": 0.1933,
+      "step": 85050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2154310422484463e-05,
+      "loss": 0.187,
+      "step": 85060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2149806323754617e-05,
+      "loss": 0.1917,
+      "step": 85070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2145302225024773e-05,
+      "loss": 0.1915,
+      "step": 85080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2140798126294928e-05,
+      "loss": 0.1965,
+      "step": 85090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2136294027565084e-05,
+      "loss": 0.1945,
+      "step": 85100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2131789928835242e-05,
+      "loss": 0.1881,
+      "step": 85110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2127285830105396e-05,
+      "loss": 0.1909,
+      "step": 85120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2122781731375553e-05,
+      "loss": 0.1913,
+      "step": 85130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2118277632645707e-05,
+      "loss": 0.1941,
+      "step": 85140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2113773533915863e-05,
+      "loss": 0.1882,
+      "step": 85150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.210926943518602e-05,
+      "loss": 0.1945,
+      "step": 85160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2104765336456175e-05,
+      "loss": 0.1881,
+      "step": 85170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2100261237726332e-05,
+      "loss": 0.1913,
+      "step": 85180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2095757138996486e-05,
+      "loss": 0.1935,
+      "step": 85190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2091253040266644e-05,
+      "loss": 0.1918,
+      "step": 85200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.20867489415368e-05,
+      "loss": 0.1899,
+      "step": 85210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2082244842806954e-05,
+      "loss": 0.1904,
+      "step": 85220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2077740744077111e-05,
+      "loss": 0.1926,
+      "step": 85230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2073236645347265e-05,
+      "loss": 0.1933,
+      "step": 85240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2068732546617423e-05,
+      "loss": 0.1869,
+      "step": 85250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2064228447887579e-05,
+      "loss": 0.1953,
+      "step": 85260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2059724349157735e-05,
+      "loss": 0.1925,
+      "step": 85270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.205522025042789e-05,
+      "loss": 0.1869,
+      "step": 85280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2050716151698044e-05,
+      "loss": 0.1874,
+      "step": 85290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2046212052968202e-05,
+      "loss": 0.1885,
+      "step": 85300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2041707954238358e-05,
+      "loss": 0.1885,
+      "step": 85310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2037203855508514e-05,
+      "loss": 0.1994,
+      "step": 85320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.203269975677867e-05,
+      "loss": 0.1921,
+      "step": 85330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2028195658048825e-05,
+      "loss": 0.1897,
+      "step": 85340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2023691559318981e-05,
+      "loss": 0.1923,
+      "step": 85350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2019187460589137e-05,
+      "loss": 0.1888,
+      "step": 85360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2014683361859293e-05,
+      "loss": 0.1907,
+      "step": 85370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2010179263129448e-05,
+      "loss": 0.1837,
+      "step": 85380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2005675164399604e-05,
+      "loss": 0.187,
+      "step": 85390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.200117106566976e-05,
+      "loss": 0.1831,
+      "step": 85400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1996666966939916e-05,
+      "loss": 0.1878,
+      "step": 85410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1992162868210071e-05,
+      "loss": 0.1909,
+      "step": 85420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1987658769480227e-05,
+      "loss": 0.1886,
+      "step": 85430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1983154670750383e-05,
+      "loss": 0.1862,
+      "step": 85440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1978650572020539e-05,
+      "loss": 0.1967,
+      "step": 85450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1974146473290695e-05,
+      "loss": 0.1848,
+      "step": 85460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.196964237456085e-05,
+      "loss": 0.1913,
+      "step": 85470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1965138275831008e-05,
+      "loss": 0.1903,
+      "step": 85480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1960634177101162e-05,
+      "loss": 0.1923,
+      "step": 85490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1956130078371318e-05,
+      "loss": 0.1959,
+      "step": 85500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1951625979641474e-05,
+      "loss": 0.1874,
+      "step": 85510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.194712188091163e-05,
+      "loss": 0.188,
+      "step": 85520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1942617782181787e-05,
+      "loss": 0.1884,
+      "step": 85530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1938113683451941e-05,
+      "loss": 0.1901,
+      "step": 85540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1933609584722099e-05,
+      "loss": 0.1905,
+      "step": 85550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1929105485992254e-05,
+      "loss": 0.1912,
+      "step": 85560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1924601387262408e-05,
+      "loss": 0.1886,
+      "step": 85570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1920097288532566e-05,
+      "loss": 0.1926,
+      "step": 85580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.191559318980272e-05,
+      "loss": 0.1911,
+      "step": 85590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1911089091072878e-05,
+      "loss": 0.1932,
+      "step": 85600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1906584992343033e-05,
+      "loss": 0.1911,
+      "step": 85610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1902080893613189e-05,
+      "loss": 0.1823,
+      "step": 85620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1897576794883345e-05,
+      "loss": 0.1964,
+      "step": 85630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1893072696153499e-05,
+      "loss": 0.1892,
+      "step": 85640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1888568597423657e-05,
+      "loss": 0.1884,
+      "step": 85650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1884064498693812e-05,
+      "loss": 0.1913,
+      "step": 85660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1879560399963968e-05,
+      "loss": 0.1912,
+      "step": 85670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1875056301234124e-05,
+      "loss": 0.1906,
+      "step": 85680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.187055220250428e-05,
+      "loss": 0.1865,
+      "step": 85690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1866048103774436e-05,
+      "loss": 0.1868,
+      "step": 85700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1861544005044591e-05,
+      "loss": 0.1901,
+      "step": 85710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1857039906314747e-05,
+      "loss": 0.1909,
+      "step": 85720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1852535807584903e-05,
+      "loss": 0.1871,
+      "step": 85730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1848031708855059e-05,
+      "loss": 0.1909,
+      "step": 85740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1843527610125215e-05,
+      "loss": 0.1916,
+      "step": 85750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.183902351139537e-05,
+      "loss": 0.1883,
+      "step": 85760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1834519412665526e-05,
+      "loss": 0.1873,
+      "step": 85770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1830015313935682e-05,
+      "loss": 0.1855,
+      "step": 85780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1825511215205838e-05,
+      "loss": 0.1904,
+      "step": 85790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1821007116475994e-05,
+      "loss": 0.1847,
+      "step": 85800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.181650301774615e-05,
+      "loss": 0.1887,
+      "step": 85810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1811998919016305e-05,
+      "loss": 0.1837,
+      "step": 85820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1807494820286463e-05,
+      "loss": 0.1865,
+      "step": 85830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1802990721556617e-05,
+      "loss": 0.1897,
+      "step": 85840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1798486622826773e-05,
+      "loss": 0.1875,
+      "step": 85850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1793982524096928e-05,
+      "loss": 0.192,
+      "step": 85860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1789478425367084e-05,
+      "loss": 0.1908,
+      "step": 85870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1784974326637242e-05,
+      "loss": 0.1898,
+      "step": 85880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1780470227907396e-05,
+      "loss": 0.1869,
+      "step": 85890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1775966129177553e-05,
+      "loss": 0.1931,
+      "step": 85900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1771462030447707e-05,
+      "loss": 0.1861,
+      "step": 85910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1766957931717863e-05,
+      "loss": 0.1855,
+      "step": 85920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.176245383298802e-05,
+      "loss": 0.1942,
+      "step": 85930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1757949734258175e-05,
+      "loss": 0.1928,
+      "step": 85940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1753445635528332e-05,
+      "loss": 0.1819,
+      "step": 85950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1748941536798486e-05,
+      "loss": 0.1886,
+      "step": 85960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1744437438068644e-05,
+      "loss": 0.1923,
+      "step": 85970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.17399333393388e-05,
+      "loss": 0.189,
+      "step": 85980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1735429240608954e-05,
+      "loss": 0.1835,
+      "step": 85990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1730925141879111e-05,
+      "loss": 0.1884,
+      "step": 86000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1726421043149265e-05,
+      "loss": 0.1891,
+      "step": 86010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1721916944419423e-05,
+      "loss": 0.193,
+      "step": 86020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1717863255562563e-05,
+      "loss": 0.1895,
+      "step": 86030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1713359156832717e-05,
+      "loss": 0.1908,
+      "step": 86040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1708855058102875e-05,
+      "loss": 0.1881,
+      "step": 86050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1704350959373029e-05,
+      "loss": 0.1898,
+      "step": 86060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1699846860643186e-05,
+      "loss": 0.1909,
+      "step": 86070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1695342761913342e-05,
+      "loss": 0.1939,
+      "step": 86080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1690838663183498e-05,
+      "loss": 0.1899,
+      "step": 86090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1686334564453654e-05,
+      "loss": 0.1911,
+      "step": 86100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1681830465723808e-05,
+      "loss": 0.1892,
+      "step": 86110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1677326366993965e-05,
+      "loss": 0.1917,
+      "step": 86120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1672822268264121e-05,
+      "loss": 0.1956,
+      "step": 86130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1668318169534277e-05,
+      "loss": 0.1876,
+      "step": 86140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1663814070804433e-05,
+      "loss": 0.1847,
+      "step": 86150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1659309972074589e-05,
+      "loss": 0.1931,
+      "step": 86160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1654805873344744e-05,
+      "loss": 0.1891,
+      "step": 86170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.16503017746149e-05,
+      "loss": 0.1836,
+      "step": 86180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1645797675885056e-05,
+      "loss": 0.1895,
+      "step": 86190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1641293577155212e-05,
+      "loss": 0.1853,
+      "step": 86200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1636789478425368e-05,
+      "loss": 0.1901,
+      "step": 86210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1632285379695523e-05,
+      "loss": 0.1904,
+      "step": 86220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.162778128096568e-05,
+      "loss": 0.1924,
+      "step": 86230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1623277182235835e-05,
+      "loss": 0.192,
+      "step": 86240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.161877308350599e-05,
+      "loss": 0.1855,
+      "step": 86250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1614268984776147e-05,
+      "loss": 0.1856,
+      "step": 86260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1609764886046302e-05,
+      "loss": 0.1891,
+      "step": 86270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.160526078731646e-05,
+      "loss": 0.1894,
+      "step": 86280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1600756688586614e-05,
+      "loss": 0.1843,
+      "step": 86290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1596252589856771e-05,
+      "loss": 0.1879,
+      "step": 86300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1591748491126926e-05,
+      "loss": 0.1857,
+      "step": 86310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1587244392397081e-05,
+      "loss": 0.1898,
+      "step": 86320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1582740293667239e-05,
+      "loss": 0.1884,
+      "step": 86330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1578236194937393e-05,
+      "loss": 0.1904,
+      "step": 86340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.157373209620755e-05,
+      "loss": 0.1882,
+      "step": 86350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1569227997477704e-05,
+      "loss": 0.1874,
+      "step": 86360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1564723898747862e-05,
+      "loss": 0.1919,
+      "step": 86370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1560219800018018e-05,
+      "loss": 0.19,
+      "step": 86380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1555715701288172e-05,
+      "loss": 0.1815,
+      "step": 86390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.155121160255833e-05,
+      "loss": 0.1833,
+      "step": 86400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1546707503828483e-05,
+      "loss": 0.1894,
+      "step": 86410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1542203405098641e-05,
+      "loss": 0.1907,
+      "step": 86420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1537699306368797e-05,
+      "loss": 0.1888,
+      "step": 86430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1533195207638953e-05,
+      "loss": 0.1886,
+      "step": 86440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1528691108909108e-05,
+      "loss": 0.1931,
+      "step": 86450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1524187010179262e-05,
+      "loss": 0.1959,
+      "step": 86460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.151968291144942e-05,
+      "loss": 0.1876,
+      "step": 86470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1515178812719576e-05,
+      "loss": 0.1856,
+      "step": 86480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1510674713989732e-05,
+      "loss": 0.189,
+      "step": 86490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1506170615259887e-05,
+      "loss": 0.1944,
+      "step": 86500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1501666516530043e-05,
+      "loss": 0.1909,
+      "step": 86510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1497162417800199e-05,
+      "loss": 0.1926,
+      "step": 86520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1492658319070355e-05,
+      "loss": 0.1861,
+      "step": 86530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.148815422034051e-05,
+      "loss": 0.1929,
+      "step": 86540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1483650121610666e-05,
+      "loss": 0.1835,
+      "step": 86550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1479146022880822e-05,
+      "loss": 0.1879,
+      "step": 86560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1474641924150978e-05,
+      "loss": 0.1857,
+      "step": 86570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1470137825421134e-05,
+      "loss": 0.1841,
+      "step": 86580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.146563372669129e-05,
+      "loss": 0.1905,
+      "step": 86590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1461129627961445e-05,
+      "loss": 0.1899,
+      "step": 86600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1456625529231601e-05,
+      "loss": 0.1851,
+      "step": 86610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1452121430501757e-05,
+      "loss": 0.187,
+      "step": 86620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1447617331771913e-05,
+      "loss": 0.1914,
+      "step": 86630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1443113233042069e-05,
+      "loss": 0.1844,
+      "step": 86640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1438609134312226e-05,
+      "loss": 0.1859,
+      "step": 86650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.143410503558238e-05,
+      "loss": 0.188,
+      "step": 86660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1429600936852536e-05,
+      "loss": 0.1959,
+      "step": 86670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1425096838122692e-05,
+      "loss": 0.1881,
+      "step": 86680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1420592739392848e-05,
+      "loss": 0.1851,
+      "step": 86690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1416088640663005e-05,
+      "loss": 0.1825,
+      "step": 86700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1411584541933159e-05,
+      "loss": 0.1888,
+      "step": 86710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1407080443203315e-05,
+      "loss": 0.1909,
+      "step": 86720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.140257634447347e-05,
+      "loss": 0.186,
+      "step": 86730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1398072245743627e-05,
+      "loss": 0.1913,
+      "step": 86740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1393568147013784e-05,
+      "loss": 0.1935,
+      "step": 86750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1389064048283938e-05,
+      "loss": 0.1962,
+      "step": 86760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1384559949554096e-05,
+      "loss": 0.1928,
+      "step": 86770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.138005585082425e-05,
+      "loss": 0.1874,
+      "step": 86780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1375551752094406e-05,
+      "loss": 0.1865,
+      "step": 86790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1371047653364563e-05,
+      "loss": 0.1965,
+      "step": 86800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1366543554634717e-05,
+      "loss": 0.1907,
+      "step": 86810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1362039455904875e-05,
+      "loss": 0.186,
+      "step": 86820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1357535357175029e-05,
+      "loss": 0.1925,
+      "step": 86830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1353031258445186e-05,
+      "loss": 0.1881,
+      "step": 86840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1348527159715342e-05,
+      "loss": 0.1885,
+      "step": 86850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1344023060985496e-05,
+      "loss": 0.1869,
+      "step": 86860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1339518962255654e-05,
+      "loss": 0.1932,
+      "step": 86870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.133501486352581e-05,
+      "loss": 0.1869,
+      "step": 86880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1330510764795965e-05,
+      "loss": 0.191,
+      "step": 86890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1326006666066121e-05,
+      "loss": 0.1876,
+      "step": 86900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1321502567336277e-05,
+      "loss": 0.1898,
+      "step": 86910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1316998468606433e-05,
+      "loss": 0.1902,
+      "step": 86920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1312494369876588e-05,
+      "loss": 0.1812,
+      "step": 86930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1307990271146744e-05,
+      "loss": 0.1944,
+      "step": 86940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.13034861724169e-05,
+      "loss": 0.1898,
+      "step": 86950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1298982073687056e-05,
+      "loss": 0.1842,
+      "step": 86960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1294477974957212e-05,
+      "loss": 0.1923,
+      "step": 86970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1289973876227367e-05,
+      "loss": 0.1874,
+      "step": 86980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1285469777497523e-05,
+      "loss": 0.1911,
+      "step": 86990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1280965678767679e-05,
+      "loss": 0.1908,
+      "step": 87000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1276461580037835e-05,
+      "loss": 0.185,
+      "step": 87010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.127195748130799e-05,
+      "loss": 0.1904,
+      "step": 87020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1267453382578146e-05,
+      "loss": 0.1913,
+      "step": 87030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1262949283848302e-05,
+      "loss": 0.1931,
+      "step": 87040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.125844518511846e-05,
+      "loss": 0.1905,
+      "step": 87050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1253941086388614e-05,
+      "loss": 0.1876,
+      "step": 87060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.124943698765877e-05,
+      "loss": 0.187,
+      "step": 87070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1244932888928925e-05,
+      "loss": 0.1922,
+      "step": 87080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1240428790199081e-05,
+      "loss": 0.1898,
+      "step": 87090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1235924691469239e-05,
+      "loss": 0.1907,
+      "step": 87100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1231420592739393e-05,
+      "loss": 0.1834,
+      "step": 87110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.122691649400955e-05,
+      "loss": 0.1866,
+      "step": 87120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1222412395279704e-05,
+      "loss": 0.1874,
+      "step": 87130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.121790829654986e-05,
+      "loss": 0.1924,
+      "step": 87140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1213404197820018e-05,
+      "loss": 0.1904,
+      "step": 87150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1208900099090172e-05,
+      "loss": 0.1922,
+      "step": 87160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.120439600036033e-05,
+      "loss": 0.1941,
+      "step": 87170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1199891901630483e-05,
+      "loss": 0.1835,
+      "step": 87180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.119538780290064e-05,
+      "loss": 0.1879,
+      "step": 87190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1190883704170797e-05,
+      "loss": 0.1869,
+      "step": 87200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.118637960544095e-05,
+      "loss": 0.1906,
+      "step": 87210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1181875506711108e-05,
+      "loss": 0.19,
+      "step": 87220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1177371407981262e-05,
+      "loss": 0.189,
+      "step": 87230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.117286730925142e-05,
+      "loss": 0.1873,
+      "step": 87240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1168363210521576e-05,
+      "loss": 0.1876,
+      "step": 87250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1163859111791731e-05,
+      "loss": 0.1892,
+      "step": 87260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1159355013061887e-05,
+      "loss": 0.1882,
+      "step": 87270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1154850914332041e-05,
+      "loss": 0.1884,
+      "step": 87280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1150346815602199e-05,
+      "loss": 0.1926,
+      "step": 87290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1145842716872355e-05,
+      "loss": 0.19,
+      "step": 87300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.114133861814251e-05,
+      "loss": 0.1841,
+      "step": 87310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1136834519412666e-05,
+      "loss": 0.1899,
+      "step": 87320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1132330420682822e-05,
+      "loss": 0.1914,
+      "step": 87330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1127826321952978e-05,
+      "loss": 0.1971,
+      "step": 87340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1123322223223134e-05,
+      "loss": 0.181,
+      "step": 87350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.111881812449329e-05,
+      "loss": 0.1813,
+      "step": 87360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1114314025763445e-05,
+      "loss": 0.1887,
+      "step": 87370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1109809927033601e-05,
+      "loss": 0.1785,
+      "step": 87380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1105305828303757e-05,
+      "loss": 0.1889,
+      "step": 87390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1100801729573913e-05,
+      "loss": 0.1863,
+      "step": 87400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1096297630844068e-05,
+      "loss": 0.1852,
+      "step": 87410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1091793532114224e-05,
+      "loss": 0.185,
+      "step": 87420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.108728943338438e-05,
+      "loss": 0.1902,
+      "step": 87430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1082785334654536e-05,
+      "loss": 0.1832,
+      "step": 87440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1078281235924692e-05,
+      "loss": 0.1821,
+      "step": 87450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1073777137194847e-05,
+      "loss": 0.1887,
+      "step": 87460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1069273038465005e-05,
+      "loss": 0.1897,
+      "step": 87470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1064768939735159e-05,
+      "loss": 0.1907,
+      "step": 87480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1060264841005315e-05,
+      "loss": 0.1849,
+      "step": 87490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.105576074227547e-05,
+      "loss": 0.1905,
+      "step": 87500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1051256643545626e-05,
+      "loss": 0.1808,
+      "step": 87510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1046752544815784e-05,
+      "loss": 0.188,
+      "step": 87520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1042248446085938e-05,
+      "loss": 0.1844,
+      "step": 87530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1037744347356095e-05,
+      "loss": 0.1867,
+      "step": 87540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.103324024862625e-05,
+      "loss": 0.1872,
+      "step": 87550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1028736149896405e-05,
+      "loss": 0.19,
+      "step": 87560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1024232051166563e-05,
+      "loss": 0.1868,
+      "step": 87570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1019727952436717e-05,
+      "loss": 0.1909,
+      "step": 87580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1015223853706874e-05,
+      "loss": 0.192,
+      "step": 87590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.101071975497703e-05,
+      "loss": 0.1842,
+      "step": 87600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1006215656247186e-05,
+      "loss": 0.1892,
+      "step": 87610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1001711557517342e-05,
+      "loss": 0.1874,
+      "step": 87620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0997207458787496e-05,
+      "loss": 0.1825,
+      "step": 87630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0992703360057653e-05,
+      "loss": 0.1894,
+      "step": 87640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.098819926132781e-05,
+      "loss": 0.1838,
+      "step": 87650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0983695162597965e-05,
+      "loss": 0.188,
+      "step": 87660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.097919106386812e-05,
+      "loss": 0.1872,
+      "step": 87670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0974686965138277e-05,
+      "loss": 0.1812,
+      "step": 87680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0970182866408432e-05,
+      "loss": 0.1916,
+      "step": 87690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0965678767678588e-05,
+      "loss": 0.1832,
+      "step": 87700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0961174668948744e-05,
+      "loss": 0.1888,
+      "step": 87710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.09566705702189e-05,
+      "loss": 0.1925,
+      "step": 87720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0952166471489056e-05,
+      "loss": 0.1837,
+      "step": 87730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0947662372759211e-05,
+      "loss": 0.1863,
+      "step": 87740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0943158274029367e-05,
+      "loss": 0.1855,
+      "step": 87750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0938654175299523e-05,
+      "loss": 0.1866,
+      "step": 87760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0934150076569679e-05,
+      "loss": 0.1901,
+      "step": 87770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0929645977839835e-05,
+      "loss": 0.1891,
+      "step": 87780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.092514187910999e-05,
+      "loss": 0.1903,
+      "step": 87790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0920637780380146e-05,
+      "loss": 0.1927,
+      "step": 87800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0916133681650302e-05,
+      "loss": 0.1855,
+      "step": 87810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.091162958292046e-05,
+      "loss": 0.194,
+      "step": 87820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0907125484190614e-05,
+      "loss": 0.1921,
+      "step": 87830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.090262138546077e-05,
+      "loss": 0.1927,
+      "step": 87840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0898117286730925e-05,
+      "loss": 0.1918,
+      "step": 87850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0893613188001081e-05,
+      "loss": 0.1857,
+      "step": 87860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0889109089271238e-05,
+      "loss": 0.184,
+      "step": 87870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0884604990541393e-05,
+      "loss": 0.1926,
+      "step": 87880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.088010089181155e-05,
+      "loss": 0.1947,
+      "step": 87890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0875596793081704e-05,
+      "loss": 0.1851,
+      "step": 87900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.087109269435186e-05,
+      "loss": 0.191,
+      "step": 87910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0866588595622017e-05,
+      "loss": 0.1915,
+      "step": 87920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0862084496892172e-05,
+      "loss": 0.1892,
+      "step": 87930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0857580398162329e-05,
+      "loss": 0.1902,
+      "step": 87940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0853076299432483e-05,
+      "loss": 0.1925,
+      "step": 87950
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.084857220070264e-05,
+      "loss": 0.1859,
+      "step": 87960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0844068101972796e-05,
+      "loss": 0.1869,
+      "step": 87970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.083956400324295e-05,
+      "loss": 0.1937,
+      "step": 87980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0835059904513108e-05,
+      "loss": 0.189,
+      "step": 87990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0830555805783262e-05,
+      "loss": 0.1923,
+      "step": 88000
+    },
+    {
+      "epoch": 0.79,
+      "eval_NEIMS_canon_smiles": 0.9387333333333333,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.05748333333333333,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.46029620466408705,
+      "eval_NEIMS_exact_mols": 0.05728333333333333,
+      "eval_NEIMS_exact_smiles": 0.05525,
+      "eval_NEIMS_loss": 0.21825504302978516,
+      "eval_NEIMS_matched_formulas": 0.5685166666666667,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.36367257426355964,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.05765,
+      "eval_NEIMS_runtime": 714.9592,
+      "eval_NEIMS_samples_per_second": 83.921,
+      "eval_NEIMS_steps_per_second": 1.312,
+      "step": 88000
+    },
+    {
+      "epoch": 0.79,
+      "eval_RASSP_canon_smiles": 0.934050155274318,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.13558954152335792,
+      "eval_RASSP_daylight_tanimoto_simil": 0.5863277874216708,
+      "eval_RASSP_exact_mols": 0.13492169499449028,
+      "eval_RASSP_exact_smiles": 0.130063111496978,
+      "eval_RASSP_loss": 0.15727342665195465,
+      "eval_RASSP_matched_formulas": 0.7488897051457575,
+      "eval_RASSP_morgan_tanimoto_simil": 0.45645551107265253,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.1350719604634855,
+      "eval_RASSP_runtime": 816.8086,
+      "eval_RASSP_samples_per_second": 73.327,
+      "eval_RASSP_steps_per_second": 1.146,
+      "step": 88000
+    },
+    {
+      "epoch": 0.79,
+      "eval_NIST_canon_smiles": 0.882315363594421,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.013166767221492707,
+      "eval_NIST_daylight_tanimoto_simil": 0.27067425568062337,
+      "eval_NIST_exact_mols": 0.01238598857223977,
+      "eval_NIST_exact_smiles": 0.01125031053696277,
+      "eval_NIST_loss": 1.4268168210983276,
+      "eval_NIST_matched_formulas": 0.08808602761117224,
+      "eval_NIST_morgan_tanimoto_simil": 0.22409104597515142,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.013095787344287894,
+      "eval_NIST_runtime": 467.8554,
+      "eval_NIST_samples_per_second": 60.226,
+      "eval_NIST_steps_per_second": 0.943,
+      "step": 88000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.082605170705342e-05,
+      "loss": 0.1873,
+      "step": 88010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0821547608323575e-05,
+      "loss": 0.1932,
+      "step": 88020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0817493919466714e-05,
+      "loss": 0.1881,
+      "step": 88030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0812989820736872e-05,
+      "loss": 0.1836,
+      "step": 88040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0808485722007026e-05,
+      "loss": 0.1893,
+      "step": 88050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0803981623277183e-05,
+      "loss": 0.1822,
+      "step": 88060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0799477524547339e-05,
+      "loss": 0.188,
+      "step": 88070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0794973425817495e-05,
+      "loss": 0.1882,
+      "step": 88080
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.079046932708765e-05,
+      "loss": 0.1881,
+      "step": 88090
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0785965228357805e-05,
+      "loss": 0.19,
+      "step": 88100
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0781461129627962e-05,
+      "loss": 0.1902,
+      "step": 88110
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0776957030898118e-05,
+      "loss": 0.1865,
+      "step": 88120
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0772452932168274e-05,
+      "loss": 0.1872,
+      "step": 88130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.076794883343843e-05,
+      "loss": 0.1871,
+      "step": 88140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0763444734708585e-05,
+      "loss": 0.1894,
+      "step": 88150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0758940635978741e-05,
+      "loss": 0.1849,
+      "step": 88160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0754436537248897e-05,
+      "loss": 0.1933,
+      "step": 88170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0749932438519053e-05,
+      "loss": 0.1912,
+      "step": 88180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0745428339789209e-05,
+      "loss": 0.1851,
+      "step": 88190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0740924241059364e-05,
+      "loss": 0.1855,
+      "step": 88200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.073642014232952e-05,
+      "loss": 0.1915,
+      "step": 88210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0731916043599676e-05,
+      "loss": 0.186,
+      "step": 88220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0727411944869832e-05,
+      "loss": 0.1833,
+      "step": 88230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0722907846139988e-05,
+      "loss": 0.1841,
+      "step": 88240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0718403747410143e-05,
+      "loss": 0.1814,
+      "step": 88250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.07138996486803e-05,
+      "loss": 0.191,
+      "step": 88260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0709395549950457e-05,
+      "loss": 0.193,
+      "step": 88270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.070489145122061e-05,
+      "loss": 0.1821,
+      "step": 88280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0700387352490768e-05,
+      "loss": 0.1886,
+      "step": 88290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0695883253760922e-05,
+      "loss": 0.1882,
+      "step": 88300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0691379155031078e-05,
+      "loss": 0.1903,
+      "step": 88310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0686875056301236e-05,
+      "loss": 0.1894,
+      "step": 88320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.068237095757139e-05,
+      "loss": 0.1875,
+      "step": 88330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0677866858841547e-05,
+      "loss": 0.1881,
+      "step": 88340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0673362760111701e-05,
+      "loss": 0.1844,
+      "step": 88350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0668858661381859e-05,
+      "loss": 0.193,
+      "step": 88360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0664354562652015e-05,
+      "loss": 0.1906,
+      "step": 88370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0659850463922169e-05,
+      "loss": 0.1888,
+      "step": 88380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0655346365192326e-05,
+      "loss": 0.1921,
+      "step": 88390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.065084226646248e-05,
+      "loss": 0.1871,
+      "step": 88400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0646338167732638e-05,
+      "loss": 0.1902,
+      "step": 88410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0641834069002794e-05,
+      "loss": 0.1787,
+      "step": 88420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.063732997027295e-05,
+      "loss": 0.1859,
+      "step": 88430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0632825871543105e-05,
+      "loss": 0.1867,
+      "step": 88440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.062832177281326e-05,
+      "loss": 0.1891,
+      "step": 88450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0623817674083417e-05,
+      "loss": 0.1886,
+      "step": 88460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0619313575353573e-05,
+      "loss": 0.1892,
+      "step": 88470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0614809476623728e-05,
+      "loss": 0.1949,
+      "step": 88480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0610305377893884e-05,
+      "loss": 0.1871,
+      "step": 88490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.060580127916404e-05,
+      "loss": 0.1892,
+      "step": 88500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0601297180434196e-05,
+      "loss": 0.1942,
+      "step": 88510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0596793081704352e-05,
+      "loss": 0.1884,
+      "step": 88520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0592288982974507e-05,
+      "loss": 0.1881,
+      "step": 88530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0587784884244663e-05,
+      "loss": 0.1867,
+      "step": 88540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0583280785514819e-05,
+      "loss": 0.1889,
+      "step": 88550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0578776686784975e-05,
+      "loss": 0.1883,
+      "step": 88560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.057427258805513e-05,
+      "loss": 0.1891,
+      "step": 88570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0569768489325286e-05,
+      "loss": 0.1867,
+      "step": 88580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0565264390595442e-05,
+      "loss": 0.1815,
+      "step": 88590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0560760291865598e-05,
+      "loss": 0.1942,
+      "step": 88600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0556256193135754e-05,
+      "loss": 0.1898,
+      "step": 88610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.055175209440591e-05,
+      "loss": 0.1882,
+      "step": 88620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0547247995676065e-05,
+      "loss": 0.1869,
+      "step": 88630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0542743896946223e-05,
+      "loss": 0.185,
+      "step": 88640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0538239798216377e-05,
+      "loss": 0.1825,
+      "step": 88650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0533735699486533e-05,
+      "loss": 0.1812,
+      "step": 88660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0529231600756689e-05,
+      "loss": 0.1876,
+      "step": 88670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0524727502026844e-05,
+      "loss": 0.1907,
+      "step": 88680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0520223403297002e-05,
+      "loss": 0.187,
+      "step": 88690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0515719304567156e-05,
+      "loss": 0.185,
+      "step": 88700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0511215205837313e-05,
+      "loss": 0.1884,
+      "step": 88710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0506711107107468e-05,
+      "loss": 0.1878,
+      "step": 88720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0502207008377623e-05,
+      "loss": 0.1895,
+      "step": 88730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0497702909647781e-05,
+      "loss": 0.1822,
+      "step": 88740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0493198810917935e-05,
+      "loss": 0.1888,
+      "step": 88750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0488694712188092e-05,
+      "loss": 0.1855,
+      "step": 88760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0484190613458247e-05,
+      "loss": 0.1955,
+      "step": 88770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0479686514728404e-05,
+      "loss": 0.1878,
+      "step": 88780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.047518241599856e-05,
+      "loss": 0.1913,
+      "step": 88790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0470678317268714e-05,
+      "loss": 0.1877,
+      "step": 88800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0466174218538871e-05,
+      "loss": 0.1861,
+      "step": 88810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0461670119809026e-05,
+      "loss": 0.1844,
+      "step": 88820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0457166021079183e-05,
+      "loss": 0.1865,
+      "step": 88830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0452661922349339e-05,
+      "loss": 0.1832,
+      "step": 88840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0448157823619495e-05,
+      "loss": 0.191,
+      "step": 88850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.044365372488965e-05,
+      "loss": 0.1827,
+      "step": 88860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0439149626159805e-05,
+      "loss": 0.179,
+      "step": 88870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0434645527429962e-05,
+      "loss": 0.1888,
+      "step": 88880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0430141428700118e-05,
+      "loss": 0.1877,
+      "step": 88890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0425637329970274e-05,
+      "loss": 0.1825,
+      "step": 88900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.042113323124043e-05,
+      "loss": 0.1883,
+      "step": 88910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0416629132510585e-05,
+      "loss": 0.1801,
+      "step": 88920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0412125033780741e-05,
+      "loss": 0.1891,
+      "step": 88930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0407620935050897e-05,
+      "loss": 0.1947,
+      "step": 88940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0403116836321053e-05,
+      "loss": 0.1834,
+      "step": 88950
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0398612737591208e-05,
+      "loss": 0.1854,
+      "step": 88960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0394108638861364e-05,
+      "loss": 0.1869,
+      "step": 88970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.038960454013152e-05,
+      "loss": 0.1883,
+      "step": 88980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0385100441401678e-05,
+      "loss": 0.1878,
+      "step": 88990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0380596342671832e-05,
+      "loss": 0.1861,
+      "step": 89000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0376092243941987e-05,
+      "loss": 0.1891,
+      "step": 89010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0371588145212143e-05,
+      "loss": 0.1872,
+      "step": 89020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0367084046482299e-05,
+      "loss": 0.1934,
+      "step": 89030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0362579947752457e-05,
+      "loss": 0.1885,
+      "step": 89040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.035807584902261e-05,
+      "loss": 0.1863,
+      "step": 89050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0353571750292768e-05,
+      "loss": 0.191,
+      "step": 89060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0349067651562922e-05,
+      "loss": 0.1886,
+      "step": 89070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0344563552833078e-05,
+      "loss": 0.1877,
+      "step": 89080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0340059454103236e-05,
+      "loss": 0.1849,
+      "step": 89090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033555535537339e-05,
+      "loss": 0.1886,
+      "step": 89100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0331051256643547e-05,
+      "loss": 0.1827,
+      "step": 89110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0326547157913701e-05,
+      "loss": 0.1885,
+      "step": 89120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0322043059183857e-05,
+      "loss": 0.1921,
+      "step": 89130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0317538960454015e-05,
+      "loss": 0.1792,
+      "step": 89140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0313034861724169e-05,
+      "loss": 0.1806,
+      "step": 89150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0308530762994326e-05,
+      "loss": 0.183,
+      "step": 89160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.030402666426448e-05,
+      "loss": 0.1861,
+      "step": 89170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0299522565534638e-05,
+      "loss": 0.1908,
+      "step": 89180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0295018466804794e-05,
+      "loss": 0.192,
+      "step": 89190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0290514368074948e-05,
+      "loss": 0.1893,
+      "step": 89200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0286010269345105e-05,
+      "loss": 0.1912,
+      "step": 89210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.028150617061526e-05,
+      "loss": 0.188,
+      "step": 89220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0277002071885417e-05,
+      "loss": 0.1816,
+      "step": 89230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0272497973155572e-05,
+      "loss": 0.1808,
+      "step": 89240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0267993874425728e-05,
+      "loss": 0.186,
+      "step": 89250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0263489775695884e-05,
+      "loss": 0.1952,
+      "step": 89260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0258985676966038e-05,
+      "loss": 0.1829,
+      "step": 89270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0254481578236196e-05,
+      "loss": 0.1875,
+      "step": 89280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0249977479506351e-05,
+      "loss": 0.1863,
+      "step": 89290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0245473380776507e-05,
+      "loss": 0.1899,
+      "step": 89300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0240969282046663e-05,
+      "loss": 0.1897,
+      "step": 89310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0236465183316819e-05,
+      "loss": 0.1886,
+      "step": 89320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0231961084586975e-05,
+      "loss": 0.1908,
+      "step": 89330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.022745698585713e-05,
+      "loss": 0.1857,
+      "step": 89340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0222952887127286e-05,
+      "loss": 0.1862,
+      "step": 89350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0218448788397442e-05,
+      "loss": 0.1841,
+      "step": 89360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0213944689667598e-05,
+      "loss": 0.187,
+      "step": 89370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0209440590937754e-05,
+      "loss": 0.1875,
+      "step": 89380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.020493649220791e-05,
+      "loss": 0.191,
+      "step": 89390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0200432393478065e-05,
+      "loss": 0.1869,
+      "step": 89400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0195928294748221e-05,
+      "loss": 0.1847,
+      "step": 89410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0191424196018377e-05,
+      "loss": 0.187,
+      "step": 89420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0186920097288533e-05,
+      "loss": 0.1888,
+      "step": 89430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0182415998558688e-05,
+      "loss": 0.1874,
+      "step": 89440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0177911899828844e-05,
+      "loss": 0.1876,
+      "step": 89450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0173407801099002e-05,
+      "loss": 0.1887,
+      "step": 89460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0168903702369156e-05,
+      "loss": 0.1885,
+      "step": 89470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0164399603639312e-05,
+      "loss": 0.1868,
+      "step": 89480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0159895504909467e-05,
+      "loss": 0.1945,
+      "step": 89490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0155391406179623e-05,
+      "loss": 0.1846,
+      "step": 89500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.015088730744978e-05,
+      "loss": 0.1838,
+      "step": 89510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0146383208719935e-05,
+      "loss": 0.1838,
+      "step": 89520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0141879109990092e-05,
+      "loss": 0.189,
+      "step": 89530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0137375011260246e-05,
+      "loss": 0.1882,
+      "step": 89540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0132870912530402e-05,
+      "loss": 0.185,
+      "step": 89550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.012836681380056e-05,
+      "loss": 0.187,
+      "step": 89560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0123862715070714e-05,
+      "loss": 0.1907,
+      "step": 89570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0119358616340871e-05,
+      "loss": 0.1836,
+      "step": 89580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0114854517611025e-05,
+      "loss": 0.1846,
+      "step": 89590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0110350418881183e-05,
+      "loss": 0.1912,
+      "step": 89600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0105846320151339e-05,
+      "loss": 0.1822,
+      "step": 89610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0101342221421493e-05,
+      "loss": 0.1891,
+      "step": 89620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.009683812269165e-05,
+      "loss": 0.1885,
+      "step": 89630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0092334023961806e-05,
+      "loss": 0.1881,
+      "step": 89640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0087829925231962e-05,
+      "loss": 0.1851,
+      "step": 89650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0083325826502118e-05,
+      "loss": 0.185,
+      "step": 89660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0078821727772274e-05,
+      "loss": 0.1907,
+      "step": 89670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.007431762904243e-05,
+      "loss": 0.1847,
+      "step": 89680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0069813530312585e-05,
+      "loss": 0.1849,
+      "step": 89690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0065309431582741e-05,
+      "loss": 0.1811,
+      "step": 89700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0060805332852897e-05,
+      "loss": 0.1882,
+      "step": 89710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0056301234123052e-05,
+      "loss": 0.1899,
+      "step": 89720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0051797135393208e-05,
+      "loss": 0.1861,
+      "step": 89730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0047293036663364e-05,
+      "loss": 0.1911,
+      "step": 89740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.004278893793352e-05,
+      "loss": 0.1835,
+      "step": 89750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0038284839203676e-05,
+      "loss": 0.1828,
+      "step": 89760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0033780740473831e-05,
+      "loss": 0.1781,
+      "step": 89770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0029276641743987e-05,
+      "loss": 0.1847,
+      "step": 89780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0024772543014143e-05,
+      "loss": 0.187,
+      "step": 89790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0020268444284299e-05,
+      "loss": 0.1815,
+      "step": 89800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0015764345554456e-05,
+      "loss": 0.187,
+      "step": 89810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.001126024682461e-05,
+      "loss": 0.1839,
+      "step": 89820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0006756148094766e-05,
+      "loss": 0.1851,
+      "step": 89830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0002252049364922e-05,
+      "loss": 0.1881,
+      "step": 89840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.997747950635078e-06,
+      "loss": 0.1877,
+      "step": 89850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.993243851905235e-06,
+      "loss": 0.1888,
+      "step": 89860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.98873975317539e-06,
+      "loss": 0.1886,
+      "step": 89870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.984235654445547e-06,
+      "loss": 0.1863,
+      "step": 89880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.979731555715701e-06,
+      "loss": 0.1812,
+      "step": 89890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.975227456985857e-06,
+      "loss": 0.1889,
+      "step": 89900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.970723358256014e-06,
+      "loss": 0.1864,
+      "step": 89910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.966219259526168e-06,
+      "loss": 0.1858,
+      "step": 89920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.961715160796326e-06,
+      "loss": 0.1842,
+      "step": 89930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.95721106206648e-06,
+      "loss": 0.1821,
+      "step": 89940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.952706963336638e-06,
+      "loss": 0.1871,
+      "step": 89950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.948202864606793e-06,
+      "loss": 0.1897,
+      "step": 89960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.943698765876947e-06,
+      "loss": 0.1877,
+      "step": 89970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.939194667147105e-06,
+      "loss": 0.1915,
+      "step": 89980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.934690568417259e-06,
+      "loss": 0.1851,
+      "step": 89990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.930186469687417e-06,
+      "loss": 0.1868,
+      "step": 90000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.925682370957572e-06,
+      "loss": 0.1856,
+      "step": 90010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.921178272227728e-06,
+      "loss": 0.183,
+      "step": 90020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.916674173497884e-06,
+      "loss": 0.1887,
+      "step": 90030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.912620484641023e-06,
+      "loss": 0.1919,
+      "step": 90040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.90811638591118e-06,
+      "loss": 0.1873,
+      "step": 90050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.903612287181336e-06,
+      "loss": 0.1842,
+      "step": 90060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.899108188451492e-06,
+      "loss": 0.1881,
+      "step": 90070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.894604089721648e-06,
+      "loss": 0.1931,
+      "step": 90080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.890099990991802e-06,
+      "loss": 0.1833,
+      "step": 90090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.885595892261959e-06,
+      "loss": 0.186,
+      "step": 90100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.881091793532115e-06,
+      "loss": 0.1874,
+      "step": 90110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.87658769480227e-06,
+      "loss": 0.1831,
+      "step": 90120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.872083596072427e-06,
+      "loss": 0.1824,
+      "step": 90130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.867579497342582e-06,
+      "loss": 0.1784,
+      "step": 90140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.863075398612738e-06,
+      "loss": 0.1826,
+      "step": 90150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.858571299882894e-06,
+      "loss": 0.1846,
+      "step": 90160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.85406720115305e-06,
+      "loss": 0.183,
+      "step": 90170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.849563102423205e-06,
+      "loss": 0.1772,
+      "step": 90180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.845059003693361e-06,
+      "loss": 0.1857,
+      "step": 90190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.840554904963517e-06,
+      "loss": 0.186,
+      "step": 90200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.836050806233673e-06,
+      "loss": 0.1863,
+      "step": 90210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.831546707503829e-06,
+      "loss": 0.1858,
+      "step": 90220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.827042608773984e-06,
+      "loss": 0.183,
+      "step": 90230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.82253851004414e-06,
+      "loss": 0.1874,
+      "step": 90240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.818034411314296e-06,
+      "loss": 0.185,
+      "step": 90250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.813530312584452e-06,
+      "loss": 0.1894,
+      "step": 90260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.809026213854608e-06,
+      "loss": 0.1919,
+      "step": 90270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.804522115124765e-06,
+      "loss": 0.1901,
+      "step": 90280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.80001801639492e-06,
+      "loss": 0.1872,
+      "step": 90290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.795513917665075e-06,
+      "loss": 0.1872,
+      "step": 90300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.791009818935233e-06,
+      "loss": 0.1879,
+      "step": 90310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.786505720205387e-06,
+      "loss": 0.1819,
+      "step": 90320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.782001621475544e-06,
+      "loss": 0.1835,
+      "step": 90330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.777497522745698e-06,
+      "loss": 0.1942,
+      "step": 90340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.772993424015856e-06,
+      "loss": 0.1884,
+      "step": 90350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.768489325286012e-06,
+      "loss": 0.1846,
+      "step": 90360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.763985226556166e-06,
+      "loss": 0.1907,
+      "step": 90370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.759481127826323e-06,
+      "loss": 0.1834,
+      "step": 90380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.754977029096477e-06,
+      "loss": 0.1865,
+      "step": 90390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.750472930366635e-06,
+      "loss": 0.1875,
+      "step": 90400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.74596883163679e-06,
+      "loss": 0.183,
+      "step": 90410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.741464732906946e-06,
+      "loss": 0.1905,
+      "step": 90420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.736960634177102e-06,
+      "loss": 0.1812,
+      "step": 90430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.732456535447256e-06,
+      "loss": 0.1863,
+      "step": 90440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.727952436717414e-06,
+      "loss": 0.1892,
+      "step": 90450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.72344833798757e-06,
+      "loss": 0.1844,
+      "step": 90460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.718944239257725e-06,
+      "loss": 0.1916,
+      "step": 90470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.714440140527881e-06,
+      "loss": 0.1823,
+      "step": 90480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.709936041798037e-06,
+      "loss": 0.1833,
+      "step": 90490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.705431943068193e-06,
+      "loss": 0.1852,
+      "step": 90500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.700927844338349e-06,
+      "loss": 0.1808,
+      "step": 90510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.696423745608504e-06,
+      "loss": 0.1836,
+      "step": 90520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.69191964687866e-06,
+      "loss": 0.1819,
+      "step": 90530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.687415548148816e-06,
+      "loss": 0.1838,
+      "step": 90540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.682911449418972e-06,
+      "loss": 0.1893,
+      "step": 90550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.678407350689128e-06,
+      "loss": 0.183,
+      "step": 90560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.673903251959283e-06,
+      "loss": 0.1858,
+      "step": 90570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.669399153229439e-06,
+      "loss": 0.1843,
+      "step": 90580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.664895054499595e-06,
+      "loss": 0.1898,
+      "step": 90590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.66039095576975e-06,
+      "loss": 0.1806,
+      "step": 90600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.655886857039907e-06,
+      "loss": 0.1906,
+      "step": 90610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.651382758310062e-06,
+      "loss": 0.1854,
+      "step": 90620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.64687865958022e-06,
+      "loss": 0.1795,
+      "step": 90630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.642374560850374e-06,
+      "loss": 0.1853,
+      "step": 90640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.63787046212053e-06,
+      "loss": 0.1858,
+      "step": 90650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.633366363390685e-06,
+      "loss": 0.1865,
+      "step": 90660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.628862264660841e-06,
+      "loss": 0.18,
+      "step": 90670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.624358165930999e-06,
+      "loss": 0.1812,
+      "step": 90680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.619854067201153e-06,
+      "loss": 0.1917,
+      "step": 90690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.61534996847131e-06,
+      "loss": 0.1863,
+      "step": 90700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.610845869741464e-06,
+      "loss": 0.1844,
+      "step": 90710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.60634177101162e-06,
+      "loss": 0.1847,
+      "step": 90720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.601837672281778e-06,
+      "loss": 0.1858,
+      "step": 90730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.597333573551932e-06,
+      "loss": 0.1825,
+      "step": 90740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.59282947482209e-06,
+      "loss": 0.1853,
+      "step": 90750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.588325376092243e-06,
+      "loss": 0.1864,
+      "step": 90760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.583821277362401e-06,
+      "loss": 0.1911,
+      "step": 90770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.579317178632557e-06,
+      "loss": 0.1859,
+      "step": 90780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.574813079902711e-06,
+      "loss": 0.1797,
+      "step": 90790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.570308981172868e-06,
+      "loss": 0.1891,
+      "step": 90800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.565804882443022e-06,
+      "loss": 0.1823,
+      "step": 90810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.56130078371318e-06,
+      "loss": 0.1834,
+      "step": 90820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.556796684983336e-06,
+      "loss": 0.1833,
+      "step": 90830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.552292586253492e-06,
+      "loss": 0.1817,
+      "step": 90840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.547788487523647e-06,
+      "loss": 0.1871,
+      "step": 90850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.543284388793801e-06,
+      "loss": 0.1898,
+      "step": 90860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.538780290063959e-06,
+      "loss": 0.1814,
+      "step": 90870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.534276191334115e-06,
+      "loss": 0.1795,
+      "step": 90880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.52977209260427e-06,
+      "loss": 0.1847,
+      "step": 90890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.525267993874426e-06,
+      "loss": 0.1803,
+      "step": 90900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.520763895144582e-06,
+      "loss": 0.182,
+      "step": 90910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.516259796414738e-06,
+      "loss": 0.1844,
+      "step": 90920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.511755697684894e-06,
+      "loss": 0.1868,
+      "step": 90930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.50725159895505e-06,
+      "loss": 0.1835,
+      "step": 90940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.502747500225205e-06,
+      "loss": 0.1835,
+      "step": 90950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.498243401495361e-06,
+      "loss": 0.1848,
+      "step": 90960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.493739302765517e-06,
+      "loss": 0.1917,
+      "step": 90970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.489235204035673e-06,
+      "loss": 0.1835,
+      "step": 90980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.484731105305829e-06,
+      "loss": 0.1874,
+      "step": 90990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.480227006575984e-06,
+      "loss": 0.1881,
+      "step": 91000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.47572290784614e-06,
+      "loss": 0.1849,
+      "step": 91010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.471218809116296e-06,
+      "loss": 0.1859,
+      "step": 91020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.466714710386453e-06,
+      "loss": 0.1846,
+      "step": 91030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.462210611656608e-06,
+      "loss": 0.1828,
+      "step": 91040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.457706512926765e-06,
+      "loss": 0.189,
+      "step": 91050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.453202414196919e-06,
+      "loss": 0.1787,
+      "step": 91060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.448698315467075e-06,
+      "loss": 0.182,
+      "step": 91070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.444194216737232e-06,
+      "loss": 0.1871,
+      "step": 91080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.439690118007387e-06,
+      "loss": 0.1862,
+      "step": 91090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.435186019277544e-06,
+      "loss": 0.1868,
+      "step": 91100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.430681920547698e-06,
+      "loss": 0.1818,
+      "step": 91110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.426177821817856e-06,
+      "loss": 0.1892,
+      "step": 91120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.421673723088011e-06,
+      "loss": 0.1899,
+      "step": 91130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.417169624358165e-06,
+      "loss": 0.1829,
+      "step": 91140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.412665525628323e-06,
+      "loss": 0.1895,
+      "step": 91150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.408161426898477e-06,
+      "loss": 0.1868,
+      "step": 91160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.403657328168635e-06,
+      "loss": 0.1876,
+      "step": 91170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.39915322943879e-06,
+      "loss": 0.1826,
+      "step": 91180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.394649130708946e-06,
+      "loss": 0.184,
+      "step": 91190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.390145031979102e-06,
+      "loss": 0.1873,
+      "step": 91200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.385640933249256e-06,
+      "loss": 0.1847,
+      "step": 91210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.381136834519414e-06,
+      "loss": 0.1918,
+      "step": 91220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.37663273578957e-06,
+      "loss": 0.1906,
+      "step": 91230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.372128637059725e-06,
+      "loss": 0.1884,
+      "step": 91240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.367624538329881e-06,
+      "loss": 0.1855,
+      "step": 91250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.363120439600037e-06,
+      "loss": 0.1892,
+      "step": 91260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.358616340870193e-06,
+      "loss": 0.1813,
+      "step": 91270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.354112242140348e-06,
+      "loss": 0.182,
+      "step": 91280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.349608143410504e-06,
+      "loss": 0.1852,
+      "step": 91290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.34510404468066e-06,
+      "loss": 0.188,
+      "step": 91300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.340599945950816e-06,
+      "loss": 0.1895,
+      "step": 91310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.336095847220972e-06,
+      "loss": 0.1943,
+      "step": 91320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.331591748491127e-06,
+      "loss": 0.1813,
+      "step": 91330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.327087649761283e-06,
+      "loss": 0.1852,
+      "step": 91340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.322583551031439e-06,
+      "loss": 0.1876,
+      "step": 91350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.318079452301595e-06,
+      "loss": 0.1783,
+      "step": 91360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.31357535357175e-06,
+      "loss": 0.1795,
+      "step": 91370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.309071254841906e-06,
+      "loss": 0.186,
+      "step": 91380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.304567156112062e-06,
+      "loss": 0.1832,
+      "step": 91390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.30006305738222e-06,
+      "loss": 0.1815,
+      "step": 91400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.295558958652374e-06,
+      "loss": 0.1855,
+      "step": 91410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.29105485992253e-06,
+      "loss": 0.182,
+      "step": 91420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.286550761192685e-06,
+      "loss": 0.1834,
+      "step": 91430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.282046662462841e-06,
+      "loss": 0.1786,
+      "step": 91440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.277542563732999e-06,
+      "loss": 0.1826,
+      "step": 91450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.273038465003153e-06,
+      "loss": 0.1843,
+      "step": 91460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.26853436627331e-06,
+      "loss": 0.1836,
+      "step": 91470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.264030267543464e-06,
+      "loss": 0.1786,
+      "step": 91480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.25952616881362e-06,
+      "loss": 0.1842,
+      "step": 91490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.255022070083778e-06,
+      "loss": 0.1866,
+      "step": 91500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.250517971353932e-06,
+      "loss": 0.1868,
+      "step": 91510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.24601387262409e-06,
+      "loss": 0.1843,
+      "step": 91520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.241509773894243e-06,
+      "loss": 0.1836,
+      "step": 91530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.237005675164399e-06,
+      "loss": 0.1813,
+      "step": 91540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.232501576434557e-06,
+      "loss": 0.1812,
+      "step": 91550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.22799747770471e-06,
+      "loss": 0.1835,
+      "step": 91560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.223493378974868e-06,
+      "loss": 0.1886,
+      "step": 91570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.218989280245022e-06,
+      "loss": 0.175,
+      "step": 91580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.21448518151518e-06,
+      "loss": 0.181,
+      "step": 91590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.209981082785336e-06,
+      "loss": 0.1744,
+      "step": 91600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.20547698405549e-06,
+      "loss": 0.18,
+      "step": 91610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.200972885325647e-06,
+      "loss": 0.18,
+      "step": 91620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.196468786595801e-06,
+      "loss": 0.1859,
+      "step": 91630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.191964687865959e-06,
+      "loss": 0.1845,
+      "step": 91640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.187460589136115e-06,
+      "loss": 0.1883,
+      "step": 91650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.18295649040627e-06,
+      "loss": 0.1886,
+      "step": 91660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.178452391676426e-06,
+      "loss": 0.1855,
+      "step": 91670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.173948292946582e-06,
+      "loss": 0.1876,
+      "step": 91680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.169444194216738e-06,
+      "loss": 0.1829,
+      "step": 91690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.164940095486894e-06,
+      "loss": 0.1818,
+      "step": 91700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.16043599675705e-06,
+      "loss": 0.1838,
+      "step": 91710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.155931898027205e-06,
+      "loss": 0.1836,
+      "step": 91720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.151427799297361e-06,
+      "loss": 0.1863,
+      "step": 91730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.146923700567517e-06,
+      "loss": 0.191,
+      "step": 91740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.142419601837673e-06,
+      "loss": 0.18,
+      "step": 91750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.137915503107828e-06,
+      "loss": 0.1856,
+      "step": 91760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.133411404377984e-06,
+      "loss": 0.1872,
+      "step": 91770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.12890730564814e-06,
+      "loss": 0.1804,
+      "step": 91780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.124403206918296e-06,
+      "loss": 0.1842,
+      "step": 91790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.119899108188453e-06,
+      "loss": 0.1834,
+      "step": 91800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.115395009458607e-06,
+      "loss": 0.1859,
+      "step": 91810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.110890910728763e-06,
+      "loss": 0.1863,
+      "step": 91820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.106386811998919e-06,
+      "loss": 0.1863,
+      "step": 91830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.101882713269075e-06,
+      "loss": 0.1885,
+      "step": 91840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.097378614539232e-06,
+      "loss": 0.1877,
+      "step": 91850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.092874515809386e-06,
+      "loss": 0.1867,
+      "step": 91860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.088370417079544e-06,
+      "loss": 0.1889,
+      "step": 91870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.083866318349698e-06,
+      "loss": 0.1877,
+      "step": 91880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.079362219619854e-06,
+      "loss": 0.1822,
+      "step": 91890
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.074858120890011e-06,
+      "loss": 0.1842,
+      "step": 91900
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.070354022160165e-06,
+      "loss": 0.1875,
+      "step": 91910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.065849923430323e-06,
+      "loss": 0.1828,
+      "step": 91920
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.061345824700477e-06,
+      "loss": 0.1886,
+      "step": 91930
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.056841725970634e-06,
+      "loss": 0.1863,
+      "step": 91940
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.05233762724079e-06,
+      "loss": 0.1826,
+      "step": 91950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.047833528510944e-06,
+      "loss": 0.1876,
+      "step": 91960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.043329429781102e-06,
+      "loss": 0.1859,
+      "step": 91970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.038825331051256e-06,
+      "loss": 0.1829,
+      "step": 91980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.034321232321413e-06,
+      "loss": 0.1816,
+      "step": 91990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.02981713359157e-06,
+      "loss": 0.1821,
+      "step": 92000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.025313034861725e-06,
+      "loss": 0.1855,
+      "step": 92010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.02080893613188e-06,
+      "loss": 0.1838,
+      "step": 92020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.016304837402035e-06,
+      "loss": 0.1864,
+      "step": 92030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.011800738672192e-06,
+      "loss": 0.1821,
+      "step": 92040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.007747049815333e-06,
+      "loss": 0.1868,
+      "step": 92050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.003242951085489e-06,
+      "loss": 0.1881,
+      "step": 92060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.998738852355644e-06,
+      "loss": 0.1794,
+      "step": 92070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.994234753625798e-06,
+      "loss": 0.1893,
+      "step": 92080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.989730654895956e-06,
+      "loss": 0.1801,
+      "step": 92090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.985226556166112e-06,
+      "loss": 0.1852,
+      "step": 92100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.980722457436268e-06,
+      "loss": 0.1827,
+      "step": 92110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.976218358706423e-06,
+      "loss": 0.1822,
+      "step": 92120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.97171425997658e-06,
+      "loss": 0.184,
+      "step": 92130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.967210161246735e-06,
+      "loss": 0.1857,
+      "step": 92140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.96270606251689e-06,
+      "loss": 0.1799,
+      "step": 92150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.958201963787047e-06,
+      "loss": 0.1818,
+      "step": 92160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.953697865057202e-06,
+      "loss": 0.1847,
+      "step": 92170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.949193766327358e-06,
+      "loss": 0.1906,
+      "step": 92180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.944689667597514e-06,
+      "loss": 0.191,
+      "step": 92190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.94018556886767e-06,
+      "loss": 0.187,
+      "step": 92200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.935681470137826e-06,
+      "loss": 0.1832,
+      "step": 92210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.931177371407981e-06,
+      "loss": 0.1851,
+      "step": 92220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.926673272678137e-06,
+      "loss": 0.1874,
+      "step": 92230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.922169173948293e-06,
+      "loss": 0.1847,
+      "step": 92240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.917665075218449e-06,
+      "loss": 0.1856,
+      "step": 92250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.913160976488605e-06,
+      "loss": 0.182,
+      "step": 92260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.908656877758762e-06,
+      "loss": 0.1862,
+      "step": 92270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.904152779028916e-06,
+      "loss": 0.189,
+      "step": 92280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.899648680299072e-06,
+      "loss": 0.18,
+      "step": 92290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.895144581569228e-06,
+      "loss": 0.1803,
+      "step": 92300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.890640482839384e-06,
+      "loss": 0.185,
+      "step": 92310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.886136384109541e-06,
+      "loss": 0.1843,
+      "step": 92320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.881632285379695e-06,
+      "loss": 0.1808,
+      "step": 92330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.877128186649853e-06,
+      "loss": 0.1811,
+      "step": 92340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.872624087920008e-06,
+      "loss": 0.1886,
+      "step": 92350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.868119989190163e-06,
+      "loss": 0.183,
+      "step": 92360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.86361589046032e-06,
+      "loss": 0.1833,
+      "step": 92370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.859111791730474e-06,
+      "loss": 0.1876,
+      "step": 92380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.854607693000632e-06,
+      "loss": 0.1865,
+      "step": 92390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.850103594270787e-06,
+      "loss": 0.1845,
+      "step": 92400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.845599495540943e-06,
+      "loss": 0.1801,
+      "step": 92410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.841095396811099e-06,
+      "loss": 0.1829,
+      "step": 92420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.836591298081253e-06,
+      "loss": 0.1819,
+      "step": 92430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.83208719935141e-06,
+      "loss": 0.1812,
+      "step": 92440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.827583100621566e-06,
+      "loss": 0.1851,
+      "step": 92450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.823079001891722e-06,
+      "loss": 0.1779,
+      "step": 92460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.818574903161878e-06,
+      "loss": 0.1855,
+      "step": 92470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.814070804432034e-06,
+      "loss": 0.1837,
+      "step": 92480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.80956670570219e-06,
+      "loss": 0.1793,
+      "step": 92490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.805062606972345e-06,
+      "loss": 0.1886,
+      "step": 92500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.800558508242501e-06,
+      "loss": 0.1811,
+      "step": 92510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.796054409512657e-06,
+      "loss": 0.1825,
+      "step": 92520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.791550310782813e-06,
+      "loss": 0.1827,
+      "step": 92530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.787046212052969e-06,
+      "loss": 0.1839,
+      "step": 92540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.782542113323124e-06,
+      "loss": 0.1839,
+      "step": 92550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.77803801459328e-06,
+      "loss": 0.1873,
+      "step": 92560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.773533915863436e-06,
+      "loss": 0.1872,
+      "step": 92570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.769029817133592e-06,
+      "loss": 0.1851,
+      "step": 92580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.764525718403748e-06,
+      "loss": 0.1828,
+      "step": 92590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.760021619673903e-06,
+      "loss": 0.1822,
+      "step": 92600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.75551752094406e-06,
+      "loss": 0.1812,
+      "step": 92610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.751013422214217e-06,
+      "loss": 0.1891,
+      "step": 92620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.74650932348437e-06,
+      "loss": 0.1839,
+      "step": 92630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.742005224754527e-06,
+      "loss": 0.1782,
+      "step": 92640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.737501126024682e-06,
+      "loss": 0.1845,
+      "step": 92650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.732997027294838e-06,
+      "loss": 0.1845,
+      "step": 92660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.728492928564996e-06,
+      "loss": 0.1818,
+      "step": 92670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.72398882983515e-06,
+      "loss": 0.1773,
+      "step": 92680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.719484731105307e-06,
+      "loss": 0.1834,
+      "step": 92690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.714980632375461e-06,
+      "loss": 0.1839,
+      "step": 92700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.710476533645617e-06,
+      "loss": 0.1868,
+      "step": 92710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.705972434915775e-06,
+      "loss": 0.1843,
+      "step": 92720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.701468336185929e-06,
+      "loss": 0.1877,
+      "step": 92730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.696964237456086e-06,
+      "loss": 0.1796,
+      "step": 92740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.69246013872624e-06,
+      "loss": 0.1794,
+      "step": 92750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.687956039996398e-06,
+      "loss": 0.1791,
+      "step": 92760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.683451941266554e-06,
+      "loss": 0.1776,
+      "step": 92770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.678947842536708e-06,
+      "loss": 0.1838,
+      "step": 92780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.674443743806865e-06,
+      "loss": 0.1814,
+      "step": 92790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.66993964507702e-06,
+      "loss": 0.184,
+      "step": 92800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.665435546347177e-06,
+      "loss": 0.181,
+      "step": 92810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.660931447617333e-06,
+      "loss": 0.1849,
+      "step": 92820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.656427348887488e-06,
+      "loss": 0.187,
+      "step": 92830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.651923250157644e-06,
+      "loss": 0.186,
+      "step": 92840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.647419151427798e-06,
+      "loss": 0.1796,
+      "step": 92850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.642915052697956e-06,
+      "loss": 0.1829,
+      "step": 92860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.638410953968112e-06,
+      "loss": 0.185,
+      "step": 92870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.633906855238267e-06,
+      "loss": 0.1766,
+      "step": 92880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.629402756508423e-06,
+      "loss": 0.1859,
+      "step": 92890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.624898657778579e-06,
+      "loss": 0.1866,
+      "step": 92900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.620394559048735e-06,
+      "loss": 0.1817,
+      "step": 92910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.61589046031889e-06,
+      "loss": 0.1829,
+      "step": 92920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.611386361589046e-06,
+      "loss": 0.1855,
+      "step": 92930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.606882262859202e-06,
+      "loss": 0.1859,
+      "step": 92940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.602378164129358e-06,
+      "loss": 0.1805,
+      "step": 92950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.597874065399514e-06,
+      "loss": 0.1878,
+      "step": 92960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.59336996666967e-06,
+      "loss": 0.1811,
+      "step": 92970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.588865867939825e-06,
+      "loss": 0.1821,
+      "step": 92980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.584361769209981e-06,
+      "loss": 0.182,
+      "step": 92990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.579857670480137e-06,
+      "loss": 0.1841,
+      "step": 93000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.575353571750293e-06,
+      "loss": 0.1848,
+      "step": 93010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.570849473020449e-06,
+      "loss": 0.1872,
+      "step": 93020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.566345374290604e-06,
+      "loss": 0.1879,
+      "step": 93030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.561841275560762e-06,
+      "loss": 0.1816,
+      "step": 93040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.557337176830916e-06,
+      "loss": 0.1836,
+      "step": 93050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.552833078101072e-06,
+      "loss": 0.1841,
+      "step": 93060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.54832897937123e-06,
+      "loss": 0.1867,
+      "step": 93070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.543824880641383e-06,
+      "loss": 0.1805,
+      "step": 93080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.539320781911541e-06,
+      "loss": 0.1785,
+      "step": 93090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.534816683181695e-06,
+      "loss": 0.1811,
+      "step": 93100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.530312584451852e-06,
+      "loss": 0.1815,
+      "step": 93110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.525808485722008e-06,
+      "loss": 0.1864,
+      "step": 93120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.521304386992162e-06,
+      "loss": 0.186,
+      "step": 93130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.51680028826232e-06,
+      "loss": 0.182,
+      "step": 93140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.512296189532474e-06,
+      "loss": 0.1837,
+      "step": 93150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.507792090802631e-06,
+      "loss": 0.1826,
+      "step": 93160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.503287992072787e-06,
+      "loss": 0.1865,
+      "step": 93170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.498783893342943e-06,
+      "loss": 0.187,
+      "step": 93180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.494279794613099e-06,
+      "loss": 0.185,
+      "step": 93190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.489775695883253e-06,
+      "loss": 0.1892,
+      "step": 93200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.48527159715341e-06,
+      "loss": 0.1801,
+      "step": 93210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.480767498423566e-06,
+      "loss": 0.1828,
+      "step": 93220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.476263399693722e-06,
+      "loss": 0.1785,
+      "step": 93230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.471759300963878e-06,
+      "loss": 0.1868,
+      "step": 93240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.467255202234034e-06,
+      "loss": 0.1862,
+      "step": 93250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.46275110350419e-06,
+      "loss": 0.1837,
+      "step": 93260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.458247004774345e-06,
+      "loss": 0.1816,
+      "step": 93270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.453742906044501e-06,
+      "loss": 0.1838,
+      "step": 93280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.449238807314657e-06,
+      "loss": 0.1854,
+      "step": 93290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.444734708584813e-06,
+      "loss": 0.1832,
+      "step": 93300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.440230609854968e-06,
+      "loss": 0.1814,
+      "step": 93310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.435726511125124e-06,
+      "loss": 0.1804,
+      "step": 93320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.43122241239528e-06,
+      "loss": 0.185,
+      "step": 93330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.426718313665436e-06,
+      "loss": 0.1805,
+      "step": 93340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.422214214935592e-06,
+      "loss": 0.1889,
+      "step": 93350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.417710116205747e-06,
+      "loss": 0.1818,
+      "step": 93360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.413206017475903e-06,
+      "loss": 0.1862,
+      "step": 93370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.408701918746059e-06,
+      "loss": 0.1845,
+      "step": 93380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.404197820016217e-06,
+      "loss": 0.1853,
+      "step": 93390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.39969372128637e-06,
+      "loss": 0.1818,
+      "step": 93400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.395189622556526e-06,
+      "loss": 0.1812,
+      "step": 93410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.390685523826682e-06,
+      "loss": 0.1866,
+      "step": 93420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.386181425096838e-06,
+      "loss": 0.1798,
+      "step": 93430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.381677326366996e-06,
+      "loss": 0.1817,
+      "step": 93440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.37717322763715e-06,
+      "loss": 0.1824,
+      "step": 93450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.372669128907307e-06,
+      "loss": 0.18,
+      "step": 93460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.368165030177461e-06,
+      "loss": 0.1838,
+      "step": 93470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.363660931447617e-06,
+      "loss": 0.181,
+      "step": 93480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.359156832717775e-06,
+      "loss": 0.1863,
+      "step": 93490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.354652733987929e-06,
+      "loss": 0.1812,
+      "step": 93500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.350148635258086e-06,
+      "loss": 0.1783,
+      "step": 93510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.34564453652824e-06,
+      "loss": 0.1863,
+      "step": 93520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.341140437798398e-06,
+      "loss": 0.1866,
+      "step": 93530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.336636339068553e-06,
+      "loss": 0.1858,
+      "step": 93540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.332132240338708e-06,
+      "loss": 0.1811,
+      "step": 93550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.327628141608865e-06,
+      "loss": 0.1855,
+      "step": 93560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.32312404287902e-06,
+      "loss": 0.184,
+      "step": 93570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.318619944149177e-06,
+      "loss": 0.1846,
+      "step": 93580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.314115845419332e-06,
+      "loss": 0.1814,
+      "step": 93590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.309611746689488e-06,
+      "loss": 0.1783,
+      "step": 93600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.305107647959644e-06,
+      "loss": 0.1776,
+      "step": 93610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.300603549229798e-06,
+      "loss": 0.1811,
+      "step": 93620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.296099450499956e-06,
+      "loss": 0.184,
+      "step": 93630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.291595351770111e-06,
+      "loss": 0.1771,
+      "step": 93640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.287091253040267e-06,
+      "loss": 0.1871,
+      "step": 93650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.282587154310423e-06,
+      "loss": 0.1844,
+      "step": 93660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.278083055580579e-06,
+      "loss": 0.1851,
+      "step": 93670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.273578956850735e-06,
+      "loss": 0.18,
+      "step": 93680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.26907485812089e-06,
+      "loss": 0.1819,
+      "step": 93690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.264570759391046e-06,
+      "loss": 0.1818,
+      "step": 93700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.260066660661202e-06,
+      "loss": 0.1862,
+      "step": 93710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.255562561931358e-06,
+      "loss": 0.1889,
+      "step": 93720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.251058463201514e-06,
+      "loss": 0.1844,
+      "step": 93730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.24655436447167e-06,
+      "loss": 0.1845,
+      "step": 93740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.242050265741825e-06,
+      "loss": 0.1812,
+      "step": 93750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.237546167011981e-06,
+      "loss": 0.1814,
+      "step": 93760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.233042068282137e-06,
+      "loss": 0.1781,
+      "step": 93770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.228537969552293e-06,
+      "loss": 0.1842,
+      "step": 93780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.22403387082245e-06,
+      "loss": 0.1867,
+      "step": 93790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.219529772092604e-06,
+      "loss": 0.183,
+      "step": 93800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.215025673362762e-06,
+      "loss": 0.1854,
+      "step": 93810
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.210521574632916e-06,
+      "loss": 0.1843,
+      "step": 93820
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.206017475903072e-06,
+      "loss": 0.185,
+      "step": 93830
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.201513377173229e-06,
+      "loss": 0.1803,
+      "step": 93840
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.197009278443383e-06,
+      "loss": 0.1815,
+      "step": 93850
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.19250517971354e-06,
+      "loss": 0.1795,
+      "step": 93860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.188001080983695e-06,
+      "loss": 0.1839,
+      "step": 93870
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.183496982253852e-06,
+      "loss": 0.1831,
+      "step": 93880
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.178992883524008e-06,
+      "loss": 0.181,
+      "step": 93890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.174488784794162e-06,
+      "loss": 0.1786,
+      "step": 93900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.16998468606432e-06,
+      "loss": 0.1811,
+      "step": 93910
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.165480587334474e-06,
+      "loss": 0.1796,
+      "step": 93920
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.160976488604631e-06,
+      "loss": 0.1832,
+      "step": 93930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.156472389874787e-06,
+      "loss": 0.1857,
+      "step": 93940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.151968291144941e-06,
+      "loss": 0.1885,
+      "step": 93950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.147464192415099e-06,
+      "loss": 0.1827,
+      "step": 93960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.142960093685253e-06,
+      "loss": 0.1793,
+      "step": 93970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.13845599495541e-06,
+      "loss": 0.1821,
+      "step": 93980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.133951896225566e-06,
+      "loss": 0.1769,
+      "step": 93990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.129447797495722e-06,
+      "loss": 0.1819,
+      "step": 94000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.124943698765878e-06,
+      "loss": 0.182,
+      "step": 94010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.120439600036032e-06,
+      "loss": 0.1848,
+      "step": 94020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.11593550130619e-06,
+      "loss": 0.1887,
+      "step": 94030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.111431402576345e-06,
+      "loss": 0.1802,
+      "step": 94040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.106927303846501e-06,
+      "loss": 0.1871,
+      "step": 94050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.102873614989641e-06,
+      "loss": 0.1785,
+      "step": 94060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.098369516259795e-06,
+      "loss": 0.1841,
+      "step": 94070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.093865417529953e-06,
+      "loss": 0.1856,
+      "step": 94080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.089361318800109e-06,
+      "loss": 0.1816,
+      "step": 94090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.084857220070264e-06,
+      "loss": 0.1818,
+      "step": 94100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.08035312134042e-06,
+      "loss": 0.1897,
+      "step": 94110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.075849022610576e-06,
+      "loss": 0.1809,
+      "step": 94120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.071344923880732e-06,
+      "loss": 0.1812,
+      "step": 94130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.066840825150888e-06,
+      "loss": 0.1857,
+      "step": 94140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.062336726421043e-06,
+      "loss": 0.176,
+      "step": 94150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.0578326276912e-06,
+      "loss": 0.1789,
+      "step": 94160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.053328528961355e-06,
+      "loss": 0.1788,
+      "step": 94170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.048824430231511e-06,
+      "loss": 0.1836,
+      "step": 94180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.044320331501667e-06,
+      "loss": 0.1818,
+      "step": 94190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.039816232771822e-06,
+      "loss": 0.1863,
+      "step": 94200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.035312134041978e-06,
+      "loss": 0.1788,
+      "step": 94210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.030808035312134e-06,
+      "loss": 0.1894,
+      "step": 94220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.02630393658229e-06,
+      "loss": 0.1864,
+      "step": 94230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.021799837852446e-06,
+      "loss": 0.183,
+      "step": 94240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.017295739122601e-06,
+      "loss": 0.1838,
+      "step": 94250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.012791640392759e-06,
+      "loss": 0.1829,
+      "step": 94260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.008287541662913e-06,
+      "loss": 0.1859,
+      "step": 94270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.003783442933069e-06,
+      "loss": 0.1835,
+      "step": 94280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.999279344203225e-06,
+      "loss": 0.1833,
+      "step": 94290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.99477524547338e-06,
+      "loss": 0.179,
+      "step": 94300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.990271146743538e-06,
+      "loss": 0.1832,
+      "step": 94310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.985767048013692e-06,
+      "loss": 0.1854,
+      "step": 94320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.98126294928385e-06,
+      "loss": 0.1861,
+      "step": 94330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.976758850554004e-06,
+      "loss": 0.1783,
+      "step": 94340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.97225475182416e-06,
+      "loss": 0.18,
+      "step": 94350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.967750653094317e-06,
+      "loss": 0.1833,
+      "step": 94360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.963246554364471e-06,
+      "loss": 0.181,
+      "step": 94370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.958742455634629e-06,
+      "loss": 0.1853,
+      "step": 94380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.954238356904784e-06,
+      "loss": 0.1818,
+      "step": 94390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.94973425817494e-06,
+      "loss": 0.1808,
+      "step": 94400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.945230159445096e-06,
+      "loss": 0.1793,
+      "step": 94410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.94072606071525e-06,
+      "loss": 0.1848,
+      "step": 94420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.936221961985408e-06,
+      "loss": 0.1846,
+      "step": 94430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.931717863255563e-06,
+      "loss": 0.187,
+      "step": 94440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.927213764525719e-06,
+      "loss": 0.183,
+      "step": 94450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.922709665795875e-06,
+      "loss": 0.185,
+      "step": 94460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.91820556706603e-06,
+      "loss": 0.1811,
+      "step": 94470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.913701468336186e-06,
+      "loss": 0.1828,
+      "step": 94480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.909197369606342e-06,
+      "loss": 0.1833,
+      "step": 94490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.904693270876498e-06,
+      "loss": 0.1844,
+      "step": 94500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.900189172146654e-06,
+      "loss": 0.1804,
+      "step": 94510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.89568507341681e-06,
+      "loss": 0.1767,
+      "step": 94520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.891180974686965e-06,
+      "loss": 0.1852,
+      "step": 94530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.886676875957121e-06,
+      "loss": 0.1815,
+      "step": 94540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.882172777227277e-06,
+      "loss": 0.1832,
+      "step": 94550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.877668678497433e-06,
+      "loss": 0.1887,
+      "step": 94560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.873614989640573e-06,
+      "loss": 0.1774,
+      "step": 94570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.869110890910729e-06,
+      "loss": 0.1844,
+      "step": 94580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.864606792180885e-06,
+      "loss": 0.1839,
+      "step": 94590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.86010269345104e-06,
+      "loss": 0.1783,
+      "step": 94600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.855598594721196e-06,
+      "loss": 0.1891,
+      "step": 94610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.851094495991352e-06,
+      "loss": 0.182,
+      "step": 94620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.846590397261508e-06,
+      "loss": 0.1829,
+      "step": 94630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.842086298531664e-06,
+      "loss": 0.1826,
+      "step": 94640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.83758219980182e-06,
+      "loss": 0.1853,
+      "step": 94650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.833078101071977e-06,
+      "loss": 0.1815,
+      "step": 94660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.828574002342131e-06,
+      "loss": 0.1814,
+      "step": 94670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.824069903612287e-06,
+      "loss": 0.1813,
+      "step": 94680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.819565804882443e-06,
+      "loss": 0.1818,
+      "step": 94690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.815061706152599e-06,
+      "loss": 0.1819,
+      "step": 94700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.810557607422756e-06,
+      "loss": 0.1835,
+      "step": 94710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.80605350869291e-06,
+      "loss": 0.1825,
+      "step": 94720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.801549409963068e-06,
+      "loss": 0.1802,
+      "step": 94730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.797045311233222e-06,
+      "loss": 0.1862,
+      "step": 94740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.792541212503378e-06,
+      "loss": 0.1815,
+      "step": 94750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.788037113773535e-06,
+      "loss": 0.1816,
+      "step": 94760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.78353301504369e-06,
+      "loss": 0.1786,
+      "step": 94770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.779028916313847e-06,
+      "loss": 0.1813,
+      "step": 94780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.774524817584e-06,
+      "loss": 0.1822,
+      "step": 94790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.770020718854158e-06,
+      "loss": 0.1817,
+      "step": 94800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.765516620124314e-06,
+      "loss": 0.18,
+      "step": 94810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.761012521394468e-06,
+      "loss": 0.1837,
+      "step": 94820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.756508422664626e-06,
+      "loss": 0.1788,
+      "step": 94830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.75200432393478e-06,
+      "loss": 0.1793,
+      "step": 94840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.747500225204937e-06,
+      "loss": 0.1819,
+      "step": 94850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.742996126475093e-06,
+      "loss": 0.1797,
+      "step": 94860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.738492027745249e-06,
+      "loss": 0.1856,
+      "step": 94870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.733987929015405e-06,
+      "loss": 0.1829,
+      "step": 94880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.729483830285559e-06,
+      "loss": 0.1875,
+      "step": 94890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.724979731555716e-06,
+      "loss": 0.1775,
+      "step": 94900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.720475632825872e-06,
+      "loss": 0.1803,
+      "step": 94910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.715971534096028e-06,
+      "loss": 0.1821,
+      "step": 94920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.711467435366184e-06,
+      "loss": 0.1799,
+      "step": 94930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.70696333663634e-06,
+      "loss": 0.1776,
+      "step": 94940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.702459237906495e-06,
+      "loss": 0.1762,
+      "step": 94950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.697955139176651e-06,
+      "loss": 0.1844,
+      "step": 94960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.693451040446807e-06,
+      "loss": 0.1793,
+      "step": 94970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.688946941716963e-06,
+      "loss": 0.1847,
+      "step": 94980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.684442842987118e-06,
+      "loss": 0.1807,
+      "step": 94990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.679938744257274e-06,
+      "loss": 0.1824,
+      "step": 95000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.67543464552743e-06,
+      "loss": 0.1811,
+      "step": 95010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.670930546797586e-06,
+      "loss": 0.1844,
+      "step": 95020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.666426448067742e-06,
+      "loss": 0.1805,
+      "step": 95030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.661922349337897e-06,
+      "loss": 0.1826,
+      "step": 95040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.657418250608053e-06,
+      "loss": 0.1805,
+      "step": 95050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.65291415187821e-06,
+      "loss": 0.18,
+      "step": 95060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.648410053148365e-06,
+      "loss": 0.1797,
+      "step": 95070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.643905954418522e-06,
+      "loss": 0.1814,
+      "step": 95080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.639401855688676e-06,
+      "loss": 0.184,
+      "step": 95090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.634897756958832e-06,
+      "loss": 0.1832,
+      "step": 95100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.63039365822899e-06,
+      "loss": 0.1797,
+      "step": 95110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.625889559499144e-06,
+      "loss": 0.1862,
+      "step": 95120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.6213854607693005e-06,
+      "loss": 0.1857,
+      "step": 95130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.616881362039456e-06,
+      "loss": 0.1764,
+      "step": 95140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.612377263309613e-06,
+      "loss": 0.1819,
+      "step": 95150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.607873164579768e-06,
+      "loss": 0.1812,
+      "step": 95160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.603369065849923e-06,
+      "loss": 0.1796,
+      "step": 95170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5988649671200795e-06,
+      "loss": 0.185,
+      "step": 95180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.594360868390235e-06,
+      "loss": 0.1846,
+      "step": 95190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.589856769660392e-06,
+      "loss": 0.1819,
+      "step": 95200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.585352670930547e-06,
+      "loss": 0.182,
+      "step": 95210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5808485722007035e-06,
+      "loss": 0.1856,
+      "step": 95220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5763444734708585e-06,
+      "loss": 0.1781,
+      "step": 95230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.571840374741014e-06,
+      "loss": 0.1823,
+      "step": 95240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.567336276011171e-06,
+      "loss": 0.182,
+      "step": 95250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.562832177281326e-06,
+      "loss": 0.1858,
+      "step": 95260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5583280785514825e-06,
+      "loss": 0.1795,
+      "step": 95270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5538239798216375e-06,
+      "loss": 0.1818,
+      "step": 95280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.549319881091794e-06,
+      "loss": 0.1813,
+      "step": 95290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.54481578236195e-06,
+      "loss": 0.1813,
+      "step": 95300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.540311683632105e-06,
+      "loss": 0.1829,
+      "step": 95310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5358075849022615e-06,
+      "loss": 0.1779,
+      "step": 95320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5313034861724165e-06,
+      "loss": 0.1835,
+      "step": 95330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.526799387442573e-06,
+      "loss": 0.1867,
+      "step": 95340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.522295288712729e-06,
+      "loss": 0.1824,
+      "step": 95350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5177911899828855e-06,
+      "loss": 0.1797,
+      "step": 95360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5132870912530405e-06,
+      "loss": 0.1828,
+      "step": 95370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.508782992523196e-06,
+      "loss": 0.1814,
+      "step": 95380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.504278893793353e-06,
+      "loss": 0.1828,
+      "step": 95390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.499774795063508e-06,
+      "loss": 0.184,
+      "step": 95400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4952706963336645e-06,
+      "loss": 0.1808,
+      "step": 95410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4907665976038195e-06,
+      "loss": 0.1768,
+      "step": 95420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.486262498873976e-06,
+      "loss": 0.1796,
+      "step": 95430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.481758400144132e-06,
+      "loss": 0.1835,
+      "step": 95440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.477254301414287e-06,
+      "loss": 0.1788,
+      "step": 95450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4727502026844435e-06,
+      "loss": 0.1857,
+      "step": 95460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4682461039545985e-06,
+      "loss": 0.1895,
+      "step": 95470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.463742005224755e-06,
+      "loss": 0.1805,
+      "step": 95480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.459237906494911e-06,
+      "loss": 0.1806,
+      "step": 95490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4547338077650676e-06,
+      "loss": 0.1795,
+      "step": 95500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4502297090352225e-06,
+      "loss": 0.1839,
+      "step": 95510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4457256103053775e-06,
+      "loss": 0.1752,
+      "step": 95520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.441221511575534e-06,
+      "loss": 0.1793,
+      "step": 95530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.43671741284569e-06,
+      "loss": 0.184,
+      "step": 95540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4322133141158466e-06,
+      "loss": 0.1919,
+      "step": 95550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4277092153860015e-06,
+      "loss": 0.1804,
+      "step": 95560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.423205116656158e-06,
+      "loss": 0.1822,
+      "step": 95570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.418701017926313e-06,
+      "loss": 0.1929,
+      "step": 95580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.414196919196469e-06,
+      "loss": 0.1808,
+      "step": 95590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4096928204666256e-06,
+      "loss": 0.1779,
+      "step": 95600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4051887217367805e-06,
+      "loss": 0.1834,
+      "step": 95610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.400684623006937e-06,
+      "loss": 0.1807,
+      "step": 95620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.396180524277092e-06,
+      "loss": 0.1842,
+      "step": 95630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.391676425547249e-06,
+      "loss": 0.1834,
+      "step": 95640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3871723268174045e-06,
+      "loss": 0.1792,
+      "step": 95650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3826682280875595e-06,
+      "loss": 0.1813,
+      "step": 95660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.378164129357716e-06,
+      "loss": 0.1797,
+      "step": 95670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.373660030627871e-06,
+      "loss": 0.1815,
+      "step": 95680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.369155931898028e-06,
+      "loss": 0.1759,
+      "step": 95690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3646518331681835e-06,
+      "loss": 0.1789,
+      "step": 95700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.36014773443834e-06,
+      "loss": 0.1873,
+      "step": 95710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.355643635708495e-06,
+      "loss": 0.1806,
+      "step": 95720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.35113953697865e-06,
+      "loss": 0.1833,
+      "step": 95730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.346635438248807e-06,
+      "loss": 0.1802,
+      "step": 95740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3421313395189625e-06,
+      "loss": 0.178,
+      "step": 95750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.337627240789119e-06,
+      "loss": 0.183,
+      "step": 95760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.333123142059274e-06,
+      "loss": 0.1776,
+      "step": 95770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.328619043329431e-06,
+      "loss": 0.1809,
+      "step": 95780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.324114944599586e-06,
+      "loss": 0.1883,
+      "step": 95790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.3196108458697415e-06,
+      "loss": 0.1844,
+      "step": 95800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.315106747139898e-06,
+      "loss": 0.1917,
+      "step": 95810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.310602648410053e-06,
+      "loss": 0.1814,
+      "step": 95820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.30609854968021e-06,
+      "loss": 0.1828,
+      "step": 95830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.301594450950365e-06,
+      "loss": 0.184,
+      "step": 95840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.297090352220521e-06,
+      "loss": 0.1853,
+      "step": 95850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.292586253490677e-06,
+      "loss": 0.1821,
+      "step": 95860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.288082154760832e-06,
+      "loss": 0.1901,
+      "step": 95870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.283578056030989e-06,
+      "loss": 0.1836,
+      "step": 95880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.279073957301144e-06,
+      "loss": 0.1813,
+      "step": 95890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.2745698585713e-06,
+      "loss": 0.1812,
+      "step": 95900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.270065759841456e-06,
+      "loss": 0.1757,
+      "step": 95910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.265561661111613e-06,
+      "loss": 0.1769,
+      "step": 95920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.261057562381768e-06,
+      "loss": 0.1892,
+      "step": 95930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.256553463651923e-06,
+      "loss": 0.1797,
+      "step": 95940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.252049364922079e-06,
+      "loss": 0.1842,
+      "step": 95950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.247545266192235e-06,
+      "loss": 0.1849,
+      "step": 95960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.243041167462392e-06,
+      "loss": 0.185,
+      "step": 95970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.238537068732547e-06,
+      "loss": 0.1832,
+      "step": 95980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.234032970002703e-06,
+      "loss": 0.1821,
+      "step": 95990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.229528871272858e-06,
+      "loss": 0.1803,
+      "step": 96000
+    },
+    {
+      "epoch": 0.86,
+      "eval_NEIMS_canon_smiles": 0.9448,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.06445,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.4672603673886566,
+      "eval_NEIMS_exact_mols": 0.0642,
+      "eval_NEIMS_exact_smiles": 0.06233333333333333,
+      "eval_NEIMS_loss": 0.21077434718608856,
+      "eval_NEIMS_matched_formulas": 0.5876333333333333,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.3718269128473269,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.06463333333333333,
+      "eval_NEIMS_runtime": 710.2144,
+      "eval_NEIMS_samples_per_second": 84.482,
+      "eval_NEIMS_steps_per_second": 1.321,
+      "step": 96000
+    },
+    {
+      "epoch": 0.86,
+      "eval_RASSP_canon_smiles": 0.9370220723277791,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.15273650115203527,
+      "eval_RASSP_daylight_tanimoto_simil": 0.5978313703107301,
+      "eval_RASSP_exact_mols": 0.15213543927605436,
+      "eval_RASSP_exact_smiles": 0.14769425985908438,
+      "eval_RASSP_loss": 0.14983013272285461,
+      "eval_RASSP_matched_formulas": 0.7636658095969546,
+      "eval_RASSP_morgan_tanimoto_simil": 0.47045272905034535,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.1522356162553845,
+      "eval_RASSP_runtime": 816.8397,
+      "eval_RASSP_samples_per_second": 73.324,
+      "eval_RASSP_steps_per_second": 1.146,
+      "step": 96000
+    },
+    {
+      "epoch": 0.86,
+      "eval_NIST_canon_smiles": 0.8822798736558186,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.01266990808105902,
+      "eval_NIST_daylight_tanimoto_simil": 0.2704161303006068,
+      "eval_NIST_exact_mols": 0.01210206906342052,
+      "eval_NIST_exact_smiles": 0.011321290414167583,
+      "eval_NIST_loss": 1.4756792783737183,
+      "eval_NIST_matched_formulas": 0.08922170564644923,
+      "eval_NIST_morgan_tanimoto_simil": 0.22269916852825272,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.01266990808105902,
+      "eval_NIST_runtime": 463.6612,
+      "eval_NIST_samples_per_second": 60.771,
+      "eval_NIST_steps_per_second": 0.951,
+      "step": 96000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.225024772543014e-06,
+      "loss": 0.1818,
+      "step": 96010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.220520673813171e-06,
+      "loss": 0.1834,
+      "step": 96020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.216016575083326e-06,
+      "loss": 0.1807,
+      "step": 96030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.211512476353482e-06,
+      "loss": 0.1799,
+      "step": 96040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.207008377623637e-06,
+      "loss": 0.1806,
+      "step": 96050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.202504278893794e-06,
+      "loss": 0.1786,
+      "step": 96060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.19800018016395e-06,
+      "loss": 0.1833,
+      "step": 96070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.193496081434105e-06,
+      "loss": 0.1819,
+      "step": 96080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.188991982704261e-06,
+      "loss": 0.1786,
+      "step": 96090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.184487883974417e-06,
+      "loss": 0.1779,
+      "step": 96100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.179983785244574e-06,
+      "loss": 0.1775,
+      "step": 96110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.175479686514729e-06,
+      "loss": 0.1814,
+      "step": 96120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.170975587784885e-06,
+      "loss": 0.1794,
+      "step": 96130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.16647148905504e-06,
+      "loss": 0.1807,
+      "step": 96140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.161967390325196e-06,
+      "loss": 0.187,
+      "step": 96150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.157463291595353e-06,
+      "loss": 0.1818,
+      "step": 96160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.152959192865508e-06,
+      "loss": 0.1824,
+      "step": 96170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.148455094135664e-06,
+      "loss": 0.1818,
+      "step": 96180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.143950995405819e-06,
+      "loss": 0.175,
+      "step": 96190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.139446896675976e-06,
+      "loss": 0.1771,
+      "step": 96200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.134942797946132e-06,
+      "loss": 0.18,
+      "step": 96210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.130438699216287e-06,
+      "loss": 0.1823,
+      "step": 96220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.125934600486443e-06,
+      "loss": 0.1817,
+      "step": 96230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.121430501756598e-06,
+      "loss": 0.1872,
+      "step": 96240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.116926403026755e-06,
+      "loss": 0.182,
+      "step": 96250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.112422304296911e-06,
+      "loss": 0.1836,
+      "step": 96260
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.107918205567067e-06,
+      "loss": 0.1832,
+      "step": 96270
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.103414106837222e-06,
+      "loss": 0.1798,
+      "step": 96280
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.098910008107377e-06,
+      "loss": 0.1781,
+      "step": 96290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.094405909377534e-06,
+      "loss": 0.1824,
+      "step": 96300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.08990181064769e-06,
+      "loss": 0.1835,
+      "step": 96310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.085397711917846e-06,
+      "loss": 0.1791,
+      "step": 96320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.080893613188001e-06,
+      "loss": 0.1769,
+      "step": 96330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.076389514458158e-06,
+      "loss": 0.1789,
+      "step": 96340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.071885415728313e-06,
+      "loss": 0.1768,
+      "step": 96350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.067381316998469e-06,
+      "loss": 0.1804,
+      "step": 96360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.062877218268625e-06,
+      "loss": 0.1873,
+      "step": 96370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.05837311953878e-06,
+      "loss": 0.1811,
+      "step": 96380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.053869020808937e-06,
+      "loss": 0.1745,
+      "step": 96390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.049364922079092e-06,
+      "loss": 0.1781,
+      "step": 96400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.044860823349248e-06,
+      "loss": 0.1797,
+      "step": 96410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.040356724619404e-06,
+      "loss": 0.1806,
+      "step": 96420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.035852625889559e-06,
+      "loss": 0.1818,
+      "step": 96430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.031348527159716e-06,
+      "loss": 0.183,
+      "step": 96440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.026844428429871e-06,
+      "loss": 0.1818,
+      "step": 96450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.022340329700028e-06,
+      "loss": 0.1822,
+      "step": 96460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.017836230970183e-06,
+      "loss": 0.175,
+      "step": 96470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.013332132240338e-06,
+      "loss": 0.1804,
+      "step": 96480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.008828033510495e-06,
+      "loss": 0.1843,
+      "step": 96490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.00432393478065e-06,
+      "loss": 0.1828,
+      "step": 96500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9998198360508066e-06,
+      "loss": 0.1809,
+      "step": 96510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.995315737320962e-06,
+      "loss": 0.1816,
+      "step": 96520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.990811638591119e-06,
+      "loss": 0.1822,
+      "step": 96530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.986307539861274e-06,
+      "loss": 0.1818,
+      "step": 96540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.981803441131429e-06,
+      "loss": 0.1789,
+      "step": 96550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9772993424015856e-06,
+      "loss": 0.1806,
+      "step": 96560
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.972795243671741e-06,
+      "loss": 0.1796,
+      "step": 96570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.968291144941898e-06,
+      "loss": 0.1828,
+      "step": 96580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.963787046212053e-06,
+      "loss": 0.1842,
+      "step": 96590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.95928294748221e-06,
+      "loss": 0.1796,
+      "step": 96600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9547788487523645e-06,
+      "loss": 0.1811,
+      "step": 96610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.95027475002252e-06,
+      "loss": 0.1778,
+      "step": 96620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.945770651292677e-06,
+      "loss": 0.1896,
+      "step": 96630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.941266552562832e-06,
+      "loss": 0.1827,
+      "step": 96640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.936762453832989e-06,
+      "loss": 0.1842,
+      "step": 96650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9322583551031435e-06,
+      "loss": 0.1812,
+      "step": 96660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9277542563733e-06,
+      "loss": 0.1847,
+      "step": 96670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.923250157643456e-06,
+      "loss": 0.1781,
+      "step": 96680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.918746058913611e-06,
+      "loss": 0.1776,
+      "step": 96690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.914241960183768e-06,
+      "loss": 0.1803,
+      "step": 96700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9097378614539225e-06,
+      "loss": 0.1756,
+      "step": 96710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.905233762724079e-06,
+      "loss": 0.1784,
+      "step": 96720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.900729663994235e-06,
+      "loss": 0.1837,
+      "step": 96730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.896225565264392e-06,
+      "loss": 0.1822,
+      "step": 96740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.8917214665345466e-06,
+      "loss": 0.1805,
+      "step": 96750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.887217367804702e-06,
+      "loss": 0.1786,
+      "step": 96760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.882713269074858e-06,
+      "loss": 0.1856,
+      "step": 96770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.878209170345014e-06,
+      "loss": 0.1833,
+      "step": 96780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.873705071615171e-06,
+      "loss": 0.1787,
+      "step": 96790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.8692009728853256e-06,
+      "loss": 0.1782,
+      "step": 96800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.864696874155482e-06,
+      "loss": 0.1777,
+      "step": 96810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.860192775425638e-06,
+      "loss": 0.1766,
+      "step": 96820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.855688676695793e-06,
+      "loss": 0.1779,
+      "step": 96830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.85118457796595e-06,
+      "loss": 0.1754,
+      "step": 96840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.8466804792361046e-06,
+      "loss": 0.1817,
+      "step": 96850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.842176380506261e-06,
+      "loss": 0.1805,
+      "step": 96860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.837672281776417e-06,
+      "loss": 0.1828,
+      "step": 96870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.833168183046574e-06,
+      "loss": 0.1776,
+      "step": 96880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.828664084316729e-06,
+      "loss": 0.1807,
+      "step": 96890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.8241599855868835e-06,
+      "loss": 0.1743,
+      "step": 96900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.81965588685704e-06,
+      "loss": 0.1834,
+      "step": 96910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.815151788127196e-06,
+      "loss": 0.1803,
+      "step": 96920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.810647689397353e-06,
+      "loss": 0.1918,
+      "step": 96930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.806143590667508e-06,
+      "loss": 0.1759,
+      "step": 96940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.801639491937664e-06,
+      "loss": 0.181,
+      "step": 96950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.797135393207819e-06,
+      "loss": 0.1773,
+      "step": 96960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.792631294477975e-06,
+      "loss": 0.1851,
+      "step": 96970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.788127195748132e-06,
+      "loss": 0.1848,
+      "step": 96980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7836230970182866e-06,
+      "loss": 0.1761,
+      "step": 96990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.779118998288443e-06,
+      "loss": 0.1815,
+      "step": 97000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.774614899558598e-06,
+      "loss": 0.1831,
+      "step": 97010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.770110800828755e-06,
+      "loss": 0.1815,
+      "step": 97020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.765606702098911e-06,
+      "loss": 0.1782,
+      "step": 97030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7611026033690656e-06,
+      "loss": 0.1872,
+      "step": 97040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.756598504639222e-06,
+      "loss": 0.1808,
+      "step": 97050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.752094405909377e-06,
+      "loss": 0.1775,
+      "step": 97060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.747590307179534e-06,
+      "loss": 0.182,
+      "step": 97070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.74308620844969e-06,
+      "loss": 0.1822,
+      "step": 97080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.738582109719846e-06,
+      "loss": 0.1836,
+      "step": 97090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.734078010990001e-06,
+      "loss": 0.1808,
+      "step": 97100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.729573912260156e-06,
+      "loss": 0.1836,
+      "step": 97110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.725069813530313e-06,
+      "loss": 0.1795,
+      "step": 97120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.720565714800469e-06,
+      "loss": 0.1816,
+      "step": 97130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.716061616070625e-06,
+      "loss": 0.1764,
+      "step": 97140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.71155751734078e-06,
+      "loss": 0.1851,
+      "step": 97150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.707053418610937e-06,
+      "loss": 0.1846,
+      "step": 97160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.702549319881092e-06,
+      "loss": 0.1814,
+      "step": 97170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.698045221151248e-06,
+      "loss": 0.1813,
+      "step": 97180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.693541122421404e-06,
+      "loss": 0.1848,
+      "step": 97190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.689037023691559e-06,
+      "loss": 0.1792,
+      "step": 97200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.684532924961716e-06,
+      "loss": 0.1793,
+      "step": 97210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.680028826231871e-06,
+      "loss": 0.1815,
+      "step": 97220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.675524727502027e-06,
+      "loss": 0.1824,
+      "step": 97230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.671020628772183e-06,
+      "loss": 0.1766,
+      "step": 97240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.666516530042338e-06,
+      "loss": 0.1842,
+      "step": 97250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.662012431312495e-06,
+      "loss": 0.183,
+      "step": 97260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.65750833258265e-06,
+      "loss": 0.1842,
+      "step": 97270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.653004233852806e-06,
+      "loss": 0.1803,
+      "step": 97280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.648500135122962e-06,
+      "loss": 0.1817,
+      "step": 97290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.643996036393119e-06,
+      "loss": 0.183,
+      "step": 97300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.639491937663274e-06,
+      "loss": 0.1823,
+      "step": 97310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.634987838933429e-06,
+      "loss": 0.1771,
+      "step": 97320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.630483740203585e-06,
+      "loss": 0.1811,
+      "step": 97330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.625979641473741e-06,
+      "loss": 0.1794,
+      "step": 97340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.621475542743898e-06,
+      "loss": 0.1847,
+      "step": 97350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.616971444014053e-06,
+      "loss": 0.1815,
+      "step": 97360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.6124673452842094e-06,
+      "loss": 0.1752,
+      "step": 97370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.607963246554364e-06,
+      "loss": 0.1842,
+      "step": 97380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.60345914782452e-06,
+      "loss": 0.1787,
+      "step": 97390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.598955049094677e-06,
+      "loss": 0.1822,
+      "step": 97400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.594450950364832e-06,
+      "loss": 0.1875,
+      "step": 97410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5899468516349884e-06,
+      "loss": 0.1736,
+      "step": 97420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.585442752905143e-06,
+      "loss": 0.1837,
+      "step": 97430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5809386541753e-06,
+      "loss": 0.1796,
+      "step": 97440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.576434555445456e-06,
+      "loss": 0.1779,
+      "step": 97450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.571930456715611e-06,
+      "loss": 0.1814,
+      "step": 97460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.567426357985767e-06,
+      "loss": 0.1778,
+      "step": 97470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.562922259255923e-06,
+      "loss": 0.1838,
+      "step": 97480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.558418160526079e-06,
+      "loss": 0.1848,
+      "step": 97490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.553914061796235e-06,
+      "loss": 0.1803,
+      "step": 97500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5494099630663915e-06,
+      "loss": 0.1838,
+      "step": 97510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.544905864336546e-06,
+      "loss": 0.1804,
+      "step": 97520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.540401765606702e-06,
+      "loss": 0.1869,
+      "step": 97530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.535897666876859e-06,
+      "loss": 0.1797,
+      "step": 97540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.531393568147014e-06,
+      "loss": 0.1798,
+      "step": 97550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5268894694171705e-06,
+      "loss": 0.176,
+      "step": 97560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.522385370687325e-06,
+      "loss": 0.1757,
+      "step": 97570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.517881271957482e-06,
+      "loss": 0.1837,
+      "step": 97580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.513377173227638e-06,
+      "loss": 0.1844,
+      "step": 97590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.508873074497793e-06,
+      "loss": 0.1793,
+      "step": 97600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5043689757679494e-06,
+      "loss": 0.18,
+      "step": 97610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.499864877038104e-06,
+      "loss": 0.1825,
+      "step": 97620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.495360778308261e-06,
+      "loss": 0.1785,
+      "step": 97630
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.490856679578417e-06,
+      "loss": 0.1807,
+      "step": 97640
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4863525808485735e-06,
+      "loss": 0.1815,
+      "step": 97650
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4818484821187284e-06,
+      "loss": 0.1792,
+      "step": 97660
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.477344383388883e-06,
+      "loss": 0.1776,
+      "step": 97670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.47284028465904e-06,
+      "loss": 0.1823,
+      "step": 97680
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.468336185929196e-06,
+      "loss": 0.178,
+      "step": 97690
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4638320871993525e-06,
+      "loss": 0.1785,
+      "step": 97700
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.459327988469507e-06,
+      "loss": 0.1856,
+      "step": 97710
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.454823889739664e-06,
+      "loss": 0.181,
+      "step": 97720
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.450319791009819e-06,
+      "loss": 0.179,
+      "step": 97730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.445815692279975e-06,
+      "loss": 0.184,
+      "step": 97740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4413115935501315e-06,
+      "loss": 0.1783,
+      "step": 97750
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.436807494820286e-06,
+      "loss": 0.1805,
+      "step": 97760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.432303396090443e-06,
+      "loss": 0.1711,
+      "step": 97770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.427799297360598e-06,
+      "loss": 0.1805,
+      "step": 97780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.423295198630755e-06,
+      "loss": 0.1795,
+      "step": 97790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4187910999009105e-06,
+      "loss": 0.1798,
+      "step": 97800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.414287001171065e-06,
+      "loss": 0.178,
+      "step": 97810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.409782902441222e-06,
+      "loss": 0.1794,
+      "step": 97820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.405278803711377e-06,
+      "loss": 0.1779,
+      "step": 97830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.400774704981534e-06,
+      "loss": 0.1766,
+      "step": 97840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3962706062516894e-06,
+      "loss": 0.1886,
+      "step": 97850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.391766507521846e-06,
+      "loss": 0.1765,
+      "step": 97860
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.387262408792001e-06,
+      "loss": 0.1774,
+      "step": 97870
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.382758310062156e-06,
+      "loss": 0.1861,
+      "step": 97880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.378254211332313e-06,
+      "loss": 0.1764,
+      "step": 97890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3737501126024684e-06,
+      "loss": 0.1826,
+      "step": 97900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.369246013872625e-06,
+      "loss": 0.181,
+      "step": 97910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.36474191514278e-06,
+      "loss": 0.1817,
+      "step": 97920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.360237816412937e-06,
+      "loss": 0.1817,
+      "step": 97930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.355733717683092e-06,
+      "loss": 0.1809,
+      "step": 97940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3512296189532474e-06,
+      "loss": 0.1806,
+      "step": 97950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.346725520223404e-06,
+      "loss": 0.1848,
+      "step": 97960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.342221421493559e-06,
+      "loss": 0.184,
+      "step": 97970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.337717322763716e-06,
+      "loss": 0.1792,
+      "step": 97980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.333213224033871e-06,
+      "loss": 0.1817,
+      "step": 97990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.328709125304027e-06,
+      "loss": 0.1792,
+      "step": 98000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.324205026574183e-06,
+      "loss": 0.1822,
+      "step": 98010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.319700927844338e-06,
+      "loss": 0.1784,
+      "step": 98020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.315196829114495e-06,
+      "loss": 0.175,
+      "step": 98030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.31069273038465e-06,
+      "loss": 0.1731,
+      "step": 98040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.306188631654806e-06,
+      "loss": 0.1778,
+      "step": 98050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.301684532924962e-06,
+      "loss": 0.181,
+      "step": 98060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.297180434195119e-06,
+      "loss": 0.184,
+      "step": 98070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.292676335465274e-06,
+      "loss": 0.1803,
+      "step": 98080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.288172236735429e-06,
+      "loss": 0.1827,
+      "step": 98090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.283668138005585e-06,
+      "loss": 0.1739,
+      "step": 98100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.279164039275741e-06,
+      "loss": 0.1771,
+      "step": 98110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.274659940545898e-06,
+      "loss": 0.1795,
+      "step": 98120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.270155841816053e-06,
+      "loss": 0.1872,
+      "step": 98130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.265651743086209e-06,
+      "loss": 0.1779,
+      "step": 98140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.261147644356364e-06,
+      "loss": 0.1794,
+      "step": 98150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.25664354562652e-06,
+      "loss": 0.1897,
+      "step": 98160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.252139446896677e-06,
+      "loss": 0.1862,
+      "step": 98170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.247635348166832e-06,
+      "loss": 0.1877,
+      "step": 98180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2431312494369874e-06,
+      "loss": 0.1791,
+      "step": 98190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.238627150707144e-06,
+      "loss": 0.182,
+      "step": 98200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2341230519773e-06,
+      "loss": 0.1842,
+      "step": 98210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.229618953247456e-06,
+      "loss": 0.1793,
+      "step": 98220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2251148545176115e-06,
+      "loss": 0.182,
+      "step": 98230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.220610755787767e-06,
+      "loss": 0.1777,
+      "step": 98240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.216106657057923e-06,
+      "loss": 0.1818,
+      "step": 98250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.211602558328079e-06,
+      "loss": 0.1791,
+      "step": 98260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.207098459598235e-06,
+      "loss": 0.1841,
+      "step": 98270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2025943608683905e-06,
+      "loss": 0.1803,
+      "step": 98280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.198090262138546e-06,
+      "loss": 0.1795,
+      "step": 98290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.193586163408702e-06,
+      "loss": 0.1765,
+      "step": 98300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.189082064678859e-06,
+      "loss": 0.1742,
+      "step": 98310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.184577965949014e-06,
+      "loss": 0.1836,
+      "step": 98320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1800738672191694e-06,
+      "loss": 0.178,
+      "step": 98330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.175569768489325e-06,
+      "loss": 0.1809,
+      "step": 98340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.171065669759481e-06,
+      "loss": 0.1827,
+      "step": 98350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.166561571029638e-06,
+      "loss": 0.1763,
+      "step": 98360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1620574722997935e-06,
+      "loss": 0.1788,
+      "step": 98370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.157553373569949e-06,
+      "loss": 0.1831,
+      "step": 98380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.153049274840104e-06,
+      "loss": 0.1852,
+      "step": 98390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.14854517611026e-06,
+      "loss": 0.1729,
+      "step": 98400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.144041077380417e-06,
+      "loss": 0.1825,
+      "step": 98410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1395369786505725e-06,
+      "loss": 0.1798,
+      "step": 98420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.135032879920728e-06,
+      "loss": 0.1755,
+      "step": 98430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.130528781190884e-06,
+      "loss": 0.1844,
+      "step": 98440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.12602468246104e-06,
+      "loss": 0.1805,
+      "step": 98450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.121520583731196e-06,
+      "loss": 0.1864,
+      "step": 98460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1170164850013515e-06,
+      "loss": 0.1849,
+      "step": 98470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.112512386271507e-06,
+      "loss": 0.1778,
+      "step": 98480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.108008287541663e-06,
+      "loss": 0.1858,
+      "step": 98490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.103504188811819e-06,
+      "loss": 0.185,
+      "step": 98500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.099000090081975e-06,
+      "loss": 0.1807,
+      "step": 98510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.094495991352131e-06,
+      "loss": 0.1793,
+      "step": 98520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.089991892622286e-06,
+      "loss": 0.1878,
+      "step": 98530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.085487793892442e-06,
+      "loss": 0.1748,
+      "step": 98540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.080983695162598e-06,
+      "loss": 0.184,
+      "step": 98550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0764795964327545e-06,
+      "loss": 0.1804,
+      "step": 98560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.07197549770291e-06,
+      "loss": 0.1797,
+      "step": 98570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.067471398973066e-06,
+      "loss": 0.1795,
+      "step": 98580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.063417710116206e-06,
+      "loss": 0.1784,
+      "step": 98590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0589136113863614e-06,
+      "loss": 0.1814,
+      "step": 98600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.054409512656518e-06,
+      "loss": 0.1798,
+      "step": 98610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.049905413926674e-06,
+      "loss": 0.1774,
+      "step": 98620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.04540131519683e-06,
+      "loss": 0.1815,
+      "step": 98630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0408972164669855e-06,
+      "loss": 0.183,
+      "step": 98640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.03639311773714e-06,
+      "loss": 0.1793,
+      "step": 98650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.031889019007297e-06,
+      "loss": 0.1774,
+      "step": 98660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.027384920277453e-06,
+      "loss": 0.1775,
+      "step": 98670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.022880821547609e-06,
+      "loss": 0.1843,
+      "step": 98680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0183767228177645e-06,
+      "loss": 0.1823,
+      "step": 98690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.01387262408792e-06,
+      "loss": 0.1738,
+      "step": 98700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.009368525358076e-06,
+      "loss": 0.1782,
+      "step": 98710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.004864426628232e-06,
+      "loss": 0.1803,
+      "step": 98720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.000360327898388e-06,
+      "loss": 0.1787,
+      "step": 98730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9958562291685435e-06,
+      "loss": 0.1795,
+      "step": 98740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.991352130438699e-06,
+      "loss": 0.1776,
+      "step": 98750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.986848031708855e-06,
+      "loss": 0.1794,
+      "step": 98760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.982343932979012e-06,
+      "loss": 0.1829,
+      "step": 98770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9778398342491675e-06,
+      "loss": 0.1804,
+      "step": 98780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9733357355193224e-06,
+      "loss": 0.1775,
+      "step": 98790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.968831636789478e-06,
+      "loss": 0.1701,
+      "step": 98800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.964327538059634e-06,
+      "loss": 0.1814,
+      "step": 98810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.959823439329791e-06,
+      "loss": 0.1782,
+      "step": 98820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9553193405999465e-06,
+      "loss": 0.1792,
+      "step": 98830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.950815241870102e-06,
+      "loss": 0.1801,
+      "step": 98840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.946311143140258e-06,
+      "loss": 0.1779,
+      "step": 98850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.941807044410414e-06,
+      "loss": 0.1794,
+      "step": 98860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.93730294568057e-06,
+      "loss": 0.1854,
+      "step": 98870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9327988469507255e-06,
+      "loss": 0.1847,
+      "step": 98880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.928294748220881e-06,
+      "loss": 0.1783,
+      "step": 98890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.923790649491037e-06,
+      "loss": 0.1825,
+      "step": 98900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.919286550761193e-06,
+      "loss": 0.174,
+      "step": 98910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9147824520313495e-06,
+      "loss": 0.1768,
+      "step": 98920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9102783533015045e-06,
+      "loss": 0.179,
+      "step": 98930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.90577425457166e-06,
+      "loss": 0.1802,
+      "step": 98940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.901270155841816e-06,
+      "loss": 0.1802,
+      "step": 98950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.896766057111972e-06,
+      "loss": 0.1813,
+      "step": 98960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8922619583821285e-06,
+      "loss": 0.1836,
+      "step": 98970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.887757859652284e-06,
+      "loss": 0.1839,
+      "step": 98980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.88325376092244e-06,
+      "loss": 0.1804,
+      "step": 98990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.878749662192595e-06,
+      "loss": 0.1784,
+      "step": 99000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.874245563462751e-06,
+      "loss": 0.1753,
+      "step": 99010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8697414647329075e-06,
+      "loss": 0.1846,
+      "step": 99020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.865237366003063e-06,
+      "loss": 0.1719,
+      "step": 99030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.860733267273219e-06,
+      "loss": 0.1822,
+      "step": 99040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.856229168543375e-06,
+      "loss": 0.1845,
+      "step": 99050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.851725069813531e-06,
+      "loss": 0.1798,
+      "step": 99060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8472209710836865e-06,
+      "loss": 0.1747,
+      "step": 99070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.842716872353842e-06,
+      "loss": 0.1762,
+      "step": 99080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.838212773623998e-06,
+      "loss": 0.1763,
+      "step": 99090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.833708674894154e-06,
+      "loss": 0.1822,
+      "step": 99100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.82920457616431e-06,
+      "loss": 0.1743,
+      "step": 99110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8247004774344655e-06,
+      "loss": 0.1797,
+      "step": 99120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.820196378704622e-06,
+      "loss": 0.1813,
+      "step": 99130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.815692279974777e-06,
+      "loss": 0.1844,
+      "step": 99140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.811188181244933e-06,
+      "loss": 0.1847,
+      "step": 99150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.806684082515089e-06,
+      "loss": 0.185,
+      "step": 99160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.8021799837852445e-06,
+      "loss": 0.1847,
+      "step": 99170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.797675885055401e-06,
+      "loss": 0.1772,
+      "step": 99180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.793171786325557e-06,
+      "loss": 0.1785,
+      "step": 99190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.788667687595713e-06,
+      "loss": 0.1739,
+      "step": 99200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.784163588865868e-06,
+      "loss": 0.1799,
+      "step": 99210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.779659490136024e-06,
+      "loss": 0.1774,
+      "step": 99220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.77515539140618e-06,
+      "loss": 0.1815,
+      "step": 99230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.770651292676336e-06,
+      "loss": 0.1824,
+      "step": 99240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.766147193946492e-06,
+      "loss": 0.174,
+      "step": 99250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7616430952166475e-06,
+      "loss": 0.1837,
+      "step": 99260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.757138996486803e-06,
+      "loss": 0.1805,
+      "step": 99270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.752634897756959e-06,
+      "loss": 0.1841,
+      "step": 99280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.748130799027115e-06,
+      "loss": 0.178,
+      "step": 99290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.743626700297271e-06,
+      "loss": 0.1804,
+      "step": 99300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7391226015674265e-06,
+      "loss": 0.1762,
+      "step": 99310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.734618502837582e-06,
+      "loss": 0.1828,
+      "step": 99320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.730114404107739e-06,
+      "loss": 0.1839,
+      "step": 99330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.725610305377895e-06,
+      "loss": 0.1782,
+      "step": 99340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.72110620664805e-06,
+      "loss": 0.1833,
+      "step": 99350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7166021079182055e-06,
+      "loss": 0.1808,
+      "step": 99360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.712098009188361e-06,
+      "loss": 0.1792,
+      "step": 99370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.707593910458518e-06,
+      "loss": 0.1844,
+      "step": 99380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.703089811728674e-06,
+      "loss": 0.181,
+      "step": 99390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6985857129988295e-06,
+      "loss": 0.1799,
+      "step": 99400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6940816142689845e-06,
+      "loss": 0.1787,
+      "step": 99410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.68957751553914e-06,
+      "loss": 0.1802,
+      "step": 99420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.685073416809297e-06,
+      "loss": 0.1797,
+      "step": 99430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.680569318079453e-06,
+      "loss": 0.1791,
+      "step": 99440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6760652193496085e-06,
+      "loss": 0.1852,
+      "step": 99450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.671561120619764e-06,
+      "loss": 0.1782,
+      "step": 99460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.66705702188992e-06,
+      "loss": 0.1778,
+      "step": 99470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.662552923160076e-06,
+      "loss": 0.1822,
+      "step": 99480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.658048824430232e-06,
+      "loss": 0.1827,
+      "step": 99490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6535447257003875e-06,
+      "loss": 0.184,
+      "step": 99500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.649040626970543e-06,
+      "loss": 0.1808,
+      "step": 99510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.644536528240699e-06,
+      "loss": 0.1817,
+      "step": 99520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.640032429510855e-06,
+      "loss": 0.1788,
+      "step": 99530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6355283307810115e-06,
+      "loss": 0.1786,
+      "step": 99540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6310242320511665e-06,
+      "loss": 0.1787,
+      "step": 99550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.626520133321322e-06,
+      "loss": 0.1821,
+      "step": 99560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.622016034591478e-06,
+      "loss": 0.1776,
+      "step": 99570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.617511935861635e-06,
+      "loss": 0.1784,
+      "step": 99580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6130078371317905e-06,
+      "loss": 0.1832,
+      "step": 99590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.608503738401946e-06,
+      "loss": 0.1788,
+      "step": 99600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.603999639672102e-06,
+      "loss": 0.1778,
+      "step": 99610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.599495540942257e-06,
+      "loss": 0.1763,
+      "step": 99620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.594991442212414e-06,
+      "loss": 0.179,
+      "step": 99630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5904873434825695e-06,
+      "loss": 0.178,
+      "step": 99640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.585983244752725e-06,
+      "loss": 0.1775,
+      "step": 99650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.581479146022881e-06,
+      "loss": 0.1816,
+      "step": 99660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.576975047293037e-06,
+      "loss": 0.1797,
+      "step": 99670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.572470948563193e-06,
+      "loss": 0.1814,
+      "step": 99680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5679668498333485e-06,
+      "loss": 0.1775,
+      "step": 99690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.563462751103504e-06,
+      "loss": 0.173,
+      "step": 99700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.55895865237366e-06,
+      "loss": 0.1786,
+      "step": 99710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.554454553643816e-06,
+      "loss": 0.1782,
+      "step": 99720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.549950454913972e-06,
+      "loss": 0.1848,
+      "step": 99730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.545446356184128e-06,
+      "loss": 0.1823,
+      "step": 99740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.540942257454284e-06,
+      "loss": 0.1757,
+      "step": 99750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.536438158724439e-06,
+      "loss": 0.1789,
+      "step": 99760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.531934059994595e-06,
+      "loss": 0.1803,
+      "step": 99770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.527429961264751e-06,
+      "loss": 0.1764,
+      "step": 99780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.522925862534907e-06,
+      "loss": 0.1779,
+      "step": 99790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.518421763805063e-06,
+      "loss": 0.1764,
+      "step": 99800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.513917665075219e-06,
+      "loss": 0.1797,
+      "step": 99810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.509413566345375e-06,
+      "loss": 0.1765,
+      "step": 99820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.50490946761553e-06,
+      "loss": 0.1846,
+      "step": 99830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.500405368885686e-06,
+      "loss": 0.1807,
+      "step": 99840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.495901270155842e-06,
+      "loss": 0.1721,
+      "step": 99850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.491397171425998e-06,
+      "loss": 0.1775,
+      "step": 99860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.486893072696154e-06,
+      "loss": 0.1735,
+      "step": 99870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4823889739663095e-06,
+      "loss": 0.1838,
+      "step": 99880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.477884875236465e-06,
+      "loss": 0.1837,
+      "step": 99890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.473380776506621e-06,
+      "loss": 0.1854,
+      "step": 99900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.468876677776777e-06,
+      "loss": 0.1764,
+      "step": 99910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.464372579046933e-06,
+      "loss": 0.1743,
+      "step": 99920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4598684803170885e-06,
+      "loss": 0.1801,
+      "step": 99930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.455364381587245e-06,
+      "loss": 0.1743,
+      "step": 99940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.450860282857401e-06,
+      "loss": 0.1754,
+      "step": 99950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.446356184127557e-06,
+      "loss": 0.1837,
+      "step": 99960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.441852085397712e-06,
+      "loss": 0.1801,
+      "step": 99970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4373479866678675e-06,
+      "loss": 0.1803,
+      "step": 99980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.432843887938024e-06,
+      "loss": 0.1768,
+      "step": 99990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.42833978920818e-06,
+      "loss": 0.1754,
+      "step": 100000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.423835690478336e-06,
+      "loss": 0.179,
+      "step": 100010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4193315917484915e-06,
+      "loss": 0.1791,
+      "step": 100020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.414827493018647e-06,
+      "loss": 0.1792,
+      "step": 100030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.410323394288803e-06,
+      "loss": 0.1774,
+      "step": 100040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.405819295558959e-06,
+      "loss": 0.1773,
+      "step": 100050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.401315196829115e-06,
+      "loss": 0.182,
+      "step": 100060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3968110980992705e-06,
+      "loss": 0.1753,
+      "step": 100070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.392306999369426e-06,
+      "loss": 0.1777,
+      "step": 100080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.387802900639582e-06,
+      "loss": 0.1753,
+      "step": 100090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.383298801909739e-06,
+      "loss": 0.1775,
+      "step": 100100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.378794703179894e-06,
+      "loss": 0.1776,
+      "step": 100110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3742906044500495e-06,
+      "loss": 0.1811,
+      "step": 100120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.369786505720205e-06,
+      "loss": 0.1775,
+      "step": 100130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.365282406990361e-06,
+      "loss": 0.1719,
+      "step": 100140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.360778308260518e-06,
+      "loss": 0.1768,
+      "step": 100150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3562742095306736e-06,
+      "loss": 0.1795,
+      "step": 100160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.351770110800829e-06,
+      "loss": 0.1822,
+      "step": 100170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.347266012070984e-06,
+      "loss": 0.178,
+      "step": 100180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.34276191334114e-06,
+      "loss": 0.178,
+      "step": 100190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.338257814611297e-06,
+      "loss": 0.1823,
+      "step": 100200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3337537158814526e-06,
+      "loss": 0.1765,
+      "step": 100210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.329249617151608e-06,
+      "loss": 0.1778,
+      "step": 100220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.324745518421764e-06,
+      "loss": 0.177,
+      "step": 100230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.32024141969192e-06,
+      "loss": 0.1741,
+      "step": 100240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.315737320962076e-06,
+      "loss": 0.1796,
+      "step": 100250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.3112332222322315e-06,
+      "loss": 0.18,
+      "step": 100260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.306729123502387e-06,
+      "loss": 0.1741,
+      "step": 100270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.302225024772543e-06,
+      "loss": 0.179,
+      "step": 100280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.297720926042699e-06,
+      "loss": 0.1803,
+      "step": 100290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.293216827312856e-06,
+      "loss": 0.1806,
+      "step": 100300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.288712728583011e-06,
+      "loss": 0.1824,
+      "step": 100310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.284208629853166e-06,
+      "loss": 0.1761,
+      "step": 100320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.279704531123322e-06,
+      "loss": 0.1809,
+      "step": 100330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.275200432393478e-06,
+      "loss": 0.1753,
+      "step": 100340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2706963336636346e-06,
+      "loss": 0.1831,
+      "step": 100350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.26619223493379e-06,
+      "loss": 0.1779,
+      "step": 100360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.261688136203946e-06,
+      "loss": 0.1796,
+      "step": 100370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.257184037474102e-06,
+      "loss": 0.1813,
+      "step": 100380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.252679938744257e-06,
+      "loss": 0.1803,
+      "step": 100390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2481758400144136e-06,
+      "loss": 0.179,
+      "step": 100400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.243671741284569e-06,
+      "loss": 0.1769,
+      "step": 100410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.239167642554725e-06,
+      "loss": 0.174,
+      "step": 100420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.234663543824881e-06,
+      "loss": 0.1794,
+      "step": 100430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.230159445095037e-06,
+      "loss": 0.1783,
+      "step": 100440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2256553463651926e-06,
+      "loss": 0.1774,
+      "step": 100450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.221151247635348e-06,
+      "loss": 0.1865,
+      "step": 100460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.216647148905504e-06,
+      "loss": 0.1736,
+      "step": 100470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.21214305017566e-06,
+      "loss": 0.177,
+      "step": 100480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.207638951445816e-06,
+      "loss": 0.1793,
+      "step": 100490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2031348527159715e-06,
+      "loss": 0.1807,
+      "step": 100500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.198630753986128e-06,
+      "loss": 0.1817,
+      "step": 100510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.194126655256284e-06,
+      "loss": 0.1801,
+      "step": 100520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.189622556526439e-06,
+      "loss": 0.175,
+      "step": 100530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.185118457796595e-06,
+      "loss": 0.178,
+      "step": 100540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1806143590667505e-06,
+      "loss": 0.1768,
+      "step": 100550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.176110260336907e-06,
+      "loss": 0.1747,
+      "step": 100560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.171606161607063e-06,
+      "loss": 0.1766,
+      "step": 100570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.167102062877219e-06,
+      "loss": 0.1714,
+      "step": 100580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.162597964147374e-06,
+      "loss": 0.1788,
+      "step": 100590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.158544275290515e-06,
+      "loss": 0.1761,
+      "step": 100600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.154040176560671e-06,
+      "loss": 0.1785,
+      "step": 100610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1495360778308266e-06,
+      "loss": 0.1805,
+      "step": 100620
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.145031979100982e-06,
+      "loss": 0.1773,
+      "step": 100630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.140527880371137e-06,
+      "loss": 0.1862,
+      "step": 100640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.136023781641294e-06,
+      "loss": 0.1844,
+      "step": 100650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.13151968291145e-06,
+      "loss": 0.181,
+      "step": 100660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1270155841816055e-06,
+      "loss": 0.1777,
+      "step": 100670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.122511485451761e-06,
+      "loss": 0.1823,
+      "step": 100680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.118007386721917e-06,
+      "loss": 0.1801,
+      "step": 100690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.113503287992073e-06,
+      "loss": 0.1748,
+      "step": 100700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.108999189262229e-06,
+      "loss": 0.1762,
+      "step": 100710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1044950905323845e-06,
+      "loss": 0.1801,
+      "step": 100720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.09999099180254e-06,
+      "loss": 0.1818,
+      "step": 100730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.095486893072696e-06,
+      "loss": 0.1772,
+      "step": 100740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.090982794342852e-06,
+      "loss": 0.1782,
+      "step": 100750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.086478695613009e-06,
+      "loss": 0.1773,
+      "step": 100760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.081974596883164e-06,
+      "loss": 0.1751,
+      "step": 100770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.077470498153319e-06,
+      "loss": 0.1803,
+      "step": 100780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.072966399423475e-06,
+      "loss": 0.1775,
+      "step": 100790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.068462300693631e-06,
+      "loss": 0.181,
+      "step": 100800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0639582019637876e-06,
+      "loss": 0.1762,
+      "step": 100810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.059454103233943e-06,
+      "loss": 0.1807,
+      "step": 100820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.054950004504099e-06,
+      "loss": 0.1812,
+      "step": 100830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.050445905774255e-06,
+      "loss": 0.183,
+      "step": 100840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.04594180704441e-06,
+      "loss": 0.1776,
+      "step": 100850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0414377083145666e-06,
+      "loss": 0.1767,
+      "step": 100860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.036933609584722e-06,
+      "loss": 0.1799,
+      "step": 100870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.032429510854878e-06,
+      "loss": 0.1713,
+      "step": 100880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.027925412125034e-06,
+      "loss": 0.1797,
+      "step": 100890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.02342131339519e-06,
+      "loss": 0.176,
+      "step": 100900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.018917214665346e-06,
+      "loss": 0.1753,
+      "step": 100910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.014413115935501e-06,
+      "loss": 0.1773,
+      "step": 100920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.009909017205657e-06,
+      "loss": 0.1777,
+      "step": 100930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.005404918475813e-06,
+      "loss": 0.175,
+      "step": 100940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.000900819745969e-06,
+      "loss": 0.1807,
+      "step": 100950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.996396721016125e-06,
+      "loss": 0.1777,
+      "step": 100960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.991892622286281e-06,
+      "loss": 0.1789,
+      "step": 100970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.987388523556437e-06,
+      "loss": 0.173,
+      "step": 100980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.982884424826592e-06,
+      "loss": 0.1754,
+      "step": 100990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.978380326096748e-06,
+      "loss": 0.179,
+      "step": 101000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.973876227366904e-06,
+      "loss": 0.1762,
+      "step": 101010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.96937212863706e-06,
+      "loss": 0.1789,
+      "step": 101020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.964868029907216e-06,
+      "loss": 0.176,
+      "step": 101030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.960363931177372e-06,
+      "loss": 0.182,
+      "step": 101040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9558598324475276e-06,
+      "loss": 0.1855,
+      "step": 101050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.951355733717683e-06,
+      "loss": 0.1749,
+      "step": 101060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.946851634987839e-06,
+      "loss": 0.1765,
+      "step": 101070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.942347536257995e-06,
+      "loss": 0.1794,
+      "step": 101080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.937843437528151e-06,
+      "loss": 0.1796,
+      "step": 101090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9333393387983066e-06,
+      "loss": 0.1827,
+      "step": 101100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.928835240068462e-06,
+      "loss": 0.1788,
+      "step": 101110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.924331141338619e-06,
+      "loss": 0.182,
+      "step": 101120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.919827042608774e-06,
+      "loss": 0.1765,
+      "step": 101130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.91532294387893e-06,
+      "loss": 0.1772,
+      "step": 101140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9108188451490856e-06,
+      "loss": 0.1808,
+      "step": 101150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.906314746419241e-06,
+      "loss": 0.1793,
+      "step": 101160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.901810647689398e-06,
+      "loss": 0.1764,
+      "step": 101170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.897306548959554e-06,
+      "loss": 0.181,
+      "step": 101180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.89280245022971e-06,
+      "loss": 0.1763,
+      "step": 101190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8882983514998645e-06,
+      "loss": 0.1756,
+      "step": 101200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.88379425277002e-06,
+      "loss": 0.175,
+      "step": 101210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.879290154040177e-06,
+      "loss": 0.1753,
+      "step": 101220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.874786055310333e-06,
+      "loss": 0.1778,
+      "step": 101230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.870281956580489e-06,
+      "loss": 0.178,
+      "step": 101240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.865777857850644e-06,
+      "loss": 0.1774,
+      "step": 101250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8612737591208e-06,
+      "loss": 0.1769,
+      "step": 101260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.856769660390956e-06,
+      "loss": 0.1772,
+      "step": 101270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.852265561661112e-06,
+      "loss": 0.1733,
+      "step": 101280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8477614629312676e-06,
+      "loss": 0.1814,
+      "step": 101290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.843257364201423e-06,
+      "loss": 0.1784,
+      "step": 101300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.838753265471579e-06,
+      "loss": 0.1813,
+      "step": 101310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.834249166741736e-06,
+      "loss": 0.1782,
+      "step": 101320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.829745068011892e-06,
+      "loss": 0.1758,
+      "step": 101330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8252409692820466e-06,
+      "loss": 0.1783,
+      "step": 101340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.820736870552202e-06,
+      "loss": 0.1726,
+      "step": 101350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.816232771822358e-06,
+      "loss": 0.174,
+      "step": 101360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.811728673092515e-06,
+      "loss": 0.1802,
+      "step": 101370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.807224574362671e-06,
+      "loss": 0.1729,
+      "step": 101380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.802720475632826e-06,
+      "loss": 0.1779,
+      "step": 101390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.798216376902982e-06,
+      "loss": 0.1795,
+      "step": 101400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.793712278173137e-06,
+      "loss": 0.1738,
+      "step": 101410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.789208179443294e-06,
+      "loss": 0.1762,
+      "step": 101420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.78470408071345e-06,
+      "loss": 0.1774,
+      "step": 101430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.780199981983605e-06,
+      "loss": 0.1741,
+      "step": 101440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.775695883253761e-06,
+      "loss": 0.1799,
+      "step": 101450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.771191784523917e-06,
+      "loss": 0.1838,
+      "step": 101460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.766687685794073e-06,
+      "loss": 0.177,
+      "step": 101470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.762183587064229e-06,
+      "loss": 0.1823,
+      "step": 101480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.757679488334384e-06,
+      "loss": 0.1777,
+      "step": 101490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.75317538960454e-06,
+      "loss": 0.1746,
+      "step": 101500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.748671290874696e-06,
+      "loss": 0.1755,
+      "step": 101510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.744167192144852e-06,
+      "loss": 0.18,
+      "step": 101520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.739663093415008e-06,
+      "loss": 0.1799,
+      "step": 101530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.735158994685164e-06,
+      "loss": 0.1742,
+      "step": 101540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.730654895955319e-06,
+      "loss": 0.1768,
+      "step": 101550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.726150797225475e-06,
+      "loss": 0.1809,
+      "step": 101560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.721646698495631e-06,
+      "loss": 0.1774,
+      "step": 101570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.717142599765787e-06,
+      "loss": 0.1754,
+      "step": 101580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.712638501035943e-06,
+      "loss": 0.1722,
+      "step": 101590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.708134402306099e-06,
+      "loss": 0.1705,
+      "step": 101600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.703630303576255e-06,
+      "loss": 0.1735,
+      "step": 101610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.699126204846411e-06,
+      "loss": 0.1763,
+      "step": 101620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.694622106116566e-06,
+      "loss": 0.1779,
+      "step": 101630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.690118007386722e-06,
+      "loss": 0.1733,
+      "step": 101640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.685613908656878e-06,
+      "loss": 0.1766,
+      "step": 101650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.681109809927034e-06,
+      "loss": 0.1794,
+      "step": 101660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.67660571119719e-06,
+      "loss": 0.1788,
+      "step": 101670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.672101612467346e-06,
+      "loss": 0.1823,
+      "step": 101680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.667597513737501e-06,
+      "loss": 0.1817,
+      "step": 101690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.663093415007657e-06,
+      "loss": 0.174,
+      "step": 101700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.658589316277813e-06,
+      "loss": 0.1728,
+      "step": 101710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.654085217547969e-06,
+      "loss": 0.1727,
+      "step": 101720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.649581118818125e-06,
+      "loss": 0.177,
+      "step": 101730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.645077020088281e-06,
+      "loss": 0.1815,
+      "step": 101740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.640572921358437e-06,
+      "loss": 0.1731,
+      "step": 101750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.636068822628592e-06,
+      "loss": 0.1837,
+      "step": 101760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6315647238987476e-06,
+      "loss": 0.1861,
+      "step": 101770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.627060625168904e-06,
+      "loss": 0.1752,
+      "step": 101780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.62255652643906e-06,
+      "loss": 0.1793,
+      "step": 101790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.618052427709216e-06,
+      "loss": 0.1778,
+      "step": 101800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.613548328979372e-06,
+      "loss": 0.1715,
+      "step": 101810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6090442302495266e-06,
+      "loss": 0.1759,
+      "step": 101820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.604540131519683e-06,
+      "loss": 0.1787,
+      "step": 101830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.600036032789839e-06,
+      "loss": 0.182,
+      "step": 101840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.595531934059995e-06,
+      "loss": 0.1752,
+      "step": 101850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.591027835330151e-06,
+      "loss": 0.1807,
+      "step": 101860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.586523736600306e-06,
+      "loss": 0.1794,
+      "step": 101870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.582019637870462e-06,
+      "loss": 0.18,
+      "step": 101880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.577515539140618e-06,
+      "loss": 0.1783,
+      "step": 101890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.573011440410774e-06,
+      "loss": 0.1772,
+      "step": 101900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.56850734168093e-06,
+      "loss": 0.1809,
+      "step": 101910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.564003242951085e-06,
+      "loss": 0.1795,
+      "step": 101920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.559499144221241e-06,
+      "loss": 0.176,
+      "step": 101930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.554995045491398e-06,
+      "loss": 0.1768,
+      "step": 101940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.550490946761554e-06,
+      "loss": 0.1765,
+      "step": 101950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.545986848031709e-06,
+      "loss": 0.1762,
+      "step": 101960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.541482749301864e-06,
+      "loss": 0.1783,
+      "step": 101970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.536978650572021e-06,
+      "loss": 0.1788,
+      "step": 101980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.532474551842177e-06,
+      "loss": 0.173,
+      "step": 101990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.527970453112333e-06,
+      "loss": 0.1766,
+      "step": 102000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.523466354382488e-06,
+      "loss": 0.1788,
+      "step": 102010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.518962255652644e-06,
+      "loss": 0.1798,
+      "step": 102020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5144581569228e-06,
+      "loss": 0.1755,
+      "step": 102030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.509954058192956e-06,
+      "loss": 0.1745,
+      "step": 102040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.505449959463112e-06,
+      "loss": 0.1783,
+      "step": 102050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.500945860733267e-06,
+      "loss": 0.1761,
+      "step": 102060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.496441762003423e-06,
+      "loss": 0.1761,
+      "step": 102070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.491937663273579e-06,
+      "loss": 0.1745,
+      "step": 102080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.487433564543736e-06,
+      "loss": 0.1811,
+      "step": 102090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.482929465813891e-06,
+      "loss": 0.1796,
+      "step": 102100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.478425367084046e-06,
+      "loss": 0.1751,
+      "step": 102110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.473921268354202e-06,
+      "loss": 0.1758,
+      "step": 102120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.469417169624358e-06,
+      "loss": 0.1796,
+      "step": 102130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.464913070894515e-06,
+      "loss": 0.1761,
+      "step": 102140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4604089721646704e-06,
+      "loss": 0.1766,
+      "step": 102150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.455904873434826e-06,
+      "loss": 0.1775,
+      "step": 102160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.451400774704981e-06,
+      "loss": 0.1754,
+      "step": 102170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.446896675975137e-06,
+      "loss": 0.1796,
+      "step": 102180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.442392577245294e-06,
+      "loss": 0.1799,
+      "step": 102190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4378884785154494e-06,
+      "loss": 0.1797,
+      "step": 102200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.433384379785605e-06,
+      "loss": 0.1749,
+      "step": 102210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.428880281055761e-06,
+      "loss": 0.173,
+      "step": 102220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.424376182325917e-06,
+      "loss": 0.1788,
+      "step": 102230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.419872083596073e-06,
+      "loss": 0.1716,
+      "step": 102240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4153679848662284e-06,
+      "loss": 0.1812,
+      "step": 102250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.410863886136384e-06,
+      "loss": 0.1733,
+      "step": 102260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.40635978740654e-06,
+      "loss": 0.1747,
+      "step": 102270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.401855688676696e-06,
+      "loss": 0.1771,
+      "step": 102280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.397351589946852e-06,
+      "loss": 0.1776,
+      "step": 102290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.392847491217008e-06,
+      "loss": 0.1773,
+      "step": 102300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.388343392487163e-06,
+      "loss": 0.1717,
+      "step": 102310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.383839293757319e-06,
+      "loss": 0.1778,
+      "step": 102320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.379335195027475e-06,
+      "loss": 0.1774,
+      "step": 102330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3748310962976315e-06,
+      "loss": 0.1813,
+      "step": 102340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.370326997567787e-06,
+      "loss": 0.1764,
+      "step": 102350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.365822898837943e-06,
+      "loss": 0.1703,
+      "step": 102360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.361318800108099e-06,
+      "loss": 0.177,
+      "step": 102370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.356814701378254e-06,
+      "loss": 0.1758,
+      "step": 102380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3523106026484104e-06,
+      "loss": 0.178,
+      "step": 102390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.347806503918566e-06,
+      "loss": 0.1844,
+      "step": 102400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.343302405188722e-06,
+      "loss": 0.1725,
+      "step": 102410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.338798306458878e-06,
+      "loss": 0.1709,
+      "step": 102420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.334294207729034e-06,
+      "loss": 0.1751,
+      "step": 102430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3297901089991894e-06,
+      "loss": 0.1722,
+      "step": 102440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.325286010269345e-06,
+      "loss": 0.1777,
+      "step": 102450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.320781911539501e-06,
+      "loss": 0.1766,
+      "step": 102460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.316277812809657e-06,
+      "loss": 0.1738,
+      "step": 102470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.311773714079813e-06,
+      "loss": 0.1812,
+      "step": 102480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.3072696153499684e-06,
+      "loss": 0.1803,
+      "step": 102490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.302765516620125e-06,
+      "loss": 0.1751,
+      "step": 102500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.298261417890281e-06,
+      "loss": 0.1809,
+      "step": 102510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.293757319160436e-06,
+      "loss": 0.1771,
+      "step": 102520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.289253220430592e-06,
+      "loss": 0.1768,
+      "step": 102530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.284749121700747e-06,
+      "loss": 0.1795,
+      "step": 102540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.280245022970904e-06,
+      "loss": 0.1808,
+      "step": 102550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.27574092424106e-06,
+      "loss": 0.1771,
+      "step": 102560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.271236825511216e-06,
+      "loss": 0.1758,
+      "step": 102570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2667327267813715e-06,
+      "loss": 0.1725,
+      "step": 102580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.262228628051526e-06,
+      "loss": 0.1814,
+      "step": 102590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.257724529321683e-06,
+      "loss": 0.1774,
+      "step": 102600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.253220430591839e-06,
+      "loss": 0.1741,
+      "step": 102610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.249166741734979e-06,
+      "loss": 0.1789,
+      "step": 102620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.244662643005135e-06,
+      "loss": 0.1767,
+      "step": 102630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.240158544275291e-06,
+      "loss": 0.1731,
+      "step": 102640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.235654445545447e-06,
+      "loss": 0.1745,
+      "step": 102650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2311503468156024e-06,
+      "loss": 0.1793,
+      "step": 102660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.226646248085758e-06,
+      "loss": 0.1778,
+      "step": 102670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.222142149355914e-06,
+      "loss": 0.1725,
+      "step": 102680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.21763805062607e-06,
+      "loss": 0.1741,
+      "step": 102690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2131339518962265e-06,
+      "loss": 0.1778,
+      "step": 102700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.208629853166381e-06,
+      "loss": 0.1779,
+      "step": 102710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.204125754436537e-06,
+      "loss": 0.1848,
+      "step": 102720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.199621655706693e-06,
+      "loss": 0.1799,
+      "step": 102730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.195117556976849e-06,
+      "loss": 0.1793,
+      "step": 102740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1906134582470055e-06,
+      "loss": 0.1828,
+      "step": 102750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.186109359517161e-06,
+      "loss": 0.1794,
+      "step": 102760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.181605260787317e-06,
+      "loss": 0.1778,
+      "step": 102770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.177101162057472e-06,
+      "loss": 0.177,
+      "step": 102780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.172597063327628e-06,
+      "loss": 0.1767,
+      "step": 102790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1680929645977845e-06,
+      "loss": 0.1786,
+      "step": 102800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.16358886586794e-06,
+      "loss": 0.1851,
+      "step": 102810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.159084767138096e-06,
+      "loss": 0.1811,
+      "step": 102820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.154580668408252e-06,
+      "loss": 0.1771,
+      "step": 102830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.150076569678408e-06,
+      "loss": 0.1828,
+      "step": 102840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1455724709485634e-06,
+      "loss": 0.1792,
+      "step": 102850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.141068372218719e-06,
+      "loss": 0.1776,
+      "step": 102860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.136564273488875e-06,
+      "loss": 0.1751,
+      "step": 102870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.132060174759031e-06,
+      "loss": 0.1762,
+      "step": 102880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.127556076029187e-06,
+      "loss": 0.1761,
+      "step": 102890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1230519772993424e-06,
+      "loss": 0.1795,
+      "step": 102900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.118547878569499e-06,
+      "loss": 0.1786,
+      "step": 102910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.114043779839654e-06,
+      "loss": 0.1704,
+      "step": 102920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.10953968110981e-06,
+      "loss": 0.1777,
+      "step": 102930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.105035582379966e-06,
+      "loss": 0.1734,
+      "step": 102940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.100531483650122e-06,
+      "loss": 0.1725,
+      "step": 102950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.096027384920278e-06,
+      "loss": 0.1764,
+      "step": 102960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.091523286190434e-06,
+      "loss": 0.1734,
+      "step": 102970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.08701918746059e-06,
+      "loss": 0.1759,
+      "step": 102980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.082515088730745e-06,
+      "loss": 0.1805,
+      "step": 102990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.078010990000901e-06,
+      "loss": 0.1717,
+      "step": 103000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.073506891271057e-06,
+      "loss": 0.18,
+      "step": 103010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.069002792541213e-06,
+      "loss": 0.1769,
+      "step": 103020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.064498693811369e-06,
+      "loss": 0.1721,
+      "step": 103030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0599945950815245e-06,
+      "loss": 0.1688,
+      "step": 103040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.05549049635168e-06,
+      "loss": 0.1744,
+      "step": 103050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.050986397621836e-06,
+      "loss": 0.1808,
+      "step": 103060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.046482298891992e-06,
+      "loss": 0.1814,
+      "step": 103070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.041978200162148e-06,
+      "loss": 0.1771,
+      "step": 103080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0374741014323034e-06,
+      "loss": 0.1726,
+      "step": 103090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.032970002702459e-06,
+      "loss": 0.1723,
+      "step": 103100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.028465903972616e-06,
+      "loss": 0.178,
+      "step": 103110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.023961805242771e-06,
+      "loss": 0.1778,
+      "step": 103120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.019457706512927e-06,
+      "loss": 0.1727,
+      "step": 103130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0149536077830824e-06,
+      "loss": 0.1861,
+      "step": 103140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.010449509053238e-06,
+      "loss": 0.1739,
+      "step": 103150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.005945410323395e-06,
+      "loss": 0.1786,
+      "step": 103160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.001441311593551e-06,
+      "loss": 0.1784,
+      "step": 103170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9969372128637065e-06,
+      "loss": 0.1811,
+      "step": 103180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9924331141338614e-06,
+      "loss": 0.1734,
+      "step": 103190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.987929015404017e-06,
+      "loss": 0.1759,
+      "step": 103200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.983424916674174e-06,
+      "loss": 0.182,
+      "step": 103210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.97892081794433e-06,
+      "loss": 0.1791,
+      "step": 103220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9744167192144855e-06,
+      "loss": 0.1779,
+      "step": 103230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.969912620484641e-06,
+      "loss": 0.1765,
+      "step": 103240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.965408521754797e-06,
+      "loss": 0.1821,
+      "step": 103250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.960904423024953e-06,
+      "loss": 0.1737,
+      "step": 103260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.956400324295109e-06,
+      "loss": 0.1752,
+      "step": 103270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9518962255652645e-06,
+      "loss": 0.1743,
+      "step": 103280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.94739212683542e-06,
+      "loss": 0.1742,
+      "step": 103290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.942888028105576e-06,
+      "loss": 0.1796,
+      "step": 103300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.938383929375733e-06,
+      "loss": 0.1733,
+      "step": 103310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9338798306458885e-06,
+      "loss": 0.1793,
+      "step": 103320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9293757319160434e-06,
+      "loss": 0.1756,
+      "step": 103330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.924871633186199e-06,
+      "loss": 0.1765,
+      "step": 103340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.920367534456355e-06,
+      "loss": 0.1807,
+      "step": 103350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.915863435726512e-06,
+      "loss": 0.1715,
+      "step": 103360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9113593369966675e-06,
+      "loss": 0.179,
+      "step": 103370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.906855238266823e-06,
+      "loss": 0.1784,
+      "step": 103380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.902351139536979e-06,
+      "loss": 0.1736,
+      "step": 103390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.897847040807134e-06,
+      "loss": 0.1793,
+      "step": 103400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.893342942077291e-06,
+      "loss": 0.1787,
+      "step": 103410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8888388433474465e-06,
+      "loss": 0.1719,
+      "step": 103420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.884334744617602e-06,
+      "loss": 0.1764,
+      "step": 103430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.879830645887758e-06,
+      "loss": 0.1755,
+      "step": 103440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.875326547157914e-06,
+      "loss": 0.1779,
+      "step": 103450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.87082244842807e-06,
+      "loss": 0.1737,
+      "step": 103460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8663183496982255e-06,
+      "loss": 0.175,
+      "step": 103470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.861814250968381e-06,
+      "loss": 0.174,
+      "step": 103480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.857310152238537e-06,
+      "loss": 0.1727,
+      "step": 103490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.852806053508693e-06,
+      "loss": 0.1695,
+      "step": 103500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.848301954778849e-06,
+      "loss": 0.1789,
+      "step": 103510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.843797856049005e-06,
+      "loss": 0.1798,
+      "step": 103520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.839293757319161e-06,
+      "loss": 0.1766,
+      "step": 103530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.834789658589316e-06,
+      "loss": 0.1772,
+      "step": 103540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.830285559859472e-06,
+      "loss": 0.1765,
+      "step": 103550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.825781461129628e-06,
+      "loss": 0.1764,
+      "step": 103560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.821277362399784e-06,
+      "loss": 0.174,
+      "step": 103570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.81677326366994e-06,
+      "loss": 0.1694,
+      "step": 103580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.812269164940096e-06,
+      "loss": 0.173,
+      "step": 103590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8077650662102517e-06,
+      "loss": 0.1792,
+      "step": 103600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.803260967480407e-06,
+      "loss": 0.1825,
+      "step": 103610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.798756868750563e-06,
+      "loss": 0.1759,
+      "step": 103620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.794252770020719e-06,
+      "loss": 0.1749,
+      "step": 103630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.789748671290875e-06,
+      "loss": 0.1775,
+      "step": 103640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7852445725610307e-06,
+      "loss": 0.1744,
+      "step": 103650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.780740473831187e-06,
+      "loss": 0.1766,
+      "step": 103660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7762363751013427e-06,
+      "loss": 0.1729,
+      "step": 103670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.771732276371498e-06,
+      "loss": 0.1764,
+      "step": 103680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.767228177641654e-06,
+      "loss": 0.1773,
+      "step": 103690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7627240789118097e-06,
+      "loss": 0.1772,
+      "step": 103700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.758219980181966e-06,
+      "loss": 0.1772,
+      "step": 103710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7537158814521217e-06,
+      "loss": 0.1761,
+      "step": 103720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7492117827222775e-06,
+      "loss": 0.1775,
+      "step": 103730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7447076839924337e-06,
+      "loss": 0.1771,
+      "step": 103740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7402035852625887e-06,
+      "loss": 0.1738,
+      "step": 103750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.735699486532745e-06,
+      "loss": 0.177,
+      "step": 103760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7311953878029007e-06,
+      "loss": 0.1776,
+      "step": 103770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7266912890730565e-06,
+      "loss": 0.1741,
+      "step": 103780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7221871903432127e-06,
+      "loss": 0.172,
+      "step": 103790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7176830916133685e-06,
+      "loss": 0.1746,
+      "step": 103800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7131789928835243e-06,
+      "loss": 0.1778,
+      "step": 103810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7086748941536797e-06,
+      "loss": 0.1783,
+      "step": 103820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.704170795423836e-06,
+      "loss": 0.1773,
+      "step": 103830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6996666966939917e-06,
+      "loss": 0.1761,
+      "step": 103840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6951625979641475e-06,
+      "loss": 0.1787,
+      "step": 103850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6906584992343037e-06,
+      "loss": 0.1726,
+      "step": 103860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6861544005044595e-06,
+      "loss": 0.1795,
+      "step": 103870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6816503017746153e-06,
+      "loss": 0.176,
+      "step": 103880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6771462030447707e-06,
+      "loss": 0.1754,
+      "step": 103890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6726421043149265e-06,
+      "loss": 0.1787,
+      "step": 103900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6681380055850827e-06,
+      "loss": 0.1826,
+      "step": 103910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6636339068552385e-06,
+      "loss": 0.1779,
+      "step": 103920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6591298081253943e-06,
+      "loss": 0.1738,
+      "step": 103930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6546257093955505e-06,
+      "loss": 0.1771,
+      "step": 103940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6501216106657063e-06,
+      "loss": 0.1765,
+      "step": 103950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6456175119358617e-06,
+      "loss": 0.1777,
+      "step": 103960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6411134132060175e-06,
+      "loss": 0.1799,
+      "step": 103970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6366093144761733e-06,
+      "loss": 0.1804,
+      "step": 103980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6321052157463295e-06,
+      "loss": 0.1772,
+      "step": 103990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6276011170164853e-06,
+      "loss": 0.1796,
+      "step": 104000
+    },
+    {
+      "epoch": 0.93,
+      "eval_NEIMS_canon_smiles": 0.9435833333333333,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.07121666666666666,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.47393896052800716,
+      "eval_NEIMS_exact_mols": 0.07103333333333334,
+      "eval_NEIMS_exact_smiles": 0.06931666666666667,
+      "eval_NEIMS_loss": 0.20564579963684082,
+      "eval_NEIMS_matched_formulas": 0.6034666666666667,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.3786909203467994,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.0714,
+      "eval_NEIMS_runtime": 714.9764,
+      "eval_NEIMS_samples_per_second": 83.919,
+      "eval_NEIMS_steps_per_second": 1.312,
+      "step": 104000
+    },
+    {
+      "epoch": 0.93,
+      "eval_RASSP_canon_smiles": 0.9392092697098207,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.1658596854442849,
+      "eval_RASSP_daylight_tanimoto_simil": 0.6079714885990567,
+      "eval_RASSP_exact_mols": 0.1652419274050823,
+      "eval_RASSP_exact_smiles": 0.16050021705012188,
+      "eval_RASSP_loss": 0.14461752772331238,
+      "eval_RASSP_matched_formulas": 0.7783584332320432,
+      "eval_RASSP_morgan_tanimoto_simil": 0.48231341268363775,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.16542558520052092,
+      "eval_RASSP_runtime": 816.0852,
+      "eval_RASSP_samples_per_second": 73.392,
+      "eval_RASSP_steps_per_second": 1.147,
+      "step": 104000
+    },
+    {
+      "epoch": 0.93,
+      "eval_NIST_canon_smiles": 0.8845867196649749,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.013805586116336019,
+      "eval_NIST_daylight_tanimoto_simil": 0.272257324898437,
+      "eval_NIST_exact_mols": 0.013486176668914362,
+      "eval_NIST_exact_smiles": 0.012598928203854207,
+      "eval_NIST_loss": 1.4478906393051147,
+      "eval_NIST_matched_formulas": 0.09422578698938851,
+      "eval_NIST_morgan_tanimoto_simil": 0.22477003460350337,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.013947545870745644,
+      "eval_NIST_runtime": 472.0033,
+      "eval_NIST_samples_per_second": 59.697,
+      "eval_NIST_steps_per_second": 0.934,
+      "step": 104000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.623097018286641e-06,
+      "loss": 0.1807,
+      "step": 104010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6185929195567973e-06,
+      "loss": 0.1725,
+      "step": 104020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6140888208269523e-06,
+      "loss": 0.172,
+      "step": 104030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6095847220971085e-06,
+      "loss": 0.1766,
+      "step": 104040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6050806233672643e-06,
+      "loss": 0.1718,
+      "step": 104050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.60057652463742e-06,
+      "loss": 0.1692,
+      "step": 104060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5960724259075763e-06,
+      "loss": 0.1778,
+      "step": 104070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.591568327177732e-06,
+      "loss": 0.1763,
+      "step": 104080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.587064228447888e-06,
+      "loss": 0.1749,
+      "step": 104090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5825601297180433e-06,
+      "loss": 0.1801,
+      "step": 104100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.578056030988199e-06,
+      "loss": 0.1785,
+      "step": 104110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5735519322583553e-06,
+      "loss": 0.1745,
+      "step": 104120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.569047833528511e-06,
+      "loss": 0.1775,
+      "step": 104130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.564543734798667e-06,
+      "loss": 0.1746,
+      "step": 104140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.560039636068823e-06,
+      "loss": 0.1785,
+      "step": 104150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.555535537338979e-06,
+      "loss": 0.1776,
+      "step": 104160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5510314386091343e-06,
+      "loss": 0.1786,
+      "step": 104170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.54652733987929e-06,
+      "loss": 0.1775,
+      "step": 104180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5420232411494463e-06,
+      "loss": 0.1851,
+      "step": 104190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.537519142419602e-06,
+      "loss": 0.1785,
+      "step": 104200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.533015043689758e-06,
+      "loss": 0.1728,
+      "step": 104210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.528510944959914e-06,
+      "loss": 0.1788,
+      "step": 104220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.524006846230069e-06,
+      "loss": 0.1729,
+      "step": 104230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5195027475002253e-06,
+      "loss": 0.1786,
+      "step": 104240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.514998648770381e-06,
+      "loss": 0.1715,
+      "step": 104250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.510494550040537e-06,
+      "loss": 0.1725,
+      "step": 104260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.505990451310693e-06,
+      "loss": 0.1725,
+      "step": 104270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.501486352580849e-06,
+      "loss": 0.1737,
+      "step": 104280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4969822538510047e-06,
+      "loss": 0.1746,
+      "step": 104290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.49247815512116e-06,
+      "loss": 0.178,
+      "step": 104300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.487974056391316e-06,
+      "loss": 0.1796,
+      "step": 104310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.483469957661472e-06,
+      "loss": 0.1763,
+      "step": 104320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.478965858931628e-06,
+      "loss": 0.1754,
+      "step": 104330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4744617602017837e-06,
+      "loss": 0.1779,
+      "step": 104340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.46995766147194e-06,
+      "loss": 0.1703,
+      "step": 104350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4654535627420957e-06,
+      "loss": 0.1688,
+      "step": 104360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.460949464012251e-06,
+      "loss": 0.1794,
+      "step": 104370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.456445365282407e-06,
+      "loss": 0.1791,
+      "step": 104380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4519412665525627e-06,
+      "loss": 0.1791,
+      "step": 104390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.447437167822719e-06,
+      "loss": 0.18,
+      "step": 104400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4429330690928747e-06,
+      "loss": 0.175,
+      "step": 104410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4384289703630305e-06,
+      "loss": 0.1716,
+      "step": 104420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4339248716331867e-06,
+      "loss": 0.1784,
+      "step": 104430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4294207729033417e-06,
+      "loss": 0.1721,
+      "step": 104440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.424916674173498e-06,
+      "loss": 0.1758,
+      "step": 104450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4204125754436537e-06,
+      "loss": 0.176,
+      "step": 104460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4159084767138095e-06,
+      "loss": 0.1774,
+      "step": 104470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4114043779839657e-06,
+      "loss": 0.1767,
+      "step": 104480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4069002792541215e-06,
+      "loss": 0.1744,
+      "step": 104490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4023961805242773e-06,
+      "loss": 0.1769,
+      "step": 104500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3978920817944327e-06,
+      "loss": 0.1719,
+      "step": 104510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.393387983064589e-06,
+      "loss": 0.1703,
+      "step": 104520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3888838843347447e-06,
+      "loss": 0.1713,
+      "step": 104530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3843797856049005e-06,
+      "loss": 0.1751,
+      "step": 104540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3798756868750567e-06,
+      "loss": 0.176,
+      "step": 104550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3753715881452125e-06,
+      "loss": 0.1689,
+      "step": 104560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3708674894153683e-06,
+      "loss": 0.1792,
+      "step": 104570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3663633906855237e-06,
+      "loss": 0.1776,
+      "step": 104580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3618592919556795e-06,
+      "loss": 0.1784,
+      "step": 104590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3573551932258357e-06,
+      "loss": 0.18,
+      "step": 104600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3528510944959915e-06,
+      "loss": 0.1768,
+      "step": 104610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.348797405639132e-06,
+      "loss": 0.1799,
+      "step": 104620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3442933069092873e-06,
+      "loss": 0.1786,
+      "step": 104630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.339789208179443e-06,
+      "loss": 0.1707,
+      "step": 104640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3352851094495993e-06,
+      "loss": 0.1723,
+      "step": 104650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.330781010719755e-06,
+      "loss": 0.1715,
+      "step": 104660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.326276911989911e-06,
+      "loss": 0.1778,
+      "step": 104670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.321772813260067e-06,
+      "loss": 0.1769,
+      "step": 104680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.317268714530223e-06,
+      "loss": 0.1799,
+      "step": 104690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3127646158003783e-06,
+      "loss": 0.1777,
+      "step": 104700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.308260517070534e-06,
+      "loss": 0.1688,
+      "step": 104710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.30375641834069e-06,
+      "loss": 0.1764,
+      "step": 104720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.299252319610846e-06,
+      "loss": 0.1744,
+      "step": 104730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.294748220881002e-06,
+      "loss": 0.1773,
+      "step": 104740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2902441221511577e-06,
+      "loss": 0.1704,
+      "step": 104750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.285740023421314e-06,
+      "loss": 0.1761,
+      "step": 104760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.281235924691469e-06,
+      "loss": 0.1753,
+      "step": 104770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.276731825961625e-06,
+      "loss": 0.1754,
+      "step": 104780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.272227727231781e-06,
+      "loss": 0.1782,
+      "step": 104790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2677236285019367e-06,
+      "loss": 0.1764,
+      "step": 104800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.263219529772093e-06,
+      "loss": 0.1761,
+      "step": 104810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2587154310422487e-06,
+      "loss": 0.1776,
+      "step": 104820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.254211332312405e-06,
+      "loss": 0.1749,
+      "step": 104830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.24970723358256e-06,
+      "loss": 0.1811,
+      "step": 104840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.245203134852716e-06,
+      "loss": 0.179,
+      "step": 104850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.240699036122872e-06,
+      "loss": 0.1727,
+      "step": 104860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2361949373930277e-06,
+      "loss": 0.1787,
+      "step": 104870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.231690838663184e-06,
+      "loss": 0.1783,
+      "step": 104880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2271867399333397e-06,
+      "loss": 0.1737,
+      "step": 104890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2226826412034955e-06,
+      "loss": 0.1714,
+      "step": 104900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.218178542473651e-06,
+      "loss": 0.1722,
+      "step": 104910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2136744437438067e-06,
+      "loss": 0.1744,
+      "step": 104920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.209170345013963e-06,
+      "loss": 0.1703,
+      "step": 104930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2046662462841187e-06,
+      "loss": 0.1741,
+      "step": 104940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2001621475542745e-06,
+      "loss": 0.1761,
+      "step": 104950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1956580488244307e-06,
+      "loss": 0.1808,
+      "step": 104960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1911539500945865e-06,
+      "loss": 0.1724,
+      "step": 104970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.186649851364742e-06,
+      "loss": 0.174,
+      "step": 104980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1821457526348977e-06,
+      "loss": 0.1745,
+      "step": 104990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1776416539050535e-06,
+      "loss": 0.1714,
+      "step": 105000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1731375551752097e-06,
+      "loss": 0.1801,
+      "step": 105010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1686334564453655e-06,
+      "loss": 0.175,
+      "step": 105020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1641293577155213e-06,
+      "loss": 0.1795,
+      "step": 105030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1596252589856776e-06,
+      "loss": 0.1786,
+      "step": 105040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1551211602558325e-06,
+      "loss": 0.1766,
+      "step": 105050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1506170615259887e-06,
+      "loss": 0.1799,
+      "step": 105060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1461129627961445e-06,
+      "loss": 0.1763,
+      "step": 105070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1416088640663003e-06,
+      "loss": 0.1763,
+      "step": 105080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1371047653364565e-06,
+      "loss": 0.1766,
+      "step": 105090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1326006666066123e-06,
+      "loss": 0.1755,
+      "step": 105100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.128096567876768e-06,
+      "loss": 0.1757,
+      "step": 105110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.123592469146924e-06,
+      "loss": 0.1768,
+      "step": 105120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1190883704170793e-06,
+      "loss": 0.1734,
+      "step": 105130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1145842716872355e-06,
+      "loss": 0.1753,
+      "step": 105140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1100801729573913e-06,
+      "loss": 0.1751,
+      "step": 105150
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.105576074227547e-06,
+      "loss": 0.1758,
+      "step": 105160
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.101071975497703e-06,
+      "loss": 0.1722,
+      "step": 105170
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0965678767678587e-06,
+      "loss": 0.1736,
+      "step": 105180
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.092063778038015e-06,
+      "loss": 0.1718,
+      "step": 105190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0875596793081703e-06,
+      "loss": 0.1784,
+      "step": 105200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0830555805783265e-06,
+      "loss": 0.1745,
+      "step": 105210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0785514818484823e-06,
+      "loss": 0.1776,
+      "step": 105220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.074047383118638e-06,
+      "loss": 0.1771,
+      "step": 105230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.069543284388794e-06,
+      "loss": 0.1738,
+      "step": 105240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0650391856589497e-06,
+      "loss": 0.1713,
+      "step": 105250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0605350869291055e-06,
+      "loss": 0.1754,
+      "step": 105260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0560309881992613e-06,
+      "loss": 0.1732,
+      "step": 105270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.051526889469417e-06,
+      "loss": 0.1752,
+      "step": 105280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0470227907395734e-06,
+      "loss": 0.1754,
+      "step": 105290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.042518692009729e-06,
+      "loss": 0.1735,
+      "step": 105300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0380145932798845e-06,
+      "loss": 0.1788,
+      "step": 105310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0335104945500408e-06,
+      "loss": 0.1766,
+      "step": 105320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0290063958201965e-06,
+      "loss": 0.1797,
+      "step": 105330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0245022970903523e-06,
+      "loss": 0.1824,
+      "step": 105340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.019998198360508e-06,
+      "loss": 0.1773,
+      "step": 105350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.015494099630664e-06,
+      "loss": 0.1785,
+      "step": 105360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.01099000090082e-06,
+      "loss": 0.1853,
+      "step": 105370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0064859021709755e-06,
+      "loss": 0.1755,
+      "step": 105380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0019818034411318e-06,
+      "loss": 0.1805,
+      "step": 105390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9974777047112876e-06,
+      "loss": 0.1744,
+      "step": 105400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.992973605981443e-06,
+      "loss": 0.1776,
+      "step": 105410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.988469507251599e-06,
+      "loss": 0.1715,
+      "step": 105420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.983965408521755e-06,
+      "loss": 0.1759,
+      "step": 105430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9794613097919108e-06,
+      "loss": 0.1757,
+      "step": 105440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9749572110620666e-06,
+      "loss": 0.1781,
+      "step": 105450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9704531123322223e-06,
+      "loss": 0.1734,
+      "step": 105460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9659490136023786e-06,
+      "loss": 0.1738,
+      "step": 105470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.961444914872534e-06,
+      "loss": 0.1733,
+      "step": 105480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9569408161426897e-06,
+      "loss": 0.1707,
+      "step": 105490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.952436717412846e-06,
+      "loss": 0.1749,
+      "step": 105500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9479326186830018e-06,
+      "loss": 0.1757,
+      "step": 105510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9434285199531576e-06,
+      "loss": 0.1777,
+      "step": 105520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9389244212233134e-06,
+      "loss": 0.1724,
+      "step": 105530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.934420322493469e-06,
+      "loss": 0.1723,
+      "step": 105540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.929916223763625e-06,
+      "loss": 0.1766,
+      "step": 105550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9254121250337808e-06,
+      "loss": 0.1779,
+      "step": 105560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.920908026303937e-06,
+      "loss": 0.1759,
+      "step": 105570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9164039275740928e-06,
+      "loss": 0.1748,
+      "step": 105580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.911899828844248e-06,
+      "loss": 0.1817,
+      "step": 105590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9073957301144044e-06,
+      "loss": 0.1733,
+      "step": 105600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.90289163138456e-06,
+      "loss": 0.1756,
+      "step": 105610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.898387532654716e-06,
+      "loss": 0.1768,
+      "step": 105620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8938834339248718e-06,
+      "loss": 0.1687,
+      "step": 105630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8893793351950276e-06,
+      "loss": 0.1727,
+      "step": 105640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8848752364651838e-06,
+      "loss": 0.1756,
+      "step": 105650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.880371137735339e-06,
+      "loss": 0.1756,
+      "step": 105660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.875867039005495e-06,
+      "loss": 0.1788,
+      "step": 105670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.871362940275651e-06,
+      "loss": 0.1743,
+      "step": 105680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8668588415458066e-06,
+      "loss": 0.1744,
+      "step": 105690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8628051526889474e-06,
+      "loss": 0.1766,
+      "step": 105700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8583010539591027e-06,
+      "loss": 0.1826,
+      "step": 105710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.853796955229259e-06,
+      "loss": 0.1782,
+      "step": 105720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8492928564994148e-06,
+      "loss": 0.1732,
+      "step": 105730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.84478875776957e-06,
+      "loss": 0.1702,
+      "step": 105740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8402846590397264e-06,
+      "loss": 0.1775,
+      "step": 105750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.835780560309882e-06,
+      "loss": 0.192,
+      "step": 105760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.831276461580038e-06,
+      "loss": 0.1789,
+      "step": 105770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8267723628501937e-06,
+      "loss": 0.1799,
+      "step": 105780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8222682641203495e-06,
+      "loss": 0.176,
+      "step": 105790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8177641653905058e-06,
+      "loss": 0.178,
+      "step": 105800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.813260066660661e-06,
+      "loss": 0.1761,
+      "step": 105810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8087559679308174e-06,
+      "loss": 0.179,
+      "step": 105820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.804251869200973e-06,
+      "loss": 0.174,
+      "step": 105830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7997477704711285e-06,
+      "loss": 0.1815,
+      "step": 105840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7952436717412848e-06,
+      "loss": 0.1758,
+      "step": 105850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7907395730114406e-06,
+      "loss": 0.1799,
+      "step": 105860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7862354742815964e-06,
+      "loss": 0.1801,
+      "step": 105870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.781731375551752e-06,
+      "loss": 0.1711,
+      "step": 105880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.777227276821908e-06,
+      "loss": 0.1732,
+      "step": 105890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.772723178092064e-06,
+      "loss": 0.1744,
+      "step": 105900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7682190793622195e-06,
+      "loss": 0.1749,
+      "step": 105910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7637149806323753e-06,
+      "loss": 0.1743,
+      "step": 105920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7592108819025316e-06,
+      "loss": 0.1742,
+      "step": 105930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7547067831726874e-06,
+      "loss": 0.1718,
+      "step": 105940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.750202684442843e-06,
+      "loss": 0.1784,
+      "step": 105950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.745698585712999e-06,
+      "loss": 0.1733,
+      "step": 105960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7411944869831548e-06,
+      "loss": 0.1755,
+      "step": 105970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7366903882533106e-06,
+      "loss": 0.1758,
+      "step": 105980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7321862895234664e-06,
+      "loss": 0.1806,
+      "step": 105990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7276821907936226e-06,
+      "loss": 0.1764,
+      "step": 106000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7231780920637784e-06,
+      "loss": 0.1744,
+      "step": 106010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7186739933339337e-06,
+      "loss": 0.1761,
+      "step": 106020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.71416989460409e-06,
+      "loss": 0.1833,
+      "step": 106030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7096657958742458e-06,
+      "loss": 0.1748,
+      "step": 106040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7051616971444016e-06,
+      "loss": 0.1752,
+      "step": 106050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7006575984145574e-06,
+      "loss": 0.1817,
+      "step": 106060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.696153499684713e-06,
+      "loss": 0.1778,
+      "step": 106070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6916494009548694e-06,
+      "loss": 0.172,
+      "step": 106080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6871453022250248e-06,
+      "loss": 0.1765,
+      "step": 106090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6826412034951806e-06,
+      "loss": 0.1786,
+      "step": 106100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6781371047653368e-06,
+      "loss": 0.1771,
+      "step": 106110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.673633006035492e-06,
+      "loss": 0.1744,
+      "step": 106120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6691289073056484e-06,
+      "loss": 0.1767,
+      "step": 106130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.664624808575804e-06,
+      "loss": 0.1759,
+      "step": 106140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.66012070984596e-06,
+      "loss": 0.1827,
+      "step": 106150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6556166111161158e-06,
+      "loss": 0.1735,
+      "step": 106160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6511125123862716e-06,
+      "loss": 0.1778,
+      "step": 106170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.646608413656428e-06,
+      "loss": 0.1818,
+      "step": 106180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.642104314926583e-06,
+      "loss": 0.1723,
+      "step": 106190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.637600216196739e-06,
+      "loss": 0.1795,
+      "step": 106200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.633096117466895e-06,
+      "loss": 0.1701,
+      "step": 106210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.628592018737051e-06,
+      "loss": 0.1792,
+      "step": 106220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6240879200072068e-06,
+      "loss": 0.1745,
+      "step": 106230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6195838212773626e-06,
+      "loss": 0.1735,
+      "step": 106240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6150797225475184e-06,
+      "loss": 0.1728,
+      "step": 106250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.610575623817674e-06,
+      "loss": 0.1787,
+      "step": 106260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.60607152508783e-06,
+      "loss": 0.1774,
+      "step": 106270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6015674263579858e-06,
+      "loss": 0.1756,
+      "step": 106280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.597063327628142e-06,
+      "loss": 0.1739,
+      "step": 106290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5925592288982974e-06,
+      "loss": 0.177,
+      "step": 106300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5880551301684536e-06,
+      "loss": 0.1729,
+      "step": 106310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5835510314386094e-06,
+      "loss": 0.1717,
+      "step": 106320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.579046932708765e-06,
+      "loss": 0.1775,
+      "step": 106330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.574542833978921e-06,
+      "loss": 0.172,
+      "step": 106340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5700387352490768e-06,
+      "loss": 0.1749,
+      "step": 106350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5655346365192326e-06,
+      "loss": 0.1766,
+      "step": 106360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5610305377893884e-06,
+      "loss": 0.1711,
+      "step": 106370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.556526439059544e-06,
+      "loss": 0.1741,
+      "step": 106380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5520223403297004e-06,
+      "loss": 0.1742,
+      "step": 106390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5475182415998558e-06,
+      "loss": 0.1749,
+      "step": 106400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.543014142870012e-06,
+      "loss": 0.1744,
+      "step": 106410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.538510044140168e-06,
+      "loss": 0.1735,
+      "step": 106420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.534005945410323e-06,
+      "loss": 0.1769,
+      "step": 106430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5295018466804794e-06,
+      "loss": 0.1725,
+      "step": 106440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.524997747950635e-06,
+      "loss": 0.1728,
+      "step": 106450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.520493649220791e-06,
+      "loss": 0.1705,
+      "step": 106460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5159895504909468e-06,
+      "loss": 0.1801,
+      "step": 106470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5114854517611026e-06,
+      "loss": 0.1749,
+      "step": 106480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.506981353031259e-06,
+      "loss": 0.1807,
+      "step": 106490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.502477254301414e-06,
+      "loss": 0.1729,
+      "step": 106500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4979731555715704e-06,
+      "loss": 0.1817,
+      "step": 106510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.493469056841726e-06,
+      "loss": 0.1814,
+      "step": 106520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.488964958111882e-06,
+      "loss": 0.179,
+      "step": 106530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.484460859382038e-06,
+      "loss": 0.1714,
+      "step": 106540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4799567606521936e-06,
+      "loss": 0.1734,
+      "step": 106550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4754526619223494e-06,
+      "loss": 0.1832,
+      "step": 106560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.470948563192505e-06,
+      "loss": 0.1764,
+      "step": 106570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.466444464462661e-06,
+      "loss": 0.1782,
+      "step": 106580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.461940365732817e-06,
+      "loss": 0.1773,
+      "step": 106590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.457436267002973e-06,
+      "loss": 0.1735,
+      "step": 106600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4529321682731284e-06,
+      "loss": 0.1765,
+      "step": 106610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4484280695432846e-06,
+      "loss": 0.1768,
+      "step": 106620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4439239708134404e-06,
+      "loss": 0.1705,
+      "step": 106630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.439419872083596e-06,
+      "loss": 0.1817,
+      "step": 106640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.434915773353752e-06,
+      "loss": 0.172,
+      "step": 106650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.430411674623908e-06,
+      "loss": 0.176,
+      "step": 106660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.425907575894064e-06,
+      "loss": 0.1693,
+      "step": 106670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4214034771642194e-06,
+      "loss": 0.174,
+      "step": 106680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4168993784343756e-06,
+      "loss": 0.1752,
+      "step": 106690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4123952797045314e-06,
+      "loss": 0.1753,
+      "step": 106700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4078911809746868e-06,
+      "loss": 0.1775,
+      "step": 106710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.403387082244843e-06,
+      "loss": 0.1743,
+      "step": 106720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.398882983514999e-06,
+      "loss": 0.1757,
+      "step": 106730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3943788847851546e-06,
+      "loss": 0.1713,
+      "step": 106740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3898747860553104e-06,
+      "loss": 0.1785,
+      "step": 106750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.385370687325466e-06,
+      "loss": 0.1768,
+      "step": 106760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3808665885956224e-06,
+      "loss": 0.1731,
+      "step": 106770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.376362489865778e-06,
+      "loss": 0.1718,
+      "step": 106780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3718583911359336e-06,
+      "loss": 0.1707,
+      "step": 106790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.36735429240609e-06,
+      "loss": 0.1824,
+      "step": 106800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3628501936762456e-06,
+      "loss": 0.1732,
+      "step": 106810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3583460949464014e-06,
+      "loss": 0.1761,
+      "step": 106820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.353841996216557e-06,
+      "loss": 0.1782,
+      "step": 106830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.349337897486713e-06,
+      "loss": 0.1733,
+      "step": 106840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.344833798756869e-06,
+      "loss": 0.1742,
+      "step": 106850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3403297000270246e-06,
+      "loss": 0.1731,
+      "step": 106860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.335825601297181e-06,
+      "loss": 0.1699,
+      "step": 106870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3313215025673366e-06,
+      "loss": 0.1753,
+      "step": 106880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.326817403837492e-06,
+      "loss": 0.1753,
+      "step": 106890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3223133051076482e-06,
+      "loss": 0.1719,
+      "step": 106900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.317809206377804e-06,
+      "loss": 0.1717,
+      "step": 106910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.31330510764796e-06,
+      "loss": 0.1741,
+      "step": 106920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3088010089181156e-06,
+      "loss": 0.1755,
+      "step": 106930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3042969101882714e-06,
+      "loss": 0.174,
+      "step": 106940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.299792811458427e-06,
+      "loss": 0.1714,
+      "step": 106950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.295288712728583e-06,
+      "loss": 0.1808,
+      "step": 106960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.290784613998739e-06,
+      "loss": 0.1744,
+      "step": 106970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.286280515268895e-06,
+      "loss": 0.1778,
+      "step": 106980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2817764165390504e-06,
+      "loss": 0.1756,
+      "step": 106990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2772723178092066e-06,
+      "loss": 0.1706,
+      "step": 107000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2727682190793624e-06,
+      "loss": 0.1753,
+      "step": 107010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.268264120349518e-06,
+      "loss": 0.18,
+      "step": 107020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.263760021619674e-06,
+      "loss": 0.1704,
+      "step": 107030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.25925592288983e-06,
+      "loss": 0.1708,
+      "step": 107040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.254751824159986e-06,
+      "loss": 0.1726,
+      "step": 107050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2502477254301414e-06,
+      "loss": 0.1713,
+      "step": 107060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.245743626700297e-06,
+      "loss": 0.1759,
+      "step": 107070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2412395279704534e-06,
+      "loss": 0.1765,
+      "step": 107080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.236735429240609e-06,
+      "loss": 0.1768,
+      "step": 107090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.232231330510765e-06,
+      "loss": 0.1756,
+      "step": 107100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.227727231780921e-06,
+      "loss": 0.1731,
+      "step": 107110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2232231330510766e-06,
+      "loss": 0.1749,
+      "step": 107120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2187190343212324e-06,
+      "loss": 0.1779,
+      "step": 107130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2142149355913882e-06,
+      "loss": 0.1772,
+      "step": 107140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.209710836861544e-06,
+      "loss": 0.1721,
+      "step": 107150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2052067381317e-06,
+      "loss": 0.1758,
+      "step": 107160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2007026394018556e-06,
+      "loss": 0.1744,
+      "step": 107170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.196198540672012e-06,
+      "loss": 0.1724,
+      "step": 107180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1916944419421676e-06,
+      "loss": 0.1692,
+      "step": 107190
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.187190343212323e-06,
+      "loss": 0.1707,
+      "step": 107200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1826862444824792e-06,
+      "loss": 0.1705,
+      "step": 107210
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.178182145752635e-06,
+      "loss": 0.1787,
+      "step": 107220
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.173678047022791e-06,
+      "loss": 0.18,
+      "step": 107230
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1691739482929466e-06,
+      "loss": 0.1719,
+      "step": 107240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1646698495631024e-06,
+      "loss": 0.1772,
+      "step": 107250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1601657508332586e-06,
+      "loss": 0.1747,
+      "step": 107260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.155661652103414e-06,
+      "loss": 0.1766,
+      "step": 107270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1511575533735702e-06,
+      "loss": 0.1773,
+      "step": 107280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.146653454643726e-06,
+      "loss": 0.1743,
+      "step": 107290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1421493559138814e-06,
+      "loss": 0.175,
+      "step": 107300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1376452571840376e-06,
+      "loss": 0.1747,
+      "step": 107310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1331411584541934e-06,
+      "loss": 0.1774,
+      "step": 107320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1286370597243492e-06,
+      "loss": 0.1829,
+      "step": 107330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.124132960994505e-06,
+      "loss": 0.1683,
+      "step": 107340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.119628862264661e-06,
+      "loss": 0.1767,
+      "step": 107350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.115124763534817e-06,
+      "loss": 0.1747,
+      "step": 107360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1106206648049724e-06,
+      "loss": 0.1758,
+      "step": 107370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1061165660751282e-06,
+      "loss": 0.1781,
+      "step": 107380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1016124673452844e-06,
+      "loss": 0.1746,
+      "step": 107390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0971083686154402e-06,
+      "loss": 0.1751,
+      "step": 107400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.092604269885596e-06,
+      "loss": 0.1779,
+      "step": 107410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.088100171155752e-06,
+      "loss": 0.1748,
+      "step": 107420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0835960724259076e-06,
+      "loss": 0.1716,
+      "step": 107430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0790919736960634e-06,
+      "loss": 0.1745,
+      "step": 107440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0745878749662192e-06,
+      "loss": 0.172,
+      "step": 107450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0700837762363755e-06,
+      "loss": 0.1734,
+      "step": 107460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0655796775065313e-06,
+      "loss": 0.1792,
+      "step": 107470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0610755787766866e-06,
+      "loss": 0.1757,
+      "step": 107480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.056571480046843e-06,
+      "loss": 0.1704,
+      "step": 107490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0520673813169986e-06,
+      "loss": 0.1754,
+      "step": 107500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0475632825871544e-06,
+      "loss": 0.1766,
+      "step": 107510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0430591838573102e-06,
+      "loss": 0.1785,
+      "step": 107520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.038555085127466e-06,
+      "loss": 0.1727,
+      "step": 107530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.034050986397622e-06,
+      "loss": 0.1745,
+      "step": 107540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0295468876677776e-06,
+      "loss": 0.1773,
+      "step": 107550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0250427889379334e-06,
+      "loss": 0.1719,
+      "step": 107560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0205386902080897e-06,
+      "loss": 0.1734,
+      "step": 107570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.016034591478245e-06,
+      "loss": 0.1776,
+      "step": 107580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0115304927484013e-06,
+      "loss": 0.1728,
+      "step": 107590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.007026394018557e-06,
+      "loss": 0.1726,
+      "step": 107600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.002522295288713e-06,
+      "loss": 0.1727,
+      "step": 107610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9980181965588686e-06,
+      "loss": 0.1748,
+      "step": 107620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9935140978290244e-06,
+      "loss": 0.175,
+      "step": 107630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9890099990991807e-06,
+      "loss": 0.171,
+      "step": 107640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.984505900369336e-06,
+      "loss": 0.174,
+      "step": 107650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.980001801639492e-06,
+      "loss": 0.1723,
+      "step": 107660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.975497702909648e-06,
+      "loss": 0.1811,
+      "step": 107670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9709936041798034e-06,
+      "loss": 0.1741,
+      "step": 107680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9664895054499597e-06,
+      "loss": 0.1802,
+      "step": 107690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9619854067201155e-06,
+      "loss": 0.1771,
+      "step": 107700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9574813079902713e-06,
+      "loss": 0.1747,
+      "step": 107710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.952977209260427e-06,
+      "loss": 0.1765,
+      "step": 107720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.948473110530583e-06,
+      "loss": 0.1783,
+      "step": 107730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9439690118007386e-06,
+      "loss": 0.1732,
+      "step": 107740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9394649130708944e-06,
+      "loss": 0.1742,
+      "step": 107750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9349608143410502e-06,
+      "loss": 0.1779,
+      "step": 107760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9304567156112065e-06,
+      "loss": 0.1776,
+      "step": 107770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9259526168813623e-06,
+      "loss": 0.1748,
+      "step": 107780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.921448518151518e-06,
+      "loss": 0.1746,
+      "step": 107790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.916944419421674e-06,
+      "loss": 0.174,
+      "step": 107800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9124403206918297e-06,
+      "loss": 0.1766,
+      "step": 107810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9079362219619855e-06,
+      "loss": 0.1742,
+      "step": 107820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9034321232321413e-06,
+      "loss": 0.1746,
+      "step": 107830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8989280245022973e-06,
+      "loss": 0.1762,
+      "step": 107840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8944239257724533e-06,
+      "loss": 0.1762,
+      "step": 107850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8899198270426089e-06,
+      "loss": 0.1746,
+      "step": 107860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8854157283127647e-06,
+      "loss": 0.1701,
+      "step": 107870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8809116295829207e-06,
+      "loss": 0.1821,
+      "step": 107880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8764075308530763e-06,
+      "loss": 0.1745,
+      "step": 107890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8719034321232323e-06,
+      "loss": 0.1744,
+      "step": 107900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.867399333393388e-06,
+      "loss": 0.1717,
+      "step": 107910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.862895234663544e-06,
+      "loss": 0.1715,
+      "step": 107920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8583911359336997e-06,
+      "loss": 0.1751,
+      "step": 107930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8538870372038557e-06,
+      "loss": 0.1783,
+      "step": 107940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8493829384740115e-06,
+      "loss": 0.1764,
+      "step": 107950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.844878839744167e-06,
+      "loss": 0.1756,
+      "step": 107960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.840374741014323e-06,
+      "loss": 0.1737,
+      "step": 107970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.835870642284479e-06,
+      "loss": 0.1782,
+      "step": 107980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8313665435546349e-06,
+      "loss": 0.1784,
+      "step": 107990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8268624448247905e-06,
+      "loss": 0.1739,
+      "step": 108000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8223583460949465e-06,
+      "loss": 0.174,
+      "step": 108010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8178542473651025e-06,
+      "loss": 0.1744,
+      "step": 108020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.813350148635258e-06,
+      "loss": 0.1739,
+      "step": 108030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.808846049905414e-06,
+      "loss": 0.1761,
+      "step": 108040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8043419511755699e-06,
+      "loss": 0.1758,
+      "step": 108050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7998378524457259e-06,
+      "loss": 0.1776,
+      "step": 108060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7953337537158815e-06,
+      "loss": 0.1697,
+      "step": 108070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7908296549860375e-06,
+      "loss": 0.1743,
+      "step": 108080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7863255562561933e-06,
+      "loss": 0.1728,
+      "step": 108090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7818214575263489e-06,
+      "loss": 0.175,
+      "step": 108100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7773173587965049e-06,
+      "loss": 0.1789,
+      "step": 108110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7728132600666609e-06,
+      "loss": 0.1723,
+      "step": 108120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7683091613368167e-06,
+      "loss": 0.1772,
+      "step": 108130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7638050626069723e-06,
+      "loss": 0.1776,
+      "step": 108140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7593009638771283e-06,
+      "loss": 0.1764,
+      "step": 108150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7547968651472843e-06,
+      "loss": 0.1767,
+      "step": 108160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7502927664174399e-06,
+      "loss": 0.1774,
+      "step": 108170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7457886676875957e-06,
+      "loss": 0.1705,
+      "step": 108180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7412845689577517e-06,
+      "loss": 0.1714,
+      "step": 108190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7367804702279073e-06,
+      "loss": 0.1778,
+      "step": 108200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7322763714980633e-06,
+      "loss": 0.1787,
+      "step": 108210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7277722727682193e-06,
+      "loss": 0.1753,
+      "step": 108220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.723268174038375e-06,
+      "loss": 0.1847,
+      "step": 108230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7187640753085307e-06,
+      "loss": 0.1689,
+      "step": 108240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7142599765786867e-06,
+      "loss": 0.1706,
+      "step": 108250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7097558778488427e-06,
+      "loss": 0.1743,
+      "step": 108260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7052517791189983e-06,
+      "loss": 0.1779,
+      "step": 108270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.700747680389154e-06,
+      "loss": 0.1739,
+      "step": 108280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.69624358165931e-06,
+      "loss": 0.1723,
+      "step": 108290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.691739482929466e-06,
+      "loss": 0.1839,
+      "step": 108300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6872353841996217e-06,
+      "loss": 0.1746,
+      "step": 108310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6827312854697775e-06,
+      "loss": 0.1741,
+      "step": 108320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6782271867399335e-06,
+      "loss": 0.1806,
+      "step": 108330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.673723088010089e-06,
+      "loss": 0.1762,
+      "step": 108340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.669218989280245e-06,
+      "loss": 0.1764,
+      "step": 108350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6647148905504009e-06,
+      "loss": 0.1735,
+      "step": 108360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6602107918205569e-06,
+      "loss": 0.1799,
+      "step": 108370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6557066930907125e-06,
+      "loss": 0.1763,
+      "step": 108380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6512025943608685e-06,
+      "loss": 0.1774,
+      "step": 108390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6466984956310245e-06,
+      "loss": 0.17,
+      "step": 108400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.64219439690118e-06,
+      "loss": 0.1753,
+      "step": 108410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6376902981713359e-06,
+      "loss": 0.1762,
+      "step": 108420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6331861994414919e-06,
+      "loss": 0.1768,
+      "step": 108430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.628682100711648e-06,
+      "loss": 0.1663,
+      "step": 108440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6241780019818035e-06,
+      "loss": 0.1734,
+      "step": 108450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6196739032519593e-06,
+      "loss": 0.1765,
+      "step": 108460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6151698045221153e-06,
+      "loss": 0.1808,
+      "step": 108470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6106657057922709e-06,
+      "loss": 0.171,
+      "step": 108480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6061616070624269e-06,
+      "loss": 0.1755,
+      "step": 108490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6016575083325827e-06,
+      "loss": 0.179,
+      "step": 108500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5971534096027387e-06,
+      "loss": 0.175,
+      "step": 108510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5926493108728943e-06,
+      "loss": 0.1762,
+      "step": 108520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5881452121430503e-06,
+      "loss": 0.1771,
+      "step": 108530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.583641113413206e-06,
+      "loss": 0.1756,
+      "step": 108540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.579137014683362e-06,
+      "loss": 0.1769,
+      "step": 108550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5746329159535177e-06,
+      "loss": 0.1751,
+      "step": 108560
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5701288172236737e-06,
+      "loss": 0.1729,
+      "step": 108570
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5656247184938297e-06,
+      "loss": 0.1756,
+      "step": 108580
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5611206197639853e-06,
+      "loss": 0.1695,
+      "step": 108590
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.556616521034141e-06,
+      "loss": 0.1721,
+      "step": 108600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.552112422304297e-06,
+      "loss": 0.1732,
+      "step": 108610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.547608323574453e-06,
+      "loss": 0.1729,
+      "step": 108620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5431042248446087e-06,
+      "loss": 0.1717,
+      "step": 108630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5386001261147645e-06,
+      "loss": 0.1764,
+      "step": 108640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5340960273849203e-06,
+      "loss": 0.177,
+      "step": 108650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.529591928655076e-06,
+      "loss": 0.176,
+      "step": 108660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5250878299252321e-06,
+      "loss": 0.1748,
+      "step": 108670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.520583731195388e-06,
+      "loss": 0.1672,
+      "step": 108680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5160796324655437e-06,
+      "loss": 0.1681,
+      "step": 108690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5115755337356995e-06,
+      "loss": 0.1721,
+      "step": 108700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5070714350058555e-06,
+      "loss": 0.1732,
+      "step": 108710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5025673362760113e-06,
+      "loss": 0.1712,
+      "step": 108720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4980632375461671e-06,
+      "loss": 0.1733,
+      "step": 108730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.493559138816323e-06,
+      "loss": 0.1715,
+      "step": 108740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4890550400864787e-06,
+      "loss": 0.1682,
+      "step": 108750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4845509413566347e-06,
+      "loss": 0.1764,
+      "step": 108760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4800468426267905e-06,
+      "loss": 0.1744,
+      "step": 108770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4755427438969463e-06,
+      "loss": 0.1734,
+      "step": 108780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4710386451671021e-06,
+      "loss": 0.1787,
+      "step": 108790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.466534546437258e-06,
+      "loss": 0.1834,
+      "step": 108800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.462030447707414e-06,
+      "loss": 0.1752,
+      "step": 108810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4575263489775697e-06,
+      "loss": 0.1741,
+      "step": 108820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4530222502477255e-06,
+      "loss": 0.1729,
+      "step": 108830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4485181515178813e-06,
+      "loss": 0.1721,
+      "step": 108840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4440140527880373e-06,
+      "loss": 0.1785,
+      "step": 108850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4395099540581931e-06,
+      "loss": 0.1732,
+      "step": 108860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4350058553283487e-06,
+      "loss": 0.1785,
+      "step": 108870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4305017565985047e-06,
+      "loss": 0.1754,
+      "step": 108880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4259976578686605e-06,
+      "loss": 0.1667,
+      "step": 108890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4214935591388165e-06,
+      "loss": 0.1707,
+      "step": 108900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4169894604089723e-06,
+      "loss": 0.1777,
+      "step": 108910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.412485361679128e-06,
+      "loss": 0.1734,
+      "step": 108920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.407981262949284e-06,
+      "loss": 0.1719,
+      "step": 108930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4034771642194397e-06,
+      "loss": 0.1717,
+      "step": 108940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3989730654895957e-06,
+      "loss": 0.1781,
+      "step": 108950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3944689667597513e-06,
+      "loss": 0.1752,
+      "step": 108960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3899648680299073e-06,
+      "loss": 0.1768,
+      "step": 108970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3854607693000631e-06,
+      "loss": 0.171,
+      "step": 108980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.380956670570219e-06,
+      "loss": 0.1736,
+      "step": 108990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.376452571840375e-06,
+      "loss": 0.1708,
+      "step": 109000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3719484731105305e-06,
+      "loss": 0.173,
+      "step": 109010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3674443743806865e-06,
+      "loss": 0.1769,
+      "step": 109020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3629402756508423e-06,
+      "loss": 0.1703,
+      "step": 109030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3584361769209983e-06,
+      "loss": 0.1757,
+      "step": 109040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353932078191154e-06,
+      "loss": 0.1766,
+      "step": 109050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3494279794613097e-06,
+      "loss": 0.1741,
+      "step": 109060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3449238807314657e-06,
+      "loss": 0.1686,
+      "step": 109070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3404197820016215e-06,
+      "loss": 0.1768,
+      "step": 109080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3359156832717775e-06,
+      "loss": 0.1762,
+      "step": 109090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3314115845419331e-06,
+      "loss": 0.1756,
+      "step": 109100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3269074858120891e-06,
+      "loss": 0.1756,
+      "step": 109110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.322403387082245e-06,
+      "loss": 0.1696,
+      "step": 109120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3178992883524007e-06,
+      "loss": 0.1778,
+      "step": 109130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3133951896225565e-06,
+      "loss": 0.1788,
+      "step": 109140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3088910908927123e-06,
+      "loss": 0.1791,
+      "step": 109150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3043869921628683e-06,
+      "loss": 0.1739,
+      "step": 109160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2998828934330241e-06,
+      "loss": 0.1776,
+      "step": 109170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2953787947031801e-06,
+      "loss": 0.1822,
+      "step": 109180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2908746959733357e-06,
+      "loss": 0.1755,
+      "step": 109190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2863705972434915e-06,
+      "loss": 0.176,
+      "step": 109200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2818664985136475e-06,
+      "loss": 0.1749,
+      "step": 109210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2773623997838033e-06,
+      "loss": 0.175,
+      "step": 109220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2728583010539591e-06,
+      "loss": 0.1765,
+      "step": 109230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.268354202324115e-06,
+      "loss": 0.1716,
+      "step": 109240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2638501035942707e-06,
+      "loss": 0.1758,
+      "step": 109250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2593460048644267e-06,
+      "loss": 0.1734,
+      "step": 109260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2548419061345825e-06,
+      "loss": 0.1703,
+      "step": 109270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2503378074047383e-06,
+      "loss": 0.1741,
+      "step": 109280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2458337086748941e-06,
+      "loss": 0.1759,
+      "step": 109290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2413296099450501e-06,
+      "loss": 0.176,
+      "step": 109300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.236825511215206e-06,
+      "loss": 0.1769,
+      "step": 109310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2323214124853617e-06,
+      "loss": 0.1774,
+      "step": 109320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2278173137555175e-06,
+      "loss": 0.1795,
+      "step": 109330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2233132150256733e-06,
+      "loss": 0.1719,
+      "step": 109340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2188091162958293e-06,
+      "loss": 0.1773,
+      "step": 109350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2143050175659851e-06,
+      "loss": 0.1728,
+      "step": 109360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.209800918836141e-06,
+      "loss": 0.1709,
+      "step": 109370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2052968201062967e-06,
+      "loss": 0.1768,
+      "step": 109380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2007927213764525e-06,
+      "loss": 0.1753,
+      "step": 109390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1962886226466085e-06,
+      "loss": 0.1801,
+      "step": 109400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1917845239167643e-06,
+      "loss": 0.1805,
+      "step": 109410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1872804251869201e-06,
+      "loss": 0.1744,
+      "step": 109420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.182776326457076e-06,
+      "loss": 0.1784,
+      "step": 109430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1787226376002161e-06,
+      "loss": 0.1724,
+      "step": 109440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1742185388703721e-06,
+      "loss": 0.1802,
+      "step": 109450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.169714440140528e-06,
+      "loss": 0.1702,
+      "step": 109460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.165210341410684e-06,
+      "loss": 0.1687,
+      "step": 109470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1607062426808395e-06,
+      "loss": 0.1785,
+      "step": 109480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1562021439509955e-06,
+      "loss": 0.1753,
+      "step": 109490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1516980452211513e-06,
+      "loss": 0.1766,
+      "step": 109500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1471939464913071e-06,
+      "loss": 0.174,
+      "step": 109510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.142689847761463e-06,
+      "loss": 0.1745,
+      "step": 109520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1381857490316187e-06,
+      "loss": 0.1733,
+      "step": 109530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1336816503017747e-06,
+      "loss": 0.1741,
+      "step": 109540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1291775515719305e-06,
+      "loss": 0.1725,
+      "step": 109550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1246734528420863e-06,
+      "loss": 0.1724,
+      "step": 109560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1201693541122421e-06,
+      "loss": 0.1758,
+      "step": 109570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.115665255382398e-06,
+      "loss": 0.1771,
+      "step": 109580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.111161156652554e-06,
+      "loss": 0.1732,
+      "step": 109590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1066570579227097e-06,
+      "loss": 0.1808,
+      "step": 109600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1021529591928655e-06,
+      "loss": 0.1731,
+      "step": 109610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0976488604630213e-06,
+      "loss": 0.1723,
+      "step": 109620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0931447617331771e-06,
+      "loss": 0.1711,
+      "step": 109630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0886406630033331e-06,
+      "loss": 0.1704,
+      "step": 109640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.084136564273489e-06,
+      "loss": 0.1712,
+      "step": 109650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0796324655436447e-06,
+      "loss": 0.1663,
+      "step": 109660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0751283668138005e-06,
+      "loss": 0.1739,
+      "step": 109670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0706242680839565e-06,
+      "loss": 0.1734,
+      "step": 109680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0661201693541123e-06,
+      "loss": 0.1727,
+      "step": 109690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0616160706242681e-06,
+      "loss": 0.172,
+      "step": 109700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.057111971894424e-06,
+      "loss": 0.1784,
+      "step": 109710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0526078731645797e-06,
+      "loss": 0.1755,
+      "step": 109720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0481037744347357e-06,
+      "loss": 0.1789,
+      "step": 109730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0435996757048915e-06,
+      "loss": 0.1778,
+      "step": 109740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0390955769750473e-06,
+      "loss": 0.1735,
+      "step": 109750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0345914782452031e-06,
+      "loss": 0.1778,
+      "step": 109760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.030087379515359e-06,
+      "loss": 0.1635,
+      "step": 109770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.025583280785515e-06,
+      "loss": 0.1769,
+      "step": 109780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0210791820556707e-06,
+      "loss": 0.1722,
+      "step": 109790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0165750833258265e-06,
+      "loss": 0.1655,
+      "step": 109800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0120709845959823e-06,
+      "loss": 0.1635,
+      "step": 109810
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0075668858661381e-06,
+      "loss": 0.1718,
+      "step": 109820
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0030627871362941e-06,
+      "loss": 0.174,
+      "step": 109830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.9855868840645e-07,
+      "loss": 0.1737,
+      "step": 109840
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.940545896766057e-07,
+      "loss": 0.175,
+      "step": 109850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.895504909467615e-07,
+      "loss": 0.1736,
+      "step": 109860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.850463922169176e-07,
+      "loss": 0.1764,
+      "step": 109870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.805422934870733e-07,
+      "loss": 0.1799,
+      "step": 109880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.760381947572291e-07,
+      "loss": 0.1747,
+      "step": 109890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.71534096027385e-07,
+      "loss": 0.173,
+      "step": 109900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.670299972975407e-07,
+      "loss": 0.176,
+      "step": 109910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.625258985676968e-07,
+      "loss": 0.1785,
+      "step": 109920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.580217998378526e-07,
+      "loss": 0.1725,
+      "step": 109930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.535177011080085e-07,
+      "loss": 0.1704,
+      "step": 109940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.490136023781641e-07,
+      "loss": 0.1812,
+      "step": 109950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.445095036483199e-07,
+      "loss": 0.1737,
+      "step": 109960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.400054049184759e-07,
+      "loss": 0.1744,
+      "step": 109970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.355013061886316e-07,
+      "loss": 0.1771,
+      "step": 109980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.309972074587876e-07,
+      "loss": 0.1756,
+      "step": 109990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.264931087289434e-07,
+      "loss": 0.1764,
+      "step": 110000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.219890099990993e-07,
+      "loss": 0.176,
+      "step": 110010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.17484911269255e-07,
+      "loss": 0.1794,
+      "step": 110020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.129808125394109e-07,
+      "loss": 0.1708,
+      "step": 110030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.084767138095668e-07,
+      "loss": 0.1685,
+      "step": 110040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.039726150797226e-07,
+      "loss": 0.1778,
+      "step": 110050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.994685163498785e-07,
+      "loss": 0.1751,
+      "step": 110060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.949644176200343e-07,
+      "loss": 0.1783,
+      "step": 110070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.904603188901902e-07,
+      "loss": 0.1731,
+      "step": 110080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.85956220160346e-07,
+      "loss": 0.1686,
+      "step": 110090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.814521214305018e-07,
+      "loss": 0.1728,
+      "step": 110100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.769480227006577e-07,
+      "loss": 0.1749,
+      "step": 110110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.724439239708135e-07,
+      "loss": 0.1722,
+      "step": 110120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.679398252409694e-07,
+      "loss": 0.1719,
+      "step": 110130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.634357265111252e-07,
+      "loss": 0.1732,
+      "step": 110140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.58931627781281e-07,
+      "loss": 0.1705,
+      "step": 110150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.544275290514369e-07,
+      "loss": 0.1764,
+      "step": 110160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.499234303215927e-07,
+      "loss": 0.1708,
+      "step": 110170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.454193315917486e-07,
+      "loss": 0.1666,
+      "step": 110180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.409152328619044e-07,
+      "loss": 0.1813,
+      "step": 110190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.364111341320603e-07,
+      "loss": 0.171,
+      "step": 110200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.319070354022161e-07,
+      "loss": 0.1749,
+      "step": 110210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.274029366723719e-07,
+      "loss": 0.1761,
+      "step": 110220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.228988379425278e-07,
+      "loss": 0.1778,
+      "step": 110230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.183947392126836e-07,
+      "loss": 0.1673,
+      "step": 110240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.138906404828395e-07,
+      "loss": 0.1733,
+      "step": 110250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.093865417529953e-07,
+      "loss": 0.1745,
+      "step": 110260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.048824430231512e-07,
+      "loss": 0.1803,
+      "step": 110270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.00378344293307e-07,
+      "loss": 0.1824,
+      "step": 110280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.958742455634628e-07,
+      "loss": 0.1752,
+      "step": 110290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.913701468336187e-07,
+      "loss": 0.1792,
+      "step": 110300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.868660481037745e-07,
+      "loss": 0.1731,
+      "step": 110310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.823619493739304e-07,
+      "loss": 0.1738,
+      "step": 110320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.778578506440862e-07,
+      "loss": 0.1771,
+      "step": 110330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.73353751914242e-07,
+      "loss": 0.1681,
+      "step": 110340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.688496531843979e-07,
+      "loss": 0.1764,
+      "step": 110350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.643455544545537e-07,
+      "loss": 0.1718,
+      "step": 110360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.598414557247096e-07,
+      "loss": 0.1703,
+      "step": 110370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.553373569948654e-07,
+      "loss": 0.1765,
+      "step": 110380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.508332582650212e-07,
+      "loss": 0.176,
+      "step": 110390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.463291595351771e-07,
+      "loss": 0.1742,
+      "step": 110400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.418250608053329e-07,
+      "loss": 0.1736,
+      "step": 110410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.373209620754888e-07,
+      "loss": 0.1749,
+      "step": 110420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.328168633456446e-07,
+      "loss": 0.1701,
+      "step": 110430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.283127646158005e-07,
+      "loss": 0.1767,
+      "step": 110440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.238086658859562e-07,
+      "loss": 0.1728,
+      "step": 110450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.193045671561121e-07,
+      "loss": 0.1782,
+      "step": 110460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.14800468426268e-07,
+      "loss": 0.1733,
+      "step": 110470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.102963696964238e-07,
+      "loss": 0.1764,
+      "step": 110480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.057922709665797e-07,
+      "loss": 0.177,
+      "step": 110490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.012881722367355e-07,
+      "loss": 0.169,
+      "step": 110500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.967840735068913e-07,
+      "loss": 0.1783,
+      "step": 110510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.922799747770471e-07,
+      "loss": 0.1767,
+      "step": 110520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.87775876047203e-07,
+      "loss": 0.1748,
+      "step": 110530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.832717773173588e-07,
+      "loss": 0.1712,
+      "step": 110540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.787676785875147e-07,
+      "loss": 0.1716,
+      "step": 110550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.742635798576706e-07,
+      "loss": 0.175,
+      "step": 110560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.697594811278264e-07,
+      "loss": 0.1754,
+      "step": 110570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.652553823979822e-07,
+      "loss": 0.1784,
+      "step": 110580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.60751283668138e-07,
+      "loss": 0.1717,
+      "step": 110590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.562471849382939e-07,
+      "loss": 0.1768,
+      "step": 110600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.517430862084497e-07,
+      "loss": 0.1772,
+      "step": 110610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.472389874786056e-07,
+      "loss": 0.1715,
+      "step": 110620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.427348887487614e-07,
+      "loss": 0.1739,
+      "step": 110630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.382307900189172e-07,
+      "loss": 0.1784,
+      "step": 110640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.337266912890731e-07,
+      "loss": 0.1755,
+      "step": 110650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.292225925592289e-07,
+      "loss": 0.1771,
+      "step": 110660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.247184938293848e-07,
+      "loss": 0.1683,
+      "step": 110670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.202143950995406e-07,
+      "loss": 0.176,
+      "step": 110680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.157102963696965e-07,
+      "loss": 0.1706,
+      "step": 110690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.112061976398523e-07,
+      "loss": 0.1759,
+      "step": 110700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.067020989100081e-07,
+      "loss": 0.1764,
+      "step": 110710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.02198000180164e-07,
+      "loss": 0.1797,
+      "step": 110720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.976939014503198e-07,
+      "loss": 0.1658,
+      "step": 110730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.931898027204757e-07,
+      "loss": 0.178,
+      "step": 110740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.886857039906315e-07,
+      "loss": 0.1756,
+      "step": 110750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.841816052607874e-07,
+      "loss": 0.1763,
+      "step": 110760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.796775065309432e-07,
+      "loss": 0.1778,
+      "step": 110770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.75173407801099e-07,
+      "loss": 0.1733,
+      "step": 110780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.706693090712549e-07,
+      "loss": 0.1725,
+      "step": 110790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.661652103414107e-07,
+      "loss": 0.1747,
+      "step": 110800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.616611116115666e-07,
+      "loss": 0.1716,
+      "step": 110810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.571570128817224e-07,
+      "loss": 0.1714,
+      "step": 110820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.526529141518783e-07,
+      "loss": 0.1681,
+      "step": 110830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.48148815422034e-07,
+      "loss": 0.1752,
+      "step": 110840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.436447166921899e-07,
+      "loss": 0.1721,
+      "step": 110850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.391406179623458e-07,
+      "loss": 0.173,
+      "step": 110860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.346365192325016e-07,
+      "loss": 0.1741,
+      "step": 110870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.301324205026575e-07,
+      "loss": 0.181,
+      "step": 110880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.256283217728133e-07,
+      "loss": 0.1715,
+      "step": 110890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.211242230429692e-07,
+      "loss": 0.1764,
+      "step": 110900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.166201243131249e-07,
+      "loss": 0.1756,
+      "step": 110910
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.121160255832808e-07,
+      "loss": 0.1811,
+      "step": 110920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.076119268534366e-07,
+      "loss": 0.1703,
+      "step": 110930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.031078281235925e-07,
+      "loss": 0.1733,
+      "step": 110940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.986037293937484e-07,
+      "loss": 0.1743,
+      "step": 110950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.940996306639042e-07,
+      "loss": 0.1727,
+      "step": 110960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8959553193406e-07,
+      "loss": 0.1749,
+      "step": 110970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.850914332042158e-07,
+      "loss": 0.1707,
+      "step": 110980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.805873344743717e-07,
+      "loss": 0.1768,
+      "step": 110990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7608323574452756e-07,
+      "loss": 0.174,
+      "step": 111000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.715791370146834e-07,
+      "loss": 0.1765,
+      "step": 111010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6707503828483926e-07,
+      "loss": 0.177,
+      "step": 111020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.625709395549951e-07,
+      "loss": 0.1735,
+      "step": 111030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5806684082515085e-07,
+      "loss": 0.1758,
+      "step": 111040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.535627420953067e-07,
+      "loss": 0.174,
+      "step": 111050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.490586433654626e-07,
+      "loss": 0.1782,
+      "step": 111060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4455454463561846e-07,
+      "loss": 0.1797,
+      "step": 111070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.400504459057743e-07,
+      "loss": 0.1744,
+      "step": 111080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3554634717593016e-07,
+      "loss": 0.172,
+      "step": 111090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.310422484460859e-07,
+      "loss": 0.1732,
+      "step": 111100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2653814971624176e-07,
+      "loss": 0.1754,
+      "step": 111110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.220340509863976e-07,
+      "loss": 0.1732,
+      "step": 111120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1752995225655346e-07,
+      "loss": 0.1726,
+      "step": 111130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.130258535267093e-07,
+      "loss": 0.1745,
+      "step": 111140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.085217547968652e-07,
+      "loss": 0.1749,
+      "step": 111150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0401765606702107e-07,
+      "loss": 0.1781,
+      "step": 111160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.995135573371768e-07,
+      "loss": 0.1734,
+      "step": 111170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9500945860733266e-07,
+      "loss": 0.1724,
+      "step": 111180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.905053598774885e-07,
+      "loss": 0.1771,
+      "step": 111190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8600126114764436e-07,
+      "loss": 0.1731,
+      "step": 111200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.814971624178002e-07,
+      "loss": 0.1678,
+      "step": 111210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7699306368795607e-07,
+      "loss": 0.1759,
+      "step": 111220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.724889649581119e-07,
+      "loss": 0.1794,
+      "step": 111230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6798486622826777e-07,
+      "loss": 0.1769,
+      "step": 111240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6348076749842357e-07,
+      "loss": 0.1762,
+      "step": 111250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.589766687685794e-07,
+      "loss": 0.1741,
+      "step": 111260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5447257003873527e-07,
+      "loss": 0.1723,
+      "step": 111270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4996847130889107e-07,
+      "loss": 0.1778,
+      "step": 111280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.454643725790469e-07,
+      "loss": 0.1814,
+      "step": 111290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.409602738492028e-07,
+      "loss": 0.1765,
+      "step": 111300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3645617511935867e-07,
+      "loss": 0.1758,
+      "step": 111310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3195207638951447e-07,
+      "loss": 0.1739,
+      "step": 111320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.274479776596703e-07,
+      "loss": 0.1771,
+      "step": 111330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2294387892982617e-07,
+      "loss": 0.1778,
+      "step": 111340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1843978019998197e-07,
+      "loss": 0.1727,
+      "step": 111350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.139356814701378e-07,
+      "loss": 0.1774,
+      "step": 111360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.094315827402937e-07,
+      "loss": 0.173,
+      "step": 111370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.049274840104495e-07,
+      "loss": 0.1745,
+      "step": 111380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.004233852806054e-07,
+      "loss": 0.1757,
+      "step": 111390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9591928655076123e-07,
+      "loss": 0.1796,
+      "step": 111400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.914151878209171e-07,
+      "loss": 0.1683,
+      "step": 111410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.869110890910729e-07,
+      "loss": 0.1757,
+      "step": 111420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8240699036122873e-07,
+      "loss": 0.1814,
+      "step": 111430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.779028916313846e-07,
+      "loss": 0.1723,
+      "step": 111440
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.733987929015404e-07,
+      "loss": 0.174,
+      "step": 111450
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.688946941716963e-07,
+      "loss": 0.1738,
+      "step": 111460
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6439059544185213e-07,
+      "loss": 0.169,
+      "step": 111470
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5988649671200793e-07,
+      "loss": 0.172,
+      "step": 111480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.553823979821638e-07,
+      "loss": 0.177,
+      "step": 111490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5087829925231963e-07,
+      "loss": 0.1788,
+      "step": 111500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4637420052247543e-07,
+      "loss": 0.1758,
+      "step": 111510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.418701017926313e-07,
+      "loss": 0.1747,
+      "step": 111520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3736600306278716e-07,
+      "loss": 0.1759,
+      "step": 111530
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.32861904332943e-07,
+      "loss": 0.1739,
+      "step": 111540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.283578056030988e-07,
+      "loss": 0.174,
+      "step": 111550
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2385370687325468e-07,
+      "loss": 0.1739,
+      "step": 111560
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1934960814341054e-07,
+      "loss": 0.1717,
+      "step": 111570
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1484550941356633e-07,
+      "loss": 0.1775,
+      "step": 111580
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1034141068372218e-07,
+      "loss": 0.1821,
+      "step": 111590
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0583731195387806e-07,
+      "loss": 0.1766,
+      "step": 111600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0133321322403386e-07,
+      "loss": 0.1834,
+      "step": 111610
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.968291144941897e-07,
+      "loss": 0.1749,
+      "step": 111620
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9232501576434556e-07,
+      "loss": 0.1676,
+      "step": 111630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8782091703450141e-07,
+      "loss": 0.1719,
+      "step": 111640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8331681830465724e-07,
+      "loss": 0.1681,
+      "step": 111650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.788127195748131e-07,
+      "loss": 0.1734,
+      "step": 111660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7430862084496891e-07,
+      "loss": 0.1816,
+      "step": 111670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.698045221151248e-07,
+      "loss": 0.1786,
+      "step": 111680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6530042338528062e-07,
+      "loss": 0.1774,
+      "step": 111690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6079632465543644e-07,
+      "loss": 0.1697,
+      "step": 111700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.562922259255923e-07,
+      "loss": 0.1793,
+      "step": 111710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5178812719574814e-07,
+      "loss": 0.1763,
+      "step": 111720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.47284028465904e-07,
+      "loss": 0.177,
+      "step": 111730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4277992973605982e-07,
+      "loss": 0.1744,
+      "step": 111740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3827583100621567e-07,
+      "loss": 0.1711,
+      "step": 111750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3377173227637152e-07,
+      "loss": 0.18,
+      "step": 111760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2926763354652734e-07,
+      "loss": 0.1758,
+      "step": 111770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2476353481668317e-07,
+      "loss": 0.1696,
+      "step": 111780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2025943608683902e-07,
+      "loss": 0.1767,
+      "step": 111790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1575533735699487e-07,
+      "loss": 0.1738,
+      "step": 111800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1125123862715072e-07,
+      "loss": 0.1728,
+      "step": 111810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0674713989730655e-07,
+      "loss": 0.1736,
+      "step": 111820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0224304116746239e-07,
+      "loss": 0.1759,
+      "step": 111830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.773894243761824e-08,
+      "loss": 0.177,
+      "step": 111840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.323484370777407e-08,
+      "loss": 0.1823,
+      "step": 111850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.873074497792991e-08,
+      "loss": 0.1713,
+      "step": 111860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.422664624808576e-08,
+      "loss": 0.1751,
+      "step": 111870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.97225475182416e-08,
+      "loss": 0.1753,
+      "step": 111880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.521844878839745e-08,
+      "loss": 0.1681,
+      "step": 111890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.071435005855329e-08,
+      "loss": 0.1769,
+      "step": 111900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.621025132870913e-08,
+      "loss": 0.1796,
+      "step": 111910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.170615259886497e-08,
+      "loss": 0.1766,
+      "step": 111920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.7202053869020816e-08,
+      "loss": 0.1833,
+      "step": 111930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.269795513917665e-08,
+      "loss": 0.1735,
+      "step": 111940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.819385640933249e-08,
+      "loss": 0.1722,
+      "step": 111950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3689757679488336e-08,
+      "loss": 0.1766,
+      "step": 111960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.918565894964418e-08,
+      "loss": 0.179,
+      "step": 111970
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.468156021980002e-08,
+      "loss": 0.1757,
+      "step": 111980
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.017746148995586e-08,
+      "loss": 0.1714,
+      "step": 111990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.56733627601117e-08,
+      "loss": 0.1766,
+      "step": 112000
+    },
+    {
+      "epoch": 1.0,
+      "eval_NEIMS_canon_smiles": 0.9432666666666667,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.07473333333333333,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.4783567710399878,
+      "eval_NEIMS_exact_mols": 0.07451666666666666,
+      "eval_NEIMS_exact_smiles": 0.0725,
+      "eval_NEIMS_loss": 0.20215021073818207,
+      "eval_NEIMS_matched_formulas": 0.6087833333333333,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.38383660502114564,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.0749,
+      "eval_NEIMS_runtime": 711.5776,
+      "eval_NEIMS_samples_per_second": 84.32,
+      "eval_NEIMS_steps_per_second": 1.318,
+      "step": 112000
+    },
+    {
+      "epoch": 1.0,
+      "eval_RASSP_canon_smiles": 0.9381574114268542,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.17480882893111163,
+      "eval_RASSP_daylight_tanimoto_simil": 0.6164718810327274,
+      "eval_RASSP_exact_mols": 0.1740575015861355,
+      "eval_RASSP_exact_smiles": 0.16873142551841586,
+      "eval_RASSP_loss": 0.1408979892730713,
+      "eval_RASSP_matched_formulas": 0.7876414999833038,
+      "eval_RASSP_morgan_tanimoto_simil": 0.49130529407527207,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.1742411593815741,
+      "eval_RASSP_runtime": 817.2133,
+      "eval_RASSP_samples_per_second": 73.291,
+      "eval_RASSP_steps_per_second": 1.145,
+      "step": 112000
+    },
+    {
+      "epoch": 1.0,
+      "eval_NIST_canon_smiles": 0.8848706391737943,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.014870284274408205,
+      "eval_NIST_daylight_tanimoto_simil": 0.27571158259102035,
+      "eval_NIST_exact_mols": 0.014124995563757676,
+      "eval_NIST_exact_smiles": 0.013202257160095113,
+      "eval_NIST_loss": 1.4582798480987549,
+      "eval_NIST_matched_formulas": 0.09571636441068956,
+      "eval_NIST_morgan_tanimoto_simil": 0.22850929188519337,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.014799304397203393,
+      "eval_NIST_runtime": 457.2134,
+      "eval_NIST_samples_per_second": 61.628,
+      "eval_NIST_steps_per_second": 0.965,
+      "step": 112000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.512151921438501e-05,
+      "loss": 0.1783,
+      "step": 112010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5119277162459082e-05,
+      "loss": 0.1825,
+      "step": 112020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.511703511053316e-05,
+      "loss": 0.1838,
+      "step": 112030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.511479305860724e-05,
+      "loss": 0.1909,
+      "step": 112040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5112551006681318e-05,
+      "loss": 0.1775,
+      "step": 112050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5110308954755397e-05,
+      "loss": 0.188,
+      "step": 112060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.510806690282947e-05,
+      "loss": 0.1879,
+      "step": 112070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5105824850903547e-05,
+      "loss": 0.1831,
+      "step": 112080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5103582798977626e-05,
+      "loss": 0.1938,
+      "step": 112090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5101340747051705e-05,
+      "loss": 0.1872,
+      "step": 112100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5099098695125777e-05,
+      "loss": 0.1933,
+      "step": 112110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5096856643199855e-05,
+      "loss": 0.1944,
+      "step": 112120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5094614591273934e-05,
+      "loss": 0.1942,
+      "step": 112130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5092372539348013e-05,
+      "loss": 0.1931,
+      "step": 112140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.509013048742209e-05,
+      "loss": 0.1868,
+      "step": 112150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5087888435496164e-05,
+      "loss": 0.189,
+      "step": 112160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5085646383570242e-05,
+      "loss": 0.1844,
+      "step": 112170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.508340433164432e-05,
+      "loss": 0.1913,
+      "step": 112180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.50811622797184e-05,
+      "loss": 0.1951,
+      "step": 112190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.507892022779248e-05,
+      "loss": 0.1938,
+      "step": 112200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5076678175866554e-05,
+      "loss": 0.1935,
+      "step": 112210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5074436123940632e-05,
+      "loss": 0.1912,
+      "step": 112220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.507219407201471e-05,
+      "loss": 0.1888,
+      "step": 112230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.506995202008879e-05,
+      "loss": 0.1921,
+      "step": 112240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5067709968162862e-05,
+      "loss": 0.1907,
+      "step": 112250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.506546791623694e-05,
+      "loss": 0.1887,
+      "step": 112260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.506322586431102e-05,
+      "loss": 0.1851,
+      "step": 112270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5060983812385098e-05,
+      "loss": 0.1924,
+      "step": 112280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5058741760459177e-05,
+      "loss": 0.1907,
+      "step": 112290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.505649970853325e-05,
+      "loss": 0.1942,
+      "step": 112300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5054257656607327e-05,
+      "loss": 0.1913,
+      "step": 112310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5052015604681406e-05,
+      "loss": 0.1896,
+      "step": 112320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5049773552755485e-05,
+      "loss": 0.1879,
+      "step": 112330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5047531500829564e-05,
+      "loss": 0.1908,
+      "step": 112340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5045289448903635e-05,
+      "loss": 0.1867,
+      "step": 112350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5043047396977714e-05,
+      "loss": 0.1942,
+      "step": 112360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5040805345051793e-05,
+      "loss": 0.1898,
+      "step": 112370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.503856329312587e-05,
+      "loss": 0.1938,
+      "step": 112380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5036321241199944e-05,
+      "loss": 0.1878,
+      "step": 112390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5034079189274022e-05,
+      "loss": 0.1905,
+      "step": 112400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.50318371373481e-05,
+      "loss": 0.1902,
+      "step": 112410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.502959508542218e-05,
+      "loss": 0.1877,
+      "step": 112420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.502735303349626e-05,
+      "loss": 0.1965,
+      "step": 112430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5025110981570334e-05,
+      "loss": 0.1977,
+      "step": 112440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.502286892964441e-05,
+      "loss": 0.1858,
+      "step": 112450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5020626877718488e-05,
+      "loss": 0.1936,
+      "step": 112460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5018384825792567e-05,
+      "loss": 0.186,
+      "step": 112470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5016142773866645e-05,
+      "loss": 0.1916,
+      "step": 112480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.501390072194072e-05,
+      "loss": 0.191,
+      "step": 112490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.50116586700148e-05,
+      "loss": 0.1903,
+      "step": 112500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5009416618088878e-05,
+      "loss": 0.1965,
+      "step": 112510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5007174566162957e-05,
+      "loss": 0.1918,
+      "step": 112520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.500493251423703e-05,
+      "loss": 0.1967,
+      "step": 112530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5002690462311107e-05,
+      "loss": 0.1908,
+      "step": 112540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5000448410385186e-05,
+      "loss": 0.1914,
+      "step": 112550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4998206358459265e-05,
+      "loss": 0.1947,
+      "step": 112560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.499596430653334e-05,
+      "loss": 0.1924,
+      "step": 112570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.499372225460742e-05,
+      "loss": 0.1873,
+      "step": 112580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4991480202681494e-05,
+      "loss": 0.1933,
+      "step": 112590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4989238150755573e-05,
+      "loss": 0.1879,
+      "step": 112600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.498699609882965e-05,
+      "loss": 0.1928,
+      "step": 112610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4984754046903727e-05,
+      "loss": 0.1926,
+      "step": 112620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4982511994977806e-05,
+      "loss": 0.1954,
+      "step": 112630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.498026994305188e-05,
+      "loss": 0.1932,
+      "step": 112640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.497802789112596e-05,
+      "loss": 0.1852,
+      "step": 112650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4975785839200035e-05,
+      "loss": 0.1891,
+      "step": 112660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4973543787274114e-05,
+      "loss": 0.1865,
+      "step": 112670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4971301735348193e-05,
+      "loss": 0.1922,
+      "step": 112680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4969059683422268e-05,
+      "loss": 0.1924,
+      "step": 112690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4966817631496347e-05,
+      "loss": 0.1909,
+      "step": 112700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4964575579570422e-05,
+      "loss": 0.1895,
+      "step": 112710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.49623335276445e-05,
+      "loss": 0.195,
+      "step": 112720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.496009147571858e-05,
+      "loss": 0.1904,
+      "step": 112730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4957849423792655e-05,
+      "loss": 0.1898,
+      "step": 112740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4955607371866733e-05,
+      "loss": 0.1909,
+      "step": 112750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4953365319940812e-05,
+      "loss": 0.1872,
+      "step": 112760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.495112326801489e-05,
+      "loss": 0.1877,
+      "step": 112770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4948881216088966e-05,
+      "loss": 0.1908,
+      "step": 112780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4946639164163045e-05,
+      "loss": 0.1866,
+      "step": 112790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.494439711223712e-05,
+      "loss": 0.1918,
+      "step": 112800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.49421550603112e-05,
+      "loss": 0.1908,
+      "step": 112810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4939913008385278e-05,
+      "loss": 0.1844,
+      "step": 112820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4937670956459353e-05,
+      "loss": 0.1958,
+      "step": 112830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.493542890453343e-05,
+      "loss": 0.191,
+      "step": 112840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4933186852607507e-05,
+      "loss": 0.1894,
+      "step": 112850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4930944800681586e-05,
+      "loss": 0.1896,
+      "step": 112860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.492870274875566e-05,
+      "loss": 0.19,
+      "step": 112870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.492646069682974e-05,
+      "loss": 0.1892,
+      "step": 112880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.492421864490382e-05,
+      "loss": 0.1851,
+      "step": 112890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4921976592977894e-05,
+      "loss": 0.1897,
+      "step": 112900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4919734541051973e-05,
+      "loss": 0.1871,
+      "step": 112910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4917492489126048e-05,
+      "loss": 0.1886,
+      "step": 112920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4915250437200127e-05,
+      "loss": 0.1909,
+      "step": 112930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4913008385274202e-05,
+      "loss": 0.19,
+      "step": 112940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.491076633334828e-05,
+      "loss": 0.1938,
+      "step": 112950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.490852428142236e-05,
+      "loss": 0.1914,
+      "step": 112960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4906282229496435e-05,
+      "loss": 0.186,
+      "step": 112970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4904040177570513e-05,
+      "loss": 0.1844,
+      "step": 112980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.490179812564459e-05,
+      "loss": 0.1922,
+      "step": 112990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4899556073718667e-05,
+      "loss": 0.193,
+      "step": 113000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4897314021792746e-05,
+      "loss": 0.192,
+      "step": 113010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4895071969866825e-05,
+      "loss": 0.1938,
+      "step": 113020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4892829917940904e-05,
+      "loss": 0.1918,
+      "step": 113030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.489058786601498e-05,
+      "loss": 0.1888,
+      "step": 113040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4888345814089058e-05,
+      "loss": 0.1916,
+      "step": 113050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4886103762163133e-05,
+      "loss": 0.1862,
+      "step": 113060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.488386171023721e-05,
+      "loss": 0.1894,
+      "step": 113070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4881619658311287e-05,
+      "loss": 0.1868,
+      "step": 113080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4879377606385366e-05,
+      "loss": 0.1855,
+      "step": 113090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.4877135554459444e-05,
+      "loss": 0.185,
+      "step": 113100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.487489350253352e-05,
+      "loss": 0.1881,
+      "step": 113110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.48726514506076e-05,
+      "loss": 0.1874,
+      "step": 113120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4870409398681674e-05,
+      "loss": 0.1918,
+      "step": 113130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4868167346755753e-05,
+      "loss": 0.1848,
+      "step": 113140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4865925294829828e-05,
+      "loss": 0.1896,
+      "step": 113150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4863683242903907e-05,
+      "loss": 0.1881,
+      "step": 113160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4861441190977985e-05,
+      "loss": 0.1885,
+      "step": 113170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.485919913905206e-05,
+      "loss": 0.1866,
+      "step": 113180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.485695708712614e-05,
+      "loss": 0.1845,
+      "step": 113190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4854715035200215e-05,
+      "loss": 0.1839,
+      "step": 113200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4852472983274293e-05,
+      "loss": 0.1851,
+      "step": 113210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.485023093134837e-05,
+      "loss": 0.1932,
+      "step": 113220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4847988879422447e-05,
+      "loss": 0.1924,
+      "step": 113230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4845746827496526e-05,
+      "loss": 0.1816,
+      "step": 113240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.48435047755706e-05,
+      "loss": 0.1942,
+      "step": 113250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.484126272364468e-05,
+      "loss": 0.1851,
+      "step": 113260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.483902067171876e-05,
+      "loss": 0.188,
+      "step": 113270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4836778619792834e-05,
+      "loss": 0.1842,
+      "step": 113280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4834536567866913e-05,
+      "loss": 0.1922,
+      "step": 113290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4832294515940992e-05,
+      "loss": 0.1902,
+      "step": 113300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.483005246401507e-05,
+      "loss": 0.1886,
+      "step": 113310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4827810412089146e-05,
+      "loss": 0.1854,
+      "step": 113320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4825568360163224e-05,
+      "loss": 0.1813,
+      "step": 113330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.48233263082373e-05,
+      "loss": 0.1875,
+      "step": 113340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.482108425631138e-05,
+      "loss": 0.1926,
+      "step": 113350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4818842204385454e-05,
+      "loss": 0.188,
+      "step": 113360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4816600152459533e-05,
+      "loss": 0.1868,
+      "step": 113370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.481435810053361e-05,
+      "loss": 0.1827,
+      "step": 113380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4812116048607687e-05,
+      "loss": 0.1909,
+      "step": 113390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4809873996681765e-05,
+      "loss": 0.1865,
+      "step": 113400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.480763194475584e-05,
+      "loss": 0.1846,
+      "step": 113410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.480538989282992e-05,
+      "loss": 0.1866,
+      "step": 113420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4803147840903995e-05,
+      "loss": 0.1907,
+      "step": 113430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4800905788978073e-05,
+      "loss": 0.1863,
+      "step": 113440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4798663737052152e-05,
+      "loss": 0.1914,
+      "step": 113450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4796421685126227e-05,
+      "loss": 0.1873,
+      "step": 113460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4794179633200306e-05,
+      "loss": 0.189,
+      "step": 113470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.479193758127438e-05,
+      "loss": 0.1872,
+      "step": 113480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.478969552934846e-05,
+      "loss": 0.1899,
+      "step": 113490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4787453477422536e-05,
+      "loss": 0.1848,
+      "step": 113500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4785211425496614e-05,
+      "loss": 0.1846,
+      "step": 113510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4782969373570693e-05,
+      "loss": 0.188,
+      "step": 113520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.478072732164477e-05,
+      "loss": 0.193,
+      "step": 113530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4778485269718847e-05,
+      "loss": 0.1886,
+      "step": 113540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4776243217792926e-05,
+      "loss": 0.1943,
+      "step": 113550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4774001165867004e-05,
+      "loss": 0.19,
+      "step": 113560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.477175911394108e-05,
+      "loss": 0.1921,
+      "step": 113570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.476951706201516e-05,
+      "loss": 0.1901,
+      "step": 113580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4767275010089237e-05,
+      "loss": 0.1941,
+      "step": 113590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4765032958163313e-05,
+      "loss": 0.1859,
+      "step": 113600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.476279090623739e-05,
+      "loss": 0.1876,
+      "step": 113610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4760548854311467e-05,
+      "loss": 0.1896,
+      "step": 113620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4758306802385545e-05,
+      "loss": 0.1924,
+      "step": 113630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.475606475045962e-05,
+      "loss": 0.1858,
+      "step": 113640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.47538226985337e-05,
+      "loss": 0.186,
+      "step": 113650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4751580646607778e-05,
+      "loss": 0.1876,
+      "step": 113660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4749338594681853e-05,
+      "loss": 0.1889,
+      "step": 113670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4747096542755932e-05,
+      "loss": 0.1932,
+      "step": 113680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4744854490830007e-05,
+      "loss": 0.1837,
+      "step": 113690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4742612438904086e-05,
+      "loss": 0.1836,
+      "step": 113700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.474037038697816e-05,
+      "loss": 0.1916,
+      "step": 113710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.473812833505224e-05,
+      "loss": 0.1857,
+      "step": 113720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.473588628312632e-05,
+      "loss": 0.1857,
+      "step": 113730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4733644231200394e-05,
+      "loss": 0.1867,
+      "step": 113740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4731402179274473e-05,
+      "loss": 0.1855,
+      "step": 113750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.472916012734855e-05,
+      "loss": 0.1839,
+      "step": 113760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4726918075422627e-05,
+      "loss": 0.1915,
+      "step": 113770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4724676023496706e-05,
+      "loss": 0.1896,
+      "step": 113780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.472243397157078e-05,
+      "loss": 0.1888,
+      "step": 113790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.472019191964486e-05,
+      "loss": 0.1903,
+      "step": 113800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.471794986771894e-05,
+      "loss": 0.1893,
+      "step": 113810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4715707815793014e-05,
+      "loss": 0.1819,
+      "step": 113820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4713465763867093e-05,
+      "loss": 0.1854,
+      "step": 113830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.471122371194117e-05,
+      "loss": 0.1869,
+      "step": 113840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.470898166001525e-05,
+      "loss": 0.1886,
+      "step": 113850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4706739608089325e-05,
+      "loss": 0.1916,
+      "step": 113860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4704497556163404e-05,
+      "loss": 0.1851,
+      "step": 113870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.470225550423748e-05,
+      "loss": 0.1873,
+      "step": 113880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4700013452311558e-05,
+      "loss": 0.1855,
+      "step": 113890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4697771400385633e-05,
+      "loss": 0.1832,
+      "step": 113900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4695529348459712e-05,
+      "loss": 0.1856,
+      "step": 113910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.469328729653379e-05,
+      "loss": 0.188,
+      "step": 113920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4691045244607866e-05,
+      "loss": 0.1885,
+      "step": 113930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4688803192681945e-05,
+      "loss": 0.1924,
+      "step": 113940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.468656114075602e-05,
+      "loss": 0.1861,
+      "step": 113950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.46843190888301e-05,
+      "loss": 0.1831,
+      "step": 113960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4682077036904174e-05,
+      "loss": 0.1837,
+      "step": 113970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4679834984978253e-05,
+      "loss": 0.1871,
+      "step": 113980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4677592933052332e-05,
+      "loss": 0.1857,
+      "step": 113990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4675350881126407e-05,
+      "loss": 0.1862,
+      "step": 114000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4673108829200486e-05,
+      "loss": 0.188,
+      "step": 114010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.467086677727456e-05,
+      "loss": 0.1889,
+      "step": 114020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.466862472534864e-05,
+      "loss": 0.1808,
+      "step": 114030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4666382673422715e-05,
+      "loss": 0.1893,
+      "step": 114040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4664140621496794e-05,
+      "loss": 0.187,
+      "step": 114050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4661898569570873e-05,
+      "loss": 0.1898,
+      "step": 114060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4659656517644948e-05,
+      "loss": 0.1872,
+      "step": 114070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4657414465719027e-05,
+      "loss": 0.1843,
+      "step": 114080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4655172413793105e-05,
+      "loss": 0.1865,
+      "step": 114090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4652930361867184e-05,
+      "loss": 0.1879,
+      "step": 114100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.465068830994126e-05,
+      "loss": 0.1866,
+      "step": 114110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4648446258015338e-05,
+      "loss": 0.1889,
+      "step": 114120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4646204206089417e-05,
+      "loss": 0.1851,
+      "step": 114130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4643962154163492e-05,
+      "loss": 0.1817,
+      "step": 114140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.464172010223757e-05,
+      "loss": 0.1885,
+      "step": 114150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4639478050311646e-05,
+      "loss": 0.1869,
+      "step": 114160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4637235998385725e-05,
+      "loss": 0.1917,
+      "step": 114170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.46349939464598e-05,
+      "loss": 0.1787,
+      "step": 114180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.463275189453388e-05,
+      "loss": 0.1861,
+      "step": 114190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4630509842607958e-05,
+      "loss": 0.1858,
+      "step": 114200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4628267790682033e-05,
+      "loss": 0.181,
+      "step": 114210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4626025738756112e-05,
+      "loss": 0.1934,
+      "step": 114220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4623783686830187e-05,
+      "loss": 0.19,
+      "step": 114230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4621541634904266e-05,
+      "loss": 0.187,
+      "step": 114240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.461929958297834e-05,
+      "loss": 0.1916,
+      "step": 114250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.461705753105242e-05,
+      "loss": 0.1862,
+      "step": 114260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.46148154791265e-05,
+      "loss": 0.1939,
+      "step": 114270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4612573427200574e-05,
+      "loss": 0.186,
+      "step": 114280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4610331375274653e-05,
+      "loss": 0.189,
+      "step": 114290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4608089323348728e-05,
+      "loss": 0.1849,
+      "step": 114300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4605847271422807e-05,
+      "loss": 0.1878,
+      "step": 114310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4603605219496882e-05,
+      "loss": 0.192,
+      "step": 114320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.460136316757096e-05,
+      "loss": 0.1833,
+      "step": 114330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.459912111564504e-05,
+      "loss": 0.187,
+      "step": 114340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4596879063719118e-05,
+      "loss": 0.1916,
+      "step": 114350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4594637011793193e-05,
+      "loss": 0.186,
+      "step": 114360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4592394959867272e-05,
+      "loss": 0.1889,
+      "step": 114370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.459015290794135e-05,
+      "loss": 0.1859,
+      "step": 114380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4587910856015426e-05,
+      "loss": 0.1827,
+      "step": 114390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4585668804089505e-05,
+      "loss": 0.1901,
+      "step": 114400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4583426752163584e-05,
+      "loss": 0.1873,
+      "step": 114410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.458118470023766e-05,
+      "loss": 0.1831,
+      "step": 114420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4578942648311738e-05,
+      "loss": 0.1844,
+      "step": 114430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4576700596385813e-05,
+      "loss": 0.1877,
+      "step": 114440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4574458544459892e-05,
+      "loss": 0.1874,
+      "step": 114450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4572216492533967e-05,
+      "loss": 0.1862,
+      "step": 114460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4569974440608046e-05,
+      "loss": 0.1843,
+      "step": 114470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4567732388682125e-05,
+      "loss": 0.1888,
+      "step": 114480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.45654903367562e-05,
+      "loss": 0.1872,
+      "step": 114490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.456324828483028e-05,
+      "loss": 0.1857,
+      "step": 114500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4561006232904354e-05,
+      "loss": 0.1832,
+      "step": 114510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4558764180978433e-05,
+      "loss": 0.185,
+      "step": 114520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4556522129052508e-05,
+      "loss": 0.1861,
+      "step": 114530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4554280077126587e-05,
+      "loss": 0.1879,
+      "step": 114540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4552038025200665e-05,
+      "loss": 0.181,
+      "step": 114550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.454979597327474e-05,
+      "loss": 0.1881,
+      "step": 114560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.454755392134882e-05,
+      "loss": 0.1849,
+      "step": 114570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4545311869422895e-05,
+      "loss": 0.1863,
+      "step": 114580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4543069817496974e-05,
+      "loss": 0.1826,
+      "step": 114590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4540827765571052e-05,
+      "loss": 0.1909,
+      "step": 114600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4538585713645128e-05,
+      "loss": 0.1799,
+      "step": 114610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4536343661719206e-05,
+      "loss": 0.1867,
+      "step": 114620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4534101609793285e-05,
+      "loss": 0.181,
+      "step": 114630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4531859557867364e-05,
+      "loss": 0.1835,
+      "step": 114640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.452961750594144e-05,
+      "loss": 0.1847,
+      "step": 114650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4527375454015518e-05,
+      "loss": 0.1796,
+      "step": 114660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4525133402089593e-05,
+      "loss": 0.1847,
+      "step": 114670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4522891350163672e-05,
+      "loss": 0.1844,
+      "step": 114680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.452064929823775e-05,
+      "loss": 0.1896,
+      "step": 114690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4518407246311826e-05,
+      "loss": 0.1871,
+      "step": 114700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4516165194385905e-05,
+      "loss": 0.1871,
+      "step": 114710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.451392314245998e-05,
+      "loss": 0.189,
+      "step": 114720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.451168109053406e-05,
+      "loss": 0.1873,
+      "step": 114730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4509439038608134e-05,
+      "loss": 0.1867,
+      "step": 114740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4507196986682213e-05,
+      "loss": 0.1844,
+      "step": 114750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.450495493475629e-05,
+      "loss": 0.1858,
+      "step": 114760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4502712882830367e-05,
+      "loss": 0.1872,
+      "step": 114770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4500470830904445e-05,
+      "loss": 0.1929,
+      "step": 114780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.449822877897852e-05,
+      "loss": 0.1862,
+      "step": 114790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.44959867270526e-05,
+      "loss": 0.1828,
+      "step": 114800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4493744675126675e-05,
+      "loss": 0.1906,
+      "step": 114810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4491502623200754e-05,
+      "loss": 0.1894,
+      "step": 114820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4489260571274832e-05,
+      "loss": 0.1844,
+      "step": 114830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4487018519348908e-05,
+      "loss": 0.185,
+      "step": 114840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4484776467422986e-05,
+      "loss": 0.1822,
+      "step": 114850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.448253441549706e-05,
+      "loss": 0.1901,
+      "step": 114860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.448029236357114e-05,
+      "loss": 0.1955,
+      "step": 114870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.447805031164522e-05,
+      "loss": 0.1848,
+      "step": 114880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4475808259719298e-05,
+      "loss": 0.1867,
+      "step": 114890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4473566207793373e-05,
+      "loss": 0.189,
+      "step": 114900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4471324155867452e-05,
+      "loss": 0.1828,
+      "step": 114910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.446908210394153e-05,
+      "loss": 0.1861,
+      "step": 114920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4466840052015606e-05,
+      "loss": 0.1856,
+      "step": 114930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4464598000089685e-05,
+      "loss": 0.1823,
+      "step": 114940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.446235594816376e-05,
+      "loss": 0.1943,
+      "step": 114950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.446011389623784e-05,
+      "loss": 0.1911,
+      "step": 114960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4457871844311917e-05,
+      "loss": 0.1874,
+      "step": 114970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4455629792385993e-05,
+      "loss": 0.1822,
+      "step": 114980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.445338774046007e-05,
+      "loss": 0.1812,
+      "step": 114990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4451145688534147e-05,
+      "loss": 0.1853,
+      "step": 115000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4448903636608225e-05,
+      "loss": 0.1895,
+      "step": 115010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.44466615846823e-05,
+      "loss": 0.1914,
+      "step": 115020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.444441953275638e-05,
+      "loss": 0.1837,
+      "step": 115030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4442177480830458e-05,
+      "loss": 0.186,
+      "step": 115040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4439935428904534e-05,
+      "loss": 0.19,
+      "step": 115050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4437693376978612e-05,
+      "loss": 0.1867,
+      "step": 115060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4435451325052688e-05,
+      "loss": 0.1827,
+      "step": 115070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4433209273126766e-05,
+      "loss": 0.1812,
+      "step": 115080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.443096722120084e-05,
+      "loss": 0.1949,
+      "step": 115090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.442872516927492e-05,
+      "loss": 0.1902,
+      "step": 115100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4426483117349e-05,
+      "loss": 0.1844,
+      "step": 115110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4424241065423074e-05,
+      "loss": 0.1847,
+      "step": 115120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4421999013497153e-05,
+      "loss": 0.1859,
+      "step": 115130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4419756961571232e-05,
+      "loss": 0.1805,
+      "step": 115140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4417514909645307e-05,
+      "loss": 0.188,
+      "step": 115150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4415272857719386e-05,
+      "loss": 0.1852,
+      "step": 115160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4413030805793465e-05,
+      "loss": 0.1863,
+      "step": 115170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4410788753867543e-05,
+      "loss": 0.1888,
+      "step": 115180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.440854670194162e-05,
+      "loss": 0.1864,
+      "step": 115190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4406304650015697e-05,
+      "loss": 0.1886,
+      "step": 115200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4404062598089773e-05,
+      "loss": 0.1856,
+      "step": 115210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.440182054616385e-05,
+      "loss": 0.1872,
+      "step": 115220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.439957849423793e-05,
+      "loss": 0.1866,
+      "step": 115230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4397336442312005e-05,
+      "loss": 0.1878,
+      "step": 115240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4395094390386084e-05,
+      "loss": 0.1894,
+      "step": 115250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.439285233846016e-05,
+      "loss": 0.1878,
+      "step": 115260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4390610286534238e-05,
+      "loss": 0.187,
+      "step": 115270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4388368234608314e-05,
+      "loss": 0.1869,
+      "step": 115280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4386126182682392e-05,
+      "loss": 0.1866,
+      "step": 115290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.438388413075647e-05,
+      "loss": 0.185,
+      "step": 115300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4381642078830546e-05,
+      "loss": 0.1863,
+      "step": 115310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4379400026904625e-05,
+      "loss": 0.1843,
+      "step": 115320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.43771579749787e-05,
+      "loss": 0.183,
+      "step": 115330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.437491592305278e-05,
+      "loss": 0.1807,
+      "step": 115340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4372673871126854e-05,
+      "loss": 0.1807,
+      "step": 115350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.4370431819200933e-05,
+      "loss": 0.1867,
+      "step": 115360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4368189767275012e-05,
+      "loss": 0.1852,
+      "step": 115370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4365947715349087e-05,
+      "loss": 0.1878,
+      "step": 115380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4363705663423166e-05,
+      "loss": 0.1841,
+      "step": 115390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.436146361149724e-05,
+      "loss": 0.183,
+      "step": 115400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.435922155957132e-05,
+      "loss": 0.18,
+      "step": 115410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.43569795076454e-05,
+      "loss": 0.1865,
+      "step": 115420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4354737455719477e-05,
+      "loss": 0.1858,
+      "step": 115430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4352495403793553e-05,
+      "loss": 0.1859,
+      "step": 115440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.435025335186763e-05,
+      "loss": 0.188,
+      "step": 115450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.434801129994171e-05,
+      "loss": 0.1854,
+      "step": 115460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4345769248015785e-05,
+      "loss": 0.1902,
+      "step": 115470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4343527196089864e-05,
+      "loss": 0.1849,
+      "step": 115480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.434128514416394e-05,
+      "loss": 0.182,
+      "step": 115490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4339043092238018e-05,
+      "loss": 0.1884,
+      "step": 115500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4336801040312097e-05,
+      "loss": 0.184,
+      "step": 115510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4334558988386172e-05,
+      "loss": 0.1832,
+      "step": 115520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.433231693646025e-05,
+      "loss": 0.1846,
+      "step": 115530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4330074884534326e-05,
+      "loss": 0.1891,
+      "step": 115540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4327832832608405e-05,
+      "loss": 0.1882,
+      "step": 115550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.432559078068248e-05,
+      "loss": 0.1848,
+      "step": 115560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.432334872875656e-05,
+      "loss": 0.1818,
+      "step": 115570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4321106676830638e-05,
+      "loss": 0.1877,
+      "step": 115580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4318864624904713e-05,
+      "loss": 0.1814,
+      "step": 115590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4316622572978792e-05,
+      "loss": 0.186,
+      "step": 115600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4314380521052867e-05,
+      "loss": 0.1868,
+      "step": 115610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4312138469126946e-05,
+      "loss": 0.1831,
+      "step": 115620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.430989641720102e-05,
+      "loss": 0.1882,
+      "step": 115630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.43076543652751e-05,
+      "loss": 0.1851,
+      "step": 115640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.430541231334918e-05,
+      "loss": 0.191,
+      "step": 115650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4303170261423254e-05,
+      "loss": 0.1804,
+      "step": 115660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4300928209497333e-05,
+      "loss": 0.1879,
+      "step": 115670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.429868615757141e-05,
+      "loss": 0.1842,
+      "step": 115680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4296444105645487e-05,
+      "loss": 0.1855,
+      "step": 115690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4294202053719565e-05,
+      "loss": 0.1895,
+      "step": 115700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4291960001793644e-05,
+      "loss": 0.1809,
+      "step": 115710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4289717949867723e-05,
+      "loss": 0.1855,
+      "step": 115720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4287475897941798e-05,
+      "loss": 0.1825,
+      "step": 115730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4285233846015877e-05,
+      "loss": 0.1905,
+      "step": 115740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4282991794089952e-05,
+      "loss": 0.1834,
+      "step": 115750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.428074974216403e-05,
+      "loss": 0.1898,
+      "step": 115760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4278507690238106e-05,
+      "loss": 0.184,
+      "step": 115770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4276265638312185e-05,
+      "loss": 0.1856,
+      "step": 115780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4274023586386264e-05,
+      "loss": 0.1916,
+      "step": 115790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.427178153446034e-05,
+      "loss": 0.1789,
+      "step": 115800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4269539482534418e-05,
+      "loss": 0.1962,
+      "step": 115810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4267297430608493e-05,
+      "loss": 0.1873,
+      "step": 115820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4265055378682572e-05,
+      "loss": 0.1891,
+      "step": 115830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4262813326756647e-05,
+      "loss": 0.1874,
+      "step": 115840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4260571274830726e-05,
+      "loss": 0.186,
+      "step": 115850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4258329222904805e-05,
+      "loss": 0.1889,
+      "step": 115860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.425608717097888e-05,
+      "loss": 0.1877,
+      "step": 115870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.425384511905296e-05,
+      "loss": 0.1924,
+      "step": 115880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4251603067127034e-05,
+      "loss": 0.1844,
+      "step": 115890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4249361015201113e-05,
+      "loss": 0.184,
+      "step": 115900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4247118963275188e-05,
+      "loss": 0.1843,
+      "step": 115910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4244876911349267e-05,
+      "loss": 0.1831,
+      "step": 115920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4242634859423345e-05,
+      "loss": 0.1806,
+      "step": 115930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.424039280749742e-05,
+      "loss": 0.1956,
+      "step": 115940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.42381507555715e-05,
+      "loss": 0.1851,
+      "step": 115950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4235908703645578e-05,
+      "loss": 0.1881,
+      "step": 115960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4233666651719657e-05,
+      "loss": 0.1825,
+      "step": 115970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4231424599793732e-05,
+      "loss": 0.1829,
+      "step": 115980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.422918254786781e-05,
+      "loss": 0.1856,
+      "step": 115990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.422694049594189e-05,
+      "loss": 0.1816,
+      "step": 116000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4224698444015965e-05,
+      "loss": 0.1826,
+      "step": 116010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4222456392090044e-05,
+      "loss": 0.1863,
+      "step": 116020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.422021434016412e-05,
+      "loss": 0.1867,
+      "step": 116030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4217972288238198e-05,
+      "loss": 0.1871,
+      "step": 116040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4215730236312273e-05,
+      "loss": 0.1892,
+      "step": 116050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4213488184386352e-05,
+      "loss": 0.1846,
+      "step": 116060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.421124613246043e-05,
+      "loss": 0.1856,
+      "step": 116070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4209004080534506e-05,
+      "loss": 0.1859,
+      "step": 116080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4206762028608585e-05,
+      "loss": 0.1906,
+      "step": 116090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.420451997668266e-05,
+      "loss": 0.1859,
+      "step": 116100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.420227792475674e-05,
+      "loss": 0.1836,
+      "step": 116110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4200035872830814e-05,
+      "loss": 0.1858,
+      "step": 116120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4197793820904893e-05,
+      "loss": 0.1835,
+      "step": 116130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.419555176897897e-05,
+      "loss": 0.1932,
+      "step": 116140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4193309717053047e-05,
+      "loss": 0.1855,
+      "step": 116150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4191067665127126e-05,
+      "loss": 0.1878,
+      "step": 116160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.41888256132012e-05,
+      "loss": 0.1798,
+      "step": 116170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.418658356127528e-05,
+      "loss": 0.1903,
+      "step": 116180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4184341509349355e-05,
+      "loss": 0.1833,
+      "step": 116190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4182099457423434e-05,
+      "loss": 0.1817,
+      "step": 116200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4179857405497512e-05,
+      "loss": 0.1846,
+      "step": 116210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.417761535357159e-05,
+      "loss": 0.19,
+      "step": 116220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4175373301645666e-05,
+      "loss": 0.1826,
+      "step": 116230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4173131249719745e-05,
+      "loss": 0.1874,
+      "step": 116240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4170889197793824e-05,
+      "loss": 0.1836,
+      "step": 116250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.41686471458679e-05,
+      "loss": 0.1787,
+      "step": 116260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4166405093941978e-05,
+      "loss": 0.1895,
+      "step": 116270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4164163042016057e-05,
+      "loss": 0.1891,
+      "step": 116280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4161920990090132e-05,
+      "loss": 0.1824,
+      "step": 116290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.415967893816421e-05,
+      "loss": 0.1919,
+      "step": 116300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4157436886238286e-05,
+      "loss": 0.1831,
+      "step": 116310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4155194834312365e-05,
+      "loss": 0.1832,
+      "step": 116320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.415295278238644e-05,
+      "loss": 0.1785,
+      "step": 116330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.415071073046052e-05,
+      "loss": 0.1869,
+      "step": 116340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4148468678534597e-05,
+      "loss": 0.1869,
+      "step": 116350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4146226626608673e-05,
+      "loss": 0.1834,
+      "step": 116360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.414398457468275e-05,
+      "loss": 0.1872,
+      "step": 116370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4141742522756827e-05,
+      "loss": 0.1867,
+      "step": 116380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4139500470830906e-05,
+      "loss": 0.1833,
+      "step": 116390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.413725841890498e-05,
+      "loss": 0.1847,
+      "step": 116400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.413501636697906e-05,
+      "loss": 0.1834,
+      "step": 116410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4132774315053138e-05,
+      "loss": 0.1842,
+      "step": 116420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4130532263127214e-05,
+      "loss": 0.1915,
+      "step": 116430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4128290211201292e-05,
+      "loss": 0.1842,
+      "step": 116440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4126048159275368e-05,
+      "loss": 0.1879,
+      "step": 116450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4123806107349446e-05,
+      "loss": 0.1873,
+      "step": 116460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4121564055423522e-05,
+      "loss": 0.1854,
+      "step": 116470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.41193220034976e-05,
+      "loss": 0.1843,
+      "step": 116480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.411707995157168e-05,
+      "loss": 0.1844,
+      "step": 116490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4114837899645758e-05,
+      "loss": 0.1776,
+      "step": 116500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4112595847719837e-05,
+      "loss": 0.1908,
+      "step": 116510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4110353795793912e-05,
+      "loss": 0.1835,
+      "step": 116520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.410811174386799e-05,
+      "loss": 0.1821,
+      "step": 116530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.410586969194207e-05,
+      "loss": 0.1805,
+      "step": 116540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4103627640016145e-05,
+      "loss": 0.1878,
+      "step": 116550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4101385588090223e-05,
+      "loss": 0.1888,
+      "step": 116560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.40991435361643e-05,
+      "loss": 0.1853,
+      "step": 116570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4096901484238377e-05,
+      "loss": 0.1912,
+      "step": 116580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4094659432312453e-05,
+      "loss": 0.1835,
+      "step": 116590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.409241738038653e-05,
+      "loss": 0.1832,
+      "step": 116600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.409017532846061e-05,
+      "loss": 0.1823,
+      "step": 116610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4087933276534686e-05,
+      "loss": 0.1857,
+      "step": 116620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4085691224608764e-05,
+      "loss": 0.1867,
+      "step": 116630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.408344917268284e-05,
+      "loss": 0.184,
+      "step": 116640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.408120712075692e-05,
+      "loss": 0.1876,
+      "step": 116650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4078965068830994e-05,
+      "loss": 0.1824,
+      "step": 116660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4076723016905072e-05,
+      "loss": 0.1828,
+      "step": 116670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.407448096497915e-05,
+      "loss": 0.1831,
+      "step": 116680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4072238913053226e-05,
+      "loss": 0.1844,
+      "step": 116690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4069996861127305e-05,
+      "loss": 0.1825,
+      "step": 116700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.406775480920138e-05,
+      "loss": 0.1877,
+      "step": 116710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.406551275727546e-05,
+      "loss": 0.1854,
+      "step": 116720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4063270705349534e-05,
+      "loss": 0.186,
+      "step": 116730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4061028653423613e-05,
+      "loss": 0.1867,
+      "step": 116740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4058786601497692e-05,
+      "loss": 0.1795,
+      "step": 116750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4056544549571767e-05,
+      "loss": 0.1808,
+      "step": 116760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4054302497645846e-05,
+      "loss": 0.1843,
+      "step": 116770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4052060445719925e-05,
+      "loss": 0.1812,
+      "step": 116780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4049818393794003e-05,
+      "loss": 0.186,
+      "step": 116790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.404757634186808e-05,
+      "loss": 0.187,
+      "step": 116800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4045334289942157e-05,
+      "loss": 0.193,
+      "step": 116810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4043092238016236e-05,
+      "loss": 0.1888,
+      "step": 116820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.404085018609031e-05,
+      "loss": 0.1812,
+      "step": 116830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.403860813416439e-05,
+      "loss": 0.1849,
+      "step": 116840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4036366082238466e-05,
+      "loss": 0.1891,
+      "step": 116850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4034124030312544e-05,
+      "loss": 0.1836,
+      "step": 116860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.403188197838662e-05,
+      "loss": 0.1837,
+      "step": 116870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.40296399264607e-05,
+      "loss": 0.1783,
+      "step": 116880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4027397874534777e-05,
+      "loss": 0.1836,
+      "step": 116890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4025155822608852e-05,
+      "loss": 0.189,
+      "step": 116900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.402291377068293e-05,
+      "loss": 0.181,
+      "step": 116910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4020671718757006e-05,
+      "loss": 0.1902,
+      "step": 116920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4018429666831085e-05,
+      "loss": 0.1827,
+      "step": 116930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.401618761490516e-05,
+      "loss": 0.1821,
+      "step": 116940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.401394556297924e-05,
+      "loss": 0.1902,
+      "step": 116950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4011703511053318e-05,
+      "loss": 0.1846,
+      "step": 116960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4009461459127393e-05,
+      "loss": 0.1862,
+      "step": 116970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4007219407201472e-05,
+      "loss": 0.1829,
+      "step": 116980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4004977355275547e-05,
+      "loss": 0.1899,
+      "step": 116990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4002735303349626e-05,
+      "loss": 0.1823,
+      "step": 117000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.40004932514237e-05,
+      "loss": 0.1852,
+      "step": 117010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.399825119949778e-05,
+      "loss": 0.1829,
+      "step": 117020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.399600914757186e-05,
+      "loss": 0.1839,
+      "step": 117030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3993767095645937e-05,
+      "loss": 0.1811,
+      "step": 117040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3991525043720016e-05,
+      "loss": 0.1868,
+      "step": 117050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.398928299179409e-05,
+      "loss": 0.1843,
+      "step": 117060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.398704093986817e-05,
+      "loss": 0.1781,
+      "step": 117070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3984798887942246e-05,
+      "loss": 0.187,
+      "step": 117080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3982556836016324e-05,
+      "loss": 0.191,
+      "step": 117090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3980314784090403e-05,
+      "loss": 0.176,
+      "step": 117100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.397807273216448e-05,
+      "loss": 0.1812,
+      "step": 117110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3975830680238557e-05,
+      "loss": 0.1776,
+      "step": 117120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3973588628312632e-05,
+      "loss": 0.1838,
+      "step": 117130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.397134657638671e-05,
+      "loss": 0.1907,
+      "step": 117140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3969104524460786e-05,
+      "loss": 0.1835,
+      "step": 117150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3966862472534865e-05,
+      "loss": 0.1923,
+      "step": 117160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3964620420608944e-05,
+      "loss": 0.1833,
+      "step": 117170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.396237836868302e-05,
+      "loss": 0.1811,
+      "step": 117180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3960136316757098e-05,
+      "loss": 0.1843,
+      "step": 117190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3957894264831173e-05,
+      "loss": 0.1879,
+      "step": 117200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3955652212905252e-05,
+      "loss": 0.186,
+      "step": 117210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3953410160979327e-05,
+      "loss": 0.184,
+      "step": 117220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3951168109053406e-05,
+      "loss": 0.1828,
+      "step": 117230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3948926057127485e-05,
+      "loss": 0.184,
+      "step": 117240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.394668400520156e-05,
+      "loss": 0.1824,
+      "step": 117250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.394444195327564e-05,
+      "loss": 0.1827,
+      "step": 117260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3942199901349714e-05,
+      "loss": 0.183,
+      "step": 117270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3939957849423793e-05,
+      "loss": 0.181,
+      "step": 117280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.393771579749787e-05,
+      "loss": 0.1849,
+      "step": 117290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3935473745571947e-05,
+      "loss": 0.1883,
+      "step": 117300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3933231693646026e-05,
+      "loss": 0.1847,
+      "step": 117310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3930989641720104e-05,
+      "loss": 0.1841,
+      "step": 117320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3928747589794183e-05,
+      "loss": 0.1831,
+      "step": 117330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.392650553786826e-05,
+      "loss": 0.1847,
+      "step": 117340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3924263485942337e-05,
+      "loss": 0.1862,
+      "step": 117350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3922021434016412e-05,
+      "loss": 0.1805,
+      "step": 117360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.391977938209049e-05,
+      "loss": 0.1805,
+      "step": 117370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.391753733016457e-05,
+      "loss": 0.1858,
+      "step": 117380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3915295278238645e-05,
+      "loss": 0.1856,
+      "step": 117390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3913053226312724e-05,
+      "loss": 0.1869,
+      "step": 117400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.39108111743868e-05,
+      "loss": 0.1895,
+      "step": 117410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3908569122460878e-05,
+      "loss": 0.1811,
+      "step": 117420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3906327070534953e-05,
+      "loss": 0.1841,
+      "step": 117430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3904085018609032e-05,
+      "loss": 0.1788,
+      "step": 117440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.390184296668311e-05,
+      "loss": 0.1839,
+      "step": 117450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3899600914757186e-05,
+      "loss": 0.1852,
+      "step": 117460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3897358862831265e-05,
+      "loss": 0.1803,
+      "step": 117470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.389511681090534e-05,
+      "loss": 0.192,
+      "step": 117480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.389287475897942e-05,
+      "loss": 0.1848,
+      "step": 117490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3890632707053494e-05,
+      "loss": 0.1849,
+      "step": 117500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3888390655127573e-05,
+      "loss": 0.1871,
+      "step": 117510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.388614860320165e-05,
+      "loss": 0.1849,
+      "step": 117520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3883906551275727e-05,
+      "loss": 0.18,
+      "step": 117530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3881664499349806e-05,
+      "loss": 0.1824,
+      "step": 117540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.387942244742388e-05,
+      "loss": 0.1901,
+      "step": 117550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.387718039549796e-05,
+      "loss": 0.1789,
+      "step": 117560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.387493834357204e-05,
+      "loss": 0.1822,
+      "step": 117570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3872696291646117e-05,
+      "loss": 0.1815,
+      "step": 117580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.3870454239720196e-05,
+      "loss": 0.1845,
+      "step": 117590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.386821218779427e-05,
+      "loss": 0.1793,
+      "step": 117600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.386597013586835e-05,
+      "loss": 0.1861,
+      "step": 117610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3863728083942425e-05,
+      "loss": 0.1886,
+      "step": 117620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3861486032016504e-05,
+      "loss": 0.1796,
+      "step": 117630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.385924398009058e-05,
+      "loss": 0.1834,
+      "step": 117640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3857001928164658e-05,
+      "loss": 0.1855,
+      "step": 117650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3854759876238737e-05,
+      "loss": 0.1823,
+      "step": 117660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3852517824312812e-05,
+      "loss": 0.1882,
+      "step": 117670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.385027577238689e-05,
+      "loss": 0.1893,
+      "step": 117680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3848033720460966e-05,
+      "loss": 0.1797,
+      "step": 117690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3845791668535045e-05,
+      "loss": 0.1853,
+      "step": 117700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.384354961660912e-05,
+      "loss": 0.1857,
+      "step": 117710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.38413075646832e-05,
+      "loss": 0.1885,
+      "step": 117720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3839065512757278e-05,
+      "loss": 0.1854,
+      "step": 117730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3836823460831353e-05,
+      "loss": 0.1817,
+      "step": 117740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.383458140890543e-05,
+      "loss": 0.1823,
+      "step": 117750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3832339356979507e-05,
+      "loss": 0.1809,
+      "step": 117760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3830097305053586e-05,
+      "loss": 0.1797,
+      "step": 117770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.382785525312766e-05,
+      "loss": 0.1903,
+      "step": 117780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.382561320120174e-05,
+      "loss": 0.1836,
+      "step": 117790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.382337114927582e-05,
+      "loss": 0.1799,
+      "step": 117800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3821129097349894e-05,
+      "loss": 0.1833,
+      "step": 117810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3818887045423972e-05,
+      "loss": 0.182,
+      "step": 117820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.381664499349805e-05,
+      "loss": 0.1813,
+      "step": 117830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3814402941572126e-05,
+      "loss": 0.1849,
+      "step": 117840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3812160889646205e-05,
+      "loss": 0.183,
+      "step": 117850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3809918837720284e-05,
+      "loss": 0.187,
+      "step": 117860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3807676785794363e-05,
+      "loss": 0.1839,
+      "step": 117870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3805434733868438e-05,
+      "loss": 0.1801,
+      "step": 117880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3803192681942517e-05,
+      "loss": 0.1835,
+      "step": 117890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3800950630016592e-05,
+      "loss": 0.1874,
+      "step": 117900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.379870857809067e-05,
+      "loss": 0.1843,
+      "step": 117910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.379646652616475e-05,
+      "loss": 0.1857,
+      "step": 117920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3794224474238825e-05,
+      "loss": 0.1854,
+      "step": 117930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3791982422312903e-05,
+      "loss": 0.1808,
+      "step": 117940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.378974037038698e-05,
+      "loss": 0.1881,
+      "step": 117950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3787498318461058e-05,
+      "loss": 0.1869,
+      "step": 117960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3785256266535133e-05,
+      "loss": 0.18,
+      "step": 117970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.378301421460921e-05,
+      "loss": 0.1781,
+      "step": 117980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.378077216268329e-05,
+      "loss": 0.1852,
+      "step": 117990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3778530110757366e-05,
+      "loss": 0.1858,
+      "step": 118000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3776288058831444e-05,
+      "loss": 0.1819,
+      "step": 118010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.377404600690552e-05,
+      "loss": 0.1807,
+      "step": 118020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.37718039549796e-05,
+      "loss": 0.187,
+      "step": 118030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3769561903053674e-05,
+      "loss": 0.1875,
+      "step": 118040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3767319851127752e-05,
+      "loss": 0.1908,
+      "step": 118050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.376507779920183e-05,
+      "loss": 0.1876,
+      "step": 118060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3762835747275906e-05,
+      "loss": 0.1816,
+      "step": 118070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3760593695349985e-05,
+      "loss": 0.18,
+      "step": 118080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.375835164342406e-05,
+      "loss": 0.1864,
+      "step": 118090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.375610959149814e-05,
+      "loss": 0.1848,
+      "step": 118100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3753867539572218e-05,
+      "loss": 0.1858,
+      "step": 118110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3751625487646297e-05,
+      "loss": 0.1845,
+      "step": 118120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3749383435720375e-05,
+      "loss": 0.1776,
+      "step": 118130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.374714138379445e-05,
+      "loss": 0.1785,
+      "step": 118140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.374489933186853e-05,
+      "loss": 0.184,
+      "step": 118150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3742657279942605e-05,
+      "loss": 0.1838,
+      "step": 118160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3740415228016684e-05,
+      "loss": 0.1826,
+      "step": 118170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.373817317609076e-05,
+      "loss": 0.187,
+      "step": 118180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3735931124164838e-05,
+      "loss": 0.1845,
+      "step": 118190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3733689072238916e-05,
+      "loss": 0.1839,
+      "step": 118200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.373144702031299e-05,
+      "loss": 0.182,
+      "step": 118210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.372920496838707e-05,
+      "loss": 0.1901,
+      "step": 118220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3726962916461146e-05,
+      "loss": 0.1886,
+      "step": 118230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3724720864535224e-05,
+      "loss": 0.1862,
+      "step": 118240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.37224788126093e-05,
+      "loss": 0.1845,
+      "step": 118250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.372023676068338e-05,
+      "loss": 0.1853,
+      "step": 118260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3717994708757457e-05,
+      "loss": 0.1813,
+      "step": 118270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3715752656831532e-05,
+      "loss": 0.1802,
+      "step": 118280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.371351060490561e-05,
+      "loss": 0.1835,
+      "step": 118290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3711268552979687e-05,
+      "loss": 0.1786,
+      "step": 118300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3709026501053765e-05,
+      "loss": 0.1894,
+      "step": 118310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.370678444912784e-05,
+      "loss": 0.1835,
+      "step": 118320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.370454239720192e-05,
+      "loss": 0.1807,
+      "step": 118330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3702300345275998e-05,
+      "loss": 0.1877,
+      "step": 118340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3700058293350073e-05,
+      "loss": 0.1813,
+      "step": 118350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3697816241424152e-05,
+      "loss": 0.1873,
+      "step": 118360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.369557418949823e-05,
+      "loss": 0.1823,
+      "step": 118370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3693332137572306e-05,
+      "loss": 0.1824,
+      "step": 118380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3691090085646385e-05,
+      "loss": 0.1875,
+      "step": 118390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3688848033720464e-05,
+      "loss": 0.1796,
+      "step": 118400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3686605981794542e-05,
+      "loss": 0.1839,
+      "step": 118410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3684363929868618e-05,
+      "loss": 0.1846,
+      "step": 118420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3682121877942696e-05,
+      "loss": 0.1839,
+      "step": 118430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.367987982601677e-05,
+      "loss": 0.1872,
+      "step": 118440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.367763777409085e-05,
+      "loss": 0.1835,
+      "step": 118450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3675395722164926e-05,
+      "loss": 0.1816,
+      "step": 118460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3673153670239004e-05,
+      "loss": 0.1842,
+      "step": 118470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3670911618313083e-05,
+      "loss": 0.1902,
+      "step": 118480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.366866956638716e-05,
+      "loss": 0.1775,
+      "step": 118490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3666427514461237e-05,
+      "loss": 0.1799,
+      "step": 118500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3664185462535312e-05,
+      "loss": 0.1836,
+      "step": 118510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.366194341060939e-05,
+      "loss": 0.1794,
+      "step": 118520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3659701358683467e-05,
+      "loss": 0.1855,
+      "step": 118530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3657459306757545e-05,
+      "loss": 0.1847,
+      "step": 118540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3655217254831624e-05,
+      "loss": 0.1863,
+      "step": 118550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.36529752029057e-05,
+      "loss": 0.1822,
+      "step": 118560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3650733150979778e-05,
+      "loss": 0.185,
+      "step": 118570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3648491099053853e-05,
+      "loss": 0.1884,
+      "step": 118580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3646249047127932e-05,
+      "loss": 0.1851,
+      "step": 118590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3644006995202007e-05,
+      "loss": 0.1812,
+      "step": 118600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3641764943276086e-05,
+      "loss": 0.1835,
+      "step": 118610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3639522891350165e-05,
+      "loss": 0.1842,
+      "step": 118620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.363728083942424e-05,
+      "loss": 0.1821,
+      "step": 118630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.363503878749832e-05,
+      "loss": 0.1798,
+      "step": 118640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3632796735572398e-05,
+      "loss": 0.1836,
+      "step": 118650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3630554683646476e-05,
+      "loss": 0.1818,
+      "step": 118660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.362831263172055e-05,
+      "loss": 0.1856,
+      "step": 118670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.362607057979463e-05,
+      "loss": 0.1866,
+      "step": 118680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.362382852786871e-05,
+      "loss": 0.1865,
+      "step": 118690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3621586475942784e-05,
+      "loss": 0.1825,
+      "step": 118700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3619344424016863e-05,
+      "loss": 0.1823,
+      "step": 118710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.361710237209094e-05,
+      "loss": 0.1841,
+      "step": 118720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3614860320165017e-05,
+      "loss": 0.1821,
+      "step": 118730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3612618268239092e-05,
+      "loss": 0.1793,
+      "step": 118740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.361037621631317e-05,
+      "loss": 0.1834,
+      "step": 118750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.360813416438725e-05,
+      "loss": 0.1872,
+      "step": 118760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3605892112461325e-05,
+      "loss": 0.1867,
+      "step": 118770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3603650060535404e-05,
+      "loss": 0.1841,
+      "step": 118780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.360140800860948e-05,
+      "loss": 0.1815,
+      "step": 118790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3599165956683558e-05,
+      "loss": 0.1821,
+      "step": 118800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3596923904757633e-05,
+      "loss": 0.1862,
+      "step": 118810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3594681852831712e-05,
+      "loss": 0.1856,
+      "step": 118820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.359243980090579e-05,
+      "loss": 0.183,
+      "step": 118830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3590197748979866e-05,
+      "loss": 0.1808,
+      "step": 118840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3587955697053945e-05,
+      "loss": 0.1889,
+      "step": 118850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.358571364512802e-05,
+      "loss": 0.1831,
+      "step": 118860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.35834715932021e-05,
+      "loss": 0.188,
+      "step": 118870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3581229541276174e-05,
+      "loss": 0.1826,
+      "step": 118880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3578987489350253e-05,
+      "loss": 0.1774,
+      "step": 118890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.357674543742433e-05,
+      "loss": 0.1819,
+      "step": 118900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.357450338549841e-05,
+      "loss": 0.1877,
+      "step": 118910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3572261333572486e-05,
+      "loss": 0.1914,
+      "step": 118920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3570019281646564e-05,
+      "loss": 0.1841,
+      "step": 118930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3567777229720643e-05,
+      "loss": 0.1883,
+      "step": 118940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.356553517779472e-05,
+      "loss": 0.1817,
+      "step": 118950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3563293125868797e-05,
+      "loss": 0.1791,
+      "step": 118960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3561051073942876e-05,
+      "loss": 0.1848,
+      "step": 118970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.355880902201695e-05,
+      "loss": 0.1817,
+      "step": 118980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.355656697009103e-05,
+      "loss": 0.1774,
+      "step": 118990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3554324918165105e-05,
+      "loss": 0.1875,
+      "step": 119000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3552082866239184e-05,
+      "loss": 0.183,
+      "step": 119010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.354984081431326e-05,
+      "loss": 0.1861,
+      "step": 119020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3547598762387338e-05,
+      "loss": 0.1825,
+      "step": 119030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3545356710461417e-05,
+      "loss": 0.1824,
+      "step": 119040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3543114658535492e-05,
+      "loss": 0.1846,
+      "step": 119050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.354087260660957e-05,
+      "loss": 0.1814,
+      "step": 119060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3538630554683646e-05,
+      "loss": 0.1837,
+      "step": 119070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3536388502757725e-05,
+      "loss": 0.1872,
+      "step": 119080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.35341464508318e-05,
+      "loss": 0.1871,
+      "step": 119090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.353190439890588e-05,
+      "loss": 0.1807,
+      "step": 119100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3529662346979958e-05,
+      "loss": 0.1741,
+      "step": 119110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3527420295054033e-05,
+      "loss": 0.1856,
+      "step": 119120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.352517824312811e-05,
+      "loss": 0.1803,
+      "step": 119130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3522936191202187e-05,
+      "loss": 0.1886,
+      "step": 119140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3520694139276266e-05,
+      "loss": 0.181,
+      "step": 119150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3518452087350344e-05,
+      "loss": 0.1814,
+      "step": 119160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.351621003542442e-05,
+      "loss": 0.1784,
+      "step": 119170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.35139679834985e-05,
+      "loss": 0.1842,
+      "step": 119180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3511725931572577e-05,
+      "loss": 0.183,
+      "step": 119190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3509483879646656e-05,
+      "loss": 0.1782,
+      "step": 119200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.350724182772073e-05,
+      "loss": 0.1832,
+      "step": 119210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.350499977579481e-05,
+      "loss": 0.1838,
+      "step": 119220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.350275772386889e-05,
+      "loss": 0.1883,
+      "step": 119230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3500515671942964e-05,
+      "loss": 0.1883,
+      "step": 119240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3498273620017043e-05,
+      "loss": 0.1852,
+      "step": 119250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3496031568091118e-05,
+      "loss": 0.1777,
+      "step": 119260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3493789516165197e-05,
+      "loss": 0.1848,
+      "step": 119270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3491547464239272e-05,
+      "loss": 0.1868,
+      "step": 119280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.348930541231335e-05,
+      "loss": 0.1813,
+      "step": 119290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.348706336038743e-05,
+      "loss": 0.1847,
+      "step": 119300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3484821308461505e-05,
+      "loss": 0.191,
+      "step": 119310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3482579256535584e-05,
+      "loss": 0.1845,
+      "step": 119320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.348033720460966e-05,
+      "loss": 0.1801,
+      "step": 119330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3478095152683738e-05,
+      "loss": 0.1867,
+      "step": 119340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3475853100757813e-05,
+      "loss": 0.1808,
+      "step": 119350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.347361104883189e-05,
+      "loss": 0.1901,
+      "step": 119360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.347136899690597e-05,
+      "loss": 0.1791,
+      "step": 119370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3469126944980046e-05,
+      "loss": 0.1884,
+      "step": 119380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3466884893054124e-05,
+      "loss": 0.1846,
+      "step": 119390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.34646428411282e-05,
+      "loss": 0.181,
+      "step": 119400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.346240078920228e-05,
+      "loss": 0.1807,
+      "step": 119410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3460158737276354e-05,
+      "loss": 0.1835,
+      "step": 119420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3457916685350433e-05,
+      "loss": 0.1834,
+      "step": 119430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.345567463342451e-05,
+      "loss": 0.1824,
+      "step": 119440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.345343258149859e-05,
+      "loss": 0.1833,
+      "step": 119450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3451190529572665e-05,
+      "loss": 0.1817,
+      "step": 119460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3448948477646744e-05,
+      "loss": 0.1855,
+      "step": 119470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3446706425720823e-05,
+      "loss": 0.1768,
+      "step": 119480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3444464373794898e-05,
+      "loss": 0.1829,
+      "step": 119490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3442222321868977e-05,
+      "loss": 0.1838,
+      "step": 119500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3439980269943056e-05,
+      "loss": 0.1795,
+      "step": 119510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.343773821801713e-05,
+      "loss": 0.1865,
+      "step": 119520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.343549616609121e-05,
+      "loss": 0.1822,
+      "step": 119530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3433254114165285e-05,
+      "loss": 0.1853,
+      "step": 119540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3431012062239364e-05,
+      "loss": 0.1861,
+      "step": 119550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.342877001031344e-05,
+      "loss": 0.18,
+      "step": 119560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3426527958387518e-05,
+      "loss": 0.1857,
+      "step": 119570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3424285906461596e-05,
+      "loss": 0.1869,
+      "step": 119580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.342204385453567e-05,
+      "loss": 0.1849,
+      "step": 119590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.341980180260975e-05,
+      "loss": 0.184,
+      "step": 119600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3417559750683826e-05,
+      "loss": 0.1851,
+      "step": 119610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3415317698757904e-05,
+      "loss": 0.1817,
+      "step": 119620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.341307564683198e-05,
+      "loss": 0.1809,
+      "step": 119630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.341083359490606e-05,
+      "loss": 0.1845,
+      "step": 119640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3408591542980137e-05,
+      "loss": 0.1831,
+      "step": 119650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3406349491054213e-05,
+      "loss": 0.1787,
+      "step": 119660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.340410743912829e-05,
+      "loss": 0.1856,
+      "step": 119670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3401865387202367e-05,
+      "loss": 0.1779,
+      "step": 119680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3399623335276445e-05,
+      "loss": 0.1847,
+      "step": 119690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3397381283350524e-05,
+      "loss": 0.1823,
+      "step": 119700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.33951392314246e-05,
+      "loss": 0.1845,
+      "step": 119710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3392897179498678e-05,
+      "loss": 0.1858,
+      "step": 119720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3390655127572757e-05,
+      "loss": 0.1853,
+      "step": 119730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3388413075646836e-05,
+      "loss": 0.1871,
+      "step": 119740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.338617102372091e-05,
+      "loss": 0.1815,
+      "step": 119750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.338392897179499e-05,
+      "loss": 0.1829,
+      "step": 119760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3381686919869065e-05,
+      "loss": 0.1852,
+      "step": 119770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3379444867943144e-05,
+      "loss": 0.18,
+      "step": 119780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3377202816017222e-05,
+      "loss": 0.1867,
+      "step": 119790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3374960764091298e-05,
+      "loss": 0.1817,
+      "step": 119800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3372718712165376e-05,
+      "loss": 0.1828,
+      "step": 119810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3370476660239452e-05,
+      "loss": 0.1824,
+      "step": 119820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.336823460831353e-05,
+      "loss": 0.1837,
+      "step": 119830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3365992556387606e-05,
+      "loss": 0.1837,
+      "step": 119840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3363750504461684e-05,
+      "loss": 0.1786,
+      "step": 119850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3361508452535763e-05,
+      "loss": 0.1794,
+      "step": 119860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.335926640060984e-05,
+      "loss": 0.1764,
+      "step": 119870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3357024348683917e-05,
+      "loss": 0.176,
+      "step": 119880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3354782296757993e-05,
+      "loss": 0.1838,
+      "step": 119890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.335254024483207e-05,
+      "loss": 0.1861,
+      "step": 119900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3350298192906147e-05,
+      "loss": 0.1781,
+      "step": 119910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3348056140980225e-05,
+      "loss": 0.1848,
+      "step": 119920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3345814089054304e-05,
+      "loss": 0.1848,
+      "step": 119930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.334357203712838e-05,
+      "loss": 0.1801,
+      "step": 119940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3341329985202458e-05,
+      "loss": 0.1853,
+      "step": 119950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3339087933276533e-05,
+      "loss": 0.1769,
+      "step": 119960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3336845881350612e-05,
+      "loss": 0.1857,
+      "step": 119970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.333460382942469e-05,
+      "loss": 0.1759,
+      "step": 119980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.333236177749877e-05,
+      "loss": 0.1838,
+      "step": 119990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3330119725572845e-05,
+      "loss": 0.1862,
+      "step": 120000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3327877673646924e-05,
+      "loss": 0.1819,
+      "step": 120010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3325635621721002e-05,
+      "loss": 0.184,
+      "step": 120020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3323393569795078e-05,
+      "loss": 0.1851,
+      "step": 120030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3321151517869156e-05,
+      "loss": 0.1835,
+      "step": 120040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3318909465943232e-05,
+      "loss": 0.1846,
+      "step": 120050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.331666741401731e-05,
+      "loss": 0.1832,
+      "step": 120060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.331442536209139e-05,
+      "loss": 0.1849,
+      "step": 120070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3312183310165464e-05,
+      "loss": 0.1797,
+      "step": 120080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3309941258239543e-05,
+      "loss": 0.1866,
+      "step": 120090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.330769920631362e-05,
+      "loss": 0.1832,
+      "step": 120100
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3305457154387697e-05,
+      "loss": 0.181,
+      "step": 120110
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3303215102461773e-05,
+      "loss": 0.1845,
+      "step": 120120
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.330097305053585e-05,
+      "loss": 0.1793,
+      "step": 120130
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.329873099860993e-05,
+      "loss": 0.1798,
+      "step": 120140
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3296488946684005e-05,
+      "loss": 0.1871,
+      "step": 120150
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3294246894758084e-05,
+      "loss": 0.181,
+      "step": 120160
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.329200484283216e-05,
+      "loss": 0.1861,
+      "step": 120170
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3289762790906238e-05,
+      "loss": 0.1851,
+      "step": 120180
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3287520738980313e-05,
+      "loss": 0.1785,
+      "step": 120190
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3285278687054392e-05,
+      "loss": 0.1815,
+      "step": 120200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.328303663512847e-05,
+      "loss": 0.189,
+      "step": 120210
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3280794583202546e-05,
+      "loss": 0.173,
+      "step": 120220
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3278552531276625e-05,
+      "loss": 0.1802,
+      "step": 120230
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3276310479350704e-05,
+      "loss": 0.1808,
+      "step": 120240
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.327406842742478e-05,
+      "loss": 0.1818,
+      "step": 120250
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3271826375498858e-05,
+      "loss": 0.1811,
+      "step": 120260
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3269584323572936e-05,
+      "loss": 0.1777,
+      "step": 120270
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3267342271647015e-05,
+      "loss": 0.1811,
+      "step": 120280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.326510021972109e-05,
+      "loss": 0.18,
+      "step": 120290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.326285816779517e-05,
+      "loss": 0.1801,
+      "step": 120300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3260616115869245e-05,
+      "loss": 0.1846,
+      "step": 120310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3258374063943323e-05,
+      "loss": 0.1836,
+      "step": 120320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.32561320120174e-05,
+      "loss": 0.1807,
+      "step": 120330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3253889960091477e-05,
+      "loss": 0.1883,
+      "step": 120340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3251647908165556e-05,
+      "loss": 0.1796,
+      "step": 120350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.324940585623963e-05,
+      "loss": 0.1797,
+      "step": 120360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.324716380431371e-05,
+      "loss": 0.1867,
+      "step": 120370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3244921752387785e-05,
+      "loss": 0.1868,
+      "step": 120380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3242679700461864e-05,
+      "loss": 0.1821,
+      "step": 120390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.324043764853594e-05,
+      "loss": 0.1845,
+      "step": 120400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3238195596610018e-05,
+      "loss": 0.1845,
+      "step": 120410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3235953544684097e-05,
+      "loss": 0.1813,
+      "step": 120420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3233711492758172e-05,
+      "loss": 0.1882,
+      "step": 120430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.323146944083225e-05,
+      "loss": 0.1763,
+      "step": 120440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3229227388906326e-05,
+      "loss": 0.1869,
+      "step": 120450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3226985336980405e-05,
+      "loss": 0.1832,
+      "step": 120460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.322474328505448e-05,
+      "loss": 0.1841,
+      "step": 120470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.322250123312856e-05,
+      "loss": 0.1808,
+      "step": 120480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3220259181202638e-05,
+      "loss": 0.1794,
+      "step": 120490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3218017129276713e-05,
+      "loss": 0.1809,
+      "step": 120500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3215775077350792e-05,
+      "loss": 0.1823,
+      "step": 120510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.321353302542487e-05,
+      "loss": 0.1797,
+      "step": 120520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.321129097349895e-05,
+      "loss": 0.1817,
+      "step": 120530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3209048921573025e-05,
+      "loss": 0.183,
+      "step": 120540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3206806869647103e-05,
+      "loss": 0.1805,
+      "step": 120550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3204564817721182e-05,
+      "loss": 0.1819,
+      "step": 120560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3202322765795257e-05,
+      "loss": 0.1856,
+      "step": 120570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3200080713869336e-05,
+      "loss": 0.1804,
+      "step": 120580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.319783866194341e-05,
+      "loss": 0.1775,
+      "step": 120590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.319559661001749e-05,
+      "loss": 0.1796,
+      "step": 120600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.319335455809157e-05,
+      "loss": 0.1873,
+      "step": 120610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3191112506165644e-05,
+      "loss": 0.1785,
+      "step": 120620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3188870454239723e-05,
+      "loss": 0.1828,
+      "step": 120630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3186628402313798e-05,
+      "loss": 0.1892,
+      "step": 120640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3184386350387877e-05,
+      "loss": 0.1811,
+      "step": 120650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3182144298461952e-05,
+      "loss": 0.1857,
+      "step": 120660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.317990224653603e-05,
+      "loss": 0.1844,
+      "step": 120670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.317766019461011e-05,
+      "loss": 0.184,
+      "step": 120680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3175418142684185e-05,
+      "loss": 0.1861,
+      "step": 120690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3173176090758264e-05,
+      "loss": 0.1808,
+      "step": 120700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.317093403883234e-05,
+      "loss": 0.1846,
+      "step": 120710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3168691986906418e-05,
+      "loss": 0.1774,
+      "step": 120720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3166449934980493e-05,
+      "loss": 0.1828,
+      "step": 120730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3164207883054572e-05,
+      "loss": 0.1746,
+      "step": 120740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.316196583112865e-05,
+      "loss": 0.1877,
+      "step": 120750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3159723779202726e-05,
+      "loss": 0.1799,
+      "step": 120760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3157481727276805e-05,
+      "loss": 0.186,
+      "step": 120770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3155239675350883e-05,
+      "loss": 0.1811,
+      "step": 120780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.315299762342496e-05,
+      "loss": 0.1809,
+      "step": 120790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3150755571499037e-05,
+      "loss": 0.185,
+      "step": 120800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3148513519573116e-05,
+      "loss": 0.1867,
+      "step": 120810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3146271467647195e-05,
+      "loss": 0.1877,
+      "step": 120820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.314402941572127e-05,
+      "loss": 0.1796,
+      "step": 120830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.314178736379535e-05,
+      "loss": 0.1864,
+      "step": 120840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3139545311869424e-05,
+      "loss": 0.1782,
+      "step": 120850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3137303259943503e-05,
+      "loss": 0.1846,
+      "step": 120860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3135061208017578e-05,
+      "loss": 0.1864,
+      "step": 120870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3132819156091657e-05,
+      "loss": 0.1853,
+      "step": 120880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3130577104165736e-05,
+      "loss": 0.1826,
+      "step": 120890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.312833505223981e-05,
+      "loss": 0.1774,
+      "step": 120900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.312609300031389e-05,
+      "loss": 0.1827,
+      "step": 120910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3123850948387965e-05,
+      "loss": 0.183,
+      "step": 120920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3121608896462044e-05,
+      "loss": 0.1852,
+      "step": 120930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.311936684453612e-05,
+      "loss": 0.1733,
+      "step": 120940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3117124792610198e-05,
+      "loss": 0.1799,
+      "step": 120950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3114882740684276e-05,
+      "loss": 0.18,
+      "step": 120960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3112640688758352e-05,
+      "loss": 0.1862,
+      "step": 120970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.311039863683243e-05,
+      "loss": 0.1826,
+      "step": 120980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3108156584906506e-05,
+      "loss": 0.1831,
+      "step": 120990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3105914532980585e-05,
+      "loss": 0.1794,
+      "step": 121000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.310367248105466e-05,
+      "loss": 0.1791,
+      "step": 121010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.310143042912874e-05,
+      "loss": 0.184,
+      "step": 121020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3099188377202817e-05,
+      "loss": 0.1815,
+      "step": 121030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3096946325276893e-05,
+      "loss": 0.1877,
+      "step": 121040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.309470427335097e-05,
+      "loss": 0.1888,
+      "step": 121050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.309246222142505e-05,
+      "loss": 0.1862,
+      "step": 121060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.309022016949913e-05,
+      "loss": 0.1794,
+      "step": 121070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3087978117573204e-05,
+      "loss": 0.188,
+      "step": 121080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3085736065647283e-05,
+      "loss": 0.1782,
+      "step": 121090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.308349401372136e-05,
+      "loss": 0.1846,
+      "step": 121100
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3081251961795437e-05,
+      "loss": 0.1809,
+      "step": 121110
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3079009909869516e-05,
+      "loss": 0.1808,
+      "step": 121120
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.307676785794359e-05,
+      "loss": 0.1778,
+      "step": 121130
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.307452580601767e-05,
+      "loss": 0.1827,
+      "step": 121140
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3072283754091745e-05,
+      "loss": 0.1835,
+      "step": 121150
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3070041702165824e-05,
+      "loss": 0.1825,
+      "step": 121160
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3067799650239902e-05,
+      "loss": 0.184,
+      "step": 121170
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3065557598313978e-05,
+      "loss": 0.1884,
+      "step": 121180
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3063315546388056e-05,
+      "loss": 0.183,
+      "step": 121190
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3061073494462132e-05,
+      "loss": 0.1851,
+      "step": 121200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.305883144253621e-05,
+      "loss": 0.1839,
+      "step": 121210
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3056589390610286e-05,
+      "loss": 0.1873,
+      "step": 121220
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3054347338684365e-05,
+      "loss": 0.1856,
+      "step": 121230
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3052105286758443e-05,
+      "loss": 0.1806,
+      "step": 121240
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.304986323483252e-05,
+      "loss": 0.1795,
+      "step": 121250
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3047621182906597e-05,
+      "loss": 0.183,
+      "step": 121260
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3045379130980673e-05,
+      "loss": 0.1834,
+      "step": 121270
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.304313707905475e-05,
+      "loss": 0.178,
+      "step": 121280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3040895027128827e-05,
+      "loss": 0.1857,
+      "step": 121290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3038652975202905e-05,
+      "loss": 0.1804,
+      "step": 121300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3036410923276984e-05,
+      "loss": 0.1854,
+      "step": 121310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3034168871351063e-05,
+      "loss": 0.1868,
+      "step": 121320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3031926819425138e-05,
+      "loss": 0.1849,
+      "step": 121330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3029684767499217e-05,
+      "loss": 0.1857,
+      "step": 121340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3027442715573296e-05,
+      "loss": 0.1844,
+      "step": 121350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.302520066364737e-05,
+      "loss": 0.1817,
+      "step": 121360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.302295861172145e-05,
+      "loss": 0.1725,
+      "step": 121370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.302071655979553e-05,
+      "loss": 0.1824,
+      "step": 121380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3018474507869604e-05,
+      "loss": 0.1825,
+      "step": 121390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3016232455943682e-05,
+      "loss": 0.1827,
+      "step": 121400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3013990404017758e-05,
+      "loss": 0.1799,
+      "step": 121410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3011748352091836e-05,
+      "loss": 0.1821,
+      "step": 121420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3009506300165912e-05,
+      "loss": 0.1778,
+      "step": 121430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.300726424823999e-05,
+      "loss": 0.1841,
+      "step": 121440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.300502219631407e-05,
+      "loss": 0.1879,
+      "step": 121450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3002780144388145e-05,
+      "loss": 0.1867,
+      "step": 121460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3000538092462223e-05,
+      "loss": 0.1774,
+      "step": 121470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.29982960405363e-05,
+      "loss": 0.1814,
+      "step": 121480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2996053988610377e-05,
+      "loss": 0.1834,
+      "step": 121490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2993811936684453e-05,
+      "loss": 0.1785,
+      "step": 121500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.299156988475853e-05,
+      "loss": 0.1787,
+      "step": 121510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.298932783283261e-05,
+      "loss": 0.1877,
+      "step": 121520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2987085780906685e-05,
+      "loss": 0.1847,
+      "step": 121530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2984843728980764e-05,
+      "loss": 0.1784,
+      "step": 121540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.298260167705484e-05,
+      "loss": 0.1824,
+      "step": 121550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2980359625128918e-05,
+      "loss": 0.1861,
+      "step": 121560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2978117573202994e-05,
+      "loss": 0.1747,
+      "step": 121570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2975875521277072e-05,
+      "loss": 0.1834,
+      "step": 121580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.297363346935115e-05,
+      "loss": 0.1802,
+      "step": 121590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.297139141742523e-05,
+      "loss": 0.1788,
+      "step": 121600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.296914936549931e-05,
+      "loss": 0.1756,
+      "step": 121610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2966907313573384e-05,
+      "loss": 0.1797,
+      "step": 121620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2964665261647462e-05,
+      "loss": 0.1827,
+      "step": 121630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2962423209721538e-05,
+      "loss": 0.1844,
+      "step": 121640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2960181157795616e-05,
+      "loss": 0.1857,
+      "step": 121650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2957939105869695e-05,
+      "loss": 0.1773,
+      "step": 121660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.295569705394377e-05,
+      "loss": 0.1816,
+      "step": 121670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.295345500201785e-05,
+      "loss": 0.1863,
+      "step": 121680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2951212950091925e-05,
+      "loss": 0.1847,
+      "step": 121690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2948970898166003e-05,
+      "loss": 0.1861,
+      "step": 121700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.294672884624008e-05,
+      "loss": 0.1823,
+      "step": 121710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2944486794314157e-05,
+      "loss": 0.1783,
+      "step": 121720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2942244742388236e-05,
+      "loss": 0.1766,
+      "step": 121730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.294000269046231e-05,
+      "loss": 0.1839,
+      "step": 121740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.293776063853639e-05,
+      "loss": 0.1823,
+      "step": 121750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2935518586610465e-05,
+      "loss": 0.1807,
+      "step": 121760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2933276534684544e-05,
+      "loss": 0.1846,
+      "step": 121770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.293103448275862e-05,
+      "loss": 0.1825,
+      "step": 121780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2928792430832698e-05,
+      "loss": 0.1792,
+      "step": 121790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2926550378906777e-05,
+      "loss": 0.1761,
+      "step": 121800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2924308326980852e-05,
+      "loss": 0.1885,
+      "step": 121810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.292206627505493e-05,
+      "loss": 0.1771,
+      "step": 121820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2919824223129006e-05,
+      "loss": 0.1831,
+      "step": 121830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2917582171203085e-05,
+      "loss": 0.1864,
+      "step": 121840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2915340119277164e-05,
+      "loss": 0.1806,
+      "step": 121850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.291309806735124e-05,
+      "loss": 0.1785,
+      "step": 121860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2910856015425318e-05,
+      "loss": 0.1783,
+      "step": 121870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2908613963499397e-05,
+      "loss": 0.1786,
+      "step": 121880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2906371911573475e-05,
+      "loss": 0.18,
+      "step": 121890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.290412985964755e-05,
+      "loss": 0.1867,
+      "step": 121900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.290188780772163e-05,
+      "loss": 0.1787,
+      "step": 121910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2899645755795708e-05,
+      "loss": 0.1821,
+      "step": 121920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2897403703869783e-05,
+      "loss": 0.1751,
+      "step": 121930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2895161651943862e-05,
+      "loss": 0.1805,
+      "step": 121940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2892919600017937e-05,
+      "loss": 0.1866,
+      "step": 121950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2890677548092016e-05,
+      "loss": 0.1861,
+      "step": 121960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.288843549616609e-05,
+      "loss": 0.1819,
+      "step": 121970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.288619344424017e-05,
+      "loss": 0.1757,
+      "step": 121980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.288395139231425e-05,
+      "loss": 0.1844,
+      "step": 121990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2881709340388324e-05,
+      "loss": 0.1823,
+      "step": 122000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2879467288462403e-05,
+      "loss": 0.181,
+      "step": 122010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2877225236536478e-05,
+      "loss": 0.1868,
+      "step": 122020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2874983184610557e-05,
+      "loss": 0.187,
+      "step": 122030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2872741132684632e-05,
+      "loss": 0.1821,
+      "step": 122040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.287049908075871e-05,
+      "loss": 0.1724,
+      "step": 122050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.286825702883279e-05,
+      "loss": 0.179,
+      "step": 122060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2866014976906865e-05,
+      "loss": 0.1795,
+      "step": 122070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2863772924980944e-05,
+      "loss": 0.1844,
+      "step": 122080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.286153087305502e-05,
+      "loss": 0.1814,
+      "step": 122090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2859288821129098e-05,
+      "loss": 0.1835,
+      "step": 122100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2857046769203173e-05,
+      "loss": 0.177,
+      "step": 122110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2854804717277252e-05,
+      "loss": 0.1798,
+      "step": 122120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.285256266535133e-05,
+      "loss": 0.1774,
+      "step": 122130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.285032061342541e-05,
+      "loss": 0.1799,
+      "step": 122140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2848078561499488e-05,
+      "loss": 0.181,
+      "step": 122150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2845836509573563e-05,
+      "loss": 0.1824,
+      "step": 122160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2843594457647642e-05,
+      "loss": 0.1826,
+      "step": 122170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2841352405721717e-05,
+      "loss": 0.1826,
+      "step": 122180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2839110353795796e-05,
+      "loss": 0.1815,
+      "step": 122190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2836868301869875e-05,
+      "loss": 0.1863,
+      "step": 122200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.283462624994395e-05,
+      "loss": 0.1825,
+      "step": 122210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.283238419801803e-05,
+      "loss": 0.1804,
+      "step": 122220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2830142146092104e-05,
+      "loss": 0.1752,
+      "step": 122230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2827900094166183e-05,
+      "loss": 0.1808,
+      "step": 122240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2825658042240258e-05,
+      "loss": 0.1795,
+      "step": 122250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2823415990314337e-05,
+      "loss": 0.1855,
+      "step": 122260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2821173938388416e-05,
+      "loss": 0.1807,
+      "step": 122270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.281893188646249e-05,
+      "loss": 0.1852,
+      "step": 122280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.281668983453657e-05,
+      "loss": 0.1889,
+      "step": 122290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2814447782610645e-05,
+      "loss": 0.1826,
+      "step": 122300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2812205730684724e-05,
+      "loss": 0.1794,
+      "step": 122310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.28099636787588e-05,
+      "loss": 0.1804,
+      "step": 122320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2807721626832878e-05,
+      "loss": 0.1806,
+      "step": 122330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2805479574906957e-05,
+      "loss": 0.1796,
+      "step": 122340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2803237522981032e-05,
+      "loss": 0.1854,
+      "step": 122350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.280099547105511e-05,
+      "loss": 0.1844,
+      "step": 122360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2798753419129186e-05,
+      "loss": 0.1819,
+      "step": 122370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2796511367203265e-05,
+      "loss": 0.182,
+      "step": 122380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2794269315277343e-05,
+      "loss": 0.1855,
+      "step": 122390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.279202726335142e-05,
+      "loss": 0.1807,
+      "step": 122400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2789785211425497e-05,
+      "loss": 0.1839,
+      "step": 122410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2787543159499576e-05,
+      "loss": 0.1803,
+      "step": 122420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2785301107573655e-05,
+      "loss": 0.1816,
+      "step": 122430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.278305905564773e-05,
+      "loss": 0.1842,
+      "step": 122440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.278081700372181e-05,
+      "loss": 0.1851,
+      "step": 122450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2778574951795884e-05,
+      "loss": 0.1906,
+      "step": 122460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2776332899869963e-05,
+      "loss": 0.1785,
+      "step": 122470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.277409084794404e-05,
+      "loss": 0.18,
+      "step": 122480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2771848796018117e-05,
+      "loss": 0.1871,
+      "step": 122490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2769606744092196e-05,
+      "loss": 0.1838,
+      "step": 122500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.276736469216627e-05,
+      "loss": 0.1808,
+      "step": 122510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.276512264024035e-05,
+      "loss": 0.1774,
+      "step": 122520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2762880588314425e-05,
+      "loss": 0.1856,
+      "step": 122530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2760638536388504e-05,
+      "loss": 0.1881,
+      "step": 122540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2758620689655175e-05,
+      "loss": 0.1813,
+      "step": 122550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.275637863772925e-05,
+      "loss": 0.1778,
+      "step": 122560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.275413658580333e-05,
+      "loss": 0.1804,
+      "step": 122570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2751894533877404e-05,
+      "loss": 0.1789,
+      "step": 122580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2749652481951483e-05,
+      "loss": 0.1828,
+      "step": 122590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.274741043002556e-05,
+      "loss": 0.1831,
+      "step": 122600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2745168378099637e-05,
+      "loss": 0.1798,
+      "step": 122610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2742926326173716e-05,
+      "loss": 0.1834,
+      "step": 122620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.274068427424779e-05,
+      "loss": 0.1836,
+      "step": 122630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.273844222232187e-05,
+      "loss": 0.1823,
+      "step": 122640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2736200170395945e-05,
+      "loss": 0.1828,
+      "step": 122650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2733958118470024e-05,
+      "loss": 0.1764,
+      "step": 122660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.27317160665441e-05,
+      "loss": 0.1795,
+      "step": 122670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2729474014618178e-05,
+      "loss": 0.1816,
+      "step": 122680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2727231962692257e-05,
+      "loss": 0.1808,
+      "step": 122690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2724989910766335e-05,
+      "loss": 0.1833,
+      "step": 122700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2722747858840414e-05,
+      "loss": 0.1823,
+      "step": 122710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.272050580691449e-05,
+      "loss": 0.1892,
+      "step": 122720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2718263754988568e-05,
+      "loss": 0.1772,
+      "step": 122730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2716021703062644e-05,
+      "loss": 0.1767,
+      "step": 122740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2713779651136722e-05,
+      "loss": 0.1817,
+      "step": 122750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.27115375992108e-05,
+      "loss": 0.1845,
+      "step": 122760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2709295547284876e-05,
+      "loss": 0.1838,
+      "step": 122770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2707053495358955e-05,
+      "loss": 0.1794,
+      "step": 122780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.270481144343303e-05,
+      "loss": 0.1794,
+      "step": 122790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.270256939150711e-05,
+      "loss": 0.1822,
+      "step": 122800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2700327339581184e-05,
+      "loss": 0.1805,
+      "step": 122810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2698085287655263e-05,
+      "loss": 0.1816,
+      "step": 122820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2695843235729342e-05,
+      "loss": 0.1864,
+      "step": 122830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2693601183803417e-05,
+      "loss": 0.1784,
+      "step": 122840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2691359131877496e-05,
+      "loss": 0.1831,
+      "step": 122850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.268911707995157e-05,
+      "loss": 0.1815,
+      "step": 122860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.268687502802565e-05,
+      "loss": 0.1853,
+      "step": 122870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.268463297609973e-05,
+      "loss": 0.1853,
+      "step": 122880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2682390924173804e-05,
+      "loss": 0.1807,
+      "step": 122890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2680148872247883e-05,
+      "loss": 0.1797,
+      "step": 122900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2677906820321958e-05,
+      "loss": 0.1768,
+      "step": 122910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2675664768396037e-05,
+      "loss": 0.1822,
+      "step": 122920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2673422716470112e-05,
+      "loss": 0.1805,
+      "step": 122930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.267118066454419e-05,
+      "loss": 0.1818,
+      "step": 122940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.266893861261827e-05,
+      "loss": 0.1778,
+      "step": 122950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2666696560692345e-05,
+      "loss": 0.1797,
+      "step": 122960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2664454508766424e-05,
+      "loss": 0.1862,
+      "step": 122970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2662212456840502e-05,
+      "loss": 0.1798,
+      "step": 122980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.265997040491458e-05,
+      "loss": 0.1842,
+      "step": 122990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2657728352988656e-05,
+      "loss": 0.1805,
+      "step": 123000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2655486301062735e-05,
+      "loss": 0.1828,
+      "step": 123010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2653244249136814e-05,
+      "loss": 0.1793,
+      "step": 123020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.265100219721089e-05,
+      "loss": 0.1728,
+      "step": 123030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2648760145284968e-05,
+      "loss": 0.184,
+      "step": 123040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2646518093359043e-05,
+      "loss": 0.1821,
+      "step": 123050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2644276041433122e-05,
+      "loss": 0.1813,
+      "step": 123060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2642033989507197e-05,
+      "loss": 0.1811,
+      "step": 123070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2639791937581276e-05,
+      "loss": 0.1842,
+      "step": 123080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2637549885655355e-05,
+      "loss": 0.1831,
+      "step": 123090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.263530783372943e-05,
+      "loss": 0.1829,
+      "step": 123100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.263306578180351e-05,
+      "loss": 0.1861,
+      "step": 123110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2630823729877584e-05,
+      "loss": 0.1816,
+      "step": 123120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2628581677951663e-05,
+      "loss": 0.1804,
+      "step": 123130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2626339626025738e-05,
+      "loss": 0.1768,
+      "step": 123140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2624097574099817e-05,
+      "loss": 0.1772,
+      "step": 123150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2621855522173895e-05,
+      "loss": 0.1802,
+      "step": 123160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.261961347024797e-05,
+      "loss": 0.1851,
+      "step": 123170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.261737141832205e-05,
+      "loss": 0.1836,
+      "step": 123180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2615129366396125e-05,
+      "loss": 0.1825,
+      "step": 123190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2612887314470204e-05,
+      "loss": 0.182,
+      "step": 123200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.261064526254428e-05,
+      "loss": 0.1855,
+      "step": 123210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2608403210618358e-05,
+      "loss": 0.1766,
+      "step": 123220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2606161158692436e-05,
+      "loss": 0.1797,
+      "step": 123230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2603919106766515e-05,
+      "loss": 0.1785,
+      "step": 123240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2601677054840594e-05,
+      "loss": 0.1765,
+      "step": 123250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.259943500291467e-05,
+      "loss": 0.1757,
+      "step": 123260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2597192950988748e-05,
+      "loss": 0.1801,
+      "step": 123270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2594950899062823e-05,
+      "loss": 0.1782,
+      "step": 123280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2592708847136902e-05,
+      "loss": 0.1841,
+      "step": 123290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.259046679521098e-05,
+      "loss": 0.182,
+      "step": 123300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2588224743285056e-05,
+      "loss": 0.1867,
+      "step": 123310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2585982691359135e-05,
+      "loss": 0.182,
+      "step": 123320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.258374063943321e-05,
+      "loss": 0.1879,
+      "step": 123330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.258149858750729e-05,
+      "loss": 0.1797,
+      "step": 123340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2579256535581364e-05,
+      "loss": 0.1795,
+      "step": 123350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2577014483655443e-05,
+      "loss": 0.1807,
+      "step": 123360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.257477243172952e-05,
+      "loss": 0.1828,
+      "step": 123370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2572530379803597e-05,
+      "loss": 0.1802,
+      "step": 123380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2570288327877675e-05,
+      "loss": 0.181,
+      "step": 123390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.256804627595175e-05,
+      "loss": 0.1817,
+      "step": 123400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.256580422402583e-05,
+      "loss": 0.1784,
+      "step": 123410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2563562172099905e-05,
+      "loss": 0.1837,
+      "step": 123420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2561320120173984e-05,
+      "loss": 0.1788,
+      "step": 123430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2559078068248062e-05,
+      "loss": 0.1801,
+      "step": 123440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2556836016322138e-05,
+      "loss": 0.1821,
+      "step": 123450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2554593964396216e-05,
+      "loss": 0.1808,
+      "step": 123460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.255235191247029e-05,
+      "loss": 0.1794,
+      "step": 123470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.255010986054437e-05,
+      "loss": 0.1857,
+      "step": 123480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.254786780861845e-05,
+      "loss": 0.1856,
+      "step": 123490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2545625756692524e-05,
+      "loss": 0.1786,
+      "step": 123500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2543383704766603e-05,
+      "loss": 0.1797,
+      "step": 123510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2541141652840682e-05,
+      "loss": 0.1835,
+      "step": 123520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.253889960091476e-05,
+      "loss": 0.1773,
+      "step": 123530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2536657548988836e-05,
+      "loss": 0.1789,
+      "step": 123540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2534415497062915e-05,
+      "loss": 0.1809,
+      "step": 123550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.253217344513699e-05,
+      "loss": 0.1825,
+      "step": 123560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.252993139321107e-05,
+      "loss": 0.1824,
+      "step": 123570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2527689341285147e-05,
+      "loss": 0.1813,
+      "step": 123580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2525447289359223e-05,
+      "loss": 0.1827,
+      "step": 123590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.25232052374333e-05,
+      "loss": 0.1825,
+      "step": 123600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2520963185507377e-05,
+      "loss": 0.1854,
+      "step": 123610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2518721133581455e-05,
+      "loss": 0.1817,
+      "step": 123620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.251647908165553e-05,
+      "loss": 0.183,
+      "step": 123630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.251423702972961e-05,
+      "loss": 0.1776,
+      "step": 123640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2511994977803688e-05,
+      "loss": 0.1812,
+      "step": 123650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2509752925877764e-05,
+      "loss": 0.1787,
+      "step": 123660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2507510873951842e-05,
+      "loss": 0.1809,
+      "step": 123670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2505268822025918e-05,
+      "loss": 0.1805,
+      "step": 123680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2503026770099996e-05,
+      "loss": 0.184,
+      "step": 123690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.250078471817407e-05,
+      "loss": 0.1801,
+      "step": 123700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.249854266624815e-05,
+      "loss": 0.1864,
+      "step": 123710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.249630061432223e-05,
+      "loss": 0.1839,
+      "step": 123720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2494058562396304e-05,
+      "loss": 0.1746,
+      "step": 123730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2491816510470383e-05,
+      "loss": 0.1782,
+      "step": 123740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.248957445854446e-05,
+      "loss": 0.1801,
+      "step": 123750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2487332406618537e-05,
+      "loss": 0.1833,
+      "step": 123760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2485090354692616e-05,
+      "loss": 0.1759,
+      "step": 123770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2482848302766695e-05,
+      "loss": 0.1808,
+      "step": 123780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2480606250840773e-05,
+      "loss": 0.1807,
+      "step": 123790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.247836419891485e-05,
+      "loss": 0.1804,
+      "step": 123800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2476122146988927e-05,
+      "loss": 0.1806,
+      "step": 123810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2473880095063003e-05,
+      "loss": 0.1765,
+      "step": 123820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.247163804313708e-05,
+      "loss": 0.1767,
+      "step": 123830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2469395991211157e-05,
+      "loss": 0.1778,
+      "step": 123840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2467153939285235e-05,
+      "loss": 0.1808,
+      "step": 123850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2464911887359314e-05,
+      "loss": 0.1825,
+      "step": 123860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.246266983543339e-05,
+      "loss": 0.1814,
+      "step": 123870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2460427783507468e-05,
+      "loss": 0.1769,
+      "step": 123880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2458185731581544e-05,
+      "loss": 0.1799,
+      "step": 123890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2455943679655622e-05,
+      "loss": 0.1861,
+      "step": 123900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2453701627729698e-05,
+      "loss": 0.1789,
+      "step": 123910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2451459575803776e-05,
+      "loss": 0.1795,
+      "step": 123920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2449217523877855e-05,
+      "loss": 0.1802,
+      "step": 123930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.244697547195193e-05,
+      "loss": 0.1789,
+      "step": 123940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.244473342002601e-05,
+      "loss": 0.1831,
+      "step": 123950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2442491368100084e-05,
+      "loss": 0.1823,
+      "step": 123960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2440249316174163e-05,
+      "loss": 0.1831,
+      "step": 123970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.243800726424824e-05,
+      "loss": 0.181,
+      "step": 123980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2435765212322317e-05,
+      "loss": 0.1826,
+      "step": 123990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2433523160396396e-05,
+      "loss": 0.1796,
+      "step": 124000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.243128110847047e-05,
+      "loss": 0.1772,
+      "step": 124010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.242903905654455e-05,
+      "loss": 0.1789,
+      "step": 124020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.242679700461863e-05,
+      "loss": 0.1794,
+      "step": 124030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2424554952692704e-05,
+      "loss": 0.177,
+      "step": 124040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2422312900766783e-05,
+      "loss": 0.1806,
+      "step": 124050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.242007084884086e-05,
+      "loss": 0.1807,
+      "step": 124060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.241782879691494e-05,
+      "loss": 0.1823,
+      "step": 124070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2415586744989016e-05,
+      "loss": 0.1787,
+      "step": 124080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2413344693063094e-05,
+      "loss": 0.1815,
+      "step": 124090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.241110264113717e-05,
+      "loss": 0.179,
+      "step": 124100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2408860589211248e-05,
+      "loss": 0.1806,
+      "step": 124110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2406618537285324e-05,
+      "loss": 0.1756,
+      "step": 124120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2404376485359402e-05,
+      "loss": 0.1895,
+      "step": 124130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.240213443343348e-05,
+      "loss": 0.1752,
+      "step": 124140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2399892381507556e-05,
+      "loss": 0.1858,
+      "step": 124150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2397650329581635e-05,
+      "loss": 0.185,
+      "step": 124160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.239540827765571e-05,
+      "loss": 0.1799,
+      "step": 124170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.239316622572979e-05,
+      "loss": 0.1806,
+      "step": 124180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2390924173803868e-05,
+      "loss": 0.1761,
+      "step": 124190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2388682121877943e-05,
+      "loss": 0.1808,
+      "step": 124200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2386440069952022e-05,
+      "loss": 0.1749,
+      "step": 124210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2384198018026097e-05,
+      "loss": 0.181,
+      "step": 124220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2381955966100176e-05,
+      "loss": 0.1859,
+      "step": 124230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.237971391417425e-05,
+      "loss": 0.1771,
+      "step": 124240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.237747186224833e-05,
+      "loss": 0.1846,
+      "step": 124250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.237522981032241e-05,
+      "loss": 0.1803,
+      "step": 124260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2372987758396484e-05,
+      "loss": 0.1827,
+      "step": 124270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2370745706470563e-05,
+      "loss": 0.1825,
+      "step": 124280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2368503654544638e-05,
+      "loss": 0.1759,
+      "step": 124290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2366261602618717e-05,
+      "loss": 0.1776,
+      "step": 124300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2364019550692796e-05,
+      "loss": 0.1796,
+      "step": 124310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2361777498766874e-05,
+      "loss": 0.1818,
+      "step": 124320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.235953544684095e-05,
+      "loss": 0.1764,
+      "step": 124330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2357293394915028e-05,
+      "loss": 0.1762,
+      "step": 124340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2355051342989107e-05,
+      "loss": 0.1795,
+      "step": 124350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2352809291063182e-05,
+      "loss": 0.1806,
+      "step": 124360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.235056723913726e-05,
+      "loss": 0.1825,
+      "step": 124370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2348325187211336e-05,
+      "loss": 0.1825,
+      "step": 124380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2346083135285415e-05,
+      "loss": 0.1783,
+      "step": 124390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2343841083359494e-05,
+      "loss": 0.1794,
+      "step": 124400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.234159903143357e-05,
+      "loss": 0.1762,
+      "step": 124410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2339356979507648e-05,
+      "loss": 0.1825,
+      "step": 124420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2337114927581723e-05,
+      "loss": 0.1811,
+      "step": 124430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2334872875655802e-05,
+      "loss": 0.1807,
+      "step": 124440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2332630823729877e-05,
+      "loss": 0.1816,
+      "step": 124450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2330388771803956e-05,
+      "loss": 0.1789,
+      "step": 124460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2328146719878035e-05,
+      "loss": 0.1794,
+      "step": 124470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.232590466795211e-05,
+      "loss": 0.183,
+      "step": 124480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.232366261602619e-05,
+      "loss": 0.1772,
+      "step": 124490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2321420564100264e-05,
+      "loss": 0.1757,
+      "step": 124500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2319178512174343e-05,
+      "loss": 0.1848,
+      "step": 124510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2316936460248418e-05,
+      "loss": 0.1809,
+      "step": 124520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2314694408322497e-05,
+      "loss": 0.1764,
+      "step": 124530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2312452356396576e-05,
+      "loss": 0.1812,
+      "step": 124540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.231021030447065e-05,
+      "loss": 0.1823,
+      "step": 124550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.230796825254473e-05,
+      "loss": 0.1791,
+      "step": 124560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2305726200618808e-05,
+      "loss": 0.177,
+      "step": 124570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2303484148692884e-05,
+      "loss": 0.1807,
+      "step": 124580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2301242096766962e-05,
+      "loss": 0.1825,
+      "step": 124590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.229900004484104e-05,
+      "loss": 0.1833,
+      "step": 124600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.229675799291512e-05,
+      "loss": 0.1829,
+      "step": 124610
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2294515940989195e-05,
+      "loss": 0.1781,
+      "step": 124620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2292273889063274e-05,
+      "loss": 0.1825,
+      "step": 124630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.229003183713735e-05,
+      "loss": 0.1843,
+      "step": 124640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2287789785211428e-05,
+      "loss": 0.183,
+      "step": 124650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2285547733285503e-05,
+      "loss": 0.1803,
+      "step": 124660
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2283305681359582e-05,
+      "loss": 0.1847,
+      "step": 124670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.228106362943366e-05,
+      "loss": 0.1809,
+      "step": 124680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2278821577507736e-05,
+      "loss": 0.1827,
+      "step": 124690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2276579525581815e-05,
+      "loss": 0.1794,
+      "step": 124700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.227433747365589e-05,
+      "loss": 0.1889,
+      "step": 124710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.227209542172997e-05,
+      "loss": 0.1815,
+      "step": 124720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2269853369804044e-05,
+      "loss": 0.1799,
+      "step": 124730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2267611317878123e-05,
+      "loss": 0.1739,
+      "step": 124740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.22653692659522e-05,
+      "loss": 0.176,
+      "step": 124750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2263127214026277e-05,
+      "loss": 0.1746,
+      "step": 124760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2260885162100356e-05,
+      "loss": 0.1777,
+      "step": 124770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.225864311017443e-05,
+      "loss": 0.1824,
+      "step": 124780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.225640105824851e-05,
+      "loss": 0.1826,
+      "step": 124790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2254159006322585e-05,
+      "loss": 0.1828,
+      "step": 124800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2251916954396664e-05,
+      "loss": 0.1781,
+      "step": 124810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2249674902470742e-05,
+      "loss": 0.176,
+      "step": 124820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2247432850544818e-05,
+      "loss": 0.18,
+      "step": 124830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2245190798618896e-05,
+      "loss": 0.188,
+      "step": 124840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2242948746692975e-05,
+      "loss": 0.1864,
+      "step": 124850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2240706694767054e-05,
+      "loss": 0.1836,
+      "step": 124860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.223846464284113e-05,
+      "loss": 0.1754,
+      "step": 124870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2236222590915208e-05,
+      "loss": 0.1806,
+      "step": 124880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2233980538989287e-05,
+      "loss": 0.1779,
+      "step": 124890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2231738487063362e-05,
+      "loss": 0.1777,
+      "step": 124900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.222949643513744e-05,
+      "loss": 0.1754,
+      "step": 124910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2227254383211516e-05,
+      "loss": 0.1787,
+      "step": 124920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2225012331285595e-05,
+      "loss": 0.1839,
+      "step": 124930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.222277027935967e-05,
+      "loss": 0.179,
+      "step": 124940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.222052822743375e-05,
+      "loss": 0.1794,
+      "step": 124950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2218286175507827e-05,
+      "loss": 0.1808,
+      "step": 124960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2216044123581903e-05,
+      "loss": 0.1811,
+      "step": 124970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.221380207165598e-05,
+      "loss": 0.1814,
+      "step": 124980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2211560019730057e-05,
+      "loss": 0.1804,
+      "step": 124990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2209317967804136e-05,
+      "loss": 0.1775,
+      "step": 125000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.220707591587821e-05,
+      "loss": 0.1848,
+      "step": 125010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.220483386395229e-05,
+      "loss": 0.1734,
+      "step": 125020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.220259181202637e-05,
+      "loss": 0.176,
+      "step": 125030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2200349760100444e-05,
+      "loss": 0.1773,
+      "step": 125040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2198107708174522e-05,
+      "loss": 0.1747,
+      "step": 125050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2195865656248598e-05,
+      "loss": 0.175,
+      "step": 125060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2193623604322676e-05,
+      "loss": 0.1808,
+      "step": 125070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2191381552396752e-05,
+      "loss": 0.1865,
+      "step": 125080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.218913950047083e-05,
+      "loss": 0.1831,
+      "step": 125090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.218689744854491e-05,
+      "loss": 0.1829,
+      "step": 125100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2184655396618988e-05,
+      "loss": 0.1717,
+      "step": 125110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2182413344693063e-05,
+      "loss": 0.1827,
+      "step": 125120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2180171292767142e-05,
+      "loss": 0.1805,
+      "step": 125130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.217792924084122e-05,
+      "loss": 0.1717,
+      "step": 125140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2175687188915296e-05,
+      "loss": 0.1819,
+      "step": 125150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2173445136989375e-05,
+      "loss": 0.1813,
+      "step": 125160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2171203085063453e-05,
+      "loss": 0.1845,
+      "step": 125170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.216896103313753e-05,
+      "loss": 0.1773,
+      "step": 125180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2166718981211607e-05,
+      "loss": 0.1813,
+      "step": 125190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2164476929285683e-05,
+      "loss": 0.1711,
+      "step": 125200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.216223487735976e-05,
+      "loss": 0.1844,
+      "step": 125210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2159992825433837e-05,
+      "loss": 0.1807,
+      "step": 125220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2157750773507916e-05,
+      "loss": 0.1796,
+      "step": 125230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2155508721581994e-05,
+      "loss": 0.184,
+      "step": 125240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.215326666965607e-05,
+      "loss": 0.1818,
+      "step": 125250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.215102461773015e-05,
+      "loss": 0.1772,
+      "step": 125260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2148782565804224e-05,
+      "loss": 0.1737,
+      "step": 125270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2146540513878302e-05,
+      "loss": 0.1815,
+      "step": 125280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2144298461952378e-05,
+      "loss": 0.1788,
+      "step": 125290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2142056410026456e-05,
+      "loss": 0.1864,
+      "step": 125300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2139814358100535e-05,
+      "loss": 0.1782,
+      "step": 125310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.213757230617461e-05,
+      "loss": 0.1792,
+      "step": 125320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.213533025424869e-05,
+      "loss": 0.1834,
+      "step": 125330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2133088202322765e-05,
+      "loss": 0.1726,
+      "step": 125340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2130846150396843e-05,
+      "loss": 0.1785,
+      "step": 125350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2128604098470922e-05,
+      "loss": 0.1739,
+      "step": 125360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2126362046544997e-05,
+      "loss": 0.1823,
+      "step": 125370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2124119994619076e-05,
+      "loss": 0.1748,
+      "step": 125380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2121877942693155e-05,
+      "loss": 0.1812,
+      "step": 125390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2119635890767233e-05,
+      "loss": 0.1737,
+      "step": 125400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.211739383884131e-05,
+      "loss": 0.1841,
+      "step": 125410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2115151786915387e-05,
+      "loss": 0.1827,
+      "step": 125420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2112909734989463e-05,
+      "loss": 0.1845,
+      "step": 125430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.211066768306354e-05,
+      "loss": 0.1779,
+      "step": 125440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.210842563113762e-05,
+      "loss": 0.1814,
+      "step": 125450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2106183579211696e-05,
+      "loss": 0.1798,
+      "step": 125460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2103941527285774e-05,
+      "loss": 0.1793,
+      "step": 125470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.210169947535985e-05,
+      "loss": 0.1792,
+      "step": 125480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.209945742343393e-05,
+      "loss": 0.1801,
+      "step": 125490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2097215371508004e-05,
+      "loss": 0.1809,
+      "step": 125500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2094973319582082e-05,
+      "loss": 0.1789,
+      "step": 125510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.209273126765616e-05,
+      "loss": 0.1769,
+      "step": 125520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2090489215730236e-05,
+      "loss": 0.1785,
+      "step": 125530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2088247163804315e-05,
+      "loss": 0.18,
+      "step": 125540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.208600511187839e-05,
+      "loss": 0.1842,
+      "step": 125550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.208376305995247e-05,
+      "loss": 0.1806,
+      "step": 125560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2081521008026548e-05,
+      "loss": 0.1789,
+      "step": 125570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2079278956100623e-05,
+      "loss": 0.1809,
+      "step": 125580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2077036904174702e-05,
+      "loss": 0.1774,
+      "step": 125590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2074794852248777e-05,
+      "loss": 0.1826,
+      "step": 125600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2072552800322856e-05,
+      "loss": 0.1803,
+      "step": 125610
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.207031074839693e-05,
+      "loss": 0.1812,
+      "step": 125620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.206806869647101e-05,
+      "loss": 0.1869,
+      "step": 125630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.206582664454509e-05,
+      "loss": 0.1853,
+      "step": 125640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2063584592619168e-05,
+      "loss": 0.1829,
+      "step": 125650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2061342540693243e-05,
+      "loss": 0.1753,
+      "step": 125660
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.205910048876732e-05,
+      "loss": 0.1788,
+      "step": 125670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.20568584368414e-05,
+      "loss": 0.1851,
+      "step": 125680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2054616384915476e-05,
+      "loss": 0.1789,
+      "step": 125690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2052374332989554e-05,
+      "loss": 0.1796,
+      "step": 125700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2050132281063633e-05,
+      "loss": 0.1819,
+      "step": 125710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.204789022913771e-05,
+      "loss": 0.1813,
+      "step": 125720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2045648177211787e-05,
+      "loss": 0.1802,
+      "step": 125730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2043406125285862e-05,
+      "loss": 0.1786,
+      "step": 125740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.204116407335994e-05,
+      "loss": 0.1793,
+      "step": 125750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2038922021434016e-05,
+      "loss": 0.183,
+      "step": 125760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2036679969508095e-05,
+      "loss": 0.1779,
+      "step": 125770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2034437917582174e-05,
+      "loss": 0.1793,
+      "step": 125780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.203219586565625e-05,
+      "loss": 0.1804,
+      "step": 125790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2029953813730328e-05,
+      "loss": 0.1779,
+      "step": 125800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2027711761804403e-05,
+      "loss": 0.1854,
+      "step": 125810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2025469709878482e-05,
+      "loss": 0.1783,
+      "step": 125820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2023227657952557e-05,
+      "loss": 0.179,
+      "step": 125830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2020985606026636e-05,
+      "loss": 0.1819,
+      "step": 125840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2018743554100715e-05,
+      "loss": 0.178,
+      "step": 125850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.201650150217479e-05,
+      "loss": 0.1802,
+      "step": 125860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.201425945024887e-05,
+      "loss": 0.1823,
+      "step": 125870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2012017398322944e-05,
+      "loss": 0.1788,
+      "step": 125880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2009775346397023e-05,
+      "loss": 0.1821,
+      "step": 125890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.20075332944711e-05,
+      "loss": 0.1755,
+      "step": 125900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2005291242545177e-05,
+      "loss": 0.1792,
+      "step": 125910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2003049190619256e-05,
+      "loss": 0.1733,
+      "step": 125920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.2000807138693334e-05,
+      "loss": 0.1796,
+      "step": 125930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1998565086767413e-05,
+      "loss": 0.1802,
+      "step": 125940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.199632303484149e-05,
+      "loss": 0.1761,
+      "step": 125950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1994080982915567e-05,
+      "loss": 0.1807,
+      "step": 125960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1992063136182235e-05,
+      "loss": 0.1852,
+      "step": 125970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1989821084256314e-05,
+      "loss": 0.1846,
+      "step": 125980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1987579032330392e-05,
+      "loss": 0.1785,
+      "step": 125990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1985336980404468e-05,
+      "loss": 0.1814,
+      "step": 126000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1983094928478546e-05,
+      "loss": 0.1833,
+      "step": 126010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1980852876552622e-05,
+      "loss": 0.1798,
+      "step": 126020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.19786108246267e-05,
+      "loss": 0.1831,
+      "step": 126030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1976368772700776e-05,
+      "loss": 0.1762,
+      "step": 126040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1974126720774854e-05,
+      "loss": 0.1771,
+      "step": 126050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1971884668848933e-05,
+      "loss": 0.176,
+      "step": 126060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.196964261692301e-05,
+      "loss": 0.1839,
+      "step": 126070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1967400564997087e-05,
+      "loss": 0.1793,
+      "step": 126080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1965158513071163e-05,
+      "loss": 0.1794,
+      "step": 126090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.196291646114524e-05,
+      "loss": 0.188,
+      "step": 126100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1960674409219317e-05,
+      "loss": 0.179,
+      "step": 126110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1958432357293395e-05,
+      "loss": 0.1781,
+      "step": 126120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1956190305367474e-05,
+      "loss": 0.1771,
+      "step": 126130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.195394825344155e-05,
+      "loss": 0.1803,
+      "step": 126140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1951706201515628e-05,
+      "loss": 0.1789,
+      "step": 126150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1949464149589703e-05,
+      "loss": 0.1779,
+      "step": 126160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1947222097663782e-05,
+      "loss": 0.1789,
+      "step": 126170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1944980045737857e-05,
+      "loss": 0.187,
+      "step": 126180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1942737993811936e-05,
+      "loss": 0.1809,
+      "step": 126190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1940495941886015e-05,
+      "loss": 0.1807,
+      "step": 126200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1938253889960094e-05,
+      "loss": 0.1759,
+      "step": 126210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.193601183803417e-05,
+      "loss": 0.1793,
+      "step": 126220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1933769786108248e-05,
+      "loss": 0.1792,
+      "step": 126230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1931527734182326e-05,
+      "loss": 0.1797,
+      "step": 126240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1929285682256402e-05,
+      "loss": 0.1813,
+      "step": 126250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.192704363033048e-05,
+      "loss": 0.1849,
+      "step": 126260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.192480157840456e-05,
+      "loss": 0.1797,
+      "step": 126270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1922559526478634e-05,
+      "loss": 0.1832,
+      "step": 126280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1920317474552713e-05,
+      "loss": 0.1812,
+      "step": 126290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.191807542262679e-05,
+      "loss": 0.1681,
+      "step": 126300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1915833370700867e-05,
+      "loss": 0.1781,
+      "step": 126310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1913591318774943e-05,
+      "loss": 0.1807,
+      "step": 126320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.191134926684902e-05,
+      "loss": 0.1729,
+      "step": 126330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.19091072149231e-05,
+      "loss": 0.1817,
+      "step": 126340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1906865162997175e-05,
+      "loss": 0.1805,
+      "step": 126350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1904623111071254e-05,
+      "loss": 0.1749,
+      "step": 126360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.190238105914533e-05,
+      "loss": 0.1752,
+      "step": 126370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1900139007219408e-05,
+      "loss": 0.1854,
+      "step": 126380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1897896955293483e-05,
+      "loss": 0.1828,
+      "step": 126390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1895654903367562e-05,
+      "loss": 0.189,
+      "step": 126400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.189341285144164e-05,
+      "loss": 0.1841,
+      "step": 126410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1891170799515716e-05,
+      "loss": 0.1799,
+      "step": 126420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1888928747589795e-05,
+      "loss": 0.1838,
+      "step": 126430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.188668669566387e-05,
+      "loss": 0.1781,
+      "step": 126440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.188444464373795e-05,
+      "loss": 0.1857,
+      "step": 126450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1882202591812028e-05,
+      "loss": 0.1765,
+      "step": 126460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1879960539886103e-05,
+      "loss": 0.1779,
+      "step": 126470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1877718487960182e-05,
+      "loss": 0.1793,
+      "step": 126480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.187547643603426e-05,
+      "loss": 0.1836,
+      "step": 126490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.187323438410834e-05,
+      "loss": 0.1803,
+      "step": 126500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1870992332182415e-05,
+      "loss": 0.1734,
+      "step": 126510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1868750280256493e-05,
+      "loss": 0.1752,
+      "step": 126520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1866508228330572e-05,
+      "loss": 0.174,
+      "step": 126530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1864266176404647e-05,
+      "loss": 0.1807,
+      "step": 126540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1862024124478726e-05,
+      "loss": 0.1814,
+      "step": 126550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.18597820725528e-05,
+      "loss": 0.1804,
+      "step": 126560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.185754002062688e-05,
+      "loss": 0.1811,
+      "step": 126570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1855297968700955e-05,
+      "loss": 0.1756,
+      "step": 126580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1853055916775034e-05,
+      "loss": 0.1812,
+      "step": 126590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1850813864849113e-05,
+      "loss": 0.1773,
+      "step": 126600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1848571812923188e-05,
+      "loss": 0.1836,
+      "step": 126610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1846329760997267e-05,
+      "loss": 0.1802,
+      "step": 126620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1844087709071342e-05,
+      "loss": 0.1778,
+      "step": 126630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.184184565714542e-05,
+      "loss": 0.1774,
+      "step": 126640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1839603605219496e-05,
+      "loss": 0.1789,
+      "step": 126650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1837361553293575e-05,
+      "loss": 0.1817,
+      "step": 126660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1835119501367654e-05,
+      "loss": 0.1795,
+      "step": 126670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.183287744944173e-05,
+      "loss": 0.1776,
+      "step": 126680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1830635397515808e-05,
+      "loss": 0.1795,
+      "step": 126690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1828393345589883e-05,
+      "loss": 0.1839,
+      "step": 126700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1826151293663962e-05,
+      "loss": 0.1799,
+      "step": 126710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1823909241738037e-05,
+      "loss": 0.1772,
+      "step": 126720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1821667189812116e-05,
+      "loss": 0.179,
+      "step": 126730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1819425137886195e-05,
+      "loss": 0.1773,
+      "step": 126740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1817183085960273e-05,
+      "loss": 0.1778,
+      "step": 126750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.181494103403435e-05,
+      "loss": 0.1765,
+      "step": 126760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1812698982108427e-05,
+      "loss": 0.1813,
+      "step": 126770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1810456930182506e-05,
+      "loss": 0.1779,
+      "step": 126780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.180821487825658e-05,
+      "loss": 0.1825,
+      "step": 126790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.180597282633066e-05,
+      "loss": 0.1803,
+      "step": 126800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.180373077440474e-05,
+      "loss": 0.1763,
+      "step": 126810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1801488722478814e-05,
+      "loss": 0.1872,
+      "step": 126820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1799246670552893e-05,
+      "loss": 0.1763,
+      "step": 126830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1797004618626968e-05,
+      "loss": 0.1837,
+      "step": 126840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1794762566701047e-05,
+      "loss": 0.181,
+      "step": 126850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1792520514775122e-05,
+      "loss": 0.1774,
+      "step": 126860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.17902784628492e-05,
+      "loss": 0.1784,
+      "step": 126870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.178803641092328e-05,
+      "loss": 0.1826,
+      "step": 126880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1785794358997355e-05,
+      "loss": 0.1806,
+      "step": 126890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1783552307071434e-05,
+      "loss": 0.1781,
+      "step": 126900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.178131025514551e-05,
+      "loss": 0.1772,
+      "step": 126910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1779068203219588e-05,
+      "loss": 0.1801,
+      "step": 126920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1776826151293663e-05,
+      "loss": 0.1788,
+      "step": 126930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1774584099367742e-05,
+      "loss": 0.1781,
+      "step": 126940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.177234204744182e-05,
+      "loss": 0.1818,
+      "step": 126950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1770099995515896e-05,
+      "loss": 0.18,
+      "step": 126960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1767857943589975e-05,
+      "loss": 0.181,
+      "step": 126970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.176561589166405e-05,
+      "loss": 0.1753,
+      "step": 126980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.176337383973813e-05,
+      "loss": 0.18,
+      "step": 126990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1761131787812207e-05,
+      "loss": 0.1797,
+      "step": 127000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1758889735886283e-05,
+      "loss": 0.181,
+      "step": 127010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.175664768396036e-05,
+      "loss": 0.1774,
+      "step": 127020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.175440563203444e-05,
+      "loss": 0.1778,
+      "step": 127030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.175216358010852e-05,
+      "loss": 0.1799,
+      "step": 127040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1749921528182594e-05,
+      "loss": 0.178,
+      "step": 127050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1747679476256673e-05,
+      "loss": 0.1787,
+      "step": 127060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1745437424330748e-05,
+      "loss": 0.1744,
+      "step": 127070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1743195372404827e-05,
+      "loss": 0.1814,
+      "step": 127080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1740953320478906e-05,
+      "loss": 0.173,
+      "step": 127090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.173871126855298e-05,
+      "loss": 0.1758,
+      "step": 127100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.173646921662706e-05,
+      "loss": 0.1783,
+      "step": 127110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1734227164701135e-05,
+      "loss": 0.1803,
+      "step": 127120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1731985112775214e-05,
+      "loss": 0.1769,
+      "step": 127130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.172974306084929e-05,
+      "loss": 0.1799,
+      "step": 127140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1727501008923368e-05,
+      "loss": 0.1774,
+      "step": 127150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1725258956997446e-05,
+      "loss": 0.1828,
+      "step": 127160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1723016905071522e-05,
+      "loss": 0.1854,
+      "step": 127170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.17207748531456e-05,
+      "loss": 0.1759,
+      "step": 127180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1718532801219676e-05,
+      "loss": 0.1815,
+      "step": 127190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1716290749293755e-05,
+      "loss": 0.1752,
+      "step": 127200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.171404869736783e-05,
+      "loss": 0.1797,
+      "step": 127210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.171180664544191e-05,
+      "loss": 0.1857,
+      "step": 127220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1709564593515987e-05,
+      "loss": 0.1806,
+      "step": 127230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1707322541590063e-05,
+      "loss": 0.1802,
+      "step": 127240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.170508048966414e-05,
+      "loss": 0.1773,
+      "step": 127250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1702838437738217e-05,
+      "loss": 0.181,
+      "step": 127260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1700596385812295e-05,
+      "loss": 0.1795,
+      "step": 127270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1698354333886374e-05,
+      "loss": 0.1826,
+      "step": 127280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1696112281960453e-05,
+      "loss": 0.1776,
+      "step": 127290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1693870230034528e-05,
+      "loss": 0.1763,
+      "step": 127300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1691628178108607e-05,
+      "loss": 0.1711,
+      "step": 127310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1689386126182686e-05,
+      "loss": 0.1795,
+      "step": 127320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.168714407425676e-05,
+      "loss": 0.1806,
+      "step": 127330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.168490202233084e-05,
+      "loss": 0.1808,
+      "step": 127340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1682659970404915e-05,
+      "loss": 0.18,
+      "step": 127350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1680417918478994e-05,
+      "loss": 0.1753,
+      "step": 127360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1678175866553072e-05,
+      "loss": 0.1802,
+      "step": 127370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1675933814627148e-05,
+      "loss": 0.1823,
+      "step": 127380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1673691762701226e-05,
+      "loss": 0.1776,
+      "step": 127390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1671449710775302e-05,
+      "loss": 0.1755,
+      "step": 127400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.166920765884938e-05,
+      "loss": 0.176,
+      "step": 127410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1666965606923456e-05,
+      "loss": 0.1823,
+      "step": 127420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1664723554997535e-05,
+      "loss": 0.1767,
+      "step": 127430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1662481503071613e-05,
+      "loss": 0.1814,
+      "step": 127440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.166023945114569e-05,
+      "loss": 0.1764,
+      "step": 127450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1657997399219767e-05,
+      "loss": 0.183,
+      "step": 127460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1655755347293843e-05,
+      "loss": 0.1797,
+      "step": 127470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.165351329536792e-05,
+      "loss": 0.1805,
+      "step": 127480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1651271243441997e-05,
+      "loss": 0.1816,
+      "step": 127490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1649029191516075e-05,
+      "loss": 0.1778,
+      "step": 127500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1646787139590154e-05,
+      "loss": 0.1784,
+      "step": 127510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.164454508766423e-05,
+      "loss": 0.1787,
+      "step": 127520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1642303035738308e-05,
+      "loss": 0.1767,
+      "step": 127530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1640060983812387e-05,
+      "loss": 0.1813,
+      "step": 127540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1637818931886462e-05,
+      "loss": 0.1831,
+      "step": 127550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.163557687996054e-05,
+      "loss": 0.186,
+      "step": 127560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.163333482803462e-05,
+      "loss": 0.1784,
+      "step": 127570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.16310927761087e-05,
+      "loss": 0.1737,
+      "step": 127580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1628850724182774e-05,
+      "loss": 0.1807,
+      "step": 127590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1626608672256852e-05,
+      "loss": 0.1774,
+      "step": 127600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1624366620330928e-05,
+      "loss": 0.1749,
+      "step": 127610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1622124568405006e-05,
+      "loss": 0.1777,
+      "step": 127620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1619882516479082e-05,
+      "loss": 0.1817,
+      "step": 127630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.161764046455316e-05,
+      "loss": 0.1783,
+      "step": 127640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.161539841262724e-05,
+      "loss": 0.1782,
+      "step": 127650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1613156360701315e-05,
+      "loss": 0.1794,
+      "step": 127660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1610914308775393e-05,
+      "loss": 0.179,
+      "step": 127670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.160867225684947e-05,
+      "loss": 0.1828,
+      "step": 127680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1606430204923547e-05,
+      "loss": 0.1813,
+      "step": 127690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1604188152997623e-05,
+      "loss": 0.1836,
+      "step": 127700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.16019461010717e-05,
+      "loss": 0.1783,
+      "step": 127710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.159970404914578e-05,
+      "loss": 0.1773,
+      "step": 127720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1597461997219855e-05,
+      "loss": 0.1794,
+      "step": 127730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1595219945293934e-05,
+      "loss": 0.1837,
+      "step": 127740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.159297789336801e-05,
+      "loss": 0.1809,
+      "step": 127750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1590735841442088e-05,
+      "loss": 0.1728,
+      "step": 127760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1588493789516164e-05,
+      "loss": 0.1761,
+      "step": 127770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1586251737590242e-05,
+      "loss": 0.1777,
+      "step": 127780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.158400968566432e-05,
+      "loss": 0.1766,
+      "step": 127790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1581767633738396e-05,
+      "loss": 0.1839,
+      "step": 127800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1579525581812475e-05,
+      "loss": 0.1785,
+      "step": 127810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1577283529886554e-05,
+      "loss": 0.1796,
+      "step": 127820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1575041477960632e-05,
+      "loss": 0.1796,
+      "step": 127830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1572799426034708e-05,
+      "loss": 0.1739,
+      "step": 127840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1570557374108787e-05,
+      "loss": 0.1786,
+      "step": 127850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1568315322182865e-05,
+      "loss": 0.1784,
+      "step": 127860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.156607327025694e-05,
+      "loss": 0.179,
+      "step": 127870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.156383121833102e-05,
+      "loss": 0.1833,
+      "step": 127880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1561589166405095e-05,
+      "loss": 0.1795,
+      "step": 127890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1559347114479173e-05,
+      "loss": 0.1841,
+      "step": 127900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1557105062553252e-05,
+      "loss": 0.1832,
+      "step": 127910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1554863010627327e-05,
+      "loss": 0.18,
+      "step": 127920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1552620958701406e-05,
+      "loss": 0.1792,
+      "step": 127930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.155037890677548e-05,
+      "loss": 0.1802,
+      "step": 127940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.154813685484956e-05,
+      "loss": 0.1784,
+      "step": 127950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1545894802923635e-05,
+      "loss": 0.1714,
+      "step": 127960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1543652750997714e-05,
+      "loss": 0.1775,
+      "step": 127970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1541410699071793e-05,
+      "loss": 0.1759,
+      "step": 127980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1539168647145868e-05,
+      "loss": 0.182,
+      "step": 127990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1536926595219947e-05,
+      "loss": 0.1842,
+      "step": 128000
+    },
+    {
+      "epoch": 0.57,
+      "eval_NEIMS_canon_smiles": 0.9442,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.060533333333333335,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.46074801317645475,
+      "eval_NEIMS_exact_mols": 0.060316666666666664,
+      "eval_NEIMS_exact_smiles": 0.05846666666666667,
+      "eval_NEIMS_loss": 0.21823979914188385,
+      "eval_NEIMS_matched_formulas": 0.5509166666666667,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.3658473136765312,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.06065,
+      "eval_NEIMS_runtime": 716.0986,
+      "eval_NEIMS_samples_per_second": 83.787,
+      "eval_NEIMS_steps_per_second": 1.31,
+      "step": 128000
+    },
+    {
+      "epoch": 0.57,
+      "eval_RASSP_canon_smiles": 0.9369385915116706,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.14238487995458643,
+      "eval_RASSP_daylight_tanimoto_simil": 0.5888624426564175,
+      "eval_RASSP_exact_mols": 0.14191738738437906,
+      "eval_RASSP_exact_smiles": 0.13750960029385248,
+      "eval_RASSP_loss": 0.15661999583244324,
+      "eval_RASSP_matched_formulas": 0.7334791464921361,
+      "eval_RASSP_morgan_tanimoto_simil": 0.46166409815266424,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.1420509566901526,
+      "eval_RASSP_runtime": 817.8885,
+      "eval_RASSP_samples_per_second": 73.23,
+      "eval_RASSP_steps_per_second": 1.144,
+      "step": 128000
+    },
+    {
+      "epoch": 0.57,
+      "eval_NIST_canon_smiles": 0.8755722752599638,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.011960109309010895,
+      "eval_NIST_daylight_tanimoto_simil": 0.2668137700125038,
+      "eval_NIST_exact_mols": 0.01153423004578202,
+      "eval_NIST_exact_smiles": 0.010859921212336303,
+      "eval_NIST_loss": 1.8035073280334473,
+      "eval_NIST_matched_formulas": 0.08265606700500408,
+      "eval_NIST_morgan_tanimoto_simil": 0.21989801486693322,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.01192461937040849,
+      "eval_NIST_runtime": 535.9421,
+      "eval_NIST_samples_per_second": 52.575,
+      "eval_NIST_steps_per_second": 0.823,
+      "step": 128000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1534684543294022e-05,
+      "loss": 0.1771,
+      "step": 128010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.15324424913681e-05,
+      "loss": 0.1828,
+      "step": 128020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1530200439442176e-05,
+      "loss": 0.1791,
+      "step": 128030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1527958387516255e-05,
+      "loss": 0.1745,
+      "step": 128040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1525716335590334e-05,
+      "loss": 0.1826,
+      "step": 128050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.152347428366441e-05,
+      "loss": 0.1768,
+      "step": 128060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1521232231738488e-05,
+      "loss": 0.1771,
+      "step": 128070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1518990179812563e-05,
+      "loss": 0.1816,
+      "step": 128080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1516748127886642e-05,
+      "loss": 0.1811,
+      "step": 128090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.151450607596072e-05,
+      "loss": 0.1751,
+      "step": 128100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.15122640240348e-05,
+      "loss": 0.1735,
+      "step": 128110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1510021972108878e-05,
+      "loss": 0.1804,
+      "step": 128120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1507779920182953e-05,
+      "loss": 0.1796,
+      "step": 128130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1505537868257032e-05,
+      "loss": 0.1776,
+      "step": 128140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1503295816331107e-05,
+      "loss": 0.1762,
+      "step": 128150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1501053764405186e-05,
+      "loss": 0.174,
+      "step": 128160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.149881171247926e-05,
+      "loss": 0.1793,
+      "step": 128170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.149656966055334e-05,
+      "loss": 0.1771,
+      "step": 128180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.149432760862742e-05,
+      "loss": 0.1841,
+      "step": 128190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1492085556701494e-05,
+      "loss": 0.1736,
+      "step": 128200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1489843504775573e-05,
+      "loss": 0.1777,
+      "step": 128210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1487601452849648e-05,
+      "loss": 0.1805,
+      "step": 128220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1485359400923727e-05,
+      "loss": 0.1798,
+      "step": 128230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1483117348997802e-05,
+      "loss": 0.1835,
+      "step": 128240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.148087529707188e-05,
+      "loss": 0.1765,
+      "step": 128250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.147863324514596e-05,
+      "loss": 0.1823,
+      "step": 128260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1476391193220035e-05,
+      "loss": 0.1779,
+      "step": 128270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1474149141294114e-05,
+      "loss": 0.1847,
+      "step": 128280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.147190708936819e-05,
+      "loss": 0.1777,
+      "step": 128290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1469665037442268e-05,
+      "loss": 0.1733,
+      "step": 128300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1467422985516343e-05,
+      "loss": 0.1779,
+      "step": 128310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1465180933590422e-05,
+      "loss": 0.184,
+      "step": 128320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.14629388816645e-05,
+      "loss": 0.177,
+      "step": 128330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1460696829738576e-05,
+      "loss": 0.1734,
+      "step": 128340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1458454777812655e-05,
+      "loss": 0.1781,
+      "step": 128350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1456212725886733e-05,
+      "loss": 0.1777,
+      "step": 128360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1453970673960812e-05,
+      "loss": 0.1785,
+      "step": 128370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1451728622034887e-05,
+      "loss": 0.178,
+      "step": 128380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1449486570108966e-05,
+      "loss": 0.182,
+      "step": 128390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1447244518183045e-05,
+      "loss": 0.1797,
+      "step": 128400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.144500246625712e-05,
+      "loss": 0.1803,
+      "step": 128410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.14427604143312e-05,
+      "loss": 0.1826,
+      "step": 128420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1440518362405274e-05,
+      "loss": 0.1813,
+      "step": 128430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1438276310479353e-05,
+      "loss": 0.1767,
+      "step": 128440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1436034258553428e-05,
+      "loss": 0.1749,
+      "step": 128450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1433792206627507e-05,
+      "loss": 0.1795,
+      "step": 128460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1431550154701586e-05,
+      "loss": 0.1743,
+      "step": 128470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.142930810277566e-05,
+      "loss": 0.1802,
+      "step": 128480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.142706605084974e-05,
+      "loss": 0.1753,
+      "step": 128490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1424823998923815e-05,
+      "loss": 0.1804,
+      "step": 128500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1422581946997894e-05,
+      "loss": 0.1829,
+      "step": 128510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.142033989507197e-05,
+      "loss": 0.1788,
+      "step": 128520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1418097843146048e-05,
+      "loss": 0.1764,
+      "step": 128530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1415855791220127e-05,
+      "loss": 0.1768,
+      "step": 128540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1413613739294202e-05,
+      "loss": 0.1746,
+      "step": 128550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.141137168736828e-05,
+      "loss": 0.1783,
+      "step": 128560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1409129635442356e-05,
+      "loss": 0.1769,
+      "step": 128570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1406887583516435e-05,
+      "loss": 0.1776,
+      "step": 128580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.140464553159051e-05,
+      "loss": 0.1677,
+      "step": 128590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.140240347966459e-05,
+      "loss": 0.1761,
+      "step": 128600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1400161427738667e-05,
+      "loss": 0.1793,
+      "step": 128610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1397919375812743e-05,
+      "loss": 0.1739,
+      "step": 128620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.139567732388682e-05,
+      "loss": 0.1792,
+      "step": 128630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.13934352719609e-05,
+      "loss": 0.1778,
+      "step": 128640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.139119322003498e-05,
+      "loss": 0.1813,
+      "step": 128650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1388951168109054e-05,
+      "loss": 0.1762,
+      "step": 128660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1386709116183133e-05,
+      "loss": 0.1746,
+      "step": 128670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.138446706425721e-05,
+      "loss": 0.1763,
+      "step": 128680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1382225012331287e-05,
+      "loss": 0.1811,
+      "step": 128690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1379982960405366e-05,
+      "loss": 0.184,
+      "step": 128700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.137774090847944e-05,
+      "loss": 0.181,
+      "step": 128710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.137549885655352e-05,
+      "loss": 0.1805,
+      "step": 128720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1373256804627595e-05,
+      "loss": 0.1736,
+      "step": 128730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1371014752701674e-05,
+      "loss": 0.1834,
+      "step": 128740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1368772700775753e-05,
+      "loss": 0.1818,
+      "step": 128750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1366530648849828e-05,
+      "loss": 0.1763,
+      "step": 128760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1364288596923907e-05,
+      "loss": 0.1794,
+      "step": 128770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1362046544997982e-05,
+      "loss": 0.1828,
+      "step": 128780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.135980449307206e-05,
+      "loss": 0.18,
+      "step": 128790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1357562441146136e-05,
+      "loss": 0.18,
+      "step": 128800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1355320389220215e-05,
+      "loss": 0.1783,
+      "step": 128810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1353078337294293e-05,
+      "loss": 0.176,
+      "step": 128820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.135083628536837e-05,
+      "loss": 0.1781,
+      "step": 128830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1348594233442447e-05,
+      "loss": 0.1814,
+      "step": 128840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1346352181516523e-05,
+      "loss": 0.1792,
+      "step": 128850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.13441101295906e-05,
+      "loss": 0.1742,
+      "step": 128860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1341868077664677e-05,
+      "loss": 0.1814,
+      "step": 128870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1339626025738756e-05,
+      "loss": 0.1825,
+      "step": 128880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1337383973812834e-05,
+      "loss": 0.1827,
+      "step": 128890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1335141921886913e-05,
+      "loss": 0.1782,
+      "step": 128900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.133289986996099e-05,
+      "loss": 0.1788,
+      "step": 128910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1330657818035067e-05,
+      "loss": 0.1758,
+      "step": 128920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1328415766109146e-05,
+      "loss": 0.1759,
+      "step": 128930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.132617371418322e-05,
+      "loss": 0.18,
+      "step": 128940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.13239316622573e-05,
+      "loss": 0.178,
+      "step": 128950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.132168961033138e-05,
+      "loss": 0.1763,
+      "step": 128960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1319447558405454e-05,
+      "loss": 0.1749,
+      "step": 128970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1317205506479533e-05,
+      "loss": 0.1753,
+      "step": 128980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1314963454553608e-05,
+      "loss": 0.1767,
+      "step": 128990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1312721402627687e-05,
+      "loss": 0.1781,
+      "step": 129000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1310479350701762e-05,
+      "loss": 0.1754,
+      "step": 129010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.130823729877584e-05,
+      "loss": 0.1786,
+      "step": 129020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.130599524684992e-05,
+      "loss": 0.1767,
+      "step": 129030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1303753194923995e-05,
+      "loss": 0.1789,
+      "step": 129040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1301511142998073e-05,
+      "loss": 0.1818,
+      "step": 129050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.129926909107215e-05,
+      "loss": 0.1775,
+      "step": 129060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1297027039146227e-05,
+      "loss": 0.1783,
+      "step": 129070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1294784987220303e-05,
+      "loss": 0.1797,
+      "step": 129080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.129254293529438e-05,
+      "loss": 0.1808,
+      "step": 129090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.129030088336846e-05,
+      "loss": 0.1789,
+      "step": 129100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1288058831442536e-05,
+      "loss": 0.1822,
+      "step": 129110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1285816779516614e-05,
+      "loss": 0.1764,
+      "step": 129120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.128357472759069e-05,
+      "loss": 0.1767,
+      "step": 129130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.128155688085736e-05,
+      "loss": 0.1802,
+      "step": 129140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.127931482893144e-05,
+      "loss": 0.179,
+      "step": 129150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1277072777005515e-05,
+      "loss": 0.1738,
+      "step": 129160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1274830725079594e-05,
+      "loss": 0.1818,
+      "step": 129170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.127258867315367e-05,
+      "loss": 0.1787,
+      "step": 129180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1270346621227748e-05,
+      "loss": 0.1781,
+      "step": 129190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1268104569301826e-05,
+      "loss": 0.1801,
+      "step": 129200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1265862517375905e-05,
+      "loss": 0.1769,
+      "step": 129210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1263620465449984e-05,
+      "loss": 0.1797,
+      "step": 129220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.126137841352406e-05,
+      "loss": 0.1772,
+      "step": 129230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1259136361598138e-05,
+      "loss": 0.1783,
+      "step": 129240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1256894309672213e-05,
+      "loss": 0.1787,
+      "step": 129250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1254652257746292e-05,
+      "loss": 0.1788,
+      "step": 129260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1252410205820367e-05,
+      "loss": 0.1701,
+      "step": 129270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1250168153894446e-05,
+      "loss": 0.1786,
+      "step": 129280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1247926101968525e-05,
+      "loss": 0.1746,
+      "step": 129290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.12456840500426e-05,
+      "loss": 0.1841,
+      "step": 129300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.124344199811668e-05,
+      "loss": 0.1824,
+      "step": 129310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1241199946190754e-05,
+      "loss": 0.1783,
+      "step": 129320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1238957894264833e-05,
+      "loss": 0.1824,
+      "step": 129330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1236715842338908e-05,
+      "loss": 0.1736,
+      "step": 129340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1234473790412987e-05,
+      "loss": 0.177,
+      "step": 129350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1232231738487065e-05,
+      "loss": 0.1794,
+      "step": 129360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.122998968656114e-05,
+      "loss": 0.1808,
+      "step": 129370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.122774763463522e-05,
+      "loss": 0.1803,
+      "step": 129380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1225505582709295e-05,
+      "loss": 0.1795,
+      "step": 129390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1223263530783374e-05,
+      "loss": 0.1771,
+      "step": 129400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.122102147885745e-05,
+      "loss": 0.1828,
+      "step": 129410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1218779426931528e-05,
+      "loss": 0.1813,
+      "step": 129420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1216537375005606e-05,
+      "loss": 0.1776,
+      "step": 129430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.121429532307968e-05,
+      "loss": 0.1693,
+      "step": 129440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.121205327115376e-05,
+      "loss": 0.1778,
+      "step": 129450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.120981121922784e-05,
+      "loss": 0.1803,
+      "step": 129460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1207569167301918e-05,
+      "loss": 0.1772,
+      "step": 129470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1205327115375993e-05,
+      "loss": 0.1794,
+      "step": 129480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1203085063450072e-05,
+      "loss": 0.1799,
+      "step": 129490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.120084301152415e-05,
+      "loss": 0.184,
+      "step": 129500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1198600959598226e-05,
+      "loss": 0.178,
+      "step": 129510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1196358907672305e-05,
+      "loss": 0.1798,
+      "step": 129520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.119411685574638e-05,
+      "loss": 0.1774,
+      "step": 129530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.119187480382046e-05,
+      "loss": 0.1787,
+      "step": 129540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1189632751894534e-05,
+      "loss": 0.1718,
+      "step": 129550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1187390699968613e-05,
+      "loss": 0.1797,
+      "step": 129560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.118514864804269e-05,
+      "loss": 0.1785,
+      "step": 129570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1182906596116767e-05,
+      "loss": 0.1842,
+      "step": 129580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1180664544190845e-05,
+      "loss": 0.1743,
+      "step": 129590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.117842249226492e-05,
+      "loss": 0.1737,
+      "step": 129600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1176180440339e-05,
+      "loss": 0.1801,
+      "step": 129610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1173938388413075e-05,
+      "loss": 0.1774,
+      "step": 129620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1171696336487154e-05,
+      "loss": 0.1784,
+      "step": 129630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1169454284561232e-05,
+      "loss": 0.1744,
+      "step": 129640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1167212232635308e-05,
+      "loss": 0.1813,
+      "step": 129650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1164970180709386e-05,
+      "loss": 0.173,
+      "step": 129660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.116272812878346e-05,
+      "loss": 0.1781,
+      "step": 129670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.116048607685754e-05,
+      "loss": 0.18,
+      "step": 129680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1158244024931616e-05,
+      "loss": 0.179,
+      "step": 129690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1156001973005694e-05,
+      "loss": 0.1732,
+      "step": 129700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1153759921079773e-05,
+      "loss": 0.1806,
+      "step": 129710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.115151786915385e-05,
+      "loss": 0.1783,
+      "step": 129720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1149275817227927e-05,
+      "loss": 0.1756,
+      "step": 129730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1147033765302006e-05,
+      "loss": 0.178,
+      "step": 129740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1144791713376085e-05,
+      "loss": 0.1767,
+      "step": 129750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.114254966145016e-05,
+      "loss": 0.1747,
+      "step": 129760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.114030760952424e-05,
+      "loss": 0.176,
+      "step": 129770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1138065557598317e-05,
+      "loss": 0.1755,
+      "step": 129780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1135823505672393e-05,
+      "loss": 0.18,
+      "step": 129790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.113358145374647e-05,
+      "loss": 0.1793,
+      "step": 129800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1131339401820547e-05,
+      "loss": 0.1767,
+      "step": 129810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1129097349894625e-05,
+      "loss": 0.1744,
+      "step": 129820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.11268552979687e-05,
+      "loss": 0.1834,
+      "step": 129830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.112461324604278e-05,
+      "loss": 0.1795,
+      "step": 129840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1122371194116858e-05,
+      "loss": 0.1734,
+      "step": 129850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1120129142190934e-05,
+      "loss": 0.1731,
+      "step": 129860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1117887090265012e-05,
+      "loss": 0.1764,
+      "step": 129870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1115645038339088e-05,
+      "loss": 0.1755,
+      "step": 129880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1113402986413166e-05,
+      "loss": 0.1766,
+      "step": 129890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.111116093448724e-05,
+      "loss": 0.1815,
+      "step": 129900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.110891888256132e-05,
+      "loss": 0.1751,
+      "step": 129910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.11066768306354e-05,
+      "loss": 0.178,
+      "step": 129920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1104434778709474e-05,
+      "loss": 0.1804,
+      "step": 129930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1102192726783553e-05,
+      "loss": 0.1776,
+      "step": 129940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.109995067485763e-05,
+      "loss": 0.1809,
+      "step": 129950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1097708622931707e-05,
+      "loss": 0.173,
+      "step": 129960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1095466571005783e-05,
+      "loss": 0.1802,
+      "step": 129970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.109322451907986e-05,
+      "loss": 0.1803,
+      "step": 129980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.109098246715394e-05,
+      "loss": 0.1777,
+      "step": 129990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.108874041522802e-05,
+      "loss": 0.1741,
+      "step": 130000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1086498363302097e-05,
+      "loss": 0.1742,
+      "step": 130010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1084256311376173e-05,
+      "loss": 0.1749,
+      "step": 130020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.108201425945025e-05,
+      "loss": 0.1786,
+      "step": 130030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1079772207524327e-05,
+      "loss": 0.1796,
+      "step": 130040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1077530155598405e-05,
+      "loss": 0.1721,
+      "step": 130050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1075288103672484e-05,
+      "loss": 0.1705,
+      "step": 130060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.107304605174656e-05,
+      "loss": 0.1758,
+      "step": 130070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1070803999820638e-05,
+      "loss": 0.1783,
+      "step": 130080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1068561947894714e-05,
+      "loss": 0.1769,
+      "step": 130090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1066319895968792e-05,
+      "loss": 0.1785,
+      "step": 130100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1064077844042868e-05,
+      "loss": 0.1767,
+      "step": 130110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1061835792116946e-05,
+      "loss": 0.1729,
+      "step": 130120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1059593740191025e-05,
+      "loss": 0.1767,
+      "step": 130130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.10573516882651e-05,
+      "loss": 0.1787,
+      "step": 130140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.105510963633918e-05,
+      "loss": 0.1804,
+      "step": 130150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1052867584413254e-05,
+      "loss": 0.1736,
+      "step": 130160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1050625532487333e-05,
+      "loss": 0.1733,
+      "step": 130170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1048383480561412e-05,
+      "loss": 0.1741,
+      "step": 130180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1046141428635487e-05,
+      "loss": 0.1764,
+      "step": 130190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1043899376709566e-05,
+      "loss": 0.1751,
+      "step": 130200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.104165732478364e-05,
+      "loss": 0.1788,
+      "step": 130210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.103941527285772e-05,
+      "loss": 0.176,
+      "step": 130220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1037173220931795e-05,
+      "loss": 0.1748,
+      "step": 130230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1034931169005874e-05,
+      "loss": 0.175,
+      "step": 130240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1032689117079953e-05,
+      "loss": 0.1741,
+      "step": 130250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1030447065154028e-05,
+      "loss": 0.1763,
+      "step": 130260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1028205013228107e-05,
+      "loss": 0.1773,
+      "step": 130270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1025962961302186e-05,
+      "loss": 0.1702,
+      "step": 130280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1023720909376264e-05,
+      "loss": 0.1777,
+      "step": 130290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.102147885745034e-05,
+      "loss": 0.178,
+      "step": 130300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1019236805524418e-05,
+      "loss": 0.1696,
+      "step": 130310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1016994753598497e-05,
+      "loss": 0.1708,
+      "step": 130320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1014752701672572e-05,
+      "loss": 0.1768,
+      "step": 130330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.101251064974665e-05,
+      "loss": 0.1775,
+      "step": 130340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1010268597820726e-05,
+      "loss": 0.1744,
+      "step": 130350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1008026545894805e-05,
+      "loss": 0.1774,
+      "step": 130360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.100578449396888e-05,
+      "loss": 0.178,
+      "step": 130370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.100354244204296e-05,
+      "loss": 0.1734,
+      "step": 130380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1001300390117038e-05,
+      "loss": 0.1811,
+      "step": 130390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0999058338191113e-05,
+      "loss": 0.1696,
+      "step": 130400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0996816286265192e-05,
+      "loss": 0.172,
+      "step": 130410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0994574234339267e-05,
+      "loss": 0.1719,
+      "step": 130420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0992332182413346e-05,
+      "loss": 0.174,
+      "step": 130430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.099009013048742e-05,
+      "loss": 0.1713,
+      "step": 130440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.09878480785615e-05,
+      "loss": 0.1752,
+      "step": 130450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.098560602663558e-05,
+      "loss": 0.1709,
+      "step": 130460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0983363974709654e-05,
+      "loss": 0.1717,
+      "step": 130470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0981121922783733e-05,
+      "loss": 0.1718,
+      "step": 130480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0978879870857808e-05,
+      "loss": 0.1726,
+      "step": 130490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0976637818931887e-05,
+      "loss": 0.1766,
+      "step": 130500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0974395767005962e-05,
+      "loss": 0.1736,
+      "step": 130510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.097215371508004e-05,
+      "loss": 0.1807,
+      "step": 130520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.096991166315412e-05,
+      "loss": 0.1777,
+      "step": 130530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0967669611228198e-05,
+      "loss": 0.1737,
+      "step": 130540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0965427559302274e-05,
+      "loss": 0.1782,
+      "step": 130550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0963185507376352e-05,
+      "loss": 0.1767,
+      "step": 130560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.096094345545043e-05,
+      "loss": 0.1772,
+      "step": 130570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0958701403524506e-05,
+      "loss": 0.1718,
+      "step": 130580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0956459351598585e-05,
+      "loss": 0.1727,
+      "step": 130590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0954217299672664e-05,
+      "loss": 0.1707,
+      "step": 130600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.095197524774674e-05,
+      "loss": 0.1701,
+      "step": 130610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0949733195820818e-05,
+      "loss": 0.168,
+      "step": 130620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0947491143894893e-05,
+      "loss": 0.175,
+      "step": 130630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0945249091968972e-05,
+      "loss": 0.1753,
+      "step": 130640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0943007040043047e-05,
+      "loss": 0.1824,
+      "step": 130650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0940764988117126e-05,
+      "loss": 0.1736,
+      "step": 130660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0938522936191205e-05,
+      "loss": 0.1749,
+      "step": 130670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.093628088426528e-05,
+      "loss": 0.1724,
+      "step": 130680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.093403883233936e-05,
+      "loss": 0.1738,
+      "step": 130690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0931796780413434e-05,
+      "loss": 0.1766,
+      "step": 130700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0929554728487513e-05,
+      "loss": 0.1718,
+      "step": 130710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0927312676561588e-05,
+      "loss": 0.1745,
+      "step": 130720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0925070624635667e-05,
+      "loss": 0.1795,
+      "step": 130730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0922828572709746e-05,
+      "loss": 0.1689,
+      "step": 130740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.092058652078382e-05,
+      "loss": 0.1707,
+      "step": 130750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.09183444688579e-05,
+      "loss": 0.1713,
+      "step": 130760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0916102416931975e-05,
+      "loss": 0.169,
+      "step": 130770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0913860365006054e-05,
+      "loss": 0.1733,
+      "step": 130780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0911618313080132e-05,
+      "loss": 0.1702,
+      "step": 130790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0909376261154208e-05,
+      "loss": 0.175,
+      "step": 130800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0907134209228286e-05,
+      "loss": 0.1754,
+      "step": 130810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0904892157302365e-05,
+      "loss": 0.1782,
+      "step": 130820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0902650105376444e-05,
+      "loss": 0.1781,
+      "step": 130830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.090040805345052e-05,
+      "loss": 0.1724,
+      "step": 130840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0898166001524598e-05,
+      "loss": 0.1682,
+      "step": 130850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0895923949598673e-05,
+      "loss": 0.1746,
+      "step": 130860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0893681897672752e-05,
+      "loss": 0.1722,
+      "step": 130870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.089143984574683e-05,
+      "loss": 0.1734,
+      "step": 130880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0889197793820906e-05,
+      "loss": 0.1686,
+      "step": 130890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0886955741894985e-05,
+      "loss": 0.1657,
+      "step": 130900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.088471368996906e-05,
+      "loss": 0.1695,
+      "step": 130910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.088247163804314e-05,
+      "loss": 0.1723,
+      "step": 130920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0880229586117214e-05,
+      "loss": 0.1701,
+      "step": 130930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0877987534191293e-05,
+      "loss": 0.1739,
+      "step": 130940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.087574548226537e-05,
+      "loss": 0.1724,
+      "step": 130950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0873503430339447e-05,
+      "loss": 0.1742,
+      "step": 130960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0871261378413526e-05,
+      "loss": 0.1705,
+      "step": 130970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.08690193264876e-05,
+      "loss": 0.1733,
+      "step": 130980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.086677727456168e-05,
+      "loss": 0.1671,
+      "step": 130990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0864535222635755e-05,
+      "loss": 0.1704,
+      "step": 131000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0862293170709834e-05,
+      "loss": 0.1702,
+      "step": 131010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0860051118783912e-05,
+      "loss": 0.1737,
+      "step": 131020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0857809066857988e-05,
+      "loss": 0.1702,
+      "step": 131030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0855567014932066e-05,
+      "loss": 0.1814,
+      "step": 131040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0853324963006142e-05,
+      "loss": 0.1681,
+      "step": 131050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.085108291108022e-05,
+      "loss": 0.1683,
+      "step": 131060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.08488408591543e-05,
+      "loss": 0.1717,
+      "step": 131070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0846598807228378e-05,
+      "loss": 0.166,
+      "step": 131080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0844356755302453e-05,
+      "loss": 0.1671,
+      "step": 131090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0842114703376532e-05,
+      "loss": 0.1717,
+      "step": 131100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.083987265145061e-05,
+      "loss": 0.1696,
+      "step": 131110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0837630599524686e-05,
+      "loss": 0.1664,
+      "step": 131120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0835388547598765e-05,
+      "loss": 0.1662,
+      "step": 131130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.083314649567284e-05,
+      "loss": 0.1681,
+      "step": 131140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.083090444374692e-05,
+      "loss": 0.171,
+      "step": 131150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0828662391820997e-05,
+      "loss": 0.1714,
+      "step": 131160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0826420339895073e-05,
+      "loss": 0.167,
+      "step": 131170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.082417828796915e-05,
+      "loss": 0.1662,
+      "step": 131180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0821936236043227e-05,
+      "loss": 0.1646,
+      "step": 131190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0819694184117306e-05,
+      "loss": 0.1672,
+      "step": 131200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.081745213219138e-05,
+      "loss": 0.1703,
+      "step": 131210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.081521008026546e-05,
+      "loss": 0.1619,
+      "step": 131220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.081296802833954e-05,
+      "loss": 0.1762,
+      "step": 131230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0810725976413614e-05,
+      "loss": 0.1633,
+      "step": 131240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0808483924487692e-05,
+      "loss": 0.1642,
+      "step": 131250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0806241872561768e-05,
+      "loss": 0.1743,
+      "step": 131260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0803999820635846e-05,
+      "loss": 0.1682,
+      "step": 131270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0801757768709922e-05,
+      "loss": 0.174,
+      "step": 131280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0799515716784e-05,
+      "loss": 0.1706,
+      "step": 131290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.079727366485808e-05,
+      "loss": 0.1692,
+      "step": 131300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0795031612932155e-05,
+      "loss": 0.1685,
+      "step": 131310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0792789561006233e-05,
+      "loss": 0.1716,
+      "step": 131320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0790547509080312e-05,
+      "loss": 0.1682,
+      "step": 131330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0788305457154387e-05,
+      "loss": 0.164,
+      "step": 131340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0786063405228466e-05,
+      "loss": 0.1694,
+      "step": 131350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0783821353302545e-05,
+      "loss": 0.169,
+      "step": 131360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0781579301376623e-05,
+      "loss": 0.1648,
+      "step": 131370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.07793372494507e-05,
+      "loss": 0.168,
+      "step": 131380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0777095197524777e-05,
+      "loss": 0.1698,
+      "step": 131390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0774853145598853e-05,
+      "loss": 0.1681,
+      "step": 131400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.077261109367293e-05,
+      "loss": 0.1712,
+      "step": 131410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0770369041747007e-05,
+      "loss": 0.1661,
+      "step": 131420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0768126989821086e-05,
+      "loss": 0.1715,
+      "step": 131430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0765884937895164e-05,
+      "loss": 0.1661,
+      "step": 131440
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.076364288596924e-05,
+      "loss": 0.1681,
+      "step": 131450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.076140083404332e-05,
+      "loss": 0.1691,
+      "step": 131460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0759158782117394e-05,
+      "loss": 0.1624,
+      "step": 131470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0756916730191472e-05,
+      "loss": 0.1666,
+      "step": 131480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.075467467826555e-05,
+      "loss": 0.1671,
+      "step": 131490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0752432626339626e-05,
+      "loss": 0.1683,
+      "step": 131500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0750190574413705e-05,
+      "loss": 0.1633,
+      "step": 131510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.074794852248778e-05,
+      "loss": 0.1666,
+      "step": 131520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.074570647056186e-05,
+      "loss": 0.1616,
+      "step": 131530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0743464418635935e-05,
+      "loss": 0.1723,
+      "step": 131540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0741222366710013e-05,
+      "loss": 0.176,
+      "step": 131550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0738980314784092e-05,
+      "loss": 0.1779,
+      "step": 131560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0736738262858167e-05,
+      "loss": 0.1761,
+      "step": 131570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0734496210932246e-05,
+      "loss": 0.1787,
+      "step": 131580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.073225415900632e-05,
+      "loss": 0.1766,
+      "step": 131590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.07300121070804e-05,
+      "loss": 0.1733,
+      "step": 131600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.072777005515448e-05,
+      "loss": 0.1785,
+      "step": 131610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0725528003228558e-05,
+      "loss": 0.1775,
+      "step": 131620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0723285951302633e-05,
+      "loss": 0.1815,
+      "step": 131630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.072104389937671e-05,
+      "loss": 0.176,
+      "step": 131640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.071880184745079e-05,
+      "loss": 0.1792,
+      "step": 131650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0716559795524866e-05,
+      "loss": 0.173,
+      "step": 131660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0714317743598944e-05,
+      "loss": 0.1763,
+      "step": 131670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.071207569167302e-05,
+      "loss": 0.1761,
+      "step": 131680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.07098336397471e-05,
+      "loss": 0.1803,
+      "step": 131690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0707591587821177e-05,
+      "loss": 0.1913,
+      "step": 131700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0705349535895252e-05,
+      "loss": 0.1834,
+      "step": 131710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.070310748396933e-05,
+      "loss": 0.1907,
+      "step": 131720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0700865432043406e-05,
+      "loss": 0.1889,
+      "step": 131730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0698623380117485e-05,
+      "loss": 0.1868,
+      "step": 131740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.069638132819156e-05,
+      "loss": 0.1913,
+      "step": 131750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.069413927626564e-05,
+      "loss": 0.1865,
+      "step": 131760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0691897224339718e-05,
+      "loss": 0.1835,
+      "step": 131770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0689655172413793e-05,
+      "loss": 0.1855,
+      "step": 131780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0687413120487872e-05,
+      "loss": 0.1823,
+      "step": 131790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0685171068561947e-05,
+      "loss": 0.1854,
+      "step": 131800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0682929016636026e-05,
+      "loss": 0.1864,
+      "step": 131810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.06806869647101e-05,
+      "loss": 0.1851,
+      "step": 131820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.067844491278418e-05,
+      "loss": 0.1817,
+      "step": 131830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.067620286085826e-05,
+      "loss": 0.1873,
+      "step": 131840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0673960808932334e-05,
+      "loss": 0.1879,
+      "step": 131850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0671718757006413e-05,
+      "loss": 0.1897,
+      "step": 131860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.066970091027308e-05,
+      "loss": 0.1871,
+      "step": 131870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.066745885834716e-05,
+      "loss": 0.1843,
+      "step": 131880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0665216806421238e-05,
+      "loss": 0.1829,
+      "step": 131890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0662974754495313e-05,
+      "loss": 0.1873,
+      "step": 131900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0660732702569392e-05,
+      "loss": 0.1866,
+      "step": 131910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.065849065064347e-05,
+      "loss": 0.184,
+      "step": 131920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.065624859871755e-05,
+      "loss": 0.189,
+      "step": 131930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0654006546791625e-05,
+      "loss": 0.1897,
+      "step": 131940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0651764494865704e-05,
+      "loss": 0.1815,
+      "step": 131950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.064952244293978e-05,
+      "loss": 0.1843,
+      "step": 131960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0647280391013858e-05,
+      "loss": 0.182,
+      "step": 131970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0645038339087936e-05,
+      "loss": 0.182,
+      "step": 131980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0642796287162012e-05,
+      "loss": 0.1843,
+      "step": 131990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.064055423523609e-05,
+      "loss": 0.1896,
+      "step": 132000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0638312183310166e-05,
+      "loss": 0.1804,
+      "step": 132010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0636070131384244e-05,
+      "loss": 0.1839,
+      "step": 132020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.063382807945832e-05,
+      "loss": 0.1837,
+      "step": 132030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.06315860275324e-05,
+      "loss": 0.1834,
+      "step": 132040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0629343975606477e-05,
+      "loss": 0.1756,
+      "step": 132050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0627101923680553e-05,
+      "loss": 0.1803,
+      "step": 132060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.062485987175463e-05,
+      "loss": 0.1818,
+      "step": 132070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0622617819828707e-05,
+      "loss": 0.1764,
+      "step": 132080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0620375767902785e-05,
+      "loss": 0.1838,
+      "step": 132090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.061813371597686e-05,
+      "loss": 0.1777,
+      "step": 132100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.061589166405094e-05,
+      "loss": 0.1836,
+      "step": 132110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0613649612125018e-05,
+      "loss": 0.181,
+      "step": 132120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0611407560199093e-05,
+      "loss": 0.1803,
+      "step": 132130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0609165508273172e-05,
+      "loss": 0.181,
+      "step": 132140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0606923456347247e-05,
+      "loss": 0.1811,
+      "step": 132150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0604681404421326e-05,
+      "loss": 0.1796,
+      "step": 132160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0602439352495405e-05,
+      "loss": 0.1831,
+      "step": 132170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0600197300569484e-05,
+      "loss": 0.1854,
+      "step": 132180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.059795524864356e-05,
+      "loss": 0.1798,
+      "step": 132190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0595713196717638e-05,
+      "loss": 0.18,
+      "step": 132200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0593471144791716e-05,
+      "loss": 0.1817,
+      "step": 132210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0591229092865792e-05,
+      "loss": 0.176,
+      "step": 132220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.058898704093987e-05,
+      "loss": 0.1743,
+      "step": 132230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0586744989013946e-05,
+      "loss": 0.1752,
+      "step": 132240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0584502937088024e-05,
+      "loss": 0.1847,
+      "step": 132250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0582260885162103e-05,
+      "loss": 0.1817,
+      "step": 132260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.058001883323618e-05,
+      "loss": 0.1824,
+      "step": 132270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0577776781310257e-05,
+      "loss": 0.179,
+      "step": 132280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0575534729384333e-05,
+      "loss": 0.1719,
+      "step": 132290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.057329267745841e-05,
+      "loss": 0.1794,
+      "step": 132300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0571050625532487e-05,
+      "loss": 0.1867,
+      "step": 132310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0568808573606565e-05,
+      "loss": 0.1805,
+      "step": 132320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0566566521680644e-05,
+      "loss": 0.1817,
+      "step": 132330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.056432446975472e-05,
+      "loss": 0.1808,
+      "step": 132340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0562082417828798e-05,
+      "loss": 0.1817,
+      "step": 132350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0559840365902873e-05,
+      "loss": 0.1763,
+      "step": 132360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0557598313976952e-05,
+      "loss": 0.1859,
+      "step": 132370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0555356262051027e-05,
+      "loss": 0.1805,
+      "step": 132380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0553114210125106e-05,
+      "loss": 0.1784,
+      "step": 132390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0550872158199185e-05,
+      "loss": 0.1773,
+      "step": 132400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.054863010627326e-05,
+      "loss": 0.1831,
+      "step": 132410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.054638805434734e-05,
+      "loss": 0.1852,
+      "step": 132420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0544146002421418e-05,
+      "loss": 0.1728,
+      "step": 132430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0541903950495493e-05,
+      "loss": 0.1762,
+      "step": 132440
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0539661898569572e-05,
+      "loss": 0.1819,
+      "step": 132450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.053741984664365e-05,
+      "loss": 0.1799,
+      "step": 132460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.053517779471773e-05,
+      "loss": 0.1748,
+      "step": 132470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0532935742791805e-05,
+      "loss": 0.1788,
+      "step": 132480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0530693690865883e-05,
+      "loss": 0.183,
+      "step": 132490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.052845163893996e-05,
+      "loss": 0.1748,
+      "step": 132500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0526209587014037e-05,
+      "loss": 0.1816,
+      "step": 132510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0523967535088116e-05,
+      "loss": 0.1863,
+      "step": 132520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.052172548316219e-05,
+      "loss": 0.1757,
+      "step": 132530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.051948343123627e-05,
+      "loss": 0.174,
+      "step": 132540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0517241379310345e-05,
+      "loss": 0.1764,
+      "step": 132550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0514999327384424e-05,
+      "loss": 0.1801,
+      "step": 132560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.05127572754585e-05,
+      "loss": 0.174,
+      "step": 132570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0510515223532578e-05,
+      "loss": 0.1777,
+      "step": 132580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0508273171606657e-05,
+      "loss": 0.1774,
+      "step": 132590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0506031119680732e-05,
+      "loss": 0.178,
+      "step": 132600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.050378906775481e-05,
+      "loss": 0.1831,
+      "step": 132610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0501547015828886e-05,
+      "loss": 0.1792,
+      "step": 132620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0499304963902965e-05,
+      "loss": 0.1786,
+      "step": 132630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.049706291197704e-05,
+      "loss": 0.1786,
+      "step": 132640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.049482086005112e-05,
+      "loss": 0.1771,
+      "step": 132650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0492578808125198e-05,
+      "loss": 0.1831,
+      "step": 132660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0490336756199273e-05,
+      "loss": 0.1763,
+      "step": 132670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0488094704273352e-05,
+      "loss": 0.1789,
+      "step": 132680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0485852652347427e-05,
+      "loss": 0.1795,
+      "step": 132690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0483610600421506e-05,
+      "loss": 0.1821,
+      "step": 132700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0481368548495585e-05,
+      "loss": 0.1762,
+      "step": 132710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0479126496569663e-05,
+      "loss": 0.1774,
+      "step": 132720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.047688444464374e-05,
+      "loss": 0.1796,
+      "step": 132730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0474642392717817e-05,
+      "loss": 0.1751,
+      "step": 132740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0472400340791896e-05,
+      "loss": 0.1769,
+      "step": 132750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.047015828886597e-05,
+      "loss": 0.1798,
+      "step": 132760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.046791623694005e-05,
+      "loss": 0.1784,
+      "step": 132770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0465674185014125e-05,
+      "loss": 0.1841,
+      "step": 132780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0463432133088204e-05,
+      "loss": 0.1756,
+      "step": 132790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0461190081162283e-05,
+      "loss": 0.1823,
+      "step": 132800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0458948029236358e-05,
+      "loss": 0.176,
+      "step": 132810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0456705977310437e-05,
+      "loss": 0.1758,
+      "step": 132820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0454463925384512e-05,
+      "loss": 0.1735,
+      "step": 132830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.045222187345859e-05,
+      "loss": 0.1713,
+      "step": 132840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0449979821532666e-05,
+      "loss": 0.1778,
+      "step": 132850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0447737769606745e-05,
+      "loss": 0.1776,
+      "step": 132860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0445495717680824e-05,
+      "loss": 0.1742,
+      "step": 132870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.04432536657549e-05,
+      "loss": 0.1734,
+      "step": 132880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0441011613828978e-05,
+      "loss": 0.1765,
+      "step": 132890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0438769561903053e-05,
+      "loss": 0.1795,
+      "step": 132900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0436527509977132e-05,
+      "loss": 0.1751,
+      "step": 132910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0434285458051207e-05,
+      "loss": 0.1816,
+      "step": 132920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0432043406125286e-05,
+      "loss": 0.18,
+      "step": 132930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0429801354199365e-05,
+      "loss": 0.1741,
+      "step": 132940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.042755930227344e-05,
+      "loss": 0.1778,
+      "step": 132950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.042531725034752e-05,
+      "loss": 0.1814,
+      "step": 132960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0423075198421597e-05,
+      "loss": 0.1755,
+      "step": 132970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0420833146495673e-05,
+      "loss": 0.1772,
+      "step": 132980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.041859109456975e-05,
+      "loss": 0.1732,
+      "step": 132990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.041634904264383e-05,
+      "loss": 0.1744,
+      "step": 133000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.041410699071791e-05,
+      "loss": 0.1795,
+      "step": 133010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0411864938791984e-05,
+      "loss": 0.1731,
+      "step": 133020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0409622886866063e-05,
+      "loss": 0.1745,
+      "step": 133030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0407380834940138e-05,
+      "loss": 0.1762,
+      "step": 133040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0405138783014217e-05,
+      "loss": 0.186,
+      "step": 133050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0402896731088292e-05,
+      "loss": 0.1762,
+      "step": 133060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.040065467916237e-05,
+      "loss": 0.1809,
+      "step": 133070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.039841262723645e-05,
+      "loss": 0.1779,
+      "step": 133080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0396170575310525e-05,
+      "loss": 0.1789,
+      "step": 133090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0393928523384604e-05,
+      "loss": 0.1742,
+      "step": 133100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.039168647145868e-05,
+      "loss": 0.179,
+      "step": 133110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0389444419532758e-05,
+      "loss": 0.1775,
+      "step": 133120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0387202367606833e-05,
+      "loss": 0.1794,
+      "step": 133130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0384960315680912e-05,
+      "loss": 0.1803,
+      "step": 133140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.038271826375499e-05,
+      "loss": 0.1781,
+      "step": 133150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0380476211829066e-05,
+      "loss": 0.1817,
+      "step": 133160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0378234159903145e-05,
+      "loss": 0.1736,
+      "step": 133170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.037599210797722e-05,
+      "loss": 0.1756,
+      "step": 133180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.03737500560513e-05,
+      "loss": 0.1846,
+      "step": 133190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0371508004125374e-05,
+      "loss": 0.1806,
+      "step": 133200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0369265952199453e-05,
+      "loss": 0.1794,
+      "step": 133210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.036702390027353e-05,
+      "loss": 0.1791,
+      "step": 133220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0364781848347607e-05,
+      "loss": 0.1753,
+      "step": 133230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0362539796421685e-05,
+      "loss": 0.1756,
+      "step": 133240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0360297744495764e-05,
+      "loss": 0.1821,
+      "step": 133250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0358055692569843e-05,
+      "loss": 0.1786,
+      "step": 133260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0355813640643918e-05,
+      "loss": 0.1751,
+      "step": 133270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0353571588717997e-05,
+      "loss": 0.1787,
+      "step": 133280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0351329536792076e-05,
+      "loss": 0.1778,
+      "step": 133290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.034908748486615e-05,
+      "loss": 0.1742,
+      "step": 133300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.034684543294023e-05,
+      "loss": 0.1782,
+      "step": 133310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0344603381014305e-05,
+      "loss": 0.1768,
+      "step": 133320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0342361329088384e-05,
+      "loss": 0.1744,
+      "step": 133330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.034011927716246e-05,
+      "loss": 0.1692,
+      "step": 133340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0337877225236538e-05,
+      "loss": 0.1804,
+      "step": 133350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0335635173310616e-05,
+      "loss": 0.1715,
+      "step": 133360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0333393121384692e-05,
+      "loss": 0.1778,
+      "step": 133370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.033115106945877e-05,
+      "loss": 0.18,
+      "step": 133380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0328909017532846e-05,
+      "loss": 0.175,
+      "step": 133390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0326666965606925e-05,
+      "loss": 0.1775,
+      "step": 133400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0324424913681e-05,
+      "loss": 0.1797,
+      "step": 133410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.032218286175508e-05,
+      "loss": 0.1756,
+      "step": 133420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0319940809829157e-05,
+      "loss": 0.1768,
+      "step": 133430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0317698757903233e-05,
+      "loss": 0.1793,
+      "step": 133440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.031545670597731e-05,
+      "loss": 0.1694,
+      "step": 133450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0313214654051387e-05,
+      "loss": 0.1725,
+      "step": 133460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0310972602125465e-05,
+      "loss": 0.1734,
+      "step": 133470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.030873055019954e-05,
+      "loss": 0.1779,
+      "step": 133480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.030648849827362e-05,
+      "loss": 0.1837,
+      "step": 133490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0304246446347698e-05,
+      "loss": 0.1709,
+      "step": 133500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0302004394421777e-05,
+      "loss": 0.1718,
+      "step": 133510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0299762342495852e-05,
+      "loss": 0.1722,
+      "step": 133520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.029752029056993e-05,
+      "loss": 0.1732,
+      "step": 133530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.029527823864401e-05,
+      "loss": 0.1719,
+      "step": 133540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0293036186718085e-05,
+      "loss": 0.178,
+      "step": 133550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0290794134792164e-05,
+      "loss": 0.1711,
+      "step": 133560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0288552082866242e-05,
+      "loss": 0.1794,
+      "step": 133570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0286310030940318e-05,
+      "loss": 0.1733,
+      "step": 133580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0284067979014396e-05,
+      "loss": 0.1875,
+      "step": 133590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0281825927088472e-05,
+      "loss": 0.1778,
+      "step": 133600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.027958387516255e-05,
+      "loss": 0.1737,
+      "step": 133610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0277341823236626e-05,
+      "loss": 0.1746,
+      "step": 133620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0275099771310705e-05,
+      "loss": 0.1812,
+      "step": 133630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0272857719384783e-05,
+      "loss": 0.1766,
+      "step": 133640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.027061566745886e-05,
+      "loss": 0.175,
+      "step": 133650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0268373615532937e-05,
+      "loss": 0.1812,
+      "step": 133660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0266131563607013e-05,
+      "loss": 0.1745,
+      "step": 133670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.026388951168109e-05,
+      "loss": 0.1786,
+      "step": 133680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0261647459755167e-05,
+      "loss": 0.1803,
+      "step": 133690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0259405407829245e-05,
+      "loss": 0.1739,
+      "step": 133700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0257163355903324e-05,
+      "loss": 0.1762,
+      "step": 133710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.02549213039774e-05,
+      "loss": 0.1749,
+      "step": 133720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0252679252051478e-05,
+      "loss": 0.1712,
+      "step": 133730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0250437200125554e-05,
+      "loss": 0.1721,
+      "step": 133740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0248195148199632e-05,
+      "loss": 0.1754,
+      "step": 133750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.024595309627371e-05,
+      "loss": 0.1773,
+      "step": 133760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0243711044347786e-05,
+      "loss": 0.1714,
+      "step": 133770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0241468992421865e-05,
+      "loss": 0.1749,
+      "step": 133780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0239226940495944e-05,
+      "loss": 0.1759,
+      "step": 133790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0236984888570022e-05,
+      "loss": 0.1742,
+      "step": 133800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0234742836644098e-05,
+      "loss": 0.1737,
+      "step": 133810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0232500784718177e-05,
+      "loss": 0.1755,
+      "step": 133820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0230258732792255e-05,
+      "loss": 0.1777,
+      "step": 133830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.022801668086633e-05,
+      "loss": 0.181,
+      "step": 133840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.022577462894041e-05,
+      "loss": 0.1731,
+      "step": 133850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0223532577014485e-05,
+      "loss": 0.1775,
+      "step": 133860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0221290525088563e-05,
+      "loss": 0.1845,
+      "step": 133870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.021904847316264e-05,
+      "loss": 0.1845,
+      "step": 133880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0216806421236717e-05,
+      "loss": 0.1774,
+      "step": 133890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0214564369310796e-05,
+      "loss": 0.1798,
+      "step": 133900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.021232231738487e-05,
+      "loss": 0.1759,
+      "step": 133910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.021008026545895e-05,
+      "loss": 0.1737,
+      "step": 133920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0207838213533025e-05,
+      "loss": 0.174,
+      "step": 133930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0205596161607104e-05,
+      "loss": 0.1755,
+      "step": 133940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.020335410968118e-05,
+      "loss": 0.1755,
+      "step": 133950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0201112057755258e-05,
+      "loss": 0.1775,
+      "step": 133960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0198870005829337e-05,
+      "loss": 0.1757,
+      "step": 133970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0196627953903412e-05,
+      "loss": 0.1729,
+      "step": 133980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.019438590197749e-05,
+      "loss": 0.1697,
+      "step": 133990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0192143850051566e-05,
+      "loss": 0.1739,
+      "step": 134000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0189901798125645e-05,
+      "loss": 0.1735,
+      "step": 134010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.018765974619972e-05,
+      "loss": 0.1665,
+      "step": 134020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.01854176942738e-05,
+      "loss": 0.1709,
+      "step": 134030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0183175642347878e-05,
+      "loss": 0.1737,
+      "step": 134040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0180933590421957e-05,
+      "loss": 0.1699,
+      "step": 134050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0178691538496032e-05,
+      "loss": 0.1789,
+      "step": 134060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0176673691762703e-05,
+      "loss": 0.1793,
+      "step": 134070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.017443163983678e-05,
+      "loss": 0.172,
+      "step": 134080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0172189587910857e-05,
+      "loss": 0.1749,
+      "step": 134090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0169947535984936e-05,
+      "loss": 0.1772,
+      "step": 134100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0167705484059015e-05,
+      "loss": 0.1729,
+      "step": 134110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.016546343213309e-05,
+      "loss": 0.1778,
+      "step": 134120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.016322138020717e-05,
+      "loss": 0.1745,
+      "step": 134130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0160979328281244e-05,
+      "loss": 0.1736,
+      "step": 134140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0158737276355323e-05,
+      "loss": 0.1802,
+      "step": 134150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0156495224429398e-05,
+      "loss": 0.1741,
+      "step": 134160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0154253172503477e-05,
+      "loss": 0.1748,
+      "step": 134170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0152011120577555e-05,
+      "loss": 0.1717,
+      "step": 134180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.014976906865163e-05,
+      "loss": 0.1753,
+      "step": 134190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.014752701672571e-05,
+      "loss": 0.181,
+      "step": 134200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0145284964799785e-05,
+      "loss": 0.176,
+      "step": 134210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0143042912873863e-05,
+      "loss": 0.178,
+      "step": 134220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.014080086094794e-05,
+      "loss": 0.1769,
+      "step": 134230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0138558809022018e-05,
+      "loss": 0.1725,
+      "step": 134240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0136316757096096e-05,
+      "loss": 0.1762,
+      "step": 134250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.013407470517017e-05,
+      "loss": 0.1746,
+      "step": 134260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.013183265324425e-05,
+      "loss": 0.1788,
+      "step": 134270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0129590601318326e-05,
+      "loss": 0.183,
+      "step": 134280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0127348549392404e-05,
+      "loss": 0.1748,
+      "step": 134290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.012510649746648e-05,
+      "loss": 0.1706,
+      "step": 134300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.012286444554056e-05,
+      "loss": 0.1787,
+      "step": 134310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0120622393614637e-05,
+      "loss": 0.1763,
+      "step": 134320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0118380341688712e-05,
+      "loss": 0.1749,
+      "step": 134330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.011613828976279e-05,
+      "loss": 0.1738,
+      "step": 134340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.011389623783687e-05,
+      "loss": 0.1735,
+      "step": 134350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.011165418591095e-05,
+      "loss": 0.1743,
+      "step": 134360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0109412133985024e-05,
+      "loss": 0.1743,
+      "step": 134370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0107170082059103e-05,
+      "loss": 0.1758,
+      "step": 134380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.010492803013318e-05,
+      "loss": 0.1737,
+      "step": 134390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0102685978207257e-05,
+      "loss": 0.1703,
+      "step": 134400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0100443926281335e-05,
+      "loss": 0.1665,
+      "step": 134410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.009820187435541e-05,
+      "loss": 0.1713,
+      "step": 134420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.009595982242949e-05,
+      "loss": 0.1744,
+      "step": 134430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0093717770503565e-05,
+      "loss": 0.1706,
+      "step": 134440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0091475718577643e-05,
+      "loss": 0.1761,
+      "step": 134450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0089233666651722e-05,
+      "loss": 0.1774,
+      "step": 134460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0086991614725798e-05,
+      "loss": 0.1768,
+      "step": 134470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0084749562799876e-05,
+      "loss": 0.1754,
+      "step": 134480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.008250751087395e-05,
+      "loss": 0.1763,
+      "step": 134490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.008026545894803e-05,
+      "loss": 0.1809,
+      "step": 134500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0078023407022106e-05,
+      "loss": 0.1741,
+      "step": 134510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0075781355096184e-05,
+      "loss": 0.178,
+      "step": 134520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0073539303170263e-05,
+      "loss": 0.1724,
+      "step": 134530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.007129725124434e-05,
+      "loss": 0.1784,
+      "step": 134540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0069055199318417e-05,
+      "loss": 0.18,
+      "step": 134550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0066813147392492e-05,
+      "loss": 0.1751,
+      "step": 134560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.006457109546657e-05,
+      "loss": 0.1761,
+      "step": 134570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0062329043540646e-05,
+      "loss": 0.1743,
+      "step": 134580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0060086991614725e-05,
+      "loss": 0.1689,
+      "step": 134590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0057844939688804e-05,
+      "loss": 0.167,
+      "step": 134600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0055602887762883e-05,
+      "loss": 0.1723,
+      "step": 134610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0053360835836958e-05,
+      "loss": 0.1718,
+      "step": 134620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0051118783911037e-05,
+      "loss": 0.1816,
+      "step": 134630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0048876731985115e-05,
+      "loss": 0.1745,
+      "step": 134640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.004663468005919e-05,
+      "loss": 0.1802,
+      "step": 134650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.004439262813327e-05,
+      "loss": 0.1782,
+      "step": 134660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0042150576207348e-05,
+      "loss": 0.1763,
+      "step": 134670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0039908524281423e-05,
+      "loss": 0.1787,
+      "step": 134680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0037666472355502e-05,
+      "loss": 0.1807,
+      "step": 134690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0035424420429578e-05,
+      "loss": 0.1807,
+      "step": 134700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0033182368503656e-05,
+      "loss": 0.1784,
+      "step": 134710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.003094031657773e-05,
+      "loss": 0.1781,
+      "step": 134720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.002869826465181e-05,
+      "loss": 0.1726,
+      "step": 134730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.002645621272589e-05,
+      "loss": 0.1721,
+      "step": 134740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0024214160799964e-05,
+      "loss": 0.1744,
+      "step": 134750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0021972108874043e-05,
+      "loss": 0.1778,
+      "step": 134760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.001973005694812e-05,
+      "loss": 0.183,
+      "step": 134770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0017488005022197e-05,
+      "loss": 0.1753,
+      "step": 134780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0015245953096276e-05,
+      "loss": 0.1782,
+      "step": 134790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.001300390117035e-05,
+      "loss": 0.1704,
+      "step": 134800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.001076184924443e-05,
+      "loss": 0.1755,
+      "step": 134810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0008519797318505e-05,
+      "loss": 0.1726,
+      "step": 134820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0006277745392584e-05,
+      "loss": 0.172,
+      "step": 134830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.000403569346666e-05,
+      "loss": 0.1788,
+      "step": 134840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0001793641540738e-05,
+      "loss": 0.176,
+      "step": 134850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9999551589614817e-05,
+      "loss": 0.1757,
+      "step": 134860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9997309537688892e-05,
+      "loss": 0.1746,
+      "step": 134870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.999506748576297e-05,
+      "loss": 0.1751,
+      "step": 134880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.999282543383705e-05,
+      "loss": 0.1789,
+      "step": 134890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9990583381911128e-05,
+      "loss": 0.1753,
+      "step": 134900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9988341329985204e-05,
+      "loss": 0.1738,
+      "step": 134910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9986099278059282e-05,
+      "loss": 0.1813,
+      "step": 134920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.998385722613336e-05,
+      "loss": 0.1762,
+      "step": 134930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9981615174207436e-05,
+      "loss": 0.176,
+      "step": 134940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9979373122281515e-05,
+      "loss": 0.1775,
+      "step": 134950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.997713107035559e-05,
+      "loss": 0.1753,
+      "step": 134960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.997488901842967e-05,
+      "loss": 0.1748,
+      "step": 134970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9972646966503744e-05,
+      "loss": 0.1802,
+      "step": 134980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9970404914577823e-05,
+      "loss": 0.1737,
+      "step": 134990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9968162862651902e-05,
+      "loss": 0.1735,
+      "step": 135000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9965920810725977e-05,
+      "loss": 0.1734,
+      "step": 135010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9963678758800056e-05,
+      "loss": 0.1711,
+      "step": 135020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.996143670687413e-05,
+      "loss": 0.1752,
+      "step": 135030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.995919465494821e-05,
+      "loss": 0.1807,
+      "step": 135040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9956952603022285e-05,
+      "loss": 0.1761,
+      "step": 135050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9954710551096364e-05,
+      "loss": 0.1747,
+      "step": 135060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9952468499170443e-05,
+      "loss": 0.172,
+      "step": 135070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9950226447244518e-05,
+      "loss": 0.1777,
+      "step": 135080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9947984395318597e-05,
+      "loss": 0.1742,
+      "step": 135090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9945742343392672e-05,
+      "loss": 0.1748,
+      "step": 135100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.994350029146675e-05,
+      "loss": 0.1736,
+      "step": 135110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9941258239540826e-05,
+      "loss": 0.1731,
+      "step": 135120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9939016187614905e-05,
+      "loss": 0.1766,
+      "step": 135130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9936774135688984e-05,
+      "loss": 0.1778,
+      "step": 135140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9934532083763062e-05,
+      "loss": 0.1734,
+      "step": 135150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9932290031837138e-05,
+      "loss": 0.168,
+      "step": 135160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9930047979911216e-05,
+      "loss": 0.1745,
+      "step": 135170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9927805927985295e-05,
+      "loss": 0.1731,
+      "step": 135180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.992556387605937e-05,
+      "loss": 0.1762,
+      "step": 135190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.992332182413345e-05,
+      "loss": 0.1754,
+      "step": 135200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9921079772207528e-05,
+      "loss": 0.1727,
+      "step": 135210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9918837720281603e-05,
+      "loss": 0.1768,
+      "step": 135220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9916595668355682e-05,
+      "loss": 0.1742,
+      "step": 135230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9914353616429757e-05,
+      "loss": 0.1758,
+      "step": 135240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9912111564503836e-05,
+      "loss": 0.1729,
+      "step": 135250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.990986951257791e-05,
+      "loss": 0.1717,
+      "step": 135260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.990762746065199e-05,
+      "loss": 0.1785,
+      "step": 135270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.990538540872607e-05,
+      "loss": 0.1708,
+      "step": 135280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9903143356800144e-05,
+      "loss": 0.1723,
+      "step": 135290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9900901304874223e-05,
+      "loss": 0.1752,
+      "step": 135300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9898659252948298e-05,
+      "loss": 0.1723,
+      "step": 135310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9896417201022377e-05,
+      "loss": 0.1744,
+      "step": 135320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9894175149096452e-05,
+      "loss": 0.1687,
+      "step": 135330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.989193309717053e-05,
+      "loss": 0.1773,
+      "step": 135340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.988969104524461e-05,
+      "loss": 0.168,
+      "step": 135350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9887448993318685e-05,
+      "loss": 0.1762,
+      "step": 135360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9885206941392764e-05,
+      "loss": 0.179,
+      "step": 135370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.988296488946684e-05,
+      "loss": 0.1773,
+      "step": 135380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9880722837540918e-05,
+      "loss": 0.1704,
+      "step": 135390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9878480785614993e-05,
+      "loss": 0.1731,
+      "step": 135400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.987623873368907e-05,
+      "loss": 0.175,
+      "step": 135410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.987399668176315e-05,
+      "loss": 0.1705,
+      "step": 135420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.987175462983723e-05,
+      "loss": 0.1728,
+      "step": 135430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9869512577911308e-05,
+      "loss": 0.1733,
+      "step": 135440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9867270525985383e-05,
+      "loss": 0.1729,
+      "step": 135450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9865028474059462e-05,
+      "loss": 0.1757,
+      "step": 135460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9862786422133537e-05,
+      "loss": 0.1724,
+      "step": 135470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9860544370207616e-05,
+      "loss": 0.1768,
+      "step": 135480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9858302318281695e-05,
+      "loss": 0.1721,
+      "step": 135490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.985606026635577e-05,
+      "loss": 0.1801,
+      "step": 135500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.985381821442985e-05,
+      "loss": 0.1712,
+      "step": 135510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9851576162503924e-05,
+      "loss": 0.173,
+      "step": 135520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9849334110578003e-05,
+      "loss": 0.1681,
+      "step": 135530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9847092058652078e-05,
+      "loss": 0.1741,
+      "step": 135540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9844850006726157e-05,
+      "loss": 0.1716,
+      "step": 135550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9842607954800235e-05,
+      "loss": 0.1822,
+      "step": 135560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.984036590287431e-05,
+      "loss": 0.176,
+      "step": 135570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.983812385094839e-05,
+      "loss": 0.1801,
+      "step": 135580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9835881799022465e-05,
+      "loss": 0.1719,
+      "step": 135590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9833639747096544e-05,
+      "loss": 0.1723,
+      "step": 135600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.983139769517062e-05,
+      "loss": 0.1752,
+      "step": 135610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9829155643244698e-05,
+      "loss": 0.1779,
+      "step": 135620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9826913591318776e-05,
+      "loss": 0.1721,
+      "step": 135630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.982467153939285e-05,
+      "loss": 0.1733,
+      "step": 135640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.982242948746693e-05,
+      "loss": 0.1713,
+      "step": 135650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9820187435541006e-05,
+      "loss": 0.1729,
+      "step": 135660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9817945383615084e-05,
+      "loss": 0.1755,
+      "step": 135670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9815703331689163e-05,
+      "loss": 0.1706,
+      "step": 135680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9813461279763242e-05,
+      "loss": 0.1761,
+      "step": 135690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9811219227837317e-05,
+      "loss": 0.1768,
+      "step": 135700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9808977175911396e-05,
+      "loss": 0.1753,
+      "step": 135710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9806735123985475e-05,
+      "loss": 0.1726,
+      "step": 135720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.980449307205955e-05,
+      "loss": 0.1767,
+      "step": 135730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.980225102013363e-05,
+      "loss": 0.1812,
+      "step": 135740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9800008968207704e-05,
+      "loss": 0.1838,
+      "step": 135750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9797766916281783e-05,
+      "loss": 0.1743,
+      "step": 135760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.979552486435586e-05,
+      "loss": 0.1733,
+      "step": 135770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9793282812429937e-05,
+      "loss": 0.1796,
+      "step": 135780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9791040760504015e-05,
+      "loss": 0.1757,
+      "step": 135790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.978879870857809e-05,
+      "loss": 0.1724,
+      "step": 135800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.978655665665217e-05,
+      "loss": 0.1751,
+      "step": 135810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9784314604726245e-05,
+      "loss": 0.1716,
+      "step": 135820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9782072552800324e-05,
+      "loss": 0.1768,
+      "step": 135830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9779830500874402e-05,
+      "loss": 0.1716,
+      "step": 135840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9777588448948478e-05,
+      "loss": 0.176,
+      "step": 135850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9775346397022556e-05,
+      "loss": 0.1782,
+      "step": 135860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.977310434509663e-05,
+      "loss": 0.1742,
+      "step": 135870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.977086229317071e-05,
+      "loss": 0.1676,
+      "step": 135880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9768620241244786e-05,
+      "loss": 0.1782,
+      "step": 135890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9766378189318864e-05,
+      "loss": 0.1708,
+      "step": 135900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9764136137392943e-05,
+      "loss": 0.1686,
+      "step": 135910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.976189408546702e-05,
+      "loss": 0.1725,
+      "step": 135920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9759652033541097e-05,
+      "loss": 0.1747,
+      "step": 135930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9757409981615173e-05,
+      "loss": 0.1764,
+      "step": 135940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.975516792968925e-05,
+      "loss": 0.1752,
+      "step": 135950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.975292587776333e-05,
+      "loss": 0.1705,
+      "step": 135960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.975068382583741e-05,
+      "loss": 0.1667,
+      "step": 135970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9748441773911487e-05,
+      "loss": 0.1751,
+      "step": 135980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9746199721985563e-05,
+      "loss": 0.1705,
+      "step": 135990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.974395767005964e-05,
+      "loss": 0.1735,
+      "step": 136000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9741715618133717e-05,
+      "loss": 0.1776,
+      "step": 136010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9739473566207795e-05,
+      "loss": 0.1762,
+      "step": 136020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.973723151428187e-05,
+      "loss": 0.1716,
+      "step": 136030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.973498946235595e-05,
+      "loss": 0.1762,
+      "step": 136040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9732747410430028e-05,
+      "loss": 0.1758,
+      "step": 136050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9730505358504104e-05,
+      "loss": 0.174,
+      "step": 136060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9728263306578182e-05,
+      "loss": 0.175,
+      "step": 136070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9726021254652258e-05,
+      "loss": 0.1705,
+      "step": 136080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.972400340791893e-05,
+      "loss": 0.1716,
+      "step": 136090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9721761355993008e-05,
+      "loss": 0.1755,
+      "step": 136100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9719519304067083e-05,
+      "loss": 0.1785,
+      "step": 136110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.971727725214116e-05,
+      "loss": 0.1712,
+      "step": 136120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9715035200215237e-05,
+      "loss": 0.1752,
+      "step": 136130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9712793148289316e-05,
+      "loss": 0.1726,
+      "step": 136140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.971055109636339e-05,
+      "loss": 0.1752,
+      "step": 136150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.970830904443747e-05,
+      "loss": 0.1757,
+      "step": 136160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.970606699251155e-05,
+      "loss": 0.1734,
+      "step": 136170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9703824940585624e-05,
+      "loss": 0.17,
+      "step": 136180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9701582888659702e-05,
+      "loss": 0.1776,
+      "step": 136190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9699340836733778e-05,
+      "loss": 0.1795,
+      "step": 136200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9697098784807856e-05,
+      "loss": 0.1712,
+      "step": 136210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9694856732881932e-05,
+      "loss": 0.1739,
+      "step": 136220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.969261468095601e-05,
+      "loss": 0.1731,
+      "step": 136230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.969037262903009e-05,
+      "loss": 0.174,
+      "step": 136240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9688130577104168e-05,
+      "loss": 0.1729,
+      "step": 136250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9685888525178243e-05,
+      "loss": 0.1721,
+      "step": 136260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9683646473252322e-05,
+      "loss": 0.1773,
+      "step": 136270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.96814044213264e-05,
+      "loss": 0.1743,
+      "step": 136280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9679162369400476e-05,
+      "loss": 0.1749,
+      "step": 136290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9676920317474555e-05,
+      "loss": 0.1741,
+      "step": 136300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9674678265548634e-05,
+      "loss": 0.174,
+      "step": 136310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.967243621362271e-05,
+      "loss": 0.1793,
+      "step": 136320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9670194161696788e-05,
+      "loss": 0.1793,
+      "step": 136330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9667952109770863e-05,
+      "loss": 0.1748,
+      "step": 136340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.966571005784494e-05,
+      "loss": 0.1739,
+      "step": 136350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9663468005919017e-05,
+      "loss": 0.1757,
+      "step": 136360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9661225953993096e-05,
+      "loss": 0.1743,
+      "step": 136370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9658983902067174e-05,
+      "loss": 0.1741,
+      "step": 136380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.965674185014125e-05,
+      "loss": 0.1752,
+      "step": 136390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.965449979821533e-05,
+      "loss": 0.1755,
+      "step": 136400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9652257746289404e-05,
+      "loss": 0.1737,
+      "step": 136410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9650015694363482e-05,
+      "loss": 0.1753,
+      "step": 136420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9647773642437558e-05,
+      "loss": 0.1758,
+      "step": 136430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9645531590511637e-05,
+      "loss": 0.1738,
+      "step": 136440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9643289538585715e-05,
+      "loss": 0.1672,
+      "step": 136450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.964104748665979e-05,
+      "loss": 0.169,
+      "step": 136460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.963880543473387e-05,
+      "loss": 0.1778,
+      "step": 136470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9636563382807945e-05,
+      "loss": 0.1761,
+      "step": 136480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9634321330882023e-05,
+      "loss": 0.173,
+      "step": 136490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.96320792789561e-05,
+      "loss": 0.1738,
+      "step": 136500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9629837227030177e-05,
+      "loss": 0.1731,
+      "step": 136510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9627595175104256e-05,
+      "loss": 0.1731,
+      "step": 136520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9625353123178335e-05,
+      "loss": 0.1748,
+      "step": 136530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9623111071252414e-05,
+      "loss": 0.1788,
+      "step": 136540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.962086901932649e-05,
+      "loss": 0.179,
+      "step": 136550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9618626967400568e-05,
+      "loss": 0.1717,
+      "step": 136560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9616384915474643e-05,
+      "loss": 0.175,
+      "step": 136570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.961414286354872e-05,
+      "loss": 0.1733,
+      "step": 136580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.96119008116228e-05,
+      "loss": 0.1745,
+      "step": 136590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9609658759696876e-05,
+      "loss": 0.1698,
+      "step": 136600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9607416707770954e-05,
+      "loss": 0.1756,
+      "step": 136610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.960517465584503e-05,
+      "loss": 0.1718,
+      "step": 136620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.960293260391911e-05,
+      "loss": 0.1721,
+      "step": 136630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9600690551993184e-05,
+      "loss": 0.174,
+      "step": 136640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9598448500067262e-05,
+      "loss": 0.1687,
+      "step": 136650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.959620644814134e-05,
+      "loss": 0.1768,
+      "step": 136660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9593964396215417e-05,
+      "loss": 0.1687,
+      "step": 136670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9591722344289495e-05,
+      "loss": 0.1702,
+      "step": 136680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.958948029236357e-05,
+      "loss": 0.1665,
+      "step": 136690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.958723824043765e-05,
+      "loss": 0.1721,
+      "step": 136700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9584996188511725e-05,
+      "loss": 0.1749,
+      "step": 136710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9582754136585803e-05,
+      "loss": 0.1832,
+      "step": 136720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9580512084659882e-05,
+      "loss": 0.1754,
+      "step": 136730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9578270032733957e-05,
+      "loss": 0.1778,
+      "step": 136740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9576027980808036e-05,
+      "loss": 0.1719,
+      "step": 136750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.957378592888211e-05,
+      "loss": 0.1734,
+      "step": 136760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.957154387695619e-05,
+      "loss": 0.1708,
+      "step": 136770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.956930182503027e-05,
+      "loss": 0.1741,
+      "step": 136780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9567059773104348e-05,
+      "loss": 0.1759,
+      "step": 136790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9564817721178423e-05,
+      "loss": 0.1749,
+      "step": 136800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.95625756692525e-05,
+      "loss": 0.1779,
+      "step": 136810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.956033361732658e-05,
+      "loss": 0.1815,
+      "step": 136820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9558091565400656e-05,
+      "loss": 0.1665,
+      "step": 136830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9555849513474734e-05,
+      "loss": 0.1743,
+      "step": 136840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.955360746154881e-05,
+      "loss": 0.1727,
+      "step": 136850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.955136540962289e-05,
+      "loss": 0.1767,
+      "step": 136860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9549123357696967e-05,
+      "loss": 0.171,
+      "step": 136870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9546881305771042e-05,
+      "loss": 0.1772,
+      "step": 136880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.954463925384512e-05,
+      "loss": 0.1712,
+      "step": 136890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9542397201919197e-05,
+      "loss": 0.177,
+      "step": 136900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9540155149993275e-05,
+      "loss": 0.1775,
+      "step": 136910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.953791309806735e-05,
+      "loss": 0.178,
+      "step": 136920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.953567104614143e-05,
+      "loss": 0.1732,
+      "step": 136930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9533428994215508e-05,
+      "loss": 0.1729,
+      "step": 136940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9531186942289583e-05,
+      "loss": 0.1697,
+      "step": 136950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9528944890363662e-05,
+      "loss": 0.1767,
+      "step": 136960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9526702838437737e-05,
+      "loss": 0.1739,
+      "step": 136970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9524460786511816e-05,
+      "loss": 0.1706,
+      "step": 136980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.952221873458589e-05,
+      "loss": 0.1739,
+      "step": 136990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.951997668265997e-05,
+      "loss": 0.1738,
+      "step": 137000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.951773463073405e-05,
+      "loss": 0.1715,
+      "step": 137010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9515492578808124e-05,
+      "loss": 0.1686,
+      "step": 137020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9513250526882203e-05,
+      "loss": 0.1736,
+      "step": 137030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9511008474956278e-05,
+      "loss": 0.1782,
+      "step": 137040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9508766423030357e-05,
+      "loss": 0.1755,
+      "step": 137050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9506524371104436e-05,
+      "loss": 0.1741,
+      "step": 137060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9504282319178514e-05,
+      "loss": 0.1752,
+      "step": 137070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9502040267252593e-05,
+      "loss": 0.1734,
+      "step": 137080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.949979821532667e-05,
+      "loss": 0.1714,
+      "step": 137090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9497556163400747e-05,
+      "loss": 0.1698,
+      "step": 137100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9495314111474823e-05,
+      "loss": 0.1736,
+      "step": 137110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.94930720595489e-05,
+      "loss": 0.1752,
+      "step": 137120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.949083000762298e-05,
+      "loss": 0.174,
+      "step": 137130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9488587955697055e-05,
+      "loss": 0.182,
+      "step": 137140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9486345903771134e-05,
+      "loss": 0.1721,
+      "step": 137150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.948410385184521e-05,
+      "loss": 0.1717,
+      "step": 137160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9481861799919288e-05,
+      "loss": 0.1779,
+      "step": 137170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9479619747993363e-05,
+      "loss": 0.1768,
+      "step": 137180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9477377696067442e-05,
+      "loss": 0.1717,
+      "step": 137190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.947513564414152e-05,
+      "loss": 0.1718,
+      "step": 137200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9472893592215596e-05,
+      "loss": 0.1676,
+      "step": 137210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9470651540289675e-05,
+      "loss": 0.1703,
+      "step": 137220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.946840948836375e-05,
+      "loss": 0.1749,
+      "step": 137230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.946616743643783e-05,
+      "loss": 0.1744,
+      "step": 137240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9463925384511904e-05,
+      "loss": 0.172,
+      "step": 137250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9461683332585983e-05,
+      "loss": 0.1765,
+      "step": 137260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.945944128066006e-05,
+      "loss": 0.1702,
+      "step": 137270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9457199228734137e-05,
+      "loss": 0.1709,
+      "step": 137280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9454957176808216e-05,
+      "loss": 0.1746,
+      "step": 137290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.945271512488229e-05,
+      "loss": 0.1739,
+      "step": 137300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.945047307295637e-05,
+      "loss": 0.1756,
+      "step": 137310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.944823102103045e-05,
+      "loss": 0.1724,
+      "step": 137320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9445988969104527e-05,
+      "loss": 0.172,
+      "step": 137330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9443746917178603e-05,
+      "loss": 0.1741,
+      "step": 137340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.944150486525268e-05,
+      "loss": 0.1717,
+      "step": 137350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.943926281332676e-05,
+      "loss": 0.1782,
+      "step": 137360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9437020761400835e-05,
+      "loss": 0.1771,
+      "step": 137370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9434778709474914e-05,
+      "loss": 0.1733,
+      "step": 137380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.943253665754899e-05,
+      "loss": 0.1762,
+      "step": 137390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9430294605623068e-05,
+      "loss": 0.1703,
+      "step": 137400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9428052553697147e-05,
+      "loss": 0.1732,
+      "step": 137410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9425810501771222e-05,
+      "loss": 0.1692,
+      "step": 137420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.94235684498453e-05,
+      "loss": 0.1762,
+      "step": 137430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9421326397919376e-05,
+      "loss": 0.1786,
+      "step": 137440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9419084345993455e-05,
+      "loss": 0.1732,
+      "step": 137450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.941684229406753e-05,
+      "loss": 0.1725,
+      "step": 137460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.941460024214161e-05,
+      "loss": 0.1731,
+      "step": 137470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9412358190215688e-05,
+      "loss": 0.175,
+      "step": 137480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9410116138289763e-05,
+      "loss": 0.1737,
+      "step": 137490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.940787408636384e-05,
+      "loss": 0.1777,
+      "step": 137500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9405632034437917e-05,
+      "loss": 0.1734,
+      "step": 137510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9403389982511996e-05,
+      "loss": 0.1728,
+      "step": 137520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.940114793058607e-05,
+      "loss": 0.1675,
+      "step": 137530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.939890587866015e-05,
+      "loss": 0.181,
+      "step": 137540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.939666382673423e-05,
+      "loss": 0.1768,
+      "step": 137550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9394421774808304e-05,
+      "loss": 0.1725,
+      "step": 137560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9392179722882383e-05,
+      "loss": 0.1769,
+      "step": 137570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9389937670956458e-05,
+      "loss": 0.1777,
+      "step": 137580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9387695619030537e-05,
+      "loss": 0.1743,
+      "step": 137590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9385453567104615e-05,
+      "loss": 0.1734,
+      "step": 137600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9383211515178694e-05,
+      "loss": 0.1773,
+      "step": 137610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9380969463252773e-05,
+      "loss": 0.1743,
+      "step": 137620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9378727411326848e-05,
+      "loss": 0.1709,
+      "step": 137630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9376485359400927e-05,
+      "loss": 0.1743,
+      "step": 137640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9374243307475002e-05,
+      "loss": 0.1783,
+      "step": 137650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.937200125554908e-05,
+      "loss": 0.1731,
+      "step": 137660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9369759203623156e-05,
+      "loss": 0.1752,
+      "step": 137670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9367517151697235e-05,
+      "loss": 0.1771,
+      "step": 137680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9365275099771314e-05,
+      "loss": 0.1695,
+      "step": 137690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.936303304784539e-05,
+      "loss": 0.1726,
+      "step": 137700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9360790995919468e-05,
+      "loss": 0.1704,
+      "step": 137710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9358548943993543e-05,
+      "loss": 0.1716,
+      "step": 137720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.935630689206762e-05,
+      "loss": 0.1746,
+      "step": 137730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9354064840141697e-05,
+      "loss": 0.1737,
+      "step": 137740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9351822788215776e-05,
+      "loss": 0.1756,
+      "step": 137750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9349580736289854e-05,
+      "loss": 0.1709,
+      "step": 137760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.934733868436393e-05,
+      "loss": 0.1714,
+      "step": 137770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.934509663243801e-05,
+      "loss": 0.1718,
+      "step": 137780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9342854580512084e-05,
+      "loss": 0.1772,
+      "step": 137790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9340612528586163e-05,
+      "loss": 0.1647,
+      "step": 137800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9338370476660238e-05,
+      "loss": 0.1764,
+      "step": 137810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9336128424734317e-05,
+      "loss": 0.1774,
+      "step": 137820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9333886372808395e-05,
+      "loss": 0.1707,
+      "step": 137830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.933164432088247e-05,
+      "loss": 0.1729,
+      "step": 137840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.932940226895655e-05,
+      "loss": 0.1767,
+      "step": 137850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9327160217030628e-05,
+      "loss": 0.1747,
+      "step": 137860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9324918165104703e-05,
+      "loss": 0.1735,
+      "step": 137870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9322676113178782e-05,
+      "loss": 0.1747,
+      "step": 137880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.932043406125286e-05,
+      "loss": 0.1694,
+      "step": 137890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.931819200932694e-05,
+      "loss": 0.1776,
+      "step": 137900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9315949957401015e-05,
+      "loss": 0.1806,
+      "step": 137910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9313707905475094e-05,
+      "loss": 0.1701,
+      "step": 137920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.931146585354917e-05,
+      "loss": 0.1761,
+      "step": 137930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9309223801623248e-05,
+      "loss": 0.1665,
+      "step": 137940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9306981749697323e-05,
+      "loss": 0.1741,
+      "step": 137950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9304739697771402e-05,
+      "loss": 0.1783,
+      "step": 137960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.930249764584548e-05,
+      "loss": 0.1741,
+      "step": 137970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9300255593919556e-05,
+      "loss": 0.1717,
+      "step": 137980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9298013541993634e-05,
+      "loss": 0.1757,
+      "step": 137990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.929577149006771e-05,
+      "loss": 0.1727,
+      "step": 138000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.929352943814179e-05,
+      "loss": 0.1711,
+      "step": 138010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9291287386215864e-05,
+      "loss": 0.1736,
+      "step": 138020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9289045334289943e-05,
+      "loss": 0.1693,
+      "step": 138030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.928680328236402e-05,
+      "loss": 0.1736,
+      "step": 138040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9284561230438097e-05,
+      "loss": 0.1696,
+      "step": 138050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9282319178512175e-05,
+      "loss": 0.1782,
+      "step": 138060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.928007712658625e-05,
+      "loss": 0.1738,
+      "step": 138070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.927783507466033e-05,
+      "loss": 0.1738,
+      "step": 138080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9275593022734405e-05,
+      "loss": 0.1715,
+      "step": 138090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9273350970808483e-05,
+      "loss": 0.1714,
+      "step": 138100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9271108918882562e-05,
+      "loss": 0.1703,
+      "step": 138110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9268866866956637e-05,
+      "loss": 0.1723,
+      "step": 138120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9266624815030716e-05,
+      "loss": 0.1697,
+      "step": 138130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9264382763104795e-05,
+      "loss": 0.1738,
+      "step": 138140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9262140711178874e-05,
+      "loss": 0.1725,
+      "step": 138150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.925989865925295e-05,
+      "loss": 0.1717,
+      "step": 138160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9257656607327028e-05,
+      "loss": 0.1702,
+      "step": 138170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9255414555401106e-05,
+      "loss": 0.1702,
+      "step": 138180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9253172503475182e-05,
+      "loss": 0.1706,
+      "step": 138190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.925093045154926e-05,
+      "loss": 0.1762,
+      "step": 138200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9248688399623336e-05,
+      "loss": 0.1771,
+      "step": 138210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9246670552890007e-05,
+      "loss": 0.1721,
+      "step": 138220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9244428500964086e-05,
+      "loss": 0.1662,
+      "step": 138230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.924218644903816e-05,
+      "loss": 0.1714,
+      "step": 138240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.923994439711224e-05,
+      "loss": 0.1727,
+      "step": 138250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9237702345186315e-05,
+      "loss": 0.1739,
+      "step": 138260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9235460293260394e-05,
+      "loss": 0.1663,
+      "step": 138270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.923321824133447e-05,
+      "loss": 0.1765,
+      "step": 138280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9230976189408548e-05,
+      "loss": 0.1759,
+      "step": 138290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9228734137482627e-05,
+      "loss": 0.1677,
+      "step": 138300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9226492085556702e-05,
+      "loss": 0.1742,
+      "step": 138310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.922425003363078e-05,
+      "loss": 0.1719,
+      "step": 138320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9222007981704856e-05,
+      "loss": 0.1766,
+      "step": 138330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9219765929778935e-05,
+      "loss": 0.1706,
+      "step": 138340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.921752387785301e-05,
+      "loss": 0.1709,
+      "step": 138350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.921528182592709e-05,
+      "loss": 0.1699,
+      "step": 138360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9213039774001167e-05,
+      "loss": 0.1708,
+      "step": 138370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9210797722075243e-05,
+      "loss": 0.1722,
+      "step": 138380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.920855567014932e-05,
+      "loss": 0.1685,
+      "step": 138390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9206313618223397e-05,
+      "loss": 0.171,
+      "step": 138400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9204071566297475e-05,
+      "loss": 0.176,
+      "step": 138410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9201829514371554e-05,
+      "loss": 0.1711,
+      "step": 138420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.919958746244563e-05,
+      "loss": 0.1754,
+      "step": 138430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9197345410519708e-05,
+      "loss": 0.1716,
+      "step": 138440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9195103358593787e-05,
+      "loss": 0.1732,
+      "step": 138450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9192861306667866e-05,
+      "loss": 0.1717,
+      "step": 138460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.919061925474194e-05,
+      "loss": 0.1709,
+      "step": 138470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.918837720281602e-05,
+      "loss": 0.1724,
+      "step": 138480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9186135150890095e-05,
+      "loss": 0.17,
+      "step": 138490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9183893098964174e-05,
+      "loss": 0.1722,
+      "step": 138500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9181651047038252e-05,
+      "loss": 0.1718,
+      "step": 138510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9179408995112328e-05,
+      "loss": 0.173,
+      "step": 138520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9177166943186407e-05,
+      "loss": 0.1739,
+      "step": 138530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9174924891260482e-05,
+      "loss": 0.1744,
+      "step": 138540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.917268283933456e-05,
+      "loss": 0.1738,
+      "step": 138550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9170440787408636e-05,
+      "loss": 0.1723,
+      "step": 138560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9168198735482715e-05,
+      "loss": 0.1708,
+      "step": 138570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9165956683556793e-05,
+      "loss": 0.1756,
+      "step": 138580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.916371463163087e-05,
+      "loss": 0.1737,
+      "step": 138590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9161472579704947e-05,
+      "loss": 0.1687,
+      "step": 138600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9159230527779023e-05,
+      "loss": 0.1714,
+      "step": 138610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.91569884758531e-05,
+      "loss": 0.1799,
+      "step": 138620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9154746423927177e-05,
+      "loss": 0.1737,
+      "step": 138630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9152504372001256e-05,
+      "loss": 0.1711,
+      "step": 138640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9150262320075334e-05,
+      "loss": 0.1745,
+      "step": 138650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.914802026814941e-05,
+      "loss": 0.1744,
+      "step": 138660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9145778216223488e-05,
+      "loss": 0.1731,
+      "step": 138670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9143536164297564e-05,
+      "loss": 0.1745,
+      "step": 138680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9141294112371642e-05,
+      "loss": 0.1732,
+      "step": 138690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.913905206044572e-05,
+      "loss": 0.1684,
+      "step": 138700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.91368100085198e-05,
+      "loss": 0.173,
+      "step": 138710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.913456795659388e-05,
+      "loss": 0.1719,
+      "step": 138720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9132325904667954e-05,
+      "loss": 0.1734,
+      "step": 138730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9130083852742033e-05,
+      "loss": 0.1734,
+      "step": 138740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9127841800816108e-05,
+      "loss": 0.1668,
+      "step": 138750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9125599748890187e-05,
+      "loss": 0.1758,
+      "step": 138760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9123357696964262e-05,
+      "loss": 0.173,
+      "step": 138770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.912111564503834e-05,
+      "loss": 0.1784,
+      "step": 138780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.911887359311242e-05,
+      "loss": 0.1796,
+      "step": 138790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9116631541186495e-05,
+      "loss": 0.1689,
+      "step": 138800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9114389489260573e-05,
+      "loss": 0.1765,
+      "step": 138810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.911214743733465e-05,
+      "loss": 0.1684,
+      "step": 138820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9109905385408727e-05,
+      "loss": 0.1705,
+      "step": 138830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9107663333482803e-05,
+      "loss": 0.1709,
+      "step": 138840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.910542128155688e-05,
+      "loss": 0.1667,
+      "step": 138850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.910317922963096e-05,
+      "loss": 0.1681,
+      "step": 138860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9100937177705036e-05,
+      "loss": 0.1758,
+      "step": 138870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9098695125779114e-05,
+      "loss": 0.1736,
+      "step": 138880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.909645307385319e-05,
+      "loss": 0.1697,
+      "step": 138890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9094211021927268e-05,
+      "loss": 0.1677,
+      "step": 138900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9091968970001344e-05,
+      "loss": 0.1724,
+      "step": 138910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9089726918075422e-05,
+      "loss": 0.1741,
+      "step": 138920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.90874848661495e-05,
+      "loss": 0.1706,
+      "step": 138930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9085242814223576e-05,
+      "loss": 0.1756,
+      "step": 138940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9083000762297655e-05,
+      "loss": 0.1684,
+      "step": 138950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9080758710371734e-05,
+      "loss": 0.1761,
+      "step": 138960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.907851665844581e-05,
+      "loss": 0.1713,
+      "step": 138970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9076274606519888e-05,
+      "loss": 0.1754,
+      "step": 138980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9074032554593967e-05,
+      "loss": 0.1769,
+      "step": 138990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9071790502668045e-05,
+      "loss": 0.1756,
+      "step": 139000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.906954845074212e-05,
+      "loss": 0.1732,
+      "step": 139010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.90673063988162e-05,
+      "loss": 0.1701,
+      "step": 139020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9065064346890275e-05,
+      "loss": 0.1705,
+      "step": 139030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9062822294964353e-05,
+      "loss": 0.1684,
+      "step": 139040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.906058024303843e-05,
+      "loss": 0.1737,
+      "step": 139050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9058338191112507e-05,
+      "loss": 0.1746,
+      "step": 139060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9056096139186586e-05,
+      "loss": 0.177,
+      "step": 139070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.905385408726066e-05,
+      "loss": 0.1717,
+      "step": 139080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.905161203533474e-05,
+      "loss": 0.1678,
+      "step": 139090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9049369983408816e-05,
+      "loss": 0.171,
+      "step": 139100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9047127931482894e-05,
+      "loss": 0.1769,
+      "step": 139110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.904488587955697e-05,
+      "loss": 0.1679,
+      "step": 139120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9042643827631048e-05,
+      "loss": 0.1739,
+      "step": 139130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9040401775705127e-05,
+      "loss": 0.1734,
+      "step": 139140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9038159723779202e-05,
+      "loss": 0.1707,
+      "step": 139150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.903591767185328e-05,
+      "loss": 0.1736,
+      "step": 139160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9033675619927356e-05,
+      "loss": 0.1781,
+      "step": 139170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9031433568001435e-05,
+      "loss": 0.1692,
+      "step": 139180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.902919151607551e-05,
+      "loss": 0.1694,
+      "step": 139190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.902694946414959e-05,
+      "loss": 0.1745,
+      "step": 139200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9024707412223668e-05,
+      "loss": 0.1747,
+      "step": 139210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9022465360297743e-05,
+      "loss": 0.1785,
+      "step": 139220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9020223308371822e-05,
+      "loss": 0.1754,
+      "step": 139230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.90179812564459e-05,
+      "loss": 0.1729,
+      "step": 139240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.901573920451998e-05,
+      "loss": 0.1727,
+      "step": 139250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9013497152594055e-05,
+      "loss": 0.1747,
+      "step": 139260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9011255100668133e-05,
+      "loss": 0.1776,
+      "step": 139270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9009013048742212e-05,
+      "loss": 0.1667,
+      "step": 139280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9006770996816287e-05,
+      "loss": 0.1733,
+      "step": 139290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9004528944890366e-05,
+      "loss": 0.1748,
+      "step": 139300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.900228689296444e-05,
+      "loss": 0.1768,
+      "step": 139310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.900004484103852e-05,
+      "loss": 0.1737,
+      "step": 139320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.89978027891126e-05,
+      "loss": 0.1731,
+      "step": 139330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8995560737186674e-05,
+      "loss": 0.1718,
+      "step": 139340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8993318685260753e-05,
+      "loss": 0.1749,
+      "step": 139350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.899107663333483e-05,
+      "loss": 0.1716,
+      "step": 139360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8988834581408907e-05,
+      "loss": 0.1727,
+      "step": 139370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8986592529482982e-05,
+      "loss": 0.1712,
+      "step": 139380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.898435047755706e-05,
+      "loss": 0.1681,
+      "step": 139390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.898210842563114e-05,
+      "loss": 0.1687,
+      "step": 139400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8979866373705215e-05,
+      "loss": 0.1716,
+      "step": 139410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8977624321779294e-05,
+      "loss": 0.1711,
+      "step": 139420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.897538226985337e-05,
+      "loss": 0.1701,
+      "step": 139430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8973140217927448e-05,
+      "loss": 0.1695,
+      "step": 139440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8970898166001523e-05,
+      "loss": 0.1771,
+      "step": 139450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8968656114075602e-05,
+      "loss": 0.1695,
+      "step": 139460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.896641406214968e-05,
+      "loss": 0.1751,
+      "step": 139470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8964172010223756e-05,
+      "loss": 0.1679,
+      "step": 139480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8961929958297835e-05,
+      "loss": 0.1702,
+      "step": 139490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8959687906371913e-05,
+      "loss": 0.174,
+      "step": 139500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.895744585444599e-05,
+      "loss": 0.1735,
+      "step": 139510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8955203802520067e-05,
+      "loss": 0.1665,
+      "step": 139520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8952961750594146e-05,
+      "loss": 0.1686,
+      "step": 139530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8950719698668225e-05,
+      "loss": 0.1703,
+      "step": 139540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.89484776467423e-05,
+      "loss": 0.1746,
+      "step": 139550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.894623559481638e-05,
+      "loss": 0.1721,
+      "step": 139560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8943993542890454e-05,
+      "loss": 0.1753,
+      "step": 139570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8941751490964533e-05,
+      "loss": 0.1776,
+      "step": 139580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.893950943903861e-05,
+      "loss": 0.1708,
+      "step": 139590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8937267387112687e-05,
+      "loss": 0.1751,
+      "step": 139600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8935025335186766e-05,
+      "loss": 0.1735,
+      "step": 139610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.893278328326084e-05,
+      "loss": 0.1727,
+      "step": 139620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.893054123133492e-05,
+      "loss": 0.1757,
+      "step": 139630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8928299179408995e-05,
+      "loss": 0.1731,
+      "step": 139640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8926057127483074e-05,
+      "loss": 0.1694,
+      "step": 139650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.892381507555715e-05,
+      "loss": 0.1691,
+      "step": 139660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8921573023631228e-05,
+      "loss": 0.1743,
+      "step": 139670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8919330971705307e-05,
+      "loss": 0.1684,
+      "step": 139680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8917088919779382e-05,
+      "loss": 0.1701,
+      "step": 139690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.891484686785346e-05,
+      "loss": 0.1723,
+      "step": 139700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8912604815927536e-05,
+      "loss": 0.1722,
+      "step": 139710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8910362764001615e-05,
+      "loss": 0.1698,
+      "step": 139720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.890812071207569e-05,
+      "loss": 0.1731,
+      "step": 139730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.890587866014977e-05,
+      "loss": 0.176,
+      "step": 139740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8903636608223847e-05,
+      "loss": 0.1723,
+      "step": 139750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8901394556297923e-05,
+      "loss": 0.1693,
+      "step": 139760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8899152504372e-05,
+      "loss": 0.1707,
+      "step": 139770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.889691045244608e-05,
+      "loss": 0.1739,
+      "step": 139780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.889466840052016e-05,
+      "loss": 0.1725,
+      "step": 139790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8892426348594234e-05,
+      "loss": 0.1705,
+      "step": 139800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8890184296668313e-05,
+      "loss": 0.173,
+      "step": 139810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8887942244742392e-05,
+      "loss": 0.1736,
+      "step": 139820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8885700192816467e-05,
+      "loss": 0.1714,
+      "step": 139830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8883458140890546e-05,
+      "loss": 0.1711,
+      "step": 139840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.888121608896462e-05,
+      "loss": 0.1758,
+      "step": 139850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.88789740370387e-05,
+      "loss": 0.1738,
+      "step": 139860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8876731985112775e-05,
+      "loss": 0.1746,
+      "step": 139870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8874489933186854e-05,
+      "loss": 0.172,
+      "step": 139880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8872247881260933e-05,
+      "loss": 0.1769,
+      "step": 139890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8870005829335008e-05,
+      "loss": 0.1754,
+      "step": 139900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8867763777409087e-05,
+      "loss": 0.1698,
+      "step": 139910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8865521725483162e-05,
+      "loss": 0.1722,
+      "step": 139920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.886327967355724e-05,
+      "loss": 0.1669,
+      "step": 139930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8861037621631316e-05,
+      "loss": 0.1695,
+      "step": 139940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8858795569705395e-05,
+      "loss": 0.1784,
+      "step": 139950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8856553517779473e-05,
+      "loss": 0.1734,
+      "step": 139960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.885431146585355e-05,
+      "loss": 0.1738,
+      "step": 139970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8852069413927627e-05,
+      "loss": 0.1724,
+      "step": 139980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8849827362001703e-05,
+      "loss": 0.1735,
+      "step": 139990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.884758531007578e-05,
+      "loss": 0.1776,
+      "step": 140000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8845343258149857e-05,
+      "loss": 0.171,
+      "step": 140010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8843101206223936e-05,
+      "loss": 0.1715,
+      "step": 140020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8840859154298014e-05,
+      "loss": 0.1704,
+      "step": 140030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8838617102372093e-05,
+      "loss": 0.174,
+      "step": 140040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.883637505044617e-05,
+      "loss": 0.1679,
+      "step": 140050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8834132998520247e-05,
+      "loss": 0.1714,
+      "step": 140060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8831890946594326e-05,
+      "loss": 0.1679,
+      "step": 140070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.88296488946684e-05,
+      "loss": 0.171,
+      "step": 140080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.882740684274248e-05,
+      "loss": 0.17,
+      "step": 140090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.882516479081656e-05,
+      "loss": 0.1722,
+      "step": 140100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8822922738890634e-05,
+      "loss": 0.1758,
+      "step": 140110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8820680686964713e-05,
+      "loss": 0.1705,
+      "step": 140120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8818438635038788e-05,
+      "loss": 0.1699,
+      "step": 140130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8816196583112867e-05,
+      "loss": 0.1742,
+      "step": 140140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8813954531186942e-05,
+      "loss": 0.1733,
+      "step": 140150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.881171247926102e-05,
+      "loss": 0.1723,
+      "step": 140160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.88094704273351e-05,
+      "loss": 0.1684,
+      "step": 140170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8807228375409175e-05,
+      "loss": 0.1727,
+      "step": 140180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8804986323483253e-05,
+      "loss": 0.1751,
+      "step": 140190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.880274427155733e-05,
+      "loss": 0.1712,
+      "step": 140200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8800502219631408e-05,
+      "loss": 0.1714,
+      "step": 140210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8798260167705483e-05,
+      "loss": 0.1649,
+      "step": 140220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.879601811577956e-05,
+      "loss": 0.174,
+      "step": 140230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.879377606385364e-05,
+      "loss": 0.1743,
+      "step": 140240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8791534011927716e-05,
+      "loss": 0.174,
+      "step": 140250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8789291960001794e-05,
+      "loss": 0.1675,
+      "step": 140260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.878704990807587e-05,
+      "loss": 0.1698,
+      "step": 140270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.878480785614995e-05,
+      "loss": 0.1626,
+      "step": 140280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8782565804224027e-05,
+      "loss": 0.1631,
+      "step": 140290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8780323752298102e-05,
+      "loss": 0.1753,
+      "step": 140300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.877808170037218e-05,
+      "loss": 0.1691,
+      "step": 140310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.877583964844626e-05,
+      "loss": 0.1752,
+      "step": 140320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.877359759652034e-05,
+      "loss": 0.1711,
+      "step": 140330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8771355544594414e-05,
+      "loss": 0.1711,
+      "step": 140340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8769113492668493e-05,
+      "loss": 0.1754,
+      "step": 140350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8766871440742568e-05,
+      "loss": 0.1684,
+      "step": 140360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8764629388816647e-05,
+      "loss": 0.167,
+      "step": 140370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8762387336890725e-05,
+      "loss": 0.1749,
+      "step": 140380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.87601452849648e-05,
+      "loss": 0.1749,
+      "step": 140390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.875790323303888e-05,
+      "loss": 0.171,
+      "step": 140400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8755661181112955e-05,
+      "loss": 0.1729,
+      "step": 140410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8753419129187033e-05,
+      "loss": 0.1739,
+      "step": 140420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.875117707726111e-05,
+      "loss": 0.1691,
+      "step": 140430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8748935025335188e-05,
+      "loss": 0.17,
+      "step": 140440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8746692973409266e-05,
+      "loss": 0.1731,
+      "step": 140450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.874445092148334e-05,
+      "loss": 0.1716,
+      "step": 140460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.874220886955742e-05,
+      "loss": 0.1705,
+      "step": 140470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8739966817631496e-05,
+      "loss": 0.1744,
+      "step": 140480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8737724765705574e-05,
+      "loss": 0.1733,
+      "step": 140490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.873548271377965e-05,
+      "loss": 0.1758,
+      "step": 140500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.873324066185373e-05,
+      "loss": 0.1699,
+      "step": 140510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8730998609927807e-05,
+      "loss": 0.1715,
+      "step": 140520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8728756558001882e-05,
+      "loss": 0.1698,
+      "step": 140530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.872651450607596e-05,
+      "loss": 0.174,
+      "step": 140540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8724272454150036e-05,
+      "loss": 0.1771,
+      "step": 140550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8722030402224115e-05,
+      "loss": 0.1735,
+      "step": 140560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8719788350298194e-05,
+      "loss": 0.1753,
+      "step": 140570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8717546298372273e-05,
+      "loss": 0.1728,
+      "step": 140580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8715304246446348e-05,
+      "loss": 0.168,
+      "step": 140590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8713062194520427e-05,
+      "loss": 0.1752,
+      "step": 140600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8710820142594505e-05,
+      "loss": 0.1734,
+      "step": 140610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.870857809066858e-05,
+      "loss": 0.1755,
+      "step": 140620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.870633603874266e-05,
+      "loss": 0.1732,
+      "step": 140630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8704093986816735e-05,
+      "loss": 0.1771,
+      "step": 140640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8701851934890813e-05,
+      "loss": 0.1725,
+      "step": 140650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8699609882964892e-05,
+      "loss": 0.167,
+      "step": 140660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8697367831038968e-05,
+      "loss": 0.1648,
+      "step": 140670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8695125779113046e-05,
+      "loss": 0.166,
+      "step": 140680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.869288372718712e-05,
+      "loss": 0.1722,
+      "step": 140690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.86906416752612e-05,
+      "loss": 0.1682,
+      "step": 140700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.868839962333528e-05,
+      "loss": 0.167,
+      "step": 140710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8686157571409354e-05,
+      "loss": 0.1646,
+      "step": 140720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8683915519483433e-05,
+      "loss": 0.1678,
+      "step": 140730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.868167346755751e-05,
+      "loss": 0.1745,
+      "step": 140740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8679431415631587e-05,
+      "loss": 0.1697,
+      "step": 140750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8677189363705662e-05,
+      "loss": 0.1713,
+      "step": 140760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.867494731177974e-05,
+      "loss": 0.1753,
+      "step": 140770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.867270525985382e-05,
+      "loss": 0.1711,
+      "step": 140780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8670463207927895e-05,
+      "loss": 0.1736,
+      "step": 140790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8668221156001974e-05,
+      "loss": 0.1664,
+      "step": 140800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.866597910407605e-05,
+      "loss": 0.1696,
+      "step": 140810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8663737052150128e-05,
+      "loss": 0.1747,
+      "step": 140820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8661495000224207e-05,
+      "loss": 0.1716,
+      "step": 140830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8659252948298282e-05,
+      "loss": 0.1703,
+      "step": 140840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.865701089637236e-05,
+      "loss": 0.1704,
+      "step": 140850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.865476884444644e-05,
+      "loss": 0.1671,
+      "step": 140860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8652526792520518e-05,
+      "loss": 0.173,
+      "step": 140870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8650284740594594e-05,
+      "loss": 0.1725,
+      "step": 140880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8648042688668672e-05,
+      "loss": 0.1671,
+      "step": 140890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8645800636742748e-05,
+      "loss": 0.1754,
+      "step": 140900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8643558584816826e-05,
+      "loss": 0.1729,
+      "step": 140910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8641316532890905e-05,
+      "loss": 0.18,
+      "step": 140920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.863907448096498e-05,
+      "loss": 0.1778,
+      "step": 140930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.863683242903906e-05,
+      "loss": 0.1741,
+      "step": 140940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8634590377113134e-05,
+      "loss": 0.1768,
+      "step": 140950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8632348325187213e-05,
+      "loss": 0.1737,
+      "step": 140960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.863010627326129e-05,
+      "loss": 0.1691,
+      "step": 140970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8627864221335367e-05,
+      "loss": 0.1617,
+      "step": 140980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8625622169409446e-05,
+      "loss": 0.1754,
+      "step": 140990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.862338011748352e-05,
+      "loss": 0.1688,
+      "step": 141000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.86211380655576e-05,
+      "loss": 0.1674,
+      "step": 141010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8618896013631675e-05,
+      "loss": 0.1697,
+      "step": 141020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8616653961705754e-05,
+      "loss": 0.1674,
+      "step": 141030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.861441190977983e-05,
+      "loss": 0.1715,
+      "step": 141040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8612169857853908e-05,
+      "loss": 0.1723,
+      "step": 141050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8609927805927987e-05,
+      "loss": 0.1723,
+      "step": 141060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8607685754002062e-05,
+      "loss": 0.1691,
+      "step": 141070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.860544370207614e-05,
+      "loss": 0.1658,
+      "step": 141080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8603201650150216e-05,
+      "loss": 0.1694,
+      "step": 141090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8600959598224295e-05,
+      "loss": 0.1709,
+      "step": 141100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8598717546298374e-05,
+      "loss": 0.1725,
+      "step": 141110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8596475494372452e-05,
+      "loss": 0.1745,
+      "step": 141120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8594233442446528e-05,
+      "loss": 0.1762,
+      "step": 141130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8591991390520606e-05,
+      "loss": 0.1717,
+      "step": 141140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8589749338594685e-05,
+      "loss": 0.1715,
+      "step": 141150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8587731491861353e-05,
+      "loss": 0.1782,
+      "step": 141160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.858548943993543e-05,
+      "loss": 0.1732,
+      "step": 141170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8583247388009507e-05,
+      "loss": 0.1688,
+      "step": 141180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8581005336083586e-05,
+      "loss": 0.1696,
+      "step": 141190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8578763284157664e-05,
+      "loss": 0.172,
+      "step": 141200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.857652123223174e-05,
+      "loss": 0.1712,
+      "step": 141210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.857427918030582e-05,
+      "loss": 0.1719,
+      "step": 141220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8572037128379894e-05,
+      "loss": 0.1736,
+      "step": 141230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8569795076453972e-05,
+      "loss": 0.1767,
+      "step": 141240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8567553024528048e-05,
+      "loss": 0.1721,
+      "step": 141250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8565310972602126e-05,
+      "loss": 0.1759,
+      "step": 141260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8563068920676205e-05,
+      "loss": 0.1671,
+      "step": 141270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.856082686875028e-05,
+      "loss": 0.1739,
+      "step": 141280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.855858481682436e-05,
+      "loss": 0.1716,
+      "step": 141290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8556342764898435e-05,
+      "loss": 0.1689,
+      "step": 141300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8554100712972513e-05,
+      "loss": 0.1754,
+      "step": 141310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.855185866104659e-05,
+      "loss": 0.1697,
+      "step": 141320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8549616609120667e-05,
+      "loss": 0.163,
+      "step": 141330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8547374557194746e-05,
+      "loss": 0.1786,
+      "step": 141340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.854513250526882e-05,
+      "loss": 0.1781,
+      "step": 141350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.85428904533429e-05,
+      "loss": 0.1697,
+      "step": 141360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8540648401416975e-05,
+      "loss": 0.1766,
+      "step": 141370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8538406349491054e-05,
+      "loss": 0.1726,
+      "step": 141380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8536164297565133e-05,
+      "loss": 0.1679,
+      "step": 141390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8533922245639208e-05,
+      "loss": 0.171,
+      "step": 141400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8531680193713287e-05,
+      "loss": 0.174,
+      "step": 141410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8529438141787366e-05,
+      "loss": 0.1697,
+      "step": 141420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8527196089861444e-05,
+      "loss": 0.1691,
+      "step": 141430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.852495403793552e-05,
+      "loss": 0.172,
+      "step": 141440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.85227119860096e-05,
+      "loss": 0.1695,
+      "step": 141450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8520469934083674e-05,
+      "loss": 0.1715,
+      "step": 141460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8518227882157752e-05,
+      "loss": 0.1714,
+      "step": 141470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.851598583023183e-05,
+      "loss": 0.1771,
+      "step": 141480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8513743778305906e-05,
+      "loss": 0.1674,
+      "step": 141490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8511501726379985e-05,
+      "loss": 0.1715,
+      "step": 141500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.850925967445406e-05,
+      "loss": 0.1668,
+      "step": 141510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.850701762252814e-05,
+      "loss": 0.1683,
+      "step": 141520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8504775570602215e-05,
+      "loss": 0.1738,
+      "step": 141530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8502533518676293e-05,
+      "loss": 0.1717,
+      "step": 141540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8500291466750372e-05,
+      "loss": 0.1703,
+      "step": 141550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8498049414824447e-05,
+      "loss": 0.1776,
+      "step": 141560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8495807362898526e-05,
+      "loss": 0.1689,
+      "step": 141570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.84935653109726e-05,
+      "loss": 0.1685,
+      "step": 141580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.849132325904668e-05,
+      "loss": 0.1634,
+      "step": 141590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8489081207120755e-05,
+      "loss": 0.1735,
+      "step": 141600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8486839155194834e-05,
+      "loss": 0.1735,
+      "step": 141610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8484597103268913e-05,
+      "loss": 0.1744,
+      "step": 141620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8482355051342988e-05,
+      "loss": 0.1662,
+      "step": 141630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8480112999417067e-05,
+      "loss": 0.1743,
+      "step": 141640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8477870947491142e-05,
+      "loss": 0.1678,
+      "step": 141650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.847562889556522e-05,
+      "loss": 0.1704,
+      "step": 141660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.84733868436393e-05,
+      "loss": 0.1712,
+      "step": 141670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.847114479171338e-05,
+      "loss": 0.1771,
+      "step": 141680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8468902739787454e-05,
+      "loss": 0.1743,
+      "step": 141690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8466660687861532e-05,
+      "loss": 0.1775,
+      "step": 141700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.846441863593561e-05,
+      "loss": 0.1735,
+      "step": 141710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8462176584009686e-05,
+      "loss": 0.1722,
+      "step": 141720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8459934532083765e-05,
+      "loss": 0.1705,
+      "step": 141730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8457692480157844e-05,
+      "loss": 0.1736,
+      "step": 141740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.845545042823192e-05,
+      "loss": 0.1702,
+      "step": 141750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8453208376305998e-05,
+      "loss": 0.1688,
+      "step": 141760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8450966324380073e-05,
+      "loss": 0.1733,
+      "step": 141770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8448724272454152e-05,
+      "loss": 0.1652,
+      "step": 141780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8446482220528227e-05,
+      "loss": 0.1708,
+      "step": 141790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8444240168602306e-05,
+      "loss": 0.1675,
+      "step": 141800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8441998116676385e-05,
+      "loss": 0.1713,
+      "step": 141810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.843975606475046e-05,
+      "loss": 0.1708,
+      "step": 141820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.843751401282454e-05,
+      "loss": 0.1753,
+      "step": 141830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8435271960898614e-05,
+      "loss": 0.1614,
+      "step": 141840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8433029908972693e-05,
+      "loss": 0.1755,
+      "step": 141850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8430787857046768e-05,
+      "loss": 0.1777,
+      "step": 141860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8428545805120847e-05,
+      "loss": 0.1724,
+      "step": 141870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8426303753194926e-05,
+      "loss": 0.1625,
+      "step": 141880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8424061701269e-05,
+      "loss": 0.1734,
+      "step": 141890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.842181964934308e-05,
+      "loss": 0.1711,
+      "step": 141900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8419577597417155e-05,
+      "loss": 0.1737,
+      "step": 141910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8417335545491234e-05,
+      "loss": 0.17,
+      "step": 141920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8415093493565312e-05,
+      "loss": 0.1694,
+      "step": 141930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8412851441639388e-05,
+      "loss": 0.1672,
+      "step": 141940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8410609389713466e-05,
+      "loss": 0.1713,
+      "step": 141950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8408367337787545e-05,
+      "loss": 0.1689,
+      "step": 141960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8406125285861624e-05,
+      "loss": 0.1734,
+      "step": 141970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.84038832339357e-05,
+      "loss": 0.1708,
+      "step": 141980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8401641182009778e-05,
+      "loss": 0.17,
+      "step": 141990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8399399130083853e-05,
+      "loss": 0.1656,
+      "step": 142000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8397157078157932e-05,
+      "loss": 0.17,
+      "step": 142010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.839491502623201e-05,
+      "loss": 0.1667,
+      "step": 142020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8392672974306086e-05,
+      "loss": 0.1705,
+      "step": 142030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8390430922380165e-05,
+      "loss": 0.1677,
+      "step": 142040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.838818887045424e-05,
+      "loss": 0.1717,
+      "step": 142050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.838594681852832e-05,
+      "loss": 0.1694,
+      "step": 142060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8383704766602394e-05,
+      "loss": 0.1685,
+      "step": 142070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8381462714676473e-05,
+      "loss": 0.1747,
+      "step": 142080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.837922066275055e-05,
+      "loss": 0.1689,
+      "step": 142090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8376978610824627e-05,
+      "loss": 0.1698,
+      "step": 142100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8374736558898706e-05,
+      "loss": 0.168,
+      "step": 142110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.837249450697278e-05,
+      "loss": 0.1741,
+      "step": 142120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.837025245504686e-05,
+      "loss": 0.1705,
+      "step": 142130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8368010403120935e-05,
+      "loss": 0.1758,
+      "step": 142140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8365768351195014e-05,
+      "loss": 0.1757,
+      "step": 142150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8363526299269092e-05,
+      "loss": 0.1664,
+      "step": 142160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8361284247343168e-05,
+      "loss": 0.1718,
+      "step": 142170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8359042195417246e-05,
+      "loss": 0.1683,
+      "step": 142180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8356800143491322e-05,
+      "loss": 0.1616,
+      "step": 142190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.83545580915654e-05,
+      "loss": 0.1712,
+      "step": 142200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.835231603963948e-05,
+      "loss": 0.1724,
+      "step": 142210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8350073987713558e-05,
+      "loss": 0.1707,
+      "step": 142220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8347831935787633e-05,
+      "loss": 0.1746,
+      "step": 142230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8345589883861712e-05,
+      "loss": 0.1685,
+      "step": 142240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.834334783193579e-05,
+      "loss": 0.1744,
+      "step": 142250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8341105780009866e-05,
+      "loss": 0.1739,
+      "step": 142260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8338863728083945e-05,
+      "loss": 0.1717,
+      "step": 142270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.833662167615802e-05,
+      "loss": 0.1737,
+      "step": 142280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.83343796242321e-05,
+      "loss": 0.1743,
+      "step": 142290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8332137572306178e-05,
+      "loss": 0.1623,
+      "step": 142300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8329895520380253e-05,
+      "loss": 0.1676,
+      "step": 142310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.832765346845433e-05,
+      "loss": 0.1681,
+      "step": 142320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8325411416528407e-05,
+      "loss": 0.1672,
+      "step": 142330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8323169364602486e-05,
+      "loss": 0.1713,
+      "step": 142340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.832092731267656e-05,
+      "loss": 0.1638,
+      "step": 142350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.831868526075064e-05,
+      "loss": 0.1727,
+      "step": 142360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.831644320882472e-05,
+      "loss": 0.1734,
+      "step": 142370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8314201156898794e-05,
+      "loss": 0.17,
+      "step": 142380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8311959104972872e-05,
+      "loss": 0.1649,
+      "step": 142390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8309717053046948e-05,
+      "loss": 0.1672,
+      "step": 142400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8307475001121027e-05,
+      "loss": 0.1639,
+      "step": 142410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8305232949195102e-05,
+      "loss": 0.1722,
+      "step": 142420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.830299089726918e-05,
+      "loss": 0.1732,
+      "step": 142430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.830074884534326e-05,
+      "loss": 0.1736,
+      "step": 142440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8298506793417335e-05,
+      "loss": 0.1648,
+      "step": 142450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8296264741491413e-05,
+      "loss": 0.1684,
+      "step": 142460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8294022689565492e-05,
+      "loss": 0.1697,
+      "step": 142470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8291780637639567e-05,
+      "loss": 0.1709,
+      "step": 142480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8289538585713646e-05,
+      "loss": 0.168,
+      "step": 142490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8287296533787725e-05,
+      "loss": 0.1671,
+      "step": 142500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8285054481861804e-05,
+      "loss": 0.1691,
+      "step": 142510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.828281242993588e-05,
+      "loss": 0.174,
+      "step": 142520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8280570378009958e-05,
+      "loss": 0.1671,
+      "step": 142530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8278328326084033e-05,
+      "loss": 0.1666,
+      "step": 142540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.827608627415811e-05,
+      "loss": 0.1709,
+      "step": 142550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8273844222232187e-05,
+      "loss": 0.1754,
+      "step": 142560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8271602170306266e-05,
+      "loss": 0.1734,
+      "step": 142570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8269360118380344e-05,
+      "loss": 0.1799,
+      "step": 142580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.826711806645442e-05,
+      "loss": 0.1732,
+      "step": 142590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.82648760145285e-05,
+      "loss": 0.176,
+      "step": 142600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8262633962602574e-05,
+      "loss": 0.1751,
+      "step": 142610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8260391910676652e-05,
+      "loss": 0.1734,
+      "step": 142620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8258149858750728e-05,
+      "loss": 0.1736,
+      "step": 142630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8255907806824807e-05,
+      "loss": 0.1728,
+      "step": 142640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8253665754898885e-05,
+      "loss": 0.1704,
+      "step": 142650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.825142370297296e-05,
+      "loss": 0.1638,
+      "step": 142660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.824918165104704e-05,
+      "loss": 0.1748,
+      "step": 142670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8246939599121115e-05,
+      "loss": 0.1727,
+      "step": 142680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8244697547195193e-05,
+      "loss": 0.176,
+      "step": 142690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.824245549526927e-05,
+      "loss": 0.1705,
+      "step": 142700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8240213443343347e-05,
+      "loss": 0.1693,
+      "step": 142710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8237971391417426e-05,
+      "loss": 0.1702,
+      "step": 142720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.82357293394915e-05,
+      "loss": 0.1709,
+      "step": 142730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.823348728756558e-05,
+      "loss": 0.172,
+      "step": 142740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.823124523563966e-05,
+      "loss": 0.1718,
+      "step": 142750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8229003183713738e-05,
+      "loss": 0.1686,
+      "step": 142760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8226761131787813e-05,
+      "loss": 0.1702,
+      "step": 142770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.822451907986189e-05,
+      "loss": 0.1711,
+      "step": 142780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.822227702793597e-05,
+      "loss": 0.1698,
+      "step": 142790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8220034976010046e-05,
+      "loss": 0.1728,
+      "step": 142800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8217792924084124e-05,
+      "loss": 0.1704,
+      "step": 142810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.82155508721582e-05,
+      "loss": 0.1714,
+      "step": 142820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.821330882023228e-05,
+      "loss": 0.1707,
+      "step": 142830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8211066768306354e-05,
+      "loss": 0.1753,
+      "step": 142840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8208824716380432e-05,
+      "loss": 0.1701,
+      "step": 142850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.820658266445451e-05,
+      "loss": 0.1712,
+      "step": 142860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8204340612528587e-05,
+      "loss": 0.1683,
+      "step": 142870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8202098560602665e-05,
+      "loss": 0.1704,
+      "step": 142880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.819985650867674e-05,
+      "loss": 0.1715,
+      "step": 142890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.819761445675082e-05,
+      "loss": 0.1689,
+      "step": 142900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8195372404824895e-05,
+      "loss": 0.1663,
+      "step": 142910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8193130352898973e-05,
+      "loss": 0.1671,
+      "step": 142920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8190888300973052e-05,
+      "loss": 0.1641,
+      "step": 142930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8188646249047127e-05,
+      "loss": 0.1716,
+      "step": 142940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8186404197121206e-05,
+      "loss": 0.1669,
+      "step": 142950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.818416214519528e-05,
+      "loss": 0.1675,
+      "step": 142960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.818192009326936e-05,
+      "loss": 0.1718,
+      "step": 142970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.817967804134344e-05,
+      "loss": 0.1749,
+      "step": 142980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8177435989417514e-05,
+      "loss": 0.1677,
+      "step": 142990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8175193937491593e-05,
+      "loss": 0.1703,
+      "step": 143000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.817295188556567e-05,
+      "loss": 0.1705,
+      "step": 143010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8170709833639747e-05,
+      "loss": 0.1702,
+      "step": 143020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8168467781713826e-05,
+      "loss": 0.1735,
+      "step": 143030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8166225729787904e-05,
+      "loss": 0.1728,
+      "step": 143040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8163983677861983e-05,
+      "loss": 0.1716,
+      "step": 143050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.816174162593606e-05,
+      "loss": 0.175,
+      "step": 143060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8159499574010137e-05,
+      "loss": 0.17,
+      "step": 143070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8157257522084213e-05,
+      "loss": 0.1706,
+      "step": 143080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.815501547015829e-05,
+      "loss": 0.178,
+      "step": 143090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8152773418232367e-05,
+      "loss": 0.1771,
+      "step": 143100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8150531366306445e-05,
+      "loss": 0.1672,
+      "step": 143110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8148289314380524e-05,
+      "loss": 0.1727,
+      "step": 143120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.81460472624546e-05,
+      "loss": 0.1694,
+      "step": 143130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8143805210528678e-05,
+      "loss": 0.1743,
+      "step": 143140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8141563158602753e-05,
+      "loss": 0.1721,
+      "step": 143150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8139321106676832e-05,
+      "loss": 0.1671,
+      "step": 143160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8137079054750907e-05,
+      "loss": 0.1728,
+      "step": 143170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8134837002824986e-05,
+      "loss": 0.1685,
+      "step": 143180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8132594950899065e-05,
+      "loss": 0.177,
+      "step": 143190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.813035289897314e-05,
+      "loss": 0.1763,
+      "step": 143200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.812811084704722e-05,
+      "loss": 0.1736,
+      "step": 143210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8125868795121294e-05,
+      "loss": 0.1732,
+      "step": 143220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8123626743195373e-05,
+      "loss": 0.1706,
+      "step": 143230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8121384691269448e-05,
+      "loss": 0.1631,
+      "step": 143240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8119142639343527e-05,
+      "loss": 0.1725,
+      "step": 143250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8116900587417606e-05,
+      "loss": 0.1732,
+      "step": 143260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.811465853549168e-05,
+      "loss": 0.1727,
+      "step": 143270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.811241648356576e-05,
+      "loss": 0.1724,
+      "step": 143280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.811017443163984e-05,
+      "loss": 0.1756,
+      "step": 143290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8107932379713917e-05,
+      "loss": 0.1768,
+      "step": 143300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8105690327787993e-05,
+      "loss": 0.1665,
+      "step": 143310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.810344827586207e-05,
+      "loss": 0.1679,
+      "step": 143320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.810120622393615e-05,
+      "loss": 0.1718,
+      "step": 143330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8098964172010225e-05,
+      "loss": 0.1692,
+      "step": 143340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8096722120084304e-05,
+      "loss": 0.1675,
+      "step": 143350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.809448006815838e-05,
+      "loss": 0.1701,
+      "step": 143360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8092238016232458e-05,
+      "loss": 0.167,
+      "step": 143370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8089995964306533e-05,
+      "loss": 0.1692,
+      "step": 143380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8087753912380612e-05,
+      "loss": 0.1715,
+      "step": 143390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.808551186045469e-05,
+      "loss": 0.1703,
+      "step": 143400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8083269808528766e-05,
+      "loss": 0.1674,
+      "step": 143410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8081027756602845e-05,
+      "loss": 0.1705,
+      "step": 143420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.807878570467692e-05,
+      "loss": 0.172,
+      "step": 143430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8076543652751e-05,
+      "loss": 0.1715,
+      "step": 143440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8074301600825074e-05,
+      "loss": 0.1702,
+      "step": 143450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8072059548899153e-05,
+      "loss": 0.1666,
+      "step": 143460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.806981749697323e-05,
+      "loss": 0.169,
+      "step": 143470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8067575445047307e-05,
+      "loss": 0.1705,
+      "step": 143480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8065333393121386e-05,
+      "loss": 0.1725,
+      "step": 143490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.806309134119546e-05,
+      "loss": 0.1702,
+      "step": 143500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.806084928926954e-05,
+      "loss": 0.1691,
+      "step": 143510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8058607237343615e-05,
+      "loss": 0.1681,
+      "step": 143520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8056365185417694e-05,
+      "loss": 0.1676,
+      "step": 143530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8054123133491773e-05,
+      "loss": 0.1732,
+      "step": 143540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.805188108156585e-05,
+      "loss": 0.1675,
+      "step": 143550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8049639029639927e-05,
+      "loss": 0.1685,
+      "step": 143560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8047396977714005e-05,
+      "loss": 0.1687,
+      "step": 143570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8045154925788084e-05,
+      "loss": 0.1725,
+      "step": 143580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.804291287386216e-05,
+      "loss": 0.1699,
+      "step": 143590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8040670821936238e-05,
+      "loss": 0.173,
+      "step": 143600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8038428770010317e-05,
+      "loss": 0.1717,
+      "step": 143610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8036186718084392e-05,
+      "loss": 0.1647,
+      "step": 143620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.803394466615847e-05,
+      "loss": 0.1754,
+      "step": 143630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8031702614232546e-05,
+      "loss": 0.1712,
+      "step": 143640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8029684767499217e-05,
+      "loss": 0.1677,
+      "step": 143650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8027442715573293e-05,
+      "loss": 0.1715,
+      "step": 143660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.802520066364737e-05,
+      "loss": 0.1696,
+      "step": 143670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.802295861172145e-05,
+      "loss": 0.1717,
+      "step": 143680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8020716559795525e-05,
+      "loss": 0.1698,
+      "step": 143690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8018474507869604e-05,
+      "loss": 0.166,
+      "step": 143700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.801623245594368e-05,
+      "loss": 0.1635,
+      "step": 143710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8013990404017758e-05,
+      "loss": 0.1664,
+      "step": 143720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8011748352091834e-05,
+      "loss": 0.1709,
+      "step": 143730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8009506300165912e-05,
+      "loss": 0.1703,
+      "step": 143740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.800726424823999e-05,
+      "loss": 0.169,
+      "step": 143750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8005022196314066e-05,
+      "loss": 0.1756,
+      "step": 143760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8002780144388145e-05,
+      "loss": 0.1756,
+      "step": 143770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.800053809246222e-05,
+      "loss": 0.1723,
+      "step": 143780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.79982960405363e-05,
+      "loss": 0.1655,
+      "step": 143790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7996053988610374e-05,
+      "loss": 0.1726,
+      "step": 143800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7993811936684453e-05,
+      "loss": 0.1688,
+      "step": 143810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7991569884758532e-05,
+      "loss": 0.1679,
+      "step": 143820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7989327832832607e-05,
+      "loss": 0.1755,
+      "step": 143830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7987085780906686e-05,
+      "loss": 0.1687,
+      "step": 143840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7984843728980765e-05,
+      "loss": 0.1702,
+      "step": 143850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7982601677054843e-05,
+      "loss": 0.17,
+      "step": 143860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.798035962512892e-05,
+      "loss": 0.1715,
+      "step": 143870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7978117573202997e-05,
+      "loss": 0.1685,
+      "step": 143880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7975875521277076e-05,
+      "loss": 0.168,
+      "step": 143890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.797363346935115e-05,
+      "loss": 0.1645,
+      "step": 143900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.797139141742523e-05,
+      "loss": 0.1707,
+      "step": 143910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7969149365499305e-05,
+      "loss": 0.1752,
+      "step": 143920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7966907313573384e-05,
+      "loss": 0.1707,
+      "step": 143930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7964665261647463e-05,
+      "loss": 0.1689,
+      "step": 143940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7962423209721538e-05,
+      "loss": 0.169,
+      "step": 143950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7960181157795617e-05,
+      "loss": 0.1727,
+      "step": 143960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7957939105869692e-05,
+      "loss": 0.1697,
+      "step": 143970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.795569705394377e-05,
+      "loss": 0.1676,
+      "step": 143980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7953455002017846e-05,
+      "loss": 0.1655,
+      "step": 143990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7951212950091925e-05,
+      "loss": 0.1744,
+      "step": 144000
+    },
+    {
+      "epoch": 0.64,
+      "eval_NEIMS_canon_smiles": 0.9400666666666667,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.07141666666666667,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.4733336833562701,
+      "eval_NEIMS_exact_mols": 0.07115,
+      "eval_NEIMS_exact_smiles": 0.0692,
+      "eval_NEIMS_loss": 0.20888929069042206,
+      "eval_NEIMS_matched_formulas": 0.5735333333333333,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.37904262966664,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.07145,
+      "eval_NEIMS_runtime": 715.6344,
+      "eval_NEIMS_samples_per_second": 83.842,
+      "eval_NEIMS_steps_per_second": 1.311,
+      "step": 144000
+    },
+    {
+      "epoch": 0.64,
+      "eval_RASSP_canon_smiles": 0.9318629578922764,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.16687815140080808,
+      "eval_RASSP_daylight_tanimoto_simil": 0.6079040598169455,
+      "eval_RASSP_exact_mols": 0.16634387417771396,
+      "eval_RASSP_exact_smiles": 0.16091762113066418,
+      "eval_RASSP_loss": 0.14781031012535095,
+      "eval_RASSP_matched_formulas": 0.7601763114836211,
+      "eval_RASSP_morgan_tanimoto_simil": 0.4826798272174804,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.16657762046281765,
+      "eval_RASSP_runtime": 820.8546,
+      "eval_RASSP_samples_per_second": 72.965,
+      "eval_RASSP_steps_per_second": 1.14,
+      "step": 144000
+    },
+    {
+      "epoch": 0.64,
+      "eval_NIST_canon_smiles": 0.8558398693970259,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.011818149554601271,
+      "eval_NIST_daylight_tanimoto_simil": 0.26327297585527626,
+      "eval_NIST_exact_mols": 0.011569719984384427,
+      "eval_NIST_exact_smiles": 0.010576001703517053,
+      "eval_NIST_loss": 1.8349838256835938,
+      "eval_NIST_matched_formulas": 0.0833303758384498,
+      "eval_NIST_morgan_tanimoto_simil": 0.2154380983344991,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.011853639493203677,
+      "eval_NIST_runtime": 498.1444,
+      "eval_NIST_samples_per_second": 56.564,
+      "eval_NIST_steps_per_second": 0.885,
+      "step": 144000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7948970898166004e-05,
+      "loss": 0.169,
+      "step": 144010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.794672884624008e-05,
+      "loss": 0.1702,
+      "step": 144020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7944486794314158e-05,
+      "loss": 0.1756,
+      "step": 144030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7942244742388233e-05,
+      "loss": 0.1717,
+      "step": 144040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7940002690462312e-05,
+      "loss": 0.1715,
+      "step": 144050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7937760638536387e-05,
+      "loss": 0.1677,
+      "step": 144060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7935518586610466e-05,
+      "loss": 0.1749,
+      "step": 144070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7933276534684545e-05,
+      "loss": 0.1718,
+      "step": 144080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.793103448275862e-05,
+      "loss": 0.1639,
+      "step": 144090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.79287924308327e-05,
+      "loss": 0.1737,
+      "step": 144100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7926550378906777e-05,
+      "loss": 0.1792,
+      "step": 144110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7924308326980853e-05,
+      "loss": 0.1682,
+      "step": 144120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.792206627505493e-05,
+      "loss": 0.1662,
+      "step": 144130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.791982422312901e-05,
+      "loss": 0.1642,
+      "step": 144140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.791758217120309e-05,
+      "loss": 0.1725,
+      "step": 144150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7915340119277164e-05,
+      "loss": 0.1689,
+      "step": 144160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7913098067351243e-05,
+      "loss": 0.1678,
+      "step": 144170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7910856015425318e-05,
+      "loss": 0.1726,
+      "step": 144180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7908613963499397e-05,
+      "loss": 0.1661,
+      "step": 144190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7906371911573472e-05,
+      "loss": 0.1692,
+      "step": 144200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.790412985964755e-05,
+      "loss": 0.1688,
+      "step": 144210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.790188780772163e-05,
+      "loss": 0.1727,
+      "step": 144220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7899645755795705e-05,
+      "loss": 0.1711,
+      "step": 144230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7897403703869784e-05,
+      "loss": 0.1691,
+      "step": 144240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.789516165194386e-05,
+      "loss": 0.1651,
+      "step": 144250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7892919600017938e-05,
+      "loss": 0.1667,
+      "step": 144260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7890677548092013e-05,
+      "loss": 0.1719,
+      "step": 144270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7888435496166092e-05,
+      "loss": 0.1658,
+      "step": 144280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.788619344424017e-05,
+      "loss": 0.1656,
+      "step": 144290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7883951392314246e-05,
+      "loss": 0.165,
+      "step": 144300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7881709340388325e-05,
+      "loss": 0.1732,
+      "step": 144310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.78794672884624e-05,
+      "loss": 0.1701,
+      "step": 144320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.787722523653648e-05,
+      "loss": 0.1732,
+      "step": 144330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7874983184610554e-05,
+      "loss": 0.1668,
+      "step": 144340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7872741132684633e-05,
+      "loss": 0.1706,
+      "step": 144350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.787049908075871e-05,
+      "loss": 0.1668,
+      "step": 144360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7868257028832787e-05,
+      "loss": 0.1693,
+      "step": 144370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7866014976906865e-05,
+      "loss": 0.1708,
+      "step": 144380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7863772924980944e-05,
+      "loss": 0.1684,
+      "step": 144390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7861530873055023e-05,
+      "loss": 0.1642,
+      "step": 144400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7859288821129098e-05,
+      "loss": 0.169,
+      "step": 144410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7857046769203177e-05,
+      "loss": 0.1744,
+      "step": 144420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7854804717277256e-05,
+      "loss": 0.1713,
+      "step": 144430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785256266535133e-05,
+      "loss": 0.1714,
+      "step": 144440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785032061342541e-05,
+      "loss": 0.178,
+      "step": 144450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7848078561499485e-05,
+      "loss": 0.1731,
+      "step": 144460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7845836509573564e-05,
+      "loss": 0.1738,
+      "step": 144470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784359445764764e-05,
+      "loss": 0.1698,
+      "step": 144480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7841352405721718e-05,
+      "loss": 0.173,
+      "step": 144490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7839110353795797e-05,
+      "loss": 0.1692,
+      "step": 144500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7836868301869872e-05,
+      "loss": 0.1719,
+      "step": 144510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.783462624994395e-05,
+      "loss": 0.1663,
+      "step": 144520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7832384198018026e-05,
+      "loss": 0.166,
+      "step": 144530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7830142146092105e-05,
+      "loss": 0.169,
+      "step": 144540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.782790009416618e-05,
+      "loss": 0.1696,
+      "step": 144550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.782565804224026e-05,
+      "loss": 0.1716,
+      "step": 144560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7823415990314337e-05,
+      "loss": 0.1698,
+      "step": 144570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7821173938388413e-05,
+      "loss": 0.1688,
+      "step": 144580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.781893188646249e-05,
+      "loss": 0.1709,
+      "step": 144590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7816689834536567e-05,
+      "loss": 0.1669,
+      "step": 144600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7814447782610645e-05,
+      "loss": 0.1657,
+      "step": 144610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.781220573068472e-05,
+      "loss": 0.1714,
+      "step": 144620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.78099636787588e-05,
+      "loss": 0.1687,
+      "step": 144630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7807721626832878e-05,
+      "loss": 0.1686,
+      "step": 144640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7805479574906954e-05,
+      "loss": 0.1643,
+      "step": 144650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7803237522981032e-05,
+      "loss": 0.1697,
+      "step": 144660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.780099547105511e-05,
+      "loss": 0.1641,
+      "step": 144670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.779875341912919e-05,
+      "loss": 0.1747,
+      "step": 144680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7796511367203265e-05,
+      "loss": 0.1716,
+      "step": 144690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7794269315277344e-05,
+      "loss": 0.163,
+      "step": 144700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7792027263351423e-05,
+      "loss": 0.1704,
+      "step": 144710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7789785211425498e-05,
+      "loss": 0.1671,
+      "step": 144720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7787543159499577e-05,
+      "loss": 0.1704,
+      "step": 144730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7785301107573652e-05,
+      "loss": 0.1724,
+      "step": 144740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778305905564773e-05,
+      "loss": 0.1689,
+      "step": 144750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7780817003721806e-05,
+      "loss": 0.1654,
+      "step": 144760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7778574951795885e-05,
+      "loss": 0.1634,
+      "step": 144770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7776332899869963e-05,
+      "loss": 0.1644,
+      "step": 144780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777409084794404e-05,
+      "loss": 0.1666,
+      "step": 144790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7771848796018117e-05,
+      "loss": 0.1643,
+      "step": 144800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7769606744092193e-05,
+      "loss": 0.164,
+      "step": 144810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.776736469216627e-05,
+      "loss": 0.1717,
+      "step": 144820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7765122640240347e-05,
+      "loss": 0.1676,
+      "step": 144830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7762880588314426e-05,
+      "loss": 0.1662,
+      "step": 144840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7760638536388504e-05,
+      "loss": 0.1613,
+      "step": 144850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775839648446258e-05,
+      "loss": 0.1725,
+      "step": 144860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7756154432536658e-05,
+      "loss": 0.1708,
+      "step": 144870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7753912380610734e-05,
+      "loss": 0.1722,
+      "step": 144880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7751670328684812e-05,
+      "loss": 0.1704,
+      "step": 144890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7749428276758888e-05,
+      "loss": 0.1712,
+      "step": 144900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7747186224832966e-05,
+      "loss": 0.1739,
+      "step": 144910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7744944172907045e-05,
+      "loss": 0.1695,
+      "step": 144920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7742702120981124e-05,
+      "loss": 0.176,
+      "step": 144930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7740460069055203e-05,
+      "loss": 0.1674,
+      "step": 144940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7738218017129278e-05,
+      "loss": 0.1678,
+      "step": 144950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7735975965203357e-05,
+      "loss": 0.1725,
+      "step": 144960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7733733913277432e-05,
+      "loss": 0.1676,
+      "step": 144970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.773149186135151e-05,
+      "loss": 0.1676,
+      "step": 144980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.772924980942559e-05,
+      "loss": 0.172,
+      "step": 144990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7727007757499665e-05,
+      "loss": 0.169,
+      "step": 145000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7724765705573743e-05,
+      "loss": 0.1684,
+      "step": 145010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.772252365364782e-05,
+      "loss": 0.169,
+      "step": 145020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7720281601721897e-05,
+      "loss": 0.1769,
+      "step": 145030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7718039549795973e-05,
+      "loss": 0.1688,
+      "step": 145040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.771579749787005e-05,
+      "loss": 0.1669,
+      "step": 145050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.771355544594413e-05,
+      "loss": 0.1708,
+      "step": 145060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7711313394018206e-05,
+      "loss": 0.1642,
+      "step": 145070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7709071342092284e-05,
+      "loss": 0.1648,
+      "step": 145080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.770682929016636e-05,
+      "loss": 0.1731,
+      "step": 145090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7704587238240438e-05,
+      "loss": 0.1635,
+      "step": 145100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7702345186314514e-05,
+      "loss": 0.1701,
+      "step": 145110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7700103134388592e-05,
+      "loss": 0.1736,
+      "step": 145120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.769786108246267e-05,
+      "loss": 0.1686,
+      "step": 145130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7695619030536746e-05,
+      "loss": 0.1701,
+      "step": 145140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7693376978610825e-05,
+      "loss": 0.1687,
+      "step": 145150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.76911349266849e-05,
+      "loss": 0.1697,
+      "step": 145160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.768889287475898e-05,
+      "loss": 0.1723,
+      "step": 145170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7686650822833058e-05,
+      "loss": 0.1704,
+      "step": 145180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7684408770907133e-05,
+      "loss": 0.1666,
+      "step": 145190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7682166718981212e-05,
+      "loss": 0.1662,
+      "step": 145200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.767992466705529e-05,
+      "loss": 0.1644,
+      "step": 145210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.767768261512937e-05,
+      "loss": 0.1677,
+      "step": 145220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7675440563203445e-05,
+      "loss": 0.1642,
+      "step": 145230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7673198511277523e-05,
+      "loss": 0.175,
+      "step": 145240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.76709564593516e-05,
+      "loss": 0.1698,
+      "step": 145250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7668714407425677e-05,
+      "loss": 0.1648,
+      "step": 145260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7666472355499756e-05,
+      "loss": 0.1691,
+      "step": 145270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.766423030357383e-05,
+      "loss": 0.1728,
+      "step": 145280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.766198825164791e-05,
+      "loss": 0.1701,
+      "step": 145290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7659746199721986e-05,
+      "loss": 0.1701,
+      "step": 145300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7657504147796064e-05,
+      "loss": 0.1683,
+      "step": 145310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7655262095870143e-05,
+      "loss": 0.167,
+      "step": 145320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.765302004394422e-05,
+      "loss": 0.1639,
+      "step": 145330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7650777992018297e-05,
+      "loss": 0.1662,
+      "step": 145340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7648535940092372e-05,
+      "loss": 0.1628,
+      "step": 145350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.764629388816645e-05,
+      "loss": 0.1688,
+      "step": 145360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7644051836240526e-05,
+      "loss": 0.1701,
+      "step": 145370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7641809784314605e-05,
+      "loss": 0.1623,
+      "step": 145380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7639567732388684e-05,
+      "loss": 0.1704,
+      "step": 145390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.763732568046276e-05,
+      "loss": 0.1667,
+      "step": 145400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7635083628536838e-05,
+      "loss": 0.1658,
+      "step": 145410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7632841576610913e-05,
+      "loss": 0.1652,
+      "step": 145420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7630599524684992e-05,
+      "loss": 0.1652,
+      "step": 145430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7628357472759067e-05,
+      "loss": 0.1651,
+      "step": 145440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7626115420833146e-05,
+      "loss": 0.167,
+      "step": 145450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7623873368907225e-05,
+      "loss": 0.1663,
+      "step": 145460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7621631316981303e-05,
+      "loss": 0.1722,
+      "step": 145470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7619389265055382e-05,
+      "loss": 0.167,
+      "step": 145480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7617147213129457e-05,
+      "loss": 0.169,
+      "step": 145490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7614905161203536e-05,
+      "loss": 0.1663,
+      "step": 145500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.761266310927761e-05,
+      "loss": 0.1772,
+      "step": 145510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.761042105735169e-05,
+      "loss": 0.1633,
+      "step": 145520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.760817900542577e-05,
+      "loss": 0.168,
+      "step": 145530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7605936953499844e-05,
+      "loss": 0.1654,
+      "step": 145540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7603694901573923e-05,
+      "loss": 0.17,
+      "step": 145550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7601452849648e-05,
+      "loss": 0.173,
+      "step": 145560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7599210797722077e-05,
+      "loss": 0.1704,
+      "step": 145570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7596968745796152e-05,
+      "loss": 0.169,
+      "step": 145580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.759472669387023e-05,
+      "loss": 0.1601,
+      "step": 145590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.759248464194431e-05,
+      "loss": 0.1633,
+      "step": 145600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7590242590018385e-05,
+      "loss": 0.1675,
+      "step": 145610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7588000538092464e-05,
+      "loss": 0.1675,
+      "step": 145620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.758575848616654e-05,
+      "loss": 0.1701,
+      "step": 145630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7583516434240618e-05,
+      "loss": 0.1705,
+      "step": 145640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7581274382314693e-05,
+      "loss": 0.1705,
+      "step": 145650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7579032330388772e-05,
+      "loss": 0.1726,
+      "step": 145660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.757679027846285e-05,
+      "loss": 0.1644,
+      "step": 145670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7574548226536926e-05,
+      "loss": 0.1666,
+      "step": 145680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7572306174611005e-05,
+      "loss": 0.1678,
+      "step": 145690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.757006412268508e-05,
+      "loss": 0.1657,
+      "step": 145700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.756782207075916e-05,
+      "loss": 0.1691,
+      "step": 145710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7565580018833237e-05,
+      "loss": 0.1691,
+      "step": 145720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7563337966907313e-05,
+      "loss": 0.1708,
+      "step": 145730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.756109591498139e-05,
+      "loss": 0.1739,
+      "step": 145740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.755885386305547e-05,
+      "loss": 0.1688,
+      "step": 145750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.755661181112955e-05,
+      "loss": 0.1678,
+      "step": 145760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7554369759203624e-05,
+      "loss": 0.1636,
+      "step": 145770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7552127707277703e-05,
+      "loss": 0.1693,
+      "step": 145780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.755010986054437e-05,
+      "loss": 0.1661,
+      "step": 145790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.754786780861845e-05,
+      "loss": 0.1699,
+      "step": 145800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7545625756692528e-05,
+      "loss": 0.1668,
+      "step": 145810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7543383704766604e-05,
+      "loss": 0.1702,
+      "step": 145820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7541141652840682e-05,
+      "loss": 0.1688,
+      "step": 145830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7538899600914758e-05,
+      "loss": 0.1724,
+      "step": 145840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7536657548988836e-05,
+      "loss": 0.1666,
+      "step": 145850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.753441549706291e-05,
+      "loss": 0.169,
+      "step": 145860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.753217344513699e-05,
+      "loss": 0.1694,
+      "step": 145870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.752993139321107e-05,
+      "loss": 0.1723,
+      "step": 145880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7527689341285144e-05,
+      "loss": 0.1732,
+      "step": 145890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7525447289359223e-05,
+      "loss": 0.1704,
+      "step": 145900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.75232052374333e-05,
+      "loss": 0.1661,
+      "step": 145910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7520963185507377e-05,
+      "loss": 0.1674,
+      "step": 145920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7518721133581453e-05,
+      "loss": 0.1708,
+      "step": 145930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.751647908165553e-05,
+      "loss": 0.1677,
+      "step": 145940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.751423702972961e-05,
+      "loss": 0.1716,
+      "step": 145950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7511994977803685e-05,
+      "loss": 0.1675,
+      "step": 145960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7509752925877764e-05,
+      "loss": 0.166,
+      "step": 145970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.750751087395184e-05,
+      "loss": 0.1677,
+      "step": 145980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7505268822025918e-05,
+      "loss": 0.1745,
+      "step": 145990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7503026770099993e-05,
+      "loss": 0.1661,
+      "step": 146000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7500784718174072e-05,
+      "loss": 0.1624,
+      "step": 146010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.749854266624815e-05,
+      "loss": 0.174,
+      "step": 146020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.749630061432223e-05,
+      "loss": 0.17,
+      "step": 146030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7494058562396308e-05,
+      "loss": 0.1737,
+      "step": 146040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7491816510470384e-05,
+      "loss": 0.1626,
+      "step": 146050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7489574458544462e-05,
+      "loss": 0.164,
+      "step": 146060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7487332406618538e-05,
+      "loss": 0.1653,
+      "step": 146070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7485090354692616e-05,
+      "loss": 0.1699,
+      "step": 146080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7482848302766695e-05,
+      "loss": 0.1675,
+      "step": 146090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.748060625084077e-05,
+      "loss": 0.1743,
+      "step": 146100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.747836419891485e-05,
+      "loss": 0.1734,
+      "step": 146110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7476122146988924e-05,
+      "loss": 0.173,
+      "step": 146120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7473880095063003e-05,
+      "loss": 0.1658,
+      "step": 146130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.747163804313708e-05,
+      "loss": 0.1665,
+      "step": 146140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7469395991211157e-05,
+      "loss": 0.1695,
+      "step": 146150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7467153939285236e-05,
+      "loss": 0.169,
+      "step": 146160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.746491188735931e-05,
+      "loss": 0.1686,
+      "step": 146170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.746266983543339e-05,
+      "loss": 0.161,
+      "step": 146180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7460427783507465e-05,
+      "loss": 0.1657,
+      "step": 146190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7458185731581544e-05,
+      "loss": 0.1715,
+      "step": 146200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.745594367965562e-05,
+      "loss": 0.17,
+      "step": 146210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7453701627729698e-05,
+      "loss": 0.1671,
+      "step": 146220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7451459575803777e-05,
+      "loss": 0.1661,
+      "step": 146230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7449217523877852e-05,
+      "loss": 0.1673,
+      "step": 146240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.744697547195193e-05,
+      "loss": 0.1683,
+      "step": 146250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7444733420026006e-05,
+      "loss": 0.1706,
+      "step": 146260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7442491368100085e-05,
+      "loss": 0.1624,
+      "step": 146270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7440249316174164e-05,
+      "loss": 0.1683,
+      "step": 146280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.743800726424824e-05,
+      "loss": 0.1683,
+      "step": 146290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7435765212322318e-05,
+      "loss": 0.1653,
+      "step": 146300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7433523160396396e-05,
+      "loss": 0.1664,
+      "step": 146310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7431281108470475e-05,
+      "loss": 0.1714,
+      "step": 146320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.742903905654455e-05,
+      "loss": 0.1715,
+      "step": 146330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.742679700461863e-05,
+      "loss": 0.1643,
+      "step": 146340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7424554952692708e-05,
+      "loss": 0.1716,
+      "step": 146350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7422312900766783e-05,
+      "loss": 0.1741,
+      "step": 146360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7420070848840862e-05,
+      "loss": 0.1697,
+      "step": 146370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7417828796914937e-05,
+      "loss": 0.1662,
+      "step": 146380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7415586744989016e-05,
+      "loss": 0.1702,
+      "step": 146390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.741334469306309e-05,
+      "loss": 0.174,
+      "step": 146400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.741110264113717e-05,
+      "loss": 0.1626,
+      "step": 146410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.740886058921125e-05,
+      "loss": 0.1683,
+      "step": 146420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7406618537285324e-05,
+      "loss": 0.1658,
+      "step": 146430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7404376485359403e-05,
+      "loss": 0.1713,
+      "step": 146440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7402134433433478e-05,
+      "loss": 0.1708,
+      "step": 146450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7399892381507557e-05,
+      "loss": 0.1685,
+      "step": 146460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7397650329581632e-05,
+      "loss": 0.1685,
+      "step": 146470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.739540827765571e-05,
+      "loss": 0.1715,
+      "step": 146480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.739316622572979e-05,
+      "loss": 0.17,
+      "step": 146490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7390924173803865e-05,
+      "loss": 0.165,
+      "step": 146500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7388682121877944e-05,
+      "loss": 0.1683,
+      "step": 146510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.738644006995202e-05,
+      "loss": 0.1698,
+      "step": 146520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7384198018026098e-05,
+      "loss": 0.1715,
+      "step": 146530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7381955966100173e-05,
+      "loss": 0.1735,
+      "step": 146540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7379713914174252e-05,
+      "loss": 0.1694,
+      "step": 146550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.737747186224833e-05,
+      "loss": 0.1657,
+      "step": 146560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.737522981032241e-05,
+      "loss": 0.1671,
+      "step": 146570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7372987758396488e-05,
+      "loss": 0.1683,
+      "step": 146580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7370745706470563e-05,
+      "loss": 0.1633,
+      "step": 146590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7368503654544642e-05,
+      "loss": 0.1721,
+      "step": 146600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7366261602618717e-05,
+      "loss": 0.1729,
+      "step": 146610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7364019550692796e-05,
+      "loss": 0.1695,
+      "step": 146620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7361777498766875e-05,
+      "loss": 0.1703,
+      "step": 146630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.735953544684095e-05,
+      "loss": 0.1693,
+      "step": 146640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.735729339491503e-05,
+      "loss": 0.1701,
+      "step": 146650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7355051342989104e-05,
+      "loss": 0.1771,
+      "step": 146660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7352809291063183e-05,
+      "loss": 0.1698,
+      "step": 146670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7350567239137258e-05,
+      "loss": 0.1685,
+      "step": 146680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7348325187211337e-05,
+      "loss": 0.1682,
+      "step": 146690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7346083135285416e-05,
+      "loss": 0.1767,
+      "step": 146700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.734384108335949e-05,
+      "loss": 0.1743,
+      "step": 146710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.734159903143357e-05,
+      "loss": 0.1622,
+      "step": 146720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7339356979507645e-05,
+      "loss": 0.1678,
+      "step": 146730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7337114927581724e-05,
+      "loss": 0.1688,
+      "step": 146740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.73348728756558e-05,
+      "loss": 0.1708,
+      "step": 146750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7332630823729878e-05,
+      "loss": 0.1658,
+      "step": 146760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7330388771803956e-05,
+      "loss": 0.1678,
+      "step": 146770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7328146719878032e-05,
+      "loss": 0.1659,
+      "step": 146780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.732590466795211e-05,
+      "loss": 0.1638,
+      "step": 146790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7323662616026186e-05,
+      "loss": 0.1706,
+      "step": 146800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7321420564100264e-05,
+      "loss": 0.1662,
+      "step": 146810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7319178512174343e-05,
+      "loss": 0.168,
+      "step": 146820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.731693646024842e-05,
+      "loss": 0.1744,
+      "step": 146830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7314694408322497e-05,
+      "loss": 0.1761,
+      "step": 146840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7312452356396576e-05,
+      "loss": 0.1705,
+      "step": 146850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7310210304470655e-05,
+      "loss": 0.1622,
+      "step": 146860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.730796825254473e-05,
+      "loss": 0.1642,
+      "step": 146870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.730572620061881e-05,
+      "loss": 0.1729,
+      "step": 146880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7303484148692884e-05,
+      "loss": 0.1681,
+      "step": 146890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7301242096766963e-05,
+      "loss": 0.1628,
+      "step": 146900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.729900004484104e-05,
+      "loss": 0.1636,
+      "step": 146910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7296757992915117e-05,
+      "loss": 0.1646,
+      "step": 146920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7294515940989196e-05,
+      "loss": 0.1722,
+      "step": 146930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.729227388906327e-05,
+      "loss": 0.1717,
+      "step": 146940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.729003183713735e-05,
+      "loss": 0.1674,
+      "step": 146950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7287789785211425e-05,
+      "loss": 0.168,
+      "step": 146960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7285547733285504e-05,
+      "loss": 0.1628,
+      "step": 146970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7283305681359582e-05,
+      "loss": 0.1673,
+      "step": 146980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7281063629433658e-05,
+      "loss": 0.1656,
+      "step": 146990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7278821577507736e-05,
+      "loss": 0.1695,
+      "step": 147000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7276579525581812e-05,
+      "loss": 0.1659,
+      "step": 147010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.727433747365589e-05,
+      "loss": 0.1654,
+      "step": 147020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7272095421729966e-05,
+      "loss": 0.1698,
+      "step": 147030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7269853369804045e-05,
+      "loss": 0.1666,
+      "step": 147040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7267611317878123e-05,
+      "loss": 0.1711,
+      "step": 147050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.72653692659522e-05,
+      "loss": 0.1701,
+      "step": 147060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7263127214026277e-05,
+      "loss": 0.1683,
+      "step": 147070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7260885162100353e-05,
+      "loss": 0.171,
+      "step": 147080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.725864311017443e-05,
+      "loss": 0.1703,
+      "step": 147090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.725640105824851e-05,
+      "loss": 0.1685,
+      "step": 147100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.725415900632259e-05,
+      "loss": 0.1731,
+      "step": 147110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7251916954396667e-05,
+      "loss": 0.1705,
+      "step": 147120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7249674902470743e-05,
+      "loss": 0.1726,
+      "step": 147130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.724743285054482e-05,
+      "loss": 0.1671,
+      "step": 147140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7245190798618897e-05,
+      "loss": 0.1701,
+      "step": 147150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7242948746692976e-05,
+      "loss": 0.1626,
+      "step": 147160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.724070669476705e-05,
+      "loss": 0.1721,
+      "step": 147170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.723846464284113e-05,
+      "loss": 0.168,
+      "step": 147180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.723622259091521e-05,
+      "loss": 0.1761,
+      "step": 147190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7233980538989284e-05,
+      "loss": 0.1667,
+      "step": 147200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7231738487063362e-05,
+      "loss": 0.1677,
+      "step": 147210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7229496435137438e-05,
+      "loss": 0.1687,
+      "step": 147220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7227254383211516e-05,
+      "loss": 0.168,
+      "step": 147230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7225012331285592e-05,
+      "loss": 0.1664,
+      "step": 147240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.722277027935967e-05,
+      "loss": 0.1652,
+      "step": 147250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.722052822743375e-05,
+      "loss": 0.1653,
+      "step": 147260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7218286175507825e-05,
+      "loss": 0.172,
+      "step": 147270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7216044123581903e-05,
+      "loss": 0.1679,
+      "step": 147280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.721380207165598e-05,
+      "loss": 0.1685,
+      "step": 147290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7211560019730057e-05,
+      "loss": 0.1775,
+      "step": 147300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7209317967804133e-05,
+      "loss": 0.1711,
+      "step": 147310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.720707591587821e-05,
+      "loss": 0.1667,
+      "step": 147320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.720483386395229e-05,
+      "loss": 0.1703,
+      "step": 147330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7202591812026365e-05,
+      "loss": 0.1678,
+      "step": 147340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7200349760100444e-05,
+      "loss": 0.1693,
+      "step": 147350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7198107708174523e-05,
+      "loss": 0.1694,
+      "step": 147360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7195865656248598e-05,
+      "loss": 0.1732,
+      "step": 147370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7193623604322677e-05,
+      "loss": 0.1662,
+      "step": 147380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7191381552396756e-05,
+      "loss": 0.1672,
+      "step": 147390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7189139500470834e-05,
+      "loss": 0.1676,
+      "step": 147400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.718689744854491e-05,
+      "loss": 0.17,
+      "step": 147410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.718465539661899e-05,
+      "loss": 0.1662,
+      "step": 147420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7182413344693064e-05,
+      "loss": 0.1636,
+      "step": 147430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7180171292767142e-05,
+      "loss": 0.1606,
+      "step": 147440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7177929240841218e-05,
+      "loss": 0.1677,
+      "step": 147450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7175687188915296e-05,
+      "loss": 0.1627,
+      "step": 147460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7173445136989375e-05,
+      "loss": 0.1702,
+      "step": 147470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.717120308506345e-05,
+      "loss": 0.1688,
+      "step": 147480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.716896103313753e-05,
+      "loss": 0.1714,
+      "step": 147490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7166718981211605e-05,
+      "loss": 0.1757,
+      "step": 147500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7164476929285683e-05,
+      "loss": 0.1685,
+      "step": 147510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.716223487735976e-05,
+      "loss": 0.1669,
+      "step": 147520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7159992825433837e-05,
+      "loss": 0.1707,
+      "step": 147530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7157750773507916e-05,
+      "loss": 0.1655,
+      "step": 147540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.715550872158199e-05,
+      "loss": 0.1629,
+      "step": 147550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.715326666965607e-05,
+      "loss": 0.1706,
+      "step": 147560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7151024617730145e-05,
+      "loss": 0.1666,
+      "step": 147570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7148782565804224e-05,
+      "loss": 0.1688,
+      "step": 147580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7146540513878303e-05,
+      "loss": 0.1684,
+      "step": 147590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7144298461952378e-05,
+      "loss": 0.1662,
+      "step": 147600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7142056410026457e-05,
+      "loss": 0.1614,
+      "step": 147610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7139814358100532e-05,
+      "loss": 0.1728,
+      "step": 147620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.713757230617461e-05,
+      "loss": 0.1699,
+      "step": 147630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.713533025424869e-05,
+      "loss": 0.1723,
+      "step": 147640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.713308820232277e-05,
+      "loss": 0.1607,
+      "step": 147650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7130846150396844e-05,
+      "loss": 0.1631,
+      "step": 147660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7128604098470922e-05,
+      "loss": 0.1714,
+      "step": 147670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7126362046545e-05,
+      "loss": 0.17,
+      "step": 147680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7124119994619076e-05,
+      "loss": 0.1659,
+      "step": 147690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7121877942693155e-05,
+      "loss": 0.1699,
+      "step": 147700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.711963589076723e-05,
+      "loss": 0.1694,
+      "step": 147710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.711739383884131e-05,
+      "loss": 0.1653,
+      "step": 147720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7115151786915388e-05,
+      "loss": 0.1732,
+      "step": 147730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7112909734989463e-05,
+      "loss": 0.1689,
+      "step": 147740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7110667683063542e-05,
+      "loss": 0.1666,
+      "step": 147750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7108425631137617e-05,
+      "loss": 0.1677,
+      "step": 147760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7106183579211696e-05,
+      "loss": 0.1679,
+      "step": 147770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.710394152728577e-05,
+      "loss": 0.1622,
+      "step": 147780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.710169947535985e-05,
+      "loss": 0.167,
+      "step": 147790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.709945742343393e-05,
+      "loss": 0.171,
+      "step": 147800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7097215371508004e-05,
+      "loss": 0.1626,
+      "step": 147810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7094973319582083e-05,
+      "loss": 0.1666,
+      "step": 147820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7092731267656158e-05,
+      "loss": 0.1631,
+      "step": 147830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7090489215730237e-05,
+      "loss": 0.1624,
+      "step": 147840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7088247163804312e-05,
+      "loss": 0.1663,
+      "step": 147850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.708600511187839e-05,
+      "loss": 0.1688,
+      "step": 147860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.708376305995247e-05,
+      "loss": 0.1671,
+      "step": 147870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7081745213219137e-05,
+      "loss": 0.1672,
+      "step": 147880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7079503161293216e-05,
+      "loss": 0.1739,
+      "step": 147890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.707726110936729e-05,
+      "loss": 0.167,
+      "step": 147900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.707501905744137e-05,
+      "loss": 0.1722,
+      "step": 147910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.707277700551545e-05,
+      "loss": 0.168,
+      "step": 147920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7070534953589524e-05,
+      "loss": 0.1694,
+      "step": 147930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7068292901663603e-05,
+      "loss": 0.1688,
+      "step": 147940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7066050849737682e-05,
+      "loss": 0.1647,
+      "step": 147950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.706380879781176e-05,
+      "loss": 0.1759,
+      "step": 147960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7061566745885836e-05,
+      "loss": 0.1668,
+      "step": 147970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7059324693959914e-05,
+      "loss": 0.1664,
+      "step": 147980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.705708264203399e-05,
+      "loss": 0.1606,
+      "step": 147990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.705484059010807e-05,
+      "loss": 0.1691,
+      "step": 148000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7052598538182147e-05,
+      "loss": 0.1704,
+      "step": 148010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7050356486256223e-05,
+      "loss": 0.1637,
+      "step": 148020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.70481144343303e-05,
+      "loss": 0.1726,
+      "step": 148030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7045872382404377e-05,
+      "loss": 0.1702,
+      "step": 148040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7043630330478455e-05,
+      "loss": 0.1712,
+      "step": 148050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.704138827855253e-05,
+      "loss": 0.163,
+      "step": 148060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.703914622662661e-05,
+      "loss": 0.1691,
+      "step": 148070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7036904174700688e-05,
+      "loss": 0.1702,
+      "step": 148080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7034662122774763e-05,
+      "loss": 0.1687,
+      "step": 148090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7032420070848842e-05,
+      "loss": 0.1717,
+      "step": 148100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7030178018922917e-05,
+      "loss": 0.163,
+      "step": 148110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7027935966996996e-05,
+      "loss": 0.1681,
+      "step": 148120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.702569391507107e-05,
+      "loss": 0.1657,
+      "step": 148130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.702345186314515e-05,
+      "loss": 0.1666,
+      "step": 148140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.702120981121923e-05,
+      "loss": 0.1646,
+      "step": 148150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7018967759293304e-05,
+      "loss": 0.1638,
+      "step": 148160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7016725707367383e-05,
+      "loss": 0.1753,
+      "step": 148170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.701448365544146e-05,
+      "loss": 0.1622,
+      "step": 148180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7012241603515537e-05,
+      "loss": 0.1664,
+      "step": 148190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7009999551589616e-05,
+      "loss": 0.1719,
+      "step": 148200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7007757499663694e-05,
+      "loss": 0.1754,
+      "step": 148210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.700551544773777e-05,
+      "loss": 0.1664,
+      "step": 148220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.700327339581185e-05,
+      "loss": 0.1629,
+      "step": 148230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7001031343885927e-05,
+      "loss": 0.1725,
+      "step": 148240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6998789291960003e-05,
+      "loss": 0.1688,
+      "step": 148250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.699654724003408e-05,
+      "loss": 0.174,
+      "step": 148260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6994305188108157e-05,
+      "loss": 0.1648,
+      "step": 148270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6992063136182235e-05,
+      "loss": 0.1694,
+      "step": 148280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6989821084256314e-05,
+      "loss": 0.1662,
+      "step": 148290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.698757903233039e-05,
+      "loss": 0.1663,
+      "step": 148300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6985336980404468e-05,
+      "loss": 0.1686,
+      "step": 148310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6983094928478543e-05,
+      "loss": 0.173,
+      "step": 148320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6980852876552622e-05,
+      "loss": 0.1722,
+      "step": 148330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6978610824626697e-05,
+      "loss": 0.1709,
+      "step": 148340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6976368772700776e-05,
+      "loss": 0.1636,
+      "step": 148350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6974126720774855e-05,
+      "loss": 0.1724,
+      "step": 148360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.697188466884893e-05,
+      "loss": 0.1655,
+      "step": 148370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.696964261692301e-05,
+      "loss": 0.1658,
+      "step": 148380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6967400564997084e-05,
+      "loss": 0.166,
+      "step": 148390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6965158513071163e-05,
+      "loss": 0.1723,
+      "step": 148400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.696291646114524e-05,
+      "loss": 0.1672,
+      "step": 148410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6960674409219317e-05,
+      "loss": 0.1673,
+      "step": 148420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6958432357293396e-05,
+      "loss": 0.1679,
+      "step": 148430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.695619030536747e-05,
+      "loss": 0.1694,
+      "step": 148440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.695394825344155e-05,
+      "loss": 0.1679,
+      "step": 148450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.695170620151563e-05,
+      "loss": 0.1622,
+      "step": 148460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6949464149589704e-05,
+      "loss": 0.1663,
+      "step": 148470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6947222097663783e-05,
+      "loss": 0.1647,
+      "step": 148480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.694498004573786e-05,
+      "loss": 0.1673,
+      "step": 148490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.694273799381194e-05,
+      "loss": 0.1703,
+      "step": 148500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6940495941886015e-05,
+      "loss": 0.1716,
+      "step": 148510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6938253889960094e-05,
+      "loss": 0.1666,
+      "step": 148520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.693601183803417e-05,
+      "loss": 0.1674,
+      "step": 148530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6933769786108248e-05,
+      "loss": 0.1676,
+      "step": 148540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6931527734182327e-05,
+      "loss": 0.1716,
+      "step": 148550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6929285682256402e-05,
+      "loss": 0.1639,
+      "step": 148560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.692704363033048e-05,
+      "loss": 0.1671,
+      "step": 148570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6924801578404556e-05,
+      "loss": 0.1723,
+      "step": 148580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6922559526478635e-05,
+      "loss": 0.1726,
+      "step": 148590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.692031747455271e-05,
+      "loss": 0.1635,
+      "step": 148600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.691807542262679e-05,
+      "loss": 0.1665,
+      "step": 148610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6915833370700868e-05,
+      "loss": 0.1637,
+      "step": 148620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6913591318774943e-05,
+      "loss": 0.1687,
+      "step": 148630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6911349266849022e-05,
+      "loss": 0.1642,
+      "step": 148640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6909107214923097e-05,
+      "loss": 0.1683,
+      "step": 148650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6906865162997176e-05,
+      "loss": 0.1648,
+      "step": 148660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.690462311107125e-05,
+      "loss": 0.1612,
+      "step": 148670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.690238105914533e-05,
+      "loss": 0.1664,
+      "step": 148680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.690013900721941e-05,
+      "loss": 0.1728,
+      "step": 148690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6897896955293484e-05,
+      "loss": 0.1657,
+      "step": 148700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6895654903367563e-05,
+      "loss": 0.169,
+      "step": 148710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6893412851441638e-05,
+      "loss": 0.1642,
+      "step": 148720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6891170799515717e-05,
+      "loss": 0.1727,
+      "step": 148730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6888928747589795e-05,
+      "loss": 0.1711,
+      "step": 148740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6886686695663874e-05,
+      "loss": 0.1742,
+      "step": 148750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.688444464373795e-05,
+      "loss": 0.1718,
+      "step": 148760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6882202591812028e-05,
+      "loss": 0.1713,
+      "step": 148770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6879960539886107e-05,
+      "loss": 0.168,
+      "step": 148780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6877718487960182e-05,
+      "loss": 0.1714,
+      "step": 148790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.687547643603426e-05,
+      "loss": 0.1701,
+      "step": 148800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6873234384108336e-05,
+      "loss": 0.1664,
+      "step": 148810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6870992332182415e-05,
+      "loss": 0.1738,
+      "step": 148820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6868750280256494e-05,
+      "loss": 0.1662,
+      "step": 148830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.686650822833057e-05,
+      "loss": 0.1605,
+      "step": 148840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6864266176404648e-05,
+      "loss": 0.1637,
+      "step": 148850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6862024124478723e-05,
+      "loss": 0.1637,
+      "step": 148860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6859782072552802e-05,
+      "loss": 0.1706,
+      "step": 148870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6857540020626877e-05,
+      "loss": 0.1694,
+      "step": 148880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6855297968700956e-05,
+      "loss": 0.1651,
+      "step": 148890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6853055916775035e-05,
+      "loss": 0.1633,
+      "step": 148900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.685081386484911e-05,
+      "loss": 0.1656,
+      "step": 148910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.684857181292319e-05,
+      "loss": 0.163,
+      "step": 148920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6846329760997264e-05,
+      "loss": 0.166,
+      "step": 148930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6844087709071343e-05,
+      "loss": 0.1653,
+      "step": 148940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6841845657145418e-05,
+      "loss": 0.1702,
+      "step": 148950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.6839603605219497e-05,
+      "loss": 0.1682,
+      "step": 148960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6837361553293575e-05,
+      "loss": 0.1683,
+      "step": 148970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.683511950136765e-05,
+      "loss": 0.1657,
+      "step": 148980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.683287744944173e-05,
+      "loss": 0.1619,
+      "step": 148990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6830635397515808e-05,
+      "loss": 0.1673,
+      "step": 149000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6828393345589883e-05,
+      "loss": 0.1691,
+      "step": 149010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6826151293663962e-05,
+      "loss": 0.165,
+      "step": 149020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.682390924173804e-05,
+      "loss": 0.165,
+      "step": 149030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.682166718981212e-05,
+      "loss": 0.168,
+      "step": 149040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6819425137886195e-05,
+      "loss": 0.1686,
+      "step": 149050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6817183085960274e-05,
+      "loss": 0.1673,
+      "step": 149060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.681494103403435e-05,
+      "loss": 0.1585,
+      "step": 149070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6812698982108428e-05,
+      "loss": 0.1692,
+      "step": 149080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6810456930182503e-05,
+      "loss": 0.1669,
+      "step": 149090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6808214878256582e-05,
+      "loss": 0.1671,
+      "step": 149100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.680597282633066e-05,
+      "loss": 0.1672,
+      "step": 149110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6803730774404736e-05,
+      "loss": 0.1673,
+      "step": 149120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6801488722478815e-05,
+      "loss": 0.1674,
+      "step": 149130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.679924667055289e-05,
+      "loss": 0.1684,
+      "step": 149140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.679700461862697e-05,
+      "loss": 0.1633,
+      "step": 149150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6794762566701044e-05,
+      "loss": 0.1692,
+      "step": 149160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6792520514775123e-05,
+      "loss": 0.1703,
+      "step": 149170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.67902784628492e-05,
+      "loss": 0.1637,
+      "step": 149180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6788036410923277e-05,
+      "loss": 0.1678,
+      "step": 149190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6785794358997355e-05,
+      "loss": 0.1694,
+      "step": 149200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.678355230707143e-05,
+      "loss": 0.1649,
+      "step": 149210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.678131025514551e-05,
+      "loss": 0.1687,
+      "step": 149220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6779068203219585e-05,
+      "loss": 0.1705,
+      "step": 149230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6776826151293663e-05,
+      "loss": 0.1659,
+      "step": 149240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6774584099367742e-05,
+      "loss": 0.1677,
+      "step": 149250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6772342047441818e-05,
+      "loss": 0.1654,
+      "step": 149260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6770099995515896e-05,
+      "loss": 0.1711,
+      "step": 149270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6767857943589975e-05,
+      "loss": 0.1696,
+      "step": 149280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6765615891664054e-05,
+      "loss": 0.1666,
+      "step": 149290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.676337383973813e-05,
+      "loss": 0.1657,
+      "step": 149300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6761131787812208e-05,
+      "loss": 0.1743,
+      "step": 149310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6758889735886286e-05,
+      "loss": 0.167,
+      "step": 149320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6756647683960362e-05,
+      "loss": 0.1677,
+      "step": 149330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.675440563203444e-05,
+      "loss": 0.1674,
+      "step": 149340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6752163580108516e-05,
+      "loss": 0.1641,
+      "step": 149350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6749921528182595e-05,
+      "loss": 0.167,
+      "step": 149360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.674767947625667e-05,
+      "loss": 0.1719,
+      "step": 149370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.674543742433075e-05,
+      "loss": 0.1617,
+      "step": 149380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6743195372404827e-05,
+      "loss": 0.1647,
+      "step": 149390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6740953320478903e-05,
+      "loss": 0.1676,
+      "step": 149400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.673871126855298e-05,
+      "loss": 0.1681,
+      "step": 149410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6736469216627057e-05,
+      "loss": 0.1699,
+      "step": 149420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6734227164701135e-05,
+      "loss": 0.1642,
+      "step": 149430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.673198511277521e-05,
+      "loss": 0.1638,
+      "step": 149440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.672974306084929e-05,
+      "loss": 0.1688,
+      "step": 149450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6727501008923368e-05,
+      "loss": 0.1717,
+      "step": 149460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6725258956997444e-05,
+      "loss": 0.1708,
+      "step": 149470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6723016905071522e-05,
+      "loss": 0.169,
+      "step": 149480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6720774853145598e-05,
+      "loss": 0.1659,
+      "step": 149490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6718532801219676e-05,
+      "loss": 0.1665,
+      "step": 149500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.671629074929375e-05,
+      "loss": 0.1665,
+      "step": 149510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.671404869736783e-05,
+      "loss": 0.1647,
+      "step": 149520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.671180664544191e-05,
+      "loss": 0.1689,
+      "step": 149530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6709564593515988e-05,
+      "loss": 0.1731,
+      "step": 149540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6707322541590063e-05,
+      "loss": 0.1704,
+      "step": 149550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6705080489664142e-05,
+      "loss": 0.1687,
+      "step": 149560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.670283843773822e-05,
+      "loss": 0.1609,
+      "step": 149570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6700596385812296e-05,
+      "loss": 0.1626,
+      "step": 149580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6698354333886375e-05,
+      "loss": 0.1631,
+      "step": 149590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6696112281960453e-05,
+      "loss": 0.1669,
+      "step": 149600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.669387023003453e-05,
+      "loss": 0.163,
+      "step": 149610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6691628178108607e-05,
+      "loss": 0.1706,
+      "step": 149620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6689386126182683e-05,
+      "loss": 0.1641,
+      "step": 149630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.668714407425676e-05,
+      "loss": 0.1681,
+      "step": 149640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6684902022330837e-05,
+      "loss": 0.1684,
+      "step": 149650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6682659970404915e-05,
+      "loss": 0.1681,
+      "step": 149660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6680417918478994e-05,
+      "loss": 0.168,
+      "step": 149670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.667817586655307e-05,
+      "loss": 0.168,
+      "step": 149680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6675933814627148e-05,
+      "loss": 0.1672,
+      "step": 149690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6673691762701224e-05,
+      "loss": 0.1611,
+      "step": 149700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6671449710775302e-05,
+      "loss": 0.1617,
+      "step": 149710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6669207658849378e-05,
+      "loss": 0.1644,
+      "step": 149720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6666965606923456e-05,
+      "loss": 0.1692,
+      "step": 149730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6664723554997535e-05,
+      "loss": 0.1741,
+      "step": 149740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.666248150307161e-05,
+      "loss": 0.1641,
+      "step": 149750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.666023945114569e-05,
+      "loss": 0.165,
+      "step": 149760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6657997399219764e-05,
+      "loss": 0.1706,
+      "step": 149770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6655755347293843e-05,
+      "loss": 0.1678,
+      "step": 149780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6653513295367922e-05,
+      "loss": 0.1715,
+      "step": 149790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6651271243441997e-05,
+      "loss": 0.1628,
+      "step": 149800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6649029191516076e-05,
+      "loss": 0.174,
+      "step": 149810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6646787139590155e-05,
+      "loss": 0.1685,
+      "step": 149820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6644545087664233e-05,
+      "loss": 0.1685,
+      "step": 149830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.664230303573831e-05,
+      "loss": 0.1713,
+      "step": 149840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6640060983812387e-05,
+      "loss": 0.1684,
+      "step": 149850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6637818931886463e-05,
+      "loss": 0.1679,
+      "step": 149860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.663557687996054e-05,
+      "loss": 0.1723,
+      "step": 149870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.663333482803462e-05,
+      "loss": 0.1681,
+      "step": 149880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6631092776108695e-05,
+      "loss": 0.1684,
+      "step": 149890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6628850724182774e-05,
+      "loss": 0.1676,
+      "step": 149900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.662660867225685e-05,
+      "loss": 0.1656,
+      "step": 149910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6624366620330928e-05,
+      "loss": 0.1664,
+      "step": 149920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6622124568405007e-05,
+      "loss": 0.1728,
+      "step": 149930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6619882516479082e-05,
+      "loss": 0.1665,
+      "step": 149940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.661764046455316e-05,
+      "loss": 0.1722,
+      "step": 149950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6615398412627236e-05,
+      "loss": 0.1613,
+      "step": 149960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6613156360701315e-05,
+      "loss": 0.1637,
+      "step": 149970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.661091430877539e-05,
+      "loss": 0.168,
+      "step": 149980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.660867225684947e-05,
+      "loss": 0.1695,
+      "step": 149990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6606430204923548e-05,
+      "loss": 0.1683,
+      "step": 150000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6604188152997623e-05,
+      "loss": 0.1644,
+      "step": 150010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6601946101071702e-05,
+      "loss": 0.1576,
+      "step": 150020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6599704049145777e-05,
+      "loss": 0.1655,
+      "step": 150030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6597461997219856e-05,
+      "loss": 0.1654,
+      "step": 150040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.659521994529393e-05,
+      "loss": 0.1719,
+      "step": 150050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.659297789336801e-05,
+      "loss": 0.1607,
+      "step": 150060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.659073584144209e-05,
+      "loss": 0.1638,
+      "step": 150070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6588493789516167e-05,
+      "loss": 0.1692,
+      "step": 150080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6586251737590243e-05,
+      "loss": 0.167,
+      "step": 150090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.658400968566432e-05,
+      "loss": 0.1626,
+      "step": 150100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.65817676337384e-05,
+      "loss": 0.1708,
+      "step": 150110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6579525581812475e-05,
+      "loss": 0.1703,
+      "step": 150120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6577283529886554e-05,
+      "loss": 0.1679,
+      "step": 150130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6575265683153225e-05,
+      "loss": 0.1631,
+      "step": 150140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.65730236312273e-05,
+      "loss": 0.1674,
+      "step": 150150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.657078157930138e-05,
+      "loss": 0.1696,
+      "step": 150160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6568539527375455e-05,
+      "loss": 0.1624,
+      "step": 150170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6566297475449533e-05,
+      "loss": 0.1692,
+      "step": 150180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.656405542352361e-05,
+      "loss": 0.1565,
+      "step": 150190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6561813371597688e-05,
+      "loss": 0.1637,
+      "step": 150200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6559571319671766e-05,
+      "loss": 0.1722,
+      "step": 150210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.655732926774584e-05,
+      "loss": 0.172,
+      "step": 150220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.655508721581992e-05,
+      "loss": 0.1629,
+      "step": 150230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6552845163893996e-05,
+      "loss": 0.1706,
+      "step": 150240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6550603111968074e-05,
+      "loss": 0.1689,
+      "step": 150250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.654836106004215e-05,
+      "loss": 0.1628,
+      "step": 150260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.654611900811623e-05,
+      "loss": 0.1631,
+      "step": 150270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6543876956190307e-05,
+      "loss": 0.1724,
+      "step": 150280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6541634904264382e-05,
+      "loss": 0.1703,
+      "step": 150290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.653939285233846e-05,
+      "loss": 0.1685,
+      "step": 150300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6537150800412536e-05,
+      "loss": 0.1695,
+      "step": 150310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6534908748486615e-05,
+      "loss": 0.1599,
+      "step": 150320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.653266669656069e-05,
+      "loss": 0.1604,
+      "step": 150330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.653042464463477e-05,
+      "loss": 0.1648,
+      "step": 150340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6528182592708848e-05,
+      "loss": 0.1599,
+      "step": 150350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6525940540782923e-05,
+      "loss": 0.1623,
+      "step": 150360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6523698488857002e-05,
+      "loss": 0.1734,
+      "step": 150370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.652145643693108e-05,
+      "loss": 0.168,
+      "step": 150380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.651921438500516e-05,
+      "loss": 0.1691,
+      "step": 150390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6516972333079235e-05,
+      "loss": 0.1656,
+      "step": 150400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6514730281153313e-05,
+      "loss": 0.1703,
+      "step": 150410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6512488229227392e-05,
+      "loss": 0.1622,
+      "step": 150420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6510246177301468e-05,
+      "loss": 0.1667,
+      "step": 150430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6508004125375546e-05,
+      "loss": 0.163,
+      "step": 150440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.650576207344962e-05,
+      "loss": 0.1701,
+      "step": 150450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.65035200215237e-05,
+      "loss": 0.1662,
+      "step": 150460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6501277969597776e-05,
+      "loss": 0.1671,
+      "step": 150470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6499035917671854e-05,
+      "loss": 0.1676,
+      "step": 150480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6496793865745933e-05,
+      "loss": 0.1729,
+      "step": 150490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.649455181382001e-05,
+      "loss": 0.168,
+      "step": 150500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6492309761894087e-05,
+      "loss": 0.1658,
+      "step": 150510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6490067709968162e-05,
+      "loss": 0.168,
+      "step": 150520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.648782565804224e-05,
+      "loss": 0.1701,
+      "step": 150530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6485583606116316e-05,
+      "loss": 0.1607,
+      "step": 150540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6483341554190395e-05,
+      "loss": 0.1654,
+      "step": 150550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6481099502264474e-05,
+      "loss": 0.1596,
+      "step": 150560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.647885745033855e-05,
+      "loss": 0.1624,
+      "step": 150570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6476615398412628e-05,
+      "loss": 0.1658,
+      "step": 150580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6474373346486703e-05,
+      "loss": 0.1639,
+      "step": 150590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6472131294560782e-05,
+      "loss": 0.1699,
+      "step": 150600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6469889242634857e-05,
+      "loss": 0.1678,
+      "step": 150610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6467647190708936e-05,
+      "loss": 0.1663,
+      "step": 150620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6465405138783015e-05,
+      "loss": 0.1645,
+      "step": 150630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6463163086857093e-05,
+      "loss": 0.1643,
+      "step": 150640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.646092103493117e-05,
+      "loss": 0.1686,
+      "step": 150650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6458678983005248e-05,
+      "loss": 0.1586,
+      "step": 150660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6456436931079326e-05,
+      "loss": 0.1683,
+      "step": 150670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.64541948791534e-05,
+      "loss": 0.1641,
+      "step": 150680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.645195282722748e-05,
+      "loss": 0.1717,
+      "step": 150690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.644971077530156e-05,
+      "loss": 0.1652,
+      "step": 150700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6447468723375634e-05,
+      "loss": 0.163,
+      "step": 150710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6445226671449713e-05,
+      "loss": 0.172,
+      "step": 150720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.644298461952379e-05,
+      "loss": 0.1676,
+      "step": 150730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6440742567597867e-05,
+      "loss": 0.1718,
+      "step": 150740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6438500515671942e-05,
+      "loss": 0.1624,
+      "step": 150750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.643625846374602e-05,
+      "loss": 0.1637,
+      "step": 150760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.64340164118201e-05,
+      "loss": 0.1697,
+      "step": 150770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6431774359894175e-05,
+      "loss": 0.1643,
+      "step": 150780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6429532307968254e-05,
+      "loss": 0.164,
+      "step": 150790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.642729025604233e-05,
+      "loss": 0.1636,
+      "step": 150800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6425048204116408e-05,
+      "loss": 0.167,
+      "step": 150810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6422806152190487e-05,
+      "loss": 0.1619,
+      "step": 150820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6420564100264562e-05,
+      "loss": 0.1691,
+      "step": 150830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.641832204833864e-05,
+      "loss": 0.1675,
+      "step": 150840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6416079996412716e-05,
+      "loss": 0.168,
+      "step": 150850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6413837944486795e-05,
+      "loss": 0.1678,
+      "step": 150860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.641159589256087e-05,
+      "loss": 0.1697,
+      "step": 150870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.640935384063495e-05,
+      "loss": 0.1673,
+      "step": 150880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6407111788709028e-05,
+      "loss": 0.1633,
+      "step": 150890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6404869736783103e-05,
+      "loss": 0.164,
+      "step": 150900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.640262768485718e-05,
+      "loss": 0.1609,
+      "step": 150910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.640038563293126e-05,
+      "loss": 0.1699,
+      "step": 150920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.639814358100534e-05,
+      "loss": 0.1672,
+      "step": 150930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6395901529079414e-05,
+      "loss": 0.1669,
+      "step": 150940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6393659477153493e-05,
+      "loss": 0.1631,
+      "step": 150950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6391417425227572e-05,
+      "loss": 0.1662,
+      "step": 150960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6389175373301647e-05,
+      "loss": 0.1695,
+      "step": 150970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6386933321375726e-05,
+      "loss": 0.1657,
+      "step": 150980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.63846912694498e-05,
+      "loss": 0.1719,
+      "step": 150990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.638244921752388e-05,
+      "loss": 0.1617,
+      "step": 151000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6380207165597955e-05,
+      "loss": 0.1639,
+      "step": 151010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6377965113672034e-05,
+      "loss": 0.1693,
+      "step": 151020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6375723061746113e-05,
+      "loss": 0.1661,
+      "step": 151030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6373481009820188e-05,
+      "loss": 0.169,
+      "step": 151040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6371238957894267e-05,
+      "loss": 0.1636,
+      "step": 151050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6368996905968342e-05,
+      "loss": 0.1667,
+      "step": 151060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.636675485404242e-05,
+      "loss": 0.1659,
+      "step": 151070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6364512802116496e-05,
+      "loss": 0.1667,
+      "step": 151080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6362270750190575e-05,
+      "loss": 0.1677,
+      "step": 151090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6360028698264654e-05,
+      "loss": 0.1623,
+      "step": 151100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.635778664633873e-05,
+      "loss": 0.164,
+      "step": 151110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6355544594412808e-05,
+      "loss": 0.1669,
+      "step": 151120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6353302542486883e-05,
+      "loss": 0.1626,
+      "step": 151130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.635106049056096e-05,
+      "loss": 0.1647,
+      "step": 151140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6348818438635037e-05,
+      "loss": 0.1734,
+      "step": 151150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6346576386709116e-05,
+      "loss": 0.1674,
+      "step": 151160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6344334334783194e-05,
+      "loss": 0.1642,
+      "step": 151170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6342092282857273e-05,
+      "loss": 0.1638,
+      "step": 151180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.633985023093135e-05,
+      "loss": 0.1636,
+      "step": 151190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.6337608179005427e-05,
+      "loss": 0.1671,
+      "step": 151200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6335366127079506e-05,
+      "loss": 0.1658,
+      "step": 151210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.633312407515358e-05,
+      "loss": 0.1688,
+      "step": 151220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.633088202322766e-05,
+      "loss": 0.1647,
+      "step": 151230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.632863997130174e-05,
+      "loss": 0.1661,
+      "step": 151240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6326397919375814e-05,
+      "loss": 0.1685,
+      "step": 151250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6324155867449893e-05,
+      "loss": 0.1674,
+      "step": 151260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6321913815523968e-05,
+      "loss": 0.1698,
+      "step": 151270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6319671763598047e-05,
+      "loss": 0.1618,
+      "step": 151280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6317429711672122e-05,
+      "loss": 0.1664,
+      "step": 151290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.63151876597462e-05,
+      "loss": 0.1658,
+      "step": 151300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.631294560782028e-05,
+      "loss": 0.1661,
+      "step": 151310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6310703555894355e-05,
+      "loss": 0.1676,
+      "step": 151320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6308461503968434e-05,
+      "loss": 0.1692,
+      "step": 151330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.630621945204251e-05,
+      "loss": 0.1687,
+      "step": 151340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6303977400116588e-05,
+      "loss": 0.1598,
+      "step": 151350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6301735348190663e-05,
+      "loss": 0.1666,
+      "step": 151360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.629949329626474e-05,
+      "loss": 0.1706,
+      "step": 151370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.629725124433882e-05,
+      "loss": 0.1619,
+      "step": 151380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6295009192412896e-05,
+      "loss": 0.1646,
+      "step": 151390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6292767140486974e-05,
+      "loss": 0.1631,
+      "step": 151400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.629052508856105e-05,
+      "loss": 0.1681,
+      "step": 151410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.628828303663513e-05,
+      "loss": 0.172,
+      "step": 151420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6286040984709207e-05,
+      "loss": 0.1593,
+      "step": 151430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6283798932783282e-05,
+      "loss": 0.1698,
+      "step": 151440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.628155688085736e-05,
+      "loss": 0.1704,
+      "step": 151450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.627931482893144e-05,
+      "loss": 0.1664,
+      "step": 151460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.627707277700552e-05,
+      "loss": 0.1633,
+      "step": 151470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6274830725079594e-05,
+      "loss": 0.1672,
+      "step": 151480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6272588673153673e-05,
+      "loss": 0.1672,
+      "step": 151490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6270346621227748e-05,
+      "loss": 0.165,
+      "step": 151500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6268104569301827e-05,
+      "loss": 0.1684,
+      "step": 151510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6265862517375905e-05,
+      "loss": 0.1746,
+      "step": 151520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.626362046544998e-05,
+      "loss": 0.1657,
+      "step": 151530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.626137841352406e-05,
+      "loss": 0.1701,
+      "step": 151540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6259136361598135e-05,
+      "loss": 0.1741,
+      "step": 151550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6256894309672214e-05,
+      "loss": 0.1701,
+      "step": 151560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.625465225774629e-05,
+      "loss": 0.1672,
+      "step": 151570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6252410205820368e-05,
+      "loss": 0.1654,
+      "step": 151580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6250168153894446e-05,
+      "loss": 0.167,
+      "step": 151590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.624792610196852e-05,
+      "loss": 0.1631,
+      "step": 151600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.62456840500426e-05,
+      "loss": 0.1647,
+      "step": 151610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6243441998116676e-05,
+      "loss": 0.1648,
+      "step": 151620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6241199946190754e-05,
+      "loss": 0.1688,
+      "step": 151630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.623895789426483e-05,
+      "loss": 0.1688,
+      "step": 151640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.623671584233891e-05,
+      "loss": 0.161,
+      "step": 151650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6234473790412987e-05,
+      "loss": 0.1647,
+      "step": 151660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6232231738487063e-05,
+      "loss": 0.1689,
+      "step": 151670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.622998968656114e-05,
+      "loss": 0.1616,
+      "step": 151680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6227747634635217e-05,
+      "loss": 0.1715,
+      "step": 151690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6225505582709295e-05,
+      "loss": 0.1629,
+      "step": 151700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6223263530783374e-05,
+      "loss": 0.1652,
+      "step": 151710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6221021478857453e-05,
+      "loss": 0.1727,
+      "step": 151720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6218779426931528e-05,
+      "loss": 0.1656,
+      "step": 151730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6216537375005607e-05,
+      "loss": 0.1625,
+      "step": 151740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6214295323079685e-05,
+      "loss": 0.1675,
+      "step": 151750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.621205327115376e-05,
+      "loss": 0.1576,
+      "step": 151760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.620981121922784e-05,
+      "loss": 0.1656,
+      "step": 151770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6207569167301915e-05,
+      "loss": 0.1675,
+      "step": 151780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6205327115375994e-05,
+      "loss": 0.1687,
+      "step": 151790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6203085063450072e-05,
+      "loss": 0.1665,
+      "step": 151800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6200843011524148e-05,
+      "loss": 0.1697,
+      "step": 151810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6198600959598226e-05,
+      "loss": 0.1664,
+      "step": 151820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.61963589076723e-05,
+      "loss": 0.1622,
+      "step": 151830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.619411685574638e-05,
+      "loss": 0.1602,
+      "step": 151840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6191874803820456e-05,
+      "loss": 0.1625,
+      "step": 151850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6189632751894534e-05,
+      "loss": 0.1656,
+      "step": 151860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6187390699968613e-05,
+      "loss": 0.167,
+      "step": 151870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.618514864804269e-05,
+      "loss": 0.1604,
+      "step": 151880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6182906596116767e-05,
+      "loss": 0.1692,
+      "step": 151890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6180664544190843e-05,
+      "loss": 0.1682,
+      "step": 151900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.617842249226492e-05,
+      "loss": 0.1636,
+      "step": 151910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6176180440338997e-05,
+      "loss": 0.1691,
+      "step": 151920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6173938388413075e-05,
+      "loss": 0.1628,
+      "step": 151930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6171696336487154e-05,
+      "loss": 0.1652,
+      "step": 151940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.616945428456123e-05,
+      "loss": 0.1673,
+      "step": 151950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6167212232635308e-05,
+      "loss": 0.1689,
+      "step": 151960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6164970180709383e-05,
+      "loss": 0.1677,
+      "step": 151970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6162728128783462e-05,
+      "loss": 0.1619,
+      "step": 151980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.616048607685754e-05,
+      "loss": 0.1645,
+      "step": 151990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.615824402493162e-05,
+      "loss": 0.1604,
+      "step": 152000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6156001973005698e-05,
+      "loss": 0.1663,
+      "step": 152010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6153759921079774e-05,
+      "loss": 0.1742,
+      "step": 152020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6151517869153852e-05,
+      "loss": 0.1622,
+      "step": 152030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6149275817227928e-05,
+      "loss": 0.1616,
+      "step": 152040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6147033765302006e-05,
+      "loss": 0.1639,
+      "step": 152050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.614479171337608e-05,
+      "loss": 0.1617,
+      "step": 152060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.614254966145016e-05,
+      "loss": 0.17,
+      "step": 152070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.614030760952424e-05,
+      "loss": 0.1604,
+      "step": 152080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6138065557598314e-05,
+      "loss": 0.1633,
+      "step": 152090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6135823505672393e-05,
+      "loss": 0.1701,
+      "step": 152100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.613358145374647e-05,
+      "loss": 0.163,
+      "step": 152110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6131339401820547e-05,
+      "loss": 0.1714,
+      "step": 152120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6129097349894623e-05,
+      "loss": 0.1664,
+      "step": 152130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.61268552979687e-05,
+      "loss": 0.1678,
+      "step": 152140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.612461324604278e-05,
+      "loss": 0.1662,
+      "step": 152150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6122371194116855e-05,
+      "loss": 0.1643,
+      "step": 152160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6120129142190934e-05,
+      "loss": 0.1621,
+      "step": 152170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.611788709026501e-05,
+      "loss": 0.1694,
+      "step": 152180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6115645038339088e-05,
+      "loss": 0.1709,
+      "step": 152190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6113402986413167e-05,
+      "loss": 0.1616,
+      "step": 152200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6111160934487242e-05,
+      "loss": 0.1622,
+      "step": 152210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.610891888256132e-05,
+      "loss": 0.1632,
+      "step": 152220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6106676830635396e-05,
+      "loss": 0.1647,
+      "step": 152230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6104434778709475e-05,
+      "loss": 0.1687,
+      "step": 152240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6102192726783554e-05,
+      "loss": 0.1641,
+      "step": 152250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6099950674857632e-05,
+      "loss": 0.1662,
+      "step": 152260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6097708622931708e-05,
+      "loss": 0.1663,
+      "step": 152270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6095466571005786e-05,
+      "loss": 0.1761,
+      "step": 152280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6093224519079865e-05,
+      "loss": 0.1627,
+      "step": 152290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.609098246715394e-05,
+      "loss": 0.1653,
+      "step": 152300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.608874041522802e-05,
+      "loss": 0.1688,
+      "step": 152310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6086498363302094e-05,
+      "loss": 0.1656,
+      "step": 152320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6084256311376173e-05,
+      "loss": 0.1642,
+      "step": 152330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6082014259450252e-05,
+      "loss": 0.1639,
+      "step": 152340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6079772207524327e-05,
+      "loss": 0.1664,
+      "step": 152350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6077530155598406e-05,
+      "loss": 0.1655,
+      "step": 152360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6075512308865074e-05,
+      "loss": 0.1674,
+      "step": 152370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6073270256939152e-05,
+      "loss": 0.167,
+      "step": 152380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6071028205013228e-05,
+      "loss": 0.1586,
+      "step": 152390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6068786153087306e-05,
+      "loss": 0.1637,
+      "step": 152400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6066544101161385e-05,
+      "loss": 0.1606,
+      "step": 152410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.606430204923546e-05,
+      "loss": 0.1649,
+      "step": 152420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.606205999730954e-05,
+      "loss": 0.1688,
+      "step": 152430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6059817945383615e-05,
+      "loss": 0.1642,
+      "step": 152440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6057575893457693e-05,
+      "loss": 0.1588,
+      "step": 152450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.605533384153177e-05,
+      "loss": 0.1697,
+      "step": 152460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6053091789605847e-05,
+      "loss": 0.1641,
+      "step": 152470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6050849737679926e-05,
+      "loss": 0.1693,
+      "step": 152480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6048607685754e-05,
+      "loss": 0.164,
+      "step": 152490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.604636563382808e-05,
+      "loss": 0.1698,
+      "step": 152500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6044123581902155e-05,
+      "loss": 0.1624,
+      "step": 152510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6041881529976234e-05,
+      "loss": 0.1663,
+      "step": 152520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6039639478050313e-05,
+      "loss": 0.1699,
+      "step": 152530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6037397426124388e-05,
+      "loss": 0.1631,
+      "step": 152540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6035155374198467e-05,
+      "loss": 0.1683,
+      "step": 152550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6032913322272546e-05,
+      "loss": 0.1665,
+      "step": 152560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6030671270346624e-05,
+      "loss": 0.171,
+      "step": 152570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.60284292184207e-05,
+      "loss": 0.1687,
+      "step": 152580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.602618716649478e-05,
+      "loss": 0.1582,
+      "step": 152590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6023945114568854e-05,
+      "loss": 0.1582,
+      "step": 152600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6021703062642932e-05,
+      "loss": 0.1623,
+      "step": 152610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.601946101071701e-05,
+      "loss": 0.1664,
+      "step": 152620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6017218958791087e-05,
+      "loss": 0.1636,
+      "step": 152630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6014976906865165e-05,
+      "loss": 0.1671,
+      "step": 152640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.601273485493924e-05,
+      "loss": 0.1657,
+      "step": 152650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.601049280301332e-05,
+      "loss": 0.1689,
+      "step": 152660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6008250751087395e-05,
+      "loss": 0.1675,
+      "step": 152670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6006008699161473e-05,
+      "loss": 0.1609,
+      "step": 152680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6003766647235552e-05,
+      "loss": 0.1603,
+      "step": 152690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.6001524595309627e-05,
+      "loss": 0.1648,
+      "step": 152700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5999282543383706e-05,
+      "loss": 0.1655,
+      "step": 152710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.599704049145778e-05,
+      "loss": 0.1681,
+      "step": 152720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.599479843953186e-05,
+      "loss": 0.165,
+      "step": 152730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5992556387605935e-05,
+      "loss": 0.1703,
+      "step": 152740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5990314335680014e-05,
+      "loss": 0.1614,
+      "step": 152750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5988072283754093e-05,
+      "loss": 0.1691,
+      "step": 152760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5985830231828168e-05,
+      "loss": 0.1722,
+      "step": 152770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5983588179902247e-05,
+      "loss": 0.1641,
+      "step": 152780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5981346127976322e-05,
+      "loss": 0.1663,
+      "step": 152790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.59791040760504e-05,
+      "loss": 0.1642,
+      "step": 152800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.597686202412448e-05,
+      "loss": 0.1662,
+      "step": 152810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.597461997219856e-05,
+      "loss": 0.1704,
+      "step": 152820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5972377920272634e-05,
+      "loss": 0.1699,
+      "step": 152830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5970135868346712e-05,
+      "loss": 0.165,
+      "step": 152840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.596789381642079e-05,
+      "loss": 0.1677,
+      "step": 152850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5965651764494867e-05,
+      "loss": 0.1674,
+      "step": 152860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5963409712568945e-05,
+      "loss": 0.1622,
+      "step": 152870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.596116766064302e-05,
+      "loss": 0.1657,
+      "step": 152880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.59589256087171e-05,
+      "loss": 0.1645,
+      "step": 152890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5956683556791178e-05,
+      "loss": 0.168,
+      "step": 152900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5954441504865253e-05,
+      "loss": 0.163,
+      "step": 152910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5952199452939332e-05,
+      "loss": 0.1694,
+      "step": 152920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5949957401013407e-05,
+      "loss": 0.1613,
+      "step": 152930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5947715349087486e-05,
+      "loss": 0.1662,
+      "step": 152940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.594547329716156e-05,
+      "loss": 0.1672,
+      "step": 152950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.594323124523564e-05,
+      "loss": 0.1638,
+      "step": 152960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.594098919330972e-05,
+      "loss": 0.1711,
+      "step": 152970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5938747141383794e-05,
+      "loss": 0.1682,
+      "step": 152980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5936505089457873e-05,
+      "loss": 0.1599,
+      "step": 152990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5934263037531948e-05,
+      "loss": 0.1658,
+      "step": 153000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5932020985606027e-05,
+      "loss": 0.1621,
+      "step": 153010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5929778933680102e-05,
+      "loss": 0.1609,
+      "step": 153020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.592753688175418e-05,
+      "loss": 0.1631,
+      "step": 153030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.592529482982826e-05,
+      "loss": 0.1668,
+      "step": 153040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5923052777902335e-05,
+      "loss": 0.1661,
+      "step": 153050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5920810725976414e-05,
+      "loss": 0.1651,
+      "step": 153060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.591856867405049e-05,
+      "loss": 0.1579,
+      "step": 153070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5916326622124568e-05,
+      "loss": 0.1628,
+      "step": 153080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5914084570198647e-05,
+      "loss": 0.1645,
+      "step": 153090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5911842518272725e-05,
+      "loss": 0.1636,
+      "step": 153100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5909600466346804e-05,
+      "loss": 0.1645,
+      "step": 153110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.590735841442088e-05,
+      "loss": 0.1659,
+      "step": 153120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5905116362494958e-05,
+      "loss": 0.1675,
+      "step": 153130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5902874310569033e-05,
+      "loss": 0.1661,
+      "step": 153140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5900632258643112e-05,
+      "loss": 0.1639,
+      "step": 153150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.589839020671719e-05,
+      "loss": 0.1672,
+      "step": 153160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5896148154791266e-05,
+      "loss": 0.1652,
+      "step": 153170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5893906102865345e-05,
+      "loss": 0.1621,
+      "step": 153180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.589166405093942e-05,
+      "loss": 0.1698,
+      "step": 153190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.58894219990135e-05,
+      "loss": 0.1667,
+      "step": 153200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5887179947087574e-05,
+      "loss": 0.1635,
+      "step": 153210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5884937895161653e-05,
+      "loss": 0.1695,
+      "step": 153220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.588269584323573e-05,
+      "loss": 0.1612,
+      "step": 153230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5880453791309807e-05,
+      "loss": 0.1658,
+      "step": 153240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5878211739383886e-05,
+      "loss": 0.1642,
+      "step": 153250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.587596968745796e-05,
+      "loss": 0.1706,
+      "step": 153260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.587372763553204e-05,
+      "loss": 0.1656,
+      "step": 153270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5871485583606115e-05,
+      "loss": 0.1665,
+      "step": 153280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5869243531680194e-05,
+      "loss": 0.1617,
+      "step": 153290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5867001479754273e-05,
+      "loss": 0.1642,
+      "step": 153300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5864759427828348e-05,
+      "loss": 0.1679,
+      "step": 153310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5862517375902427e-05,
+      "loss": 0.1682,
+      "step": 153320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5860275323976502e-05,
+      "loss": 0.1685,
+      "step": 153330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.585803327205058e-05,
+      "loss": 0.1703,
+      "step": 153340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.585579122012466e-05,
+      "loss": 0.1648,
+      "step": 153350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5853549168198738e-05,
+      "loss": 0.1605,
+      "step": 153360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5851307116272813e-05,
+      "loss": 0.1622,
+      "step": 153370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5849065064346892e-05,
+      "loss": 0.1738,
+      "step": 153380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.584682301242097e-05,
+      "loss": 0.1681,
+      "step": 153390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5844580960495046e-05,
+      "loss": 0.1688,
+      "step": 153400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5842338908569125e-05,
+      "loss": 0.1664,
+      "step": 153410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.58400968566432e-05,
+      "loss": 0.1672,
+      "step": 153420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.583785480471728e-05,
+      "loss": 0.1664,
+      "step": 153430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5835612752791358e-05,
+      "loss": 0.1709,
+      "step": 153440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5833370700865433e-05,
+      "loss": 0.1617,
+      "step": 153450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.583112864893951e-05,
+      "loss": 0.1674,
+      "step": 153460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5828886597013587e-05,
+      "loss": 0.1683,
+      "step": 153470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5826644545087666e-05,
+      "loss": 0.1604,
+      "step": 153480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.582440249316174e-05,
+      "loss": 0.1675,
+      "step": 153490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.582216044123582e-05,
+      "loss": 0.1617,
+      "step": 153500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.58199183893099e-05,
+      "loss": 0.161,
+      "step": 153510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5817676337383974e-05,
+      "loss": 0.1695,
+      "step": 153520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5815434285458053e-05,
+      "loss": 0.1642,
+      "step": 153530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5813192233532128e-05,
+      "loss": 0.171,
+      "step": 153540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5810950181606207e-05,
+      "loss": 0.1631,
+      "step": 153550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5808708129680282e-05,
+      "loss": 0.1718,
+      "step": 153560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.580646607775436e-05,
+      "loss": 0.1703,
+      "step": 153570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.580422402582844e-05,
+      "loss": 0.1594,
+      "step": 153580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5801981973902515e-05,
+      "loss": 0.1652,
+      "step": 153590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5799739921976593e-05,
+      "loss": 0.1612,
+      "step": 153600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.579749787005067e-05,
+      "loss": 0.1709,
+      "step": 153610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5795255818124747e-05,
+      "loss": 0.1645,
+      "step": 153620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5793013766198826e-05,
+      "loss": 0.1627,
+      "step": 153630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5790771714272905e-05,
+      "loss": 0.1619,
+      "step": 153640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5788529662346984e-05,
+      "loss": 0.1674,
+      "step": 153650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.578628761042106e-05,
+      "loss": 0.1674,
+      "step": 153660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5784045558495138e-05,
+      "loss": 0.165,
+      "step": 153670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5781803506569213e-05,
+      "loss": 0.1678,
+      "step": 153680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.577956145464329e-05,
+      "loss": 0.1669,
+      "step": 153690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5777319402717367e-05,
+      "loss": 0.1713,
+      "step": 153700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5775077350791446e-05,
+      "loss": 0.1625,
+      "step": 153710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5772835298865524e-05,
+      "loss": 0.1657,
+      "step": 153720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.57705932469396e-05,
+      "loss": 0.1635,
+      "step": 153730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.576835119501368e-05,
+      "loss": 0.1599,
+      "step": 153740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5766109143087754e-05,
+      "loss": 0.1628,
+      "step": 153750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5763867091161833e-05,
+      "loss": 0.1623,
+      "step": 153760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5761625039235908e-05,
+      "loss": 0.1663,
+      "step": 153770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5759382987309987e-05,
+      "loss": 0.1674,
+      "step": 153780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5757140935384065e-05,
+      "loss": 0.1672,
+      "step": 153790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.575489888345814e-05,
+      "loss": 0.1666,
+      "step": 153800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.575265683153222e-05,
+      "loss": 0.1671,
+      "step": 153810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5750414779606295e-05,
+      "loss": 0.1673,
+      "step": 153820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5748172727680373e-05,
+      "loss": 0.1651,
+      "step": 153830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.574593067575445e-05,
+      "loss": 0.1637,
+      "step": 153840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5743688623828527e-05,
+      "loss": 0.1664,
+      "step": 153850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5741446571902606e-05,
+      "loss": 0.1615,
+      "step": 153860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.573920451997668e-05,
+      "loss": 0.1607,
+      "step": 153870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.573696246805076e-05,
+      "loss": 0.1671,
+      "step": 153880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.573472041612484e-05,
+      "loss": 0.1617,
+      "step": 153890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5732478364198918e-05,
+      "loss": 0.1628,
+      "step": 153900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5730236312272993e-05,
+      "loss": 0.1705,
+      "step": 153910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5727994260347072e-05,
+      "loss": 0.166,
+      "step": 153920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.572575220842115e-05,
+      "loss": 0.168,
+      "step": 153930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5723510156495226e-05,
+      "loss": 0.1604,
+      "step": 153940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5721268104569304e-05,
+      "loss": 0.1629,
+      "step": 153950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.571902605264338e-05,
+      "loss": 0.1631,
+      "step": 153960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.571678400071746e-05,
+      "loss": 0.1722,
+      "step": 153970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5714541948791534e-05,
+      "loss": 0.1648,
+      "step": 153980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5712299896865613e-05,
+      "loss": 0.1677,
+      "step": 153990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.571005784493969e-05,
+      "loss": 0.1673,
+      "step": 154000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5707815793013767e-05,
+      "loss": 0.1603,
+      "step": 154010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5705573741087845e-05,
+      "loss": 0.1622,
+      "step": 154020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.570333168916192e-05,
+      "loss": 0.1654,
+      "step": 154030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5701089637236e-05,
+      "loss": 0.1691,
+      "step": 154040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5698847585310075e-05,
+      "loss": 0.1612,
+      "step": 154050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5696605533384153e-05,
+      "loss": 0.1717,
+      "step": 154060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5694363481458232e-05,
+      "loss": 0.1717,
+      "step": 154070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5692121429532307e-05,
+      "loss": 0.1656,
+      "step": 154080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5689879377606386e-05,
+      "loss": 0.1678,
+      "step": 154090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.568763732568046e-05,
+      "loss": 0.1577,
+      "step": 154100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.568539527375454e-05,
+      "loss": 0.1595,
+      "step": 154110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5683153221828616e-05,
+      "loss": 0.1676,
+      "step": 154120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5680911169902694e-05,
+      "loss": 0.1729,
+      "step": 154130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5678669117976773e-05,
+      "loss": 0.1646,
+      "step": 154140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.567642706605085e-05,
+      "loss": 0.1727,
+      "step": 154150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5674185014124927e-05,
+      "loss": 0.1681,
+      "step": 154160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5671942962199006e-05,
+      "loss": 0.1653,
+      "step": 154170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5669700910273084e-05,
+      "loss": 0.1636,
+      "step": 154180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.566745885834716e-05,
+      "loss": 0.1632,
+      "step": 154190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.566521680642124e-05,
+      "loss": 0.165,
+      "step": 154200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5662974754495317e-05,
+      "loss": 0.1675,
+      "step": 154210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5660732702569393e-05,
+      "loss": 0.1646,
+      "step": 154220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.565849065064347e-05,
+      "loss": 0.1599,
+      "step": 154230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5656248598717547e-05,
+      "loss": 0.1627,
+      "step": 154240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5654006546791625e-05,
+      "loss": 0.1647,
+      "step": 154250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.56517644948657e-05,
+      "loss": 0.1632,
+      "step": 154260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.564952244293978e-05,
+      "loss": 0.1617,
+      "step": 154270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5647280391013858e-05,
+      "loss": 0.1642,
+      "step": 154280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5645038339087933e-05,
+      "loss": 0.1654,
+      "step": 154290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5642796287162012e-05,
+      "loss": 0.1666,
+      "step": 154300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5640554235236087e-05,
+      "loss": 0.1669,
+      "step": 154310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5638312183310166e-05,
+      "loss": 0.1672,
+      "step": 154320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.563607013138424e-05,
+      "loss": 0.1637,
+      "step": 154330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.563382807945832e-05,
+      "loss": 0.1636,
+      "step": 154340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.56315860275324e-05,
+      "loss": 0.1659,
+      "step": 154350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5629343975606474e-05,
+      "loss": 0.1684,
+      "step": 154360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5627101923680553e-05,
+      "loss": 0.1639,
+      "step": 154370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.562508407694722e-05,
+      "loss": 0.1625,
+      "step": 154380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.56228420250213e-05,
+      "loss": 0.1636,
+      "step": 154390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5620599973095378e-05,
+      "loss": 0.1631,
+      "step": 154400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5618357921169454e-05,
+      "loss": 0.1621,
+      "step": 154410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5616115869243532e-05,
+      "loss": 0.1616,
+      "step": 154420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5613873817317608e-05,
+      "loss": 0.1654,
+      "step": 154430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5611631765391686e-05,
+      "loss": 0.1631,
+      "step": 154440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5609389713465765e-05,
+      "loss": 0.1621,
+      "step": 154450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5607147661539844e-05,
+      "loss": 0.1614,
+      "step": 154460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.560490560961392e-05,
+      "loss": 0.1608,
+      "step": 154470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5602663557687998e-05,
+      "loss": 0.165,
+      "step": 154480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5600421505762077e-05,
+      "loss": 0.1645,
+      "step": 154490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5598179453836152e-05,
+      "loss": 0.167,
+      "step": 154500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.559593740191023e-05,
+      "loss": 0.1649,
+      "step": 154510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5593695349984306e-05,
+      "loss": 0.1688,
+      "step": 154520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5591453298058385e-05,
+      "loss": 0.1584,
+      "step": 154530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5589211246132463e-05,
+      "loss": 0.1592,
+      "step": 154540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.558696919420654e-05,
+      "loss": 0.1633,
+      "step": 154550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5584727142280617e-05,
+      "loss": 0.1672,
+      "step": 154560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5582485090354693e-05,
+      "loss": 0.1654,
+      "step": 154570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.558024303842877e-05,
+      "loss": 0.166,
+      "step": 154580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5578000986502847e-05,
+      "loss": 0.1648,
+      "step": 154590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5575758934576925e-05,
+      "loss": 0.1664,
+      "step": 154600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5573516882651004e-05,
+      "loss": 0.1653,
+      "step": 154610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.557127483072508e-05,
+      "loss": 0.1647,
+      "step": 154620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5569032778799158e-05,
+      "loss": 0.165,
+      "step": 154630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5566790726873234e-05,
+      "loss": 0.1596,
+      "step": 154640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5564548674947312e-05,
+      "loss": 0.1677,
+      "step": 154650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5562306623021388e-05,
+      "loss": 0.163,
+      "step": 154660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5560064571095466e-05,
+      "loss": 0.1654,
+      "step": 154670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5557822519169545e-05,
+      "loss": 0.1656,
+      "step": 154680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.555558046724362e-05,
+      "loss": 0.165,
+      "step": 154690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.55533384153177e-05,
+      "loss": 0.1604,
+      "step": 154700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5551096363391774e-05,
+      "loss": 0.1629,
+      "step": 154710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5548854311465853e-05,
+      "loss": 0.1635,
+      "step": 154720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5546612259539932e-05,
+      "loss": 0.1578,
+      "step": 154730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.554437020761401e-05,
+      "loss": 0.1694,
+      "step": 154740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.554212815568809e-05,
+      "loss": 0.1659,
+      "step": 154750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5539886103762165e-05,
+      "loss": 0.1615,
+      "step": 154760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5537644051836243e-05,
+      "loss": 0.1664,
+      "step": 154770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.553540199991032e-05,
+      "loss": 0.1594,
+      "step": 154780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5533159947984397e-05,
+      "loss": 0.1692,
+      "step": 154790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5530917896058473e-05,
+      "loss": 0.16,
+      "step": 154800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.552867584413255e-05,
+      "loss": 0.1612,
+      "step": 154810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.552643379220663e-05,
+      "loss": 0.163,
+      "step": 154820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5524191740280706e-05,
+      "loss": 0.1666,
+      "step": 154830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5521949688354784e-05,
+      "loss": 0.1657,
+      "step": 154840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.551970763642886e-05,
+      "loss": 0.1607,
+      "step": 154850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5517465584502938e-05,
+      "loss": 0.1685,
+      "step": 154860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5515223532577014e-05,
+      "loss": 0.1594,
+      "step": 154870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5512981480651092e-05,
+      "loss": 0.1615,
+      "step": 154880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.551073942872517e-05,
+      "loss": 0.1679,
+      "step": 154890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5508497376799246e-05,
+      "loss": 0.161,
+      "step": 154900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5506255324873325e-05,
+      "loss": 0.1593,
+      "step": 154910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.55040132729474e-05,
+      "loss": 0.1633,
+      "step": 154920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.550177122102148e-05,
+      "loss": 0.1605,
+      "step": 154930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5499529169095554e-05,
+      "loss": 0.1655,
+      "step": 154940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5497287117169633e-05,
+      "loss": 0.1666,
+      "step": 154950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5495045065243712e-05,
+      "loss": 0.1627,
+      "step": 154960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5492803013317787e-05,
+      "loss": 0.167,
+      "step": 154970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5490560961391866e-05,
+      "loss": 0.1611,
+      "step": 154980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5488318909465945e-05,
+      "loss": 0.157,
+      "step": 154990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5486076857540023e-05,
+      "loss": 0.1654,
+      "step": 155000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.54838348056141e-05,
+      "loss": 0.1612,
+      "step": 155010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5481592753688177e-05,
+      "loss": 0.1604,
+      "step": 155020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5479350701762256e-05,
+      "loss": 0.1623,
+      "step": 155030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.547710864983633e-05,
+      "loss": 0.1654,
+      "step": 155040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.547486659791041e-05,
+      "loss": 0.1617,
+      "step": 155050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5472624545984486e-05,
+      "loss": 0.165,
+      "step": 155060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5470382494058564e-05,
+      "loss": 0.1654,
+      "step": 155070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.546814044213264e-05,
+      "loss": 0.1625,
+      "step": 155080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5465898390206718e-05,
+      "loss": 0.1687,
+      "step": 155090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5463656338280797e-05,
+      "loss": 0.1672,
+      "step": 155100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5461414286354872e-05,
+      "loss": 0.1656,
+      "step": 155110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.545917223442895e-05,
+      "loss": 0.1668,
+      "step": 155120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5456930182503026e-05,
+      "loss": 0.1636,
+      "step": 155130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5454688130577105e-05,
+      "loss": 0.1723,
+      "step": 155140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.545244607865118e-05,
+      "loss": 0.1646,
+      "step": 155150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.545020402672526e-05,
+      "loss": 0.1564,
+      "step": 155160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5447961974799338e-05,
+      "loss": 0.1621,
+      "step": 155170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5445719922873413e-05,
+      "loss": 0.1635,
+      "step": 155180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5443477870947492e-05,
+      "loss": 0.1647,
+      "step": 155190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5441235819021567e-05,
+      "loss": 0.1606,
+      "step": 155200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5438993767095646e-05,
+      "loss": 0.164,
+      "step": 155210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543675171516972e-05,
+      "loss": 0.1638,
+      "step": 155220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.54345096632438e-05,
+      "loss": 0.1695,
+      "step": 155230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543226761131788e-05,
+      "loss": 0.164,
+      "step": 155240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5430025559391954e-05,
+      "loss": 0.1626,
+      "step": 155250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5427783507466033e-05,
+      "loss": 0.1654,
+      "step": 155260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542554145554011e-05,
+      "loss": 0.1664,
+      "step": 155270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542329940361419e-05,
+      "loss": 0.1602,
+      "step": 155280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5421057351688266e-05,
+      "loss": 0.1621,
+      "step": 155290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5418815299762344e-05,
+      "loss": 0.1669,
+      "step": 155300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5416573247836423e-05,
+      "loss": 0.1664,
+      "step": 155310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5414331195910498e-05,
+      "loss": 0.1676,
+      "step": 155320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5412089143984577e-05,
+      "loss": 0.1658,
+      "step": 155330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5409847092058652e-05,
+      "loss": 0.1699,
+      "step": 155340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540760504013273e-05,
+      "loss": 0.1609,
+      "step": 155350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5405362988206806e-05,
+      "loss": 0.163,
+      "step": 155360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5403120936280885e-05,
+      "loss": 0.1659,
+      "step": 155370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5400878884354964e-05,
+      "loss": 0.1678,
+      "step": 155380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539863683242904e-05,
+      "loss": 0.1637,
+      "step": 155390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5396394780503118e-05,
+      "loss": 0.1636,
+      "step": 155400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5394152728577193e-05,
+      "loss": 0.1629,
+      "step": 155410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5391910676651272e-05,
+      "loss": 0.1614,
+      "step": 155420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538966862472535e-05,
+      "loss": 0.1609,
+      "step": 155430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5387426572799426e-05,
+      "loss": 0.1637,
+      "step": 155440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5385184520873505e-05,
+      "loss": 0.1663,
+      "step": 155450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538294246894758e-05,
+      "loss": 0.1642,
+      "step": 155460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538070041702166e-05,
+      "loss": 0.1655,
+      "step": 155470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5378458365095734e-05,
+      "loss": 0.1625,
+      "step": 155480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5376216313169813e-05,
+      "loss": 0.1671,
+      "step": 155490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537397426124389e-05,
+      "loss": 0.166,
+      "step": 155500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5371732209317967e-05,
+      "loss": 0.1637,
+      "step": 155510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5369490157392046e-05,
+      "loss": 0.1647,
+      "step": 155520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5367248105466124e-05,
+      "loss": 0.1645,
+      "step": 155530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.53650060535402e-05,
+      "loss": 0.1689,
+      "step": 155540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.536276400161428e-05,
+      "loss": 0.1686,
+      "step": 155550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5360521949688357e-05,
+      "loss": 0.1624,
+      "step": 155560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5358279897762436e-05,
+      "loss": 0.1556,
+      "step": 155570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.535603784583651e-05,
+      "loss": 0.1646,
+      "step": 155580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.535379579391059e-05,
+      "loss": 0.168,
+      "step": 155590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5351553741984665e-05,
+      "loss": 0.1627,
+      "step": 155600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5349311690058744e-05,
+      "loss": 0.1586,
+      "step": 155610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.534706963813282e-05,
+      "loss": 0.166,
+      "step": 155620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5344827586206898e-05,
+      "loss": 0.1584,
+      "step": 155630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5342585534280977e-05,
+      "loss": 0.1624,
+      "step": 155640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5340343482355052e-05,
+      "loss": 0.1681,
+      "step": 155650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.533810143042913e-05,
+      "loss": 0.1651,
+      "step": 155660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5335859378503206e-05,
+      "loss": 0.1609,
+      "step": 155670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5333617326577285e-05,
+      "loss": 0.1605,
+      "step": 155680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533137527465136e-05,
+      "loss": 0.1621,
+      "step": 155690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532913322272544e-05,
+      "loss": 0.1645,
+      "step": 155700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5326891170799517e-05,
+      "loss": 0.1617,
+      "step": 155710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5324649118873593e-05,
+      "loss": 0.161,
+      "step": 155720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532240706694767e-05,
+      "loss": 0.1591,
+      "step": 155730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5320165015021747e-05,
+      "loss": 0.1669,
+      "step": 155740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5317922963095826e-05,
+      "loss": 0.1639,
+      "step": 155750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53156809111699e-05,
+      "loss": 0.1634,
+      "step": 155760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531343885924398e-05,
+      "loss": 0.1698,
+      "step": 155770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531119680731806e-05,
+      "loss": 0.1607,
+      "step": 155780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5308954755392134e-05,
+      "loss": 0.1591,
+      "step": 155790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5306712703466212e-05,
+      "loss": 0.1613,
+      "step": 155800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530447065154029e-05,
+      "loss": 0.1587,
+      "step": 155810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530222859961437e-05,
+      "loss": 0.1654,
+      "step": 155820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5299986547688445e-05,
+      "loss": 0.1701,
+      "step": 155830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5297744495762524e-05,
+      "loss": 0.1619,
+      "step": 155840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5295502443836603e-05,
+      "loss": 0.1679,
+      "step": 155850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5293260391910678e-05,
+      "loss": 0.1614,
+      "step": 155860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5291018339984757e-05,
+      "loss": 0.1617,
+      "step": 155870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5288776288058832e-05,
+      "loss": 0.1634,
+      "step": 155880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.528653423613291e-05,
+      "loss": 0.1662,
+      "step": 155890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5284292184206986e-05,
+      "loss": 0.1692,
+      "step": 155900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5282050132281065e-05,
+      "loss": 0.1643,
+      "step": 155910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5279808080355143e-05,
+      "loss": 0.1647,
+      "step": 155920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.527756602842922e-05,
+      "loss": 0.1706,
+      "step": 155930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5275323976503297e-05,
+      "loss": 0.1615,
+      "step": 155940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5273081924577373e-05,
+      "loss": 0.1648,
+      "step": 155950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.527083987265145e-05,
+      "loss": 0.1667,
+      "step": 155960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5268597820725527e-05,
+      "loss": 0.1663,
+      "step": 155970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5266355768799606e-05,
+      "loss": 0.164,
+      "step": 155980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5264113716873684e-05,
+      "loss": 0.1621,
+      "step": 155990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.526187166494776e-05,
+      "loss": 0.1644,
+      "step": 156000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.525962961302184e-05,
+      "loss": 0.1692,
+      "step": 156010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5257387561095915e-05,
+      "loss": 0.1682,
+      "step": 156020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5255145509169994e-05,
+      "loss": 0.1591,
+      "step": 156030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.525290345724407e-05,
+      "loss": 0.1609,
+      "step": 156040
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5250661405318148e-05,
+      "loss": 0.1646,
+      "step": 156050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5248419353392227e-05,
+      "loss": 0.1634,
+      "step": 156060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5246177301466302e-05,
+      "loss": 0.1618,
+      "step": 156070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5243935249540381e-05,
+      "loss": 0.1654,
+      "step": 156080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5241693197614456e-05,
+      "loss": 0.163,
+      "step": 156090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5239451145688535e-05,
+      "loss": 0.1636,
+      "step": 156100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.523720909376261e-05,
+      "loss": 0.1645,
+      "step": 156110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5234967041836689e-05,
+      "loss": 0.1633,
+      "step": 156120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5232724989910768e-05,
+      "loss": 0.1705,
+      "step": 156130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5230482937984845e-05,
+      "loss": 0.1633,
+      "step": 156140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5228240886058923e-05,
+      "loss": 0.1684,
+      "step": 156150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5225998834132999e-05,
+      "loss": 0.1656,
+      "step": 156160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5223756782207078e-05,
+      "loss": 0.1599,
+      "step": 156170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5221514730281153e-05,
+      "loss": 0.1639,
+      "step": 156180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5219272678355232e-05,
+      "loss": 0.1616,
+      "step": 156190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.521703062642931e-05,
+      "loss": 0.1614,
+      "step": 156200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5214788574503386e-05,
+      "loss": 0.1664,
+      "step": 156210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5212546522577464e-05,
+      "loss": 0.1641,
+      "step": 156220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.521030447065154e-05,
+      "loss": 0.1623,
+      "step": 156230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5208062418725618e-05,
+      "loss": 0.1587,
+      "step": 156240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5205820366799695e-05,
+      "loss": 0.16,
+      "step": 156250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5203578314873772e-05,
+      "loss": 0.1684,
+      "step": 156260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5201336262947851e-05,
+      "loss": 0.1615,
+      "step": 156270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5199094211021928e-05,
+      "loss": 0.1656,
+      "step": 156280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5196852159096007e-05,
+      "loss": 0.1659,
+      "step": 156290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5194610107170082e-05,
+      "loss": 0.1668,
+      "step": 156300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5192368055244161e-05,
+      "loss": 0.1721,
+      "step": 156310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5190126003318236e-05,
+      "loss": 0.1641,
+      "step": 156320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5187883951392315e-05,
+      "loss": 0.1615,
+      "step": 156330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5185641899466394e-05,
+      "loss": 0.1676,
+      "step": 156340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5183399847540469e-05,
+      "loss": 0.159,
+      "step": 156350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5181157795614548e-05,
+      "loss": 0.1583,
+      "step": 156360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5178915743688623e-05,
+      "loss": 0.1675,
+      "step": 156370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5176673691762702e-05,
+      "loss": 0.1668,
+      "step": 156380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5174431639836779e-05,
+      "loss": 0.1626,
+      "step": 156390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5172189587910856e-05,
+      "loss": 0.1652,
+      "step": 156400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5169947535984935e-05,
+      "loss": 0.1629,
+      "step": 156410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5167705484059012e-05,
+      "loss": 0.1603,
+      "step": 156420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.516546343213309e-05,
+      "loss": 0.1636,
+      "step": 156430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5163221380207166e-05,
+      "loss": 0.166,
+      "step": 156440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5160979328281244e-05,
+      "loss": 0.1706,
+      "step": 156450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.515873727635532e-05,
+      "loss": 0.164,
+      "step": 156460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5156495224429398e-05,
+      "loss": 0.1676,
+      "step": 156470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5154253172503477e-05,
+      "loss": 0.1644,
+      "step": 156480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5152011120577552e-05,
+      "loss": 0.1636,
+      "step": 156490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5149769068651631e-05,
+      "loss": 0.1636,
+      "step": 156500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5147527016725706e-05,
+      "loss": 0.1641,
+      "step": 156510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5145284964799785e-05,
+      "loss": 0.1614,
+      "step": 156520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5143042912873862e-05,
+      "loss": 0.1622,
+      "step": 156530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5140800860947941e-05,
+      "loss": 0.1694,
+      "step": 156540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5138558809022018e-05,
+      "loss": 0.1568,
+      "step": 156550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5136316757096095e-05,
+      "loss": 0.1618,
+      "step": 156560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5134074705170174e-05,
+      "loss": 0.1609,
+      "step": 156570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5131832653244249e-05,
+      "loss": 0.1631,
+      "step": 156580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5129590601318328e-05,
+      "loss": 0.1594,
+      "step": 156590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5127348549392403e-05,
+      "loss": 0.1635,
+      "step": 156600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5125106497466482e-05,
+      "loss": 0.1604,
+      "step": 156610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.512286444554056e-05,
+      "loss": 0.1652,
+      "step": 156620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5120622393614636e-05,
+      "loss": 0.1623,
+      "step": 156630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5118380341688715e-05,
+      "loss": 0.1656,
+      "step": 156640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.511613828976279e-05,
+      "loss": 0.1673,
+      "step": 156650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5113896237836869e-05,
+      "loss": 0.1724,
+      "step": 156660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5111654185910946e-05,
+      "loss": 0.1638,
+      "step": 156670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5109412133985024e-05,
+      "loss": 0.1708,
+      "step": 156680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5107170082059103e-05,
+      "loss": 0.1675,
+      "step": 156690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5104928030133178e-05,
+      "loss": 0.1593,
+      "step": 156700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5102685978207257e-05,
+      "loss": 0.1586,
+      "step": 156710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5100443926281332e-05,
+      "loss": 0.1639,
+      "step": 156720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5098201874355411e-05,
+      "loss": 0.1628,
+      "step": 156730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5095959822429486e-05,
+      "loss": 0.1615,
+      "step": 156740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5093717770503565e-05,
+      "loss": 0.1649,
+      "step": 156750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5091475718577644e-05,
+      "loss": 0.1704,
+      "step": 156760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.508923366665172e-05,
+      "loss": 0.1645,
+      "step": 156770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5086991614725798e-05,
+      "loss": 0.1665,
+      "step": 156780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5084749562799875e-05,
+      "loss": 0.1579,
+      "step": 156790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5082507510873952e-05,
+      "loss": 0.1634,
+      "step": 156800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.508026545894803e-05,
+      "loss": 0.1593,
+      "step": 156810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5078023407022108e-05,
+      "loss": 0.1663,
+      "step": 156820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5075781355096186e-05,
+      "loss": 0.167,
+      "step": 156830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5073539303170262e-05,
+      "loss": 0.1654,
+      "step": 156840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.507129725124434e-05,
+      "loss": 0.1609,
+      "step": 156850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5069055199318416e-05,
+      "loss": 0.1624,
+      "step": 156860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5066813147392495e-05,
+      "loss": 0.164,
+      "step": 156870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5064571095466573e-05,
+      "loss": 0.1597,
+      "step": 156880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5062329043540649e-05,
+      "loss": 0.162,
+      "step": 156890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5060086991614727e-05,
+      "loss": 0.166,
+      "step": 156900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5057844939688803e-05,
+      "loss": 0.1681,
+      "step": 156910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5055602887762881e-05,
+      "loss": 0.1712,
+      "step": 156920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.505358504102955e-05,
+      "loss": 0.1668,
+      "step": 156930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5051342989103628e-05,
+      "loss": 0.1647,
+      "step": 156940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5049100937177707e-05,
+      "loss": 0.1609,
+      "step": 156950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5046858885251782e-05,
+      "loss": 0.1657,
+      "step": 156960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.504461683332586e-05,
+      "loss": 0.1584,
+      "step": 156970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5042374781399938e-05,
+      "loss": 0.1658,
+      "step": 156980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5040132729474016e-05,
+      "loss": 0.1679,
+      "step": 156990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5037890677548092e-05,
+      "loss": 0.1628,
+      "step": 157000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.503564862562217e-05,
+      "loss": 0.1618,
+      "step": 157010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.503340657369625e-05,
+      "loss": 0.1599,
+      "step": 157020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5031164521770324e-05,
+      "loss": 0.1648,
+      "step": 157030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5028922469844403e-05,
+      "loss": 0.1608,
+      "step": 157040
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5026680417918479e-05,
+      "loss": 0.1655,
+      "step": 157050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5024438365992557e-05,
+      "loss": 0.1636,
+      "step": 157060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5022196314066633e-05,
+      "loss": 0.1626,
+      "step": 157070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5019954262140711e-05,
+      "loss": 0.1591,
+      "step": 157080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.501771221021479e-05,
+      "loss": 0.1586,
+      "step": 157090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5015470158288867e-05,
+      "loss": 0.1646,
+      "step": 157100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5013228106362944e-05,
+      "loss": 0.1623,
+      "step": 157110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5010986054437021e-05,
+      "loss": 0.1632,
+      "step": 157120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.50087440025111e-05,
+      "loss": 0.1648,
+      "step": 157130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5006501950585175e-05,
+      "loss": 0.1639,
+      "step": 157140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5004259898659254e-05,
+      "loss": 0.1612,
+      "step": 157150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5002017846733333e-05,
+      "loss": 0.1641,
+      "step": 157160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4999775794807408e-05,
+      "loss": 0.1632,
+      "step": 157170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4997533742881487e-05,
+      "loss": 0.1588,
+      "step": 157180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4995291690955562e-05,
+      "loss": 0.1609,
+      "step": 157190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.499304963902964e-05,
+      "loss": 0.1593,
+      "step": 157200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4990807587103716e-05,
+      "loss": 0.1655,
+      "step": 157210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4988565535177795e-05,
+      "loss": 0.164,
+      "step": 157220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4986323483251873e-05,
+      "loss": 0.1659,
+      "step": 157230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.498408143132595e-05,
+      "loss": 0.1546,
+      "step": 157240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.498183937940003e-05,
+      "loss": 0.1636,
+      "step": 157250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4979597327474105e-05,
+      "loss": 0.1588,
+      "step": 157260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4977355275548183e-05,
+      "loss": 0.1686,
+      "step": 157270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4975113223622259e-05,
+      "loss": 0.1634,
+      "step": 157280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4972871171696337e-05,
+      "loss": 0.1648,
+      "step": 157290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4970629119770416e-05,
+      "loss": 0.1599,
+      "step": 157300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4968387067844491e-05,
+      "loss": 0.1597,
+      "step": 157310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.496614501591857e-05,
+      "loss": 0.168,
+      "step": 157320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4963902963992645e-05,
+      "loss": 0.1594,
+      "step": 157330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4961660912066724e-05,
+      "loss": 0.159,
+      "step": 157340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4959418860140801e-05,
+      "loss": 0.1596,
+      "step": 157350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4957176808214878e-05,
+      "loss": 0.1656,
+      "step": 157360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4954934756288957e-05,
+      "loss": 0.1638,
+      "step": 157370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4952692704363034e-05,
+      "loss": 0.1617,
+      "step": 157380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4950450652437113e-05,
+      "loss": 0.1594,
+      "step": 157390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4948208600511188e-05,
+      "loss": 0.1578,
+      "step": 157400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4945966548585267e-05,
+      "loss": 0.1667,
+      "step": 157410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4943724496659342e-05,
+      "loss": 0.1627,
+      "step": 157420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.494148244473342e-05,
+      "loss": 0.1677,
+      "step": 157430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.49392403928075e-05,
+      "loss": 0.1601,
+      "step": 157440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4936998340881575e-05,
+      "loss": 0.1584,
+      "step": 157450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4934756288955653e-05,
+      "loss": 0.163,
+      "step": 157460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4932514237029729e-05,
+      "loss": 0.1597,
+      "step": 157470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4930272185103808e-05,
+      "loss": 0.1668,
+      "step": 157480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4928030133177885e-05,
+      "loss": 0.1609,
+      "step": 157490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4925788081251962e-05,
+      "loss": 0.1609,
+      "step": 157500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.492354602932604e-05,
+      "loss": 0.1609,
+      "step": 157510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4921303977400117e-05,
+      "loss": 0.162,
+      "step": 157520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4919061925474196e-05,
+      "loss": 0.1638,
+      "step": 157530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4916819873548271e-05,
+      "loss": 0.1603,
+      "step": 157540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.491457782162235e-05,
+      "loss": 0.1616,
+      "step": 157550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4912335769696425e-05,
+      "loss": 0.163,
+      "step": 157560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4910093717770504e-05,
+      "loss": 0.1668,
+      "step": 157570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4907851665844583e-05,
+      "loss": 0.1689,
+      "step": 157580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4905609613918658e-05,
+      "loss": 0.1647,
+      "step": 157590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4903367561992737e-05,
+      "loss": 0.1602,
+      "step": 157600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4901125510066812e-05,
+      "loss": 0.1629,
+      "step": 157610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4898883458140891e-05,
+      "loss": 0.1652,
+      "step": 157620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4896641406214968e-05,
+      "loss": 0.1626,
+      "step": 157630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4894399354289047e-05,
+      "loss": 0.1625,
+      "step": 157640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4892157302363124e-05,
+      "loss": 0.1591,
+      "step": 157650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.48899152504372e-05,
+      "loss": 0.1614,
+      "step": 157660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.488767319851128e-05,
+      "loss": 0.1629,
+      "step": 157670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4885431146585355e-05,
+      "loss": 0.1572,
+      "step": 157680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4883189094659433e-05,
+      "loss": 0.1682,
+      "step": 157690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4880947042733509e-05,
+      "loss": 0.1617,
+      "step": 157700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4878704990807588e-05,
+      "loss": 0.1622,
+      "step": 157710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4876462938881666e-05,
+      "loss": 0.1607,
+      "step": 157720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4874220886955742e-05,
+      "loss": 0.1578,
+      "step": 157730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.487197883502982e-05,
+      "loss": 0.159,
+      "step": 157740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4869736783103896e-05,
+      "loss": 0.1606,
+      "step": 157750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4867494731177974e-05,
+      "loss": 0.163,
+      "step": 157760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4865252679252053e-05,
+      "loss": 0.1627,
+      "step": 157770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.486301062732613e-05,
+      "loss": 0.1658,
+      "step": 157780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4860768575400209e-05,
+      "loss": 0.1612,
+      "step": 157790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4858526523474284e-05,
+      "loss": 0.1694,
+      "step": 157800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4856284471548363e-05,
+      "loss": 0.1703,
+      "step": 157810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4854042419622438e-05,
+      "loss": 0.1635,
+      "step": 157820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4851800367696517e-05,
+      "loss": 0.1684,
+      "step": 157830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4849558315770596e-05,
+      "loss": 0.158,
+      "step": 157840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4847316263844671e-05,
+      "loss": 0.1645,
+      "step": 157850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.484507421191875e-05,
+      "loss": 0.1683,
+      "step": 157860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4842832159992825e-05,
+      "loss": 0.1627,
+      "step": 157870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4840590108066904e-05,
+      "loss": 0.1671,
+      "step": 157880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.483834805614098e-05,
+      "loss": 0.1656,
+      "step": 157890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4836106004215058e-05,
+      "loss": 0.1625,
+      "step": 157900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4833863952289136e-05,
+      "loss": 0.1656,
+      "step": 157910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.4831621900363213e-05,
+      "loss": 0.1588,
+      "step": 157920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4829379848437292e-05,
+      "loss": 0.1652,
+      "step": 157930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4827137796511368e-05,
+      "loss": 0.1612,
+      "step": 157940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4824895744585446e-05,
+      "loss": 0.1555,
+      "step": 157950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4822653692659522e-05,
+      "loss": 0.1623,
+      "step": 157960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.48204116407336e-05,
+      "loss": 0.1624,
+      "step": 157970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4818169588807679e-05,
+      "loss": 0.167,
+      "step": 157980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4815927536881754e-05,
+      "loss": 0.1652,
+      "step": 157990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4813685484955833e-05,
+      "loss": 0.1627,
+      "step": 158000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4811443433029908e-05,
+      "loss": 0.1615,
+      "step": 158010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4809201381103987e-05,
+      "loss": 0.1698,
+      "step": 158020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4806959329178064e-05,
+      "loss": 0.1601,
+      "step": 158030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4804717277252141e-05,
+      "loss": 0.162,
+      "step": 158040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.480247522532622e-05,
+      "loss": 0.1628,
+      "step": 158050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4800233173400297e-05,
+      "loss": 0.1643,
+      "step": 158060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4797991121474376e-05,
+      "loss": 0.1632,
+      "step": 158070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4795749069548451e-05,
+      "loss": 0.1682,
+      "step": 158080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.479350701762253e-05,
+      "loss": 0.1624,
+      "step": 158090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4791264965696605e-05,
+      "loss": 0.1595,
+      "step": 158100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4789022913770684e-05,
+      "loss": 0.1609,
+      "step": 158110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4786780861844762e-05,
+      "loss": 0.162,
+      "step": 158120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4784538809918838e-05,
+      "loss": 0.166,
+      "step": 158130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4782296757992916e-05,
+      "loss": 0.1589,
+      "step": 158140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4780054706066992e-05,
+      "loss": 0.1658,
+      "step": 158150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.477781265414107e-05,
+      "loss": 0.1615,
+      "step": 158160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4775570602215148e-05,
+      "loss": 0.1624,
+      "step": 158170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4773328550289226e-05,
+      "loss": 0.1635,
+      "step": 158180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4771086498363303e-05,
+      "loss": 0.1628,
+      "step": 158190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.476884444643738e-05,
+      "loss": 0.1581,
+      "step": 158200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4766602394511459e-05,
+      "loss": 0.1707,
+      "step": 158210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4764360342585534e-05,
+      "loss": 0.1651,
+      "step": 158220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4762118290659613e-05,
+      "loss": 0.1626,
+      "step": 158230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4759876238733688e-05,
+      "loss": 0.1635,
+      "step": 158240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4757634186807767e-05,
+      "loss": 0.1616,
+      "step": 158250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4755392134881846e-05,
+      "loss": 0.163,
+      "step": 158260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4753150082955921e-05,
+      "loss": 0.167,
+      "step": 158270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.475090803103e-05,
+      "loss": 0.1558,
+      "step": 158280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4748665979104075e-05,
+      "loss": 0.1616,
+      "step": 158290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4746423927178154e-05,
+      "loss": 0.1634,
+      "step": 158300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4744181875252231e-05,
+      "loss": 0.1641,
+      "step": 158310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.474193982332631e-05,
+      "loss": 0.1663,
+      "step": 158320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4739697771400387e-05,
+      "loss": 0.161,
+      "step": 158330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4737455719474464e-05,
+      "loss": 0.163,
+      "step": 158340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4735213667548542e-05,
+      "loss": 0.1641,
+      "step": 158350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4732971615622618e-05,
+      "loss": 0.1633,
+      "step": 158360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4730729563696696e-05,
+      "loss": 0.1637,
+      "step": 158370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4728487511770772e-05,
+      "loss": 0.1606,
+      "step": 158380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.472624545984485e-05,
+      "loss": 0.1599,
+      "step": 158390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.472400340791893e-05,
+      "loss": 0.1664,
+      "step": 158400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4721761355993005e-05,
+      "loss": 0.1632,
+      "step": 158410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4719519304067083e-05,
+      "loss": 0.1632,
+      "step": 158420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.471727725214116e-05,
+      "loss": 0.1674,
+      "step": 158430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4715035200215237e-05,
+      "loss": 0.162,
+      "step": 158440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4712793148289314e-05,
+      "loss": 0.167,
+      "step": 158450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4710551096363393e-05,
+      "loss": 0.1598,
+      "step": 158460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4708309044437472e-05,
+      "loss": 0.1609,
+      "step": 158470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4706066992511547e-05,
+      "loss": 0.1622,
+      "step": 158480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4703824940585626e-05,
+      "loss": 0.1675,
+      "step": 158490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4701582888659701e-05,
+      "loss": 0.1645,
+      "step": 158500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.469934083673378e-05,
+      "loss": 0.164,
+      "step": 158510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4697098784807855e-05,
+      "loss": 0.1636,
+      "step": 158520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4694856732881934e-05,
+      "loss": 0.165,
+      "step": 158530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4692614680956013e-05,
+      "loss": 0.16,
+      "step": 158540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4690372629030088e-05,
+      "loss": 0.1637,
+      "step": 158550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4688130577104167e-05,
+      "loss": 0.158,
+      "step": 158560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4685888525178244e-05,
+      "loss": 0.158,
+      "step": 158570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.468364647325232e-05,
+      "loss": 0.1626,
+      "step": 158580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4681404421326398e-05,
+      "loss": 0.1628,
+      "step": 158590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4679162369400477e-05,
+      "loss": 0.1646,
+      "step": 158600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4676920317474555e-05,
+      "loss": 0.1598,
+      "step": 158610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.467467826554863e-05,
+      "loss": 0.1577,
+      "step": 158620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.467243621362271e-05,
+      "loss": 0.1577,
+      "step": 158630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4670194161696785e-05,
+      "loss": 0.1596,
+      "step": 158640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4667952109770863e-05,
+      "loss": 0.1607,
+      "step": 158650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4665710057844939e-05,
+      "loss": 0.162,
+      "step": 158660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4663468005919017e-05,
+      "loss": 0.1609,
+      "step": 158670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4661225953993096e-05,
+      "loss": 0.1641,
+      "step": 158680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4658983902067171e-05,
+      "loss": 0.1557,
+      "step": 158690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.465674185014125e-05,
+      "loss": 0.1629,
+      "step": 158700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4654499798215327e-05,
+      "loss": 0.1591,
+      "step": 158710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4652257746289406e-05,
+      "loss": 0.1659,
+      "step": 158720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4650015694363481e-05,
+      "loss": 0.1618,
+      "step": 158730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.464777364243756e-05,
+      "loss": 0.1645,
+      "step": 158740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4645531590511639e-05,
+      "loss": 0.1668,
+      "step": 158750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4643289538585714e-05,
+      "loss": 0.1637,
+      "step": 158760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4641047486659793e-05,
+      "loss": 0.1635,
+      "step": 158770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4638805434733868e-05,
+      "loss": 0.1662,
+      "step": 158780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4636563382807947e-05,
+      "loss": 0.1664,
+      "step": 158790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4634321330882022e-05,
+      "loss": 0.159,
+      "step": 158800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.46320792789561e-05,
+      "loss": 0.1709,
+      "step": 158810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.462983722703018e-05,
+      "loss": 0.1636,
+      "step": 158820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4627595175104255e-05,
+      "loss": 0.1599,
+      "step": 158830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4625353123178334e-05,
+      "loss": 0.1622,
+      "step": 158840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.462311107125241e-05,
+      "loss": 0.1634,
+      "step": 158850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.462086901932649e-05,
+      "loss": 0.1626,
+      "step": 158860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4618626967400565e-05,
+      "loss": 0.1673,
+      "step": 158870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4616384915474643e-05,
+      "loss": 0.1609,
+      "step": 158880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4614142863548722e-05,
+      "loss": 0.1573,
+      "step": 158890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4611900811622797e-05,
+      "loss": 0.1683,
+      "step": 158900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4609658759696876e-05,
+      "loss": 0.1635,
+      "step": 158910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4607416707770951e-05,
+      "loss": 0.1611,
+      "step": 158920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.460517465584503e-05,
+      "loss": 0.1572,
+      "step": 158930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4602932603919105e-05,
+      "loss": 0.1665,
+      "step": 158940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4600690551993184e-05,
+      "loss": 0.159,
+      "step": 158950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4598448500067263e-05,
+      "loss": 0.1661,
+      "step": 158960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4596206448141338e-05,
+      "loss": 0.1585,
+      "step": 158970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4593964396215417e-05,
+      "loss": 0.1554,
+      "step": 158980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4591722344289494e-05,
+      "loss": 0.1585,
+      "step": 158990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4589480292363573e-05,
+      "loss": 0.1596,
+      "step": 159000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4587238240437648e-05,
+      "loss": 0.1633,
+      "step": 159010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4584996188511727e-05,
+      "loss": 0.1634,
+      "step": 159020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4582754136585805e-05,
+      "loss": 0.1631,
+      "step": 159030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.458051208465988e-05,
+      "loss": 0.1619,
+      "step": 159040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.457827003273396e-05,
+      "loss": 0.1582,
+      "step": 159050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4576027980808035e-05,
+      "loss": 0.1614,
+      "step": 159060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4573785928882114e-05,
+      "loss": 0.1643,
+      "step": 159070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4571543876956192e-05,
+      "loss": 0.1586,
+      "step": 159080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4569301825030268e-05,
+      "loss": 0.1661,
+      "step": 159090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4567059773104346e-05,
+      "loss": 0.1673,
+      "step": 159100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4564817721178423e-05,
+      "loss": 0.1614,
+      "step": 159110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.45625756692525e-05,
+      "loss": 0.1645,
+      "step": 159120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4560333617326577e-05,
+      "loss": 0.1627,
+      "step": 159130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4558091565400656e-05,
+      "loss": 0.1599,
+      "step": 159140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4555849513474735e-05,
+      "loss": 0.1621,
+      "step": 159150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.455360746154881e-05,
+      "loss": 0.1626,
+      "step": 159160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4551365409622889e-05,
+      "loss": 0.1656,
+      "step": 159170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4549123357696964e-05,
+      "loss": 0.1585,
+      "step": 159180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4546881305771043e-05,
+      "loss": 0.1612,
+      "step": 159190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4544639253845118e-05,
+      "loss": 0.1615,
+      "step": 159200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4542397201919197e-05,
+      "loss": 0.1608,
+      "step": 159210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4540155149993276e-05,
+      "loss": 0.1649,
+      "step": 159220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4537913098067351e-05,
+      "loss": 0.1641,
+      "step": 159230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.453567104614143e-05,
+      "loss": 0.1565,
+      "step": 159240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4533428994215507e-05,
+      "loss": 0.1599,
+      "step": 159250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4531186942289585e-05,
+      "loss": 0.1579,
+      "step": 159260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.452894489036366e-05,
+      "loss": 0.1593,
+      "step": 159270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.452670283843774e-05,
+      "loss": 0.1549,
+      "step": 159280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4524460786511818e-05,
+      "loss": 0.1655,
+      "step": 159290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4522218734585894e-05,
+      "loss": 0.1614,
+      "step": 159300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4519976682659972e-05,
+      "loss": 0.1604,
+      "step": 159310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.451795883592664e-05,
+      "loss": 0.1627,
+      "step": 159320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4515716784000719e-05,
+      "loss": 0.1645,
+      "step": 159330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4513474732074794e-05,
+      "loss": 0.1667,
+      "step": 159340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4511232680148873e-05,
+      "loss": 0.1646,
+      "step": 159350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4508990628222952e-05,
+      "loss": 0.1601,
+      "step": 159360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4506748576297027e-05,
+      "loss": 0.1608,
+      "step": 159370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4504506524371106e-05,
+      "loss": 0.165,
+      "step": 159380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4502264472445181e-05,
+      "loss": 0.161,
+      "step": 159390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.450002242051926e-05,
+      "loss": 0.1655,
+      "step": 159400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4497780368593337e-05,
+      "loss": 0.1635,
+      "step": 159410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4495538316667415e-05,
+      "loss": 0.1622,
+      "step": 159420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4493296264741492e-05,
+      "loss": 0.1641,
+      "step": 159430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.449105421281557e-05,
+      "loss": 0.167,
+      "step": 159440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4488812160889648e-05,
+      "loss": 0.1634,
+      "step": 159450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4486570108963724e-05,
+      "loss": 0.161,
+      "step": 159460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4484328057037802e-05,
+      "loss": 0.1634,
+      "step": 159470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4482086005111878e-05,
+      "loss": 0.1629,
+      "step": 159480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4479843953185956e-05,
+      "loss": 0.1577,
+      "step": 159490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4477601901260035e-05,
+      "loss": 0.1651,
+      "step": 159500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.447535984933411e-05,
+      "loss": 0.167,
+      "step": 159510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4473117797408189e-05,
+      "loss": 0.1639,
+      "step": 159520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4470875745482264e-05,
+      "loss": 0.1627,
+      "step": 159530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4468633693556343e-05,
+      "loss": 0.1571,
+      "step": 159540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.446639164163042e-05,
+      "loss": 0.1622,
+      "step": 159550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4464149589704499e-05,
+      "loss": 0.1604,
+      "step": 159560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4461907537778578e-05,
+      "loss": 0.1617,
+      "step": 159570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4459665485852653e-05,
+      "loss": 0.1635,
+      "step": 159580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4457423433926732e-05,
+      "loss": 0.1642,
+      "step": 159590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4455181382000807e-05,
+      "loss": 0.1641,
+      "step": 159600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4452939330074886e-05,
+      "loss": 0.1605,
+      "step": 159610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4450697278148961e-05,
+      "loss": 0.1628,
+      "step": 159620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.444845522622304e-05,
+      "loss": 0.1594,
+      "step": 159630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4446213174297118e-05,
+      "loss": 0.1654,
+      "step": 159640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4443971122371194e-05,
+      "loss": 0.1629,
+      "step": 159650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4441729070445272e-05,
+      "loss": 0.1578,
+      "step": 159660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.443948701851935e-05,
+      "loss": 0.1578,
+      "step": 159670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4437244966593426e-05,
+      "loss": 0.1586,
+      "step": 159680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4435002914667504e-05,
+      "loss": 0.1668,
+      "step": 159690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4432760862741582e-05,
+      "loss": 0.1625,
+      "step": 159700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4430518810815661e-05,
+      "loss": 0.1593,
+      "step": 159710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4428276758889736e-05,
+      "loss": 0.1647,
+      "step": 159720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4426034706963815e-05,
+      "loss": 0.1625,
+      "step": 159730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.442379265503789e-05,
+      "loss": 0.1659,
+      "step": 159740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4421550603111969e-05,
+      "loss": 0.1566,
+      "step": 159750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4419308551186044e-05,
+      "loss": 0.1601,
+      "step": 159760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4417066499260123e-05,
+      "loss": 0.1616,
+      "step": 159770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4414824447334202e-05,
+      "loss": 0.1598,
+      "step": 159780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4412582395408277e-05,
+      "loss": 0.1669,
+      "step": 159790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4410340343482356e-05,
+      "loss": 0.1601,
+      "step": 159800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4408098291556433e-05,
+      "loss": 0.1586,
+      "step": 159810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4405856239630512e-05,
+      "loss": 0.1628,
+      "step": 159820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4403614187704587e-05,
+      "loss": 0.1615,
+      "step": 159830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4401372135778666e-05,
+      "loss": 0.1648,
+      "step": 159840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4399130083852744e-05,
+      "loss": 0.1615,
+      "step": 159850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.439688803192682e-05,
+      "loss": 0.1646,
+      "step": 159860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4394645980000898e-05,
+      "loss": 0.1579,
+      "step": 159870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4392403928074974e-05,
+      "loss": 0.1568,
+      "step": 159880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4390161876149052e-05,
+      "loss": 0.1588,
+      "step": 159890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4387919824223128e-05,
+      "loss": 0.1634,
+      "step": 159900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4385677772297207e-05,
+      "loss": 0.1634,
+      "step": 159910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4383435720371285e-05,
+      "loss": 0.164,
+      "step": 159920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.438119366844536e-05,
+      "loss": 0.1646,
+      "step": 159930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.437895161651944e-05,
+      "loss": 0.1658,
+      "step": 159940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4376709564593516e-05,
+      "loss": 0.1664,
+      "step": 159950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4374467512667595e-05,
+      "loss": 0.1611,
+      "step": 159960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.437222546074167e-05,
+      "loss": 0.1603,
+      "step": 159970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4369983408815749e-05,
+      "loss": 0.1598,
+      "step": 159980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4367741356889828e-05,
+      "loss": 0.1605,
+      "step": 159990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4365499304963903e-05,
+      "loss": 0.1636,
+      "step": 160000
+    },
+    {
+      "epoch": 0.71,
+      "eval_NEIMS_canon_smiles": 0.9451666666666667,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.08185,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.4856654689384945,
+      "eval_NEIMS_exact_mols": 0.08156666666666666,
+      "eval_NEIMS_exact_smiles": 0.07951666666666667,
+      "eval_NEIMS_loss": 0.19957083463668823,
+      "eval_NEIMS_matched_formulas": 0.5902833333333334,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.39239375453426634,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.08201666666666667,
+      "eval_NEIMS_runtime": 714.7921,
+      "eval_NEIMS_samples_per_second": 83.94,
+      "eval_NEIMS_steps_per_second": 1.312,
+      "step": 160000
+    },
+    {
+      "epoch": 0.71,
+      "eval_RASSP_canon_smiles": 0.938975523424717,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.18973519885130397,
+      "eval_RASSP_daylight_tanimoto_simil": 0.627314988259476,
+      "eval_RASSP_exact_mols": 0.18903395999599293,
+      "eval_RASSP_exact_smiles": 0.1828897719304104,
+      "eval_RASSP_loss": 0.1380533128976822,
+      "eval_RASSP_matched_formulas": 0.767706281096604,
+      "eval_RASSP_morgan_tanimoto_simil": 0.504474684110161,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.1892677062810966,
+      "eval_RASSP_runtime": 815.587,
+      "eval_RASSP_samples_per_second": 73.437,
+      "eval_RASSP_steps_per_second": 1.148,
+      "step": 160000
+    },
+    {
+      "epoch": 0.71,
+      "eval_NIST_canon_smiles": 0.8837704510771196,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.013663626361926393,
+      "eval_NIST_daylight_tanimoto_simil": 0.27592106839424707,
+      "eval_NIST_exact_mols": 0.0131312772828903,
+      "eval_NIST_exact_smiles": 0.012066579124818113,
+      "eval_NIST_loss": 1.8635547161102295,
+      "eval_NIST_matched_formulas": 0.08971856478688292,
+      "eval_NIST_morgan_tanimoto_simil": 0.22686512865065844,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.013663626361926393,
+      "eval_NIST_runtime": 464.2492,
+      "eval_NIST_samples_per_second": 60.694,
+      "eval_NIST_steps_per_second": 0.95,
+      "step": 160000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4363257253037982e-05,
+      "loss": 0.1618,
+      "step": 160010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4361015201112057e-05,
+      "loss": 0.1555,
+      "step": 160020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4358773149186136e-05,
+      "loss": 0.1601,
+      "step": 160030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4356531097260215e-05,
+      "loss": 0.1633,
+      "step": 160040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.435428904533429e-05,
+      "loss": 0.1618,
+      "step": 160050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4352046993408369e-05,
+      "loss": 0.1623,
+      "step": 160060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4349804941482444e-05,
+      "loss": 0.1599,
+      "step": 160070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4347562889556523e-05,
+      "loss": 0.1586,
+      "step": 160080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.43453208376306e-05,
+      "loss": 0.161,
+      "step": 160090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4343078785704678e-05,
+      "loss": 0.1637,
+      "step": 160100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4340836733778757e-05,
+      "loss": 0.1613,
+      "step": 160110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4338594681852832e-05,
+      "loss": 0.1649,
+      "step": 160120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4336352629926911e-05,
+      "loss": 0.1546,
+      "step": 160130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4334110578000987e-05,
+      "loss": 0.1662,
+      "step": 160140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4331868526075065e-05,
+      "loss": 0.1599,
+      "step": 160150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.432962647414914e-05,
+      "loss": 0.1635,
+      "step": 160160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.432738442222322e-05,
+      "loss": 0.1629,
+      "step": 160170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4325142370297298e-05,
+      "loss": 0.1675,
+      "step": 160180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4322900318371373e-05,
+      "loss": 0.1623,
+      "step": 160190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4320658266445452e-05,
+      "loss": 0.159,
+      "step": 160200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4318416214519529e-05,
+      "loss": 0.1602,
+      "step": 160210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4316174162593606e-05,
+      "loss": 0.16,
+      "step": 160220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4313932110667683e-05,
+      "loss": 0.1583,
+      "step": 160230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4311690058741762e-05,
+      "loss": 0.1615,
+      "step": 160240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.430944800681584e-05,
+      "loss": 0.1623,
+      "step": 160250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4307205954889916e-05,
+      "loss": 0.1621,
+      "step": 160260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4304963902963995e-05,
+      "loss": 0.1643,
+      "step": 160270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.430272185103807e-05,
+      "loss": 0.1671,
+      "step": 160280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4300479799112149e-05,
+      "loss": 0.1551,
+      "step": 160290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4298237747186224e-05,
+      "loss": 0.1607,
+      "step": 160300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4295995695260303e-05,
+      "loss": 0.1573,
+      "step": 160310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4293753643334381e-05,
+      "loss": 0.1606,
+      "step": 160320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4291511591408457e-05,
+      "loss": 0.1622,
+      "step": 160330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4289269539482535e-05,
+      "loss": 0.1641,
+      "step": 160340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4287027487556612e-05,
+      "loss": 0.1664,
+      "step": 160350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4284785435630691e-05,
+      "loss": 0.1651,
+      "step": 160360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4282543383704767e-05,
+      "loss": 0.1608,
+      "step": 160370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4280301331778845e-05,
+      "loss": 0.1622,
+      "step": 160380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4278059279852924e-05,
+      "loss": 0.1593,
+      "step": 160390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4275817227927e-05,
+      "loss": 0.1602,
+      "step": 160400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4273575176001078e-05,
+      "loss": 0.1627,
+      "step": 160410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4271333124075153e-05,
+      "loss": 0.1612,
+      "step": 160420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4269091072149232e-05,
+      "loss": 0.1626,
+      "step": 160430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4266849020223307e-05,
+      "loss": 0.1614,
+      "step": 160440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4264606968297386e-05,
+      "loss": 0.1666,
+      "step": 160450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4262364916371465e-05,
+      "loss": 0.163,
+      "step": 160460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.426012286444554e-05,
+      "loss": 0.1571,
+      "step": 160470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4257880812519619e-05,
+      "loss": 0.1583,
+      "step": 160480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4255638760593696e-05,
+      "loss": 0.1635,
+      "step": 160490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4253396708667775e-05,
+      "loss": 0.1588,
+      "step": 160500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.425115465674185e-05,
+      "loss": 0.1701,
+      "step": 160510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4248912604815929e-05,
+      "loss": 0.1629,
+      "step": 160520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4246670552890007e-05,
+      "loss": 0.1621,
+      "step": 160530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4244428500964083e-05,
+      "loss": 0.1666,
+      "step": 160540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4242186449038161e-05,
+      "loss": 0.1625,
+      "step": 160550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4239944397112237e-05,
+      "loss": 0.1644,
+      "step": 160560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4237702345186315e-05,
+      "loss": 0.1672,
+      "step": 160570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.423546029326039e-05,
+      "loss": 0.1637,
+      "step": 160580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.423321824133447e-05,
+      "loss": 0.1684,
+      "step": 160590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4230976189408548e-05,
+      "loss": 0.1614,
+      "step": 160600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4228734137482624e-05,
+      "loss": 0.1649,
+      "step": 160610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4226492085556702e-05,
+      "loss": 0.1573,
+      "step": 160620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.422425003363078e-05,
+      "loss": 0.1613,
+      "step": 160630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4222007981704858e-05,
+      "loss": 0.1575,
+      "step": 160640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4219765929778933e-05,
+      "loss": 0.1553,
+      "step": 160650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4217523877853012e-05,
+      "loss": 0.1596,
+      "step": 160660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.421528182592709e-05,
+      "loss": 0.1698,
+      "step": 160670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4213039774001166e-05,
+      "loss": 0.1607,
+      "step": 160680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4210797722075245e-05,
+      "loss": 0.1568,
+      "step": 160690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.420855567014932e-05,
+      "loss": 0.1622,
+      "step": 160700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4206313618223399e-05,
+      "loss": 0.1675,
+      "step": 160710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4204071566297474e-05,
+      "loss": 0.1583,
+      "step": 160720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4201829514371553e-05,
+      "loss": 0.1668,
+      "step": 160730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4199587462445632e-05,
+      "loss": 0.1602,
+      "step": 160740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4197345410519709e-05,
+      "loss": 0.1642,
+      "step": 160750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4195103358593786e-05,
+      "loss": 0.165,
+      "step": 160760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4192861306667863e-05,
+      "loss": 0.1669,
+      "step": 160770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4190619254741941e-05,
+      "loss": 0.1618,
+      "step": 160780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4188377202816017e-05,
+      "loss": 0.1574,
+      "step": 160790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4186135150890096e-05,
+      "loss": 0.1615,
+      "step": 160800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4183893098964174e-05,
+      "loss": 0.1593,
+      "step": 160810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.418165104703825e-05,
+      "loss": 0.1589,
+      "step": 160820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4179408995112328e-05,
+      "loss": 0.162,
+      "step": 160830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4177166943186404e-05,
+      "loss": 0.1594,
+      "step": 160840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4174924891260482e-05,
+      "loss": 0.1605,
+      "step": 160850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4172682839334558e-05,
+      "loss": 0.1633,
+      "step": 160860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4170440787408636e-05,
+      "loss": 0.1594,
+      "step": 160870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4168198735482715e-05,
+      "loss": 0.1622,
+      "step": 160880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4165956683556792e-05,
+      "loss": 0.1647,
+      "step": 160890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.416371463163087e-05,
+      "loss": 0.1621,
+      "step": 160900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4161472579704946e-05,
+      "loss": 0.1602,
+      "step": 160910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4159230527779025e-05,
+      "loss": 0.1632,
+      "step": 160920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.41569884758531e-05,
+      "loss": 0.1656,
+      "step": 160930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4154746423927179e-05,
+      "loss": 0.1617,
+      "step": 160940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4152504372001258e-05,
+      "loss": 0.1595,
+      "step": 160950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4150262320075333e-05,
+      "loss": 0.1617,
+      "step": 160960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4148020268149412e-05,
+      "loss": 0.1601,
+      "step": 160970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4145778216223487e-05,
+      "loss": 0.1656,
+      "step": 160980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4143536164297566e-05,
+      "loss": 0.1595,
+      "step": 160990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4141294112371643e-05,
+      "loss": 0.1548,
+      "step": 161000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.413905206044572e-05,
+      "loss": 0.1642,
+      "step": 161010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4136810008519798e-05,
+      "loss": 0.1624,
+      "step": 161020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4134567956593876e-05,
+      "loss": 0.1623,
+      "step": 161030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4132325904667954e-05,
+      "loss": 0.1632,
+      "step": 161040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.413008385274203e-05,
+      "loss": 0.159,
+      "step": 161050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4127841800816108e-05,
+      "loss": 0.1673,
+      "step": 161060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4125599748890184e-05,
+      "loss": 0.1608,
+      "step": 161070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4123357696964262e-05,
+      "loss": 0.1573,
+      "step": 161080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4121115645038341e-05,
+      "loss": 0.1593,
+      "step": 161090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4118873593112416e-05,
+      "loss": 0.1582,
+      "step": 161100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4116631541186495e-05,
+      "loss": 0.166,
+      "step": 161110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.411438948926057e-05,
+      "loss": 0.1655,
+      "step": 161120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4112147437334649e-05,
+      "loss": 0.1611,
+      "step": 161130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4109905385408726e-05,
+      "loss": 0.1633,
+      "step": 161140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4107663333482803e-05,
+      "loss": 0.1651,
+      "step": 161150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4105421281556882e-05,
+      "loss": 0.1621,
+      "step": 161160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4103179229630959e-05,
+      "loss": 0.1644,
+      "step": 161170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4100937177705038e-05,
+      "loss": 0.1655,
+      "step": 161180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4098695125779113e-05,
+      "loss": 0.1624,
+      "step": 161190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4096453073853192e-05,
+      "loss": 0.1644,
+      "step": 161200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4094211021927267e-05,
+      "loss": 0.1636,
+      "step": 161210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4091968970001346e-05,
+      "loss": 0.1644,
+      "step": 161220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4089726918075424e-05,
+      "loss": 0.1593,
+      "step": 161230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.40874848661495e-05,
+      "loss": 0.1609,
+      "step": 161240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4085242814223579e-05,
+      "loss": 0.1617,
+      "step": 161250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4083000762297654e-05,
+      "loss": 0.1609,
+      "step": 161260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4080758710371733e-05,
+      "loss": 0.1594,
+      "step": 161270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.407851665844581e-05,
+      "loss": 0.161,
+      "step": 161280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4076274606519888e-05,
+      "loss": 0.1602,
+      "step": 161290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4074032554593965e-05,
+      "loss": 0.1521,
+      "step": 161300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4071790502668042e-05,
+      "loss": 0.1575,
+      "step": 161310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4069548450742121e-05,
+      "loss": 0.1595,
+      "step": 161320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4067306398816196e-05,
+      "loss": 0.1607,
+      "step": 161330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4065064346890275e-05,
+      "loss": 0.1588,
+      "step": 161340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.406282229496435e-05,
+      "loss": 0.1596,
+      "step": 161350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.406058024303843e-05,
+      "loss": 0.1631,
+      "step": 161360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4058338191112508e-05,
+      "loss": 0.1602,
+      "step": 161370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4056096139186583e-05,
+      "loss": 0.1599,
+      "step": 161380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4053854087260662e-05,
+      "loss": 0.1536,
+      "step": 161390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4051612035334737e-05,
+      "loss": 0.1579,
+      "step": 161400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4049369983408816e-05,
+      "loss": 0.1592,
+      "step": 161410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4047127931482895e-05,
+      "loss": 0.1555,
+      "step": 161420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4044885879556972e-05,
+      "loss": 0.1686,
+      "step": 161430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4042643827631049e-05,
+      "loss": 0.1662,
+      "step": 161440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4040401775705126e-05,
+      "loss": 0.1615,
+      "step": 161450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4038159723779204e-05,
+      "loss": 0.1627,
+      "step": 161460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.403591767185328e-05,
+      "loss": 0.1675,
+      "step": 161470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4033675619927359e-05,
+      "loss": 0.1564,
+      "step": 161480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4031433568001437e-05,
+      "loss": 0.1642,
+      "step": 161490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4029191516075513e-05,
+      "loss": 0.1606,
+      "step": 161500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4026949464149591e-05,
+      "loss": 0.1662,
+      "step": 161510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4024707412223667e-05,
+      "loss": 0.1623,
+      "step": 161520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4022465360297745e-05,
+      "loss": 0.1631,
+      "step": 161530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4020223308371822e-05,
+      "loss": 0.166,
+      "step": 161540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.40179812564459e-05,
+      "loss": 0.1639,
+      "step": 161550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4015739204519978e-05,
+      "loss": 0.1587,
+      "step": 161560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4013497152594055e-05,
+      "loss": 0.1585,
+      "step": 161570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4011255100668134e-05,
+      "loss": 0.1604,
+      "step": 161580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.400901304874221e-05,
+      "loss": 0.1627,
+      "step": 161590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4006770996816288e-05,
+      "loss": 0.163,
+      "step": 161600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4004528944890363e-05,
+      "loss": 0.1655,
+      "step": 161610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4002286892964442e-05,
+      "loss": 0.165,
+      "step": 161620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.400004484103852e-05,
+      "loss": 0.1589,
+      "step": 161630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3997802789112596e-05,
+      "loss": 0.1575,
+      "step": 161640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3995560737186675e-05,
+      "loss": 0.1613,
+      "step": 161650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.399331868526075e-05,
+      "loss": 0.1622,
+      "step": 161660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3991076633334829e-05,
+      "loss": 0.1592,
+      "step": 161670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3988834581408906e-05,
+      "loss": 0.1608,
+      "step": 161680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3986592529482983e-05,
+      "loss": 0.1628,
+      "step": 161690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3984350477557062e-05,
+      "loss": 0.1592,
+      "step": 161700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3982108425631139e-05,
+      "loss": 0.1647,
+      "step": 161710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3979866373705217e-05,
+      "loss": 0.1556,
+      "step": 161720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3977624321779293e-05,
+      "loss": 0.1592,
+      "step": 161730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3975382269853371e-05,
+      "loss": 0.1575,
+      "step": 161740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3973140217927447e-05,
+      "loss": 0.1617,
+      "step": 161750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3970898166001525e-05,
+      "loss": 0.1626,
+      "step": 161760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3968656114075604e-05,
+      "loss": 0.1622,
+      "step": 161770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.396641406214968e-05,
+      "loss": 0.1616,
+      "step": 161780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3964172010223758e-05,
+      "loss": 0.1667,
+      "step": 161790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3961929958297833e-05,
+      "loss": 0.1598,
+      "step": 161800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3959687906371912e-05,
+      "loss": 0.1612,
+      "step": 161810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.395744585444599e-05,
+      "loss": 0.1636,
+      "step": 161820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3955203802520068e-05,
+      "loss": 0.1594,
+      "step": 161830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3952961750594145e-05,
+      "loss": 0.1553,
+      "step": 161840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3950719698668222e-05,
+      "loss": 0.1581,
+      "step": 161850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.39484776467423e-05,
+      "loss": 0.1565,
+      "step": 161860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3946235594816376e-05,
+      "loss": 0.1693,
+      "step": 161870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3944217748083047e-05,
+      "loss": 0.17,
+      "step": 161880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3941975696157123e-05,
+      "loss": 0.1648,
+      "step": 161890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3939733644231201e-05,
+      "loss": 0.155,
+      "step": 161900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.393749159230528e-05,
+      "loss": 0.1604,
+      "step": 161910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3935249540379355e-05,
+      "loss": 0.1575,
+      "step": 161920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3933007488453434e-05,
+      "loss": 0.1634,
+      "step": 161930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.393076543652751e-05,
+      "loss": 0.1623,
+      "step": 161940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3928523384601588e-05,
+      "loss": 0.1578,
+      "step": 161950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3926281332675663e-05,
+      "loss": 0.1589,
+      "step": 161960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3924039280749742e-05,
+      "loss": 0.1589,
+      "step": 161970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.392179722882382e-05,
+      "loss": 0.1619,
+      "step": 161980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3919555176897898e-05,
+      "loss": 0.1604,
+      "step": 161990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3917313124971975e-05,
+      "loss": 0.1583,
+      "step": 162000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3915071073046052e-05,
+      "loss": 0.1613,
+      "step": 162010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.391282902112013e-05,
+      "loss": 0.1617,
+      "step": 162020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3910586969194206e-05,
+      "loss": 0.1636,
+      "step": 162030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3908344917268285e-05,
+      "loss": 0.1568,
+      "step": 162040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3906102865342363e-05,
+      "loss": 0.1595,
+      "step": 162050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3903860813416439e-05,
+      "loss": 0.1584,
+      "step": 162060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3901618761490517e-05,
+      "loss": 0.1593,
+      "step": 162070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3899376709564593e-05,
+      "loss": 0.1589,
+      "step": 162080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3897134657638671e-05,
+      "loss": 0.1575,
+      "step": 162090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3894892605712748e-05,
+      "loss": 0.1574,
+      "step": 162100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3892650553786826e-05,
+      "loss": 0.1596,
+      "step": 162110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3890408501860904e-05,
+      "loss": 0.1596,
+      "step": 162120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3888166449934981e-05,
+      "loss": 0.161,
+      "step": 162130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.388592439800906e-05,
+      "loss": 0.1595,
+      "step": 162140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3883682346083135e-05,
+      "loss": 0.1572,
+      "step": 162150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3881440294157214e-05,
+      "loss": 0.1642,
+      "step": 162160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.387919824223129e-05,
+      "loss": 0.1612,
+      "step": 162170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3876956190305368e-05,
+      "loss": 0.1665,
+      "step": 162180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3874714138379447e-05,
+      "loss": 0.1589,
+      "step": 162190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3872472086453522e-05,
+      "loss": 0.1574,
+      "step": 162200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.38702300345276e-05,
+      "loss": 0.1601,
+      "step": 162210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3867987982601676e-05,
+      "loss": 0.1614,
+      "step": 162220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3865745930675755e-05,
+      "loss": 0.1608,
+      "step": 162230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3863503878749832e-05,
+      "loss": 0.158,
+      "step": 162240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3861261826823909e-05,
+      "loss": 0.1636,
+      "step": 162250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3859019774897988e-05,
+      "loss": 0.1612,
+      "step": 162260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3856777722972065e-05,
+      "loss": 0.1505,
+      "step": 162270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3854535671046143e-05,
+      "loss": 0.162,
+      "step": 162280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3852293619120219e-05,
+      "loss": 0.157,
+      "step": 162290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3850051567194297e-05,
+      "loss": 0.1585,
+      "step": 162300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3847809515268376e-05,
+      "loss": 0.1567,
+      "step": 162310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3845567463342451e-05,
+      "loss": 0.1572,
+      "step": 162320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.384332541141653e-05,
+      "loss": 0.1576,
+      "step": 162330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3841083359490606e-05,
+      "loss": 0.1671,
+      "step": 162340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3838841307564684e-05,
+      "loss": 0.1609,
+      "step": 162350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.383659925563876e-05,
+      "loss": 0.1598,
+      "step": 162360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3834357203712838e-05,
+      "loss": 0.1585,
+      "step": 162370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3832115151786917e-05,
+      "loss": 0.1614,
+      "step": 162380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3829873099860994e-05,
+      "loss": 0.162,
+      "step": 162390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3827631047935071e-05,
+      "loss": 0.1578,
+      "step": 162400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3825388996009148e-05,
+      "loss": 0.1539,
+      "step": 162410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3823146944083227e-05,
+      "loss": 0.165,
+      "step": 162420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3820904892157302e-05,
+      "loss": 0.1622,
+      "step": 162430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.381866284023138e-05,
+      "loss": 0.1676,
+      "step": 162440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.381642078830546e-05,
+      "loss": 0.1646,
+      "step": 162450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3814178736379535e-05,
+      "loss": 0.1656,
+      "step": 162460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3811936684453614e-05,
+      "loss": 0.1634,
+      "step": 162470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3809694632527689e-05,
+      "loss": 0.1614,
+      "step": 162480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3807452580601768e-05,
+      "loss": 0.1614,
+      "step": 162490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3805210528675843e-05,
+      "loss": 0.1639,
+      "step": 162500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3802968476749922e-05,
+      "loss": 0.1591,
+      "step": 162510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3800726424824e-05,
+      "loss": 0.1633,
+      "step": 162520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3798484372898077e-05,
+      "loss": 0.1613,
+      "step": 162530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3796242320972154e-05,
+      "loss": 0.163,
+      "step": 162540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3794000269046231e-05,
+      "loss": 0.1617,
+      "step": 162550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.379175821712031e-05,
+      "loss": 0.1606,
+      "step": 162560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3789516165194386e-05,
+      "loss": 0.1587,
+      "step": 162570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3787274113268464e-05,
+      "loss": 0.1636,
+      "step": 162580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3785032061342543e-05,
+      "loss": 0.1634,
+      "step": 162590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3782790009416618e-05,
+      "loss": 0.1627,
+      "step": 162600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3780547957490697e-05,
+      "loss": 0.1637,
+      "step": 162610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3778305905564772e-05,
+      "loss": 0.1647,
+      "step": 162620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3776063853638851e-05,
+      "loss": 0.1615,
+      "step": 162630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3773821801712928e-05,
+      "loss": 0.1558,
+      "step": 162640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3771579749787005e-05,
+      "loss": 0.1569,
+      "step": 162650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3769337697861084e-05,
+      "loss": 0.1592,
+      "step": 162660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3767095645935161e-05,
+      "loss": 0.1588,
+      "step": 162670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.376485359400924e-05,
+      "loss": 0.1559,
+      "step": 162680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3762611542083315e-05,
+      "loss": 0.1653,
+      "step": 162690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3760369490157394e-05,
+      "loss": 0.1609,
+      "step": 162700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3758127438231469e-05,
+      "loss": 0.161,
+      "step": 162710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3755885386305548e-05,
+      "loss": 0.1524,
+      "step": 162720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3753643334379626e-05,
+      "loss": 0.1607,
+      "step": 162730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3751401282453702e-05,
+      "loss": 0.158,
+      "step": 162740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.374915923052778e-05,
+      "loss": 0.1659,
+      "step": 162750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3746917178601856e-05,
+      "loss": 0.1625,
+      "step": 162760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3744675126675934e-05,
+      "loss": 0.165,
+      "step": 162770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3742433074750012e-05,
+      "loss": 0.1602,
+      "step": 162780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3740191022824089e-05,
+      "loss": 0.1597,
+      "step": 162790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3737948970898167e-05,
+      "loss": 0.1594,
+      "step": 162800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3735706918972244e-05,
+      "loss": 0.1589,
+      "step": 162810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3733464867046323e-05,
+      "loss": 0.1605,
+      "step": 162820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3731222815120398e-05,
+      "loss": 0.1591,
+      "step": 162830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3728980763194477e-05,
+      "loss": 0.1555,
+      "step": 162840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3726738711268552e-05,
+      "loss": 0.1617,
+      "step": 162850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3724496659342631e-05,
+      "loss": 0.1609,
+      "step": 162860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.372225460741671e-05,
+      "loss": 0.1658,
+      "step": 162870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3720012555490785e-05,
+      "loss": 0.1605,
+      "step": 162880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3717770503564864e-05,
+      "loss": 0.1533,
+      "step": 162890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.371552845163894e-05,
+      "loss": 0.1591,
+      "step": 162900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3713286399713018e-05,
+      "loss": 0.1603,
+      "step": 162910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3711044347787095e-05,
+      "loss": 0.1608,
+      "step": 162920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3708802295861174e-05,
+      "loss": 0.1612,
+      "step": 162930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.370656024393525e-05,
+      "loss": 0.1558,
+      "step": 162940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3704318192009328e-05,
+      "loss": 0.1605,
+      "step": 162950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3702076140083406e-05,
+      "loss": 0.1585,
+      "step": 162960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3699834088157482e-05,
+      "loss": 0.1594,
+      "step": 162970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.369759203623156e-05,
+      "loss": 0.1626,
+      "step": 162980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3695349984305636e-05,
+      "loss": 0.1601,
+      "step": 162990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3693107932379714e-05,
+      "loss": 0.1641,
+      "step": 163000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3690865880453793e-05,
+      "loss": 0.1617,
+      "step": 163010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3688623828527869e-05,
+      "loss": 0.1627,
+      "step": 163020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3686381776601947e-05,
+      "loss": 0.1629,
+      "step": 163030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3684139724676023e-05,
+      "loss": 0.1573,
+      "step": 163040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3681897672750101e-05,
+      "loss": 0.1602,
+      "step": 163050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3679655620824178e-05,
+      "loss": 0.1524,
+      "step": 163060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3677413568898257e-05,
+      "loss": 0.1659,
+      "step": 163070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3675171516972334e-05,
+      "loss": 0.162,
+      "step": 163080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3672929465046411e-05,
+      "loss": 0.1571,
+      "step": 163090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.367068741312049e-05,
+      "loss": 0.1606,
+      "step": 163100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3668445361194565e-05,
+      "loss": 0.163,
+      "step": 163110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3666203309268644e-05,
+      "loss": 0.1605,
+      "step": 163120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.366396125734272e-05,
+      "loss": 0.1622,
+      "step": 163130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3661719205416798e-05,
+      "loss": 0.1694,
+      "step": 163140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3659477153490877e-05,
+      "loss": 0.1584,
+      "step": 163150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3657235101564952e-05,
+      "loss": 0.1644,
+      "step": 163160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.365499304963903e-05,
+      "loss": 0.1612,
+      "step": 163170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3652750997713106e-05,
+      "loss": 0.1591,
+      "step": 163180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3650508945787185e-05,
+      "loss": 0.1558,
+      "step": 163190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3648266893861262e-05,
+      "loss": 0.1613,
+      "step": 163200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.364602484193534e-05,
+      "loss": 0.1619,
+      "step": 163210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.364378279000942e-05,
+      "loss": 0.1578,
+      "step": 163220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3641540738083495e-05,
+      "loss": 0.1617,
+      "step": 163230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3639298686157573e-05,
+      "loss": 0.162,
+      "step": 163240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3637056634231649e-05,
+      "loss": 0.1617,
+      "step": 163250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3634814582305727e-05,
+      "loss": 0.1584,
+      "step": 163260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3632572530379803e-05,
+      "loss": 0.1571,
+      "step": 163270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3630330478453881e-05,
+      "loss": 0.163,
+      "step": 163280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.362808842652796e-05,
+      "loss": 0.1606,
+      "step": 163290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3625846374602035e-05,
+      "loss": 0.1616,
+      "step": 163300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3623604322676114e-05,
+      "loss": 0.1602,
+      "step": 163310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3621362270750191e-05,
+      "loss": 0.1588,
+      "step": 163320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3619120218824268e-05,
+      "loss": 0.1574,
+      "step": 163330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3616878166898345e-05,
+      "loss": 0.1578,
+      "step": 163340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3614636114972424e-05,
+      "loss": 0.1636,
+      "step": 163350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3612394063046503e-05,
+      "loss": 0.1575,
+      "step": 163360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3610152011120578e-05,
+      "loss": 0.162,
+      "step": 163370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3607909959194657e-05,
+      "loss": 0.1599,
+      "step": 163380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3605667907268732e-05,
+      "loss": 0.1602,
+      "step": 163390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.360342585534281e-05,
+      "loss": 0.1615,
+      "step": 163400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3601183803416886e-05,
+      "loss": 0.1635,
+      "step": 163410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3598941751490965e-05,
+      "loss": 0.1563,
+      "step": 163420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3596699699565043e-05,
+      "loss": 0.1587,
+      "step": 163430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3594457647639119e-05,
+      "loss": 0.1565,
+      "step": 163440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3592215595713198e-05,
+      "loss": 0.1613,
+      "step": 163450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3589973543787275e-05,
+      "loss": 0.1634,
+      "step": 163460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3587731491861353e-05,
+      "loss": 0.1624,
+      "step": 163470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3585489439935429e-05,
+      "loss": 0.1585,
+      "step": 163480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3583247388009507e-05,
+      "loss": 0.1642,
+      "step": 163490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3581005336083586e-05,
+      "loss": 0.1615,
+      "step": 163500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3578763284157661e-05,
+      "loss": 0.1567,
+      "step": 163510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.357652123223174e-05,
+      "loss": 0.1622,
+      "step": 163520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3574279180305815e-05,
+      "loss": 0.1569,
+      "step": 163530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3572037128379894e-05,
+      "loss": 0.1639,
+      "step": 163540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.356979507645397e-05,
+      "loss": 0.157,
+      "step": 163550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3567553024528048e-05,
+      "loss": 0.1627,
+      "step": 163560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3565310972602127e-05,
+      "loss": 0.1646,
+      "step": 163570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3563068920676202e-05,
+      "loss": 0.164,
+      "step": 163580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3560826868750281e-05,
+      "loss": 0.1587,
+      "step": 163590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3558584816824358e-05,
+      "loss": 0.1592,
+      "step": 163600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3556342764898437e-05,
+      "loss": 0.1574,
+      "step": 163610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3554100712972512e-05,
+      "loss": 0.1579,
+      "step": 163620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.355185866104659e-05,
+      "loss": 0.1659,
+      "step": 163630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.354961660912067e-05,
+      "loss": 0.1599,
+      "step": 163640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3547374557194745e-05,
+      "loss": 0.1605,
+      "step": 163650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3545132505268823e-05,
+      "loss": 0.1583,
+      "step": 163660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3542890453342899e-05,
+      "loss": 0.1584,
+      "step": 163670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3540648401416978e-05,
+      "loss": 0.1595,
+      "step": 163680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3538406349491056e-05,
+      "loss": 0.1607,
+      "step": 163690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3536164297565132e-05,
+      "loss": 0.1588,
+      "step": 163700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.353392224563921e-05,
+      "loss": 0.1575,
+      "step": 163710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3531680193713286e-05,
+      "loss": 0.1626,
+      "step": 163720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3529438141787364e-05,
+      "loss": 0.1604,
+      "step": 163730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3527196089861441e-05,
+      "loss": 0.1606,
+      "step": 163740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.352495403793552e-05,
+      "loss": 0.1603,
+      "step": 163750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3522711986009599e-05,
+      "loss": 0.1635,
+      "step": 163760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3520469934083674e-05,
+      "loss": 0.1577,
+      "step": 163770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3518227882157753e-05,
+      "loss": 0.1592,
+      "step": 163780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3515985830231828e-05,
+      "loss": 0.1608,
+      "step": 163790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3513743778305907e-05,
+      "loss": 0.1657,
+      "step": 163800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3511501726379982e-05,
+      "loss": 0.1583,
+      "step": 163810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3509259674454061e-05,
+      "loss": 0.1666,
+      "step": 163820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.350701762252814e-05,
+      "loss": 0.1619,
+      "step": 163830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3504775570602215e-05,
+      "loss": 0.1698,
+      "step": 163840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3502533518676294e-05,
+      "loss": 0.1626,
+      "step": 163850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.350029146675037e-05,
+      "loss": 0.1595,
+      "step": 163860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3498049414824448e-05,
+      "loss": 0.1673,
+      "step": 163870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3495807362898525e-05,
+      "loss": 0.1602,
+      "step": 163880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3493565310972603e-05,
+      "loss": 0.1629,
+      "step": 163890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3491323259046682e-05,
+      "loss": 0.1576,
+      "step": 163900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3489081207120758e-05,
+      "loss": 0.1531,
+      "step": 163910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3486839155194836e-05,
+      "loss": 0.1655,
+      "step": 163920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3484597103268912e-05,
+      "loss": 0.1523,
+      "step": 163930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.348235505134299e-05,
+      "loss": 0.1589,
+      "step": 163940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3480112999417066e-05,
+      "loss": 0.163,
+      "step": 163950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3477870947491144e-05,
+      "loss": 0.164,
+      "step": 163960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3475628895565223e-05,
+      "loss": 0.1626,
+      "step": 163970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3473386843639298e-05,
+      "loss": 0.1556,
+      "step": 163980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3471144791713377e-05,
+      "loss": 0.1572,
+      "step": 163990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3468902739787454e-05,
+      "loss": 0.1569,
+      "step": 164000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3466660687861533e-05,
+      "loss": 0.1578,
+      "step": 164010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3464418635935608e-05,
+      "loss": 0.1606,
+      "step": 164020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3462176584009687e-05,
+      "loss": 0.1614,
+      "step": 164030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3459934532083766e-05,
+      "loss": 0.1576,
+      "step": 164040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3457692480157841e-05,
+      "loss": 0.1616,
+      "step": 164050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.345545042823192e-05,
+      "loss": 0.1679,
+      "step": 164060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3453208376305995e-05,
+      "loss": 0.1599,
+      "step": 164070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3450966324380074e-05,
+      "loss": 0.1622,
+      "step": 164080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3448724272454149e-05,
+      "loss": 0.1666,
+      "step": 164090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3446482220528228e-05,
+      "loss": 0.1633,
+      "step": 164100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3444240168602306e-05,
+      "loss": 0.1673,
+      "step": 164110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3441998116676382e-05,
+      "loss": 0.1556,
+      "step": 164120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.343975606475046e-05,
+      "loss": 0.1604,
+      "step": 164130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3437514012824538e-05,
+      "loss": 0.1624,
+      "step": 164140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3435271960898616e-05,
+      "loss": 0.1667,
+      "step": 164150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3433029908972692e-05,
+      "loss": 0.1588,
+      "step": 164160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.343078785704677e-05,
+      "loss": 0.1622,
+      "step": 164170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3428545805120849e-05,
+      "loss": 0.157,
+      "step": 164180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3426303753194924e-05,
+      "loss": 0.1611,
+      "step": 164190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3424061701269003e-05,
+      "loss": 0.1588,
+      "step": 164200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3421819649343078e-05,
+      "loss": 0.1641,
+      "step": 164210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3419577597417157e-05,
+      "loss": 0.1609,
+      "step": 164220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3417335545491232e-05,
+      "loss": 0.16,
+      "step": 164230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3415093493565311e-05,
+      "loss": 0.1582,
+      "step": 164240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.341285144163939e-05,
+      "loss": 0.1597,
+      "step": 164250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3410609389713465e-05,
+      "loss": 0.1589,
+      "step": 164260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3408367337787544e-05,
+      "loss": 0.1613,
+      "step": 164270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3406125285861621e-05,
+      "loss": 0.1575,
+      "step": 164280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.34038832339357e-05,
+      "loss": 0.1593,
+      "step": 164290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3401641182009775e-05,
+      "loss": 0.1663,
+      "step": 164300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3399399130083854e-05,
+      "loss": 0.1608,
+      "step": 164310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3397157078157932e-05,
+      "loss": 0.1636,
+      "step": 164320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3394915026232008e-05,
+      "loss": 0.1569,
+      "step": 164330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3392672974306086e-05,
+      "loss": 0.1632,
+      "step": 164340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3390430922380162e-05,
+      "loss": 0.1649,
+      "step": 164350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.338818887045424e-05,
+      "loss": 0.1616,
+      "step": 164360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3385946818528316e-05,
+      "loss": 0.1652,
+      "step": 164370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3383704766602395e-05,
+      "loss": 0.1603,
+      "step": 164380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3381462714676473e-05,
+      "loss": 0.1615,
+      "step": 164390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.337922066275055e-05,
+      "loss": 0.1556,
+      "step": 164400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3376978610824627e-05,
+      "loss": 0.1583,
+      "step": 164410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3374736558898704e-05,
+      "loss": 0.1644,
+      "step": 164420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3372494506972783e-05,
+      "loss": 0.1691,
+      "step": 164430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3370252455046858e-05,
+      "loss": 0.1586,
+      "step": 164440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3368010403120937e-05,
+      "loss": 0.1579,
+      "step": 164450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3365768351195016e-05,
+      "loss": 0.157,
+      "step": 164460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3363526299269091e-05,
+      "loss": 0.1555,
+      "step": 164470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.336128424734317e-05,
+      "loss": 0.1582,
+      "step": 164480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3359266400609838e-05,
+      "loss": 0.1649,
+      "step": 164490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3357024348683916e-05,
+      "loss": 0.1575,
+      "step": 164500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3354782296757992e-05,
+      "loss": 0.1651,
+      "step": 164510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.335254024483207e-05,
+      "loss": 0.1578,
+      "step": 164520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.335029819290615e-05,
+      "loss": 0.1605,
+      "step": 164530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3348056140980225e-05,
+      "loss": 0.1581,
+      "step": 164540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3345814089054303e-05,
+      "loss": 0.158,
+      "step": 164550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.334357203712838e-05,
+      "loss": 0.1579,
+      "step": 164560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3341329985202459e-05,
+      "loss": 0.1514,
+      "step": 164570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3339087933276534e-05,
+      "loss": 0.1557,
+      "step": 164580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3336845881350613e-05,
+      "loss": 0.1501,
+      "step": 164590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3334603829424692e-05,
+      "loss": 0.1559,
+      "step": 164600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3332361777498767e-05,
+      "loss": 0.1574,
+      "step": 164610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3330119725572846e-05,
+      "loss": 0.1573,
+      "step": 164620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3327877673646921e-05,
+      "loss": 0.1607,
+      "step": 164630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3325635621721e-05,
+      "loss": 0.1632,
+      "step": 164640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3323393569795079e-05,
+      "loss": 0.1586,
+      "step": 164650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3321151517869154e-05,
+      "loss": 0.161,
+      "step": 164660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3318909465943233e-05,
+      "loss": 0.1591,
+      "step": 164670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3316667414017308e-05,
+      "loss": 0.1657,
+      "step": 164680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3314425362091387e-05,
+      "loss": 0.1649,
+      "step": 164690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3312183310165464e-05,
+      "loss": 0.1612,
+      "step": 164700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3309941258239542e-05,
+      "loss": 0.1583,
+      "step": 164710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.330769920631362e-05,
+      "loss": 0.1581,
+      "step": 164720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3305457154387696e-05,
+      "loss": 0.163,
+      "step": 164730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3303215102461775e-05,
+      "loss": 0.1552,
+      "step": 164740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.330097305053585e-05,
+      "loss": 0.1615,
+      "step": 164750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.329873099860993e-05,
+      "loss": 0.1643,
+      "step": 164760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3296488946684005e-05,
+      "loss": 0.1595,
+      "step": 164770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3294246894758083e-05,
+      "loss": 0.161,
+      "step": 164780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3292004842832162e-05,
+      "loss": 0.1643,
+      "step": 164790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3289762790906237e-05,
+      "loss": 0.1631,
+      "step": 164800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3287520738980316e-05,
+      "loss": 0.1674,
+      "step": 164810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3285278687054391e-05,
+      "loss": 0.1613,
+      "step": 164820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.328303663512847e-05,
+      "loss": 0.158,
+      "step": 164830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3280794583202547e-05,
+      "loss": 0.1635,
+      "step": 164840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3278552531276626e-05,
+      "loss": 0.1621,
+      "step": 164850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3276310479350705e-05,
+      "loss": 0.1625,
+      "step": 164860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.327406842742478e-05,
+      "loss": 0.1613,
+      "step": 164870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3271826375498859e-05,
+      "loss": 0.1607,
+      "step": 164880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3269584323572934e-05,
+      "loss": 0.1652,
+      "step": 164890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3267342271647013e-05,
+      "loss": 0.1598,
+      "step": 164900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3265100219721088e-05,
+      "loss": 0.1596,
+      "step": 164910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3262858167795167e-05,
+      "loss": 0.1594,
+      "step": 164920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3260616115869245e-05,
+      "loss": 0.1554,
+      "step": 164930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.325837406394332e-05,
+      "loss": 0.162,
+      "step": 164940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.32561320120174e-05,
+      "loss": 0.163,
+      "step": 164950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3253889960091476e-05,
+      "loss": 0.1611,
+      "step": 164960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3251647908165553e-05,
+      "loss": 0.1624,
+      "step": 164970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.324940585623963e-05,
+      "loss": 0.1644,
+      "step": 164980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.324716380431371e-05,
+      "loss": 0.1542,
+      "step": 164990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3244921752387788e-05,
+      "loss": 0.1576,
+      "step": 165000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3242679700461863e-05,
+      "loss": 0.163,
+      "step": 165010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3240437648535942e-05,
+      "loss": 0.1594,
+      "step": 165020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3238195596610017e-05,
+      "loss": 0.1613,
+      "step": 165030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3235953544684096e-05,
+      "loss": 0.1573,
+      "step": 165040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3233711492758171e-05,
+      "loss": 0.1621,
+      "step": 165050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.323146944083225e-05,
+      "loss": 0.1575,
+      "step": 165060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3229227388906329e-05,
+      "loss": 0.1636,
+      "step": 165070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3226985336980404e-05,
+      "loss": 0.1594,
+      "step": 165080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3224743285054483e-05,
+      "loss": 0.1608,
+      "step": 165090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.322250123312856e-05,
+      "loss": 0.1494,
+      "step": 165100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3220259181202639e-05,
+      "loss": 0.1559,
+      "step": 165110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3218017129276714e-05,
+      "loss": 0.1634,
+      "step": 165120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3215775077350793e-05,
+      "loss": 0.1627,
+      "step": 165130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3213533025424871e-05,
+      "loss": 0.1602,
+      "step": 165140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3211290973498947e-05,
+      "loss": 0.1608,
+      "step": 165150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3209048921573025e-05,
+      "loss": 0.1582,
+      "step": 165160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.32068068696471e-05,
+      "loss": 0.1568,
+      "step": 165170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.320456481772118e-05,
+      "loss": 0.1603,
+      "step": 165180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3202322765795255e-05,
+      "loss": 0.1601,
+      "step": 165190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3200080713869333e-05,
+      "loss": 0.1644,
+      "step": 165200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3197838661943412e-05,
+      "loss": 0.1642,
+      "step": 165210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3195596610017488e-05,
+      "loss": 0.1624,
+      "step": 165220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3193354558091566e-05,
+      "loss": 0.1667,
+      "step": 165230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3191112506165643e-05,
+      "loss": 0.1617,
+      "step": 165240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3188870454239722e-05,
+      "loss": 0.161,
+      "step": 165250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3186628402313797e-05,
+      "loss": 0.1591,
+      "step": 165260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3184386350387876e-05,
+      "loss": 0.1574,
+      "step": 165270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3182144298461955e-05,
+      "loss": 0.1574,
+      "step": 165280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.317990224653603e-05,
+      "loss": 0.1582,
+      "step": 165290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3177660194610109e-05,
+      "loss": 0.1637,
+      "step": 165300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3175418142684184e-05,
+      "loss": 0.1605,
+      "step": 165310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3173176090758263e-05,
+      "loss": 0.1595,
+      "step": 165320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3170934038832338e-05,
+      "loss": 0.1626,
+      "step": 165330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3168691986906417e-05,
+      "loss": 0.1579,
+      "step": 165340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3166449934980496e-05,
+      "loss": 0.1578,
+      "step": 165350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3164207883054571e-05,
+      "loss": 0.1539,
+      "step": 165360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.316196583112865e-05,
+      "loss": 0.1598,
+      "step": 165370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3159723779202727e-05,
+      "loss": 0.1592,
+      "step": 165380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3157481727276805e-05,
+      "loss": 0.1589,
+      "step": 165390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.315523967535088e-05,
+      "loss": 0.1624,
+      "step": 165400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.315299762342496e-05,
+      "loss": 0.1589,
+      "step": 165410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3150755571499038e-05,
+      "loss": 0.1536,
+      "step": 165420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3148513519573114e-05,
+      "loss": 0.1558,
+      "step": 165430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3146271467647192e-05,
+      "loss": 0.161,
+      "step": 165440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3144029415721268e-05,
+      "loss": 0.1648,
+      "step": 165450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3141787363795346e-05,
+      "loss": 0.1576,
+      "step": 165460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3139545311869422e-05,
+      "loss": 0.1647,
+      "step": 165470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.31373032599435e-05,
+      "loss": 0.1533,
+      "step": 165480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3135061208017579e-05,
+      "loss": 0.1606,
+      "step": 165490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3132819156091656e-05,
+      "loss": 0.1513,
+      "step": 165500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3130577104165733e-05,
+      "loss": 0.1671,
+      "step": 165510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.312833505223981e-05,
+      "loss": 0.1586,
+      "step": 165520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3126093000313889e-05,
+      "loss": 0.1621,
+      "step": 165530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3123850948387964e-05,
+      "loss": 0.1629,
+      "step": 165540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3121608896462043e-05,
+      "loss": 0.1614,
+      "step": 165550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3119366844536122e-05,
+      "loss": 0.1536,
+      "step": 165560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3117124792610197e-05,
+      "loss": 0.16,
+      "step": 165570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3114882740684276e-05,
+      "loss": 0.1617,
+      "step": 165580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3112640688758351e-05,
+      "loss": 0.1608,
+      "step": 165590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.311039863683243e-05,
+      "loss": 0.1567,
+      "step": 165600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3108156584906505e-05,
+      "loss": 0.1601,
+      "step": 165610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3105914532980584e-05,
+      "loss": 0.1608,
+      "step": 165620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3103672481054662e-05,
+      "loss": 0.1604,
+      "step": 165630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.310143042912874e-05,
+      "loss": 0.1616,
+      "step": 165640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3099188377202816e-05,
+      "loss": 0.1574,
+      "step": 165650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3096946325276894e-05,
+      "loss": 0.1575,
+      "step": 165660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3094704273350972e-05,
+      "loss": 0.1551,
+      "step": 165670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3092462221425048e-05,
+      "loss": 0.1589,
+      "step": 165680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3090220169499126e-05,
+      "loss": 0.1613,
+      "step": 165690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3087978117573205e-05,
+      "loss": 0.1558,
+      "step": 165700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.308573606564728e-05,
+      "loss": 0.1603,
+      "step": 165710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3083494013721359e-05,
+      "loss": 0.1545,
+      "step": 165720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3081251961795434e-05,
+      "loss": 0.1579,
+      "step": 165730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3079009909869513e-05,
+      "loss": 0.164,
+      "step": 165740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.307676785794359e-05,
+      "loss": 0.1609,
+      "step": 165750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3074525806017667e-05,
+      "loss": 0.1676,
+      "step": 165760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3072283754091746e-05,
+      "loss": 0.1658,
+      "step": 165770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3070041702165823e-05,
+      "loss": 0.1632,
+      "step": 165780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3067799650239902e-05,
+      "loss": 0.1619,
+      "step": 165790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3065557598313977e-05,
+      "loss": 0.1617,
+      "step": 165800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3063315546388056e-05,
+      "loss": 0.1559,
+      "step": 165810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3061073494462131e-05,
+      "loss": 0.1568,
+      "step": 165820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.305883144253621e-05,
+      "loss": 0.1635,
+      "step": 165830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3056589390610288e-05,
+      "loss": 0.1615,
+      "step": 165840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3054347338684364e-05,
+      "loss": 0.1605,
+      "step": 165850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3052105286758442e-05,
+      "loss": 0.1629,
+      "step": 165860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3049863234832518e-05,
+      "loss": 0.1655,
+      "step": 165870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3047621182906597e-05,
+      "loss": 0.1602,
+      "step": 165880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3045379130980674e-05,
+      "loss": 0.1602,
+      "step": 165890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.304313707905475e-05,
+      "loss": 0.1652,
+      "step": 165900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.304089502712883e-05,
+      "loss": 0.1616,
+      "step": 165910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3038652975202906e-05,
+      "loss": 0.1596,
+      "step": 165920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3036410923276985e-05,
+      "loss": 0.1638,
+      "step": 165930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.303416887135106e-05,
+      "loss": 0.1578,
+      "step": 165940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3031926819425139e-05,
+      "loss": 0.1658,
+      "step": 165950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3029684767499218e-05,
+      "loss": 0.1608,
+      "step": 165960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3027442715573293e-05,
+      "loss": 0.1554,
+      "step": 165970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3025200663647372e-05,
+      "loss": 0.1627,
+      "step": 165980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3022958611721447e-05,
+      "loss": 0.1595,
+      "step": 165990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3020716559795526e-05,
+      "loss": 0.1619,
+      "step": 166000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3018474507869601e-05,
+      "loss": 0.1585,
+      "step": 166010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.301623245594368e-05,
+      "loss": 0.1613,
+      "step": 166020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3013990404017759e-05,
+      "loss": 0.1653,
+      "step": 166030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3011748352091836e-05,
+      "loss": 0.1563,
+      "step": 166040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3009506300165913e-05,
+      "loss": 0.1575,
+      "step": 166050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.300726424823999e-05,
+      "loss": 0.1594,
+      "step": 166060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3005022196314068e-05,
+      "loss": 0.16,
+      "step": 166070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3002780144388144e-05,
+      "loss": 0.159,
+      "step": 166080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3000538092462222e-05,
+      "loss": 0.1612,
+      "step": 166090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2998296040536301e-05,
+      "loss": 0.1548,
+      "step": 166100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2996053988610377e-05,
+      "loss": 0.1601,
+      "step": 166110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2993811936684455e-05,
+      "loss": 0.1516,
+      "step": 166120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.299156988475853e-05,
+      "loss": 0.1594,
+      "step": 166130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.298932783283261e-05,
+      "loss": 0.1581,
+      "step": 166140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2987085780906685e-05,
+      "loss": 0.1589,
+      "step": 166150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2984843728980763e-05,
+      "loss": 0.164,
+      "step": 166160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2982601677054842e-05,
+      "loss": 0.1572,
+      "step": 166170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2980359625128919e-05,
+      "loss": 0.1603,
+      "step": 166180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2978117573202996e-05,
+      "loss": 0.1566,
+      "step": 166190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2975875521277073e-05,
+      "loss": 0.1608,
+      "step": 166200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2973633469351152e-05,
+      "loss": 0.1616,
+      "step": 166210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2971391417425227e-05,
+      "loss": 0.1583,
+      "step": 166220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2969149365499306e-05,
+      "loss": 0.1621,
+      "step": 166230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2966907313573385e-05,
+      "loss": 0.1575,
+      "step": 166240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.296466526164746e-05,
+      "loss": 0.1618,
+      "step": 166250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2962423209721539e-05,
+      "loss": 0.1596,
+      "step": 166260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2960181157795614e-05,
+      "loss": 0.1683,
+      "step": 166270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2957939105869693e-05,
+      "loss": 0.1615,
+      "step": 166280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2955697053943768e-05,
+      "loss": 0.1617,
+      "step": 166290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2953455002017847e-05,
+      "loss": 0.1648,
+      "step": 166300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2951212950091925e-05,
+      "loss": 0.1588,
+      "step": 166310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2948970898166002e-05,
+      "loss": 0.1573,
+      "step": 166320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2946728846240081e-05,
+      "loss": 0.1613,
+      "step": 166330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2944486794314157e-05,
+      "loss": 0.1604,
+      "step": 166340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2942244742388235e-05,
+      "loss": 0.1593,
+      "step": 166350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.294000269046231e-05,
+      "loss": 0.1557,
+      "step": 166360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.293776063853639e-05,
+      "loss": 0.1574,
+      "step": 166370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2935518586610468e-05,
+      "loss": 0.1598,
+      "step": 166380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2933276534684543e-05,
+      "loss": 0.1567,
+      "step": 166390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2931034482758622e-05,
+      "loss": 0.156,
+      "step": 166400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2928792430832697e-05,
+      "loss": 0.1614,
+      "step": 166410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2926550378906776e-05,
+      "loss": 0.1562,
+      "step": 166420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2924308326980853e-05,
+      "loss": 0.16,
+      "step": 166430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.292206627505493e-05,
+      "loss": 0.1555,
+      "step": 166440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2919824223129009e-05,
+      "loss": 0.1622,
+      "step": 166450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2917582171203086e-05,
+      "loss": 0.1592,
+      "step": 166460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2915340119277165e-05,
+      "loss": 0.1573,
+      "step": 166470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.291309806735124e-05,
+      "loss": 0.1529,
+      "step": 166480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2910856015425319e-05,
+      "loss": 0.1547,
+      "step": 166490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2908613963499394e-05,
+      "loss": 0.1611,
+      "step": 166500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2906371911573473e-05,
+      "loss": 0.1671,
+      "step": 166510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2904129859647551e-05,
+      "loss": 0.1626,
+      "step": 166520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2901887807721627e-05,
+      "loss": 0.1578,
+      "step": 166530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2899645755795705e-05,
+      "loss": 0.1605,
+      "step": 166540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.289740370386978e-05,
+      "loss": 0.1545,
+      "step": 166550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.289516165194386e-05,
+      "loss": 0.1628,
+      "step": 166560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2892919600017937e-05,
+      "loss": 0.1571,
+      "step": 166570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2890677548092015e-05,
+      "loss": 0.1604,
+      "step": 166580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2888435496166092e-05,
+      "loss": 0.1605,
+      "step": 166590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.288619344424017e-05,
+      "loss": 0.1561,
+      "step": 166600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2883951392314248e-05,
+      "loss": 0.159,
+      "step": 166610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2881709340388323e-05,
+      "loss": 0.1632,
+      "step": 166620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2879467288462402e-05,
+      "loss": 0.1584,
+      "step": 166630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2877225236536477e-05,
+      "loss": 0.1608,
+      "step": 166640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2874983184610556e-05,
+      "loss": 0.1605,
+      "step": 166650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2872741132684635e-05,
+      "loss": 0.16,
+      "step": 166660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.287049908075871e-05,
+      "loss": 0.1555,
+      "step": 166670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2868257028832789e-05,
+      "loss": 0.1653,
+      "step": 166680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2866014976906864e-05,
+      "loss": 0.1609,
+      "step": 166690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2863772924980943e-05,
+      "loss": 0.1587,
+      "step": 166700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.286153087305502e-05,
+      "loss": 0.1573,
+      "step": 166710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2859288821129099e-05,
+      "loss": 0.1618,
+      "step": 166720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2857046769203176e-05,
+      "loss": 0.159,
+      "step": 166730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2854804717277253e-05,
+      "loss": 0.1561,
+      "step": 166740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2852562665351331e-05,
+      "loss": 0.1496,
+      "step": 166750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2850320613425407e-05,
+      "loss": 0.16,
+      "step": 166760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2848078561499485e-05,
+      "loss": 0.1631,
+      "step": 166770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.284583650957356e-05,
+      "loss": 0.16,
+      "step": 166780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.284359445764764e-05,
+      "loss": 0.1638,
+      "step": 166790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2841352405721718e-05,
+      "loss": 0.1576,
+      "step": 166800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2839110353795794e-05,
+      "loss": 0.162,
+      "step": 166810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2836868301869872e-05,
+      "loss": 0.1602,
+      "step": 166820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2834626249943948e-05,
+      "loss": 0.1577,
+      "step": 166830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2832384198018026e-05,
+      "loss": 0.1558,
+      "step": 166840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2830142146092103e-05,
+      "loss": 0.1584,
+      "step": 166850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2827900094166182e-05,
+      "loss": 0.1587,
+      "step": 166860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.282565804224026e-05,
+      "loss": 0.1622,
+      "step": 166870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2823415990314336e-05,
+      "loss": 0.1566,
+      "step": 166880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2821173938388415e-05,
+      "loss": 0.1599,
+      "step": 166890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.281893188646249e-05,
+      "loss": 0.1542,
+      "step": 166900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2816689834536569e-05,
+      "loss": 0.1586,
+      "step": 166910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2814447782610644e-05,
+      "loss": 0.1593,
+      "step": 166920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2812205730684723e-05,
+      "loss": 0.1539,
+      "step": 166930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2809963678758802e-05,
+      "loss": 0.1633,
+      "step": 166940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2807721626832877e-05,
+      "loss": 0.1581,
+      "step": 166950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2805479574906956e-05,
+      "loss": 0.1661,
+      "step": 166960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2803237522981033e-05,
+      "loss": 0.1584,
+      "step": 166970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.280099547105511e-05,
+      "loss": 0.1625,
+      "step": 166980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2798753419129187e-05,
+      "loss": 0.1583,
+      "step": 166990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2796511367203266e-05,
+      "loss": 0.1543,
+      "step": 167000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2794269315277344e-05,
+      "loss": 0.1612,
+      "step": 167010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.279202726335142e-05,
+      "loss": 0.1593,
+      "step": 167020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2789785211425498e-05,
+      "loss": 0.1572,
+      "step": 167030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2787543159499574e-05,
+      "loss": 0.1565,
+      "step": 167040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2785301107573652e-05,
+      "loss": 0.1613,
+      "step": 167050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2783059055647728e-05,
+      "loss": 0.1587,
+      "step": 167060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2780817003721806e-05,
+      "loss": 0.1645,
+      "step": 167070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2778574951795885e-05,
+      "loss": 0.1562,
+      "step": 167080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.277633289986996e-05,
+      "loss": 0.1533,
+      "step": 167090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2774090847944039e-05,
+      "loss": 0.1554,
+      "step": 167100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2771848796018116e-05,
+      "loss": 0.1592,
+      "step": 167110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2769606744092195e-05,
+      "loss": 0.1628,
+      "step": 167120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.276736469216627e-05,
+      "loss": 0.1637,
+      "step": 167130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2765122640240349e-05,
+      "loss": 0.16,
+      "step": 167140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2762880588314428e-05,
+      "loss": 0.1559,
+      "step": 167150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2760638536388503e-05,
+      "loss": 0.1599,
+      "step": 167160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2758396484462582e-05,
+      "loss": 0.156,
+      "step": 167170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2756154432536657e-05,
+      "loss": 0.1614,
+      "step": 167180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2753912380610736e-05,
+      "loss": 0.1594,
+      "step": 167190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2751670328684811e-05,
+      "loss": 0.1607,
+      "step": 167200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.274942827675889e-05,
+      "loss": 0.1605,
+      "step": 167210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2747186224832969e-05,
+      "loss": 0.1597,
+      "step": 167220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2744944172907044e-05,
+      "loss": 0.1598,
+      "step": 167230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2742702120981123e-05,
+      "loss": 0.1544,
+      "step": 167240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.27404600690552e-05,
+      "loss": 0.1579,
+      "step": 167250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2738218017129278e-05,
+      "loss": 0.1618,
+      "step": 167260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2735975965203354e-05,
+      "loss": 0.1553,
+      "step": 167270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2733733913277432e-05,
+      "loss": 0.1504,
+      "step": 167280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2731491861351511e-05,
+      "loss": 0.1647,
+      "step": 167290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2729249809425586e-05,
+      "loss": 0.1603,
+      "step": 167300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2727231962692258e-05,
+      "loss": 0.1608,
+      "step": 167310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2724989910766333e-05,
+      "loss": 0.1573,
+      "step": 167320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2722747858840412e-05,
+      "loss": 0.1608,
+      "step": 167330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.272050580691449e-05,
+      "loss": 0.1599,
+      "step": 167340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2718263754988566e-05,
+      "loss": 0.1544,
+      "step": 167350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2716021703062644e-05,
+      "loss": 0.1593,
+      "step": 167360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.271377965113672e-05,
+      "loss": 0.1578,
+      "step": 167370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2711537599210798e-05,
+      "loss": 0.1585,
+      "step": 167380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2709295547284874e-05,
+      "loss": 0.1601,
+      "step": 167390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2707053495358952e-05,
+      "loss": 0.1629,
+      "step": 167400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2704811443433031e-05,
+      "loss": 0.1556,
+      "step": 167410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2702569391507108e-05,
+      "loss": 0.1612,
+      "step": 167420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2700327339581187e-05,
+      "loss": 0.1555,
+      "step": 167430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2698085287655262e-05,
+      "loss": 0.1586,
+      "step": 167440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2695843235729341e-05,
+      "loss": 0.1623,
+      "step": 167450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2693601183803416e-05,
+      "loss": 0.1552,
+      "step": 167460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2691359131877495e-05,
+      "loss": 0.1536,
+      "step": 167470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2689117079951574e-05,
+      "loss": 0.1583,
+      "step": 167480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2686875028025649e-05,
+      "loss": 0.1576,
+      "step": 167490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2684632976099728e-05,
+      "loss": 0.1587,
+      "step": 167500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2682390924173803e-05,
+      "loss": 0.1594,
+      "step": 167510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2680148872247882e-05,
+      "loss": 0.1562,
+      "step": 167520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2677906820321959e-05,
+      "loss": 0.163,
+      "step": 167530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2675664768396036e-05,
+      "loss": 0.1584,
+      "step": 167540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2673422716470115e-05,
+      "loss": 0.1543,
+      "step": 167550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2671180664544192e-05,
+      "loss": 0.1546,
+      "step": 167560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.266893861261827e-05,
+      "loss": 0.1655,
+      "step": 167570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2666696560692346e-05,
+      "loss": 0.1592,
+      "step": 167580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2664454508766424e-05,
+      "loss": 0.152,
+      "step": 167590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.26622124568405e-05,
+      "loss": 0.1564,
+      "step": 167600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2659970404914578e-05,
+      "loss": 0.1609,
+      "step": 167610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2657728352988657e-05,
+      "loss": 0.1589,
+      "step": 167620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2655486301062732e-05,
+      "loss": 0.1582,
+      "step": 167630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2653244249136811e-05,
+      "loss": 0.1622,
+      "step": 167640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2651002197210887e-05,
+      "loss": 0.156,
+      "step": 167650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2648760145284965e-05,
+      "loss": 0.1565,
+      "step": 167660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2646518093359042e-05,
+      "loss": 0.1625,
+      "step": 167670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2644276041433121e-05,
+      "loss": 0.165,
+      "step": 167680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2642033989507198e-05,
+      "loss": 0.1612,
+      "step": 167690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2639791937581275e-05,
+      "loss": 0.1586,
+      "step": 167700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2637549885655354e-05,
+      "loss": 0.159,
+      "step": 167710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2635307833729429e-05,
+      "loss": 0.1592,
+      "step": 167720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2633065781803508e-05,
+      "loss": 0.1622,
+      "step": 167730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2630823729877583e-05,
+      "loss": 0.158,
+      "step": 167740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2628581677951662e-05,
+      "loss": 0.1632,
+      "step": 167750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.262633962602574e-05,
+      "loss": 0.1614,
+      "step": 167760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2624097574099816e-05,
+      "loss": 0.1551,
+      "step": 167770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2621855522173895e-05,
+      "loss": 0.1566,
+      "step": 167780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.261961347024797e-05,
+      "loss": 0.1577,
+      "step": 167790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2617371418322049e-05,
+      "loss": 0.1616,
+      "step": 167800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2615129366396126e-05,
+      "loss": 0.159,
+      "step": 167810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2612887314470204e-05,
+      "loss": 0.1543,
+      "step": 167820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2610645262544281e-05,
+      "loss": 0.1639,
+      "step": 167830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2608403210618358e-05,
+      "loss": 0.1571,
+      "step": 167840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2606161158692437e-05,
+      "loss": 0.1579,
+      "step": 167850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2603919106766513e-05,
+      "loss": 0.1597,
+      "step": 167860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2601677054840591e-05,
+      "loss": 0.1539,
+      "step": 167870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2599435002914667e-05,
+      "loss": 0.1606,
+      "step": 167880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2597192950988745e-05,
+      "loss": 0.1619,
+      "step": 167890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2594950899062824e-05,
+      "loss": 0.1551,
+      "step": 167900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.25927088471369e-05,
+      "loss": 0.1612,
+      "step": 167910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2590466795210978e-05,
+      "loss": 0.156,
+      "step": 167920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2588224743285053e-05,
+      "loss": 0.1567,
+      "step": 167930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2585982691359132e-05,
+      "loss": 0.1602,
+      "step": 167940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2583740639433209e-05,
+      "loss": 0.1608,
+      "step": 167950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2581498587507288e-05,
+      "loss": 0.1591,
+      "step": 167960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2579256535581367e-05,
+      "loss": 0.1597,
+      "step": 167970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2577014483655442e-05,
+      "loss": 0.1504,
+      "step": 167980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.257477243172952e-05,
+      "loss": 0.1571,
+      "step": 167990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2572530379803596e-05,
+      "loss": 0.1638,
+      "step": 168000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2570288327877675e-05,
+      "loss": 0.1641,
+      "step": 168010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.256804627595175e-05,
+      "loss": 0.1588,
+      "step": 168020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2565804224025829e-05,
+      "loss": 0.1604,
+      "step": 168030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2563562172099907e-05,
+      "loss": 0.1558,
+      "step": 168040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2561320120173983e-05,
+      "loss": 0.1611,
+      "step": 168050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2559078068248061e-05,
+      "loss": 0.1552,
+      "step": 168060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2556836016322138e-05,
+      "loss": 0.1554,
+      "step": 168070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2554593964396216e-05,
+      "loss": 0.1578,
+      "step": 168080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2552351912470293e-05,
+      "loss": 0.1572,
+      "step": 168090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2550109860544371e-05,
+      "loss": 0.162,
+      "step": 168100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.254786780861845e-05,
+      "loss": 0.1579,
+      "step": 168110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2545625756692525e-05,
+      "loss": 0.1577,
+      "step": 168120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2543383704766604e-05,
+      "loss": 0.1606,
+      "step": 168130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.254114165284068e-05,
+      "loss": 0.1545,
+      "step": 168140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2538899600914758e-05,
+      "loss": 0.1612,
+      "step": 168150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2536657548988833e-05,
+      "loss": 0.1601,
+      "step": 168160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2534415497062912e-05,
+      "loss": 0.1619,
+      "step": 168170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.253217344513699e-05,
+      "loss": 0.1589,
+      "step": 168180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2529931393211066e-05,
+      "loss": 0.1542,
+      "step": 168190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2527689341285145e-05,
+      "loss": 0.1614,
+      "step": 168200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2525447289359222e-05,
+      "loss": 0.1559,
+      "step": 168210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.25232052374333e-05,
+      "loss": 0.1554,
+      "step": 168220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2520963185507376e-05,
+      "loss": 0.1566,
+      "step": 168230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2518721133581455e-05,
+      "loss": 0.1587,
+      "step": 168240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2516479081655533e-05,
+      "loss": 0.1567,
+      "step": 168250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2514237029729609e-05,
+      "loss": 0.1594,
+      "step": 168260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2511994977803687e-05,
+      "loss": 0.1539,
+      "step": 168270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2509752925877763e-05,
+      "loss": 0.1585,
+      "step": 168280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2507510873951841e-05,
+      "loss": 0.1601,
+      "step": 168290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.250526882202592e-05,
+      "loss": 0.1578,
+      "step": 168300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2503026770099996e-05,
+      "loss": 0.157,
+      "step": 168310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2500784718174074e-05,
+      "loss": 0.1567,
+      "step": 168320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.249854266624815e-05,
+      "loss": 0.1567,
+      "step": 168330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2496300614322228e-05,
+      "loss": 0.1566,
+      "step": 168340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2494058562396305e-05,
+      "loss": 0.1626,
+      "step": 168350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2491816510470384e-05,
+      "loss": 0.1548,
+      "step": 168360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2489574458544461e-05,
+      "loss": 0.1536,
+      "step": 168370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2487332406618538e-05,
+      "loss": 0.1556,
+      "step": 168380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2485090354692615e-05,
+      "loss": 0.1564,
+      "step": 168390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2482848302766692e-05,
+      "loss": 0.1595,
+      "step": 168400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.248060625084077e-05,
+      "loss": 0.16,
+      "step": 168410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2478364198914848e-05,
+      "loss": 0.161,
+      "step": 168420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2476122146988925e-05,
+      "loss": 0.1582,
+      "step": 168430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2473880095063002e-05,
+      "loss": 0.1565,
+      "step": 168440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2471638043137079e-05,
+      "loss": 0.159,
+      "step": 168450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2469395991211156e-05,
+      "loss": 0.1637,
+      "step": 168460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2467153939285233e-05,
+      "loss": 0.1628,
+      "step": 168470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2464911887359312e-05,
+      "loss": 0.1602,
+      "step": 168480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2462669835433389e-05,
+      "loss": 0.1571,
+      "step": 168490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2460427783507467e-05,
+      "loss": 0.161,
+      "step": 168500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2458185731581544e-05,
+      "loss": 0.1587,
+      "step": 168510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2455943679655621e-05,
+      "loss": 0.1567,
+      "step": 168520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2453701627729699e-05,
+      "loss": 0.1594,
+      "step": 168530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2451459575803776e-05,
+      "loss": 0.1484,
+      "step": 168540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2449217523877854e-05,
+      "loss": 0.1567,
+      "step": 168550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2446975471951931e-05,
+      "loss": 0.1545,
+      "step": 168560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2444733420026008e-05,
+      "loss": 0.1579,
+      "step": 168570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2442491368100085e-05,
+      "loss": 0.1512,
+      "step": 168580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2440249316174162e-05,
+      "loss": 0.1579,
+      "step": 168590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.243800726424824e-05,
+      "loss": 0.1592,
+      "step": 168600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2435765212322318e-05,
+      "loss": 0.1536,
+      "step": 168610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2433523160396395e-05,
+      "loss": 0.1663,
+      "step": 168620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2431281108470474e-05,
+      "loss": 0.1647,
+      "step": 168630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2429039056544551e-05,
+      "loss": 0.1564,
+      "step": 168640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2426797004618628e-05,
+      "loss": 0.1597,
+      "step": 168650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2424554952692705e-05,
+      "loss": 0.1548,
+      "step": 168660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2422312900766782e-05,
+      "loss": 0.1625,
+      "step": 168670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.242007084884086e-05,
+      "loss": 0.1593,
+      "step": 168680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2417828796914938e-05,
+      "loss": 0.1571,
+      "step": 168690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2415586744989015e-05,
+      "loss": 0.1544,
+      "step": 168700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2413344693063092e-05,
+      "loss": 0.1612,
+      "step": 168710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2411102641137169e-05,
+      "loss": 0.1615,
+      "step": 168720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2408860589211246e-05,
+      "loss": 0.1618,
+      "step": 168730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2406618537285323e-05,
+      "loss": 0.1591,
+      "step": 168740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2404376485359401e-05,
+      "loss": 0.153,
+      "step": 168750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2402134433433479e-05,
+      "loss": 0.1547,
+      "step": 168760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2399892381507557e-05,
+      "loss": 0.1581,
+      "step": 168770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2397650329581634e-05,
+      "loss": 0.1575,
+      "step": 168780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2395408277655711e-05,
+      "loss": 0.1532,
+      "step": 168790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2393166225729788e-05,
+      "loss": 0.1549,
+      "step": 168800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2390924173803865e-05,
+      "loss": 0.1588,
+      "step": 168810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2388682121877944e-05,
+      "loss": 0.1538,
+      "step": 168820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2386440069952021e-05,
+      "loss": 0.1549,
+      "step": 168830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2384198018026098e-05,
+      "loss": 0.1564,
+      "step": 168840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2381955966100175e-05,
+      "loss": 0.1565,
+      "step": 168850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2379713914174252e-05,
+      "loss": 0.1593,
+      "step": 168860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.237747186224833e-05,
+      "loss": 0.1558,
+      "step": 168870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2375229810322408e-05,
+      "loss": 0.1556,
+      "step": 168880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2372987758396485e-05,
+      "loss": 0.1578,
+      "step": 168890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2370745706470564e-05,
+      "loss": 0.1608,
+      "step": 168900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.236850365454464e-05,
+      "loss": 0.1541,
+      "step": 168910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2366261602618718e-05,
+      "loss": 0.1602,
+      "step": 168920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2364019550692795e-05,
+      "loss": 0.1564,
+      "step": 168930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2361777498766872e-05,
+      "loss": 0.16,
+      "step": 168940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2359535446840949e-05,
+      "loss": 0.1569,
+      "step": 168950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2357293394915027e-05,
+      "loss": 0.1523,
+      "step": 168960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2355051342989104e-05,
+      "loss": 0.1541,
+      "step": 168970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2352809291063182e-05,
+      "loss": 0.1618,
+      "step": 168980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2350567239137259e-05,
+      "loss": 0.1525,
+      "step": 168990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2348325187211336e-05,
+      "loss": 0.1602,
+      "step": 169000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2346083135285413e-05,
+      "loss": 0.1587,
+      "step": 169010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2343841083359491e-05,
+      "loss": 0.1619,
+      "step": 169020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2341599031433568e-05,
+      "loss": 0.1535,
+      "step": 169030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2339356979507647e-05,
+      "loss": 0.1599,
+      "step": 169040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2337114927581724e-05,
+      "loss": 0.1601,
+      "step": 169050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2334872875655801e-05,
+      "loss": 0.1565,
+      "step": 169060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2332630823729878e-05,
+      "loss": 0.1609,
+      "step": 169070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2330388771803955e-05,
+      "loss": 0.1553,
+      "step": 169080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2328146719878032e-05,
+      "loss": 0.156,
+      "step": 169090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2325904667952111e-05,
+      "loss": 0.1591,
+      "step": 169100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2323662616026188e-05,
+      "loss": 0.1553,
+      "step": 169110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2321420564100265e-05,
+      "loss": 0.1643,
+      "step": 169120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2319178512174342e-05,
+      "loss": 0.1603,
+      "step": 169130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2316936460248419e-05,
+      "loss": 0.156,
+      "step": 169140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2314694408322498e-05,
+      "loss": 0.1638,
+      "step": 169150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2312452356396575e-05,
+      "loss": 0.161,
+      "step": 169160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2310210304470653e-05,
+      "loss": 0.1611,
+      "step": 169170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.230796825254473e-05,
+      "loss": 0.16,
+      "step": 169180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2305726200618807e-05,
+      "loss": 0.1608,
+      "step": 169190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2303484148692885e-05,
+      "loss": 0.1632,
+      "step": 169200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2301242096766962e-05,
+      "loss": 0.1577,
+      "step": 169210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2299000044841039e-05,
+      "loss": 0.1546,
+      "step": 169220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2296757992915116e-05,
+      "loss": 0.1574,
+      "step": 169230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2294515940989194e-05,
+      "loss": 0.1591,
+      "step": 169240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2292273889063271e-05,
+      "loss": 0.1607,
+      "step": 169250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2290031837137348e-05,
+      "loss": 0.1581,
+      "step": 169260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2287789785211425e-05,
+      "loss": 0.1504,
+      "step": 169270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2285547733285502e-05,
+      "loss": 0.156,
+      "step": 169280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2283305681359581e-05,
+      "loss": 0.1604,
+      "step": 169290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2281063629433658e-05,
+      "loss": 0.1603,
+      "step": 169300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2278821577507737e-05,
+      "loss": 0.1532,
+      "step": 169310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2276579525581814e-05,
+      "loss": 0.1619,
+      "step": 169320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2274337473655891e-05,
+      "loss": 0.1628,
+      "step": 169330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2272095421729968e-05,
+      "loss": 0.1574,
+      "step": 169340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2269853369804045e-05,
+      "loss": 0.1629,
+      "step": 169350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2267611317878122e-05,
+      "loss": 0.1613,
+      "step": 169360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.22653692659522e-05,
+      "loss": 0.1584,
+      "step": 169370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2263127214026278e-05,
+      "loss": 0.1579,
+      "step": 169380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2260885162100355e-05,
+      "loss": 0.1607,
+      "step": 169390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2258643110174432e-05,
+      "loss": 0.1554,
+      "step": 169400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2256401058248509e-05,
+      "loss": 0.1546,
+      "step": 169410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2254159006322587e-05,
+      "loss": 0.1611,
+      "step": 169420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2251916954396665e-05,
+      "loss": 0.1562,
+      "step": 169430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2249674902470743e-05,
+      "loss": 0.1573,
+      "step": 169440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.224743285054482e-05,
+      "loss": 0.1572,
+      "step": 169450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2245190798618897e-05,
+      "loss": 0.1593,
+      "step": 169460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2242948746692974e-05,
+      "loss": 0.1612,
+      "step": 169470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2240706694767051e-05,
+      "loss": 0.1616,
+      "step": 169480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2238464642841128e-05,
+      "loss": 0.1683,
+      "step": 169490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2236222590915205e-05,
+      "loss": 0.1614,
+      "step": 169500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2233980538989284e-05,
+      "loss": 0.1542,
+      "step": 169510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2231962692255954e-05,
+      "loss": 0.1563,
+      "step": 169520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.222972064033003e-05,
+      "loss": 0.1531,
+      "step": 169530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2227478588404108e-05,
+      "loss": 0.1598,
+      "step": 169540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2225236536478185e-05,
+      "loss": 0.1611,
+      "step": 169550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2222994484552262e-05,
+      "loss": 0.1594,
+      "step": 169560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.222075243262634e-05,
+      "loss": 0.1592,
+      "step": 169570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2218510380700417e-05,
+      "loss": 0.1567,
+      "step": 169580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2216268328774494e-05,
+      "loss": 0.1585,
+      "step": 169590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2214026276848573e-05,
+      "loss": 0.1582,
+      "step": 169600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.221178422492265e-05,
+      "loss": 0.1584,
+      "step": 169610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2209542172996727e-05,
+      "loss": 0.1541,
+      "step": 169620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2207300121070804e-05,
+      "loss": 0.1575,
+      "step": 169630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2205058069144883e-05,
+      "loss": 0.1571,
+      "step": 169640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.220281601721896e-05,
+      "loss": 0.1572,
+      "step": 169650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2200573965293037e-05,
+      "loss": 0.1531,
+      "step": 169660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2198331913367114e-05,
+      "loss": 0.1534,
+      "step": 169670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2196089861441191e-05,
+      "loss": 0.1517,
+      "step": 169680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2193847809515268e-05,
+      "loss": 0.1599,
+      "step": 169690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2191605757589345e-05,
+      "loss": 0.158,
+      "step": 169700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2189363705663424e-05,
+      "loss": 0.1628,
+      "step": 169710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.21871216537375e-05,
+      "loss": 0.1521,
+      "step": 169720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.218487960181158e-05,
+      "loss": 0.1546,
+      "step": 169730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2182637549885657e-05,
+      "loss": 0.1603,
+      "step": 169740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2180395497959734e-05,
+      "loss": 0.1587,
+      "step": 169750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.217815344603381e-05,
+      "loss": 0.1601,
+      "step": 169760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2175911394107888e-05,
+      "loss": 0.1593,
+      "step": 169770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2173669342181966e-05,
+      "loss": 0.154,
+      "step": 169780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2171427290256043e-05,
+      "loss": 0.1579,
+      "step": 169790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.216918523833012e-05,
+      "loss": 0.1561,
+      "step": 169800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2166943186404197e-05,
+      "loss": 0.1581,
+      "step": 169810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2164701134478274e-05,
+      "loss": 0.1579,
+      "step": 169820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2162459082552351e-05,
+      "loss": 0.1562,
+      "step": 169830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2160217030626429e-05,
+      "loss": 0.1528,
+      "step": 169840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2157974978700507e-05,
+      "loss": 0.1611,
+      "step": 169850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2155732926774584e-05,
+      "loss": 0.1576,
+      "step": 169860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2153490874848663e-05,
+      "loss": 0.1553,
+      "step": 169870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.215124882292274e-05,
+      "loss": 0.1572,
+      "step": 169880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2149006770996817e-05,
+      "loss": 0.1561,
+      "step": 169890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2146764719070894e-05,
+      "loss": 0.1567,
+      "step": 169900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2144522667144971e-05,
+      "loss": 0.1562,
+      "step": 169910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.214228061521905e-05,
+      "loss": 0.1579,
+      "step": 169920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2140038563293127e-05,
+      "loss": 0.1628,
+      "step": 169930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2137796511367204e-05,
+      "loss": 0.1612,
+      "step": 169940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2135554459441281e-05,
+      "loss": 0.1565,
+      "step": 169950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2133312407515358e-05,
+      "loss": 0.1588,
+      "step": 169960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2131070355589435e-05,
+      "loss": 0.157,
+      "step": 169970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2128828303663514e-05,
+      "loss": 0.1581,
+      "step": 169980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.212658625173759e-05,
+      "loss": 0.1596,
+      "step": 169990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.212434419981167e-05,
+      "loss": 0.1587,
+      "step": 170000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2122102147885746e-05,
+      "loss": 0.1657,
+      "step": 170010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2119860095959823e-05,
+      "loss": 0.1617,
+      "step": 170020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.21176180440339e-05,
+      "loss": 0.1573,
+      "step": 170030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2115375992107977e-05,
+      "loss": 0.1517,
+      "step": 170040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2113133940182054e-05,
+      "loss": 0.1643,
+      "step": 170050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2110891888256133e-05,
+      "loss": 0.1644,
+      "step": 170060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.210864983633021e-05,
+      "loss": 0.1613,
+      "step": 170070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2106407784404287e-05,
+      "loss": 0.1619,
+      "step": 170080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2104165732478364e-05,
+      "loss": 0.1567,
+      "step": 170090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2101923680552441e-05,
+      "loss": 0.1557,
+      "step": 170100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2099681628626518e-05,
+      "loss": 0.165,
+      "step": 170110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2097439576700597e-05,
+      "loss": 0.16,
+      "step": 170120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2095197524774674e-05,
+      "loss": 0.1625,
+      "step": 170130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2092955472848753e-05,
+      "loss": 0.1542,
+      "step": 170140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.209071342092283e-05,
+      "loss": 0.1628,
+      "step": 170150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2088471368996907e-05,
+      "loss": 0.1593,
+      "step": 170160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2086229317070984e-05,
+      "loss": 0.164,
+      "step": 170170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2083987265145061e-05,
+      "loss": 0.1602,
+      "step": 170180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2081745213219138e-05,
+      "loss": 0.1538,
+      "step": 170190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2079503161293217e-05,
+      "loss": 0.1619,
+      "step": 170200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2077261109367294e-05,
+      "loss": 0.1621,
+      "step": 170210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.207501905744137e-05,
+      "loss": 0.1586,
+      "step": 170220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2072777005515448e-05,
+      "loss": 0.1559,
+      "step": 170230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2070534953589525e-05,
+      "loss": 0.1585,
+      "step": 170240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2068292901663603e-05,
+      "loss": 0.1588,
+      "step": 170250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.206605084973768e-05,
+      "loss": 0.1548,
+      "step": 170260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.206380879781176e-05,
+      "loss": 0.1611,
+      "step": 170270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2061566745885836e-05,
+      "loss": 0.1644,
+      "step": 170280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2059324693959913e-05,
+      "loss": 0.1623,
+      "step": 170290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.205708264203399e-05,
+      "loss": 0.1573,
+      "step": 170300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2054840590108067e-05,
+      "loss": 0.1548,
+      "step": 170310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2052598538182144e-05,
+      "loss": 0.1654,
+      "step": 170320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2050356486256223e-05,
+      "loss": 0.1644,
+      "step": 170330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.20481144343303e-05,
+      "loss": 0.1625,
+      "step": 170340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2045872382404377e-05,
+      "loss": 0.1517,
+      "step": 170350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2043630330478454e-05,
+      "loss": 0.1576,
+      "step": 170360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2041388278552531e-05,
+      "loss": 0.1589,
+      "step": 170370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2039146226626608e-05,
+      "loss": 0.1579,
+      "step": 170380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2036904174700687e-05,
+      "loss": 0.1622,
+      "step": 170390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2034662122774764e-05,
+      "loss": 0.1612,
+      "step": 170400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2032420070848843e-05,
+      "loss": 0.1565,
+      "step": 170410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.203017801892292e-05,
+      "loss": 0.1556,
+      "step": 170420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2027935966996997e-05,
+      "loss": 0.1603,
+      "step": 170430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2025693915071074e-05,
+      "loss": 0.1633,
+      "step": 170440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.202345186314515e-05,
+      "loss": 0.1622,
+      "step": 170450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2021209811219228e-05,
+      "loss": 0.1639,
+      "step": 170460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2018967759293306e-05,
+      "loss": 0.1603,
+      "step": 170470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2016725707367383e-05,
+      "loss": 0.163,
+      "step": 170480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.201448365544146e-05,
+      "loss": 0.1636,
+      "step": 170490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2012241603515537e-05,
+      "loss": 0.1588,
+      "step": 170500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2009999551589615e-05,
+      "loss": 0.1612,
+      "step": 170510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2007757499663693e-05,
+      "loss": 0.1567,
+      "step": 170520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.200551544773777e-05,
+      "loss": 0.1603,
+      "step": 170530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2003273395811849e-05,
+      "loss": 0.1546,
+      "step": 170540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2001031343885926e-05,
+      "loss": 0.1567,
+      "step": 170550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1998789291960003e-05,
+      "loss": 0.1602,
+      "step": 170560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.199654724003408e-05,
+      "loss": 0.1619,
+      "step": 170570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1994305188108157e-05,
+      "loss": 0.1571,
+      "step": 170580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1992063136182234e-05,
+      "loss": 0.1557,
+      "step": 170590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1989821084256311e-05,
+      "loss": 0.1572,
+      "step": 170600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.198757903233039e-05,
+      "loss": 0.1578,
+      "step": 170610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1985336980404467e-05,
+      "loss": 0.1599,
+      "step": 170620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1983094928478544e-05,
+      "loss": 0.163,
+      "step": 170630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1980852876552621e-05,
+      "loss": 0.1585,
+      "step": 170640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1978610824626698e-05,
+      "loss": 0.1631,
+      "step": 170650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1976368772700777e-05,
+      "loss": 0.1575,
+      "step": 170660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1974126720774854e-05,
+      "loss": 0.1507,
+      "step": 170670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1971884668848932e-05,
+      "loss": 0.1575,
+      "step": 170680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.196964261692301e-05,
+      "loss": 0.1547,
+      "step": 170690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1967400564997086e-05,
+      "loss": 0.1558,
+      "step": 170700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1965158513071163e-05,
+      "loss": 0.1566,
+      "step": 170710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.196291646114524e-05,
+      "loss": 0.1601,
+      "step": 170720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1960674409219317e-05,
+      "loss": 0.1563,
+      "step": 170730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1958432357293395e-05,
+      "loss": 0.1546,
+      "step": 170740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1956190305367473e-05,
+      "loss": 0.1553,
+      "step": 170750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.195394825344155e-05,
+      "loss": 0.1558,
+      "step": 170760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1951706201515627e-05,
+      "loss": 0.1511,
+      "step": 170770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1949464149589704e-05,
+      "loss": 0.1555,
+      "step": 170780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1947222097663783e-05,
+      "loss": 0.1626,
+      "step": 170790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.194498004573786e-05,
+      "loss": 0.1544,
+      "step": 170800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1942737993811937e-05,
+      "loss": 0.1595,
+      "step": 170810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1940495941886016e-05,
+      "loss": 0.1541,
+      "step": 170820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1938253889960093e-05,
+      "loss": 0.1594,
+      "step": 170830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.193601183803417e-05,
+      "loss": 0.1534,
+      "step": 170840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1933769786108247e-05,
+      "loss": 0.1556,
+      "step": 170850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1931527734182324e-05,
+      "loss": 0.1557,
+      "step": 170860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1929285682256401e-05,
+      "loss": 0.1581,
+      "step": 170870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1927043630330478e-05,
+      "loss": 0.1559,
+      "step": 170880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1924801578404557e-05,
+      "loss": 0.1497,
+      "step": 170890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1922559526478634e-05,
+      "loss": 0.1552,
+      "step": 170900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.192031747455271e-05,
+      "loss": 0.162,
+      "step": 170910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1918075422626788e-05,
+      "loss": 0.1571,
+      "step": 170920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1915833370700866e-05,
+      "loss": 0.1616,
+      "step": 170930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1913591318774943e-05,
+      "loss": 0.1596,
+      "step": 170940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1911349266849022e-05,
+      "loss": 0.1614,
+      "step": 170950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.19091072149231e-05,
+      "loss": 0.1552,
+      "step": 170960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1906865162997176e-05,
+      "loss": 0.1631,
+      "step": 170970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1904623111071253e-05,
+      "loss": 0.159,
+      "step": 170980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.190238105914533e-05,
+      "loss": 0.1526,
+      "step": 170990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1900139007219407e-05,
+      "loss": 0.1629,
+      "step": 171000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1897896955293484e-05,
+      "loss": 0.1629,
+      "step": 171010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1895654903367563e-05,
+      "loss": 0.16,
+      "step": 171020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.189341285144164e-05,
+      "loss": 0.1529,
+      "step": 171030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1891170799515717e-05,
+      "loss": 0.1591,
+      "step": 171040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1888928747589794e-05,
+      "loss": 0.1596,
+      "step": 171050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1886686695663873e-05,
+      "loss": 0.1653,
+      "step": 171060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.188444464373795e-05,
+      "loss": 0.1596,
+      "step": 171070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1882202591812027e-05,
+      "loss": 0.1539,
+      "step": 171080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1879960539886106e-05,
+      "loss": 0.1529,
+      "step": 171090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1877718487960183e-05,
+      "loss": 0.1622,
+      "step": 171100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.187547643603426e-05,
+      "loss": 0.1541,
+      "step": 171110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1873234384108337e-05,
+      "loss": 0.1547,
+      "step": 171120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1870992332182414e-05,
+      "loss": 0.1552,
+      "step": 171130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.186875028025649e-05,
+      "loss": 0.1559,
+      "step": 171140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1866508228330568e-05,
+      "loss": 0.1594,
+      "step": 171150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1864266176404646e-05,
+      "loss": 0.156,
+      "step": 171160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1862024124478723e-05,
+      "loss": 0.1567,
+      "step": 171170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.18597820725528e-05,
+      "loss": 0.1569,
+      "step": 171180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1857540020626878e-05,
+      "loss": 0.1553,
+      "step": 171190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1855297968700956e-05,
+      "loss": 0.1554,
+      "step": 171200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1853055916775033e-05,
+      "loss": 0.1529,
+      "step": 171210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.185081386484911e-05,
+      "loss": 0.154,
+      "step": 171220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1848571812923189e-05,
+      "loss": 0.1588,
+      "step": 171230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1846329760997266e-05,
+      "loss": 0.1591,
+      "step": 171240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1844087709071343e-05,
+      "loss": 0.1532,
+      "step": 171250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.184184565714542e-05,
+      "loss": 0.1612,
+      "step": 171260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1839603605219497e-05,
+      "loss": 0.1585,
+      "step": 171270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1837361553293574e-05,
+      "loss": 0.156,
+      "step": 171280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1835119501367651e-05,
+      "loss": 0.1576,
+      "step": 171290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.183287744944173e-05,
+      "loss": 0.1556,
+      "step": 171300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1830635397515807e-05,
+      "loss": 0.1621,
+      "step": 171310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1828393345589884e-05,
+      "loss": 0.1576,
+      "step": 171320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1826151293663963e-05,
+      "loss": 0.1612,
+      "step": 171330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.182390924173804e-05,
+      "loss": 0.1586,
+      "step": 171340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1821667189812117e-05,
+      "loss": 0.1588,
+      "step": 171350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.1819425137886194e-05,
+      "loss": 0.1574,
+      "step": 171360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1817183085960272e-05,
+      "loss": 0.1545,
+      "step": 171370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.181494103403435e-05,
+      "loss": 0.1585,
+      "step": 171380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1812698982108426e-05,
+      "loss": 0.1585,
+      "step": 171390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1810456930182503e-05,
+      "loss": 0.167,
+      "step": 171400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.180821487825658e-05,
+      "loss": 0.1551,
+      "step": 171410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1805972826330658e-05,
+      "loss": 0.1529,
+      "step": 171420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1803730774404735e-05,
+      "loss": 0.156,
+      "step": 171430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1801488722478813e-05,
+      "loss": 0.1602,
+      "step": 171440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.179924667055289e-05,
+      "loss": 0.1599,
+      "step": 171450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1797004618626967e-05,
+      "loss": 0.1577,
+      "step": 171460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1794762566701046e-05,
+      "loss": 0.1602,
+      "step": 171470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1792520514775123e-05,
+      "loss": 0.1559,
+      "step": 171480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.17902784628492e-05,
+      "loss": 0.1568,
+      "step": 171490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1788036410923277e-05,
+      "loss": 0.1571,
+      "step": 171500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1785794358997356e-05,
+      "loss": 0.1475,
+      "step": 171510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1783552307071433e-05,
+      "loss": 0.156,
+      "step": 171520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.178131025514551e-05,
+      "loss": 0.1564,
+      "step": 171530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1779068203219587e-05,
+      "loss": 0.1484,
+      "step": 171540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1776826151293664e-05,
+      "loss": 0.1625,
+      "step": 171550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1774584099367741e-05,
+      "loss": 0.1595,
+      "step": 171560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.177234204744182e-05,
+      "loss": 0.1656,
+      "step": 171570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1770099995515897e-05,
+      "loss": 0.1599,
+      "step": 171580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1767857943589974e-05,
+      "loss": 0.1562,
+      "step": 171590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.176561589166405e-05,
+      "loss": 0.1579,
+      "step": 171600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.176337383973813e-05,
+      "loss": 0.1622,
+      "step": 171610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1761131787812206e-05,
+      "loss": 0.1611,
+      "step": 171620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1758889735886284e-05,
+      "loss": 0.1563,
+      "step": 171630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1756647683960362e-05,
+      "loss": 0.1516,
+      "step": 171640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.175440563203444e-05,
+      "loss": 0.1649,
+      "step": 171650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1752163580108516e-05,
+      "loss": 0.1534,
+      "step": 171660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1749921528182593e-05,
+      "loss": 0.1618,
+      "step": 171670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.174767947625667e-05,
+      "loss": 0.1579,
+      "step": 171680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1745437424330747e-05,
+      "loss": 0.159,
+      "step": 171690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1743195372404824e-05,
+      "loss": 0.157,
+      "step": 171700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1740953320478903e-05,
+      "loss": 0.1627,
+      "step": 171710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.173871126855298e-05,
+      "loss": 0.159,
+      "step": 171720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1736469216627057e-05,
+      "loss": 0.1615,
+      "step": 171730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1734227164701136e-05,
+      "loss": 0.1641,
+      "step": 171740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1731985112775213e-05,
+      "loss": 0.1592,
+      "step": 171750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.172974306084929e-05,
+      "loss": 0.1607,
+      "step": 171760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1727501008923367e-05,
+      "loss": 0.1551,
+      "step": 171770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1725258956997446e-05,
+      "loss": 0.1594,
+      "step": 171780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1723016905071523e-05,
+      "loss": 0.1518,
+      "step": 171790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.17207748531456e-05,
+      "loss": 0.1611,
+      "step": 171800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1718532801219677e-05,
+      "loss": 0.158,
+      "step": 171810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1716290749293754e-05,
+      "loss": 0.156,
+      "step": 171820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.171404869736783e-05,
+      "loss": 0.1551,
+      "step": 171830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1711806645441908e-05,
+      "loss": 0.1563,
+      "step": 171840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1709564593515987e-05,
+      "loss": 0.155,
+      "step": 171850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1707322541590064e-05,
+      "loss": 0.1566,
+      "step": 171860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.170508048966414e-05,
+      "loss": 0.1604,
+      "step": 171870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.170283843773822e-05,
+      "loss": 0.1577,
+      "step": 171880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1700596385812296e-05,
+      "loss": 0.1585,
+      "step": 171890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1698354333886373e-05,
+      "loss": 0.1604,
+      "step": 171900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.169611228196045e-05,
+      "loss": 0.1686,
+      "step": 171910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1693870230034529e-05,
+      "loss": 0.1578,
+      "step": 171920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1691628178108606e-05,
+      "loss": 0.1562,
+      "step": 171930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1689386126182683e-05,
+      "loss": 0.1531,
+      "step": 171940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.168714407425676e-05,
+      "loss": 0.1565,
+      "step": 171950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1684902022330837e-05,
+      "loss": 0.1527,
+      "step": 171960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1682659970404914e-05,
+      "loss": 0.1523,
+      "step": 171970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1680417918478991e-05,
+      "loss": 0.1542,
+      "step": 171980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.167817586655307e-05,
+      "loss": 0.1566,
+      "step": 171990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1675933814627147e-05,
+      "loss": 0.1565,
+      "step": 172000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1673691762701226e-05,
+      "loss": 0.1553,
+      "step": 172010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1671449710775303e-05,
+      "loss": 0.1563,
+      "step": 172020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.166920765884938e-05,
+      "loss": 0.1583,
+      "step": 172030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1666965606923457e-05,
+      "loss": 0.1543,
+      "step": 172040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1664723554997534e-05,
+      "loss": 0.1536,
+      "step": 172050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1662481503071612e-05,
+      "loss": 0.1567,
+      "step": 172060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.166023945114569e-05,
+      "loss": 0.1575,
+      "step": 172070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1657997399219767e-05,
+      "loss": 0.1642,
+      "step": 172080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1655755347293844e-05,
+      "loss": 0.1558,
+      "step": 172090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.165351329536792e-05,
+      "loss": 0.1565,
+      "step": 172100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1651271243441998e-05,
+      "loss": 0.1586,
+      "step": 172110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1649029191516075e-05,
+      "loss": 0.1546,
+      "step": 172120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1646787139590153e-05,
+      "loss": 0.1548,
+      "step": 172130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.164454508766423e-05,
+      "loss": 0.1585,
+      "step": 172140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1642303035738309e-05,
+      "loss": 0.1641,
+      "step": 172150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1640060983812386e-05,
+      "loss": 0.1575,
+      "step": 172160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1637818931886463e-05,
+      "loss": 0.1539,
+      "step": 172170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.163557687996054e-05,
+      "loss": 0.1569,
+      "step": 172180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1633334828034617e-05,
+      "loss": 0.1593,
+      "step": 172190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1631092776108696e-05,
+      "loss": 0.1592,
+      "step": 172200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1628850724182773e-05,
+      "loss": 0.1591,
+      "step": 172210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.162660867225685e-05,
+      "loss": 0.1568,
+      "step": 172220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1624366620330927e-05,
+      "loss": 0.1511,
+      "step": 172230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1622124568405004e-05,
+      "loss": 0.1566,
+      "step": 172240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1619882516479081e-05,
+      "loss": 0.1585,
+      "step": 172250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.161764046455316e-05,
+      "loss": 0.1535,
+      "step": 172260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1615398412627237e-05,
+      "loss": 0.1614,
+      "step": 172270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1613156360701315e-05,
+      "loss": 0.1628,
+      "step": 172280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1610914308775392e-05,
+      "loss": 0.157,
+      "step": 172290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1608896462042062e-05,
+      "loss": 0.1534,
+      "step": 172300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1606654410116139e-05,
+      "loss": 0.1554,
+      "step": 172310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1604412358190216e-05,
+      "loss": 0.1627,
+      "step": 172320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1602170306264295e-05,
+      "loss": 0.1543,
+      "step": 172330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1599928254338372e-05,
+      "loss": 0.1596,
+      "step": 172340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1597686202412449e-05,
+      "loss": 0.1648,
+      "step": 172350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1595444150486526e-05,
+      "loss": 0.1547,
+      "step": 172360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1593202098560603e-05,
+      "loss": 0.1546,
+      "step": 172370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.159096004663468e-05,
+      "loss": 0.1581,
+      "step": 172380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1588717994708757e-05,
+      "loss": 0.1588,
+      "step": 172390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1586475942782836e-05,
+      "loss": 0.1589,
+      "step": 172400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1584233890856913e-05,
+      "loss": 0.1565,
+      "step": 172410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.158199183893099e-05,
+      "loss": 0.1599,
+      "step": 172420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1579749787005067e-05,
+      "loss": 0.1548,
+      "step": 172430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1577507735079145e-05,
+      "loss": 0.1548,
+      "step": 172440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1575265683153222e-05,
+      "loss": 0.1584,
+      "step": 172450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.15730236312273e-05,
+      "loss": 0.1624,
+      "step": 172460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1570781579301378e-05,
+      "loss": 0.1553,
+      "step": 172470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1568539527375455e-05,
+      "loss": 0.1563,
+      "step": 172480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1566297475449532e-05,
+      "loss": 0.1565,
+      "step": 172490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.156405542352361e-05,
+      "loss": 0.1587,
+      "step": 172500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1561813371597686e-05,
+      "loss": 0.1589,
+      "step": 172510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1559571319671763e-05,
+      "loss": 0.1545,
+      "step": 172520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1557329267745842e-05,
+      "loss": 0.1613,
+      "step": 172530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1555087215819919e-05,
+      "loss": 0.1569,
+      "step": 172540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1552845163893996e-05,
+      "loss": 0.1538,
+      "step": 172550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1550603111968073e-05,
+      "loss": 0.1545,
+      "step": 172560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1548361060042152e-05,
+      "loss": 0.1565,
+      "step": 172570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1546119008116229e-05,
+      "loss": 0.158,
+      "step": 172580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1543876956190306e-05,
+      "loss": 0.1571,
+      "step": 172590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1541634904264385e-05,
+      "loss": 0.1534,
+      "step": 172600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1539392852338462e-05,
+      "loss": 0.1635,
+      "step": 172610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1537150800412539e-05,
+      "loss": 0.1576,
+      "step": 172620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1534908748486616e-05,
+      "loss": 0.1569,
+      "step": 172630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1532666696560693e-05,
+      "loss": 0.1551,
+      "step": 172640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.153042464463477e-05,
+      "loss": 0.1583,
+      "step": 172650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1528182592708847e-05,
+      "loss": 0.1531,
+      "step": 172660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1525940540782925e-05,
+      "loss": 0.155,
+      "step": 172670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1523698488857002e-05,
+      "loss": 0.1653,
+      "step": 172680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.152145643693108e-05,
+      "loss": 0.1577,
+      "step": 172690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1519214385005156e-05,
+      "loss": 0.1562,
+      "step": 172700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1516972333079235e-05,
+      "loss": 0.1634,
+      "step": 172710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1514730281153312e-05,
+      "loss": 0.1615,
+      "step": 172720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.151248822922739e-05,
+      "loss": 0.1593,
+      "step": 172730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1510246177301468e-05,
+      "loss": 0.148,
+      "step": 172740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1508004125375545e-05,
+      "loss": 0.1583,
+      "step": 172750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1505762073449622e-05,
+      "loss": 0.1569,
+      "step": 172760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1503520021523699e-05,
+      "loss": 0.1576,
+      "step": 172770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1501277969597776e-05,
+      "loss": 0.1543,
+      "step": 172780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1499035917671853e-05,
+      "loss": 0.1597,
+      "step": 172790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.149679386574593e-05,
+      "loss": 0.1603,
+      "step": 172800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1494551813820009e-05,
+      "loss": 0.1549,
+      "step": 172810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1492309761894086e-05,
+      "loss": 0.1596,
+      "step": 172820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1490067709968163e-05,
+      "loss": 0.1547,
+      "step": 172830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1487825658042242e-05,
+      "loss": 0.1559,
+      "step": 172840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1485583606116319e-05,
+      "loss": 0.1572,
+      "step": 172850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1483341554190396e-05,
+      "loss": 0.162,
+      "step": 172860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1481099502264473e-05,
+      "loss": 0.1568,
+      "step": 172870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1478857450338551e-05,
+      "loss": 0.1597,
+      "step": 172880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1476615398412628e-05,
+      "loss": 0.1607,
+      "step": 172890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1474373346486705e-05,
+      "loss": 0.1555,
+      "step": 172900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1472131294560782e-05,
+      "loss": 0.1575,
+      "step": 172910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.146988924263486e-05,
+      "loss": 0.1533,
+      "step": 172920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1467647190708936e-05,
+      "loss": 0.1577,
+      "step": 172930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1465405138783014e-05,
+      "loss": 0.1516,
+      "step": 172940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1463163086857092e-05,
+      "loss": 0.1545,
+      "step": 172950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.146092103493117e-05,
+      "loss": 0.1561,
+      "step": 172960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1458678983005246e-05,
+      "loss": 0.1595,
+      "step": 172970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1456436931079325e-05,
+      "loss": 0.1542,
+      "step": 172980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1454194879153402e-05,
+      "loss": 0.1614,
+      "step": 172990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1451952827227479e-05,
+      "loss": 0.1557,
+      "step": 173000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1449710775301556e-05,
+      "loss": 0.1576,
+      "step": 173010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1447468723375635e-05,
+      "loss": 0.1576,
+      "step": 173020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1445226671449712e-05,
+      "loss": 0.1621,
+      "step": 173030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1442984619523789e-05,
+      "loss": 0.1564,
+      "step": 173040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1440742567597866e-05,
+      "loss": 0.1564,
+      "step": 173050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1438500515671943e-05,
+      "loss": 0.1576,
+      "step": 173060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.143625846374602e-05,
+      "loss": 0.1488,
+      "step": 173070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1434016411820097e-05,
+      "loss": 0.1592,
+      "step": 173080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1431774359894176e-05,
+      "loss": 0.157,
+      "step": 173090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1429532307968253e-05,
+      "loss": 0.1654,
+      "step": 173100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1427290256042331e-05,
+      "loss": 0.1561,
+      "step": 173110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1425048204116408e-05,
+      "loss": 0.1539,
+      "step": 173120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1422806152190485e-05,
+      "loss": 0.1559,
+      "step": 173130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1420564100264562e-05,
+      "loss": 0.1564,
+      "step": 173140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.141832204833864e-05,
+      "loss": 0.1539,
+      "step": 173150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1416079996412718e-05,
+      "loss": 0.1555,
+      "step": 173160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1413837944486795e-05,
+      "loss": 0.1538,
+      "step": 173170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1411595892560872e-05,
+      "loss": 0.1545,
+      "step": 173180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.140935384063495e-05,
+      "loss": 0.159,
+      "step": 173190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1407111788709026e-05,
+      "loss": 0.1637,
+      "step": 173200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1404869736783103e-05,
+      "loss": 0.1525,
+      "step": 173210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1402627684857182e-05,
+      "loss": 0.1593,
+      "step": 173220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1400385632931259e-05,
+      "loss": 0.1506,
+      "step": 173230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1398143581005336e-05,
+      "loss": 0.1557,
+      "step": 173240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1395901529079415e-05,
+      "loss": 0.1611,
+      "step": 173250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1393659477153492e-05,
+      "loss": 0.1613,
+      "step": 173260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1391417425227569e-05,
+      "loss": 0.1524,
+      "step": 173270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1389175373301646e-05,
+      "loss": 0.161,
+      "step": 173280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1386933321375725e-05,
+      "loss": 0.1612,
+      "step": 173290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1384691269449802e-05,
+      "loss": 0.1525,
+      "step": 173300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1382449217523879e-05,
+      "loss": 0.1562,
+      "step": 173310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1380207165597956e-05,
+      "loss": 0.1559,
+      "step": 173320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1377965113672033e-05,
+      "loss": 0.1589,
+      "step": 173330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.137572306174611e-05,
+      "loss": 0.1585,
+      "step": 173340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1373481009820187e-05,
+      "loss": 0.1511,
+      "step": 173350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1371238957894265e-05,
+      "loss": 0.1542,
+      "step": 173360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1368996905968342e-05,
+      "loss": 0.1579,
+      "step": 173370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1366754854042421e-05,
+      "loss": 0.1555,
+      "step": 173380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1364512802116498e-05,
+      "loss": 0.1626,
+      "step": 173390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1362270750190575e-05,
+      "loss": 0.1551,
+      "step": 173400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1360028698264652e-05,
+      "loss": 0.1519,
+      "step": 173410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.135778664633873e-05,
+      "loss": 0.1525,
+      "step": 173420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1355544594412808e-05,
+      "loss": 0.1507,
+      "step": 173430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1353302542486885e-05,
+      "loss": 0.1574,
+      "step": 173440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1351060490560962e-05,
+      "loss": 0.1616,
+      "step": 173450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1348818438635039e-05,
+      "loss": 0.1516,
+      "step": 173460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1346576386709116e-05,
+      "loss": 0.1557,
+      "step": 173470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1344334334783193e-05,
+      "loss": 0.1609,
+      "step": 173480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.134209228285727e-05,
+      "loss": 0.1587,
+      "step": 173490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1339850230931349e-05,
+      "loss": 0.1509,
+      "step": 173500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1337608179005426e-05,
+      "loss": 0.1523,
+      "step": 173510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1335366127079505e-05,
+      "loss": 0.162,
+      "step": 173520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1333124075153582e-05,
+      "loss": 0.1572,
+      "step": 173530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1330882023227659e-05,
+      "loss": 0.1541,
+      "step": 173540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1328639971301736e-05,
+      "loss": 0.1598,
+      "step": 173550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1326397919375813e-05,
+      "loss": 0.1638,
+      "step": 173560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1324155867449891e-05,
+      "loss": 0.1597,
+      "step": 173570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1321913815523968e-05,
+      "loss": 0.1579,
+      "step": 173580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1319671763598045e-05,
+      "loss": 0.1574,
+      "step": 173590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.1317429711672122e-05,
+      "loss": 0.1522,
+      "step": 173600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.13151876597462e-05,
+      "loss": 0.1623,
+      "step": 173610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1312945607820277e-05,
+      "loss": 0.1521,
+      "step": 173620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1310703555894355e-05,
+      "loss": 0.1531,
+      "step": 173630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1308461503968432e-05,
+      "loss": 0.1555,
+      "step": 173640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1306219452042511e-05,
+      "loss": 0.1541,
+      "step": 173650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1303977400116588e-05,
+      "loss": 0.1576,
+      "step": 173660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1301735348190665e-05,
+      "loss": 0.1573,
+      "step": 173670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1299493296264742e-05,
+      "loss": 0.1556,
+      "step": 173680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1297251244338819e-05,
+      "loss": 0.1628,
+      "step": 173690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1295009192412896e-05,
+      "loss": 0.1536,
+      "step": 173700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1292767140486975e-05,
+      "loss": 0.1581,
+      "step": 173710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1290525088561052e-05,
+      "loss": 0.1583,
+      "step": 173720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1288283036635129e-05,
+      "loss": 0.1616,
+      "step": 173730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1286040984709206e-05,
+      "loss": 0.1619,
+      "step": 173740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1283798932783283e-05,
+      "loss": 0.1507,
+      "step": 173750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.128155688085736e-05,
+      "loss": 0.1574,
+      "step": 173760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1279314828931439e-05,
+      "loss": 0.1606,
+      "step": 173770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1277072777005516e-05,
+      "loss": 0.1609,
+      "step": 173780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1274830725079594e-05,
+      "loss": 0.1603,
+      "step": 173790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1272588673153671e-05,
+      "loss": 0.154,
+      "step": 173800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1270346621227748e-05,
+      "loss": 0.1516,
+      "step": 173810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1268104569301825e-05,
+      "loss": 0.1541,
+      "step": 173820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1265862517375903e-05,
+      "loss": 0.1589,
+      "step": 173830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.126362046544998e-05,
+      "loss": 0.1618,
+      "step": 173840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1261378413524058e-05,
+      "loss": 0.1524,
+      "step": 173850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1259136361598135e-05,
+      "loss": 0.1619,
+      "step": 173860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1256894309672212e-05,
+      "loss": 0.1568,
+      "step": 173870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.125465225774629e-05,
+      "loss": 0.1534,
+      "step": 173880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1252410205820366e-05,
+      "loss": 0.1622,
+      "step": 173890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1250168153894445e-05,
+      "loss": 0.152,
+      "step": 173900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1247926101968522e-05,
+      "loss": 0.1557,
+      "step": 173910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.12456840500426e-05,
+      "loss": 0.1526,
+      "step": 173920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1243441998116678e-05,
+      "loss": 0.1592,
+      "step": 173930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1241199946190755e-05,
+      "loss": 0.1589,
+      "step": 173940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1238957894264832e-05,
+      "loss": 0.1567,
+      "step": 173950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1236715842338909e-05,
+      "loss": 0.1566,
+      "step": 173960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1234473790412986e-05,
+      "loss": 0.1568,
+      "step": 173970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1232231738487065e-05,
+      "loss": 0.1543,
+      "step": 173980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1229989686561142e-05,
+      "loss": 0.1529,
+      "step": 173990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1227747634635219e-05,
+      "loss": 0.1627,
+      "step": 174000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1225505582709296e-05,
+      "loss": 0.1594,
+      "step": 174010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1223263530783373e-05,
+      "loss": 0.158,
+      "step": 174020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.122102147885745e-05,
+      "loss": 0.1538,
+      "step": 174030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1218779426931528e-05,
+      "loss": 0.1545,
+      "step": 174040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1216537375005605e-05,
+      "loss": 0.1559,
+      "step": 174050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1214295323079684e-05,
+      "loss": 0.1604,
+      "step": 174060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1212053271153761e-05,
+      "loss": 0.1561,
+      "step": 174070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1209811219227838e-05,
+      "loss": 0.1591,
+      "step": 174080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1207569167301915e-05,
+      "loss": 0.1557,
+      "step": 174090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1205327115375992e-05,
+      "loss": 0.1536,
+      "step": 174100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.120308506345007e-05,
+      "loss": 0.1569,
+      "step": 174110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1200843011524148e-05,
+      "loss": 0.1613,
+      "step": 174120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1198600959598225e-05,
+      "loss": 0.1526,
+      "step": 174130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1196358907672302e-05,
+      "loss": 0.1577,
+      "step": 174140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1194116855746379e-05,
+      "loss": 0.1554,
+      "step": 174150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1191874803820456e-05,
+      "loss": 0.1609,
+      "step": 174160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1189632751894535e-05,
+      "loss": 0.154,
+      "step": 174170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1187390699968612e-05,
+      "loss": 0.1584,
+      "step": 174180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.118514864804269e-05,
+      "loss": 0.1538,
+      "step": 174190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1182906596116768e-05,
+      "loss": 0.1514,
+      "step": 174200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1180664544190845e-05,
+      "loss": 0.1548,
+      "step": 174210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1178422492264922e-05,
+      "loss": 0.1533,
+      "step": 174220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1176180440338999e-05,
+      "loss": 0.1585,
+      "step": 174230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1173938388413076e-05,
+      "loss": 0.1577,
+      "step": 174240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1171696336487153e-05,
+      "loss": 0.1581,
+      "step": 174250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1169454284561231e-05,
+      "loss": 0.1578,
+      "step": 174260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1167212232635308e-05,
+      "loss": 0.1559,
+      "step": 174270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1164970180709386e-05,
+      "loss": 0.1496,
+      "step": 174280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1162728128783463e-05,
+      "loss": 0.1638,
+      "step": 174290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.116048607685754e-05,
+      "loss": 0.1593,
+      "step": 174300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1158244024931618e-05,
+      "loss": 0.1597,
+      "step": 174310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1156001973005695e-05,
+      "loss": 0.1533,
+      "step": 174320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1153759921079774e-05,
+      "loss": 0.1582,
+      "step": 174330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1151517869153851e-05,
+      "loss": 0.1541,
+      "step": 174340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1149275817227928e-05,
+      "loss": 0.1511,
+      "step": 174350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1147033765302005e-05,
+      "loss": 0.1552,
+      "step": 174360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1144791713376082e-05,
+      "loss": 0.1519,
+      "step": 174370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1142549661450159e-05,
+      "loss": 0.1555,
+      "step": 174380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1140307609524236e-05,
+      "loss": 0.1573,
+      "step": 174390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1138065557598315e-05,
+      "loss": 0.1665,
+      "step": 174400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1135823505672392e-05,
+      "loss": 0.1547,
+      "step": 174410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1133581453746469e-05,
+      "loss": 0.1551,
+      "step": 174420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1131339401820546e-05,
+      "loss": 0.1552,
+      "step": 174430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1129097349894625e-05,
+      "loss": 0.1569,
+      "step": 174440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1126855297968702e-05,
+      "loss": 0.1585,
+      "step": 174450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1124613246042779e-05,
+      "loss": 0.1546,
+      "step": 174460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1122371194116857e-05,
+      "loss": 0.1569,
+      "step": 174470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1120129142190934e-05,
+      "loss": 0.1564,
+      "step": 174480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1117887090265011e-05,
+      "loss": 0.1494,
+      "step": 174490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1115645038339089e-05,
+      "loss": 0.1572,
+      "step": 174500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1113402986413166e-05,
+      "loss": 0.1595,
+      "step": 174510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1111160934487243e-05,
+      "loss": 0.1576,
+      "step": 174520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.110891888256132e-05,
+      "loss": 0.1543,
+      "step": 174530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1106676830635398e-05,
+      "loss": 0.1591,
+      "step": 174540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1104434778709475e-05,
+      "loss": 0.1545,
+      "step": 174550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1102192726783552e-05,
+      "loss": 0.1548,
+      "step": 174560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.109995067485763e-05,
+      "loss": 0.1574,
+      "step": 174570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1097708622931708e-05,
+      "loss": 0.1508,
+      "step": 174580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1095466571005785e-05,
+      "loss": 0.1554,
+      "step": 174590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1093224519079864e-05,
+      "loss": 0.1553,
+      "step": 174600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.109098246715394e-05,
+      "loss": 0.1556,
+      "step": 174610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1088740415228018e-05,
+      "loss": 0.1566,
+      "step": 174620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1086722568494687e-05,
+      "loss": 0.1574,
+      "step": 174630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1084480516568764e-05,
+      "loss": 0.1574,
+      "step": 174640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1082238464642841e-05,
+      "loss": 0.1586,
+      "step": 174650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1079996412716918e-05,
+      "loss": 0.1607,
+      "step": 174660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1077754360790997e-05,
+      "loss": 0.1591,
+      "step": 174670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1075512308865074e-05,
+      "loss": 0.1556,
+      "step": 174680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1073270256939151e-05,
+      "loss": 0.151,
+      "step": 174690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1071028205013228e-05,
+      "loss": 0.1572,
+      "step": 174700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1068786153087305e-05,
+      "loss": 0.157,
+      "step": 174710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1066544101161382e-05,
+      "loss": 0.1579,
+      "step": 174720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1064302049235461e-05,
+      "loss": 0.1566,
+      "step": 174730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1062059997309538e-05,
+      "loss": 0.1558,
+      "step": 174740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1059817945383617e-05,
+      "loss": 0.16,
+      "step": 174750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1057575893457694e-05,
+      "loss": 0.154,
+      "step": 174760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.105533384153177e-05,
+      "loss": 0.156,
+      "step": 174770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1053091789605848e-05,
+      "loss": 0.1572,
+      "step": 174780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1050849737679925e-05,
+      "loss": 0.1597,
+      "step": 174790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1048607685754002e-05,
+      "loss": 0.1579,
+      "step": 174800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.104636563382808e-05,
+      "loss": 0.155,
+      "step": 174810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1044123581902158e-05,
+      "loss": 0.1558,
+      "step": 174820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1041881529976235e-05,
+      "loss": 0.1603,
+      "step": 174830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1039639478050312e-05,
+      "loss": 0.152,
+      "step": 174840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1037397426124389e-05,
+      "loss": 0.156,
+      "step": 174850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1035155374198466e-05,
+      "loss": 0.154,
+      "step": 174860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1032913322272544e-05,
+      "loss": 0.1549,
+      "step": 174870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1030671270346621e-05,
+      "loss": 0.1543,
+      "step": 174880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.10284292184207e-05,
+      "loss": 0.1524,
+      "step": 174890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1026187166494777e-05,
+      "loss": 0.1513,
+      "step": 174900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1023945114568854e-05,
+      "loss": 0.1645,
+      "step": 174910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1021703062642931e-05,
+      "loss": 0.1571,
+      "step": 174920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1019461010717008e-05,
+      "loss": 0.1509,
+      "step": 174930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1017218958791087e-05,
+      "loss": 0.1624,
+      "step": 174940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1014976906865164e-05,
+      "loss": 0.1552,
+      "step": 174950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1012734854939241e-05,
+      "loss": 0.158,
+      "step": 174960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1010492803013318e-05,
+      "loss": 0.1569,
+      "step": 174970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1008250751087395e-05,
+      "loss": 0.153,
+      "step": 174980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1006008699161472e-05,
+      "loss": 0.1523,
+      "step": 174990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.100376664723555e-05,
+      "loss": 0.156,
+      "step": 175000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1001524595309628e-05,
+      "loss": 0.159,
+      "step": 175010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0999282543383707e-05,
+      "loss": 0.1525,
+      "step": 175020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0997040491457784e-05,
+      "loss": 0.162,
+      "step": 175030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.099479843953186e-05,
+      "loss": 0.15,
+      "step": 175040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0992556387605938e-05,
+      "loss": 0.1568,
+      "step": 175050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0990314335680015e-05,
+      "loss": 0.1552,
+      "step": 175060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0988072283754092e-05,
+      "loss": 0.1541,
+      "step": 175070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.098583023182817e-05,
+      "loss": 0.1534,
+      "step": 175080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0983588179902247e-05,
+      "loss": 0.1551,
+      "step": 175090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0981346127976324e-05,
+      "loss": 0.1538,
+      "step": 175100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0979104076050401e-05,
+      "loss": 0.1572,
+      "step": 175110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0976862024124478e-05,
+      "loss": 0.1546,
+      "step": 175120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0974619972198555e-05,
+      "loss": 0.1584,
+      "step": 175130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0972377920272634e-05,
+      "loss": 0.1542,
+      "step": 175140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0970135868346711e-05,
+      "loss": 0.1535,
+      "step": 175150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.096789381642079e-05,
+      "loss": 0.1575,
+      "step": 175160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0965651764494867e-05,
+      "loss": 0.1542,
+      "step": 175170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0963409712568944e-05,
+      "loss": 0.1542,
+      "step": 175180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0961167660643021e-05,
+      "loss": 0.1525,
+      "step": 175190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0958925608717098e-05,
+      "loss": 0.1591,
+      "step": 175200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0956683556791175e-05,
+      "loss": 0.1546,
+      "step": 175210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0954441504865254e-05,
+      "loss": 0.1567,
+      "step": 175220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.095219945293933e-05,
+      "loss": 0.1538,
+      "step": 175230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0949957401013408e-05,
+      "loss": 0.1592,
+      "step": 175240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0947715349087485e-05,
+      "loss": 0.1566,
+      "step": 175250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0945473297161562e-05,
+      "loss": 0.1536,
+      "step": 175260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.094323124523564e-05,
+      "loss": 0.1555,
+      "step": 175270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0940989193309718e-05,
+      "loss": 0.1526,
+      "step": 175280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0938747141383796e-05,
+      "loss": 0.1623,
+      "step": 175290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0936505089457873e-05,
+      "loss": 0.1584,
+      "step": 175300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.093426303753195e-05,
+      "loss": 0.1497,
+      "step": 175310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0932020985606027e-05,
+      "loss": 0.1541,
+      "step": 175320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0929778933680104e-05,
+      "loss": 0.1527,
+      "step": 175330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0927536881754181e-05,
+      "loss": 0.1565,
+      "step": 175340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0925294829828258e-05,
+      "loss": 0.1575,
+      "step": 175350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0923052777902337e-05,
+      "loss": 0.1524,
+      "step": 175360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0920810725976414e-05,
+      "loss": 0.153,
+      "step": 175370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0918568674050491e-05,
+      "loss": 0.1557,
+      "step": 175380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0916326622124568e-05,
+      "loss": 0.1556,
+      "step": 175390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0914084570198645e-05,
+      "loss": 0.1534,
+      "step": 175400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0911842518272724e-05,
+      "loss": 0.1576,
+      "step": 175410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0909600466346801e-05,
+      "loss": 0.1565,
+      "step": 175420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.090735841442088e-05,
+      "loss": 0.1563,
+      "step": 175430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0905116362494957e-05,
+      "loss": 0.1557,
+      "step": 175440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0902874310569034e-05,
+      "loss": 0.1576,
+      "step": 175450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.090063225864311e-05,
+      "loss": 0.1535,
+      "step": 175460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0898390206717188e-05,
+      "loss": 0.1546,
+      "step": 175470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0896148154791265e-05,
+      "loss": 0.1459,
+      "step": 175480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0893906102865344e-05,
+      "loss": 0.1469,
+      "step": 175490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.089166405093942e-05,
+      "loss": 0.1509,
+      "step": 175500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0889421999013498e-05,
+      "loss": 0.1561,
+      "step": 175510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0887179947087575e-05,
+      "loss": 0.1541,
+      "step": 175520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0884937895161652e-05,
+      "loss": 0.15,
+      "step": 175530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0882695843235729e-05,
+      "loss": 0.1523,
+      "step": 175540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0880453791309807e-05,
+      "loss": 0.1557,
+      "step": 175550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0878211739383886e-05,
+      "loss": 0.1551,
+      "step": 175560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0875969687457963e-05,
+      "loss": 0.1591,
+      "step": 175570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.087372763553204e-05,
+      "loss": 0.1586,
+      "step": 175580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0871485583606117e-05,
+      "loss": 0.1599,
+      "step": 175590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0869243531680194e-05,
+      "loss": 0.1537,
+      "step": 175600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0867001479754271e-05,
+      "loss": 0.1566,
+      "step": 175610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0864759427828348e-05,
+      "loss": 0.1554,
+      "step": 175620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0862517375902427e-05,
+      "loss": 0.1593,
+      "step": 175630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0860275323976504e-05,
+      "loss": 0.1595,
+      "step": 175640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0858033272050581e-05,
+      "loss": 0.1539,
+      "step": 175650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0855791220124658e-05,
+      "loss": 0.1528,
+      "step": 175660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0853549168198735e-05,
+      "loss": 0.1551,
+      "step": 175670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0851307116272814e-05,
+      "loss": 0.1552,
+      "step": 175680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.084906506434689e-05,
+      "loss": 0.1553,
+      "step": 175690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.084682301242097e-05,
+      "loss": 0.1608,
+      "step": 175700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0844580960495047e-05,
+      "loss": 0.1557,
+      "step": 175710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0842338908569124e-05,
+      "loss": 0.1533,
+      "step": 175720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.08400968566432e-05,
+      "loss": 0.1552,
+      "step": 175730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0837854804717278e-05,
+      "loss": 0.1512,
+      "step": 175740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0835612752791355e-05,
+      "loss": 0.1572,
+      "step": 175750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0833370700865432e-05,
+      "loss": 0.1561,
+      "step": 175760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.083112864893951e-05,
+      "loss": 0.1531,
+      "step": 175770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0828886597013587e-05,
+      "loss": 0.1532,
+      "step": 175780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0826644545087664e-05,
+      "loss": 0.1546,
+      "step": 175790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0824402493161741e-05,
+      "loss": 0.1588,
+      "step": 175800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0822160441235819e-05,
+      "loss": 0.1569,
+      "step": 175810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0819918389309897e-05,
+      "loss": 0.1543,
+      "step": 175820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0817676337383974e-05,
+      "loss": 0.1537,
+      "step": 175830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.0815434285458053e-05,
+      "loss": 0.1545,
+      "step": 175840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.081319223353213e-05,
+      "loss": 0.1507,
+      "step": 175850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0810950181606207e-05,
+      "loss": 0.1534,
+      "step": 175860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0808708129680284e-05,
+      "loss": 0.1521,
+      "step": 175870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0806466077754361e-05,
+      "loss": 0.1615,
+      "step": 175880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0804224025828438e-05,
+      "loss": 0.1552,
+      "step": 175890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0801981973902515e-05,
+      "loss": 0.1576,
+      "step": 175900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0799739921976594e-05,
+      "loss": 0.1525,
+      "step": 175910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0797497870050671e-05,
+      "loss": 0.15,
+      "step": 175920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0795255818124748e-05,
+      "loss": 0.1589,
+      "step": 175930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0793013766198825e-05,
+      "loss": 0.1511,
+      "step": 175940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0790771714272904e-05,
+      "loss": 0.1451,
+      "step": 175950
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.078852966234698e-05,
+      "loss": 0.1567,
+      "step": 175960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0786287610421058e-05,
+      "loss": 0.1548,
+      "step": 175970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0784045558495136e-05,
+      "loss": 0.1517,
+      "step": 175980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0781803506569213e-05,
+      "loss": 0.1536,
+      "step": 175990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.077956145464329e-05,
+      "loss": 0.1501,
+      "step": 176000
+    },
+    {
+      "epoch": 0.79,
+      "eval_NEIMS_canon_smiles": 0.9436333333333333,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.09533333333333334,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.49817997067988024,
+      "eval_NEIMS_exact_mols": 0.09498333333333334,
+      "eval_NEIMS_exact_smiles": 0.09238333333333333,
+      "eval_NEIMS_loss": 0.19087590277194977,
+      "eval_NEIMS_matched_formulas": 0.6240333333333333,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.40670017593076974,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.09548333333333334,
+      "eval_NEIMS_runtime": 712.1843,
+      "eval_NEIMS_samples_per_second": 84.248,
+      "eval_NEIMS_steps_per_second": 1.317,
+      "step": 176000
+    },
+    {
+      "epoch": 0.79,
+      "eval_RASSP_canon_smiles": 0.9381741075900758,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.2163321868634588,
+      "eval_RASSP_daylight_tanimoto_simil": 0.6437398086932788,
+      "eval_RASSP_exact_mols": 0.21576451731392127,
+      "eval_RASSP_exact_smiles": 0.2090526596988012,
+      "eval_RASSP_loss": 0.1295212209224701,
+      "eval_RASSP_matched_formulas": 0.8016328847630815,
+      "eval_RASSP_morgan_tanimoto_simil": 0.5252261223042175,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.21589808661969478,
+      "eval_RASSP_runtime": 814.9341,
+      "eval_RASSP_samples_per_second": 73.496,
+      "eval_RASSP_steps_per_second": 1.149,
+      "step": 176000
+    },
+    {
+      "epoch": 0.79,
+      "eval_NIST_canon_smiles": 0.8716328920750968,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.016786740958938142,
+      "eval_NIST_daylight_tanimoto_simil": 0.2798117873592377,
+      "eval_NIST_exact_mols": 0.016076942186890018,
+      "eval_NIST_exact_smiles": 0.014763814458600987,
+      "eval_NIST_loss": 1.874320149421692,
+      "eval_NIST_matched_formulas": 0.09791674060403875,
+      "eval_NIST_morgan_tanimoto_simil": 0.23257471064607163,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.01660929126592611,
+      "eval_NIST_runtime": 472.9919,
+      "eval_NIST_samples_per_second": 59.572,
+      "eval_NIST_steps_per_second": 0.932,
+      "step": 176000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0777319402717367e-05,
+      "loss": 0.1522,
+      "step": 176010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0775077350791444e-05,
+      "loss": 0.1591,
+      "step": 176020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0772835298865521e-05,
+      "loss": 0.1574,
+      "step": 176030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0770593246939599e-05,
+      "loss": 0.1516,
+      "step": 176040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0768351195013677e-05,
+      "loss": 0.1573,
+      "step": 176050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0766109143087754e-05,
+      "loss": 0.1584,
+      "step": 176060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0763867091161831e-05,
+      "loss": 0.1541,
+      "step": 176070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0761625039235908e-05,
+      "loss": 0.1561,
+      "step": 176080
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0759382987309987e-05,
+      "loss": 0.1553,
+      "step": 176090
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0757140935384064e-05,
+      "loss": 0.1535,
+      "step": 176100
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0754898883458141e-05,
+      "loss": 0.1575,
+      "step": 176110
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.075265683153222e-05,
+      "loss": 0.151,
+      "step": 176120
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0750414779606297e-05,
+      "loss": 0.1564,
+      "step": 176130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0748172727680374e-05,
+      "loss": 0.1537,
+      "step": 176140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0745930675754451e-05,
+      "loss": 0.1568,
+      "step": 176150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0743688623828528e-05,
+      "loss": 0.1555,
+      "step": 176160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0741446571902605e-05,
+      "loss": 0.1529,
+      "step": 176170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0739204519976684e-05,
+      "loss": 0.153,
+      "step": 176180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.073696246805076e-05,
+      "loss": 0.1569,
+      "step": 176190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0734720416124838e-05,
+      "loss": 0.1518,
+      "step": 176200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0732478364198915e-05,
+      "loss": 0.1544,
+      "step": 176210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0730236312272993e-05,
+      "loss": 0.1561,
+      "step": 176220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.072799426034707e-05,
+      "loss": 0.1606,
+      "step": 176230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0725752208421147e-05,
+      "loss": 0.1526,
+      "step": 176240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0723510156495226e-05,
+      "loss": 0.1534,
+      "step": 176250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0721268104569303e-05,
+      "loss": 0.1523,
+      "step": 176260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.071902605264338e-05,
+      "loss": 0.1562,
+      "step": 176270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0716784000717457e-05,
+      "loss": 0.1533,
+      "step": 176280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0714541948791534e-05,
+      "loss": 0.1625,
+      "step": 176290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0712299896865611e-05,
+      "loss": 0.1514,
+      "step": 176300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0710057844939688e-05,
+      "loss": 0.1511,
+      "step": 176310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0707815793013767e-05,
+      "loss": 0.1537,
+      "step": 176320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0705573741087844e-05,
+      "loss": 0.1525,
+      "step": 176330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0703331689161921e-05,
+      "loss": 0.1539,
+      "step": 176340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0701089637235998e-05,
+      "loss": 0.1597,
+      "step": 176350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0698847585310077e-05,
+      "loss": 0.151,
+      "step": 176360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0696605533384154e-05,
+      "loss": 0.1551,
+      "step": 176370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0694363481458231e-05,
+      "loss": 0.1544,
+      "step": 176380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.069212142953231e-05,
+      "loss": 0.1557,
+      "step": 176390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0689879377606387e-05,
+      "loss": 0.1536,
+      "step": 176400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0687637325680464e-05,
+      "loss": 0.1526,
+      "step": 176410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.068539527375454e-05,
+      "loss": 0.1531,
+      "step": 176420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0683153221828618e-05,
+      "loss": 0.1518,
+      "step": 176430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0680911169902695e-05,
+      "loss": 0.1535,
+      "step": 176440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0678669117976772e-05,
+      "loss": 0.1541,
+      "step": 176450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.067642706605085e-05,
+      "loss": 0.1606,
+      "step": 176460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0674185014124927e-05,
+      "loss": 0.1598,
+      "step": 176470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0671942962199005e-05,
+      "loss": 0.154,
+      "step": 176480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0669700910273083e-05,
+      "loss": 0.1592,
+      "step": 176490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.066745885834716e-05,
+      "loss": 0.1542,
+      "step": 176500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0665216806421237e-05,
+      "loss": 0.1589,
+      "step": 176510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0662974754495314e-05,
+      "loss": 0.1602,
+      "step": 176520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0660732702569393e-05,
+      "loss": 0.1484,
+      "step": 176530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.065849065064347e-05,
+      "loss": 0.1574,
+      "step": 176540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0656248598717547e-05,
+      "loss": 0.15,
+      "step": 176550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0654006546791624e-05,
+      "loss": 0.1577,
+      "step": 176560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0651764494865701e-05,
+      "loss": 0.1595,
+      "step": 176570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0649522442939778e-05,
+      "loss": 0.1525,
+      "step": 176580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0647280391013855e-05,
+      "loss": 0.1503,
+      "step": 176590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0645038339087934e-05,
+      "loss": 0.1486,
+      "step": 176600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0642796287162011e-05,
+      "loss": 0.1514,
+      "step": 176610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0640554235236088e-05,
+      "loss": 0.156,
+      "step": 176620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0638312183310167e-05,
+      "loss": 0.1574,
+      "step": 176630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0636070131384244e-05,
+      "loss": 0.1542,
+      "step": 176640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.063382807945832e-05,
+      "loss": 0.1578,
+      "step": 176650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0631586027532398e-05,
+      "loss": 0.1455,
+      "step": 176660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0629343975606476e-05,
+      "loss": 0.1526,
+      "step": 176670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0627101923680553e-05,
+      "loss": 0.1543,
+      "step": 176680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.062485987175463e-05,
+      "loss": 0.152,
+      "step": 176690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0622617819828707e-05,
+      "loss": 0.1632,
+      "step": 176700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0620375767902785e-05,
+      "loss": 0.1519,
+      "step": 176710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0618133715976862e-05,
+      "loss": 0.1537,
+      "step": 176720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0615891664050939e-05,
+      "loss": 0.1523,
+      "step": 176730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0613649612125017e-05,
+      "loss": 0.153,
+      "step": 176740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0611407560199094e-05,
+      "loss": 0.1497,
+      "step": 176750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0609165508273173e-05,
+      "loss": 0.1576,
+      "step": 176760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.060692345634725e-05,
+      "loss": 0.1581,
+      "step": 176770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0604681404421327e-05,
+      "loss": 0.1537,
+      "step": 176780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0602439352495404e-05,
+      "loss": 0.1515,
+      "step": 176790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0600197300569481e-05,
+      "loss": 0.153,
+      "step": 176800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.059795524864356e-05,
+      "loss": 0.1583,
+      "step": 176810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0595713196717637e-05,
+      "loss": 0.1501,
+      "step": 176820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0593471144791714e-05,
+      "loss": 0.1514,
+      "step": 176830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0591229092865791e-05,
+      "loss": 0.1567,
+      "step": 176840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0588987040939868e-05,
+      "loss": 0.1553,
+      "step": 176850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0586744989013945e-05,
+      "loss": 0.1553,
+      "step": 176860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0584502937088024e-05,
+      "loss": 0.1546,
+      "step": 176870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.05822608851621e-05,
+      "loss": 0.1517,
+      "step": 176880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0580018833236178e-05,
+      "loss": 0.1546,
+      "step": 176890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0577776781310256e-05,
+      "loss": 0.1491,
+      "step": 176900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0575534729384333e-05,
+      "loss": 0.1485,
+      "step": 176910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.057329267745841e-05,
+      "loss": 0.1518,
+      "step": 176920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0571050625532488e-05,
+      "loss": 0.152,
+      "step": 176930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0568808573606566e-05,
+      "loss": 0.1537,
+      "step": 176940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0566566521680643e-05,
+      "loss": 0.1473,
+      "step": 176950
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.056432446975472e-05,
+      "loss": 0.1552,
+      "step": 176960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0562082417828797e-05,
+      "loss": 0.1531,
+      "step": 176970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0559840365902874e-05,
+      "loss": 0.1532,
+      "step": 176980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0557598313976951e-05,
+      "loss": 0.1494,
+      "step": 176990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0555356262051028e-05,
+      "loss": 0.1564,
+      "step": 177000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0553114210125107e-05,
+      "loss": 0.1516,
+      "step": 177010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0550872158199184e-05,
+      "loss": 0.1563,
+      "step": 177020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0548630106273263e-05,
+      "loss": 0.1518,
+      "step": 177030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.054638805434734e-05,
+      "loss": 0.1568,
+      "step": 177040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0544146002421417e-05,
+      "loss": 0.1525,
+      "step": 177050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0541903950495494e-05,
+      "loss": 0.1472,
+      "step": 177060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0539661898569571e-05,
+      "loss": 0.1558,
+      "step": 177070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.053741984664365e-05,
+      "loss": 0.152,
+      "step": 177080
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0535177794717727e-05,
+      "loss": 0.1518,
+      "step": 177090
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0532935742791804e-05,
+      "loss": 0.1534,
+      "step": 177100
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.053069369086588e-05,
+      "loss": 0.155,
+      "step": 177110
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0528451638939958e-05,
+      "loss": 0.1591,
+      "step": 177120
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0526209587014035e-05,
+      "loss": 0.1589,
+      "step": 177130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0523967535088112e-05,
+      "loss": 0.1585,
+      "step": 177140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.052172548316219e-05,
+      "loss": 0.1495,
+      "step": 177150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0519483431236268e-05,
+      "loss": 0.1503,
+      "step": 177160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0517241379310346e-05,
+      "loss": 0.1512,
+      "step": 177170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0514999327384423e-05,
+      "loss": 0.1519,
+      "step": 177180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.05127572754585e-05,
+      "loss": 0.1494,
+      "step": 177190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0510515223532577e-05,
+      "loss": 0.1513,
+      "step": 177200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0508273171606654e-05,
+      "loss": 0.1551,
+      "step": 177210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0506031119680733e-05,
+      "loss": 0.1524,
+      "step": 177220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.050378906775481e-05,
+      "loss": 0.157,
+      "step": 177230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0501547015828887e-05,
+      "loss": 0.1486,
+      "step": 177240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0499304963902964e-05,
+      "loss": 0.1589,
+      "step": 177250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0497062911977041e-05,
+      "loss": 0.1523,
+      "step": 177260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0494820860051118e-05,
+      "loss": 0.152,
+      "step": 177270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0492578808125197e-05,
+      "loss": 0.1523,
+      "step": 177280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0490336756199274e-05,
+      "loss": 0.1573,
+      "step": 177290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0488094704273353e-05,
+      "loss": 0.1548,
+      "step": 177300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.048585265234743e-05,
+      "loss": 0.1562,
+      "step": 177310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0483610600421507e-05,
+      "loss": 0.1498,
+      "step": 177320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0481368548495584e-05,
+      "loss": 0.1545,
+      "step": 177330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.047912649656966e-05,
+      "loss": 0.1536,
+      "step": 177340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0476884444643738e-05,
+      "loss": 0.153,
+      "step": 177350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0474642392717816e-05,
+      "loss": 0.1508,
+      "step": 177360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0472400340791893e-05,
+      "loss": 0.1552,
+      "step": 177370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.047015828886597e-05,
+      "loss": 0.1512,
+      "step": 177380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0467916236940048e-05,
+      "loss": 0.1526,
+      "step": 177390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0465674185014125e-05,
+      "loss": 0.1533,
+      "step": 177400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0463432133088202e-05,
+      "loss": 0.158,
+      "step": 177410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0461414286354873e-05,
+      "loss": 0.1537,
+      "step": 177420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.045917223442895e-05,
+      "loss": 0.1537,
+      "step": 177430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0456930182503027e-05,
+      "loss": 0.1488,
+      "step": 177440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0454688130577104e-05,
+      "loss": 0.1461,
+      "step": 177450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0452446078651183e-05,
+      "loss": 0.152,
+      "step": 177460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.045020402672526e-05,
+      "loss": 0.1524,
+      "step": 177470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0447961974799337e-05,
+      "loss": 0.1509,
+      "step": 177480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0445719922873415e-05,
+      "loss": 0.1453,
+      "step": 177490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0443477870947492e-05,
+      "loss": 0.1531,
+      "step": 177500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.044123581902157e-05,
+      "loss": 0.1499,
+      "step": 177510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0438993767095646e-05,
+      "loss": 0.1459,
+      "step": 177520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0436751715169723e-05,
+      "loss": 0.1503,
+      "step": 177530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.04345096632438e-05,
+      "loss": 0.1588,
+      "step": 177540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0432267611317877e-05,
+      "loss": 0.1492,
+      "step": 177550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0430025559391956e-05,
+      "loss": 0.1443,
+      "step": 177560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0427783507466033e-05,
+      "loss": 0.1479,
+      "step": 177570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.042554145554011e-05,
+      "loss": 0.1482,
+      "step": 177580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0423299403614189e-05,
+      "loss": 0.1535,
+      "step": 177590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0421057351688266e-05,
+      "loss": 0.1531,
+      "step": 177600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0418815299762343e-05,
+      "loss": 0.1515,
+      "step": 177610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.041657324783642e-05,
+      "loss": 0.1524,
+      "step": 177620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0414331195910499e-05,
+      "loss": 0.1412,
+      "step": 177630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0412089143984576e-05,
+      "loss": 0.1456,
+      "step": 177640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0409847092058653e-05,
+      "loss": 0.1503,
+      "step": 177650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.040760504013273e-05,
+      "loss": 0.1507,
+      "step": 177660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0405362988206807e-05,
+      "loss": 0.154,
+      "step": 177670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0403120936280884e-05,
+      "loss": 0.1562,
+      "step": 177680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0400878884354961e-05,
+      "loss": 0.1493,
+      "step": 177690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.039863683242904e-05,
+      "loss": 0.1492,
+      "step": 177700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0396394780503117e-05,
+      "loss": 0.1496,
+      "step": 177710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0394152728577194e-05,
+      "loss": 0.1506,
+      "step": 177720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0391910676651272e-05,
+      "loss": 0.1499,
+      "step": 177730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.038966862472535e-05,
+      "loss": 0.1496,
+      "step": 177740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0387426572799426e-05,
+      "loss": 0.1491,
+      "step": 177750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0385184520873503e-05,
+      "loss": 0.1523,
+      "step": 177760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0382942468947582e-05,
+      "loss": 0.1539,
+      "step": 177770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.038070041702166e-05,
+      "loss": 0.1528,
+      "step": 177780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0378458365095736e-05,
+      "loss": 0.1456,
+      "step": 177790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0376216313169813e-05,
+      "loss": 0.152,
+      "step": 177800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.037397426124389e-05,
+      "loss": 0.1524,
+      "step": 177810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0371732209317967e-05,
+      "loss": 0.1531,
+      "step": 177820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0369490157392046e-05,
+      "loss": 0.1476,
+      "step": 177830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0367248105466123e-05,
+      "loss": 0.1539,
+      "step": 177840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.03650060535402e-05,
+      "loss": 0.1543,
+      "step": 177850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0362764001614279e-05,
+      "loss": 0.1535,
+      "step": 177860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0360521949688356e-05,
+      "loss": 0.1511,
+      "step": 177870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0358279897762433e-05,
+      "loss": 0.1523,
+      "step": 177880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.035603784583651e-05,
+      "loss": 0.1516,
+      "step": 177890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0353795793910589e-05,
+      "loss": 0.1517,
+      "step": 177900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0351553741984666e-05,
+      "loss": 0.147,
+      "step": 177910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0349311690058743e-05,
+      "loss": 0.1559,
+      "step": 177920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.034706963813282e-05,
+      "loss": 0.1478,
+      "step": 177930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0344827586206897e-05,
+      "loss": 0.1524,
+      "step": 177940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0342585534280974e-05,
+      "loss": 0.1557,
+      "step": 177950
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.034034348235505e-05,
+      "loss": 0.1477,
+      "step": 177960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.033810143042913e-05,
+      "loss": 0.1523,
+      "step": 177970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0335859378503206e-05,
+      "loss": 0.1534,
+      "step": 177980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0333617326577283e-05,
+      "loss": 0.1527,
+      "step": 177990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0331375274651362e-05,
+      "loss": 0.1464,
+      "step": 178000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.032913322272544e-05,
+      "loss": 0.1527,
+      "step": 178010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0326891170799516e-05,
+      "loss": 0.1536,
+      "step": 178020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0324649118873593e-05,
+      "loss": 0.1547,
+      "step": 178030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0322407066947672e-05,
+      "loss": 0.1481,
+      "step": 178040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0320165015021749e-05,
+      "loss": 0.1536,
+      "step": 178050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0317922963095826e-05,
+      "loss": 0.1504,
+      "step": 178060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0315680911169903e-05,
+      "loss": 0.1539,
+      "step": 178070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.031343885924398e-05,
+      "loss": 0.15,
+      "step": 178080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0311196807318057e-05,
+      "loss": 0.1448,
+      "step": 178090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0308954755392134e-05,
+      "loss": 0.1512,
+      "step": 178100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0306712703466213e-05,
+      "loss": 0.145,
+      "step": 178110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.030447065154029e-05,
+      "loss": 0.1578,
+      "step": 178120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0302228599614369e-05,
+      "loss": 0.1501,
+      "step": 178130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0299986547688446e-05,
+      "loss": 0.1443,
+      "step": 178140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0297744495762523e-05,
+      "loss": 0.1499,
+      "step": 178150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.02955024438366e-05,
+      "loss": 0.1487,
+      "step": 178160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0293260391910677e-05,
+      "loss": 0.1509,
+      "step": 178170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0291018339984755e-05,
+      "loss": 0.1497,
+      "step": 178180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0288776288058832e-05,
+      "loss": 0.1467,
+      "step": 178190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.028653423613291e-05,
+      "loss": 0.15,
+      "step": 178200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0284292184206986e-05,
+      "loss": 0.1475,
+      "step": 178210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0282050132281063e-05,
+      "loss": 0.1474,
+      "step": 178220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.027980808035514e-05,
+      "loss": 0.1485,
+      "step": 178230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0277566028429218e-05,
+      "loss": 0.1464,
+      "step": 178240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0275323976503296e-05,
+      "loss": 0.1417,
+      "step": 178250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0273081924577373e-05,
+      "loss": 0.151,
+      "step": 178260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0270839872651452e-05,
+      "loss": 0.1577,
+      "step": 178270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0268597820725529e-05,
+      "loss": 0.1471,
+      "step": 178280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0266355768799606e-05,
+      "loss": 0.1472,
+      "step": 178290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0264113716873683e-05,
+      "loss": 0.1506,
+      "step": 178300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.026187166494776e-05,
+      "loss": 0.1538,
+      "step": 178310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0259629613021839e-05,
+      "loss": 0.1513,
+      "step": 178320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0257387561095916e-05,
+      "loss": 0.1514,
+      "step": 178330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0255145509169993e-05,
+      "loss": 0.1481,
+      "step": 178340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.025290345724407e-05,
+      "loss": 0.1496,
+      "step": 178350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0250661405318147e-05,
+      "loss": 0.1454,
+      "step": 178360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0248419353392224e-05,
+      "loss": 0.148,
+      "step": 178370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0246177301466303e-05,
+      "loss": 0.1472,
+      "step": 178380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.024393524954038e-05,
+      "loss": 0.1499,
+      "step": 178390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0241693197614458e-05,
+      "loss": 0.15,
+      "step": 178400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0239451145688535e-05,
+      "loss": 0.1561,
+      "step": 178410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0237209093762612e-05,
+      "loss": 0.1435,
+      "step": 178420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.023496704183669e-05,
+      "loss": 0.1504,
+      "step": 178430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0232724989910766e-05,
+      "loss": 0.1493,
+      "step": 178440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0230482937984843e-05,
+      "loss": 0.1489,
+      "step": 178450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0228240886058922e-05,
+      "loss": 0.1469,
+      "step": 178460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0225998834133e-05,
+      "loss": 0.1514,
+      "step": 178470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0223756782207076e-05,
+      "loss": 0.1435,
+      "step": 178480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0221514730281153e-05,
+      "loss": 0.1516,
+      "step": 178490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.021927267835523e-05,
+      "loss": 0.1513,
+      "step": 178500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0217030626429307e-05,
+      "loss": 0.1468,
+      "step": 178510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0214788574503386e-05,
+      "loss": 0.1488,
+      "step": 178520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0212546522577463e-05,
+      "loss": 0.1481,
+      "step": 178530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0210304470651542e-05,
+      "loss": 0.1489,
+      "step": 178540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0208062418725619e-05,
+      "loss": 0.1534,
+      "step": 178550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0205820366799696e-05,
+      "loss": 0.149,
+      "step": 178560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0203578314873773e-05,
+      "loss": 0.1501,
+      "step": 178570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.020133626294785e-05,
+      "loss": 0.1526,
+      "step": 178580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0199094211021929e-05,
+      "loss": 0.1513,
+      "step": 178590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0196852159096006e-05,
+      "loss": 0.1484,
+      "step": 178600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0194610107170083e-05,
+      "loss": 0.1504,
+      "step": 178610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.019236805524416e-05,
+      "loss": 0.1497,
+      "step": 178620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0190126003318237e-05,
+      "loss": 0.1555,
+      "step": 178630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0187883951392314e-05,
+      "loss": 0.148,
+      "step": 178640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0185641899466392e-05,
+      "loss": 0.1512,
+      "step": 178650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.018339984754047e-05,
+      "loss": 0.144,
+      "step": 178660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0181157795614548e-05,
+      "loss": 0.1525,
+      "step": 178670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0178915743688625e-05,
+      "loss": 0.1508,
+      "step": 178680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0176673691762702e-05,
+      "loss": 0.1522,
+      "step": 178690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.017443163983678e-05,
+      "loss": 0.1501,
+      "step": 178700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0172189587910856e-05,
+      "loss": 0.1478,
+      "step": 178710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0169947535984933e-05,
+      "loss": 0.1481,
+      "step": 178720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0167705484059012e-05,
+      "loss": 0.1481,
+      "step": 178730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0165463432133089e-05,
+      "loss": 0.1463,
+      "step": 178740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0163221380207166e-05,
+      "loss": 0.1432,
+      "step": 178750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0160979328281243e-05,
+      "loss": 0.1464,
+      "step": 178760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.015873727635532e-05,
+      "loss": 0.1533,
+      "step": 178770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0156495224429397e-05,
+      "loss": 0.1486,
+      "step": 178780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0154253172503476e-05,
+      "loss": 0.1468,
+      "step": 178790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0152011120577553e-05,
+      "loss": 0.1445,
+      "step": 178800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0149769068651632e-05,
+      "loss": 0.1529,
+      "step": 178810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0147527016725709e-05,
+      "loss": 0.1554,
+      "step": 178820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0145284964799786e-05,
+      "loss": 0.1462,
+      "step": 178830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0143042912873863e-05,
+      "loss": 0.149,
+      "step": 178840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.014080086094794e-05,
+      "loss": 0.1518,
+      "step": 178850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0138558809022017e-05,
+      "loss": 0.1452,
+      "step": 178860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0136316757096095e-05,
+      "loss": 0.1465,
+      "step": 178870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0134074705170172e-05,
+      "loss": 0.1479,
+      "step": 178880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.013183265324425e-05,
+      "loss": 0.1422,
+      "step": 178890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0129590601318326e-05,
+      "loss": 0.1428,
+      "step": 178900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0127348549392404e-05,
+      "loss": 0.1508,
+      "step": 178910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.012510649746648e-05,
+      "loss": 0.1418,
+      "step": 178920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.012286444554056e-05,
+      "loss": 0.147,
+      "step": 178930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0120622393614638e-05,
+      "loss": 0.148,
+      "step": 178940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0118380341688715e-05,
+      "loss": 0.1474,
+      "step": 178950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0116138289762792e-05,
+      "loss": 0.1513,
+      "step": 178960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0113896237836869e-05,
+      "loss": 0.1491,
+      "step": 178970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0111654185910946e-05,
+      "loss": 0.1527,
+      "step": 178980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0109412133985023e-05,
+      "loss": 0.152,
+      "step": 178990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.01071700820591e-05,
+      "loss": 0.1484,
+      "step": 179000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0104928030133179e-05,
+      "loss": 0.1478,
+      "step": 179010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0102685978207256e-05,
+      "loss": 0.1475,
+      "step": 179020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0100443926281333e-05,
+      "loss": 0.1468,
+      "step": 179030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.009820187435541e-05,
+      "loss": 0.1515,
+      "step": 179040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0095959822429487e-05,
+      "loss": 0.1511,
+      "step": 179050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0093717770503566e-05,
+      "loss": 0.1504,
+      "step": 179060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0091475718577643e-05,
+      "loss": 0.1487,
+      "step": 179070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0089233666651721e-05,
+      "loss": 0.1443,
+      "step": 179080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0086991614725798e-05,
+      "loss": 0.1499,
+      "step": 179090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0084749562799875e-05,
+      "loss": 0.1516,
+      "step": 179100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0082507510873952e-05,
+      "loss": 0.1462,
+      "step": 179110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.008026545894803e-05,
+      "loss": 0.1476,
+      "step": 179120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0078023407022107e-05,
+      "loss": 0.1577,
+      "step": 179130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0075781355096184e-05,
+      "loss": 0.1508,
+      "step": 179140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0073539303170262e-05,
+      "loss": 0.1489,
+      "step": 179150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.007129725124434e-05,
+      "loss": 0.1435,
+      "step": 179160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0069055199318416e-05,
+      "loss": 0.1479,
+      "step": 179170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0066813147392493e-05,
+      "loss": 0.1559,
+      "step": 179180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.006457109546657e-05,
+      "loss": 0.1488,
+      "step": 179190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0062329043540649e-05,
+      "loss": 0.1557,
+      "step": 179200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0060086991614728e-05,
+      "loss": 0.1468,
+      "step": 179210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0057844939688805e-05,
+      "loss": 0.1511,
+      "step": 179220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0055602887762882e-05,
+      "loss": 0.1469,
+      "step": 179230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0053360835836959e-05,
+      "loss": 0.1497,
+      "step": 179240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0051118783911036e-05,
+      "loss": 0.1493,
+      "step": 179250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0048876731985113e-05,
+      "loss": 0.1466,
+      "step": 179260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.004663468005919e-05,
+      "loss": 0.1488,
+      "step": 179270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0044392628133269e-05,
+      "loss": 0.1471,
+      "step": 179280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0042150576207346e-05,
+      "loss": 0.1528,
+      "step": 179290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0039908524281423e-05,
+      "loss": 0.149,
+      "step": 179300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.00376664723555e-05,
+      "loss": 0.1513,
+      "step": 179310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0035424420429577e-05,
+      "loss": 0.1514,
+      "step": 179320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0033182368503655e-05,
+      "loss": 0.1451,
+      "step": 179330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0030940316577732e-05,
+      "loss": 0.1413,
+      "step": 179340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0028698264651811e-05,
+      "loss": 0.1489,
+      "step": 179350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0026456212725888e-05,
+      "loss": 0.1526,
+      "step": 179360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0024214160799965e-05,
+      "loss": 0.1493,
+      "step": 179370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0021972108874042e-05,
+      "loss": 0.1497,
+      "step": 179380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.001973005694812e-05,
+      "loss": 0.1442,
+      "step": 179390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0017488005022196e-05,
+      "loss": 0.1511,
+      "step": 179400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0015245953096273e-05,
+      "loss": 0.1485,
+      "step": 179410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0013003901170352e-05,
+      "loss": 0.1517,
+      "step": 179420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0010761849244429e-05,
+      "loss": 0.1501,
+      "step": 179430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0008519797318506e-05,
+      "loss": 0.1452,
+      "step": 179440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0006277745392583e-05,
+      "loss": 0.1431,
+      "step": 179450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.000403569346666e-05,
+      "loss": 0.1531,
+      "step": 179460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0001793641540739e-05,
+      "loss": 0.148,
+      "step": 179470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.999551589614816e-06,
+      "loss": 0.1483,
+      "step": 179480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.997309537688895e-06,
+      "loss": 0.1421,
+      "step": 179490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.995067485762972e-06,
+      "loss": 0.1521,
+      "step": 179500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.992825433837049e-06,
+      "loss": 0.1457,
+      "step": 179510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.990583381911126e-06,
+      "loss": 0.1461,
+      "step": 179520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.988341329985203e-06,
+      "loss": 0.1436,
+      "step": 179530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.98609927805928e-06,
+      "loss": 0.1439,
+      "step": 179540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.983857226133357e-06,
+      "loss": 0.1461,
+      "step": 179550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.981615174207435e-06,
+      "loss": 0.1479,
+      "step": 179560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.979373122281512e-06,
+      "loss": 0.1462,
+      "step": 179570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.97713107035559e-06,
+      "loss": 0.1508,
+      "step": 179580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.974889018429667e-06,
+      "loss": 0.1505,
+      "step": 179590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.972646966503745e-06,
+      "loss": 0.1509,
+      "step": 179600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.970404914577822e-06,
+      "loss": 0.1501,
+      "step": 179610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.9681628626519e-06,
+      "loss": 0.1529,
+      "step": 179620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.965920810725978e-06,
+      "loss": 0.145,
+      "step": 179630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.963678758800055e-06,
+      "loss": 0.1516,
+      "step": 179640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.961436706874132e-06,
+      "loss": 0.1438,
+      "step": 179650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.959194654948209e-06,
+      "loss": 0.1497,
+      "step": 179660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.956952603022286e-06,
+      "loss": 0.1427,
+      "step": 179670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.954710551096363e-06,
+      "loss": 0.1456,
+      "step": 179680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.95246849917044e-06,
+      "loss": 0.1517,
+      "step": 179690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.950226447244519e-06,
+      "loss": 0.149,
+      "step": 179700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.947984395318596e-06,
+      "loss": 0.1442,
+      "step": 179710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.945742343392673e-06,
+      "loss": 0.1498,
+      "step": 179720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.94350029146675e-06,
+      "loss": 0.1461,
+      "step": 179730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.941258239540829e-06,
+      "loss": 0.1528,
+      "step": 179740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.939016187614906e-06,
+      "loss": 0.1492,
+      "step": 179750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.936774135688983e-06,
+      "loss": 0.1545,
+      "step": 179760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.934532083763061e-06,
+      "loss": 0.1459,
+      "step": 179770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.932290031837138e-06,
+      "loss": 0.1498,
+      "step": 179780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.930047979911215e-06,
+      "loss": 0.1521,
+      "step": 179790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.927805927985293e-06,
+      "loss": 0.1498,
+      "step": 179800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.92556387605937e-06,
+      "loss": 0.1495,
+      "step": 179810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.923321824133447e-06,
+      "loss": 0.1527,
+      "step": 179820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.921079772207525e-06,
+      "loss": 0.1505,
+      "step": 179830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.918837720281602e-06,
+      "loss": 0.145,
+      "step": 179840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.91659566835568e-06,
+      "loss": 0.1506,
+      "step": 179850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.914353616429756e-06,
+      "loss": 0.1526,
+      "step": 179860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.912111564503835e-06,
+      "loss": 0.1452,
+      "step": 179870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.909869512577912e-06,
+      "loss": 0.1475,
+      "step": 179880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.907627460651989e-06,
+      "loss": 0.155,
+      "step": 179890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.905385408726068e-06,
+      "loss": 0.1447,
+      "step": 179900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.903143356800145e-06,
+      "loss": 0.1429,
+      "step": 179910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.900901304874222e-06,
+      "loss": 0.143,
+      "step": 179920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.898659252948299e-06,
+      "loss": 0.1491,
+      "step": 179930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.896417201022376e-06,
+      "loss": 0.1437,
+      "step": 179940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.894175149096453e-06,
+      "loss": 0.153,
+      "step": 179950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.89193309717053e-06,
+      "loss": 0.1467,
+      "step": 179960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.889691045244609e-06,
+      "loss": 0.1511,
+      "step": 179970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.887448993318686e-06,
+      "loss": 0.1507,
+      "step": 179980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.885206941392763e-06,
+      "loss": 0.1516,
+      "step": 179990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.88296488946684e-06,
+      "loss": 0.1465,
+      "step": 180000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.880722837540918e-06,
+      "loss": 0.1479,
+      "step": 180010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.878480785614995e-06,
+      "loss": 0.1466,
+      "step": 180020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.876238733689073e-06,
+      "loss": 0.1473,
+      "step": 180030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.873996681763151e-06,
+      "loss": 0.1451,
+      "step": 180040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.871754629837228e-06,
+      "loss": 0.1455,
+      "step": 180050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.869512577911305e-06,
+      "loss": 0.1451,
+      "step": 180060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.867270525985382e-06,
+      "loss": 0.1508,
+      "step": 180070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.86502847405946e-06,
+      "loss": 0.1477,
+      "step": 180080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.862786422133536e-06,
+      "loss": 0.1432,
+      "step": 180090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.860544370207613e-06,
+      "loss": 0.1528,
+      "step": 180100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.858302318281692e-06,
+      "loss": 0.1491,
+      "step": 180110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.856060266355769e-06,
+      "loss": 0.1455,
+      "step": 180120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.853818214429846e-06,
+      "loss": 0.1509,
+      "step": 180130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.851576162503925e-06,
+      "loss": 0.1445,
+      "step": 180140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.849334110578002e-06,
+      "loss": 0.1489,
+      "step": 180150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.847092058652079e-06,
+      "loss": 0.1477,
+      "step": 180160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.844850006726156e-06,
+      "loss": 0.1496,
+      "step": 180170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.842607954800235e-06,
+      "loss": 0.1513,
+      "step": 180180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.840365902874312e-06,
+      "loss": 0.1474,
+      "step": 180190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.838123850948389e-06,
+      "loss": 0.1485,
+      "step": 180200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.835881799022466e-06,
+      "loss": 0.1477,
+      "step": 180210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.833639747096543e-06,
+      "loss": 0.1545,
+      "step": 180220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.83139769517062e-06,
+      "loss": 0.1473,
+      "step": 180230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.829155643244697e-06,
+      "loss": 0.1493,
+      "step": 180240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.826913591318776e-06,
+      "loss": 0.1441,
+      "step": 180250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.824671539392853e-06,
+      "loss": 0.1414,
+      "step": 180260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.82242948746693e-06,
+      "loss": 0.142,
+      "step": 180270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.820187435541008e-06,
+      "loss": 0.1488,
+      "step": 180280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.817945383615085e-06,
+      "loss": 0.1488,
+      "step": 180290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.815703331689162e-06,
+      "loss": 0.1496,
+      "step": 180300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.81346127976324e-06,
+      "loss": 0.1496,
+      "step": 180310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.811219227837318e-06,
+      "loss": 0.1474,
+      "step": 180320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.808977175911395e-06,
+      "loss": 0.1482,
+      "step": 180330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.806735123985472e-06,
+      "loss": 0.1442,
+      "step": 180340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.804493072059549e-06,
+      "loss": 0.1453,
+      "step": 180350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.802251020133626e-06,
+      "loss": 0.1492,
+      "step": 180360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.800008968207703e-06,
+      "loss": 0.1478,
+      "step": 180370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.79776691628178e-06,
+      "loss": 0.1407,
+      "step": 180380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.795524864355859e-06,
+      "loss": 0.1493,
+      "step": 180390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.793282812429936e-06,
+      "loss": 0.1505,
+      "step": 180400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.791040760504015e-06,
+      "loss": 0.1469,
+      "step": 180410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.788798708578092e-06,
+      "loss": 0.1461,
+      "step": 180420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.786556656652169e-06,
+      "loss": 0.1496,
+      "step": 180430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.784314604726246e-06,
+      "loss": 0.1461,
+      "step": 180440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.782072552800323e-06,
+      "loss": 0.1462,
+      "step": 180450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.779830500874401e-06,
+      "loss": 0.1474,
+      "step": 180460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.777588448948478e-06,
+      "loss": 0.1524,
+      "step": 180470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.775346397022556e-06,
+      "loss": 0.1533,
+      "step": 180480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.773104345096633e-06,
+      "loss": 0.1501,
+      "step": 180490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.77086229317071e-06,
+      "loss": 0.1507,
+      "step": 180500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.768620241244787e-06,
+      "loss": 0.1461,
+      "step": 180510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.766378189318865e-06,
+      "loss": 0.1455,
+      "step": 180520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.764136137392942e-06,
+      "loss": 0.1471,
+      "step": 180530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.76189408546702e-06,
+      "loss": 0.1484,
+      "step": 180540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.759652033541098e-06,
+      "loss": 0.1492,
+      "step": 180550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.757409981615175e-06,
+      "loss": 0.1524,
+      "step": 180560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.755167929689252e-06,
+      "loss": 0.1472,
+      "step": 180570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.752925877763329e-06,
+      "loss": 0.1506,
+      "step": 180580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.750683825837408e-06,
+      "loss": 0.1479,
+      "step": 180590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.748441773911485e-06,
+      "loss": 0.1437,
+      "step": 180600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.746199721985562e-06,
+      "loss": 0.1509,
+      "step": 180610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.743957670059639e-06,
+      "loss": 0.1479,
+      "step": 180620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.741715618133716e-06,
+      "loss": 0.1414,
+      "step": 180630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.739473566207793e-06,
+      "loss": 0.1467,
+      "step": 180640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.73723151428187e-06,
+      "loss": 0.1506,
+      "step": 180650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.734989462355949e-06,
+      "loss": 0.1431,
+      "step": 180660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.732747410430026e-06,
+      "loss": 0.149,
+      "step": 180670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.730505358504104e-06,
+      "loss": 0.1522,
+      "step": 180680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.728263306578181e-06,
+      "loss": 0.1512,
+      "step": 180690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.726021254652259e-06,
+      "loss": 0.1454,
+      "step": 180700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.723779202726336e-06,
+      "loss": 0.1443,
+      "step": 180710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.721537150800413e-06,
+      "loss": 0.148,
+      "step": 180720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.719295098874491e-06,
+      "loss": 0.1469,
+      "step": 180730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.717053046948568e-06,
+      "loss": 0.1474,
+      "step": 180740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.714810995022645e-06,
+      "loss": 0.1452,
+      "step": 180750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.712568943096722e-06,
+      "loss": 0.1478,
+      "step": 180760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.7103268911708e-06,
+      "loss": 0.1467,
+      "step": 180770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.708084839244876e-06,
+      "loss": 0.1483,
+      "step": 180780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.705842787318953e-06,
+      "loss": 0.1485,
+      "step": 180790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.703600735393032e-06,
+      "loss": 0.1475,
+      "step": 180800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.70135868346711e-06,
+      "loss": 0.1444,
+      "step": 180810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.699116631541188e-06,
+      "loss": 0.147,
+      "step": 180820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.696874579615265e-06,
+      "loss": 0.1432,
+      "step": 180830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.694632527689342e-06,
+      "loss": 0.144,
+      "step": 180840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.692390475763419e-06,
+      "loss": 0.1507,
+      "step": 180850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.690148423837496e-06,
+      "loss": 0.1459,
+      "step": 180860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.687906371911575e-06,
+      "loss": 0.143,
+      "step": 180870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.685664319985652e-06,
+      "loss": 0.1417,
+      "step": 180880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.683422268059729e-06,
+      "loss": 0.1496,
+      "step": 180890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.681180216133806e-06,
+      "loss": 0.1458,
+      "step": 180900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.678938164207883e-06,
+      "loss": 0.1442,
+      "step": 180910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.67669611228196e-06,
+      "loss": 0.1423,
+      "step": 180920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.674454060356039e-06,
+      "loss": 0.1391,
+      "step": 180930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.672212008430116e-06,
+      "loss": 0.1496,
+      "step": 180940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.669969956504194e-06,
+      "loss": 0.1398,
+      "step": 180950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.667727904578271e-06,
+      "loss": 0.1482,
+      "step": 180960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.665485852652348e-06,
+      "loss": 0.1502,
+      "step": 180970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.663243800726425e-06,
+      "loss": 0.147,
+      "step": 180980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.661001748800502e-06,
+      "loss": 0.1492,
+      "step": 180990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.65875969687458e-06,
+      "loss": 0.1505,
+      "step": 181000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.656517644948658e-06,
+      "loss": 0.1461,
+      "step": 181010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.654275593022735e-06,
+      "loss": 0.1518,
+      "step": 181020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.652033541096812e-06,
+      "loss": 0.1442,
+      "step": 181030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.64979148917089e-06,
+      "loss": 0.1527,
+      "step": 181040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.647549437244966e-06,
+      "loss": 0.1428,
+      "step": 181050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.645307385319043e-06,
+      "loss": 0.1416,
+      "step": 181060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.643065333393122e-06,
+      "loss": 0.1402,
+      "step": 181070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.640823281467199e-06,
+      "loss": 0.1472,
+      "step": 181080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.638581229541278e-06,
+      "loss": 0.1496,
+      "step": 181090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.636339177615355e-06,
+      "loss": 0.1412,
+      "step": 181100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.634097125689432e-06,
+      "loss": 0.1506,
+      "step": 181110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.631855073763509e-06,
+      "loss": 0.1496,
+      "step": 181120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.629613021837586e-06,
+      "loss": 0.1474,
+      "step": 181130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.627370969911663e-06,
+      "loss": 0.1532,
+      "step": 181140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.625128917985742e-06,
+      "loss": 0.1428,
+      "step": 181150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.622886866059819e-06,
+      "loss": 0.149,
+      "step": 181160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.620644814133896e-06,
+      "loss": 0.1456,
+      "step": 181170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.618402762207973e-06,
+      "loss": 0.1466,
+      "step": 181180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.61616071028205e-06,
+      "loss": 0.1481,
+      "step": 181190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.613918658356127e-06,
+      "loss": 0.1461,
+      "step": 181200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.611676606430205e-06,
+      "loss": 0.1473,
+      "step": 181210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.609434554504284e-06,
+      "loss": 0.1458,
+      "step": 181220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.607192502578361e-06,
+      "loss": 0.1459,
+      "step": 181230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.604950450652438e-06,
+      "loss": 0.1512,
+      "step": 181240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.602708398726515e-06,
+      "loss": 0.1387,
+      "step": 181250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.600466346800592e-06,
+      "loss": 0.1514,
+      "step": 181260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.59822429487467e-06,
+      "loss": 0.149,
+      "step": 181270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.595982242948748e-06,
+      "loss": 0.1445,
+      "step": 181280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.593740191022825e-06,
+      "loss": 0.1419,
+      "step": 181290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.591498139096902e-06,
+      "loss": 0.1494,
+      "step": 181300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.589256087170979e-06,
+      "loss": 0.1484,
+      "step": 181310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.587014035245056e-06,
+      "loss": 0.143,
+      "step": 181320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.584771983319133e-06,
+      "loss": 0.1423,
+      "step": 181330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.582529931393212e-06,
+      "loss": 0.152,
+      "step": 181340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.580287879467289e-06,
+      "loss": 0.1442,
+      "step": 181350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.578045827541367e-06,
+      "loss": 0.143,
+      "step": 181360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.575803775615445e-06,
+      "loss": 0.1436,
+      "step": 181370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.573561723689522e-06,
+      "loss": 0.1483,
+      "step": 181380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.571319671763599e-06,
+      "loss": 0.15,
+      "step": 181390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.569077619837676e-06,
+      "loss": 0.1482,
+      "step": 181400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.566835567911753e-06,
+      "loss": 0.1475,
+      "step": 181410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.564817721178424e-06,
+      "loss": 0.1467,
+      "step": 181420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.5625756692525e-06,
+      "loss": 0.1448,
+      "step": 181430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.560333617326578e-06,
+      "loss": 0.1484,
+      "step": 181440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.558091565400655e-06,
+      "loss": 0.1481,
+      "step": 181450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.555849513474732e-06,
+      "loss": 0.1479,
+      "step": 181460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.553607461548809e-06,
+      "loss": 0.148,
+      "step": 181470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.551365409622888e-06,
+      "loss": 0.1402,
+      "step": 181480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.549123357696965e-06,
+      "loss": 0.1465,
+      "step": 181490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.546881305771042e-06,
+      "loss": 0.149,
+      "step": 181500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.54463925384512e-06,
+      "loss": 0.1476,
+      "step": 181510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.542397201919197e-06,
+      "loss": 0.1527,
+      "step": 181520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.540155149993274e-06,
+      "loss": 0.1466,
+      "step": 181530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.537913098067351e-06,
+      "loss": 0.1445,
+      "step": 181540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.53567104614143e-06,
+      "loss": 0.1462,
+      "step": 181550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.533428994215507e-06,
+      "loss": 0.1494,
+      "step": 181560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.531186942289584e-06,
+      "loss": 0.1473,
+      "step": 181570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.528944890363661e-06,
+      "loss": 0.1426,
+      "step": 181580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.526702838437738e-06,
+      "loss": 0.1486,
+      "step": 181590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.524460786511815e-06,
+      "loss": 0.1462,
+      "step": 181600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.522218734585892e-06,
+      "loss": 0.152,
+      "step": 181610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.519976682659971e-06,
+      "loss": 0.15,
+      "step": 181620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.517734630734048e-06,
+      "loss": 0.1425,
+      "step": 181630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.515492578808125e-06,
+      "loss": 0.144,
+      "step": 181640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.513250526882204e-06,
+      "loss": 0.1432,
+      "step": 181650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.51100847495628e-06,
+      "loss": 0.1473,
+      "step": 181660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.508766423030358e-06,
+      "loss": 0.1436,
+      "step": 181670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.506524371104435e-06,
+      "loss": 0.1453,
+      "step": 181680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.504282319178514e-06,
+      "loss": 0.1437,
+      "step": 181690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.50204026725259e-06,
+      "loss": 0.1444,
+      "step": 181700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.499798215326668e-06,
+      "loss": 0.1501,
+      "step": 181710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.497556163400745e-06,
+      "loss": 0.1474,
+      "step": 181720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.495314111474822e-06,
+      "loss": 0.1483,
+      "step": 181730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.493072059548899e-06,
+      "loss": 0.1501,
+      "step": 181740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.490830007622976e-06,
+      "loss": 0.1479,
+      "step": 181750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.488587955697054e-06,
+      "loss": 0.1468,
+      "step": 181760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.486345903771131e-06,
+      "loss": 0.1468,
+      "step": 181770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.48410385184521e-06,
+      "loss": 0.1472,
+      "step": 181780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.481861799919287e-06,
+      "loss": 0.1443,
+      "step": 181790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.479619747993364e-06,
+      "loss": 0.1411,
+      "step": 181800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.477377696067441e-06,
+      "loss": 0.1501,
+      "step": 181810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.475135644141518e-06,
+      "loss": 0.1461,
+      "step": 181820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.472893592215597e-06,
+      "loss": 0.1494,
+      "step": 181830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.470651540289674e-06,
+      "loss": 0.1506,
+      "step": 181840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.468409488363751e-06,
+      "loss": 0.1509,
+      "step": 181850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.466167436437828e-06,
+      "loss": 0.1522,
+      "step": 181860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.463925384511905e-06,
+      "loss": 0.1444,
+      "step": 181870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.461683332585982e-06,
+      "loss": 0.1396,
+      "step": 181880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.459441280660059e-06,
+      "loss": 0.1448,
+      "step": 181890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.457199228734138e-06,
+      "loss": 0.1412,
+      "step": 181900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.454957176808215e-06,
+      "loss": 0.1476,
+      "step": 181910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.452715124882294e-06,
+      "loss": 0.1384,
+      "step": 181920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.45047307295637e-06,
+      "loss": 0.1423,
+      "step": 181930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.448231021030448e-06,
+      "loss": 0.1441,
+      "step": 181940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.445988969104525e-06,
+      "loss": 0.1484,
+      "step": 181950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.443746917178602e-06,
+      "loss": 0.1468,
+      "step": 181960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.44150486525268e-06,
+      "loss": 0.1421,
+      "step": 181970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.439262813326757e-06,
+      "loss": 0.148,
+      "step": 181980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.437020761400834e-06,
+      "loss": 0.1503,
+      "step": 181990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.434778709474911e-06,
+      "loss": 0.1429,
+      "step": 182000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.432536657548989e-06,
+      "loss": 0.1513,
+      "step": 182010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.430294605623066e-06,
+      "loss": 0.1507,
+      "step": 182020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.428052553697143e-06,
+      "loss": 0.1458,
+      "step": 182030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.425810501771221e-06,
+      "loss": 0.1462,
+      "step": 182040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.4235684498453e-06,
+      "loss": 0.1474,
+      "step": 182050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.421326397919377e-06,
+      "loss": 0.1445,
+      "step": 182060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.419084345993454e-06,
+      "loss": 0.1482,
+      "step": 182070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.416842294067531e-06,
+      "loss": 0.1448,
+      "step": 182080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.414600242141608e-06,
+      "loss": 0.1489,
+      "step": 182090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.412358190215685e-06,
+      "loss": 0.146,
+      "step": 182100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.410116138289764e-06,
+      "loss": 0.1432,
+      "step": 182110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.407874086363841e-06,
+      "loss": 0.1476,
+      "step": 182120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.405632034437918e-06,
+      "loss": 0.1464,
+      "step": 182130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.403389982511995e-06,
+      "loss": 0.151,
+      "step": 182140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.401147930586072e-06,
+      "loss": 0.1445,
+      "step": 182150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.398905878660149e-06,
+      "loss": 0.148,
+      "step": 182160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.396663826734228e-06,
+      "loss": 0.1443,
+      "step": 182170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.394421774808305e-06,
+      "loss": 0.1464,
+      "step": 182180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.392179722882383e-06,
+      "loss": 0.1429,
+      "step": 182190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.38993767095646e-06,
+      "loss": 0.1487,
+      "step": 182200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.387695619030537e-06,
+      "loss": 0.1457,
+      "step": 182210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.385453567104614e-06,
+      "loss": 0.1494,
+      "step": 182220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.383211515178692e-06,
+      "loss": 0.1492,
+      "step": 182230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.38096946325277e-06,
+      "loss": 0.1449,
+      "step": 182240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.378727411326847e-06,
+      "loss": 0.1489,
+      "step": 182250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.376485359400924e-06,
+      "loss": 0.1524,
+      "step": 182260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.374243307475001e-06,
+      "loss": 0.1484,
+      "step": 182270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.372001255549078e-06,
+      "loss": 0.1465,
+      "step": 182280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.369759203623155e-06,
+      "loss": 0.1461,
+      "step": 182290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.367517151697232e-06,
+      "loss": 0.145,
+      "step": 182300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.365275099771311e-06,
+      "loss": 0.1481,
+      "step": 182310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.36303304784539e-06,
+      "loss": 0.1443,
+      "step": 182320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.360790995919467e-06,
+      "loss": 0.146,
+      "step": 182330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.358548943993544e-06,
+      "loss": 0.1472,
+      "step": 182340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.356306892067621e-06,
+      "loss": 0.1461,
+      "step": 182350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.354064840141698e-06,
+      "loss": 0.1524,
+      "step": 182360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.351822788215775e-06,
+      "loss": 0.1494,
+      "step": 182370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.349580736289854e-06,
+      "loss": 0.1477,
+      "step": 182380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.34733868436393e-06,
+      "loss": 0.1465,
+      "step": 182390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.345096632438008e-06,
+      "loss": 0.1473,
+      "step": 182400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.342854580512085e-06,
+      "loss": 0.1453,
+      "step": 182410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.340612528586162e-06,
+      "loss": 0.1417,
+      "step": 182420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.338370476660239e-06,
+      "loss": 0.1443,
+      "step": 182430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.336128424734317e-06,
+      "loss": 0.1447,
+      "step": 182440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.333886372808394e-06,
+      "loss": 0.1481,
+      "step": 182450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.331644320882473e-06,
+      "loss": 0.1433,
+      "step": 182460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.32940226895655e-06,
+      "loss": 0.1454,
+      "step": 182470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.327160217030627e-06,
+      "loss": 0.1405,
+      "step": 182480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.324918165104704e-06,
+      "loss": 0.1432,
+      "step": 182490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.322676113178781e-06,
+      "loss": 0.1492,
+      "step": 182500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.320434061252858e-06,
+      "loss": 0.1449,
+      "step": 182510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.318192009326937e-06,
+      "loss": 0.147,
+      "step": 182520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.315949957401014e-06,
+      "loss": 0.1464,
+      "step": 182530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.313707905475091e-06,
+      "loss": 0.1483,
+      "step": 182540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.311465853549168e-06,
+      "loss": 0.1409,
+      "step": 182550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.309223801623245e-06,
+      "loss": 0.1433,
+      "step": 182560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.306981749697322e-06,
+      "loss": 0.1507,
+      "step": 182570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.304739697771401e-06,
+      "loss": 0.147,
+      "step": 182580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.30249764584548e-06,
+      "loss": 0.1522,
+      "step": 182590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.300255593919557e-06,
+      "loss": 0.1523,
+      "step": 182600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.298013541993634e-06,
+      "loss": 0.1469,
+      "step": 182610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.29577149006771e-06,
+      "loss": 0.1479,
+      "step": 182620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.293529438141788e-06,
+      "loss": 0.1473,
+      "step": 182630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.291287386215865e-06,
+      "loss": 0.1482,
+      "step": 182640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.289045334289942e-06,
+      "loss": 0.1405,
+      "step": 182650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.28680328236402e-06,
+      "loss": 0.1442,
+      "step": 182660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.284561230438097e-06,
+      "loss": 0.1499,
+      "step": 182670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.282319178512175e-06,
+      "loss": 0.1461,
+      "step": 182680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.280077126586252e-06,
+      "loss": 0.1461,
+      "step": 182690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.277835074660329e-06,
+      "loss": 0.149,
+      "step": 182700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.275593022734407e-06,
+      "loss": 0.1451,
+      "step": 182710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.273350970808484e-06,
+      "loss": 0.1474,
+      "step": 182720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.271108918882563e-06,
+      "loss": 0.1489,
+      "step": 182730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.26886686695664e-06,
+      "loss": 0.1403,
+      "step": 182740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.266624815030717e-06,
+      "loss": 0.1458,
+      "step": 182750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.264382763104794e-06,
+      "loss": 0.1492,
+      "step": 182760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.262140711178871e-06,
+      "loss": 0.1476,
+      "step": 182770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.259898659252948e-06,
+      "loss": 0.1494,
+      "step": 182780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.257656607327025e-06,
+      "loss": 0.1482,
+      "step": 182790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.255414555401104e-06,
+      "loss": 0.1489,
+      "step": 182800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.253172503475181e-06,
+      "loss": 0.146,
+      "step": 182810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.250930451549258e-06,
+      "loss": 0.1413,
+      "step": 182820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.248688399623335e-06,
+      "loss": 0.1436,
+      "step": 182830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.246446347697412e-06,
+      "loss": 0.1433,
+      "step": 182840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.24420429577149e-06,
+      "loss": 0.1444,
+      "step": 182850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.24196224384557e-06,
+      "loss": 0.1428,
+      "step": 182860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.239720191919646e-06,
+      "loss": 0.1484,
+      "step": 182870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.237478139993723e-06,
+      "loss": 0.1474,
+      "step": 182880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.2352360880678e-06,
+      "loss": 0.1499,
+      "step": 182890
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.232994036141878e-06,
+      "loss": 0.1483,
+      "step": 182900
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.230751984215955e-06,
+      "loss": 0.1401,
+      "step": 182910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.228509932290032e-06,
+      "loss": 0.1444,
+      "step": 182920
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.22626788036411e-06,
+      "loss": 0.1482,
+      "step": 182930
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.224025828438187e-06,
+      "loss": 0.1491,
+      "step": 182940
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.221783776512264e-06,
+      "loss": 0.1475,
+      "step": 182950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.219541724586341e-06,
+      "loss": 0.1523,
+      "step": 182960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.217299672660418e-06,
+      "loss": 0.1438,
+      "step": 182970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.215057620734497e-06,
+      "loss": 0.1492,
+      "step": 182980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.212815568808574e-06,
+      "loss": 0.1484,
+      "step": 182990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.210573516882653e-06,
+      "loss": 0.1435,
+      "step": 183000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.20833146495673e-06,
+      "loss": 0.1409,
+      "step": 183010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.206089413030807e-06,
+      "loss": 0.1416,
+      "step": 183020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.203847361104884e-06,
+      "loss": 0.1446,
+      "step": 183030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.201605309178961e-06,
+      "loss": 0.1452,
+      "step": 183040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.199363257253038e-06,
+      "loss": 0.1419,
+      "step": 183050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.197121205327115e-06,
+      "loss": 0.1465,
+      "step": 183060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.194879153401194e-06,
+      "loss": 0.1455,
+      "step": 183070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.19263710147527e-06,
+      "loss": 0.1519,
+      "step": 183080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.190395049549348e-06,
+      "loss": 0.1508,
+      "step": 183090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.188152997623425e-06,
+      "loss": 0.1429,
+      "step": 183100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.185910945697502e-06,
+      "loss": 0.1486,
+      "step": 183110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.18366889377158e-06,
+      "loss": 0.1467,
+      "step": 183120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.181426841845658e-06,
+      "loss": 0.1477,
+      "step": 183130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.179184789919736e-06,
+      "loss": 0.1463,
+      "step": 183140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.176942737993813e-06,
+      "loss": 0.1403,
+      "step": 183150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.17470068606789e-06,
+      "loss": 0.1503,
+      "step": 183160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.172458634141967e-06,
+      "loss": 0.141,
+      "step": 183170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.170216582216044e-06,
+      "loss": 0.146,
+      "step": 183180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.167974530290121e-06,
+      "loss": 0.1478,
+      "step": 183190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.165732478364198e-06,
+      "loss": 0.1418,
+      "step": 183200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.163490426438277e-06,
+      "loss": 0.1465,
+      "step": 183210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.161248374512354e-06,
+      "loss": 0.1434,
+      "step": 183220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.159006322586431e-06,
+      "loss": 0.1508,
+      "step": 183230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.156764270660508e-06,
+      "loss": 0.1467,
+      "step": 183240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.154522218734587e-06,
+      "loss": 0.1411,
+      "step": 183250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.152280166808664e-06,
+      "loss": 0.1407,
+      "step": 183260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.150038114882741e-06,
+      "loss": 0.1449,
+      "step": 183270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.14779606295682e-06,
+      "loss": 0.1481,
+      "step": 183280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.145554011030897e-06,
+      "loss": 0.1422,
+      "step": 183290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.143311959104974e-06,
+      "loss": 0.147,
+      "step": 183300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.14106990717905e-06,
+      "loss": 0.1409,
+      "step": 183310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.138827855253128e-06,
+      "loss": 0.1442,
+      "step": 183320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.136585803327205e-06,
+      "loss": 0.1442,
+      "step": 183330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.134343751401282e-06,
+      "loss": 0.1456,
+      "step": 183340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.13210169947536e-06,
+      "loss": 0.1428,
+      "step": 183350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.129859647549438e-06,
+      "loss": 0.1473,
+      "step": 183360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.127617595623515e-06,
+      "loss": 0.1492,
+      "step": 183370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.125375543697592e-06,
+      "loss": 0.1427,
+      "step": 183380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.12313349177167e-06,
+      "loss": 0.1398,
+      "step": 183390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.120891439845747e-06,
+      "loss": 0.1421,
+      "step": 183400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.118649387919824e-06,
+      "loss": 0.1425,
+      "step": 183410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.116407335993903e-06,
+      "loss": 0.1408,
+      "step": 183420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.114389489260573e-06,
+      "loss": 0.1448,
+      "step": 183430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.11214743733465e-06,
+      "loss": 0.1455,
+      "step": 183440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.109905385408727e-06,
+      "loss": 0.1457,
+      "step": 183450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.107663333482804e-06,
+      "loss": 0.1435,
+      "step": 183460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.10542128155688e-06,
+      "loss": 0.1461,
+      "step": 183470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.10317922963096e-06,
+      "loss": 0.1438,
+      "step": 183480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.100937177705036e-06,
+      "loss": 0.1446,
+      "step": 183490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.098695125779113e-06,
+      "loss": 0.1451,
+      "step": 183500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.09645307385319e-06,
+      "loss": 0.1431,
+      "step": 183510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.094211021927267e-06,
+      "loss": 0.1438,
+      "step": 183520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.091968970001344e-06,
+      "loss": 0.1458,
+      "step": 183530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.089726918075423e-06,
+      "loss": 0.1457,
+      "step": 183540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.0874848661495e-06,
+      "loss": 0.1449,
+      "step": 183550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.085242814223579e-06,
+      "loss": 0.1472,
+      "step": 183560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.083000762297656e-06,
+      "loss": 0.1485,
+      "step": 183570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.080758710371733e-06,
+      "loss": 0.1432,
+      "step": 183580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.07851665844581e-06,
+      "loss": 0.1487,
+      "step": 183590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.076274606519887e-06,
+      "loss": 0.1447,
+      "step": 183600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.074032554593964e-06,
+      "loss": 0.1379,
+      "step": 183610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.071790502668043e-06,
+      "loss": 0.1446,
+      "step": 183620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.06954845074212e-06,
+      "loss": 0.1468,
+      "step": 183630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.067306398816197e-06,
+      "loss": 0.1435,
+      "step": 183640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.065064346890274e-06,
+      "loss": 0.149,
+      "step": 183650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.062822294964351e-06,
+      "loss": 0.1459,
+      "step": 183660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.060580243038428e-06,
+      "loss": 0.1446,
+      "step": 183670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.058338191112507e-06,
+      "loss": 0.1499,
+      "step": 183680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.056096139186585e-06,
+      "loss": 0.1446,
+      "step": 183690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.053854087260662e-06,
+      "loss": 0.1419,
+      "step": 183700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.05161203533474e-06,
+      "loss": 0.1438,
+      "step": 183710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.049369983408816e-06,
+      "loss": 0.1399,
+      "step": 183720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.047127931482893e-06,
+      "loss": 0.1409,
+      "step": 183730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.04488587955697e-06,
+      "loss": 0.1456,
+      "step": 183740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.04264382763105e-06,
+      "loss": 0.1431,
+      "step": 183750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.040401775705126e-06,
+      "loss": 0.1401,
+      "step": 183760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.038159723779203e-06,
+      "loss": 0.1461,
+      "step": 183770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.03591767185328e-06,
+      "loss": 0.1508,
+      "step": 183780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.033675619927357e-06,
+      "loss": 0.1544,
+      "step": 183790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.031433568001434e-06,
+      "loss": 0.1396,
+      "step": 183800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.029191516075513e-06,
+      "loss": 0.1466,
+      "step": 183810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.02694946414959e-06,
+      "loss": 0.1464,
+      "step": 183820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.024707412223669e-06,
+      "loss": 0.1483,
+      "step": 183830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.022465360297746e-06,
+      "loss": 0.1461,
+      "step": 183840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.020223308371823e-06,
+      "loss": 0.1469,
+      "step": 183850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.0179812564459e-06,
+      "loss": 0.1446,
+      "step": 183860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.015739204519977e-06,
+      "loss": 0.1438,
+      "step": 183870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.013497152594054e-06,
+      "loss": 0.1437,
+      "step": 183880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.011255100668133e-06,
+      "loss": 0.145,
+      "step": 183890
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.00901304874221e-06,
+      "loss": 0.1481,
+      "step": 183900
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.006770996816287e-06,
+      "loss": 0.1445,
+      "step": 183910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.004528944890364e-06,
+      "loss": 0.1484,
+      "step": 183920
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.00228689296444e-06,
+      "loss": 0.1469,
+      "step": 183930
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.000044841038518e-06,
+      "loss": 0.144,
+      "step": 183940
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.997802789112596e-06,
+      "loss": 0.1474,
+      "step": 183950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.995560737186675e-06,
+      "loss": 0.1453,
+      "step": 183960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.993318685260752e-06,
+      "loss": 0.1444,
+      "step": 183970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.99107663333483e-06,
+      "loss": 0.1481,
+      "step": 183980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.988834581408906e-06,
+      "loss": 0.1476,
+      "step": 183990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.986592529482983e-06,
+      "loss": 0.1527,
+      "step": 184000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.98435047755706e-06,
+      "loss": 0.1444,
+      "step": 184010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.982108425631137e-06,
+      "loss": 0.1487,
+      "step": 184020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.979866373705216e-06,
+      "loss": 0.1433,
+      "step": 184030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.977624321779293e-06,
+      "loss": 0.1402,
+      "step": 184040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.97538226985337e-06,
+      "loss": 0.142,
+      "step": 184050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.973140217927447e-06,
+      "loss": 0.1446,
+      "step": 184060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.970898166001524e-06,
+      "loss": 0.1473,
+      "step": 184070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.968656114075603e-06,
+      "loss": 0.1393,
+      "step": 184080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.96641406214968e-06,
+      "loss": 0.1441,
+      "step": 184090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.964172010223759e-06,
+      "loss": 0.1443,
+      "step": 184100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.962154163490426e-06,
+      "loss": 0.1477,
+      "step": 184110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.959912111564505e-06,
+      "loss": 0.1446,
+      "step": 184120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.957670059638582e-06,
+      "loss": 0.146,
+      "step": 184130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.955428007712659e-06,
+      "loss": 0.1446,
+      "step": 184140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.953185955786736e-06,
+      "loss": 0.1429,
+      "step": 184150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.950943903860815e-06,
+      "loss": 0.143,
+      "step": 184160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.948701851934892e-06,
+      "loss": 0.1489,
+      "step": 184170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.946459800008969e-06,
+      "loss": 0.1454,
+      "step": 184180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.944217748083046e-06,
+      "loss": 0.1423,
+      "step": 184190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.941975696157123e-06,
+      "loss": 0.1479,
+      "step": 184200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.9397336442312e-06,
+      "loss": 0.1498,
+      "step": 184210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.937491592305277e-06,
+      "loss": 0.1433,
+      "step": 184220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.935249540379356e-06,
+      "loss": 0.1476,
+      "step": 184230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.933007488453433e-06,
+      "loss": 0.1379,
+      "step": 184240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.930765436527511e-06,
+      "loss": 0.1431,
+      "step": 184250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.928523384601588e-06,
+      "loss": 0.1396,
+      "step": 184260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.926281332675666e-06,
+      "loss": 0.1459,
+      "step": 184270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.924039280749743e-06,
+      "loss": 0.1447,
+      "step": 184280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.92179722882382e-06,
+      "loss": 0.1416,
+      "step": 184290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.919555176897898e-06,
+      "loss": 0.1439,
+      "step": 184300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.917313124971975e-06,
+      "loss": 0.1485,
+      "step": 184310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.915071073046052e-06,
+      "loss": 0.1465,
+      "step": 184320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.91282902112013e-06,
+      "loss": 0.1465,
+      "step": 184330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.910586969194206e-06,
+      "loss": 0.1464,
+      "step": 184340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.908344917268283e-06,
+      "loss": 0.1409,
+      "step": 184350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.90610286534236e-06,
+      "loss": 0.1414,
+      "step": 184360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.903860813416439e-06,
+      "loss": 0.1435,
+      "step": 184370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.901618761490516e-06,
+      "loss": 0.1465,
+      "step": 184380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.899376709564595e-06,
+      "loss": 0.1453,
+      "step": 184390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.897134657638672e-06,
+      "loss": 0.1462,
+      "step": 184400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.894892605712749e-06,
+      "loss": 0.1429,
+      "step": 184410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.892650553786826e-06,
+      "loss": 0.1455,
+      "step": 184420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.890408501860903e-06,
+      "loss": 0.1442,
+      "step": 184430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.888166449934982e-06,
+      "loss": 0.1482,
+      "step": 184440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.885924398009059e-06,
+      "loss": 0.1458,
+      "step": 184450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.883682346083136e-06,
+      "loss": 0.1454,
+      "step": 184460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.881440294157213e-06,
+      "loss": 0.1429,
+      "step": 184470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.87919824223129e-06,
+      "loss": 0.1428,
+      "step": 184480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.876956190305367e-06,
+      "loss": 0.1399,
+      "step": 184490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.874714138379444e-06,
+      "loss": 0.1465,
+      "step": 184500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.872472086453523e-06,
+      "loss": 0.1419,
+      "step": 184510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.870230034527601e-06,
+      "loss": 0.1429,
+      "step": 184520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.867987982601678e-06,
+      "loss": 0.1403,
+      "step": 184530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.865745930675755e-06,
+      "loss": 0.148,
+      "step": 184540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.863503878749832e-06,
+      "loss": 0.1491,
+      "step": 184550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.86126182682391e-06,
+      "loss": 0.1422,
+      "step": 184560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.859019774897986e-06,
+      "loss": 0.1438,
+      "step": 184570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.856777722972065e-06,
+      "loss": 0.1404,
+      "step": 184580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.854535671046142e-06,
+      "loss": 0.1457,
+      "step": 184590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.852293619120219e-06,
+      "loss": 0.1391,
+      "step": 184600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.850051567194296e-06,
+      "loss": 0.146,
+      "step": 184610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.847809515268373e-06,
+      "loss": 0.1487,
+      "step": 184620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.84556746334245e-06,
+      "loss": 0.1449,
+      "step": 184630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.843325411416529e-06,
+      "loss": 0.1396,
+      "step": 184640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.841083359490606e-06,
+      "loss": 0.1448,
+      "step": 184650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.838841307564685e-06,
+      "loss": 0.1427,
+      "step": 184660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.836599255638762e-06,
+      "loss": 0.1438,
+      "step": 184670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.834357203712839e-06,
+      "loss": 0.1493,
+      "step": 184680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.832115151786916e-06,
+      "loss": 0.1415,
+      "step": 184690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.829873099860993e-06,
+      "loss": 0.1445,
+      "step": 184700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.827631047935071e-06,
+      "loss": 0.1427,
+      "step": 184710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.825388996009149e-06,
+      "loss": 0.1392,
+      "step": 184720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.823146944083226e-06,
+      "loss": 0.1479,
+      "step": 184730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.820904892157303e-06,
+      "loss": 0.1475,
+      "step": 184740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.81866284023138e-06,
+      "loss": 0.1478,
+      "step": 184750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.816420788305457e-06,
+      "loss": 0.148,
+      "step": 184760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.814178736379534e-06,
+      "loss": 0.1422,
+      "step": 184770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.811936684453612e-06,
+      "loss": 0.1435,
+      "step": 184780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.809694632527691e-06,
+      "loss": 0.1415,
+      "step": 184790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.807452580601768e-06,
+      "loss": 0.1472,
+      "step": 184800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.805210528675845e-06,
+      "loss": 0.1425,
+      "step": 184810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.802968476749922e-06,
+      "loss": 0.1371,
+      "step": 184820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.800726424824e-06,
+      "loss": 0.1422,
+      "step": 184830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.798484372898076e-06,
+      "loss": 0.1428,
+      "step": 184840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.796242320972155e-06,
+      "loss": 0.1397,
+      "step": 184850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.794000269046232e-06,
+      "loss": 0.1451,
+      "step": 184860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.791758217120309e-06,
+      "loss": 0.1452,
+      "step": 184870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.789516165194386e-06,
+      "loss": 0.1437,
+      "step": 184880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.787274113268463e-06,
+      "loss": 0.1469,
+      "step": 184890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.78503206134254e-06,
+      "loss": 0.1465,
+      "step": 184900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.782790009416619e-06,
+      "loss": 0.1459,
+      "step": 184910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.780547957490696e-06,
+      "loss": 0.1459,
+      "step": 184920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.778305905564774e-06,
+      "loss": 0.1434,
+      "step": 184930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.776063853638852e-06,
+      "loss": 0.1441,
+      "step": 184940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.773821801712929e-06,
+      "loss": 0.1504,
+      "step": 184950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.771579749787006e-06,
+      "loss": 0.1443,
+      "step": 184960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.769337697861083e-06,
+      "loss": 0.1454,
+      "step": 184970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.76709564593516e-06,
+      "loss": 0.1505,
+      "step": 184980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.764853594009238e-06,
+      "loss": 0.1429,
+      "step": 184990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.762611542083315e-06,
+      "loss": 0.1438,
+      "step": 185000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.760369490157392e-06,
+      "loss": 0.1427,
+      "step": 185010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.75812743823147e-06,
+      "loss": 0.146,
+      "step": 185020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.755885386305546e-06,
+      "loss": 0.146,
+      "step": 185030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.753643334379623e-06,
+      "loss": 0.1406,
+      "step": 185040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.751401282453702e-06,
+      "loss": 0.1427,
+      "step": 185050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.749159230527781e-06,
+      "loss": 0.1452,
+      "step": 185060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.746917178601858e-06,
+      "loss": 0.1419,
+      "step": 185070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.744675126675935e-06,
+      "loss": 0.1472,
+      "step": 185080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.742433074750012e-06,
+      "loss": 0.1477,
+      "step": 185090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.740191022824089e-06,
+      "loss": 0.1393,
+      "step": 185100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.737948970898166e-06,
+      "loss": 0.15,
+      "step": 185110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.735706918972243e-06,
+      "loss": 0.1478,
+      "step": 185120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.733464867046322e-06,
+      "loss": 0.1424,
+      "step": 185130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.731222815120399e-06,
+      "loss": 0.1424,
+      "step": 185140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.728980763194476e-06,
+      "loss": 0.1401,
+      "step": 185150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.726738711268553e-06,
+      "loss": 0.1443,
+      "step": 185160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.72449665934263e-06,
+      "loss": 0.1392,
+      "step": 185170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.722254607416709e-06,
+      "loss": 0.1448,
+      "step": 185180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.720012555490786e-06,
+      "loss": 0.1474,
+      "step": 185190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.717770503564864e-06,
+      "loss": 0.1405,
+      "step": 185200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.715528451638941e-06,
+      "loss": 0.1447,
+      "step": 185210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.713286399713018e-06,
+      "loss": 0.1461,
+      "step": 185220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.711044347787095e-06,
+      "loss": 0.1456,
+      "step": 185230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.708802295861172e-06,
+      "loss": 0.1444,
+      "step": 185240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.70656024393525e-06,
+      "loss": 0.1448,
+      "step": 185250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.704318192009326e-06,
+      "loss": 0.1457,
+      "step": 185260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.702076140083405e-06,
+      "loss": 0.1473,
+      "step": 185270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.699834088157482e-06,
+      "loss": 0.1375,
+      "step": 185280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.69759203623156e-06,
+      "loss": 0.1414,
+      "step": 185290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.695349984305636e-06,
+      "loss": 0.1438,
+      "step": 185300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.693107932379713e-06,
+      "loss": 0.1412,
+      "step": 185310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.690865880453792e-06,
+      "loss": 0.1469,
+      "step": 185320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.688623828527869e-06,
+      "loss": 0.1456,
+      "step": 185330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.686381776601948e-06,
+      "loss": 0.1447,
+      "step": 185340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.684139724676025e-06,
+      "loss": 0.1411,
+      "step": 185350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.681897672750102e-06,
+      "loss": 0.1451,
+      "step": 185360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.679655620824179e-06,
+      "loss": 0.1426,
+      "step": 185370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.677413568898256e-06,
+      "loss": 0.1428,
+      "step": 185380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.675171516972333e-06,
+      "loss": 0.1497,
+      "step": 185390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.672929465046412e-06,
+      "loss": 0.1438,
+      "step": 185400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.670687413120489e-06,
+      "loss": 0.1489,
+      "step": 185410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.668445361194566e-06,
+      "loss": 0.1457,
+      "step": 185420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.666203309268643e-06,
+      "loss": 0.1436,
+      "step": 185430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.66396125734272e-06,
+      "loss": 0.1438,
+      "step": 185440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.661719205416798e-06,
+      "loss": 0.1421,
+      "step": 185450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.659477153490875e-06,
+      "loss": 0.1389,
+      "step": 185460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.657235101564954e-06,
+      "loss": 0.1416,
+      "step": 185470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.654993049639031e-06,
+      "loss": 0.1422,
+      "step": 185480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.652750997713108e-06,
+      "loss": 0.1429,
+      "step": 185490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.650508945787185e-06,
+      "loss": 0.1444,
+      "step": 185500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.648266893861262e-06,
+      "loss": 0.1398,
+      "step": 185510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.64602484193534e-06,
+      "loss": 0.146,
+      "step": 185520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.643782790009416e-06,
+      "loss": 0.139,
+      "step": 185530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.641540738083495e-06,
+      "loss": 0.1458,
+      "step": 185540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.639298686157572e-06,
+      "loss": 0.1457,
+      "step": 185550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.637056634231649e-06,
+      "loss": 0.1402,
+      "step": 185560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.634814582305726e-06,
+      "loss": 0.1448,
+      "step": 185570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.632572530379803e-06,
+      "loss": 0.1468,
+      "step": 185580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.630330478453882e-06,
+      "loss": 0.1482,
+      "step": 185590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.628088426527959e-06,
+      "loss": 0.1431,
+      "step": 185600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.625846374602038e-06,
+      "loss": 0.1442,
+      "step": 185610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.623604322676115e-06,
+      "loss": 0.1451,
+      "step": 185620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.621362270750192e-06,
+      "loss": 0.1477,
+      "step": 185630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.619120218824269e-06,
+      "loss": 0.1476,
+      "step": 185640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.616878166898346e-06,
+      "loss": 0.1458,
+      "step": 185650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.614636114972423e-06,
+      "loss": 0.1414,
+      "step": 185660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.6123940630465e-06,
+      "loss": 0.1461,
+      "step": 185670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.610152011120578e-06,
+      "loss": 0.1422,
+      "step": 185680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.607909959194655e-06,
+      "loss": 0.1411,
+      "step": 185690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.605667907268732e-06,
+      "loss": 0.146,
+      "step": 185700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.60342585534281e-06,
+      "loss": 0.1452,
+      "step": 185710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.601183803416888e-06,
+      "loss": 0.1491,
+      "step": 185720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.598941751490965e-06,
+      "loss": 0.1455,
+      "step": 185730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.596699699565042e-06,
+      "loss": 0.1454,
+      "step": 185740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.594457647639121e-06,
+      "loss": 0.1401,
+      "step": 185750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.592215595713198e-06,
+      "loss": 0.144,
+      "step": 185760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.589973543787275e-06,
+      "loss": 0.1489,
+      "step": 185770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.587731491861352e-06,
+      "loss": 0.1446,
+      "step": 185780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.585489439935429e-06,
+      "loss": 0.1447,
+      "step": 185790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.583247388009506e-06,
+      "loss": 0.146,
+      "step": 185800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.581005336083583e-06,
+      "loss": 0.1464,
+      "step": 185810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.578763284157662e-06,
+      "loss": 0.1463,
+      "step": 185820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.576521232231739e-06,
+      "loss": 0.1436,
+      "step": 185830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.574279180305816e-06,
+      "loss": 0.1405,
+      "step": 185840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.572037128379893e-06,
+      "loss": 0.141,
+      "step": 185850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.569795076453972e-06,
+      "loss": 0.1499,
+      "step": 185860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.567553024528049e-06,
+      "loss": 0.1438,
+      "step": 185870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.565310972602126e-06,
+      "loss": 0.1413,
+      "step": 185880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.563068920676204e-06,
+      "loss": 0.14,
+      "step": 185890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.560826868750281e-06,
+      "loss": 0.1406,
+      "step": 185900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.558584816824358e-06,
+      "loss": 0.145,
+      "step": 185910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.556342764898435e-06,
+      "loss": 0.1465,
+      "step": 185920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.554100712972512e-06,
+      "loss": 0.1448,
+      "step": 185930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.55185866104659e-06,
+      "loss": 0.1481,
+      "step": 185940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.549616609120666e-06,
+      "loss": 0.1409,
+      "step": 185950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.547374557194745e-06,
+      "loss": 0.142,
+      "step": 185960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.545132505268822e-06,
+      "loss": 0.143,
+      "step": 185970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.5428904533429e-06,
+      "loss": 0.1449,
+      "step": 185980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.540648401416978e-06,
+      "loss": 0.1442,
+      "step": 185990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.538406349491055e-06,
+      "loss": 0.1469,
+      "step": 186000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.536164297565132e-06,
+      "loss": 0.144,
+      "step": 186010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.533922245639209e-06,
+      "loss": 0.1426,
+      "step": 186020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.531680193713288e-06,
+      "loss": 0.1446,
+      "step": 186030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.529438141787365e-06,
+      "loss": 0.1471,
+      "step": 186040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.527196089861442e-06,
+      "loss": 0.147,
+      "step": 186050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.524954037935519e-06,
+      "loss": 0.1433,
+      "step": 186060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.522711986009596e-06,
+      "loss": 0.1424,
+      "step": 186070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.520469934083673e-06,
+      "loss": 0.1425,
+      "step": 186080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.518227882157752e-06,
+      "loss": 0.1419,
+      "step": 186090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.515985830231829e-06,
+      "loss": 0.1426,
+      "step": 186100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.513743778305906e-06,
+      "loss": 0.145,
+      "step": 186110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.511501726379983e-06,
+      "loss": 0.142,
+      "step": 186120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.509259674454061e-06,
+      "loss": 0.1404,
+      "step": 186130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.507017622528138e-06,
+      "loss": 0.1428,
+      "step": 186140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.504775570602215e-06,
+      "loss": 0.1422,
+      "step": 186150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.502533518676294e-06,
+      "loss": 0.1422,
+      "step": 186160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.500291466750371e-06,
+      "loss": 0.1428,
+      "step": 186170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.498049414824448e-06,
+      "loss": 0.1374,
+      "step": 186180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.495807362898525e-06,
+      "loss": 0.1497,
+      "step": 186190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.493565310972602e-06,
+      "loss": 0.1414,
+      "step": 186200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.49132325904668e-06,
+      "loss": 0.1439,
+      "step": 186210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.489081207120756e-06,
+      "loss": 0.1451,
+      "step": 186220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.486839155194835e-06,
+      "loss": 0.1433,
+      "step": 186230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.484597103268912e-06,
+      "loss": 0.1428,
+      "step": 186240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.482355051342989e-06,
+      "loss": 0.1433,
+      "step": 186250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.480112999417068e-06,
+      "loss": 0.1419,
+      "step": 186260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.477870947491145e-06,
+      "loss": 0.1444,
+      "step": 186270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.475628895565222e-06,
+      "loss": 0.1431,
+      "step": 186280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.473386843639299e-06,
+      "loss": 0.1445,
+      "step": 186290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.471144791713378e-06,
+      "loss": 0.1438,
+      "step": 186300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.468902739787455e-06,
+      "loss": 0.1455,
+      "step": 186310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.466660687861532e-06,
+      "loss": 0.1388,
+      "step": 186320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.464418635935609e-06,
+      "loss": 0.1427,
+      "step": 186330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.462176584009686e-06,
+      "loss": 0.1423,
+      "step": 186340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.459934532083763e-06,
+      "loss": 0.1456,
+      "step": 186350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.45769248015784e-06,
+      "loss": 0.1445,
+      "step": 186360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.455450428231918e-06,
+      "loss": 0.1444,
+      "step": 186370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.453208376305995e-06,
+      "loss": 0.1392,
+      "step": 186380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.450966324380072e-06,
+      "loss": 0.1393,
+      "step": 186390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.448724272454151e-06,
+      "loss": 0.139,
+      "step": 186400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.446482220528228e-06,
+      "loss": 0.1415,
+      "step": 186410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.444240168602305e-06,
+      "loss": 0.1446,
+      "step": 186420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.441998116676382e-06,
+      "loss": 0.1457,
+      "step": 186430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.439756064750461e-06,
+      "loss": 0.1448,
+      "step": 186440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.437514012824538e-06,
+      "loss": 0.145,
+      "step": 186450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.435271960898615e-06,
+      "loss": 0.1378,
+      "step": 186460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.433029908972692e-06,
+      "loss": 0.1382,
+      "step": 186470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.430787857046769e-06,
+      "loss": 0.1496,
+      "step": 186480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.428545805120846e-06,
+      "loss": 0.1389,
+      "step": 186490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.426303753194923e-06,
+      "loss": 0.1384,
+      "step": 186500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.424061701269002e-06,
+      "loss": 0.1478,
+      "step": 186510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.421819649343079e-06,
+      "loss": 0.1494,
+      "step": 186520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.419577597417158e-06,
+      "loss": 0.145,
+      "step": 186530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.417335545491235e-06,
+      "loss": 0.1502,
+      "step": 186540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.415093493565312e-06,
+      "loss": 0.1504,
+      "step": 186550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.412851441639389e-06,
+      "loss": 0.1487,
+      "step": 186560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.410609389713466e-06,
+      "loss": 0.1402,
+      "step": 186570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.408367337787544e-06,
+      "loss": 0.1462,
+      "step": 186580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.406125285861621e-06,
+      "loss": 0.1415,
+      "step": 186590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.403883233935698e-06,
+      "loss": 0.1429,
+      "step": 186600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.401641182009775e-06,
+      "loss": 0.1427,
+      "step": 186610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.399399130083852e-06,
+      "loss": 0.1432,
+      "step": 186620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.39715707815793e-06,
+      "loss": 0.1458,
+      "step": 186630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.394915026232007e-06,
+      "loss": 0.1398,
+      "step": 186640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.392672974306085e-06,
+      "loss": 0.1443,
+      "step": 186650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.390430922380162e-06,
+      "loss": 0.1442,
+      "step": 186660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.388188870454241e-06,
+      "loss": 0.1417,
+      "step": 186670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.385946818528318e-06,
+      "loss": 0.1418,
+      "step": 186680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.383704766602395e-06,
+      "loss": 0.1463,
+      "step": 186690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.381462714676472e-06,
+      "loss": 0.1449,
+      "step": 186700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.379220662750549e-06,
+      "loss": 0.1386,
+      "step": 186710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.376978610824628e-06,
+      "loss": 0.1475,
+      "step": 186720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.374736558898705e-06,
+      "loss": 0.1425,
+      "step": 186730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.372494506972782e-06,
+      "loss": 0.1462,
+      "step": 186740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.370252455046859e-06,
+      "loss": 0.1466,
+      "step": 186750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.368010403120936e-06,
+      "loss": 0.1454,
+      "step": 186760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.365768351195013e-06,
+      "loss": 0.1453,
+      "step": 186770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.363526299269092e-06,
+      "loss": 0.1436,
+      "step": 186780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.361284247343169e-06,
+      "loss": 0.1458,
+      "step": 186790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.359042195417247e-06,
+      "loss": 0.1457,
+      "step": 186800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.356800143491324e-06,
+      "loss": 0.1441,
+      "step": 186810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.354558091565401e-06,
+      "loss": 0.14,
+      "step": 186820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.352316039639478e-06,
+      "loss": 0.1486,
+      "step": 186830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.350073987713555e-06,
+      "loss": 0.1473,
+      "step": 186840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.347831935787634e-06,
+      "loss": 0.145,
+      "step": 186850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.345589883861711e-06,
+      "loss": 0.1415,
+      "step": 186860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.343347831935788e-06,
+      "loss": 0.1449,
+      "step": 186870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.341105780009865e-06,
+      "loss": 0.1489,
+      "step": 186880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.338863728083942e-06,
+      "loss": 0.1428,
+      "step": 186890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.33662167615802e-06,
+      "loss": 0.1393,
+      "step": 186900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.334379624232096e-06,
+      "loss": 0.1428,
+      "step": 186910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.332137572306175e-06,
+      "loss": 0.1367,
+      "step": 186920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.329895520380252e-06,
+      "loss": 0.1438,
+      "step": 186930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.32765346845433e-06,
+      "loss": 0.1463,
+      "step": 186940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.325411416528408e-06,
+      "loss": 0.1394,
+      "step": 186950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.323169364602485e-06,
+      "loss": 0.1457,
+      "step": 186960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.320927312676562e-06,
+      "loss": 0.1444,
+      "step": 186970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.318685260750639e-06,
+      "loss": 0.1389,
+      "step": 186980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.316443208824718e-06,
+      "loss": 0.1468,
+      "step": 186990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.314201156898795e-06,
+      "loss": 0.1453,
+      "step": 187000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.311959104972872e-06,
+      "loss": 0.146,
+      "step": 187010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.309717053046949e-06,
+      "loss": 0.1397,
+      "step": 187020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.307475001121026e-06,
+      "loss": 0.1424,
+      "step": 187030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.305232949195103e-06,
+      "loss": 0.1426,
+      "step": 187040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.30299089726918e-06,
+      "loss": 0.148,
+      "step": 187050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.300748845343258e-06,
+      "loss": 0.14,
+      "step": 187060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.298506793417337e-06,
+      "loss": 0.1365,
+      "step": 187070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.296264741491414e-06,
+      "loss": 0.1493,
+      "step": 187080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.294022689565491e-06,
+      "loss": 0.1395,
+      "step": 187090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.291780637639568e-06,
+      "loss": 0.1384,
+      "step": 187100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.289538585713645e-06,
+      "loss": 0.1392,
+      "step": 187110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.287296533787722e-06,
+      "loss": 0.1443,
+      "step": 187120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.285054481861801e-06,
+      "loss": 0.1478,
+      "step": 187130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.282812429935878e-06,
+      "loss": 0.1449,
+      "step": 187140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.280570378009955e-06,
+      "loss": 0.1448,
+      "step": 187150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.278328326084032e-06,
+      "loss": 0.1399,
+      "step": 187160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.276086274158109e-06,
+      "loss": 0.1455,
+      "step": 187170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.273844222232186e-06,
+      "loss": 0.1512,
+      "step": 187180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.271602170306265e-06,
+      "loss": 0.1433,
+      "step": 187190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.269360118380342e-06,
+      "loss": 0.1437,
+      "step": 187200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.26711806645442e-06,
+      "loss": 0.1431,
+      "step": 187210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.264876014528498e-06,
+      "loss": 0.1411,
+      "step": 187220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.262633962602575e-06,
+      "loss": 0.1402,
+      "step": 187230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.260391910676652e-06,
+      "loss": 0.1456,
+      "step": 187240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.258149858750729e-06,
+      "loss": 0.1491,
+      "step": 187250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.255907806824806e-06,
+      "loss": 0.1425,
+      "step": 187260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.253665754898884e-06,
+      "loss": 0.1446,
+      "step": 187270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.251423702972961e-06,
+      "loss": 0.1367,
+      "step": 187280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.249181651047038e-06,
+      "loss": 0.1407,
+      "step": 187290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.246939599121115e-06,
+      "loss": 0.142,
+      "step": 187300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.244697547195193e-06,
+      "loss": 0.1455,
+      "step": 187310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.24245549526927e-06,
+      "loss": 0.1412,
+      "step": 187320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.240213443343348e-06,
+      "loss": 0.1472,
+      "step": 187330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.237971391417427e-06,
+      "loss": 0.1429,
+      "step": 187340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.235729339491504e-06,
+      "loss": 0.1446,
+      "step": 187350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.233487287565581e-06,
+      "loss": 0.1438,
+      "step": 187360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.231245235639658e-06,
+      "loss": 0.1412,
+      "step": 187370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.229003183713735e-06,
+      "loss": 0.1433,
+      "step": 187380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.226761131787812e-06,
+      "loss": 0.1394,
+      "step": 187390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.22451907986189e-06,
+      "loss": 0.1466,
+      "step": 187400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.222277027935968e-06,
+      "loss": 0.1476,
+      "step": 187410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.220034976010045e-06,
+      "loss": 0.1433,
+      "step": 187420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.217792924084122e-06,
+      "loss": 0.1451,
+      "step": 187430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.215550872158199e-06,
+      "loss": 0.143,
+      "step": 187440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.213308820232276e-06,
+      "loss": 0.1373,
+      "step": 187450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.211066768306355e-06,
+      "loss": 0.1418,
+      "step": 187460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.208824716380432e-06,
+      "loss": 0.1454,
+      "step": 187470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.20658266445451e-06,
+      "loss": 0.1459,
+      "step": 187480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.204340612528587e-06,
+      "loss": 0.14,
+      "step": 187490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.202098560602664e-06,
+      "loss": 0.1427,
+      "step": 187500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.199856508676741e-06,
+      "loss": 0.1456,
+      "step": 187510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.197614456750818e-06,
+      "loss": 0.146,
+      "step": 187520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.195372404824896e-06,
+      "loss": 0.1332,
+      "step": 187530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.193130352898974e-06,
+      "loss": 0.1489,
+      "step": 187540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.190888300973051e-06,
+      "loss": 0.1419,
+      "step": 187550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.188646249047128e-06,
+      "loss": 0.1471,
+      "step": 187560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.186404197121205e-06,
+      "loss": 0.1424,
+      "step": 187570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.184162145195282e-06,
+      "loss": 0.139,
+      "step": 187580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.18192009326936e-06,
+      "loss": 0.1458,
+      "step": 187590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.179678041343438e-06,
+      "loss": 0.1406,
+      "step": 187600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.177435989417517e-06,
+      "loss": 0.1401,
+      "step": 187610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.175193937491594e-06,
+      "loss": 0.1456,
+      "step": 187620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.17295188556567e-06,
+      "loss": 0.1456,
+      "step": 187630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.170709833639748e-06,
+      "loss": 0.1465,
+      "step": 187640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.168467781713825e-06,
+      "loss": 0.1468,
+      "step": 187650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.166225729787902e-06,
+      "loss": 0.1443,
+      "step": 187660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.163983677861979e-06,
+      "loss": 0.1441,
+      "step": 187670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.161741625936058e-06,
+      "loss": 0.1375,
+      "step": 187680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.159499574010135e-06,
+      "loss": 0.1481,
+      "step": 187690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.157257522084212e-06,
+      "loss": 0.1445,
+      "step": 187700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.155015470158289e-06,
+      "loss": 0.1432,
+      "step": 187710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.152773418232366e-06,
+      "loss": 0.1375,
+      "step": 187720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.150531366306444e-06,
+      "loss": 0.1426,
+      "step": 187730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.148289314380521e-06,
+      "loss": 0.1417,
+      "step": 187740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.1460472624546e-06,
+      "loss": 0.1433,
+      "step": 187750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.143805210528677e-06,
+      "loss": 0.1426,
+      "step": 187760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.141563158602754e-06,
+      "loss": 0.1421,
+      "step": 187770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.139321106676831e-06,
+      "loss": 0.1495,
+      "step": 187780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.137079054750908e-06,
+      "loss": 0.1427,
+      "step": 187790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.134837002824985e-06,
+      "loss": 0.1463,
+      "step": 187800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.132594950899062e-06,
+      "loss": 0.139,
+      "step": 187810
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.130352898973141e-06,
+      "loss": 0.1495,
+      "step": 187820
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.128110847047218e-06,
+      "loss": 0.1455,
+      "step": 187830
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.125868795121295e-06,
+      "loss": 0.1435,
+      "step": 187840
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.123626743195372e-06,
+      "loss": 0.1496,
+      "step": 187850
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.121384691269449e-06,
+      "loss": 0.1497,
+      "step": 187860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.119142639343528e-06,
+      "loss": 0.1387,
+      "step": 187870
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.116900587417605e-06,
+      "loss": 0.1383,
+      "step": 187880
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.114658535491684e-06,
+      "loss": 0.1381,
+      "step": 187890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.11241648356576e-06,
+      "loss": 0.1439,
+      "step": 187900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.110174431639838e-06,
+      "loss": 0.141,
+      "step": 187910
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.107932379713915e-06,
+      "loss": 0.1425,
+      "step": 187920
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.105690327787992e-06,
+      "loss": 0.1423,
+      "step": 187930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.103448275862069e-06,
+      "loss": 0.1406,
+      "step": 187940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.101206223936146e-06,
+      "loss": 0.1412,
+      "step": 187950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.098964172010224e-06,
+      "loss": 0.1406,
+      "step": 187960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.096722120084301e-06,
+      "loss": 0.1422,
+      "step": 187970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.094480068158379e-06,
+      "loss": 0.1444,
+      "step": 187980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.092238016232456e-06,
+      "loss": 0.1444,
+      "step": 187990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.089995964306534e-06,
+      "loss": 0.146,
+      "step": 188000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.087753912380611e-06,
+      "loss": 0.1464,
+      "step": 188010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.085511860454688e-06,
+      "loss": 0.1465,
+      "step": 188020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.083269808528767e-06,
+      "loss": 0.1407,
+      "step": 188030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.081027756602844e-06,
+      "loss": 0.1404,
+      "step": 188040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.078785704676921e-06,
+      "loss": 0.1458,
+      "step": 188050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.076543652750998e-06,
+      "loss": 0.1451,
+      "step": 188060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.074301600825075e-06,
+      "loss": 0.1455,
+      "step": 188070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.072059548899152e-06,
+      "loss": 0.1398,
+      "step": 188080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.069817496973231e-06,
+      "loss": 0.1421,
+      "step": 188090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.067575445047308e-06,
+      "loss": 0.1423,
+      "step": 188100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.065333393121385e-06,
+      "loss": 0.1414,
+      "step": 188110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.063315546388054e-06,
+      "loss": 0.1405,
+      "step": 188120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.061073494462131e-06,
+      "loss": 0.1457,
+      "step": 188130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.058831442536208e-06,
+      "loss": 0.1354,
+      "step": 188140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.056589390610285e-06,
+      "loss": 0.1388,
+      "step": 188150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.054347338684364e-06,
+      "loss": 0.1393,
+      "step": 188160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.052105286758443e-06,
+      "loss": 0.1377,
+      "step": 188170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.04986323483252e-06,
+      "loss": 0.1391,
+      "step": 188180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.047621182906597e-06,
+      "loss": 0.1413,
+      "step": 188190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.045379130980674e-06,
+      "loss": 0.1375,
+      "step": 188200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.043137079054751e-06,
+      "loss": 0.1401,
+      "step": 188210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.040895027128828e-06,
+      "loss": 0.1436,
+      "step": 188220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.038652975202907e-06,
+      "loss": 0.1495,
+      "step": 188230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.036410923276984e-06,
+      "loss": 0.1493,
+      "step": 188240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.03416887135106e-06,
+      "loss": 0.139,
+      "step": 188250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.031926819425138e-06,
+      "loss": 0.141,
+      "step": 188260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.029684767499215e-06,
+      "loss": 0.1443,
+      "step": 188270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.027442715573292e-06,
+      "loss": 0.1487,
+      "step": 188280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.02520066364737e-06,
+      "loss": 0.1375,
+      "step": 188290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.022958611721448e-06,
+      "loss": 0.1416,
+      "step": 188300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.020716559795526e-06,
+      "loss": 0.1454,
+      "step": 188310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.018474507869603e-06,
+      "loss": 0.1439,
+      "step": 188320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.01623245594368e-06,
+      "loss": 0.1461,
+      "step": 188330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.013990404017757e-06,
+      "loss": 0.1404,
+      "step": 188340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.011748352091834e-06,
+      "loss": 0.1387,
+      "step": 188350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.009506300165913e-06,
+      "loss": 0.1467,
+      "step": 188360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.00726424823999e-06,
+      "loss": 0.1393,
+      "step": 188370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.005022196314067e-06,
+      "loss": 0.1424,
+      "step": 188380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.002780144388144e-06,
+      "loss": 0.1442,
+      "step": 188390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.000538092462221e-06,
+      "loss": 0.1408,
+      "step": 188400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.998296040536298e-06,
+      "loss": 0.1422,
+      "step": 188410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.996053988610375e-06,
+      "loss": 0.1484,
+      "step": 188420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.993811936684454e-06,
+      "loss": 0.1442,
+      "step": 188430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.991569884758531e-06,
+      "loss": 0.1409,
+      "step": 188440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.98932783283261e-06,
+      "loss": 0.1417,
+      "step": 188450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.987085780906687e-06,
+      "loss": 0.1435,
+      "step": 188460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.984843728980764e-06,
+      "loss": 0.1468,
+      "step": 188470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.98260167705484e-06,
+      "loss": 0.1392,
+      "step": 188480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.980359625128918e-06,
+      "loss": 0.1443,
+      "step": 188490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.978117573202997e-06,
+      "loss": 0.1455,
+      "step": 188500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.975875521277074e-06,
+      "loss": 0.1425,
+      "step": 188510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.97363346935115e-06,
+      "loss": 0.1467,
+      "step": 188520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.971391417425228e-06,
+      "loss": 0.139,
+      "step": 188530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.969149365499305e-06,
+      "loss": 0.1421,
+      "step": 188540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.966907313573382e-06,
+      "loss": 0.142,
+      "step": 188550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.96466526164746e-06,
+      "loss": 0.1348,
+      "step": 188560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.962423209721537e-06,
+      "loss": 0.1404,
+      "step": 188570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.960181157795616e-06,
+      "loss": 0.1442,
+      "step": 188580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.957939105869693e-06,
+      "loss": 0.1433,
+      "step": 188590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.95569705394377e-06,
+      "loss": 0.1422,
+      "step": 188600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.953455002017847e-06,
+      "loss": 0.1432,
+      "step": 188610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.951212950091924e-06,
+      "loss": 0.139,
+      "step": 188620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.948970898166001e-06,
+      "loss": 0.1458,
+      "step": 188630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.94672884624008e-06,
+      "loss": 0.1446,
+      "step": 188640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.944486794314157e-06,
+      "loss": 0.1433,
+      "step": 188650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.942244742388234e-06,
+      "loss": 0.1419,
+      "step": 188660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.940002690462311e-06,
+      "loss": 0.1413,
+      "step": 188670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.937760638536388e-06,
+      "loss": 0.145,
+      "step": 188680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.935518586610465e-06,
+      "loss": 0.1423,
+      "step": 188690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.933276534684544e-06,
+      "loss": 0.1468,
+      "step": 188700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.93103448275862e-06,
+      "loss": 0.1442,
+      "step": 188710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.9287924308327e-06,
+      "loss": 0.1462,
+      "step": 188720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.926550378906777e-06,
+      "loss": 0.1496,
+      "step": 188730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.924308326980854e-06,
+      "loss": 0.1488,
+      "step": 188740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.92206627505493e-06,
+      "loss": 0.1451,
+      "step": 188750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.919824223129008e-06,
+      "loss": 0.1438,
+      "step": 188760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.917582171203085e-06,
+      "loss": 0.1457,
+      "step": 188770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.915340119277163e-06,
+      "loss": 0.1417,
+      "step": 188780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.91309806735124e-06,
+      "loss": 0.1377,
+      "step": 188790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.910856015425317e-06,
+      "loss": 0.1456,
+      "step": 188800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.908613963499394e-06,
+      "loss": 0.1452,
+      "step": 188810
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.906371911573471e-06,
+      "loss": 0.1465,
+      "step": 188820
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.90412985964755e-06,
+      "loss": 0.1362,
+      "step": 188830
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.901887807721627e-06,
+      "loss": 0.1429,
+      "step": 188840
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.899645755795706e-06,
+      "loss": 0.1458,
+      "step": 188850
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.897403703869783e-06,
+      "loss": 0.1414,
+      "step": 188860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.89516165194386e-06,
+      "loss": 0.1455,
+      "step": 188870
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.892919600017937e-06,
+      "loss": 0.1457,
+      "step": 188880
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.890677548092014e-06,
+      "loss": 0.1485,
+      "step": 188890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.888435496166091e-06,
+      "loss": 0.1408,
+      "step": 188900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.886193444240168e-06,
+      "loss": 0.1487,
+      "step": 188910
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.883951392314247e-06,
+      "loss": 0.143,
+      "step": 188920
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.881709340388324e-06,
+      "loss": 0.145,
+      "step": 188930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.8794672884624e-06,
+      "loss": 0.1485,
+      "step": 188940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.877225236536478e-06,
+      "loss": 0.1439,
+      "step": 188950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.874983184610555e-06,
+      "loss": 0.1454,
+      "step": 188960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.872741132684634e-06,
+      "loss": 0.1462,
+      "step": 188970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.87049908075871e-06,
+      "loss": 0.1461,
+      "step": 188980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.86825702883279e-06,
+      "loss": 0.1409,
+      "step": 188990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.866014976906866e-06,
+      "loss": 0.1425,
+      "step": 189000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.863772924980943e-06,
+      "loss": 0.1456,
+      "step": 189010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.86153087305502e-06,
+      "loss": 0.1484,
+      "step": 189020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.859288821129097e-06,
+      "loss": 0.1437,
+      "step": 189030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.857046769203174e-06,
+      "loss": 0.136,
+      "step": 189040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.854804717277253e-06,
+      "loss": 0.1439,
+      "step": 189050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.85256266535133e-06,
+      "loss": 0.1456,
+      "step": 189060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.850320613425407e-06,
+      "loss": 0.1369,
+      "step": 189070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.848078561499484e-06,
+      "loss": 0.1401,
+      "step": 189080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.845836509573561e-06,
+      "loss": 0.1414,
+      "step": 189090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.84359445764764e-06,
+      "loss": 0.1457,
+      "step": 189100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.841352405721717e-06,
+      "loss": 0.1412,
+      "step": 189110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.839110353795796e-06,
+      "loss": 0.1448,
+      "step": 189120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.836868301869873e-06,
+      "loss": 0.1441,
+      "step": 189130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.83462624994395e-06,
+      "loss": 0.1402,
+      "step": 189140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.832384198018027e-06,
+      "loss": 0.1417,
+      "step": 189150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.830142146092104e-06,
+      "loss": 0.1444,
+      "step": 189160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.82790009416618e-06,
+      "loss": 0.1387,
+      "step": 189170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.825658042240258e-06,
+      "loss": 0.1425,
+      "step": 189180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.823415990314337e-06,
+      "loss": 0.1426,
+      "step": 189190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.821173938388414e-06,
+      "loss": 0.1431,
+      "step": 189200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.81893188646249e-06,
+      "loss": 0.1442,
+      "step": 189210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.816689834536568e-06,
+      "loss": 0.1415,
+      "step": 189220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.814447782610645e-06,
+      "loss": 0.1424,
+      "step": 189230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.812205730684723e-06,
+      "loss": 0.1443,
+      "step": 189240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.8099636787588e-06,
+      "loss": 0.146,
+      "step": 189250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.807721626832879e-06,
+      "loss": 0.1402,
+      "step": 189260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.805479574906956e-06,
+      "loss": 0.1399,
+      "step": 189270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.803237522981033e-06,
+      "loss": 0.1453,
+      "step": 189280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.80099547105511e-06,
+      "loss": 0.1469,
+      "step": 189290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.798753419129187e-06,
+      "loss": 0.142,
+      "step": 189300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.796511367203264e-06,
+      "loss": 0.1416,
+      "step": 189310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.794269315277341e-06,
+      "loss": 0.1475,
+      "step": 189320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.79202726335142e-06,
+      "loss": 0.1392,
+      "step": 189330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.789785211425497e-06,
+      "loss": 0.1437,
+      "step": 189340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.787543159499574e-06,
+      "loss": 0.143,
+      "step": 189350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.785301107573651e-06,
+      "loss": 0.1424,
+      "step": 189360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.78305905564773e-06,
+      "loss": 0.1482,
+      "step": 189370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.780817003721807e-06,
+      "loss": 0.1432,
+      "step": 189380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.778574951795884e-06,
+      "loss": 0.1449,
+      "step": 189390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.776332899869963e-06,
+      "loss": 0.1446,
+      "step": 189400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.77409084794404e-06,
+      "loss": 0.1433,
+      "step": 189410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.771848796018117e-06,
+      "loss": 0.143,
+      "step": 189420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.769606744092194e-06,
+      "loss": 0.1414,
+      "step": 189430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.76736469216627e-06,
+      "loss": 0.147,
+      "step": 189440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.765122640240348e-06,
+      "loss": 0.1395,
+      "step": 189450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.762880588314425e-06,
+      "loss": 0.1401,
+      "step": 189460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.760638536388503e-06,
+      "loss": 0.1434,
+      "step": 189470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.75839648446258e-06,
+      "loss": 0.145,
+      "step": 189480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.756154432536657e-06,
+      "loss": 0.1406,
+      "step": 189490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.753912380610734e-06,
+      "loss": 0.1373,
+      "step": 189500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.751670328684813e-06,
+      "loss": 0.1371,
+      "step": 189510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.74942827675889e-06,
+      "loss": 0.1456,
+      "step": 189520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.747186224832967e-06,
+      "loss": 0.1393,
+      "step": 189530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.744944172907046e-06,
+      "loss": 0.1398,
+      "step": 189540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.742702120981123e-06,
+      "loss": 0.1381,
+      "step": 189550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.7404600690552e-06,
+      "loss": 0.1433,
+      "step": 189560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.738218017129277e-06,
+      "loss": 0.1417,
+      "step": 189570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.735975965203354e-06,
+      "loss": 0.1392,
+      "step": 189580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.733733913277431e-06,
+      "loss": 0.1429,
+      "step": 189590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.731491861351508e-06,
+      "loss": 0.1427,
+      "step": 189600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.729249809425587e-06,
+      "loss": 0.1438,
+      "step": 189610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.727007757499664e-06,
+      "loss": 0.1454,
+      "step": 189620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.724765705573741e-06,
+      "loss": 0.1428,
+      "step": 189630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.72252365364782e-06,
+      "loss": 0.1416,
+      "step": 189640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.720281601721897e-06,
+      "loss": 0.144,
+      "step": 189650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.718039549795974e-06,
+      "loss": 0.1371,
+      "step": 189660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.71579749787005e-06,
+      "loss": 0.1411,
+      "step": 189670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.71355544594413e-06,
+      "loss": 0.1439,
+      "step": 189680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.711313394018206e-06,
+      "loss": 0.1426,
+      "step": 189690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.709071342092283e-06,
+      "loss": 0.1391,
+      "step": 189700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.70682929016636e-06,
+      "loss": 0.1387,
+      "step": 189710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.704587238240437e-06,
+      "loss": 0.1465,
+      "step": 189720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.702345186314514e-06,
+      "loss": 0.1438,
+      "step": 189730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.700103134388593e-06,
+      "loss": 0.1435,
+      "step": 189740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.69786108246267e-06,
+      "loss": 0.1429,
+      "step": 189750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.695619030536747e-06,
+      "loss": 0.1409,
+      "step": 189760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.693376978610824e-06,
+      "loss": 0.1441,
+      "step": 189770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.691134926684903e-06,
+      "loss": 0.1406,
+      "step": 189780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.68889287475898e-06,
+      "loss": 0.1448,
+      "step": 189790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.686650822833057e-06,
+      "loss": 0.1469,
+      "step": 189800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.684408770907136e-06,
+      "loss": 0.1442,
+      "step": 189810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.682166718981213e-06,
+      "loss": 0.1424,
+      "step": 189820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.67992466705529e-06,
+      "loss": 0.1411,
+      "step": 189830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.677682615129367e-06,
+      "loss": 0.1446,
+      "step": 189840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.675440563203444e-06,
+      "loss": 0.1422,
+      "step": 189850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.673198511277521e-06,
+      "loss": 0.1385,
+      "step": 189860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.670956459351598e-06,
+      "loss": 0.1419,
+      "step": 189870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.668714407425677e-06,
+      "loss": 0.1442,
+      "step": 189880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.666472355499754e-06,
+      "loss": 0.1419,
+      "step": 189890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.66423030357383e-06,
+      "loss": 0.1453,
+      "step": 189900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.66198825164791e-06,
+      "loss": 0.1444,
+      "step": 189910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.659746199721986e-06,
+      "loss": 0.1366,
+      "step": 189920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.657504147796063e-06,
+      "loss": 0.141,
+      "step": 189930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.65526209587014e-06,
+      "loss": 0.1412,
+      "step": 189940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.65302004394422e-06,
+      "loss": 0.1413,
+      "step": 189950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.650777992018296e-06,
+      "loss": 0.1416,
+      "step": 189960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.648535940092373e-06,
+      "loss": 0.1409,
+      "step": 189970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.64629388816645e-06,
+      "loss": 0.1404,
+      "step": 189980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.644051836240527e-06,
+      "loss": 0.1426,
+      "step": 189990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.641809784314604e-06,
+      "loss": 0.1373,
+      "step": 190000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.639567732388681e-06,
+      "loss": 0.1419,
+      "step": 190010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.63732568046276e-06,
+      "loss": 0.1394,
+      "step": 190020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.635083628536837e-06,
+      "loss": 0.1467,
+      "step": 190030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.632841576610914e-06,
+      "loss": 0.1424,
+      "step": 190040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.630599524684993e-06,
+      "loss": 0.1433,
+      "step": 190050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.628357472759069e-06,
+      "loss": 0.1389,
+      "step": 190060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.626115420833147e-06,
+      "loss": 0.1412,
+      "step": 190070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.623873368907224e-06,
+      "loss": 0.1399,
+      "step": 190080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.621631316981303e-06,
+      "loss": 0.1428,
+      "step": 190090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.61938926505538e-06,
+      "loss": 0.1395,
+      "step": 190100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.617147213129457e-06,
+      "loss": 0.1473,
+      "step": 190110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.614905161203534e-06,
+      "loss": 0.1436,
+      "step": 190120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.612663109277611e-06,
+      "loss": 0.1392,
+      "step": 190130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.6104210573516886e-06,
+      "loss": 0.1407,
+      "step": 190140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.608179005425766e-06,
+      "loss": 0.1423,
+      "step": 190150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.605936953499844e-06,
+      "loss": 0.1466,
+      "step": 190160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.603919106766513e-06,
+      "loss": 0.141,
+      "step": 190170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.601677054840591e-06,
+      "loss": 0.1424,
+      "step": 190180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.599435002914668e-06,
+      "loss": 0.1449,
+      "step": 190190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.597192950988745e-06,
+      "loss": 0.1456,
+      "step": 190200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.594950899062822e-06,
+      "loss": 0.1378,
+      "step": 190210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.592708847136901e-06,
+      "loss": 0.1477,
+      "step": 190220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.590466795210978e-06,
+      "loss": 0.1488,
+      "step": 190230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.588224743285055e-06,
+      "loss": 0.1439,
+      "step": 190240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5859826913591325e-06,
+      "loss": 0.1423,
+      "step": 190250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5837406394332095e-06,
+      "loss": 0.1424,
+      "step": 190260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5814985875072866e-06,
+      "loss": 0.1436,
+      "step": 190270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.579256535581364e-06,
+      "loss": 0.143,
+      "step": 190280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.577014483655442e-06,
+      "loss": 0.1461,
+      "step": 190290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.574772431729519e-06,
+      "loss": 0.1483,
+      "step": 190300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.572530379803597e-06,
+      "loss": 0.1408,
+      "step": 190310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.570288327877674e-06,
+      "loss": 0.141,
+      "step": 190320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.568046275951751e-06,
+      "loss": 0.1439,
+      "step": 190330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.565804224025828e-06,
+      "loss": 0.1407,
+      "step": 190340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.563562172099905e-06,
+      "loss": 0.141,
+      "step": 190350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.561320120173984e-06,
+      "loss": 0.1455,
+      "step": 190360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.559078068248061e-06,
+      "loss": 0.1405,
+      "step": 190370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.556836016322139e-06,
+      "loss": 0.1393,
+      "step": 190380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.554593964396216e-06,
+      "loss": 0.1383,
+      "step": 190390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.552351912470293e-06,
+      "loss": 0.1424,
+      "step": 190400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.55010986054437e-06,
+      "loss": 0.1392,
+      "step": 190410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.547867808618447e-06,
+      "loss": 0.1453,
+      "step": 190420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.545625756692526e-06,
+      "loss": 0.1436,
+      "step": 190430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.543383704766603e-06,
+      "loss": 0.1408,
+      "step": 190440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.541141652840681e-06,
+      "loss": 0.1399,
+      "step": 190450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.538899600914758e-06,
+      "loss": 0.1447,
+      "step": 190460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.536657548988835e-06,
+      "loss": 0.1478,
+      "step": 190470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.534415497062912e-06,
+      "loss": 0.1414,
+      "step": 190480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.532173445136989e-06,
+      "loss": 0.1405,
+      "step": 190490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5299313932110674e-06,
+      "loss": 0.142,
+      "step": 190500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5276893412851445e-06,
+      "loss": 0.1453,
+      "step": 190510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.525447289359222e-06,
+      "loss": 0.1396,
+      "step": 190520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.523205237433299e-06,
+      "loss": 0.1434,
+      "step": 190530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.520963185507376e-06,
+      "loss": 0.1453,
+      "step": 190540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.518721133581453e-06,
+      "loss": 0.1383,
+      "step": 190550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.5164790816555304e-06,
+      "loss": 0.1435,
+      "step": 190560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.514237029729609e-06,
+      "loss": 0.1457,
+      "step": 190570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.511994977803687e-06,
+      "loss": 0.1453,
+      "step": 190580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.509752925877764e-06,
+      "loss": 0.1442,
+      "step": 190590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.507510873951841e-06,
+      "loss": 0.1369,
+      "step": 190600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.505268822025918e-06,
+      "loss": 0.1483,
+      "step": 190610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.503026770099995e-06,
+      "loss": 0.1441,
+      "step": 190620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.500784718174073e-06,
+      "loss": 0.1395,
+      "step": 190630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.498542666248151e-06,
+      "loss": 0.1428,
+      "step": 190640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.496300614322229e-06,
+      "loss": 0.1388,
+      "step": 190650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.494058562396306e-06,
+      "loss": 0.1435,
+      "step": 190660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.491816510470383e-06,
+      "loss": 0.1486,
+      "step": 190670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.48957445854446e-06,
+      "loss": 0.1369,
+      "step": 190680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.487332406618537e-06,
+      "loss": 0.1462,
+      "step": 190690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4850903546926155e-06,
+      "loss": 0.1461,
+      "step": 190700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4828483027666926e-06,
+      "loss": 0.1409,
+      "step": 190710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4806062508407704e-06,
+      "loss": 0.1398,
+      "step": 190720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4783641989148474e-06,
+      "loss": 0.1475,
+      "step": 190730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4761221469889245e-06,
+      "loss": 0.1414,
+      "step": 190740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4738800950630015e-06,
+      "loss": 0.1449,
+      "step": 190750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4716380431370785e-06,
+      "loss": 0.1423,
+      "step": 190760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.469395991211157e-06,
+      "loss": 0.146,
+      "step": 190770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.467153939285234e-06,
+      "loss": 0.1441,
+      "step": 190780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.464911887359312e-06,
+      "loss": 0.1386,
+      "step": 190790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.462669835433389e-06,
+      "loss": 0.1431,
+      "step": 190800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.460427783507466e-06,
+      "loss": 0.137,
+      "step": 190810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.458185731581543e-06,
+      "loss": 0.1422,
+      "step": 190820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.45594367965562e-06,
+      "loss": 0.1443,
+      "step": 190830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.453701627729699e-06,
+      "loss": 0.139,
+      "step": 190840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.451459575803777e-06,
+      "loss": 0.1388,
+      "step": 190850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.449217523877854e-06,
+      "loss": 0.1393,
+      "step": 190860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.446975471951931e-06,
+      "loss": 0.1421,
+      "step": 190870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.444733420026008e-06,
+      "loss": 0.1396,
+      "step": 190880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.442491368100085e-06,
+      "loss": 0.1409,
+      "step": 190890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.440249316174163e-06,
+      "loss": 0.1377,
+      "step": 190900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.438007264248241e-06,
+      "loss": 0.1394,
+      "step": 190910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4357652123223185e-06,
+      "loss": 0.1376,
+      "step": 190920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4335231603963955e-06,
+      "loss": 0.1401,
+      "step": 190930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4312811084704726e-06,
+      "loss": 0.1417,
+      "step": 190940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.42903905654455e-06,
+      "loss": 0.144,
+      "step": 190950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.426797004618627e-06,
+      "loss": 0.14,
+      "step": 190960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.4245549526927045e-06,
+      "loss": 0.1433,
+      "step": 190970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.422312900766782e-06,
+      "loss": 0.1434,
+      "step": 190980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.42007084884086e-06,
+      "loss": 0.1413,
+      "step": 190990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.417828796914937e-06,
+      "loss": 0.143,
+      "step": 191000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.415586744989014e-06,
+      "loss": 0.1421,
+      "step": 191010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.413344693063091e-06,
+      "loss": 0.142,
+      "step": 191020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.411102641137168e-06,
+      "loss": 0.1383,
+      "step": 191030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.408860589211246e-06,
+      "loss": 0.1464,
+      "step": 191040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.406618537285324e-06,
+      "loss": 0.1407,
+      "step": 191050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.404376485359402e-06,
+      "loss": 0.1423,
+      "step": 191060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.402134433433479e-06,
+      "loss": 0.1394,
+      "step": 191070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.399892381507556e-06,
+      "loss": 0.1423,
+      "step": 191080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.397650329581633e-06,
+      "loss": 0.1377,
+      "step": 191090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.39540827765571e-06,
+      "loss": 0.1386,
+      "step": 191100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.393166225729788e-06,
+      "loss": 0.1401,
+      "step": 191110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.390924173803867e-06,
+      "loss": 0.1382,
+      "step": 191120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.388682121877944e-06,
+      "loss": 0.1449,
+      "step": 191130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.386440069952021e-06,
+      "loss": 0.14,
+      "step": 191140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.384198018026098e-06,
+      "loss": 0.1436,
+      "step": 191150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.381955966100175e-06,
+      "loss": 0.1446,
+      "step": 191160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.379713914174253e-06,
+      "loss": 0.1413,
+      "step": 191170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.37747186224833e-06,
+      "loss": 0.1423,
+      "step": 191180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.375229810322408e-06,
+      "loss": 0.147,
+      "step": 191190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.372987758396485e-06,
+      "loss": 0.1407,
+      "step": 191200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.370745706470562e-06,
+      "loss": 0.1471,
+      "step": 191210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.368503654544639e-06,
+      "loss": 0.1403,
+      "step": 191220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.366261602618716e-06,
+      "loss": 0.1435,
+      "step": 191230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.364019550692794e-06,
+      "loss": 0.1404,
+      "step": 191240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.361777498766871e-06,
+      "loss": 0.145,
+      "step": 191250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.35953544684095e-06,
+      "loss": 0.1473,
+      "step": 191260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.357293394915027e-06,
+      "loss": 0.1402,
+      "step": 191270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.355051342989104e-06,
+      "loss": 0.1478,
+      "step": 191280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.352809291063181e-06,
+      "loss": 0.1433,
+      "step": 191290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.350567239137258e-06,
+      "loss": 0.141,
+      "step": 191300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.348325187211336e-06,
+      "loss": 0.1424,
+      "step": 191310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.346083135285414e-06,
+      "loss": 0.1406,
+      "step": 191320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.343841083359492e-06,
+      "loss": 0.1391,
+      "step": 191330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.341599031433569e-06,
+      "loss": 0.141,
+      "step": 191340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.339356979507646e-06,
+      "loss": 0.1473,
+      "step": 191350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.337114927581723e-06,
+      "loss": 0.146,
+      "step": 191360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3348728756558e-06,
+      "loss": 0.1374,
+      "step": 191370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.332630823729878e-06,
+      "loss": 0.1416,
+      "step": 191380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3303887718039564e-06,
+      "loss": 0.1388,
+      "step": 191390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3281467198780334e-06,
+      "loss": 0.1442,
+      "step": 191400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3259046679521105e-06,
+      "loss": 0.1432,
+      "step": 191410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3236626160261875e-06,
+      "loss": 0.1444,
+      "step": 191420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.3214205641002645e-06,
+      "loss": 0.1406,
+      "step": 191430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.319178512174342e-06,
+      "loss": 0.1382,
+      "step": 191440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.316936460248419e-06,
+      "loss": 0.1464,
+      "step": 191450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.314694408322498e-06,
+      "loss": 0.1437,
+      "step": 191460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.312452356396575e-06,
+      "loss": 0.1407,
+      "step": 191470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.310210304470652e-06,
+      "loss": 0.1435,
+      "step": 191480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.307968252544729e-06,
+      "loss": 0.141,
+      "step": 191490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.305726200618806e-06,
+      "loss": 0.1415,
+      "step": 191500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.303484148692884e-06,
+      "loss": 0.1387,
+      "step": 191510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 7.301242096766961e-06,
+      "loss": 0.1439,
+      "step": 191520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.29900004484104e-06,
+      "loss": 0.1398,
+      "step": 191530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.296757992915117e-06,
+      "loss": 0.1407,
+      "step": 191540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.294515940989194e-06,
+      "loss": 0.1401,
+      "step": 191550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.292273889063271e-06,
+      "loss": 0.1447,
+      "step": 191560
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.290031837137348e-06,
+      "loss": 0.1441,
+      "step": 191570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.287789785211426e-06,
+      "loss": 0.1418,
+      "step": 191580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.285547733285503e-06,
+      "loss": 0.1451,
+      "step": 191590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.2833056813595815e-06,
+      "loss": 0.1464,
+      "step": 191600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.2810636294336586e-06,
+      "loss": 0.1394,
+      "step": 191610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.278821577507736e-06,
+      "loss": 0.1386,
+      "step": 191620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.276579525581813e-06,
+      "loss": 0.1369,
+      "step": 191630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.27433747365589e-06,
+      "loss": 0.1424,
+      "step": 191640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.2720954217299675e-06,
+      "loss": 0.1408,
+      "step": 191650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.2698533698040445e-06,
+      "loss": 0.1399,
+      "step": 191660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.267611317878123e-06,
+      "loss": 0.1438,
+      "step": 191670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.2653692659522e-06,
+      "loss": 0.1421,
+      "step": 191680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.263127214026277e-06,
+      "loss": 0.1452,
+      "step": 191690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.260885162100354e-06,
+      "loss": 0.1473,
+      "step": 191700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.258643110174432e-06,
+      "loss": 0.1422,
+      "step": 191710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.256401058248509e-06,
+      "loss": 0.1433,
+      "step": 191720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.254159006322586e-06,
+      "loss": 0.1372,
+      "step": 191730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.251916954396665e-06,
+      "loss": 0.1469,
+      "step": 191740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.249674902470742e-06,
+      "loss": 0.1405,
+      "step": 191750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.247432850544819e-06,
+      "loss": 0.1467,
+      "step": 191760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.245190798618896e-06,
+      "loss": 0.1387,
+      "step": 191770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.242948746692974e-06,
+      "loss": 0.1358,
+      "step": 191780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.240706694767051e-06,
+      "loss": 0.1401,
+      "step": 191790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.238464642841128e-06,
+      "loss": 0.1413,
+      "step": 191800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.236222590915207e-06,
+      "loss": 0.1412,
+      "step": 191810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.233980538989284e-06,
+      "loss": 0.1439,
+      "step": 191820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.231738487063361e-06,
+      "loss": 0.1442,
+      "step": 191830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.229496435137438e-06,
+      "loss": 0.1432,
+      "step": 191840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.227254383211516e-06,
+      "loss": 0.1412,
+      "step": 191850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.225012331285593e-06,
+      "loss": 0.1411,
+      "step": 191860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.22277027935967e-06,
+      "loss": 0.1396,
+      "step": 191870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.220528227433748e-06,
+      "loss": 0.1356,
+      "step": 191880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.218286175507825e-06,
+      "loss": 0.1449,
+      "step": 191890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.216044123581902e-06,
+      "loss": 0.1427,
+      "step": 191900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.2138020716559794e-06,
+      "loss": 0.1424,
+      "step": 191910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.211560019730057e-06,
+      "loss": 0.1407,
+      "step": 191920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.209317967804134e-06,
+      "loss": 0.1415,
+      "step": 191930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.207075915878211e-06,
+      "loss": 0.1427,
+      "step": 191940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.20483386395229e-06,
+      "loss": 0.1391,
+      "step": 191950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.202591812026367e-06,
+      "loss": 0.1429,
+      "step": 191960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.200349760100444e-06,
+      "loss": 0.135,
+      "step": 191970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.198107708174521e-06,
+      "loss": 0.1435,
+      "step": 191980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.195865656248599e-06,
+      "loss": 0.1426,
+      "step": 191990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.193623604322676e-06,
+      "loss": 0.1439,
+      "step": 192000
+    },
+    {
+      "epoch": 0.86,
+      "eval_NEIMS_canon_smiles": 0.9442666666666667,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.10665,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.508517053716457,
+      "eval_NEIMS_exact_mols": 0.10645,
+      "eval_NEIMS_exact_smiles": 0.10338333333333333,
+      "eval_NEIMS_loss": 0.18446384370326996,
+      "eval_NEIMS_matched_formulas": 0.6333,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.4186830011326453,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.10686666666666667,
+      "eval_NEIMS_runtime": 715.0375,
+      "eval_NEIMS_samples_per_second": 83.912,
+      "eval_NEIMS_steps_per_second": 1.312,
+      "step": 192000
+    },
+    {
+      "epoch": 0.86,
+      "eval_RASSP_canon_smiles": 0.9397268507696931,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.23414699302100378,
+      "eval_RASSP_daylight_tanimoto_simil": 0.6567451166528004,
+      "eval_RASSP_exact_mols": 0.23341236183924935,
+      "eval_RASSP_exact_smiles": 0.22701773132534145,
+      "eval_RASSP_loss": 0.12375875562429428,
+      "eval_RASSP_matched_formulas": 0.8093131198450596,
+      "eval_RASSP_morgan_tanimoto_simil": 0.5400474171561165,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.2336961966140181,
+      "eval_RASSP_runtime": 816.4158,
+      "eval_RASSP_samples_per_second": 73.362,
+      "eval_RASSP_steps_per_second": 1.146,
+      "step": 192000
+    },
+    {
+      "epoch": 0.86,
+      "eval_NIST_canon_smiles": 0.8672321396883983,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.016786740958938142,
+      "eval_NIST_daylight_tanimoto_simil": 0.279486469735125,
+      "eval_NIST_exact_mols": 0.015934982432480392,
+      "eval_NIST_exact_smiles": 0.014799304397203393,
+      "eval_NIST_loss": 1.8838388919830322,
+      "eval_NIST_matched_formulas": 0.10249494268374916,
+      "eval_NIST_morgan_tanimoto_simil": 0.23107522621292642,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.016680271143130924,
+      "eval_NIST_runtime": 492.9095,
+      "eval_NIST_samples_per_second": 57.165,
+      "eval_NIST_steps_per_second": 0.895,
+      "step": 192000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.191381552396755e-06,
+      "loss": 0.1417,
+      "step": 192010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.189139500470832e-06,
+      "loss": 0.1425,
+      "step": 192020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.186897448544909e-06,
+      "loss": 0.1431,
+      "step": 192030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.184655396618986e-06,
+      "loss": 0.1396,
+      "step": 192040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.182413344693064e-06,
+      "loss": 0.1441,
+      "step": 192050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.180171292767141e-06,
+      "loss": 0.1385,
+      "step": 192060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.177929240841218e-06,
+      "loss": 0.1424,
+      "step": 192070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1756871889152965e-06,
+      "loss": 0.1426,
+      "step": 192080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1734451369893735e-06,
+      "loss": 0.1449,
+      "step": 192090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1712030850634505e-06,
+      "loss": 0.1369,
+      "step": 192100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1689610331375275e-06,
+      "loss": 0.1428,
+      "step": 192110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.166718981211605e-06,
+      "loss": 0.1427,
+      "step": 192120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1644769292856824e-06,
+      "loss": 0.143,
+      "step": 192130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1622348773597595e-06,
+      "loss": 0.1421,
+      "step": 192140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.159992825433838e-06,
+      "loss": 0.1422,
+      "step": 192150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.157750773507915e-06,
+      "loss": 0.1451,
+      "step": 192160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.155732926774585e-06,
+      "loss": 0.1407,
+      "step": 192170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.153490874848662e-06,
+      "loss": 0.1379,
+      "step": 192180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.151248822922739e-06,
+      "loss": 0.146,
+      "step": 192190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.149006770996816e-06,
+      "loss": 0.1454,
+      "step": 192200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.146764719070894e-06,
+      "loss": 0.1399,
+      "step": 192210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.144522667144972e-06,
+      "loss": 0.1414,
+      "step": 192220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.142280615219049e-06,
+      "loss": 0.1423,
+      "step": 192230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.140038563293126e-06,
+      "loss": 0.1459,
+      "step": 192240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.137796511367203e-06,
+      "loss": 0.1385,
+      "step": 192250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1355544594412804e-06,
+      "loss": 0.1412,
+      "step": 192260
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.133312407515358e-06,
+      "loss": 0.1477,
+      "step": 192270
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.131070355589436e-06,
+      "loss": 0.1436,
+      "step": 192280
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.128828303663514e-06,
+      "loss": 0.1422,
+      "step": 192290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.126586251737591e-06,
+      "loss": 0.14,
+      "step": 192300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.124344199811668e-06,
+      "loss": 0.1433,
+      "step": 192310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.122102147885745e-06,
+      "loss": 0.1397,
+      "step": 192320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.119860095959822e-06,
+      "loss": 0.1382,
+      "step": 192330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.1176180440339e-06,
+      "loss": 0.1408,
+      "step": 192340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.115375992107978e-06,
+      "loss": 0.1458,
+      "step": 192350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.113133940182056e-06,
+      "loss": 0.1402,
+      "step": 192360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.110891888256133e-06,
+      "loss": 0.1411,
+      "step": 192370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.10864983633021e-06,
+      "loss": 0.1448,
+      "step": 192380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.106407784404287e-06,
+      "loss": 0.1437,
+      "step": 192390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.104165732478364e-06,
+      "loss": 0.139,
+      "step": 192400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.101923680552442e-06,
+      "loss": 0.1381,
+      "step": 192410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.09968162862652e-06,
+      "loss": 0.1377,
+      "step": 192420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0974395767005975e-06,
+      "loss": 0.1394,
+      "step": 192430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0951975247746745e-06,
+      "loss": 0.1394,
+      "step": 192440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0929554728487515e-06,
+      "loss": 0.1489,
+      "step": 192450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0907134209228285e-06,
+      "loss": 0.136,
+      "step": 192460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0884713689969056e-06,
+      "loss": 0.1388,
+      "step": 192470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0862293170709834e-06,
+      "loss": 0.1429,
+      "step": 192480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.083987265145061e-06,
+      "loss": 0.1471,
+      "step": 192490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.081745213219139e-06,
+      "loss": 0.1448,
+      "step": 192500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.079503161293216e-06,
+      "loss": 0.1471,
+      "step": 192510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.077261109367293e-06,
+      "loss": 0.147,
+      "step": 192520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.07501905744137e-06,
+      "loss": 0.1462,
+      "step": 192530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.072777005515448e-06,
+      "loss": 0.1397,
+      "step": 192540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.070534953589525e-06,
+      "loss": 0.1435,
+      "step": 192550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.068292901663604e-06,
+      "loss": 0.1411,
+      "step": 192560
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.066050849737681e-06,
+      "loss": 0.1363,
+      "step": 192570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.063808797811758e-06,
+      "loss": 0.1406,
+      "step": 192580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.061566745885835e-06,
+      "loss": 0.142,
+      "step": 192590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.059324693959912e-06,
+      "loss": 0.1428,
+      "step": 192600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.05708264203399e-06,
+      "loss": 0.1409,
+      "step": 192610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.054840590108067e-06,
+      "loss": 0.1383,
+      "step": 192620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0525985381821456e-06,
+      "loss": 0.1414,
+      "step": 192630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.050356486256223e-06,
+      "loss": 0.1428,
+      "step": 192640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0481144343303e-06,
+      "loss": 0.1409,
+      "step": 192650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.045872382404377e-06,
+      "loss": 0.1416,
+      "step": 192660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.043630330478454e-06,
+      "loss": 0.147,
+      "step": 192670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0413882785525315e-06,
+      "loss": 0.1369,
+      "step": 192680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.0391462266266086e-06,
+      "loss": 0.1382,
+      "step": 192690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.036904174700687e-06,
+      "loss": 0.1429,
+      "step": 192700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.034662122774764e-06,
+      "loss": 0.1395,
+      "step": 192710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.032420070848841e-06,
+      "loss": 0.1381,
+      "step": 192720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.030178018922918e-06,
+      "loss": 0.1405,
+      "step": 192730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.027935966996995e-06,
+      "loss": 0.143,
+      "step": 192740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.025693915071073e-06,
+      "loss": 0.1431,
+      "step": 192750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.02345186314515e-06,
+      "loss": 0.135,
+      "step": 192760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.021209811219229e-06,
+      "loss": 0.1417,
+      "step": 192770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.018967759293306e-06,
+      "loss": 0.1403,
+      "step": 192780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.016725707367383e-06,
+      "loss": 0.1435,
+      "step": 192790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.01448365544146e-06,
+      "loss": 0.1441,
+      "step": 192800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.012241603515537e-06,
+      "loss": 0.134,
+      "step": 192810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.009999551589615e-06,
+      "loss": 0.1379,
+      "step": 192820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.007757499663692e-06,
+      "loss": 0.1422,
+      "step": 192830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.005515447737771e-06,
+      "loss": 0.1364,
+      "step": 192840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.003273395811848e-06,
+      "loss": 0.1431,
+      "step": 192850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 7.001031343885925e-06,
+      "loss": 0.1422,
+      "step": 192860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.998789291960002e-06,
+      "loss": 0.1343,
+      "step": 192870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.99654724003408e-06,
+      "loss": 0.1426,
+      "step": 192880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.994305188108157e-06,
+      "loss": 0.1436,
+      "step": 192890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.992063136182234e-06,
+      "loss": 0.1437,
+      "step": 192900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.989821084256312e-06,
+      "loss": 0.1463,
+      "step": 192910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.987579032330389e-06,
+      "loss": 0.1392,
+      "step": 192920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9853369804044664e-06,
+      "loss": 0.1449,
+      "step": 192930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9830949284785435e-06,
+      "loss": 0.1409,
+      "step": 192940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.980852876552621e-06,
+      "loss": 0.1387,
+      "step": 192950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.978610824626698e-06,
+      "loss": 0.1452,
+      "step": 192960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.976368772700777e-06,
+      "loss": 0.1443,
+      "step": 192970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.974126720774854e-06,
+      "loss": 0.1406,
+      "step": 192980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.971884668848931e-06,
+      "loss": 0.1345,
+      "step": 192990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.969642616923008e-06,
+      "loss": 0.1415,
+      "step": 193000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.967400564997085e-06,
+      "loss": 0.1474,
+      "step": 193010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.965158513071163e-06,
+      "loss": 0.1429,
+      "step": 193020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.96291646114524e-06,
+      "loss": 0.1417,
+      "step": 193030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.960674409219319e-06,
+      "loss": 0.1379,
+      "step": 193040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.958432357293396e-06,
+      "loss": 0.1424,
+      "step": 193050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.956190305367473e-06,
+      "loss": 0.1415,
+      "step": 193060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.95394825344155e-06,
+      "loss": 0.1475,
+      "step": 193070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.951706201515627e-06,
+      "loss": 0.1407,
+      "step": 193080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.949464149589705e-06,
+      "loss": 0.1435,
+      "step": 193090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.947222097663782e-06,
+      "loss": 0.1449,
+      "step": 193100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9449800457378605e-06,
+      "loss": 0.1434,
+      "step": 193110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9427379938119375e-06,
+      "loss": 0.1438,
+      "step": 193120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9404959418860145e-06,
+      "loss": 0.1436,
+      "step": 193130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9382538899600916e-06,
+      "loss": 0.1392,
+      "step": 193140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9360118380341694e-06,
+      "loss": 0.1418,
+      "step": 193150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9337697861082465e-06,
+      "loss": 0.1454,
+      "step": 193160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.9315277341823235e-06,
+      "loss": 0.1467,
+      "step": 193170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.929285682256402e-06,
+      "loss": 0.1416,
+      "step": 193180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.927043630330479e-06,
+      "loss": 0.1383,
+      "step": 193190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.924801578404556e-06,
+      "loss": 0.1398,
+      "step": 193200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.922559526478633e-06,
+      "loss": 0.1471,
+      "step": 193210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.920317474552711e-06,
+      "loss": 0.14,
+      "step": 193220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.918075422626788e-06,
+      "loss": 0.1427,
+      "step": 193230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.915833370700865e-06,
+      "loss": 0.1397,
+      "step": 193240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.913591318774944e-06,
+      "loss": 0.1404,
+      "step": 193250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.911349266849021e-06,
+      "loss": 0.14,
+      "step": 193260
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.909107214923098e-06,
+      "loss": 0.144,
+      "step": 193270
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.906865162997175e-06,
+      "loss": 0.1388,
+      "step": 193280
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.904623111071253e-06,
+      "loss": 0.141,
+      "step": 193290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.90238105914533e-06,
+      "loss": 0.1431,
+      "step": 193300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.900139007219407e-06,
+      "loss": 0.1416,
+      "step": 193310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.897896955293486e-06,
+      "loss": 0.1441,
+      "step": 193320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.895654903367563e-06,
+      "loss": 0.1387,
+      "step": 193330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.89341285144164e-06,
+      "loss": 0.1369,
+      "step": 193340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.891170799515717e-06,
+      "loss": 0.1379,
+      "step": 193350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.8889287475897946e-06,
+      "loss": 0.1391,
+      "step": 193360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.886686695663872e-06,
+      "loss": 0.1418,
+      "step": 193370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.884444643737949e-06,
+      "loss": 0.1413,
+      "step": 193380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.882202591812027e-06,
+      "loss": 0.1476,
+      "step": 193390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.879960539886104e-06,
+      "loss": 0.1402,
+      "step": 193400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.877718487960181e-06,
+      "loss": 0.1395,
+      "step": 193410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.875476436034259e-06,
+      "loss": 0.1396,
+      "step": 193420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.873234384108336e-06,
+      "loss": 0.1383,
+      "step": 193430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.870992332182413e-06,
+      "loss": 0.1446,
+      "step": 193440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.86875028025649e-06,
+      "loss": 0.1423,
+      "step": 193450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.866508228330569e-06,
+      "loss": 0.1411,
+      "step": 193460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.864266176404646e-06,
+      "loss": 0.1382,
+      "step": 193470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.862024124478723e-06,
+      "loss": 0.1419,
+      "step": 193480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.859782072552801e-06,
+      "loss": 0.1409,
+      "step": 193490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.857540020626878e-06,
+      "loss": 0.1418,
+      "step": 193500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.855297968700955e-06,
+      "loss": 0.1467,
+      "step": 193510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.853055916775032e-06,
+      "loss": 0.148,
+      "step": 193520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.850813864849111e-06,
+      "loss": 0.1431,
+      "step": 193530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.848571812923188e-06,
+      "loss": 0.1363,
+      "step": 193540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.846329760997265e-06,
+      "loss": 0.1408,
+      "step": 193550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.844087709071343e-06,
+      "loss": 0.1378,
+      "step": 193560
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.84184565714542e-06,
+      "loss": 0.1371,
+      "step": 193570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.839603605219497e-06,
+      "loss": 0.1418,
+      "step": 193580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.837361553293574e-06,
+      "loss": 0.1419,
+      "step": 193590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.8351195013676524e-06,
+      "loss": 0.1468,
+      "step": 193600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.8328774494417295e-06,
+      "loss": 0.1363,
+      "step": 193610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.8306353975158065e-06,
+      "loss": 0.1371,
+      "step": 193620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.828393345589884e-06,
+      "loss": 0.1469,
+      "step": 193630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.826151293663961e-06,
+      "loss": 0.138,
+      "step": 193640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.823909241738038e-06,
+      "loss": 0.1417,
+      "step": 193650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.821667189812117e-06,
+      "loss": 0.1418,
+      "step": 193660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.819425137886194e-06,
+      "loss": 0.1393,
+      "step": 193670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.817183085960271e-06,
+      "loss": 0.1392,
+      "step": 193680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.814941034034349e-06,
+      "loss": 0.1404,
+      "step": 193690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.812698982108426e-06,
+      "loss": 0.138,
+      "step": 193700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.810456930182503e-06,
+      "loss": 0.1414,
+      "step": 193710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.80821487825658e-06,
+      "loss": 0.1467,
+      "step": 193720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.805972826330659e-06,
+      "loss": 0.1429,
+      "step": 193730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.803730774404736e-06,
+      "loss": 0.1438,
+      "step": 193740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.801488722478813e-06,
+      "loss": 0.1407,
+      "step": 193750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.799246670552891e-06,
+      "loss": 0.1398,
+      "step": 193760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.797004618626968e-06,
+      "loss": 0.1366,
+      "step": 193770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.794762566701045e-06,
+      "loss": 0.1433,
+      "step": 193780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.792520514775122e-06,
+      "loss": 0.1466,
+      "step": 193790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7902784628492005e-06,
+      "loss": 0.1403,
+      "step": 193800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7880364109232776e-06,
+      "loss": 0.1391,
+      "step": 193810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.785794358997355e-06,
+      "loss": 0.1402,
+      "step": 193820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7835523070714325e-06,
+      "loss": 0.1416,
+      "step": 193830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7813102551455095e-06,
+      "loss": 0.1396,
+      "step": 193840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7790682032195865e-06,
+      "loss": 0.1406,
+      "step": 193850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7768261512936635e-06,
+      "loss": 0.1421,
+      "step": 193860
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.774584099367742e-06,
+      "loss": 0.1425,
+      "step": 193870
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.772342047441819e-06,
+      "loss": 0.1462,
+      "step": 193880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.770099995515896e-06,
+      "loss": 0.1386,
+      "step": 193890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.767857943589974e-06,
+      "loss": 0.1415,
+      "step": 193900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.765615891664051e-06,
+      "loss": 0.1403,
+      "step": 193910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.763373839738128e-06,
+      "loss": 0.1409,
+      "step": 193920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.761131787812205e-06,
+      "loss": 0.1428,
+      "step": 193930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.758889735886284e-06,
+      "loss": 0.1382,
+      "step": 193940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.756647683960361e-06,
+      "loss": 0.1413,
+      "step": 193950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.754405632034439e-06,
+      "loss": 0.143,
+      "step": 193960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.752163580108516e-06,
+      "loss": 0.1411,
+      "step": 193970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.749921528182593e-06,
+      "loss": 0.1421,
+      "step": 193980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.74767947625667e-06,
+      "loss": 0.1426,
+      "step": 193990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.745437424330747e-06,
+      "loss": 0.1431,
+      "step": 194000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.743195372404826e-06,
+      "loss": 0.1415,
+      "step": 194010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.740953320478903e-06,
+      "loss": 0.1456,
+      "step": 194020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7387112685529806e-06,
+      "loss": 0.1378,
+      "step": 194030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7364692166270576e-06,
+      "loss": 0.142,
+      "step": 194040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.734227164701135e-06,
+      "loss": 0.137,
+      "step": 194050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.731985112775212e-06,
+      "loss": 0.1386,
+      "step": 194060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.729743060849289e-06,
+      "loss": 0.142,
+      "step": 194070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.727501008923367e-06,
+      "loss": 0.1403,
+      "step": 194080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.725258956997444e-06,
+      "loss": 0.1395,
+      "step": 194090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.723016905071522e-06,
+      "loss": 0.1407,
+      "step": 194100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.720774853145599e-06,
+      "loss": 0.1399,
+      "step": 194110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.718532801219676e-06,
+      "loss": 0.1439,
+      "step": 194120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.716290749293753e-06,
+      "loss": 0.1392,
+      "step": 194130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.71404869736783e-06,
+      "loss": 0.1449,
+      "step": 194140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.711806645441909e-06,
+      "loss": 0.1436,
+      "step": 194150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.709564593515986e-06,
+      "loss": 0.1383,
+      "step": 194160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.707322541590064e-06,
+      "loss": 0.1361,
+      "step": 194170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.705080489664141e-06,
+      "loss": 0.1445,
+      "step": 194180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7030626429308105e-06,
+      "loss": 0.1409,
+      "step": 194190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.7008205910048875e-06,
+      "loss": 0.1397,
+      "step": 194200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.698578539078966e-06,
+      "loss": 0.1442,
+      "step": 194210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.696336487153043e-06,
+      "loss": 0.1429,
+      "step": 194220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.69409443522712e-06,
+      "loss": 0.1452,
+      "step": 194230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.691852383301197e-06,
+      "loss": 0.137,
+      "step": 194240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.689610331375275e-06,
+      "loss": 0.145,
+      "step": 194250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.687368279449352e-06,
+      "loss": 0.1371,
+      "step": 194260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.685126227523429e-06,
+      "loss": 0.1432,
+      "step": 194270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.682884175597508e-06,
+      "loss": 0.1449,
+      "step": 194280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.680642123671585e-06,
+      "loss": 0.1406,
+      "step": 194290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.678400071745662e-06,
+      "loss": 0.1409,
+      "step": 194300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.676158019819739e-06,
+      "loss": 0.1411,
+      "step": 194310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.673915967893817e-06,
+      "loss": 0.1382,
+      "step": 194320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.671673915967894e-06,
+      "loss": 0.1384,
+      "step": 194330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.669431864041971e-06,
+      "loss": 0.1398,
+      "step": 194340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.66718981211605e-06,
+      "loss": 0.1434,
+      "step": 194350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.664947760190127e-06,
+      "loss": 0.1383,
+      "step": 194360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.662705708264204e-06,
+      "loss": 0.1386,
+      "step": 194370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.660463656338281e-06,
+      "loss": 0.1391,
+      "step": 194380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.658221604412359e-06,
+      "loss": 0.1391,
+      "step": 194390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.655979552486436e-06,
+      "loss": 0.139,
+      "step": 194400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.653737500560513e-06,
+      "loss": 0.1388,
+      "step": 194410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.651495448634591e-06,
+      "loss": 0.1388,
+      "step": 194420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.649253396708668e-06,
+      "loss": 0.1414,
+      "step": 194430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.647011344782745e-06,
+      "loss": 0.1455,
+      "step": 194440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.644769292856822e-06,
+      "loss": 0.1408,
+      "step": 194450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.6425272409309e-06,
+      "loss": 0.1483,
+      "step": 194460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.640285189004977e-06,
+      "loss": 0.1426,
+      "step": 194470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.638043137079054e-06,
+      "loss": 0.1418,
+      "step": 194480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.635801085153133e-06,
+      "loss": 0.1355,
+      "step": 194490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.63355903322721e-06,
+      "loss": 0.1398,
+      "step": 194500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.631316981301287e-06,
+      "loss": 0.1421,
+      "step": 194510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.629074929375365e-06,
+      "loss": 0.1403,
+      "step": 194520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.626832877449442e-06,
+      "loss": 0.1383,
+      "step": 194530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.624590825523519e-06,
+      "loss": 0.1392,
+      "step": 194540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.622348773597596e-06,
+      "loss": 0.1397,
+      "step": 194550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.620106721671675e-06,
+      "loss": 0.1382,
+      "step": 194560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.617864669745752e-06,
+      "loss": 0.1401,
+      "step": 194570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.615622617819829e-06,
+      "loss": 0.1428,
+      "step": 194580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.613380565893907e-06,
+      "loss": 0.1382,
+      "step": 194590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.611138513967984e-06,
+      "loss": 0.1384,
+      "step": 194600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.608896462042061e-06,
+      "loss": 0.1361,
+      "step": 194610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.6066544101161394e-06,
+      "loss": 0.1397,
+      "step": 194620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.6044123581902165e-06,
+      "loss": 0.1435,
+      "step": 194630
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.6021703062642935e-06,
+      "loss": 0.141,
+      "step": 194640
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5999282543383705e-06,
+      "loss": 0.1461,
+      "step": 194650
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.597686202412448e-06,
+      "loss": 0.1411,
+      "step": 194660
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.595444150486525e-06,
+      "loss": 0.1443,
+      "step": 194670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5932020985606024e-06,
+      "loss": 0.1432,
+      "step": 194680
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.590960046634681e-06,
+      "loss": 0.1367,
+      "step": 194690
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.588717994708758e-06,
+      "loss": 0.1408,
+      "step": 194700
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.586475942782835e-06,
+      "loss": 0.1399,
+      "step": 194710
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.584233890856912e-06,
+      "loss": 0.14,
+      "step": 194720
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.58199183893099e-06,
+      "loss": 0.1419,
+      "step": 194730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.579749787005067e-06,
+      "loss": 0.143,
+      "step": 194740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.577507735079144e-06,
+      "loss": 0.1448,
+      "step": 194750
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.575265683153223e-06,
+      "loss": 0.1423,
+      "step": 194760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5730236312273e-06,
+      "loss": 0.1384,
+      "step": 194770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.570781579301377e-06,
+      "loss": 0.1351,
+      "step": 194780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.568539527375455e-06,
+      "loss": 0.1428,
+      "step": 194790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.566297475449532e-06,
+      "loss": 0.1351,
+      "step": 194800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.564055423523609e-06,
+      "loss": 0.1368,
+      "step": 194810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.561813371597686e-06,
+      "loss": 0.1423,
+      "step": 194820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5595713196717646e-06,
+      "loss": 0.139,
+      "step": 194830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.557329267745842e-06,
+      "loss": 0.1421,
+      "step": 194840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.555087215819919e-06,
+      "loss": 0.142,
+      "step": 194850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5528451638939965e-06,
+      "loss": 0.1422,
+      "step": 194860
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5506031119680735e-06,
+      "loss": 0.1407,
+      "step": 194870
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5483610600421505e-06,
+      "loss": 0.1378,
+      "step": 194880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5461190081162275e-06,
+      "loss": 0.1353,
+      "step": 194890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.543876956190306e-06,
+      "loss": 0.1398,
+      "step": 194900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.541634904264383e-06,
+      "loss": 0.1404,
+      "step": 194910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.53939285233846e-06,
+      "loss": 0.1425,
+      "step": 194920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.537150800412538e-06,
+      "loss": 0.1379,
+      "step": 194930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.534908748486615e-06,
+      "loss": 0.1408,
+      "step": 194940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.532666696560692e-06,
+      "loss": 0.1408,
+      "step": 194950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.530424644634769e-06,
+      "loss": 0.1437,
+      "step": 194960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.528182592708848e-06,
+      "loss": 0.1405,
+      "step": 194970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.525940540782925e-06,
+      "loss": 0.1358,
+      "step": 194980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.523698488857002e-06,
+      "loss": 0.1408,
+      "step": 194990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.52145643693108e-06,
+      "loss": 0.1366,
+      "step": 195000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.519214385005157e-06,
+      "loss": 0.1401,
+      "step": 195010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.516972333079234e-06,
+      "loss": 0.137,
+      "step": 195020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.514730281153311e-06,
+      "loss": 0.1405,
+      "step": 195030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.51248822922739e-06,
+      "loss": 0.1412,
+      "step": 195040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.510246177301467e-06,
+      "loss": 0.1401,
+      "step": 195050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.5080041253755446e-06,
+      "loss": 0.1389,
+      "step": 195060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.505762073449622e-06,
+      "loss": 0.1401,
+      "step": 195070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.503520021523699e-06,
+      "loss": 0.1413,
+      "step": 195080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.501277969597776e-06,
+      "loss": 0.1421,
+      "step": 195090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.499035917671853e-06,
+      "loss": 0.1392,
+      "step": 195100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.496793865745931e-06,
+      "loss": 0.1391,
+      "step": 195110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.494551813820008e-06,
+      "loss": 0.1458,
+      "step": 195120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.492309761894086e-06,
+      "loss": 0.1358,
+      "step": 195130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.490067709968163e-06,
+      "loss": 0.1402,
+      "step": 195140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.48782565804224e-06,
+      "loss": 0.1398,
+      "step": 195150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.485583606116317e-06,
+      "loss": 0.1414,
+      "step": 195160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.483341554190394e-06,
+      "loss": 0.1392,
+      "step": 195170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.481099502264473e-06,
+      "loss": 0.1415,
+      "step": 195180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.47885745033855e-06,
+      "loss": 0.1372,
+      "step": 195190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.476615398412628e-06,
+      "loss": 0.1371,
+      "step": 195200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.474373346486705e-06,
+      "loss": 0.1366,
+      "step": 195210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.472131294560782e-06,
+      "loss": 0.1415,
+      "step": 195220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.469889242634859e-06,
+      "loss": 0.1398,
+      "step": 195230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.467647190708938e-06,
+      "loss": 0.1379,
+      "step": 195240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.465405138783015e-06,
+      "loss": 0.1373,
+      "step": 195250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.463163086857092e-06,
+      "loss": 0.1419,
+      "step": 195260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.46092103493117e-06,
+      "loss": 0.1428,
+      "step": 195270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.458678983005247e-06,
+      "loss": 0.1354,
+      "step": 195280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.456436931079324e-06,
+      "loss": 0.1394,
+      "step": 195290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.454194879153401e-06,
+      "loss": 0.1394,
+      "step": 195300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4519528272274795e-06,
+      "loss": 0.1412,
+      "step": 195310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4497107753015565e-06,
+      "loss": 0.1371,
+      "step": 195320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.447468723375634e-06,
+      "loss": 0.1397,
+      "step": 195330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.445226671449711e-06,
+      "loss": 0.1446,
+      "step": 195340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4429846195237884e-06,
+      "loss": 0.1403,
+      "step": 195350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4407425675978654e-06,
+      "loss": 0.1347,
+      "step": 195360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4385005156719425e-06,
+      "loss": 0.1401,
+      "step": 195370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.436258463746021e-06,
+      "loss": 0.1346,
+      "step": 195380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.434016411820098e-06,
+      "loss": 0.1395,
+      "step": 195390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.431774359894176e-06,
+      "loss": 0.1418,
+      "step": 195400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.429532307968253e-06,
+      "loss": 0.1381,
+      "step": 195410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.42729025604233e-06,
+      "loss": 0.1364,
+      "step": 195420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.425048204116407e-06,
+      "loss": 0.1349,
+      "step": 195430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.422806152190484e-06,
+      "loss": 0.1404,
+      "step": 195440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.420564100264563e-06,
+      "loss": 0.1361,
+      "step": 195450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.41832204833864e-06,
+      "loss": 0.1406,
+      "step": 195460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.416079996412718e-06,
+      "loss": 0.1413,
+      "step": 195470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.413837944486795e-06,
+      "loss": 0.1391,
+      "step": 195480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.411595892560872e-06,
+      "loss": 0.1399,
+      "step": 195490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.409353840634949e-06,
+      "loss": 0.1334,
+      "step": 195500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.407111788709026e-06,
+      "loss": 0.1385,
+      "step": 195510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.404869736783105e-06,
+      "loss": 0.1426,
+      "step": 195520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.402627684857182e-06,
+      "loss": 0.1343,
+      "step": 195530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.4003856329312595e-06,
+      "loss": 0.1409,
+      "step": 195540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3981435810053365e-06,
+      "loss": 0.1407,
+      "step": 195550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3959015290794135e-06,
+      "loss": 0.1325,
+      "step": 195560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3936594771534906e-06,
+      "loss": 0.1383,
+      "step": 195570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.391417425227568e-06,
+      "loss": 0.1398,
+      "step": 195580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.389175373301646e-06,
+      "loss": 0.1406,
+      "step": 195590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.386933321375723e-06,
+      "loss": 0.1382,
+      "step": 195600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.384691269449801e-06,
+      "loss": 0.1403,
+      "step": 195610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.382449217523878e-06,
+      "loss": 0.1406,
+      "step": 195620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.380207165597955e-06,
+      "loss": 0.137,
+      "step": 195630
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.377965113672032e-06,
+      "loss": 0.143,
+      "step": 195640
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.37572306174611e-06,
+      "loss": 0.1385,
+      "step": 195650
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.373481009820188e-06,
+      "loss": 0.1366,
+      "step": 195660
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.371238957894266e-06,
+      "loss": 0.1375,
+      "step": 195670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.368996905968343e-06,
+      "loss": 0.1405,
+      "step": 195680
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.36675485404242e-06,
+      "loss": 0.1399,
+      "step": 195690
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.364512802116497e-06,
+      "loss": 0.138,
+      "step": 195700
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.362270750190574e-06,
+      "loss": 0.1451,
+      "step": 195710
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.360028698264652e-06,
+      "loss": 0.1427,
+      "step": 195720
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.35778664633873e-06,
+      "loss": 0.1351,
+      "step": 195730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.355544594412808e-06,
+      "loss": 0.1402,
+      "step": 195740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.353302542486885e-06,
+      "loss": 0.1431,
+      "step": 195750
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.351060490560962e-06,
+      "loss": 0.1448,
+      "step": 195760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.348818438635039e-06,
+      "loss": 0.1467,
+      "step": 195770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.346576386709116e-06,
+      "loss": 0.1447,
+      "step": 195780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3443343347831936e-06,
+      "loss": 0.1457,
+      "step": 195790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.3420922828572714e-06,
+      "loss": 0.1432,
+      "step": 195800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.339850230931349e-06,
+      "loss": 0.135,
+      "step": 195810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.337608179005426e-06,
+      "loss": 0.1409,
+      "step": 195820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.335366127079503e-06,
+      "loss": 0.1407,
+      "step": 195830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.33312407515358e-06,
+      "loss": 0.1434,
+      "step": 195840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.330882023227657e-06,
+      "loss": 0.1366,
+      "step": 195850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.328639971301735e-06,
+      "loss": 0.1358,
+      "step": 195860
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.326397919375813e-06,
+      "loss": 0.1372,
+      "step": 195870
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.324155867449891e-06,
+      "loss": 0.1414,
+      "step": 195880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.321913815523968e-06,
+      "loss": 0.1395,
+      "step": 195890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.319671763598045e-06,
+      "loss": 0.1422,
+      "step": 195900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.317429711672122e-06,
+      "loss": 0.1427,
+      "step": 195910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.315187659746199e-06,
+      "loss": 0.1423,
+      "step": 195920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.312945607820278e-06,
+      "loss": 0.1428,
+      "step": 195930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.310703555894356e-06,
+      "loss": 0.1365,
+      "step": 195940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.308461503968433e-06,
+      "loss": 0.1394,
+      "step": 195950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.30621945204251e-06,
+      "loss": 0.1445,
+      "step": 195960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.303977400116587e-06,
+      "loss": 0.1421,
+      "step": 195970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.301735348190664e-06,
+      "loss": 0.1393,
+      "step": 195980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.299493296264742e-06,
+      "loss": 0.1405,
+      "step": 195990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.2972512443388195e-06,
+      "loss": 0.1395,
+      "step": 196000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.295009192412897e-06,
+      "loss": 0.1354,
+      "step": 196010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2927671404869744e-06,
+      "loss": 0.1379,
+      "step": 196020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2905250885610514e-06,
+      "loss": 0.1409,
+      "step": 196030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2882830366351285e-06,
+      "loss": 0.1431,
+      "step": 196040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2860409847092055e-06,
+      "loss": 0.1333,
+      "step": 196050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.283798932783283e-06,
+      "loss": 0.1374,
+      "step": 196060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.281556880857361e-06,
+      "loss": 0.137,
+      "step": 196070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.279314828931439e-06,
+      "loss": 0.149,
+      "step": 196080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.277072777005516e-06,
+      "loss": 0.1391,
+      "step": 196090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.274830725079593e-06,
+      "loss": 0.1404,
+      "step": 196100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.27258867315367e-06,
+      "loss": 0.1389,
+      "step": 196110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.270346621227747e-06,
+      "loss": 0.1415,
+      "step": 196120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.268104569301825e-06,
+      "loss": 0.1362,
+      "step": 196130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.265862517375903e-06,
+      "loss": 0.1405,
+      "step": 196140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.263620465449981e-06,
+      "loss": 0.1404,
+      "step": 196150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.261378413524058e-06,
+      "loss": 0.1396,
+      "step": 196160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.259136361598135e-06,
+      "loss": 0.1369,
+      "step": 196170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.256894309672212e-06,
+      "loss": 0.1385,
+      "step": 196180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.254876462938881e-06,
+      "loss": 0.1398,
+      "step": 196190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.25263441101296e-06,
+      "loss": 0.14,
+      "step": 196200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.250392359087037e-06,
+      "loss": 0.1386,
+      "step": 196210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.248150307161114e-06,
+      "loss": 0.143,
+      "step": 196220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.245908255235192e-06,
+      "loss": 0.1411,
+      "step": 196230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.243666203309269e-06,
+      "loss": 0.1398,
+      "step": 196240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.241424151383346e-06,
+      "loss": 0.136,
+      "step": 196250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.239182099457424e-06,
+      "loss": 0.1347,
+      "step": 196260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.236940047531501e-06,
+      "loss": 0.1368,
+      "step": 196270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.234697995605578e-06,
+      "loss": 0.1392,
+      "step": 196280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.232455943679656e-06,
+      "loss": 0.1404,
+      "step": 196290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.230213891753734e-06,
+      "loss": 0.1414,
+      "step": 196300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.227971839827811e-06,
+      "loss": 0.1397,
+      "step": 196310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.225729787901888e-06,
+      "loss": 0.1383,
+      "step": 196320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.223487735975966e-06,
+      "loss": 0.1396,
+      "step": 196330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.221245684050043e-06,
+      "loss": 0.1383,
+      "step": 196340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.21900363212412e-06,
+      "loss": 0.1331,
+      "step": 196350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2167615801981976e-06,
+      "loss": 0.1394,
+      "step": 196360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2145195282722754e-06,
+      "loss": 0.1348,
+      "step": 196370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.2122774763463524e-06,
+      "loss": 0.1429,
+      "step": 196380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.21003542442043e-06,
+      "loss": 0.1392,
+      "step": 196390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.207793372494507e-06,
+      "loss": 0.138,
+      "step": 196400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.205551320568584e-06,
+      "loss": 0.1398,
+      "step": 196410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.203309268642661e-06,
+      "loss": 0.1389,
+      "step": 196420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.201067216716739e-06,
+      "loss": 0.1371,
+      "step": 196430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.198825164790817e-06,
+      "loss": 0.1362,
+      "step": 196440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.196583112864894e-06,
+      "loss": 0.1329,
+      "step": 196450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.194341060938972e-06,
+      "loss": 0.1405,
+      "step": 196460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.192099009013049e-06,
+      "loss": 0.136,
+      "step": 196470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.189856957087126e-06,
+      "loss": 0.138,
+      "step": 196480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.187614905161204e-06,
+      "loss": 0.1407,
+      "step": 196490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.185372853235282e-06,
+      "loss": 0.141,
+      "step": 196500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.183130801309359e-06,
+      "loss": 0.1394,
+      "step": 196510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.180888749383436e-06,
+      "loss": 0.1348,
+      "step": 196520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.178646697457514e-06,
+      "loss": 0.1364,
+      "step": 196530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.176404645531591e-06,
+      "loss": 0.1425,
+      "step": 196540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.174162593605668e-06,
+      "loss": 0.1347,
+      "step": 196550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.171920541679746e-06,
+      "loss": 0.1401,
+      "step": 196560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1696784897538235e-06,
+      "loss": 0.1365,
+      "step": 196570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1674364378279005e-06,
+      "loss": 0.1357,
+      "step": 196580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1651943859019776e-06,
+      "loss": 0.1373,
+      "step": 196590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1629523339760554e-06,
+      "loss": 0.1352,
+      "step": 196600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1607102820501325e-06,
+      "loss": 0.138,
+      "step": 196610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1584682301242095e-06,
+      "loss": 0.1355,
+      "step": 196620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.156226178198287e-06,
+      "loss": 0.1362,
+      "step": 196630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.153984126272365e-06,
+      "loss": 0.1307,
+      "step": 196640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.151742074346442e-06,
+      "loss": 0.1403,
+      "step": 196650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.149500022420519e-06,
+      "loss": 0.136,
+      "step": 196660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.147257970494597e-06,
+      "loss": 0.1338,
+      "step": 196670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.145015918568674e-06,
+      "loss": 0.1359,
+      "step": 196680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.142773866642751e-06,
+      "loss": 0.1368,
+      "step": 196690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.140531814716829e-06,
+      "loss": 0.1366,
+      "step": 196700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.138289762790907e-06,
+      "loss": 0.1392,
+      "step": 196710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.136047710864984e-06,
+      "loss": 0.1341,
+      "step": 196720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.133805658939061e-06,
+      "loss": 0.1303,
+      "step": 196730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.131563607013139e-06,
+      "loss": 0.1321,
+      "step": 196740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.129321555087216e-06,
+      "loss": 0.1381,
+      "step": 196750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.127079503161294e-06,
+      "loss": 0.1366,
+      "step": 196760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.124837451235372e-06,
+      "loss": 0.1369,
+      "step": 196770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.122595399309449e-06,
+      "loss": 0.1378,
+      "step": 196780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.120353347383526e-06,
+      "loss": 0.1403,
+      "step": 196790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.118111295457603e-06,
+      "loss": 0.1341,
+      "step": 196800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1158692435316806e-06,
+      "loss": 0.1395,
+      "step": 196810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.113627191605758e-06,
+      "loss": 0.1401,
+      "step": 196820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.1113851396798355e-06,
+      "loss": 0.1408,
+      "step": 196830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.109143087753913e-06,
+      "loss": 0.1316,
+      "step": 196840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.10690103582799e-06,
+      "loss": 0.1326,
+      "step": 196850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.104658983902067e-06,
+      "loss": 0.1384,
+      "step": 196860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.102416931976144e-06,
+      "loss": 0.1318,
+      "step": 196870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.100174880050222e-06,
+      "loss": 0.1417,
+      "step": 196880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.097932828124299e-06,
+      "loss": 0.1345,
+      "step": 196890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.095690776198377e-06,
+      "loss": 0.1371,
+      "step": 196900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.093448724272455e-06,
+      "loss": 0.1334,
+      "step": 196910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.091206672346532e-06,
+      "loss": 0.1391,
+      "step": 196920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.088964620420609e-06,
+      "loss": 0.1342,
+      "step": 196930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.086722568494686e-06,
+      "loss": 0.137,
+      "step": 196940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.084480516568764e-06,
+      "loss": 0.1359,
+      "step": 196950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.082238464642841e-06,
+      "loss": 0.135,
+      "step": 196960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.079996412716919e-06,
+      "loss": 0.1347,
+      "step": 196970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.077754360790997e-06,
+      "loss": 0.1347,
+      "step": 196980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.075512308865074e-06,
+      "loss": 0.1373,
+      "step": 196990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.073270256939151e-06,
+      "loss": 0.1352,
+      "step": 197000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.071028205013228e-06,
+      "loss": 0.1369,
+      "step": 197010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.068786153087306e-06,
+      "loss": 0.1329,
+      "step": 197020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0665441011613836e-06,
+      "loss": 0.1414,
+      "step": 197030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0643020492354606e-06,
+      "loss": 0.1328,
+      "step": 197040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0620599973095384e-06,
+      "loss": 0.1318,
+      "step": 197050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0598179453836155e-06,
+      "loss": 0.1364,
+      "step": 197060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0575758934576925e-06,
+      "loss": 0.1345,
+      "step": 197070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.05533384153177e-06,
+      "loss": 0.1347,
+      "step": 197080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.053091789605847e-06,
+      "loss": 0.1381,
+      "step": 197090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.050849737679925e-06,
+      "loss": 0.1371,
+      "step": 197100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.048607685754002e-06,
+      "loss": 0.1349,
+      "step": 197110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.04636563382808e-06,
+      "loss": 0.1316,
+      "step": 197120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.044123581902157e-06,
+      "loss": 0.1366,
+      "step": 197130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.041881529976234e-06,
+      "loss": 0.1355,
+      "step": 197140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.039639478050312e-06,
+      "loss": 0.1305,
+      "step": 197150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.037397426124389e-06,
+      "loss": 0.1338,
+      "step": 197160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.035155374198467e-06,
+      "loss": 0.1379,
+      "step": 197170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.032913322272544e-06,
+      "loss": 0.1306,
+      "step": 197180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.030671270346622e-06,
+      "loss": 0.1382,
+      "step": 197190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.028429218420699e-06,
+      "loss": 0.1379,
+      "step": 197200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.026187166494776e-06,
+      "loss": 0.1345,
+      "step": 197210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.023945114568854e-06,
+      "loss": 0.132,
+      "step": 197220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.021703062642931e-06,
+      "loss": 0.1385,
+      "step": 197230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.019461010717009e-06,
+      "loss": 0.1296,
+      "step": 197240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.017218958791086e-06,
+      "loss": 0.1333,
+      "step": 197250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0149769068651636e-06,
+      "loss": 0.138,
+      "step": 197260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.012734854939241e-06,
+      "loss": 0.1381,
+      "step": 197270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.010492803013318e-06,
+      "loss": 0.1308,
+      "step": 197280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.0082507510873955e-06,
+      "loss": 0.1355,
+      "step": 197290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.006008699161473e-06,
+      "loss": 0.138,
+      "step": 197300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.00376664723555e-06,
+      "loss": 0.1431,
+      "step": 197310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.001524595309627e-06,
+      "loss": 0.1378,
+      "step": 197320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.999282543383705e-06,
+      "loss": 0.1383,
+      "step": 197330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.997040491457782e-06,
+      "loss": 0.1387,
+      "step": 197340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.994798439531859e-06,
+      "loss": 0.1319,
+      "step": 197350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.992556387605937e-06,
+      "loss": 0.1408,
+      "step": 197360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.990314335680015e-06,
+      "loss": 0.1396,
+      "step": 197370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.988072283754092e-06,
+      "loss": 0.139,
+      "step": 197380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.98583023182817e-06,
+      "loss": 0.1376,
+      "step": 197390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.983588179902247e-06,
+      "loss": 0.1464,
+      "step": 197400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.981346127976324e-06,
+      "loss": 0.1402,
+      "step": 197410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.979104076050401e-06,
+      "loss": 0.1421,
+      "step": 197420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.976862024124479e-06,
+      "loss": 0.1351,
+      "step": 197430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.974619972198557e-06,
+      "loss": 0.1451,
+      "step": 197440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.972377920272634e-06,
+      "loss": 0.1353,
+      "step": 197450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.970135868346712e-06,
+      "loss": 0.1416,
+      "step": 197460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.967893816420789e-06,
+      "loss": 0.1377,
+      "step": 197470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.965651764494866e-06,
+      "loss": 0.1384,
+      "step": 197480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.963409712568943e-06,
+      "loss": 0.1386,
+      "step": 197490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.961167660643021e-06,
+      "loss": 0.1371,
+      "step": 197500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9589256087170985e-06,
+      "loss": 0.1378,
+      "step": 197510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9566835567911755e-06,
+      "loss": 0.1405,
+      "step": 197520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.954441504865253e-06,
+      "loss": 0.1381,
+      "step": 197530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.95219945293933e-06,
+      "loss": 0.137,
+      "step": 197540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.949957401013407e-06,
+      "loss": 0.1386,
+      "step": 197550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9477153490874844e-06,
+      "loss": 0.1497,
+      "step": 197560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.945473297161562e-06,
+      "loss": 0.1418,
+      "step": 197570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.94323124523564e-06,
+      "loss": 0.1448,
+      "step": 197580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.940989193309717e-06,
+      "loss": 0.1402,
+      "step": 197590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.938747141383795e-06,
+      "loss": 0.1448,
+      "step": 197600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.936505089457872e-06,
+      "loss": 0.1395,
+      "step": 197610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.934263037531949e-06,
+      "loss": 0.1484,
+      "step": 197620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.932020985606027e-06,
+      "loss": 0.1472,
+      "step": 197630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.929778933680105e-06,
+      "loss": 0.1424,
+      "step": 197640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.927536881754182e-06,
+      "loss": 0.1396,
+      "step": 197650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.925294829828259e-06,
+      "loss": 0.1455,
+      "step": 197660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.923052777902337e-06,
+      "loss": 0.1444,
+      "step": 197670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.920810725976414e-06,
+      "loss": 0.143,
+      "step": 197680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.918568674050491e-06,
+      "loss": 0.1484,
+      "step": 197690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.916326622124569e-06,
+      "loss": 0.146,
+      "step": 197700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9140845701986466e-06,
+      "loss": 0.1454,
+      "step": 197710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.911842518272724e-06,
+      "loss": 0.1466,
+      "step": 197720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.909600466346801e-06,
+      "loss": 0.144,
+      "step": 197730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9073584144208785e-06,
+      "loss": 0.1441,
+      "step": 197740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9051163624949555e-06,
+      "loss": 0.1458,
+      "step": 197750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.9028743105690325e-06,
+      "loss": 0.1413,
+      "step": 197760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.90063225864311e-06,
+      "loss": 0.1372,
+      "step": 197770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.898390206717188e-06,
+      "loss": 0.1435,
+      "step": 197780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.896148154791265e-06,
+      "loss": 0.1436,
+      "step": 197790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.893906102865342e-06,
+      "loss": 0.1428,
+      "step": 197800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.89166405093942e-06,
+      "loss": 0.1445,
+      "step": 197810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.889421999013497e-06,
+      "loss": 0.142,
+      "step": 197820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.887179947087574e-06,
+      "loss": 0.1454,
+      "step": 197830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.884937895161652e-06,
+      "loss": 0.145,
+      "step": 197840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.88269584323573e-06,
+      "loss": 0.1406,
+      "step": 197850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.880453791309807e-06,
+      "loss": 0.1421,
+      "step": 197860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.878211739383884e-06,
+      "loss": 0.1428,
+      "step": 197870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.875969687457962e-06,
+      "loss": 0.1447,
+      "step": 197880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.873727635532039e-06,
+      "loss": 0.1481,
+      "step": 197890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.871485583606117e-06,
+      "loss": 0.1467,
+      "step": 197900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.869243531680195e-06,
+      "loss": 0.1469,
+      "step": 197910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.867001479754272e-06,
+      "loss": 0.1447,
+      "step": 197920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.864759427828349e-06,
+      "loss": 0.1456,
+      "step": 197930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.862517375902426e-06,
+      "loss": 0.1419,
+      "step": 197940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.860275323976504e-06,
+      "loss": 0.141,
+      "step": 197950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.858033272050581e-06,
+      "loss": 0.1469,
+      "step": 197960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8557912201246585e-06,
+      "loss": 0.1514,
+      "step": 197970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.853549168198736e-06,
+      "loss": 0.1389,
+      "step": 197980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.851307116272813e-06,
+      "loss": 0.1409,
+      "step": 197990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8490650643468904e-06,
+      "loss": 0.1362,
+      "step": 198000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8468230124209674e-06,
+      "loss": 0.1432,
+      "step": 198010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.844580960495045e-06,
+      "loss": 0.1464,
+      "step": 198020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.842338908569122e-06,
+      "loss": 0.1426,
+      "step": 198030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8400968566432e-06,
+      "loss": 0.1351,
+      "step": 198040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.837854804717278e-06,
+      "loss": 0.145,
+      "step": 198050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.835612752791355e-06,
+      "loss": 0.1408,
+      "step": 198060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.833370700865432e-06,
+      "loss": 0.1507,
+      "step": 198070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.83112864893951e-06,
+      "loss": 0.1418,
+      "step": 198080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.828886597013587e-06,
+      "loss": 0.1423,
+      "step": 198090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.826644545087664e-06,
+      "loss": 0.1394,
+      "step": 198100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.824402493161742e-06,
+      "loss": 0.1367,
+      "step": 198110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.82216044123582e-06,
+      "loss": 0.1426,
+      "step": 198120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.819918389309897e-06,
+      "loss": 0.1405,
+      "step": 198130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.817676337383974e-06,
+      "loss": 0.1445,
+      "step": 198140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.815434285458052e-06,
+      "loss": 0.1425,
+      "step": 198150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.813192233532129e-06,
+      "loss": 0.1396,
+      "step": 198160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.810950181606207e-06,
+      "loss": 0.1468,
+      "step": 198170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.808708129680284e-06,
+      "loss": 0.1428,
+      "step": 198180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.8064660777543615e-06,
+      "loss": 0.1416,
+      "step": 198190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.804448231021031e-06,
+      "loss": 0.1368,
+      "step": 198200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.802206179095108e-06,
+      "loss": 0.1418,
+      "step": 198210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.799964127169186e-06,
+      "loss": 0.1411,
+      "step": 198220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.797722075243263e-06,
+      "loss": 0.1367,
+      "step": 198230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.79548002331734e-06,
+      "loss": 0.1468,
+      "step": 198240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.793237971391418e-06,
+      "loss": 0.1395,
+      "step": 198250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.790995919465495e-06,
+      "loss": 0.1395,
+      "step": 198260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.788753867539573e-06,
+      "loss": 0.1394,
+      "step": 198270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.78651181561365e-06,
+      "loss": 0.1383,
+      "step": 198280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.784269763687728e-06,
+      "loss": 0.1415,
+      "step": 198290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.782027711761805e-06,
+      "loss": 0.1431,
+      "step": 198300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.779785659835882e-06,
+      "loss": 0.1438,
+      "step": 198310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7775436079099595e-06,
+      "loss": 0.1389,
+      "step": 198320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7753015559840365e-06,
+      "loss": 0.1465,
+      "step": 198330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.773059504058114e-06,
+      "loss": 0.1429,
+      "step": 198340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.770817452132192e-06,
+      "loss": 0.1448,
+      "step": 198350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.768575400206269e-06,
+      "loss": 0.1425,
+      "step": 198360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.766333348280346e-06,
+      "loss": 0.1419,
+      "step": 198370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.764091296354423e-06,
+      "loss": 0.1418,
+      "step": 198380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.761849244428501e-06,
+      "loss": 0.1394,
+      "step": 198390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.759607192502578e-06,
+      "loss": 0.1414,
+      "step": 198400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.757365140576656e-06,
+      "loss": 0.1414,
+      "step": 198410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.755123088650734e-06,
+      "loss": 0.1398,
+      "step": 198420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.752881036724811e-06,
+      "loss": 0.1414,
+      "step": 198430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.750638984798888e-06,
+      "loss": 0.1353,
+      "step": 198440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.748396932872965e-06,
+      "loss": 0.1417,
+      "step": 198450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.746154880947043e-06,
+      "loss": 0.1409,
+      "step": 198460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.743912829021121e-06,
+      "loss": 0.139,
+      "step": 198470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.741670777095198e-06,
+      "loss": 0.1388,
+      "step": 198480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.739428725169276e-06,
+      "loss": 0.1449,
+      "step": 198490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.737186673243353e-06,
+      "loss": 0.1363,
+      "step": 198500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.73494462131743e-06,
+      "loss": 0.1415,
+      "step": 198510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.732702569391507e-06,
+      "loss": 0.1441,
+      "step": 198520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.730460517465585e-06,
+      "loss": 0.1323,
+      "step": 198530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7282184655396625e-06,
+      "loss": 0.1417,
+      "step": 198540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7259764136137395e-06,
+      "loss": 0.1434,
+      "step": 198550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.723734361687817e-06,
+      "loss": 0.1411,
+      "step": 198560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.721492309761894e-06,
+      "loss": 0.1404,
+      "step": 198570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7192502578359714e-06,
+      "loss": 0.1449,
+      "step": 198580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7170082059100485e-06,
+      "loss": 0.1457,
+      "step": 198590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.714766153984126e-06,
+      "loss": 0.1425,
+      "step": 198600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.712524102058204e-06,
+      "loss": 0.1405,
+      "step": 198610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.710282050132281e-06,
+      "loss": 0.141,
+      "step": 198620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.708039998206359e-06,
+      "loss": 0.1447,
+      "step": 198630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.705797946280436e-06,
+      "loss": 0.1434,
+      "step": 198640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.703555894354513e-06,
+      "loss": 0.1363,
+      "step": 198650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.701313842428591e-06,
+      "loss": 0.1377,
+      "step": 198660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.699071790502668e-06,
+      "loss": 0.1363,
+      "step": 198670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.696829738576746e-06,
+      "loss": 0.144,
+      "step": 198680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.694587686650823e-06,
+      "loss": 0.1409,
+      "step": 198690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.692345634724901e-06,
+      "loss": 0.139,
+      "step": 198700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.690103582798978e-06,
+      "loss": 0.1416,
+      "step": 198710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.687861530873055e-06,
+      "loss": 0.1435,
+      "step": 198720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.685619478947133e-06,
+      "loss": 0.138,
+      "step": 198730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.683377427021211e-06,
+      "loss": 0.1414,
+      "step": 198740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.681135375095288e-06,
+      "loss": 0.14,
+      "step": 198750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.678893323169365e-06,
+      "loss": 0.1467,
+      "step": 198760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6766512712434425e-06,
+      "loss": 0.1379,
+      "step": 198770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6744092193175195e-06,
+      "loss": 0.1378,
+      "step": 198780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6721671673915966e-06,
+      "loss": 0.1417,
+      "step": 198790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6699251154656744e-06,
+      "loss": 0.1431,
+      "step": 198800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.667683063539752e-06,
+      "loss": 0.1436,
+      "step": 198810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.665441011613829e-06,
+      "loss": 0.1393,
+      "step": 198820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.663198959687906e-06,
+      "loss": 0.1394,
+      "step": 198830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.660956907761984e-06,
+      "loss": 0.1396,
+      "step": 198840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.658714855836061e-06,
+      "loss": 0.1431,
+      "step": 198850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.656472803910138e-06,
+      "loss": 0.1418,
+      "step": 198860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.654230751984216e-06,
+      "loss": 0.1427,
+      "step": 198870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.651988700058294e-06,
+      "loss": 0.1406,
+      "step": 198880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.649746648132371e-06,
+      "loss": 0.1433,
+      "step": 198890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.647504596206448e-06,
+      "loss": 0.1352,
+      "step": 198900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.645262544280526e-06,
+      "loss": 0.1412,
+      "step": 198910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.643020492354603e-06,
+      "loss": 0.1421,
+      "step": 198920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.64077844042868e-06,
+      "loss": 0.1415,
+      "step": 198930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.638536388502758e-06,
+      "loss": 0.1431,
+      "step": 198940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.636294336576836e-06,
+      "loss": 0.1422,
+      "step": 198950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.634052284650913e-06,
+      "loss": 0.1414,
+      "step": 198960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.63181023272499e-06,
+      "loss": 0.1382,
+      "step": 198970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.629568180799068e-06,
+      "loss": 0.141,
+      "step": 198980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.627326128873145e-06,
+      "loss": 0.1405,
+      "step": 198990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6250840769472225e-06,
+      "loss": 0.1426,
+      "step": 199000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6228420250213e-06,
+      "loss": 0.1414,
+      "step": 199010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6205999730953774e-06,
+      "loss": 0.1454,
+      "step": 199020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6183579211694544e-06,
+      "loss": 0.1419,
+      "step": 199030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.616115869243532e-06,
+      "loss": 0.1372,
+      "step": 199040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.613873817317609e-06,
+      "loss": 0.1433,
+      "step": 199050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.611631765391686e-06,
+      "loss": 0.1398,
+      "step": 199060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.609389713465764e-06,
+      "loss": 0.138,
+      "step": 199070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.607147661539842e-06,
+      "loss": 0.1442,
+      "step": 199080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.604905609613919e-06,
+      "loss": 0.1381,
+      "step": 199090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.602663557687996e-06,
+      "loss": 0.1414,
+      "step": 199100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.600421505762074e-06,
+      "loss": 0.1461,
+      "step": 199110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.598179453836151e-06,
+      "loss": 0.1429,
+      "step": 199120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.595937401910228e-06,
+      "loss": 0.1414,
+      "step": 199130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.593695349984306e-06,
+      "loss": 0.1369,
+      "step": 199140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.591453298058384e-06,
+      "loss": 0.1389,
+      "step": 199150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.589211246132461e-06,
+      "loss": 0.1336,
+      "step": 199160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.586969194206538e-06,
+      "loss": 0.1403,
+      "step": 199170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.584727142280616e-06,
+      "loss": 0.1385,
+      "step": 199180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.582485090354693e-06,
+      "loss": 0.1451,
+      "step": 199190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.58024303842877e-06,
+      "loss": 0.1458,
+      "step": 199200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.578000986502848e-06,
+      "loss": 0.1376,
+      "step": 199210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5757589345769255e-06,
+      "loss": 0.1361,
+      "step": 199220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5735168826510025e-06,
+      "loss": 0.1386,
+      "step": 199230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5712748307250796e-06,
+      "loss": 0.1359,
+      "step": 199240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5690327787991574e-06,
+      "loss": 0.142,
+      "step": 199250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5667907268732345e-06,
+      "loss": 0.1405,
+      "step": 199260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.564548674947312e-06,
+      "loss": 0.1392,
+      "step": 199270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.562306623021389e-06,
+      "loss": 0.1389,
+      "step": 199280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.560064571095467e-06,
+      "loss": 0.1395,
+      "step": 199290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.557822519169544e-06,
+      "loss": 0.1417,
+      "step": 199300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.555580467243621e-06,
+      "loss": 0.1423,
+      "step": 199310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.553338415317699e-06,
+      "loss": 0.1384,
+      "step": 199320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.551096363391776e-06,
+      "loss": 0.1407,
+      "step": 199330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.548854311465854e-06,
+      "loss": 0.1398,
+      "step": 199340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.546612259539932e-06,
+      "loss": 0.1429,
+      "step": 199350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.544370207614009e-06,
+      "loss": 0.1455,
+      "step": 199360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.542128155688086e-06,
+      "loss": 0.1415,
+      "step": 199370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.539886103762163e-06,
+      "loss": 0.1382,
+      "step": 199380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.537644051836241e-06,
+      "loss": 0.1342,
+      "step": 199390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.535401999910318e-06,
+      "loss": 0.1401,
+      "step": 199400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.533159947984396e-06,
+      "loss": 0.1386,
+      "step": 199410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.530917896058474e-06,
+      "loss": 0.1406,
+      "step": 199420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.528675844132551e-06,
+      "loss": 0.1419,
+      "step": 199430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.526433792206628e-06,
+      "loss": 0.1426,
+      "step": 199440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.524191740280705e-06,
+      "loss": 0.1383,
+      "step": 199450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5219496883547826e-06,
+      "loss": 0.1386,
+      "step": 199460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.51970763642886e-06,
+      "loss": 0.1418,
+      "step": 199470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5174655845029375e-06,
+      "loss": 0.1371,
+      "step": 199480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.515223532577015e-06,
+      "loss": 0.1388,
+      "step": 199490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.512981480651092e-06,
+      "loss": 0.1367,
+      "step": 199500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.510739428725169e-06,
+      "loss": 0.1462,
+      "step": 199510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.508497376799246e-06,
+      "loss": 0.1393,
+      "step": 199520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.506255324873324e-06,
+      "loss": 0.134,
+      "step": 199530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.504013272947401e-06,
+      "loss": 0.143,
+      "step": 199540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.501771221021479e-06,
+      "loss": 0.1419,
+      "step": 199550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.499529169095557e-06,
+      "loss": 0.1415,
+      "step": 199560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.497287117169634e-06,
+      "loss": 0.1425,
+      "step": 199570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.495045065243711e-06,
+      "loss": 0.1395,
+      "step": 199580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.492803013317788e-06,
+      "loss": 0.1417,
+      "step": 199590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.490560961391866e-06,
+      "loss": 0.138,
+      "step": 199600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.488318909465944e-06,
+      "loss": 0.1398,
+      "step": 199610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.486076857540021e-06,
+      "loss": 0.1397,
+      "step": 199620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.483834805614099e-06,
+      "loss": 0.1391,
+      "step": 199630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.481592753688176e-06,
+      "loss": 0.1386,
+      "step": 199640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.479350701762253e-06,
+      "loss": 0.1439,
+      "step": 199650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.477108649836331e-06,
+      "loss": 0.1371,
+      "step": 199660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.474866597910408e-06,
+      "loss": 0.1459,
+      "step": 199670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4726245459844856e-06,
+      "loss": 0.1423,
+      "step": 199680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4703824940585626e-06,
+      "loss": 0.1428,
+      "step": 199690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4681404421326404e-06,
+      "loss": 0.145,
+      "step": 199700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4658983902067175e-06,
+      "loss": 0.1408,
+      "step": 199710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4636563382807945e-06,
+      "loss": 0.1368,
+      "step": 199720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.461414286354872e-06,
+      "loss": 0.1421,
+      "step": 199730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.459172234428949e-06,
+      "loss": 0.1399,
+      "step": 199740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.456930182503027e-06,
+      "loss": 0.1403,
+      "step": 199750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.454688130577104e-06,
+      "loss": 0.138,
+      "step": 199760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.452446078651182e-06,
+      "loss": 0.1349,
+      "step": 199770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.450204026725259e-06,
+      "loss": 0.1423,
+      "step": 199780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.447961974799336e-06,
+      "loss": 0.1373,
+      "step": 199790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.445719922873414e-06,
+      "loss": 0.1438,
+      "step": 199800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.443477870947491e-06,
+      "loss": 0.1411,
+      "step": 199810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.441235819021569e-06,
+      "loss": 0.1401,
+      "step": 199820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.438993767095646e-06,
+      "loss": 0.1418,
+      "step": 199830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.436751715169724e-06,
+      "loss": 0.1355,
+      "step": 199840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.434509663243801e-06,
+      "loss": 0.1378,
+      "step": 199850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.432267611317878e-06,
+      "loss": 0.1382,
+      "step": 199860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.430025559391956e-06,
+      "loss": 0.1447,
+      "step": 199870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.427783507466034e-06,
+      "loss": 0.1372,
+      "step": 199880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.425541455540111e-06,
+      "loss": 0.1405,
+      "step": 199890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.423299403614188e-06,
+      "loss": 0.1411,
+      "step": 199900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4210573516882656e-06,
+      "loss": 0.1353,
+      "step": 199910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.418815299762343e-06,
+      "loss": 0.1402,
+      "step": 199920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.41657324783642e-06,
+      "loss": 0.1366,
+      "step": 199930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.4143311959104975e-06,
+      "loss": 0.1351,
+      "step": 199940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.412089143984575e-06,
+      "loss": 0.1477,
+      "step": 199950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.409847092058652e-06,
+      "loss": 0.1378,
+      "step": 199960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.407605040132729e-06,
+      "loss": 0.1435,
+      "step": 199970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.405362988206807e-06,
+      "loss": 0.1381,
+      "step": 199980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.403120936280884e-06,
+      "loss": 0.1443,
+      "step": 199990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.400878884354961e-06,
+      "loss": 0.1394,
+      "step": 200000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.398636832429039e-06,
+      "loss": 0.1394,
+      "step": 200010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.396394780503117e-06,
+      "loss": 0.1418,
+      "step": 200020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.394152728577194e-06,
+      "loss": 0.1381,
+      "step": 200030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.391910676651272e-06,
+      "loss": 0.1422,
+      "step": 200040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.389668624725349e-06,
+      "loss": 0.1457,
+      "step": 200050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.387426572799426e-06,
+      "loss": 0.1424,
+      "step": 200060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.385184520873503e-06,
+      "loss": 0.1383,
+      "step": 200070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.382942468947581e-06,
+      "loss": 0.1367,
+      "step": 200080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.380700417021659e-06,
+      "loss": 0.1489,
+      "step": 200090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.378458365095736e-06,
+      "loss": 0.1379,
+      "step": 200100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.376216313169814e-06,
+      "loss": 0.1382,
+      "step": 200110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.373974261243891e-06,
+      "loss": 0.1402,
+      "step": 200120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.371732209317968e-06,
+      "loss": 0.1351,
+      "step": 200130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.369490157392046e-06,
+      "loss": 0.1359,
+      "step": 200140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3672481054661235e-06,
+      "loss": 0.1361,
+      "step": 200150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3650060535402005e-06,
+      "loss": 0.1392,
+      "step": 200160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3627640016142775e-06,
+      "loss": 0.1411,
+      "step": 200170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.360521949688355e-06,
+      "loss": 0.138,
+      "step": 200180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.358279897762432e-06,
+      "loss": 0.1419,
+      "step": 200190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.356037845836509e-06,
+      "loss": 0.145,
+      "step": 200200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.35401999910318e-06,
+      "loss": 0.1459,
+      "step": 200210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.351777947177257e-06,
+      "loss": 0.1381,
+      "step": 200220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.349535895251334e-06,
+      "loss": 0.1417,
+      "step": 200230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.347293843325412e-06,
+      "loss": 0.1415,
+      "step": 200240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3450517913994895e-06,
+      "loss": 0.1382,
+      "step": 200250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3428097394735666e-06,
+      "loss": 0.1409,
+      "step": 200260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.340567687547644e-06,
+      "loss": 0.1407,
+      "step": 200270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3383256356217215e-06,
+      "loss": 0.1371,
+      "step": 200280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3360835836957985e-06,
+      "loss": 0.1388,
+      "step": 200290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.3338415317698755e-06,
+      "loss": 0.1432,
+      "step": 200300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.331599479843953e-06,
+      "loss": 0.1416,
+      "step": 200310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.329357427918031e-06,
+      "loss": 0.138,
+      "step": 200320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.327115375992108e-06,
+      "loss": 0.1393,
+      "step": 200330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.324873324066185e-06,
+      "loss": 0.1419,
+      "step": 200340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.322631272140263e-06,
+      "loss": 0.1399,
+      "step": 200350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.32038922021434e-06,
+      "loss": 0.1356,
+      "step": 200360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.318147168288417e-06,
+      "loss": 0.1381,
+      "step": 200370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.315905116362496e-06,
+      "loss": 0.1317,
+      "step": 200380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.313663064436573e-06,
+      "loss": 0.1417,
+      "step": 200390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.31142101251065e-06,
+      "loss": 0.143,
+      "step": 200400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.309178960584727e-06,
+      "loss": 0.1359,
+      "step": 200410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.306936908658805e-06,
+      "loss": 0.1451,
+      "step": 200420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.304694856732882e-06,
+      "loss": 0.1402,
+      "step": 200430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.30245280480696e-06,
+      "loss": 0.1386,
+      "step": 200440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.300210752881038e-06,
+      "loss": 0.1377,
+      "step": 200450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.297968700955115e-06,
+      "loss": 0.1413,
+      "step": 200460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.295726649029192e-06,
+      "loss": 0.1386,
+      "step": 200470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.293484597103269e-06,
+      "loss": 0.1387,
+      "step": 200480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.291242545177347e-06,
+      "loss": 0.1387,
+      "step": 200490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.289000493251424e-06,
+      "loss": 0.1392,
+      "step": 200500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2867584413255015e-06,
+      "loss": 0.1371,
+      "step": 200510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.284516389399579e-06,
+      "loss": 0.1407,
+      "step": 200520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.282274337473656e-06,
+      "loss": 0.1388,
+      "step": 200530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.280032285547733e-06,
+      "loss": 0.1384,
+      "step": 200540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.27779023362181e-06,
+      "loss": 0.142,
+      "step": 200550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.275548181695888e-06,
+      "loss": 0.1347,
+      "step": 200560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.273306129769965e-06,
+      "loss": 0.1405,
+      "step": 200570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.271064077844043e-06,
+      "loss": 0.1372,
+      "step": 200580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.268822025918121e-06,
+      "loss": 0.1406,
+      "step": 200590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.266579973992198e-06,
+      "loss": 0.1366,
+      "step": 200600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.264337922066275e-06,
+      "loss": 0.1367,
+      "step": 200610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.262095870140353e-06,
+      "loss": 0.1375,
+      "step": 200620
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.25985381821443e-06,
+      "loss": 0.1394,
+      "step": 200630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.257611766288507e-06,
+      "loss": 0.1388,
+      "step": 200640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.255369714362585e-06,
+      "loss": 0.1398,
+      "step": 200650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.253127662436663e-06,
+      "loss": 0.144,
+      "step": 200660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.25088561051074e-06,
+      "loss": 0.138,
+      "step": 200670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.248643558584817e-06,
+      "loss": 0.1343,
+      "step": 200680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.246401506658895e-06,
+      "loss": 0.1396,
+      "step": 200690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.244159454732972e-06,
+      "loss": 0.14,
+      "step": 200700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2419174028070496e-06,
+      "loss": 0.134,
+      "step": 200710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.239675350881127e-06,
+      "loss": 0.1371,
+      "step": 200720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2374332989552045e-06,
+      "loss": 0.1359,
+      "step": 200730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2351912470292815e-06,
+      "loss": 0.1451,
+      "step": 200740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2329491951033585e-06,
+      "loss": 0.1402,
+      "step": 200750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.230707143177436e-06,
+      "loss": 0.1398,
+      "step": 200760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.228465091251513e-06,
+      "loss": 0.1379,
+      "step": 200770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.226223039325591e-06,
+      "loss": 0.1408,
+      "step": 200780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.223980987399668e-06,
+      "loss": 0.1441,
+      "step": 200790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.221738935473746e-06,
+      "loss": 0.1408,
+      "step": 200800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.219496883547823e-06,
+      "loss": 0.138,
+      "step": 200810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2172548316219e-06,
+      "loss": 0.137,
+      "step": 200820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.215012779695978e-06,
+      "loss": 0.1356,
+      "step": 200830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.212770727770055e-06,
+      "loss": 0.1418,
+      "step": 200840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.210528675844133e-06,
+      "loss": 0.1417,
+      "step": 200850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.20828662391821e-06,
+      "loss": 0.1413,
+      "step": 200860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.206044571992288e-06,
+      "loss": 0.1406,
+      "step": 200870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.203802520066365e-06,
+      "loss": 0.1399,
+      "step": 200880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.201560468140442e-06,
+      "loss": 0.1368,
+      "step": 200890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.19931841621452e-06,
+      "loss": 0.1378,
+      "step": 200900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.197076364288597e-06,
+      "loss": 0.1395,
+      "step": 200910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.194834312362675e-06,
+      "loss": 0.1418,
+      "step": 200920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.192592260436752e-06,
+      "loss": 0.1426,
+      "step": 200930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.19035020851083e-06,
+      "loss": 0.1412,
+      "step": 200940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.188108156584907e-06,
+      "loss": 0.1377,
+      "step": 200950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.185866104658984e-06,
+      "loss": 0.1424,
+      "step": 200960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1836240527330615e-06,
+      "loss": 0.1402,
+      "step": 200970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.181382000807139e-06,
+      "loss": 0.1394,
+      "step": 200980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.179139948881216e-06,
+      "loss": 0.1367,
+      "step": 200990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.176897896955294e-06,
+      "loss": 0.1362,
+      "step": 201000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.174655845029371e-06,
+      "loss": 0.1322,
+      "step": 201010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.172413793103448e-06,
+      "loss": 0.1433,
+      "step": 201020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.170171741177525e-06,
+      "loss": 0.1367,
+      "step": 201030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.167929689251603e-06,
+      "loss": 0.1382,
+      "step": 201040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.165687637325681e-06,
+      "loss": 0.1392,
+      "step": 201050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.163445585399758e-06,
+      "loss": 0.1404,
+      "step": 201060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.161203533473836e-06,
+      "loss": 0.1332,
+      "step": 201070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.158961481547913e-06,
+      "loss": 0.1379,
+      "step": 201080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.15671942962199e-06,
+      "loss": 0.1435,
+      "step": 201090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.154477377696067e-06,
+      "loss": 0.1423,
+      "step": 201100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.152235325770145e-06,
+      "loss": 0.1401,
+      "step": 201110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.149993273844223e-06,
+      "loss": 0.1434,
+      "step": 201120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1477512219183e-06,
+      "loss": 0.1385,
+      "step": 201130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.145509169992378e-06,
+      "loss": 0.1429,
+      "step": 201140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.143267118066455e-06,
+      "loss": 0.137,
+      "step": 201150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.141025066140532e-06,
+      "loss": 0.1404,
+      "step": 201160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.138783014214609e-06,
+      "loss": 0.1356,
+      "step": 201170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.136540962288687e-06,
+      "loss": 0.1325,
+      "step": 201180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1342989103627645e-06,
+      "loss": 0.135,
+      "step": 201190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1320568584368415e-06,
+      "loss": 0.1352,
+      "step": 201200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.129814806510919e-06,
+      "loss": 0.1427,
+      "step": 201210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.127572754584996e-06,
+      "loss": 0.1412,
+      "step": 201220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1253307026590734e-06,
+      "loss": 0.1441,
+      "step": 201230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.123088650733151e-06,
+      "loss": 0.1459,
+      "step": 201240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.120846598807229e-06,
+      "loss": 0.1394,
+      "step": 201250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.118604546881306e-06,
+      "loss": 0.1415,
+      "step": 201260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.116362494955383e-06,
+      "loss": 0.142,
+      "step": 201270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.114120443029461e-06,
+      "loss": 0.1327,
+      "step": 201280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.111878391103538e-06,
+      "loss": 0.1381,
+      "step": 201290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.109636339177615e-06,
+      "loss": 0.1366,
+      "step": 201300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.107394287251693e-06,
+      "loss": 0.1329,
+      "step": 201310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.105152235325771e-06,
+      "loss": 0.1396,
+      "step": 201320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.102910183399848e-06,
+      "loss": 0.1355,
+      "step": 201330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.100668131473925e-06,
+      "loss": 0.1399,
+      "step": 201340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.098426079548003e-06,
+      "loss": 0.1435,
+      "step": 201350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.09618402762208e-06,
+      "loss": 0.1422,
+      "step": 201360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.093941975696157e-06,
+      "loss": 0.1371,
+      "step": 201370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.091699923770235e-06,
+      "loss": 0.1424,
+      "step": 201380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.089457871844313e-06,
+      "loss": 0.1332,
+      "step": 201390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.08721581991839e-06,
+      "loss": 0.1438,
+      "step": 201400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.084973767992467e-06,
+      "loss": 0.1394,
+      "step": 201410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0827317160665445e-06,
+      "loss": 0.1423,
+      "step": 201420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0804896641406215e-06,
+      "loss": 0.141,
+      "step": 201430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0782476122146986e-06,
+      "loss": 0.1403,
+      "step": 201440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0760055602887764e-06,
+      "loss": 0.1397,
+      "step": 201450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.073763508362854e-06,
+      "loss": 0.1384,
+      "step": 201460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.071521456436931e-06,
+      "loss": 0.1338,
+      "step": 201470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.069279404511008e-06,
+      "loss": 0.1419,
+      "step": 201480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.067037352585086e-06,
+      "loss": 0.1381,
+      "step": 201490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.064795300659163e-06,
+      "loss": 0.1434,
+      "step": 201500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.06255324873324e-06,
+      "loss": 0.1399,
+      "step": 201510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.060311196807319e-06,
+      "loss": 0.1406,
+      "step": 201520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.058069144881396e-06,
+      "loss": 0.136,
+      "step": 201530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.055827092955473e-06,
+      "loss": 0.1369,
+      "step": 201540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.05358504102955e-06,
+      "loss": 0.1378,
+      "step": 201550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.051342989103628e-06,
+      "loss": 0.1396,
+      "step": 201560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.049100937177705e-06,
+      "loss": 0.145,
+      "step": 201570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.046858885251783e-06,
+      "loss": 0.1372,
+      "step": 201580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.044616833325861e-06,
+      "loss": 0.146,
+      "step": 201590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.042374781399938e-06,
+      "loss": 0.1429,
+      "step": 201600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.040132729474015e-06,
+      "loss": 0.138,
+      "step": 201610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.037890677548092e-06,
+      "loss": 0.136,
+      "step": 201620
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.03564862562217e-06,
+      "loss": 0.138,
+      "step": 201630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.033406573696247e-06,
+      "loss": 0.142,
+      "step": 201640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0311645217703245e-06,
+      "loss": 0.136,
+      "step": 201650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.028922469844402e-06,
+      "loss": 0.137,
+      "step": 201660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.026680417918479e-06,
+      "loss": 0.1385,
+      "step": 201670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0244383659925564e-06,
+      "loss": 0.137,
+      "step": 201680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.022196314066634e-06,
+      "loss": 0.1407,
+      "step": 201690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.019954262140711e-06,
+      "loss": 0.1421,
+      "step": 201700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.017712210214788e-06,
+      "loss": 0.1384,
+      "step": 201710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.015470158288866e-06,
+      "loss": 0.1424,
+      "step": 201720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.013228106362944e-06,
+      "loss": 0.1406,
+      "step": 201730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.010986054437021e-06,
+      "loss": 0.1412,
+      "step": 201740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.008744002511098e-06,
+      "loss": 0.1387,
+      "step": 201750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.006501950585176e-06,
+      "loss": 0.1387,
+      "step": 201760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.004259898659253e-06,
+      "loss": 0.1384,
+      "step": 201770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.00201784673333e-06,
+      "loss": 0.1435,
+      "step": 201780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.999775794807408e-06,
+      "loss": 0.1407,
+      "step": 201790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.997533742881486e-06,
+      "loss": 0.1384,
+      "step": 201800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.995291690955563e-06,
+      "loss": 0.1413,
+      "step": 201810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.99304963902964e-06,
+      "loss": 0.1377,
+      "step": 201820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.990807587103718e-06,
+      "loss": 0.1431,
+      "step": 201830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.988565535177795e-06,
+      "loss": 0.1333,
+      "step": 201840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.986323483251873e-06,
+      "loss": 0.1357,
+      "step": 201850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.98408143132595e-06,
+      "loss": 0.1419,
+      "step": 201860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9818393794000275e-06,
+      "loss": 0.1433,
+      "step": 201870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9795973274741045e-06,
+      "loss": 0.1366,
+      "step": 201880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9773552755481816e-06,
+      "loss": 0.133,
+      "step": 201890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9751132236222594e-06,
+      "loss": 0.1428,
+      "step": 201900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9728711716963365e-06,
+      "loss": 0.1376,
+      "step": 201910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.970629119770414e-06,
+      "loss": 0.1385,
+      "step": 201920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.968387067844491e-06,
+      "loss": 0.1408,
+      "step": 201930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.966145015918569e-06,
+      "loss": 0.1362,
+      "step": 201940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.963902963992646e-06,
+      "loss": 0.133,
+      "step": 201950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.961660912066723e-06,
+      "loss": 0.1356,
+      "step": 201960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.959418860140801e-06,
+      "loss": 0.1372,
+      "step": 201970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.957176808214878e-06,
+      "loss": 0.1402,
+      "step": 201980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.954934756288956e-06,
+      "loss": 0.1374,
+      "step": 201990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.952692704363034e-06,
+      "loss": 0.14,
+      "step": 202000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.950450652437111e-06,
+      "loss": 0.1389,
+      "step": 202010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.948208600511188e-06,
+      "loss": 0.1454,
+      "step": 202020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.945966548585265e-06,
+      "loss": 0.1385,
+      "step": 202030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.943724496659343e-06,
+      "loss": 0.1335,
+      "step": 202040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.94148244473342e-06,
+      "loss": 0.137,
+      "step": 202050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.939240392807498e-06,
+      "loss": 0.1372,
+      "step": 202060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.936998340881576e-06,
+      "loss": 0.1369,
+      "step": 202070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.934756288955653e-06,
+      "loss": 0.1394,
+      "step": 202080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.93251423702973e-06,
+      "loss": 0.1446,
+      "step": 202090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.930272185103807e-06,
+      "loss": 0.1361,
+      "step": 202100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9280301331778846e-06,
+      "loss": 0.1388,
+      "step": 202110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9257880812519624e-06,
+      "loss": 0.136,
+      "step": 202120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9235460293260395e-06,
+      "loss": 0.1362,
+      "step": 202130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.921303977400117e-06,
+      "loss": 0.1353,
+      "step": 202140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.919061925474194e-06,
+      "loss": 0.1393,
+      "step": 202150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.916819873548271e-06,
+      "loss": 0.1368,
+      "step": 202160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.914577821622348e-06,
+      "loss": 0.1344,
+      "step": 202170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.912335769696426e-06,
+      "loss": 0.1302,
+      "step": 202180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.910093717770504e-06,
+      "loss": 0.1428,
+      "step": 202190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.907851665844581e-06,
+      "loss": 0.1385,
+      "step": 202200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.905609613918659e-06,
+      "loss": 0.1426,
+      "step": 202210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.903367561992736e-06,
+      "loss": 0.1369,
+      "step": 202220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.901125510066813e-06,
+      "loss": 0.139,
+      "step": 202230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.89888345814089e-06,
+      "loss": 0.14,
+      "step": 202240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8968656114075604e-06,
+      "loss": 0.1372,
+      "step": 202250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8946235594816375e-06,
+      "loss": 0.1364,
+      "step": 202260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.892381507555715e-06,
+      "loss": 0.1355,
+      "step": 202270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.890139455629792e-06,
+      "loss": 0.132,
+      "step": 202280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.88789740370387e-06,
+      "loss": 0.1374,
+      "step": 202290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.885655351777947e-06,
+      "loss": 0.1343,
+      "step": 202300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.883413299852025e-06,
+      "loss": 0.1403,
+      "step": 202310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.881171247926102e-06,
+      "loss": 0.1395,
+      "step": 202320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.878929196000179e-06,
+      "loss": 0.1404,
+      "step": 202330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.876687144074257e-06,
+      "loss": 0.1411,
+      "step": 202340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.874445092148335e-06,
+      "loss": 0.1368,
+      "step": 202350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.872203040222412e-06,
+      "loss": 0.1414,
+      "step": 202360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.869960988296489e-06,
+      "loss": 0.1377,
+      "step": 202370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.867718936370567e-06,
+      "loss": 0.1386,
+      "step": 202380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.865476884444644e-06,
+      "loss": 0.1384,
+      "step": 202390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.863234832518721e-06,
+      "loss": 0.1364,
+      "step": 202400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.860992780592799e-06,
+      "loss": 0.1345,
+      "step": 202410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.858750728666877e-06,
+      "loss": 0.1369,
+      "step": 202420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.856508676740954e-06,
+      "loss": 0.137,
+      "step": 202430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.854266624815031e-06,
+      "loss": 0.1373,
+      "step": 202440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8520245728891085e-06,
+      "loss": 0.1401,
+      "step": 202450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8497825209631856e-06,
+      "loss": 0.1349,
+      "step": 202460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.847540469037263e-06,
+      "loss": 0.1425,
+      "step": 202470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8452984171113405e-06,
+      "loss": 0.1422,
+      "step": 202480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.843056365185418e-06,
+      "loss": 0.1425,
+      "step": 202490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.840814313259495e-06,
+      "loss": 0.1368,
+      "step": 202500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.838572261333572e-06,
+      "loss": 0.1342,
+      "step": 202510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.83633020940765e-06,
+      "loss": 0.1416,
+      "step": 202520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.834088157481727e-06,
+      "loss": 0.1408,
+      "step": 202530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.831846105555804e-06,
+      "loss": 0.1399,
+      "step": 202540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.829604053629882e-06,
+      "loss": 0.1372,
+      "step": 202550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.82736200170396e-06,
+      "loss": 0.1377,
+      "step": 202560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.825119949778037e-06,
+      "loss": 0.145,
+      "step": 202570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.822877897852115e-06,
+      "loss": 0.1438,
+      "step": 202580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.820635845926192e-06,
+      "loss": 0.1338,
+      "step": 202590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.818393794000269e-06,
+      "loss": 0.1448,
+      "step": 202600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.816151742074346e-06,
+      "loss": 0.1414,
+      "step": 202610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.813909690148425e-06,
+      "loss": 0.1352,
+      "step": 202620
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.811667638222502e-06,
+      "loss": 0.1428,
+      "step": 202630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.809425586296579e-06,
+      "loss": 0.1346,
+      "step": 202640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.807183534370657e-06,
+      "loss": 0.1422,
+      "step": 202650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.804941482444734e-06,
+      "loss": 0.142,
+      "step": 202660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.802699430518811e-06,
+      "loss": 0.1394,
+      "step": 202670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8004573785928886e-06,
+      "loss": 0.1426,
+      "step": 202680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.798215326666966e-06,
+      "loss": 0.1393,
+      "step": 202690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7959732747410434e-06,
+      "loss": 0.1361,
+      "step": 202700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7937312228151205e-06,
+      "loss": 0.1368,
+      "step": 202710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.791489170889198e-06,
+      "loss": 0.1391,
+      "step": 202720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.789247118963275e-06,
+      "loss": 0.1373,
+      "step": 202730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.787005067037352e-06,
+      "loss": 0.1405,
+      "step": 202740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.78476301511143e-06,
+      "loss": 0.1335,
+      "step": 202750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.782520963185508e-06,
+      "loss": 0.1374,
+      "step": 202760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.780278911259585e-06,
+      "loss": 0.1349,
+      "step": 202770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.778036859333662e-06,
+      "loss": 0.1393,
+      "step": 202780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.77579480740774e-06,
+      "loss": 0.1374,
+      "step": 202790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.773552755481817e-06,
+      "loss": 0.1389,
+      "step": 202800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.771310703555894e-06,
+      "loss": 0.1375,
+      "step": 202810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.769068651629972e-06,
+      "loss": 0.1391,
+      "step": 202820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.76682659970405e-06,
+      "loss": 0.136,
+      "step": 202830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.764584547778127e-06,
+      "loss": 0.1392,
+      "step": 202840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.762342495852204e-06,
+      "loss": 0.1405,
+      "step": 202850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.760100443926282e-06,
+      "loss": 0.1393,
+      "step": 202860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.757858392000359e-06,
+      "loss": 0.1408,
+      "step": 202870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.755616340074436e-06,
+      "loss": 0.139,
+      "step": 202880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.753374288148514e-06,
+      "loss": 0.1397,
+      "step": 202890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.7511322362225915e-06,
+      "loss": 0.1428,
+      "step": 202900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.7488901842966686e-06,
+      "loss": 0.136,
+      "step": 202910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.746648132370746e-06,
+      "loss": 0.1403,
+      "step": 202920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.7444060804448235e-06,
+      "loss": 0.1398,
+      "step": 202930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.7421640285189005e-06,
+      "loss": 0.1399,
+      "step": 202940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.739921976592978e-06,
+      "loss": 0.137,
+      "step": 202950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.737679924667056e-06,
+      "loss": 0.1373,
+      "step": 202960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.735437872741133e-06,
+      "loss": 0.1375,
+      "step": 202970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.73319582081521e-06,
+      "loss": 0.1432,
+      "step": 202980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.730953768889287e-06,
+      "loss": 0.1356,
+      "step": 202990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.728711716963365e-06,
+      "loss": 0.1371,
+      "step": 203000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.726469665037442e-06,
+      "loss": 0.1369,
+      "step": 203010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.72422761311152e-06,
+      "loss": 0.1346,
+      "step": 203020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.721985561185598e-06,
+      "loss": 0.1362,
+      "step": 203030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.719743509259675e-06,
+      "loss": 0.1342,
+      "step": 203040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.717501457333752e-06,
+      "loss": 0.1347,
+      "step": 203050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.715259405407829e-06,
+      "loss": 0.1383,
+      "step": 203060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.713017353481907e-06,
+      "loss": 0.1367,
+      "step": 203070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.710775301555984e-06,
+      "loss": 0.1356,
+      "step": 203080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.708533249630062e-06,
+      "loss": 0.1382,
+      "step": 203090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.70629119770414e-06,
+      "loss": 0.1365,
+      "step": 203100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.704049145778217e-06,
+      "loss": 0.1375,
+      "step": 203110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.701807093852294e-06,
+      "loss": 0.1376,
+      "step": 203120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.699565041926371e-06,
+      "loss": 0.1377,
+      "step": 203130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.697322990000449e-06,
+      "loss": 0.1392,
+      "step": 203140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.695080938074526e-06,
+      "loss": 0.1417,
+      "step": 203150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6928388861486035e-06,
+      "loss": 0.1429,
+      "step": 203160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.690596834222681e-06,
+      "loss": 0.1361,
+      "step": 203170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.688354782296758e-06,
+      "loss": 0.1393,
+      "step": 203180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.686112730370835e-06,
+      "loss": 0.133,
+      "step": 203190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.683870678444912e-06,
+      "loss": 0.1334,
+      "step": 203200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.68162862651899e-06,
+      "loss": 0.1361,
+      "step": 203210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.679386574593068e-06,
+      "loss": 0.1428,
+      "step": 203220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.677144522667145e-06,
+      "loss": 0.1317,
+      "step": 203230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.674902470741223e-06,
+      "loss": 0.1369,
+      "step": 203240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6726604188153e-06,
+      "loss": 0.1391,
+      "step": 203250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.670418366889377e-06,
+      "loss": 0.1367,
+      "step": 203260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.668176314963455e-06,
+      "loss": 0.14,
+      "step": 203270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.665934263037532e-06,
+      "loss": 0.1347,
+      "step": 203280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.66369221111161e-06,
+      "loss": 0.1406,
+      "step": 203290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.661450159185687e-06,
+      "loss": 0.1367,
+      "step": 203300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.659208107259765e-06,
+      "loss": 0.139,
+      "step": 203310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.656966055333842e-06,
+      "loss": 0.1384,
+      "step": 203320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.654724003407919e-06,
+      "loss": 0.1419,
+      "step": 203330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.652481951481997e-06,
+      "loss": 0.1427,
+      "step": 203340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.650239899556074e-06,
+      "loss": 0.1374,
+      "step": 203350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6479978476301516e-06,
+      "loss": 0.1427,
+      "step": 203360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.645755795704229e-06,
+      "loss": 0.1405,
+      "step": 203370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6435137437783065e-06,
+      "loss": 0.1327,
+      "step": 203380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6412716918523835e-06,
+      "loss": 0.1385,
+      "step": 203390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6390296399264605e-06,
+      "loss": 0.1332,
+      "step": 203400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.636787588000538e-06,
+      "loss": 0.1437,
+      "step": 203410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.634545536074615e-06,
+      "loss": 0.1396,
+      "step": 203420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.632303484148693e-06,
+      "loss": 0.1377,
+      "step": 203430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.63006143222277e-06,
+      "loss": 0.1398,
+      "step": 203440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.627819380296848e-06,
+      "loss": 0.1388,
+      "step": 203450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.625577328370925e-06,
+      "loss": 0.1408,
+      "step": 203460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.623335276445002e-06,
+      "loss": 0.138,
+      "step": 203470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.62109322451908e-06,
+      "loss": 0.1383,
+      "step": 203480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.618851172593158e-06,
+      "loss": 0.1425,
+      "step": 203490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.616609120667235e-06,
+      "loss": 0.1351,
+      "step": 203500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.614367068741312e-06,
+      "loss": 0.1402,
+      "step": 203510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.612349222007982e-06,
+      "loss": 0.1388,
+      "step": 203520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.610107170082059e-06,
+      "loss": 0.1438,
+      "step": 203530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.607865118156137e-06,
+      "loss": 0.1433,
+      "step": 203540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.605623066230214e-06,
+      "loss": 0.1407,
+      "step": 203550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.603381014304291e-06,
+      "loss": 0.1359,
+      "step": 203560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.601138962378368e-06,
+      "loss": 0.1342,
+      "step": 203570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.598896910452446e-06,
+      "loss": 0.1346,
+      "step": 203580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.596654858526524e-06,
+      "loss": 0.1423,
+      "step": 203590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.594412806600601e-06,
+      "loss": 0.1403,
+      "step": 203600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.592170754674679e-06,
+      "loss": 0.1428,
+      "step": 203610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.589928702748756e-06,
+      "loss": 0.1347,
+      "step": 203620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.587686650822833e-06,
+      "loss": 0.1451,
+      "step": 203630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.58544459889691e-06,
+      "loss": 0.1347,
+      "step": 203640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.583202546970988e-06,
+      "loss": 0.137,
+      "step": 203650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.580960495045066e-06,
+      "loss": 0.1421,
+      "step": 203660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.578718443119143e-06,
+      "loss": 0.1426,
+      "step": 203670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.576476391193221e-06,
+      "loss": 0.1359,
+      "step": 203680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.574234339267298e-06,
+      "loss": 0.1411,
+      "step": 203690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.571992287341375e-06,
+      "loss": 0.1385,
+      "step": 203700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.569750235415452e-06,
+      "loss": 0.1334,
+      "step": 203710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5675081834895304e-06,
+      "loss": 0.1361,
+      "step": 203720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5652661315636075e-06,
+      "loss": 0.1359,
+      "step": 203730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5630240796376845e-06,
+      "loss": 0.1379,
+      "step": 203740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.560782027711762e-06,
+      "loss": 0.1406,
+      "step": 203750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.558539975785839e-06,
+      "loss": 0.1402,
+      "step": 203760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.556297923859916e-06,
+      "loss": 0.1377,
+      "step": 203770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.554055871933994e-06,
+      "loss": 0.1354,
+      "step": 203780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.551813820008072e-06,
+      "loss": 0.1368,
+      "step": 203790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.549571768082149e-06,
+      "loss": 0.1417,
+      "step": 203800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.547329716156226e-06,
+      "loss": 0.1366,
+      "step": 203810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.545087664230304e-06,
+      "loss": 0.1364,
+      "step": 203820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.542845612304381e-06,
+      "loss": 0.1402,
+      "step": 203830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.540603560378458e-06,
+      "loss": 0.1383,
+      "step": 203840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.538361508452536e-06,
+      "loss": 0.1409,
+      "step": 203850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.536119456526614e-06,
+      "loss": 0.1341,
+      "step": 203860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.533877404600691e-06,
+      "loss": 0.1334,
+      "step": 203870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.531635352674768e-06,
+      "loss": 0.1372,
+      "step": 203880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.529393300748846e-06,
+      "loss": 0.1415,
+      "step": 203890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.527151248822923e-06,
+      "loss": 0.1444,
+      "step": 203900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.524909196897e-06,
+      "loss": 0.1336,
+      "step": 203910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.522667144971078e-06,
+      "loss": 0.136,
+      "step": 203920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5204250930451556e-06,
+      "loss": 0.1405,
+      "step": 203930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.518183041119233e-06,
+      "loss": 0.137,
+      "step": 203940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.51594098919331e-06,
+      "loss": 0.1352,
+      "step": 203950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5136989372673875e-06,
+      "loss": 0.1392,
+      "step": 203960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5114568853414645e-06,
+      "loss": 0.1476,
+      "step": 203970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5092148334155415e-06,
+      "loss": 0.136,
+      "step": 203980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.506972781489619e-06,
+      "loss": 0.1342,
+      "step": 203990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.504730729563697e-06,
+      "loss": 0.1418,
+      "step": 204000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.502488677637774e-06,
+      "loss": 0.1355,
+      "step": 204010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.500246625711851e-06,
+      "loss": 0.138,
+      "step": 204020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.498004573785929e-06,
+      "loss": 0.1343,
+      "step": 204030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.495762521860006e-06,
+      "loss": 0.1429,
+      "step": 204040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.493520469934084e-06,
+      "loss": 0.135,
+      "step": 204050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.491278418008162e-06,
+      "loss": 0.1408,
+      "step": 204060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.489036366082239e-06,
+      "loss": 0.1392,
+      "step": 204070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.486794314156316e-06,
+      "loss": 0.1374,
+      "step": 204080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.484552262230393e-06,
+      "loss": 0.1386,
+      "step": 204090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.482310210304471e-06,
+      "loss": 0.1344,
+      "step": 204100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.480068158378548e-06,
+      "loss": 0.1409,
+      "step": 204110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.477826106452626e-06,
+      "loss": 0.1419,
+      "step": 204120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.475584054526704e-06,
+      "loss": 0.1374,
+      "step": 204130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.473342002600781e-06,
+      "loss": 0.1392,
+      "step": 204140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.471099950674858e-06,
+      "loss": 0.137,
+      "step": 204150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.468857898748935e-06,
+      "loss": 0.1361,
+      "step": 204160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.466615846823013e-06,
+      "loss": 0.139,
+      "step": 204170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.46437379489709e-06,
+      "loss": 0.1367,
+      "step": 204180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4621317429711675e-06,
+      "loss": 0.1426,
+      "step": 204190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.459889691045245e-06,
+      "loss": 0.1414,
+      "step": 204200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.457647639119322e-06,
+      "loss": 0.1406,
+      "step": 204210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.455405587193399e-06,
+      "loss": 0.1375,
+      "step": 204220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.453163535267477e-06,
+      "loss": 0.1396,
+      "step": 204230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.450921483341554e-06,
+      "loss": 0.1344,
+      "step": 204240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.448679431415631e-06,
+      "loss": 0.1377,
+      "step": 204250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.446437379489709e-06,
+      "loss": 0.1394,
+      "step": 204260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.444195327563787e-06,
+      "loss": 0.1437,
+      "step": 204270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.441953275637864e-06,
+      "loss": 0.1376,
+      "step": 204280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.439711223711941e-06,
+      "loss": 0.1383,
+      "step": 204290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.437469171786019e-06,
+      "loss": 0.1314,
+      "step": 204300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.435227119860096e-06,
+      "loss": 0.1355,
+      "step": 204310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.432985067934174e-06,
+      "loss": 0.1367,
+      "step": 204320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.430743016008251e-06,
+      "loss": 0.1418,
+      "step": 204330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.428500964082329e-06,
+      "loss": 0.1432,
+      "step": 204340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.426258912156406e-06,
+      "loss": 0.1378,
+      "step": 204350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.424016860230483e-06,
+      "loss": 0.1346,
+      "step": 204360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.421774808304561e-06,
+      "loss": 0.1425,
+      "step": 204370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.419532756378638e-06,
+      "loss": 0.141,
+      "step": 204380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.417290704452716e-06,
+      "loss": 0.1391,
+      "step": 204390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.415048652526793e-06,
+      "loss": 0.1412,
+      "step": 204400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4128066006008705e-06,
+      "loss": 0.1401,
+      "step": 204410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4105645486749475e-06,
+      "loss": 0.1384,
+      "step": 204420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4083224967490245e-06,
+      "loss": 0.1358,
+      "step": 204430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.406080444823102e-06,
+      "loss": 0.1346,
+      "step": 204440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4038383928971794e-06,
+      "loss": 0.14,
+      "step": 204450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.401596340971257e-06,
+      "loss": 0.1385,
+      "step": 204460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.399354289045334e-06,
+      "loss": 0.1398,
+      "step": 204470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.397112237119412e-06,
+      "loss": 0.1382,
+      "step": 204480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.394870185193489e-06,
+      "loss": 0.1417,
+      "step": 204490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.392628133267566e-06,
+      "loss": 0.1357,
+      "step": 204500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.390386081341644e-06,
+      "loss": 0.1435,
+      "step": 204510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.388144029415721e-06,
+      "loss": 0.1369,
+      "step": 204520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.385901977489799e-06,
+      "loss": 0.1368,
+      "step": 204530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.383659925563877e-06,
+      "loss": 0.1399,
+      "step": 204540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.381417873637954e-06,
+      "loss": 0.1352,
+      "step": 204550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.379175821712031e-06,
+      "loss": 0.1376,
+      "step": 204560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.376933769786108e-06,
+      "loss": 0.1393,
+      "step": 204570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.374691717860186e-06,
+      "loss": 0.1369,
+      "step": 204580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.372449665934264e-06,
+      "loss": 0.1428,
+      "step": 204590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.370207614008341e-06,
+      "loss": 0.1373,
+      "step": 204600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.367965562082419e-06,
+      "loss": 0.1362,
+      "step": 204610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.365723510156496e-06,
+      "loss": 0.1389,
+      "step": 204620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.363481458230573e-06,
+      "loss": 0.1365,
+      "step": 204630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.36123940630465e-06,
+      "loss": 0.1355,
+      "step": 204640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3589973543787275e-06,
+      "loss": 0.1338,
+      "step": 204650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.356755302452805e-06,
+      "loss": 0.1379,
+      "step": 204660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.354513250526882e-06,
+      "loss": 0.1398,
+      "step": 204670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.35227119860096e-06,
+      "loss": 0.1385,
+      "step": 204680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.350029146675037e-06,
+      "loss": 0.1371,
+      "step": 204690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.347787094749114e-06,
+      "loss": 0.1382,
+      "step": 204700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.345545042823191e-06,
+      "loss": 0.1365,
+      "step": 204710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.343302990897269e-06,
+      "loss": 0.1344,
+      "step": 204720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.341060938971347e-06,
+      "loss": 0.1385,
+      "step": 204730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.338818887045424e-06,
+      "loss": 0.1437,
+      "step": 204740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.336576835119502e-06,
+      "loss": 0.1383,
+      "step": 204750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.334334783193579e-06,
+      "loss": 0.1373,
+      "step": 204760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.332092731267656e-06,
+      "loss": 0.137,
+      "step": 204770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.329850679341733e-06,
+      "loss": 0.1349,
+      "step": 204780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.327608627415811e-06,
+      "loss": 0.1372,
+      "step": 204790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.325366575489889e-06,
+      "loss": 0.1346,
+      "step": 204800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.323124523563966e-06,
+      "loss": 0.14,
+      "step": 204810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.320882471638044e-06,
+      "loss": 0.1466,
+      "step": 204820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.318640419712121e-06,
+      "loss": 0.1334,
+      "step": 204830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.316398367786198e-06,
+      "loss": 0.1363,
+      "step": 204840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.314156315860275e-06,
+      "loss": 0.1376,
+      "step": 204850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3119142639343535e-06,
+      "loss": 0.1391,
+      "step": 204860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3096722120084305e-06,
+      "loss": 0.1361,
+      "step": 204870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3074301600825075e-06,
+      "loss": 0.1408,
+      "step": 204880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.305188108156585e-06,
+      "loss": 0.1386,
+      "step": 204890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3029460562306624e-06,
+      "loss": 0.1417,
+      "step": 204900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3007040043047395e-06,
+      "loss": 0.1343,
+      "step": 204910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.298461952378817e-06,
+      "loss": 0.1375,
+      "step": 204920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.296219900452895e-06,
+      "loss": 0.1394,
+      "step": 204930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.293977848526972e-06,
+      "loss": 0.1395,
+      "step": 204940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.291735796601049e-06,
+      "loss": 0.1379,
+      "step": 204950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.289493744675127e-06,
+      "loss": 0.1352,
+      "step": 204960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.287251692749204e-06,
+      "loss": 0.138,
+      "step": 204970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.285009640823281e-06,
+      "loss": 0.1367,
+      "step": 204980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.282767588897359e-06,
+      "loss": 0.1343,
+      "step": 204990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.280525536971437e-06,
+      "loss": 0.1402,
+      "step": 205000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.278283485045514e-06,
+      "loss": 0.1349,
+      "step": 205010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.276041433119591e-06,
+      "loss": 0.1348,
+      "step": 205020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.273799381193669e-06,
+      "loss": 0.1324,
+      "step": 205030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.271557329267746e-06,
+      "loss": 0.133,
+      "step": 205040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.269315277341823e-06,
+      "loss": 0.1325,
+      "step": 205050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.267073225415901e-06,
+      "loss": 0.1376,
+      "step": 205060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.264831173489979e-06,
+      "loss": 0.146,
+      "step": 205070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.262589121564056e-06,
+      "loss": 0.1354,
+      "step": 205080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.260347069638133e-06,
+      "loss": 0.1361,
+      "step": 205090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2581050177122105e-06,
+      "loss": 0.1375,
+      "step": 205100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2558629657862876e-06,
+      "loss": 0.1441,
+      "step": 205110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.253620913860365e-06,
+      "loss": 0.1388,
+      "step": 205120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2513788619344425e-06,
+      "loss": 0.1359,
+      "step": 205130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.24913681000852e-06,
+      "loss": 0.1384,
+      "step": 205140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.246894758082597e-06,
+      "loss": 0.1416,
+      "step": 205150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.244652706156674e-06,
+      "loss": 0.1387,
+      "step": 205160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.242410654230752e-06,
+      "loss": 0.1351,
+      "step": 205170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.240168602304829e-06,
+      "loss": 0.1441,
+      "step": 205180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.237926550378907e-06,
+      "loss": 0.1392,
+      "step": 205190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.235684498452985e-06,
+      "loss": 0.1341,
+      "step": 205200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.233442446527062e-06,
+      "loss": 0.1382,
+      "step": 205210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.231200394601139e-06,
+      "loss": 0.1384,
+      "step": 205220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.228958342675217e-06,
+      "loss": 0.1396,
+      "step": 205230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.226716290749294e-06,
+      "loss": 0.1362,
+      "step": 205240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.224474238823371e-06,
+      "loss": 0.1357,
+      "step": 205250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.222232186897449e-06,
+      "loss": 0.1407,
+      "step": 205260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.219990134971527e-06,
+      "loss": 0.1371,
+      "step": 205270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.217748083045604e-06,
+      "loss": 0.1406,
+      "step": 205280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.215506031119681e-06,
+      "loss": 0.1443,
+      "step": 205290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.213263979193759e-06,
+      "loss": 0.1357,
+      "step": 205300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.211021927267836e-06,
+      "loss": 0.1353,
+      "step": 205310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.208779875341913e-06,
+      "loss": 0.1374,
+      "step": 205320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2065378234159905e-06,
+      "loss": 0.1383,
+      "step": 205330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.204295771490068e-06,
+      "loss": 0.14,
+      "step": 205340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2020537195641454e-06,
+      "loss": 0.1383,
+      "step": 205350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1998116676382225e-06,
+      "loss": 0.1399,
+      "step": 205360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1975696157123e-06,
+      "loss": 0.1407,
+      "step": 205370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.195327563786377e-06,
+      "loss": 0.1366,
+      "step": 205380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.193085511860454e-06,
+      "loss": 0.142,
+      "step": 205390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.190843459934532e-06,
+      "loss": 0.135,
+      "step": 205400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.18860140800861e-06,
+      "loss": 0.1363,
+      "step": 205410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.186359356082687e-06,
+      "loss": 0.1409,
+      "step": 205420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.184117304156764e-06,
+      "loss": 0.1367,
+      "step": 205430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.181875252230842e-06,
+      "loss": 0.1366,
+      "step": 205440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.179633200304919e-06,
+      "loss": 0.14,
+      "step": 205450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.177391148378997e-06,
+      "loss": 0.138,
+      "step": 205460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.175149096453074e-06,
+      "loss": 0.136,
+      "step": 205470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.172907044527152e-06,
+      "loss": 0.1402,
+      "step": 205480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.170664992601229e-06,
+      "loss": 0.1457,
+      "step": 205490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.168422940675306e-06,
+      "loss": 0.1352,
+      "step": 205500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.166180888749384e-06,
+      "loss": 0.1355,
+      "step": 205510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.163938836823461e-06,
+      "loss": 0.1348,
+      "step": 205520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.161696784897539e-06,
+      "loss": 0.141,
+      "step": 205530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.159454732971616e-06,
+      "loss": 0.136,
+      "step": 205540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1572126810456935e-06,
+      "loss": 0.1343,
+      "step": 205550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1549706291197706e-06,
+      "loss": 0.1432,
+      "step": 205560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.152728577193848e-06,
+      "loss": 0.1309,
+      "step": 205570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1504865252679255e-06,
+      "loss": 0.14,
+      "step": 205580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1482444733420025e-06,
+      "loss": 0.1382,
+      "step": 205590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.14600242141608e-06,
+      "loss": 0.1392,
+      "step": 205600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.143760369490158e-06,
+      "loss": 0.1426,
+      "step": 205610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.141518317564235e-06,
+      "loss": 0.1386,
+      "step": 205620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.139276265638312e-06,
+      "loss": 0.1402,
+      "step": 205630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.137034213712389e-06,
+      "loss": 0.1383,
+      "step": 205640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.134792161786467e-06,
+      "loss": 0.1356,
+      "step": 205650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.132550109860544e-06,
+      "loss": 0.1386,
+      "step": 205660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.130308057934622e-06,
+      "loss": 0.133,
+      "step": 205670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1280660060087e-06,
+      "loss": 0.1354,
+      "step": 205680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.125823954082777e-06,
+      "loss": 0.1452,
+      "step": 205690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.123581902156854e-06,
+      "loss": 0.1382,
+      "step": 205700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.121339850230931e-06,
+      "loss": 0.1394,
+      "step": 205710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.119097798305009e-06,
+      "loss": 0.143,
+      "step": 205720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.116855746379087e-06,
+      "loss": 0.136,
+      "step": 205730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.114613694453164e-06,
+      "loss": 0.1387,
+      "step": 205740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.112371642527242e-06,
+      "loss": 0.1367,
+      "step": 205750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.110129590601319e-06,
+      "loss": 0.1361,
+      "step": 205760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.107887538675396e-06,
+      "loss": 0.1422,
+      "step": 205770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.105645486749473e-06,
+      "loss": 0.1335,
+      "step": 205780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.103403434823551e-06,
+      "loss": 0.13,
+      "step": 205790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1011613828976285e-06,
+      "loss": 0.1374,
+      "step": 205800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0989193309717055e-06,
+      "loss": 0.1372,
+      "step": 205810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.096677279045783e-06,
+      "loss": 0.1337,
+      "step": 205820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.09443522711986e-06,
+      "loss": 0.1343,
+      "step": 205830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.092193175193937e-06,
+      "loss": 0.1415,
+      "step": 205840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.089951123268014e-06,
+      "loss": 0.1378,
+      "step": 205850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.087709071342092e-06,
+      "loss": 0.136,
+      "step": 205860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.08546701941617e-06,
+      "loss": 0.1422,
+      "step": 205870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.083224967490247e-06,
+      "loss": 0.1422,
+      "step": 205880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.080982915564325e-06,
+      "loss": 0.1362,
+      "step": 205890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.078740863638402e-06,
+      "loss": 0.134,
+      "step": 205900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.076498811712479e-06,
+      "loss": 0.1359,
+      "step": 205910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.074256759786557e-06,
+      "loss": 0.1432,
+      "step": 205920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.072014707860634e-06,
+      "loss": 0.136,
+      "step": 205930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.069772655934712e-06,
+      "loss": 0.1354,
+      "step": 205940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.067530604008789e-06,
+      "loss": 0.1335,
+      "step": 205950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.065288552082867e-06,
+      "loss": 0.1366,
+      "step": 205960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.063046500156944e-06,
+      "loss": 0.137,
+      "step": 205970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.060804448231021e-06,
+      "loss": 0.1439,
+      "step": 205980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.058562396305099e-06,
+      "loss": 0.1383,
+      "step": 205990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0563203443791765e-06,
+      "loss": 0.138,
+      "step": 206000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0540782924532536e-06,
+      "loss": 0.1465,
+      "step": 206010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.051836240527331e-06,
+      "loss": 0.1396,
+      "step": 206020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0495941886014085e-06,
+      "loss": 0.1404,
+      "step": 206030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0473521366754855e-06,
+      "loss": 0.1379,
+      "step": 206040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0451100847495625e-06,
+      "loss": 0.1381,
+      "step": 206050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.04286803282364e-06,
+      "loss": 0.1395,
+      "step": 206060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.040625980897718e-06,
+      "loss": 0.1398,
+      "step": 206070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.038383928971795e-06,
+      "loss": 0.1367,
+      "step": 206080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.036141877045872e-06,
+      "loss": 0.1343,
+      "step": 206090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.03389982511995e-06,
+      "loss": 0.1342,
+      "step": 206100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.031657773194027e-06,
+      "loss": 0.1336,
+      "step": 206110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.029415721268104e-06,
+      "loss": 0.1377,
+      "step": 206120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.027173669342182e-06,
+      "loss": 0.1341,
+      "step": 206130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.02493161741626e-06,
+      "loss": 0.1415,
+      "step": 206140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.022689565490337e-06,
+      "loss": 0.1382,
+      "step": 206150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.020447513564414e-06,
+      "loss": 0.1336,
+      "step": 206160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.018205461638492e-06,
+      "loss": 0.1335,
+      "step": 206170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.015963409712569e-06,
+      "loss": 0.1382,
+      "step": 206180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.013721357786646e-06,
+      "loss": 0.135,
+      "step": 206190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.011479305860724e-06,
+      "loss": 0.1351,
+      "step": 206200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.009237253934802e-06,
+      "loss": 0.1393,
+      "step": 206210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.006995202008879e-06,
+      "loss": 0.1359,
+      "step": 206220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0047531500829566e-06,
+      "loss": 0.1373,
+      "step": 206230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.002511098157034e-06,
+      "loss": 0.141,
+      "step": 206240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.000269046231111e-06,
+      "loss": 0.135,
+      "step": 206250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.998026994305188e-06,
+      "loss": 0.1401,
+      "step": 206260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9957849423792655e-06,
+      "loss": 0.1381,
+      "step": 206270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.993542890453343e-06,
+      "loss": 0.1382,
+      "step": 206280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.99130083852742e-06,
+      "loss": 0.1424,
+      "step": 206290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.989058786601498e-06,
+      "loss": 0.1431,
+      "step": 206300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.986816734675575e-06,
+      "loss": 0.1388,
+      "step": 206310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.984574682749652e-06,
+      "loss": 0.135,
+      "step": 206320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.98233263082373e-06,
+      "loss": 0.1388,
+      "step": 206330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.980090578897808e-06,
+      "loss": 0.1394,
+      "step": 206340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.977848526971885e-06,
+      "loss": 0.1349,
+      "step": 206350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.975606475045962e-06,
+      "loss": 0.1399,
+      "step": 206360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.97336442312004e-06,
+      "loss": 0.1388,
+      "step": 206370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.971122371194117e-06,
+      "loss": 0.1412,
+      "step": 206380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.968880319268194e-06,
+      "loss": 0.1415,
+      "step": 206390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.966638267342272e-06,
+      "loss": 0.1375,
+      "step": 206400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.96439621541635e-06,
+      "loss": 0.1409,
+      "step": 206410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.962154163490427e-06,
+      "loss": 0.1346,
+      "step": 206420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.959912111564504e-06,
+      "loss": 0.1394,
+      "step": 206430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.957670059638582e-06,
+      "loss": 0.1348,
+      "step": 206440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.955428007712659e-06,
+      "loss": 0.1367,
+      "step": 206450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.953185955786736e-06,
+      "loss": 0.138,
+      "step": 206460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.950943903860814e-06,
+      "loss": 0.1366,
+      "step": 206470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9487018519348915e-06,
+      "loss": 0.1317,
+      "step": 206480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9464598000089685e-06,
+      "loss": 0.1349,
+      "step": 206490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9442177480830455e-06,
+      "loss": 0.134,
+      "step": 206500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.941975696157123e-06,
+      "loss": 0.1405,
+      "step": 206510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9397336442312e-06,
+      "loss": 0.1452,
+      "step": 206520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9374915923052774e-06,
+      "loss": 0.1358,
+      "step": 206530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.935249540379355e-06,
+      "loss": 0.1405,
+      "step": 206540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.933007488453433e-06,
+      "loss": 0.1341,
+      "step": 206550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.93076543652751e-06,
+      "loss": 0.1383,
+      "step": 206560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.928523384601587e-06,
+      "loss": 0.1372,
+      "step": 206570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.926281332675665e-06,
+      "loss": 0.147,
+      "step": 206580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.924039280749742e-06,
+      "loss": 0.1369,
+      "step": 206590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.92179722882382e-06,
+      "loss": 0.1286,
+      "step": 206600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.919555176897898e-06,
+      "loss": 0.1383,
+      "step": 206610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.917313124971975e-06,
+      "loss": 0.1333,
+      "step": 206620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.915071073046052e-06,
+      "loss": 0.1315,
+      "step": 206630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.912829021120129e-06,
+      "loss": 0.1365,
+      "step": 206640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.910586969194207e-06,
+      "loss": 0.1332,
+      "step": 206650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.908344917268284e-06,
+      "loss": 0.1354,
+      "step": 206660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.906102865342362e-06,
+      "loss": 0.1425,
+      "step": 206670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9038608134164396e-06,
+      "loss": 0.1442,
+      "step": 206680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.901618761490517e-06,
+      "loss": 0.1363,
+      "step": 206690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.899376709564594e-06,
+      "loss": 0.1344,
+      "step": 206700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.897134657638671e-06,
+      "loss": 0.1347,
+      "step": 206710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8948926057127485e-06,
+      "loss": 0.1404,
+      "step": 206720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8926505537868255e-06,
+      "loss": 0.1376,
+      "step": 206730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.890408501860903e-06,
+      "loss": 0.133,
+      "step": 206740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.888166449934981e-06,
+      "loss": 0.1418,
+      "step": 206750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.885924398009058e-06,
+      "loss": 0.14,
+      "step": 206760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.883682346083135e-06,
+      "loss": 0.1373,
+      "step": 206770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.881440294157212e-06,
+      "loss": 0.1369,
+      "step": 206780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.87919824223129e-06,
+      "loss": 0.132,
+      "step": 206790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.876956190305367e-06,
+      "loss": 0.1395,
+      "step": 206800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.874714138379445e-06,
+      "loss": 0.1342,
+      "step": 206810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.872472086453523e-06,
+      "loss": 0.1337,
+      "step": 206820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8702300345276e-06,
+      "loss": 0.1401,
+      "step": 206830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.867987982601677e-06,
+      "loss": 0.1382,
+      "step": 206840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.865745930675754e-06,
+      "loss": 0.1389,
+      "step": 206850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.863503878749832e-06,
+      "loss": 0.1332,
+      "step": 206860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.86126182682391e-06,
+      "loss": 0.1348,
+      "step": 206870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.859019774897987e-06,
+      "loss": 0.1387,
+      "step": 206880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.856777722972065e-06,
+      "loss": 0.1345,
+      "step": 206890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.854535671046142e-06,
+      "loss": 0.1387,
+      "step": 206900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.852293619120219e-06,
+      "loss": 0.1339,
+      "step": 206910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.850051567194297e-06,
+      "loss": 0.1339,
+      "step": 206920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.847809515268374e-06,
+      "loss": 0.1377,
+      "step": 206930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8455674633424515e-06,
+      "loss": 0.1364,
+      "step": 206940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8433254114165285e-06,
+      "loss": 0.1397,
+      "step": 206950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.841083359490606e-06,
+      "loss": 0.1363,
+      "step": 206960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.838841307564683e-06,
+      "loss": 0.139,
+      "step": 206970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8365992556387604e-06,
+      "loss": 0.1452,
+      "step": 206980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.834357203712838e-06,
+      "loss": 0.141,
+      "step": 206990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.832115151786915e-06,
+      "loss": 0.1395,
+      "step": 207000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.829873099860993e-06,
+      "loss": 0.14,
+      "step": 207010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.82763104793507e-06,
+      "loss": 0.1416,
+      "step": 207020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.825388996009148e-06,
+      "loss": 0.1378,
+      "step": 207030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.823146944083225e-06,
+      "loss": 0.1426,
+      "step": 207040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.820904892157302e-06,
+      "loss": 0.1405,
+      "step": 207050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.81866284023138e-06,
+      "loss": 0.1397,
+      "step": 207060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.816420788305457e-06,
+      "loss": 0.1359,
+      "step": 207070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8141787363795345e-06,
+      "loss": 0.1338,
+      "step": 207080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.811936684453612e-06,
+      "loss": 0.1376,
+      "step": 207090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.80969463252769e-06,
+      "loss": 0.141,
+      "step": 207100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.807452580601767e-06,
+      "loss": 0.1343,
+      "step": 207110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.8052105286758443e-06,
+      "loss": 0.1397,
+      "step": 207120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.802968476749922e-06,
+      "loss": 0.1366,
+      "step": 207130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.800726424823999e-06,
+      "loss": 0.1347,
+      "step": 207140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.798484372898076e-06,
+      "loss": 0.1393,
+      "step": 207150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7962423209721536e-06,
+      "loss": 0.1444,
+      "step": 207160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7940002690462315e-06,
+      "loss": 0.1383,
+      "step": 207170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7917582171203085e-06,
+      "loss": 0.1395,
+      "step": 207180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.789516165194386e-06,
+      "loss": 0.1399,
+      "step": 207190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.787274113268464e-06,
+      "loss": 0.1378,
+      "step": 207200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.785032061342541e-06,
+      "loss": 0.1319,
+      "step": 207210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7827900094166183e-06,
+      "loss": 0.135,
+      "step": 207220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7805479574906954e-06,
+      "loss": 0.134,
+      "step": 207230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7783059055647732e-06,
+      "loss": 0.1356,
+      "step": 207240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7760638536388502e-06,
+      "loss": 0.1357,
+      "step": 207250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7738218017129277e-06,
+      "loss": 0.1336,
+      "step": 207260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7715797497870056e-06,
+      "loss": 0.1366,
+      "step": 207270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7693376978610826e-06,
+      "loss": 0.1361,
+      "step": 207280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.76709564593516e-06,
+      "loss": 0.1378,
+      "step": 207290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.764853594009238e-06,
+      "loss": 0.1371,
+      "step": 207300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.762611542083315e-06,
+      "loss": 0.1353,
+      "step": 207310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.760369490157392e-06,
+      "loss": 0.1382,
+      "step": 207320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7581274382314694e-06,
+      "loss": 0.1393,
+      "step": 207330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7558853863055473e-06,
+      "loss": 0.1341,
+      "step": 207340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7536433343796243e-06,
+      "loss": 0.1391,
+      "step": 207350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7514012824537017e-06,
+      "loss": 0.1345,
+      "step": 207360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7491592305277796e-06,
+      "loss": 0.1367,
+      "step": 207370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7469171786018566e-06,
+      "loss": 0.1424,
+      "step": 207380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.744675126675934e-06,
+      "loss": 0.1431,
+      "step": 207390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.742433074750011e-06,
+      "loss": 0.1439,
+      "step": 207400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.740191022824089e-06,
+      "loss": 0.1358,
+      "step": 207410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.737948970898166e-06,
+      "loss": 0.135,
+      "step": 207420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7357069189722435e-06,
+      "loss": 0.1351,
+      "step": 207430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7334648670463213e-06,
+      "loss": 0.1411,
+      "step": 207440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7312228151203983e-06,
+      "loss": 0.1421,
+      "step": 207450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.728980763194476e-06,
+      "loss": 0.1395,
+      "step": 207460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.726738711268553e-06,
+      "loss": 0.1414,
+      "step": 207470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7244966593426307e-06,
+      "loss": 0.1382,
+      "step": 207480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.722254607416708e-06,
+      "loss": 0.1403,
+      "step": 207490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.720012555490785e-06,
+      "loss": 0.1428,
+      "step": 207500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.717770503564863e-06,
+      "loss": 0.1372,
+      "step": 207510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.71552845163894e-06,
+      "loss": 0.1373,
+      "step": 207520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7135106049056095e-06,
+      "loss": 0.1322,
+      "step": 207530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7112685529796874e-06,
+      "loss": 0.141,
+      "step": 207540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7090265010537644e-06,
+      "loss": 0.1381,
+      "step": 207550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.706784449127842e-06,
+      "loss": 0.1422,
+      "step": 207560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7045423972019198e-06,
+      "loss": 0.1347,
+      "step": 207570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7023003452759968e-06,
+      "loss": 0.1337,
+      "step": 207580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7000582933500742e-06,
+      "loss": 0.1371,
+      "step": 207590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6978162414241512e-06,
+      "loss": 0.1341,
+      "step": 207600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.695574189498229e-06,
+      "loss": 0.1372,
+      "step": 207610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6933321375723066e-06,
+      "loss": 0.136,
+      "step": 207620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6910900856463836e-06,
+      "loss": 0.14,
+      "step": 207630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6888480337204615e-06,
+      "loss": 0.1316,
+      "step": 207640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6866059817945385e-06,
+      "loss": 0.137,
+      "step": 207650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.684363929868616e-06,
+      "loss": 0.1412,
+      "step": 207660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.682121877942693e-06,
+      "loss": 0.1357,
+      "step": 207670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.679879826016771e-06,
+      "loss": 0.1353,
+      "step": 207680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6776377740908483e-06,
+      "loss": 0.1363,
+      "step": 207690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6753957221649253e-06,
+      "loss": 0.135,
+      "step": 207700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.673153670239003e-06,
+      "loss": 0.1382,
+      "step": 207710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.67091161831308e-06,
+      "loss": 0.1383,
+      "step": 207720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6686695663871576e-06,
+      "loss": 0.1368,
+      "step": 207730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6664275144612347e-06,
+      "loss": 0.1342,
+      "step": 207740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6641854625353125e-06,
+      "loss": 0.1376,
+      "step": 207750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.66194341060939e-06,
+      "loss": 0.1389,
+      "step": 207760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.659701358683467e-06,
+      "loss": 0.1418,
+      "step": 207770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.657459306757545e-06,
+      "loss": 0.1393,
+      "step": 207780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6552172548316223e-06,
+      "loss": 0.138,
+      "step": 207790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6529752029056993e-06,
+      "loss": 0.137,
+      "step": 207800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6507331509797764e-06,
+      "loss": 0.1353,
+      "step": 207810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6484910990538542e-06,
+      "loss": 0.138,
+      "step": 207820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6462490471279317e-06,
+      "loss": 0.1422,
+      "step": 207830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6440069952020087e-06,
+      "loss": 0.135,
+      "step": 207840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6417649432760866e-06,
+      "loss": 0.1349,
+      "step": 207850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.639522891350164e-06,
+      "loss": 0.1378,
+      "step": 207860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.637280839424241e-06,
+      "loss": 0.1334,
+      "step": 207870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.635038787498319e-06,
+      "loss": 0.1376,
+      "step": 207880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6327967355723964e-06,
+      "loss": 0.1341,
+      "step": 207890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6305546836464734e-06,
+      "loss": 0.1455,
+      "step": 207900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6283126317205504e-06,
+      "loss": 0.136,
+      "step": 207910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6260705797946283e-06,
+      "loss": 0.1384,
+      "step": 207920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6238285278687057e-06,
+      "loss": 0.1376,
+      "step": 207930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6215864759427828e-06,
+      "loss": 0.1344,
+      "step": 207940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6193444240168606e-06,
+      "loss": 0.1338,
+      "step": 207950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.617102372090938e-06,
+      "loss": 0.1356,
+      "step": 207960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.614860320165015e-06,
+      "loss": 0.138,
+      "step": 207970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.612618268239092e-06,
+      "loss": 0.137,
+      "step": 207980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.61037621631317e-06,
+      "loss": 0.143,
+      "step": 207990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6081341643872474e-06,
+      "loss": 0.1383,
+      "step": 208000
+    },
+    {
+      "epoch": 0.93,
+      "eval_NEIMS_canon_smiles": 0.9464833333333333,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.11943333333333334,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.5189043382837861,
+      "eval_NEIMS_exact_mols": 0.119,
+      "eval_NEIMS_exact_smiles": 0.11601666666666667,
+      "eval_NEIMS_loss": 0.17722082138061523,
+      "eval_NEIMS_matched_formulas": 0.6513,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.4303859638570458,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.11956666666666667,
+      "eval_NEIMS_runtime": 713.3866,
+      "eval_NEIMS_samples_per_second": 84.106,
+      "eval_NEIMS_steps_per_second": 1.315,
+      "step": 208000
+    },
+    {
+      "epoch": 0.93,
+      "eval_RASSP_canon_smiles": 0.9441847263498848,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.2553344241493305,
+      "eval_RASSP_daylight_tanimoto_simil": 0.6690008143553688,
+      "eval_RASSP_exact_mols": 0.25459979296757607,
+      "eval_RASSP_exact_smiles": 0.2474705312719137,
+      "eval_RASSP_loss": 0.11752571910619736,
+      "eval_RASSP_matched_formulas": 0.8234714662570541,
+      "eval_RASSP_morgan_tanimoto_simil": 0.5549820144524215,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.25480014692623637,
+      "eval_RASSP_runtime": 817.2827,
+      "eval_RASSP_samples_per_second": 73.284,
+      "eval_RASSP_steps_per_second": 1.145,
+      "step": 208000
+    },
+    {
+      "epoch": 0.93,
+      "eval_NIST_canon_smiles": 0.8780920609007347,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.01838378819604642,
+      "eval_NIST_daylight_tanimoto_simil": 0.28151468193484197,
+      "eval_NIST_exact_mols": 0.017390069915179048,
+      "eval_NIST_exact_smiles": 0.016360861695709265,
+      "eval_NIST_loss": 1.9125394821166992,
+      "eval_NIST_matched_formulas": 0.10363062071902615,
+      "eval_NIST_morgan_tanimoto_simil": 0.23322152811744992,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.018064378748624765,
+      "eval_NIST_runtime": 478.1073,
+      "eval_NIST_samples_per_second": 58.934,
+      "eval_NIST_steps_per_second": 0.922,
+      "step": 208000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6058921124613245e-06,
+      "loss": 0.1409,
+      "step": 208010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6036500605354023e-06,
+      "loss": 0.1395,
+      "step": 208020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6014080086094798e-06,
+      "loss": 0.1331,
+      "step": 208030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.599165956683557e-06,
+      "loss": 0.1416,
+      "step": 208040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5969239047576343e-06,
+      "loss": 0.136,
+      "step": 208050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.594681852831712e-06,
+      "loss": 0.1382,
+      "step": 208060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.592439800905789e-06,
+      "loss": 0.1379,
+      "step": 208070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.590197748979866e-06,
+      "loss": 0.1356,
+      "step": 208080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.587955697053944e-06,
+      "loss": 0.1367,
+      "step": 208090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5857136451280215e-06,
+      "loss": 0.136,
+      "step": 208100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5834715932020985e-06,
+      "loss": 0.1353,
+      "step": 208110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.581229541276176e-06,
+      "loss": 0.1361,
+      "step": 208120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.578987489350254e-06,
+      "loss": 0.1416,
+      "step": 208130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.576745437424331e-06,
+      "loss": 0.1328,
+      "step": 208140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.574503385498408e-06,
+      "loss": 0.1382,
+      "step": 208150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.572261333572486e-06,
+      "loss": 0.1443,
+      "step": 208160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.570019281646563e-06,
+      "loss": 0.1417,
+      "step": 208170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5677772297206402e-06,
+      "loss": 0.1339,
+      "step": 208180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.565535177794718e-06,
+      "loss": 0.1386,
+      "step": 208190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5632931258687955e-06,
+      "loss": 0.1392,
+      "step": 208200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5610510739428726e-06,
+      "loss": 0.1381,
+      "step": 208210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.55880902201695e-06,
+      "loss": 0.1354,
+      "step": 208220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.556566970091028e-06,
+      "loss": 0.1403,
+      "step": 208230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.554324918165105e-06,
+      "loss": 0.1368,
+      "step": 208240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.552082866239182e-06,
+      "loss": 0.142,
+      "step": 208250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.54984081431326e-06,
+      "loss": 0.1355,
+      "step": 208260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5475987623873372e-06,
+      "loss": 0.1402,
+      "step": 208270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5453567104614143e-06,
+      "loss": 0.1408,
+      "step": 208280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5431146585354917e-06,
+      "loss": 0.1351,
+      "step": 208290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5408726066095696e-06,
+      "loss": 0.1313,
+      "step": 208300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5386305546836466e-06,
+      "loss": 0.1365,
+      "step": 208310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.536388502757724e-06,
+      "loss": 0.14,
+      "step": 208320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.534146450831802e-06,
+      "loss": 0.1364,
+      "step": 208330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.531904398905879e-06,
+      "loss": 0.1391,
+      "step": 208340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.529662346979956e-06,
+      "loss": 0.1355,
+      "step": 208350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5274202950540334e-06,
+      "loss": 0.1349,
+      "step": 208360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5251782431281113e-06,
+      "loss": 0.1405,
+      "step": 208370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5229361912021883e-06,
+      "loss": 0.1375,
+      "step": 208380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5206941392762658e-06,
+      "loss": 0.1381,
+      "step": 208390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5184520873503436e-06,
+      "loss": 0.1352,
+      "step": 208400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5162100354244207e-06,
+      "loss": 0.1409,
+      "step": 208410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5139679834984977e-06,
+      "loss": 0.1382,
+      "step": 208420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.511725931572575e-06,
+      "loss": 0.1338,
+      "step": 208430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.509483879646653e-06,
+      "loss": 0.1375,
+      "step": 208440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.50724182772073e-06,
+      "loss": 0.1346,
+      "step": 208450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5049997757948075e-06,
+      "loss": 0.1387,
+      "step": 208460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5027577238688853e-06,
+      "loss": 0.1367,
+      "step": 208470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5005156719429624e-06,
+      "loss": 0.1383,
+      "step": 208480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.49827362001704e-06,
+      "loss": 0.136,
+      "step": 208490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.496031568091117e-06,
+      "loss": 0.1361,
+      "step": 208500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4937895161651947e-06,
+      "loss": 0.1374,
+      "step": 208510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4915474642392717e-06,
+      "loss": 0.1382,
+      "step": 208520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.489305412313349e-06,
+      "loss": 0.1427,
+      "step": 208530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.487063360387427e-06,
+      "loss": 0.1353,
+      "step": 208540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.484821308461504e-06,
+      "loss": 0.1365,
+      "step": 208550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4825792565355815e-06,
+      "loss": 0.1374,
+      "step": 208560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4803372046096594e-06,
+      "loss": 0.1369,
+      "step": 208570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4780951526837364e-06,
+      "loss": 0.1419,
+      "step": 208580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4758531007578134e-06,
+      "loss": 0.1425,
+      "step": 208590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.473611048831891e-06,
+      "loss": 0.1407,
+      "step": 208600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4713689969059688e-06,
+      "loss": 0.1402,
+      "step": 208610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4691269449800458e-06,
+      "loss": 0.1361,
+      "step": 208620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4668848930541232e-06,
+      "loss": 0.137,
+      "step": 208630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.464642841128201e-06,
+      "loss": 0.1354,
+      "step": 208640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.462400789202278e-06,
+      "loss": 0.1392,
+      "step": 208650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4601587372763556e-06,
+      "loss": 0.1319,
+      "step": 208660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4579166853504326e-06,
+      "loss": 0.1422,
+      "step": 208670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4556746334245105e-06,
+      "loss": 0.1355,
+      "step": 208680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4534325814985875e-06,
+      "loss": 0.1353,
+      "step": 208690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.451190529572665e-06,
+      "loss": 0.1424,
+      "step": 208700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.448948477646743e-06,
+      "loss": 0.1368,
+      "step": 208710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.44670642572082e-06,
+      "loss": 0.1388,
+      "step": 208720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4444643737948973e-06,
+      "loss": 0.1304,
+      "step": 208730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4422223218689743e-06,
+      "loss": 0.1421,
+      "step": 208740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.439980269943052e-06,
+      "loss": 0.1322,
+      "step": 208750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4377382180171296e-06,
+      "loss": 0.1366,
+      "step": 208760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4354961660912066e-06,
+      "loss": 0.1372,
+      "step": 208770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4332541141652845e-06,
+      "loss": 0.1356,
+      "step": 208780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4310120622393615e-06,
+      "loss": 0.1358,
+      "step": 208790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.428770010313439e-06,
+      "loss": 0.1396,
+      "step": 208800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.426527958387516e-06,
+      "loss": 0.1389,
+      "step": 208810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.424285906461594e-06,
+      "loss": 0.1382,
+      "step": 208820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4220438545356713e-06,
+      "loss": 0.1421,
+      "step": 208830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4198018026097483e-06,
+      "loss": 0.1417,
+      "step": 208840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4175597506838262e-06,
+      "loss": 0.1365,
+      "step": 208850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4153176987579032e-06,
+      "loss": 0.1296,
+      "step": 208860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4130756468319807e-06,
+      "loss": 0.1427,
+      "step": 208870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4108335949060586e-06,
+      "loss": 0.142,
+      "step": 208880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4085915429801356e-06,
+      "loss": 0.1385,
+      "step": 208890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.406349491054213e-06,
+      "loss": 0.1368,
+      "step": 208900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.40410743912829e-06,
+      "loss": 0.1423,
+      "step": 208910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.401865387202368e-06,
+      "loss": 0.1395,
+      "step": 208920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3996233352764454e-06,
+      "loss": 0.1346,
+      "step": 208930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3973812833505224e-06,
+      "loss": 0.1335,
+      "step": 208940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3951392314246003e-06,
+      "loss": 0.1365,
+      "step": 208950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3928971794986773e-06,
+      "loss": 0.1445,
+      "step": 208960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3906551275727547e-06,
+      "loss": 0.1307,
+      "step": 208970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3884130756468318e-06,
+      "loss": 0.1372,
+      "step": 208980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3861710237209096e-06,
+      "loss": 0.1333,
+      "step": 208990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.383928971794987e-06,
+      "loss": 0.1372,
+      "step": 209000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.381686919869064e-06,
+      "loss": 0.1435,
+      "step": 209010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.379444867943142e-06,
+      "loss": 0.1399,
+      "step": 209020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3772028160172194e-06,
+      "loss": 0.1383,
+      "step": 209030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3749607640912964e-06,
+      "loss": 0.1395,
+      "step": 209040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3727187121653735e-06,
+      "loss": 0.1406,
+      "step": 209050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3704766602394513e-06,
+      "loss": 0.1335,
+      "step": 209060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3682346083135288e-06,
+      "loss": 0.1366,
+      "step": 209070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.365992556387606e-06,
+      "loss": 0.131,
+      "step": 209080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3637505044616837e-06,
+      "loss": 0.1377,
+      "step": 209090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.361508452535761e-06,
+      "loss": 0.1386,
+      "step": 209100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.359266400609838e-06,
+      "loss": 0.1415,
+      "step": 209110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.357024348683915e-06,
+      "loss": 0.1379,
+      "step": 209120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.354782296757993e-06,
+      "loss": 0.1345,
+      "step": 209130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3525402448320705e-06,
+      "loss": 0.1403,
+      "step": 209140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3502981929061475e-06,
+      "loss": 0.1375,
+      "step": 209150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3480561409802254e-06,
+      "loss": 0.1358,
+      "step": 209160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.345814089054303e-06,
+      "loss": 0.1402,
+      "step": 209170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.34357203712838e-06,
+      "loss": 0.1416,
+      "step": 209180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3413299852024573e-06,
+      "loss": 0.138,
+      "step": 209190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.339087933276535e-06,
+      "loss": 0.1321,
+      "step": 209200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.336845881350612e-06,
+      "loss": 0.1321,
+      "step": 209210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3346038294246892e-06,
+      "loss": 0.1414,
+      "step": 209220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.332361777498767e-06,
+      "loss": 0.1344,
+      "step": 209230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3301197255728445e-06,
+      "loss": 0.1364,
+      "step": 209240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3278776736469216e-06,
+      "loss": 0.1349,
+      "step": 209250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3256356217209994e-06,
+      "loss": 0.1411,
+      "step": 209260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.323393569795077e-06,
+      "loss": 0.1394,
+      "step": 209270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.321151517869154e-06,
+      "loss": 0.1398,
+      "step": 209280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.318909465943231e-06,
+      "loss": 0.1369,
+      "step": 209290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3166674140173092e-06,
+      "loss": 0.1374,
+      "step": 209300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3144253620913862e-06,
+      "loss": 0.1423,
+      "step": 209310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3121833101654633e-06,
+      "loss": 0.1381,
+      "step": 209320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.309941258239541e-06,
+      "loss": 0.1393,
+      "step": 209330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3076992063136186e-06,
+      "loss": 0.1362,
+      "step": 209340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3054571543876956e-06,
+      "loss": 0.1357,
+      "step": 209350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.303215102461773e-06,
+      "loss": 0.1366,
+      "step": 209360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.300973050535851e-06,
+      "loss": 0.1368,
+      "step": 209370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.298730998609928e-06,
+      "loss": 0.1389,
+      "step": 209380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.296488946684005e-06,
+      "loss": 0.1334,
+      "step": 209390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.294246894758083e-06,
+      "loss": 0.1391,
+      "step": 209400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2920048428321603e-06,
+      "loss": 0.1383,
+      "step": 209410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2897627909062373e-06,
+      "loss": 0.1364,
+      "step": 209420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2875207389803148e-06,
+      "loss": 0.1398,
+      "step": 209430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2852786870543926e-06,
+      "loss": 0.1399,
+      "step": 209440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2830366351284697e-06,
+      "loss": 0.1389,
+      "step": 209450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.280794583202547e-06,
+      "loss": 0.1344,
+      "step": 209460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.278552531276625e-06,
+      "loss": 0.1377,
+      "step": 209470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.276310479350702e-06,
+      "loss": 0.1405,
+      "step": 209480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.274068427424779e-06,
+      "loss": 0.1445,
+      "step": 209490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2718263754988565e-06,
+      "loss": 0.1369,
+      "step": 209500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2695843235729343e-06,
+      "loss": 0.1375,
+      "step": 209510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2673422716470114e-06,
+      "loss": 0.1409,
+      "step": 209520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.265100219721089e-06,
+      "loss": 0.1401,
+      "step": 209530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2630823729877587e-06,
+      "loss": 0.134,
+      "step": 209540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2608403210618358e-06,
+      "loss": 0.1363,
+      "step": 209550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.258598269135913e-06,
+      "loss": 0.1383,
+      "step": 209560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.256356217209991e-06,
+      "loss": 0.1416,
+      "step": 209570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.254114165284068e-06,
+      "loss": 0.1341,
+      "step": 209580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2518721133581455e-06,
+      "loss": 0.1399,
+      "step": 209590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2496300614322234e-06,
+      "loss": 0.1329,
+      "step": 209600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2473880095063004e-06,
+      "loss": 0.1362,
+      "step": 209610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2451459575803775e-06,
+      "loss": 0.1351,
+      "step": 209620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.242903905654455e-06,
+      "loss": 0.1385,
+      "step": 209630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2406618537285328e-06,
+      "loss": 0.1362,
+      "step": 209640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.23841980180261e-06,
+      "loss": 0.1368,
+      "step": 209650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2361777498766873e-06,
+      "loss": 0.1379,
+      "step": 209660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.233935697950765e-06,
+      "loss": 0.1345,
+      "step": 209670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.231693646024842e-06,
+      "loss": 0.1341,
+      "step": 209680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.229451594098919e-06,
+      "loss": 0.1416,
+      "step": 209690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2272095421729966e-06,
+      "loss": 0.1423,
+      "step": 209700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2249674902470745e-06,
+      "loss": 0.1362,
+      "step": 209710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2227254383211515e-06,
+      "loss": 0.1387,
+      "step": 209720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.220483386395229e-06,
+      "loss": 0.1375,
+      "step": 209730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.218241334469307e-06,
+      "loss": 0.1364,
+      "step": 209740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.215999282543384e-06,
+      "loss": 0.138,
+      "step": 209750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2137572306174613e-06,
+      "loss": 0.1444,
+      "step": 209760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2115151786915383e-06,
+      "loss": 0.1408,
+      "step": 209770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.209273126765616e-06,
+      "loss": 0.1366,
+      "step": 209780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2070310748396932e-06,
+      "loss": 0.14,
+      "step": 209790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2047890229137707e-06,
+      "loss": 0.1323,
+      "step": 209800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2025469709878485e-06,
+      "loss": 0.1329,
+      "step": 209810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.2003049190619256e-06,
+      "loss": 0.133,
+      "step": 209820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.198062867136003e-06,
+      "loss": 0.1349,
+      "step": 209830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.195820815210081e-06,
+      "loss": 0.137,
+      "step": 209840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.193578763284158e-06,
+      "loss": 0.1371,
+      "step": 209850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1913367113582353e-06,
+      "loss": 0.1405,
+      "step": 209860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1890946594323124e-06,
+      "loss": 0.1363,
+      "step": 209870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1868526075063902e-06,
+      "loss": 0.1373,
+      "step": 209880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1846105555804673e-06,
+      "loss": 0.1406,
+      "step": 209890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1823685036545447e-06,
+      "loss": 0.1349,
+      "step": 209900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1801264517286226e-06,
+      "loss": 0.1389,
+      "step": 209910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1778843998026996e-06,
+      "loss": 0.1393,
+      "step": 209920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.175642347876777e-06,
+      "loss": 0.1361,
+      "step": 209930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.173400295950854e-06,
+      "loss": 0.1346,
+      "step": 209940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.171158244024932e-06,
+      "loss": 0.1369,
+      "step": 209950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.168916192099009e-06,
+      "loss": 0.1358,
+      "step": 209960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1666741401730864e-06,
+      "loss": 0.1373,
+      "step": 209970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1644320882471643e-06,
+      "loss": 0.1398,
+      "step": 209980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1621900363212413e-06,
+      "loss": 0.1351,
+      "step": 209990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1599479843953188e-06,
+      "loss": 0.1321,
+      "step": 210000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1577059324693958e-06,
+      "loss": 0.1385,
+      "step": 210010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1554638805434737e-06,
+      "loss": 0.1401,
+      "step": 210020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.153221828617551e-06,
+      "loss": 0.1335,
+      "step": 210030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.150979776691628e-06,
+      "loss": 0.1353,
+      "step": 210040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.148737724765706e-06,
+      "loss": 0.1376,
+      "step": 210050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.146495672839783e-06,
+      "loss": 0.1346,
+      "step": 210060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1442536209138605e-06,
+      "loss": 0.1343,
+      "step": 210070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1420115689879375e-06,
+      "loss": 0.1324,
+      "step": 210080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1397695170620154e-06,
+      "loss": 0.1412,
+      "step": 210090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.137527465136093e-06,
+      "loss": 0.1343,
+      "step": 210100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.13528541321017e-06,
+      "loss": 0.133,
+      "step": 210110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1330433612842477e-06,
+      "loss": 0.1384,
+      "step": 210120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.130801309358325e-06,
+      "loss": 0.1361,
+      "step": 210130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.128559257432402e-06,
+      "loss": 0.1361,
+      "step": 210140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.12631720550648e-06,
+      "loss": 0.1408,
+      "step": 210150
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.124075153580557e-06,
+      "loss": 0.1346,
+      "step": 210160
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1218331016546345e-06,
+      "loss": 0.1357,
+      "step": 210170
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.119591049728712e-06,
+      "loss": 0.139,
+      "step": 210180
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.117348997802789e-06,
+      "loss": 0.1394,
+      "step": 210190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.115106945876867e-06,
+      "loss": 0.1361,
+      "step": 210200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.112864893950944e-06,
+      "loss": 0.1388,
+      "step": 210210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1106228420250213e-06,
+      "loss": 0.1423,
+      "step": 210220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1083807900990988e-06,
+      "loss": 0.1374,
+      "step": 210230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1061387381731762e-06,
+      "loss": 0.1403,
+      "step": 210240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1038966862472537e-06,
+      "loss": 0.1341,
+      "step": 210250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1016546343213307e-06,
+      "loss": 0.1382,
+      "step": 210260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0994125823954086e-06,
+      "loss": 0.1364,
+      "step": 210270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.097170530469486e-06,
+      "loss": 0.1361,
+      "step": 210280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.094928478543563e-06,
+      "loss": 0.1367,
+      "step": 210290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.092686426617641e-06,
+      "loss": 0.1398,
+      "step": 210300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.090444374691718e-06,
+      "loss": 0.1342,
+      "step": 210310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0882023227657954e-06,
+      "loss": 0.1296,
+      "step": 210320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.085960270839873e-06,
+      "loss": 0.1336,
+      "step": 210330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0837182189139503e-06,
+      "loss": 0.1417,
+      "step": 210340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0814761669880277e-06,
+      "loss": 0.1405,
+      "step": 210350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0792341150621047e-06,
+      "loss": 0.1343,
+      "step": 210360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0769920631361826e-06,
+      "loss": 0.1361,
+      "step": 210370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0747500112102596e-06,
+      "loss": 0.137,
+      "step": 210380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.072507959284337e-06,
+      "loss": 0.1432,
+      "step": 210390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0702659073584145e-06,
+      "loss": 0.1342,
+      "step": 210400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.068023855432492e-06,
+      "loss": 0.1353,
+      "step": 210410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0657818035065694e-06,
+      "loss": 0.1367,
+      "step": 210420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.063539751580647e-06,
+      "loss": 0.1333,
+      "step": 210430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0612976996547243e-06,
+      "loss": 0.1404,
+      "step": 210440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0590556477288013e-06,
+      "loss": 0.1366,
+      "step": 210450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.056813595802879e-06,
+      "loss": 0.138,
+      "step": 210460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0545715438769567e-06,
+      "loss": 0.1377,
+      "step": 210470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0523294919510337e-06,
+      "loss": 0.1363,
+      "step": 210480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.050087440025111e-06,
+      "loss": 0.1371,
+      "step": 210490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0478453880991886e-06,
+      "loss": 0.142,
+      "step": 210500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.045603336173266e-06,
+      "loss": 0.1338,
+      "step": 210510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.043361284247343e-06,
+      "loss": 0.1443,
+      "step": 210520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0411192323214205e-06,
+      "loss": 0.1364,
+      "step": 210530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0388771803954984e-06,
+      "loss": 0.138,
+      "step": 210540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0366351284695754e-06,
+      "loss": 0.139,
+      "step": 210550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.034393076543653e-06,
+      "loss": 0.1373,
+      "step": 210560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0321510246177303e-06,
+      "loss": 0.1325,
+      "step": 210570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0299089726918077e-06,
+      "loss": 0.1412,
+      "step": 210580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.027666920765885e-06,
+      "loss": 0.1405,
+      "step": 210590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0254248688399626e-06,
+      "loss": 0.1388,
+      "step": 210600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.02318281691404e-06,
+      "loss": 0.1331,
+      "step": 210610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.020940764988117e-06,
+      "loss": 0.1428,
+      "step": 210620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0186987130621945e-06,
+      "loss": 0.1364,
+      "step": 210630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.016456661136272e-06,
+      "loss": 0.1321,
+      "step": 210640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0142146092103494e-06,
+      "loss": 0.1375,
+      "step": 210650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.011972557284427e-06,
+      "loss": 0.1353,
+      "step": 210660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0097305053585043e-06,
+      "loss": 0.1387,
+      "step": 210670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0074884534325818e-06,
+      "loss": 0.1379,
+      "step": 210680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.005246401506659e-06,
+      "loss": 0.1379,
+      "step": 210690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0030043495807367e-06,
+      "loss": 0.1315,
+      "step": 210700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0007622976548137e-06,
+      "loss": 0.1338,
+      "step": 210710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.998520245728891e-06,
+      "loss": 0.1394,
+      "step": 210720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9962781938029686e-06,
+      "loss": 0.1331,
+      "step": 210730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.994036141877046e-06,
+      "loss": 0.1319,
+      "step": 210740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9917940899511235e-06,
+      "loss": 0.1345,
+      "step": 210750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9895520380252005e-06,
+      "loss": 0.1314,
+      "step": 210760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9873099860992784e-06,
+      "loss": 0.143,
+      "step": 210770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.985067934173356e-06,
+      "loss": 0.1397,
+      "step": 210780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.982825882247433e-06,
+      "loss": 0.137,
+      "step": 210790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9805838303215103e-06,
+      "loss": 0.1408,
+      "step": 210800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9783417783955878e-06,
+      "loss": 0.1366,
+      "step": 210810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.976099726469665e-06,
+      "loss": 0.1402,
+      "step": 210820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9738576745437422e-06,
+      "loss": 0.1362,
+      "step": 210830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.97161562261782e-06,
+      "loss": 0.1351,
+      "step": 210840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9693735706918975e-06,
+      "loss": 0.1351,
+      "step": 210850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9671315187659746e-06,
+      "loss": 0.1374,
+      "step": 210860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9648894668400524e-06,
+      "loss": 0.1432,
+      "step": 210870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9626474149141295e-06,
+      "loss": 0.1333,
+      "step": 210880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.960405362988207e-06,
+      "loss": 0.1311,
+      "step": 210890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9581633110622844e-06,
+      "loss": 0.1374,
+      "step": 210900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.955921259136362e-06,
+      "loss": 0.1341,
+      "step": 210910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9536792072104392e-06,
+      "loss": 0.1358,
+      "step": 210920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9514371552845163e-06,
+      "loss": 0.1396,
+      "step": 210930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.949195103358594e-06,
+      "loss": 0.1326,
+      "step": 210940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.946953051432671e-06,
+      "loss": 0.1369,
+      "step": 210950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9447109995067486e-06,
+      "loss": 0.1338,
+      "step": 210960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.942468947580826e-06,
+      "loss": 0.1359,
+      "step": 210970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9402268956549035e-06,
+      "loss": 0.1348,
+      "step": 210980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.937984843728981e-06,
+      "loss": 0.1335,
+      "step": 210990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9357427918030584e-06,
+      "loss": 0.1367,
+      "step": 211000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.933500739877136e-06,
+      "loss": 0.136,
+      "step": 211010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.931258687951213e-06,
+      "loss": 0.1409,
+      "step": 211020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9290166360252903e-06,
+      "loss": 0.1398,
+      "step": 211030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.926774584099368e-06,
+      "loss": 0.138,
+      "step": 211040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9245325321734452e-06,
+      "loss": 0.1359,
+      "step": 211050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9222904802475227e-06,
+      "loss": 0.1352,
+      "step": 211060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9200484283216e-06,
+      "loss": 0.1394,
+      "step": 211070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9178063763956776e-06,
+      "loss": 0.1376,
+      "step": 211080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.915564324469755e-06,
+      "loss": 0.1396,
+      "step": 211090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.913322272543832e-06,
+      "loss": 0.1391,
+      "step": 211100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.91108022061791e-06,
+      "loss": 0.1328,
+      "step": 211110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.908838168691987e-06,
+      "loss": 0.1268,
+      "step": 211120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9065961167660644e-06,
+      "loss": 0.1367,
+      "step": 211130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.904354064840142e-06,
+      "loss": 0.1401,
+      "step": 211140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9021120129142193e-06,
+      "loss": 0.1339,
+      "step": 211150
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8998699609882967e-06,
+      "loss": 0.1349,
+      "step": 211160
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.897627909062374e-06,
+      "loss": 0.1393,
+      "step": 211170
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8953858571364516e-06,
+      "loss": 0.1345,
+      "step": 211180
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8931438052105286e-06,
+      "loss": 0.1392,
+      "step": 211190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.890901753284606e-06,
+      "loss": 0.1395,
+      "step": 211200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8886597013586835e-06,
+      "loss": 0.1337,
+      "step": 211210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.886417649432761e-06,
+      "loss": 0.1385,
+      "step": 211220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8841755975068384e-06,
+      "loss": 0.1428,
+      "step": 211230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.881933545580916e-06,
+      "loss": 0.1291,
+      "step": 211240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8796914936549933e-06,
+      "loss": 0.1387,
+      "step": 211250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8774494417290703e-06,
+      "loss": 0.1349,
+      "step": 211260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.875207389803148e-06,
+      "loss": 0.1401,
+      "step": 211270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8729653378772257e-06,
+      "loss": 0.1375,
+      "step": 211280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8707232859513027e-06,
+      "loss": 0.1359,
+      "step": 211290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.86848123402538e-06,
+      "loss": 0.1379,
+      "step": 211300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8662391820994576e-06,
+      "loss": 0.131,
+      "step": 211310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.863997130173535e-06,
+      "loss": 0.14,
+      "step": 211320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.861755078247612e-06,
+      "loss": 0.1355,
+      "step": 211330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.85951302632169e-06,
+      "loss": 0.1393,
+      "step": 211340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8572709743957674e-06,
+      "loss": 0.1342,
+      "step": 211350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8550289224698444e-06,
+      "loss": 0.1377,
+      "step": 211360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.852786870543922e-06,
+      "loss": 0.1345,
+      "step": 211370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8505448186179993e-06,
+      "loss": 0.1359,
+      "step": 211380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8483027666920767e-06,
+      "loss": 0.1314,
+      "step": 211390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8460607147661537e-06,
+      "loss": 0.1356,
+      "step": 211400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8438186628402316e-06,
+      "loss": 0.1346,
+      "step": 211410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.841576610914309e-06,
+      "loss": 0.1314,
+      "step": 211420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.839334558988386e-06,
+      "loss": 0.135,
+      "step": 211430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.837092507062464e-06,
+      "loss": 0.1355,
+      "step": 211440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.834850455136541e-06,
+      "loss": 0.1339,
+      "step": 211450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8326084032106184e-06,
+      "loss": 0.139,
+      "step": 211460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.830366351284696e-06,
+      "loss": 0.1419,
+      "step": 211470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8281242993587733e-06,
+      "loss": 0.1428,
+      "step": 211480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8258822474328508e-06,
+      "loss": 0.1363,
+      "step": 211490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.823640195506928e-06,
+      "loss": 0.143,
+      "step": 211500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8213981435810057e-06,
+      "loss": 0.14,
+      "step": 211510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8191560916550827e-06,
+      "loss": 0.1369,
+      "step": 211520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.81691403972916e-06,
+      "loss": 0.1366,
+      "step": 211530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8146719878032376e-06,
+      "loss": 0.1402,
+      "step": 211540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8126541410699075e-06,
+      "loss": 0.1399,
+      "step": 211550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8104120891439845e-06,
+      "loss": 0.1433,
+      "step": 211560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8081700372180624e-06,
+      "loss": 0.1363,
+      "step": 211570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8059279852921394e-06,
+      "loss": 0.1382,
+      "step": 211580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.803685933366217e-06,
+      "loss": 0.1332,
+      "step": 211590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8014438814402943e-06,
+      "loss": 0.1406,
+      "step": 211600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7992018295143718e-06,
+      "loss": 0.137,
+      "step": 211610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.796959777588449e-06,
+      "loss": 0.1377,
+      "step": 211620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7947177256625262e-06,
+      "loss": 0.1355,
+      "step": 211630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.792475673736604e-06,
+      "loss": 0.1409,
+      "step": 211640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.790233621810681e-06,
+      "loss": 0.1364,
+      "step": 211650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7879915698847586e-06,
+      "loss": 0.1394,
+      "step": 211660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.785749517958836e-06,
+      "loss": 0.1357,
+      "step": 211670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7835074660329135e-06,
+      "loss": 0.1415,
+      "step": 211680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.781265414106991e-06,
+      "loss": 0.1395,
+      "step": 211690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7790233621810684e-06,
+      "loss": 0.1406,
+      "step": 211700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.776781310255146e-06,
+      "loss": 0.1391,
+      "step": 211710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.774539258329223e-06,
+      "loss": 0.1317,
+      "step": 211720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7722972064033003e-06,
+      "loss": 0.1397,
+      "step": 211730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.770055154477378e-06,
+      "loss": 0.137,
+      "step": 211740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.767813102551455e-06,
+      "loss": 0.1348,
+      "step": 211750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7655710506255326e-06,
+      "loss": 0.1372,
+      "step": 211760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.76332899869961e-06,
+      "loss": 0.1375,
+      "step": 211770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7610869467736875e-06,
+      "loss": 0.14,
+      "step": 211780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7588448948477645e-06,
+      "loss": 0.14,
+      "step": 211790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.756602842921842e-06,
+      "loss": 0.1327,
+      "step": 211800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.75436079099592e-06,
+      "loss": 0.1417,
+      "step": 211810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.752118739069997e-06,
+      "loss": 0.1377,
+      "step": 211820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7498766871440743e-06,
+      "loss": 0.1347,
+      "step": 211830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7476346352181518e-06,
+      "loss": 0.1341,
+      "step": 211840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7453925832922292e-06,
+      "loss": 0.1383,
+      "step": 211850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7431505313663067e-06,
+      "loss": 0.137,
+      "step": 211860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.740908479440384e-06,
+      "loss": 0.1356,
+      "step": 211870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7386664275144616e-06,
+      "loss": 0.1412,
+      "step": 211880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7364243755885386e-06,
+      "loss": 0.1386,
+      "step": 211890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.734182323662616e-06,
+      "loss": 0.1377,
+      "step": 211900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7319402717366935e-06,
+      "loss": 0.1386,
+      "step": 211910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.729698219810771e-06,
+      "loss": 0.1326,
+      "step": 211920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7274561678848484e-06,
+      "loss": 0.1361,
+      "step": 211930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.725214115958926e-06,
+      "loss": 0.1358,
+      "step": 211940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7229720640330033e-06,
+      "loss": 0.1365,
+      "step": 211950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7207300121070803e-06,
+      "loss": 0.1384,
+      "step": 211960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718487960181158e-06,
+      "loss": 0.1412,
+      "step": 211970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.716245908255235e-06,
+      "loss": 0.1365,
+      "step": 211980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7140038563293126e-06,
+      "loss": 0.14,
+      "step": 211990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.71176180440339e-06,
+      "loss": 0.1362,
+      "step": 212000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7095197524774675e-06,
+      "loss": 0.1373,
+      "step": 212010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.707277700551545e-06,
+      "loss": 0.1406,
+      "step": 212020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.705035648625622e-06,
+      "loss": 0.1386,
+      "step": 212030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7027935966997e-06,
+      "loss": 0.1383,
+      "step": 212040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7005515447737773e-06,
+      "loss": 0.1357,
+      "step": 212050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6983094928478543e-06,
+      "loss": 0.1357,
+      "step": 212060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6960674409219318e-06,
+      "loss": 0.1408,
+      "step": 212070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6938253889960092e-06,
+      "loss": 0.1408,
+      "step": 212080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6915833370700867e-06,
+      "loss": 0.1374,
+      "step": 212090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.689341285144164e-06,
+      "loss": 0.1377,
+      "step": 212100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6870992332182416e-06,
+      "loss": 0.1388,
+      "step": 212110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.684857181292319e-06,
+      "loss": 0.137,
+      "step": 212120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.682615129366396e-06,
+      "loss": 0.1337,
+      "step": 212130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.680373077440474e-06,
+      "loss": 0.1344,
+      "step": 212140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.678131025514551e-06,
+      "loss": 0.1355,
+      "step": 212150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6758889735886284e-06,
+      "loss": 0.1321,
+      "step": 212160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.673646921662706e-06,
+      "loss": 0.1346,
+      "step": 212170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6714048697367833e-06,
+      "loss": 0.1284,
+      "step": 212180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6691628178108607e-06,
+      "loss": 0.1346,
+      "step": 212190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6669207658849378e-06,
+      "loss": 0.1378,
+      "step": 212200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6646787139590156e-06,
+      "loss": 0.132,
+      "step": 212210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6624366620330926e-06,
+      "loss": 0.1371,
+      "step": 212220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.66019461010717e-06,
+      "loss": 0.1375,
+      "step": 212230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.657952558181248e-06,
+      "loss": 0.1379,
+      "step": 212240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.655710506255325e-06,
+      "loss": 0.137,
+      "step": 212250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6534684543294024e-06,
+      "loss": 0.1381,
+      "step": 212260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.65122640240348e-06,
+      "loss": 0.1369,
+      "step": 212270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6489843504775573e-06,
+      "loss": 0.142,
+      "step": 212280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6467422985516344e-06,
+      "loss": 0.1319,
+      "step": 212290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.644500246625712e-06,
+      "loss": 0.139,
+      "step": 212300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6422581946997897e-06,
+      "loss": 0.1368,
+      "step": 212310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6400161427738667e-06,
+      "loss": 0.1401,
+      "step": 212320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.637774090847944e-06,
+      "loss": 0.1412,
+      "step": 212330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6355320389220216e-06,
+      "loss": 0.1335,
+      "step": 212340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.633289986996099e-06,
+      "loss": 0.1347,
+      "step": 212350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6310479350701765e-06,
+      "loss": 0.1344,
+      "step": 212360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6288058831442535e-06,
+      "loss": 0.1398,
+      "step": 212370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6265638312183314e-06,
+      "loss": 0.1357,
+      "step": 212380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6243217792924084e-06,
+      "loss": 0.137,
+      "step": 212390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.622079727366486e-06,
+      "loss": 0.1362,
+      "step": 212400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6198376754405633e-06,
+      "loss": 0.1391,
+      "step": 212410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6175956235146407e-06,
+      "loss": 0.1393,
+      "step": 212420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.615353571588718e-06,
+      "loss": 0.1384,
+      "step": 212430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6131115196627956e-06,
+      "loss": 0.1379,
+      "step": 212440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.610869467736873e-06,
+      "loss": 0.1359,
+      "step": 212450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.60862741581095e-06,
+      "loss": 0.1373,
+      "step": 212460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6063853638850276e-06,
+      "loss": 0.1414,
+      "step": 212470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.604143311959105e-06,
+      "loss": 0.1382,
+      "step": 212480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6019012600331825e-06,
+      "loss": 0.1361,
+      "step": 212490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.59965920810726e-06,
+      "loss": 0.1383,
+      "step": 212500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5974171561813373e-06,
+      "loss": 0.1357,
+      "step": 212510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.595175104255415e-06,
+      "loss": 0.1406,
+      "step": 212520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.592933052329492e-06,
+      "loss": 0.138,
+      "step": 212530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5906910004035697e-06,
+      "loss": 0.1342,
+      "step": 212540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.588448948477647e-06,
+      "loss": 0.1352,
+      "step": 212550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.586206896551724e-06,
+      "loss": 0.1388,
+      "step": 212560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5839648446258016e-06,
+      "loss": 0.1394,
+      "step": 212570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.581722792699879e-06,
+      "loss": 0.1402,
+      "step": 212580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5794807407739565e-06,
+      "loss": 0.1339,
+      "step": 212590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5772386888480335e-06,
+      "loss": 0.134,
+      "step": 212600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5749966369221114e-06,
+      "loss": 0.1394,
+      "step": 212610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.572754584996189e-06,
+      "loss": 0.1423,
+      "step": 212620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.570512533070266e-06,
+      "loss": 0.142,
+      "step": 212630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5682704811443433e-06,
+      "loss": 0.1375,
+      "step": 212640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5660284292184208e-06,
+      "loss": 0.1371,
+      "step": 212650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.563786377292498e-06,
+      "loss": 0.1417,
+      "step": 212660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5615443253665757e-06,
+      "loss": 0.1372,
+      "step": 212670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.559302273440653e-06,
+      "loss": 0.1387,
+      "step": 212680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5570602215147305e-06,
+      "loss": 0.1389,
+      "step": 212690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5548181695888076e-06,
+      "loss": 0.1353,
+      "step": 212700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5525761176628854e-06,
+      "loss": 0.1374,
+      "step": 212710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5503340657369625e-06,
+      "loss": 0.1354,
+      "step": 212720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.54809201381104e-06,
+      "loss": 0.1338,
+      "step": 212730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5458499618851174e-06,
+      "loss": 0.1359,
+      "step": 212740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.543607909959195e-06,
+      "loss": 0.1355,
+      "step": 212750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5413658580332723e-06,
+      "loss": 0.1407,
+      "step": 212760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5391238061073493e-06,
+      "loss": 0.1356,
+      "step": 212770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.536881754181427e-06,
+      "loss": 0.1344,
+      "step": 212780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.534639702255504e-06,
+      "loss": 0.1402,
+      "step": 212790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5323976503295816e-06,
+      "loss": 0.1395,
+      "step": 212800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5301555984036595e-06,
+      "loss": 0.1388,
+      "step": 212810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5279135464777365e-06,
+      "loss": 0.1353,
+      "step": 212820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.525671494551814e-06,
+      "loss": 0.137,
+      "step": 212830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5234294426258914e-06,
+      "loss": 0.1375,
+      "step": 212840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.521187390699969e-06,
+      "loss": 0.1402,
+      "step": 212850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.518945338774046e-06,
+      "loss": 0.134,
+      "step": 212860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5167032868481233e-06,
+      "loss": 0.1365,
+      "step": 212870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.514461234922201e-06,
+      "loss": 0.1391,
+      "step": 212880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5122191829962782e-06,
+      "loss": 0.1368,
+      "step": 212890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5099771310703557e-06,
+      "loss": 0.137,
+      "step": 212900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.507735079144433e-06,
+      "loss": 0.1369,
+      "step": 212910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.5054930272185106e-06,
+      "loss": 0.1411,
+      "step": 212920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.503250975292588e-06,
+      "loss": 0.1387,
+      "step": 212930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.501008923366665e-06,
+      "loss": 0.1378,
+      "step": 212940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.498766871440743e-06,
+      "loss": 0.141,
+      "step": 212950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.49652481951482e-06,
+      "loss": 0.1367,
+      "step": 212960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4942827675888974e-06,
+      "loss": 0.133,
+      "step": 212970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.492040715662975e-06,
+      "loss": 0.142,
+      "step": 212980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4897986637370523e-06,
+      "loss": 0.1411,
+      "step": 212990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4875566118111297e-06,
+      "loss": 0.133,
+      "step": 213000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.485314559885207e-06,
+      "loss": 0.1427,
+      "step": 213010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4830725079592846e-06,
+      "loss": 0.1395,
+      "step": 213020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4808304560333616e-06,
+      "loss": 0.1411,
+      "step": 213030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.478588404107439e-06,
+      "loss": 0.1319,
+      "step": 213040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.476346352181517e-06,
+      "loss": 0.1404,
+      "step": 213050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.474104300255594e-06,
+      "loss": 0.1437,
+      "step": 213060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4718622483296714e-06,
+      "loss": 0.1357,
+      "step": 213070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.469620196403749e-06,
+      "loss": 0.1379,
+      "step": 213080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4673781444778263e-06,
+      "loss": 0.1395,
+      "step": 213090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4651360925519033e-06,
+      "loss": 0.1326,
+      "step": 213100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4628940406259812e-06,
+      "loss": 0.137,
+      "step": 213110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4606519887000587e-06,
+      "loss": 0.1364,
+      "step": 213120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4584099367741357e-06,
+      "loss": 0.1392,
+      "step": 213130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.456167884848213e-06,
+      "loss": 0.1362,
+      "step": 213140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4539258329222906e-06,
+      "loss": 0.1336,
+      "step": 213150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.451683780996368e-06,
+      "loss": 0.1343,
+      "step": 213160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.449441729070445e-06,
+      "loss": 0.1385,
+      "step": 213170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.447199677144523e-06,
+      "loss": 0.1391,
+      "step": 213180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4449576252186004e-06,
+      "loss": 0.1395,
+      "step": 213190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4427155732926774e-06,
+      "loss": 0.1353,
+      "step": 213200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.440473521366755e-06,
+      "loss": 0.1399,
+      "step": 213210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4382314694408323e-06,
+      "loss": 0.1351,
+      "step": 213220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4359894175149097e-06,
+      "loss": 0.1398,
+      "step": 213230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.433747365588987e-06,
+      "loss": 0.1346,
+      "step": 213240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4315053136630646e-06,
+      "loss": 0.1355,
+      "step": 213250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.429263261737142e-06,
+      "loss": 0.1389,
+      "step": 213260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.427021209811219e-06,
+      "loss": 0.1357,
+      "step": 213270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.424779157885297e-06,
+      "loss": 0.1422,
+      "step": 213280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.422537105959374e-06,
+      "loss": 0.135,
+      "step": 213290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4202950540334514e-06,
+      "loss": 0.141,
+      "step": 213300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.418053002107529e-06,
+      "loss": 0.137,
+      "step": 213310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4158109501816063e-06,
+      "loss": 0.1353,
+      "step": 213320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4135688982556838e-06,
+      "loss": 0.1296,
+      "step": 213330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.411326846329761e-06,
+      "loss": 0.1362,
+      "step": 213340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4090847944038387e-06,
+      "loss": 0.1322,
+      "step": 213350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.4068427424779157e-06,
+      "loss": 0.135,
+      "step": 213360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.404600690551993e-06,
+      "loss": 0.133,
+      "step": 213370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.402358638626071e-06,
+      "loss": 0.1362,
+      "step": 213380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.400116586700148e-06,
+      "loss": 0.1331,
+      "step": 213390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3978745347742255e-06,
+      "loss": 0.1358,
+      "step": 213400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.395632482848303e-06,
+      "loss": 0.1349,
+      "step": 213410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3933904309223804e-06,
+      "loss": 0.1386,
+      "step": 213420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.391148378996458e-06,
+      "loss": 0.1338,
+      "step": 213430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.388906327070535e-06,
+      "loss": 0.1354,
+      "step": 213440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3866642751446127e-06,
+      "loss": 0.1391,
+      "step": 213450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3844222232186897e-06,
+      "loss": 0.135,
+      "step": 213460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.382180171292767e-06,
+      "loss": 0.1402,
+      "step": 213470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3799381193668446e-06,
+      "loss": 0.1401,
+      "step": 213480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.377696067440922e-06,
+      "loss": 0.1379,
+      "step": 213490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3754540155149995e-06,
+      "loss": 0.1373,
+      "step": 213500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3732119635890766e-06,
+      "loss": 0.1331,
+      "step": 213510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3709699116631544e-06,
+      "loss": 0.1341,
+      "step": 213520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3687278597372315e-06,
+      "loss": 0.1368,
+      "step": 213530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.366485807811309e-06,
+      "loss": 0.1324,
+      "step": 213540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3642437558853868e-06,
+      "loss": 0.1373,
+      "step": 213550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.362001703959464e-06,
+      "loss": 0.1374,
+      "step": 213560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3597596520335412e-06,
+      "loss": 0.1378,
+      "step": 213570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.357741805300211e-06,
+      "loss": 0.1394,
+      "step": 213580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.355499753374288e-06,
+      "loss": 0.1429,
+      "step": 213590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3532577014483656e-06,
+      "loss": 0.1328,
+      "step": 213600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.351015649522443e-06,
+      "loss": 0.1416,
+      "step": 213610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3487735975965205e-06,
+      "loss": 0.1375,
+      "step": 213620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3465315456705975e-06,
+      "loss": 0.136,
+      "step": 213630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3442894937446754e-06,
+      "loss": 0.142,
+      "step": 213640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.342047441818753e-06,
+      "loss": 0.1345,
+      "step": 213650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.33980538989283e-06,
+      "loss": 0.1387,
+      "step": 213660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3375633379669073e-06,
+      "loss": 0.1363,
+      "step": 213670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3353212860409848e-06,
+      "loss": 0.1342,
+      "step": 213680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3330792341150622e-06,
+      "loss": 0.138,
+      "step": 213690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3308371821891397e-06,
+      "loss": 0.1393,
+      "step": 213700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.328595130263217e-06,
+      "loss": 0.1379,
+      "step": 213710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3263530783372946e-06,
+      "loss": 0.1399,
+      "step": 213720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3241110264113716e-06,
+      "loss": 0.1361,
+      "step": 213730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.321868974485449e-06,
+      "loss": 0.1356,
+      "step": 213740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3196269225595265e-06,
+      "loss": 0.1348,
+      "step": 213750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.317384870633604e-06,
+      "loss": 0.1398,
+      "step": 213760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3151428187076814e-06,
+      "loss": 0.1377,
+      "step": 213770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.312900766781759e-06,
+      "loss": 0.1379,
+      "step": 213780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3106587148558363e-06,
+      "loss": 0.136,
+      "step": 213790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3084166629299133e-06,
+      "loss": 0.136,
+      "step": 213800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.306174611003991e-06,
+      "loss": 0.1441,
+      "step": 213810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3039325590780686e-06,
+      "loss": 0.1364,
+      "step": 213820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.3016905071521456e-06,
+      "loss": 0.1335,
+      "step": 213830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.299448455226223e-06,
+      "loss": 0.1363,
+      "step": 213840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2972064033003005e-06,
+      "loss": 0.1343,
+      "step": 213850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.294964351374378e-06,
+      "loss": 0.1343,
+      "step": 213860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.292722299448455e-06,
+      "loss": 0.1361,
+      "step": 213870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.290480247522533e-06,
+      "loss": 0.1413,
+      "step": 213880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2882381955966103e-06,
+      "loss": 0.1341,
+      "step": 213890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2859961436706874e-06,
+      "loss": 0.1369,
+      "step": 213900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2837540917447652e-06,
+      "loss": 0.1373,
+      "step": 213910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2815120398188422e-06,
+      "loss": 0.1408,
+      "step": 213920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2792699878929197e-06,
+      "loss": 0.1415,
+      "step": 213930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.277027935966997e-06,
+      "loss": 0.1355,
+      "step": 213940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2747858840410746e-06,
+      "loss": 0.1377,
+      "step": 213950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.272543832115152e-06,
+      "loss": 0.1391,
+      "step": 213960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.270301780189229e-06,
+      "loss": 0.1358,
+      "step": 213970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.268059728263307e-06,
+      "loss": 0.1348,
+      "step": 213980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.265817676337384e-06,
+      "loss": 0.1328,
+      "step": 213990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2635756244114614e-06,
+      "loss": 0.1325,
+      "step": 214000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.261333572485539e-06,
+      "loss": 0.1433,
+      "step": 214010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2590915205596163e-06,
+      "loss": 0.1313,
+      "step": 214020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2568494686336937e-06,
+      "loss": 0.1342,
+      "step": 214030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2546074167077708e-06,
+      "loss": 0.1338,
+      "step": 214040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2523653647818486e-06,
+      "loss": 0.1382,
+      "step": 214050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2501233128559257e-06,
+      "loss": 0.1329,
+      "step": 214060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.247881260930003e-06,
+      "loss": 0.1427,
+      "step": 214070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.245639209004081e-06,
+      "loss": 0.1379,
+      "step": 214080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.243397157078158e-06,
+      "loss": 0.1323,
+      "step": 214090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2411551051522354e-06,
+      "loss": 0.1372,
+      "step": 214100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.238913053226313e-06,
+      "loss": 0.1373,
+      "step": 214110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2366710013003903e-06,
+      "loss": 0.1332,
+      "step": 214120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2344289493744674e-06,
+      "loss": 0.1369,
+      "step": 214130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.232186897448545e-06,
+      "loss": 0.1395,
+      "step": 214140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2299448455226227e-06,
+      "loss": 0.1407,
+      "step": 214150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2277027935966997e-06,
+      "loss": 0.1374,
+      "step": 214160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.225460741670777e-06,
+      "loss": 0.1362,
+      "step": 214170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2232186897448546e-06,
+      "loss": 0.1418,
+      "step": 214180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.220976637818932e-06,
+      "loss": 0.1411,
+      "step": 214190
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2187345858930095e-06,
+      "loss": 0.1429,
+      "step": 214200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.216492533967087e-06,
+      "loss": 0.1342,
+      "step": 214210
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2142504820411644e-06,
+      "loss": 0.134,
+      "step": 214220
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2120084301152414e-06,
+      "loss": 0.1365,
+      "step": 214230
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.209766378189319e-06,
+      "loss": 0.1347,
+      "step": 214240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2075243262633963e-06,
+      "loss": 0.1383,
+      "step": 214250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2052822743374738e-06,
+      "loss": 0.1364,
+      "step": 214260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.203040222411551e-06,
+      "loss": 0.1363,
+      "step": 214270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2007981704856287e-06,
+      "loss": 0.1364,
+      "step": 214280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.198556118559706e-06,
+      "loss": 0.1336,
+      "step": 214290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.196314066633783e-06,
+      "loss": 0.1321,
+      "step": 214300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1940720147078606e-06,
+      "loss": 0.1364,
+      "step": 214310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1918299627819384e-06,
+      "loss": 0.1361,
+      "step": 214320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1895879108560155e-06,
+      "loss": 0.14,
+      "step": 214330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.187345858930093e-06,
+      "loss": 0.1337,
+      "step": 214340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1851038070041704e-06,
+      "loss": 0.1301,
+      "step": 214350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.182861755078248e-06,
+      "loss": 0.1407,
+      "step": 214360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.180619703152325e-06,
+      "loss": 0.1333,
+      "step": 214370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1783776512264027e-06,
+      "loss": 0.1422,
+      "step": 214380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.17613559930048e-06,
+      "loss": 0.1371,
+      "step": 214390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.173893547374557e-06,
+      "loss": 0.1374,
+      "step": 214400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1716514954486346e-06,
+      "loss": 0.1425,
+      "step": 214410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.169409443522712e-06,
+      "loss": 0.134,
+      "step": 214420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1671673915967895e-06,
+      "loss": 0.1344,
+      "step": 214430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1649253396708665e-06,
+      "loss": 0.1322,
+      "step": 214440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1626832877449444e-06,
+      "loss": 0.133,
+      "step": 214450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.160441235819022e-06,
+      "loss": 0.1366,
+      "step": 214460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.158199183893099e-06,
+      "loss": 0.1386,
+      "step": 214470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1559571319671767e-06,
+      "loss": 0.1322,
+      "step": 214480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1537150800412538e-06,
+      "loss": 0.138,
+      "step": 214490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1514730281153312e-06,
+      "loss": 0.1317,
+      "step": 214500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1492309761894087e-06,
+      "loss": 0.1362,
+      "step": 214510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.146988924263486e-06,
+      "loss": 0.1353,
+      "step": 214520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1447468723375636e-06,
+      "loss": 0.1364,
+      "step": 214530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1425048204116406e-06,
+      "loss": 0.1367,
+      "step": 214540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1402627684857185e-06,
+      "loss": 0.1383,
+      "step": 214550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1380207165597955e-06,
+      "loss": 0.1355,
+      "step": 214560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.135778664633873e-06,
+      "loss": 0.1401,
+      "step": 214570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1335366127079504e-06,
+      "loss": 0.1346,
+      "step": 214580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.131294560782028e-06,
+      "loss": 0.1352,
+      "step": 214590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1290525088561053e-06,
+      "loss": 0.1369,
+      "step": 214600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1268104569301823e-06,
+      "loss": 0.132,
+      "step": 214610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.12456840500426e-06,
+      "loss": 0.1365,
+      "step": 214620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.122326353078337e-06,
+      "loss": 0.1377,
+      "step": 214630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1200843011524146e-06,
+      "loss": 0.1375,
+      "step": 214640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1178422492264925e-06,
+      "loss": 0.132,
+      "step": 214650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1156001973005695e-06,
+      "loss": 0.1372,
+      "step": 214660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.113358145374647e-06,
+      "loss": 0.1361,
+      "step": 214670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1111160934487244e-06,
+      "loss": 0.1355,
+      "step": 214680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.108874041522802e-06,
+      "loss": 0.1377,
+      "step": 214690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1066319895968793e-06,
+      "loss": 0.1324,
+      "step": 214700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1043899376709563e-06,
+      "loss": 0.1368,
+      "step": 214710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.102147885745034e-06,
+      "loss": 0.1304,
+      "step": 214720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0999058338191112e-06,
+      "loss": 0.1372,
+      "step": 214730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0976637818931887e-06,
+      "loss": 0.1336,
+      "step": 214740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.095421729967266e-06,
+      "loss": 0.1386,
+      "step": 214750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0931796780413436e-06,
+      "loss": 0.1402,
+      "step": 214760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.090937626115421e-06,
+      "loss": 0.1376,
+      "step": 214770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0886955741894985e-06,
+      "loss": 0.1333,
+      "step": 214780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.086453522263576e-06,
+      "loss": 0.1393,
+      "step": 214790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.084211470337653e-06,
+      "loss": 0.1333,
+      "step": 214800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0819694184117304e-06,
+      "loss": 0.1305,
+      "step": 214810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.079727366485808e-06,
+      "loss": 0.1371,
+      "step": 214820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0774853145598853e-06,
+      "loss": 0.1352,
+      "step": 214830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0752432626339627e-06,
+      "loss": 0.1375,
+      "step": 214840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.07300121070804e-06,
+      "loss": 0.1354,
+      "step": 214850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0707591587821176e-06,
+      "loss": 0.1363,
+      "step": 214860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0685171068561946e-06,
+      "loss": 0.1376,
+      "step": 214870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.066275054930272e-06,
+      "loss": 0.1356,
+      "step": 214880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.06403300300435e-06,
+      "loss": 0.138,
+      "step": 214890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.061790951078427e-06,
+      "loss": 0.1353,
+      "step": 214900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0595488991525044e-06,
+      "loss": 0.1399,
+      "step": 214910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.057306847226582e-06,
+      "loss": 0.14,
+      "step": 214920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0550647953006593e-06,
+      "loss": 0.1328,
+      "step": 214930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0528227433747364e-06,
+      "loss": 0.1366,
+      "step": 214940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0505806914488142e-06,
+      "loss": 0.1382,
+      "step": 214950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0483386395228917e-06,
+      "loss": 0.1397,
+      "step": 214960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0460965875969687e-06,
+      "loss": 0.1388,
+      "step": 214970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.043854535671046e-06,
+      "loss": 0.1398,
+      "step": 214980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0416124837451236e-06,
+      "loss": 0.1349,
+      "step": 214990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.039370431819201e-06,
+      "loss": 0.1335,
+      "step": 215000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0371283798932785e-06,
+      "loss": 0.1326,
+      "step": 215010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.034886327967356e-06,
+      "loss": 0.1363,
+      "step": 215020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0326442760414334e-06,
+      "loss": 0.1373,
+      "step": 215030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0304022241155104e-06,
+      "loss": 0.1352,
+      "step": 215040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0281601721895883e-06,
+      "loss": 0.133,
+      "step": 215050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0259181202636653e-06,
+      "loss": 0.1336,
+      "step": 215060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0236760683377427e-06,
+      "loss": 0.1372,
+      "step": 215070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.02143401641182e-06,
+      "loss": 0.1334,
+      "step": 215080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0191919644858976e-06,
+      "loss": 0.1285,
+      "step": 215090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.016949912559975e-06,
+      "loss": 0.1345,
+      "step": 215100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.014707860634052e-06,
+      "loss": 0.1362,
+      "step": 215110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.01246580870813e-06,
+      "loss": 0.1304,
+      "step": 215120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.010223756782207e-06,
+      "loss": 0.1358,
+      "step": 215130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0079817048562845e-06,
+      "loss": 0.136,
+      "step": 215140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.005739652930362e-06,
+      "loss": 0.1412,
+      "step": 215150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0034976010044393e-06,
+      "loss": 0.1405,
+      "step": 215160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.001255549078517e-06,
+      "loss": 0.1355,
+      "step": 215170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.999013497152594e-06,
+      "loss": 0.1364,
+      "step": 215180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9967714452266717e-06,
+      "loss": 0.1375,
+      "step": 215190
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.994529393300749e-06,
+      "loss": 0.1367,
+      "step": 215200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.992287341374826e-06,
+      "loss": 0.1388,
+      "step": 215210
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.990045289448904e-06,
+      "loss": 0.1398,
+      "step": 215220
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.987803237522981e-06,
+      "loss": 0.1381,
+      "step": 215230
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9855611855970585e-06,
+      "loss": 0.142,
+      "step": 215240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.983319133671136e-06,
+      "loss": 0.1363,
+      "step": 215250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9810770817452134e-06,
+      "loss": 0.1385,
+      "step": 215260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.978835029819291e-06,
+      "loss": 0.1353,
+      "step": 215270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.976592977893368e-06,
+      "loss": 0.1296,
+      "step": 215280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9743509259674457e-06,
+      "loss": 0.1346,
+      "step": 215290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9721088740415228e-06,
+      "loss": 0.1389,
+      "step": 215300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9698668221156e-06,
+      "loss": 0.1339,
+      "step": 215310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9676247701896777e-06,
+      "loss": 0.138,
+      "step": 215320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.965382718263755e-06,
+      "loss": 0.1338,
+      "step": 215330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9631406663378325e-06,
+      "loss": 0.1355,
+      "step": 215340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.96089861441191e-06,
+      "loss": 0.1378,
+      "step": 215350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9586565624859874e-06,
+      "loss": 0.1388,
+      "step": 215360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9564145105600645e-06,
+      "loss": 0.139,
+      "step": 215370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.954172458634142e-06,
+      "loss": 0.1361,
+      "step": 215380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9519304067082198e-06,
+      "loss": 0.137,
+      "step": 215390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.949688354782297e-06,
+      "loss": 0.1395,
+      "step": 215400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9474463028563743e-06,
+      "loss": 0.1351,
+      "step": 215410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9452042509304517e-06,
+      "loss": 0.1327,
+      "step": 215420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.942962199004529e-06,
+      "loss": 0.1364,
+      "step": 215430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.940720147078606e-06,
+      "loss": 0.1328,
+      "step": 215440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9384780951526836e-06,
+      "loss": 0.1386,
+      "step": 215450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9362360432267615e-06,
+      "loss": 0.137,
+      "step": 215460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9339939913008385e-06,
+      "loss": 0.1384,
+      "step": 215470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.931751939374916e-06,
+      "loss": 0.1369,
+      "step": 215480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9295098874489934e-06,
+      "loss": 0.1376,
+      "step": 215490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.927267835523071e-06,
+      "loss": 0.1314,
+      "step": 215500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9250257835971483e-06,
+      "loss": 0.14,
+      "step": 215510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9227837316712258e-06,
+      "loss": 0.1407,
+      "step": 215520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.920541679745303e-06,
+      "loss": 0.1392,
+      "step": 215530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9182996278193802e-06,
+      "loss": 0.1335,
+      "step": 215540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9160575758934577e-06,
+      "loss": 0.1389,
+      "step": 215550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.913815523967535e-06,
+      "loss": 0.1388,
+      "step": 215560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9115734720416126e-06,
+      "loss": 0.1322,
+      "step": 215570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.90933142011569e-06,
+      "loss": 0.1342,
+      "step": 215580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9070893681897672e-06,
+      "loss": 0.1378,
+      "step": 215590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.905071521456437e-06,
+      "loss": 0.142,
+      "step": 215600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9028294695305146e-06,
+      "loss": 0.138,
+      "step": 215610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9005874176045916e-06,
+      "loss": 0.1355,
+      "step": 215620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8983453656786693e-06,
+      "loss": 0.1415,
+      "step": 215630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8961033137527465e-06,
+      "loss": 0.1366,
+      "step": 215640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.893861261826824e-06,
+      "loss": 0.1326,
+      "step": 215650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8916192099009016e-06,
+      "loss": 0.1353,
+      "step": 215660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8893771579749787e-06,
+      "loss": 0.1382,
+      "step": 215670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8871351060490563e-06,
+      "loss": 0.1344,
+      "step": 215680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8848930541231335e-06,
+      "loss": 0.1326,
+      "step": 215690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.882651002197211e-06,
+      "loss": 0.1342,
+      "step": 215700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8804089502712882e-06,
+      "loss": 0.1385,
+      "step": 215710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8781668983453657e-06,
+      "loss": 0.1329,
+      "step": 215720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8759248464194433e-06,
+      "loss": 0.1391,
+      "step": 215730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8736827944935206e-06,
+      "loss": 0.1351,
+      "step": 215740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.871440742567598e-06,
+      "loss": 0.1442,
+      "step": 215750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8691986906416753e-06,
+      "loss": 0.1382,
+      "step": 215760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8669566387157527e-06,
+      "loss": 0.1356,
+      "step": 215770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8647145867898304e-06,
+      "loss": 0.132,
+      "step": 215780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8624725348639076e-06,
+      "loss": 0.1357,
+      "step": 215790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.860230482937985e-06,
+      "loss": 0.1411,
+      "step": 215800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8579884310120623e-06,
+      "loss": 0.1385,
+      "step": 215810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8557463790861397e-06,
+      "loss": 0.1375,
+      "step": 215820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.853504327160217e-06,
+      "loss": 0.1352,
+      "step": 215830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8512622752342946e-06,
+      "loss": 0.136,
+      "step": 215840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.849020223308372e-06,
+      "loss": 0.1359,
+      "step": 215850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8467781713824493e-06,
+      "loss": 0.1346,
+      "step": 215860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8445361194565268e-06,
+      "loss": 0.1451,
+      "step": 215870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.842294067530604e-06,
+      "loss": 0.139,
+      "step": 215880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8400520156046814e-06,
+      "loss": 0.1356,
+      "step": 215890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8378099636787587e-06,
+      "loss": 0.1334,
+      "step": 215900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8355679117528363e-06,
+      "loss": 0.1384,
+      "step": 215910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8333258598269138e-06,
+      "loss": 0.1369,
+      "step": 215920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.831083807900991e-06,
+      "loss": 0.1367,
+      "step": 215930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8288417559750685e-06,
+      "loss": 0.1325,
+      "step": 215940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8265997040491457e-06,
+      "loss": 0.1337,
+      "step": 215950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8243576521232234e-06,
+      "loss": 0.1374,
+      "step": 215960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8221156001973008e-06,
+      "loss": 0.1415,
+      "step": 215970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.819873548271378e-06,
+      "loss": 0.1393,
+      "step": 215980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8176314963454555e-06,
+      "loss": 0.1376,
+      "step": 215990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8153894444195327e-06,
+      "loss": 0.1387,
+      "step": 216000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8131473924936104e-06,
+      "loss": 0.14,
+      "step": 216010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8109053405676874e-06,
+      "loss": 0.1366,
+      "step": 216020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.808663288641765e-06,
+      "loss": 0.1395,
+      "step": 216030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8064212367158425e-06,
+      "loss": 0.1418,
+      "step": 216040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8041791847899197e-06,
+      "loss": 0.1338,
+      "step": 216050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8019371328639974e-06,
+      "loss": 0.1378,
+      "step": 216060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7996950809380744e-06,
+      "loss": 0.1357,
+      "step": 216070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.797453029012152e-06,
+      "loss": 0.1374,
+      "step": 216080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7952109770862293e-06,
+      "loss": 0.1345,
+      "step": 216090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7929689251603068e-06,
+      "loss": 0.1342,
+      "step": 216100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7907268732343842e-06,
+      "loss": 0.1327,
+      "step": 216110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7884848213084615e-06,
+      "loss": 0.1398,
+      "step": 216120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7862427693825391e-06,
+      "loss": 0.1382,
+      "step": 216130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7840007174566163e-06,
+      "loss": 0.1356,
+      "step": 216140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7817586655306938e-06,
+      "loss": 0.1371,
+      "step": 216150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7795166136047712e-06,
+      "loss": 0.1361,
+      "step": 216160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7772745616788485e-06,
+      "loss": 0.1332,
+      "step": 216170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7750325097529261e-06,
+      "loss": 0.1433,
+      "step": 216180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7727904578270032e-06,
+      "loss": 0.1428,
+      "step": 216190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7705484059010808e-06,
+      "loss": 0.1394,
+      "step": 216200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.768306353975158e-06,
+      "loss": 0.1373,
+      "step": 216210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7660643020492355e-06,
+      "loss": 0.1352,
+      "step": 216220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7638222501233132e-06,
+      "loss": 0.1358,
+      "step": 216230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7615801981973902e-06,
+      "loss": 0.1389,
+      "step": 216240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7593381462714678e-06,
+      "loss": 0.1387,
+      "step": 216250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.757096094345545e-06,
+      "loss": 0.1381,
+      "step": 216260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7548540424196225e-06,
+      "loss": 0.1431,
+      "step": 216270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7526119904937002e-06,
+      "loss": 0.1413,
+      "step": 216280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7503699385677772e-06,
+      "loss": 0.1406,
+      "step": 216290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7481278866418549e-06,
+      "loss": 0.1355,
+      "step": 216300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.745885834715932e-06,
+      "loss": 0.1358,
+      "step": 216310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7436437827900095e-06,
+      "loss": 0.1428,
+      "step": 216320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7414017308640868e-06,
+      "loss": 0.1381,
+      "step": 216330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7391596789381642e-06,
+      "loss": 0.1405,
+      "step": 216340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7369176270122419e-06,
+      "loss": 0.1342,
+      "step": 216350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7346755750863191e-06,
+      "loss": 0.1377,
+      "step": 216360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7324335231603966e-06,
+      "loss": 0.1322,
+      "step": 216370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7301914712344738e-06,
+      "loss": 0.1382,
+      "step": 216380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7279494193085513e-06,
+      "loss": 0.1368,
+      "step": 216390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7257073673826285e-06,
+      "loss": 0.1291,
+      "step": 216400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7234653154567061e-06,
+      "loss": 0.1359,
+      "step": 216410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7212232635307836e-06,
+      "loss": 0.1389,
+      "step": 216420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7189812116048608e-06,
+      "loss": 0.134,
+      "step": 216430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7167391596789383e-06,
+      "loss": 0.1366,
+      "step": 216440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7144971077530155e-06,
+      "loss": 0.1341,
+      "step": 216450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.712255055827093e-06,
+      "loss": 0.1382,
+      "step": 216460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7100130039011706e-06,
+      "loss": 0.1412,
+      "step": 216470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7077709519752479e-06,
+      "loss": 0.1346,
+      "step": 216480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7055289000493253e-06,
+      "loss": 0.1337,
+      "step": 216490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7032868481234025e-06,
+      "loss": 0.1339,
+      "step": 216500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.70104479619748e-06,
+      "loss": 0.1381,
+      "step": 216510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6988027442715572e-06,
+      "loss": 0.1319,
+      "step": 216520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6965606923456349e-06,
+      "loss": 0.1368,
+      "step": 216530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6943186404197123e-06,
+      "loss": 0.1415,
+      "step": 216540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6920765884937896e-06,
+      "loss": 0.1349,
+      "step": 216550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.689834536567867e-06,
+      "loss": 0.1345,
+      "step": 216560
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6875924846419442e-06,
+      "loss": 0.1369,
+      "step": 216570
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.685350432716022e-06,
+      "loss": 0.1331,
+      "step": 216580
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.683108380790099e-06,
+      "loss": 0.1326,
+      "step": 216590
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6808663288641766e-06,
+      "loss": 0.1363,
+      "step": 216600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.678624276938254e-06,
+      "loss": 0.139,
+      "step": 216610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6763822250123313e-06,
+      "loss": 0.1353,
+      "step": 216620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.674140173086409e-06,
+      "loss": 0.1341,
+      "step": 216630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.671898121160486e-06,
+      "loss": 0.1365,
+      "step": 216640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6696560692345636e-06,
+      "loss": 0.1376,
+      "step": 216650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.667414017308641e-06,
+      "loss": 0.1383,
+      "step": 216660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6651719653827183e-06,
+      "loss": 0.138,
+      "step": 216670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6629299134567957e-06,
+      "loss": 0.1363,
+      "step": 216680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.660687861530873e-06,
+      "loss": 0.1397,
+      "step": 216690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6584458096049506e-06,
+      "loss": 0.1341,
+      "step": 216700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6562037576790279e-06,
+      "loss": 0.138,
+      "step": 216710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6539617057531053e-06,
+      "loss": 0.1379,
+      "step": 216720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6517196538271828e-06,
+      "loss": 0.1331,
+      "step": 216730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.64947760190126e-06,
+      "loss": 0.1354,
+      "step": 216740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6472355499753377e-06,
+      "loss": 0.1418,
+      "step": 216750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6449934980494147e-06,
+      "loss": 0.134,
+      "step": 216760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6427514461234923e-06,
+      "loss": 0.139,
+      "step": 216770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6405093941975696e-06,
+      "loss": 0.14,
+      "step": 216780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.638267342271647e-06,
+      "loss": 0.1341,
+      "step": 216790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6360252903457247e-06,
+      "loss": 0.1388,
+      "step": 216800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6337832384198017e-06,
+      "loss": 0.1358,
+      "step": 216810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6315411864938794e-06,
+      "loss": 0.1383,
+      "step": 216820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6292991345679566e-06,
+      "loss": 0.1361,
+      "step": 216830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.627057082642034e-06,
+      "loss": 0.1292,
+      "step": 216840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6248150307161117e-06,
+      "loss": 0.1328,
+      "step": 216850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6225729787901887e-06,
+      "loss": 0.1327,
+      "step": 216860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6203309268642664e-06,
+      "loss": 0.131,
+      "step": 216870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6180888749383436e-06,
+      "loss": 0.1376,
+      "step": 216880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.615846823012421e-06,
+      "loss": 0.1386,
+      "step": 216890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6136047710864983e-06,
+      "loss": 0.1366,
+      "step": 216900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6113627191605758e-06,
+      "loss": 0.1331,
+      "step": 216910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6091206672346534e-06,
+      "loss": 0.1404,
+      "step": 216920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6068786153087306e-06,
+      "loss": 0.1391,
+      "step": 216930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.604636563382808e-06,
+      "loss": 0.1392,
+      "step": 216940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6023945114568853e-06,
+      "loss": 0.1369,
+      "step": 216950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6001524595309628e-06,
+      "loss": 0.135,
+      "step": 216960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5979104076050404e-06,
+      "loss": 0.1388,
+      "step": 216970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5956683556791177e-06,
+      "loss": 0.1359,
+      "step": 216980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5934263037531951e-06,
+      "loss": 0.1376,
+      "step": 216990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5911842518272724e-06,
+      "loss": 0.1333,
+      "step": 217000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5889421999013498e-06,
+      "loss": 0.1331,
+      "step": 217010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.586700147975427e-06,
+      "loss": 0.1333,
+      "step": 217020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5844580960495045e-06,
+      "loss": 0.1353,
+      "step": 217030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5822160441235821e-06,
+      "loss": 0.1317,
+      "step": 217040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5799739921976594e-06,
+      "loss": 0.1365,
+      "step": 217050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5777319402717368e-06,
+      "loss": 0.1414,
+      "step": 217060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.575489888345814e-06,
+      "loss": 0.1328,
+      "step": 217070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5732478364198915e-06,
+      "loss": 0.1417,
+      "step": 217080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5710057844939687e-06,
+      "loss": 0.1398,
+      "step": 217090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5687637325680464e-06,
+      "loss": 0.1406,
+      "step": 217100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5665216806421239e-06,
+      "loss": 0.1428,
+      "step": 217110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.564279628716201e-06,
+      "loss": 0.1379,
+      "step": 217120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5620375767902785e-06,
+      "loss": 0.1402,
+      "step": 217130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.559795524864356e-06,
+      "loss": 0.132,
+      "step": 217140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5575534729384334e-06,
+      "loss": 0.14,
+      "step": 217150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5553114210125107e-06,
+      "loss": 0.1384,
+      "step": 217160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5530693690865881e-06,
+      "loss": 0.1343,
+      "step": 217170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5508273171606653e-06,
+      "loss": 0.1365,
+      "step": 217180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.548585265234743e-06,
+      "loss": 0.1385,
+      "step": 217190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5463432133088205e-06,
+      "loss": 0.1307,
+      "step": 217200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5441011613828977e-06,
+      "loss": 0.1323,
+      "step": 217210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5418591094569751e-06,
+      "loss": 0.135,
+      "step": 217220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5396170575310524e-06,
+      "loss": 0.1375,
+      "step": 217230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5373750056051298e-06,
+      "loss": 0.1379,
+      "step": 217240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5351329536792073e-06,
+      "loss": 0.1371,
+      "step": 217250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5328909017532847e-06,
+      "loss": 0.1389,
+      "step": 217260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5306488498273622e-06,
+      "loss": 0.1394,
+      "step": 217270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5284067979014394e-06,
+      "loss": 0.1353,
+      "step": 217280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5261647459755168e-06,
+      "loss": 0.1366,
+      "step": 217290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5239226940495943e-06,
+      "loss": 0.1336,
+      "step": 217300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5216806421236715e-06,
+      "loss": 0.1361,
+      "step": 217310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5194385901977492e-06,
+      "loss": 0.1356,
+      "step": 217320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5171965382718264e-06,
+      "loss": 0.1359,
+      "step": 217330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5149544863459039e-06,
+      "loss": 0.1402,
+      "step": 217340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5127124344199813e-06,
+      "loss": 0.1396,
+      "step": 217350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5104703824940586e-06,
+      "loss": 0.1383,
+      "step": 217360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.508228330568136e-06,
+      "loss": 0.1305,
+      "step": 217370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5059862786422134e-06,
+      "loss": 0.1408,
+      "step": 217380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5037442267162909e-06,
+      "loss": 0.1324,
+      "step": 217390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5015021747903683e-06,
+      "loss": 0.1309,
+      "step": 217400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4992601228644456e-06,
+      "loss": 0.1395,
+      "step": 217410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.497018070938523e-06,
+      "loss": 0.14,
+      "step": 217420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4947760190126003e-06,
+      "loss": 0.1372,
+      "step": 217430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.492533967086678e-06,
+      "loss": 0.1311,
+      "step": 217440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4902919151607552e-06,
+      "loss": 0.1358,
+      "step": 217450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4880498632348326e-06,
+      "loss": 0.1346,
+      "step": 217460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.48580781130891e-06,
+      "loss": 0.139,
+      "step": 217470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4835657593829873e-06,
+      "loss": 0.135,
+      "step": 217480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4813237074570647e-06,
+      "loss": 0.1299,
+      "step": 217490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4790816555311422e-06,
+      "loss": 0.142,
+      "step": 217500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4768396036052196e-06,
+      "loss": 0.1379,
+      "step": 217510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.474597551679297e-06,
+      "loss": 0.1399,
+      "step": 217520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4723554997533743e-06,
+      "loss": 0.1333,
+      "step": 217530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4701134478274518e-06,
+      "loss": 0.1375,
+      "step": 217540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4678713959015292e-06,
+      "loss": 0.1391,
+      "step": 217550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4656293439756064e-06,
+      "loss": 0.1336,
+      "step": 217560
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.463387292049684e-06,
+      "loss": 0.1381,
+      "step": 217570
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4611452401237613e-06,
+      "loss": 0.1313,
+      "step": 217580
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4589031881978388e-06,
+      "loss": 0.1427,
+      "step": 217590
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.456661136271916e-06,
+      "loss": 0.1318,
+      "step": 217600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4544190843459935e-06,
+      "loss": 0.1353,
+      "step": 217610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4524012376126632e-06,
+      "loss": 0.1374,
+      "step": 217620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4501591856867406e-06,
+      "loss": 0.137,
+      "step": 217630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4479171337608178e-06,
+      "loss": 0.1373,
+      "step": 217640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4456750818348955e-06,
+      "loss": 0.1374,
+      "step": 217650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4434330299089727e-06,
+      "loss": 0.1393,
+      "step": 217660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4411909779830502e-06,
+      "loss": 0.1347,
+      "step": 217670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4389489260571276e-06,
+      "loss": 0.1351,
+      "step": 217680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4367068741312049e-06,
+      "loss": 0.1268,
+      "step": 217690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4344648222052823e-06,
+      "loss": 0.1371,
+      "step": 217700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4322227702793598e-06,
+      "loss": 0.1366,
+      "step": 217710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4299807183534372e-06,
+      "loss": 0.1409,
+      "step": 217720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4277386664275147e-06,
+      "loss": 0.1316,
+      "step": 217730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4254966145015919e-06,
+      "loss": 0.1328,
+      "step": 217740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4232545625756693e-06,
+      "loss": 0.1343,
+      "step": 217750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4210125106497466e-06,
+      "loss": 0.1338,
+      "step": 217760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.418770458723824e-06,
+      "loss": 0.1408,
+      "step": 217770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4165284067979015e-06,
+      "loss": 0.1343,
+      "step": 217780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.414286354871979e-06,
+      "loss": 0.1371,
+      "step": 217790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4120443029460564e-06,
+      "loss": 0.1378,
+      "step": 217800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4098022510201336e-06,
+      "loss": 0.1347,
+      "step": 217810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.407560199094211e-06,
+      "loss": 0.1447,
+      "step": 217820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4053181471682885e-06,
+      "loss": 0.1379,
+      "step": 217830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.403076095242366e-06,
+      "loss": 0.131,
+      "step": 217840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4008340433164434e-06,
+      "loss": 0.133,
+      "step": 217850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3985919913905206e-06,
+      "loss": 0.135,
+      "step": 217860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.396349939464598e-06,
+      "loss": 0.1359,
+      "step": 217870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3941078875386755e-06,
+      "loss": 0.1326,
+      "step": 217880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3918658356127528e-06,
+      "loss": 0.132,
+      "step": 217890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3896237836868304e-06,
+      "loss": 0.1377,
+      "step": 217900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3873817317609076e-06,
+      "loss": 0.1331,
+      "step": 217910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.385139679834985e-06,
+      "loss": 0.1297,
+      "step": 217920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3828976279090623e-06,
+      "loss": 0.1433,
+      "step": 217930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3806555759831398e-06,
+      "loss": 0.1415,
+      "step": 217940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3784135240572172e-06,
+      "loss": 0.1345,
+      "step": 217950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3761714721312947e-06,
+      "loss": 0.1308,
+      "step": 217960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3739294202053721e-06,
+      "loss": 0.1384,
+      "step": 217970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3716873682794494e-06,
+      "loss": 0.1361,
+      "step": 217980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3694453163535268e-06,
+      "loss": 0.1406,
+      "step": 217990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3672032644276042e-06,
+      "loss": 0.1398,
+      "step": 218000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3649612125016815e-06,
+      "loss": 0.14,
+      "step": 218010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.362719160575759e-06,
+      "loss": 0.1404,
+      "step": 218020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3604771086498364e-06,
+      "loss": 0.1426,
+      "step": 218030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3582350567239138e-06,
+      "loss": 0.1328,
+      "step": 218040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3559930047979913e-06,
+      "loss": 0.135,
+      "step": 218050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3537509528720685e-06,
+      "loss": 0.1343,
+      "step": 218060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351508900946146e-06,
+      "loss": 0.137,
+      "step": 218070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3492668490202232e-06,
+      "loss": 0.1404,
+      "step": 218080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3470247970943009e-06,
+      "loss": 0.1382,
+      "step": 218090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3447827451683783e-06,
+      "loss": 0.1361,
+      "step": 218100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3425406932424555e-06,
+      "loss": 0.1403,
+      "step": 218110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.340298641316533e-06,
+      "loss": 0.138,
+      "step": 218120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3380565893906102e-06,
+      "loss": 0.1371,
+      "step": 218130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3358145374646877e-06,
+      "loss": 0.1391,
+      "step": 218140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3335724855387651e-06,
+      "loss": 0.1382,
+      "step": 218150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3313304336128426e-06,
+      "loss": 0.137,
+      "step": 218160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.32908838168692e-06,
+      "loss": 0.1315,
+      "step": 218170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3268463297609972e-06,
+      "loss": 0.1352,
+      "step": 218180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3246042778350747e-06,
+      "loss": 0.1337,
+      "step": 218190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3223622259091521e-06,
+      "loss": 0.1333,
+      "step": 218200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3201201739832296e-06,
+      "loss": 0.1378,
+      "step": 218210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.317878122057307e-06,
+      "loss": 0.1371,
+      "step": 218220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3156360701313843e-06,
+      "loss": 0.1394,
+      "step": 218230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3133940182054617e-06,
+      "loss": 0.1365,
+      "step": 218240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3111519662795392e-06,
+      "loss": 0.1351,
+      "step": 218250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3089099143536164e-06,
+      "loss": 0.1384,
+      "step": 218260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3066678624276938e-06,
+      "loss": 0.1413,
+      "step": 218270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3044258105017713e-06,
+      "loss": 0.1393,
+      "step": 218280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3021837585758487e-06,
+      "loss": 0.1364,
+      "step": 218290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2999417066499262e-06,
+      "loss": 0.1308,
+      "step": 218300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2976996547240034e-06,
+      "loss": 0.1327,
+      "step": 218310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2954576027980809e-06,
+      "loss": 0.1359,
+      "step": 218320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.293215550872158e-06,
+      "loss": 0.137,
+      "step": 218330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2909734989462358e-06,
+      "loss": 0.1304,
+      "step": 218340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.288731447020313e-06,
+      "loss": 0.1402,
+      "step": 218350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2864893950943904e-06,
+      "loss": 0.1383,
+      "step": 218360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2842473431684679e-06,
+      "loss": 0.1327,
+      "step": 218370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2820052912425451e-06,
+      "loss": 0.136,
+      "step": 218380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2797632393166226e-06,
+      "loss": 0.134,
+      "step": 218390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2775211873907e-06,
+      "loss": 0.134,
+      "step": 218400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2752791354647775e-06,
+      "loss": 0.1402,
+      "step": 218410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.273037083538855e-06,
+      "loss": 0.1351,
+      "step": 218420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2707950316129322e-06,
+      "loss": 0.1358,
+      "step": 218430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2685529796870096e-06,
+      "loss": 0.1382,
+      "step": 218440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.266310927761087e-06,
+      "loss": 0.1368,
+      "step": 218450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2640688758351645e-06,
+      "loss": 0.13,
+      "step": 218460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.261826823909242e-06,
+      "loss": 0.1377,
+      "step": 218470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2595847719833192e-06,
+      "loss": 0.1366,
+      "step": 218480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2573427200573966e-06,
+      "loss": 0.1369,
+      "step": 218490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2551006681314739e-06,
+      "loss": 0.1414,
+      "step": 218500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2528586162055513e-06,
+      "loss": 0.1386,
+      "step": 218510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2506165642796288e-06,
+      "loss": 0.1379,
+      "step": 218520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2483745123537062e-06,
+      "loss": 0.1343,
+      "step": 218530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2461324604277836e-06,
+      "loss": 0.133,
+      "step": 218540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2438904085018609e-06,
+      "loss": 0.132,
+      "step": 218550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2416483565759383e-06,
+      "loss": 0.1337,
+      "step": 218560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2394063046500158e-06,
+      "loss": 0.1327,
+      "step": 218570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.237164252724093e-06,
+      "loss": 0.139,
+      "step": 218580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2349222007981707e-06,
+      "loss": 0.138,
+      "step": 218590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.232680148872248e-06,
+      "loss": 0.1337,
+      "step": 218600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2304380969463254e-06,
+      "loss": 0.1337,
+      "step": 218610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2281960450204028e-06,
+      "loss": 0.1394,
+      "step": 218620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.22595399309448e-06,
+      "loss": 0.1387,
+      "step": 218630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2237119411685575e-06,
+      "loss": 0.1372,
+      "step": 218640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.221469889242635e-06,
+      "loss": 0.1376,
+      "step": 218650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2192278373167124e-06,
+      "loss": 0.1339,
+      "step": 218660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2169857853907898e-06,
+      "loss": 0.1329,
+      "step": 218670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.214743733464867e-06,
+      "loss": 0.1384,
+      "step": 218680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2125016815389445e-06,
+      "loss": 0.1369,
+      "step": 218690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2102596296130217e-06,
+      "loss": 0.1372,
+      "step": 218700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2080175776870994e-06,
+      "loss": 0.1326,
+      "step": 218710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2057755257611768e-06,
+      "loss": 0.1313,
+      "step": 218720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.203533473835254e-06,
+      "loss": 0.1357,
+      "step": 218730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2012914219093315e-06,
+      "loss": 0.1367,
+      "step": 218740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1990493699834088e-06,
+      "loss": 0.1382,
+      "step": 218750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1968073180574862e-06,
+      "loss": 0.1472,
+      "step": 218760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1945652661315637e-06,
+      "loss": 0.134,
+      "step": 218770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1923232142056411e-06,
+      "loss": 0.1321,
+      "step": 218780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1900811622797186e-06,
+      "loss": 0.1361,
+      "step": 218790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1878391103537958e-06,
+      "loss": 0.1344,
+      "step": 218800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1855970584278732e-06,
+      "loss": 0.1333,
+      "step": 218810
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1833550065019507e-06,
+      "loss": 0.1329,
+      "step": 218820
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.181112954576028e-06,
+      "loss": 0.1362,
+      "step": 218830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1788709026501056e-06,
+      "loss": 0.1399,
+      "step": 218840
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1766288507241828e-06,
+      "loss": 0.1399,
+      "step": 218850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1743867987982603e-06,
+      "loss": 0.1374,
+      "step": 218860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1721447468723377e-06,
+      "loss": 0.139,
+      "step": 218870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.169902694946415e-06,
+      "loss": 0.1398,
+      "step": 218880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1676606430204924e-06,
+      "loss": 0.1401,
+      "step": 218890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1654185910945698e-06,
+      "loss": 0.1354,
+      "step": 218900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1631765391686473e-06,
+      "loss": 0.1378,
+      "step": 218910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1609344872427245e-06,
+      "loss": 0.1411,
+      "step": 218920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.158692435316802e-06,
+      "loss": 0.1337,
+      "step": 218930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1564503833908794e-06,
+      "loss": 0.1326,
+      "step": 218940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1542083314649567e-06,
+      "loss": 0.1348,
+      "step": 218950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1519662795390343e-06,
+      "loss": 0.134,
+      "step": 218960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1497242276131115e-06,
+      "loss": 0.1403,
+      "step": 218970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.147482175687189e-06,
+      "loss": 0.1366,
+      "step": 218980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1452401237612664e-06,
+      "loss": 0.1359,
+      "step": 218990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1429980718353437e-06,
+      "loss": 0.1394,
+      "step": 219000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1407560199094211e-06,
+      "loss": 0.1319,
+      "step": 219010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1385139679834986e-06,
+      "loss": 0.1368,
+      "step": 219020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.136271916057576e-06,
+      "loss": 0.1322,
+      "step": 219030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1340298641316535e-06,
+      "loss": 0.1407,
+      "step": 219040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1317878122057307e-06,
+      "loss": 0.1356,
+      "step": 219050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1295457602798081e-06,
+      "loss": 0.1378,
+      "step": 219060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1273037083538854e-06,
+      "loss": 0.1359,
+      "step": 219070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1250616564279628e-06,
+      "loss": 0.1422,
+      "step": 219080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1228196045020405e-06,
+      "loss": 0.135,
+      "step": 219090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1205775525761177e-06,
+      "loss": 0.1367,
+      "step": 219100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1183355006501952e-06,
+      "loss": 0.1406,
+      "step": 219110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1160934487242724e-06,
+      "loss": 0.139,
+      "step": 219120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1138513967983499e-06,
+      "loss": 0.1374,
+      "step": 219130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1116093448724273e-06,
+      "loss": 0.1408,
+      "step": 219140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1093672929465047e-06,
+      "loss": 0.1371,
+      "step": 219150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1071252410205822e-06,
+      "loss": 0.1369,
+      "step": 219160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1048831890946594e-06,
+      "loss": 0.1324,
+      "step": 219170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1026411371687369e-06,
+      "loss": 0.1368,
+      "step": 219180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1003990852428143e-06,
+      "loss": 0.1332,
+      "step": 219190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0981570333168916e-06,
+      "loss": 0.1301,
+      "step": 219200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0959149813909692e-06,
+      "loss": 0.1394,
+      "step": 219210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0936729294650465e-06,
+      "loss": 0.1407,
+      "step": 219220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.091430877539124e-06,
+      "loss": 0.1384,
+      "step": 219230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0891888256132013e-06,
+      "loss": 0.1378,
+      "step": 219240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0869467736872786e-06,
+      "loss": 0.1353,
+      "step": 219250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.084704721761356e-06,
+      "loss": 0.1385,
+      "step": 219260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0824626698354333e-06,
+      "loss": 0.1385,
+      "step": 219270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.080220617909511e-06,
+      "loss": 0.1377,
+      "step": 219280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0779785659835884e-06,
+      "loss": 0.1397,
+      "step": 219290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0757365140576656e-06,
+      "loss": 0.1322,
+      "step": 219300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.073494462131743e-06,
+      "loss": 0.1408,
+      "step": 219310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0712524102058203e-06,
+      "loss": 0.1356,
+      "step": 219320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0690103582798977e-06,
+      "loss": 0.1392,
+      "step": 219330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0667683063539752e-06,
+      "loss": 0.1374,
+      "step": 219340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0645262544280526e-06,
+      "loss": 0.1328,
+      "step": 219350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.06228420250213e-06,
+      "loss": 0.139,
+      "step": 219360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0600421505762073e-06,
+      "loss": 0.1329,
+      "step": 219370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0578000986502848e-06,
+      "loss": 0.1347,
+      "step": 219380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0555580467243622e-06,
+      "loss": 0.13,
+      "step": 219390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0533159947984397e-06,
+      "loss": 0.1415,
+      "step": 219400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.051073942872517e-06,
+      "loss": 0.1416,
+      "step": 219410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0488318909465943e-06,
+      "loss": 0.1372,
+      "step": 219420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0465898390206718e-06,
+      "loss": 0.134,
+      "step": 219430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0443477870947492e-06,
+      "loss": 0.134,
+      "step": 219440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0421057351688265e-06,
+      "loss": 0.1338,
+      "step": 219450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.039863683242904e-06,
+      "loss": 0.1375,
+      "step": 219460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0376216313169814e-06,
+      "loss": 0.1372,
+      "step": 219470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0353795793910588e-06,
+      "loss": 0.1397,
+      "step": 219480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.033137527465136e-06,
+      "loss": 0.1379,
+      "step": 219490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0308954755392135e-06,
+      "loss": 0.1371,
+      "step": 219500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.028653423613291e-06,
+      "loss": 0.1355,
+      "step": 219510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0264113716873682e-06,
+      "loss": 0.137,
+      "step": 219520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0241693197614458e-06,
+      "loss": 0.1414,
+      "step": 219530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.021927267835523e-06,
+      "loss": 0.1373,
+      "step": 219540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0196852159096005e-06,
+      "loss": 0.1345,
+      "step": 219550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.017443163983678e-06,
+      "loss": 0.1354,
+      "step": 219560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0152011120577552e-06,
+      "loss": 0.1387,
+      "step": 219570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0129590601318326e-06,
+      "loss": 0.1394,
+      "step": 219580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.01071700820591e-06,
+      "loss": 0.1323,
+      "step": 219590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0084749562799875e-06,
+      "loss": 0.1386,
+      "step": 219600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.006232904354065e-06,
+      "loss": 0.1338,
+      "step": 219610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0042150576207347e-06,
+      "loss": 0.1369,
+      "step": 219620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.001973005694812e-06,
+      "loss": 0.1349,
+      "step": 219630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.997309537688894e-07,
+      "loss": 0.1382,
+      "step": 219640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.974889018429666e-07,
+      "loss": 0.1361,
+      "step": 219650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.95246849917044e-07,
+      "loss": 0.1397,
+      "step": 219660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.930047979911215e-07,
+      "loss": 0.1324,
+      "step": 219670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.90762746065199e-07,
+      "loss": 0.1343,
+      "step": 219680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.885206941392764e-07,
+      "loss": 0.1411,
+      "step": 219690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.862786422133536e-07,
+      "loss": 0.1402,
+      "step": 219700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.84036590287431e-07,
+      "loss": 0.134,
+      "step": 219710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.817945383615085e-07,
+      "loss": 0.1419,
+      "step": 219720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.795524864355858e-07,
+      "loss": 0.1379,
+      "step": 219730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.773104345096634e-07,
+      "loss": 0.1358,
+      "step": 219740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.750683825837407e-07,
+      "loss": 0.1377,
+      "step": 219750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.728263306578181e-07,
+      "loss": 0.1326,
+      "step": 219760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.705842787318956e-07,
+      "loss": 0.1336,
+      "step": 219770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.683422268059728e-07,
+      "loss": 0.1387,
+      "step": 219780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.661001748800502e-07,
+      "loss": 0.1382,
+      "step": 219790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.638581229541277e-07,
+      "loss": 0.1384,
+      "step": 219800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.616160710282051e-07,
+      "loss": 0.1306,
+      "step": 219810
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.593740191022824e-07,
+      "loss": 0.1397,
+      "step": 219820
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.571319671763598e-07,
+      "loss": 0.1424,
+      "step": 219830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.548899152504373e-07,
+      "loss": 0.1346,
+      "step": 219840
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.526478633245146e-07,
+      "loss": 0.1371,
+      "step": 219850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.504058113985922e-07,
+      "loss": 0.1347,
+      "step": 219860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.481637594726695e-07,
+      "loss": 0.1372,
+      "step": 219870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.459217075467468e-07,
+      "loss": 0.1393,
+      "step": 219880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.436796556208242e-07,
+      "loss": 0.1376,
+      "step": 219890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.414376036949016e-07,
+      "loss": 0.1342,
+      "step": 219900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.39195551768979e-07,
+      "loss": 0.1347,
+      "step": 219910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.369534998430565e-07,
+      "loss": 0.1349,
+      "step": 219920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.347114479171339e-07,
+      "loss": 0.1338,
+      "step": 219930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.324693959912112e-07,
+      "loss": 0.1329,
+      "step": 219940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.302273440652885e-07,
+      "loss": 0.1334,
+      "step": 219950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.27985292139366e-07,
+      "loss": 0.1308,
+      "step": 219960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.257432402134433e-07,
+      "loss": 0.1379,
+      "step": 219970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.235011882875207e-07,
+      "loss": 0.1327,
+      "step": 219980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.212591363615982e-07,
+      "loss": 0.1364,
+      "step": 219990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.190170844356756e-07,
+      "loss": 0.1332,
+      "step": 220000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.16775032509753e-07,
+      "loss": 0.1371,
+      "step": 220010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.145329805838304e-07,
+      "loss": 0.1402,
+      "step": 220020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.122909286579077e-07,
+      "loss": 0.1341,
+      "step": 220030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.10048876731985e-07,
+      "loss": 0.1384,
+      "step": 220040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.078068248060626e-07,
+      "loss": 0.1363,
+      "step": 220050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.0556477288014e-07,
+      "loss": 0.1347,
+      "step": 220060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.033227209542174e-07,
+      "loss": 0.1314,
+      "step": 220070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.010806690282947e-07,
+      "loss": 0.1379,
+      "step": 220080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.988386171023721e-07,
+      "loss": 0.1391,
+      "step": 220090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.965965651764495e-07,
+      "loss": 0.1402,
+      "step": 220100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.94354513250527e-07,
+      "loss": 0.1409,
+      "step": 220110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.921124613246044e-07,
+      "loss": 0.1354,
+      "step": 220120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.898704093986817e-07,
+      "loss": 0.133,
+      "step": 220130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.876283574727591e-07,
+      "loss": 0.1385,
+      "step": 220140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.853863055468364e-07,
+      "loss": 0.1356,
+      "step": 220150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.831442536209139e-07,
+      "loss": 0.1355,
+      "step": 220160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.809022016949914e-07,
+      "loss": 0.1356,
+      "step": 220170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.786601497690688e-07,
+      "loss": 0.1365,
+      "step": 220180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.764180978431461e-07,
+      "loss": 0.1392,
+      "step": 220190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.741760459172235e-07,
+      "loss": 0.1363,
+      "step": 220200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.719339939913009e-07,
+      "loss": 0.1345,
+      "step": 220210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.696919420653782e-07,
+      "loss": 0.1353,
+      "step": 220220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.674498901394556e-07,
+      "loss": 0.1312,
+      "step": 220230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.652078382135331e-07,
+      "loss": 0.1426,
+      "step": 220240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.629657862876105e-07,
+      "loss": 0.1438,
+      "step": 220250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.607237343616878e-07,
+      "loss": 0.1375,
+      "step": 220260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.584816824357653e-07,
+      "loss": 0.1363,
+      "step": 220270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.562396305098426e-07,
+      "loss": 0.1377,
+      "step": 220280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.5399757858392e-07,
+      "loss": 0.1357,
+      "step": 220290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.517555266579975e-07,
+      "loss": 0.1384,
+      "step": 220300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.495134747320748e-07,
+      "loss": 0.1401,
+      "step": 220310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.472714228061523e-07,
+      "loss": 0.1329,
+      "step": 220320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.450293708802296e-07,
+      "loss": 0.1377,
+      "step": 220330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.42787318954307e-07,
+      "loss": 0.1353,
+      "step": 220340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.405452670283843e-07,
+      "loss": 0.1329,
+      "step": 220350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.383032151024619e-07,
+      "loss": 0.14,
+      "step": 220360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.360611631765392e-07,
+      "loss": 0.1332,
+      "step": 220370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.338191112506167e-07,
+      "loss": 0.132,
+      "step": 220380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.31577059324694e-07,
+      "loss": 0.1405,
+      "step": 220390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.293350073987713e-07,
+      "loss": 0.1382,
+      "step": 220400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.270929554728487e-07,
+      "loss": 0.1412,
+      "step": 220410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.248509035469262e-07,
+      "loss": 0.1367,
+      "step": 220420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.226088516210037e-07,
+      "loss": 0.1408,
+      "step": 220430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.20366799695081e-07,
+      "loss": 0.1368,
+      "step": 220440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.181247477691584e-07,
+      "loss": 0.1438,
+      "step": 220450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.158826958432357e-07,
+      "loss": 0.1394,
+      "step": 220460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.136406439173132e-07,
+      "loss": 0.1419,
+      "step": 220470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.113985919913905e-07,
+      "loss": 0.1317,
+      "step": 220480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.09156540065468e-07,
+      "loss": 0.1352,
+      "step": 220490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.069144881395454e-07,
+      "loss": 0.1377,
+      "step": 220500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.046724362136227e-07,
+      "loss": 0.1345,
+      "step": 220510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.024303842877001e-07,
+      "loss": 0.139,
+      "step": 220520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.001883323617775e-07,
+      "loss": 0.1395,
+      "step": 220530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.979462804358549e-07,
+      "loss": 0.1342,
+      "step": 220540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.957042285099324e-07,
+      "loss": 0.1398,
+      "step": 220550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.934621765840098e-07,
+      "loss": 0.131,
+      "step": 220560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.912201246580871e-07,
+      "loss": 0.1283,
+      "step": 220570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.889780727321645e-07,
+      "loss": 0.1307,
+      "step": 220580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.867360208062419e-07,
+      "loss": 0.1378,
+      "step": 220590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.844939688803192e-07,
+      "loss": 0.1395,
+      "step": 220600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.822519169543968e-07,
+      "loss": 0.1357,
+      "step": 220610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.80009865028474e-07,
+      "loss": 0.1364,
+      "step": 220620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.777678131025516e-07,
+      "loss": 0.1432,
+      "step": 220630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.755257611766289e-07,
+      "loss": 0.1404,
+      "step": 220640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.732837092507062e-07,
+      "loss": 0.1388,
+      "step": 220650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.710416573247837e-07,
+      "loss": 0.1357,
+      "step": 220660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.68799605398861e-07,
+      "loss": 0.1394,
+      "step": 220670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.665575534729385e-07,
+      "loss": 0.1381,
+      "step": 220680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.643155015470159e-07,
+      "loss": 0.1345,
+      "step": 220690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.620734496210933e-07,
+      "loss": 0.1337,
+      "step": 220700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.598313976951706e-07,
+      "loss": 0.1381,
+      "step": 220710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.575893457692481e-07,
+      "loss": 0.1321,
+      "step": 220720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.553472938433254e-07,
+      "loss": 0.132,
+      "step": 220730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.531052419174028e-07,
+      "loss": 0.1352,
+      "step": 220740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.508631899914803e-07,
+      "loss": 0.1337,
+      "step": 220750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.486211380655576e-07,
+      "loss": 0.1381,
+      "step": 220760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.46379086139635e-07,
+      "loss": 0.1388,
+      "step": 220770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.441370342137124e-07,
+      "loss": 0.1417,
+      "step": 220780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.418949822877899e-07,
+      "loss": 0.1404,
+      "step": 220790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.396529303618672e-07,
+      "loss": 0.1325,
+      "step": 220800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.374108784359447e-07,
+      "loss": 0.1321,
+      "step": 220810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.35168826510022e-07,
+      "loss": 0.1371,
+      "step": 220820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.329267745840993e-07,
+      "loss": 0.1345,
+      "step": 220830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.306847226581768e-07,
+      "loss": 0.1355,
+      "step": 220840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.284426707322542e-07,
+      "loss": 0.1396,
+      "step": 220850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.262006188063316e-07,
+      "loss": 0.1353,
+      "step": 220860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.239585668804089e-07,
+      "loss": 0.1404,
+      "step": 220870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.217165149544864e-07,
+      "loss": 0.1351,
+      "step": 220880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.194744630285638e-07,
+      "loss": 0.138,
+      "step": 220890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.172324111026412e-07,
+      "loss": 0.1371,
+      "step": 220900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.149903591767186e-07,
+      "loss": 0.1379,
+      "step": 220910
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.127483072507959e-07,
+      "loss": 0.137,
+      "step": 220920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.105062553248733e-07,
+      "loss": 0.1394,
+      "step": 220930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.082642033989507e-07,
+      "loss": 0.1374,
+      "step": 220940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.060221514730282e-07,
+      "loss": 0.1389,
+      "step": 220950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.037800995471055e-07,
+      "loss": 0.1357,
+      "step": 220960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.01538047621183e-07,
+      "loss": 0.1364,
+      "step": 220970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.992959956952603e-07,
+      "loss": 0.1396,
+      "step": 220980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.970539437693378e-07,
+      "loss": 0.1397,
+      "step": 220990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.948118918434152e-07,
+      "loss": 0.1325,
+      "step": 221000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.925698399174925e-07,
+      "loss": 0.1338,
+      "step": 221010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.903277879915699e-07,
+      "loss": 0.131,
+      "step": 221020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.880857360656473e-07,
+      "loss": 0.1348,
+      "step": 221030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.858436841397247e-07,
+      "loss": 0.1373,
+      "step": 221040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.836016322138021e-07,
+      "loss": 0.138,
+      "step": 221050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.813595802878795e-07,
+      "loss": 0.1393,
+      "step": 221060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.791175283619569e-07,
+      "loss": 0.136,
+      "step": 221070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.768754764360343e-07,
+      "loss": 0.136,
+      "step": 221080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.746334245101116e-07,
+      "loss": 0.1372,
+      "step": 221090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.723913725841891e-07,
+      "loss": 0.1417,
+      "step": 221100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.701493206582665e-07,
+      "loss": 0.1348,
+      "step": 221110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.679072687323438e-07,
+      "loss": 0.1334,
+      "step": 221120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.656652168064213e-07,
+      "loss": 0.1239,
+      "step": 221130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.634231648804986e-07,
+      "loss": 0.1413,
+      "step": 221140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.611811129545761e-07,
+      "loss": 0.1367,
+      "step": 221150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.589390610286535e-07,
+      "loss": 0.1371,
+      "step": 221160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.566970091027309e-07,
+      "loss": 0.1356,
+      "step": 221170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.544549571768082e-07,
+      "loss": 0.1389,
+      "step": 221180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.522129052508856e-07,
+      "loss": 0.1389,
+      "step": 221190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.499708533249631e-07,
+      "loss": 0.1394,
+      "step": 221200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.477288013990404e-07,
+      "loss": 0.1352,
+      "step": 221210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.454867494731179e-07,
+      "loss": 0.1378,
+      "step": 221220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.432446975471952e-07,
+      "loss": 0.1336,
+      "step": 221230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.410026456212726e-07,
+      "loss": 0.1349,
+      "step": 221240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.3876059369535e-07,
+      "loss": 0.1423,
+      "step": 221250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.365185417694275e-07,
+      "loss": 0.136,
+      "step": 221260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.342764898435048e-07,
+      "loss": 0.133,
+      "step": 221270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.320344379175822e-07,
+      "loss": 0.1324,
+      "step": 221280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.297923859916596e-07,
+      "loss": 0.1403,
+      "step": 221290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.275503340657369e-07,
+      "loss": 0.138,
+      "step": 221300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.253082821398144e-07,
+      "loss": 0.1356,
+      "step": 221310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.230662302138918e-07,
+      "loss": 0.1331,
+      "step": 221320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.208241782879692e-07,
+      "loss": 0.138,
+      "step": 221330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.185821263620465e-07,
+      "loss": 0.139,
+      "step": 221340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.16340074436124e-07,
+      "loss": 0.136,
+      "step": 221350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.140980225102014e-07,
+      "loss": 0.1371,
+      "step": 221360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.118559705842787e-07,
+      "loss": 0.1346,
+      "step": 221370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.096139186583562e-07,
+      "loss": 0.1368,
+      "step": 221380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.073718667324335e-07,
+      "loss": 0.1397,
+      "step": 221390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.051298148065109e-07,
+      "loss": 0.1337,
+      "step": 221400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.028877628805884e-07,
+      "loss": 0.1288,
+      "step": 221410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.006457109546658e-07,
+      "loss": 0.1357,
+      "step": 221420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.984036590287431e-07,
+      "loss": 0.1339,
+      "step": 221430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.961616071028206e-07,
+      "loss": 0.1386,
+      "step": 221440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.939195551768979e-07,
+      "loss": 0.1337,
+      "step": 221450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.916775032509753e-07,
+      "loss": 0.1371,
+      "step": 221460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.894354513250528e-07,
+      "loss": 0.136,
+      "step": 221470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.871933993991301e-07,
+      "loss": 0.1394,
+      "step": 221480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.849513474732075e-07,
+      "loss": 0.1424,
+      "step": 221490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.827092955472849e-07,
+      "loss": 0.1338,
+      "step": 221500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.804672436213623e-07,
+      "loss": 0.138,
+      "step": 221510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.782251916954397e-07,
+      "loss": 0.1413,
+      "step": 221520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.759831397695172e-07,
+      "loss": 0.137,
+      "step": 221530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.737410878435945e-07,
+      "loss": 0.1414,
+      "step": 221540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.714990359176718e-07,
+      "loss": 0.1371,
+      "step": 221550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.692569839917493e-07,
+      "loss": 0.1322,
+      "step": 221560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.670149320658267e-07,
+      "loss": 0.1372,
+      "step": 221570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.647728801399041e-07,
+      "loss": 0.1382,
+      "step": 221580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.625308282139814e-07,
+      "loss": 0.1347,
+      "step": 221590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.602887762880589e-07,
+      "loss": 0.1392,
+      "step": 221600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.580467243621362e-07,
+      "loss": 0.1395,
+      "step": 221610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.558046724362137e-07,
+      "loss": 0.1342,
+      "step": 221620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.535626205102911e-07,
+      "loss": 0.1352,
+      "step": 221630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.513205685843684e-07,
+      "loss": 0.1401,
+      "step": 221640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.493027218510381e-07,
+      "loss": 0.1417,
+      "step": 221650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.470606699251155e-07,
+      "loss": 0.1335,
+      "step": 221660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.448186179991928e-07,
+      "loss": 0.1368,
+      "step": 221670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.425765660732703e-07,
+      "loss": 0.1389,
+      "step": 221680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.403345141473477e-07,
+      "loss": 0.1326,
+      "step": 221690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.380924622214251e-07,
+      "loss": 0.1359,
+      "step": 221700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.358504102955025e-07,
+      "loss": 0.1314,
+      "step": 221710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.336083583695798e-07,
+      "loss": 0.1374,
+      "step": 221720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.313663064436572e-07,
+      "loss": 0.136,
+      "step": 221730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.291242545177347e-07,
+      "loss": 0.1412,
+      "step": 221740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.268822025918121e-07,
+      "loss": 0.1376,
+      "step": 221750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.246401506658894e-07,
+      "loss": 0.1337,
+      "step": 221760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.223980987399669e-07,
+      "loss": 0.1396,
+      "step": 221770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.201560468140442e-07,
+      "loss": 0.1383,
+      "step": 221780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.179139948881217e-07,
+      "loss": 0.1367,
+      "step": 221790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.156719429621991e-07,
+      "loss": 0.1353,
+      "step": 221800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.134298910362764e-07,
+      "loss": 0.1396,
+      "step": 221810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.111878391103538e-07,
+      "loss": 0.136,
+      "step": 221820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.089457871844311e-07,
+      "loss": 0.1339,
+      "step": 221830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.067037352585086e-07,
+      "loss": 0.1361,
+      "step": 221840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.04461683332586e-07,
+      "loss": 0.1338,
+      "step": 221850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.022196314066634e-07,
+      "loss": 0.1367,
+      "step": 221860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.999775794807408e-07,
+      "loss": 0.1418,
+      "step": 221870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.977355275548182e-07,
+      "loss": 0.1415,
+      "step": 221880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.954934756288956e-07,
+      "loss": 0.1385,
+      "step": 221890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.93251423702973e-07,
+      "loss": 0.1405,
+      "step": 221900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.910093717770504e-07,
+      "loss": 0.1378,
+      "step": 221910
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.887673198511277e-07,
+      "loss": 0.1362,
+      "step": 221920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.865252679252052e-07,
+      "loss": 0.1379,
+      "step": 221930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.842832159992825e-07,
+      "loss": 0.1355,
+      "step": 221940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8204116407336e-07,
+      "loss": 0.138,
+      "step": 221950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.797991121474374e-07,
+      "loss": 0.1344,
+      "step": 221960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.775570602215148e-07,
+      "loss": 0.1359,
+      "step": 221970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7531500829559215e-07,
+      "loss": 0.1328,
+      "step": 221980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.730729563696696e-07,
+      "loss": 0.1375,
+      "step": 221990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7083090444374694e-07,
+      "loss": 0.1401,
+      "step": 222000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6858885251782434e-07,
+      "loss": 0.1411,
+      "step": 222010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.663468005919018e-07,
+      "loss": 0.1379,
+      "step": 222020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.641047486659791e-07,
+      "loss": 0.1397,
+      "step": 222030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.618626967400565e-07,
+      "loss": 0.1335,
+      "step": 222040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5962064481413386e-07,
+      "loss": 0.134,
+      "step": 222050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.573785928882113e-07,
+      "loss": 0.133,
+      "step": 222060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.551365409622887e-07,
+      "loss": 0.1366,
+      "step": 222070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5289448903636604e-07,
+      "loss": 0.131,
+      "step": 222080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5065243711044354e-07,
+      "loss": 0.1367,
+      "step": 222090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.484103851845209e-07,
+      "loss": 0.1286,
+      "step": 222100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.461683332585983e-07,
+      "loss": 0.1289,
+      "step": 222110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.439262813326757e-07,
+      "loss": 0.1343,
+      "step": 222120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4168422940675307e-07,
+      "loss": 0.1301,
+      "step": 222130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3944217748083046e-07,
+      "loss": 0.1382,
+      "step": 222140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.372001255549079e-07,
+      "loss": 0.1355,
+      "step": 222150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3495807362898525e-07,
+      "loss": 0.1341,
+      "step": 222160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3271602170306264e-07,
+      "loss": 0.1364,
+      "step": 222170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.304739697771401e-07,
+      "loss": 0.1362,
+      "step": 222180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2823191785121743e-07,
+      "loss": 0.1424,
+      "step": 222190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.259898659252948e-07,
+      "loss": 0.137,
+      "step": 222200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2374781399937227e-07,
+      "loss": 0.1382,
+      "step": 222210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2150576207344967e-07,
+      "loss": 0.1329,
+      "step": 222220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.19263710147527e-07,
+      "loss": 0.1376,
+      "step": 222230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1702165822160446e-07,
+      "loss": 0.1339,
+      "step": 222240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1477960629568185e-07,
+      "loss": 0.1388,
+      "step": 222250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.125375543697592e-07,
+      "loss": 0.138,
+      "step": 222260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1029550244383664e-07,
+      "loss": 0.133,
+      "step": 222270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0805345051791403e-07,
+      "loss": 0.1332,
+      "step": 222280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.058113985919914e-07,
+      "loss": 0.1382,
+      "step": 222290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0356934666606877e-07,
+      "loss": 0.1357,
+      "step": 222300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.013272947401462e-07,
+      "loss": 0.1331,
+      "step": 222310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.990852428142236e-07,
+      "loss": 0.137,
+      "step": 222320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9684319088830095e-07,
+      "loss": 0.1326,
+      "step": 222330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.946011389623784e-07,
+      "loss": 0.1342,
+      "step": 222340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.923590870364558e-07,
+      "loss": 0.1373,
+      "step": 222350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.901170351105332e-07,
+      "loss": 0.1375,
+      "step": 222360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.878749831846106e-07,
+      "loss": 0.1396,
+      "step": 222370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.85632931258688e-07,
+      "loss": 0.1429,
+      "step": 222380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8339087933276537e-07,
+      "loss": 0.1333,
+      "step": 222390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8114882740684276e-07,
+      "loss": 0.1354,
+      "step": 222400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7890677548092016e-07,
+      "loss": 0.1362,
+      "step": 222410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7666472355499755e-07,
+      "loss": 0.1385,
+      "step": 222420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7442267162907495e-07,
+      "loss": 0.1357,
+      "step": 222430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7218061970315234e-07,
+      "loss": 0.1398,
+      "step": 222440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6993856777722974e-07,
+      "loss": 0.1345,
+      "step": 222450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6769651585130713e-07,
+      "loss": 0.142,
+      "step": 222460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.654544639253845e-07,
+      "loss": 0.1385,
+      "step": 222470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.632124119994619e-07,
+      "loss": 0.1349,
+      "step": 222480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.609703600735393e-07,
+      "loss": 0.1407,
+      "step": 222490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.587283081476167e-07,
+      "loss": 0.1313,
+      "step": 222500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.564862562216941e-07,
+      "loss": 0.1397,
+      "step": 222510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.542442042957715e-07,
+      "loss": 0.1348,
+      "step": 222520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5200215236984894e-07,
+      "loss": 0.1371,
+      "step": 222530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.497601004439263e-07,
+      "loss": 0.1336,
+      "step": 222540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.475180485180037e-07,
+      "loss": 0.1354,
+      "step": 222550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.452759965920811e-07,
+      "loss": 0.1325,
+      "step": 222560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4303394466615847e-07,
+      "loss": 0.1399,
+      "step": 222570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4079189274023586e-07,
+      "loss": 0.1332,
+      "step": 222580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3854984081431326e-07,
+      "loss": 0.1336,
+      "step": 222590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3630778888839065e-07,
+      "loss": 0.1371,
+      "step": 222600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.340657369624681e-07,
+      "loss": 0.1365,
+      "step": 222610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3182368503654544e-07,
+      "loss": 0.1367,
+      "step": 222620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2958163311062283e-07,
+      "loss": 0.1329,
+      "step": 222630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.273395811847003e-07,
+      "loss": 0.1363,
+      "step": 222640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.250975292587776e-07,
+      "loss": 0.1342,
+      "step": 222650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2285547733285507e-07,
+      "loss": 0.1369,
+      "step": 222660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2061342540693246e-07,
+      "loss": 0.1364,
+      "step": 222670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.183713734810098e-07,
+      "loss": 0.1413,
+      "step": 222680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1612932155508725e-07,
+      "loss": 0.1441,
+      "step": 222690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1388726962916464e-07,
+      "loss": 0.1364,
+      "step": 222700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1164521770324204e-07,
+      "loss": 0.1419,
+      "step": 222710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0940316577731943e-07,
+      "loss": 0.1378,
+      "step": 222720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.071611138513968e-07,
+      "loss": 0.1389,
+      "step": 222730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.049190619254742e-07,
+      "loss": 0.1331,
+      "step": 222740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.026770099995516e-07,
+      "loss": 0.1372,
+      "step": 222750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0043495807362896e-07,
+      "loss": 0.1371,
+      "step": 222760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.981929061477064e-07,
+      "loss": 0.1363,
+      "step": 222770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.959508542217838e-07,
+      "loss": 0.1345,
+      "step": 222780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.937088022958612e-07,
+      "loss": 0.1396,
+      "step": 222790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.914667503699386e-07,
+      "loss": 0.1432,
+      "step": 222800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.89224698444016e-07,
+      "loss": 0.1297,
+      "step": 222810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.869826465180934e-07,
+      "loss": 0.1337,
+      "step": 222820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8474059459217077e-07,
+      "loss": 0.1365,
+      "step": 222830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8249854266624816e-07,
+      "loss": 0.134,
+      "step": 222840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8025649074032556e-07,
+      "loss": 0.1389,
+      "step": 222850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7801443881440295e-07,
+      "loss": 0.1394,
+      "step": 222860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7577238688848035e-07,
+      "loss": 0.1327,
+      "step": 222870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7353033496255774e-07,
+      "loss": 0.1369,
+      "step": 222880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7128828303663514e-07,
+      "loss": 0.1322,
+      "step": 222890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6904623111071253e-07,
+      "loss": 0.1374,
+      "step": 222900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.668041791847899e-07,
+      "loss": 0.1381,
+      "step": 222910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6456212725886737e-07,
+      "loss": 0.1368,
+      "step": 222920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.623200753329447e-07,
+      "loss": 0.1381,
+      "step": 222930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.600780234070221e-07,
+      "loss": 0.1359,
+      "step": 222940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5783597148109955e-07,
+      "loss": 0.1386,
+      "step": 222950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.555939195551769e-07,
+      "loss": 0.1368,
+      "step": 222960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.533518676292543e-07,
+      "loss": 0.1378,
+      "step": 222970
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.511098157033317e-07,
+      "loss": 0.1285,
+      "step": 222980
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.488677637774091e-07,
+      "loss": 0.1316,
+      "step": 222990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.466257118514865e-07,
+      "loss": 0.1374,
+      "step": 223000
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4438365992556387e-07,
+      "loss": 0.1357,
+      "step": 223010
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4214160799964126e-07,
+      "loss": 0.1325,
+      "step": 223020
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.398995560737187e-07,
+      "loss": 0.1386,
+      "step": 223030
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3765750414779608e-07,
+      "loss": 0.1413,
+      "step": 223040
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3541545222187347e-07,
+      "loss": 0.1331,
+      "step": 223050
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.331734002959509e-07,
+      "loss": 0.1335,
+      "step": 223060
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3093134837002826e-07,
+      "loss": 0.1399,
+      "step": 223070
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2868929644410565e-07,
+      "loss": 0.1432,
+      "step": 223080
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2644724451818302e-07,
+      "loss": 0.1359,
+      "step": 223090
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2420519259226044e-07,
+      "loss": 0.132,
+      "step": 223100
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2196314066633786e-07,
+      "loss": 0.1388,
+      "step": 223110
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1972108874041523e-07,
+      "loss": 0.1335,
+      "step": 223120
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1747903681449262e-07,
+      "loss": 0.1365,
+      "step": 223130
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1523698488857005e-07,
+      "loss": 0.1298,
+      "step": 223140
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.129949329626474e-07,
+      "loss": 0.1376,
+      "step": 223150
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1075288103672483e-07,
+      "loss": 0.1358,
+      "step": 223160
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0851082911080223e-07,
+      "loss": 0.1333,
+      "step": 223170
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.062687771848796e-07,
+      "loss": 0.1378,
+      "step": 223180
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0402672525895702e-07,
+      "loss": 0.1398,
+      "step": 223190
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0178467333303438e-07,
+      "loss": 0.1362,
+      "step": 223200
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.995426214071118e-07,
+      "loss": 0.1384,
+      "step": 223210
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.973005694811892e-07,
+      "loss": 0.1376,
+      "step": 223220
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.950585175552666e-07,
+      "loss": 0.1454,
+      "step": 223230
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.92816465629344e-07,
+      "loss": 0.1316,
+      "step": 223240
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9057441370342138e-07,
+      "loss": 0.1364,
+      "step": 223250
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8833236177749878e-07,
+      "loss": 0.1382,
+      "step": 223260
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8609030985157617e-07,
+      "loss": 0.1298,
+      "step": 223270
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8384825792565356e-07,
+      "loss": 0.1305,
+      "step": 223280
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8160620599973096e-07,
+      "loss": 0.1433,
+      "step": 223290
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7936415407380835e-07,
+      "loss": 0.1381,
+      "step": 223300
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7712210214788575e-07,
+      "loss": 0.1372,
+      "step": 223310
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7488005022196314e-07,
+      "loss": 0.1379,
+      "step": 223320
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7263799829604056e-07,
+      "loss": 0.1394,
+      "step": 223330
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7039594637011793e-07,
+      "loss": 0.1409,
+      "step": 223340
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6815389444419532e-07,
+      "loss": 0.1407,
+      "step": 223350
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6591184251827272e-07,
+      "loss": 0.1362,
+      "step": 223360
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6366979059235014e-07,
+      "loss": 0.1316,
+      "step": 223370
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6142773866642753e-07,
+      "loss": 0.1373,
+      "step": 223380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.591856867405049e-07,
+      "loss": 0.1378,
+      "step": 223390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5694363481458232e-07,
+      "loss": 0.1321,
+      "step": 223400
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5470158288865972e-07,
+      "loss": 0.1408,
+      "step": 223410
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.524595309627371e-07,
+      "loss": 0.1345,
+      "step": 223420
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5021747903681448e-07,
+      "loss": 0.1416,
+      "step": 223430
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.479754271108919e-07,
+      "loss": 0.1344,
+      "step": 223440
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.457333751849693e-07,
+      "loss": 0.1374,
+      "step": 223450
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.434913232590467e-07,
+      "loss": 0.138,
+      "step": 223460
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4124927133312408e-07,
+      "loss": 0.1327,
+      "step": 223470
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3900721940720148e-07,
+      "loss": 0.1426,
+      "step": 223480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3676516748127887e-07,
+      "loss": 0.1332,
+      "step": 223490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3452311555535627e-07,
+      "loss": 0.1307,
+      "step": 223500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3228106362943369e-07,
+      "loss": 0.1424,
+      "step": 223510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3003901170351105e-07,
+      "loss": 0.1412,
+      "step": 223520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2779695977758845e-07,
+      "loss": 0.1376,
+      "step": 223530
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2555490785166584e-07,
+      "loss": 0.1438,
+      "step": 223540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2331285592574326e-07,
+      "loss": 0.1394,
+      "step": 223550
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2107080399982063e-07,
+      "loss": 0.1354,
+      "step": 223560
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1882875207389804e-07,
+      "loss": 0.1355,
+      "step": 223570
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1658670014797545e-07,
+      "loss": 0.1333,
+      "step": 223580
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1434464822205283e-07,
+      "loss": 0.1331,
+      "step": 223590
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1210259629613022e-07,
+      "loss": 0.14,
+      "step": 223600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0986054437020762e-07,
+      "loss": 0.1418,
+      "step": 223610
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0761849244428502e-07,
+      "loss": 0.135,
+      "step": 223620
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0537644051836242e-07,
+      "loss": 0.1392,
+      "step": 223630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.031343885924398e-07,
+      "loss": 0.1416,
+      "step": 223640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0089233666651719e-07,
+      "loss": 0.1352,
+      "step": 223650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.887448993318685e-08,
+      "loss": 0.1394,
+      "step": 223660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.663243800726425e-08,
+      "loss": 0.1359,
+      "step": 223670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.439038608134166e-08,
+      "loss": 0.1388,
+      "step": 223680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.214833415541904e-08,
+      "loss": 0.1367,
+      "step": 223690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.990628222949645e-08,
+      "loss": 0.1406,
+      "step": 223700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.766423030357383e-08,
+      "loss": 0.1442,
+      "step": 223710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.542217837765123e-08,
+      "loss": 0.1386,
+      "step": 223720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.318012645172863e-08,
+      "loss": 0.1391,
+      "step": 223730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.093807452580602e-08,
+      "loss": 0.1375,
+      "step": 223740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.869602259988342e-08,
+      "loss": 0.1374,
+      "step": 223750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.645397067396081e-08,
+      "loss": 0.1409,
+      "step": 223760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.421191874803822e-08,
+      "loss": 0.1325,
+      "step": 223770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.19698668221156e-08,
+      "loss": 0.1333,
+      "step": 223780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.972781489619301e-08,
+      "loss": 0.1353,
+      "step": 223790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.748576297027039e-08,
+      "loss": 0.1394,
+      "step": 223800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.52437110443478e-08,
+      "loss": 0.136,
+      "step": 223810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.300165911842518e-08,
+      "loss": 0.1379,
+      "step": 223820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.075960719250258e-08,
+      "loss": 0.1371,
+      "step": 223830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.851755526657997e-08,
+      "loss": 0.1419,
+      "step": 223840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.627550334065737e-08,
+      "loss": 0.1372,
+      "step": 223850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.403345141473477e-08,
+      "loss": 0.136,
+      "step": 223860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.179139948881216e-08,
+      "loss": 0.1376,
+      "step": 223870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.954934756288956e-08,
+      "loss": 0.1394,
+      "step": 223880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7307295636966956e-08,
+      "loss": 0.1361,
+      "step": 223890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.506524371104435e-08,
+      "loss": 0.1393,
+      "step": 223900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2823191785121744e-08,
+      "loss": 0.1369,
+      "step": 223910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.058113985919914e-08,
+      "loss": 0.1375,
+      "step": 223920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.833908793327654e-08,
+      "loss": 0.137,
+      "step": 223930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.6097036007353934e-08,
+      "loss": 0.1411,
+      "step": 223940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.385498408143133e-08,
+      "loss": 0.138,
+      "step": 223950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.161293215550872e-08,
+      "loss": 0.1336,
+      "step": 223960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9370880229586117e-08,
+      "loss": 0.1389,
+      "step": 223970
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7128828303663518e-08,
+      "loss": 0.1345,
+      "step": 223980
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4886776377740912e-08,
+      "loss": 0.1394,
+      "step": 223990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2644724451818306e-08,
+      "loss": 0.1415,
+      "step": 224000
+    },
+    {
+      "epoch": 1.0,
+      "eval_NEIMS_canon_smiles": 0.94685,
+      "eval_NEIMS_daylight_tanimoto_hits_equals_1": 0.12601666666666667,
+      "eval_NEIMS_daylight_tanimoto_simil": 0.5248372483121932,
+      "eval_NEIMS_exact_mols": 0.12576666666666667,
+      "eval_NEIMS_exact_smiles": 0.12238333333333333,
+      "eval_NEIMS_loss": 0.17296098172664642,
+      "eval_NEIMS_matched_formulas": 0.65695,
+      "eval_NEIMS_morgan_tanimoto_simil": 0.4371484987482767,
+      "eval_NEIMS_morgan_tanimoto_simil_equals_1": 0.12631666666666666,
+      "eval_NEIMS_runtime": 713.0516,
+      "eval_NEIMS_samples_per_second": 84.145,
+      "eval_NEIMS_steps_per_second": 1.315,
+      "step": 224000
+    },
+    {
+      "epoch": 1.0,
+      "eval_RASSP_canon_smiles": 0.9435669683106822,
+      "eval_RASSP_daylight_tanimoto_hits_equals_1": 0.26875813937957055,
+      "eval_RASSP_daylight_tanimoto_simil": 0.6792164058093763,
+      "eval_RASSP_exact_mols": 0.2678064580759342,
+      "eval_RASSP_exact_smiles": 0.2604434500951681,
+      "eval_RASSP_loss": 0.11311110854148865,
+      "eval_RASSP_matched_formulas": 0.8309012588907069,
+      "eval_RASSP_morgan_tanimoto_simil": 0.567031216286806,
+      "eval_RASSP_morgan_tanimoto_simil_equals_1": 0.2680402043610378,
+      "eval_RASSP_runtime": 816.712,
+      "eval_RASSP_samples_per_second": 73.336,
+      "eval_RASSP_steps_per_second": 1.146,
+      "step": 224000
+    },
+    {
+      "epoch": 1.0,
+      "eval_NIST_canon_smiles": 0.8784114703481563,
+      "eval_NIST_daylight_tanimoto_hits_equals_1": 0.0188451573978777,
+      "eval_NIST_daylight_tanimoto_simil": 0.28303764326500064,
+      "eval_NIST_exact_mols": 0.018028888810022358,
+      "eval_NIST_exact_smiles": 0.016680271143130924,
+      "eval_NIST_loss": 1.8960157632827759,
+      "eval_NIST_matched_formulas": 0.10586648685097774,
+      "eval_NIST_morgan_tanimoto_simil": 0.235590992457229,
+      "eval_NIST_morgan_tanimoto_simil_equals_1": 0.018525747950456046,
+      "eval_NIST_runtime": 476.0566,
+      "eval_NIST_samples_per_second": 59.188,
+      "eval_NIST_steps_per_second": 0.926,
+      "step": 224000
+    }
+  ],
+  "max_steps": 224010,
+  "num_train_epochs": 9223372036854775807,
+  "total_flos": 1.6768754049282736e+19,
+  "trial_name": null,
+  "trial_params": null
+}