diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5920 @@
+{
+  "best_metric": 1.71484375,
+  "best_model_checkpoint": "bert_12_layer_model_v1_complete_training/checkpoint-450000",
+  "epoch": 5.0,
+  "global_step": 457720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3715366784496464e-05,
+      "loss": 7.3181,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.748913176609214e-05,
+      "loss": 6.6228,
+      "step": 1000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.9686644919105525e-05,
+      "loss": 6.5274,
+      "step": 1500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.12520067118919e-05,
+      "loss": 6.4586,
+      "step": 2000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.246120561428155e-05,
+      "loss": 6.4182,
+      "step": 2500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.345314745008792e-05,
+      "loss": 6.3793,
+      "step": 3000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.428842793791915e-05,
+      "loss": 6.3585,
+      "step": 3500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.501488165769166e-05,
+      "loss": 6.3289,
+      "step": 4000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.5653079262541536e-05,
+      "loss": 6.3082,
+      "step": 4500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.622625682029237e-05,
+      "loss": 6.2991,
+      "step": 5000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.6742676283733686e-05,
+      "loss": 6.2733,
+      "step": 5500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7216022395887685e-05,
+      "loss": 6.2641,
+      "step": 6000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.7649711804211924e-05,
+      "loss": 6.2454,
+      "step": 6500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.805285726627034e-05,
+      "loss": 6.2317,
+      "step": 7000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.842667223534965e-05,
+      "loss": 6.225,
+      "step": 7500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8777756603491424e-05,
+      "loss": 6.2133,
+      "step": 8000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.910622835154613e-05,
+      "loss": 6.2112,
+      "step": 8500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.941716313408369e-05,
+      "loss": 6.2017,
+      "step": 9000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.971010421672582e-05,
+      "loss": 6.1863,
+      "step": 9500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9989131766092136e-05,
+      "loss": 6.1779,
+      "step": 10000
+    },
+    {
+      "epoch": 0.11,
+      "eval_accuracy": 0.1486846615077511,
+      "eval_loss": 6.171875,
+      "eval_runtime": 1132.2269,
+      "eval_samples_per_second": 272.345,
+      "eval_steps_per_second": 4.256,
+      "step": 10000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.994673009916912e-05,
+      "loss": 6.1746,
+      "step": 10500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9890891628696506e-05,
+      "loss": 6.1631,
+      "step": 11000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.983527651210578e-05,
+      "loss": 6.1478,
+      "step": 11500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.9779438041633164e-05,
+      "loss": 6.122,
+      "step": 12000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.972382292504244e-05,
+      "loss": 6.0631,
+      "step": 12500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.966798445456982e-05,
+      "loss": 5.9931,
+      "step": 13000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.96123693379791e-05,
+      "loss": 5.9445,
+      "step": 13500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.955653086750648e-05,
+      "loss": 5.9015,
+      "step": 14000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9500915750915755e-05,
+      "loss": 5.8657,
+      "step": 14500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.944507728044314e-05,
+      "loss": 5.8324,
+      "step": 15000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.938946216385241e-05,
+      "loss": 5.7922,
+      "step": 15500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.9333623693379796e-05,
+      "loss": 5.7422,
+      "step": 16000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.9278008576789064e-05,
+      "loss": 5.656,
+      "step": 16500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.9222170106316454e-05,
+      "loss": 5.4694,
+      "step": 17000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.916655498972572e-05,
+      "loss": 5.2826,
+      "step": 17500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.9110716519253105e-05,
+      "loss": 5.1224,
+      "step": 18000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.905510140266238e-05,
+      "loss": 4.9978,
+      "step": 18500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.8999262932189763e-05,
+      "loss": 4.8805,
+      "step": 19000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.894364781559904e-05,
+      "loss": 4.7842,
+      "step": 19500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.888780934512642e-05,
+      "loss": 4.6914,
+      "step": 20000
+    },
+    {
+      "epoch": 0.22,
+      "eval_accuracy": 0.31781246345401765,
+      "eval_loss": 4.50390625,
+      "eval_runtime": 1170.7642,
+      "eval_samples_per_second": 263.38,
+      "eval_steps_per_second": 4.116,
+      "step": 20000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.8832194228535696e-05,
+      "loss": 4.5744,
+      "step": 20500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.877635575806307e-05,
+      "loss": 4.429,
+      "step": 21000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.8720740641472354e-05,
+      "loss": 4.2966,
+      "step": 21500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.866490217099973e-05,
+      "loss": 4.1822,
+      "step": 22000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.8609287054409006e-05,
+      "loss": 4.0959,
+      "step": 22500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.855344858393639e-05,
+      "loss": 4.0111,
+      "step": 23000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.8497833467345664e-05,
+      "loss": 3.9325,
+      "step": 23500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.844199499687305e-05,
+      "loss": 3.8522,
+      "step": 24000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.838637988028232e-05,
+      "loss": 3.7721,
+      "step": 24500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.8330541409809705e-05,
+      "loss": 3.6958,
+      "step": 25000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.827492629321898e-05,
+      "loss": 3.6262,
+      "step": 25500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.821908782274636e-05,
+      "loss": 3.5627,
+      "step": 26000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.816347270615564e-05,
+      "loss": 3.5099,
+      "step": 26500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.810763423568302e-05,
+      "loss": 3.4574,
+      "step": 27000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.805201911909229e-05,
+      "loss": 3.4123,
+      "step": 27500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.799618064861968e-05,
+      "loss": 3.3626,
+      "step": 28000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.794056553202895e-05,
+      "loss": 3.3279,
+      "step": 28500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.788472706155634e-05,
+      "loss": 3.2847,
+      "step": 29000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.7829111944965605e-05,
+      "loss": 3.2545,
+      "step": 29500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.777327347449299e-05,
+      "loss": 3.2325,
+      "step": 30000
+    },
+    {
+      "epoch": 0.33,
+      "eval_accuracy": 0.4772067036025076,
+      "eval_loss": 3.09765625,
+      "eval_runtime": 1211.4769,
+      "eval_samples_per_second": 254.529,
+      "eval_steps_per_second": 3.978,
+      "step": 30000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.771765835790226e-05,
+      "loss": 3.1891,
+      "step": 30500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.7661819887429646e-05,
+      "loss": 3.1672,
+      "step": 31000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.760620477083892e-05,
+      "loss": 3.1438,
+      "step": 31500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.75503663003663e-05,
+      "loss": 3.1156,
+      "step": 32000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.749475118377558e-05,
+      "loss": 3.0926,
+      "step": 32500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.7438912713302956e-05,
+      "loss": 3.0655,
+      "step": 33000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.738329759671224e-05,
+      "loss": 3.0478,
+      "step": 33500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.7327459126239614e-05,
+      "loss": 3.0258,
+      "step": 34000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.727184400964889e-05,
+      "loss": 2.9978,
+      "step": 34500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.721600553917627e-05,
+      "loss": 2.9823,
+      "step": 35000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.7160390422585546e-05,
+      "loss": 2.967,
+      "step": 35500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.710455195211293e-05,
+      "loss": 2.945,
+      "step": 36000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.7048936835522204e-05,
+      "loss": 2.9286,
+      "step": 36500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.699309836504959e-05,
+      "loss": 2.9146,
+      "step": 37000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.6937483248458856e-05,
+      "loss": 2.8956,
+      "step": 37500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.6881644777986246e-05,
+      "loss": 2.8819,
+      "step": 38000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.6826029661395514e-05,
+      "loss": 2.8697,
+      "step": 38500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.6770191190922904e-05,
+      "loss": 2.8571,
+      "step": 39000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.671457607433217e-05,
+      "loss": 2.8425,
+      "step": 39500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.665873760385956e-05,
+      "loss": 2.831,
+      "step": 40000
+    },
+    {
+      "epoch": 0.44,
+      "eval_accuracy": 0.5223783031240715,
+      "eval_loss": 2.7265625,
+      "eval_runtime": 1229.432,
+      "eval_samples_per_second": 250.812,
+      "eval_steps_per_second": 3.92,
+      "step": 40000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.660312248726883e-05,
+      "loss": 2.8148,
+      "step": 40500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.654728401679622e-05,
+      "loss": 2.8,
+      "step": 41000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.649166890020549e-05,
+      "loss": 2.7909,
+      "step": 41500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.643583042973287e-05,
+      "loss": 2.7752,
+      "step": 42000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.6380215313142146e-05,
+      "loss": 2.7642,
+      "step": 42500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.632437684266953e-05,
+      "loss": 2.7494,
+      "step": 43000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.6268761726078804e-05,
+      "loss": 2.7433,
+      "step": 43500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.621292325560618e-05,
+      "loss": 2.7307,
+      "step": 44000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.615730813901546e-05,
+      "loss": 2.7294,
+      "step": 44500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.610146966854284e-05,
+      "loss": 2.7154,
+      "step": 45000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.604585455195212e-05,
+      "loss": 2.7007,
+      "step": 45500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.5990016081479496e-05,
+      "loss": 2.6921,
+      "step": 46000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.593440096488877e-05,
+      "loss": 2.6842,
+      "step": 46500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.5878562494416154e-05,
+      "loss": 2.6767,
+      "step": 47000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.582294737782543e-05,
+      "loss": 2.6696,
+      "step": 47500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.576710890735281e-05,
+      "loss": 2.6598,
+      "step": 48000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.571149379076208e-05,
+      "loss": 2.6462,
+      "step": 48500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.565565532028947e-05,
+      "loss": 2.6451,
+      "step": 49000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.560004020369874e-05,
+      "loss": 2.6378,
+      "step": 49500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.554420173322613e-05,
+      "loss": 2.6262,
+      "step": 50000
+    },
+    {
+      "epoch": 0.55,
+      "eval_accuracy": 0.545525046728465,
+      "eval_loss": 2.537109375,
+      "eval_runtime": 1265.0645,
+      "eval_samples_per_second": 243.747,
+      "eval_steps_per_second": 3.809,
+      "step": 50000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5488586616635397e-05,
+      "loss": 2.616,
+      "step": 50500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5432748146162787e-05,
+      "loss": 2.6109,
+      "step": 51000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5377133029572055e-05,
+      "loss": 2.6142,
+      "step": 51500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.5321294559099445e-05,
+      "loss": 2.5987,
+      "step": 52000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.526567944250871e-05,
+      "loss": 2.5894,
+      "step": 52500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.5209840972036096e-05,
+      "loss": 2.5782,
+      "step": 53000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.515422585544537e-05,
+      "loss": 2.5747,
+      "step": 53500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.5098387384972754e-05,
+      "loss": 2.5671,
+      "step": 54000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.504277226838203e-05,
+      "loss": 2.565,
+      "step": 54500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.498693379790941e-05,
+      "loss": 2.5533,
+      "step": 55000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.493131868131869e-05,
+      "loss": 2.5541,
+      "step": 55500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.487548021084606e-05,
+      "loss": 2.5446,
+      "step": 56000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.4819865094255345e-05,
+      "loss": 2.5328,
+      "step": 56500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.476402662378272e-05,
+      "loss": 2.5298,
+      "step": 57000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.4708411507192e-05,
+      "loss": 2.5278,
+      "step": 57500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.465257303671938e-05,
+      "loss": 2.5095,
+      "step": 58000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.4596957920128654e-05,
+      "loss": 2.5112,
+      "step": 58500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.454111944965604e-05,
+      "loss": 2.5037,
+      "step": 59000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.448550433306531e-05,
+      "loss": 2.5049,
+      "step": 59500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.4429665862592695e-05,
+      "loss": 2.5006,
+      "step": 60000
+    },
+    {
+      "epoch": 0.66,
+      "eval_accuracy": 0.561378824934529,
+      "eval_loss": 2.4140625,
+      "eval_runtime": 1237.7775,
+      "eval_samples_per_second": 249.121,
+      "eval_steps_per_second": 3.893,
+      "step": 60000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.437405074600196e-05,
+      "loss": 2.4905,
+      "step": 60500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.431821227552935e-05,
+      "loss": 2.4854,
+      "step": 61000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.426259715893862e-05,
+      "loss": 2.4858,
+      "step": 61500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.420675868846601e-05,
+      "loss": 2.472,
+      "step": 62000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.415114357187528e-05,
+      "loss": 2.4715,
+      "step": 62500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.409530510140266e-05,
+      "loss": 2.4633,
+      "step": 63000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.403968998481194e-05,
+      "loss": 2.4549,
+      "step": 63500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.398385151433932e-05,
+      "loss": 2.4558,
+      "step": 64000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3928236397748595e-05,
+      "loss": 2.4529,
+      "step": 64500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.387239792727598e-05,
+      "loss": 2.4473,
+      "step": 65000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.3816782810685253e-05,
+      "loss": 2.4402,
+      "step": 65500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.376094434021264e-05,
+      "loss": 2.441,
+      "step": 66000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.370532922362191e-05,
+      "loss": 2.4376,
+      "step": 66500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.3649490753149295e-05,
+      "loss": 2.4264,
+      "step": 67000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.359387563655857e-05,
+      "loss": 2.4239,
+      "step": 67500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.3538037166085946e-05,
+      "loss": 2.4179,
+      "step": 68000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.348242204949523e-05,
+      "loss": 2.4189,
+      "step": 68500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3426583579022604e-05,
+      "loss": 2.416,
+      "step": 69000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.337096846243188e-05,
+      "loss": 2.4088,
+      "step": 69500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.331512999195926e-05,
+      "loss": 2.4062,
+      "step": 70000
+    },
+    {
+      "epoch": 0.76,
+      "eval_accuracy": 0.5734350887860105,
+      "eval_loss": 2.32421875,
+      "eval_runtime": 1141.7265,
+      "eval_samples_per_second": 270.079,
+      "eval_steps_per_second": 4.221,
+      "step": 70000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.325951487536854e-05,
+      "loss": 2.4003,
+      "step": 70500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.320367640489592e-05,
+      "loss": 2.3966,
+      "step": 71000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.3148061288305195e-05,
+      "loss": 2.3965,
+      "step": 71500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.309222281783258e-05,
+      "loss": 2.389,
+      "step": 72000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.3036607701241846e-05,
+      "loss": 2.3833,
+      "step": 72500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2980769230769236e-05,
+      "loss": 2.3841,
+      "step": 73000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2925154114178504e-05,
+      "loss": 2.3791,
+      "step": 73500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.286931564370589e-05,
+      "loss": 2.378,
+      "step": 74000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.281370052711516e-05,
+      "loss": 2.3719,
+      "step": 74500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2757862056642545e-05,
+      "loss": 2.3649,
+      "step": 75000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.270224694005182e-05,
+      "loss": 2.3667,
+      "step": 75500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.2646408469579203e-05,
+      "loss": 2.3602,
+      "step": 76000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.259079335298848e-05,
+      "loss": 2.3529,
+      "step": 76500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.253495488251586e-05,
+      "loss": 2.3573,
+      "step": 77000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.2479339765925136e-05,
+      "loss": 2.3537,
+      "step": 77500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.242350129545252e-05,
+      "loss": 2.3471,
+      "step": 78000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.2367886178861794e-05,
+      "loss": 2.3454,
+      "step": 78500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.231204770838918e-05,
+      "loss": 2.3423,
+      "step": 79000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2256432591798446e-05,
+      "loss": 2.3349,
+      "step": 79500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.220059412132583e-05,
+      "loss": 2.3338,
+      "step": 80000
+    },
+    {
+      "epoch": 0.87,
+      "eval_accuracy": 0.5822560488428162,
+      "eval_loss": 2.25390625,
+      "eval_runtime": 1241.5767,
+      "eval_samples_per_second": 248.358,
+      "eval_steps_per_second": 3.881,
+      "step": 80000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2144979004735104e-05,
+      "loss": 2.3318,
+      "step": 80500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.208914053426249e-05,
+      "loss": 2.3315,
+      "step": 81000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.203352541767176e-05,
+      "loss": 2.3286,
+      "step": 81500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.1977686947199145e-05,
+      "loss": 2.3203,
+      "step": 82000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.192207183060842e-05,
+      "loss": 2.3236,
+      "step": 82500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.18662333601358e-05,
+      "loss": 2.3147,
+      "step": 83000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.181061824354508e-05,
+      "loss": 2.3194,
+      "step": 83500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.1754779773072454e-05,
+      "loss": 2.3037,
+      "step": 84000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.169916465648173e-05,
+      "loss": 2.3121,
+      "step": 84500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.164332618600911e-05,
+      "loss": 2.302,
+      "step": 85000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.158771106941839e-05,
+      "loss": 2.3057,
+      "step": 85500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.153187259894577e-05,
+      "loss": 2.2998,
+      "step": 86000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.1476257482355045e-05,
+      "loss": 2.3006,
+      "step": 86500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.142041901188243e-05,
+      "loss": 2.2962,
+      "step": 87000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.13648038952917e-05,
+      "loss": 2.2941,
+      "step": 87500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1308965424819086e-05,
+      "loss": 2.2915,
+      "step": 88000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.125335030822836e-05,
+      "loss": 2.2857,
+      "step": 88500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1197511837755744e-05,
+      "loss": 2.2876,
+      "step": 89000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.114189672116502e-05,
+      "loss": 2.2776,
+      "step": 89500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.10860582506924e-05,
+      "loss": 2.2838,
+      "step": 90000
+    },
+    {
+      "epoch": 0.98,
+      "eval_accuracy": 0.5894123096001136,
+      "eval_loss": 2.201171875,
+      "eval_runtime": 1183.2684,
+      "eval_samples_per_second": 260.597,
+      "eval_steps_per_second": 4.073,
+      "step": 90000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.103044313410167e-05,
+      "loss": 2.2785,
+      "step": 90500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.097460466362906e-05,
+      "loss": 2.2731,
+      "step": 91000
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.091898954703833e-05,
+      "loss": 2.2686,
+      "step": 91500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.086315107656571e-05,
+      "loss": 2.2635,
+      "step": 92000
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.0807535959974986e-05,
+      "loss": 2.2623,
+      "step": 92500
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.075169748950237e-05,
+      "loss": 2.2522,
+      "step": 93000
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.0696082372911645e-05,
+      "loss": 2.2606,
+      "step": 93500
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.064024390243902e-05,
+      "loss": 2.2587,
+      "step": 94000
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.05846287858483e-05,
+      "loss": 2.258,
+      "step": 94500
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.052879031537568e-05,
+      "loss": 2.2499,
+      "step": 95000
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.047317519878496e-05,
+      "loss": 2.2513,
+      "step": 95500
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.041733672831234e-05,
+      "loss": 2.2464,
+      "step": 96000
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.036172161172161e-05,
+      "loss": 2.2443,
+      "step": 96500
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.0305883141248995e-05,
+      "loss": 2.2454,
+      "step": 97000
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.025026802465827e-05,
+      "loss": 2.2452,
+      "step": 97500
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.019442955418565e-05,
+      "loss": 2.2395,
+      "step": 98000
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.013881443759493e-05,
+      "loss": 2.2376,
+      "step": 98500
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.008297596712231e-05,
+      "loss": 2.2295,
+      "step": 99000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.0027360850531586e-05,
+      "loss": 2.2414,
+      "step": 99500
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.997152238005897e-05,
+      "loss": 2.231,
+      "step": 100000
+    },
+    {
+      "epoch": 1.09,
+      "eval_accuracy": 0.5959185956012876,
+      "eval_loss": 2.150390625,
+      "eval_runtime": 1176.5848,
+      "eval_samples_per_second": 262.077,
+      "eval_steps_per_second": 4.096,
+      "step": 100000
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.991590726346824e-05,
+      "loss": 2.2314,
+      "step": 100500
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.986006879299563e-05,
+      "loss": 2.2268,
+      "step": 101000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.9804453676404895e-05,
+      "loss": 2.231,
+      "step": 101500
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.9748615205932285e-05,
+      "loss": 2.2243,
+      "step": 102000
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.969300008934155e-05,
+      "loss": 2.2207,
+      "step": 102500
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.963716161886894e-05,
+      "loss": 2.2192,
+      "step": 103000
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.958154650227821e-05,
+      "loss": 2.2114,
+      "step": 103500
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.9525708031805594e-05,
+      "loss": 2.2099,
+      "step": 104000
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.947009291521487e-05,
+      "loss": 2.2115,
+      "step": 104500
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.9414254444742246e-05,
+      "loss": 2.2105,
+      "step": 105000
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.935863932815153e-05,
+      "loss": 2.2079,
+      "step": 105500
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.9302800857678904e-05,
+      "loss": 2.2013,
+      "step": 106000
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.9247185741088185e-05,
+      "loss": 2.2021,
+      "step": 106500
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.919134727061556e-05,
+      "loss": 2.2035,
+      "step": 107000
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.9135732154024843e-05,
+      "loss": 2.1966,
+      "step": 107500
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.907989368355222e-05,
+      "loss": 2.2014,
+      "step": 108000
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.9024278566961495e-05,
+      "loss": 2.1986,
+      "step": 108500
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.896844009648888e-05,
+      "loss": 2.1917,
+      "step": 109000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.891282497989815e-05,
+      "loss": 2.1919,
+      "step": 109500
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.8856986509425536e-05,
+      "loss": 2.1903,
+      "step": 110000
+    },
+    {
+      "epoch": 1.2,
+      "eval_accuracy": 0.6009023831034535,
+      "eval_loss": 2.11328125,
+      "eval_runtime": 1234.1136,
+      "eval_samples_per_second": 249.86,
+      "eval_steps_per_second": 3.905,
+      "step": 110000
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.8801371392834804e-05,
+      "loss": 2.189,
+      "step": 110500
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.8745532922362194e-05,
+      "loss": 2.1858,
+      "step": 111000
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.868991780577146e-05,
+      "loss": 2.1867,
+      "step": 111500
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.863407933529885e-05,
+      "loss": 2.1871,
+      "step": 112000
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.857846421870812e-05,
+      "loss": 2.1823,
+      "step": 112500
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.852262574823551e-05,
+      "loss": 2.1827,
+      "step": 113000
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.846701063164478e-05,
+      "loss": 2.1797,
+      "step": 113500
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.841117216117217e-05,
+      "loss": 2.1761,
+      "step": 114000
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.8355557044581436e-05,
+      "loss": 2.1804,
+      "step": 114500
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.829971857410882e-05,
+      "loss": 2.1727,
+      "step": 115000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.8244103457518094e-05,
+      "loss": 2.1697,
+      "step": 115500
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.818826498704548e-05,
+      "loss": 2.1735,
+      "step": 116000
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.813264987045475e-05,
+      "loss": 2.1701,
+      "step": 116500
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.807681139998213e-05,
+      "loss": 2.1679,
+      "step": 117000
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.802119628339141e-05,
+      "loss": 2.1651,
+      "step": 117500
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.7965357812918787e-05,
+      "loss": 2.1638,
+      "step": 118000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.790974269632807e-05,
+      "loss": 2.1653,
+      "step": 118500
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.7853904225855445e-05,
+      "loss": 2.1604,
+      "step": 119000
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.7798289109264726e-05,
+      "loss": 2.1616,
+      "step": 119500
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.77424506387921e-05,
+      "loss": 2.1594,
+      "step": 120000
+    },
+    {
+      "epoch": 1.31,
+      "eval_accuracy": 0.6053601315304664,
+      "eval_loss": 2.080078125,
+      "eval_runtime": 1170.3916,
+      "eval_samples_per_second": 263.464,
+      "eval_steps_per_second": 4.117,
+      "step": 120000
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.768683552220138e-05,
+      "loss": 2.1564,
+      "step": 120500
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.763099705172876e-05,
+      "loss": 2.1504,
+      "step": 121000
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.757538193513803e-05,
+      "loss": 2.1511,
+      "step": 121500
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.751954346466542e-05,
+      "loss": 2.1462,
+      "step": 122000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.746392834807469e-05,
+      "loss": 2.1409,
+      "step": 122500
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.740808987760208e-05,
+      "loss": 2.1509,
+      "step": 123000
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.7352474761011345e-05,
+      "loss": 2.1453,
+      "step": 123500
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.7296636290538735e-05,
+      "loss": 2.1465,
+      "step": 124000
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.7241021173948e-05,
+      "loss": 2.141,
+      "step": 124500
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.718518270347539e-05,
+      "loss": 2.144,
+      "step": 125000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.712956758688466e-05,
+      "loss": 2.1372,
+      "step": 125500
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.7073729116412044e-05,
+      "loss": 2.14,
+      "step": 126000
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.701811399982132e-05,
+      "loss": 2.1332,
+      "step": 126500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.69622755293487e-05,
+      "loss": 2.1398,
+      "step": 127000
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.690666041275798e-05,
+      "loss": 2.13,
+      "step": 127500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.685082194228536e-05,
+      "loss": 2.1305,
+      "step": 128000
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.6795206825694635e-05,
+      "loss": 2.1296,
+      "step": 128500
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.673936835522201e-05,
+      "loss": 2.1309,
+      "step": 129000
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.668375323863129e-05,
+      "loss": 2.1274,
+      "step": 129500
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 3.662791476815867e-05,
+      "loss": 2.1307,
+      "step": 130000
+    },
+    {
+      "epoch": 1.42,
+      "eval_accuracy": 0.609508441276654,
+      "eval_loss": 2.048828125,
+      "eval_runtime": 1327.7196,
+      "eval_samples_per_second": 232.245,
+      "eval_steps_per_second": 3.63,
+      "step": 130000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.657229965156795e-05,
+      "loss": 2.123,
+      "step": 130500
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.651646118109533e-05,
+      "loss": 2.1239,
+      "step": 131000
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.64608460645046e-05,
+      "loss": 2.1248,
+      "step": 131500
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.6405007594031985e-05,
+      "loss": 2.1189,
+      "step": 132000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.634939247744126e-05,
+      "loss": 2.1214,
+      "step": 132500
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.6293554006968643e-05,
+      "loss": 2.1213,
+      "step": 133000
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.623793889037791e-05,
+      "loss": 2.1205,
+      "step": 133500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.61821004199053e-05,
+      "loss": 2.1236,
+      "step": 134000
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.612648530331457e-05,
+      "loss": 2.1142,
+      "step": 134500
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.607064683284196e-05,
+      "loss": 2.1223,
+      "step": 135000
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.601503171625123e-05,
+      "loss": 2.11,
+      "step": 135500
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.595919324577861e-05,
+      "loss": 2.1085,
+      "step": 136000
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.5903578129187886e-05,
+      "loss": 2.1053,
+      "step": 136500
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.584773965871527e-05,
+      "loss": 2.1034,
+      "step": 137000
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.5792124542124544e-05,
+      "loss": 2.1058,
+      "step": 137500
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.573628607165193e-05,
+      "loss": 2.1057,
+      "step": 138000
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.56806709550612e-05,
+      "loss": 2.1048,
+      "step": 138500
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.5624832484588585e-05,
+      "loss": 2.1043,
+      "step": 139000
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.556921736799786e-05,
+      "loss": 2.1041,
+      "step": 139500
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.551337889752524e-05,
+      "loss": 2.0948,
+      "step": 140000
+    },
+    {
+      "epoch": 1.53,
+      "eval_accuracy": 0.6133498725939791,
+      "eval_loss": 2.0234375,
+      "eval_runtime": 1220.9619,
+      "eval_samples_per_second": 252.552,
+      "eval_steps_per_second": 3.947,
+      "step": 140000
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.545776378093452e-05,
+      "loss": 2.0964,
+      "step": 140500
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.5401925310461894e-05,
+      "loss": 2.1022,
+      "step": 141000
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.5346310193871176e-05,
+      "loss": 2.0929,
+      "step": 141500
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.529047172339855e-05,
+      "loss": 2.0938,
+      "step": 142000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.523485660680783e-05,
+      "loss": 2.0947,
+      "step": 142500
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.517901813633521e-05,
+      "loss": 2.0924,
+      "step": 143000
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.5123403019744485e-05,
+      "loss": 2.096,
+      "step": 143500
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.506756454927187e-05,
+      "loss": 2.0918,
+      "step": 144000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.501194943268114e-05,
+      "loss": 2.0927,
+      "step": 144500
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.4956110962208526e-05,
+      "loss": 2.0903,
+      "step": 145000
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.4900495845617794e-05,
+      "loss": 2.0883,
+      "step": 145500
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.4844657375145184e-05,
+      "loss": 2.0866,
+      "step": 146000
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.478904225855445e-05,
+      "loss": 2.0836,
+      "step": 146500
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.4733203788081836e-05,
+      "loss": 2.0867,
+      "step": 147000
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.467758867149111e-05,
+      "loss": 2.0763,
+      "step": 147500
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.4621750201018494e-05,
+      "loss": 2.0821,
+      "step": 148000
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.456613508442777e-05,
+      "loss": 2.0852,
+      "step": 148500
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.451029661395515e-05,
+      "loss": 2.081,
+      "step": 149000
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.4454681497364427e-05,
+      "loss": 2.0729,
+      "step": 149500
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.439884302689181e-05,
+      "loss": 2.0748,
+      "step": 150000
+    },
+    {
+      "epoch": 1.64,
+      "eval_accuracy": 0.6168509909854293,
+      "eval_loss": 1.998046875,
+      "eval_runtime": 1143.5239,
+      "eval_samples_per_second": 269.654,
+      "eval_steps_per_second": 4.214,
+      "step": 150000
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.4343227910301085e-05,
+      "loss": 2.0759,
+      "step": 150500
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.428738943982847e-05,
+      "loss": 2.0725,
+      "step": 151000
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.423177432323774e-05,
+      "loss": 2.0669,
+      "step": 151500
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4175935852765126e-05,
+      "loss": 2.0715,
+      "step": 152000
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4120320736174394e-05,
+      "loss": 2.0689,
+      "step": 152500
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.406448226570178e-05,
+      "loss": 2.0695,
+      "step": 153000
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.400886714911105e-05,
+      "loss": 2.0675,
+      "step": 153500
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3953028678638435e-05,
+      "loss": 2.0664,
+      "step": 154000
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.389741356204771e-05,
+      "loss": 2.0692,
+      "step": 154500
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.384157509157509e-05,
+      "loss": 2.0679,
+      "step": 155000
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.378595997498437e-05,
+      "loss": 2.0619,
+      "step": 155500
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.373012150451175e-05,
+      "loss": 2.0656,
+      "step": 156000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.3674506387921026e-05,
+      "loss": 2.0608,
+      "step": 156500
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.36186679174484e-05,
+      "loss": 2.056,
+      "step": 157000
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.356305280085768e-05,
+      "loss": 2.0616,
+      "step": 157500
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.350721433038506e-05,
+      "loss": 2.0541,
+      "step": 158000
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.3451599213794335e-05,
+      "loss": 2.0608,
+      "step": 158500
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.339576074332172e-05,
+      "loss": 2.0552,
+      "step": 159000
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.334014562673099e-05,
+      "loss": 2.0551,
+      "step": 159500
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.3284307156258376e-05,
+      "loss": 2.0572,
+      "step": 160000
+    },
+    {
+      "epoch": 1.75,
+      "eval_accuracy": 0.6195138509619146,
+      "eval_loss": 1.9755859375,
+      "eval_runtime": 1218.1327,
+      "eval_samples_per_second": 253.138,
+      "eval_steps_per_second": 3.956,
+      "step": 160000
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.322869203966765e-05,
+      "loss": 2.0571,
+      "step": 160500
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.3172853569195035e-05,
+      "loss": 2.0501,
+      "step": 161000
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.311723845260431e-05,
+      "loss": 2.0531,
+      "step": 161500
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.306139998213169e-05,
+      "loss": 2.0503,
+      "step": 162000
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.300578486554097e-05,
+      "loss": 2.0486,
+      "step": 162500
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.294994639506835e-05,
+      "loss": 2.0453,
+      "step": 163000
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.289433127847762e-05,
+      "loss": 2.0495,
+      "step": 163500
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.283849280800501e-05,
+      "loss": 2.0512,
+      "step": 164000
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.278287769141428e-05,
+      "loss": 2.0421,
+      "step": 164500
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.272703922094166e-05,
+      "loss": 2.0428,
+      "step": 165000
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.2671424104350935e-05,
+      "loss": 2.0441,
+      "step": 165500
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.261558563387832e-05,
+      "loss": 2.0437,
+      "step": 166000
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.255997051728759e-05,
+      "loss": 2.0416,
+      "step": 166500
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.2504132046814976e-05,
+      "loss": 2.0395,
+      "step": 167000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.244851693022425e-05,
+      "loss": 2.0431,
+      "step": 167500
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.239267845975163e-05,
+      "loss": 2.0385,
+      "step": 168000
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.233706334316091e-05,
+      "loss": 2.0385,
+      "step": 168500
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.2281224872688285e-05,
+      "loss": 2.0369,
+      "step": 169000
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.222560975609756e-05,
+      "loss": 2.0395,
+      "step": 169500
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.216977128562494e-05,
+      "loss": 2.0359,
+      "step": 170000
+    },
+    {
+      "epoch": 1.86,
+      "eval_accuracy": 0.6225490703978083,
+      "eval_loss": 1.955078125,
+      "eval_runtime": 1231.0639,
+      "eval_samples_per_second": 250.479,
+      "eval_steps_per_second": 3.915,
+      "step": 170000
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.211415616903422e-05,
+      "loss": 2.0289,
+      "step": 170500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.20583176985616e-05,
+      "loss": 2.0304,
+      "step": 171000
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.2002702581970876e-05,
+      "loss": 2.0313,
+      "step": 171500
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.194686411149826e-05,
+      "loss": 2.0323,
+      "step": 172000
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.1891248994907534e-05,
+      "loss": 2.0289,
+      "step": 172500
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.183541052443492e-05,
+      "loss": 2.0291,
+      "step": 173000
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.1779795407844185e-05,
+      "loss": 2.0292,
+      "step": 173500
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.1723956937371575e-05,
+      "loss": 2.0294,
+      "step": 174000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.1668341820780843e-05,
+      "loss": 2.0227,
+      "step": 174500
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.1612503350308233e-05,
+      "loss": 2.0217,
+      "step": 175000
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.15568882337175e-05,
+      "loss": 2.0263,
+      "step": 175500
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.150104976324489e-05,
+      "loss": 2.0237,
+      "step": 176000
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.144543464665416e-05,
+      "loss": 2.0225,
+      "step": 176500
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.138959617618154e-05,
+      "loss": 2.0178,
+      "step": 177000
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.133398105959082e-05,
+      "loss": 2.0148,
+      "step": 177500
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.12781425891182e-05,
+      "loss": 2.0175,
+      "step": 178000
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.1222527472527476e-05,
+      "loss": 2.0107,
+      "step": 178500
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.116668900205485e-05,
+      "loss": 2.0172,
+      "step": 179000
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.1111073885464134e-05,
+      "loss": 2.0201,
+      "step": 179500
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.105523541499151e-05,
+      "loss": 2.0148,
+      "step": 180000
+    },
+    {
+      "epoch": 1.97,
+      "eval_accuracy": 0.6250524947954239,
+      "eval_loss": 1.9384765625,
+      "eval_runtime": 1280.5256,
+      "eval_samples_per_second": 240.804,
+      "eval_steps_per_second": 3.763,
+      "step": 180000
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.099962029840079e-05,
+      "loss": 2.0144,
+      "step": 180500
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.094378182792817e-05,
+      "loss": 2.0074,
+      "step": 181000
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.088816671133744e-05,
+      "loss": 2.0143,
+      "step": 181500
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.0832328240864826e-05,
+      "loss": 2.0134,
+      "step": 182000
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.07767131242741e-05,
+      "loss": 2.0085,
+      "step": 182500
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.0720874653801484e-05,
+      "loss": 2.0101,
+      "step": 183000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.066525953721076e-05,
+      "loss": 2.004,
+      "step": 183500
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.060942106673814e-05,
+      "loss": 2.0083,
+      "step": 184000
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.055380595014741e-05,
+      "loss": 2.0067,
+      "step": 184500
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.04979674796748e-05,
+      "loss": 2.0051,
+      "step": 185000
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.044235236308407e-05,
+      "loss": 2.0021,
+      "step": 185500
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.0386513892611458e-05,
+      "loss": 2.0043,
+      "step": 186000
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.0330898776020726e-05,
+      "loss": 2.0063,
+      "step": 186500
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.0275060305548113e-05,
+      "loss": 1.9993,
+      "step": 187000
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.0219445188957384e-05,
+      "loss": 1.9985,
+      "step": 187500
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.0163606718484767e-05,
+      "loss": 1.998,
+      "step": 188000
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.0107991601894042e-05,
+      "loss": 1.9991,
+      "step": 188500
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.0052153131421422e-05,
+      "loss": 1.9955,
+      "step": 189000
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.99965380148307e-05,
+      "loss": 1.9956,
+      "step": 189500
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.994069954435808e-05,
+      "loss": 1.9994,
+      "step": 190000
+    },
+    {
+      "epoch": 2.08,
+      "eval_accuracy": 0.6274383613077523,
+      "eval_loss": 1.921875,
+      "eval_runtime": 1238.0942,
+      "eval_samples_per_second": 249.057,
+      "eval_steps_per_second": 3.892,
+      "step": 190000
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.988508442776736e-05,
+      "loss": 1.9938,
+      "step": 190500
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.9829245957294738e-05,
+      "loss": 1.9881,
+      "step": 191000
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.9773630840704013e-05,
+      "loss": 1.9959,
+      "step": 191500
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.9717792370231396e-05,
+      "loss": 1.9872,
+      "step": 192000
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.966217725364067e-05,
+      "loss": 1.9908,
+      "step": 192500
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.9606338783168054e-05,
+      "loss": 1.9894,
+      "step": 193000
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.955072366657733e-05,
+      "loss": 1.9939,
+      "step": 193500
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.949488519610471e-05,
+      "loss": 1.995,
+      "step": 194000
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.943927007951398e-05,
+      "loss": 1.9915,
+      "step": 194500
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.9383431609041367e-05,
+      "loss": 1.9917,
+      "step": 195000
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.932781649245064e-05,
+      "loss": 1.9832,
+      "step": 195500
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.9271978021978025e-05,
+      "loss": 1.9897,
+      "step": 196000
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.9216362905387296e-05,
+      "loss": 1.9838,
+      "step": 196500
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.9160524434914683e-05,
+      "loss": 1.986,
+      "step": 197000
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.9104909318323954e-05,
+      "loss": 1.9864,
+      "step": 197500
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.904907084785134e-05,
+      "loss": 1.9837,
+      "step": 198000
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.899345573126061e-05,
+      "loss": 1.9817,
+      "step": 198500
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.8937617260787992e-05,
+      "loss": 1.9793,
+      "step": 199000
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.8882002144197267e-05,
+      "loss": 1.9817,
+      "step": 199500
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.882616367372465e-05,
+      "loss": 1.9769,
+      "step": 200000
+    },
+    {
+      "epoch": 2.18,
+      "eval_accuracy": 0.629700250907056,
+      "eval_loss": 1.904296875,
+      "eval_runtime": 1276.1988,
+      "eval_samples_per_second": 241.621,
+      "eval_steps_per_second": 3.776,
+      "step": 200000
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.8770548557133925e-05,
+      "loss": 1.9835,
+      "step": 200500
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.8714710086661305e-05,
+      "loss": 1.9843,
+      "step": 201000
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.8659094970070583e-05,
+      "loss": 1.9766,
+      "step": 201500
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.8603256499597963e-05,
+      "loss": 1.9834,
+      "step": 202000
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.854764138300724e-05,
+      "loss": 1.9767,
+      "step": 202500
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.849180291253462e-05,
+      "loss": 1.9767,
+      "step": 203000
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.8436187795943896e-05,
+      "loss": 1.975,
+      "step": 203500
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.838034932547128e-05,
+      "loss": 1.9758,
+      "step": 204000
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.832473420888055e-05,
+      "loss": 1.9763,
+      "step": 204500
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.8268895738407937e-05,
+      "loss": 1.969,
+      "step": 205000
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.8213280621817205e-05,
+      "loss": 1.9721,
+      "step": 205500
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8157442151344592e-05,
+      "loss": 1.9696,
+      "step": 206000
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8101827034753863e-05,
+      "loss": 1.9675,
+      "step": 206500
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.804598856428125e-05,
+      "loss": 1.9738,
+      "step": 207000
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.799037344769052e-05,
+      "loss": 1.9686,
+      "step": 207500
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7934534977217908e-05,
+      "loss": 1.9699,
+      "step": 208000
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.787891986062718e-05,
+      "loss": 1.9712,
+      "step": 208500
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.782308139015456e-05,
+      "loss": 1.9679,
+      "step": 209000
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.7767466273563837e-05,
+      "loss": 1.9679,
+      "step": 209500
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.7711627803091217e-05,
+      "loss": 1.9705,
+      "step": 210000
+    },
+    {
+      "epoch": 2.29,
+      "eval_accuracy": 0.6316547044841081,
+      "eval_loss": 1.8916015625,
+      "eval_runtime": 1400.444,
+      "eval_samples_per_second": 220.184,
+      "eval_steps_per_second": 3.441,
+      "step": 210000
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.7656012686500492e-05,
+      "loss": 1.9671,
+      "step": 210500
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.7600174216027875e-05,
+      "loss": 1.9677,
+      "step": 211000
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.754455909943715e-05,
+      "loss": 1.9629,
+      "step": 211500
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.7488720628964533e-05,
+      "loss": 1.9631,
+      "step": 212000
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.7433105512373808e-05,
+      "loss": 1.967,
+      "step": 212500
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.7377267041901188e-05,
+      "loss": 1.9629,
+      "step": 213000
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.7321651925310466e-05,
+      "loss": 1.9636,
+      "step": 213500
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.7265813454837846e-05,
+      "loss": 1.9572,
+      "step": 214000
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.7210198338247124e-05,
+      "loss": 1.9611,
+      "step": 214500
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.7154359867774504e-05,
+      "loss": 1.9565,
+      "step": 215000
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.7098744751183775e-05,
+      "loss": 1.9583,
+      "step": 215500
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.7042906280711162e-05,
+      "loss": 1.9562,
+      "step": 216000
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.6987291164120433e-05,
+      "loss": 1.9561,
+      "step": 216500
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.693145269364782e-05,
+      "loss": 1.9557,
+      "step": 217000
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.6875837577057088e-05,
+      "loss": 1.9502,
+      "step": 217500
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.6819999106584475e-05,
+      "loss": 1.9551,
+      "step": 218000
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.6764383989993746e-05,
+      "loss": 1.9553,
+      "step": 218500
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.6708545519521133e-05,
+      "loss": 1.9523,
+      "step": 219000
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.6652930402930404e-05,
+      "loss": 1.9503,
+      "step": 219500
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.6597091932457784e-05,
+      "loss": 1.9557,
+      "step": 220000
+    },
+    {
+      "epoch": 2.4,
+      "eval_accuracy": 0.6337785581735893,
+      "eval_loss": 1.8779296875,
+      "eval_runtime": 1159.2961,
+      "eval_samples_per_second": 265.986,
+      "eval_steps_per_second": 4.157,
+      "step": 220000
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.6541476815867062e-05,
+      "loss": 1.9482,
+      "step": 220500
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.6485638345394442e-05,
+      "loss": 1.9543,
+      "step": 221000
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.643002322880372e-05,
+      "loss": 1.9518,
+      "step": 221500
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.63741847583311e-05,
+      "loss": 1.9449,
+      "step": 222000
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.6318569641740375e-05,
+      "loss": 1.9454,
+      "step": 222500
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6262731171267758e-05,
+      "loss": 1.9453,
+      "step": 223000
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.6207116054677033e-05,
+      "loss": 1.9497,
+      "step": 223500
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.6151277584204416e-05,
+      "loss": 1.9492,
+      "step": 224000
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.609566246761369e-05,
+      "loss": 1.9471,
+      "step": 224500
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.603982399714107e-05,
+      "loss": 1.9475,
+      "step": 225000
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.5984208880550342e-05,
+      "loss": 1.9405,
+      "step": 225500
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.592837041007773e-05,
+      "loss": 1.9433,
+      "step": 226000
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.5872755293487e-05,
+      "loss": 1.9449,
+      "step": 226500
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.5816916823014387e-05,
+      "loss": 1.9424,
+      "step": 227000
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.5761301706423658e-05,
+      "loss": 1.9457,
+      "step": 227500
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.5705463235951045e-05,
+      "loss": 1.9398,
+      "step": 228000
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.5649848119360316e-05,
+      "loss": 1.9355,
+      "step": 228500
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.5594009648887703e-05,
+      "loss": 1.9387,
+      "step": 229000
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.553839453229697e-05,
+      "loss": 1.9432,
+      "step": 229500
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.5482556061824354e-05,
+      "loss": 1.9407,
+      "step": 230000
+    },
+    {
+      "epoch": 2.51,
+      "eval_accuracy": 0.6353719982942662,
+      "eval_loss": 1.8642578125,
+      "eval_runtime": 1486.9925,
+      "eval_samples_per_second": 207.369,
+      "eval_steps_per_second": 3.241,
+      "step": 230000
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.542694094523363e-05,
+      "loss": 1.9397,
+      "step": 230500
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.5371102474761012e-05,
+      "loss": 1.9367,
+      "step": 231000
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.5315487358170287e-05,
+      "loss": 1.9436,
+      "step": 231500
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.5259648887697667e-05,
+      "loss": 1.9399,
+      "step": 232000
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.5204033771106945e-05,
+      "loss": 1.9365,
+      "step": 232500
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.5148195300634325e-05,
+      "loss": 1.9371,
+      "step": 233000
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.5092580184043603e-05,
+      "loss": 1.9315,
+      "step": 233500
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.5036741713570983e-05,
+      "loss": 1.9344,
+      "step": 234000
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.4981126596980258e-05,
+      "loss": 1.9403,
+      "step": 234500
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.492528812650764e-05,
+      "loss": 1.9395,
+      "step": 235000
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.4869673009916912e-05,
+      "loss": 1.938,
+      "step": 235500
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.48138345394443e-05,
+      "loss": 1.9306,
+      "step": 236000
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.475821942285357e-05,
+      "loss": 1.93,
+      "step": 236500
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.4702380952380953e-05,
+      "loss": 1.9341,
+      "step": 237000
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.4646765835790228e-05,
+      "loss": 1.9329,
+      "step": 237500
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.4590927365317608e-05,
+      "loss": 1.9327,
+      "step": 238000
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.4535312248726883e-05,
+      "loss": 1.9281,
+      "step": 238500
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.4479473778254266e-05,
+      "loss": 1.9326,
+      "step": 239000
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.442385866166354e-05,
+      "loss": 1.927,
+      "step": 239500
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.4368020191190924e-05,
+      "loss": 1.9307,
+      "step": 240000
+    },
+    {
+      "epoch": 2.62,
+      "eval_accuracy": 0.6371893215095429,
+      "eval_loss": 1.8525390625,
+      "eval_runtime": 1314.8093,
+      "eval_samples_per_second": 234.525,
+      "eval_steps_per_second": 3.665,
+      "step": 240000
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.43124050746002e-05,
+      "loss": 1.9308,
+      "step": 240500
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4256566604127582e-05,
+      "loss": 1.9299,
+      "step": 241000
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4200951487536854e-05,
+      "loss": 1.9282,
+      "step": 241500
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.414511301706424e-05,
+      "loss": 1.9239,
+      "step": 242000
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.408949790047351e-05,
+      "loss": 1.9211,
+      "step": 242500
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.4033659430000895e-05,
+      "loss": 1.9273,
+      "step": 243000
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.397804431341017e-05,
+      "loss": 1.9244,
+      "step": 243500
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.392220584293755e-05,
+      "loss": 1.9234,
+      "step": 244000
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.3866590726346824e-05,
+      "loss": 1.9232,
+      "step": 244500
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.3810752255874208e-05,
+      "loss": 1.9286,
+      "step": 245000
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.3755137139283482e-05,
+      "loss": 1.9197,
+      "step": 245500
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.3699298668810866e-05,
+      "loss": 1.9198,
+      "step": 246000
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.364368355222014e-05,
+      "loss": 1.9207,
+      "step": 246500
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.3587845081747524e-05,
+      "loss": 1.9256,
+      "step": 247000
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.3532229965156795e-05,
+      "loss": 1.9135,
+      "step": 247500
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.347639149468418e-05,
+      "loss": 1.9166,
+      "step": 248000
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.3420776378093453e-05,
+      "loss": 1.9219,
+      "step": 248500
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.3364937907620836e-05,
+      "loss": 1.9193,
+      "step": 249000
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.3309322791030108e-05,
+      "loss": 1.9187,
+      "step": 249500
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.325348432055749e-05,
+      "loss": 1.9186,
+      "step": 250000
+    },
+    {
+      "epoch": 2.73,
+      "eval_accuracy": 0.6388078472748526,
+      "eval_loss": 1.8408203125,
+      "eval_runtime": 1109.3427,
+      "eval_samples_per_second": 277.963,
+      "eval_steps_per_second": 4.344,
+      "step": 250000
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.3197645850084877e-05,
+      "loss": 1.9138,
+      "step": 250500
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.314203073349415e-05,
+      "loss": 1.9199,
+      "step": 251000
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.3086192263021532e-05,
+      "loss": 1.9143,
+      "step": 251500
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.3030577146430807e-05,
+      "loss": 1.911,
+      "step": 252000
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.297473867595819e-05,
+      "loss": 1.9132,
+      "step": 252500
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.2919123559367465e-05,
+      "loss": 1.9185,
+      "step": 253000
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.2863285088894845e-05,
+      "loss": 1.9123,
+      "step": 253500
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.280766997230412e-05,
+      "loss": 1.9163,
+      "step": 254000
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.2751831501831503e-05,
+      "loss": 1.9149,
+      "step": 254500
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.2696216385240778e-05,
+      "loss": 1.9139,
+      "step": 255000
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.264037791476816e-05,
+      "loss": 1.9168,
+      "step": 255500
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2584762798177432e-05,
+      "loss": 1.9118,
+      "step": 256000
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.252892432770482e-05,
+      "loss": 1.9113,
+      "step": 256500
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.247330921111409e-05,
+      "loss": 1.9096,
+      "step": 257000
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.2417470740641474e-05,
+      "loss": 1.9077,
+      "step": 257500
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.236185562405075e-05,
+      "loss": 1.9144,
+      "step": 258000
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.2306017153578128e-05,
+      "loss": 1.9067,
+      "step": 258500
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.2250402036987403e-05,
+      "loss": 1.9135,
+      "step": 259000
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.2194563566514786e-05,
+      "loss": 1.9104,
+      "step": 259500
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.213894844992406e-05,
+      "loss": 1.9114,
+      "step": 260000
+    },
+    {
+      "epoch": 2.84,
+      "eval_accuracy": 0.6401361780429158,
+      "eval_loss": 1.83203125,
+      "eval_runtime": 1109.6609,
+      "eval_samples_per_second": 277.883,
+      "eval_steps_per_second": 4.343,
+      "step": 260000
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.2083109979451444e-05,
+      "loss": 1.905,
+      "step": 260500
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.202749486286072e-05,
+      "loss": 1.9056,
+      "step": 261000
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1971656392388102e-05,
+      "loss": 1.9107,
+      "step": 261500
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1916041275797374e-05,
+      "loss": 1.9076,
+      "step": 262000
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.186020280532476e-05,
+      "loss": 1.9032,
+      "step": 262500
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1804587688734032e-05,
+      "loss": 1.8933,
+      "step": 263000
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1748749218261415e-05,
+      "loss": 1.9051,
+      "step": 263500
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1693134101670686e-05,
+      "loss": 1.902,
+      "step": 264000
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.163729563119807e-05,
+      "loss": 1.898,
+      "step": 264500
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1581680514607344e-05,
+      "loss": 1.9062,
+      "step": 265000
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1525842044134728e-05,
+      "loss": 1.9064,
+      "step": 265500
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1470226927544002e-05,
+      "loss": 1.9038,
+      "step": 266000
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1414388457071386e-05,
+      "loss": 1.8987,
+      "step": 266500
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.135877334048066e-05,
+      "loss": 1.8978,
+      "step": 267000
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1302934870008044e-05,
+      "loss": 1.8959,
+      "step": 267500
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1247319753417315e-05,
+      "loss": 1.8968,
+      "step": 268000
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.11914812829447e-05,
+      "loss": 1.8985,
+      "step": 268500
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.113586616635397e-05,
+      "loss": 1.899,
+      "step": 269000
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1080027695881356e-05,
+      "loss": 1.8932,
+      "step": 269500
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1024412579290628e-05,
+      "loss": 1.896,
+      "step": 270000
+    },
+    {
+      "epoch": 2.95,
+      "eval_accuracy": 0.6418558502033883,
+      "eval_loss": 1.8212890625,
+      "eval_runtime": 1100.3529,
+      "eval_samples_per_second": 280.234,
+      "eval_steps_per_second": 4.38,
+      "step": 270000
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.096857410881801e-05,
+      "loss": 1.901,
+      "step": 270500
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.0912958992227286e-05,
+      "loss": 1.8951,
+      "step": 271000
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.085712052175467e-05,
+      "loss": 1.8921,
+      "step": 271500
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.0801505405163944e-05,
+      "loss": 1.8882,
+      "step": 272000
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.0745666934691327e-05,
+      "loss": 1.89,
+      "step": 272500
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.0690051818100602e-05,
+      "loss": 1.8966,
+      "step": 273000
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.0634213347627982e-05,
+      "loss": 1.8921,
+      "step": 273500
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.0578598231037257e-05,
+      "loss": 1.893,
+      "step": 274000
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.052275976056464e-05,
+      "loss": 1.892,
+      "step": 274500
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.046714464397391e-05,
+      "loss": 1.8935,
+      "step": 275000
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0411306173501298e-05,
+      "loss": 1.886,
+      "step": 275500
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.035569105691057e-05,
+      "loss": 1.8924,
+      "step": 276000
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0299852586437952e-05,
+      "loss": 1.8896,
+      "step": 276500
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0244237469847227e-05,
+      "loss": 1.887,
+      "step": 277000
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.018839899937461e-05,
+      "loss": 1.8863,
+      "step": 277500
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0132783882783885e-05,
+      "loss": 1.8886,
+      "step": 278000
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.007694541231127e-05,
+      "loss": 1.8867,
+      "step": 278500
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0021330295720543e-05,
+      "loss": 1.8913,
+      "step": 279000
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 1.9965491825247923e-05,
+      "loss": 1.8878,
+      "step": 279500
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 1.9909876708657198e-05,
+      "loss": 1.8857,
+      "step": 280000
+    },
+    {
+      "epoch": 3.06,
+      "eval_accuracy": 0.6433153812116338,
+      "eval_loss": 1.8115234375,
+      "eval_runtime": 1118.8118,
+      "eval_samples_per_second": 275.61,
+      "eval_steps_per_second": 4.307,
+      "step": 280000
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 1.985403823818458e-05,
+      "loss": 1.8914,
+      "step": 280500
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 1.9798423121593853e-05,
+      "loss": 1.8881,
+      "step": 281000
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 1.974258465112124e-05,
+      "loss": 1.8848,
+      "step": 281500
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 1.968696953453051e-05,
+      "loss": 1.8866,
+      "step": 282000
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 1.9631131064057894e-05,
+      "loss": 1.8775,
+      "step": 282500
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 1.957551594746717e-05,
+      "loss": 1.879,
+      "step": 283000
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 1.9519677476994552e-05,
+      "loss": 1.8776,
+      "step": 283500
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 1.9464062360403827e-05,
+      "loss": 1.8828,
+      "step": 284000
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 1.9408223889931207e-05,
+      "loss": 1.8768,
+      "step": 284500
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 1.935260877334048e-05,
+      "loss": 1.8815,
+      "step": 285000
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 1.9296770302867865e-05,
+      "loss": 1.884,
+      "step": 285500
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 1.924115518627714e-05,
+      "loss": 1.8801,
+      "step": 286000
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 1.9185316715804523e-05,
+      "loss": 1.8835,
+      "step": 286500
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 1.9129701599213794e-05,
+      "loss": 1.8803,
+      "step": 287000
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 1.907386312874118e-05,
+      "loss": 1.8807,
+      "step": 287500
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 1.9018248012150452e-05,
+      "loss": 1.8786,
+      "step": 288000
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 1.8962409541677835e-05,
+      "loss": 1.8704,
+      "step": 288500
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 1.890679442508711e-05,
+      "loss": 1.876,
+      "step": 289000
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 1.885095595461449e-05,
+      "loss": 1.8795,
+      "step": 289500
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 1.8795340838023765e-05,
+      "loss": 1.8752,
+      "step": 290000
+    },
+    {
+      "epoch": 3.17,
+      "eval_accuracy": 0.6442635719569795,
+      "eval_loss": 1.8037109375,
+      "eval_runtime": 1091.3905,
+      "eval_samples_per_second": 282.535,
+      "eval_steps_per_second": 4.415,
+      "step": 290000
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 1.8739502367551148e-05,
+      "loss": 1.8821,
+      "step": 290500
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 1.8683887250960423e-05,
+      "loss": 1.8784,
+      "step": 291000
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 1.8628048780487806e-05,
+      "loss": 1.8773,
+      "step": 291500
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 1.857243366389708e-05,
+      "loss": 1.8748,
+      "step": 292000
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 1.8516595193424464e-05,
+      "loss": 1.8694,
+      "step": 292500
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 1.8460980076833735e-05,
+      "loss": 1.8677,
+      "step": 293000
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 1.8405141606361122e-05,
+      "loss": 1.8778,
+      "step": 293500
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 1.8349526489770393e-05,
+      "loss": 1.8749,
+      "step": 294000
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 1.8293688019297773e-05,
+      "loss": 1.8698,
+      "step": 294500
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 1.823807290270705e-05,
+      "loss": 1.8729,
+      "step": 295000
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 1.818223443223443e-05,
+      "loss": 1.8699,
+      "step": 295500
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 1.8126619315643706e-05,
+      "loss": 1.872,
+      "step": 296000
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 1.807078084517109e-05,
+      "loss": 1.8648,
+      "step": 296500
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 1.8015165728580364e-05,
+      "loss": 1.8712,
+      "step": 297000
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 1.7959327258107747e-05,
+      "loss": 1.8692,
+      "step": 297500
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 1.7903712141517022e-05,
+      "loss": 1.8731,
+      "step": 298000
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 1.7847873671044405e-05,
+      "loss": 1.8671,
+      "step": 298500
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 1.7792258554453677e-05,
+      "loss": 1.8672,
+      "step": 299000
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 1.7736420083981063e-05,
+      "loss": 1.8684,
+      "step": 299500
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 1.7680804967390335e-05,
+      "loss": 1.8662,
+      "step": 300000
+    },
+    {
+      "epoch": 3.28,
+      "eval_accuracy": 0.6457191174875389,
+      "eval_loss": 1.794921875,
+      "eval_runtime": 1092.3358,
+      "eval_samples_per_second": 282.29,
+      "eval_steps_per_second": 4.412,
+      "step": 300000
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 1.7624966496917715e-05,
+      "loss": 1.8683,
+      "step": 300500
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 1.756935138032699e-05,
+      "loss": 1.8664,
+      "step": 301000
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 1.7513512909854373e-05,
+      "loss": 1.8733,
+      "step": 301500
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 1.7457897793263648e-05,
+      "loss": 1.867,
+      "step": 302000
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 1.740205932279103e-05,
+      "loss": 1.8667,
+      "step": 302500
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 1.7346444206200306e-05,
+      "loss": 1.8672,
+      "step": 303000
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 1.729060573572769e-05,
+      "loss": 1.87,
+      "step": 303500
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 1.7234990619136964e-05,
+      "loss": 1.8707,
+      "step": 304000
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 1.7179152148664347e-05,
+      "loss": 1.8679,
+      "step": 304500
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 1.7123537032073618e-05,
+      "loss": 1.8665,
+      "step": 305000
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.7067698561601e-05,
+      "loss": 1.8603,
+      "step": 305500
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.7012083445010273e-05,
+      "loss": 1.8664,
+      "step": 306000
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.6956244974537656e-05,
+      "loss": 1.8612,
+      "step": 306500
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.690062985794693e-05,
+      "loss": 1.8667,
+      "step": 307000
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.6844791387474314e-05,
+      "loss": 1.8608,
+      "step": 307500
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.678917627088359e-05,
+      "loss": 1.8606,
+      "step": 308000
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.6733337800410972e-05,
+      "loss": 1.8592,
+      "step": 308500
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.6677722683820247e-05,
+      "loss": 1.8641,
+      "step": 309000
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.662188421334763e-05,
+      "loss": 1.8584,
+      "step": 309500
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.6566269096756905e-05,
+      "loss": 1.8575,
+      "step": 310000
+    },
+    {
+      "epoch": 3.39,
+      "eval_accuracy": 0.6469689531293227,
+      "eval_loss": 1.787109375,
+      "eval_runtime": 1090.9668,
+      "eval_samples_per_second": 282.645,
+      "eval_steps_per_second": 4.417,
+      "step": 310000
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.6510430626284285e-05,
+      "loss": 1.8563,
+      "step": 310500
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.6454815509693556e-05,
+      "loss": 1.8623,
+      "step": 311000
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.6398977039220943e-05,
+      "loss": 1.8601,
+      "step": 311500
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.6343361922630214e-05,
+      "loss": 1.8568,
+      "step": 312000
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.6287523452157598e-05,
+      "loss": 1.8626,
+      "step": 312500
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.6231908335566872e-05,
+      "loss": 1.8553,
+      "step": 313000
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.6176069865094256e-05,
+      "loss": 1.8546,
+      "step": 313500
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.612045474850353e-05,
+      "loss": 1.8576,
+      "step": 314000
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.6064616278030914e-05,
+      "loss": 1.8591,
+      "step": 314500
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.600900116144019e-05,
+      "loss": 1.8626,
+      "step": 315000
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.5953162690967568e-05,
+      "loss": 1.8535,
+      "step": 315500
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.5897547574376843e-05,
+      "loss": 1.856,
+      "step": 316000
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.5841709103904226e-05,
+      "loss": 1.8562,
+      "step": 316500
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.5786093987313498e-05,
+      "loss": 1.8535,
+      "step": 317000
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.5730255516840884e-05,
+      "loss": 1.858,
+      "step": 317500
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.5674640400250156e-05,
+      "loss": 1.8552,
+      "step": 318000
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.561880192977754e-05,
+      "loss": 1.852,
+      "step": 318500
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.5563186813186814e-05,
+      "loss": 1.851,
+      "step": 319000
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.5507348342714197e-05,
+      "loss": 1.849,
+      "step": 319500
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.5451733226123472e-05,
+      "loss": 1.8538,
+      "step": 320000
+    },
+    {
+      "epoch": 3.5,
+      "eval_accuracy": 0.6478183167003876,
+      "eval_loss": 1.779296875,
+      "eval_runtime": 1093.454,
+      "eval_samples_per_second": 282.002,
+      "eval_steps_per_second": 4.407,
+      "step": 320000
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.539589475565085e-05,
+      "loss": 1.8531,
+      "step": 320500
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.534027963906013e-05,
+      "loss": 1.8541,
+      "step": 321000
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.528444116858751e-05,
+      "loss": 1.8532,
+      "step": 321500
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.5228826051996783e-05,
+      "loss": 1.8527,
+      "step": 322000
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.5172987581524168e-05,
+      "loss": 1.855,
+      "step": 322500
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.511737246493344e-05,
+      "loss": 1.8474,
+      "step": 323000
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.5061533994460824e-05,
+      "loss": 1.8451,
+      "step": 323500
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.5005918877870099e-05,
+      "loss": 1.8498,
+      "step": 324000
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.4950080407397482e-05,
+      "loss": 1.8517,
+      "step": 324500
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.4894465290806755e-05,
+      "loss": 1.8478,
+      "step": 325000
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.483862682033414e-05,
+      "loss": 1.8477,
+      "step": 325500
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.4783011703743413e-05,
+      "loss": 1.851,
+      "step": 326000
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.4727173233270795e-05,
+      "loss": 1.8511,
+      "step": 326500
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.4671558116680068e-05,
+      "loss": 1.8536,
+      "step": 327000
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.4615719646207451e-05,
+      "loss": 1.847,
+      "step": 327500
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.4560104529616724e-05,
+      "loss": 1.8487,
+      "step": 328000
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.4504266059144109e-05,
+      "loss": 1.8507,
+      "step": 328500
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.4448650942553382e-05,
+      "loss": 1.8466,
+      "step": 329000
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.4392812472080765e-05,
+      "loss": 1.8448,
+      "step": 329500
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.433719735549004e-05,
+      "loss": 1.8426,
+      "step": 330000
+    },
+    {
+      "epoch": 3.6,
+      "eval_accuracy": 0.6488575128878505,
+      "eval_loss": 1.7734375,
+      "eval_runtime": 1118.1827,
+      "eval_samples_per_second": 275.765,
+      "eval_steps_per_second": 4.31,
+      "step": 330000
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.4281358885017423e-05,
+      "loss": 1.8477,
+      "step": 330500
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.4225743768426697e-05,
+      "loss": 1.8491,
+      "step": 331000
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.4169905297954078e-05,
+      "loss": 1.8402,
+      "step": 331500
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.4114290181363351e-05,
+      "loss": 1.843,
+      "step": 332000
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.4058451710890736e-05,
+      "loss": 1.8444,
+      "step": 332500
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.400283659430001e-05,
+      "loss": 1.8424,
+      "step": 333000
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.3946998123827392e-05,
+      "loss": 1.8499,
+      "step": 333500
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.3891383007236666e-05,
+      "loss": 1.8437,
+      "step": 334000
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.383554453676405e-05,
+      "loss": 1.8441,
+      "step": 334500
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.3779929420173324e-05,
+      "loss": 1.8397,
+      "step": 335000
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.3724090949700707e-05,
+      "loss": 1.8444,
+      "step": 335500
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.3668475833109982e-05,
+      "loss": 1.8394,
+      "step": 336000
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.3612637362637362e-05,
+      "loss": 1.8431,
+      "step": 336500
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.3557022246046636e-05,
+      "loss": 1.8396,
+      "step": 337000
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.350118377557402e-05,
+      "loss": 1.8363,
+      "step": 337500
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.3445568658983293e-05,
+      "loss": 1.8372,
+      "step": 338000
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.3389730188510678e-05,
+      "loss": 1.8385,
+      "step": 338500
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.333411507191995e-05,
+      "loss": 1.8399,
+      "step": 339000
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.3278276601447334e-05,
+      "loss": 1.8401,
+      "step": 339500
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.3222661484856607e-05,
+      "loss": 1.8389,
+      "step": 340000
+    },
+    {
+      "epoch": 3.71,
+      "eval_accuracy": 0.650145600351507,
+      "eval_loss": 1.7646484375,
+      "eval_runtime": 1118.7806,
+      "eval_samples_per_second": 275.618,
+      "eval_steps_per_second": 4.307,
+      "step": 340000
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.3166823014383992e-05,
+      "loss": 1.8357,
+      "step": 340500
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.3111207897793265e-05,
+      "loss": 1.8402,
+      "step": 341000
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.3055369427320647e-05,
+      "loss": 1.8349,
+      "step": 341500
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.2999754310729923e-05,
+      "loss": 1.8359,
+      "step": 342000
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.2943915840257303e-05,
+      "loss": 1.839,
+      "step": 342500
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.2888300723666578e-05,
+      "loss": 1.8331,
+      "step": 343000
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.2832462253193961e-05,
+      "loss": 1.8348,
+      "step": 343500
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.2776847136603234e-05,
+      "loss": 1.8394,
+      "step": 344000
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.2721008666130619e-05,
+      "loss": 1.8347,
+      "step": 344500
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.2665393549539892e-05,
+      "loss": 1.8368,
+      "step": 345000
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.2609555079067275e-05,
+      "loss": 1.8331,
+      "step": 345500
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.2553939962476548e-05,
+      "loss": 1.8302,
+      "step": 346000
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.2498101492003932e-05,
+      "loss": 1.836,
+      "step": 346500
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.2442486375413205e-05,
+      "loss": 1.835,
+      "step": 347000
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.238664790494059e-05,
+      "loss": 1.8351,
+      "step": 347500
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.2331032788349863e-05,
+      "loss": 1.8323,
+      "step": 348000
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.2275194317877244e-05,
+      "loss": 1.831,
+      "step": 348500
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.221957920128652e-05,
+      "loss": 1.8349,
+      "step": 349000
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.2163740730813902e-05,
+      "loss": 1.8326,
+      "step": 349500
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.2108125614223176e-05,
+      "loss": 1.8278,
+      "step": 350000
+    },
+    {
+      "epoch": 3.82,
+      "eval_accuracy": 0.651100408667841,
+      "eval_loss": 1.759765625,
+      "eval_runtime": 1097.038,
+      "eval_samples_per_second": 281.081,
+      "eval_steps_per_second": 4.393,
+      "step": 350000
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.2052287143750559e-05,
+      "loss": 1.8426,
+      "step": 350500
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.1996672027159834e-05,
+      "loss": 1.8314,
+      "step": 351000
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.1940833556687215e-05,
+      "loss": 1.8348,
+      "step": 351500
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.188521844009649e-05,
+      "loss": 1.827,
+      "step": 352000
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.1829379969623873e-05,
+      "loss": 1.828,
+      "step": 352500
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.1773764853033146e-05,
+      "loss": 1.8306,
+      "step": 353000
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.171792638256053e-05,
+      "loss": 1.8321,
+      "step": 353500
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.1662311265969803e-05,
+      "loss": 1.8293,
+      "step": 354000
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.1606472795497186e-05,
+      "loss": 1.8292,
+      "step": 354500
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.155085767890646e-05,
+      "loss": 1.8267,
+      "step": 355000
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.1495019208433844e-05,
+      "loss": 1.8349,
+      "step": 355500
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.1439404091843117e-05,
+      "loss": 1.8303,
+      "step": 356000
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.13835656213705e-05,
+      "loss": 1.8295,
+      "step": 356500
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.1327950504779773e-05,
+      "loss": 1.8285,
+      "step": 357000
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.1272112034307156e-05,
+      "loss": 1.8326,
+      "step": 357500
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.1216496917716431e-05,
+      "loss": 1.8239,
+      "step": 358000
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.1160658447243813e-05,
+      "loss": 1.8253,
+      "step": 358500
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.1105043330653088e-05,
+      "loss": 1.824,
+      "step": 359000
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.104920486018047e-05,
+      "loss": 1.831,
+      "step": 359500
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.0993589743589744e-05,
+      "loss": 1.8319,
+      "step": 360000
+    },
+    {
+      "epoch": 3.93,
+      "eval_accuracy": 0.6520117131015779,
+      "eval_loss": 1.7529296875,
+      "eval_runtime": 1095.3232,
+      "eval_samples_per_second": 281.521,
+      "eval_steps_per_second": 4.4,
+      "step": 360000
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.0937751273117127e-05,
+      "loss": 1.8286,
+      "step": 360500
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.0882136156526402e-05,
+      "loss": 1.8253,
+      "step": 361000
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.0826297686053783e-05,
+      "loss": 1.8312,
+      "step": 361500
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.0770682569463057e-05,
+      "loss": 1.8271,
+      "step": 362000
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.0714844098990442e-05,
+      "loss": 1.8258,
+      "step": 362500
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.0659228982399715e-05,
+      "loss": 1.8274,
+      "step": 363000
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.0603390511927098e-05,
+      "loss": 1.8201,
+      "step": 363500
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.0547775395336373e-05,
+      "loss": 1.8269,
+      "step": 364000
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.0491936924863754e-05,
+      "loss": 1.826,
+      "step": 364500
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.0436321808273027e-05,
+      "loss": 1.8206,
+      "step": 365000
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.0380483337800412e-05,
+      "loss": 1.8279,
+      "step": 365500
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.0324868221209685e-05,
+      "loss": 1.8242,
+      "step": 366000
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.0269029750737069e-05,
+      "loss": 1.8204,
+      "step": 366500
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.0213414634146342e-05,
+      "loss": 1.8198,
+      "step": 367000
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.0157576163673725e-05,
+      "loss": 1.8219,
+      "step": 367500
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.0101961047082998e-05,
+      "loss": 1.8195,
+      "step": 368000
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.0046122576610383e-05,
+      "loss": 1.8181,
+      "step": 368500
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 9.990507460019656e-06,
+      "loss": 1.8245,
+      "step": 369000
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 9.93466898954704e-06,
+      "loss": 1.8179,
+      "step": 369500
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 9.879053872956312e-06,
+      "loss": 1.8203,
+      "step": 370000
+    },
+    {
+      "epoch": 4.04,
+      "eval_accuracy": 0.6527260129714422,
+      "eval_loss": 1.7470703125,
+      "eval_runtime": 1100.1141,
+      "eval_samples_per_second": 280.295,
+      "eval_steps_per_second": 4.38,
+      "step": 370000
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 9.823215402483696e-06,
+      "loss": 1.8153,
+      "step": 370500
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 9.767600285892969e-06,
+      "loss": 1.8176,
+      "step": 371000
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 9.711761815420352e-06,
+      "loss": 1.8204,
+      "step": 371500
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 9.656146698829627e-06,
+      "loss": 1.8176,
+      "step": 372000
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 9.60030822835701e-06,
+      "loss": 1.8161,
+      "step": 372500
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 9.544693111766283e-06,
+      "loss": 1.8186,
+      "step": 373000
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 9.488854641293666e-06,
+      "loss": 1.8183,
+      "step": 373500
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 9.43323952470294e-06,
+      "loss": 1.8181,
+      "step": 374000
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 9.377401054230323e-06,
+      "loss": 1.8131,
+      "step": 374500
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 9.321785937639596e-06,
+      "loss": 1.8174,
+      "step": 375000
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 9.26594746716698e-06,
+      "loss": 1.8157,
+      "step": 375500
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 9.210332350576254e-06,
+      "loss": 1.8124,
+      "step": 376000
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 9.154493880103637e-06,
+      "loss": 1.8173,
+      "step": 376500
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 9.09887876351291e-06,
+      "loss": 1.8158,
+      "step": 377000
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 9.043040293040293e-06,
+      "loss": 1.8163,
+      "step": 377500
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 8.987425176449567e-06,
+      "loss": 1.8144,
+      "step": 378000
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 8.931586705976951e-06,
+      "loss": 1.8201,
+      "step": 378500
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 8.875971589386225e-06,
+      "loss": 1.8151,
+      "step": 379000
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 8.820133118913606e-06,
+      "loss": 1.8133,
+      "step": 379500
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 8.764518002322881e-06,
+      "loss": 1.8162,
+      "step": 380000
+    },
+    {
+      "epoch": 4.15,
+      "eval_accuracy": 0.6536149151745975,
+      "eval_loss": 1.7412109375,
+      "eval_runtime": 1101.5449,
+      "eval_samples_per_second": 279.93,
+      "eval_steps_per_second": 4.375,
+      "step": 380000
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 8.708679531850264e-06,
+      "loss": 1.8092,
+      "step": 380500
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 8.653064415259537e-06,
+      "loss": 1.8136,
+      "step": 381000
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.597225944786922e-06,
+      "loss": 1.8141,
+      "step": 381500
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.541610828196195e-06,
+      "loss": 1.8116,
+      "step": 382000
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 8.485772357723577e-06,
+      "loss": 1.8128,
+      "step": 382500
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 8.430157241132852e-06,
+      "loss": 1.8134,
+      "step": 383000
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 8.374318770660235e-06,
+      "loss": 1.8128,
+      "step": 383500
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 8.318703654069508e-06,
+      "loss": 1.8073,
+      "step": 384000
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 8.262865183596891e-06,
+      "loss": 1.8102,
+      "step": 384500
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 8.207250067006166e-06,
+      "loss": 1.8114,
+      "step": 385000
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 8.151411596533547e-06,
+      "loss": 1.8091,
+      "step": 385500
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 8.095796479942822e-06,
+      "loss": 1.8082,
+      "step": 386000
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 8.039958009470205e-06,
+      "loss": 1.8159,
+      "step": 386500
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 7.984342892879479e-06,
+      "loss": 1.8124,
+      "step": 387000
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 7.928504422406862e-06,
+      "loss": 1.8109,
+      "step": 387500
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 7.872889305816135e-06,
+      "loss": 1.8148,
+      "step": 388000
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 7.817050835343518e-06,
+      "loss": 1.813,
+      "step": 388500
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 7.761435718752793e-06,
+      "loss": 1.8152,
+      "step": 389000
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 7.705597248280176e-06,
+      "loss": 1.8099,
+      "step": 389500
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 7.64998213168945e-06,
+      "loss": 1.8113,
+      "step": 390000
+    },
+    {
+      "epoch": 4.26,
+      "eval_accuracy": 0.6542793430067383,
+      "eval_loss": 1.7373046875,
+      "eval_runtime": 1096.3677,
+      "eval_samples_per_second": 281.252,
+      "eval_steps_per_second": 4.395,
+      "step": 390000
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 7.594143661216832e-06,
+      "loss": 1.8111,
+      "step": 390500
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 7.538528544626106e-06,
+      "loss": 1.8098,
+      "step": 391000
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 7.48269007415349e-06,
+      "loss": 1.807,
+      "step": 391500
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 7.427074957562763e-06,
+      "loss": 1.8063,
+      "step": 392000
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 7.371236487090145e-06,
+      "loss": 1.8079,
+      "step": 392500
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 7.31562137049942e-06,
+      "loss": 1.8064,
+      "step": 393000
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 7.259782900026802e-06,
+      "loss": 1.8082,
+      "step": 393500
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 7.204167783436076e-06,
+      "loss": 1.8075,
+      "step": 394000
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 7.1483293129634604e-06,
+      "loss": 1.8078,
+      "step": 394500
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 7.092714196372734e-06,
+      "loss": 1.8005,
+      "step": 395000
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 7.036875725900116e-06,
+      "loss": 1.8048,
+      "step": 395500
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 6.98126060930939e-06,
+      "loss": 1.8076,
+      "step": 396000
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 6.925422138836773e-06,
+      "loss": 1.8055,
+      "step": 396500
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 6.869807022246047e-06,
+      "loss": 1.805,
+      "step": 397000
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 6.8139685517734294e-06,
+      "loss": 1.8104,
+      "step": 397500
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 6.758353435182704e-06,
+      "loss": 1.812,
+      "step": 398000
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 6.702514964710087e-06,
+      "loss": 1.8072,
+      "step": 398500
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 6.646899848119361e-06,
+      "loss": 1.8032,
+      "step": 399000
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 6.591061377646744e-06,
+      "loss": 1.8022,
+      "step": 399500
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 6.535446261056018e-06,
+      "loss": 1.8055,
+      "step": 400000
+    },
+    {
+      "epoch": 4.37,
+      "eval_accuracy": 0.655123594491739,
+      "eval_loss": 1.732421875,
+      "eval_runtime": 1175.7931,
+      "eval_samples_per_second": 262.254,
+      "eval_steps_per_second": 4.099,
+      "step": 400000
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 6.4796077905834e-06,
+      "loss": 1.806,
+      "step": 400500
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 6.423992673992673e-06,
+      "loss": 1.8084,
+      "step": 401000
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 6.368154203520057e-06,
+      "loss": 1.8053,
+      "step": 401500
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 6.312539086929331e-06,
+      "loss": 1.8048,
+      "step": 402000
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 6.2567006164567145e-06,
+      "loss": 1.8099,
+      "step": 402500
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 6.201085499865988e-06,
+      "loss": 1.8034,
+      "step": 403000
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 6.145247029393371e-06,
+      "loss": 1.8049,
+      "step": 403500
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 6.089631912802645e-06,
+      "loss": 1.8046,
+      "step": 404000
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 6.033793442330028e-06,
+      "loss": 1.8008,
+      "step": 404500
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.978178325739302e-06,
+      "loss": 1.8016,
+      "step": 405000
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.922339855266685e-06,
+      "loss": 1.8048,
+      "step": 405500
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.866724738675958e-06,
+      "loss": 1.8014,
+      "step": 406000
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.8108862682033416e-06,
+      "loss": 1.8031,
+      "step": 406500
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.755271151612615e-06,
+      "loss": 1.8013,
+      "step": 407000
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.699432681139998e-06,
+      "loss": 1.7957,
+      "step": 407500
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.643817564549273e-06,
+      "loss": 1.8082,
+      "step": 408000
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.587979094076656e-06,
+      "loss": 1.8064,
+      "step": 408500
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.532363977485929e-06,
+      "loss": 1.7994,
+      "step": 409000
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.476525507013312e-06,
+      "loss": 1.8027,
+      "step": 409500
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.4209103904225854e-06,
+      "loss": 1.7991,
+      "step": 410000
+    },
+    {
+      "epoch": 4.48,
+      "eval_accuracy": 0.6556022179984572,
+      "eval_loss": 1.728515625,
+      "eval_runtime": 1174.9994,
+      "eval_samples_per_second": 262.431,
+      "eval_steps_per_second": 4.101,
+      "step": 410000
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.365071919949969e-06,
+      "loss": 1.8046,
+      "step": 410500
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.309456803359243e-06,
+      "loss": 1.8017,
+      "step": 411000
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.253618332886626e-06,
+      "loss": 1.8038,
+      "step": 411500
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.198003216295899e-06,
+      "loss": 1.8046,
+      "step": 412000
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.142164745823282e-06,
+      "loss": 1.8046,
+      "step": 412500
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.086549629232556e-06,
+      "loss": 1.7988,
+      "step": 413000
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.030711158759939e-06,
+      "loss": 1.7947,
+      "step": 413500
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 4.975096042169213e-06,
+      "loss": 1.7999,
+      "step": 414000
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 4.9192575716965965e-06,
+      "loss": 1.7909,
+      "step": 414500
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 4.86364245510587e-06,
+      "loss": 1.8029,
+      "step": 415000
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.807803984633253e-06,
+      "loss": 1.8017,
+      "step": 415500
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.752188868042527e-06,
+      "loss": 1.7924,
+      "step": 416000
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 4.69635039756991e-06,
+      "loss": 1.797,
+      "step": 416500
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 4.640735280979184e-06,
+      "loss": 1.7941,
+      "step": 417000
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 4.584896810506567e-06,
+      "loss": 1.7972,
+      "step": 417500
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 4.52928169391584e-06,
+      "loss": 1.8002,
+      "step": 418000
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 4.4734432234432235e-06,
+      "loss": 1.796,
+      "step": 418500
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 4.4178281068524975e-06,
+      "loss": 1.8012,
+      "step": 419000
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 4.361989636379881e-06,
+      "loss": 1.7975,
+      "step": 419500
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.306374519789154e-06,
+      "loss": 1.7965,
+      "step": 420000
+    },
+    {
+      "epoch": 4.59,
+      "eval_accuracy": 0.6561621453610571,
+      "eval_loss": 1.724609375,
+      "eval_runtime": 1175.4978,
+      "eval_samples_per_second": 262.32,
+      "eval_steps_per_second": 4.1,
+      "step": 420000
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.250536049316537e-06,
+      "loss": 1.7974,
+      "step": 420500
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 4.194920932725811e-06,
+      "loss": 1.7986,
+      "step": 421000
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 4.139082462253194e-06,
+      "loss": 1.7967,
+      "step": 421500
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 4.083467345662468e-06,
+      "loss": 1.7991,
+      "step": 422000
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 4.0276288751898514e-06,
+      "loss": 1.7968,
+      "step": 422500
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.972013758599125e-06,
+      "loss": 1.7939,
+      "step": 423000
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.916175288126508e-06,
+      "loss": 1.7957,
+      "step": 423500
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.860560171535781e-06,
+      "loss": 1.7962,
+      "step": 424000
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.8047217010631645e-06,
+      "loss": 1.7939,
+      "step": 424500
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.7491065844724385e-06,
+      "loss": 1.7937,
+      "step": 425000
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.6932681139998217e-06,
+      "loss": 1.7939,
+      "step": 425500
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.6376529974090953e-06,
+      "loss": 1.7954,
+      "step": 426000
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.5818145269364785e-06,
+      "loss": 1.7946,
+      "step": 426500
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.526199410345752e-06,
+      "loss": 1.7933,
+      "step": 427000
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.4703609398731352e-06,
+      "loss": 1.7967,
+      "step": 427500
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.414745823282409e-06,
+      "loss": 1.7946,
+      "step": 428000
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.358907352809792e-06,
+      "loss": 1.7916,
+      "step": 428500
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.303292236219065e-06,
+      "loss": 1.7982,
+      "step": 429000
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.2474537657464483e-06,
+      "loss": 1.7951,
+      "step": 429500
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.1918386491557228e-06,
+      "loss": 1.7938,
+      "step": 430000
+    },
+    {
+      "epoch": 4.7,
+      "eval_accuracy": 0.6567150458487134,
+      "eval_loss": 1.720703125,
+      "eval_runtime": 1096.463,
+      "eval_samples_per_second": 281.228,
+      "eval_steps_per_second": 4.395,
+      "step": 430000
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.136000178683106e-06,
+      "loss": 1.7908,
+      "step": 430500
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.0803850620923795e-06,
+      "loss": 1.798,
+      "step": 431000
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.0245465916197627e-06,
+      "loss": 1.796,
+      "step": 431500
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.968931475029036e-06,
+      "loss": 1.7931,
+      "step": 432000
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.913093004556419e-06,
+      "loss": 1.7915,
+      "step": 432500
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 2.857477887965693e-06,
+      "loss": 1.7932,
+      "step": 433000
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.8016394174930762e-06,
+      "loss": 1.7867,
+      "step": 433500
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.74602430090235e-06,
+      "loss": 1.7906,
+      "step": 434000
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.690185830429733e-06,
+      "loss": 1.7915,
+      "step": 434500
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.6345707138390066e-06,
+      "loss": 1.7906,
+      "step": 435000
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.5787322433663898e-06,
+      "loss": 1.7956,
+      "step": 435500
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.5231171267756633e-06,
+      "loss": 1.793,
+      "step": 436000
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.4672786563030465e-06,
+      "loss": 1.7897,
+      "step": 436500
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.4116635397123205e-06,
+      "loss": 1.7912,
+      "step": 437000
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.3558250692397037e-06,
+      "loss": 1.7917,
+      "step": 437500
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.3002099526489773e-06,
+      "loss": 1.7924,
+      "step": 438000
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.2443714821763605e-06,
+      "loss": 1.7868,
+      "step": 438500
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.188756365585634e-06,
+      "loss": 1.7911,
+      "step": 439000
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.1329178951130172e-06,
+      "loss": 1.7874,
+      "step": 439500
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.077302778522291e-06,
+      "loss": 1.793,
+      "step": 440000
+    },
+    {
+      "epoch": 4.81,
+      "eval_accuracy": 0.6571231862326748,
+      "eval_loss": 1.7177734375,
+      "eval_runtime": 1080.1698,
+      "eval_samples_per_second": 285.47,
+      "eval_steps_per_second": 4.461,
+      "step": 440000
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.021464308049674e-06,
+      "loss": 1.784,
+      "step": 440500
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.965849191458948e-06,
+      "loss": 1.7901,
+      "step": 441000
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.910010720986331e-06,
+      "loss": 1.7949,
+      "step": 441500
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.8543956043956045e-06,
+      "loss": 1.7868,
+      "step": 442000
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.7985571339229877e-06,
+      "loss": 1.789,
+      "step": 442500
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.742942017332261e-06,
+      "loss": 1.7878,
+      "step": 443000
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.6871035468596447e-06,
+      "loss": 1.7864,
+      "step": 443500
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.6314884302689183e-06,
+      "loss": 1.7892,
+      "step": 444000
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5756499597963012e-06,
+      "loss": 1.794,
+      "step": 444500
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.520034843205575e-06,
+      "loss": 1.7899,
+      "step": 445000
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.4641963727329582e-06,
+      "loss": 1.788,
+      "step": 445500
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.4085812561422318e-06,
+      "loss": 1.7869,
+      "step": 446000
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.352742785669615e-06,
+      "loss": 1.7861,
+      "step": 446500
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.2971276690788888e-06,
+      "loss": 1.7895,
+      "step": 447000
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.241289198606272e-06,
+      "loss": 1.791,
+      "step": 447500
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.1856740820155455e-06,
+      "loss": 1.7922,
+      "step": 448000
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.1298356115429287e-06,
+      "loss": 1.7853,
+      "step": 448500
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.0742204949522025e-06,
+      "loss": 1.7862,
+      "step": 449000
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.0183820244795855e-06,
+      "loss": 1.7849,
+      "step": 449500
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 9.62766907888859e-07,
+      "loss": 1.7848,
+      "step": 450000
+    },
+    {
+      "epoch": 4.92,
+      "eval_accuracy": 0.6576050004678714,
+      "eval_loss": 1.71484375,
+      "eval_runtime": 1081.0209,
+      "eval_samples_per_second": 285.245,
+      "eval_steps_per_second": 4.458,
+      "step": 450000
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 9.069284374162424e-07,
+      "loss": 1.787,
+      "step": 450500
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 8.513133208255159e-07,
+      "loss": 1.7882,
+      "step": 451000
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 7.954748503528992e-07,
+      "loss": 1.7849,
+      "step": 451500
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 7.398597337621729e-07,
+      "loss": 1.7889,
+      "step": 452000
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 6.84021263289556e-07,
+      "loss": 1.79,
+      "step": 452500
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 6.284061466988297e-07,
+      "loss": 1.7851,
+      "step": 453000
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.725676762262128e-07,
+      "loss": 1.7859,
+      "step": 453500
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.169525596354864e-07,
+      "loss": 1.7877,
+      "step": 454000
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 4.611140891628697e-07,
+      "loss": 1.7828,
+      "step": 454500
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.054989725721433e-07,
+      "loss": 1.7882,
+      "step": 455000
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.496605020995265e-07,
+      "loss": 1.7847,
+      "step": 455500
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.9404538550880015e-07,
+      "loss": 1.7876,
+      "step": 456000
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.3820691503618333e-07,
+      "loss": 1.7887,
+      "step": 456500
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.82591798445457e-07,
+      "loss": 1.7844,
+      "step": 457000
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.2675332797284018e-07,
+      "loss": 1.7896,
+      "step": 457500
+    },
+    {
+      "epoch": 5.0,
+      "step": 457720,
+      "total_flos": 7.401345611745722e+18,
+      "train_loss": 2.2305684480447927,
+      "train_runtime": 322220.8599,
+      "train_samples_per_second": 90.912,
+      "train_steps_per_second": 1.421
+    }
+  ],
+  "max_steps": 457720,
+  "num_train_epochs": 5,
+  "total_flos": 7.401345611745722e+18,
+  "trial_name": null,
+  "trial_params": null
+}