diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5079 @@
+{
+  "best_metric": 0.8791208791208791,
+  "best_model_checkpoint": "vit-msn-small-lateral_flow_ivalidation_train_test_6/checkpoint-318",
+  "epoch": 92.3076923076923,
+  "eval_steps": 500,
+  "global_step": 600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.15384615384615385,
+      "grad_norm": 8.29874038696289,
+      "learning_rate": 2.7777777777777776e-09,
+      "loss": 0.6813,
+      "step": 1
+    },
+    {
+      "epoch": 0.3076923076923077,
+      "grad_norm": 14.01459789276123,
+      "learning_rate": 5.555555555555555e-09,
+      "loss": 0.6845,
+      "step": 2
+    },
+    {
+      "epoch": 0.46153846153846156,
+      "grad_norm": 12.911795616149902,
+      "learning_rate": 8.333333333333334e-09,
+      "loss": 0.6836,
+      "step": 3
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 6.274200439453125,
+      "learning_rate": 1.111111111111111e-08,
+      "loss": 0.6521,
+      "step": 4
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 8.031621932983398,
+      "learning_rate": 1.3888888888888887e-08,
+      "loss": 0.6781,
+      "step": 5
+    },
+    {
+      "epoch": 0.9230769230769231,
+      "grad_norm": 7.054372787475586,
+      "learning_rate": 1.6666666666666667e-08,
+      "loss": 0.6672,
+      "step": 6
+    },
+    {
+      "epoch": 0.9230769230769231,
+      "eval_accuracy": 0.42124542124542125,
+      "eval_loss": 0.6979768872261047,
+      "eval_runtime": 1.1041,
+      "eval_samples_per_second": 247.254,
+      "eval_steps_per_second": 4.528,
+      "step": 6
+    },
+    {
+      "epoch": 1.0769230769230769,
+      "grad_norm": 7.790348529815674,
+      "learning_rate": 1.9444444444444445e-08,
+      "loss": 0.6725,
+      "step": 7
+    },
+    {
+      "epoch": 1.2307692307692308,
+      "grad_norm": 13.414072036743164,
+      "learning_rate": 2.222222222222222e-08,
+      "loss": 0.6652,
+      "step": 8
+    },
+    {
+      "epoch": 1.3846153846153846,
+      "grad_norm": 8.81810474395752,
+      "learning_rate": 2.5e-08,
+      "loss": 0.6698,
+      "step": 9
+    },
+    {
+      "epoch": 1.5384615384615383,
+      "grad_norm": 7.751201152801514,
+      "learning_rate": 2.7777777777777774e-08,
+      "loss": 0.6817,
+      "step": 10
+    },
+    {
+      "epoch": 1.6923076923076923,
+      "grad_norm": 8.985770225524902,
+      "learning_rate": 3.0555555555555556e-08,
+      "loss": 0.6661,
+      "step": 11
+    },
+    {
+      "epoch": 1.8461538461538463,
+      "grad_norm": 6.888631820678711,
+      "learning_rate": 3.3333333333333334e-08,
+      "loss": 0.6712,
+      "step": 12
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 8.5791597366333,
+      "learning_rate": 3.6111111111111106e-08,
+      "loss": 0.6617,
+      "step": 13
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.4249084249084249,
+      "eval_loss": 0.6965357065200806,
+      "eval_runtime": 1.1521,
+      "eval_samples_per_second": 236.95,
+      "eval_steps_per_second": 4.34,
+      "step": 13
+    },
+    {
+      "epoch": 2.1538461538461537,
+      "grad_norm": 9.721835136413574,
+      "learning_rate": 3.888888888888889e-08,
+      "loss": 0.665,
+      "step": 14
+    },
+    {
+      "epoch": 2.3076923076923075,
+      "grad_norm": 10.525633811950684,
+      "learning_rate": 4.166666666666666e-08,
+      "loss": 0.6629,
+      "step": 15
+    },
+    {
+      "epoch": 2.4615384615384617,
+      "grad_norm": 5.860161304473877,
+      "learning_rate": 4.444444444444444e-08,
+      "loss": 0.686,
+      "step": 16
+    },
+    {
+      "epoch": 2.6153846153846154,
+      "grad_norm": 7.980571269989014,
+      "learning_rate": 4.722222222222222e-08,
+      "loss": 0.6638,
+      "step": 17
+    },
+    {
+      "epoch": 2.769230769230769,
+      "grad_norm": 6.91188383102417,
+      "learning_rate": 5e-08,
+      "loss": 0.6586,
+      "step": 18
+    },
+    {
+      "epoch": 2.9230769230769234,
+      "grad_norm": 8.952598571777344,
+      "learning_rate": 5.2777777777777776e-08,
+      "loss": 0.6699,
+      "step": 19
+    },
+    {
+      "epoch": 2.9230769230769234,
+      "eval_accuracy": 0.43956043956043955,
+      "eval_loss": 0.6944313049316406,
+      "eval_runtime": 1.1568,
+      "eval_samples_per_second": 235.989,
+      "eval_steps_per_second": 4.322,
+      "step": 19
+    },
+    {
+      "epoch": 3.076923076923077,
+      "grad_norm": 16.572885513305664,
+      "learning_rate": 5.555555555555555e-08,
+      "loss": 0.6761,
+      "step": 20
+    },
+    {
+      "epoch": 3.230769230769231,
+      "grad_norm": 10.379450798034668,
+      "learning_rate": 5.833333333333333e-08,
+      "loss": 0.6757,
+      "step": 21
+    },
+    {
+      "epoch": 3.3846153846153846,
+      "grad_norm": 8.093114852905273,
+      "learning_rate": 6.111111111111111e-08,
+      "loss": 0.6572,
+      "step": 22
+    },
+    {
+      "epoch": 3.5384615384615383,
+      "grad_norm": 12.009868621826172,
+      "learning_rate": 6.388888888888888e-08,
+      "loss": 0.6781,
+      "step": 23
+    },
+    {
+      "epoch": 3.6923076923076925,
+      "grad_norm": 6.616479396820068,
+      "learning_rate": 6.666666666666667e-08,
+      "loss": 0.6519,
+      "step": 24
+    },
+    {
+      "epoch": 3.8461538461538463,
+      "grad_norm": 4.752511024475098,
+      "learning_rate": 6.944444444444444e-08,
+      "loss": 0.669,
+      "step": 25
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 6.52732515335083,
+      "learning_rate": 7.222222222222221e-08,
+      "loss": 0.662,
+      "step": 26
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.43956043956043955,
+      "eval_loss": 0.6909541487693787,
+      "eval_runtime": 1.1432,
+      "eval_samples_per_second": 238.799,
+      "eval_steps_per_second": 4.374,
+      "step": 26
+    },
+    {
+      "epoch": 4.153846153846154,
+      "grad_norm": 10.048100471496582,
+      "learning_rate": 7.5e-08,
+      "loss": 0.6616,
+      "step": 27
+    },
+    {
+      "epoch": 4.3076923076923075,
+      "grad_norm": 9.803552627563477,
+      "learning_rate": 7.777777777777778e-08,
+      "loss": 0.672,
+      "step": 28
+    },
+    {
+      "epoch": 4.461538461538462,
+      "grad_norm": 4.989064693450928,
+      "learning_rate": 8.055555555555555e-08,
+      "loss": 0.6615,
+      "step": 29
+    },
+    {
+      "epoch": 4.615384615384615,
+      "grad_norm": 9.584320068359375,
+      "learning_rate": 8.333333333333333e-08,
+      "loss": 0.6534,
+      "step": 30
+    },
+    {
+      "epoch": 4.769230769230769,
+      "grad_norm": 8.697301864624023,
+      "learning_rate": 8.611111111111111e-08,
+      "loss": 0.6592,
+      "step": 31
+    },
+    {
+      "epoch": 4.923076923076923,
+      "grad_norm": 6.305507183074951,
+      "learning_rate": 8.888888888888888e-08,
+      "loss": 0.6548,
+      "step": 32
+    },
+    {
+      "epoch": 4.923076923076923,
+      "eval_accuracy": 0.45787545787545786,
+      "eval_loss": 0.6873242855072021,
+      "eval_runtime": 1.0879,
+      "eval_samples_per_second": 250.944,
+      "eval_steps_per_second": 4.596,
+      "step": 32
+    },
+    {
+      "epoch": 5.076923076923077,
+      "grad_norm": 8.161693572998047,
+      "learning_rate": 9.166666666666665e-08,
+      "loss": 0.6463,
+      "step": 33
+    },
+    {
+      "epoch": 5.230769230769231,
+      "grad_norm": 5.462063789367676,
+      "learning_rate": 9.444444444444444e-08,
+      "loss": 0.6532,
+      "step": 34
+    },
+    {
+      "epoch": 5.384615384615385,
+      "grad_norm": 7.397779941558838,
+      "learning_rate": 9.722222222222222e-08,
+      "loss": 0.6447,
+      "step": 35
+    },
+    {
+      "epoch": 5.538461538461538,
+      "grad_norm": 9.939643859863281,
+      "learning_rate": 1e-07,
+      "loss": 0.6484,
+      "step": 36
+    },
+    {
+      "epoch": 5.6923076923076925,
+      "grad_norm": 6.816234111785889,
+      "learning_rate": 1.0277777777777777e-07,
+      "loss": 0.6558,
+      "step": 37
+    },
+    {
+      "epoch": 5.846153846153846,
+      "grad_norm": 11.238946914672852,
+      "learning_rate": 1.0555555555555555e-07,
+      "loss": 0.6669,
+      "step": 38
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 8.005824089050293,
+      "learning_rate": 1.0833333333333334e-07,
+      "loss": 0.6541,
+      "step": 39
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.4835164835164835,
+      "eval_loss": 0.6825475692749023,
+      "eval_runtime": 1.1582,
+      "eval_samples_per_second": 235.72,
+      "eval_steps_per_second": 4.317,
+      "step": 39
+    },
+    {
+      "epoch": 6.153846153846154,
+      "grad_norm": 6.5383524894714355,
+      "learning_rate": 1.111111111111111e-07,
+      "loss": 0.6317,
+      "step": 40
+    },
+    {
+      "epoch": 6.3076923076923075,
+      "grad_norm": 4.98549747467041,
+      "learning_rate": 1.1388888888888888e-07,
+      "loss": 0.6609,
+      "step": 41
+    },
+    {
+      "epoch": 6.461538461538462,
+      "grad_norm": 6.8932037353515625,
+      "learning_rate": 1.1666666666666667e-07,
+      "loss": 0.6459,
+      "step": 42
+    },
+    {
+      "epoch": 6.615384615384615,
+      "grad_norm": 6.804798126220703,
+      "learning_rate": 1.1944444444444445e-07,
+      "loss": 0.6628,
+      "step": 43
+    },
+    {
+      "epoch": 6.769230769230769,
+      "grad_norm": 7.5995707511901855,
+      "learning_rate": 1.2222222222222222e-07,
+      "loss": 0.6501,
+      "step": 44
+    },
+    {
+      "epoch": 6.923076923076923,
+      "grad_norm": 10.426745414733887,
+      "learning_rate": 1.25e-07,
+      "loss": 0.6222,
+      "step": 45
+    },
+    {
+      "epoch": 6.923076923076923,
+      "eval_accuracy": 0.5311355311355311,
+      "eval_loss": 0.6776841282844543,
+      "eval_runtime": 1.1748,
+      "eval_samples_per_second": 232.383,
+      "eval_steps_per_second": 4.256,
+      "step": 45
+    },
+    {
+      "epoch": 7.076923076923077,
+      "grad_norm": 6.766184329986572,
+      "learning_rate": 1.2777777777777777e-07,
+      "loss": 0.6452,
+      "step": 46
+    },
+    {
+      "epoch": 7.230769230769231,
+      "grad_norm": 10.556381225585938,
+      "learning_rate": 1.3055555555555556e-07,
+      "loss": 0.6157,
+      "step": 47
+    },
+    {
+      "epoch": 7.384615384615385,
+      "grad_norm": 9.157587051391602,
+      "learning_rate": 1.3333333333333334e-07,
+      "loss": 0.6313,
+      "step": 48
+    },
+    {
+      "epoch": 7.538461538461538,
+      "grad_norm": 6.186281204223633,
+      "learning_rate": 1.3611111111111108e-07,
+      "loss": 0.6773,
+      "step": 49
+    },
+    {
+      "epoch": 7.6923076923076925,
+      "grad_norm": 5.614987373352051,
+      "learning_rate": 1.3888888888888888e-07,
+      "loss": 0.6474,
+      "step": 50
+    },
+    {
+      "epoch": 7.846153846153846,
+      "grad_norm": 5.66176700592041,
+      "learning_rate": 1.4166666666666665e-07,
+      "loss": 0.6463,
+      "step": 51
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 4.733225345611572,
+      "learning_rate": 1.4444444444444442e-07,
+      "loss": 0.6555,
+      "step": 52
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.5421245421245421,
+      "eval_loss": 0.6718742251396179,
+      "eval_runtime": 1.1693,
+      "eval_samples_per_second": 233.475,
+      "eval_steps_per_second": 4.276,
+      "step": 52
+    },
+    {
+      "epoch": 8.153846153846153,
+      "grad_norm": 5.067059516906738,
+      "learning_rate": 1.4722222222222222e-07,
+      "loss": 0.6388,
+      "step": 53
+    },
+    {
+      "epoch": 8.307692307692308,
+      "grad_norm": 7.690587520599365,
+      "learning_rate": 1.5e-07,
+      "loss": 0.6139,
+      "step": 54
+    },
+    {
+      "epoch": 8.461538461538462,
+      "grad_norm": 4.471611976623535,
+      "learning_rate": 1.527777777777778e-07,
+      "loss": 0.64,
+      "step": 55
+    },
+    {
+      "epoch": 8.615384615384615,
+      "grad_norm": 6.964099884033203,
+      "learning_rate": 1.5555555555555556e-07,
+      "loss": 0.6399,
+      "step": 56
+    },
+    {
+      "epoch": 8.76923076923077,
+      "grad_norm": 4.763670444488525,
+      "learning_rate": 1.583333333333333e-07,
+      "loss": 0.6176,
+      "step": 57
+    },
+    {
+      "epoch": 8.923076923076923,
+      "grad_norm": 7.8895063400268555,
+      "learning_rate": 1.611111111111111e-07,
+      "loss": 0.6226,
+      "step": 58
+    },
+    {
+      "epoch": 8.923076923076923,
+      "eval_accuracy": 0.5860805860805861,
+      "eval_loss": 0.6665313839912415,
+      "eval_runtime": 1.1415,
+      "eval_samples_per_second": 239.164,
+      "eval_steps_per_second": 4.38,
+      "step": 58
+    },
+    {
+      "epoch": 9.076923076923077,
+      "grad_norm": 5.872027397155762,
+      "learning_rate": 1.6388888888888888e-07,
+      "loss": 0.6097,
+      "step": 59
+    },
+    {
+      "epoch": 9.23076923076923,
+      "grad_norm": 4.506241798400879,
+      "learning_rate": 1.6666666666666665e-07,
+      "loss": 0.6481,
+      "step": 60
+    },
+    {
+      "epoch": 9.384615384615385,
+      "grad_norm": 6.888943672180176,
+      "learning_rate": 1.6944444444444445e-07,
+      "loss": 0.6135,
+      "step": 61
+    },
+    {
+      "epoch": 9.538461538461538,
+      "grad_norm": 5.079667568206787,
+      "learning_rate": 1.7222222222222222e-07,
+      "loss": 0.6268,
+      "step": 62
+    },
+    {
+      "epoch": 9.692307692307692,
+      "grad_norm": 5.653709411621094,
+      "learning_rate": 1.75e-07,
+      "loss": 0.6259,
+      "step": 63
+    },
+    {
+      "epoch": 9.846153846153847,
+      "grad_norm": 5.669986724853516,
+      "learning_rate": 1.7777777777777776e-07,
+      "loss": 0.6195,
+      "step": 64
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 5.195782661437988,
+      "learning_rate": 1.8055555555555554e-07,
+      "loss": 0.5989,
+      "step": 65
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.6153846153846154,
+      "eval_loss": 0.6603101491928101,
+      "eval_runtime": 1.1943,
+      "eval_samples_per_second": 228.592,
+      "eval_steps_per_second": 4.187,
+      "step": 65
+    },
+    {
+      "epoch": 10.153846153846153,
+      "grad_norm": 5.164191722869873,
+      "learning_rate": 1.833333333333333e-07,
+      "loss": 0.6329,
+      "step": 66
+    },
+    {
+      "epoch": 10.307692307692308,
+      "grad_norm": 5.711167812347412,
+      "learning_rate": 1.861111111111111e-07,
+      "loss": 0.6198,
+      "step": 67
+    },
+    {
+      "epoch": 10.461538461538462,
+      "grad_norm": 5.415841102600098,
+      "learning_rate": 1.8888888888888888e-07,
+      "loss": 0.6162,
+      "step": 68
+    },
+    {
+      "epoch": 10.615384615384615,
+      "grad_norm": 5.584977626800537,
+      "learning_rate": 1.9166666666666668e-07,
+      "loss": 0.626,
+      "step": 69
+    },
+    {
+      "epoch": 10.76923076923077,
+      "grad_norm": 4.149146556854248,
+      "learning_rate": 1.9444444444444445e-07,
+      "loss": 0.6246,
+      "step": 70
+    },
+    {
+      "epoch": 10.923076923076923,
+      "grad_norm": 8.436751365661621,
+      "learning_rate": 1.9722222222222222e-07,
+      "loss": 0.5754,
+      "step": 71
+    },
+    {
+      "epoch": 10.923076923076923,
+      "eval_accuracy": 0.6263736263736264,
+      "eval_loss": 0.6555379629135132,
+      "eval_runtime": 1.1482,
+      "eval_samples_per_second": 237.759,
+      "eval_steps_per_second": 4.355,
+      "step": 71
+    },
+    {
+      "epoch": 11.076923076923077,
+      "grad_norm": 5.09835147857666,
+      "learning_rate": 2e-07,
+      "loss": 0.6153,
+      "step": 72
+    },
+    {
+      "epoch": 11.23076923076923,
+      "grad_norm": 6.399817943572998,
+      "learning_rate": 2.0277777777777776e-07,
+      "loss": 0.6001,
+      "step": 73
+    },
+    {
+      "epoch": 11.384615384615385,
+      "grad_norm": 4.717789173126221,
+      "learning_rate": 2.0555555555555553e-07,
+      "loss": 0.6009,
+      "step": 74
+    },
+    {
+      "epoch": 11.538461538461538,
+      "grad_norm": 5.399771213531494,
+      "learning_rate": 2.0833333333333333e-07,
+      "loss": 0.5998,
+      "step": 75
+    },
+    {
+      "epoch": 11.692307692307692,
+      "grad_norm": 5.429381370544434,
+      "learning_rate": 2.111111111111111e-07,
+      "loss": 0.5892,
+      "step": 76
+    },
+    {
+      "epoch": 11.846153846153847,
+      "grad_norm": 4.105190277099609,
+      "learning_rate": 2.1388888888888888e-07,
+      "loss": 0.6111,
+      "step": 77
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 6.455173969268799,
+      "learning_rate": 2.1666666666666667e-07,
+      "loss": 0.6251,
+      "step": 78
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.6483516483516484,
+      "eval_loss": 0.6492875218391418,
+      "eval_runtime": 1.1131,
+      "eval_samples_per_second": 245.251,
+      "eval_steps_per_second": 4.492,
+      "step": 78
+    },
+    {
+      "epoch": 12.153846153846153,
+      "grad_norm": 6.023905277252197,
+      "learning_rate": 2.1944444444444442e-07,
+      "loss": 0.6211,
+      "step": 79
+    },
+    {
+      "epoch": 12.307692307692308,
+      "grad_norm": 4.803109645843506,
+      "learning_rate": 2.222222222222222e-07,
+      "loss": 0.5972,
+      "step": 80
+    },
+    {
+      "epoch": 12.461538461538462,
+      "grad_norm": 4.598735332489014,
+      "learning_rate": 2.25e-07,
+      "loss": 0.5978,
+      "step": 81
+    },
+    {
+      "epoch": 12.615384615384615,
+      "grad_norm": 5.137476444244385,
+      "learning_rate": 2.2777777777777776e-07,
+      "loss": 0.5878,
+      "step": 82
+    },
+    {
+      "epoch": 12.76923076923077,
+      "grad_norm": 5.255553245544434,
+      "learning_rate": 2.3055555555555556e-07,
+      "loss": 0.59,
+      "step": 83
+    },
+    {
+      "epoch": 12.923076923076923,
+      "grad_norm": 4.83677864074707,
+      "learning_rate": 2.3333333333333333e-07,
+      "loss": 0.5796,
+      "step": 84
+    },
+    {
+      "epoch": 12.923076923076923,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 0.6446050405502319,
+      "eval_runtime": 1.1065,
+      "eval_samples_per_second": 246.726,
+      "eval_steps_per_second": 4.519,
+      "step": 84
+    },
+    {
+      "epoch": 13.076923076923077,
+      "grad_norm": 5.2046403884887695,
+      "learning_rate": 2.361111111111111e-07,
+      "loss": 0.5808,
+      "step": 85
+    },
+    {
+      "epoch": 13.23076923076923,
+      "grad_norm": 5.6977057456970215,
+      "learning_rate": 2.388888888888889e-07,
+      "loss": 0.5921,
+      "step": 86
+    },
+    {
+      "epoch": 13.384615384615385,
+      "grad_norm": 4.75449800491333,
+      "learning_rate": 2.4166666666666665e-07,
+      "loss": 0.576,
+      "step": 87
+    },
+    {
+      "epoch": 13.538461538461538,
+      "grad_norm": 4.761056423187256,
+      "learning_rate": 2.4444444444444445e-07,
+      "loss": 0.6126,
+      "step": 88
+    },
+    {
+      "epoch": 13.692307692307692,
+      "grad_norm": 4.913057327270508,
+      "learning_rate": 2.4722222222222224e-07,
+      "loss": 0.5814,
+      "step": 89
+    },
+    {
+      "epoch": 13.846153846153847,
+      "grad_norm": 7.290613651275635,
+      "learning_rate": 2.5e-07,
+      "loss": 0.5803,
+      "step": 90
+    },
+    {
+      "epoch": 14.0,
+      "grad_norm": 6.033799171447754,
+      "learning_rate": 2.5277777777777773e-07,
+      "loss": 0.5763,
+      "step": 91
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.6666666666666666,
+      "eval_loss": 0.6390318274497986,
+      "eval_runtime": 1.1176,
+      "eval_samples_per_second": 244.273,
+      "eval_steps_per_second": 4.474,
+      "step": 91
+    },
+    {
+      "epoch": 14.153846153846153,
+      "grad_norm": 5.3400678634643555,
+      "learning_rate": 2.5555555555555553e-07,
+      "loss": 0.5817,
+      "step": 92
+    },
+    {
+      "epoch": 14.307692307692308,
+      "grad_norm": 3.9350779056549072,
+      "learning_rate": 2.5833333333333333e-07,
+      "loss": 0.588,
+      "step": 93
+    },
+    {
+      "epoch": 14.461538461538462,
+      "grad_norm": 4.339548110961914,
+      "learning_rate": 2.6111111111111113e-07,
+      "loss": 0.5964,
+      "step": 94
+    },
+    {
+      "epoch": 14.615384615384615,
+      "grad_norm": 5.889684677124023,
+      "learning_rate": 2.638888888888889e-07,
+      "loss": 0.5636,
+      "step": 95
+    },
+    {
+      "epoch": 14.76923076923077,
+      "grad_norm": 4.385285377502441,
+      "learning_rate": 2.6666666666666667e-07,
+      "loss": 0.5898,
+      "step": 96
+    },
+    {
+      "epoch": 14.923076923076923,
+      "grad_norm": 7.42651891708374,
+      "learning_rate": 2.694444444444444e-07,
+      "loss": 0.5952,
+      "step": 97
+    },
+    {
+      "epoch": 14.923076923076923,
+      "eval_accuracy": 0.684981684981685,
+      "eval_loss": 0.6333425045013428,
+      "eval_runtime": 1.0931,
+      "eval_samples_per_second": 249.747,
+      "eval_steps_per_second": 4.574,
+      "step": 97
+    },
+    {
+      "epoch": 15.076923076923077,
+      "grad_norm": 7.099740028381348,
+      "learning_rate": 2.7222222222222216e-07,
+      "loss": 0.5412,
+      "step": 98
+    },
+    {
+      "epoch": 15.23076923076923,
+      "grad_norm": 5.428436279296875,
+      "learning_rate": 2.75e-07,
+      "loss": 0.5645,
+      "step": 99
+    },
+    {
+      "epoch": 15.384615384615385,
+      "grad_norm": 5.187060832977295,
+      "learning_rate": 2.7777777777777776e-07,
+      "loss": 0.5795,
+      "step": 100
+    },
+    {
+      "epoch": 15.538461538461538,
+      "grad_norm": 4.902520656585693,
+      "learning_rate": 2.8055555555555556e-07,
+      "loss": 0.5459,
+      "step": 101
+    },
+    {
+      "epoch": 15.692307692307692,
+      "grad_norm": 4.782113075256348,
+      "learning_rate": 2.833333333333333e-07,
+      "loss": 0.577,
+      "step": 102
+    },
+    {
+      "epoch": 15.846153846153847,
+      "grad_norm": 6.622628688812256,
+      "learning_rate": 2.861111111111111e-07,
+      "loss": 0.5583,
+      "step": 103
+    },
+    {
+      "epoch": 16.0,
+      "grad_norm": 4.653256416320801,
+      "learning_rate": 2.8888888888888885e-07,
+      "loss": 0.5675,
+      "step": 104
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.7032967032967034,
+      "eval_loss": 0.6269450783729553,
+      "eval_runtime": 1.2331,
+      "eval_samples_per_second": 221.389,
+      "eval_steps_per_second": 4.055,
+      "step": 104
+    },
+    {
+      "epoch": 16.153846153846153,
+      "grad_norm": 5.335786819458008,
+      "learning_rate": 2.916666666666667e-07,
+      "loss": 0.5492,
+      "step": 105
+    },
+    {
+      "epoch": 16.307692307692307,
+      "grad_norm": 6.031639575958252,
+      "learning_rate": 2.9444444444444444e-07,
+      "loss": 0.5475,
+      "step": 106
+    },
+    {
+      "epoch": 16.46153846153846,
+      "grad_norm": 8.427881240844727,
+      "learning_rate": 2.972222222222222e-07,
+      "loss": 0.5995,
+      "step": 107
+    },
+    {
+      "epoch": 16.615384615384617,
+      "grad_norm": 4.41181755065918,
+      "learning_rate": 3e-07,
+      "loss": 0.5458,
+      "step": 108
+    },
+    {
+      "epoch": 16.76923076923077,
+      "grad_norm": 5.0706257820129395,
+      "learning_rate": 3.0277777777777773e-07,
+      "loss": 0.5739,
+      "step": 109
+    },
+    {
+      "epoch": 16.923076923076923,
+      "grad_norm": 5.834512233734131,
+      "learning_rate": 3.055555555555556e-07,
+      "loss": 0.5453,
+      "step": 110
+    },
+    {
+      "epoch": 16.923076923076923,
+      "eval_accuracy": 0.7106227106227107,
+      "eval_loss": 0.6210848689079285,
+      "eval_runtime": 1.1361,
+      "eval_samples_per_second": 240.302,
+      "eval_steps_per_second": 4.401,
+      "step": 110
+    },
+    {
+      "epoch": 17.076923076923077,
+      "grad_norm": 4.99503231048584,
+      "learning_rate": 3.0833333333333333e-07,
+      "loss": 0.5517,
+      "step": 111
+    },
+    {
+      "epoch": 17.23076923076923,
+      "grad_norm": 4.0909833908081055,
+      "learning_rate": 3.111111111111111e-07,
+      "loss": 0.5664,
+      "step": 112
+    },
+    {
+      "epoch": 17.384615384615383,
+      "grad_norm": 4.5490851402282715,
+      "learning_rate": 3.1388888888888887e-07,
+      "loss": 0.541,
+      "step": 113
+    },
+    {
+      "epoch": 17.53846153846154,
+      "grad_norm": 4.016046524047852,
+      "learning_rate": 3.166666666666666e-07,
+      "loss": 0.5788,
+      "step": 114
+    },
+    {
+      "epoch": 17.692307692307693,
+      "grad_norm": 5.587233066558838,
+      "learning_rate": 3.194444444444444e-07,
+      "loss": 0.5155,
+      "step": 115
+    },
+    {
+      "epoch": 17.846153846153847,
+      "grad_norm": 5.020046234130859,
+      "learning_rate": 3.222222222222222e-07,
+      "loss": 0.5386,
+      "step": 116
+    },
+    {
+      "epoch": 18.0,
+      "grad_norm": 6.417337417602539,
+      "learning_rate": 3.25e-07,
+      "loss": 0.5199,
+      "step": 117
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.7142857142857143,
+      "eval_loss": 0.6150110363960266,
+      "eval_runtime": 1.1307,
+      "eval_samples_per_second": 241.446,
+      "eval_steps_per_second": 4.422,
+      "step": 117
+    },
+    {
+      "epoch": 18.153846153846153,
+      "grad_norm": 5.794677257537842,
+      "learning_rate": 3.2777777777777776e-07,
+      "loss": 0.5343,
+      "step": 118
+    },
+    {
+      "epoch": 18.307692307692307,
+      "grad_norm": 4.628763198852539,
+      "learning_rate": 3.3055555555555556e-07,
+      "loss": 0.5275,
+      "step": 119
+    },
+    {
+      "epoch": 18.46153846153846,
+      "grad_norm": 7.958808422088623,
+      "learning_rate": 3.333333333333333e-07,
+      "loss": 0.5252,
+      "step": 120
+    },
+    {
+      "epoch": 18.615384615384617,
+      "grad_norm": 4.4327826499938965,
+      "learning_rate": 3.361111111111111e-07,
+      "loss": 0.5233,
+      "step": 121
+    },
+    {
+      "epoch": 18.76923076923077,
+      "grad_norm": 5.422006130218506,
+      "learning_rate": 3.388888888888889e-07,
+      "loss": 0.5646,
+      "step": 122
+    },
+    {
+      "epoch": 18.923076923076923,
+      "grad_norm": 4.5244975090026855,
+      "learning_rate": 3.4166666666666664e-07,
+      "loss": 0.541,
+      "step": 123
+    },
+    {
+      "epoch": 18.923076923076923,
+      "eval_accuracy": 0.7216117216117216,
+      "eval_loss": 0.6090343594551086,
+      "eval_runtime": 1.2237,
+      "eval_samples_per_second": 223.1,
+      "eval_steps_per_second": 4.086,
+      "step": 123
+    },
+    {
+      "epoch": 19.076923076923077,
+      "grad_norm": 6.068373680114746,
+      "learning_rate": 3.4444444444444444e-07,
+      "loss": 0.5179,
+      "step": 124
+    },
+    {
+      "epoch": 19.23076923076923,
+      "grad_norm": 4.209527492523193,
+      "learning_rate": 3.472222222222222e-07,
+      "loss": 0.5533,
+      "step": 125
+    },
+    {
+      "epoch": 19.384615384615383,
+      "grad_norm": 5.68998384475708,
+      "learning_rate": 3.5e-07,
+      "loss": 0.5219,
+      "step": 126
+    },
+    {
+      "epoch": 19.53846153846154,
+      "grad_norm": 5.829248428344727,
+      "learning_rate": 3.527777777777778e-07,
+      "loss": 0.503,
+      "step": 127
+    },
+    {
+      "epoch": 19.692307692307693,
+      "grad_norm": 4.40165376663208,
+      "learning_rate": 3.5555555555555553e-07,
+      "loss": 0.5483,
+      "step": 128
+    },
+    {
+      "epoch": 19.846153846153847,
+      "grad_norm": 4.628623962402344,
+      "learning_rate": 3.583333333333333e-07,
+      "loss": 0.5249,
+      "step": 129
+    },
+    {
+      "epoch": 20.0,
+      "grad_norm": 4.771780014038086,
+      "learning_rate": 3.6111111111111107e-07,
+      "loss": 0.5273,
+      "step": 130
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.7289377289377289,
+      "eval_loss": 0.6007034182548523,
+      "eval_runtime": 1.1281,
+      "eval_samples_per_second": 241.989,
+      "eval_steps_per_second": 4.432,
+      "step": 130
+    },
+    {
+      "epoch": 20.153846153846153,
+      "grad_norm": 4.505945682525635,
+      "learning_rate": 3.6388888888888887e-07,
+      "loss": 0.515,
+      "step": 131
+    },
+    {
+      "epoch": 20.307692307692307,
+      "grad_norm": 7.482431411743164,
+      "learning_rate": 3.666666666666666e-07,
+      "loss": 0.4948,
+      "step": 132
+    },
+    {
+      "epoch": 20.46153846153846,
+      "grad_norm": 4.583889007568359,
+      "learning_rate": 3.6944444444444447e-07,
+      "loss": 0.5215,
+      "step": 133
+    },
+    {
+      "epoch": 20.615384615384617,
+      "grad_norm": 5.050055980682373,
+      "learning_rate": 3.722222222222222e-07,
+      "loss": 0.502,
+      "step": 134
+    },
+    {
+      "epoch": 20.76923076923077,
+      "grad_norm": 5.0981245040893555,
+      "learning_rate": 3.75e-07,
+      "loss": 0.5343,
+      "step": 135
+    },
+    {
+      "epoch": 20.923076923076923,
+      "grad_norm": 7.6524529457092285,
+      "learning_rate": 3.7777777777777775e-07,
+      "loss": 0.495,
+      "step": 136
+    },
+    {
+      "epoch": 20.923076923076923,
+      "eval_accuracy": 0.7289377289377289,
+      "eval_loss": 0.5934087634086609,
+      "eval_runtime": 1.1036,
+      "eval_samples_per_second": 247.368,
+      "eval_steps_per_second": 4.531,
+      "step": 136
+    },
+    {
+      "epoch": 21.076923076923077,
+      "grad_norm": 5.747613906860352,
+      "learning_rate": 3.805555555555555e-07,
+      "loss": 0.4929,
+      "step": 137
+    },
+    {
+      "epoch": 21.23076923076923,
+      "grad_norm": 8.28075122833252,
+      "learning_rate": 3.8333333333333335e-07,
+      "loss": 0.5302,
+      "step": 138
+    },
+    {
+      "epoch": 21.384615384615383,
+      "grad_norm": 5.376802921295166,
+      "learning_rate": 3.861111111111111e-07,
+      "loss": 0.547,
+      "step": 139
+    },
+    {
+      "epoch": 21.53846153846154,
+      "grad_norm": 4.9347968101501465,
+      "learning_rate": 3.888888888888889e-07,
+      "loss": 0.4807,
+      "step": 140
+    },
+    {
+      "epoch": 21.692307692307693,
+      "grad_norm": 6.574237823486328,
+      "learning_rate": 3.9166666666666664e-07,
+      "loss": 0.4839,
+      "step": 141
+    },
+    {
+      "epoch": 21.846153846153847,
+      "grad_norm": 4.990074157714844,
+      "learning_rate": 3.9444444444444444e-07,
+      "loss": 0.4937,
+      "step": 142
+    },
+    {
+      "epoch": 22.0,
+      "grad_norm": 6.261688232421875,
+      "learning_rate": 3.972222222222222e-07,
+      "loss": 0.4855,
+      "step": 143
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.7472527472527473,
+      "eval_loss": 0.5855301022529602,
+      "eval_runtime": 1.1327,
+      "eval_samples_per_second": 241.009,
+      "eval_steps_per_second": 4.414,
+      "step": 143
+    },
+    {
+      "epoch": 22.153846153846153,
+      "grad_norm": 5.447592258453369,
+      "learning_rate": 4e-07,
+      "loss": 0.4225,
+      "step": 144
+    },
+    {
+      "epoch": 22.307692307692307,
+      "grad_norm": 4.675624847412109,
+      "learning_rate": 4.027777777777778e-07,
+      "loss": 0.4982,
+      "step": 145
+    },
+    {
+      "epoch": 22.46153846153846,
+      "grad_norm": 4.06153678894043,
+      "learning_rate": 4.055555555555555e-07,
+      "loss": 0.4751,
+      "step": 146
+    },
+    {
+      "epoch": 22.615384615384617,
+      "grad_norm": 7.109158039093018,
+      "learning_rate": 4.083333333333333e-07,
+      "loss": 0.4232,
+      "step": 147
+    },
+    {
+      "epoch": 22.76923076923077,
+      "grad_norm": 7.506172180175781,
+      "learning_rate": 4.1111111111111107e-07,
+      "loss": 0.4935,
+      "step": 148
+    },
+    {
+      "epoch": 22.923076923076923,
+      "grad_norm": 4.346311092376709,
+      "learning_rate": 4.1388888888888887e-07,
+      "loss": 0.4763,
+      "step": 149
+    },
+    {
+      "epoch": 22.923076923076923,
+      "eval_accuracy": 0.7362637362637363,
+      "eval_loss": 0.5787296295166016,
+      "eval_runtime": 1.163,
+      "eval_samples_per_second": 234.74,
+      "eval_steps_per_second": 4.299,
+      "step": 149
+    },
+    {
+      "epoch": 23.076923076923077,
+      "grad_norm": 7.535876750946045,
+      "learning_rate": 4.1666666666666667e-07,
+      "loss": 0.4539,
+      "step": 150
+    },
+    {
+      "epoch": 23.23076923076923,
+      "grad_norm": 4.603224277496338,
+      "learning_rate": 4.194444444444444e-07,
+      "loss": 0.491,
+      "step": 151
+    },
+    {
+      "epoch": 23.384615384615383,
+      "grad_norm": 4.099734783172607,
+      "learning_rate": 4.222222222222222e-07,
+      "loss": 0.4877,
+      "step": 152
+    },
+    {
+      "epoch": 23.53846153846154,
+      "grad_norm": 6.270420551300049,
+      "learning_rate": 4.2499999999999995e-07,
+      "loss": 0.4791,
+      "step": 153
+    },
+    {
+      "epoch": 23.692307692307693,
+      "grad_norm": 5.5756330490112305,
+      "learning_rate": 4.2777777777777775e-07,
+      "loss": 0.4192,
+      "step": 154
+    },
+    {
+      "epoch": 23.846153846153847,
+      "grad_norm": 4.492015838623047,
+      "learning_rate": 4.3055555555555555e-07,
+      "loss": 0.429,
+      "step": 155
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 5.119944095611572,
+      "learning_rate": 4.3333333333333335e-07,
+      "loss": 0.4287,
+      "step": 156
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.7509157509157509,
+      "eval_loss": 0.5693350434303284,
+      "eval_runtime": 1.2614,
+      "eval_samples_per_second": 216.434,
+      "eval_steps_per_second": 3.964,
+      "step": 156
+    },
+    {
+      "epoch": 24.153846153846153,
+      "grad_norm": 6.68319034576416,
+      "learning_rate": 4.361111111111111e-07,
+      "loss": 0.4549,
+      "step": 157
+    },
+    {
+      "epoch": 24.307692307692307,
+      "grad_norm": 6.93166446685791,
+      "learning_rate": 4.3888888888888884e-07,
+      "loss": 0.4878,
+      "step": 158
+    },
+    {
+      "epoch": 24.46153846153846,
+      "grad_norm": 4.761252403259277,
+      "learning_rate": 4.4166666666666664e-07,
+      "loss": 0.463,
+      "step": 159
+    },
+    {
+      "epoch": 24.615384615384617,
+      "grad_norm": 5.771098613739014,
+      "learning_rate": 4.444444444444444e-07,
+      "loss": 0.4418,
+      "step": 160
+    },
+    {
+      "epoch": 24.76923076923077,
+      "grad_norm": 5.692720413208008,
+      "learning_rate": 4.4722222222222223e-07,
+      "loss": 0.4211,
+      "step": 161
+    },
+    {
+      "epoch": 24.923076923076923,
+      "grad_norm": 10.276296615600586,
+      "learning_rate": 4.5e-07,
+      "loss": 0.445,
+      "step": 162
+    },
+    {
+      "epoch": 24.923076923076923,
+      "eval_accuracy": 0.7692307692307693,
+      "eval_loss": 0.5618556141853333,
+      "eval_runtime": 1.2168,
+      "eval_samples_per_second": 224.355,
+      "eval_steps_per_second": 4.109,
+      "step": 162
+    },
+    {
+      "epoch": 25.076923076923077,
+      "grad_norm": 6.790480613708496,
+      "learning_rate": 4.527777777777778e-07,
+      "loss": 0.4313,
+      "step": 163
+    },
+    {
+      "epoch": 25.23076923076923,
+      "grad_norm": 4.737351417541504,
+      "learning_rate": 4.555555555555555e-07,
+      "loss": 0.4303,
+      "step": 164
+    },
+    {
+      "epoch": 25.384615384615383,
+      "grad_norm": 4.519160747528076,
+      "learning_rate": 4.5833333333333327e-07,
+      "loss": 0.455,
+      "step": 165
+    },
+    {
+      "epoch": 25.53846153846154,
+      "grad_norm": 5.933927536010742,
+      "learning_rate": 4.611111111111111e-07,
+      "loss": 0.4461,
+      "step": 166
+    },
+    {
+      "epoch": 25.692307692307693,
+      "grad_norm": 7.9362664222717285,
+      "learning_rate": 4.6388888888888886e-07,
+      "loss": 0.3951,
+      "step": 167
+    },
+    {
+      "epoch": 25.846153846153847,
+      "grad_norm": 9.995780944824219,
+      "learning_rate": 4.6666666666666666e-07,
+      "loss": 0.4142,
+      "step": 168
+    },
+    {
+      "epoch": 26.0,
+      "grad_norm": 6.652438163757324,
+      "learning_rate": 4.694444444444444e-07,
+      "loss": 0.4343,
+      "step": 169
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.7802197802197802,
+      "eval_loss": 0.5539770126342773,
+      "eval_runtime": 1.1366,
+      "eval_samples_per_second": 240.189,
+      "eval_steps_per_second": 4.399,
+      "step": 169
+    },
+    {
+      "epoch": 26.153846153846153,
+      "grad_norm": 4.573406219482422,
+      "learning_rate": 4.722222222222222e-07,
+      "loss": 0.4279,
+      "step": 170
+    },
+    {
+      "epoch": 26.307692307692307,
+      "grad_norm": 6.939181327819824,
+      "learning_rate": 4.7499999999999995e-07,
+      "loss": 0.443,
+      "step": 171
+    },
+    {
+      "epoch": 26.46153846153846,
+      "grad_norm": 10.387896537780762,
+      "learning_rate": 4.777777777777778e-07,
+      "loss": 0.4462,
+      "step": 172
+    },
+    {
+      "epoch": 26.615384615384617,
+      "grad_norm": 6.387257099151611,
+      "learning_rate": 4.805555555555555e-07,
+      "loss": 0.4221,
+      "step": 173
+    },
+    {
+      "epoch": 26.76923076923077,
+      "grad_norm": 5.768893241882324,
+      "learning_rate": 4.833333333333333e-07,
+      "loss": 0.4048,
+      "step": 174
+    },
+    {
+      "epoch": 26.923076923076923,
+      "grad_norm": 6.23684024810791,
+      "learning_rate": 4.861111111111111e-07,
+      "loss": 0.3748,
+      "step": 175
+    },
+    {
+      "epoch": 26.923076923076923,
+      "eval_accuracy": 0.7875457875457875,
+      "eval_loss": 0.5466815829277039,
+      "eval_runtime": 1.1181,
+      "eval_samples_per_second": 244.171,
+      "eval_steps_per_second": 4.472,
+      "step": 175
+    },
+    {
+      "epoch": 27.076923076923077,
+      "grad_norm": 6.4479146003723145,
+      "learning_rate": 4.888888888888889e-07,
+      "loss": 0.3923,
+      "step": 176
+    },
+    {
+      "epoch": 27.23076923076923,
+      "grad_norm": 8.554872512817383,
+      "learning_rate": 4.916666666666666e-07,
+      "loss": 0.4114,
+      "step": 177
+    },
+    {
+      "epoch": 27.384615384615383,
+      "grad_norm": 6.017930030822754,
+      "learning_rate": 4.944444444444445e-07,
+      "loss": 0.3798,
+      "step": 178
+    },
+    {
+      "epoch": 27.53846153846154,
+      "grad_norm": 8.284281730651855,
+      "learning_rate": 4.972222222222222e-07,
+      "loss": 0.4146,
+      "step": 179
+    },
+    {
+      "epoch": 27.692307692307693,
+      "grad_norm": 9.119588851928711,
+      "learning_rate": 5e-07,
+      "loss": 0.4436,
+      "step": 180
+    },
+    {
+      "epoch": 27.846153846153847,
+      "grad_norm": 9.632242202758789,
+      "learning_rate": 4.988095238095238e-07,
+      "loss": 0.4031,
+      "step": 181
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 16.935251235961914,
+      "learning_rate": 4.976190476190476e-07,
+      "loss": 0.4041,
+      "step": 182
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.8021978021978022,
+      "eval_loss": 0.5421282052993774,
+      "eval_runtime": 1.1482,
+      "eval_samples_per_second": 237.764,
+      "eval_steps_per_second": 4.355,
+      "step": 182
+    },
+    {
+      "epoch": 28.153846153846153,
+      "grad_norm": 5.241950511932373,
+      "learning_rate": 4.964285714285715e-07,
+      "loss": 0.4528,
+      "step": 183
+    },
+    {
+      "epoch": 28.307692307692307,
+      "grad_norm": 11.666001319885254,
+      "learning_rate": 4.952380952380952e-07,
+      "loss": 0.3889,
+      "step": 184
+    },
+    {
+      "epoch": 28.46153846153846,
+      "grad_norm": 9.493650436401367,
+      "learning_rate": 4.94047619047619e-07,
+      "loss": 0.3956,
+      "step": 185
+    },
+    {
+      "epoch": 28.615384615384617,
+      "grad_norm": 11.480545043945312,
+      "learning_rate": 4.928571428571429e-07,
+      "loss": 0.3756,
+      "step": 186
+    },
+    {
+      "epoch": 28.76923076923077,
+      "grad_norm": 18.468042373657227,
+      "learning_rate": 4.916666666666666e-07,
+      "loss": 0.401,
+      "step": 187
+    },
+    {
+      "epoch": 28.923076923076923,
+      "grad_norm": 8.127854347229004,
+      "learning_rate": 4.904761904761904e-07,
+      "loss": 0.3543,
+      "step": 188
+    },
+    {
+      "epoch": 28.923076923076923,
+      "eval_accuracy": 0.8205128205128205,
+      "eval_loss": 0.529083788394928,
+      "eval_runtime": 1.1719,
+      "eval_samples_per_second": 232.959,
+      "eval_steps_per_second": 4.267,
+      "step": 188
+    },
+    {
+      "epoch": 29.076923076923077,
+      "grad_norm": 12.729475975036621,
+      "learning_rate": 4.892857142857142e-07,
+      "loss": 0.4255,
+      "step": 189
+    },
+    {
+      "epoch": 29.23076923076923,
+      "grad_norm": 6.080268383026123,
+      "learning_rate": 4.880952380952381e-07,
+      "loss": 0.4233,
+      "step": 190
+    },
+    {
+      "epoch": 29.384615384615383,
+      "grad_norm": 9.651863098144531,
+      "learning_rate": 4.869047619047619e-07,
+      "loss": 0.3753,
+      "step": 191
+    },
+    {
+      "epoch": 29.53846153846154,
+      "grad_norm": 12.342060089111328,
+      "learning_rate": 4.857142857142857e-07,
+      "loss": 0.392,
+      "step": 192
+    },
+    {
+      "epoch": 29.692307692307693,
+      "grad_norm": 5.0623779296875,
+      "learning_rate": 4.845238095238095e-07,
+      "loss": 0.3942,
+      "step": 193
+    },
+    {
+      "epoch": 29.846153846153847,
+      "grad_norm": 18.967161178588867,
+      "learning_rate": 4.833333333333333e-07,
+      "loss": 0.3374,
+      "step": 194
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 13.24194622039795,
+      "learning_rate": 4.821428571428571e-07,
+      "loss": 0.3972,
+      "step": 195
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.8278388278388278,
+      "eval_loss": 0.5134266018867493,
+      "eval_runtime": 1.2397,
+      "eval_samples_per_second": 220.215,
+      "eval_steps_per_second": 4.033,
+      "step": 195
+    },
+    {
+      "epoch": 30.153846153846153,
+      "grad_norm": 12.518712997436523,
+      "learning_rate": 4.80952380952381e-07,
+      "loss": 0.4197,
+      "step": 196
+    },
+    {
+      "epoch": 30.307692307692307,
+      "grad_norm": 15.362881660461426,
+      "learning_rate": 4.797619047619048e-07,
+      "loss": 0.4056,
+      "step": 197
+    },
+    {
+      "epoch": 30.46153846153846,
+      "grad_norm": 17.072725296020508,
+      "learning_rate": 4.785714285714286e-07,
+      "loss": 0.3856,
+      "step": 198
+    },
+    {
+      "epoch": 30.615384615384617,
+      "grad_norm": 6.637291431427002,
+      "learning_rate": 4.773809523809523e-07,
+      "loss": 0.3627,
+      "step": 199
+    },
+    {
+      "epoch": 30.76923076923077,
+      "grad_norm": 8.751256942749023,
+      "learning_rate": 4.761904761904761e-07,
+      "loss": 0.3637,
+      "step": 200
+    },
+    {
+      "epoch": 30.923076923076923,
+      "grad_norm": 5.588964462280273,
+      "learning_rate": 4.7499999999999995e-07,
+      "loss": 0.3716,
+      "step": 201
+    },
+    {
+      "epoch": 30.923076923076923,
+      "eval_accuracy": 0.8241758241758241,
+      "eval_loss": 0.5149514079093933,
+      "eval_runtime": 1.1326,
+      "eval_samples_per_second": 241.045,
+      "eval_steps_per_second": 4.415,
+      "step": 201
+    },
+    {
+      "epoch": 31.076923076923077,
+      "grad_norm": 8.8916654586792,
+      "learning_rate": 4.738095238095238e-07,
+      "loss": 0.362,
+      "step": 202
+    },
+    {
+      "epoch": 31.23076923076923,
+      "grad_norm": 7.172123908996582,
+      "learning_rate": 4.7261904761904756e-07,
+      "loss": 0.3541,
+      "step": 203
+    },
+    {
+      "epoch": 31.384615384615383,
+      "grad_norm": 9.399160385131836,
+      "learning_rate": 4.714285714285714e-07,
+      "loss": 0.3629,
+      "step": 204
+    },
+    {
+      "epoch": 31.53846153846154,
+      "grad_norm": 12.05125904083252,
+      "learning_rate": 4.702380952380952e-07,
+      "loss": 0.3414,
+      "step": 205
+    },
+    {
+      "epoch": 31.692307692307693,
+      "grad_norm": 6.808493137359619,
+      "learning_rate": 4.69047619047619e-07,
+      "loss": 0.372,
+      "step": 206
+    },
+    {
+      "epoch": 31.846153846153847,
+      "grad_norm": 12.759626388549805,
+      "learning_rate": 4.6785714285714283e-07,
+      "loss": 0.3574,
+      "step": 207
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 9.680743217468262,
+      "learning_rate": 4.6666666666666666e-07,
+      "loss": 0.3871,
+      "step": 208
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.8315018315018315,
+      "eval_loss": 0.5100430846214294,
+      "eval_runtime": 1.1335,
+      "eval_samples_per_second": 240.838,
+      "eval_steps_per_second": 4.411,
+      "step": 208
+    },
+    {
+      "epoch": 32.15384615384615,
+      "grad_norm": 13.226339340209961,
+      "learning_rate": 4.6547619047619044e-07,
+      "loss": 0.3623,
+      "step": 209
+    },
+    {
+      "epoch": 32.30769230769231,
+      "grad_norm": 9.359790802001953,
+      "learning_rate": 4.6428571428571427e-07,
+      "loss": 0.4045,
+      "step": 210
+    },
+    {
+      "epoch": 32.46153846153846,
+      "grad_norm": 4.744467258453369,
+      "learning_rate": 4.630952380952381e-07,
+      "loss": 0.3852,
+      "step": 211
+    },
+    {
+      "epoch": 32.61538461538461,
+      "grad_norm": 9.221460342407227,
+      "learning_rate": 4.6190476190476193e-07,
+      "loss": 0.3267,
+      "step": 212
+    },
+    {
+      "epoch": 32.76923076923077,
+      "grad_norm": 5.6791510581970215,
+      "learning_rate": 4.6071428571428566e-07,
+      "loss": 0.3969,
+      "step": 213
+    },
+    {
+      "epoch": 32.92307692307692,
+      "grad_norm": 13.916045188903809,
+      "learning_rate": 4.595238095238095e-07,
+      "loss": 0.3729,
+      "step": 214
+    },
+    {
+      "epoch": 32.92307692307692,
+      "eval_accuracy": 0.8351648351648352,
+      "eval_loss": 0.4985570013523102,
+      "eval_runtime": 1.1523,
+      "eval_samples_per_second": 236.921,
+      "eval_steps_per_second": 4.339,
+      "step": 214
+    },
+    {
+      "epoch": 33.07692307692308,
+      "grad_norm": 20.20587158203125,
+      "learning_rate": 4.5833333333333327e-07,
+      "loss": 0.3999,
+      "step": 215
+    },
+    {
+      "epoch": 33.23076923076923,
+      "grad_norm": 5.628478050231934,
+      "learning_rate": 4.571428571428571e-07,
+      "loss": 0.333,
+      "step": 216
+    },
+    {
+      "epoch": 33.38461538461539,
+      "grad_norm": 9.061274528503418,
+      "learning_rate": 4.5595238095238093e-07,
+      "loss": 0.3807,
+      "step": 217
+    },
+    {
+      "epoch": 33.53846153846154,
+      "grad_norm": 7.302306652069092,
+      "learning_rate": 4.5476190476190476e-07,
+      "loss": 0.3309,
+      "step": 218
+    },
+    {
+      "epoch": 33.69230769230769,
+      "grad_norm": 17.617534637451172,
+      "learning_rate": 4.5357142857142854e-07,
+      "loss": 0.319,
+      "step": 219
+    },
+    {
+      "epoch": 33.84615384615385,
+      "grad_norm": 9.831780433654785,
+      "learning_rate": 4.5238095238095237e-07,
+      "loss": 0.3849,
+      "step": 220
+    },
+    {
+      "epoch": 34.0,
+      "grad_norm": 10.667181968688965,
+      "learning_rate": 4.511904761904762e-07,
+      "loss": 0.3286,
+      "step": 221
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.8461538461538461,
+      "eval_loss": 0.4945792853832245,
+      "eval_runtime": 1.1795,
+      "eval_samples_per_second": 231.451,
+      "eval_steps_per_second": 4.239,
+      "step": 221
+    },
+    {
+      "epoch": 34.15384615384615,
+      "grad_norm": 6.82171106338501,
+      "learning_rate": 4.5e-07,
+      "loss": 0.3129,
+      "step": 222
+    },
+    {
+      "epoch": 34.30769230769231,
+      "grad_norm": 12.462568283081055,
+      "learning_rate": 4.488095238095238e-07,
+      "loss": 0.2947,
+      "step": 223
+    },
+    {
+      "epoch": 34.46153846153846,
+      "grad_norm": 14.821395874023438,
+      "learning_rate": 4.4761904761904764e-07,
+      "loss": 0.381,
+      "step": 224
+    },
+    {
+      "epoch": 34.61538461538461,
+      "grad_norm": 11.688921928405762,
+      "learning_rate": 4.464285714285714e-07,
+      "loss": 0.392,
+      "step": 225
+    },
+    {
+      "epoch": 34.76923076923077,
+      "grad_norm": 16.245473861694336,
+      "learning_rate": 4.452380952380952e-07,
+      "loss": 0.3356,
+      "step": 226
+    },
+    {
+      "epoch": 34.92307692307692,
+      "grad_norm": 8.85981559753418,
+      "learning_rate": 4.4404761904761903e-07,
+      "loss": 0.4261,
+      "step": 227
+    },
+    {
+      "epoch": 34.92307692307692,
+      "eval_accuracy": 0.8388278388278388,
+      "eval_loss": 0.49569690227508545,
+      "eval_runtime": 1.1615,
+      "eval_samples_per_second": 235.046,
+      "eval_steps_per_second": 4.305,
+      "step": 227
+    },
+    {
+      "epoch": 35.07692307692308,
+      "grad_norm": 14.444450378417969,
+      "learning_rate": 4.428571428571428e-07,
+      "loss": 0.3065,
+      "step": 228
+    },
+    {
+      "epoch": 35.23076923076923,
+      "grad_norm": 18.541534423828125,
+      "learning_rate": 4.4166666666666664e-07,
+      "loss": 0.319,
+      "step": 229
+    },
+    {
+      "epoch": 35.38461538461539,
+      "grad_norm": 14.79796314239502,
+      "learning_rate": 4.4047619047619047e-07,
+      "loss": 0.3511,
+      "step": 230
+    },
+    {
+      "epoch": 35.53846153846154,
+      "grad_norm": 14.025275230407715,
+      "learning_rate": 4.3928571428571425e-07,
+      "loss": 0.351,
+      "step": 231
+    },
+    {
+      "epoch": 35.69230769230769,
+      "grad_norm": 9.698802947998047,
+      "learning_rate": 4.380952380952381e-07,
+      "loss": 0.3369,
+      "step": 232
+    },
+    {
+      "epoch": 35.84615384615385,
+      "grad_norm": 10.56847858428955,
+      "learning_rate": 4.369047619047619e-07,
+      "loss": 0.2976,
+      "step": 233
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 15.778355598449707,
+      "learning_rate": 4.357142857142857e-07,
+      "loss": 0.4014,
+      "step": 234
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.8534798534798534,
+      "eval_loss": 0.48495998978614807,
+      "eval_runtime": 1.1929,
+      "eval_samples_per_second": 228.863,
+      "eval_steps_per_second": 4.192,
+      "step": 234
+    },
+    {
+      "epoch": 36.15384615384615,
+      "grad_norm": 16.71240234375,
+      "learning_rate": 4.345238095238095e-07,
+      "loss": 0.3749,
+      "step": 235
+    },
+    {
+      "epoch": 36.30769230769231,
+      "grad_norm": 13.584702491760254,
+      "learning_rate": 4.3333333333333335e-07,
+      "loss": 0.3812,
+      "step": 236
+    },
+    {
+      "epoch": 36.46153846153846,
+      "grad_norm": 8.108072280883789,
+      "learning_rate": 4.3214285714285713e-07,
+      "loss": 0.3024,
+      "step": 237
+    },
+    {
+      "epoch": 36.61538461538461,
+      "grad_norm": 9.233282089233398,
+      "learning_rate": 4.3095238095238096e-07,
+      "loss": 0.3413,
+      "step": 238
+    },
+    {
+      "epoch": 36.76923076923077,
+      "grad_norm": 13.080716133117676,
+      "learning_rate": 4.297619047619048e-07,
+      "loss": 0.2792,
+      "step": 239
+    },
+    {
+      "epoch": 36.92307692307692,
+      "grad_norm": 10.88381576538086,
+      "learning_rate": 4.285714285714285e-07,
+      "loss": 0.3514,
+      "step": 240
+    },
+    {
+      "epoch": 36.92307692307692,
+      "eval_accuracy": 0.8534798534798534,
+      "eval_loss": 0.4806550443172455,
+      "eval_runtime": 1.1462,
+      "eval_samples_per_second": 238.179,
+      "eval_steps_per_second": 4.362,
+      "step": 240
+    },
+    {
+      "epoch": 37.07692307692308,
+      "grad_norm": 13.345056533813477,
+      "learning_rate": 4.2738095238095235e-07,
+      "loss": 0.318,
+      "step": 241
+    },
+    {
+      "epoch": 37.23076923076923,
+      "grad_norm": 9.593791961669922,
+      "learning_rate": 4.261904761904762e-07,
+      "loss": 0.3487,
+      "step": 242
+    },
+    {
+      "epoch": 37.38461538461539,
+      "grad_norm": 9.83149242401123,
+      "learning_rate": 4.2499999999999995e-07,
+      "loss": 0.3283,
+      "step": 243
+    },
+    {
+      "epoch": 37.53846153846154,
+      "grad_norm": 11.976517677307129,
+      "learning_rate": 4.238095238095238e-07,
+      "loss": 0.407,
+      "step": 244
+    },
+    {
+      "epoch": 37.69230769230769,
+      "grad_norm": 7.24540901184082,
+      "learning_rate": 4.226190476190476e-07,
+      "loss": 0.3899,
+      "step": 245
+    },
+    {
+      "epoch": 37.84615384615385,
+      "grad_norm": 11.755511283874512,
+      "learning_rate": 4.214285714285714e-07,
+      "loss": 0.247,
+      "step": 246
+    },
+    {
+      "epoch": 38.0,
+      "grad_norm": 14.939422607421875,
+      "learning_rate": 4.202380952380952e-07,
+      "loss": 0.3883,
+      "step": 247
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.8534798534798534,
+      "eval_loss": 0.47668036818504333,
+      "eval_runtime": 1.2822,
+      "eval_samples_per_second": 212.923,
+      "eval_steps_per_second": 3.9,
+      "step": 247
+    },
+    {
+      "epoch": 38.15384615384615,
+      "grad_norm": 7.6719279289245605,
+      "learning_rate": 4.1904761904761906e-07,
+      "loss": 0.3579,
+      "step": 248
+    },
+    {
+      "epoch": 38.30769230769231,
+      "grad_norm": 18.015718460083008,
+      "learning_rate": 4.1785714285714283e-07,
+      "loss": 0.3072,
+      "step": 249
+    },
+    {
+      "epoch": 38.46153846153846,
+      "grad_norm": 13.246123313903809,
+      "learning_rate": 4.1666666666666667e-07,
+      "loss": 0.3756,
+      "step": 250
+    },
+    {
+      "epoch": 38.61538461538461,
+      "grad_norm": 7.806217670440674,
+      "learning_rate": 4.154761904761905e-07,
+      "loss": 0.3919,
+      "step": 251
+    },
+    {
+      "epoch": 38.76923076923077,
+      "grad_norm": 5.912841796875,
+      "learning_rate": 4.142857142857143e-07,
+      "loss": 0.3079,
+      "step": 252
+    },
+    {
+      "epoch": 38.92307692307692,
+      "grad_norm": 7.757283687591553,
+      "learning_rate": 4.1309523809523805e-07,
+      "loss": 0.3219,
+      "step": 253
+    },
+    {
+      "epoch": 38.92307692307692,
+      "eval_accuracy": 0.8534798534798534,
+      "eval_loss": 0.4762944281101227,
+      "eval_runtime": 1.2149,
+      "eval_samples_per_second": 224.702,
+      "eval_steps_per_second": 4.115,
+      "step": 253
+    },
+    {
+      "epoch": 39.07692307692308,
+      "grad_norm": 7.3901238441467285,
+      "learning_rate": 4.119047619047619e-07,
+      "loss": 0.2908,
+      "step": 254
+    },
+    {
+      "epoch": 39.23076923076923,
+      "grad_norm": 6.3056535720825195,
+      "learning_rate": 4.1071428571428566e-07,
+      "loss": 0.2982,
+      "step": 255
+    },
+    {
+      "epoch": 39.38461538461539,
+      "grad_norm": 9.348855018615723,
+      "learning_rate": 4.095238095238095e-07,
+      "loss": 0.3693,
+      "step": 256
+    },
+    {
+      "epoch": 39.53846153846154,
+      "grad_norm": 7.346530914306641,
+      "learning_rate": 4.083333333333333e-07,
+      "loss": 0.3475,
+      "step": 257
+    },
+    {
+      "epoch": 39.69230769230769,
+      "grad_norm": 5.94871711730957,
+      "learning_rate": 4.071428571428571e-07,
+      "loss": 0.3274,
+      "step": 258
+    },
+    {
+      "epoch": 39.84615384615385,
+      "grad_norm": 35.71484375,
+      "learning_rate": 4.0595238095238093e-07,
+      "loss": 0.3718,
+      "step": 259
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 13.31654167175293,
+      "learning_rate": 4.0476190476190476e-07,
+      "loss": 0.4351,
+      "step": 260
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 0.47377410531044006,
+      "eval_runtime": 1.1651,
+      "eval_samples_per_second": 234.318,
+      "eval_steps_per_second": 4.292,
+      "step": 260
+    },
+    {
+      "epoch": 40.15384615384615,
+      "grad_norm": 8.30079460144043,
+      "learning_rate": 4.0357142857142854e-07,
+      "loss": 0.4084,
+      "step": 261
+    },
+    {
+      "epoch": 40.30769230769231,
+      "grad_norm": 11.171014785766602,
+      "learning_rate": 4.0238095238095237e-07,
+      "loss": 0.2589,
+      "step": 262
+    },
+    {
+      "epoch": 40.46153846153846,
+      "grad_norm": 12.395658493041992,
+      "learning_rate": 4.011904761904762e-07,
+      "loss": 0.3645,
+      "step": 263
+    },
+    {
+      "epoch": 40.61538461538461,
+      "grad_norm": 12.52223014831543,
+      "learning_rate": 4e-07,
+      "loss": 0.3063,
+      "step": 264
+    },
+    {
+      "epoch": 40.76923076923077,
+      "grad_norm": 22.095630645751953,
+      "learning_rate": 3.988095238095238e-07,
+      "loss": 0.2982,
+      "step": 265
+    },
+    {
+      "epoch": 40.92307692307692,
+      "grad_norm": 19.723215103149414,
+      "learning_rate": 3.976190476190476e-07,
+      "loss": 0.3068,
+      "step": 266
+    },
+    {
+      "epoch": 40.92307692307692,
+      "eval_accuracy": 0.8644688644688645,
+      "eval_loss": 0.46877339482307434,
+      "eval_runtime": 1.1413,
+      "eval_samples_per_second": 239.204,
+      "eval_steps_per_second": 4.381,
+      "step": 266
+    },
+    {
+      "epoch": 41.07692307692308,
+      "grad_norm": 14.390270233154297,
+      "learning_rate": 3.9642857142857137e-07,
+      "loss": 0.3193,
+      "step": 267
+    },
+    {
+      "epoch": 41.23076923076923,
+      "grad_norm": 14.494707107543945,
+      "learning_rate": 3.952380952380952e-07,
+      "loss": 0.3274,
+      "step": 268
+    },
+    {
+      "epoch": 41.38461538461539,
+      "grad_norm": 9.31578540802002,
+      "learning_rate": 3.9404761904761903e-07,
+      "loss": 0.2905,
+      "step": 269
+    },
+    {
+      "epoch": 41.53846153846154,
+      "grad_norm": 11.39842700958252,
+      "learning_rate": 3.928571428571428e-07,
+      "loss": 0.2591,
+      "step": 270
+    },
+    {
+      "epoch": 41.69230769230769,
+      "grad_norm": 12.236638069152832,
+      "learning_rate": 3.9166666666666664e-07,
+      "loss": 0.3198,
+      "step": 271
+    },
+    {
+      "epoch": 41.84615384615385,
+      "grad_norm": 14.803117752075195,
+      "learning_rate": 3.9047619047619047e-07,
+      "loss": 0.3718,
+      "step": 272
+    },
+    {
+      "epoch": 42.0,
+      "grad_norm": 12.712557792663574,
+      "learning_rate": 3.8928571428571425e-07,
+      "loss": 0.3356,
+      "step": 273
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.8644688644688645,
+      "eval_loss": 0.45851626992225647,
+      "eval_runtime": 1.178,
+      "eval_samples_per_second": 231.744,
+      "eval_steps_per_second": 4.244,
+      "step": 273
+    },
+    {
+      "epoch": 42.15384615384615,
+      "grad_norm": 5.399446964263916,
+      "learning_rate": 3.880952380952381e-07,
+      "loss": 0.3028,
+      "step": 274
+    },
+    {
+      "epoch": 42.30769230769231,
+      "grad_norm": 9.010210990905762,
+      "learning_rate": 3.869047619047619e-07,
+      "loss": 0.2897,
+      "step": 275
+    },
+    {
+      "epoch": 42.46153846153846,
+      "grad_norm": 8.666064262390137,
+      "learning_rate": 3.857142857142857e-07,
+      "loss": 0.3349,
+      "step": 276
+    },
+    {
+      "epoch": 42.61538461538461,
+      "grad_norm": 25.002635955810547,
+      "learning_rate": 3.845238095238095e-07,
+      "loss": 0.3166,
+      "step": 277
+    },
+    {
+      "epoch": 42.76923076923077,
+      "grad_norm": 13.861302375793457,
+      "learning_rate": 3.8333333333333335e-07,
+      "loss": 0.3475,
+      "step": 278
+    },
+    {
+      "epoch": 42.92307692307692,
+      "grad_norm": 11.409740447998047,
+      "learning_rate": 3.821428571428571e-07,
+      "loss": 0.345,
+      "step": 279
+    },
+    {
+      "epoch": 42.92307692307692,
+      "eval_accuracy": 0.8681318681318682,
+      "eval_loss": 0.4540693759918213,
+      "eval_runtime": 1.1168,
+      "eval_samples_per_second": 244.446,
+      "eval_steps_per_second": 4.477,
+      "step": 279
+    },
+    {
+      "epoch": 43.07692307692308,
+      "grad_norm": 21.283527374267578,
+      "learning_rate": 3.809523809523809e-07,
+      "loss": 0.3316,
+      "step": 280
+    },
+    {
+      "epoch": 43.23076923076923,
+      "grad_norm": 7.409358978271484,
+      "learning_rate": 3.7976190476190474e-07,
+      "loss": 0.3305,
+      "step": 281
+    },
+    {
+      "epoch": 43.38461538461539,
+      "grad_norm": 11.763964653015137,
+      "learning_rate": 3.785714285714285e-07,
+      "loss": 0.3293,
+      "step": 282
+    },
+    {
+      "epoch": 43.53846153846154,
+      "grad_norm": 5.29448127746582,
+      "learning_rate": 3.7738095238095235e-07,
+      "loss": 0.2883,
+      "step": 283
+    },
+    {
+      "epoch": 43.69230769230769,
+      "grad_norm": 16.18635368347168,
+      "learning_rate": 3.761904761904762e-07,
+      "loss": 0.3549,
+      "step": 284
+    },
+    {
+      "epoch": 43.84615384615385,
+      "grad_norm": 17.321565628051758,
+      "learning_rate": 3.75e-07,
+      "loss": 0.3016,
+      "step": 285
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 8.436405181884766,
+      "learning_rate": 3.738095238095238e-07,
+      "loss": 0.3254,
+      "step": 286
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.8644688644688645,
+      "eval_loss": 0.45843759179115295,
+      "eval_runtime": 1.1597,
+      "eval_samples_per_second": 235.408,
+      "eval_steps_per_second": 4.311,
+      "step": 286
+    },
+    {
+      "epoch": 44.15384615384615,
+      "grad_norm": 11.177824974060059,
+      "learning_rate": 3.726190476190476e-07,
+      "loss": 0.3276,
+      "step": 287
+    },
+    {
+      "epoch": 44.30769230769231,
+      "grad_norm": 10.337651252746582,
+      "learning_rate": 3.7142857142857145e-07,
+      "loss": 0.3419,
+      "step": 288
+    },
+    {
+      "epoch": 44.46153846153846,
+      "grad_norm": 21.42737579345703,
+      "learning_rate": 3.7023809523809523e-07,
+      "loss": 0.2689,
+      "step": 289
+    },
+    {
+      "epoch": 44.61538461538461,
+      "grad_norm": 11.9132661819458,
+      "learning_rate": 3.6904761904761906e-07,
+      "loss": 0.3776,
+      "step": 290
+    },
+    {
+      "epoch": 44.76923076923077,
+      "grad_norm": 14.86318302154541,
+      "learning_rate": 3.678571428571429e-07,
+      "loss": 0.3984,
+      "step": 291
+    },
+    {
+      "epoch": 44.92307692307692,
+      "grad_norm": 15.56070327758789,
+      "learning_rate": 3.666666666666666e-07,
+      "loss": 0.3164,
+      "step": 292
+    },
+    {
+      "epoch": 44.92307692307692,
+      "eval_accuracy": 0.8571428571428571,
+      "eval_loss": 0.4591527581214905,
+      "eval_runtime": 1.1076,
+      "eval_samples_per_second": 246.479,
+      "eval_steps_per_second": 4.514,
+      "step": 292
+    },
+    {
+      "epoch": 45.07692307692308,
+      "grad_norm": 14.275500297546387,
+      "learning_rate": 3.6547619047619045e-07,
+      "loss": 0.2905,
+      "step": 293
+    },
+    {
+      "epoch": 45.23076923076923,
+      "grad_norm": 11.206035614013672,
+      "learning_rate": 3.642857142857143e-07,
+      "loss": 0.2992,
+      "step": 294
+    },
+    {
+      "epoch": 45.38461538461539,
+      "grad_norm": 10.007750511169434,
+      "learning_rate": 3.6309523809523805e-07,
+      "loss": 0.3662,
+      "step": 295
+    },
+    {
+      "epoch": 45.53846153846154,
+      "grad_norm": 13.443836212158203,
+      "learning_rate": 3.619047619047619e-07,
+      "loss": 0.3129,
+      "step": 296
+    },
+    {
+      "epoch": 45.69230769230769,
+      "grad_norm": 22.08678436279297,
+      "learning_rate": 3.607142857142857e-07,
+      "loss": 0.3904,
+      "step": 297
+    },
+    {
+      "epoch": 45.84615384615385,
+      "grad_norm": 24.27136993408203,
+      "learning_rate": 3.595238095238095e-07,
+      "loss": 0.3193,
+      "step": 298
+    },
+    {
+      "epoch": 46.0,
+      "grad_norm": 9.215178489685059,
+      "learning_rate": 3.583333333333333e-07,
+      "loss": 0.3657,
+      "step": 299
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.8608058608058609,
+      "eval_loss": 0.4533578157424927,
+      "eval_runtime": 1.1346,
+      "eval_samples_per_second": 240.624,
+      "eval_steps_per_second": 4.407,
+      "step": 299
+    },
+    {
+      "epoch": 46.15384615384615,
+      "grad_norm": 13.869656562805176,
+      "learning_rate": 3.5714285714285716e-07,
+      "loss": 0.3241,
+      "step": 300
+    },
+    {
+      "epoch": 46.30769230769231,
+      "grad_norm": 13.870816230773926,
+      "learning_rate": 3.5595238095238094e-07,
+      "loss": 0.294,
+      "step": 301
+    },
+    {
+      "epoch": 46.46153846153846,
+      "grad_norm": 5.2440338134765625,
+      "learning_rate": 3.5476190476190477e-07,
+      "loss": 0.3046,
+      "step": 302
+    },
+    {
+      "epoch": 46.61538461538461,
+      "grad_norm": 11.387068748474121,
+      "learning_rate": 3.535714285714286e-07,
+      "loss": 0.4067,
+      "step": 303
+    },
+    {
+      "epoch": 46.76923076923077,
+      "grad_norm": 10.643122673034668,
+      "learning_rate": 3.523809523809524e-07,
+      "loss": 0.3217,
+      "step": 304
+    },
+    {
+      "epoch": 46.92307692307692,
+      "grad_norm": 21.845155715942383,
+      "learning_rate": 3.5119047619047615e-07,
+      "loss": 0.2655,
+      "step": 305
+    },
+    {
+      "epoch": 46.92307692307692,
+      "eval_accuracy": 0.8644688644688645,
+      "eval_loss": 0.4501632750034332,
+      "eval_runtime": 1.1413,
+      "eval_samples_per_second": 239.204,
+      "eval_steps_per_second": 4.381,
+      "step": 305
+    },
+    {
+      "epoch": 47.07692307692308,
+      "grad_norm": 12.1947603225708,
+      "learning_rate": 3.5e-07,
+      "loss": 0.2994,
+      "step": 306
+    },
+    {
+      "epoch": 47.23076923076923,
+      "grad_norm": 21.314617156982422,
+      "learning_rate": 3.4880952380952376e-07,
+      "loss": 0.2751,
+      "step": 307
+    },
+    {
+      "epoch": 47.38461538461539,
+      "grad_norm": 5.636424541473389,
+      "learning_rate": 3.476190476190476e-07,
+      "loss": 0.292,
+      "step": 308
+    },
+    {
+      "epoch": 47.53846153846154,
+      "grad_norm": 11.540352821350098,
+      "learning_rate": 3.464285714285714e-07,
+      "loss": 0.3693,
+      "step": 309
+    },
+    {
+      "epoch": 47.69230769230769,
+      "grad_norm": 9.435784339904785,
+      "learning_rate": 3.452380952380952e-07,
+      "loss": 0.334,
+      "step": 310
+    },
+    {
+      "epoch": 47.84615384615385,
+      "grad_norm": 14.798314094543457,
+      "learning_rate": 3.4404761904761903e-07,
+      "loss": 0.2237,
+      "step": 311
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 10.26159954071045,
+      "learning_rate": 3.4285714285714286e-07,
+      "loss": 0.2981,
+      "step": 312
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.8644688644688645,
+      "eval_loss": 0.4451758563518524,
+      "eval_runtime": 1.2162,
+      "eval_samples_per_second": 224.473,
+      "eval_steps_per_second": 4.111,
+      "step": 312
+    },
+    {
+      "epoch": 48.15384615384615,
+      "grad_norm": 19.33696937561035,
+      "learning_rate": 3.4166666666666664e-07,
+      "loss": 0.2414,
+      "step": 313
+    },
+    {
+      "epoch": 48.30769230769231,
+      "grad_norm": 8.318500518798828,
+      "learning_rate": 3.4047619047619047e-07,
+      "loss": 0.3193,
+      "step": 314
+    },
+    {
+      "epoch": 48.46153846153846,
+      "grad_norm": 19.92133140563965,
+      "learning_rate": 3.392857142857143e-07,
+      "loss": 0.3218,
+      "step": 315
+    },
+    {
+      "epoch": 48.61538461538461,
+      "grad_norm": 18.465848922729492,
+      "learning_rate": 3.380952380952381e-07,
+      "loss": 0.3046,
+      "step": 316
+    },
+    {
+      "epoch": 48.76923076923077,
+      "grad_norm": 20.254159927368164,
+      "learning_rate": 3.369047619047619e-07,
+      "loss": 0.3651,
+      "step": 317
+    },
+    {
+      "epoch": 48.92307692307692,
+      "grad_norm": 9.707018852233887,
+      "learning_rate": 3.357142857142857e-07,
+      "loss": 0.3508,
+      "step": 318
+    },
+    {
+      "epoch": 48.92307692307692,
+      "eval_accuracy": 0.8791208791208791,
+      "eval_loss": 0.4371393620967865,
+      "eval_runtime": 1.2035,
+      "eval_samples_per_second": 226.845,
+      "eval_steps_per_second": 4.155,
+      "step": 318
+    },
+    {
+      "epoch": 49.07692307692308,
+      "grad_norm": 9.588353157043457,
+      "learning_rate": 3.3452380952380947e-07,
+      "loss": 0.4015,
+      "step": 319
+    },
+    {
+      "epoch": 49.23076923076923,
+      "grad_norm": 19.106985092163086,
+      "learning_rate": 3.333333333333333e-07,
+      "loss": 0.2461,
+      "step": 320
+    },
+    {
+      "epoch": 49.38461538461539,
+      "grad_norm": 18.668371200561523,
+      "learning_rate": 3.3214285714285713e-07,
+      "loss": 0.3201,
+      "step": 321
+    },
+    {
+      "epoch": 49.53846153846154,
+      "grad_norm": 22.97618865966797,
+      "learning_rate": 3.309523809523809e-07,
+      "loss": 0.2973,
+      "step": 322
+    },
+    {
+      "epoch": 49.69230769230769,
+      "grad_norm": 15.057040214538574,
+      "learning_rate": 3.2976190476190474e-07,
+      "loss": 0.2931,
+      "step": 323
+    },
+    {
+      "epoch": 49.84615384615385,
+      "grad_norm": 9.635611534118652,
+      "learning_rate": 3.2857142857142857e-07,
+      "loss": 0.3173,
+      "step": 324
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 8.364943504333496,
+      "learning_rate": 3.2738095238095235e-07,
+      "loss": 0.3419,
+      "step": 325
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.43940499424934387,
+      "eval_runtime": 1.1837,
+      "eval_samples_per_second": 230.634,
+      "eval_steps_per_second": 4.224,
+      "step": 325
+    },
+    {
+      "epoch": 50.15384615384615,
+      "grad_norm": 10.205245018005371,
+      "learning_rate": 3.261904761904762e-07,
+      "loss": 0.3744,
+      "step": 326
+    },
+    {
+      "epoch": 50.30769230769231,
+      "grad_norm": 8.429767608642578,
+      "learning_rate": 3.25e-07,
+      "loss": 0.2852,
+      "step": 327
+    },
+    {
+      "epoch": 50.46153846153846,
+      "grad_norm": 19.509811401367188,
+      "learning_rate": 3.238095238095238e-07,
+      "loss": 0.283,
+      "step": 328
+    },
+    {
+      "epoch": 50.61538461538461,
+      "grad_norm": 12.072210311889648,
+      "learning_rate": 3.226190476190476e-07,
+      "loss": 0.2955,
+      "step": 329
+    },
+    {
+      "epoch": 50.76923076923077,
+      "grad_norm": 19.032461166381836,
+      "learning_rate": 3.2142857142857145e-07,
+      "loss": 0.351,
+      "step": 330
+    },
+    {
+      "epoch": 50.92307692307692,
+      "grad_norm": 24.8001708984375,
+      "learning_rate": 3.202380952380952e-07,
+      "loss": 0.2668,
+      "step": 331
+    },
+    {
+      "epoch": 50.92307692307692,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.4430113434791565,
+      "eval_runtime": 1.1828,
+      "eval_samples_per_second": 230.809,
+      "eval_steps_per_second": 4.227,
+      "step": 331
+    },
+    {
+      "epoch": 51.07692307692308,
+      "grad_norm": 7.619977951049805,
+      "learning_rate": 3.19047619047619e-07,
+      "loss": 0.2316,
+      "step": 332
+    },
+    {
+      "epoch": 51.23076923076923,
+      "grad_norm": 15.534941673278809,
+      "learning_rate": 3.1785714285714284e-07,
+      "loss": 0.2948,
+      "step": 333
+    },
+    {
+      "epoch": 51.38461538461539,
+      "grad_norm": 6.369411945343018,
+      "learning_rate": 3.166666666666666e-07,
+      "loss": 0.2319,
+      "step": 334
+    },
+    {
+      "epoch": 51.53846153846154,
+      "grad_norm": 12.510886192321777,
+      "learning_rate": 3.1547619047619045e-07,
+      "loss": 0.3389,
+      "step": 335
+    },
+    {
+      "epoch": 51.69230769230769,
+      "grad_norm": 9.731184005737305,
+      "learning_rate": 3.142857142857143e-07,
+      "loss": 0.2822,
+      "step": 336
+    },
+    {
+      "epoch": 51.84615384615385,
+      "grad_norm": 6.847411155700684,
+      "learning_rate": 3.1309523809523806e-07,
+      "loss": 0.3447,
+      "step": 337
+    },
+    {
+      "epoch": 52.0,
+      "grad_norm": 15.400504112243652,
+      "learning_rate": 3.119047619047619e-07,
+      "loss": 0.2972,
+      "step": 338
+    },
+    {
+      "epoch": 52.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.43954789638519287,
+      "eval_runtime": 1.1577,
+      "eval_samples_per_second": 235.817,
+      "eval_steps_per_second": 4.319,
+      "step": 338
+    },
+    {
+      "epoch": 52.15384615384615,
+      "grad_norm": 14.40497875213623,
+      "learning_rate": 3.107142857142857e-07,
+      "loss": 0.2947,
+      "step": 339
+    },
+    {
+      "epoch": 52.30769230769231,
+      "grad_norm": 12.60912799835205,
+      "learning_rate": 3.095238095238095e-07,
+      "loss": 0.2866,
+      "step": 340
+    },
+    {
+      "epoch": 52.46153846153846,
+      "grad_norm": 10.782893180847168,
+      "learning_rate": 3.0833333333333333e-07,
+      "loss": 0.275,
+      "step": 341
+    },
+    {
+      "epoch": 52.61538461538461,
+      "grad_norm": 14.848359107971191,
+      "learning_rate": 3.0714285714285716e-07,
+      "loss": 0.3377,
+      "step": 342
+    },
+    {
+      "epoch": 52.76923076923077,
+      "grad_norm": 16.875308990478516,
+      "learning_rate": 3.0595238095238094e-07,
+      "loss": 0.2871,
+      "step": 343
+    },
+    {
+      "epoch": 52.92307692307692,
+      "grad_norm": 10.62590217590332,
+      "learning_rate": 3.0476190476190477e-07,
+      "loss": 0.3514,
+      "step": 344
+    },
+    {
+      "epoch": 52.92307692307692,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.437090665102005,
+      "eval_runtime": 1.1411,
+      "eval_samples_per_second": 239.24,
+      "eval_steps_per_second": 4.382,
+      "step": 344
+    },
+    {
+      "epoch": 53.07692307692308,
+      "grad_norm": 19.662609100341797,
+      "learning_rate": 3.0357142857142855e-07,
+      "loss": 0.2457,
+      "step": 345
+    },
+    {
+      "epoch": 53.23076923076923,
+      "grad_norm": 10.951351165771484,
+      "learning_rate": 3.023809523809523e-07,
+      "loss": 0.2542,
+      "step": 346
+    },
+    {
+      "epoch": 53.38461538461539,
+      "grad_norm": 6.810473918914795,
+      "learning_rate": 3.0119047619047616e-07,
+      "loss": 0.2873,
+      "step": 347
+    },
+    {
+      "epoch": 53.53846153846154,
+      "grad_norm": 11.747807502746582,
+      "learning_rate": 3e-07,
+      "loss": 0.3965,
+      "step": 348
+    },
+    {
+      "epoch": 53.69230769230769,
+      "grad_norm": 12.671740531921387,
+      "learning_rate": 2.9880952380952376e-07,
+      "loss": 0.3395,
+      "step": 349
+    },
+    {
+      "epoch": 53.84615384615385,
+      "grad_norm": 10.57718276977539,
+      "learning_rate": 2.976190476190476e-07,
+      "loss": 0.3071,
+      "step": 350
+    },
+    {
+      "epoch": 54.0,
+      "grad_norm": 7.219900131225586,
+      "learning_rate": 2.9642857142857143e-07,
+      "loss": 0.3012,
+      "step": 351
+    },
+    {
+      "epoch": 54.0,
+      "eval_accuracy": 0.8791208791208791,
+      "eval_loss": 0.43296605348587036,
+      "eval_runtime": 1.1556,
+      "eval_samples_per_second": 236.25,
+      "eval_steps_per_second": 4.327,
+      "step": 351
+    },
+    {
+      "epoch": 54.15384615384615,
+      "grad_norm": 7.588340759277344,
+      "learning_rate": 2.952380952380952e-07,
+      "loss": 0.3188,
+      "step": 352
+    },
+    {
+      "epoch": 54.30769230769231,
+      "grad_norm": 20.165128707885742,
+      "learning_rate": 2.9404761904761904e-07,
+      "loss": 0.377,
+      "step": 353
+    },
+    {
+      "epoch": 54.46153846153846,
+      "grad_norm": 9.232548713684082,
+      "learning_rate": 2.9285714285714287e-07,
+      "loss": 0.276,
+      "step": 354
+    },
+    {
+      "epoch": 54.61538461538461,
+      "grad_norm": 8.916671752929688,
+      "learning_rate": 2.916666666666667e-07,
+      "loss": 0.3599,
+      "step": 355
+    },
+    {
+      "epoch": 54.76923076923077,
+      "grad_norm": 11.789280891418457,
+      "learning_rate": 2.904761904761905e-07,
+      "loss": 0.3004,
+      "step": 356
+    },
+    {
+      "epoch": 54.92307692307692,
+      "grad_norm": 8.527569770812988,
+      "learning_rate": 2.892857142857143e-07,
+      "loss": 0.2725,
+      "step": 357
+    },
+    {
+      "epoch": 54.92307692307692,
+      "eval_accuracy": 0.8791208791208791,
+      "eval_loss": 0.4297783672809601,
+      "eval_runtime": 1.1219,
+      "eval_samples_per_second": 243.346,
+      "eval_steps_per_second": 4.457,
+      "step": 357
+    },
+    {
+      "epoch": 55.07692307692308,
+      "grad_norm": 9.423765182495117,
+      "learning_rate": 2.8809523809523803e-07,
+      "loss": 0.2731,
+      "step": 358
+    },
+    {
+      "epoch": 55.23076923076923,
+      "grad_norm": 13.653554916381836,
+      "learning_rate": 2.8690476190476186e-07,
+      "loss": 0.4064,
+      "step": 359
+    },
+    {
+      "epoch": 55.38461538461539,
+      "grad_norm": 7.217813491821289,
+      "learning_rate": 2.857142857142857e-07,
+      "loss": 0.305,
+      "step": 360
+    },
+    {
+      "epoch": 55.53846153846154,
+      "grad_norm": 7.166869640350342,
+      "learning_rate": 2.845238095238095e-07,
+      "loss": 0.2488,
+      "step": 361
+    },
+    {
+      "epoch": 55.69230769230769,
+      "grad_norm": 6.370850086212158,
+      "learning_rate": 2.833333333333333e-07,
+      "loss": 0.2946,
+      "step": 362
+    },
+    {
+      "epoch": 55.84615384615385,
+      "grad_norm": 19.36492919921875,
+      "learning_rate": 2.8214285714285713e-07,
+      "loss": 0.3314,
+      "step": 363
+    },
+    {
+      "epoch": 56.0,
+      "grad_norm": 10.563324928283691,
+      "learning_rate": 2.8095238095238096e-07,
+      "loss": 0.2547,
+      "step": 364
+    },
+    {
+      "epoch": 56.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4288838803768158,
+      "eval_runtime": 1.1679,
+      "eval_samples_per_second": 233.746,
+      "eval_steps_per_second": 4.281,
+      "step": 364
+    },
+    {
+      "epoch": 56.15384615384615,
+      "grad_norm": 10.190791130065918,
+      "learning_rate": 2.7976190476190474e-07,
+      "loss": 0.2646,
+      "step": 365
+    },
+    {
+      "epoch": 56.30769230769231,
+      "grad_norm": 16.170412063598633,
+      "learning_rate": 2.785714285714286e-07,
+      "loss": 0.3392,
+      "step": 366
+    },
+    {
+      "epoch": 56.46153846153846,
+      "grad_norm": 7.313807964324951,
+      "learning_rate": 2.773809523809524e-07,
+      "loss": 0.2909,
+      "step": 367
+    },
+    {
+      "epoch": 56.61538461538461,
+      "grad_norm": 9.653914451599121,
+      "learning_rate": 2.761904761904762e-07,
+      "loss": 0.3295,
+      "step": 368
+    },
+    {
+      "epoch": 56.76923076923077,
+      "grad_norm": 6.966893672943115,
+      "learning_rate": 2.75e-07,
+      "loss": 0.2956,
+      "step": 369
+    },
+    {
+      "epoch": 56.92307692307692,
+      "grad_norm": 15.753593444824219,
+      "learning_rate": 2.7380952380952385e-07,
+      "loss": 0.2896,
+      "step": 370
+    },
+    {
+      "epoch": 56.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4282112419605255,
+      "eval_runtime": 1.1522,
+      "eval_samples_per_second": 236.937,
+      "eval_steps_per_second": 4.339,
+      "step": 370
+    },
+    {
+      "epoch": 57.07692307692308,
+      "grad_norm": 13.775495529174805,
+      "learning_rate": 2.7261904761904757e-07,
+      "loss": 0.386,
+      "step": 371
+    },
+    {
+      "epoch": 57.23076923076923,
+      "grad_norm": 6.003649711608887,
+      "learning_rate": 2.714285714285714e-07,
+      "loss": 0.2473,
+      "step": 372
+    },
+    {
+      "epoch": 57.38461538461539,
+      "grad_norm": 22.614078521728516,
+      "learning_rate": 2.7023809523809523e-07,
+      "loss": 0.4098,
+      "step": 373
+    },
+    {
+      "epoch": 57.53846153846154,
+      "grad_norm": 15.326905250549316,
+      "learning_rate": 2.69047619047619e-07,
+      "loss": 0.2628,
+      "step": 374
+    },
+    {
+      "epoch": 57.69230769230769,
+      "grad_norm": 6.482524871826172,
+      "learning_rate": 2.6785714285714284e-07,
+      "loss": 0.3312,
+      "step": 375
+    },
+    {
+      "epoch": 57.84615384615385,
+      "grad_norm": 19.687318801879883,
+      "learning_rate": 2.6666666666666667e-07,
+      "loss": 0.3625,
+      "step": 376
+    },
+    {
+      "epoch": 58.0,
+      "grad_norm": 11.662421226501465,
+      "learning_rate": 2.6547619047619045e-07,
+      "loss": 0.3469,
+      "step": 377
+    },
+    {
+      "epoch": 58.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4272923469543457,
+      "eval_runtime": 1.1242,
+      "eval_samples_per_second": 242.848,
+      "eval_steps_per_second": 4.448,
+      "step": 377
+    },
+    {
+      "epoch": 58.15384615384615,
+      "grad_norm": 8.805383682250977,
+      "learning_rate": 2.642857142857143e-07,
+      "loss": 0.3987,
+      "step": 378
+    },
+    {
+      "epoch": 58.30769230769231,
+      "grad_norm": 11.661012649536133,
+      "learning_rate": 2.630952380952381e-07,
+      "loss": 0.2147,
+      "step": 379
+    },
+    {
+      "epoch": 58.46153846153846,
+      "grad_norm": 12.969446182250977,
+      "learning_rate": 2.619047619047619e-07,
+      "loss": 0.3074,
+      "step": 380
+    },
+    {
+      "epoch": 58.61538461538461,
+      "grad_norm": 9.435002326965332,
+      "learning_rate": 2.607142857142857e-07,
+      "loss": 0.2626,
+      "step": 381
+    },
+    {
+      "epoch": 58.76923076923077,
+      "grad_norm": 14.217181205749512,
+      "learning_rate": 2.5952380952380955e-07,
+      "loss": 0.2264,
+      "step": 382
+    },
+    {
+      "epoch": 58.92307692307692,
+      "grad_norm": 13.90890884399414,
+      "learning_rate": 2.5833333333333333e-07,
+      "loss": 0.3528,
+      "step": 383
+    },
+    {
+      "epoch": 58.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4268935024738312,
+      "eval_runtime": 1.1071,
+      "eval_samples_per_second": 246.586,
+      "eval_steps_per_second": 4.516,
+      "step": 383
+    },
+    {
+      "epoch": 59.07692307692308,
+      "grad_norm": 6.627048492431641,
+      "learning_rate": 2.571428571428571e-07,
+      "loss": 0.2552,
+      "step": 384
+    },
+    {
+      "epoch": 59.23076923076923,
+      "grad_norm": 14.275843620300293,
+      "learning_rate": 2.5595238095238094e-07,
+      "loss": 0.2876,
+      "step": 385
+    },
+    {
+      "epoch": 59.38461538461539,
+      "grad_norm": 15.865604400634766,
+      "learning_rate": 2.547619047619047e-07,
+      "loss": 0.3701,
+      "step": 386
+    },
+    {
+      "epoch": 59.53846153846154,
+      "grad_norm": 12.051728248596191,
+      "learning_rate": 2.5357142857142855e-07,
+      "loss": 0.2598,
+      "step": 387
+    },
+    {
+      "epoch": 59.69230769230769,
+      "grad_norm": 11.886255264282227,
+      "learning_rate": 2.523809523809524e-07,
+      "loss": 0.238,
+      "step": 388
+    },
+    {
+      "epoch": 59.84615384615385,
+      "grad_norm": 12.55905818939209,
+      "learning_rate": 2.5119047619047616e-07,
+      "loss": 0.2893,
+      "step": 389
+    },
+    {
+      "epoch": 60.0,
+      "grad_norm": 18.840225219726562,
+      "learning_rate": 2.5e-07,
+      "loss": 0.2552,
+      "step": 390
+    },
+    {
+      "epoch": 60.0,
+      "eval_accuracy": 0.8681318681318682,
+      "eval_loss": 0.4324240982532501,
+      "eval_runtime": 1.1584,
+      "eval_samples_per_second": 235.676,
+      "eval_steps_per_second": 4.316,
+      "step": 390
+    },
+    {
+      "epoch": 60.15384615384615,
+      "grad_norm": 7.739254474639893,
+      "learning_rate": 2.488095238095238e-07,
+      "loss": 0.3825,
+      "step": 391
+    },
+    {
+      "epoch": 60.30769230769231,
+      "grad_norm": 27.177335739135742,
+      "learning_rate": 2.476190476190476e-07,
+      "loss": 0.3577,
+      "step": 392
+    },
+    {
+      "epoch": 60.46153846153846,
+      "grad_norm": 8.35522747039795,
+      "learning_rate": 2.4642857142857143e-07,
+      "loss": 0.296,
+      "step": 393
+    },
+    {
+      "epoch": 60.61538461538461,
+      "grad_norm": 12.505022048950195,
+      "learning_rate": 2.452380952380952e-07,
+      "loss": 0.2922,
+      "step": 394
+    },
+    {
+      "epoch": 60.76923076923077,
+      "grad_norm": 8.860665321350098,
+      "learning_rate": 2.4404761904761904e-07,
+      "loss": 0.3924,
+      "step": 395
+    },
+    {
+      "epoch": 60.92307692307692,
+      "grad_norm": 13.714400291442871,
+      "learning_rate": 2.4285714285714287e-07,
+      "loss": 0.239,
+      "step": 396
+    },
+    {
+      "epoch": 60.92307692307692,
+      "eval_accuracy": 0.8644688644688645,
+      "eval_loss": 0.4319455325603485,
+      "eval_runtime": 1.158,
+      "eval_samples_per_second": 235.758,
+      "eval_steps_per_second": 4.318,
+      "step": 396
+    },
+    {
+      "epoch": 61.07692307692308,
+      "grad_norm": 11.82459831237793,
+      "learning_rate": 2.4166666666666665e-07,
+      "loss": 0.2555,
+      "step": 397
+    },
+    {
+      "epoch": 61.23076923076923,
+      "grad_norm": 13.206809997558594,
+      "learning_rate": 2.404761904761905e-07,
+      "loss": 0.3353,
+      "step": 398
+    },
+    {
+      "epoch": 61.38461538461539,
+      "grad_norm": 11.129719734191895,
+      "learning_rate": 2.392857142857143e-07,
+      "loss": 0.313,
+      "step": 399
+    },
+    {
+      "epoch": 61.53846153846154,
+      "grad_norm": 19.73814582824707,
+      "learning_rate": 2.3809523809523806e-07,
+      "loss": 0.2591,
+      "step": 400
+    },
+    {
+      "epoch": 61.69230769230769,
+      "grad_norm": 10.5856351852417,
+      "learning_rate": 2.369047619047619e-07,
+      "loss": 0.3416,
+      "step": 401
+    },
+    {
+      "epoch": 61.84615384615385,
+      "grad_norm": 12.684653282165527,
+      "learning_rate": 2.357142857142857e-07,
+      "loss": 0.2315,
+      "step": 402
+    },
+    {
+      "epoch": 62.0,
+      "grad_norm": 8.558398246765137,
+      "learning_rate": 2.345238095238095e-07,
+      "loss": 0.3321,
+      "step": 403
+    },
+    {
+      "epoch": 62.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.42702218890190125,
+      "eval_runtime": 1.1757,
+      "eval_samples_per_second": 232.2,
+      "eval_steps_per_second": 4.253,
+      "step": 403
+    },
+    {
+      "epoch": 62.15384615384615,
+      "grad_norm": 9.598026275634766,
+      "learning_rate": 2.3333333333333333e-07,
+      "loss": 0.3119,
+      "step": 404
+    },
+    {
+      "epoch": 62.30769230769231,
+      "grad_norm": 13.107952117919922,
+      "learning_rate": 2.3214285714285714e-07,
+      "loss": 0.3379,
+      "step": 405
+    },
+    {
+      "epoch": 62.46153846153846,
+      "grad_norm": 18.639419555664062,
+      "learning_rate": 2.3095238095238097e-07,
+      "loss": 0.2689,
+      "step": 406
+    },
+    {
+      "epoch": 62.61538461538461,
+      "grad_norm": 17.175498962402344,
+      "learning_rate": 2.2976190476190475e-07,
+      "loss": 0.3154,
+      "step": 407
+    },
+    {
+      "epoch": 62.76923076923077,
+      "grad_norm": 11.377558708190918,
+      "learning_rate": 2.2857142857142855e-07,
+      "loss": 0.2969,
+      "step": 408
+    },
+    {
+      "epoch": 62.92307692307692,
+      "grad_norm": 12.16077709197998,
+      "learning_rate": 2.2738095238095238e-07,
+      "loss": 0.3115,
+      "step": 409
+    },
+    {
+      "epoch": 62.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.41838309168815613,
+      "eval_runtime": 1.133,
+      "eval_samples_per_second": 240.951,
+      "eval_steps_per_second": 4.413,
+      "step": 409
+    },
+    {
+      "epoch": 63.07692307692308,
+      "grad_norm": 29.19352912902832,
+      "learning_rate": 2.2619047619047619e-07,
+      "loss": 0.327,
+      "step": 410
+    },
+    {
+      "epoch": 63.23076923076923,
+      "grad_norm": 21.762849807739258,
+      "learning_rate": 2.25e-07,
+      "loss": 0.3142,
+      "step": 411
+    },
+    {
+      "epoch": 63.38461538461539,
+      "grad_norm": 20.668453216552734,
+      "learning_rate": 2.2380952380952382e-07,
+      "loss": 0.2705,
+      "step": 412
+    },
+    {
+      "epoch": 63.53846153846154,
+      "grad_norm": 10.485206604003906,
+      "learning_rate": 2.226190476190476e-07,
+      "loss": 0.2635,
+      "step": 413
+    },
+    {
+      "epoch": 63.69230769230769,
+      "grad_norm": 5.819842338562012,
+      "learning_rate": 2.214285714285714e-07,
+      "loss": 0.281,
+      "step": 414
+    },
+    {
+      "epoch": 63.84615384615385,
+      "grad_norm": 8.578193664550781,
+      "learning_rate": 2.2023809523809523e-07,
+      "loss": 0.2981,
+      "step": 415
+    },
+    {
+      "epoch": 64.0,
+      "grad_norm": 14.02076244354248,
+      "learning_rate": 2.1904761904761904e-07,
+      "loss": 0.306,
+      "step": 416
+    },
+    {
+      "epoch": 64.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4168645739555359,
+      "eval_runtime": 1.1838,
+      "eval_samples_per_second": 230.615,
+      "eval_steps_per_second": 4.224,
+      "step": 416
+    },
+    {
+      "epoch": 64.15384615384616,
+      "grad_norm": 8.580323219299316,
+      "learning_rate": 2.1785714285714284e-07,
+      "loss": 0.2481,
+      "step": 417
+    },
+    {
+      "epoch": 64.3076923076923,
+      "grad_norm": 12.701449394226074,
+      "learning_rate": 2.1666666666666667e-07,
+      "loss": 0.2978,
+      "step": 418
+    },
+    {
+      "epoch": 64.46153846153847,
+      "grad_norm": 7.2950544357299805,
+      "learning_rate": 2.1547619047619048e-07,
+      "loss": 0.2778,
+      "step": 419
+    },
+    {
+      "epoch": 64.61538461538461,
+      "grad_norm": 14.55117416381836,
+      "learning_rate": 2.1428571428571426e-07,
+      "loss": 0.3589,
+      "step": 420
+    },
+    {
+      "epoch": 64.76923076923077,
+      "grad_norm": 15.672528266906738,
+      "learning_rate": 2.130952380952381e-07,
+      "loss": 0.3638,
+      "step": 421
+    },
+    {
+      "epoch": 64.92307692307692,
+      "grad_norm": 8.848112106323242,
+      "learning_rate": 2.119047619047619e-07,
+      "loss": 0.3086,
+      "step": 422
+    },
+    {
+      "epoch": 64.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.41758179664611816,
+      "eval_runtime": 1.1479,
+      "eval_samples_per_second": 237.817,
+      "eval_steps_per_second": 4.356,
+      "step": 422
+    },
+    {
+      "epoch": 65.07692307692308,
+      "grad_norm": 20.250051498413086,
+      "learning_rate": 2.107142857142857e-07,
+      "loss": 0.2965,
+      "step": 423
+    },
+    {
+      "epoch": 65.23076923076923,
+      "grad_norm": 8.315861701965332,
+      "learning_rate": 2.0952380952380953e-07,
+      "loss": 0.3119,
+      "step": 424
+    },
+    {
+      "epoch": 65.38461538461539,
+      "grad_norm": 7.086258888244629,
+      "learning_rate": 2.0833333333333333e-07,
+      "loss": 0.2904,
+      "step": 425
+    },
+    {
+      "epoch": 65.53846153846153,
+      "grad_norm": 9.35214900970459,
+      "learning_rate": 2.0714285714285714e-07,
+      "loss": 0.2787,
+      "step": 426
+    },
+    {
+      "epoch": 65.6923076923077,
+      "grad_norm": 11.061731338500977,
+      "learning_rate": 2.0595238095238094e-07,
+      "loss": 0.287,
+      "step": 427
+    },
+    {
+      "epoch": 65.84615384615384,
+      "grad_norm": 17.736583709716797,
+      "learning_rate": 2.0476190476190475e-07,
+      "loss": 0.3101,
+      "step": 428
+    },
+    {
+      "epoch": 66.0,
+      "grad_norm": 20.51115608215332,
+      "learning_rate": 2.0357142857142855e-07,
+      "loss": 0.4256,
+      "step": 429
+    },
+    {
+      "epoch": 66.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4195899963378906,
+      "eval_runtime": 1.2137,
+      "eval_samples_per_second": 224.925,
+      "eval_steps_per_second": 4.119,
+      "step": 429
+    },
+    {
+      "epoch": 66.15384615384616,
+      "grad_norm": 7.543363094329834,
+      "learning_rate": 2.0238095238095238e-07,
+      "loss": 0.1944,
+      "step": 430
+    },
+    {
+      "epoch": 66.3076923076923,
+      "grad_norm": 15.504809379577637,
+      "learning_rate": 2.0119047619047619e-07,
+      "loss": 0.2767,
+      "step": 431
+    },
+    {
+      "epoch": 66.46153846153847,
+      "grad_norm": 12.454192161560059,
+      "learning_rate": 2e-07,
+      "loss": 0.2697,
+      "step": 432
+    },
+    {
+      "epoch": 66.61538461538461,
+      "grad_norm": 11.19575023651123,
+      "learning_rate": 1.988095238095238e-07,
+      "loss": 0.2626,
+      "step": 433
+    },
+    {
+      "epoch": 66.76923076923077,
+      "grad_norm": 8.080245971679688,
+      "learning_rate": 1.976190476190476e-07,
+      "loss": 0.3262,
+      "step": 434
+    },
+    {
+      "epoch": 66.92307692307692,
+      "grad_norm": 16.002864837646484,
+      "learning_rate": 1.964285714285714e-07,
+      "loss": 0.2798,
+      "step": 435
+    },
+    {
+      "epoch": 66.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4219285249710083,
+      "eval_runtime": 1.1309,
+      "eval_samples_per_second": 241.401,
+      "eval_steps_per_second": 4.421,
+      "step": 435
+    },
+    {
+      "epoch": 67.07692307692308,
+      "grad_norm": 9.127429008483887,
+      "learning_rate": 1.9523809523809524e-07,
+      "loss": 0.2828,
+      "step": 436
+    },
+    {
+      "epoch": 67.23076923076923,
+      "grad_norm": 9.812334060668945,
+      "learning_rate": 1.9404761904761904e-07,
+      "loss": 0.3028,
+      "step": 437
+    },
+    {
+      "epoch": 67.38461538461539,
+      "grad_norm": 16.504629135131836,
+      "learning_rate": 1.9285714285714284e-07,
+      "loss": 0.2382,
+      "step": 438
+    },
+    {
+      "epoch": 67.53846153846153,
+      "grad_norm": 15.225912094116211,
+      "learning_rate": 1.9166666666666668e-07,
+      "loss": 0.3387,
+      "step": 439
+    },
+    {
+      "epoch": 67.6923076923077,
+      "grad_norm": 32.0360221862793,
+      "learning_rate": 1.9047619047619045e-07,
+      "loss": 0.252,
+      "step": 440
+    },
+    {
+      "epoch": 67.84615384615384,
+      "grad_norm": 10.604074478149414,
+      "learning_rate": 1.8928571428571426e-07,
+      "loss": 0.3423,
+      "step": 441
+    },
+    {
+      "epoch": 68.0,
+      "grad_norm": 11.707740783691406,
+      "learning_rate": 1.880952380952381e-07,
+      "loss": 0.3016,
+      "step": 442
+    },
+    {
+      "epoch": 68.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4223931133747101,
+      "eval_runtime": 1.1901,
+      "eval_samples_per_second": 229.39,
+      "eval_steps_per_second": 4.201,
+      "step": 442
+    },
+    {
+      "epoch": 68.15384615384616,
+      "grad_norm": 15.720952987670898,
+      "learning_rate": 1.869047619047619e-07,
+      "loss": 0.2488,
+      "step": 443
+    },
+    {
+      "epoch": 68.3076923076923,
+      "grad_norm": 21.468849182128906,
+      "learning_rate": 1.8571428571428572e-07,
+      "loss": 0.3655,
+      "step": 444
+    },
+    {
+      "epoch": 68.46153846153847,
+      "grad_norm": 7.170112133026123,
+      "learning_rate": 1.8452380952380953e-07,
+      "loss": 0.2293,
+      "step": 445
+    },
+    {
+      "epoch": 68.61538461538461,
+      "grad_norm": 13.677178382873535,
+      "learning_rate": 1.833333333333333e-07,
+      "loss": 0.3512,
+      "step": 446
+    },
+    {
+      "epoch": 68.76923076923077,
+      "grad_norm": 15.128756523132324,
+      "learning_rate": 1.8214285714285714e-07,
+      "loss": 0.3291,
+      "step": 447
+    },
+    {
+      "epoch": 68.92307692307692,
+      "grad_norm": 9.91322135925293,
+      "learning_rate": 1.8095238095238094e-07,
+      "loss": 0.2791,
+      "step": 448
+    },
+    {
+      "epoch": 68.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.42070242762565613,
+      "eval_runtime": 1.1329,
+      "eval_samples_per_second": 240.966,
+      "eval_steps_per_second": 4.413,
+      "step": 448
+    },
+    {
+      "epoch": 69.07692307692308,
+      "grad_norm": 18.428592681884766,
+      "learning_rate": 1.7976190476190475e-07,
+      "loss": 0.2857,
+      "step": 449
+    },
+    {
+      "epoch": 69.23076923076923,
+      "grad_norm": 13.563444137573242,
+      "learning_rate": 1.7857142857142858e-07,
+      "loss": 0.3579,
+      "step": 450
+    },
+    {
+      "epoch": 69.38461538461539,
+      "grad_norm": 7.071059226989746,
+      "learning_rate": 1.7738095238095238e-07,
+      "loss": 0.288,
+      "step": 451
+    },
+    {
+      "epoch": 69.53846153846153,
+      "grad_norm": 13.733859062194824,
+      "learning_rate": 1.761904761904762e-07,
+      "loss": 0.3107,
+      "step": 452
+    },
+    {
+      "epoch": 69.6923076923077,
+      "grad_norm": 24.475296020507812,
+      "learning_rate": 1.75e-07,
+      "loss": 0.3036,
+      "step": 453
+    },
+    {
+      "epoch": 69.84615384615384,
+      "grad_norm": 8.03947925567627,
+      "learning_rate": 1.738095238095238e-07,
+      "loss": 0.3046,
+      "step": 454
+    },
+    {
+      "epoch": 70.0,
+      "grad_norm": 12.644840240478516,
+      "learning_rate": 1.726190476190476e-07,
+      "loss": 0.2651,
+      "step": 455
+    },
+    {
+      "epoch": 70.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4188561737537384,
+      "eval_runtime": 1.1824,
+      "eval_samples_per_second": 230.892,
+      "eval_steps_per_second": 4.229,
+      "step": 455
+    },
+    {
+      "epoch": 70.15384615384616,
+      "grad_norm": 10.41529655456543,
+      "learning_rate": 1.7142857142857143e-07,
+      "loss": 0.2741,
+      "step": 456
+    },
+    {
+      "epoch": 70.3076923076923,
+      "grad_norm": 13.780343055725098,
+      "learning_rate": 1.7023809523809524e-07,
+      "loss": 0.1817,
+      "step": 457
+    },
+    {
+      "epoch": 70.46153846153847,
+      "grad_norm": 15.142984390258789,
+      "learning_rate": 1.6904761904761904e-07,
+      "loss": 0.2511,
+      "step": 458
+    },
+    {
+      "epoch": 70.61538461538461,
+      "grad_norm": 11.113754272460938,
+      "learning_rate": 1.6785714285714285e-07,
+      "loss": 0.3617,
+      "step": 459
+    },
+    {
+      "epoch": 70.76923076923077,
+      "grad_norm": 9.247007369995117,
+      "learning_rate": 1.6666666666666665e-07,
+      "loss": 0.3423,
+      "step": 460
+    },
+    {
+      "epoch": 70.92307692307692,
+      "grad_norm": 10.317591667175293,
+      "learning_rate": 1.6547619047619045e-07,
+      "loss": 0.2466,
+      "step": 461
+    },
+    {
+      "epoch": 70.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4177640378475189,
+      "eval_runtime": 1.1382,
+      "eval_samples_per_second": 239.846,
+      "eval_steps_per_second": 4.393,
+      "step": 461
+    },
+    {
+      "epoch": 71.07692307692308,
+      "grad_norm": 12.4044771194458,
+      "learning_rate": 1.6428571428571429e-07,
+      "loss": 0.274,
+      "step": 462
+    },
+    {
+      "epoch": 71.23076923076923,
+      "grad_norm": 12.682540893554688,
+      "learning_rate": 1.630952380952381e-07,
+      "loss": 0.3048,
+      "step": 463
+    },
+    {
+      "epoch": 71.38461538461539,
+      "grad_norm": 25.84153175354004,
+      "learning_rate": 1.619047619047619e-07,
+      "loss": 0.2463,
+      "step": 464
+    },
+    {
+      "epoch": 71.53846153846153,
+      "grad_norm": 13.235491752624512,
+      "learning_rate": 1.6071428571428573e-07,
+      "loss": 0.415,
+      "step": 465
+    },
+    {
+      "epoch": 71.6923076923077,
+      "grad_norm": 6.873939514160156,
+      "learning_rate": 1.595238095238095e-07,
+      "loss": 0.3163,
+      "step": 466
+    },
+    {
+      "epoch": 71.84615384615384,
+      "grad_norm": 16.569108963012695,
+      "learning_rate": 1.583333333333333e-07,
+      "loss": 0.3067,
+      "step": 467
+    },
+    {
+      "epoch": 72.0,
+      "grad_norm": 18.71702003479004,
+      "learning_rate": 1.5714285714285714e-07,
+      "loss": 0.1913,
+      "step": 468
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4177253544330597,
+      "eval_runtime": 1.1798,
+      "eval_samples_per_second": 231.388,
+      "eval_steps_per_second": 4.238,
+      "step": 468
+    },
+    {
+      "epoch": 72.15384615384616,
+      "grad_norm": 21.70823097229004,
+      "learning_rate": 1.5595238095238094e-07,
+      "loss": 0.3477,
+      "step": 469
+    },
+    {
+      "epoch": 72.3076923076923,
+      "grad_norm": 18.4373779296875,
+      "learning_rate": 1.5476190476190475e-07,
+      "loss": 0.3226,
+      "step": 470
+    },
+    {
+      "epoch": 72.46153846153847,
+      "grad_norm": 12.795662879943848,
+      "learning_rate": 1.5357142857142858e-07,
+      "loss": 0.36,
+      "step": 471
+    },
+    {
+      "epoch": 72.61538461538461,
+      "grad_norm": 6.41522741317749,
+      "learning_rate": 1.5238095238095238e-07,
+      "loss": 0.2615,
+      "step": 472
+    },
+    {
+      "epoch": 72.76923076923077,
+      "grad_norm": 11.777397155761719,
+      "learning_rate": 1.5119047619047616e-07,
+      "loss": 0.2648,
+      "step": 473
+    },
+    {
+      "epoch": 72.92307692307692,
+      "grad_norm": 14.508996963500977,
+      "learning_rate": 1.5e-07,
+      "loss": 0.2719,
+      "step": 474
+    },
+    {
+      "epoch": 72.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4163550138473511,
+      "eval_runtime": 1.1181,
+      "eval_samples_per_second": 244.158,
+      "eval_steps_per_second": 4.472,
+      "step": 474
+    },
+    {
+      "epoch": 73.07692307692308,
+      "grad_norm": 15.088624000549316,
+      "learning_rate": 1.488095238095238e-07,
+      "loss": 0.3153,
+      "step": 475
+    },
+    {
+      "epoch": 73.23076923076923,
+      "grad_norm": 6.610665798187256,
+      "learning_rate": 1.476190476190476e-07,
+      "loss": 0.2292,
+      "step": 476
+    },
+    {
+      "epoch": 73.38461538461539,
+      "grad_norm": 19.090049743652344,
+      "learning_rate": 1.4642857142857143e-07,
+      "loss": 0.2765,
+      "step": 477
+    },
+    {
+      "epoch": 73.53846153846153,
+      "grad_norm": 19.875932693481445,
+      "learning_rate": 1.4523809523809524e-07,
+      "loss": 0.2797,
+      "step": 478
+    },
+    {
+      "epoch": 73.6923076923077,
+      "grad_norm": 21.46002960205078,
+      "learning_rate": 1.4404761904761902e-07,
+      "loss": 0.2846,
+      "step": 479
+    },
+    {
+      "epoch": 73.84615384615384,
+      "grad_norm": 9.745357513427734,
+      "learning_rate": 1.4285714285714285e-07,
+      "loss": 0.3138,
+      "step": 480
+    },
+    {
+      "epoch": 74.0,
+      "grad_norm": 9.963276863098145,
+      "learning_rate": 1.4166666666666665e-07,
+      "loss": 0.3364,
+      "step": 481
+    },
+    {
+      "epoch": 74.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.41662341356277466,
+      "eval_runtime": 1.1761,
+      "eval_samples_per_second": 232.123,
+      "eval_steps_per_second": 4.251,
+      "step": 481
+    },
+    {
+      "epoch": 74.15384615384616,
+      "grad_norm": 12.587307929992676,
+      "learning_rate": 1.4047619047619048e-07,
+      "loss": 0.3476,
+      "step": 482
+    },
+    {
+      "epoch": 74.3076923076923,
+      "grad_norm": 15.544804573059082,
+      "learning_rate": 1.392857142857143e-07,
+      "loss": 0.298,
+      "step": 483
+    },
+    {
+      "epoch": 74.46153846153847,
+      "grad_norm": 16.263813018798828,
+      "learning_rate": 1.380952380952381e-07,
+      "loss": 0.3103,
+      "step": 484
+    },
+    {
+      "epoch": 74.61538461538461,
+      "grad_norm": 15.350561141967773,
+      "learning_rate": 1.3690476190476192e-07,
+      "loss": 0.3083,
+      "step": 485
+    },
+    {
+      "epoch": 74.76923076923077,
+      "grad_norm": 18.922351837158203,
+      "learning_rate": 1.357142857142857e-07,
+      "loss": 0.2371,
+      "step": 486
+    },
+    {
+      "epoch": 74.92307692307692,
+      "grad_norm": 10.000033378601074,
+      "learning_rate": 1.345238095238095e-07,
+      "loss": 0.283,
+      "step": 487
+    },
+    {
+      "epoch": 74.92307692307692,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.4179239273071289,
+      "eval_runtime": 1.1425,
+      "eval_samples_per_second": 238.952,
+      "eval_steps_per_second": 4.376,
+      "step": 487
+    },
+    {
+      "epoch": 75.07692307692308,
+      "grad_norm": 11.999493598937988,
+      "learning_rate": 1.3333333333333334e-07,
+      "loss": 0.2356,
+      "step": 488
+    },
+    {
+      "epoch": 75.23076923076923,
+      "grad_norm": 15.124945640563965,
+      "learning_rate": 1.3214285714285714e-07,
+      "loss": 0.2473,
+      "step": 489
+    },
+    {
+      "epoch": 75.38461538461539,
+      "grad_norm": 5.885743618011475,
+      "learning_rate": 1.3095238095238095e-07,
+      "loss": 0.2623,
+      "step": 490
+    },
+    {
+      "epoch": 75.53846153846153,
+      "grad_norm": 17.72136116027832,
+      "learning_rate": 1.2976190476190478e-07,
+      "loss": 0.2667,
+      "step": 491
+    },
+    {
+      "epoch": 75.6923076923077,
+      "grad_norm": 17.649593353271484,
+      "learning_rate": 1.2857142857142855e-07,
+      "loss": 0.3016,
+      "step": 492
+    },
+    {
+      "epoch": 75.84615384615384,
+      "grad_norm": 15.614336013793945,
+      "learning_rate": 1.2738095238095236e-07,
+      "loss": 0.2771,
+      "step": 493
+    },
+    {
+      "epoch": 76.0,
+      "grad_norm": 6.446508884429932,
+      "learning_rate": 1.261904761904762e-07,
+      "loss": 0.2891,
+      "step": 494
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.4174346923828125,
+      "eval_runtime": 1.1254,
+      "eval_samples_per_second": 242.591,
+      "eval_steps_per_second": 4.443,
+      "step": 494
+    },
+    {
+      "epoch": 76.15384615384616,
+      "grad_norm": 8.419719696044922,
+      "learning_rate": 1.25e-07,
+      "loss": 0.2639,
+      "step": 495
+    },
+    {
+      "epoch": 76.3076923076923,
+      "grad_norm": 8.500795364379883,
+      "learning_rate": 1.238095238095238e-07,
+      "loss": 0.2902,
+      "step": 496
+    },
+    {
+      "epoch": 76.46153846153847,
+      "grad_norm": 9.533052444458008,
+      "learning_rate": 1.226190476190476e-07,
+      "loss": 0.3342,
+      "step": 497
+    },
+    {
+      "epoch": 76.61538461538461,
+      "grad_norm": 32.33898162841797,
+      "learning_rate": 1.2142857142857143e-07,
+      "loss": 0.3059,
+      "step": 498
+    },
+    {
+      "epoch": 76.76923076923077,
+      "grad_norm": 14.258476257324219,
+      "learning_rate": 1.2023809523809524e-07,
+      "loss": 0.249,
+      "step": 499
+    },
+    {
+      "epoch": 76.92307692307692,
+      "grad_norm": 28.383731842041016,
+      "learning_rate": 1.1904761904761903e-07,
+      "loss": 0.2625,
+      "step": 500
+    },
+    {
+      "epoch": 76.92307692307692,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.4179657995700836,
+      "eval_runtime": 1.0921,
+      "eval_samples_per_second": 249.988,
+      "eval_steps_per_second": 4.579,
+      "step": 500
+    },
+    {
+      "epoch": 77.07692307692308,
+      "grad_norm": 14.453038215637207,
+      "learning_rate": 1.1785714285714285e-07,
+      "loss": 0.3501,
+      "step": 501
+    },
+    {
+      "epoch": 77.23076923076923,
+      "grad_norm": 15.141488075256348,
+      "learning_rate": 1.1666666666666667e-07,
+      "loss": 0.2959,
+      "step": 502
+    },
+    {
+      "epoch": 77.38461538461539,
+      "grad_norm": 27.595386505126953,
+      "learning_rate": 1.1547619047619048e-07,
+      "loss": 0.3883,
+      "step": 503
+    },
+    {
+      "epoch": 77.53846153846153,
+      "grad_norm": 7.041699409484863,
+      "learning_rate": 1.1428571428571427e-07,
+      "loss": 0.2567,
+      "step": 504
+    },
+    {
+      "epoch": 77.6923076923077,
+      "grad_norm": 5.878727912902832,
+      "learning_rate": 1.1309523809523809e-07,
+      "loss": 0.1855,
+      "step": 505
+    },
+    {
+      "epoch": 77.84615384615384,
+      "grad_norm": 9.228104591369629,
+      "learning_rate": 1.1190476190476191e-07,
+      "loss": 0.2607,
+      "step": 506
+    },
+    {
+      "epoch": 78.0,
+      "grad_norm": 10.124485969543457,
+      "learning_rate": 1.107142857142857e-07,
+      "loss": 0.2843,
+      "step": 507
+    },
+    {
+      "epoch": 78.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4184325635433197,
+      "eval_runtime": 1.2364,
+      "eval_samples_per_second": 220.81,
+      "eval_steps_per_second": 4.044,
+      "step": 507
+    },
+    {
+      "epoch": 78.15384615384616,
+      "grad_norm": 32.02544403076172,
+      "learning_rate": 1.0952380952380952e-07,
+      "loss": 0.2707,
+      "step": 508
+    },
+    {
+      "epoch": 78.3076923076923,
+      "grad_norm": 8.309106826782227,
+      "learning_rate": 1.0833333333333334e-07,
+      "loss": 0.3501,
+      "step": 509
+    },
+    {
+      "epoch": 78.46153846153847,
+      "grad_norm": 10.85938549041748,
+      "learning_rate": 1.0714285714285713e-07,
+      "loss": 0.3379,
+      "step": 510
+    },
+    {
+      "epoch": 78.61538461538461,
+      "grad_norm": 14.523594856262207,
+      "learning_rate": 1.0595238095238095e-07,
+      "loss": 0.2191,
+      "step": 511
+    },
+    {
+      "epoch": 78.76923076923077,
+      "grad_norm": 16.100353240966797,
+      "learning_rate": 1.0476190476190476e-07,
+      "loss": 0.3082,
+      "step": 512
+    },
+    {
+      "epoch": 78.92307692307692,
+      "grad_norm": 10.894936561584473,
+      "learning_rate": 1.0357142857142857e-07,
+      "loss": 0.375,
+      "step": 513
+    },
+    {
+      "epoch": 78.92307692307692,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.41671693325042725,
+      "eval_runtime": 1.116,
+      "eval_samples_per_second": 244.622,
+      "eval_steps_per_second": 4.48,
+      "step": 513
+    },
+    {
+      "epoch": 79.07692307692308,
+      "grad_norm": 10.38732624053955,
+      "learning_rate": 1.0238095238095237e-07,
+      "loss": 0.2812,
+      "step": 514
+    },
+    {
+      "epoch": 79.23076923076923,
+      "grad_norm": 15.733484268188477,
+      "learning_rate": 1.0119047619047619e-07,
+      "loss": 0.3247,
+      "step": 515
+    },
+    {
+      "epoch": 79.38461538461539,
+      "grad_norm": 6.772809028625488,
+      "learning_rate": 1e-07,
+      "loss": 0.221,
+      "step": 516
+    },
+    {
+      "epoch": 79.53846153846153,
+      "grad_norm": 8.480406761169434,
+      "learning_rate": 9.88095238095238e-08,
+      "loss": 0.2817,
+      "step": 517
+    },
+    {
+      "epoch": 79.6923076923077,
+      "grad_norm": 8.911038398742676,
+      "learning_rate": 9.761904761904762e-08,
+      "loss": 0.3514,
+      "step": 518
+    },
+    {
+      "epoch": 79.84615384615384,
+      "grad_norm": 18.952770233154297,
+      "learning_rate": 9.642857142857142e-08,
+      "loss": 0.2881,
+      "step": 519
+    },
+    {
+      "epoch": 80.0,
+      "grad_norm": 6.310261249542236,
+      "learning_rate": 9.523809523809523e-08,
+      "loss": 0.3107,
+      "step": 520
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.41499239206314087,
+      "eval_runtime": 1.191,
+      "eval_samples_per_second": 229.225,
+      "eval_steps_per_second": 4.198,
+      "step": 520
+    },
+    {
+      "epoch": 80.15384615384616,
+      "grad_norm": 17.95909309387207,
+      "learning_rate": 9.404761904761904e-08,
+      "loss": 0.2551,
+      "step": 521
+    },
+    {
+      "epoch": 80.3076923076923,
+      "grad_norm": 27.223407745361328,
+      "learning_rate": 9.285714285714286e-08,
+      "loss": 0.3356,
+      "step": 522
+    },
+    {
+      "epoch": 80.46153846153847,
+      "grad_norm": 6.595218658447266,
+      "learning_rate": 9.166666666666665e-08,
+      "loss": 0.2383,
+      "step": 523
+    },
+    {
+      "epoch": 80.61538461538461,
+      "grad_norm": 20.459001541137695,
+      "learning_rate": 9.047619047619047e-08,
+      "loss": 0.3265,
+      "step": 524
+    },
+    {
+      "epoch": 80.76923076923077,
+      "grad_norm": 15.349759101867676,
+      "learning_rate": 8.928571428571429e-08,
+      "loss": 0.2763,
+      "step": 525
+    },
+    {
+      "epoch": 80.92307692307692,
+      "grad_norm": 10.789344787597656,
+      "learning_rate": 8.80952380952381e-08,
+      "loss": 0.3742,
+      "step": 526
+    },
+    {
+      "epoch": 80.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4144986569881439,
+      "eval_runtime": 1.1418,
+      "eval_samples_per_second": 239.087,
+      "eval_steps_per_second": 4.379,
+      "step": 526
+    },
+    {
+      "epoch": 81.07692307692308,
+      "grad_norm": 13.075380325317383,
+      "learning_rate": 8.69047619047619e-08,
+      "loss": 0.2957,
+      "step": 527
+    },
+    {
+      "epoch": 81.23076923076923,
+      "grad_norm": 10.450531959533691,
+      "learning_rate": 8.571428571428572e-08,
+      "loss": 0.3245,
+      "step": 528
+    },
+    {
+      "epoch": 81.38461538461539,
+      "grad_norm": 20.416603088378906,
+      "learning_rate": 8.452380952380952e-08,
+      "loss": 0.2355,
+      "step": 529
+    },
+    {
+      "epoch": 81.53846153846153,
+      "grad_norm": 11.954909324645996,
+      "learning_rate": 8.333333333333333e-08,
+      "loss": 0.2847,
+      "step": 530
+    },
+    {
+      "epoch": 81.6923076923077,
+      "grad_norm": 10.024072647094727,
+      "learning_rate": 8.214285714285714e-08,
+      "loss": 0.3908,
+      "step": 531
+    },
+    {
+      "epoch": 81.84615384615384,
+      "grad_norm": 13.764399528503418,
+      "learning_rate": 8.095238095238095e-08,
+      "loss": 0.2062,
+      "step": 532
+    },
+    {
+      "epoch": 82.0,
+      "grad_norm": 18.268247604370117,
+      "learning_rate": 7.976190476190475e-08,
+      "loss": 0.2574,
+      "step": 533
+    },
+    {
+      "epoch": 82.0,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.4145357608795166,
+      "eval_runtime": 1.2255,
+      "eval_samples_per_second": 222.772,
+      "eval_steps_per_second": 4.08,
+      "step": 533
+    },
+    {
+      "epoch": 82.15384615384616,
+      "grad_norm": 11.204299926757812,
+      "learning_rate": 7.857142857142857e-08,
+      "loss": 0.294,
+      "step": 534
+    },
+    {
+      "epoch": 82.3076923076923,
+      "grad_norm": 12.602919578552246,
+      "learning_rate": 7.738095238095237e-08,
+      "loss": 0.3379,
+      "step": 535
+    },
+    {
+      "epoch": 82.46153846153847,
+      "grad_norm": 17.001785278320312,
+      "learning_rate": 7.619047619047619e-08,
+      "loss": 0.2501,
+      "step": 536
+    },
+    {
+      "epoch": 82.61538461538461,
+      "grad_norm": 10.472540855407715,
+      "learning_rate": 7.5e-08,
+      "loss": 0.2673,
+      "step": 537
+    },
+    {
+      "epoch": 82.76923076923077,
+      "grad_norm": 12.93094539642334,
+      "learning_rate": 7.38095238095238e-08,
+      "loss": 0.3303,
+      "step": 538
+    },
+    {
+      "epoch": 82.92307692307692,
+      "grad_norm": 15.572615623474121,
+      "learning_rate": 7.261904761904762e-08,
+      "loss": 0.329,
+      "step": 539
+    },
+    {
+      "epoch": 82.92307692307692,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.41488900780677795,
+      "eval_runtime": 1.1463,
+      "eval_samples_per_second": 238.152,
+      "eval_steps_per_second": 4.362,
+      "step": 539
+    },
+    {
+      "epoch": 83.07692307692308,
+      "grad_norm": 8.833599090576172,
+      "learning_rate": 7.142857142857142e-08,
+      "loss": 0.2862,
+      "step": 540
+    },
+    {
+      "epoch": 83.23076923076923,
+      "grad_norm": 7.227090358734131,
+      "learning_rate": 7.023809523809524e-08,
+      "loss": 0.2553,
+      "step": 541
+    },
+    {
+      "epoch": 83.38461538461539,
+      "grad_norm": 16.44085693359375,
+      "learning_rate": 6.904761904761905e-08,
+      "loss": 0.3129,
+      "step": 542
+    },
+    {
+      "epoch": 83.53846153846153,
+      "grad_norm": 13.633960723876953,
+      "learning_rate": 6.785714285714285e-08,
+      "loss": 0.3134,
+      "step": 543
+    },
+    {
+      "epoch": 83.6923076923077,
+      "grad_norm": 16.555570602416992,
+      "learning_rate": 6.666666666666667e-08,
+      "loss": 0.2504,
+      "step": 544
+    },
+    {
+      "epoch": 83.84615384615384,
+      "grad_norm": 7.340324878692627,
+      "learning_rate": 6.547619047619047e-08,
+      "loss": 0.2966,
+      "step": 545
+    },
+    {
+      "epoch": 84.0,
+      "grad_norm": 10.442778587341309,
+      "learning_rate": 6.428571428571428e-08,
+      "loss": 0.2727,
+      "step": 546
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.4145146608352661,
+      "eval_runtime": 1.1279,
+      "eval_samples_per_second": 242.042,
+      "eval_steps_per_second": 4.433,
+      "step": 546
+    },
+    {
+      "epoch": 84.15384615384616,
+      "grad_norm": 9.072919845581055,
+      "learning_rate": 6.30952380952381e-08,
+      "loss": 0.2461,
+      "step": 547
+    },
+    {
+      "epoch": 84.3076923076923,
+      "grad_norm": 8.624760627746582,
+      "learning_rate": 6.19047619047619e-08,
+      "loss": 0.2812,
+      "step": 548
+    },
+    {
+      "epoch": 84.46153846153847,
+      "grad_norm": 8.95349407196045,
+      "learning_rate": 6.071428571428572e-08,
+      "loss": 0.2835,
+      "step": 549
+    },
+    {
+      "epoch": 84.61538461538461,
+      "grad_norm": 18.060441970825195,
+      "learning_rate": 5.9523809523809515e-08,
+      "loss": 0.2697,
+      "step": 550
+    },
+    {
+      "epoch": 84.76923076923077,
+      "grad_norm": 15.820292472839355,
+      "learning_rate": 5.833333333333333e-08,
+      "loss": 0.3266,
+      "step": 551
+    },
+    {
+      "epoch": 84.92307692307692,
+      "grad_norm": 10.503725051879883,
+      "learning_rate": 5.714285714285714e-08,
+      "loss": 0.2977,
+      "step": 552
+    },
+    {
+      "epoch": 84.92307692307692,
+      "eval_accuracy": 0.8754578754578755,
+      "eval_loss": 0.41494670510292053,
+      "eval_runtime": 1.1373,
+      "eval_samples_per_second": 240.052,
+      "eval_steps_per_second": 4.397,
+      "step": 552
+    },
+    {
+      "epoch": 85.07692307692308,
+      "grad_norm": 9.66882610321045,
+      "learning_rate": 5.5952380952380955e-08,
+      "loss": 0.3452,
+      "step": 553
+    },
+    {
+      "epoch": 85.23076923076923,
+      "grad_norm": 9.360061645507812,
+      "learning_rate": 5.476190476190476e-08,
+      "loss": 0.3033,
+      "step": 554
+    },
+    {
+      "epoch": 85.38461538461539,
+      "grad_norm": 10.878594398498535,
+      "learning_rate": 5.3571428571428564e-08,
+      "loss": 0.323,
+      "step": 555
+    },
+    {
+      "epoch": 85.53846153846153,
+      "grad_norm": 9.788655281066895,
+      "learning_rate": 5.238095238095238e-08,
+      "loss": 0.2796,
+      "step": 556
+    },
+    {
+      "epoch": 85.6923076923077,
+      "grad_norm": 11.249568939208984,
+      "learning_rate": 5.1190476190476187e-08,
+      "loss": 0.2381,
+      "step": 557
+    },
+    {
+      "epoch": 85.84615384615384,
+      "grad_norm": 22.544105529785156,
+      "learning_rate": 5e-08,
+      "loss": 0.3593,
+      "step": 558
+    },
+    {
+      "epoch": 86.0,
+      "grad_norm": 9.197429656982422,
+      "learning_rate": 4.880952380952381e-08,
+      "loss": 0.2611,
+      "step": 559
+    },
+    {
+      "epoch": 86.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4160268008708954,
+      "eval_runtime": 1.1319,
+      "eval_samples_per_second": 241.183,
+      "eval_steps_per_second": 4.417,
+      "step": 559
+    },
+    {
+      "epoch": 86.15384615384616,
+      "grad_norm": 10.480982780456543,
+      "learning_rate": 4.7619047619047613e-08,
+      "loss": 0.2919,
+      "step": 560
+    },
+    {
+      "epoch": 86.3076923076923,
+      "grad_norm": 15.81564712524414,
+      "learning_rate": 4.642857142857143e-08,
+      "loss": 0.3395,
+      "step": 561
+    },
+    {
+      "epoch": 86.46153846153847,
+      "grad_norm": 25.986629486083984,
+      "learning_rate": 4.5238095238095236e-08,
+      "loss": 0.3436,
+      "step": 562
+    },
+    {
+      "epoch": 86.61538461538461,
+      "grad_norm": 20.077136993408203,
+      "learning_rate": 4.404761904761905e-08,
+      "loss": 0.363,
+      "step": 563
+    },
+    {
+      "epoch": 86.76923076923077,
+      "grad_norm": 16.67424774169922,
+      "learning_rate": 4.285714285714286e-08,
+      "loss": 0.2731,
+      "step": 564
+    },
+    {
+      "epoch": 86.92307692307692,
+      "grad_norm": 16.004140853881836,
+      "learning_rate": 4.166666666666666e-08,
+      "loss": 0.2542,
+      "step": 565
+    },
+    {
+      "epoch": 86.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4169901907444,
+      "eval_runtime": 1.1368,
+      "eval_samples_per_second": 240.144,
+      "eval_steps_per_second": 4.398,
+      "step": 565
+    },
+    {
+      "epoch": 87.07692307692308,
+      "grad_norm": 18.212156295776367,
+      "learning_rate": 4.0476190476190474e-08,
+      "loss": 0.3265,
+      "step": 566
+    },
+    {
+      "epoch": 87.23076923076923,
+      "grad_norm": 12.494462966918945,
+      "learning_rate": 3.9285714285714285e-08,
+      "loss": 0.3226,
+      "step": 567
+    },
+    {
+      "epoch": 87.38461538461539,
+      "grad_norm": 6.6637773513793945,
+      "learning_rate": 3.8095238095238096e-08,
+      "loss": 0.2262,
+      "step": 568
+    },
+    {
+      "epoch": 87.53846153846153,
+      "grad_norm": 12.791160583496094,
+      "learning_rate": 3.69047619047619e-08,
+      "loss": 0.3541,
+      "step": 569
+    },
+    {
+      "epoch": 87.6923076923077,
+      "grad_norm": 9.732351303100586,
+      "learning_rate": 3.571428571428571e-08,
+      "loss": 0.2551,
+      "step": 570
+    },
+    {
+      "epoch": 87.84615384615384,
+      "grad_norm": 21.216760635375977,
+      "learning_rate": 3.452380952380952e-08,
+      "loss": 0.2629,
+      "step": 571
+    },
+    {
+      "epoch": 88.0,
+      "grad_norm": 15.293636322021484,
+      "learning_rate": 3.3333333333333334e-08,
+      "loss": 0.2665,
+      "step": 572
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.41707876324653625,
+      "eval_runtime": 1.1463,
+      "eval_samples_per_second": 238.151,
+      "eval_steps_per_second": 4.362,
+      "step": 572
+    },
+    {
+      "epoch": 88.15384615384616,
+      "grad_norm": 11.668240547180176,
+      "learning_rate": 3.214285714285714e-08,
+      "loss": 0.3315,
+      "step": 573
+    },
+    {
+      "epoch": 88.3076923076923,
+      "grad_norm": 28.14773178100586,
+      "learning_rate": 3.095238095238095e-08,
+      "loss": 0.3098,
+      "step": 574
+    },
+    {
+      "epoch": 88.46153846153847,
+      "grad_norm": 13.815803527832031,
+      "learning_rate": 2.9761904761904758e-08,
+      "loss": 0.3381,
+      "step": 575
+    },
+    {
+      "epoch": 88.61538461538461,
+      "grad_norm": 11.341737747192383,
+      "learning_rate": 2.857142857142857e-08,
+      "loss": 0.2995,
+      "step": 576
+    },
+    {
+      "epoch": 88.76923076923077,
+      "grad_norm": 9.191329002380371,
+      "learning_rate": 2.738095238095238e-08,
+      "loss": 0.2485,
+      "step": 577
+    },
+    {
+      "epoch": 88.92307692307692,
+      "grad_norm": 12.781599998474121,
+      "learning_rate": 2.619047619047619e-08,
+      "loss": 0.2654,
+      "step": 578
+    },
+    {
+      "epoch": 88.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4170469343662262,
+      "eval_runtime": 1.1144,
+      "eval_samples_per_second": 244.964,
+      "eval_steps_per_second": 4.487,
+      "step": 578
+    },
+    {
+      "epoch": 89.07692307692308,
+      "grad_norm": 18.813859939575195,
+      "learning_rate": 2.5e-08,
+      "loss": 0.4085,
+      "step": 579
+    },
+    {
+      "epoch": 89.23076923076923,
+      "grad_norm": 9.056866645812988,
+      "learning_rate": 2.3809523809523807e-08,
+      "loss": 0.2377,
+      "step": 580
+    },
+    {
+      "epoch": 89.38461538461539,
+      "grad_norm": 21.75194549560547,
+      "learning_rate": 2.2619047619047618e-08,
+      "loss": 0.2511,
+      "step": 581
+    },
+    {
+      "epoch": 89.53846153846153,
+      "grad_norm": 24.87982177734375,
+      "learning_rate": 2.142857142857143e-08,
+      "loss": 0.3468,
+      "step": 582
+    },
+    {
+      "epoch": 89.6923076923077,
+      "grad_norm": 13.90104866027832,
+      "learning_rate": 2.0238095238095237e-08,
+      "loss": 0.2001,
+      "step": 583
+    },
+    {
+      "epoch": 89.84615384615384,
+      "grad_norm": 13.514042854309082,
+      "learning_rate": 1.9047619047619048e-08,
+      "loss": 0.2673,
+      "step": 584
+    },
+    {
+      "epoch": 90.0,
+      "grad_norm": 17.871490478515625,
+      "learning_rate": 1.7857142857142856e-08,
+      "loss": 0.3059,
+      "step": 585
+    },
+    {
+      "epoch": 90.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4172230660915375,
+      "eval_runtime": 1.1156,
+      "eval_samples_per_second": 244.721,
+      "eval_steps_per_second": 4.482,
+      "step": 585
+    },
+    {
+      "epoch": 90.15384615384616,
+      "grad_norm": 9.703531265258789,
+      "learning_rate": 1.6666666666666667e-08,
+      "loss": 0.1731,
+      "step": 586
+    },
+    {
+      "epoch": 90.3076923076923,
+      "grad_norm": 13.011953353881836,
+      "learning_rate": 1.5476190476190475e-08,
+      "loss": 0.3008,
+      "step": 587
+    },
+    {
+      "epoch": 90.46153846153847,
+      "grad_norm": 13.331348419189453,
+      "learning_rate": 1.4285714285714284e-08,
+      "loss": 0.2402,
+      "step": 588
+    },
+    {
+      "epoch": 90.61538461538461,
+      "grad_norm": 9.919706344604492,
+      "learning_rate": 1.3095238095238096e-08,
+      "loss": 0.3376,
+      "step": 589
+    },
+    {
+      "epoch": 90.76923076923077,
+      "grad_norm": 20.184898376464844,
+      "learning_rate": 1.1904761904761903e-08,
+      "loss": 0.3106,
+      "step": 590
+    },
+    {
+      "epoch": 90.92307692307692,
+      "grad_norm": 6.9222588539123535,
+      "learning_rate": 1.0714285714285715e-08,
+      "loss": 0.2377,
+      "step": 591
+    },
+    {
+      "epoch": 90.92307692307692,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.417271226644516,
+      "eval_runtime": 1.1559,
+      "eval_samples_per_second": 236.178,
+      "eval_steps_per_second": 4.326,
+      "step": 591
+    },
+    {
+      "epoch": 91.07692307692308,
+      "grad_norm": 10.644214630126953,
+      "learning_rate": 9.523809523809524e-09,
+      "loss": 0.2738,
+      "step": 592
+    },
+    {
+      "epoch": 91.23076923076923,
+      "grad_norm": 12.530488014221191,
+      "learning_rate": 8.333333333333334e-09,
+      "loss": 0.3082,
+      "step": 593
+    },
+    {
+      "epoch": 91.38461538461539,
+      "grad_norm": 10.068582534790039,
+      "learning_rate": 7.142857142857142e-09,
+      "loss": 0.2678,
+      "step": 594
+    },
+    {
+      "epoch": 91.53846153846153,
+      "grad_norm": 8.5696439743042,
+      "learning_rate": 5.952380952380952e-09,
+      "loss": 0.2902,
+      "step": 595
+    },
+    {
+      "epoch": 91.6923076923077,
+      "grad_norm": 22.662599563598633,
+      "learning_rate": 4.761904761904762e-09,
+      "loss": 0.2826,
+      "step": 596
+    },
+    {
+      "epoch": 91.84615384615384,
+      "grad_norm": 19.666378021240234,
+      "learning_rate": 3.571428571428571e-09,
+      "loss": 0.3109,
+      "step": 597
+    },
+    {
+      "epoch": 92.0,
+      "grad_norm": 15.434282302856445,
+      "learning_rate": 2.380952380952381e-09,
+      "loss": 0.2896,
+      "step": 598
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4172247350215912,
+      "eval_runtime": 1.1283,
+      "eval_samples_per_second": 241.962,
+      "eval_steps_per_second": 4.432,
+      "step": 598
+    },
+    {
+      "epoch": 92.15384615384616,
+      "grad_norm": 21.7716064453125,
+      "learning_rate": 1.1904761904761905e-09,
+      "loss": 0.3775,
+      "step": 599
+    },
+    {
+      "epoch": 92.3076923076923,
+      "grad_norm": 9.635214805603027,
+      "learning_rate": 0.0,
+      "loss": 0.3133,
+      "step": 600
+    },
+    {
+      "epoch": 92.3076923076923,
+      "eval_accuracy": 0.8717948717948718,
+      "eval_loss": 0.4172203838825226,
+      "eval_runtime": 1.1209,
+      "eval_samples_per_second": 243.565,
+      "eval_steps_per_second": 4.461,
+      "step": 600
+    },
+    {
+      "epoch": 92.3076923076923,
+      "step": 600,
+      "total_flos": 1.4722503891660472e+18,
+      "train_loss": 0.3897706772387028,
+      "train_runtime": 753.6703,
+      "train_samples_per_second": 108.137,
+      "train_steps_per_second": 0.796
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 600,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4722503891660472e+18,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}