diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,3505 +1,11665 @@
 {
-  "best_metric": 0.7222222222222222,
-  "best_model_checkpoint": "D:\\HDD Tiziana\\Output-prova_melanoma\\checkpoint-212",
-  "epoch": 39.24528301886792,
-  "global_step": 520,
+  "best_metric": 0.9733333333333334,
+  "best_model_checkpoint": "D:\\HDD Tiziana\\Output-prova_melanoma\\checkpoint-94",
+  "epoch": 40.0,
+  "global_step": 1880,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.08,
-      "learning_rate": 1.9230769230769234e-07,
-      "loss": 1.482,
+      "epoch": 0.02,
+      "learning_rate": 5.319148936170213e-08,
+      "loss": 0.7312,
       "step": 1
     },
     {
-      "epoch": 0.15,
-      "learning_rate": 3.846153846153847e-07,
-      "loss": 1.3753,
+      "epoch": 0.04,
+      "learning_rate": 1.0638297872340426e-07,
+      "loss": 0.396,
       "step": 2
     },
     {
-      "epoch": 0.23,
-      "learning_rate": 5.76923076923077e-07,
-      "loss": 1.5431,
+      "epoch": 0.06,
+      "learning_rate": 1.5957446808510638e-07,
+      "loss": 0.6299,
       "step": 3
     },
     {
-      "epoch": 0.3,
-      "learning_rate": 7.692307692307694e-07,
-      "loss": 1.6805,
+      "epoch": 0.09,
+      "learning_rate": 2.1276595744680852e-07,
+      "loss": 0.4952,
       "step": 4
     },
     {
-      "epoch": 0.38,
-      "learning_rate": 9.615384615384617e-07,
-      "loss": 1.4787,
+      "epoch": 0.11,
+      "learning_rate": 2.6595744680851066e-07,
+      "loss": 0.5529,
       "step": 5
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 1.153846153846154e-06,
-      "loss": 1.6381,
+      "epoch": 0.13,
+      "learning_rate": 3.1914893617021275e-07,
+      "loss": 0.3591,
       "step": 6
     },
     {
-      "epoch": 0.53,
-      "learning_rate": 1.3461538461538462e-06,
-      "loss": 1.5637,
+      "epoch": 0.15,
+      "learning_rate": 3.723404255319149e-07,
+      "loss": 0.9101,
       "step": 7
     },
     {
-      "epoch": 0.6,
-      "learning_rate": 1.5384615384615387e-06,
-      "loss": 1.6039,
+      "epoch": 0.17,
+      "learning_rate": 4.2553191489361704e-07,
+      "loss": 0.6084,
       "step": 8
     },
     {
-      "epoch": 0.68,
-      "learning_rate": 1.7307692307692308e-06,
-      "loss": 1.2823,
+      "epoch": 0.19,
+      "learning_rate": 4.787234042553192e-07,
+      "loss": 0.4291,
       "step": 9
     },
     {
-      "epoch": 0.75,
-      "learning_rate": 1.9230769230769234e-06,
-      "loss": 1.4396,
+      "epoch": 0.21,
+      "learning_rate": 5.319148936170213e-07,
+      "loss": 0.5454,
       "step": 10
     },
     {
-      "epoch": 0.83,
-      "learning_rate": 2.1153846153846155e-06,
-      "loss": 1.352,
+      "epoch": 0.23,
+      "learning_rate": 5.851063829787235e-07,
+      "loss": 0.2325,
       "step": 11
     },
     {
-      "epoch": 0.91,
-      "learning_rate": 2.307692307692308e-06,
-      "loss": 1.2464,
+      "epoch": 0.26,
+      "learning_rate": 6.382978723404255e-07,
+      "loss": 0.3134,
       "step": 12
     },
     {
-      "epoch": 0.98,
-      "learning_rate": 2.5e-06,
-      "loss": 1.3691,
-      "step": 13
-    },
-    {
-      "epoch": 0.98,
-      "eval_accuracy": 0.4074074074074074,
-      "eval_loss": 1.263465166091919,
-      "eval_runtime": 1.145,
-      "eval_samples_per_second": 47.162,
-      "eval_steps_per_second": 3.493,
+      "epoch": 0.28,
+      "learning_rate": 6.914893617021278e-07,
+      "loss": 0.5571,
       "step": 13
     },
     {
-      "epoch": 1.06,
-      "learning_rate": 2.6923076923076923e-06,
-      "loss": 1.0168,
+      "epoch": 0.3,
+      "learning_rate": 7.446808510638298e-07,
+      "loss": 0.8986,
       "step": 14
     },
     {
-      "epoch": 1.13,
-      "learning_rate": 2.8846153846153845e-06,
-      "loss": 1.1878,
+      "epoch": 0.32,
+      "learning_rate": 7.97872340425532e-07,
+      "loss": 0.5506,
       "step": 15
     },
     {
-      "epoch": 1.21,
-      "learning_rate": 3.0769230769230774e-06,
-      "loss": 1.2215,
+      "epoch": 0.34,
+      "learning_rate": 8.510638297872341e-07,
+      "loss": 0.9027,
       "step": 16
     },
     {
-      "epoch": 1.28,
-      "learning_rate": 3.2692307692307696e-06,
-      "loss": 1.4159,
+      "epoch": 0.36,
+      "learning_rate": 9.042553191489363e-07,
+      "loss": 0.7196,
       "step": 17
     },
     {
-      "epoch": 1.36,
-      "learning_rate": 3.4615384615384617e-06,
-      "loss": 1.3136,
+      "epoch": 0.38,
+      "learning_rate": 9.574468085106384e-07,
+      "loss": 0.569,
       "step": 18
     },
     {
-      "epoch": 1.43,
-      "learning_rate": 3.653846153846154e-06,
-      "loss": 1.3873,
+      "epoch": 0.4,
+      "learning_rate": 1.0106382978723404e-06,
+      "loss": 0.542,
       "step": 19
     },
     {
-      "epoch": 1.51,
-      "learning_rate": 3.846153846153847e-06,
-      "loss": 1.1608,
+      "epoch": 0.43,
+      "learning_rate": 1.0638297872340427e-06,
+      "loss": 0.5898,
       "step": 20
     },
     {
-      "epoch": 1.58,
-      "learning_rate": 4.0384615384615385e-06,
-      "loss": 1.313,
+      "epoch": 0.45,
+      "learning_rate": 1.1170212765957447e-06,
+      "loss": 0.8707,
       "step": 21
     },
     {
-      "epoch": 1.66,
-      "learning_rate": 4.230769230769231e-06,
-      "loss": 1.1513,
+      "epoch": 0.47,
+      "learning_rate": 1.170212765957447e-06,
+      "loss": 0.2164,
       "step": 22
     },
     {
-      "epoch": 1.74,
-      "learning_rate": 4.423076923076924e-06,
-      "loss": 1.3801,
+      "epoch": 0.49,
+      "learning_rate": 1.223404255319149e-06,
+      "loss": 0.5452,
       "step": 23
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 4.615384615384616e-06,
-      "loss": 0.8811,
+      "epoch": 0.51,
+      "learning_rate": 1.276595744680851e-06,
+      "loss": 0.524,
       "step": 24
     },
     {
-      "epoch": 1.89,
-      "learning_rate": 4.807692307692308e-06,
-      "loss": 1.0194,
+      "epoch": 0.53,
+      "learning_rate": 1.3297872340425533e-06,
+      "loss": 0.5463,
       "step": 25
     },
     {
-      "epoch": 1.96,
-      "learning_rate": 5e-06,
-      "loss": 1.0847,
-      "step": 26
-    },
-    {
-      "epoch": 1.96,
-      "eval_accuracy": 0.5,
-      "eval_loss": 1.0837514400482178,
-      "eval_runtime": 0.934,
-      "eval_samples_per_second": 57.816,
-      "eval_steps_per_second": 4.283,
+      "epoch": 0.55,
+      "learning_rate": 1.3829787234042555e-06,
+      "loss": 0.469,
       "step": 26
     },
     {
-      "epoch": 2.04,
-      "learning_rate": 5.192307692307693e-06,
-      "loss": 0.9396,
+      "epoch": 0.57,
+      "learning_rate": 1.4361702127659578e-06,
+      "loss": 0.6523,
       "step": 27
     },
     {
-      "epoch": 2.11,
-      "learning_rate": 5.384615384615385e-06,
-      "loss": 0.9428,
+      "epoch": 0.6,
+      "learning_rate": 1.4893617021276596e-06,
+      "loss": 0.394,
       "step": 28
     },
     {
-      "epoch": 2.19,
-      "learning_rate": 5.576923076923077e-06,
-      "loss": 1.0722,
+      "epoch": 0.62,
+      "learning_rate": 1.5425531914893618e-06,
+      "loss": 0.1193,
       "step": 29
     },
     {
-      "epoch": 2.26,
-      "learning_rate": 5.769230769230769e-06,
-      "loss": 1.1936,
+      "epoch": 0.64,
+      "learning_rate": 1.595744680851064e-06,
+      "loss": 0.3905,
       "step": 30
     },
     {
-      "epoch": 2.34,
-      "learning_rate": 5.961538461538462e-06,
-      "loss": 1.1763,
+      "epoch": 0.66,
+      "learning_rate": 1.648936170212766e-06,
+      "loss": 0.5521,
       "step": 31
     },
     {
-      "epoch": 2.42,
-      "learning_rate": 6.153846153846155e-06,
-      "loss": 1.1016,
+      "epoch": 0.68,
+      "learning_rate": 1.7021276595744682e-06,
+      "loss": 0.3829,
       "step": 32
     },
     {
-      "epoch": 2.49,
-      "learning_rate": 6.3461538461538466e-06,
-      "loss": 1.0138,
+      "epoch": 0.7,
+      "learning_rate": 1.7553191489361704e-06,
+      "loss": 0.5464,
       "step": 33
     },
     {
-      "epoch": 2.57,
-      "learning_rate": 6.538461538461539e-06,
-      "loss": 0.9713,
+      "epoch": 0.72,
+      "learning_rate": 1.8085106382978727e-06,
+      "loss": 0.4134,
       "step": 34
     },
     {
-      "epoch": 2.64,
-      "learning_rate": 6.730769230769232e-06,
-      "loss": 1.0227,
+      "epoch": 0.74,
+      "learning_rate": 1.8617021276595745e-06,
+      "loss": 0.3238,
       "step": 35
     },
     {
-      "epoch": 2.72,
-      "learning_rate": 6.923076923076923e-06,
-      "loss": 1.0324,
+      "epoch": 0.77,
+      "learning_rate": 1.9148936170212767e-06,
+      "loss": 0.3327,
       "step": 36
     },
     {
-      "epoch": 2.79,
-      "learning_rate": 7.115384615384616e-06,
-      "loss": 1.0444,
+      "epoch": 0.79,
+      "learning_rate": 1.968085106382979e-06,
+      "loss": 0.6116,
       "step": 37
     },
     {
-      "epoch": 2.87,
-      "learning_rate": 7.307692307692308e-06,
-      "loss": 1.0133,
+      "epoch": 0.81,
+      "learning_rate": 2.021276595744681e-06,
+      "loss": 0.5777,
       "step": 38
     },
     {
-      "epoch": 2.94,
-      "learning_rate": 7.500000000000001e-06,
-      "loss": 0.9902,
-      "step": 39
-    },
-    {
-      "epoch": 2.94,
-      "eval_accuracy": 0.5185185185185185,
-      "eval_loss": 1.013327956199646,
-      "eval_runtime": 0.886,
-      "eval_samples_per_second": 60.948,
-      "eval_steps_per_second": 4.515,
+      "epoch": 0.83,
+      "learning_rate": 2.074468085106383e-06,
+      "loss": 0.4258,
       "step": 39
     },
     {
-      "epoch": 3.02,
-      "learning_rate": 7.692307692307694e-06,
-      "loss": 0.9945,
+      "epoch": 0.85,
+      "learning_rate": 2.1276595744680853e-06,
+      "loss": 0.4682,
       "step": 40
     },
     {
-      "epoch": 3.09,
-      "learning_rate": 7.884615384615384e-06,
-      "loss": 1.0984,
+      "epoch": 0.87,
+      "learning_rate": 2.1808510638297876e-06,
+      "loss": 0.4169,
       "step": 41
     },
     {
-      "epoch": 3.17,
-      "learning_rate": 8.076923076923077e-06,
-      "loss": 1.0475,
+      "epoch": 0.89,
+      "learning_rate": 2.2340425531914894e-06,
+      "loss": 0.45,
       "step": 42
     },
     {
-      "epoch": 3.25,
-      "learning_rate": 8.26923076923077e-06,
-      "loss": 1.005,
+      "epoch": 0.91,
+      "learning_rate": 2.2872340425531916e-06,
+      "loss": 0.3447,
       "step": 43
     },
     {
-      "epoch": 3.32,
-      "learning_rate": 8.461538461538462e-06,
-      "loss": 0.9797,
+      "epoch": 0.94,
+      "learning_rate": 2.340425531914894e-06,
+      "loss": 0.5947,
       "step": 44
     },
     {
-      "epoch": 3.4,
-      "learning_rate": 8.653846153846155e-06,
-      "loss": 1.0645,
+      "epoch": 0.96,
+      "learning_rate": 2.393617021276596e-06,
+      "loss": 0.6954,
       "step": 45
     },
     {
-      "epoch": 3.47,
-      "learning_rate": 8.846153846153847e-06,
-      "loss": 1.0021,
+      "epoch": 0.98,
+      "learning_rate": 2.446808510638298e-06,
+      "loss": 0.301,
       "step": 46
     },
     {
-      "epoch": 3.55,
-      "learning_rate": 9.03846153846154e-06,
-      "loss": 0.8505,
+      "epoch": 1.0,
+      "learning_rate": 2.5e-06,
+      "loss": 0.5144,
       "step": 47
     },
     {
-      "epoch": 3.62,
-      "learning_rate": 9.230769230769232e-06,
-      "loss": 1.0448,
+      "epoch": 1.0,
+      "eval_accuracy": 0.9666666666666667,
+      "eval_loss": 0.2076544165611267,
+      "eval_runtime": 30.3266,
+      "eval_samples_per_second": 4.946,
+      "eval_steps_per_second": 0.33,
+      "step": 47
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.553191489361702e-06,
+      "loss": 0.1981,
       "step": 48
     },
     {
-      "epoch": 3.7,
-      "learning_rate": 9.423076923076923e-06,
-      "loss": 0.9469,
+      "epoch": 1.04,
+      "learning_rate": 2.6063829787234047e-06,
+      "loss": 0.4807,
       "step": 49
     },
     {
-      "epoch": 3.77,
-      "learning_rate": 9.615384615384616e-06,
-      "loss": 0.954,
+      "epoch": 1.06,
+      "learning_rate": 2.6595744680851065e-06,
+      "loss": 0.257,
       "step": 50
     },
     {
-      "epoch": 3.85,
-      "learning_rate": 9.807692307692308e-06,
-      "loss": 1.0331,
+      "epoch": 1.09,
+      "learning_rate": 2.7127659574468084e-06,
+      "loss": 0.4549,
       "step": 51
     },
     {
-      "epoch": 3.92,
-      "learning_rate": 1e-05,
-      "loss": 0.9454,
+      "epoch": 1.11,
+      "learning_rate": 2.765957446808511e-06,
+      "loss": 0.7708,
       "step": 52
     },
     {
-      "epoch": 4.0,
-      "learning_rate": 9.97863247863248e-06,
-      "loss": 1.081,
-      "step": 53
-    },
-    {
-      "epoch": 4.0,
-      "eval_accuracy": 0.5370370370370371,
-      "eval_loss": 0.9912998080253601,
-      "eval_runtime": 0.897,
-      "eval_samples_per_second": 60.201,
-      "eval_steps_per_second": 4.459,
+      "epoch": 1.13,
+      "learning_rate": 2.819148936170213e-06,
+      "loss": 0.6563,
       "step": 53
     },
     {
-      "epoch": 4.08,
-      "learning_rate": 9.957264957264958e-06,
-      "loss": 0.9973,
+      "epoch": 1.15,
+      "learning_rate": 2.8723404255319155e-06,
+      "loss": 0.6773,
       "step": 54
     },
     {
-      "epoch": 4.15,
-      "learning_rate": 9.935897435897437e-06,
-      "loss": 0.9586,
+      "epoch": 1.17,
+      "learning_rate": 2.9255319148936174e-06,
+      "loss": 0.5255,
       "step": 55
     },
     {
-      "epoch": 4.23,
-      "learning_rate": 9.914529914529915e-06,
-      "loss": 0.9395,
+      "epoch": 1.19,
+      "learning_rate": 2.978723404255319e-06,
+      "loss": 0.162,
       "step": 56
     },
     {
-      "epoch": 4.3,
-      "learning_rate": 9.893162393162394e-06,
-      "loss": 0.9566,
+      "epoch": 1.21,
+      "learning_rate": 3.031914893617022e-06,
+      "loss": 0.2103,
       "step": 57
     },
     {
-      "epoch": 4.38,
-      "learning_rate": 9.871794871794872e-06,
-      "loss": 0.9207,
+      "epoch": 1.23,
+      "learning_rate": 3.0851063829787237e-06,
+      "loss": 0.55,
       "step": 58
     },
     {
-      "epoch": 4.45,
-      "learning_rate": 9.850427350427351e-06,
-      "loss": 0.9926,
+      "epoch": 1.26,
+      "learning_rate": 3.1382978723404255e-06,
+      "loss": 0.311,
       "step": 59
     },
     {
-      "epoch": 4.53,
-      "learning_rate": 9.82905982905983e-06,
-      "loss": 1.0408,
+      "epoch": 1.28,
+      "learning_rate": 3.191489361702128e-06,
+      "loss": 0.6394,
       "step": 60
     },
     {
-      "epoch": 4.6,
-      "learning_rate": 9.807692307692308e-06,
-      "loss": 0.9548,
+      "epoch": 1.3,
+      "learning_rate": 3.24468085106383e-06,
+      "loss": 0.5727,
       "step": 61
     },
     {
-      "epoch": 4.68,
-      "learning_rate": 9.786324786324787e-06,
-      "loss": 1.0464,
+      "epoch": 1.32,
+      "learning_rate": 3.297872340425532e-06,
+      "loss": 0.4183,
       "step": 62
     },
     {
-      "epoch": 4.75,
-      "learning_rate": 9.764957264957265e-06,
-      "loss": 1.0078,
+      "epoch": 1.34,
+      "learning_rate": 3.3510638297872345e-06,
+      "loss": 0.4414,
       "step": 63
     },
     {
-      "epoch": 4.83,
-      "learning_rate": 9.743589743589744e-06,
-      "loss": 1.0497,
+      "epoch": 1.36,
+      "learning_rate": 3.4042553191489363e-06,
+      "loss": 0.2669,
       "step": 64
     },
     {
-      "epoch": 4.91,
-      "learning_rate": 9.722222222222223e-06,
-      "loss": 0.9383,
+      "epoch": 1.38,
+      "learning_rate": 3.457446808510639e-06,
+      "loss": 0.4992,
       "step": 65
     },
     {
-      "epoch": 4.98,
-      "learning_rate": 9.700854700854701e-06,
-      "loss": 0.9947,
-      "step": 66
-    },
-    {
-      "epoch": 4.98,
-      "eval_accuracy": 0.5555555555555556,
-      "eval_loss": 0.9545946717262268,
-      "eval_runtime": 0.961,
-      "eval_samples_per_second": 56.191,
-      "eval_steps_per_second": 4.162,
+      "epoch": 1.4,
+      "learning_rate": 3.510638297872341e-06,
+      "loss": 0.2593,
       "step": 66
     },
     {
-      "epoch": 5.06,
-      "learning_rate": 9.67948717948718e-06,
-      "loss": 0.913,
+      "epoch": 1.43,
+      "learning_rate": 3.5638297872340426e-06,
+      "loss": 0.4269,
       "step": 67
     },
     {
-      "epoch": 5.13,
-      "learning_rate": 9.658119658119659e-06,
-      "loss": 1.011,
+      "epoch": 1.45,
+      "learning_rate": 3.6170212765957453e-06,
+      "loss": 0.6559,
       "step": 68
     },
     {
-      "epoch": 5.21,
-      "learning_rate": 9.636752136752137e-06,
-      "loss": 1.0539,
+      "epoch": 1.47,
+      "learning_rate": 3.670212765957447e-06,
+      "loss": 0.1297,
       "step": 69
     },
     {
-      "epoch": 5.28,
-      "learning_rate": 9.615384615384616e-06,
-      "loss": 1.0097,
+      "epoch": 1.49,
+      "learning_rate": 3.723404255319149e-06,
+      "loss": 0.7374,
       "step": 70
     },
     {
-      "epoch": 5.36,
-      "learning_rate": 9.594017094017094e-06,
-      "loss": 0.8825,
+      "epoch": 1.51,
+      "learning_rate": 3.7765957446808516e-06,
+      "loss": 0.5733,
       "step": 71
     },
     {
-      "epoch": 5.43,
-      "learning_rate": 9.572649572649575e-06,
-      "loss": 0.8532,
+      "epoch": 1.53,
+      "learning_rate": 3.8297872340425535e-06,
+      "loss": 0.4708,
       "step": 72
     },
     {
-      "epoch": 5.51,
-      "learning_rate": 9.551282051282053e-06,
-      "loss": 1.0515,
+      "epoch": 1.55,
+      "learning_rate": 3.882978723404256e-06,
+      "loss": 0.7006,
       "step": 73
     },
     {
-      "epoch": 5.58,
-      "learning_rate": 9.52991452991453e-06,
-      "loss": 0.9984,
+      "epoch": 1.57,
+      "learning_rate": 3.936170212765958e-06,
+      "loss": 0.4236,
       "step": 74
     },
     {
-      "epoch": 5.66,
-      "learning_rate": 9.508547008547009e-06,
-      "loss": 1.034,
+      "epoch": 1.6,
+      "learning_rate": 3.98936170212766e-06,
+      "loss": 0.208,
       "step": 75
     },
     {
-      "epoch": 5.74,
-      "learning_rate": 9.487179487179487e-06,
-      "loss": 0.9349,
+      "epoch": 1.62,
+      "learning_rate": 4.042553191489362e-06,
+      "loss": 0.256,
       "step": 76
     },
     {
-      "epoch": 5.81,
-      "learning_rate": 9.465811965811966e-06,
-      "loss": 1.077,
+      "epoch": 1.64,
+      "learning_rate": 4.095744680851064e-06,
+      "loss": 0.6313,
       "step": 77
     },
     {
-      "epoch": 5.89,
-      "learning_rate": 9.444444444444445e-06,
-      "loss": 1.0075,
+      "epoch": 1.66,
+      "learning_rate": 4.148936170212766e-06,
+      "loss": 0.3943,
       "step": 78
     },
     {
-      "epoch": 5.96,
-      "learning_rate": 9.423076923076923e-06,
-      "loss": 0.893,
-      "step": 79
-    },
-    {
-      "epoch": 5.96,
-      "eval_accuracy": 0.5555555555555556,
-      "eval_loss": 0.9382752180099487,
-      "eval_runtime": 0.992,
-      "eval_samples_per_second": 54.436,
-      "eval_steps_per_second": 4.032,
+      "epoch": 1.68,
+      "learning_rate": 4.202127659574468e-06,
+      "loss": 0.257,
       "step": 79
     },
     {
-      "epoch": 6.04,
-      "learning_rate": 9.401709401709402e-06,
-      "loss": 0.9507,
+      "epoch": 1.7,
+      "learning_rate": 4.255319148936171e-06,
+      "loss": 0.3632,
       "step": 80
     },
     {
-      "epoch": 6.11,
-      "learning_rate": 9.38034188034188e-06,
-      "loss": 0.8549,
+      "epoch": 1.72,
+      "learning_rate": 4.308510638297873e-06,
+      "loss": 0.5641,
       "step": 81
     },
     {
-      "epoch": 6.19,
-      "learning_rate": 9.358974358974359e-06,
-      "loss": 0.974,
+      "epoch": 1.74,
+      "learning_rate": 4.361702127659575e-06,
+      "loss": 0.3837,
       "step": 82
     },
     {
-      "epoch": 6.26,
-      "learning_rate": 9.33760683760684e-06,
-      "loss": 1.0265,
+      "epoch": 1.77,
+      "learning_rate": 4.414893617021277e-06,
+      "loss": 0.4097,
       "step": 83
     },
     {
-      "epoch": 6.34,
-      "learning_rate": 9.316239316239318e-06,
-      "loss": 1.022,
+      "epoch": 1.79,
+      "learning_rate": 4.468085106382979e-06,
+      "loss": 0.3017,
       "step": 84
     },
     {
-      "epoch": 6.42,
-      "learning_rate": 9.294871794871796e-06,
-      "loss": 0.9804,
+      "epoch": 1.81,
+      "learning_rate": 4.521276595744681e-06,
+      "loss": 0.1729,
       "step": 85
     },
     {
-      "epoch": 6.49,
-      "learning_rate": 9.273504273504275e-06,
-      "loss": 0.9192,
+      "epoch": 1.83,
+      "learning_rate": 4.574468085106383e-06,
+      "loss": 0.7201,
       "step": 86
     },
     {
-      "epoch": 6.57,
-      "learning_rate": 9.252136752136754e-06,
-      "loss": 0.9648,
+      "epoch": 1.85,
+      "learning_rate": 4.6276595744680855e-06,
+      "loss": 0.4627,
       "step": 87
     },
     {
-      "epoch": 6.64,
-      "learning_rate": 9.230769230769232e-06,
-      "loss": 0.9896,
+      "epoch": 1.87,
+      "learning_rate": 4.680851063829788e-06,
+      "loss": 0.3441,
       "step": 88
     },
     {
-      "epoch": 6.72,
-      "learning_rate": 9.20940170940171e-06,
-      "loss": 1.0195,
+      "epoch": 1.89,
+      "learning_rate": 4.73404255319149e-06,
+      "loss": 0.2474,
       "step": 89
     },
     {
-      "epoch": 6.79,
-      "learning_rate": 9.188034188034188e-06,
-      "loss": 0.9377,
+      "epoch": 1.91,
+      "learning_rate": 4.787234042553192e-06,
+      "loss": 0.3293,
       "step": 90
     },
     {
-      "epoch": 6.87,
-      "learning_rate": 9.166666666666666e-06,
-      "loss": 1.0336,
+      "epoch": 1.94,
+      "learning_rate": 4.840425531914894e-06,
+      "loss": 0.2504,
       "step": 91
     },
     {
-      "epoch": 6.94,
-      "learning_rate": 9.145299145299145e-06,
-      "loss": 1.0001,
+      "epoch": 1.96,
+      "learning_rate": 4.893617021276596e-06,
+      "loss": 0.5434,
       "step": 92
     },
     {
-      "epoch": 6.94,
-      "eval_accuracy": 0.5925925925925926,
-      "eval_loss": 0.934367299079895,
-      "eval_runtime": 0.919,
-      "eval_samples_per_second": 58.76,
-      "eval_steps_per_second": 4.353,
-      "step": 92
+      "epoch": 1.98,
+      "learning_rate": 4.946808510638298e-06,
+      "loss": 0.1894,
+      "step": 93
     },
     {
-      "epoch": 7.02,
-      "learning_rate": 9.123931623931624e-06,
-      "loss": 1.0079,
-      "step": 93
+      "epoch": 2.0,
+      "learning_rate": 5e-06,
+      "loss": 0.6803,
+      "step": 94
     },
     {
-      "epoch": 7.09,
-      "learning_rate": 9.102564102564104e-06,
-      "loss": 0.9886,
+      "epoch": 2.0,
+      "eval_accuracy": 0.9733333333333334,
+      "eval_loss": 0.1974334716796875,
+      "eval_runtime": 25.5602,
+      "eval_samples_per_second": 5.868,
+      "eval_steps_per_second": 0.391,
       "step": 94
     },
     {
-      "epoch": 7.17,
-      "learning_rate": 9.081196581196583e-06,
-      "loss": 0.9617,
+      "epoch": 2.02,
+      "learning_rate": 5.053191489361703e-06,
+      "loss": 0.3462,
       "step": 95
     },
     {
-      "epoch": 7.25,
-      "learning_rate": 9.059829059829061e-06,
-      "loss": 0.9792,
+      "epoch": 2.04,
+      "learning_rate": 5.106382978723404e-06,
+      "loss": 0.317,
       "step": 96
     },
     {
-      "epoch": 7.32,
-      "learning_rate": 9.03846153846154e-06,
-      "loss": 1.0374,
+      "epoch": 2.06,
+      "learning_rate": 5.159574468085107e-06,
+      "loss": 0.3044,
       "step": 97
     },
     {
-      "epoch": 7.4,
-      "learning_rate": 9.017094017094018e-06,
-      "loss": 0.9013,
+      "epoch": 2.09,
+      "learning_rate": 5.212765957446809e-06,
+      "loss": 0.3918,
       "step": 98
     },
     {
-      "epoch": 7.47,
-      "learning_rate": 8.995726495726497e-06,
-      "loss": 0.9687,
+      "epoch": 2.11,
+      "learning_rate": 5.265957446808511e-06,
+      "loss": 0.3004,
       "step": 99
     },
     {
-      "epoch": 7.55,
-      "learning_rate": 8.974358974358976e-06,
-      "loss": 1.0754,
+      "epoch": 2.13,
+      "learning_rate": 5.319148936170213e-06,
+      "loss": 0.3588,
       "step": 100
     },
     {
-      "epoch": 7.62,
-      "learning_rate": 8.952991452991454e-06,
-      "loss": 0.9205,
+      "epoch": 2.15,
+      "learning_rate": 5.372340425531915e-06,
+      "loss": 0.1615,
       "step": 101
     },
     {
-      "epoch": 7.7,
-      "learning_rate": 8.931623931623933e-06,
-      "loss": 0.8129,
+      "epoch": 2.17,
+      "learning_rate": 5.425531914893617e-06,
+      "loss": 0.1751,
       "step": 102
     },
     {
-      "epoch": 7.77,
-      "learning_rate": 8.910256410256411e-06,
-      "loss": 1.0025,
+      "epoch": 2.19,
+      "learning_rate": 5.47872340425532e-06,
+      "loss": 0.3063,
       "step": 103
     },
     {
-      "epoch": 7.85,
-      "learning_rate": 8.888888888888888e-06,
-      "loss": 0.9074,
+      "epoch": 2.21,
+      "learning_rate": 5.531914893617022e-06,
+      "loss": 0.3351,
       "step": 104
     },
     {
-      "epoch": 7.92,
-      "learning_rate": 8.867521367521369e-06,
-      "loss": 0.9338,
+      "epoch": 2.23,
+      "learning_rate": 5.5851063829787235e-06,
+      "loss": 0.3217,
       "step": 105
     },
     {
-      "epoch": 8.0,
-      "learning_rate": 8.846153846153847e-06,
-      "loss": 0.8996,
-      "step": 106
-    },
-    {
-      "epoch": 8.0,
-      "eval_accuracy": 0.5370370370370371,
-      "eval_loss": 0.9466460943222046,
-      "eval_runtime": 0.891,
-      "eval_samples_per_second": 60.606,
-      "eval_steps_per_second": 4.489,
+      "epoch": 2.26,
+      "learning_rate": 5.638297872340426e-06,
+      "loss": 0.3238,
       "step": 106
     },
     {
-      "epoch": 8.08,
-      "learning_rate": 8.824786324786326e-06,
-      "loss": 0.8974,
+      "epoch": 2.28,
+      "learning_rate": 5.691489361702128e-06,
+      "loss": 0.4434,
       "step": 107
     },
     {
-      "epoch": 8.15,
-      "learning_rate": 8.803418803418804e-06,
-      "loss": 0.9618,
+      "epoch": 2.3,
+      "learning_rate": 5.744680851063831e-06,
+      "loss": 0.224,
       "step": 108
     },
     {
-      "epoch": 8.23,
-      "learning_rate": 8.782051282051283e-06,
-      "loss": 1.0464,
+      "epoch": 2.32,
+      "learning_rate": 5.7978723404255325e-06,
+      "loss": 0.3132,
       "step": 109
     },
     {
-      "epoch": 8.3,
-      "learning_rate": 8.760683760683762e-06,
-      "loss": 0.9708,
+      "epoch": 2.34,
+      "learning_rate": 5.851063829787235e-06,
+      "loss": 0.3991,
       "step": 110
     },
     {
-      "epoch": 8.38,
-      "learning_rate": 8.73931623931624e-06,
-      "loss": 1.0742,
+      "epoch": 2.36,
+      "learning_rate": 5.904255319148937e-06,
+      "loss": 0.3712,
       "step": 111
     },
     {
-      "epoch": 8.45,
-      "learning_rate": 8.717948717948719e-06,
-      "loss": 1.0627,
+      "epoch": 2.38,
+      "learning_rate": 5.957446808510638e-06,
+      "loss": 0.1702,
       "step": 112
     },
     {
-      "epoch": 8.53,
-      "learning_rate": 8.696581196581197e-06,
-      "loss": 0.9862,
+      "epoch": 2.4,
+      "learning_rate": 6.010638297872341e-06,
+      "loss": 0.4146,
       "step": 113
     },
     {
-      "epoch": 8.6,
-      "learning_rate": 8.675213675213676e-06,
-      "loss": 0.9358,
+      "epoch": 2.43,
+      "learning_rate": 6.063829787234044e-06,
+      "loss": 0.5029,
       "step": 114
     },
     {
-      "epoch": 8.68,
-      "learning_rate": 8.653846153846155e-06,
-      "loss": 0.9234,
+      "epoch": 2.45,
+      "learning_rate": 6.117021276595745e-06,
+      "loss": 0.1649,
       "step": 115
     },
     {
-      "epoch": 8.75,
-      "learning_rate": 8.632478632478633e-06,
-      "loss": 0.9072,
+      "epoch": 2.47,
+      "learning_rate": 6.170212765957447e-06,
+      "loss": 0.1779,
       "step": 116
     },
     {
-      "epoch": 8.83,
-      "learning_rate": 8.611111111111112e-06,
-      "loss": 0.9825,
+      "epoch": 2.49,
+      "learning_rate": 6.22340425531915e-06,
+      "loss": 0.4024,
       "step": 117
     },
     {
-      "epoch": 8.91,
-      "learning_rate": 8.58974358974359e-06,
-      "loss": 0.8957,
+      "epoch": 2.51,
+      "learning_rate": 6.276595744680851e-06,
+      "loss": 0.3909,
       "step": 118
     },
     {
-      "epoch": 8.98,
-      "learning_rate": 8.568376068376069e-06,
-      "loss": 0.8788,
-      "step": 119
-    },
-    {
-      "epoch": 8.98,
-      "eval_accuracy": 0.5555555555555556,
-      "eval_loss": 0.9110423922538757,
-      "eval_runtime": 1.006,
-      "eval_samples_per_second": 53.678,
-      "eval_steps_per_second": 3.976,
+      "epoch": 2.53,
+      "learning_rate": 6.329787234042554e-06,
+      "loss": 0.3265,
       "step": 119
     },
     {
-      "epoch": 9.06,
-      "learning_rate": 8.547008547008548e-06,
-      "loss": 0.809,
+      "epoch": 2.55,
+      "learning_rate": 6.382978723404256e-06,
+      "loss": 0.3744,
       "step": 120
     },
     {
-      "epoch": 9.13,
-      "learning_rate": 8.525641025641026e-06,
-      "loss": 0.9355,
+      "epoch": 2.57,
+      "learning_rate": 6.436170212765958e-06,
+      "loss": 0.3084,
       "step": 121
     },
     {
-      "epoch": 9.21,
-      "learning_rate": 8.504273504273505e-06,
-      "loss": 0.9886,
+      "epoch": 2.6,
+      "learning_rate": 6.48936170212766e-06,
+      "loss": 0.1718,
       "step": 122
     },
     {
-      "epoch": 9.28,
-      "learning_rate": 8.482905982905983e-06,
-      "loss": 0.9477,
+      "epoch": 2.62,
+      "learning_rate": 6.542553191489362e-06,
+      "loss": 0.3708,
       "step": 123
     },
     {
-      "epoch": 9.36,
-      "learning_rate": 8.461538461538462e-06,
-      "loss": 0.9334,
+      "epoch": 2.64,
+      "learning_rate": 6.595744680851064e-06,
+      "loss": 0.1746,
       "step": 124
     },
     {
-      "epoch": 9.43,
-      "learning_rate": 8.44017094017094e-06,
-      "loss": 1.0755,
+      "epoch": 2.66,
+      "learning_rate": 6.648936170212767e-06,
+      "loss": 0.3368,
       "step": 125
     },
     {
-      "epoch": 9.51,
-      "learning_rate": 8.41880341880342e-06,
-      "loss": 0.9141,
+      "epoch": 2.68,
+      "learning_rate": 6.702127659574469e-06,
+      "loss": 0.1741,
       "step": 126
     },
     {
-      "epoch": 9.58,
-      "learning_rate": 8.397435897435898e-06,
-      "loss": 0.8678,
+      "epoch": 2.7,
+      "learning_rate": 6.75531914893617e-06,
+      "loss": 0.4534,
       "step": 127
     },
     {
-      "epoch": 9.66,
-      "learning_rate": 8.376068376068377e-06,
-      "loss": 0.9334,
+      "epoch": 2.72,
+      "learning_rate": 6.808510638297873e-06,
+      "loss": 0.319,
       "step": 128
     },
     {
-      "epoch": 9.74,
-      "learning_rate": 8.354700854700855e-06,
-      "loss": 0.9729,
+      "epoch": 2.74,
+      "learning_rate": 6.861702127659575e-06,
+      "loss": 0.281,
       "step": 129
     },
     {
-      "epoch": 9.81,
-      "learning_rate": 8.333333333333334e-06,
-      "loss": 0.9144,
+      "epoch": 2.77,
+      "learning_rate": 6.914893617021278e-06,
+      "loss": 0.4105,
       "step": 130
     },
     {
-      "epoch": 9.89,
-      "learning_rate": 8.311965811965812e-06,
-      "loss": 0.8939,
+      "epoch": 2.79,
+      "learning_rate": 6.968085106382979e-06,
+      "loss": 0.1162,
       "step": 131
     },
     {
-      "epoch": 9.96,
-      "learning_rate": 8.290598290598293e-06,
-      "loss": 0.9626,
-      "step": 132
-    },
-    {
-      "epoch": 9.96,
-      "eval_accuracy": 0.5925925925925926,
-      "eval_loss": 0.8969193696975708,
-      "eval_runtime": 0.897,
-      "eval_samples_per_second": 60.201,
-      "eval_steps_per_second": 4.459,
+      "epoch": 2.81,
+      "learning_rate": 7.021276595744682e-06,
+      "loss": 0.2702,
       "step": 132
     },
     {
-      "epoch": 10.04,
-      "learning_rate": 8.26923076923077e-06,
-      "loss": 0.8669,
+      "epoch": 2.83,
+      "learning_rate": 7.074468085106384e-06,
+      "loss": 0.4616,
       "step": 133
     },
     {
-      "epoch": 10.11,
-      "learning_rate": 8.247863247863248e-06,
-      "loss": 0.9141,
+      "epoch": 2.85,
+      "learning_rate": 7.127659574468085e-06,
+      "loss": 0.5854,
       "step": 134
     },
     {
-      "epoch": 10.19,
-      "learning_rate": 8.226495726495727e-06,
-      "loss": 0.9647,
+      "epoch": 2.87,
+      "learning_rate": 7.1808510638297875e-06,
+      "loss": 0.438,
       "step": 135
     },
     {
-      "epoch": 10.26,
-      "learning_rate": 8.205128205128205e-06,
-      "loss": 0.9626,
+      "epoch": 2.89,
+      "learning_rate": 7.234042553191491e-06,
+      "loss": 0.1504,
       "step": 136
     },
     {
-      "epoch": 10.34,
-      "learning_rate": 8.183760683760684e-06,
-      "loss": 1.031,
+      "epoch": 2.91,
+      "learning_rate": 7.287234042553192e-06,
+      "loss": 0.2831,
       "step": 137
     },
     {
-      "epoch": 10.42,
-      "learning_rate": 8.162393162393163e-06,
-      "loss": 0.951,
+      "epoch": 2.94,
+      "learning_rate": 7.340425531914894e-06,
+      "loss": 0.2162,
       "step": 138
     },
     {
-      "epoch": 10.49,
-      "learning_rate": 8.141025641025641e-06,
-      "loss": 0.9853,
+      "epoch": 2.96,
+      "learning_rate": 7.3936170212765965e-06,
+      "loss": 0.3657,
       "step": 139
     },
     {
-      "epoch": 10.57,
-      "learning_rate": 8.11965811965812e-06,
-      "loss": 1.0287,
+      "epoch": 2.98,
+      "learning_rate": 7.446808510638298e-06,
+      "loss": 0.2698,
       "step": 140
     },
     {
-      "epoch": 10.64,
-      "learning_rate": 8.098290598290598e-06,
-      "loss": 0.896,
+      "epoch": 3.0,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.6678,
       "step": 141
     },
     {
-      "epoch": 10.72,
-      "learning_rate": 8.076923076923077e-06,
-      "loss": 0.9207,
+      "epoch": 3.0,
+      "eval_accuracy": 0.9733333333333334,
+      "eval_loss": 0.1826505810022354,
+      "eval_runtime": 28.848,
+      "eval_samples_per_second": 5.2,
+      "eval_steps_per_second": 0.347,
+      "step": 141
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.553191489361703e-06,
+      "loss": 0.2271,
       "step": 142
     },
     {
-      "epoch": 10.79,
-      "learning_rate": 8.055555555555557e-06,
-      "loss": 0.8187,
+      "epoch": 3.04,
+      "learning_rate": 7.606382978723405e-06,
+      "loss": 0.16,
       "step": 143
     },
     {
-      "epoch": 10.87,
-      "learning_rate": 8.034188034188036e-06,
-      "loss": 0.9038,
+      "epoch": 3.06,
+      "learning_rate": 7.659574468085107e-06,
+      "loss": 0.1593,
       "step": 144
     },
     {
-      "epoch": 10.94,
-      "learning_rate": 8.012820512820515e-06,
-      "loss": 0.8905,
-      "step": 145
-    },
-    {
-      "epoch": 10.94,
-      "eval_accuracy": 0.5925925925925926,
-      "eval_loss": 0.8684723377227783,
-      "eval_runtime": 0.948,
-      "eval_samples_per_second": 56.962,
-      "eval_steps_per_second": 4.219,
+      "epoch": 3.09,
+      "learning_rate": 7.71276595744681e-06,
+      "loss": 0.4488,
       "step": 145
     },
     {
-      "epoch": 11.02,
-      "learning_rate": 7.991452991452993e-06,
-      "loss": 0.7527,
+      "epoch": 3.11,
+      "learning_rate": 7.765957446808511e-06,
+      "loss": 0.2494,
       "step": 146
     },
     {
-      "epoch": 11.09,
-      "learning_rate": 7.970085470085472e-06,
-      "loss": 0.9236,
+      "epoch": 3.13,
+      "learning_rate": 7.819148936170213e-06,
+      "loss": 0.263,
       "step": 147
     },
     {
-      "epoch": 11.17,
-      "learning_rate": 7.948717948717949e-06,
-      "loss": 1.0422,
+      "epoch": 3.15,
+      "learning_rate": 7.872340425531916e-06,
+      "loss": 0.2376,
       "step": 148
     },
     {
-      "epoch": 11.25,
-      "learning_rate": 7.927350427350427e-06,
-      "loss": 0.8892,
+      "epoch": 3.17,
+      "learning_rate": 7.925531914893617e-06,
+      "loss": 0.3901,
       "step": 149
     },
     {
-      "epoch": 11.32,
-      "learning_rate": 7.905982905982906e-06,
-      "loss": 1.0063,
+      "epoch": 3.19,
+      "learning_rate": 7.97872340425532e-06,
+      "loss": 0.3074,
       "step": 150
     },
     {
-      "epoch": 11.4,
-      "learning_rate": 7.884615384615384e-06,
-      "loss": 0.9112,
+      "epoch": 3.21,
+      "learning_rate": 8.031914893617022e-06,
+      "loss": 0.1844,
       "step": 151
     },
     {
-      "epoch": 11.47,
-      "learning_rate": 7.863247863247863e-06,
-      "loss": 0.9458,
+      "epoch": 3.23,
+      "learning_rate": 8.085106382978723e-06,
+      "loss": 0.2638,
       "step": 152
     },
     {
-      "epoch": 11.55,
-      "learning_rate": 7.841880341880342e-06,
-      "loss": 0.8673,
+      "epoch": 3.26,
+      "learning_rate": 8.138297872340426e-06,
+      "loss": 0.3396,
       "step": 153
     },
     {
-      "epoch": 11.62,
-      "learning_rate": 7.820512820512822e-06,
-      "loss": 0.9388,
+      "epoch": 3.28,
+      "learning_rate": 8.191489361702128e-06,
+      "loss": 0.5432,
       "step": 154
     },
     {
-      "epoch": 11.7,
-      "learning_rate": 7.7991452991453e-06,
-      "loss": 0.9432,
+      "epoch": 3.3,
+      "learning_rate": 8.24468085106383e-06,
+      "loss": 0.2503,
       "step": 155
     },
     {
-      "epoch": 11.77,
-      "learning_rate": 7.77777777777778e-06,
-      "loss": 0.8781,
+      "epoch": 3.32,
+      "learning_rate": 8.297872340425532e-06,
+      "loss": 0.1522,
       "step": 156
     },
     {
-      "epoch": 11.85,
-      "learning_rate": 7.756410256410258e-06,
-      "loss": 0.8851,
+      "epoch": 3.34,
+      "learning_rate": 8.351063829787235e-06,
+      "loss": 0.3099,
       "step": 157
     },
     {
-      "epoch": 11.92,
-      "learning_rate": 7.735042735042736e-06,
-      "loss": 0.8097,
+      "epoch": 3.36,
+      "learning_rate": 8.404255319148937e-06,
+      "loss": 0.2561,
       "step": 158
     },
     {
-      "epoch": 12.0,
-      "learning_rate": 7.713675213675215e-06,
-      "loss": 0.9012,
-      "step": 159
-    },
-    {
-      "epoch": 12.0,
-      "eval_accuracy": 0.5925925925925926,
-      "eval_loss": 0.8695763349533081,
-      "eval_runtime": 0.911,
-      "eval_samples_per_second": 59.276,
-      "eval_steps_per_second": 4.391,
+      "epoch": 3.38,
+      "learning_rate": 8.457446808510638e-06,
+      "loss": 0.6562,
       "step": 159
     },
     {
-      "epoch": 12.08,
-      "learning_rate": 7.692307692307694e-06,
-      "loss": 0.7753,
+      "epoch": 3.4,
+      "learning_rate": 8.510638297872341e-06,
+      "loss": 0.1523,
       "step": 160
     },
     {
-      "epoch": 12.15,
-      "learning_rate": 7.670940170940172e-06,
-      "loss": 0.9428,
+      "epoch": 3.43,
+      "learning_rate": 8.563829787234044e-06,
+      "loss": 0.2853,
       "step": 161
     },
     {
-      "epoch": 12.23,
-      "learning_rate": 7.649572649572649e-06,
-      "loss": 0.7526,
+      "epoch": 3.45,
+      "learning_rate": 8.617021276595746e-06,
+      "loss": 0.3714,
       "step": 162
     },
     {
-      "epoch": 12.3,
-      "learning_rate": 7.6282051282051286e-06,
-      "loss": 0.9018,
+      "epoch": 3.47,
+      "learning_rate": 8.670212765957447e-06,
+      "loss": 0.3193,
       "step": 163
     },
     {
-      "epoch": 12.38,
-      "learning_rate": 7.606837606837607e-06,
-      "loss": 0.9759,
+      "epoch": 3.49,
+      "learning_rate": 8.72340425531915e-06,
+      "loss": 0.4281,
       "step": 164
     },
     {
-      "epoch": 12.45,
-      "learning_rate": 7.585470085470086e-06,
-      "loss": 0.9882,
+      "epoch": 3.51,
+      "learning_rate": 8.776595744680852e-06,
+      "loss": 0.2143,
       "step": 165
     },
     {
-      "epoch": 12.53,
-      "learning_rate": 7.564102564102564e-06,
-      "loss": 0.8978,
+      "epoch": 3.53,
+      "learning_rate": 8.829787234042555e-06,
+      "loss": 0.2817,
       "step": 166
     },
     {
-      "epoch": 12.6,
-      "learning_rate": 7.542735042735043e-06,
-      "loss": 0.9279,
+      "epoch": 3.55,
+      "learning_rate": 8.882978723404256e-06,
+      "loss": 0.2315,
       "step": 167
     },
     {
-      "epoch": 12.68,
-      "learning_rate": 7.521367521367522e-06,
-      "loss": 0.8842,
+      "epoch": 3.57,
+      "learning_rate": 8.936170212765958e-06,
+      "loss": 0.3649,
       "step": 168
     },
     {
-      "epoch": 12.75,
-      "learning_rate": 7.500000000000001e-06,
-      "loss": 0.8827,
+      "epoch": 3.6,
+      "learning_rate": 8.98936170212766e-06,
+      "loss": 0.2411,
       "step": 169
     },
     {
-      "epoch": 12.83,
-      "learning_rate": 7.47863247863248e-06,
-      "loss": 0.8801,
+      "epoch": 3.62,
+      "learning_rate": 9.042553191489362e-06,
+      "loss": 0.246,
       "step": 170
     },
     {
-      "epoch": 12.91,
-      "learning_rate": 7.457264957264958e-06,
-      "loss": 0.8813,
+      "epoch": 3.64,
+      "learning_rate": 9.095744680851063e-06,
+      "loss": 0.4795,
       "step": 171
     },
     {
-      "epoch": 12.98,
-      "learning_rate": 7.435897435897437e-06,
-      "loss": 1.0567,
-      "step": 172
-    },
-    {
-      "epoch": 12.98,
-      "eval_accuracy": 0.48148148148148145,
-      "eval_loss": 0.8707943558692932,
-      "eval_runtime": 0.9703,
-      "eval_samples_per_second": 55.654,
-      "eval_steps_per_second": 4.123,
+      "epoch": 3.66,
+      "learning_rate": 9.148936170212767e-06,
+      "loss": 0.3337,
       "step": 172
     },
     {
-      "epoch": 13.06,
-      "learning_rate": 7.4145299145299155e-06,
-      "loss": 0.8535,
+      "epoch": 3.68,
+      "learning_rate": 9.20212765957447e-06,
+      "loss": 0.4258,
       "step": 173
     },
     {
-      "epoch": 13.13,
-      "learning_rate": 7.393162393162394e-06,
-      "loss": 0.838,
+      "epoch": 3.7,
+      "learning_rate": 9.255319148936171e-06,
+      "loss": 0.4237,
       "step": 174
     },
     {
-      "epoch": 13.21,
-      "learning_rate": 7.371794871794873e-06,
-      "loss": 0.7775,
+      "epoch": 3.72,
+      "learning_rate": 9.308510638297872e-06,
+      "loss": 0.2356,
       "step": 175
     },
     {
-      "epoch": 13.28,
-      "learning_rate": 7.350427350427351e-06,
-      "loss": 0.9342,
+      "epoch": 3.74,
+      "learning_rate": 9.361702127659576e-06,
+      "loss": 0.2685,
       "step": 176
     },
     {
-      "epoch": 13.36,
-      "learning_rate": 7.329059829059829e-06,
-      "loss": 0.8631,
+      "epoch": 3.77,
+      "learning_rate": 9.414893617021279e-06,
+      "loss": 0.2206,
       "step": 177
     },
     {
-      "epoch": 13.43,
-      "learning_rate": 7.307692307692308e-06,
-      "loss": 0.7946,
+      "epoch": 3.79,
+      "learning_rate": 9.46808510638298e-06,
+      "loss": 0.3711,
       "step": 178
     },
     {
-      "epoch": 13.51,
-      "learning_rate": 7.286324786324786e-06,
-      "loss": 1.0964,
+      "epoch": 3.81,
+      "learning_rate": 9.521276595744681e-06,
+      "loss": 0.3985,
       "step": 179
     },
     {
-      "epoch": 13.58,
-      "learning_rate": 7.264957264957266e-06,
-      "loss": 0.778,
+      "epoch": 3.83,
+      "learning_rate": 9.574468085106385e-06,
+      "loss": 0.1682,
       "step": 180
     },
     {
-      "epoch": 13.66,
-      "learning_rate": 7.243589743589744e-06,
-      "loss": 0.927,
+      "epoch": 3.85,
+      "learning_rate": 9.627659574468086e-06,
+      "loss": 0.1705,
       "step": 181
     },
     {
-      "epoch": 13.74,
-      "learning_rate": 7.222222222222223e-06,
-      "loss": 0.9679,
+      "epoch": 3.87,
+      "learning_rate": 9.680851063829787e-06,
+      "loss": 0.2066,
       "step": 182
     },
     {
-      "epoch": 13.81,
-      "learning_rate": 7.2008547008547015e-06,
-      "loss": 0.9749,
+      "epoch": 3.89,
+      "learning_rate": 9.73404255319149e-06,
+      "loss": 0.2966,
       "step": 183
     },
     {
-      "epoch": 13.89,
-      "learning_rate": 7.17948717948718e-06,
-      "loss": 0.9278,
+      "epoch": 3.91,
+      "learning_rate": 9.787234042553192e-06,
+      "loss": 0.3677,
       "step": 184
     },
     {
-      "epoch": 13.96,
-      "learning_rate": 7.158119658119659e-06,
-      "loss": 0.8742,
-      "step": 185
-    },
-    {
-      "epoch": 13.96,
-      "eval_accuracy": 0.6111111111111112,
-      "eval_loss": 0.8361570835113525,
-      "eval_runtime": 0.901,
-      "eval_samples_per_second": 59.933,
-      "eval_steps_per_second": 4.44,
+      "epoch": 3.94,
+      "learning_rate": 9.840425531914895e-06,
+      "loss": 0.154,
       "step": 185
     },
     {
-      "epoch": 14.04,
-      "learning_rate": 7.136752136752137e-06,
-      "loss": 0.8267,
+      "epoch": 3.96,
+      "learning_rate": 9.893617021276596e-06,
+      "loss": 0.3762,
       "step": 186
     },
     {
-      "epoch": 14.11,
-      "learning_rate": 7.115384615384616e-06,
-      "loss": 0.8073,
+      "epoch": 3.98,
+      "learning_rate": 9.946808510638298e-06,
+      "loss": 0.2466,
       "step": 187
     },
     {
-      "epoch": 14.19,
-      "learning_rate": 7.0940170940170945e-06,
-      "loss": 0.979,
+      "epoch": 4.0,
+      "learning_rate": 1e-05,
+      "loss": 0.081,
       "step": 188
     },
     {
-      "epoch": 14.26,
-      "learning_rate": 7.072649572649574e-06,
-      "loss": 0.7523,
+      "epoch": 4.0,
+      "eval_accuracy": 0.9666666666666667,
+      "eval_loss": 0.15473408997058868,
+      "eval_runtime": 27.8581,
+      "eval_samples_per_second": 5.384,
+      "eval_steps_per_second": 0.359,
+      "step": 188
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 9.994089834515367e-06,
+      "loss": 0.2889,
       "step": 189
     },
     {
-      "epoch": 14.34,
-      "learning_rate": 7.051282051282053e-06,
-      "loss": 0.9303,
+      "epoch": 4.04,
+      "learning_rate": 9.988179669030735e-06,
+      "loss": 0.1362,
       "step": 190
     },
     {
-      "epoch": 14.42,
-      "learning_rate": 7.02991452991453e-06,
-      "loss": 0.9685,
+      "epoch": 4.06,
+      "learning_rate": 9.982269503546101e-06,
+      "loss": 0.2528,
       "step": 191
     },
     {
-      "epoch": 14.49,
-      "learning_rate": 7.008547008547009e-06,
-      "loss": 0.8666,
+      "epoch": 4.09,
+      "learning_rate": 9.976359338061467e-06,
+      "loss": 0.3166,
       "step": 192
     },
     {
-      "epoch": 14.57,
-      "learning_rate": 6.9871794871794876e-06,
-      "loss": 1.0187,
+      "epoch": 4.11,
+      "learning_rate": 9.970449172576833e-06,
+      "loss": 0.2386,
       "step": 193
     },
     {
-      "epoch": 14.64,
-      "learning_rate": 6.965811965811966e-06,
-      "loss": 0.8791,
+      "epoch": 4.13,
+      "learning_rate": 9.9645390070922e-06,
+      "loss": 0.2631,
       "step": 194
     },
     {
-      "epoch": 14.72,
-      "learning_rate": 6.944444444444445e-06,
-      "loss": 0.787,
+      "epoch": 4.15,
+      "learning_rate": 9.958628841607565e-06,
+      "loss": 0.2089,
       "step": 195
     },
     {
-      "epoch": 14.79,
-      "learning_rate": 6.923076923076923e-06,
-      "loss": 0.933,
+      "epoch": 4.17,
+      "learning_rate": 9.952718676122932e-06,
+      "loss": 0.1153,
       "step": 196
     },
     {
-      "epoch": 14.87,
-      "learning_rate": 6.901709401709402e-06,
-      "loss": 0.9288,
+      "epoch": 4.19,
+      "learning_rate": 9.946808510638298e-06,
+      "loss": 0.2531,
       "step": 197
     },
     {
-      "epoch": 14.94,
-      "learning_rate": 6.880341880341881e-06,
-      "loss": 0.8001,
-      "step": 198
-    },
-    {
-      "epoch": 14.94,
-      "eval_accuracy": 0.5925925925925926,
-      "eval_loss": 0.8309553861618042,
-      "eval_runtime": 0.886,
-      "eval_samples_per_second": 60.948,
-      "eval_steps_per_second": 4.515,
+      "epoch": 4.21,
+      "learning_rate": 9.940898345153666e-06,
+      "loss": 0.4003,
       "step": 198
     },
     {
-      "epoch": 15.02,
-      "learning_rate": 6.858974358974359e-06,
-      "loss": 0.7482,
+      "epoch": 4.23,
+      "learning_rate": 9.934988179669032e-06,
+      "loss": 0.3401,
       "step": 199
     },
     {
-      "epoch": 15.09,
-      "learning_rate": 6.837606837606839e-06,
-      "loss": 0.9188,
+      "epoch": 4.26,
+      "learning_rate": 9.929078014184398e-06,
+      "loss": 0.2579,
       "step": 200
     },
     {
-      "epoch": 15.17,
-      "learning_rate": 6.816239316239317e-06,
-      "loss": 0.8423,
+      "epoch": 4.28,
+      "learning_rate": 9.923167848699764e-06,
+      "loss": 0.2276,
       "step": 201
     },
     {
-      "epoch": 15.25,
-      "learning_rate": 6.794871794871796e-06,
-      "loss": 0.8319,
+      "epoch": 4.3,
+      "learning_rate": 9.917257683215132e-06,
+      "loss": 0.1539,
       "step": 202
     },
     {
-      "epoch": 15.32,
-      "learning_rate": 6.7735042735042745e-06,
-      "loss": 0.9125,
+      "epoch": 4.32,
+      "learning_rate": 9.911347517730498e-06,
+      "loss": 0.1157,
       "step": 203
     },
     {
-      "epoch": 15.4,
-      "learning_rate": 6.752136752136753e-06,
-      "loss": 0.9435,
+      "epoch": 4.34,
+      "learning_rate": 9.905437352245864e-06,
+      "loss": 0.2896,
       "step": 204
     },
     {
-      "epoch": 15.47,
-      "learning_rate": 6.730769230769232e-06,
-      "loss": 0.8235,
+      "epoch": 4.36,
+      "learning_rate": 9.89952718676123e-06,
+      "loss": 0.1379,
       "step": 205
     },
     {
-      "epoch": 15.55,
-      "learning_rate": 6.7094017094017094e-06,
-      "loss": 0.934,
+      "epoch": 4.38,
+      "learning_rate": 9.893617021276596e-06,
+      "loss": 0.1045,
       "step": 206
     },
     {
-      "epoch": 15.62,
-      "learning_rate": 6.688034188034188e-06,
-      "loss": 0.8989,
+      "epoch": 4.4,
+      "learning_rate": 9.887706855791962e-06,
+      "loss": 0.2013,
       "step": 207
     },
     {
-      "epoch": 15.7,
-      "learning_rate": 6.666666666666667e-06,
-      "loss": 0.8772,
+      "epoch": 4.43,
+      "learning_rate": 9.881796690307329e-06,
+      "loss": 0.1514,
       "step": 208
     },
     {
-      "epoch": 15.77,
-      "learning_rate": 6.645299145299145e-06,
-      "loss": 0.8303,
+      "epoch": 4.45,
+      "learning_rate": 9.875886524822695e-06,
+      "loss": 0.2668,
       "step": 209
     },
     {
-      "epoch": 15.85,
-      "learning_rate": 6.623931623931624e-06,
-      "loss": 0.8535,
+      "epoch": 4.47,
+      "learning_rate": 9.869976359338063e-06,
+      "loss": 0.2831,
       "step": 210
     },
     {
-      "epoch": 15.92,
-      "learning_rate": 6.602564102564103e-06,
-      "loss": 0.8901,
+      "epoch": 4.49,
+      "learning_rate": 9.864066193853429e-06,
+      "loss": 0.1787,
       "step": 211
     },
     {
-      "epoch": 16.0,
-      "learning_rate": 6.581196581196582e-06,
-      "loss": 0.9412,
-      "step": 212
-    },
-    {
-      "epoch": 16.0,
-      "eval_accuracy": 0.7222222222222222,
-      "eval_loss": 0.8012484908103943,
-      "eval_runtime": 0.898,
-      "eval_samples_per_second": 60.134,
-      "eval_steps_per_second": 4.454,
+      "epoch": 4.51,
+      "learning_rate": 9.858156028368795e-06,
+      "loss": 0.2138,
       "step": 212
     },
     {
-      "epoch": 16.08,
-      "learning_rate": 6.5598290598290605e-06,
-      "loss": 0.9037,
+      "epoch": 4.53,
+      "learning_rate": 9.852245862884161e-06,
+      "loss": 0.2498,
       "step": 213
     },
     {
-      "epoch": 16.15,
-      "learning_rate": 6.538461538461539e-06,
-      "loss": 0.7875,
+      "epoch": 4.55,
+      "learning_rate": 9.846335697399529e-06,
+      "loss": 0.1772,
       "step": 214
     },
     {
-      "epoch": 16.23,
-      "learning_rate": 6.517094017094018e-06,
-      "loss": 0.7622,
+      "epoch": 4.57,
+      "learning_rate": 9.840425531914895e-06,
+      "loss": 0.3668,
       "step": 215
     },
     {
-      "epoch": 16.3,
-      "learning_rate": 6.495726495726496e-06,
-      "loss": 0.7955,
+      "epoch": 4.6,
+      "learning_rate": 9.834515366430261e-06,
+      "loss": 0.1898,
       "step": 216
     },
     {
-      "epoch": 16.38,
-      "learning_rate": 6.474358974358975e-06,
-      "loss": 0.9151,
+      "epoch": 4.62,
+      "learning_rate": 9.828605200945627e-06,
+      "loss": 0.1715,
       "step": 217
     },
     {
-      "epoch": 16.45,
-      "learning_rate": 6.4529914529914535e-06,
-      "loss": 0.8582,
+      "epoch": 4.64,
+      "learning_rate": 9.822695035460993e-06,
+      "loss": 0.1061,
       "step": 218
     },
     {
-      "epoch": 16.53,
-      "learning_rate": 6.431623931623933e-06,
-      "loss": 0.9383,
+      "epoch": 4.66,
+      "learning_rate": 9.81678486997636e-06,
+      "loss": 0.1456,
       "step": 219
     },
     {
-      "epoch": 16.6,
-      "learning_rate": 6.410256410256412e-06,
-      "loss": 0.8632,
+      "epoch": 4.68,
+      "learning_rate": 9.810874704491726e-06,
+      "loss": 0.3272,
       "step": 220
     },
     {
-      "epoch": 16.68,
-      "learning_rate": 6.3888888888888885e-06,
-      "loss": 0.9965,
+      "epoch": 4.7,
+      "learning_rate": 9.804964539007093e-06,
+      "loss": 0.2013,
       "step": 221
     },
     {
-      "epoch": 16.75,
-      "learning_rate": 6.367521367521368e-06,
-      "loss": 0.7761,
+      "epoch": 4.72,
+      "learning_rate": 9.79905437352246e-06,
+      "loss": 0.2834,
       "step": 222
     },
     {
-      "epoch": 16.83,
-      "learning_rate": 6.3461538461538466e-06,
-      "loss": 0.8543,
+      "epoch": 4.74,
+      "learning_rate": 9.793144208037826e-06,
+      "loss": 0.2331,
       "step": 223
     },
     {
-      "epoch": 16.91,
-      "learning_rate": 6.324786324786325e-06,
-      "loss": 0.9711,
+      "epoch": 4.77,
+      "learning_rate": 9.787234042553192e-06,
+      "loss": 0.1753,
       "step": 224
     },
     {
-      "epoch": 16.98,
-      "learning_rate": 6.303418803418804e-06,
-      "loss": 0.909,
-      "step": 225
-    },
-    {
-      "epoch": 16.98,
-      "eval_accuracy": 0.6481481481481481,
-      "eval_loss": 0.8124245405197144,
-      "eval_runtime": 0.899,
-      "eval_samples_per_second": 60.067,
-      "eval_steps_per_second": 4.449,
+      "epoch": 4.79,
+      "learning_rate": 9.78132387706856e-06,
+      "loss": 0.2296,
       "step": 225
     },
     {
-      "epoch": 17.06,
-      "learning_rate": 6.282051282051282e-06,
-      "loss": 1.0041,
+      "epoch": 4.81,
+      "learning_rate": 9.775413711583926e-06,
+      "loss": 0.2541,
       "step": 226
     },
     {
-      "epoch": 17.13,
-      "learning_rate": 6.260683760683761e-06,
-      "loss": 0.7288,
+      "epoch": 4.83,
+      "learning_rate": 9.769503546099292e-06,
+      "loss": 0.2262,
       "step": 227
     },
     {
-      "epoch": 17.21,
-      "learning_rate": 6.23931623931624e-06,
-      "loss": 0.8377,
+      "epoch": 4.85,
+      "learning_rate": 9.763593380614658e-06,
+      "loss": 0.2869,
       "step": 228
     },
     {
-      "epoch": 17.28,
-      "learning_rate": 6.217948717948718e-06,
-      "loss": 0.8346,
+      "epoch": 4.87,
+      "learning_rate": 9.757683215130024e-06,
+      "loss": 0.2128,
       "step": 229
     },
     {
-      "epoch": 17.36,
-      "learning_rate": 6.196581196581198e-06,
-      "loss": 0.8166,
+      "epoch": 4.89,
+      "learning_rate": 9.75177304964539e-06,
+      "loss": 0.3023,
       "step": 230
     },
     {
-      "epoch": 17.43,
-      "learning_rate": 6.175213675213676e-06,
-      "loss": 0.7913,
+      "epoch": 4.91,
+      "learning_rate": 9.745862884160756e-06,
+      "loss": 0.1332,
       "step": 231
     },
     {
-      "epoch": 17.51,
-      "learning_rate": 6.153846153846155e-06,
-      "loss": 0.8084,
+      "epoch": 4.94,
+      "learning_rate": 9.739952718676123e-06,
+      "loss": 0.4152,
       "step": 232
     },
     {
-      "epoch": 17.58,
-      "learning_rate": 6.1324786324786335e-06,
-      "loss": 0.8618,
+      "epoch": 4.96,
+      "learning_rate": 9.73404255319149e-06,
+      "loss": 0.2543,
       "step": 233
     },
     {
-      "epoch": 17.66,
-      "learning_rate": 6.111111111111112e-06,
-      "loss": 0.9029,
+      "epoch": 4.98,
+      "learning_rate": 9.728132387706857e-06,
+      "loss": 0.3129,
       "step": 234
     },
     {
-      "epoch": 17.74,
-      "learning_rate": 6.08974358974359e-06,
-      "loss": 0.8843,
+      "epoch": 5.0,
+      "learning_rate": 9.722222222222223e-06,
+      "loss": 0.2983,
       "step": 235
     },
     {
-      "epoch": 17.81,
-      "learning_rate": 6.0683760683760684e-06,
-      "loss": 0.9124,
-      "step": 236
+      "epoch": 5.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.14884129166603088,
+      "eval_runtime": 27.4292,
+      "eval_samples_per_second": 5.469,
+      "eval_steps_per_second": 0.365,
+      "step": 235
     },
     {
-      "epoch": 17.89,
-      "learning_rate": 6.047008547008547e-06,
-      "loss": 0.9024,
-      "step": 237
+      "epoch": 5.02,
+      "learning_rate": 9.716312056737589e-06,
+      "loss": 0.2603,
+      "step": 236
     },
     {
-      "epoch": 17.96,
-      "learning_rate": 6.025641025641026e-06,
-      "loss": 0.9637,
-      "step": 238
+      "epoch": 5.04,
+      "learning_rate": 9.710401891252957e-06,
+      "loss": 0.2247,
+      "step": 237
     },
     {
-      "epoch": 17.96,
-      "eval_accuracy": 0.6481481481481481,
-      "eval_loss": 0.8133196234703064,
-      "eval_runtime": 1.011,
-      "eval_samples_per_second": 53.413,
-      "eval_steps_per_second": 3.956,
+      "epoch": 5.06,
+      "learning_rate": 9.704491725768323e-06,
+      "loss": 0.3716,
       "step": 238
     },
     {
-      "epoch": 18.04,
-      "learning_rate": 6.004273504273504e-06,
-      "loss": 0.7778,
+      "epoch": 5.09,
+      "learning_rate": 9.698581560283689e-06,
+      "loss": 0.1204,
       "step": 239
     },
     {
-      "epoch": 18.11,
-      "learning_rate": 5.982905982905983e-06,
-      "loss": 0.8647,
+      "epoch": 5.11,
+      "learning_rate": 9.692671394799055e-06,
+      "loss": 0.1341,
       "step": 240
     },
     {
-      "epoch": 18.19,
-      "learning_rate": 5.961538461538462e-06,
-      "loss": 0.8894,
+      "epoch": 5.13,
+      "learning_rate": 9.686761229314421e-06,
+      "loss": 0.3148,
       "step": 241
     },
     {
-      "epoch": 18.26,
-      "learning_rate": 5.940170940170941e-06,
-      "loss": 0.8028,
+      "epoch": 5.15,
+      "learning_rate": 9.680851063829787e-06,
+      "loss": 0.1362,
       "step": 242
     },
     {
-      "epoch": 18.34,
-      "learning_rate": 5.9188034188034195e-06,
-      "loss": 0.8028,
+      "epoch": 5.17,
+      "learning_rate": 9.674940898345153e-06,
+      "loss": 0.3036,
       "step": 243
     },
     {
-      "epoch": 18.42,
-      "learning_rate": 5.897435897435898e-06,
-      "loss": 0.7512,
+      "epoch": 5.19,
+      "learning_rate": 9.66903073286052e-06,
+      "loss": 0.1807,
       "step": 244
     },
     {
-      "epoch": 18.49,
-      "learning_rate": 5.876068376068377e-06,
-      "loss": 0.9288,
+      "epoch": 5.21,
+      "learning_rate": 9.663120567375887e-06,
+      "loss": 0.2095,
       "step": 245
     },
     {
-      "epoch": 18.57,
-      "learning_rate": 5.854700854700855e-06,
-      "loss": 0.8492,
+      "epoch": 5.23,
+      "learning_rate": 9.657210401891254e-06,
+      "loss": 0.1633,
       "step": 246
     },
     {
-      "epoch": 18.64,
-      "learning_rate": 5.833333333333334e-06,
-      "loss": 0.7186,
+      "epoch": 5.26,
+      "learning_rate": 9.65130023640662e-06,
+      "loss": 0.2447,
       "step": 247
     },
     {
-      "epoch": 18.72,
-      "learning_rate": 5.8119658119658126e-06,
-      "loss": 0.8551,
+      "epoch": 5.28,
+      "learning_rate": 9.645390070921986e-06,
+      "loss": 0.1574,
       "step": 248
     },
     {
-      "epoch": 18.79,
-      "learning_rate": 5.790598290598292e-06,
-      "loss": 0.8594,
+      "epoch": 5.3,
+      "learning_rate": 9.639479905437354e-06,
+      "loss": 0.1104,
       "step": 249
     },
     {
-      "epoch": 18.87,
-      "learning_rate": 5.769230769230769e-06,
-      "loss": 0.75,
+      "epoch": 5.32,
+      "learning_rate": 9.63356973995272e-06,
+      "loss": 0.16,
       "step": 250
     },
     {
-      "epoch": 18.94,
-      "learning_rate": 5.7478632478632475e-06,
-      "loss": 0.8025,
-      "step": 251
-    },
-    {
-      "epoch": 18.94,
-      "eval_accuracy": 0.7037037037037037,
-      "eval_loss": 0.78453528881073,
-      "eval_runtime": 0.925,
-      "eval_samples_per_second": 58.378,
-      "eval_steps_per_second": 4.324,
+      "epoch": 5.34,
+      "learning_rate": 9.627659574468086e-06,
+      "loss": 0.3077,
       "step": 251
     },
     {
-      "epoch": 19.02,
-      "learning_rate": 5.726495726495727e-06,
-      "loss": 0.831,
+      "epoch": 5.36,
+      "learning_rate": 9.621749408983452e-06,
+      "loss": 0.2495,
       "step": 252
     },
     {
-      "epoch": 19.09,
-      "learning_rate": 5.705128205128206e-06,
-      "loss": 0.7797,
+      "epoch": 5.38,
+      "learning_rate": 9.615839243498818e-06,
+      "loss": 0.1471,
       "step": 253
     },
     {
-      "epoch": 19.17,
-      "learning_rate": 5.683760683760684e-06,
-      "loss": 0.6896,
+      "epoch": 5.4,
+      "learning_rate": 9.609929078014186e-06,
+      "loss": 0.1242,
       "step": 254
     },
     {
-      "epoch": 19.25,
-      "learning_rate": 5.662393162393163e-06,
-      "loss": 0.8953,
+      "epoch": 5.43,
+      "learning_rate": 9.604018912529552e-06,
+      "loss": 0.26,
       "step": 255
     },
     {
-      "epoch": 19.32,
-      "learning_rate": 5.641025641025641e-06,
-      "loss": 0.7566,
+      "epoch": 5.45,
+      "learning_rate": 9.598108747044918e-06,
+      "loss": 0.1284,
       "step": 256
     },
     {
-      "epoch": 19.4,
-      "learning_rate": 5.61965811965812e-06,
-      "loss": 0.7492,
+      "epoch": 5.47,
+      "learning_rate": 9.592198581560284e-06,
+      "loss": 0.2278,
       "step": 257
     },
     {
-      "epoch": 19.47,
-      "learning_rate": 5.598290598290599e-06,
-      "loss": 0.936,
+      "epoch": 5.49,
+      "learning_rate": 9.58628841607565e-06,
+      "loss": 0.2207,
       "step": 258
     },
     {
-      "epoch": 19.55,
-      "learning_rate": 5.576923076923077e-06,
-      "loss": 0.8861,
+      "epoch": 5.51,
+      "learning_rate": 9.580378250591017e-06,
+      "loss": 0.2141,
       "step": 259
     },
     {
-      "epoch": 19.62,
-      "learning_rate": 5.555555555555557e-06,
-      "loss": 0.8464,
+      "epoch": 5.53,
+      "learning_rate": 9.574468085106385e-06,
+      "loss": 0.1358,
       "step": 260
     },
     {
-      "epoch": 19.7,
-      "learning_rate": 5.534188034188035e-06,
-      "loss": 0.8636,
+      "epoch": 5.55,
+      "learning_rate": 9.56855791962175e-06,
+      "loss": 0.1367,
       "step": 261
     },
     {
-      "epoch": 19.77,
-      "learning_rate": 5.512820512820514e-06,
-      "loss": 0.7882,
+      "epoch": 5.57,
+      "learning_rate": 9.562647754137117e-06,
+      "loss": 0.195,
       "step": 262
     },
     {
-      "epoch": 19.85,
-      "learning_rate": 5.4914529914529925e-06,
-      "loss": 0.9639,
+      "epoch": 5.6,
+      "learning_rate": 9.556737588652483e-06,
+      "loss": 0.1402,
       "step": 263
     },
     {
-      "epoch": 19.92,
-      "learning_rate": 5.470085470085471e-06,
-      "loss": 0.7772,
+      "epoch": 5.62,
+      "learning_rate": 9.55082742316785e-06,
+      "loss": 0.1701,
       "step": 264
     },
     {
-      "epoch": 20.0,
-      "learning_rate": 5.448717948717949e-06,
-      "loss": 0.855,
-      "step": 265
-    },
-    {
-      "epoch": 20.0,
-      "eval_accuracy": 0.6481481481481481,
-      "eval_loss": 0.7630794644355774,
-      "eval_runtime": 0.893,
-      "eval_samples_per_second": 60.47,
-      "eval_steps_per_second": 4.479,
+      "epoch": 5.64,
+      "learning_rate": 9.544917257683217e-06,
+      "loss": 0.1882,
       "step": 265
     },
     {
-      "epoch": 20.08,
-      "learning_rate": 5.4273504273504275e-06,
-      "loss": 0.7891,
+      "epoch": 5.66,
+      "learning_rate": 9.539007092198583e-06,
+      "loss": 0.1325,
       "step": 266
     },
     {
-      "epoch": 20.15,
-      "learning_rate": 5.405982905982906e-06,
-      "loss": 0.8162,
+      "epoch": 5.68,
+      "learning_rate": 9.53309692671395e-06,
+      "loss": 0.2147,
       "step": 267
     },
     {
-      "epoch": 20.23,
-      "learning_rate": 5.384615384615385e-06,
-      "loss": 0.844,
+      "epoch": 5.7,
+      "learning_rate": 9.527186761229315e-06,
+      "loss": 0.2866,
       "step": 268
     },
     {
-      "epoch": 20.3,
-      "learning_rate": 5.363247863247863e-06,
-      "loss": 0.8381,
+      "epoch": 5.72,
+      "learning_rate": 9.521276595744681e-06,
+      "loss": 0.2633,
       "step": 269
     },
     {
-      "epoch": 20.38,
-      "learning_rate": 5.341880341880342e-06,
-      "loss": 0.6331,
+      "epoch": 5.74,
+      "learning_rate": 9.515366430260048e-06,
+      "loss": 0.1095,
       "step": 270
     },
     {
-      "epoch": 20.45,
-      "learning_rate": 5.320512820512821e-06,
-      "loss": 0.7966,
+      "epoch": 5.77,
+      "learning_rate": 9.509456264775414e-06,
+      "loss": 0.2193,
       "step": 271
     },
     {
-      "epoch": 20.53,
-      "learning_rate": 5.2991452991453e-06,
-      "loss": 0.6957,
+      "epoch": 5.79,
+      "learning_rate": 9.503546099290782e-06,
+      "loss": 0.2552,
       "step": 272
     },
     {
-      "epoch": 20.6,
-      "learning_rate": 5.2777777777777785e-06,
-      "loss": 0.8695,
+      "epoch": 5.81,
+      "learning_rate": 9.497635933806148e-06,
+      "loss": 0.0807,
       "step": 273
     },
     {
-      "epoch": 20.68,
-      "learning_rate": 5.256410256410257e-06,
-      "loss": 0.8797,
+      "epoch": 5.83,
+      "learning_rate": 9.491725768321514e-06,
+      "loss": 0.2423,
       "step": 274
     },
     {
-      "epoch": 20.75,
-      "learning_rate": 5.235042735042736e-06,
-      "loss": 0.8812,
+      "epoch": 5.85,
+      "learning_rate": 9.48581560283688e-06,
+      "loss": 0.0809,
       "step": 275
     },
     {
-      "epoch": 20.83,
-      "learning_rate": 5.213675213675214e-06,
-      "loss": 0.8978,
+      "epoch": 5.87,
+      "learning_rate": 9.479905437352248e-06,
+      "loss": 0.2793,
       "step": 276
     },
     {
-      "epoch": 20.91,
-      "learning_rate": 5.192307692307693e-06,
-      "loss": 0.7805,
+      "epoch": 5.89,
+      "learning_rate": 9.473995271867614e-06,
+      "loss": 0.1313,
       "step": 277
     },
     {
-      "epoch": 20.98,
-      "learning_rate": 5.1709401709401716e-06,
-      "loss": 0.8325,
-      "step": 278
-    },
-    {
-      "epoch": 20.98,
-      "eval_accuracy": 0.6666666666666666,
-      "eval_loss": 0.7550090551376343,
-      "eval_runtime": 0.906,
-      "eval_samples_per_second": 59.603,
-      "eval_steps_per_second": 4.415,
+      "epoch": 5.91,
+      "learning_rate": 9.46808510638298e-06,
+      "loss": 0.2163,
       "step": 278
     },
     {
-      "epoch": 21.06,
-      "learning_rate": 5.149572649572649e-06,
-      "loss": 0.8514,
+      "epoch": 5.94,
+      "learning_rate": 9.462174940898346e-06,
+      "loss": 0.2081,
       "step": 279
     },
     {
-      "epoch": 21.13,
-      "learning_rate": 5.128205128205128e-06,
-      "loss": 0.7688,
+      "epoch": 5.96,
+      "learning_rate": 9.456264775413712e-06,
+      "loss": 0.1641,
       "step": 280
     },
     {
-      "epoch": 21.21,
-      "learning_rate": 5.1068376068376065e-06,
-      "loss": 0.9067,
+      "epoch": 5.98,
+      "learning_rate": 9.450354609929078e-06,
+      "loss": 0.1622,
       "step": 281
     },
     {
-      "epoch": 21.28,
-      "learning_rate": 5.085470085470086e-06,
-      "loss": 0.8122,
+      "epoch": 6.0,
+      "learning_rate": 9.444444444444445e-06,
+      "loss": 0.4184,
       "step": 282
     },
     {
-      "epoch": 21.36,
-      "learning_rate": 5.064102564102565e-06,
-      "loss": 0.7124,
+      "epoch": 6.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.1980644017457962,
+      "eval_runtime": 27.368,
+      "eval_samples_per_second": 5.481,
+      "eval_steps_per_second": 0.365,
+      "step": 282
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 9.43853427895981e-06,
+      "loss": 0.1959,
       "step": 283
     },
     {
-      "epoch": 21.43,
-      "learning_rate": 5.042735042735043e-06,
-      "loss": 0.8195,
+      "epoch": 6.04,
+      "learning_rate": 9.432624113475179e-06,
+      "loss": 0.2509,
       "step": 284
     },
     {
-      "epoch": 21.51,
-      "learning_rate": 5.021367521367522e-06,
-      "loss": 0.8771,
+      "epoch": 6.06,
+      "learning_rate": 9.426713947990545e-06,
+      "loss": 0.1872,
       "step": 285
     },
     {
-      "epoch": 21.58,
-      "learning_rate": 5e-06,
-      "loss": 0.8645,
+      "epoch": 6.09,
+      "learning_rate": 9.42080378250591e-06,
+      "loss": 0.4202,
       "step": 286
     },
     {
-      "epoch": 21.66,
-      "learning_rate": 4.978632478632479e-06,
-      "loss": 0.8883,
+      "epoch": 6.11,
+      "learning_rate": 9.414893617021279e-06,
+      "loss": 0.2385,
       "step": 287
     },
     {
-      "epoch": 21.74,
-      "learning_rate": 4.957264957264958e-06,
-      "loss": 0.7989,
+      "epoch": 6.13,
+      "learning_rate": 9.408983451536645e-06,
+      "loss": 0.2406,
       "step": 288
     },
     {
-      "epoch": 21.81,
-      "learning_rate": 4.935897435897436e-06,
-      "loss": 0.9106,
+      "epoch": 6.15,
+      "learning_rate": 9.403073286052011e-06,
+      "loss": 0.3377,
       "step": 289
     },
     {
-      "epoch": 21.89,
-      "learning_rate": 4.914529914529915e-06,
-      "loss": 0.6479,
+      "epoch": 6.17,
+      "learning_rate": 9.397163120567377e-06,
+      "loss": 0.129,
       "step": 290
     },
     {
-      "epoch": 21.96,
-      "learning_rate": 4.8931623931623934e-06,
-      "loss": 0.7347,
-      "step": 291
-    },
-    {
-      "epoch": 21.96,
-      "eval_accuracy": 0.6296296296296297,
-      "eval_loss": 0.7868622541427612,
-      "eval_runtime": 1.002,
-      "eval_samples_per_second": 53.892,
-      "eval_steps_per_second": 3.992,
+      "epoch": 6.19,
+      "learning_rate": 9.391252955082743e-06,
+      "loss": 0.1488,
       "step": 291
     },
     {
-      "epoch": 22.04,
-      "learning_rate": 4.871794871794872e-06,
-      "loss": 0.8206,
+      "epoch": 6.21,
+      "learning_rate": 9.38534278959811e-06,
+      "loss": 0.3337,
       "step": 292
     },
     {
-      "epoch": 22.11,
-      "learning_rate": 4.850427350427351e-06,
-      "loss": 0.7932,
+      "epoch": 6.23,
+      "learning_rate": 9.379432624113475e-06,
+      "loss": 0.1531,
       "step": 293
     },
     {
-      "epoch": 22.19,
-      "learning_rate": 4.829059829059829e-06,
-      "loss": 0.8516,
+      "epoch": 6.26,
+      "learning_rate": 9.373522458628842e-06,
+      "loss": 0.2482,
       "step": 294
     },
     {
-      "epoch": 22.26,
-      "learning_rate": 4.807692307692308e-06,
-      "loss": 0.8488,
+      "epoch": 6.28,
+      "learning_rate": 9.36761229314421e-06,
+      "loss": 0.1532,
       "step": 295
     },
     {
-      "epoch": 22.34,
-      "learning_rate": 4.786324786324787e-06,
-      "loss": 0.7658,
+      "epoch": 6.3,
+      "learning_rate": 9.361702127659576e-06,
+      "loss": 0.177,
       "step": 296
     },
     {
-      "epoch": 22.42,
-      "learning_rate": 4.764957264957265e-06,
-      "loss": 0.8099,
+      "epoch": 6.32,
+      "learning_rate": 9.355791962174942e-06,
+      "loss": 0.23,
       "step": 297
     },
     {
-      "epoch": 22.49,
-      "learning_rate": 4.743589743589744e-06,
-      "loss": 0.7266,
+      "epoch": 6.34,
+      "learning_rate": 9.349881796690308e-06,
+      "loss": 0.1723,
       "step": 298
     },
     {
-      "epoch": 22.57,
-      "learning_rate": 4.722222222222222e-06,
-      "loss": 0.7245,
+      "epoch": 6.36,
+      "learning_rate": 9.343971631205676e-06,
+      "loss": 0.245,
       "step": 299
     },
     {
-      "epoch": 22.64,
-      "learning_rate": 4.700854700854701e-06,
-      "loss": 0.8976,
+      "epoch": 6.38,
+      "learning_rate": 9.338061465721042e-06,
+      "loss": 0.1958,
       "step": 300
     },
     {
-      "epoch": 22.72,
-      "learning_rate": 4.6794871794871795e-06,
-      "loss": 0.8127,
+      "epoch": 6.4,
+      "learning_rate": 9.332151300236408e-06,
+      "loss": 0.1538,
       "step": 301
     },
     {
-      "epoch": 22.79,
-      "learning_rate": 4.658119658119659e-06,
-      "loss": 0.7858,
+      "epoch": 6.43,
+      "learning_rate": 9.326241134751774e-06,
+      "loss": 0.2182,
       "step": 302
     },
     {
-      "epoch": 22.87,
-      "learning_rate": 4.6367521367521375e-06,
-      "loss": 0.8944,
+      "epoch": 6.45,
+      "learning_rate": 9.32033096926714e-06,
+      "loss": 0.3064,
       "step": 303
     },
     {
-      "epoch": 22.94,
-      "learning_rate": 4.615384615384616e-06,
-      "loss": 0.8302,
-      "step": 304
-    },
-    {
-      "epoch": 22.94,
-      "eval_accuracy": 0.6481481481481481,
-      "eval_loss": 0.8143013715744019,
-      "eval_runtime": 0.955,
-      "eval_samples_per_second": 56.545,
-      "eval_steps_per_second": 4.188,
+      "epoch": 6.47,
+      "learning_rate": 9.314420803782506e-06,
+      "loss": 0.2012,
       "step": 304
     },
     {
-      "epoch": 23.02,
-      "learning_rate": 4.594017094017094e-06,
-      "loss": 0.8172,
+      "epoch": 6.49,
+      "learning_rate": 9.308510638297872e-06,
+      "loss": 0.1923,
       "step": 305
     },
     {
-      "epoch": 23.09,
-      "learning_rate": 4.5726495726495725e-06,
-      "loss": 0.7397,
+      "epoch": 6.51,
+      "learning_rate": 9.302600472813239e-06,
+      "loss": 0.2688,
       "step": 306
     },
     {
-      "epoch": 23.17,
-      "learning_rate": 4.551282051282052e-06,
-      "loss": 0.7199,
+      "epoch": 6.53,
+      "learning_rate": 9.296690307328606e-06,
+      "loss": 0.2508,
       "step": 307
     },
     {
-      "epoch": 23.25,
-      "learning_rate": 4.5299145299145306e-06,
-      "loss": 0.7816,
+      "epoch": 6.55,
+      "learning_rate": 9.290780141843973e-06,
+      "loss": 0.195,
       "step": 308
     },
     {
-      "epoch": 23.32,
-      "learning_rate": 4.508547008547009e-06,
-      "loss": 0.9711,
+      "epoch": 6.57,
+      "learning_rate": 9.284869976359339e-06,
+      "loss": 0.1546,
       "step": 309
     },
     {
-      "epoch": 23.4,
-      "learning_rate": 4.487179487179488e-06,
-      "loss": 0.829,
+      "epoch": 6.6,
+      "learning_rate": 9.278959810874705e-06,
+      "loss": 0.2169,
       "step": 310
     },
     {
-      "epoch": 23.47,
-      "learning_rate": 4.465811965811966e-06,
-      "loss": 0.8121,
+      "epoch": 6.62,
+      "learning_rate": 9.273049645390073e-06,
+      "loss": 0.117,
       "step": 311
     },
     {
-      "epoch": 23.55,
-      "learning_rate": 4.444444444444444e-06,
-      "loss": 0.9511,
+      "epoch": 6.64,
+      "learning_rate": 9.267139479905439e-06,
+      "loss": 0.1128,
       "step": 312
     },
     {
-      "epoch": 23.62,
-      "learning_rate": 4.423076923076924e-06,
-      "loss": 0.8192,
+      "epoch": 6.66,
+      "learning_rate": 9.261229314420805e-06,
+      "loss": 0.2209,
       "step": 313
     },
     {
-      "epoch": 23.7,
-      "learning_rate": 4.401709401709402e-06,
-      "loss": 0.9055,
+      "epoch": 6.68,
+      "learning_rate": 9.255319148936171e-06,
+      "loss": 0.1076,
       "step": 314
     },
     {
-      "epoch": 23.77,
-      "learning_rate": 4.380341880341881e-06,
-      "loss": 0.7471,
+      "epoch": 6.7,
+      "learning_rate": 9.249408983451537e-06,
+      "loss": 0.1119,
       "step": 315
     },
     {
-      "epoch": 23.85,
-      "learning_rate": 4.358974358974359e-06,
-      "loss": 0.8605,
+      "epoch": 6.72,
+      "learning_rate": 9.243498817966903e-06,
+      "loss": 0.1008,
       "step": 316
     },
     {
-      "epoch": 23.92,
-      "learning_rate": 4.337606837606838e-06,
-      "loss": 0.9495,
+      "epoch": 6.74,
+      "learning_rate": 9.23758865248227e-06,
+      "loss": 0.1756,
       "step": 317
     },
     {
-      "epoch": 24.0,
-      "learning_rate": 4.316239316239317e-06,
-      "loss": 0.9766,
-      "step": 318
-    },
-    {
-      "epoch": 24.0,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7929366230964661,
-      "eval_runtime": 0.896,
-      "eval_samples_per_second": 60.268,
-      "eval_steps_per_second": 4.464,
+      "epoch": 6.77,
+      "learning_rate": 9.231678486997636e-06,
+      "loss": 0.194,
       "step": 318
     },
     {
-      "epoch": 24.08,
-      "learning_rate": 4.294871794871795e-06,
-      "loss": 0.7964,
+      "epoch": 6.79,
+      "learning_rate": 9.225768321513003e-06,
+      "loss": 0.2079,
       "step": 319
     },
     {
-      "epoch": 24.15,
-      "learning_rate": 4.273504273504274e-06,
-      "loss": 0.7545,
+      "epoch": 6.81,
+      "learning_rate": 9.21985815602837e-06,
+      "loss": 0.2568,
       "step": 320
     },
     {
-      "epoch": 24.23,
-      "learning_rate": 4.2521367521367524e-06,
-      "loss": 0.7971,
+      "epoch": 6.83,
+      "learning_rate": 9.213947990543736e-06,
+      "loss": 0.3773,
       "step": 321
     },
     {
-      "epoch": 24.3,
-      "learning_rate": 4.230769230769231e-06,
-      "loss": 0.7687,
+      "epoch": 6.85,
+      "learning_rate": 9.208037825059103e-06,
+      "loss": 0.1122,
       "step": 322
     },
     {
-      "epoch": 24.38,
-      "learning_rate": 4.20940170940171e-06,
-      "loss": 0.867,
+      "epoch": 6.87,
+      "learning_rate": 9.20212765957447e-06,
+      "loss": 0.14,
       "step": 323
     },
     {
-      "epoch": 24.45,
-      "learning_rate": 4.188034188034188e-06,
-      "loss": 0.7875,
+      "epoch": 6.89,
+      "learning_rate": 9.196217494089836e-06,
+      "loss": 0.2066,
       "step": 324
     },
     {
-      "epoch": 24.53,
-      "learning_rate": 4.166666666666667e-06,
-      "loss": 0.812,
+      "epoch": 6.91,
+      "learning_rate": 9.190307328605202e-06,
+      "loss": 0.1965,
       "step": 325
     },
     {
-      "epoch": 24.6,
-      "learning_rate": 4.145299145299146e-06,
-      "loss": 0.8408,
+      "epoch": 6.94,
+      "learning_rate": 9.184397163120568e-06,
+      "loss": 0.2624,
       "step": 326
     },
     {
-      "epoch": 24.68,
-      "learning_rate": 4.123931623931624e-06,
-      "loss": 0.8721,
+      "epoch": 6.96,
+      "learning_rate": 9.178486997635934e-06,
+      "loss": 0.435,
       "step": 327
     },
     {
-      "epoch": 24.75,
-      "learning_rate": 4.102564102564103e-06,
-      "loss": 0.7369,
+      "epoch": 6.98,
+      "learning_rate": 9.1725768321513e-06,
+      "loss": 0.277,
       "step": 328
     },
     {
-      "epoch": 24.83,
-      "learning_rate": 4.081196581196581e-06,
-      "loss": 0.6444,
+      "epoch": 7.0,
+      "learning_rate": 9.166666666666666e-06,
+      "loss": 0.1672,
       "step": 329
     },
     {
-      "epoch": 24.91,
-      "learning_rate": 4.05982905982906e-06,
-      "loss": 0.7127,
-      "step": 330
+      "epoch": 7.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.197462260723114,
+      "eval_runtime": 28.6202,
+      "eval_samples_per_second": 5.241,
+      "eval_steps_per_second": 0.349,
+      "step": 329
     },
     {
-      "epoch": 24.98,
-      "learning_rate": 4.0384615384615385e-06,
-      "loss": 0.7516,
-      "step": 331
+      "epoch": 7.02,
+      "learning_rate": 9.160756501182034e-06,
+      "loss": 0.052,
+      "step": 330
     },
     {
-      "epoch": 24.98,
-      "eval_accuracy": 0.6296296296296297,
-      "eval_loss": 0.8089188933372498,
-      "eval_runtime": 0.9192,
-      "eval_samples_per_second": 58.747,
-      "eval_steps_per_second": 4.352,
+      "epoch": 7.04,
+      "learning_rate": 9.1548463356974e-06,
+      "loss": 0.1349,
       "step": 331
     },
     {
-      "epoch": 25.06,
-      "learning_rate": 4.017094017094018e-06,
-      "loss": 0.9607,
+      "epoch": 7.06,
+      "learning_rate": 9.148936170212767e-06,
+      "loss": 0.1376,
       "step": 332
     },
     {
-      "epoch": 25.13,
-      "learning_rate": 3.9957264957264966e-06,
-      "loss": 0.7029,
+      "epoch": 7.09,
+      "learning_rate": 9.143026004728133e-06,
+      "loss": 0.1323,
       "step": 333
     },
     {
-      "epoch": 25.21,
-      "learning_rate": 3.974358974358974e-06,
-      "loss": 0.7801,
+      "epoch": 7.11,
+      "learning_rate": 9.1371158392435e-06,
+      "loss": 0.1828,
       "step": 334
     },
     {
-      "epoch": 25.28,
-      "learning_rate": 3.952991452991453e-06,
-      "loss": 0.8308,
+      "epoch": 7.13,
+      "learning_rate": 9.131205673758867e-06,
+      "loss": 0.2104,
       "step": 335
     },
     {
-      "epoch": 25.36,
-      "learning_rate": 3.9316239316239315e-06,
-      "loss": 0.853,
+      "epoch": 7.15,
+      "learning_rate": 9.125295508274233e-06,
+      "loss": 0.11,
       "step": 336
     },
     {
-      "epoch": 25.43,
-      "learning_rate": 3.910256410256411e-06,
-      "loss": 0.7821,
+      "epoch": 7.17,
+      "learning_rate": 9.119385342789599e-06,
+      "loss": 0.1149,
       "step": 337
     },
     {
-      "epoch": 25.51,
-      "learning_rate": 3.88888888888889e-06,
-      "loss": 0.871,
+      "epoch": 7.19,
+      "learning_rate": 9.113475177304965e-06,
+      "loss": 0.1864,
       "step": 338
     },
     {
-      "epoch": 25.58,
-      "learning_rate": 3.867521367521368e-06,
-      "loss": 0.7356,
+      "epoch": 7.21,
+      "learning_rate": 9.107565011820331e-06,
+      "loss": 0.1546,
       "step": 339
     },
     {
-      "epoch": 25.66,
-      "learning_rate": 3.846153846153847e-06,
-      "loss": 0.6685,
+      "epoch": 7.23,
+      "learning_rate": 9.101654846335697e-06,
+      "loss": 0.2051,
       "step": 340
     },
     {
-      "epoch": 25.74,
-      "learning_rate": 3.8247863247863246e-06,
-      "loss": 0.8098,
+      "epoch": 7.26,
+      "learning_rate": 9.095744680851063e-06,
+      "loss": 0.1523,
       "step": 341
     },
     {
-      "epoch": 25.81,
-      "learning_rate": 3.8034188034188036e-06,
-      "loss": 0.7344,
+      "epoch": 7.28,
+      "learning_rate": 9.089834515366431e-06,
+      "loss": 0.17,
       "step": 342
     },
     {
-      "epoch": 25.89,
-      "learning_rate": 3.782051282051282e-06,
-      "loss": 0.944,
+      "epoch": 7.3,
+      "learning_rate": 9.083924349881797e-06,
+      "loss": 0.1032,
       "step": 343
     },
     {
-      "epoch": 25.96,
-      "learning_rate": 3.760683760683761e-06,
-      "loss": 0.7784,
-      "step": 344
-    },
-    {
-      "epoch": 25.96,
-      "eval_accuracy": 0.6296296296296297,
-      "eval_loss": 0.7877990007400513,
-      "eval_runtime": 0.884,
-      "eval_samples_per_second": 61.086,
-      "eval_steps_per_second": 4.525,
+      "epoch": 7.32,
+      "learning_rate": 9.078014184397164e-06,
+      "loss": 0.1639,
       "step": 344
     },
     {
-      "epoch": 26.04,
-      "learning_rate": 3.73931623931624e-06,
-      "loss": 0.7081,
+      "epoch": 7.34,
+      "learning_rate": 9.07210401891253e-06,
+      "loss": 0.2212,
       "step": 345
     },
     {
-      "epoch": 26.11,
-      "learning_rate": 3.7179487179487184e-06,
-      "loss": 0.7859,
+      "epoch": 7.36,
+      "learning_rate": 9.066193853427897e-06,
+      "loss": 0.137,
       "step": 346
     },
     {
-      "epoch": 26.19,
-      "learning_rate": 3.696581196581197e-06,
-      "loss": 0.8629,
+      "epoch": 7.38,
+      "learning_rate": 9.060283687943264e-06,
+      "loss": 0.0877,
       "step": 347
     },
     {
-      "epoch": 26.26,
-      "learning_rate": 3.6752136752136756e-06,
-      "loss": 0.93,
+      "epoch": 7.4,
+      "learning_rate": 9.05437352245863e-06,
+      "loss": 0.1001,
       "step": 348
     },
     {
-      "epoch": 26.34,
-      "learning_rate": 3.653846153846154e-06,
-      "loss": 0.7968,
+      "epoch": 7.43,
+      "learning_rate": 9.048463356973996e-06,
+      "loss": 0.1186,
       "step": 349
     },
     {
-      "epoch": 26.42,
-      "learning_rate": 3.632478632478633e-06,
-      "loss": 0.7629,
+      "epoch": 7.45,
+      "learning_rate": 9.042553191489362e-06,
+      "loss": 0.2238,
       "step": 350
     },
     {
-      "epoch": 26.49,
-      "learning_rate": 3.6111111111111115e-06,
-      "loss": 0.7159,
+      "epoch": 7.47,
+      "learning_rate": 9.036643026004728e-06,
+      "loss": 0.1258,
       "step": 351
     },
     {
-      "epoch": 26.57,
-      "learning_rate": 3.58974358974359e-06,
-      "loss": 0.8303,
+      "epoch": 7.49,
+      "learning_rate": 9.030732860520094e-06,
+      "loss": 0.2327,
       "step": 352
     },
     {
-      "epoch": 26.64,
-      "learning_rate": 3.5683760683760687e-06,
-      "loss": 0.7054,
+      "epoch": 7.51,
+      "learning_rate": 9.024822695035462e-06,
+      "loss": 0.2076,
       "step": 353
     },
     {
-      "epoch": 26.72,
-      "learning_rate": 3.5470085470085473e-06,
-      "loss": 0.7428,
+      "epoch": 7.53,
+      "learning_rate": 9.018912529550828e-06,
+      "loss": 0.1542,
       "step": 354
     },
     {
-      "epoch": 26.79,
-      "learning_rate": 3.5256410256410263e-06,
-      "loss": 0.7547,
+      "epoch": 7.55,
+      "learning_rate": 9.013002364066194e-06,
+      "loss": 0.1837,
       "step": 355
     },
     {
-      "epoch": 26.87,
-      "learning_rate": 3.5042735042735045e-06,
-      "loss": 0.8752,
+      "epoch": 7.57,
+      "learning_rate": 9.00709219858156e-06,
+      "loss": 0.1446,
       "step": 356
     },
     {
-      "epoch": 26.94,
-      "learning_rate": 3.482905982905983e-06,
-      "loss": 0.6438,
-      "step": 357
-    },
-    {
-      "epoch": 26.94,
-      "eval_accuracy": 0.7037037037037037,
-      "eval_loss": 0.7568920254707336,
-      "eval_runtime": 0.971,
-      "eval_samples_per_second": 55.613,
-      "eval_steps_per_second": 4.119,
+      "epoch": 7.6,
+      "learning_rate": 9.001182033096928e-06,
+      "loss": 0.1783,
       "step": 357
     },
     {
-      "epoch": 27.02,
-      "learning_rate": 3.4615384615384617e-06,
-      "loss": 0.8912,
+      "epoch": 7.62,
+      "learning_rate": 8.995271867612294e-06,
+      "loss": 0.0919,
       "step": 358
     },
     {
-      "epoch": 27.09,
-      "learning_rate": 3.4401709401709403e-06,
-      "loss": 0.8638,
+      "epoch": 7.64,
+      "learning_rate": 8.98936170212766e-06,
+      "loss": 0.2306,
       "step": 359
     },
     {
-      "epoch": 27.17,
-      "learning_rate": 3.4188034188034193e-06,
-      "loss": 0.7835,
+      "epoch": 7.66,
+      "learning_rate": 8.983451536643027e-06,
+      "loss": 0.1372,
       "step": 360
     },
     {
-      "epoch": 27.25,
-      "learning_rate": 3.397435897435898e-06,
-      "loss": 0.8035,
+      "epoch": 7.68,
+      "learning_rate": 8.977541371158393e-06,
+      "loss": 0.2437,
       "step": 361
     },
     {
-      "epoch": 27.32,
-      "learning_rate": 3.3760683760683765e-06,
-      "loss": 0.7462,
+      "epoch": 7.7,
+      "learning_rate": 8.971631205673759e-06,
+      "loss": 0.2218,
       "step": 362
     },
     {
-      "epoch": 27.4,
-      "learning_rate": 3.3547008547008547e-06,
-      "loss": 0.8455,
+      "epoch": 7.72,
+      "learning_rate": 8.965721040189125e-06,
+      "loss": 0.1947,
       "step": 363
     },
     {
-      "epoch": 27.47,
-      "learning_rate": 3.3333333333333333e-06,
-      "loss": 0.8631,
+      "epoch": 7.74,
+      "learning_rate": 8.959810874704491e-06,
+      "loss": 0.3182,
       "step": 364
     },
     {
-      "epoch": 27.55,
-      "learning_rate": 3.311965811965812e-06,
-      "loss": 0.7807,
+      "epoch": 7.77,
+      "learning_rate": 8.953900709219859e-06,
+      "loss": 0.0986,
       "step": 365
     },
     {
-      "epoch": 27.62,
-      "learning_rate": 3.290598290598291e-06,
-      "loss": 0.6779,
+      "epoch": 7.79,
+      "learning_rate": 8.947990543735225e-06,
+      "loss": 0.3112,
       "step": 366
     },
     {
-      "epoch": 27.7,
-      "learning_rate": 3.2692307692307696e-06,
-      "loss": 0.7527,
+      "epoch": 7.81,
+      "learning_rate": 8.942080378250591e-06,
+      "loss": 0.2532,
       "step": 367
     },
     {
-      "epoch": 27.77,
-      "learning_rate": 3.247863247863248e-06,
-      "loss": 0.9228,
+      "epoch": 7.83,
+      "learning_rate": 8.936170212765958e-06,
+      "loss": 0.1206,
       "step": 368
     },
     {
-      "epoch": 27.85,
-      "learning_rate": 3.2264957264957268e-06,
-      "loss": 0.8497,
+      "epoch": 7.85,
+      "learning_rate": 8.930260047281325e-06,
+      "loss": 0.118,
       "step": 369
     },
     {
-      "epoch": 27.92,
-      "learning_rate": 3.205128205128206e-06,
-      "loss": 0.7918,
+      "epoch": 7.87,
+      "learning_rate": 8.924349881796691e-06,
+      "loss": 0.186,
       "step": 370
     },
     {
-      "epoch": 28.0,
-      "learning_rate": 3.183760683760684e-06,
-      "loss": 0.5649,
-      "step": 371
-    },
-    {
-      "epoch": 28.0,
-      "eval_accuracy": 0.7037037037037037,
-      "eval_loss": 0.7707066535949707,
-      "eval_runtime": 0.909,
-      "eval_samples_per_second": 59.406,
-      "eval_steps_per_second": 4.4,
+      "epoch": 7.89,
+      "learning_rate": 8.918439716312058e-06,
+      "loss": 0.2026,
       "step": 371
     },
     {
-      "epoch": 28.08,
-      "learning_rate": 3.1623931623931626e-06,
-      "loss": 0.7562,
+      "epoch": 7.91,
+      "learning_rate": 8.912529550827424e-06,
+      "loss": 0.1106,
       "step": 372
     },
     {
-      "epoch": 28.15,
-      "learning_rate": 3.141025641025641e-06,
-      "loss": 0.7619,
+      "epoch": 7.94,
+      "learning_rate": 8.90661938534279e-06,
+      "loss": 0.1839,
       "step": 373
     },
     {
-      "epoch": 28.23,
-      "learning_rate": 3.11965811965812e-06,
-      "loss": 0.812,
+      "epoch": 7.96,
+      "learning_rate": 8.900709219858156e-06,
+      "loss": 0.1336,
       "step": 374
     },
     {
-      "epoch": 28.3,
-      "learning_rate": 3.098290598290599e-06,
-      "loss": 0.7089,
+      "epoch": 7.98,
+      "learning_rate": 8.894799054373522e-06,
+      "loss": 0.1515,
       "step": 375
     },
     {
-      "epoch": 28.38,
-      "learning_rate": 3.0769230769230774e-06,
-      "loss": 0.7866,
+      "epoch": 8.0,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.0593,
       "step": 376
     },
     {
-      "epoch": 28.45,
-      "learning_rate": 3.055555555555556e-06,
-      "loss": 0.765,
+      "epoch": 8.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.19359709322452545,
+      "eval_runtime": 27.3282,
+      "eval_samples_per_second": 5.489,
+      "eval_steps_per_second": 0.366,
+      "step": 376
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 8.882978723404256e-06,
+      "loss": 0.236,
       "step": 377
     },
     {
-      "epoch": 28.53,
-      "learning_rate": 3.0341880341880342e-06,
-      "loss": 0.74,
+      "epoch": 8.04,
+      "learning_rate": 8.877068557919622e-06,
+      "loss": 0.1811,
       "step": 378
     },
     {
-      "epoch": 28.6,
-      "learning_rate": 3.012820512820513e-06,
-      "loss": 0.68,
+      "epoch": 8.06,
+      "learning_rate": 8.871158392434988e-06,
+      "loss": 0.2372,
       "step": 379
     },
     {
-      "epoch": 28.68,
-      "learning_rate": 2.9914529914529914e-06,
-      "loss": 0.6915,
+      "epoch": 8.09,
+      "learning_rate": 8.865248226950355e-06,
+      "loss": 0.1679,
       "step": 380
     },
     {
-      "epoch": 28.75,
-      "learning_rate": 2.9700854700854705e-06,
-      "loss": 0.7603,
+      "epoch": 8.11,
+      "learning_rate": 8.859338061465722e-06,
+      "loss": 0.1808,
       "step": 381
     },
     {
-      "epoch": 28.83,
-      "learning_rate": 2.948717948717949e-06,
-      "loss": 0.8074,
+      "epoch": 8.13,
+      "learning_rate": 8.853427895981088e-06,
+      "loss": 0.0963,
       "step": 382
     },
     {
-      "epoch": 28.91,
-      "learning_rate": 2.9273504273504277e-06,
-      "loss": 0.7192,
+      "epoch": 8.15,
+      "learning_rate": 8.847517730496455e-06,
+      "loss": 0.1729,
       "step": 383
     },
     {
-      "epoch": 28.98,
-      "learning_rate": 2.9059829059829063e-06,
-      "loss": 0.8274,
-      "step": 384
-    },
-    {
-      "epoch": 28.98,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7647761702537537,
-      "eval_runtime": 0.949,
-      "eval_samples_per_second": 56.902,
-      "eval_steps_per_second": 4.215,
+      "epoch": 8.17,
+      "learning_rate": 8.84160756501182e-06,
+      "loss": 0.2334,
       "step": 384
     },
     {
-      "epoch": 29.06,
-      "learning_rate": 2.8846153846153845e-06,
-      "loss": 0.7109,
+      "epoch": 8.19,
+      "learning_rate": 8.835697399527189e-06,
+      "loss": 0.1737,
       "step": 385
     },
     {
-      "epoch": 29.13,
-      "learning_rate": 2.8632478632478635e-06,
-      "loss": 0.7341,
+      "epoch": 8.21,
+      "learning_rate": 8.829787234042555e-06,
+      "loss": 0.1748,
       "step": 386
     },
     {
-      "epoch": 29.21,
-      "learning_rate": 2.841880341880342e-06,
-      "loss": 0.7888,
+      "epoch": 8.23,
+      "learning_rate": 8.823877068557921e-06,
+      "loss": 0.2874,
       "step": 387
     },
     {
-      "epoch": 29.28,
-      "learning_rate": 2.8205128205128207e-06,
-      "loss": 0.8613,
+      "epoch": 8.26,
+      "learning_rate": 8.817966903073287e-06,
+      "loss": 0.1395,
       "step": 388
     },
     {
-      "epoch": 29.36,
-      "learning_rate": 2.7991452991452993e-06,
-      "loss": 0.714,
+      "epoch": 8.28,
+      "learning_rate": 8.812056737588653e-06,
+      "loss": 0.1517,
       "step": 389
     },
     {
-      "epoch": 29.43,
-      "learning_rate": 2.7777777777777783e-06,
-      "loss": 0.8172,
+      "epoch": 8.3,
+      "learning_rate": 8.80614657210402e-06,
+      "loss": 0.1729,
       "step": 390
     },
     {
-      "epoch": 29.51,
-      "learning_rate": 2.756410256410257e-06,
-      "loss": 0.7677,
+      "epoch": 8.32,
+      "learning_rate": 8.800236406619385e-06,
+      "loss": 0.1086,
       "step": 391
     },
     {
-      "epoch": 29.58,
-      "learning_rate": 2.7350427350427355e-06,
-      "loss": 0.7619,
+      "epoch": 8.34,
+      "learning_rate": 8.794326241134753e-06,
+      "loss": 0.1598,
       "step": 392
     },
     {
-      "epoch": 29.66,
-      "learning_rate": 2.7136752136752137e-06,
-      "loss": 0.7238,
+      "epoch": 8.36,
+      "learning_rate": 8.78841607565012e-06,
+      "loss": 0.1477,
       "step": 393
     },
     {
-      "epoch": 29.74,
-      "learning_rate": 2.6923076923076923e-06,
-      "loss": 0.7523,
+      "epoch": 8.38,
+      "learning_rate": 8.782505910165485e-06,
+      "loss": 0.203,
       "step": 394
     },
     {
-      "epoch": 29.81,
-      "learning_rate": 2.670940170940171e-06,
-      "loss": 0.9351,
+      "epoch": 8.4,
+      "learning_rate": 8.776595744680852e-06,
+      "loss": 0.2094,
       "step": 395
     },
     {
-      "epoch": 29.89,
-      "learning_rate": 2.64957264957265e-06,
-      "loss": 0.6977,
+      "epoch": 8.43,
+      "learning_rate": 8.77068557919622e-06,
+      "loss": 0.1914,
       "step": 396
     },
     {
-      "epoch": 29.96,
-      "learning_rate": 2.6282051282051286e-06,
-      "loss": 0.7754,
-      "step": 397
-    },
-    {
-      "epoch": 29.96,
-      "eval_accuracy": 0.7037037037037037,
-      "eval_loss": 0.753173291683197,
-      "eval_runtime": 0.9079,
-      "eval_samples_per_second": 59.478,
-      "eval_steps_per_second": 4.406,
+      "epoch": 8.45,
+      "learning_rate": 8.764775413711586e-06,
+      "loss": 0.0927,
       "step": 397
     },
     {
-      "epoch": 30.04,
-      "learning_rate": 2.606837606837607e-06,
-      "loss": 0.7449,
+      "epoch": 8.47,
+      "learning_rate": 8.758865248226952e-06,
+      "loss": 0.3336,
       "step": 398
     },
     {
-      "epoch": 30.11,
-      "learning_rate": 2.5854700854700858e-06,
-      "loss": 0.8368,
+      "epoch": 8.49,
+      "learning_rate": 8.752955082742318e-06,
+      "loss": 0.1669,
       "step": 399
     },
     {
-      "epoch": 30.19,
-      "learning_rate": 2.564102564102564e-06,
-      "loss": 0.8493,
+      "epoch": 8.51,
+      "learning_rate": 8.747044917257684e-06,
+      "loss": 0.2994,
       "step": 400
     },
     {
-      "epoch": 30.26,
-      "learning_rate": 2.542735042735043e-06,
-      "loss": 0.7002,
+      "epoch": 8.53,
+      "learning_rate": 8.74113475177305e-06,
+      "loss": 0.3159,
       "step": 401
     },
     {
-      "epoch": 30.34,
-      "learning_rate": 2.5213675213675216e-06,
-      "loss": 0.6892,
+      "epoch": 8.55,
+      "learning_rate": 8.735224586288416e-06,
+      "loss": 0.1439,
       "step": 402
     },
     {
-      "epoch": 30.42,
-      "learning_rate": 2.5e-06,
-      "loss": 0.8284,
+      "epoch": 8.57,
+      "learning_rate": 8.729314420803782e-06,
+      "loss": 0.1997,
       "step": 403
     },
     {
-      "epoch": 30.49,
-      "learning_rate": 2.478632478632479e-06,
-      "loss": 0.6511,
+      "epoch": 8.6,
+      "learning_rate": 8.72340425531915e-06,
+      "loss": 0.0435,
       "step": 404
     },
     {
-      "epoch": 30.57,
-      "learning_rate": 2.4572649572649574e-06,
-      "loss": 0.8352,
+      "epoch": 8.62,
+      "learning_rate": 8.717494089834516e-06,
+      "loss": 0.2095,
       "step": 405
     },
     {
-      "epoch": 30.64,
-      "learning_rate": 2.435897435897436e-06,
-      "loss": 0.6737,
+      "epoch": 8.64,
+      "learning_rate": 8.711583924349882e-06,
+      "loss": 0.0567,
       "step": 406
     },
     {
-      "epoch": 30.72,
-      "learning_rate": 2.4145299145299146e-06,
-      "loss": 0.7473,
+      "epoch": 8.66,
+      "learning_rate": 8.705673758865249e-06,
+      "loss": 0.1495,
       "step": 407
     },
     {
-      "epoch": 30.79,
-      "learning_rate": 2.3931623931623937e-06,
-      "loss": 0.8458,
+      "epoch": 8.68,
+      "learning_rate": 8.699763593380616e-06,
+      "loss": 0.2779,
       "step": 408
     },
     {
-      "epoch": 30.87,
-      "learning_rate": 2.371794871794872e-06,
-      "loss": 0.5765,
+      "epoch": 8.7,
+      "learning_rate": 8.693853427895983e-06,
+      "loss": 0.1556,
       "step": 409
     },
     {
-      "epoch": 30.94,
-      "learning_rate": 2.3504273504273504e-06,
-      "loss": 0.7434,
-      "step": 410
-    },
-    {
-      "epoch": 30.94,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7607016563415527,
-      "eval_runtime": 0.908,
-      "eval_samples_per_second": 59.471,
-      "eval_steps_per_second": 4.405,
+      "epoch": 8.72,
+      "learning_rate": 8.687943262411349e-06,
+      "loss": 0.1986,
       "step": 410
     },
     {
-      "epoch": 31.02,
-      "learning_rate": 2.3290598290598295e-06,
-      "loss": 0.6838,
+      "epoch": 8.74,
+      "learning_rate": 8.682033096926715e-06,
+      "loss": 0.2002,
       "step": 411
     },
     {
-      "epoch": 31.09,
-      "learning_rate": 2.307692307692308e-06,
-      "loss": 0.7299,
+      "epoch": 8.77,
+      "learning_rate": 8.676122931442081e-06,
+      "loss": 0.0976,
       "step": 412
     },
     {
-      "epoch": 31.17,
-      "learning_rate": 2.2863247863247863e-06,
-      "loss": 0.5945,
+      "epoch": 8.79,
+      "learning_rate": 8.670212765957447e-06,
+      "loss": 0.162,
       "step": 413
     },
     {
-      "epoch": 31.25,
-      "learning_rate": 2.2649572649572653e-06,
-      "loss": 0.9878,
+      "epoch": 8.81,
+      "learning_rate": 8.664302600472813e-06,
+      "loss": 0.1384,
       "step": 414
     },
     {
-      "epoch": 31.32,
-      "learning_rate": 2.243589743589744e-06,
-      "loss": 0.5763,
+      "epoch": 8.83,
+      "learning_rate": 8.65839243498818e-06,
+      "loss": 0.124,
       "step": 415
     },
     {
-      "epoch": 31.4,
-      "learning_rate": 2.222222222222222e-06,
-      "loss": 0.6778,
+      "epoch": 8.85,
+      "learning_rate": 8.652482269503547e-06,
+      "loss": 0.2278,
       "step": 416
     },
     {
-      "epoch": 31.47,
-      "learning_rate": 2.200854700854701e-06,
-      "loss": 0.7268,
+      "epoch": 8.87,
+      "learning_rate": 8.646572104018913e-06,
+      "loss": 0.1431,
       "step": 417
     },
     {
-      "epoch": 31.55,
-      "learning_rate": 2.1794871794871797e-06,
-      "loss": 0.6828,
+      "epoch": 8.89,
+      "learning_rate": 8.64066193853428e-06,
+      "loss": 0.1597,
       "step": 418
     },
     {
-      "epoch": 31.62,
-      "learning_rate": 2.1581196581196583e-06,
-      "loss": 0.64,
+      "epoch": 8.91,
+      "learning_rate": 8.634751773049647e-06,
+      "loss": 0.0942,
       "step": 419
     },
     {
-      "epoch": 31.7,
-      "learning_rate": 2.136752136752137e-06,
-      "loss": 0.6549,
+      "epoch": 8.94,
+      "learning_rate": 8.628841607565013e-06,
+      "loss": 0.1807,
       "step": 420
     },
     {
-      "epoch": 31.77,
-      "learning_rate": 2.1153846153846155e-06,
-      "loss": 0.7474,
+      "epoch": 8.96,
+      "learning_rate": 8.62293144208038e-06,
+      "loss": 0.1897,
       "step": 421
     },
     {
-      "epoch": 31.85,
-      "learning_rate": 2.094017094017094e-06,
-      "loss": 0.5879,
+      "epoch": 8.98,
+      "learning_rate": 8.617021276595746e-06,
+      "loss": 0.1379,
       "step": 422
     },
     {
-      "epoch": 31.92,
-      "learning_rate": 2.072649572649573e-06,
-      "loss": 0.8655,
+      "epoch": 9.0,
+      "learning_rate": 8.611111111111112e-06,
+      "loss": 0.1465,
       "step": 423
     },
     {
-      "epoch": 32.0,
-      "learning_rate": 2.0512820512820513e-06,
-      "loss": 0.8634,
-      "step": 424
+      "epoch": 9.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.20118573307991028,
+      "eval_runtime": 28.5572,
+      "eval_samples_per_second": 5.253,
+      "eval_steps_per_second": 0.35,
+      "step": 423
     },
     {
-      "epoch": 32.0,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7515384554862976,
-      "eval_runtime": 0.97,
-      "eval_samples_per_second": 55.67,
-      "eval_steps_per_second": 4.124,
+      "epoch": 9.02,
+      "learning_rate": 8.605200945626478e-06,
+      "loss": 0.2322,
       "step": 424
     },
     {
-      "epoch": 32.08,
-      "learning_rate": 2.02991452991453e-06,
-      "loss": 0.8906,
+      "epoch": 9.04,
+      "learning_rate": 8.599290780141844e-06,
+      "loss": 0.2463,
       "step": 425
     },
     {
-      "epoch": 32.15,
-      "learning_rate": 2.008547008547009e-06,
-      "loss": 0.7675,
+      "epoch": 9.06,
+      "learning_rate": 8.59338061465721e-06,
+      "loss": 0.1512,
       "step": 426
     },
     {
-      "epoch": 32.23,
-      "learning_rate": 1.987179487179487e-06,
-      "loss": 0.7077,
+      "epoch": 9.09,
+      "learning_rate": 8.587470449172578e-06,
+      "loss": 0.1272,
       "step": 427
     },
     {
-      "epoch": 32.3,
-      "learning_rate": 1.9658119658119658e-06,
-      "loss": 0.8138,
+      "epoch": 9.11,
+      "learning_rate": 8.581560283687944e-06,
+      "loss": 0.2729,
       "step": 428
     },
     {
-      "epoch": 32.38,
-      "learning_rate": 1.944444444444445e-06,
-      "loss": 0.7514,
+      "epoch": 9.13,
+      "learning_rate": 8.57565011820331e-06,
+      "loss": 0.1044,
       "step": 429
     },
     {
-      "epoch": 32.45,
-      "learning_rate": 1.9230769230769234e-06,
-      "loss": 0.8134,
+      "epoch": 9.15,
+      "learning_rate": 8.569739952718676e-06,
+      "loss": 0.1095,
       "step": 430
     },
     {
-      "epoch": 32.53,
-      "learning_rate": 1.9017094017094018e-06,
-      "loss": 0.6852,
+      "epoch": 9.17,
+      "learning_rate": 8.563829787234044e-06,
+      "loss": 0.2102,
       "step": 431
     },
     {
-      "epoch": 32.6,
-      "learning_rate": 1.8803418803418804e-06,
-      "loss": 0.7071,
+      "epoch": 9.19,
+      "learning_rate": 8.55791962174941e-06,
+      "loss": 0.292,
       "step": 432
     },
     {
-      "epoch": 32.68,
-      "learning_rate": 1.8589743589743592e-06,
-      "loss": 0.8496,
+      "epoch": 9.21,
+      "learning_rate": 8.552009456264777e-06,
+      "loss": 0.2335,
       "step": 433
     },
     {
-      "epoch": 32.75,
-      "learning_rate": 1.8376068376068378e-06,
-      "loss": 0.7524,
+      "epoch": 9.23,
+      "learning_rate": 8.546099290780143e-06,
+      "loss": 0.1425,
       "step": 434
     },
     {
-      "epoch": 32.83,
-      "learning_rate": 1.8162393162393164e-06,
-      "loss": 0.6203,
+      "epoch": 9.26,
+      "learning_rate": 8.540189125295509e-06,
+      "loss": 0.095,
       "step": 435
     },
     {
-      "epoch": 32.91,
-      "learning_rate": 1.794871794871795e-06,
-      "loss": 0.8117,
+      "epoch": 9.28,
+      "learning_rate": 8.534278959810875e-06,
+      "loss": 0.1466,
       "step": 436
     },
     {
-      "epoch": 32.98,
-      "learning_rate": 1.7735042735042736e-06,
-      "loss": 0.7391,
-      "step": 437
-    },
-    {
-      "epoch": 32.98,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7634187340736389,
-      "eval_runtime": 0.969,
-      "eval_samples_per_second": 55.728,
-      "eval_steps_per_second": 4.128,
+      "epoch": 9.3,
+      "learning_rate": 8.528368794326241e-06,
+      "loss": 0.1495,
       "step": 437
     },
     {
-      "epoch": 33.06,
-      "learning_rate": 1.7521367521367522e-06,
-      "loss": 0.7485,
+      "epoch": 9.32,
+      "learning_rate": 8.522458628841607e-06,
+      "loss": 0.161,
       "step": 438
     },
     {
-      "epoch": 33.13,
-      "learning_rate": 1.7307692307692308e-06,
-      "loss": 0.8215,
+      "epoch": 9.34,
+      "learning_rate": 8.516548463356975e-06,
+      "loss": 0.1862,
       "step": 439
     },
     {
-      "epoch": 33.21,
-      "learning_rate": 1.7094017094017097e-06,
-      "loss": 0.7559,
+      "epoch": 9.36,
+      "learning_rate": 8.510638297872341e-06,
+      "loss": 0.0323,
       "step": 440
     },
     {
-      "epoch": 33.28,
-      "learning_rate": 1.6880341880341883e-06,
-      "loss": 0.7406,
+      "epoch": 9.38,
+      "learning_rate": 8.504728132387707e-06,
+      "loss": 0.3312,
       "step": 441
     },
     {
-      "epoch": 33.36,
-      "learning_rate": 1.6666666666666667e-06,
-      "loss": 0.7202,
+      "epoch": 9.4,
+      "learning_rate": 8.498817966903073e-06,
+      "loss": 0.2288,
       "step": 442
     },
     {
-      "epoch": 33.43,
-      "learning_rate": 1.6452991452991455e-06,
-      "loss": 0.7759,
+      "epoch": 9.43,
+      "learning_rate": 8.492907801418441e-06,
+      "loss": 0.0697,
       "step": 443
     },
     {
-      "epoch": 33.51,
-      "learning_rate": 1.623931623931624e-06,
-      "loss": 0.809,
+      "epoch": 9.45,
+      "learning_rate": 8.486997635933807e-06,
+      "loss": 0.1052,
       "step": 444
     },
     {
-      "epoch": 33.58,
-      "learning_rate": 1.602564102564103e-06,
-      "loss": 0.6217,
+      "epoch": 9.47,
+      "learning_rate": 8.481087470449174e-06,
+      "loss": 0.1446,
       "step": 445
     },
     {
-      "epoch": 33.66,
-      "learning_rate": 1.5811965811965813e-06,
-      "loss": 0.8605,
+      "epoch": 9.49,
+      "learning_rate": 8.47517730496454e-06,
+      "loss": 0.2471,
       "step": 446
     },
     {
-      "epoch": 33.74,
-      "learning_rate": 1.55982905982906e-06,
-      "loss": 0.7522,
+      "epoch": 9.51,
+      "learning_rate": 8.469267139479906e-06,
+      "loss": 0.2052,
       "step": 447
     },
     {
-      "epoch": 33.81,
-      "learning_rate": 1.5384615384615387e-06,
-      "loss": 0.781,
+      "epoch": 9.53,
+      "learning_rate": 8.463356973995272e-06,
+      "loss": 0.1399,
       "step": 448
     },
     {
-      "epoch": 33.89,
-      "learning_rate": 1.5170940170940171e-06,
-      "loss": 0.8538,
+      "epoch": 9.55,
+      "learning_rate": 8.457446808510638e-06,
+      "loss": 0.1688,
       "step": 449
     },
     {
-      "epoch": 33.96,
-      "learning_rate": 1.4957264957264957e-06,
-      "loss": 0.7261,
-      "step": 450
-    },
-    {
-      "epoch": 33.96,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.766499400138855,
-      "eval_runtime": 0.894,
-      "eval_samples_per_second": 60.403,
-      "eval_steps_per_second": 4.474,
+      "epoch": 9.57,
+      "learning_rate": 8.451536643026006e-06,
+      "loss": 0.2278,
       "step": 450
     },
     {
-      "epoch": 34.04,
-      "learning_rate": 1.4743589743589745e-06,
-      "loss": 0.6699,
+      "epoch": 9.6,
+      "learning_rate": 8.445626477541372e-06,
+      "loss": 0.3461,
       "step": 451
     },
     {
-      "epoch": 34.11,
-      "learning_rate": 1.4529914529914531e-06,
-      "loss": 0.5943,
+      "epoch": 9.62,
+      "learning_rate": 8.439716312056738e-06,
+      "loss": 0.2647,
       "step": 452
     },
     {
-      "epoch": 34.19,
-      "learning_rate": 1.4316239316239317e-06,
-      "loss": 0.7945,
+      "epoch": 9.64,
+      "learning_rate": 8.433806146572104e-06,
+      "loss": 0.1682,
       "step": 453
     },
     {
-      "epoch": 34.26,
-      "learning_rate": 1.4102564102564104e-06,
-      "loss": 0.6278,
+      "epoch": 9.66,
+      "learning_rate": 8.427895981087472e-06,
+      "loss": 0.2095,
       "step": 454
     },
     {
-      "epoch": 34.34,
-      "learning_rate": 1.3888888888888892e-06,
-      "loss": 0.6077,
+      "epoch": 9.68,
+      "learning_rate": 8.421985815602838e-06,
+      "loss": 0.3544,
       "step": 455
     },
     {
-      "epoch": 34.42,
-      "learning_rate": 1.3675213675213678e-06,
-      "loss": 0.7242,
+      "epoch": 9.7,
+      "learning_rate": 8.416075650118204e-06,
+      "loss": 0.2699,
       "step": 456
     },
     {
-      "epoch": 34.49,
-      "learning_rate": 1.3461538461538462e-06,
-      "loss": 0.634,
+      "epoch": 9.72,
+      "learning_rate": 8.41016548463357e-06,
+      "loss": 0.2197,
       "step": 457
     },
     {
-      "epoch": 34.57,
-      "learning_rate": 1.324786324786325e-06,
-      "loss": 0.8309,
+      "epoch": 9.74,
+      "learning_rate": 8.404255319148937e-06,
+      "loss": 0.171,
       "step": 458
     },
     {
-      "epoch": 34.64,
-      "learning_rate": 1.3034188034188036e-06,
-      "loss": 0.8022,
+      "epoch": 9.77,
+      "learning_rate": 8.398345153664303e-06,
+      "loss": 0.2125,
       "step": 459
     },
     {
-      "epoch": 34.72,
-      "learning_rate": 1.282051282051282e-06,
-      "loss": 0.7771,
+      "epoch": 9.79,
+      "learning_rate": 8.392434988179669e-06,
+      "loss": 0.2334,
       "step": 460
     },
     {
-      "epoch": 34.79,
-      "learning_rate": 1.2606837606837608e-06,
-      "loss": 0.7012,
+      "epoch": 9.81,
+      "learning_rate": 8.386524822695035e-06,
+      "loss": 0.1981,
       "step": 461
     },
     {
-      "epoch": 34.87,
-      "learning_rate": 1.2393162393162394e-06,
-      "loss": 0.8051,
+      "epoch": 9.83,
+      "learning_rate": 8.380614657210403e-06,
+      "loss": 0.2872,
       "step": 462
     },
     {
-      "epoch": 34.94,
-      "learning_rate": 1.217948717948718e-06,
-      "loss": 0.6929,
-      "step": 463
-    },
-    {
-      "epoch": 34.94,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7704557776451111,
-      "eval_runtime": 0.907,
-      "eval_samples_per_second": 59.537,
-      "eval_steps_per_second": 4.41,
+      "epoch": 9.85,
+      "learning_rate": 8.374704491725769e-06,
+      "loss": 0.0991,
       "step": 463
     },
     {
-      "epoch": 35.02,
-      "learning_rate": 1.1965811965811968e-06,
-      "loss": 0.623,
+      "epoch": 9.87,
+      "learning_rate": 8.368794326241135e-06,
+      "loss": 0.126,
       "step": 464
     },
     {
-      "epoch": 35.09,
-      "learning_rate": 1.1752136752136752e-06,
-      "loss": 0.6496,
+      "epoch": 9.89,
+      "learning_rate": 8.362884160756501e-06,
+      "loss": 0.0847,
       "step": 465
     },
     {
-      "epoch": 35.17,
-      "learning_rate": 1.153846153846154e-06,
-      "loss": 0.6899,
+      "epoch": 9.91,
+      "learning_rate": 8.35697399527187e-06,
+      "loss": 0.1583,
       "step": 466
     },
     {
-      "epoch": 35.25,
-      "learning_rate": 1.1324786324786326e-06,
-      "loss": 0.7069,
+      "epoch": 9.94,
+      "learning_rate": 8.351063829787235e-06,
+      "loss": 0.18,
       "step": 467
     },
     {
-      "epoch": 35.32,
-      "learning_rate": 1.111111111111111e-06,
-      "loss": 0.7134,
+      "epoch": 9.96,
+      "learning_rate": 8.345153664302601e-06,
+      "loss": 0.2639,
       "step": 468
     },
     {
-      "epoch": 35.4,
-      "learning_rate": 1.0897435897435899e-06,
-      "loss": 0.7735,
+      "epoch": 9.98,
+      "learning_rate": 8.339243498817968e-06,
+      "loss": 0.2598,
       "step": 469
     },
     {
-      "epoch": 35.47,
-      "learning_rate": 1.0683760683760685e-06,
-      "loss": 0.8744,
+      "epoch": 10.0,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.1286,
       "step": 470
     },
     {
-      "epoch": 35.55,
-      "learning_rate": 1.047008547008547e-06,
-      "loss": 0.6951,
+      "epoch": 10.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.18934974074363708,
+      "eval_runtime": 27.6433,
+      "eval_samples_per_second": 5.426,
+      "eval_steps_per_second": 0.362,
+      "step": 470
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 8.3274231678487e-06,
+      "loss": 0.1011,
       "step": 471
     },
     {
-      "epoch": 35.62,
-      "learning_rate": 1.0256410256410257e-06,
-      "loss": 0.6363,
+      "epoch": 10.04,
+      "learning_rate": 8.321513002364066e-06,
+      "loss": 0.1658,
       "step": 472
     },
     {
-      "epoch": 35.7,
-      "learning_rate": 1.0042735042735045e-06,
-      "loss": 0.6599,
+      "epoch": 10.06,
+      "learning_rate": 8.315602836879432e-06,
+      "loss": 0.1703,
       "step": 473
     },
     {
-      "epoch": 35.77,
-      "learning_rate": 9.829059829059829e-07,
-      "loss": 0.5542,
+      "epoch": 10.09,
+      "learning_rate": 8.3096926713948e-06,
+      "loss": 0.0641,
       "step": 474
     },
     {
-      "epoch": 35.85,
-      "learning_rate": 9.615384615384617e-07,
-      "loss": 0.833,
+      "epoch": 10.11,
+      "learning_rate": 8.303782505910166e-06,
+      "loss": 0.1319,
       "step": 475
     },
     {
-      "epoch": 35.92,
-      "learning_rate": 9.401709401709402e-07,
-      "loss": 0.6738,
+      "epoch": 10.13,
+      "learning_rate": 8.297872340425532e-06,
+      "loss": 0.148,
       "step": 476
     },
     {
-      "epoch": 36.0,
-      "learning_rate": 9.188034188034189e-07,
-      "loss": 0.655,
-      "step": 477
-    },
-    {
-      "epoch": 36.0,
-      "eval_accuracy": 0.6666666666666666,
-      "eval_loss": 0.7655048370361328,
-      "eval_runtime": 0.898,
-      "eval_samples_per_second": 60.133,
-      "eval_steps_per_second": 4.454,
+      "epoch": 10.15,
+      "learning_rate": 8.291962174940898e-06,
+      "loss": 0.1989,
       "step": 477
     },
     {
-      "epoch": 36.08,
-      "learning_rate": 8.974358974358975e-07,
-      "loss": 0.6654,
+      "epoch": 10.17,
+      "learning_rate": 8.286052009456266e-06,
+      "loss": 0.3026,
       "step": 478
     },
     {
-      "epoch": 36.15,
-      "learning_rate": 8.760683760683761e-07,
-      "loss": 0.6554,
+      "epoch": 10.19,
+      "learning_rate": 8.280141843971632e-06,
+      "loss": 0.1093,
       "step": 479
     },
     {
-      "epoch": 36.23,
-      "learning_rate": 8.547008547008548e-07,
-      "loss": 0.8517,
+      "epoch": 10.21,
+      "learning_rate": 8.274231678486998e-06,
+      "loss": 0.1618,
       "step": 480
     },
     {
-      "epoch": 36.3,
-      "learning_rate": 8.333333333333333e-07,
-      "loss": 0.6938,
+      "epoch": 10.23,
+      "learning_rate": 8.268321513002365e-06,
+      "loss": 0.2696,
       "step": 481
     },
     {
-      "epoch": 36.38,
-      "learning_rate": 8.11965811965812e-07,
-      "loss": 0.8445,
+      "epoch": 10.26,
+      "learning_rate": 8.26241134751773e-06,
+      "loss": 0.1018,
       "step": 482
     },
     {
-      "epoch": 36.45,
-      "learning_rate": 7.905982905982906e-07,
-      "loss": 0.5875,
+      "epoch": 10.28,
+      "learning_rate": 8.256501182033097e-06,
+      "loss": 0.114,
       "step": 483
     },
     {
-      "epoch": 36.53,
-      "learning_rate": 7.692307692307694e-07,
-      "loss": 0.6948,
+      "epoch": 10.3,
+      "learning_rate": 8.250591016548463e-06,
+      "loss": 0.1685,
       "step": 484
     },
     {
-      "epoch": 36.6,
-      "learning_rate": 7.478632478632479e-07,
-      "loss": 0.674,
+      "epoch": 10.32,
+      "learning_rate": 8.24468085106383e-06,
+      "loss": 0.1016,
       "step": 485
     },
     {
-      "epoch": 36.68,
-      "learning_rate": 7.264957264957266e-07,
-      "loss": 0.6888,
+      "epoch": 10.34,
+      "learning_rate": 8.238770685579197e-06,
+      "loss": 0.2505,
       "step": 486
     },
     {
-      "epoch": 36.75,
-      "learning_rate": 7.051282051282052e-07,
-      "loss": 0.784,
+      "epoch": 10.36,
+      "learning_rate": 8.232860520094563e-06,
+      "loss": 0.1744,
       "step": 487
     },
     {
-      "epoch": 36.83,
-      "learning_rate": 6.837606837606839e-07,
-      "loss": 0.8358,
+      "epoch": 10.38,
+      "learning_rate": 8.22695035460993e-06,
+      "loss": 0.0732,
       "step": 488
     },
     {
-      "epoch": 36.91,
-      "learning_rate": 6.623931623931625e-07,
-      "loss": 0.7664,
+      "epoch": 10.4,
+      "learning_rate": 8.221040189125297e-06,
+      "loss": 0.1577,
       "step": 489
     },
     {
-      "epoch": 36.98,
-      "learning_rate": 6.41025641025641e-07,
-      "loss": 0.707,
-      "step": 490
-    },
-    {
-      "epoch": 36.98,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7611520290374756,
-      "eval_runtime": 0.915,
-      "eval_samples_per_second": 59.016,
-      "eval_steps_per_second": 4.372,
+      "epoch": 10.43,
+      "learning_rate": 8.215130023640663e-06,
+      "loss": 0.1691,
       "step": 490
     },
     {
-      "epoch": 37.06,
-      "learning_rate": 6.196581196581197e-07,
-      "loss": 0.6732,
+      "epoch": 10.45,
+      "learning_rate": 8.20921985815603e-06,
+      "loss": 0.0536,
       "step": 491
     },
     {
-      "epoch": 37.13,
-      "learning_rate": 5.982905982905984e-07,
-      "loss": 0.7807,
+      "epoch": 10.47,
+      "learning_rate": 8.203309692671395e-06,
+      "loss": 0.2361,
       "step": 492
     },
     {
-      "epoch": 37.21,
-      "learning_rate": 5.76923076923077e-07,
-      "loss": 0.8527,
+      "epoch": 10.49,
+      "learning_rate": 8.197399527186762e-06,
+      "loss": 0.1096,
       "step": 493
     },
     {
-      "epoch": 37.28,
-      "learning_rate": 5.555555555555555e-07,
-      "loss": 0.8081,
+      "epoch": 10.51,
+      "learning_rate": 8.191489361702128e-06,
+      "loss": 0.2123,
       "step": 494
     },
     {
-      "epoch": 37.36,
-      "learning_rate": 5.341880341880342e-07,
-      "loss": 0.6997,
+      "epoch": 10.53,
+      "learning_rate": 8.185579196217494e-06,
+      "loss": 0.0969,
       "step": 495
     },
     {
-      "epoch": 37.43,
-      "learning_rate": 5.128205128205128e-07,
-      "loss": 0.6318,
+      "epoch": 10.55,
+      "learning_rate": 8.17966903073286e-06,
+      "loss": 0.2472,
       "step": 496
     },
     {
-      "epoch": 37.51,
-      "learning_rate": 4.914529914529914e-07,
-      "loss": 0.7318,
+      "epoch": 10.57,
+      "learning_rate": 8.173758865248228e-06,
+      "loss": 0.1411,
       "step": 497
     },
     {
-      "epoch": 37.58,
-      "learning_rate": 4.700854700854701e-07,
-      "loss": 0.7661,
+      "epoch": 10.6,
+      "learning_rate": 8.167848699763594e-06,
+      "loss": 0.2842,
       "step": 498
     },
     {
-      "epoch": 37.66,
-      "learning_rate": 4.4871794871794876e-07,
-      "loss": 0.8482,
+      "epoch": 10.62,
+      "learning_rate": 8.16193853427896e-06,
+      "loss": 0.1568,
       "step": 499
     },
     {
-      "epoch": 37.74,
-      "learning_rate": 4.273504273504274e-07,
-      "loss": 0.6658,
+      "epoch": 10.64,
+      "learning_rate": 8.156028368794326e-06,
+      "loss": 0.1802,
       "step": 500
     },
     {
-      "epoch": 37.81,
-      "learning_rate": 4.05982905982906e-07,
-      "loss": 0.5801,
+      "epoch": 10.66,
+      "learning_rate": 8.150118203309694e-06,
+      "loss": 0.2008,
       "step": 501
     },
     {
-      "epoch": 37.89,
-      "learning_rate": 3.846153846153847e-07,
-      "loss": 0.6643,
+      "epoch": 10.68,
+      "learning_rate": 8.14420803782506e-06,
+      "loss": 0.1095,
       "step": 502
     },
     {
-      "epoch": 37.96,
-      "learning_rate": 3.632478632478633e-07,
-      "loss": 0.7946,
-      "step": 503
-    },
-    {
-      "epoch": 37.96,
-      "eval_accuracy": 0.6481481481481481,
-      "eval_loss": 0.7685385942459106,
-      "eval_runtime": 0.9,
-      "eval_samples_per_second": 60.0,
-      "eval_steps_per_second": 4.444,
+      "epoch": 10.7,
+      "learning_rate": 8.138297872340426e-06,
+      "loss": 0.2468,
       "step": 503
     },
     {
-      "epoch": 38.04,
-      "learning_rate": 3.4188034188034194e-07,
-      "loss": 0.6904,
+      "epoch": 10.72,
+      "learning_rate": 8.132387706855792e-06,
+      "loss": 0.0797,
       "step": 504
     },
     {
-      "epoch": 38.11,
-      "learning_rate": 3.205128205128205e-07,
-      "loss": 0.6983,
+      "epoch": 10.74,
+      "learning_rate": 8.126477541371159e-06,
+      "loss": 0.1783,
       "step": 505
     },
     {
-      "epoch": 38.19,
-      "learning_rate": 2.991452991452992e-07,
-      "loss": 0.7762,
+      "epoch": 10.77,
+      "learning_rate": 8.120567375886525e-06,
+      "loss": 0.05,
       "step": 506
     },
     {
-      "epoch": 38.26,
-      "learning_rate": 2.7777777777777776e-07,
-      "loss": 0.613,
+      "epoch": 10.79,
+      "learning_rate": 8.114657210401891e-06,
+      "loss": 0.2518,
       "step": 507
     },
     {
-      "epoch": 38.34,
-      "learning_rate": 2.564102564102564e-07,
-      "loss": 0.7969,
+      "epoch": 10.81,
+      "learning_rate": 8.108747044917257e-06,
+      "loss": 0.2543,
       "step": 508
     },
     {
-      "epoch": 38.42,
-      "learning_rate": 2.3504273504273505e-07,
-      "loss": 0.7437,
+      "epoch": 10.83,
+      "learning_rate": 8.102836879432625e-06,
+      "loss": 0.1675,
       "step": 509
     },
     {
-      "epoch": 38.49,
-      "learning_rate": 2.136752136752137e-07,
-      "loss": 0.6025,
+      "epoch": 10.85,
+      "learning_rate": 8.096926713947991e-06,
+      "loss": 0.0669,
       "step": 510
     },
     {
-      "epoch": 38.57,
-      "learning_rate": 1.9230769230769234e-07,
-      "loss": 0.6432,
+      "epoch": 10.87,
+      "learning_rate": 8.091016548463357e-06,
+      "loss": 0.1285,
       "step": 511
     },
     {
-      "epoch": 38.64,
-      "learning_rate": 1.7094017094017097e-07,
-      "loss": 0.7152,
+      "epoch": 10.89,
+      "learning_rate": 8.085106382978723e-06,
+      "loss": 0.0718,
       "step": 512
     },
     {
-      "epoch": 38.72,
-      "learning_rate": 1.495726495726496e-07,
-      "loss": 0.9015,
+      "epoch": 10.91,
+      "learning_rate": 8.079196217494091e-06,
+      "loss": 0.149,
       "step": 513
     },
     {
-      "epoch": 38.79,
-      "learning_rate": 1.282051282051282e-07,
-      "loss": 0.6963,
+      "epoch": 10.94,
+      "learning_rate": 8.073286052009457e-06,
+      "loss": 0.0915,
       "step": 514
     },
     {
-      "epoch": 38.87,
-      "learning_rate": 1.0683760683760685e-07,
-      "loss": 0.6298,
+      "epoch": 10.96,
+      "learning_rate": 8.067375886524823e-06,
+      "loss": 0.2115,
       "step": 515
     },
     {
-      "epoch": 38.94,
-      "learning_rate": 8.547008547008549e-08,
-      "loss": 0.7646,
+      "epoch": 10.98,
+      "learning_rate": 8.061465721040191e-06,
+      "loss": 0.223,
       "step": 516
     },
     {
-      "epoch": 38.94,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7679648399353027,
-      "eval_runtime": 0.925,
-      "eval_samples_per_second": 58.378,
-      "eval_steps_per_second": 4.324,
-      "step": 516
+      "epoch": 11.0,
+      "learning_rate": 8.055555555555557e-06,
+      "loss": 0.2288,
+      "step": 517
     },
     {
-      "epoch": 39.02,
-      "learning_rate": 6.41025641025641e-08,
-      "loss": 0.5088,
+      "epoch": 11.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.21527163684368134,
+      "eval_runtime": 28.218,
+      "eval_samples_per_second": 5.316,
+      "eval_steps_per_second": 0.354,
       "step": 517
     },
     {
-      "epoch": 39.09,
-      "learning_rate": 4.273504273504274e-08,
-      "loss": 0.7081,
+      "epoch": 11.02,
+      "learning_rate": 8.049645390070923e-06,
+      "loss": 0.233,
       "step": 518
     },
     {
-      "epoch": 39.17,
-      "learning_rate": 2.136752136752137e-08,
-      "loss": 0.649,
+      "epoch": 11.04,
+      "learning_rate": 8.04373522458629e-06,
+      "loss": 0.1299,
       "step": 519
     },
     {
-      "epoch": 39.25,
-      "learning_rate": 0.0,
-      "loss": 0.7836,
+      "epoch": 11.06,
+      "learning_rate": 8.037825059101656e-06,
+      "loss": 0.2992,
       "step": 520
     },
     {
-      "epoch": 39.25,
-      "eval_accuracy": 0.6851851851851852,
-      "eval_loss": 0.7684235572814941,
-      "eval_runtime": 0.91,
-      "eval_samples_per_second": 59.341,
-      "eval_steps_per_second": 4.396,
-      "step": 520
+      "epoch": 11.09,
+      "learning_rate": 8.031914893617022e-06,
+      "loss": 0.146,
+      "step": 521
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 8.026004728132388e-06,
+      "loss": 0.1264,
+      "step": 522
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 8.020094562647754e-06,
+      "loss": 0.2332,
+      "step": 523
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 8.014184397163122e-06,
+      "loss": 0.1551,
+      "step": 524
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 8.008274231678488e-06,
+      "loss": 0.1627,
+      "step": 525
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 8.002364066193854e-06,
+      "loss": 0.1528,
+      "step": 526
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 7.99645390070922e-06,
+      "loss": 0.1529,
+      "step": 527
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 7.990543735224588e-06,
+      "loss": 0.2731,
+      "step": 528
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 7.984633569739954e-06,
+      "loss": 0.1777,
+      "step": 529
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 7.97872340425532e-06,
+      "loss": 0.1944,
+      "step": 530
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 7.972813238770687e-06,
+      "loss": 0.2423,
+      "step": 531
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 7.966903073286053e-06,
+      "loss": 0.1179,
+      "step": 532
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 7.960992907801419e-06,
+      "loss": 0.1424,
+      "step": 533
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 7.955082742316785e-06,
+      "loss": 0.1547,
+      "step": 534
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 7.949172576832151e-06,
+      "loss": 0.0886,
+      "step": 535
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 7.943262411347519e-06,
+      "loss": 0.0665,
+      "step": 536
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 7.937352245862885e-06,
+      "loss": 0.1836,
+      "step": 537
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 7.931442080378251e-06,
+      "loss": 0.0566,
+      "step": 538
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 7.925531914893617e-06,
+      "loss": 0.1242,
+      "step": 539
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 7.919621749408985e-06,
+      "loss": 0.1796,
+      "step": 540
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 7.913711583924351e-06,
+      "loss": 0.2005,
+      "step": 541
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 7.907801418439717e-06,
+      "loss": 0.1064,
+      "step": 542
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 7.901891252955084e-06,
+      "loss": 0.0471,
+      "step": 543
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 7.89598108747045e-06,
+      "loss": 0.1079,
+      "step": 544
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 7.890070921985816e-06,
+      "loss": 0.2049,
+      "step": 545
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 7.884160756501182e-06,
+      "loss": 0.1901,
+      "step": 546
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 7.87825059101655e-06,
+      "loss": 0.0452,
+      "step": 547
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 7.872340425531916e-06,
+      "loss": 0.1931,
+      "step": 548
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 7.866430260047282e-06,
+      "loss": 0.1954,
+      "step": 549
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 7.860520094562648e-06,
+      "loss": 0.064,
+      "step": 550
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 7.854609929078016e-06,
+      "loss": 0.1997,
+      "step": 551
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 7.848699763593382e-06,
+      "loss": 0.1362,
+      "step": 552
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 7.842789598108748e-06,
+      "loss": 0.0822,
+      "step": 553
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 7.836879432624114e-06,
+      "loss": 0.1183,
+      "step": 554
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 7.83096926713948e-06,
+      "loss": 0.147,
+      "step": 555
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 7.825059101654847e-06,
+      "loss": 0.1189,
+      "step": 556
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 7.819148936170213e-06,
+      "loss": 0.1825,
+      "step": 557
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 7.813238770685579e-06,
+      "loss": 0.2104,
+      "step": 558
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 7.807328605200947e-06,
+      "loss": 0.0653,
+      "step": 559
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 7.801418439716313e-06,
+      "loss": 0.2191,
+      "step": 560
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 7.795508274231679e-06,
+      "loss": 0.1731,
+      "step": 561
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 7.789598108747045e-06,
+      "loss": 0.1446,
+      "step": 562
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 7.783687943262413e-06,
+      "loss": 0.093,
+      "step": 563
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 7.77777777777778e-06,
+      "loss": 0.0711,
+      "step": 564
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.2097337543964386,
+      "eval_runtime": 28.4719,
+      "eval_samples_per_second": 5.268,
+      "eval_steps_per_second": 0.351,
+      "step": 564
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 7.771867612293145e-06,
+      "loss": 0.109,
+      "step": 565
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 7.765957446808511e-06,
+      "loss": 0.1399,
+      "step": 566
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 7.760047281323878e-06,
+      "loss": 0.1897,
+      "step": 567
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 7.754137115839244e-06,
+      "loss": 0.0678,
+      "step": 568
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 7.74822695035461e-06,
+      "loss": 0.1767,
+      "step": 569
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 7.742316784869976e-06,
+      "loss": 0.063,
+      "step": 570
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 7.736406619385344e-06,
+      "loss": 0.2719,
+      "step": 571
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 7.73049645390071e-06,
+      "loss": 0.0452,
+      "step": 572
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 7.724586288416076e-06,
+      "loss": 0.1834,
+      "step": 573
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 7.718676122931442e-06,
+      "loss": 0.2154,
+      "step": 574
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 7.71276595744681e-06,
+      "loss": 0.1692,
+      "step": 575
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 7.706855791962176e-06,
+      "loss": 0.1082,
+      "step": 576
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 7.700945626477542e-06,
+      "loss": 0.146,
+      "step": 577
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 7.695035460992908e-06,
+      "loss": 0.2234,
+      "step": 578
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 7.689125295508275e-06,
+      "loss": 0.0576,
+      "step": 579
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 7.68321513002364e-06,
+      "loss": 0.2184,
+      "step": 580
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 7.677304964539007e-06,
+      "loss": 0.1373,
+      "step": 581
     },
     {
-      "epoch": 39.25,
-      "step": 520,
-      "total_flos": 2.658396678267912e+19,
-      "train_loss": 0.8652688105519001,
-      "train_runtime": 1463.2419,
-      "train_samples_per_second": 22.826,
-      "train_steps_per_second": 0.355
+      "epoch": 12.38,
+      "learning_rate": 7.671394799054375e-06,
+      "loss": 0.035,
+      "step": 582
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 7.66548463356974e-06,
+      "loss": 0.1055,
+      "step": 583
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 7.659574468085107e-06,
+      "loss": 0.1071,
+      "step": 584
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 7.653664302600473e-06,
+      "loss": 0.2821,
+      "step": 585
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 7.647754137115841e-06,
+      "loss": 0.1227,
+      "step": 586
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 7.641843971631207e-06,
+      "loss": 0.1901,
+      "step": 587
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 7.635933806146573e-06,
+      "loss": 0.2457,
+      "step": 588
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 7.63002364066194e-06,
+      "loss": 0.1206,
+      "step": 589
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 7.624113475177306e-06,
+      "loss": 0.1682,
+      "step": 590
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 7.618203309692672e-06,
+      "loss": 0.0998,
+      "step": 591
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 7.6122931442080386e-06,
+      "loss": 0.1231,
+      "step": 592
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 7.606382978723405e-06,
+      "loss": 0.2388,
+      "step": 593
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 7.600472813238772e-06,
+      "loss": 0.1429,
+      "step": 594
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 7.594562647754138e-06,
+      "loss": 0.1066,
+      "step": 595
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 7.588652482269504e-06,
+      "loss": 0.1267,
+      "step": 596
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 7.58274231678487e-06,
+      "loss": 0.1373,
+      "step": 597
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 7.576832151300237e-06,
+      "loss": 0.0704,
+      "step": 598
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 7.570921985815603e-06,
+      "loss": 0.1648,
+      "step": 599
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 7.565011820330969e-06,
+      "loss": 0.2569,
+      "step": 600
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 7.5591016548463355e-06,
+      "loss": 0.1815,
+      "step": 601
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 7.553191489361703e-06,
+      "loss": 0.3306,
+      "step": 602
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 7.547281323877069e-06,
+      "loss": 0.2071,
+      "step": 603
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 7.5413711583924356e-06,
+      "loss": 0.1832,
+      "step": 604
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 7.535460992907802e-06,
+      "loss": 0.1183,
+      "step": 605
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 7.529550827423169e-06,
+      "loss": 0.1158,
+      "step": 606
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 7.523640661938535e-06,
+      "loss": 0.1866,
+      "step": 607
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 7.517730496453901e-06,
+      "loss": 0.2165,
+      "step": 608
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 7.511820330969267e-06,
+      "loss": 0.1728,
+      "step": 609
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 7.505910165484634e-06,
+      "loss": 0.1713,
+      "step": 610
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.0688,
+      "step": 611
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.25344350934028625,
+      "eval_runtime": 27.5716,
+      "eval_samples_per_second": 5.44,
+      "eval_steps_per_second": 0.363,
+      "step": 611
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 7.494089834515367e-06,
+      "loss": 0.1552,
+      "step": 612
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 7.488179669030734e-06,
+      "loss": 0.2189,
+      "step": 613
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 7.4822695035461e-06,
+      "loss": 0.0961,
+      "step": 614
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 7.476359338061466e-06,
+      "loss": 0.1522,
+      "step": 615
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 7.4704491725768326e-06,
+      "loss": 0.0806,
+      "step": 616
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 7.4645390070921995e-06,
+      "loss": 0.1632,
+      "step": 617
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 7.458628841607566e-06,
+      "loss": 0.1551,
+      "step": 618
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 7.452718676122932e-06,
+      "loss": 0.1856,
+      "step": 619
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 7.446808510638298e-06,
+      "loss": 0.2835,
+      "step": 620
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 7.440898345153666e-06,
+      "loss": 0.1258,
+      "step": 621
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 7.434988179669032e-06,
+      "loss": 0.2093,
+      "step": 622
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 7.429078014184398e-06,
+      "loss": 0.2895,
+      "step": 623
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 7.423167848699764e-06,
+      "loss": 0.1813,
+      "step": 624
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 7.417257683215131e-06,
+      "loss": 0.2003,
+      "step": 625
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 7.411347517730497e-06,
+      "loss": 0.1096,
+      "step": 626
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 7.405437352245863e-06,
+      "loss": 0.1164,
+      "step": 627
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 7.3995271867612296e-06,
+      "loss": 0.1951,
+      "step": 628
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 7.3936170212765965e-06,
+      "loss": 0.1516,
+      "step": 629
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 7.387706855791963e-06,
+      "loss": 0.1575,
+      "step": 630
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 7.381796690307329e-06,
+      "loss": 0.2507,
+      "step": 631
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 7.375886524822695e-06,
+      "loss": 0.1353,
+      "step": 632
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 7.369976359338063e-06,
+      "loss": 0.176,
+      "step": 633
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 7.364066193853429e-06,
+      "loss": 0.0907,
+      "step": 634
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 7.358156028368795e-06,
+      "loss": 0.1603,
+      "step": 635
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 7.352245862884161e-06,
+      "loss": 0.0918,
+      "step": 636
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 7.346335697399528e-06,
+      "loss": 0.2853,
+      "step": 637
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 7.340425531914894e-06,
+      "loss": 0.216,
+      "step": 638
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 7.3345153664302604e-06,
+      "loss": 0.1424,
+      "step": 639
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 7.3286052009456266e-06,
+      "loss": 0.0872,
+      "step": 640
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 7.3226950354609935e-06,
+      "loss": 0.2051,
+      "step": 641
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 7.31678486997636e-06,
+      "loss": 0.0977,
+      "step": 642
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 7.310874704491726e-06,
+      "loss": 0.194,
+      "step": 643
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 7.304964539007094e-06,
+      "loss": 0.2274,
+      "step": 644
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 7.29905437352246e-06,
+      "loss": 0.1563,
+      "step": 645
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 7.293144208037826e-06,
+      "loss": 0.2597,
+      "step": 646
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 7.287234042553192e-06,
+      "loss": 0.0707,
+      "step": 647
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 7.281323877068559e-06,
+      "loss": 0.0694,
+      "step": 648
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 7.275413711583925e-06,
+      "loss": 0.201,
+      "step": 649
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 7.269503546099291e-06,
+      "loss": 0.1036,
+      "step": 650
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 7.2635933806146574e-06,
+      "loss": 0.0878,
+      "step": 651
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 7.257683215130024e-06,
+      "loss": 0.0523,
+      "step": 652
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 7.2517730496453905e-06,
+      "loss": 0.1671,
+      "step": 653
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 7.245862884160757e-06,
+      "loss": 0.1418,
+      "step": 654
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 7.239952718676123e-06,
+      "loss": 0.0759,
+      "step": 655
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 7.234042553191491e-06,
+      "loss": 0.1376,
+      "step": 656
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 7.228132387706857e-06,
+      "loss": 0.2008,
+      "step": 657
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 7.222222222222223e-06,
+      "loss": 0.1424,
+      "step": 658
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.22436822950839996,
+      "eval_runtime": 28.2598,
+      "eval_samples_per_second": 5.308,
+      "eval_steps_per_second": 0.354,
+      "step": 658
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 7.216312056737589e-06,
+      "loss": 0.1279,
+      "step": 659
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 7.210401891252956e-06,
+      "loss": 0.1566,
+      "step": 660
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 7.204491725768322e-06,
+      "loss": 0.1552,
+      "step": 661
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 7.198581560283688e-06,
+      "loss": 0.1049,
+      "step": 662
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 7.1926713947990544e-06,
+      "loss": 0.1719,
+      "step": 663
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 7.186761229314421e-06,
+      "loss": 0.103,
+      "step": 664
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 7.1808510638297875e-06,
+      "loss": 0.212,
+      "step": 665
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 7.174940898345154e-06,
+      "loss": 0.2466,
+      "step": 666
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 7.16903073286052e-06,
+      "loss": 0.2435,
+      "step": 667
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 7.163120567375888e-06,
+      "loss": 0.1409,
+      "step": 668
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 7.157210401891254e-06,
+      "loss": 0.2337,
+      "step": 669
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 7.15130023640662e-06,
+      "loss": 0.2439,
+      "step": 670
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 7.145390070921986e-06,
+      "loss": 0.3125,
+      "step": 671
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 7.139479905437353e-06,
+      "loss": 0.3175,
+      "step": 672
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 7.133569739952719e-06,
+      "loss": 0.1032,
+      "step": 673
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 7.127659574468085e-06,
+      "loss": 0.2648,
+      "step": 674
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 7.1217494089834514e-06,
+      "loss": 0.0668,
+      "step": 675
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 7.115839243498818e-06,
+      "loss": 0.0643,
+      "step": 676
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 7.109929078014185e-06,
+      "loss": 0.1022,
+      "step": 677
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 7.1040189125295515e-06,
+      "loss": 0.1391,
+      "step": 678
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 7.0981087470449185e-06,
+      "loss": 0.0788,
+      "step": 679
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 7.092198581560285e-06,
+      "loss": 0.0903,
+      "step": 680
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 7.086288416075651e-06,
+      "loss": 0.1402,
+      "step": 681
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 7.080378250591017e-06,
+      "loss": 0.1051,
+      "step": 682
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 7.074468085106384e-06,
+      "loss": 0.1206,
+      "step": 683
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 7.06855791962175e-06,
+      "loss": 0.2082,
+      "step": 684
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 7.062647754137116e-06,
+      "loss": 0.113,
+      "step": 685
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 7.056737588652482e-06,
+      "loss": 0.0997,
+      "step": 686
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 7.05082742316785e-06,
+      "loss": 0.0556,
+      "step": 687
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 7.044917257683216e-06,
+      "loss": 0.3704,
+      "step": 688
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 7.039007092198582e-06,
+      "loss": 0.1763,
+      "step": 689
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 7.0330969267139485e-06,
+      "loss": 0.0409,
+      "step": 690
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 7.0271867612293155e-06,
+      "loss": 0.1376,
+      "step": 691
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 7.021276595744682e-06,
+      "loss": 0.164,
+      "step": 692
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 7.015366430260048e-06,
+      "loss": 0.0949,
+      "step": 693
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 7.009456264775414e-06,
+      "loss": 0.0457,
+      "step": 694
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 7.003546099290781e-06,
+      "loss": 0.1031,
+      "step": 695
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 6.997635933806147e-06,
+      "loss": 0.0282,
+      "step": 696
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 6.991725768321513e-06,
+      "loss": 0.1204,
+      "step": 697
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 6.985815602836879e-06,
+      "loss": 0.2717,
+      "step": 698
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 6.979905437352247e-06,
+      "loss": 0.1839,
+      "step": 699
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 6.973995271867613e-06,
+      "loss": 0.1715,
+      "step": 700
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 6.968085106382979e-06,
+      "loss": 0.128,
+      "step": 701
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 6.9621749408983455e-06,
+      "loss": 0.1608,
+      "step": 702
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 6.9562647754137125e-06,
+      "loss": 0.1562,
+      "step": 703
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 6.950354609929079e-06,
+      "loss": 0.1449,
+      "step": 704
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 6.944444444444445e-06,
+      "loss": 0.0376,
+      "step": 705
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.1792953461408615,
+      "eval_runtime": 27.7526,
+      "eval_samples_per_second": 5.405,
+      "eval_steps_per_second": 0.36,
+      "step": 705
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 6.938534278959811e-06,
+      "loss": 0.2765,
+      "step": 706
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 6.932624113475178e-06,
+      "loss": 0.1466,
+      "step": 707
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 6.926713947990544e-06,
+      "loss": 0.1807,
+      "step": 708
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 6.92080378250591e-06,
+      "loss": 0.0304,
+      "step": 709
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 6.914893617021278e-06,
+      "loss": 0.1558,
+      "step": 710
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 6.908983451536644e-06,
+      "loss": 0.0908,
+      "step": 711
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 6.90307328605201e-06,
+      "loss": 0.2957,
+      "step": 712
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 6.897163120567376e-06,
+      "loss": 0.0472,
+      "step": 713
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 6.891252955082743e-06,
+      "loss": 0.2326,
+      "step": 714
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 6.8853427895981095e-06,
+      "loss": 0.0773,
+      "step": 715
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 6.879432624113476e-06,
+      "loss": 0.1124,
+      "step": 716
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 6.873522458628842e-06,
+      "loss": 0.2249,
+      "step": 717
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 6.867612293144209e-06,
+      "loss": 0.0923,
+      "step": 718
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 6.861702127659575e-06,
+      "loss": 0.2394,
+      "step": 719
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 6.855791962174941e-06,
+      "loss": 0.0876,
+      "step": 720
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 6.849881796690307e-06,
+      "loss": 0.2303,
+      "step": 721
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 6.843971631205675e-06,
+      "loss": 0.1628,
+      "step": 722
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 6.838061465721041e-06,
+      "loss": 0.1544,
+      "step": 723
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 6.832151300236407e-06,
+      "loss": 0.238,
+      "step": 724
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 6.826241134751773e-06,
+      "loss": 0.1024,
+      "step": 725
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 6.82033096926714e-06,
+      "loss": 0.1822,
+      "step": 726
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 6.8144208037825065e-06,
+      "loss": 0.1514,
+      "step": 727
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 6.808510638297873e-06,
+      "loss": 0.1045,
+      "step": 728
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 6.802600472813239e-06,
+      "loss": 0.0577,
+      "step": 729
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 6.796690307328606e-06,
+      "loss": 0.0805,
+      "step": 730
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 6.790780141843972e-06,
+      "loss": 0.0701,
+      "step": 731
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.784869976359338e-06,
+      "loss": 0.1419,
+      "step": 732
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.778959810874704e-06,
+      "loss": 0.213,
+      "step": 733
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.773049645390072e-06,
+      "loss": 0.2103,
+      "step": 734
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.767139479905438e-06,
+      "loss": 0.0962,
+      "step": 735
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.761229314420804e-06,
+      "loss": 0.0706,
+      "step": 736
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.75531914893617e-06,
+      "loss": 0.1622,
+      "step": 737
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.749408983451537e-06,
+      "loss": 0.0692,
+      "step": 738
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.7434988179669035e-06,
+      "loss": 0.1209,
+      "step": 739
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.73758865248227e-06,
+      "loss": 0.0459,
+      "step": 740
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.731678486997636e-06,
+      "loss": 0.1919,
+      "step": 741
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.725768321513003e-06,
+      "loss": 0.1962,
+      "step": 742
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.71985815602837e-06,
+      "loss": 0.0693,
+      "step": 743
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.713947990543736e-06,
+      "loss": 0.1344,
+      "step": 744
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.708037825059103e-06,
+      "loss": 0.1271,
+      "step": 745
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.702127659574469e-06,
+      "loss": 0.047,
+      "step": 746
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.696217494089835e-06,
+      "loss": 0.2394,
+      "step": 747
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.690307328605201e-06,
+      "loss": 0.1064,
+      "step": 748
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.684397163120568e-06,
+      "loss": 0.0945,
+      "step": 749
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.678486997635934e-06,
+      "loss": 0.1546,
+      "step": 750
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.6725768321513005e-06,
+      "loss": 0.1956,
+      "step": 751
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.1185,
+      "step": 752
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.9666666666666667,
+      "eval_loss": 0.15507754683494568,
+      "eval_runtime": 29.0994,
+      "eval_samples_per_second": 5.155,
+      "eval_steps_per_second": 0.344,
+      "step": 752
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.6607565011820345e-06,
+      "loss": 0.1892,
+      "step": 753
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.654846335697401e-06,
+      "loss": 0.1414,
+      "step": 754
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.648936170212767e-06,
+      "loss": 0.1081,
+      "step": 755
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.643026004728133e-06,
+      "loss": 0.2987,
+      "step": 756
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.6371158392435e-06,
+      "loss": 0.0667,
+      "step": 757
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.631205673758866e-06,
+      "loss": 0.131,
+      "step": 758
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.625295508274232e-06,
+      "loss": 0.0547,
+      "step": 759
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.619385342789598e-06,
+      "loss": 0.1274,
+      "step": 760
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.613475177304965e-06,
+      "loss": 0.1664,
+      "step": 761
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 6.607565011820331e-06,
+      "loss": 0.164,
+      "step": 762
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 6.6016548463356975e-06,
+      "loss": 0.1083,
+      "step": 763
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 6.595744680851064e-06,
+      "loss": 0.0636,
+      "step": 764
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 6.5898345153664315e-06,
+      "loss": 0.1676,
+      "step": 765
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 6.583924349881798e-06,
+      "loss": 0.1677,
+      "step": 766
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 6.578014184397164e-06,
+      "loss": 0.0899,
+      "step": 767
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 6.57210401891253e-06,
+      "loss": 0.2001,
+      "step": 768
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 6.566193853427897e-06,
+      "loss": 0.0928,
+      "step": 769
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 6.560283687943263e-06,
+      "loss": 0.1827,
+      "step": 770
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 6.554373522458629e-06,
+      "loss": 0.1222,
+      "step": 771
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 6.548463356973995e-06,
+      "loss": 0.0772,
+      "step": 772
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 6.542553191489362e-06,
+      "loss": 0.2411,
+      "step": 773
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 6.536643026004728e-06,
+      "loss": 0.1582,
+      "step": 774
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 6.5307328605200945e-06,
+      "loss": 0.2083,
+      "step": 775
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 6.524822695035462e-06,
+      "loss": 0.0924,
+      "step": 776
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 6.5189125295508285e-06,
+      "loss": 0.1137,
+      "step": 777
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 6.513002364066195e-06,
+      "loss": 0.0822,
+      "step": 778
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 6.507092198581561e-06,
+      "loss": 0.1544,
+      "step": 779
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 6.501182033096928e-06,
+      "loss": 0.0848,
+      "step": 780
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 6.495271867612294e-06,
+      "loss": 0.1115,
+      "step": 781
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 6.48936170212766e-06,
+      "loss": 0.123,
+      "step": 782
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 6.483451536643026e-06,
+      "loss": 0.0808,
+      "step": 783
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 6.477541371158393e-06,
+      "loss": 0.1448,
+      "step": 784
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 6.471631205673759e-06,
+      "loss": 0.246,
+      "step": 785
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 6.465721040189125e-06,
+      "loss": 0.1282,
+      "step": 786
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 6.4598108747044915e-06,
+      "loss": 0.1451,
+      "step": 787
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 6.453900709219859e-06,
+      "loss": 0.1156,
+      "step": 788
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 6.4479905437352255e-06,
+      "loss": 0.1425,
+      "step": 789
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 6.442080378250592e-06,
+      "loss": 0.1463,
+      "step": 790
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 6.436170212765958e-06,
+      "loss": 0.0584,
+      "step": 791
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 6.430260047281325e-06,
+      "loss": 0.2019,
+      "step": 792
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 6.424349881796691e-06,
+      "loss": 0.1082,
+      "step": 793
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 6.418439716312057e-06,
+      "loss": 0.0866,
+      "step": 794
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 6.412529550827423e-06,
+      "loss": 0.1448,
+      "step": 795
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 6.40661938534279e-06,
+      "loss": 0.1286,
+      "step": 796
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 6.400709219858156e-06,
+      "loss": 0.0802,
+      "step": 797
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 6.394799054373522e-06,
+      "loss": 0.1393,
+      "step": 798
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 6.3888888888888885e-06,
+      "loss": 0.103,
+      "step": 799
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.2279948741197586,
+      "eval_runtime": 28.1284,
+      "eval_samples_per_second": 5.333,
+      "eval_steps_per_second": 0.356,
+      "step": 799
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 6.382978723404256e-06,
+      "loss": 0.2635,
+      "step": 800
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 6.3770685579196225e-06,
+      "loss": 0.0591,
+      "step": 801
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 6.371158392434989e-06,
+      "loss": 0.0733,
+      "step": 802
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 6.365248226950355e-06,
+      "loss": 0.1501,
+      "step": 803
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 6.359338061465722e-06,
+      "loss": 0.2733,
+      "step": 804
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 6.353427895981088e-06,
+      "loss": 0.0687,
+      "step": 805
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 6.347517730496454e-06,
+      "loss": 0.0686,
+      "step": 806
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 6.341607565011821e-06,
+      "loss": 0.1015,
+      "step": 807
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 6.335697399527188e-06,
+      "loss": 0.1556,
+      "step": 808
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.329787234042554e-06,
+      "loss": 0.1103,
+      "step": 809
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.32387706855792e-06,
+      "loss": 0.0845,
+      "step": 810
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.317966903073287e-06,
+      "loss": 0.1207,
+      "step": 811
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.312056737588653e-06,
+      "loss": 0.1115,
+      "step": 812
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.3061465721040195e-06,
+      "loss": 0.1794,
+      "step": 813
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.300236406619386e-06,
+      "loss": 0.1251,
+      "step": 814
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.294326241134753e-06,
+      "loss": 0.2005,
+      "step": 815
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.288416075650119e-06,
+      "loss": 0.2608,
+      "step": 816
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.282505910165485e-06,
+      "loss": 0.0828,
+      "step": 817
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.276595744680851e-06,
+      "loss": 0.1229,
+      "step": 818
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.270685579196219e-06,
+      "loss": 0.1358,
+      "step": 819
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.264775413711585e-06,
+      "loss": 0.1636,
+      "step": 820
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.258865248226951e-06,
+      "loss": 0.0838,
+      "step": 821
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.252955082742317e-06,
+      "loss": 0.1774,
+      "step": 822
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.247044917257684e-06,
+      "loss": 0.1998,
+      "step": 823
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.24113475177305e-06,
+      "loss": 0.05,
+      "step": 824
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.2352245862884165e-06,
+      "loss": 0.0612,
+      "step": 825
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.229314420803783e-06,
+      "loss": 0.0819,
+      "step": 826
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.22340425531915e-06,
+      "loss": 0.216,
+      "step": 827
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 6.217494089834516e-06,
+      "loss": 0.0731,
+      "step": 828
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 6.211583924349882e-06,
+      "loss": 0.0816,
+      "step": 829
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 6.205673758865248e-06,
+      "loss": 0.2151,
+      "step": 830
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 6.199763593380616e-06,
+      "loss": 0.0517,
+      "step": 831
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 6.193853427895982e-06,
+      "loss": 0.2564,
+      "step": 832
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 6.187943262411348e-06,
+      "loss": 0.1289,
+      "step": 833
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 6.182033096926714e-06,
+      "loss": 0.1269,
+      "step": 834
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 6.176122931442081e-06,
+      "loss": 0.2925,
+      "step": 835
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 6.170212765957447e-06,
+      "loss": 0.1634,
+      "step": 836
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 6.1643026004728135e-06,
+      "loss": 0.2278,
+      "step": 837
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 6.15839243498818e-06,
+      "loss": 0.1462,
+      "step": 838
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 6.152482269503547e-06,
+      "loss": 0.185,
+      "step": 839
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 6.146572104018913e-06,
+      "loss": 0.1466,
+      "step": 840
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 6.140661938534279e-06,
+      "loss": 0.0962,
+      "step": 841
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 6.134751773049647e-06,
+      "loss": 0.2847,
+      "step": 842
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 6.128841607565013e-06,
+      "loss": 0.1037,
+      "step": 843
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 6.122931442080379e-06,
+      "loss": 0.2013,
+      "step": 844
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 6.117021276595745e-06,
+      "loss": 0.079,
+      "step": 845
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 6.111111111111112e-06,
+      "loss": 0.2761,
+      "step": 846
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.18010368943214417,
+      "eval_runtime": 27.2389,
+      "eval_samples_per_second": 5.507,
+      "eval_steps_per_second": 0.367,
+      "step": 846
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 6.105200945626478e-06,
+      "loss": 0.1611,
+      "step": 847
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 6.099290780141844e-06,
+      "loss": 0.1067,
+      "step": 848
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 6.0933806146572105e-06,
+      "loss": 0.1764,
+      "step": 849
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 6.0874704491725775e-06,
+      "loss": 0.0898,
+      "step": 850
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 6.081560283687944e-06,
+      "loss": 0.2354,
+      "step": 851
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 6.07565011820331e-06,
+      "loss": 0.2319,
+      "step": 852
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 6.069739952718676e-06,
+      "loss": 0.1434,
+      "step": 853
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 6.063829787234044e-06,
+      "loss": 0.156,
+      "step": 854
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 6.05791962174941e-06,
+      "loss": 0.0485,
+      "step": 855
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 6.052009456264776e-06,
+      "loss": 0.1808,
+      "step": 856
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 6.046099290780142e-06,
+      "loss": 0.1715,
+      "step": 857
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 6.040189125295509e-06,
+      "loss": 0.0668,
+      "step": 858
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 6.034278959810875e-06,
+      "loss": 0.2415,
+      "step": 859
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 6.028368794326241e-06,
+      "loss": 0.1369,
+      "step": 860
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 6.0224586288416075e-06,
+      "loss": 0.1057,
+      "step": 861
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 6.0165484633569745e-06,
+      "loss": 0.1945,
+      "step": 862
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 6.010638297872341e-06,
+      "loss": 0.0814,
+      "step": 863
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 6.004728132387707e-06,
+      "loss": 0.1266,
+      "step": 864
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 5.998817966903073e-06,
+      "loss": 0.1774,
+      "step": 865
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 5.992907801418441e-06,
+      "loss": 0.2278,
+      "step": 866
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 5.986997635933807e-06,
+      "loss": 0.1181,
+      "step": 867
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 5.981087470449173e-06,
+      "loss": 0.1224,
+      "step": 868
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 5.975177304964539e-06,
+      "loss": 0.213,
+      "step": 869
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 5.969267139479906e-06,
+      "loss": 0.1497,
+      "step": 870
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 5.963356973995272e-06,
+      "loss": 0.1088,
+      "step": 871
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 5.957446808510638e-06,
+      "loss": 0.1364,
+      "step": 872
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 5.951536643026005e-06,
+      "loss": 0.145,
+      "step": 873
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 5.945626477541372e-06,
+      "loss": 0.1556,
+      "step": 874
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 5.9397163120567384e-06,
+      "loss": 0.3101,
+      "step": 875
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 5.933806146572105e-06,
+      "loss": 0.1418,
+      "step": 876
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 5.9278959810874716e-06,
+      "loss": 0.1525,
+      "step": 877
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 5.921985815602838e-06,
+      "loss": 0.1399,
+      "step": 878
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 5.916075650118204e-06,
+      "loss": 0.2677,
+      "step": 879
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 5.91016548463357e-06,
+      "loss": 0.1908,
+      "step": 880
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 5.904255319148937e-06,
+      "loss": 0.1428,
+      "step": 881
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 5.898345153664303e-06,
+      "loss": 0.0872,
+      "step": 882
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 5.892434988179669e-06,
+      "loss": 0.084,
+      "step": 883
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 5.886524822695035e-06,
+      "loss": 0.1621,
+      "step": 884
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 5.880614657210403e-06,
+      "loss": 0.1544,
+      "step": 885
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 5.874704491725769e-06,
+      "loss": 0.0779,
+      "step": 886
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 5.8687943262411355e-06,
+      "loss": 0.1479,
+      "step": 887
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 5.862884160756502e-06,
+      "loss": 0.1279,
+      "step": 888
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 5.8569739952718686e-06,
+      "loss": 0.1327,
+      "step": 889
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 5.851063829787235e-06,
+      "loss": 0.1567,
+      "step": 890
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 5.845153664302601e-06,
+      "loss": 0.1174,
+      "step": 891
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 5.839243498817967e-06,
+      "loss": 0.0869,
+      "step": 892
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 5.833333333333334e-06,
+      "loss": 0.331,
+      "step": 893
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.18192118406295776,
+      "eval_runtime": 28.7638,
+      "eval_samples_per_second": 5.215,
+      "eval_steps_per_second": 0.348,
+      "step": 893
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 5.8274231678487e-06,
+      "loss": 0.0696,
+      "step": 894
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 5.821513002364066e-06,
+      "loss": 0.3089,
+      "step": 895
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 5.815602836879432e-06,
+      "loss": 0.2479,
+      "step": 896
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 5.8096926713948e-06,
+      "loss": 0.119,
+      "step": 897
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 5.803782505910166e-06,
+      "loss": 0.1195,
+      "step": 898
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 5.7978723404255325e-06,
+      "loss": 0.1582,
+      "step": 899
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 5.791962174940899e-06,
+      "loss": 0.0841,
+      "step": 900
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 5.7860520094562656e-06,
+      "loss": 0.1501,
+      "step": 901
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 5.780141843971632e-06,
+      "loss": 0.1332,
+      "step": 902
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 5.774231678486998e-06,
+      "loss": 0.0599,
+      "step": 903
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 5.768321513002365e-06,
+      "loss": 0.0981,
+      "step": 904
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 5.762411347517731e-06,
+      "loss": 0.1576,
+      "step": 905
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 5.756501182033097e-06,
+      "loss": 0.1011,
+      "step": 906
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 5.750591016548463e-06,
+      "loss": 0.1679,
+      "step": 907
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 5.744680851063831e-06,
+      "loss": 0.1249,
+      "step": 908
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 5.738770685579197e-06,
+      "loss": 0.0783,
+      "step": 909
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 5.732860520094563e-06,
+      "loss": 0.2152,
+      "step": 910
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 5.7269503546099295e-06,
+      "loss": 0.1904,
+      "step": 911
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 5.7210401891252964e-06,
+      "loss": 0.2753,
+      "step": 912
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 5.7151300236406626e-06,
+      "loss": 0.2328,
+      "step": 913
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 5.709219858156029e-06,
+      "loss": 0.1073,
+      "step": 914
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 5.703309692671395e-06,
+      "loss": 0.0632,
+      "step": 915
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 5.697399527186762e-06,
+      "loss": 0.0974,
+      "step": 916
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 5.691489361702128e-06,
+      "loss": 0.2008,
+      "step": 917
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 5.685579196217494e-06,
+      "loss": 0.0444,
+      "step": 918
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 5.67966903073286e-06,
+      "loss": 0.0833,
+      "step": 919
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 5.673758865248228e-06,
+      "loss": 0.0686,
+      "step": 920
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 5.667848699763594e-06,
+      "loss": 0.1525,
+      "step": 921
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 5.66193853427896e-06,
+      "loss": 0.075,
+      "step": 922
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.6560283687943265e-06,
+      "loss": 0.0902,
+      "step": 923
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.6501182033096934e-06,
+      "loss": 0.1806,
+      "step": 924
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.6442080378250596e-06,
+      "loss": 0.136,
+      "step": 925
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 5.638297872340426e-06,
+      "loss": 0.3517,
+      "step": 926
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 5.632387706855792e-06,
+      "loss": 0.1229,
+      "step": 927
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 5.626477541371159e-06,
+      "loss": 0.2154,
+      "step": 928
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.620567375886525e-06,
+      "loss": 0.2972,
+      "step": 929
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.614657210401891e-06,
+      "loss": 0.1307,
+      "step": 930
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 5.608747044917257e-06,
+      "loss": 0.0522,
+      "step": 931
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 5.602836879432625e-06,
+      "loss": 0.1353,
+      "step": 932
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 5.596926713947991e-06,
+      "loss": 0.0459,
+      "step": 933
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 5.591016548463357e-06,
+      "loss": 0.0752,
+      "step": 934
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 5.5851063829787235e-06,
+      "loss": 0.2326,
+      "step": 935
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 5.5791962174940904e-06,
+      "loss": 0.1292,
+      "step": 936
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 5.5732860520094566e-06,
+      "loss": 0.0813,
+      "step": 937
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 5.567375886524823e-06,
+      "loss": 0.1031,
+      "step": 938
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.5614657210401905e-06,
+      "loss": 0.2272,
+      "step": 939
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 5.555555555555557e-06,
+      "loss": 0.1042,
+      "step": 940
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.19954480230808258,
+      "eval_runtime": 27.9281,
+      "eval_samples_per_second": 5.371,
+      "eval_steps_per_second": 0.358,
+      "step": 940
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 5.549645390070923e-06,
+      "loss": 0.2128,
+      "step": 941
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 5.543735224586289e-06,
+      "loss": 0.1099,
+      "step": 942
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 5.537825059101656e-06,
+      "loss": 0.1214,
+      "step": 943
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 5.531914893617022e-06,
+      "loss": 0.0531,
+      "step": 944
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 5.526004728132388e-06,
+      "loss": 0.1288,
+      "step": 945
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 5.520094562647754e-06,
+      "loss": 0.1089,
+      "step": 946
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 5.514184397163121e-06,
+      "loss": 0.0964,
+      "step": 947
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 5.5082742316784874e-06,
+      "loss": 0.1162,
+      "step": 948
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 5.5023640661938536e-06,
+      "loss": 0.185,
+      "step": 949
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 5.49645390070922e-06,
+      "loss": 0.195,
+      "step": 950
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 5.4905437352245875e-06,
+      "loss": 0.2165,
+      "step": 951
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 5.484633569739954e-06,
+      "loss": 0.139,
+      "step": 952
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 5.47872340425532e-06,
+      "loss": 0.1136,
+      "step": 953
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 5.472813238770686e-06,
+      "loss": 0.2375,
+      "step": 954
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 5.466903073286053e-06,
+      "loss": 0.05,
+      "step": 955
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 5.460992907801419e-06,
+      "loss": 0.1751,
+      "step": 956
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 5.455082742316785e-06,
+      "loss": 0.0755,
+      "step": 957
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 5.449172576832151e-06,
+      "loss": 0.1111,
+      "step": 958
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 5.443262411347518e-06,
+      "loss": 0.1234,
+      "step": 959
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 5.4373522458628844e-06,
+      "loss": 0.1906,
+      "step": 960
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 5.4314420803782506e-06,
+      "loss": 0.0711,
+      "step": 961
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 5.425531914893617e-06,
+      "loss": 0.1555,
+      "step": 962
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 5.4196217494089845e-06,
+      "loss": 0.2301,
+      "step": 963
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 5.413711583924351e-06,
+      "loss": 0.1399,
+      "step": 964
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 5.407801418439717e-06,
+      "loss": 0.0951,
+      "step": 965
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 5.401891252955083e-06,
+      "loss": 0.1925,
+      "step": 966
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 5.39598108747045e-06,
+      "loss": 0.1502,
+      "step": 967
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 5.390070921985816e-06,
+      "loss": 0.2213,
+      "step": 968
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 5.384160756501182e-06,
+      "loss": 0.1141,
+      "step": 969
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 5.378250591016549e-06,
+      "loss": 0.2517,
+      "step": 970
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 5.372340425531915e-06,
+      "loss": 0.0779,
+      "step": 971
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 5.3664302600472814e-06,
+      "loss": 0.0553,
+      "step": 972
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 5.3605200945626476e-06,
+      "loss": 0.1373,
+      "step": 973
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 5.354609929078015e-06,
+      "loss": 0.0989,
+      "step": 974
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 5.3486997635933815e-06,
+      "loss": 0.0927,
+      "step": 975
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 5.342789598108748e-06,
+      "loss": 0.0509,
+      "step": 976
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 5.336879432624114e-06,
+      "loss": 0.1441,
+      "step": 977
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 5.330969267139481e-06,
+      "loss": 0.0645,
+      "step": 978
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 5.325059101654847e-06,
+      "loss": 0.1518,
+      "step": 979
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 5.319148936170213e-06,
+      "loss": 0.0602,
+      "step": 980
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 5.313238770685579e-06,
+      "loss": 0.0645,
+      "step": 981
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 5.307328605200946e-06,
+      "loss": 0.1791,
+      "step": 982
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 5.301418439716312e-06,
+      "loss": 0.1088,
+      "step": 983
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 5.2955082742316784e-06,
+      "loss": 0.1369,
+      "step": 984
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 5.2895981087470446e-06,
+      "loss": 0.0793,
+      "step": 985
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 5.283687943262412e-06,
+      "loss": 0.0488,
+      "step": 986
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 5.2777777777777785e-06,
+      "loss": 0.218,
+      "step": 987
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.2048463374376297,
+      "eval_runtime": 28.1571,
+      "eval_samples_per_second": 5.327,
+      "eval_steps_per_second": 0.355,
+      "step": 987
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 5.271867612293145e-06,
+      "loss": 0.1207,
+      "step": 988
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 5.265957446808511e-06,
+      "loss": 0.0732,
+      "step": 989
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 5.260047281323878e-06,
+      "loss": 0.0743,
+      "step": 990
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 5.254137115839244e-06,
+      "loss": 0.0231,
+      "step": 991
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 5.24822695035461e-06,
+      "loss": 0.1706,
+      "step": 992
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 5.242316784869976e-06,
+      "loss": 0.1212,
+      "step": 993
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 5.236406619385343e-06,
+      "loss": 0.0977,
+      "step": 994
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 5.230496453900709e-06,
+      "loss": 0.1252,
+      "step": 995
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 5.2245862884160754e-06,
+      "loss": 0.0965,
+      "step": 996
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 5.218676122931442e-06,
+      "loss": 0.0741,
+      "step": 997
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 5.212765957446809e-06,
+      "loss": 0.2761,
+      "step": 998
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 5.2068557919621755e-06,
+      "loss": 0.0862,
+      "step": 999
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 5.200945626477542e-06,
+      "loss": 0.0869,
+      "step": 1000
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 5.195035460992908e-06,
+      "loss": 0.1431,
+      "step": 1001
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 5.189125295508275e-06,
+      "loss": 0.113,
+      "step": 1002
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 5.183215130023641e-06,
+      "loss": 0.0402,
+      "step": 1003
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 5.177304964539007e-06,
+      "loss": 0.2286,
+      "step": 1004
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 5.171394799054375e-06,
+      "loss": 0.1708,
+      "step": 1005
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 5.165484633569741e-06,
+      "loss": 0.1398,
+      "step": 1006
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 5.159574468085107e-06,
+      "loss": 0.2739,
+      "step": 1007
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 5.153664302600473e-06,
+      "loss": 0.0718,
+      "step": 1008
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 5.14775413711584e-06,
+      "loss": 0.121,
+      "step": 1009
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 5.141843971631206e-06,
+      "loss": 0.027,
+      "step": 1010
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 5.1359338061465725e-06,
+      "loss": 0.0362,
+      "step": 1011
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 5.130023640661939e-06,
+      "loss": 0.0415,
+      "step": 1012
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 5.124113475177306e-06,
+      "loss": 0.1507,
+      "step": 1013
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 5.118203309692672e-06,
+      "loss": 0.1205,
+      "step": 1014
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 5.112293144208038e-06,
+      "loss": 0.2023,
+      "step": 1015
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 5.106382978723404e-06,
+      "loss": 0.1421,
+      "step": 1016
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 5.100472813238772e-06,
+      "loss": 0.0935,
+      "step": 1017
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 5.094562647754138e-06,
+      "loss": 0.2314,
+      "step": 1018
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 5.088652482269504e-06,
+      "loss": 0.209,
+      "step": 1019
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 5.08274231678487e-06,
+      "loss": 0.0834,
+      "step": 1020
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 5.076832151300237e-06,
+      "loss": 0.091,
+      "step": 1021
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 5.070921985815603e-06,
+      "loss": 0.0392,
+      "step": 1022
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 5.0650118203309695e-06,
+      "loss": 0.1495,
+      "step": 1023
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 5.059101654846336e-06,
+      "loss": 0.1024,
+      "step": 1024
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 5.053191489361703e-06,
+      "loss": 0.162,
+      "step": 1025
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 5.047281323877069e-06,
+      "loss": 0.1167,
+      "step": 1026
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 5.041371158392435e-06,
+      "loss": 0.1394,
+      "step": 1027
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 5.035460992907801e-06,
+      "loss": 0.1382,
+      "step": 1028
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 5.029550827423169e-06,
+      "loss": 0.1269,
+      "step": 1029
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 5.023640661938535e-06,
+      "loss": 0.1219,
+      "step": 1030
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 5.017730496453901e-06,
+      "loss": 0.0926,
+      "step": 1031
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 5.011820330969267e-06,
+      "loss": 0.0577,
+      "step": 1032
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 5.005910165484634e-06,
+      "loss": 0.1622,
+      "step": 1033
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 5e-06,
+      "loss": 0.1605,
+      "step": 1034
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.2818104922771454,
+      "eval_runtime": 29.1233,
+      "eval_samples_per_second": 5.151,
+      "eval_steps_per_second": 0.343,
+      "step": 1034
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 4.994089834515367e-06,
+      "loss": 0.1548,
+      "step": 1035
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 4.9881796690307335e-06,
+      "loss": 0.089,
+      "step": 1036
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 4.9822695035461e-06,
+      "loss": 0.2128,
+      "step": 1037
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 4.976359338061466e-06,
+      "loss": 0.0399,
+      "step": 1038
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 4.970449172576833e-06,
+      "loss": 0.0874,
+      "step": 1039
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 4.964539007092199e-06,
+      "loss": 0.2084,
+      "step": 1040
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 4.958628841607566e-06,
+      "loss": 0.1339,
+      "step": 1041
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 4.952718676122932e-06,
+      "loss": 0.1327,
+      "step": 1042
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 4.946808510638298e-06,
+      "loss": 0.1134,
+      "step": 1043
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 4.940898345153664e-06,
+      "loss": 0.1062,
+      "step": 1044
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 4.934988179669031e-06,
+      "loss": 0.097,
+      "step": 1045
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 4.929078014184397e-06,
+      "loss": 0.1738,
+      "step": 1046
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 4.923167848699764e-06,
+      "loss": 0.1329,
+      "step": 1047
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 4.9172576832151305e-06,
+      "loss": 0.0517,
+      "step": 1048
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 4.911347517730497e-06,
+      "loss": 0.1219,
+      "step": 1049
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 4.905437352245863e-06,
+      "loss": 0.2236,
+      "step": 1050
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 4.89952718676123e-06,
+      "loss": 0.0675,
+      "step": 1051
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 4.893617021276596e-06,
+      "loss": 0.2771,
+      "step": 1052
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 4.887706855791963e-06,
+      "loss": 0.1065,
+      "step": 1053
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 4.881796690307329e-06,
+      "loss": 0.064,
+      "step": 1054
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 4.875886524822695e-06,
+      "loss": 0.0501,
+      "step": 1055
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 4.869976359338061e-06,
+      "loss": 0.178,
+      "step": 1056
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 4.864066193853428e-06,
+      "loss": 0.0726,
+      "step": 1057
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 4.858156028368794e-06,
+      "loss": 0.1107,
+      "step": 1058
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 4.852245862884161e-06,
+      "loss": 0.1439,
+      "step": 1059
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 4.8463356973995275e-06,
+      "loss": 0.1327,
+      "step": 1060
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 4.840425531914894e-06,
+      "loss": 0.1563,
+      "step": 1061
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 4.83451536643026e-06,
+      "loss": 0.1419,
+      "step": 1062
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 4.828605200945627e-06,
+      "loss": 0.1597,
+      "step": 1063
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 4.822695035460993e-06,
+      "loss": 0.0764,
+      "step": 1064
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 4.81678486997636e-06,
+      "loss": 0.0779,
+      "step": 1065
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 4.810874704491726e-06,
+      "loss": 0.0822,
+      "step": 1066
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 4.804964539007093e-06,
+      "loss": 0.1178,
+      "step": 1067
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 4.799054373522459e-06,
+      "loss": 0.099,
+      "step": 1068
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 4.793144208037825e-06,
+      "loss": 0.1089,
+      "step": 1069
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 4.787234042553192e-06,
+      "loss": 0.1218,
+      "step": 1070
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 4.781323877068558e-06,
+      "loss": 0.2479,
+      "step": 1071
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 4.775413711583925e-06,
+      "loss": 0.1294,
+      "step": 1072
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 4.7695035460992915e-06,
+      "loss": 0.1751,
+      "step": 1073
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 4.763593380614658e-06,
+      "loss": 0.1034,
+      "step": 1074
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 4.757683215130024e-06,
+      "loss": 0.2087,
+      "step": 1075
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 4.751773049645391e-06,
+      "loss": 0.0676,
+      "step": 1076
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 4.745862884160757e-06,
+      "loss": 0.0755,
+      "step": 1077
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 4.739952718676124e-06,
+      "loss": 0.0997,
+      "step": 1078
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 4.73404255319149e-06,
+      "loss": 0.0459,
+      "step": 1079
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 4.728132387706856e-06,
+      "loss": 0.2043,
+      "step": 1080
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 4.722222222222222e-06,
+      "loss": 0.1551,
+      "step": 1081
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.24055908620357513,
+      "eval_runtime": 28.2087,
+      "eval_samples_per_second": 5.318,
+      "eval_steps_per_second": 0.355,
+      "step": 1081
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 4.716312056737589e-06,
+      "loss": 0.202,
+      "step": 1082
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 4.710401891252955e-06,
+      "loss": 0.1006,
+      "step": 1083
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 4.704491725768322e-06,
+      "loss": 0.1791,
+      "step": 1084
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 4.6985815602836885e-06,
+      "loss": 0.1852,
+      "step": 1085
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 4.692671394799055e-06,
+      "loss": 0.1216,
+      "step": 1086
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 4.686761229314421e-06,
+      "loss": 0.0893,
+      "step": 1087
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 4.680851063829788e-06,
+      "loss": 0.2344,
+      "step": 1088
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 4.674940898345154e-06,
+      "loss": 0.2026,
+      "step": 1089
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 4.669030732860521e-06,
+      "loss": 0.2164,
+      "step": 1090
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 4.663120567375887e-06,
+      "loss": 0.1566,
+      "step": 1091
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 4.657210401891253e-06,
+      "loss": 0.1407,
+      "step": 1092
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 4.651300236406619e-06,
+      "loss": 0.1073,
+      "step": 1093
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 4.645390070921986e-06,
+      "loss": 0.0446,
+      "step": 1094
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 4.639479905437352e-06,
+      "loss": 0.1398,
+      "step": 1095
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 4.633569739952719e-06,
+      "loss": 0.0538,
+      "step": 1096
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 4.6276595744680855e-06,
+      "loss": 0.1449,
+      "step": 1097
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 4.621749408983452e-06,
+      "loss": 0.2262,
+      "step": 1098
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 4.615839243498818e-06,
+      "loss": 0.125,
+      "step": 1099
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 4.609929078014185e-06,
+      "loss": 0.0953,
+      "step": 1100
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 4.604018912529552e-06,
+      "loss": 0.1365,
+      "step": 1101
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 4.598108747044918e-06,
+      "loss": 0.0789,
+      "step": 1102
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 4.592198581560284e-06,
+      "loss": 0.0961,
+      "step": 1103
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 4.58628841607565e-06,
+      "loss": 0.0873,
+      "step": 1104
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 4.580378250591017e-06,
+      "loss": 0.036,
+      "step": 1105
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 4.574468085106383e-06,
+      "loss": 0.1164,
+      "step": 1106
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 4.56855791962175e-06,
+      "loss": 0.1239,
+      "step": 1107
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 4.562647754137116e-06,
+      "loss": 0.12,
+      "step": 1108
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 4.5567375886524825e-06,
+      "loss": 0.158,
+      "step": 1109
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 4.550827423167849e-06,
+      "loss": 0.0835,
+      "step": 1110
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 4.544917257683216e-06,
+      "loss": 0.1776,
+      "step": 1111
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 4.539007092198582e-06,
+      "loss": 0.1395,
+      "step": 1112
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 4.533096926713949e-06,
+      "loss": 0.0552,
+      "step": 1113
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 4.527186761229315e-06,
+      "loss": 0.1748,
+      "step": 1114
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 4.521276595744681e-06,
+      "loss": 0.0891,
+      "step": 1115
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 4.515366430260047e-06,
+      "loss": 0.1899,
+      "step": 1116
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 4.509456264775414e-06,
+      "loss": 0.0773,
+      "step": 1117
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 4.50354609929078e-06,
+      "loss": 0.1076,
+      "step": 1118
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 4.497635933806147e-06,
+      "loss": 0.0285,
+      "step": 1119
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 4.491725768321513e-06,
+      "loss": 0.2005,
+      "step": 1120
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 4.4858156028368795e-06,
+      "loss": 0.0516,
+      "step": 1121
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 4.479905437352246e-06,
+      "loss": 0.1147,
+      "step": 1122
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 4.473995271867613e-06,
+      "loss": 0.0674,
+      "step": 1123
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 4.468085106382979e-06,
+      "loss": 0.0741,
+      "step": 1124
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 4.462174940898346e-06,
+      "loss": 0.0902,
+      "step": 1125
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 4.456264775413712e-06,
+      "loss": 0.0974,
+      "step": 1126
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 4.450354609929078e-06,
+      "loss": 0.1643,
+      "step": 1127
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.1109,
+      "step": 1128
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.18501786887645721,
+      "eval_runtime": 27.9288,
+      "eval_samples_per_second": 5.371,
+      "eval_steps_per_second": 0.358,
+      "step": 1128
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 4.438534278959811e-06,
+      "loss": 0.1377,
+      "step": 1129
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 4.432624113475177e-06,
+      "loss": 0.1532,
+      "step": 1130
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 4.426713947990544e-06,
+      "loss": 0.1822,
+      "step": 1131
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 4.42080378250591e-06,
+      "loss": 0.0808,
+      "step": 1132
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 4.414893617021277e-06,
+      "loss": 0.0935,
+      "step": 1133
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 4.4089834515366435e-06,
+      "loss": 0.0386,
+      "step": 1134
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 4.40307328605201e-06,
+      "loss": 0.156,
+      "step": 1135
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 4.397163120567377e-06,
+      "loss": 0.1532,
+      "step": 1136
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 4.391252955082743e-06,
+      "loss": 0.1389,
+      "step": 1137
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 4.38534278959811e-06,
+      "loss": 0.0726,
+      "step": 1138
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 4.379432624113476e-06,
+      "loss": 0.1013,
+      "step": 1139
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 4.373522458628842e-06,
+      "loss": 0.0999,
+      "step": 1140
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 4.367612293144208e-06,
+      "loss": 0.0918,
+      "step": 1141
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 4.361702127659575e-06,
+      "loss": 0.0476,
+      "step": 1142
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 4.355791962174941e-06,
+      "loss": 0.0313,
+      "step": 1143
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 4.349881796690308e-06,
+      "loss": 0.1126,
+      "step": 1144
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 4.343971631205674e-06,
+      "loss": 0.1825,
+      "step": 1145
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 4.3380614657210405e-06,
+      "loss": 0.1783,
+      "step": 1146
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 4.332151300236407e-06,
+      "loss": 0.0545,
+      "step": 1147
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 4.326241134751774e-06,
+      "loss": 0.1425,
+      "step": 1148
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 4.32033096926714e-06,
+      "loss": 0.1151,
+      "step": 1149
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 4.314420803782507e-06,
+      "loss": 0.0882,
+      "step": 1150
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 4.308510638297873e-06,
+      "loss": 0.085,
+      "step": 1151
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 4.302600472813239e-06,
+      "loss": 0.2536,
+      "step": 1152
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 4.296690307328605e-06,
+      "loss": 0.0977,
+      "step": 1153
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 4.290780141843972e-06,
+      "loss": 0.1242,
+      "step": 1154
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 4.284869976359338e-06,
+      "loss": 0.1048,
+      "step": 1155
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 4.278959810874705e-06,
+      "loss": 0.0692,
+      "step": 1156
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 4.273049645390071e-06,
+      "loss": 0.1447,
+      "step": 1157
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 4.2671394799054375e-06,
+      "loss": 0.1317,
+      "step": 1158
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 4.261229314420804e-06,
+      "loss": 0.2195,
+      "step": 1159
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 4.255319148936171e-06,
+      "loss": 0.1035,
+      "step": 1160
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 4.249408983451537e-06,
+      "loss": 0.068,
+      "step": 1161
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 4.243498817966904e-06,
+      "loss": 0.1618,
+      "step": 1162
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 4.23758865248227e-06,
+      "loss": 0.0644,
+      "step": 1163
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 4.231678486997636e-06,
+      "loss": 0.1419,
+      "step": 1164
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 4.225768321513003e-06,
+      "loss": 0.1503,
+      "step": 1165
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 4.219858156028369e-06,
+      "loss": 0.1438,
+      "step": 1166
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 4.213947990543736e-06,
+      "loss": 0.1886,
+      "step": 1167
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 4.208037825059102e-06,
+      "loss": 0.0759,
+      "step": 1168
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 4.202127659574468e-06,
+      "loss": 0.09,
+      "step": 1169
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 4.1962174940898345e-06,
+      "loss": 0.2031,
+      "step": 1170
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 4.1903073286052015e-06,
+      "loss": 0.1132,
+      "step": 1171
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 4.184397163120568e-06,
+      "loss": 0.1027,
+      "step": 1172
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 4.178486997635935e-06,
+      "loss": 0.0526,
+      "step": 1173
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 4.172576832151301e-06,
+      "loss": 0.0405,
+      "step": 1174
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 0.2726,
+      "step": 1175
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.20962999761104584,
+      "eval_runtime": 27.6607,
+      "eval_samples_per_second": 5.423,
+      "eval_steps_per_second": 0.362,
+      "step": 1175
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 4.160756501182033e-06,
+      "loss": 0.0859,
+      "step": 1176
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 4.1548463356974e-06,
+      "loss": 0.0819,
+      "step": 1177
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 4.148936170212766e-06,
+      "loss": 0.0858,
+      "step": 1178
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 4.143026004728133e-06,
+      "loss": 0.1408,
+      "step": 1179
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 4.137115839243499e-06,
+      "loss": 0.1143,
+      "step": 1180
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 4.131205673758865e-06,
+      "loss": 0.0642,
+      "step": 1181
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 4.1252955082742315e-06,
+      "loss": 0.0556,
+      "step": 1182
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 4.1193853427895985e-06,
+      "loss": 0.1235,
+      "step": 1183
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 4.113475177304965e-06,
+      "loss": 0.093,
+      "step": 1184
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 4.107565011820332e-06,
+      "loss": 0.0835,
+      "step": 1185
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 4.101654846335698e-06,
+      "loss": 0.048,
+      "step": 1186
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 4.095744680851064e-06,
+      "loss": 0.0586,
+      "step": 1187
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 4.08983451536643e-06,
+      "loss": 0.0718,
+      "step": 1188
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 4.083924349881797e-06,
+      "loss": 0.0602,
+      "step": 1189
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 4.078014184397163e-06,
+      "loss": 0.0671,
+      "step": 1190
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 4.07210401891253e-06,
+      "loss": 0.0949,
+      "step": 1191
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 4.066193853427896e-06,
+      "loss": 0.1588,
+      "step": 1192
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 4.060283687943262e-06,
+      "loss": 0.1036,
+      "step": 1193
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 4.0543735224586285e-06,
+      "loss": 0.1236,
+      "step": 1194
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 4.0484633569739955e-06,
+      "loss": 0.0811,
+      "step": 1195
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 4.042553191489362e-06,
+      "loss": 0.0559,
+      "step": 1196
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 4.036643026004729e-06,
+      "loss": 0.0554,
+      "step": 1197
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 4.030732860520096e-06,
+      "loss": 0.1683,
+      "step": 1198
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 4.024822695035462e-06,
+      "loss": 0.1212,
+      "step": 1199
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 4.018912529550828e-06,
+      "loss": 0.1792,
+      "step": 1200
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 4.013002364066194e-06,
+      "loss": 0.1181,
+      "step": 1201
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 4.007092198581561e-06,
+      "loss": 0.0747,
+      "step": 1202
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 4.001182033096927e-06,
+      "loss": 0.1001,
+      "step": 1203
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.995271867612294e-06,
+      "loss": 0.1918,
+      "step": 1204
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.98936170212766e-06,
+      "loss": 0.107,
+      "step": 1205
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.983451536643026e-06,
+      "loss": 0.1343,
+      "step": 1206
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.9775413711583925e-06,
+      "loss": 0.1849,
+      "step": 1207
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.9716312056737595e-06,
+      "loss": 0.1238,
+      "step": 1208
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.965721040189126e-06,
+      "loss": 0.1323,
+      "step": 1209
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.959810874704493e-06,
+      "loss": 0.1411,
+      "step": 1210
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.953900709219859e-06,
+      "loss": 0.1956,
+      "step": 1211
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.947990543735225e-06,
+      "loss": 0.0644,
+      "step": 1212
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.942080378250591e-06,
+      "loss": 0.1501,
+      "step": 1213
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.936170212765958e-06,
+      "loss": 0.0763,
+      "step": 1214
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.930260047281324e-06,
+      "loss": 0.1151,
+      "step": 1215
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.924349881796691e-06,
+      "loss": 0.1943,
+      "step": 1216
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 3.918439716312057e-06,
+      "loss": 0.1411,
+      "step": 1217
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.912529550827423e-06,
+      "loss": 0.1352,
+      "step": 1218
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.9066193853427895e-06,
+      "loss": 0.1556,
+      "step": 1219
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.9007092198581565e-06,
+      "loss": 0.1769,
+      "step": 1220
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.894799054373523e-06,
+      "loss": 0.0837,
+      "step": 1221
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 3.88888888888889e-06,
+      "loss": 0.1511,
+      "step": 1222
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.23165231943130493,
+      "eval_runtime": 27.749,
+      "eval_samples_per_second": 5.406,
+      "eval_steps_per_second": 0.36,
+      "step": 1222
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.882978723404256e-06,
+      "loss": 0.0731,
+      "step": 1223
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.877068557919622e-06,
+      "loss": 0.0898,
+      "step": 1224
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.871158392434988e-06,
+      "loss": 0.1353,
+      "step": 1225
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.865248226950355e-06,
+      "loss": 0.0625,
+      "step": 1226
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.859338061465721e-06,
+      "loss": 0.0461,
+      "step": 1227
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.853427895981088e-06,
+      "loss": 0.1641,
+      "step": 1228
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.847517730496454e-06,
+      "loss": 0.0789,
+      "step": 1229
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.84160756501182e-06,
+      "loss": 0.1789,
+      "step": 1230
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.835697399527187e-06,
+      "loss": 0.2413,
+      "step": 1231
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.8297872340425535e-06,
+      "loss": 0.1122,
+      "step": 1232
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 3.8238770685579205e-06,
+      "loss": 0.2099,
+      "step": 1233
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.817966903073287e-06,
+      "loss": 0.061,
+      "step": 1234
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.812056737588653e-06,
+      "loss": 0.1362,
+      "step": 1235
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.8061465721040193e-06,
+      "loss": 0.1498,
+      "step": 1236
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.800236406619386e-06,
+      "loss": 0.1084,
+      "step": 1237
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.794326241134752e-06,
+      "loss": 0.1348,
+      "step": 1238
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.7884160756501185e-06,
+      "loss": 0.1376,
+      "step": 1239
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.7825059101654847e-06,
+      "loss": 0.188,
+      "step": 1240
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.7765957446808516e-06,
+      "loss": 0.1272,
+      "step": 1241
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.7706855791962178e-06,
+      "loss": 0.1375,
+      "step": 1242
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.7647754137115843e-06,
+      "loss": 0.2204,
+      "step": 1243
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.7588652482269505e-06,
+      "loss": 0.1515,
+      "step": 1244
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.752955082742317e-06,
+      "loss": 0.0913,
+      "step": 1245
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.7470449172576836e-06,
+      "loss": 0.1928,
+      "step": 1246
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.74113475177305e-06,
+      "loss": 0.0805,
+      "step": 1247
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.7352245862884163e-06,
+      "loss": 0.0536,
+      "step": 1248
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.729314420803783e-06,
+      "loss": 0.2634,
+      "step": 1249
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.723404255319149e-06,
+      "loss": 0.1905,
+      "step": 1250
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.717494089834516e-06,
+      "loss": 0.1861,
+      "step": 1251
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.711583924349882e-06,
+      "loss": 0.1037,
+      "step": 1252
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.7056737588652486e-06,
+      "loss": 0.2239,
+      "step": 1253
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.6997635933806148e-06,
+      "loss": 0.077,
+      "step": 1254
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.6938534278959813e-06,
+      "loss": 0.1357,
+      "step": 1255
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.6879432624113475e-06,
+      "loss": 0.1156,
+      "step": 1256
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.6820330969267145e-06,
+      "loss": 0.0922,
+      "step": 1257
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.6761229314420806e-06,
+      "loss": 0.0914,
+      "step": 1258
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.670212765957447e-06,
+      "loss": 0.1876,
+      "step": 1259
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.6643026004728133e-06,
+      "loss": 0.108,
+      "step": 1260
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.65839243498818e-06,
+      "loss": 0.0766,
+      "step": 1261
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.652482269503547e-06,
+      "loss": 0.0656,
+      "step": 1262
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.646572104018913e-06,
+      "loss": 0.1737,
+      "step": 1263
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 3.6406619385342795e-06,
+      "loss": 0.1517,
+      "step": 1264
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.6347517730496456e-06,
+      "loss": 0.1011,
+      "step": 1265
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.628841607565012e-06,
+      "loss": 0.0891,
+      "step": 1266
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.6229314420803783e-06,
+      "loss": 0.1164,
+      "step": 1267
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.6170212765957453e-06,
+      "loss": 0.0854,
+      "step": 1268
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.6111111111111115e-06,
+      "loss": 0.2063,
+      "step": 1269
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.2068907916545868,
+      "eval_runtime": 31.4797,
+      "eval_samples_per_second": 4.765,
+      "eval_steps_per_second": 0.318,
+      "step": 1269
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.605200945626478e-06,
+      "loss": 0.1211,
+      "step": 1270
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.599290780141844e-06,
+      "loss": 0.0374,
+      "step": 1271
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.5933806146572107e-06,
+      "loss": 0.0942,
+      "step": 1272
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.587470449172577e-06,
+      "loss": 0.0989,
+      "step": 1273
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.581560283687944e-06,
+      "loss": 0.108,
+      "step": 1274
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.57565011820331e-06,
+      "loss": 0.1878,
+      "step": 1275
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.5697399527186765e-06,
+      "loss": 0.0534,
+      "step": 1276
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.5638297872340426e-06,
+      "loss": 0.157,
+      "step": 1277
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.557919621749409e-06,
+      "loss": 0.0569,
+      "step": 1278
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.5520094562647758e-06,
+      "loss": 0.1313,
+      "step": 1279
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.5460992907801423e-06,
+      "loss": 0.1062,
+      "step": 1280
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.5401891252955085e-06,
+      "loss": 0.0707,
+      "step": 1281
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.534278959810875e-06,
+      "loss": 0.1848,
+      "step": 1282
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.528368794326241e-06,
+      "loss": 0.1236,
+      "step": 1283
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.522458628841608e-06,
+      "loss": 0.047,
+      "step": 1284
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.5165484633569743e-06,
+      "loss": 0.1346,
+      "step": 1285
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.510638297872341e-06,
+      "loss": 0.1902,
+      "step": 1286
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.504728132387707e-06,
+      "loss": 0.1967,
+      "step": 1287
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.4988179669030735e-06,
+      "loss": 0.2004,
+      "step": 1288
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.4929078014184396e-06,
+      "loss": 0.0804,
+      "step": 1289
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.4869976359338066e-06,
+      "loss": 0.0884,
+      "step": 1290
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.4810874704491728e-06,
+      "loss": 0.0945,
+      "step": 1291
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.4751773049645393e-06,
+      "loss": 0.0473,
+      "step": 1292
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.4692671394799055e-06,
+      "loss": 0.0797,
+      "step": 1293
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.463356973995272e-06,
+      "loss": 0.0599,
+      "step": 1294
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.457446808510639e-06,
+      "loss": 0.1282,
+      "step": 1295
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.451536643026005e-06,
+      "loss": 0.0599,
+      "step": 1296
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.4456264775413717e-06,
+      "loss": 0.1494,
+      "step": 1297
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.439716312056738e-06,
+      "loss": 0.164,
+      "step": 1298
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.4338061465721044e-06,
+      "loss": 0.1015,
+      "step": 1299
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.4278959810874705e-06,
+      "loss": 0.1259,
+      "step": 1300
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.4219858156028375e-06,
+      "loss": 0.1226,
+      "step": 1301
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.4160756501182036e-06,
+      "loss": 0.1062,
+      "step": 1302
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.41016548463357e-06,
+      "loss": 0.1546,
+      "step": 1303
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.4042553191489363e-06,
+      "loss": 0.0529,
+      "step": 1304
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.398345153664303e-06,
+      "loss": 0.1495,
+      "step": 1305
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.392434988179669e-06,
+      "loss": 0.1642,
+      "step": 1306
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.386524822695036e-06,
+      "loss": 0.1288,
+      "step": 1307
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.380614657210402e-06,
+      "loss": 0.1162,
+      "step": 1308
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.3747044917257687e-06,
+      "loss": 0.0531,
+      "step": 1309
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.368794326241135e-06,
+      "loss": 0.1363,
+      "step": 1310
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.3628841607565014e-06,
+      "loss": 0.1148,
+      "step": 1311
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.356973995271868e-06,
+      "loss": 0.1352,
+      "step": 1312
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.3510638297872345e-06,
+      "loss": 0.0842,
+      "step": 1313
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.3451536643026006e-06,
+      "loss": 0.1115,
+      "step": 1314
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.339243498817967e-06,
+      "loss": 0.1996,
+      "step": 1315
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.1137,
+      "step": 1316
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.9266666666666666,
+      "eval_loss": 0.22749407589435577,
+      "eval_runtime": 29.4353,
+      "eval_samples_per_second": 5.096,
+      "eval_steps_per_second": 0.34,
+      "step": 1316
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.3274231678487003e-06,
+      "loss": 0.0581,
+      "step": 1317
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.3215130023640664e-06,
+      "loss": 0.1556,
+      "step": 1318
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.315602836879433e-06,
+      "loss": 0.166,
+      "step": 1319
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.309692671394799e-06,
+      "loss": 0.1916,
+      "step": 1320
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.3037825059101657e-06,
+      "loss": 0.0268,
+      "step": 1321
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.297872340425532e-06,
+      "loss": 0.1913,
+      "step": 1322
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.291962174940899e-06,
+      "loss": 0.1247,
+      "step": 1323
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.286052009456265e-06,
+      "loss": 0.1408,
+      "step": 1324
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.2801418439716315e-06,
+      "loss": 0.1061,
+      "step": 1325
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 3.2742316784869976e-06,
+      "loss": 0.0646,
+      "step": 1326
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.268321513002364e-06,
+      "loss": 0.0968,
+      "step": 1327
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.262411347517731e-06,
+      "loss": 0.1353,
+      "step": 1328
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.2565011820330973e-06,
+      "loss": 0.0151,
+      "step": 1329
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.250591016548464e-06,
+      "loss": 0.1191,
+      "step": 1330
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 3.24468085106383e-06,
+      "loss": 0.1207,
+      "step": 1331
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.2387706855791966e-06,
+      "loss": 0.0849,
+      "step": 1332
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.2328605200945627e-06,
+      "loss": 0.1479,
+      "step": 1333
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.2269503546099297e-06,
+      "loss": 0.1848,
+      "step": 1334
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.221040189125296e-06,
+      "loss": 0.1349,
+      "step": 1335
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 3.2151300236406624e-06,
+      "loss": 0.2443,
+      "step": 1336
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.2092198581560285e-06,
+      "loss": 0.0991,
+      "step": 1337
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.203309692671395e-06,
+      "loss": 0.0309,
+      "step": 1338
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.197399527186761e-06,
+      "loss": 0.0964,
+      "step": 1339
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.191489361702128e-06,
+      "loss": 0.1839,
+      "step": 1340
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.1855791962174943e-06,
+      "loss": 0.099,
+      "step": 1341
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.179669030732861e-06,
+      "loss": 0.3016,
+      "step": 1342
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.173758865248227e-06,
+      "loss": 0.05,
+      "step": 1343
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.167848699763594e-06,
+      "loss": 0.0993,
+      "step": 1344
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.16193853427896e-06,
+      "loss": 0.0543,
+      "step": 1345
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.1560283687943267e-06,
+      "loss": 0.094,
+      "step": 1346
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 3.150118203309693e-06,
+      "loss": 0.1189,
+      "step": 1347
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.1442080378250594e-06,
+      "loss": 0.0278,
+      "step": 1348
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.1382978723404255e-06,
+      "loss": 0.0852,
+      "step": 1349
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.1323877068557925e-06,
+      "loss": 0.0542,
+      "step": 1350
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.1264775413711586e-06,
+      "loss": 0.1148,
+      "step": 1351
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 3.120567375886525e-06,
+      "loss": 0.0649,
+      "step": 1352
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.1146572104018913e-06,
+      "loss": 0.105,
+      "step": 1353
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.108747044917258e-06,
+      "loss": 0.095,
+      "step": 1354
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.102836879432624e-06,
+      "loss": 0.0901,
+      "step": 1355
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.096926713947991e-06,
+      "loss": 0.0679,
+      "step": 1356
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.091016548463357e-06,
+      "loss": 0.1174,
+      "step": 1357
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.0851063829787237e-06,
+      "loss": 0.0849,
+      "step": 1358
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.07919621749409e-06,
+      "loss": 0.111,
+      "step": 1359
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.0732860520094564e-06,
+      "loss": 0.0732,
+      "step": 1360
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.0673758865248233e-06,
+      "loss": 0.2551,
+      "step": 1361
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.0614657210401895e-06,
+      "loss": 0.1047,
+      "step": 1362
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.055555555555556e-06,
+      "loss": 0.1694,
+      "step": 1363
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.20361079275608063,
+      "eval_runtime": 30.8167,
+      "eval_samples_per_second": 4.867,
+      "eval_steps_per_second": 0.324,
+      "step": 1363
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.049645390070922e-06,
+      "loss": 0.1365,
+      "step": 1364
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.0437352245862887e-06,
+      "loss": 0.0676,
+      "step": 1365
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.037825059101655e-06,
+      "loss": 0.099,
+      "step": 1366
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.031914893617022e-06,
+      "loss": 0.1443,
+      "step": 1367
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.026004728132388e-06,
+      "loss": 0.0455,
+      "step": 1368
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.0200945626477545e-06,
+      "loss": 0.0583,
+      "step": 1369
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.0141843971631207e-06,
+      "loss": 0.2367,
+      "step": 1370
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.0082742316784872e-06,
+      "loss": 0.1293,
+      "step": 1371
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.0023640661938534e-06,
+      "loss": 0.1683,
+      "step": 1372
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 2.9964539007092203e-06,
+      "loss": 0.0519,
+      "step": 1373
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 2.9905437352245865e-06,
+      "loss": 0.1117,
+      "step": 1374
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 2.984633569739953e-06,
+      "loss": 0.2395,
+      "step": 1375
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 2.978723404255319e-06,
+      "loss": 0.3579,
+      "step": 1376
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 2.972813238770686e-06,
+      "loss": 0.1681,
+      "step": 1377
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 2.9669030732860523e-06,
+      "loss": 0.1236,
+      "step": 1378
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 2.960992907801419e-06,
+      "loss": 0.0733,
+      "step": 1379
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 2.955082742316785e-06,
+      "loss": 0.166,
+      "step": 1380
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 2.9491725768321515e-06,
+      "loss": 0.1433,
+      "step": 1381
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 2.9432624113475177e-06,
+      "loss": 0.0828,
+      "step": 1382
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 2.9373522458628847e-06,
+      "loss": 0.1157,
+      "step": 1383
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 2.931442080378251e-06,
+      "loss": 0.0189,
+      "step": 1384
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 2.9255319148936174e-06,
+      "loss": 0.0496,
+      "step": 1385
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 2.9196217494089835e-06,
+      "loss": 0.0162,
+      "step": 1386
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 2.91371158392435e-06,
+      "loss": 0.0628,
+      "step": 1387
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 2.907801418439716e-06,
+      "loss": 0.112,
+      "step": 1388
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 2.901891252955083e-06,
+      "loss": 0.1458,
+      "step": 1389
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 2.8959810874704493e-06,
+      "loss": 0.0983,
+      "step": 1390
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 2.890070921985816e-06,
+      "loss": 0.0666,
+      "step": 1391
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 2.8841607565011824e-06,
+      "loss": 0.1232,
+      "step": 1392
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 2.8782505910165485e-06,
+      "loss": 0.0838,
+      "step": 1393
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 2.8723404255319155e-06,
+      "loss": 0.1409,
+      "step": 1394
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 2.8664302600472817e-06,
+      "loss": 0.0307,
+      "step": 1395
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 2.8605200945626482e-06,
+      "loss": 0.1027,
+      "step": 1396
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 2.8546099290780144e-06,
+      "loss": 0.1163,
+      "step": 1397
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 2.848699763593381e-06,
+      "loss": 0.1074,
+      "step": 1398
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 2.842789598108747e-06,
+      "loss": 0.1586,
+      "step": 1399
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 2.836879432624114e-06,
+      "loss": 0.279,
+      "step": 1400
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 2.83096926713948e-06,
+      "loss": 0.0282,
+      "step": 1401
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 2.8250591016548467e-06,
+      "loss": 0.0512,
+      "step": 1402
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 2.819148936170213e-06,
+      "loss": 0.0916,
+      "step": 1403
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 2.8132387706855794e-06,
+      "loss": 0.0527,
+      "step": 1404
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 2.8073286052009455e-06,
+      "loss": 0.0178,
+      "step": 1405
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 2.8014184397163125e-06,
+      "loss": 0.0459,
+      "step": 1406
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 2.7955082742316787e-06,
+      "loss": 0.1192,
+      "step": 1407
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 2.7895981087470452e-06,
+      "loss": 0.1251,
+      "step": 1408
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 2.7836879432624114e-06,
+      "loss": 0.1609,
+      "step": 1409
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 2.7777777777777783e-06,
+      "loss": 0.0561,
+      "step": 1410
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.21250122785568237,
+      "eval_runtime": 28.9261,
+      "eval_samples_per_second": 5.186,
+      "eval_steps_per_second": 0.346,
+      "step": 1410
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 2.7718676122931445e-06,
+      "loss": 0.1186,
+      "step": 1411
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 2.765957446808511e-06,
+      "loss": 0.074,
+      "step": 1412
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 2.760047281323877e-06,
+      "loss": 0.0515,
+      "step": 1413
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 2.7541371158392437e-06,
+      "loss": 0.1098,
+      "step": 1414
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 2.74822695035461e-06,
+      "loss": 0.0518,
+      "step": 1415
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 2.742316784869977e-06,
+      "loss": 0.2821,
+      "step": 1416
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 2.736406619385343e-06,
+      "loss": 0.1009,
+      "step": 1417
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 2.7304964539007095e-06,
+      "loss": 0.0495,
+      "step": 1418
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 2.7245862884160757e-06,
+      "loss": 0.1478,
+      "step": 1419
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 2.7186761229314422e-06,
+      "loss": 0.1491,
+      "step": 1420
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 2.7127659574468084e-06,
+      "loss": 0.0497,
+      "step": 1421
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 2.7068557919621753e-06,
+      "loss": 0.1341,
+      "step": 1422
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 2.7009456264775415e-06,
+      "loss": 0.0698,
+      "step": 1423
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 2.695035460992908e-06,
+      "loss": 0.1336,
+      "step": 1424
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 2.6891252955082746e-06,
+      "loss": 0.1709,
+      "step": 1425
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 2.6832151300236407e-06,
+      "loss": 0.1385,
+      "step": 1426
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 2.6773049645390077e-06,
+      "loss": 0.1465,
+      "step": 1427
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 2.671394799054374e-06,
+      "loss": 0.1178,
+      "step": 1428
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 2.6654846335697404e-06,
+      "loss": 0.0821,
+      "step": 1429
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 2.6595744680851065e-06,
+      "loss": 0.0225,
+      "step": 1430
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 2.653664302600473e-06,
+      "loss": 0.2123,
+      "step": 1431
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 2.6477541371158392e-06,
+      "loss": 0.1411,
+      "step": 1432
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 2.641843971631206e-06,
+      "loss": 0.0585,
+      "step": 1433
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 2.6359338061465723e-06,
+      "loss": 0.0914,
+      "step": 1434
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 2.630023640661939e-06,
+      "loss": 0.1812,
+      "step": 1435
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 2.624113475177305e-06,
+      "loss": 0.056,
+      "step": 1436
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 2.6182033096926716e-06,
+      "loss": 0.1424,
+      "step": 1437
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 2.6122931442080377e-06,
+      "loss": 0.0334,
+      "step": 1438
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 2.6063829787234047e-06,
+      "loss": 0.152,
+      "step": 1439
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 2.600472813238771e-06,
+      "loss": 0.1057,
+      "step": 1440
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 2.5945626477541374e-06,
+      "loss": 0.0527,
+      "step": 1441
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 2.5886524822695035e-06,
+      "loss": 0.108,
+      "step": 1442
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 2.5827423167848705e-06,
+      "loss": 0.1776,
+      "step": 1443
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 2.5768321513002366e-06,
+      "loss": 0.1777,
+      "step": 1444
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 2.570921985815603e-06,
+      "loss": 0.0754,
+      "step": 1445
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 2.5650118203309693e-06,
+      "loss": 0.1294,
+      "step": 1446
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 2.559101654846336e-06,
+      "loss": 0.1404,
+      "step": 1447
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 2.553191489361702e-06,
+      "loss": 0.0767,
+      "step": 1448
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 2.547281323877069e-06,
+      "loss": 0.2017,
+      "step": 1449
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 2.541371158392435e-06,
+      "loss": 0.1782,
+      "step": 1450
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 2.5354609929078017e-06,
+      "loss": 0.0856,
+      "step": 1451
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 2.529550827423168e-06,
+      "loss": 0.1707,
+      "step": 1452
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 2.5236406619385344e-06,
+      "loss": 0.1326,
+      "step": 1453
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 2.5177304964539005e-06,
+      "loss": 0.1536,
+      "step": 1454
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 2.5118203309692675e-06,
+      "loss": 0.1727,
+      "step": 1455
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 2.5059101654846336e-06,
+      "loss": 0.0578,
+      "step": 1456
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 2.5e-06,
+      "loss": 0.076,
+      "step": 1457
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.21977651119232178,
+      "eval_runtime": 30.711,
+      "eval_samples_per_second": 4.884,
+      "eval_steps_per_second": 0.326,
+      "step": 1457
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 2.4940898345153668e-06,
+      "loss": 0.0916,
+      "step": 1458
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 2.488179669030733e-06,
+      "loss": 0.1417,
+      "step": 1459
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 2.4822695035460995e-06,
+      "loss": 0.0267,
+      "step": 1460
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 2.476359338061466e-06,
+      "loss": 0.1474,
+      "step": 1461
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 2.470449172576832e-06,
+      "loss": 0.0832,
+      "step": 1462
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 2.4645390070921987e-06,
+      "loss": 0.0405,
+      "step": 1463
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 2.4586288416075653e-06,
+      "loss": 0.0599,
+      "step": 1464
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 2.4527186761229314e-06,
+      "loss": 0.1223,
+      "step": 1465
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 2.446808510638298e-06,
+      "loss": 0.0781,
+      "step": 1466
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 2.4408983451536645e-06,
+      "loss": 0.1001,
+      "step": 1467
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 2.4349881796690306e-06,
+      "loss": 0.1537,
+      "step": 1468
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 2.429078014184397e-06,
+      "loss": 0.1405,
+      "step": 1469
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 2.4231678486997638e-06,
+      "loss": 0.2278,
+      "step": 1470
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 2.41725768321513e-06,
+      "loss": 0.1465,
+      "step": 1471
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 2.4113475177304965e-06,
+      "loss": 0.0776,
+      "step": 1472
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 2.405437352245863e-06,
+      "loss": 0.1287,
+      "step": 1473
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 2.3995271867612296e-06,
+      "loss": 0.2023,
+      "step": 1474
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 2.393617021276596e-06,
+      "loss": 0.1138,
+      "step": 1475
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 2.3877068557919627e-06,
+      "loss": 0.0971,
+      "step": 1476
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 2.381796690307329e-06,
+      "loss": 0.1333,
+      "step": 1477
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 2.3758865248226954e-06,
+      "loss": 0.1307,
+      "step": 1478
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 2.369976359338062e-06,
+      "loss": 0.1668,
+      "step": 1479
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 2.364066193853428e-06,
+      "loss": 0.1154,
+      "step": 1480
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 2.3581560283687946e-06,
+      "loss": 0.1263,
+      "step": 1481
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 2.352245862884161e-06,
+      "loss": 0.0738,
+      "step": 1482
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 2.3463356973995273e-06,
+      "loss": 0.1134,
+      "step": 1483
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 2.340425531914894e-06,
+      "loss": 0.1297,
+      "step": 1484
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 2.3345153664302604e-06,
+      "loss": 0.2019,
+      "step": 1485
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 2.3286052009456266e-06,
+      "loss": 0.0438,
+      "step": 1486
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 2.322695035460993e-06,
+      "loss": 0.0363,
+      "step": 1487
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 2.3167848699763597e-06,
+      "loss": 0.0488,
+      "step": 1488
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 2.310874704491726e-06,
+      "loss": 0.0958,
+      "step": 1489
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 2.3049645390070924e-06,
+      "loss": 0.1228,
+      "step": 1490
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 2.299054373522459e-06,
+      "loss": 0.0581,
+      "step": 1491
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 2.293144208037825e-06,
+      "loss": 0.1054,
+      "step": 1492
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 2.2872340425531916e-06,
+      "loss": 0.2084,
+      "step": 1493
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 2.281323877068558e-06,
+      "loss": 0.0994,
+      "step": 1494
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 2.2754137115839243e-06,
+      "loss": 0.0138,
+      "step": 1495
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 2.269503546099291e-06,
+      "loss": 0.0961,
+      "step": 1496
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 2.2635933806146574e-06,
+      "loss": 0.0359,
+      "step": 1497
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 2.2576832151300236e-06,
+      "loss": 0.0527,
+      "step": 1498
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 2.25177304964539e-06,
+      "loss": 0.1778,
+      "step": 1499
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 2.2458628841607567e-06,
+      "loss": 0.0936,
+      "step": 1500
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 2.239952718676123e-06,
+      "loss": 0.0624,
+      "step": 1501
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 2.2340425531914894e-06,
+      "loss": 0.0442,
+      "step": 1502
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 2.228132387706856e-06,
+      "loss": 0.1033,
+      "step": 1503
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.0575,
+      "step": 1504
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.1989830583333969,
+      "eval_runtime": 28.5011,
+      "eval_samples_per_second": 5.263,
+      "eval_steps_per_second": 0.351,
+      "step": 1504
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 2.2163120567375886e-06,
+      "loss": 0.0872,
+      "step": 1505
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 2.210401891252955e-06,
+      "loss": 0.2034,
+      "step": 1506
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 2.2044917257683217e-06,
+      "loss": 0.1135,
+      "step": 1507
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 2.1985815602836883e-06,
+      "loss": 0.0755,
+      "step": 1508
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 2.192671394799055e-06,
+      "loss": 0.0839,
+      "step": 1509
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 2.186761229314421e-06,
+      "loss": 0.0365,
+      "step": 1510
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 2.1808510638297876e-06,
+      "loss": 0.045,
+      "step": 1511
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 2.174940898345154e-06,
+      "loss": 0.0871,
+      "step": 1512
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 2.1690307328605202e-06,
+      "loss": 0.0796,
+      "step": 1513
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 2.163120567375887e-06,
+      "loss": 0.0438,
+      "step": 1514
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 2.1572104018912534e-06,
+      "loss": 0.2849,
+      "step": 1515
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 2.1513002364066195e-06,
+      "loss": 0.1946,
+      "step": 1516
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 2.145390070921986e-06,
+      "loss": 0.0323,
+      "step": 1517
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 2.1394799054373526e-06,
+      "loss": 0.1012,
+      "step": 1518
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 2.1335697399527187e-06,
+      "loss": 0.1563,
+      "step": 1519
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 2.1276595744680853e-06,
+      "loss": 0.1287,
+      "step": 1520
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 2.121749408983452e-06,
+      "loss": 0.0585,
+      "step": 1521
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 2.115839243498818e-06,
+      "loss": 0.1837,
+      "step": 1522
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 2.1099290780141846e-06,
+      "loss": 0.1131,
+      "step": 1523
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 2.104018912529551e-06,
+      "loss": 0.1241,
+      "step": 1524
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 2.0981087470449173e-06,
+      "loss": 0.1577,
+      "step": 1525
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 2.092198581560284e-06,
+      "loss": 0.1185,
+      "step": 1526
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 2.0862884160756504e-06,
+      "loss": 0.0713,
+      "step": 1527
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 2.0803782505910165e-06,
+      "loss": 0.1414,
+      "step": 1528
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 2.074468085106383e-06,
+      "loss": 0.0833,
+      "step": 1529
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 2.0685579196217496e-06,
+      "loss": 0.0737,
+      "step": 1530
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 2.0626477541371158e-06,
+      "loss": 0.0679,
+      "step": 1531
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 2.0567375886524823e-06,
+      "loss": 0.1382,
+      "step": 1532
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 2.050827423167849e-06,
+      "loss": 0.0568,
+      "step": 1533
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 2.044917257683215e-06,
+      "loss": 0.076,
+      "step": 1534
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 2.0390070921985816e-06,
+      "loss": 0.0723,
+      "step": 1535
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 2.033096926713948e-06,
+      "loss": 0.1098,
+      "step": 1536
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 2.0271867612293143e-06,
+      "loss": 0.1371,
+      "step": 1537
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 2.021276595744681e-06,
+      "loss": 0.1218,
+      "step": 1538
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 2.015366430260048e-06,
+      "loss": 0.1195,
+      "step": 1539
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 2.009456264775414e-06,
+      "loss": 0.0855,
+      "step": 1540
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 2.0035460992907805e-06,
+      "loss": 0.0413,
+      "step": 1541
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 1.997635933806147e-06,
+      "loss": 0.1806,
+      "step": 1542
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 1.991725768321513e-06,
+      "loss": 0.1767,
+      "step": 1543
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 1.9858156028368797e-06,
+      "loss": 0.0408,
+      "step": 1544
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 1.9799054373522463e-06,
+      "loss": 0.0825,
+      "step": 1545
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 1.9739952718676124e-06,
+      "loss": 0.157,
+      "step": 1546
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 1.968085106382979e-06,
+      "loss": 0.1717,
+      "step": 1547
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 1.9621749408983455e-06,
+      "loss": 0.0413,
+      "step": 1548
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 1.9562647754137117e-06,
+      "loss": 0.0961,
+      "step": 1549
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 1.9503546099290782e-06,
+      "loss": 0.0795,
+      "step": 1550
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 1.944444444444445e-06,
+      "loss": 0.1484,
+      "step": 1551
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.9533333333333334,
+      "eval_loss": 0.18401113152503967,
+      "eval_runtime": 28.103,
+      "eval_samples_per_second": 5.338,
+      "eval_steps_per_second": 0.356,
+      "step": 1551
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 1.938534278959811e-06,
+      "loss": 0.073,
+      "step": 1552
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 1.9326241134751775e-06,
+      "loss": 0.0744,
+      "step": 1553
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 1.926713947990544e-06,
+      "loss": 0.0608,
+      "step": 1554
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 1.92080378250591e-06,
+      "loss": 0.1015,
+      "step": 1555
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 1.9148936170212767e-06,
+      "loss": 0.0964,
+      "step": 1556
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 1.9089834515366433e-06,
+      "loss": 0.0788,
+      "step": 1557
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 1.9030732860520096e-06,
+      "loss": 0.0661,
+      "step": 1558
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 1.897163120567376e-06,
+      "loss": 0.0476,
+      "step": 1559
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 1.8912529550827423e-06,
+      "loss": 0.0365,
+      "step": 1560
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 1.8853427895981089e-06,
+      "loss": 0.0341,
+      "step": 1561
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 1.8794326241134752e-06,
+      "loss": 0.0306,
+      "step": 1562
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 1.8735224586288418e-06,
+      "loss": 0.1458,
+      "step": 1563
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 1.8676122931442081e-06,
+      "loss": 0.0864,
+      "step": 1564
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 1.8617021276595745e-06,
+      "loss": 0.0459,
+      "step": 1565
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 1.855791962174941e-06,
+      "loss": 0.0901,
+      "step": 1566
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 1.8498817966903074e-06,
+      "loss": 0.1727,
+      "step": 1567
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 1.8439716312056737e-06,
+      "loss": 0.1663,
+      "step": 1568
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 1.8380614657210403e-06,
+      "loss": 0.106,
+      "step": 1569
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 1.8321513002364066e-06,
+      "loss": 0.0519,
+      "step": 1570
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 1.8262411347517734e-06,
+      "loss": 0.0207,
+      "step": 1571
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 1.8203309692671398e-06,
+      "loss": 0.0608,
+      "step": 1572
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 1.814420803782506e-06,
+      "loss": 0.1035,
+      "step": 1573
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 1.8085106382978727e-06,
+      "loss": 0.0596,
+      "step": 1574
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 1.802600472813239e-06,
+      "loss": 0.0841,
+      "step": 1575
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 1.7966903073286054e-06,
+      "loss": 0.1228,
+      "step": 1576
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 1.790780141843972e-06,
+      "loss": 0.1126,
+      "step": 1577
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 1.7848699763593383e-06,
+      "loss": 0.1051,
+      "step": 1578
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 1.7789598108747046e-06,
+      "loss": 0.1162,
+      "step": 1579
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 1.7730496453900712e-06,
+      "loss": 0.1586,
+      "step": 1580
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 1.7671394799054375e-06,
+      "loss": 0.089,
+      "step": 1581
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 1.761229314420804e-06,
+      "loss": 0.166,
+      "step": 1582
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 1.7553191489361704e-06,
+      "loss": 0.0845,
+      "step": 1583
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 1.7494089834515368e-06,
+      "loss": 0.0579,
+      "step": 1584
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 1.7434988179669033e-06,
+      "loss": 0.1985,
+      "step": 1585
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 1.7375886524822697e-06,
+      "loss": 0.0717,
+      "step": 1586
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 1.731678486997636e-06,
+      "loss": 0.1178,
+      "step": 1587
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 1.7257683215130026e-06,
+      "loss": 0.1241,
+      "step": 1588
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 1.719858156028369e-06,
+      "loss": 0.0214,
+      "step": 1589
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 1.7139479905437353e-06,
+      "loss": 0.0814,
+      "step": 1590
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 1.7080378250591018e-06,
+      "loss": 0.0727,
+      "step": 1591
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 1.7021276595744682e-06,
+      "loss": 0.1495,
+      "step": 1592
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 1.6962174940898345e-06,
+      "loss": 0.0435,
+      "step": 1593
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 1.690307328605201e-06,
+      "loss": 0.0983,
+      "step": 1594
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 1.6843971631205674e-06,
+      "loss": 0.0769,
+      "step": 1595
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 1.678486997635934e-06,
+      "loss": 0.0541,
+      "step": 1596
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 1.6725768321513003e-06,
+      "loss": 0.085,
+      "step": 1597
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 0.127,
+      "step": 1598
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.20804789662361145,
+      "eval_runtime": 27.2652,
+      "eval_samples_per_second": 5.502,
+      "eval_steps_per_second": 0.367,
+      "step": 1598
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 1.6607565011820332e-06,
+      "loss": 0.0302,
+      "step": 1599
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 1.6548463356973996e-06,
+      "loss": 0.0802,
+      "step": 1600
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 1.648936170212766e-06,
+      "loss": 0.1473,
+      "step": 1601
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 1.6430260047281325e-06,
+      "loss": 0.0942,
+      "step": 1602
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 1.6371158392434988e-06,
+      "loss": 0.0526,
+      "step": 1603
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 1.6312056737588656e-06,
+      "loss": 0.074,
+      "step": 1604
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 1.625295508274232e-06,
+      "loss": 0.1139,
+      "step": 1605
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 1.6193853427895983e-06,
+      "loss": 0.0913,
+      "step": 1606
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 1.6134751773049648e-06,
+      "loss": 0.1,
+      "step": 1607
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 1.6075650118203312e-06,
+      "loss": 0.1364,
+      "step": 1608
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 1.6016548463356975e-06,
+      "loss": 0.0926,
+      "step": 1609
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 1.595744680851064e-06,
+      "loss": 0.1021,
+      "step": 1610
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 1.5898345153664304e-06,
+      "loss": 0.0479,
+      "step": 1611
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 1.583924349881797e-06,
+      "loss": 0.2439,
+      "step": 1612
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 1.5780141843971633e-06,
+      "loss": 0.0291,
+      "step": 1613
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 1.5721040189125297e-06,
+      "loss": 0.0438,
+      "step": 1614
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 1.5661938534278962e-06,
+      "loss": 0.1005,
+      "step": 1615
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 1.5602836879432626e-06,
+      "loss": 0.0744,
+      "step": 1616
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 1.554373522458629e-06,
+      "loss": 0.0645,
+      "step": 1617
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 1.5484633569739955e-06,
+      "loss": 0.1591,
+      "step": 1618
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 1.5425531914893618e-06,
+      "loss": 0.1109,
+      "step": 1619
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 1.5366430260047282e-06,
+      "loss": 0.1086,
+      "step": 1620
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 1.5307328605200947e-06,
+      "loss": 0.0859,
+      "step": 1621
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 1.524822695035461e-06,
+      "loss": 0.1015,
+      "step": 1622
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 1.5189125295508274e-06,
+      "loss": 0.0672,
+      "step": 1623
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 1.513002364066194e-06,
+      "loss": 0.1186,
+      "step": 1624
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 1.5070921985815603e-06,
+      "loss": 0.1482,
+      "step": 1625
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 1.5011820330969267e-06,
+      "loss": 0.0755,
+      "step": 1626
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 1.4952718676122932e-06,
+      "loss": 0.0709,
+      "step": 1627
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 1.4893617021276596e-06,
+      "loss": 0.0685,
+      "step": 1628
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 1.4834515366430261e-06,
+      "loss": 0.133,
+      "step": 1629
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 1.4775413711583925e-06,
+      "loss": 0.0971,
+      "step": 1630
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 1.4716312056737588e-06,
+      "loss": 0.1336,
+      "step": 1631
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 1.4657210401891254e-06,
+      "loss": 0.0576,
+      "step": 1632
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 1.4598108747044917e-06,
+      "loss": 0.1251,
+      "step": 1633
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 1.453900709219858e-06,
+      "loss": 0.1915,
+      "step": 1634
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 1.4479905437352246e-06,
+      "loss": 0.1301,
+      "step": 1635
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 1.4420803782505912e-06,
+      "loss": 0.1445,
+      "step": 1636
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 1.4361702127659578e-06,
+      "loss": 0.0925,
+      "step": 1637
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 1.4302600472813241e-06,
+      "loss": 0.1404,
+      "step": 1638
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 1.4243498817966905e-06,
+      "loss": 0.04,
+      "step": 1639
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 1.418439716312057e-06,
+      "loss": 0.0277,
+      "step": 1640
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 1.4125295508274234e-06,
+      "loss": 0.1181,
+      "step": 1641
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 1.4066193853427897e-06,
+      "loss": 0.1559,
+      "step": 1642
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 1.4007092198581563e-06,
+      "loss": 0.1345,
+      "step": 1643
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 1.3947990543735226e-06,
+      "loss": 0.0852,
+      "step": 1644
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 1.3888888888888892e-06,
+      "loss": 0.1309,
+      "step": 1645
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.21183215081691742,
+      "eval_runtime": 28.4896,
+      "eval_samples_per_second": 5.265,
+      "eval_steps_per_second": 0.351,
+      "step": 1645
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 1.3829787234042555e-06,
+      "loss": 0.0925,
+      "step": 1646
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 1.3770685579196219e-06,
+      "loss": 0.0346,
+      "step": 1647
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 1.3711583924349884e-06,
+      "loss": 0.1104,
+      "step": 1648
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 1.3652482269503548e-06,
+      "loss": 0.2196,
+      "step": 1649
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 1.3593380614657211e-06,
+      "loss": 0.1117,
+      "step": 1650
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 1.3534278959810877e-06,
+      "loss": 0.0881,
+      "step": 1651
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 1.347517730496454e-06,
+      "loss": 0.0738,
+      "step": 1652
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 1.3416075650118204e-06,
+      "loss": 0.0646,
+      "step": 1653
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 1.335697399527187e-06,
+      "loss": 0.0561,
+      "step": 1654
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 1.3297872340425533e-06,
+      "loss": 0.0984,
+      "step": 1655
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 1.3238770685579196e-06,
+      "loss": 0.0491,
+      "step": 1656
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 1.3179669030732862e-06,
+      "loss": 0.1372,
+      "step": 1657
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 1.3120567375886525e-06,
+      "loss": 0.1142,
+      "step": 1658
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 1.3061465721040189e-06,
+      "loss": 0.1807,
+      "step": 1659
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 1.3002364066193854e-06,
+      "loss": 0.0575,
+      "step": 1660
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 1.2943262411347518e-06,
+      "loss": 0.1205,
+      "step": 1661
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 1.2884160756501183e-06,
+      "loss": 0.1414,
+      "step": 1662
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 1.2825059101654847e-06,
+      "loss": 0.1014,
+      "step": 1663
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 1.276595744680851e-06,
+      "loss": 0.1625,
+      "step": 1664
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 1.2706855791962176e-06,
+      "loss": 0.089,
+      "step": 1665
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 1.264775413711584e-06,
+      "loss": 0.1479,
+      "step": 1666
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 1.2588652482269503e-06,
+      "loss": 0.1924,
+      "step": 1667
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 1.2529550827423168e-06,
+      "loss": 0.1019,
+      "step": 1668
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 1.2470449172576834e-06,
+      "loss": 0.104,
+      "step": 1669
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 1.2411347517730497e-06,
+      "loss": 0.1407,
+      "step": 1670
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 1.235224586288416e-06,
+      "loss": 0.1395,
+      "step": 1671
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 1.2293144208037826e-06,
+      "loss": 0.1047,
+      "step": 1672
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 1.223404255319149e-06,
+      "loss": 0.0904,
+      "step": 1673
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 1.2174940898345153e-06,
+      "loss": 0.0968,
+      "step": 1674
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 1.2115839243498819e-06,
+      "loss": 0.0279,
+      "step": 1675
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 1.2056737588652482e-06,
+      "loss": 0.2175,
+      "step": 1676
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 1.1997635933806148e-06,
+      "loss": 0.0774,
+      "step": 1677
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 1.1938534278959813e-06,
+      "loss": 0.1148,
+      "step": 1678
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 1.1879432624113477e-06,
+      "loss": 0.0871,
+      "step": 1679
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 1.182033096926714e-06,
+      "loss": 0.1023,
+      "step": 1680
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 1.1761229314420806e-06,
+      "loss": 0.096,
+      "step": 1681
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 1.170212765957447e-06,
+      "loss": 0.0944,
+      "step": 1682
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 1.1643026004728133e-06,
+      "loss": 0.0433,
+      "step": 1683
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 1.1583924349881798e-06,
+      "loss": 0.1532,
+      "step": 1684
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 1.1524822695035462e-06,
+      "loss": 0.0751,
+      "step": 1685
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 1.1465721040189125e-06,
+      "loss": 0.2811,
+      "step": 1686
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 1.140661938534279e-06,
+      "loss": 0.0281,
+      "step": 1687
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 1.1347517730496454e-06,
+      "loss": 0.061,
+      "step": 1688
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 1.1288416075650118e-06,
+      "loss": 0.0607,
+      "step": 1689
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 1.1229314420803783e-06,
+      "loss": 0.1055,
+      "step": 1690
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 1.1170212765957447e-06,
+      "loss": 0.1605,
+      "step": 1691
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 1.111111111111111e-06,
+      "loss": 0.0675,
+      "step": 1692
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.22731274366378784,
+      "eval_runtime": 28.8146,
+      "eval_samples_per_second": 5.206,
+      "eval_steps_per_second": 0.347,
+      "step": 1692
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 1.1052009456264776e-06,
+      "loss": 0.2093,
+      "step": 1693
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 1.0992907801418442e-06,
+      "loss": 0.0379,
+      "step": 1694
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 1.0933806146572105e-06,
+      "loss": 0.0562,
+      "step": 1695
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 1.087470449172577e-06,
+      "loss": 0.0937,
+      "step": 1696
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 1.0815602836879434e-06,
+      "loss": 0.1512,
+      "step": 1697
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 1.0756501182033097e-06,
+      "loss": 0.1174,
+      "step": 1698
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 1.0697399527186763e-06,
+      "loss": 0.1466,
+      "step": 1699
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 1.0638297872340427e-06,
+      "loss": 0.1895,
+      "step": 1700
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 1.057919621749409e-06,
+      "loss": 0.0691,
+      "step": 1701
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 1.0520094562647756e-06,
+      "loss": 0.2142,
+      "step": 1702
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 1.046099290780142e-06,
+      "loss": 0.1616,
+      "step": 1703
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 1.0401891252955083e-06,
+      "loss": 0.0837,
+      "step": 1704
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 1.0342789598108748e-06,
+      "loss": 0.0858,
+      "step": 1705
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 1.0283687943262412e-06,
+      "loss": 0.0369,
+      "step": 1706
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 1.0224586288416075e-06,
+      "loss": 0.0924,
+      "step": 1707
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 1.016548463356974e-06,
+      "loss": 0.1045,
+      "step": 1708
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 1.0106382978723404e-06,
+      "loss": 0.0304,
+      "step": 1709
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 1.004728132387707e-06,
+      "loss": 0.1583,
+      "step": 1710
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 9.988179669030735e-07,
+      "loss": 0.1646,
+      "step": 1711
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 9.929078014184399e-07,
+      "loss": 0.0915,
+      "step": 1712
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 9.869976359338062e-07,
+      "loss": 0.1368,
+      "step": 1713
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 9.810874704491728e-07,
+      "loss": 0.0446,
+      "step": 1714
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 9.751773049645391e-07,
+      "loss": 0.2204,
+      "step": 1715
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 9.692671394799055e-07,
+      "loss": 0.1251,
+      "step": 1716
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 9.63356973995272e-07,
+      "loss": 0.0724,
+      "step": 1717
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 9.574468085106384e-07,
+      "loss": 0.2293,
+      "step": 1718
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 9.515366430260048e-07,
+      "loss": 0.0778,
+      "step": 1719
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 9.456264775413712e-07,
+      "loss": 0.0392,
+      "step": 1720
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 9.397163120567376e-07,
+      "loss": 0.0576,
+      "step": 1721
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 9.338061465721041e-07,
+      "loss": 0.0636,
+      "step": 1722
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 9.278959810874705e-07,
+      "loss": 0.2375,
+      "step": 1723
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 9.219858156028369e-07,
+      "loss": 0.1318,
+      "step": 1724
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 9.160756501182033e-07,
+      "loss": 0.1168,
+      "step": 1725
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 9.101654846335699e-07,
+      "loss": 0.0768,
+      "step": 1726
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 9.042553191489363e-07,
+      "loss": 0.0879,
+      "step": 1727
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 8.983451536643027e-07,
+      "loss": 0.1622,
+      "step": 1728
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 8.924349881796691e-07,
+      "loss": 0.0293,
+      "step": 1729
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 8.865248226950356e-07,
+      "loss": 0.0452,
+      "step": 1730
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 8.80614657210402e-07,
+      "loss": 0.0954,
+      "step": 1731
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 8.747044917257684e-07,
+      "loss": 0.0797,
+      "step": 1732
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 8.687943262411348e-07,
+      "loss": 0.1064,
+      "step": 1733
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 8.628841607565013e-07,
+      "loss": 0.0448,
+      "step": 1734
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 8.569739952718676e-07,
+      "loss": 0.0459,
+      "step": 1735
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 8.510638297872341e-07,
+      "loss": 0.066,
+      "step": 1736
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 8.451536643026005e-07,
+      "loss": 0.205,
+      "step": 1737
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 8.39243498817967e-07,
+      "loss": 0.1716,
+      "step": 1738
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 8.333333333333333e-07,
+      "loss": 0.0809,
+      "step": 1739
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.20993804931640625,
+      "eval_runtime": 29.8684,
+      "eval_samples_per_second": 5.022,
+      "eval_steps_per_second": 0.335,
+      "step": 1739
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 8.274231678486998e-07,
+      "loss": 0.1138,
+      "step": 1740
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 8.215130023640662e-07,
+      "loss": 0.1418,
+      "step": 1741
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 8.156028368794328e-07,
+      "loss": 0.1194,
+      "step": 1742
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 8.096926713947991e-07,
+      "loss": 0.048,
+      "step": 1743
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 8.037825059101656e-07,
+      "loss": 0.1289,
+      "step": 1744
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 7.97872340425532e-07,
+      "loss": 0.1016,
+      "step": 1745
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 7.919621749408985e-07,
+      "loss": 0.138,
+      "step": 1746
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 7.860520094562648e-07,
+      "loss": 0.0653,
+      "step": 1747
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 7.801418439716313e-07,
+      "loss": 0.0319,
+      "step": 1748
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 7.742316784869977e-07,
+      "loss": 0.1363,
+      "step": 1749
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 7.683215130023641e-07,
+      "loss": 0.1069,
+      "step": 1750
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 7.624113475177305e-07,
+      "loss": 0.0395,
+      "step": 1751
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 7.56501182033097e-07,
+      "loss": 0.085,
+      "step": 1752
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 7.505910165484633e-07,
+      "loss": 0.0521,
+      "step": 1753
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 7.446808510638298e-07,
+      "loss": 0.1596,
+      "step": 1754
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 7.387706855791962e-07,
+      "loss": 0.0545,
+      "step": 1755
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 7.328605200945627e-07,
+      "loss": 0.0757,
+      "step": 1756
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 7.26950354609929e-07,
+      "loss": 0.0266,
+      "step": 1757
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 7.210401891252956e-07,
+      "loss": 0.119,
+      "step": 1758
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 7.151300236406621e-07,
+      "loss": 0.1604,
+      "step": 1759
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 7.092198581560285e-07,
+      "loss": 0.1227,
+      "step": 1760
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 7.033096926713949e-07,
+      "loss": 0.0958,
+      "step": 1761
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 6.973995271867613e-07,
+      "loss": 0.1965,
+      "step": 1762
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 6.914893617021278e-07,
+      "loss": 0.1495,
+      "step": 1763
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 6.855791962174942e-07,
+      "loss": 0.0408,
+      "step": 1764
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 6.796690307328606e-07,
+      "loss": 0.0476,
+      "step": 1765
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 6.73758865248227e-07,
+      "loss": 0.1118,
+      "step": 1766
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 6.678486997635935e-07,
+      "loss": 0.0388,
+      "step": 1767
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 6.619385342789598e-07,
+      "loss": 0.1512,
+      "step": 1768
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 6.560283687943263e-07,
+      "loss": 0.1096,
+      "step": 1769
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 6.501182033096927e-07,
+      "loss": 0.1148,
+      "step": 1770
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 6.442080378250592e-07,
+      "loss": 0.2244,
+      "step": 1771
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 6.382978723404255e-07,
+      "loss": 0.2085,
+      "step": 1772
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 6.32387706855792e-07,
+      "loss": 0.2136,
+      "step": 1773
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 6.264775413711584e-07,
+      "loss": 0.1876,
+      "step": 1774
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 6.205673758865249e-07,
+      "loss": 0.1343,
+      "step": 1775
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 6.146572104018913e-07,
+      "loss": 0.0488,
+      "step": 1776
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 6.087470449172577e-07,
+      "loss": 0.1072,
+      "step": 1777
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 6.028368794326241e-07,
+      "loss": 0.064,
+      "step": 1778
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 5.969267139479907e-07,
+      "loss": 0.1474,
+      "step": 1779
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 5.91016548463357e-07,
+      "loss": 0.191,
+      "step": 1780
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 5.851063829787235e-07,
+      "loss": 0.0595,
+      "step": 1781
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 5.791962174940899e-07,
+      "loss": 0.0838,
+      "step": 1782
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 5.732860520094563e-07,
+      "loss": 0.1127,
+      "step": 1783
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 5.673758865248227e-07,
+      "loss": 0.1537,
+      "step": 1784
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 5.614657210401892e-07,
+      "loss": 0.1813,
+      "step": 1785
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 5.555555555555555e-07,
+      "loss": 0.1067,
+      "step": 1786
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.96,
+      "eval_loss": 0.19049935042858124,
+      "eval_runtime": 29.1257,
+      "eval_samples_per_second": 5.15,
+      "eval_steps_per_second": 0.343,
+      "step": 1786
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 5.496453900709221e-07,
+      "loss": 0.1782,
+      "step": 1787
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 5.437352245862885e-07,
+      "loss": 0.0702,
+      "step": 1788
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 5.378250591016549e-07,
+      "loss": 0.0907,
+      "step": 1789
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 5.319148936170213e-07,
+      "loss": 0.121,
+      "step": 1790
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 5.260047281323878e-07,
+      "loss": 0.0904,
+      "step": 1791
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 5.200945626477541e-07,
+      "loss": 0.0139,
+      "step": 1792
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 5.141843971631206e-07,
+      "loss": 0.123,
+      "step": 1793
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 5.08274231678487e-07,
+      "loss": 0.0546,
+      "step": 1794
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 5.023640661938535e-07,
+      "loss": 0.0923,
+      "step": 1795
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 4.964539007092199e-07,
+      "loss": 0.1283,
+      "step": 1796
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 4.905437352245864e-07,
+      "loss": 0.1077,
+      "step": 1797
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 4.846335697399527e-07,
+      "loss": 0.0083,
+      "step": 1798
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 4.787234042553192e-07,
+      "loss": 0.1163,
+      "step": 1799
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 4.728132387706856e-07,
+      "loss": 0.1415,
+      "step": 1800
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 4.6690307328605203e-07,
+      "loss": 0.1157,
+      "step": 1801
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 4.6099290780141843e-07,
+      "loss": 0.0275,
+      "step": 1802
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 4.5508274231678494e-07,
+      "loss": 0.1601,
+      "step": 1803
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 4.4917257683215134e-07,
+      "loss": 0.0757,
+      "step": 1804
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 4.432624113475178e-07,
+      "loss": 0.0584,
+      "step": 1805
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 4.373522458628842e-07,
+      "loss": 0.1073,
+      "step": 1806
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 4.3144208037825064e-07,
+      "loss": 0.1501,
+      "step": 1807
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 4.2553191489361704e-07,
+      "loss": 0.136,
+      "step": 1808
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 4.196217494089835e-07,
+      "loss": 0.1956,
+      "step": 1809
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 4.137115839243499e-07,
+      "loss": 0.0667,
+      "step": 1810
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 4.078014184397164e-07,
+      "loss": 0.0804,
+      "step": 1811
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 4.018912529550828e-07,
+      "loss": 0.1036,
+      "step": 1812
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.9598108747044925e-07,
+      "loss": 0.11,
+      "step": 1813
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.9007092198581565e-07,
+      "loss": 0.0792,
+      "step": 1814
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.8416075650118205e-07,
+      "loss": 0.0577,
+      "step": 1815
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.782505910165485e-07,
+      "loss": 0.148,
+      "step": 1816
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.723404255319149e-07,
+      "loss": 0.06,
+      "step": 1817
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.6643026004728135e-07,
+      "loss": 0.0755,
+      "step": 1818
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 3.605200945626478e-07,
+      "loss": 0.1108,
+      "step": 1819
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.5460992907801425e-07,
+      "loss": 0.0556,
+      "step": 1820
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.4869976359338065e-07,
+      "loss": 0.1119,
+      "step": 1821
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.427895981087471e-07,
+      "loss": 0.2159,
+      "step": 1822
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.368794326241135e-07,
+      "loss": 0.2013,
+      "step": 1823
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.309692671394799e-07,
+      "loss": 0.0403,
+      "step": 1824
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.2505910165484635e-07,
+      "loss": 0.1251,
+      "step": 1825
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.1914893617021275e-07,
+      "loss": 0.1151,
+      "step": 1826
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.132387706855792e-07,
+      "loss": 0.1483,
+      "step": 1827
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.0732860520094566e-07,
+      "loss": 0.0987,
+      "step": 1828
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.0141843971631206e-07,
+      "loss": 0.0604,
+      "step": 1829
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 2.955082742316785e-07,
+      "loss": 0.0416,
+      "step": 1830
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 2.8959810874704496e-07,
+      "loss": 0.1791,
+      "step": 1831
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 2.8368794326241136e-07,
+      "loss": 0.0535,
+      "step": 1832
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 2.7777777777777776e-07,
+      "loss": 0.0636,
+      "step": 1833
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.9466666666666667,
+      "eval_loss": 0.19790515303611755,
+      "eval_runtime": 28.6757,
+      "eval_samples_per_second": 5.231,
+      "eval_steps_per_second": 0.349,
+      "step": 1833
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 2.7186761229314426e-07,
+      "loss": 0.0251,
+      "step": 1834
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 2.6595744680851066e-07,
+      "loss": 0.2527,
+      "step": 1835
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 2.6004728132387706e-07,
+      "loss": 0.1374,
+      "step": 1836
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 2.541371158392435e-07,
+      "loss": 0.0723,
+      "step": 1837
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 2.4822695035460997e-07,
+      "loss": 0.0708,
+      "step": 1838
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 2.4231678486997637e-07,
+      "loss": 0.0547,
+      "step": 1839
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 2.364066193853428e-07,
+      "loss": 0.1723,
+      "step": 1840
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 2.3049645390070922e-07,
+      "loss": 0.0192,
+      "step": 1841
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 2.2458628841607567e-07,
+      "loss": 0.1793,
+      "step": 1842
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 2.186761229314421e-07,
+      "loss": 0.0421,
+      "step": 1843
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 2.1276595744680852e-07,
+      "loss": 0.0718,
+      "step": 1844
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 2.0685579196217495e-07,
+      "loss": 0.0543,
+      "step": 1845
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 2.009456264775414e-07,
+      "loss": 0.0871,
+      "step": 1846
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 1.9503546099290782e-07,
+      "loss": 0.0271,
+      "step": 1847
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 1.8912529550827425e-07,
+      "loss": 0.0955,
+      "step": 1848
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 1.8321513002364067e-07,
+      "loss": 0.1236,
+      "step": 1849
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 1.7730496453900713e-07,
+      "loss": 0.1269,
+      "step": 1850
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 1.7139479905437355e-07,
+      "loss": 0.0969,
+      "step": 1851
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 1.6548463356973995e-07,
+      "loss": 0.0403,
+      "step": 1852
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 1.5957446808510638e-07,
+      "loss": 0.1651,
+      "step": 1853
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 1.5366430260047283e-07,
+      "loss": 0.1306,
+      "step": 1854
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 1.4775413711583925e-07,
+      "loss": 0.1473,
+      "step": 1855
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 1.4184397163120568e-07,
+      "loss": 0.0832,
+      "step": 1856
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 1.3593380614657213e-07,
+      "loss": 0.1157,
+      "step": 1857
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 1.3002364066193853e-07,
+      "loss": 0.1489,
+      "step": 1858
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 1.2411347517730498e-07,
+      "loss": 0.1923,
+      "step": 1859
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 1.182033096926714e-07,
+      "loss": 0.049,
+      "step": 1860
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 1.1229314420803783e-07,
+      "loss": 0.1184,
+      "step": 1861
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 1.0638297872340426e-07,
+      "loss": 0.0994,
+      "step": 1862
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 1.004728132387707e-07,
+      "loss": 0.1859,
+      "step": 1863
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 9.456264775413712e-08,
+      "loss": 0.1702,
+      "step": 1864
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 8.865248226950356e-08,
+      "loss": 0.1668,
+      "step": 1865
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 8.274231678486998e-08,
+      "loss": 0.0321,
+      "step": 1866
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 7.683215130023641e-08,
+      "loss": 0.0753,
+      "step": 1867
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 7.092198581560284e-08,
+      "loss": 0.058,
+      "step": 1868
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 6.501182033096927e-08,
+      "loss": 0.1605,
+      "step": 1869
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 5.91016548463357e-08,
+      "loss": 0.2168,
+      "step": 1870
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 5.319148936170213e-08,
+      "loss": 0.102,
+      "step": 1871
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 4.728132387706856e-08,
+      "loss": 0.1099,
+      "step": 1872
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 4.137115839243499e-08,
+      "loss": 0.0837,
+      "step": 1873
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.546099290780142e-08,
+      "loss": 0.1766,
+      "step": 1874
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 2.955082742316785e-08,
+      "loss": 0.1302,
+      "step": 1875
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 2.364066193853428e-08,
+      "loss": 0.0703,
+      "step": 1876
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 1.773049645390071e-08,
+      "loss": 0.0305,
+      "step": 1877
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 1.182033096926714e-08,
+      "loss": 0.0858,
+      "step": 1878
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 5.91016548463357e-09,
+      "loss": 0.0925,
+      "step": 1879
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 0.0,
+      "loss": 0.0764,
+      "step": 1880
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.94,
+      "eval_loss": 0.20296132564544678,
+      "eval_runtime": 27.9673,
+      "eval_samples_per_second": 5.363,
+      "eval_steps_per_second": 0.358,
+      "step": 1880
+    },
+    {
+      "epoch": 40.0,
+      "step": 1880,
+      "total_flos": 9.727364893598024e+19,
+      "train_loss": 0.16249114074397514,
+      "train_runtime": 15279.6367,
+      "train_samples_per_second": 7.848,
+      "train_steps_per_second": 0.123
     }
   ],
-  "max_steps": 520,
+  "max_steps": 1880,
   "num_train_epochs": 40,
-  "total_flos": 2.658396678267912e+19,
+  "total_flos": 9.727364893598024e+19,
   "trial_name": null,
   "trial_params": null
 }